Management of amphibians, reptiles, and small mammals in North America : proceedings of the symposium, July 19-21, 1988, Flagstaff, Arizona

Historic, Archive Document

Do not assume content reflects current
scientific knowledge, policies, or practices.

■1

sou

United States ^
; Department of
Agriculture

Forest Service

Rocky Mountain
Forest and Range
Experiment Station

Fort Collins,
Colorado 80526

General Technical
Report RM-166

Management of
Amphibians, Reptiles,
and Small Mammals ir
North America

Proceedings of the Symposium

July 19-21,1988
Flagstaff, Arizona

■■■■■■ "-T^^j^-^ .■■^'■i^^<'r~^

ACKNOWLEDGEMENTS

This meering owes its success to sev-
eral organizations and individuals.
First, we thank the sponsoring or-
ganizations (listed on the title page)
whose financial support and encour-
agement helped make the conference
a reality. The local committee on ar-
rangements, J. Kevin Aitkin, Marga-
ret Bailey, Tom Britt, Roxanne Britt,
Charles BuUington, Glen Dickens,
and Katherine Holly did a superb job
of handling room setup, registration,
providing rides, and running the
slide projector.

We are especially grateful to the
session chairman, K. Bruce Jones,
George Dalrymple, Robert
M,Closkey, David Germano,
Winifred Sidle, Constantine Slobod-
chikoff, Michael Morrison, Gregory
Adler, Martin Raphael, and Ray-
mond Dueser, for their help and for
keeping the meeting on schedule.
Our thanks to those who attended
for their enthusiastic participation.
We thank Randall Babb for the line
drawings in the proceedings and De-
borah Johnson and J. Kevin Aitkin
for their help in organizing manu-
script files and standardizing word
processing formats.

We would like to extend our sin-
cere thanks to the following peer re-
viewers who generously gave their
time to improve the quality of this
proceedings: Gregory H. Adler,
Stanley H. Anderson, Michael J.
Armbruster, David M. Armstrong,
Walter Auffenberg, Keith B. Aubry,
Gary C. Bateman, Ronald E.
Beiswenger, Kristin H. Berry, Wil-
liam M. Block, Michael A. Bowers,
Richard C. Bruce, James H. Brown,
K. A. Buhlmann, Russell Burke, R.
Bruce Bury, Ronald K. Chesser, Ste-
ven P. Christman, Tim W. Clark,
Tames P. Collins, Stephen Corn,
Stephen P. Cross, George Dalrymple,
Joan E. Diemer, James G. Dickson, C.
Kenneth Dodd, Jr., Raymond D.
Dueser, Gary M. Fellers , Henry S.
Fitch, Jerran Flinders, Vagn F. Flyger,
Kenneth Feluso, Richard Fitzner,
David J. Germano, Lowell L. Getz,
William E. Grant, Patrick T. Gregory,
Marc P. Hayes, Clyde Jones, K. Bruce
Jones, Donald W. Kaufman, Brian J.
Klatt, Thomas Kunz, J. Larry Lan-
ders, James N. Layne, Harvey B. Lil-
lywhite, Raymond Linder, William
Mannan, S. Clark Martin, Robert T.
M'Closkey, David A. McCullough,

Gary K. Meefe, Joseph C. Mitchell,
Paul E. Moler, Henry R. Mushinsky,
Thomas J. O'Shea, William S. Parker,
Kenneth H. Pollock, Mary V. Price,
Martin G. Raphael, O. J. Reichman,
Fred B. Samson, D. J. Schmidly, Nor-
man Scott, Steven W. Seagle, Ray-
mond D. Semlitsch, Henry L. Short,
Lee H. Simons, Graham W. Smith,
Hobart M. Smith, Dan Speake, James
R. Spotila, Judy A. Stamps, Thomas
P. Sullivan, Daniel W. Uresk, Laurie
J. Vitt, Peter D. Weigl, Gary C. White,
Daniel F. Williams, Richard G.
Zweifel.

Finally, we thank the speakers for
following our schedule for submit-
ting the various stages of their manu-
scripts and providing us with excel-
lent manuscripts in computer format
to expedite and enhance the publica-
tion of the proceedings. The opinions
expressed in these papers are the au-
thors' and do not necessarily reflect
those of the U.S. Department of Agri-
culture.

30301 Baltimore SVd
Beltsv/lle. MO 20705.2351

ral Lfbrary

USDA Forest Service

General Technical Report RM-166

November 1988

Management of Amphibians, Reptiles, and
Small Mammals in North America

Proceedings of the Symposium

July 19-21, 1988
Flagstaff, Arizona

Robert C. Szaro, Kieth E. Severson, and David R. Patton

technical coordinators^

Sponsored by:
Arizona Chapter of the Wildlife Society
Arizona Game and Fish Department
Northern Arizona University, School of Forestry
USDA Forest Service, Rocky Mountain Forest and Range Experiment Station
USDA Forest Service, National Wildlife and Fish Ecology Program
USDA Forest Service, Southwestern Region

'Szaro and Severson are with the USDA Forest Service, Rocky Mountain Forest and Range
Experiment Station, at the Station's Research VJork Unit in fempe, in cooperation with Arizona
State University. Patton is with the School of Forestry. Northern Arizona University, Flagstaff.

The Management of Amphibians, Reptiles and Small Mammals
in North America: Historical Perspective and Objectives

Robert C. Szaro 1

The Management of Amphibians, Reptiles and Small Mammals
in North America: The Need for an Environmental Attitude

J. Whitfield Gibbons 4

Douglas-Fir Forests in the Oregon and Washington Cascades:
Relation of the Herpetofauna to Stand Age and Moisture

R. Bruce Bury and Paul Stephen Corn 11

Long-Term Trends in Abundance of Amphibians, Reptiles, and
Mammals in Douglas-Fir Forests of Northwestern California

Martin G. Raphael 23

Use of Woody Debris by Plethodontid Salamanders in Douglas-
Fir in Washington

Keith B. Aubry, Lawrence L C. Jones, and Patricia A. Hall 32

Forestry Operations and Terrestrial Salamanders: Techniques in
a Study of the Cow Knob Salamander, Plethodon
punctafus

Kurt A. Buhlmann, Christopher A. Pague, Joseph C. r\/litchelL

and Robert B. Glasgow 38

Conserving Genetically Distinctive Populations: The Case of the
Huachuca Tiger Salamander (Ambystoma tighnum
sfebbinsi Lowe)

James P. Collins, Thomas R. Jones, and Howard J. Berna 45

Habitat Requirements of New Mexico's Endangered
Salamanders

Cynthia A. Ramotnik and Norman J. Scott, Jr. 54

Utilization of Abandoned Mine Drifts and Fracture Caves By Bats
and Salamanders: Unique Subterranean Habitat in the
Ouachita Mountains

David A. Saugey, Gary A. Heidt, and Darrell R. Heath 64

The Herpetofauna of Long Pine Key, Everglades National Park,
in Relation to Vegetation and Hydrology

George H. Dalrymple 72

The Herpetofaunal Community of Temporary Ponds in North
Florida Sandhills: Species Composition, Temporal Use, and
Management Implications

C. Kenneth Dodd, Jr. and Bert G. Charest 87

(Continued)

Management of Amphibians, Reptiles, and Small Mammals in
Xeric Pinelands of Peninsular Florida

/. Jock Stout, Donold R. Richordson, ond Richord E. Roberts 98

Distribution and Habitat Associations of Herpetofauna in
Arizona: Comparisons by Habitat Type

K. Bruce Jones 109

Multivariate Analysis of thie Summer Habitat Structure of Rana
pipiens Sctireber, in Lac Saint Pierre (Quebec, Canada)

N. Beouregord ond R. Lecloir Jr. 1 29

Habitat Correlates of Distribution of the California Red-Legged
Frog (Rona aurora draytonii) and the Foothill Yellow-
Legged Frog (Rana boylii): Implications for Management

More P. Hoyes ond Mork R. Jennings 144

Integrating Anuran Amphibian Species into Environmental
Assessment Programs

Ronold E. Beiswenger 159

PrelimirKjry Report on Effect of Bullfrogs on Wetland
Herpetofaunas in Southeastern Arizona

Cecil R. Schwolbe ond Philip C. Rosen 166

Developing Management Guidelines for Snapping Turtles

Ronold J. Brooks, Dovid A. Golbroith, E. Grohom Noncekiveli,

ond Christine A. Bishop 1 74

Spatial Distribution of Desert Tortoises (Gopherus agassizii) at
Twentynine Palms, California: Implications for Relocations

Ronold J. Boxter 180

Changes in a Desert Tortoise (Gopherus agassizii) Population
After a Period of High Mortality

Dovid J. Germono ond Michele A. Joyner 190

A Survey Method for Measuring Gopher Tortoise Density and
Habitat Distribution

Doniel M. Spillers ond Don W. Speoke 199

Evaluation and Review of Field Techniques Used to Study and
Manage Gopher Tortoises

Russell L Burke ond Jomes Cox 205

Talus Use by Amphibians and Reptiles in the Pacific Northwest

Robert E. Herrington 216

Comparison of Herpetofounas of a Natural and Altered Riparian

Ecosystem

K. Bruce Jones 222

Critical Habitat, Predator Pressures, and the Management of
Epicrates monoensis (Serpentes: Boidae) on the Puerto
Rico Bank: A Multivariate Analysis

Peter J. Tolson 228

The Use of Timed Fixed-Area Plots and a Mark-Recapture

Technique in Assessing Riparian Garter Snake Populations

Robert C. Szaro, Scott C. Belfit, J. Kevin Aitkin, and

Randall D. Babb 239

Design Considerations for the Study of Amphibians, Reptiles and
Small Mammals in California's Oak Woodlands: Temporal
and Spatial Patterns

William M. Block, Michael L Morrison, John C. Slaymaker,

and Gwen Jongejan 247

The Importance of Biological Surveys in Managing Public Lands
in the Western United States

Michael A. Began, Robert B. Finley, Jr, and

Stephen J. Petersburg 254

Sampling Problems in Estimating Small Mammal Population Size

George E. Menkens, Jr. and Stanley H. Anderson 262

The Design and Importance of Long-Term Ecological Studies:
Analysis of Vertebrates in the Inyo-White Mountains,
California

Michael L. Morrison 267

An Ecological Problem-Solving Process for Managing Special-
Interest Species

Henry L. Short and Samuel C. Williamson 276

Comparative Effectiveness of Pitfalls and Live-Traps in
Measuring Small Mammal Community Structure

Robert C. Szaro, Lee H. Simons, and Scott C. Belfit 282

The Role of Habitat Structure in Organizing Small Mammal
Populations and Communities

Gregory H. Adier 289

Microhabitat as a Template for the Organization of a Desert
Rodent Community

Michael A. Bowers and Christine A. Flanagan 300

(Continued)

Response of Small Mammal Communities to Silvicultural

Treatments in Eastem Hardwood Forests of West Virginia
and Massachusetts

Robert T. Brooks and William M. Healy 313

Habitat Structure and ttie Distribution of Small Mammals in a
Northiern Hardwoods Forest

JefferyA. Gore 319

Thie Value of Rocky Mountain Juniper (Juniperus scopulorum)
Woodlands in Soutti Dakota as Small Mammal Habitat

Carolyn Hull Sieg 328

Postfire Rodent Succession Following Prescribed Fire in Southern
California Chaparral

William O. Wirfz, II, David Hoekman, John R. Muhm, and

Sherrie L Sauza 333

Douglas-Fir Forests in the Cascade Mountains of Oregon and
Washington: Is the Abundance of Small Mammals Related
to Stand Age and Moisture?

Paul Stephen Corn, R. Bruce Bury, and Thomas A. Spies 340

Evaluation of Small Mammals as Ecological Indicators of Old-
Growth Conditions

Kirk A. Nordyke and Steven W. Buskirk 353

Habitat Associations of Small Mammals in a Subalpine Forest,
Southeastern Wyoming

Martin G. Raphael 359

Differences in the Ability of Vegetation Models to Predict Small
Mammal Abundance in Different Aged Douglas- Fir Forests

Cathy A. Taylor, C. John Ralph, andArlene T. Doyle 368

Small Mammals in Streamside Management Zones in Pine
Plantations

James G. Dickson and J. Howard Williamson 375

Patterns of Relative Diversity Within Riparian Small Mammal
Communities, Platte River Watershed, Colorado

Thomas E. Olson and Fritz L. Knopf 379

Estimated Carrying Capacity for Cattle Competing with Prairie
Dogs and Forage Utilization in Western South Dakota

Daniel W. Uresk and Deborah D. Paulson 387

(Continued)

Cattle Grazing and Small Mannmals on the Sheldon National
Wildlife Refuge, Nevada

John L Oldemeyer and Lydia R. Allen-Johnson 391

Effect of Seed Size on Removal by Rodents

William G. Standley 399

Habitat Use by Gunnison's Prairie Dogs

C, N. Slobodchikoff, Anthony Robinson, and Clark Schaack 403

Environmental Contaminants and the Management of Bat
Populations in the United States

Donald R. Clark, Jr. 409

Habitat Structure, Forest Composition and Landscape

Dimensions as Components of Habitat Suitability for the
Delmarva Fox Squirrel

Raymond D. Dueser, James L. Dooley, Jr., and Gary J. Taylor ....414

Effects of Treating Creosotebush with Tebuthiuron on Rodents

William G. Standley and Norman S. Smith 422

Foraging Patterns of Tassel-Eared Squirrels in Selected
Ponderosa Pine Stands

Jacks States, William S. Gaud, W. Sylvester Allred, and

William J. Austin 425

Small Mammal Response to the Introduction of Cattle into a
Cottonwood Floodplain

Fred B. Samson, Fritz L. Knopf, and Lisa B. Mass 432

Old Growth Forests and the Distribution of the Terrestrial
Herpetofauna

HartwellH. Welsh, Jr. and Amy L. Lind 439

The Management of
Amphibians, Reptiles and
Small Mammals in North
America: Historical
Perspective and Objectives^

Robert C. Szaro^

Historically the management of pub-
lic lands from a multiple use perspec-
tive has led to a system that empha-
sizes those habitat components or
faunal elements that primarily re-
sulted in some sort of definable eco-
nomic value. While this often benefit-
ted other species that were not even
considered in the original prescrip-
tions, it also negatively impacted oth-
ers. We no longer can afford to take
this simplistic view of ecosystem
management. We need to use a more
holistic approach where ecological
landscapes are considered as units,
and land management practices in-
corporate all elements into an inte-
grated policy. This includes examin-
ing the impacts of proposed land
uses on amphibian, reptile, and small
mammal populations.

With the passage of the National
Forest Management Act of 1976, the
monitoring of all renewable natural
resources became law. Even with this
legislation, most emphasis by Na-
tional Forests in the United States has
been placed on big game, other game
species, or threatened and endan-
gered species. Yet, the act lists five

'Paper presented at symposium. Man-
agement of Amphibians, Reptiles, and
Small Mammals in Nortt) America. (Flag-
staff, AZ, July 19-21 1988).

^Robert C. Szaro is Researct) Wildlife Bi-
ologist, USDA Forest Service, Rocky Moun-
tain Forest and Range Experiment Station,
at thie Station's Research) Work Unit in
Tempo, in cooperation with Arizona State
University. Station Headquarters is in Fort
Collins, in cooperation with Colorado State
University.

categories of management indicator
species: (1) endangered and threat-
ened plants and animals; (2) species
with special habitat needs; (3) species
commonly hunted, fished, or
trapped; (4) nongame species of spe-
cial interest; and (5) plant and animal
species selected because their popu-
lation changes are believed to indi-
cate the effects of management activi-
ties on other species of selected ma-
jor biological communities or on wa-
ter quality.

Nongame birds have been the first
group to benefit from changing man-
agement practices and public con-
cern. The management of nongame
birds within the National Forest Sys-
tem received a big boost from the
"Symposium on Management of For-
est and Range Habitats for Nongame
Birds" held in Tucson in May 1975
(Smith 1975). Since that initial sym-
posium, four regional workshops
were held emphasizing the manage-
ment of nongame birds in forest and
range habitats (Degraaf 1978a, 1978b;
Degraaf and Evans 1979; Degraaf
and Tilghman 1980). There have also
been Forest Service sponsored sym-
posia targeting specific bird groups
such as owls (Nero et al. 1987) and
birds using specific habitat features
such as snags (Davis et al. 1983).

Only recently has the management
of other nongame species gained in-
creased recognition. The landmark
symposium on "Herpetological
Communities" held in Lawrence,
Kansas, August 1977, as part of the

joint meeting of the Herpetologists'
League and the Society for the Study
of Amphibians and Reptiles, was the
first attempt to organize a vehicle for
the incorporation of papers dealing
with herpetological communities
(Scott 1982). Yet, as Gibbons (this
volume) clearly shows, little progress
has been made in the recognition of
amphibians, reptiles, and small
mammals as being important focal
points for research and management
efforts. It is encouraging that recent
comprehensive symposia have incor-
porated papers dealing with these
groups. There was an entire session
on Amphibians and Reptiles in the
symposium "Riparian Ecosystems
and Their Management" (Johnson et
al. 1985), and almost 30% of the
Southern Evaluation Project Work-
shop reports work on amphibians,
reptiles, and small mammals (Pear-
son et al. 1987).

The intent of this symposium was
to bring scientists and managers to-
gether to exchange knowledge and
ideas on habitat requirements, man-
agement needs, and other informa-
tion on these often overlooked com-
ponents of North American fauna.
Another purpose was to summarize
the state-of-the-science of habitats
and habitat requirements of species
within these groups. Of particular
interest were papers emphasizing
habitat models, habitat requirements,
sampling techniques and problems,
community dynamics, and manage-
ment recommendations.

1

The overwhelming response to
our announcement for papers was
unexpected. More than 60 abstracts
were originally submitted for presen-
tation. In order to overcome recent
criticism concerning so-called "gray"
literature (Bart and Anderson 1981,
Capen 1982, Finch et al. 1982, Scott
and Ralph 1988), we made every ef-
fort to improve the quality of the
symposium and its subsequent pro-
ceedings. All authors were required
to submit their first drafts 5 months
prior to the meeting in order to en-
sure adequate time for peer review
and editing. Each manuscript was
reviewed by two experts familiar
with the topic, and edited for style
and content by one of the sympo-
sium editors.

We found the meeting itself to be a
fertile exchange of ideas and tech-
niques between managers and re-
searchers from all over the country.
Those attending found the meeting
extremely enlightening both for re-
searchers and managers because of
their exposure to new viewpoints. It
is a testament to those attending and
the quality of the presentations that
very little discussion occurred out-
side the meeting hall when papers
were in progress. Virtually all partici-
pants were present throughout the
symposium, from the first session to
the last.

We hope this symposium will
prove to be the boost that these fau-
nal groups need to get increased re-
search and management recognition.
For only with an adequate data base
can models be developed that predict
diversity in relation to natural or
man-made disturbance of ecosys-
tems. These holistic models are of the
utmost importance for the mainte-
nance of worldwide biodiversity
(Wilson and Peters 1988). Ecosystem
diversity is a key correlate with bio-
logical productivity and has recently
attracted considerable interest both
from theoreticians and from profes-
sionals concerned with management
of land and water systems (Suffling
et al. 1988). We feel that amphibians.

reptiles, and small mammal popula-
tions may prove to be the ultimate
indicators of habitat quality and
health, because of their sedentary
characteristics which make them
much more susceptible to manage-
ment activities than do highly mobile
bird species and ubiquitous species
such as deer and turkey.

Literature Cited

Bart,], and D. R. Anderson. 1981.
The case against publishing sym-
posia proceedings. Wildlife Soci-
ety Bulletin 9:201-202.

Capen, David E. 1982. Publishing
symposia proceedings: another
viewpoint. Wildlife Society Bulle-
tin 10:183-184.

Davis, Jerry W., Gregory A. Good-
win, and Richard A. Ockenfeis
(Technical Coordinators). 1983.
Snag habitat management: Pro-
ceedings of the symposium.
USDA Forest Service General
Technical Report RM-99. Rocky
Mountain Forest and Range Ex-
periment Station, Ft. Collins, Colo.
226 p.

Degraaf, Richard M. (Technical Coor-
dinator). 1978a. Proceedings of the
workshop on nongame bird habi-
tat management in the coniferous
forests of the western United
States. USDA Forest Service Gen-
eral Technical Report PNW-64.
Pacific Northwest Forest and
Range Experiment Station, Port-
land, Oregon. 100 p.

Degraaf, Richard M. (Technical Coor-
dinator). 1978b. Proceedings of the
workshop: Management of south-
ern forests for nongame birds.
USDA Forest Service General
Technical Report SE-14. Southeast-
ern Forest Experiment Station,
Asheville, North Carolina. 176 p.

Degraaf, Richard M. and Keith E.
Evans (Proceedings Compilers).
1979. Management of north central
and northeastern forests for
nongame birds. USDA Forest
Service General Technical Report

NC-51. North Central Forest Ex-
periment Station, St. Paul, Minn.
268 p.

Degraaf, Richard M. and Nancy G.
Tilghman (Proceedings Compil-
ers). 1980. Workshop proceedings:
Management of western forests
and grasslands for nongame birds.
USDA Forest Service General
Technical Report INT-86. Inter-
mountain Forest and Range Ex-
periment Station, Ogden, Utah.
535 p.

Finch, Deborah M., A. Lauren Ward,
and Robert H. Hamre. 1982. Com-
ments in defense of symposium
proceedings: response to Bart and
Anderson. Wildlife Society Bulle-
tin 10:181-183.

Johnson, R. Roy, Charles D. Ziebel,
David R. Patton, Peter F. Ffolliott,
and Robert H. Hamre (Technical
Coordinators). 1985. Riparian eco-
systems and their management:
reconciling conflicting uses. First
North American Riparian Confer-
ence. USDA Forest Service Gen-
eral Technical Report RM-120.
Rocky Mountain Forest and Range
Experiment Station, Ft. Collins,
Colo. 523 p.

Nero, Robert W., Richard J. Clark,
Richard J. Knapton, and R. H.
Hamre (Editors). 1987. Biology
and conservation of northern for-
est owls. USDA Forest Service
General Technical Report RM-142.
Rocky Mountain Forest and Range
Experiment Station, Ft. Collins,
Colo. 309 p.

Pearson, Henry A., Fred E. Smeins,
and Ronald E. Thill (Proceedings
Compilers). 1987. Ecological,
physical, and socioeconomic rela-
tionships within southern national
forests: Proceedings of the south-
ern evaluation workshop. USDA
Forest Service General Technical
Report SO-68. Southern Forest Ex-
periment Station, New Orleans,
Louisiana. 293 p.

Scott, J. Michael and C. John Ralph.
1988. Quality control of symposia
and their published proceedings.
Wildlife Society Bulletin 16:68-74.

2

Scott, Norman J., Jr. 1982. Herpeto-
logical communities. USDI Fish
and Wildlife Service, Wildlife Re-
search Report 13. 239 p.

Smith, Dixie R. (Technical Coordina-
tor). 1975. Proceedings of the sym-
posium on management of forest
and range habitats for nongame
birds. USDA Forest Service Gen-
eral Technical Report WO-1.
Washington, D.C. 343 p.

Suffling, Roger, Catherine Lihou, and
Yvette Morand. 1988. Control of
landscape diversity by cata-
strophic disturbance: a theory and
a case study in a Canadian Boreal
Forest. Environmental Manage-
ment 12:73-78.

Wilson, E. O. (Editor) and Frances M.
Peter (Associate Editor). 1988. Bio-
diversity. National Academy
Press, Washington, D.C. 521 p.

■-. ;..

3

The Management of
Amphibians, Reptiles and
Small Mammals in North
America: The Need for an
Environmental Attitude
Adjustment^

Abstract.— Amphibians, reptiles, and smail
mammals need special consideration in
environmental management and conservation
because (1) they are significant biotic components
in terrestrial and freshwater habitats; (2) research
and management efforts have lagged behind those
on other vertebrates; (3) a stronger understanding
of their ecology and life history is needed to guide
management decisions; and (4) their importance
has not been promoted satisfactorily to develop the
proper public attitude.

J. Whitfield Gibbons^

My objective is to provide an over-
view and perspective of the amphibi-
ans, reptiles, and small mammals of
North America as a group that de-
serves more careful consideration
from an environmental management
and conservation standpoint. The
justification of the need for and time-
liness of a careful examination of am-
phibian, reptile, and small mammal
assemblages is based on the premises
stated below. One intent is to bring
the problem into focus so that both
scientists and managers can identify
problem areas and conjoin in an ef-
fort that will result in the manage-
ment of these animals in North
America in a prudent and far-sighted
manner.

I offer four premises to support
the contention that amphibians, rep-
tiles, and small mammals deserve
special attention with regard to man-
agement considerations:

1. Amphibians, rephles, and
small mammals are a signifi-
cant and important wildlife
component of the fauna in
most terrestrial and freshwa-
ter habitats in North Amer-
ica.

'Paper presented at symposium. Man-
agement of Amphibians, Reptiles, and
Small Mammals in North America. (Flag-
staff, AZ, July 19-21,1988).

'J. Whitfield Gibbons, Head, Division of
Stress and Wildlife Ecology, Savannah River
Ecology Laboratory, Drawer E, Aiken, SC
29801.

2. Research and management
publication efforts as well as
funding have lagged behind
those of many of the more
obvious faunal components
(e.g., game species of large
mammals, birds and fishes,
and many insects, because of
their importance as pests).

3. The direct empirical meas-
urements of habitat require-
ments, species interactions,
and life history patterns
needed for proper manage-
ment are often lacking for
amphibians, reptiles, and
small mammals.

4. An attitude that amphibians,
reptiles, and small mammals
should be of concern in envi-
ronmental management deci-
sions has not been satisfacto-
rily instilled among some
managers, the general public,
and political officials.

Support for Premises

Premise 1 —Amphibians, reptiles,
and small mammals are a
significant and important wildlife
component in North American
ecosystems.

One way for a taxonomic group or
species assemblage to qualify as im-

portant to an environmental manager
is to be identified as making a major
contribution to biological complexity
in terms of species diversity, trophic
dynamics, and interactions within
communities. Some groups clearly
have the potential for overall com-
munity influence by virtue of abun-
dance. Salamanders at Hubbard
Brook were demonstrated to have a
higher biomass than other vertebrate
groups (Burton and Likens 1975).
The capture of as many as 88,000
amphibians in one year (SREL Report
1980) and large numbers in most
years (Pechmann et al. 1988) at a
1 ha temporary pond in South Caro-
lina suggest that they dominate the
higher trophic level in some habitats.
Other studies support the postula-
tion that amphibians are often the
top predators in some aquatic sys-
tems (Taylor et al. in press). Freshwa-
ter turtles represent the majority of
vertebrate biomass in many aquatic
habitats (Congdon, Greene, and Gib-
bons 1986), and their potential sig-
nificance as vectors for seeds and
parasites among temporary aquatic
habitats has been suggested (Cong-
don and Gibbons 1988). Box turtles
(Terrapene Carolina) have also been
implicated as seed vectors (Braun
and Brooks 1987). Small rodents are
noted for their impact on plant com-
munities under certain environ-
mental conditions (Hayward and
Phillipson 1979); desert granivores
affect the density, biomass, and com-
position of annual plants (Brown et

4

Table 1 .—Publications on different taxonomic groups in major North American
journals in general ecology and wildlife ecology. Issues from 1983-1988 were
selected at random until 200 titles were chosen. Assignment to taxonomic
categories was based on the appearance of study organism names in the
titles. Not all papers used in tabulation were based on North American fauna.
The definition of small mammals is that used in this Symposium.

JOURNAL
(total)

A

R

S

L

F

B

1

(218) AMN

9

12

43

14

43

18

61

(201) ECOL

10

22

28

10

24

50

58

(213) CJZ

8

9

36

39

34

53

34

(614) Total

27

42

107

63

101

121

153

%

4

7

17

10

16

20

25

(139) Hsr

1

4

4

13

50

67

(204) JWM

0

2

6

103

93

(343) Total

1

6

110

126

50

160

%

<1

2

3

37

15

47

A= Amphibians

R = Reptiles

S = Small mammals

L = Large mammals

F = Fishes

B = Birds

I = Insects

General Ecology

AMN = American Midland Naturalist
ECOL = Ecology

CJZ = Canadian Journal of Zoology
Wildlife Ecology

HSI = U.S. Fish and Wildlife Service Habitat

Suitability Index Models
JWM = Journal of Wildlife Management

*Only 139 fifles were available.

al. 1986). These represent only a few
of the available examples for am-
phibians, reptiles, and small mam-
mals; however, many more studies
are needed that document the role
and importance of species in these
groups in enhancing biological com-
plexity.

Another way for a group to as-
sume importance is for it to have a
direct, measurable economic value or
impact. Several examples can be
given of the importance of amphibi-
ans, reptiles, and small mammals
from the economic perspective, but
their impact has been trivial in com-
parison to large game mammals or
insect pests, and controls and regula-
tions have been comparatively loose.
The limited economic importance of
most small terrestrial or semi-aquatic
vertebrates is presumably one expla-
nation for their being given minimal
attention in many management
schemes. A few species such as
American alligators (Joanen and
McNease 1987), bullfrogs (Shifter
1987), and snapping turtles (Bushey,
no date) are commercially important
as human food items. Other species
assume an economic value in the le-
gal pet trade (Conant 1975) or as re-
search animals sold by biological
supply houses (Carolina Biological
Supply 1987). Some venomous
snakes, especially eastern (Crotalus
adamanteus) and western (C. atrox)
diamondback rattlesnakes, are an
economic irony in that the venom is
necessary to make antivenin (Parrish
1980). Of course, such species
achieve some level of importance
simply by being potentially injurious.
Small mammals have been indicted
in a variety of situations for negative
economic impacts, such as prairie
dog damage (Walker 1983), rabies in
bats (Constantine 1970), and grain-
eating by rodents (Rowe 1981).

Another measure of importance of
some species or groups is the intan-
gible aesthetic value that some
people place on them. Many species
assume an undeniable importance to
many people and may ultimately ac-

quire protected status. Legal protec-
tion of "the species" often provides
protection to certain habitats. This
circle of protection is a factor that can
work to great advantage for those
persons interested in preservation —
the species is protected because it is
important (aesthetic) and becomes
even more important (legal) because
it is protected and results in preser-
vation of the habitat. For example,
the legal status offered the desert tor-
toise {Xerobates agassizi; Luckenbach
1982) and the Morro Bay kangaroo
rat (Dipodomys heermanni morroensis;
USDI 1980) in California or the
American crocodile (Crocodylus
acutus; Kushlan and Mazzotti 1986)
in Rorida serves to provide some
level of environmental protection for
the entire community where they oc-
cur. The protection given the black
footed ferret has resulted in protec-
tion of its prey. The World Wildlife

Fund recognizes this effect in its con-
servation programs by designating
"flagship" species such as great apes
or monkeys, for which funds are
more easily raised, in order to pro-
tect entire communities or ecosys-
tems.

Premise 2— Ecological research on
herpetofouna and small
mammals has lagged behind that
of other animal groups.

Support for the contention that the
level of ecological research on am-
phibians, reptiles, and small mam-
mals is lower than that of certain
other animal groups can be given in
several ways. These include annual
publications on particular groups
(table 1) and the proportion of
funded grants that fall into each cate-
gory (table 2).

5

The reasons for the lower levels of
publication and funding in research
on amphibians, reptiles, and small
mammals are varied and in part con-
jectural. One seemingly obvious rea-
son is that most species in these
groups have low profile in health,
hunting, agricultural, or other eco-
nomic issues and therefore receive
minimal attention from some quar-
ters. The comparatively low level of
attention given to small, non-game
terrestrial and semi-aquatic verte-
brates by certain sectors of society is
reflected in lower overall funding
and subsequently in fewer general
publications.

Research funding is inequitable
because of the emphasis on species
that have important economic status;
thus, the life history and ecology of
even moderately abundant herpe-
tofaunal or small mammal species
are seldom understood at a level that
would permit prudent management.
Even those with potential economic
importance receive less emphasis
than many birds, large mammals,
and fish. As an example, the Ameri-
can alligator represents a reptile spe-
cies of vital concern from a manage-

ment standpoint, yet the number of
publications that focus on the life his-
tory, ecology, behavior, and genetics
of the sp)ecies is limited (see Brisbin
et al. 1985) compared to the hun-
dreds on large mammal game species
such as white-tailed deer (Halls 1984;
Johns and Smith 1985).

Premise 3— The basic ecological
and life history information
necessary to make thoughtful
environmental management
decisions is often absent for many
of the amphibians, reptiles, and
small mammals in a community.

As indicated above, the research ef-
fort directed toward amphibians,
reptiles, and small mammals by
ecologists appears to be below that
for other vertebrate groups. Al-
though difficult to measure, it would
also be expected that the fundamen-
tal data bases necessary for thought-
ful management decisions would ex-
ist in lower proportions for herpe-
tofaunal and small mammal species.
One reason is that, compared to
many large mammals, birds, and

Table 2.— Number of grant proposals funded by selected U.S. granting agen-
cies on particular groups of animals.

A

R

S

L

F

B

1

NSF(1987)

1

3

4

2

11

7

24

Sigma Xi

4

8

9

7

9

24

10

(March 1987)

National Geograpl-iic

0

3

2

28

8

15

14

(1988)

World Wildlife Fund

0

15

0

49

0

14

1

(1987-1988)

Total

5

29

15

86

28

60

49

% 2

11

6

32

10

22

18

A=Amphibians

R=Reptiles

S=Small mammals

L=Large mammals

F=Fishes

B=Birds

l=lnsects

fishes, certain aspects of field studies
on many of the amphibians, reptiles,
and small mammals are sometimes
perceived as being more difficult be-
cause of factors such as small body
or population sizes, fossorial or cryp-
tic habits, patchy distribution, and
unpredictable seasonality. Conse-
quently, fewer papers are likely to be
published in general ecology journals
that expect quantitative ecological
and life history research results
rather than ones that are descriptive
and qualitative. An exception to this
may be manipulative field experi-
ments in which small rodents have
been used in almost half of the stud-
ies involving vertebrates.

The actual or apparent rarity or
unpredictability of occurrence of
many amphibian, reptile, and small
mammal species makes it difficult or
impossible for the research ecologist
to gather useful data without a fund-
ing base that is accepting of the un-
certainty of whether data will actu-
ally be forthcoming in a particular
year. The environmental manager in
turn cannot incorporate such species
into a management plan, and thus
their perceived imp>ortance is dimin-
ished. The unpredictability of occur-
rence of some species can be demon-
strated with amphibians and reptiles
on the Savannah River Plant (SRP) in
South Carolina. In spite of more than
a quarter of a century of field studies
and the capture of more than half a
million reptiles and amphibians
across all available habitats, species
previously unreported from the SRP
continue to be discovered (Gibbons
and Semlitsch 1988; Young 1988). Or,
some species have gone for intervals
as long as one decade (e.g., pickerel
frog. Ram palustris) or two (e.g.,
glossy water snake, Regina rigida) be-
tween sightings (Gibbons and Sem-
litsch 1988). Clearly, developing a
basic ecological field study on such
species in a region is not feasible un-
der typical funding situations.

Resolutions to the problem of gar-
nering information about rare sp>ecies
include intensifying survey efforts in

6

geographic regions of interest by
supporting long-term research pro-
grams that can ultimately reveal the
presence of rare or fossorial species.
Once a species is identified to be
present in a habitat, the decision
should be made on whether an eco-
logical research effort is warranted.

Long-term studies may be neces-
sary to reveal certain life history
traits, even about common species,
because of the inherent variability in
some life history features that can
result from natural environmental
variation (Semhtsch et al. 1988). Such
studies may be essential to identify
the extent of variability due to an-
nual weather patterns and climatic
variation (Semlitsch 1985; Pechmann
et al. 1988). Long-term research pro-
grams may be needed because some
species are long- lived, or in the case
of many, because the potential lon-
gevity is great but unknown (Gib-
bons 1987).

For many species that have eco-
nomic value (e.g., snapping turtle,
Chelydra serpentina; Congdon et al.
1987), the impact of harvesting has
not been properly assessed. Because
of the limited baseline ecological and
life history data for most species, a
priority goal should be the establish-
ment of a moratorium on the whole-
sale removal of all native species of
amphibians, reptiles, and small
mammals until it can be verified that
regional populations can sustain the
removal rate. State permits should be
required of, and possession limits
should be set for, all commercial col-
lectors for all species of amphibians,
reptiles, and small mammals.

Today's emphasis should be on
protection of each species until con-
vincing evidence is supplied that har-
vesting has no long-term impact,
rather than placing the burden on
herpetologists and mammalogists to
demonstrate population irrecovera-
bility before harvesting is discontin-
ued. The negative consequences of
the latter, and current, approach (i.e.,
demonstrating the impact of removal
while harvesting is in progress) is

that some populations will be re-
duced to the point of no recovery be-
fore the necessary evidence can be
collected. Each species should be
protected until proven harvestable.
The appropriate basic research
should be conducted by scientists
with no economic or emotional in-
vestment in the outcome. Research
support should be provided by state
or federal agencies and by special
interest groups that have no influ-
ence over the final management deci-
sions. The ideal approach is that sci-
entists would gather the facts and
that environmental managers would
interpret them in the context of har-
vesting quotas. The development
and use of predation (Holling 1966)
or harvest (Ricker 1975) models may
be effective approaches for address-
ing the issue of human predation
(i.e., harvestability by man).

One area that deserves attention in
strengthening the study of small ter-
restrial or semi-aquatic vertebrates is
the use of innovative techniques to
address physiological, ecological,
and behavioral questions under natu-
ral conditions. Non-destructive field
sampling techniques are critical in
the study of both rare and endan-
gered species but are also important
for preserving the integrity of any
study population. These include
techniques for capture, field identifi-
cation of individuals, non-disruptive
handling or observation, recapture,
and the acquisition of non- destruc-
tive physiological, genetic, behav-
ioral, and life history data. Some ex-
amples include radiography (Gib-
bons and Greene 1978) or sonogra-
phy for determination of clutch sizes,
blood sampling for genetic and
hormonal analyses (Scribner et al.
1986), and cyclopropane for measur-
ing lipid levels (Peterson 1988). A
broader use of such techniques in
field studies could strengthen the
foundation of ecological and life his-
tory understanding that is necessary
for environmental management.

A direct contribution to environ-
mental managers could be achieved

by attempts to verify the several am-
phibian, reptile, and small mammal
Habitat Suitability Index models of
the U.S. Fish and Wildlife Depart-
ment. The concept has the potential
value of providing an initial quanti-
tative approach that gives a tangible
product. However, to be of greatest
value, the HSI models must be evalu-
ated and modified as appropriate. It
is perhaps noteworthy that the HSI
models prepared for amphibians (1),
reptiles (4), and small mammals (4)
collectively represent only 6% of the
139 that have been completed on ver-
tebrates (table 1). For these to be-
come an effective tool in manage-
ment of herpetofauna and small
mammals, more herpetologists and
mammalogists need to volunteer to
develop HSI models for these
groups.

A distinction must be made be-
tween (1) problem oriented applied
research on specific systems that re-
lies on qualitative assessments or in-
direct measurements of variables
with minimal inference power and
(2) basic research that is founded on
quantitative or direct measurements
of variables, has a conceptual or
theoretical base or orientation, and
can be strongly inferential through
general field or laboratory experi-
ments. The latter approach will be
necessary if environmental managers
are to have a reliable data base that is
founded on broad applicability, lev-
els of predictability, and clear direc-
tions for future research.

Premise 4— The attitude of most
people in North America toward
most amphibians, reptiles, and
small mammals is either negative
or neutral, in part because efforts
to develop an attitude change
have been insufficient or
ineffective.

Although documentation is difficult,
it would appear that in North Amer-
ica we are far from a suitable accep-
tance level toward these groups of

7

organisms. People still try to run
over snakes on highways, have little
awareness that many conspicuous
predators rely on small mammals for
their basic diets, and give no thought
to how many small vertebrates will
be eliminated by the draining of a
swamp or damming of a stream. I
think the situation is an embarrassing
one for the scientists and general
public of a nation that espouses edu-
cation and knowledge.

Evidence that a more positive atti-
tude and less environmental leniency
has developed over the last several
years is the recent federal listings of
snakes (e.g., indigo snake, Dry-
marchon corias; San Francisco garter
snake, Thamnophis sirtalis tetrataenia)
and small rodents (e.g., Utah prairie
dog, Cynomys parvidens; salt marsh
harvest mouse, Reithrodontomys
raviventris; Key Largo cotton mouse,
Peromyscus gossypinus allapaticola) as
protected species. However, many of
the listings involving amphibians,
reptiles, and small mammals have
been hard fought ones against public
and political opinions that such spe-
cies hardly deserve such concessions.
The failed efforts at protection far
outnumber the successful ones. The
attitude that these animals are unim-
portant is pervasive throughout the
general public, politicians, and even
some environmental managers. The
basic responsibility for eliminating
ignorance and effecting the proper
environmental attitude adjustment
must start with the scientist.

It is my firm opinion that many
scientists have lost sight of who their
patrons are (for most of us, the U.S.
taxpayers) and of their responsibility
to communicate findings to all levels
of society. This communication proc-
ess entails a level of cooperation and
an educational spirit that allows each
individual to contribute in the most
effective manner. However, we must
all accept and work toward the com-
mon goals of establishing a thorough
and general foundation of ecological
information for amphibians, reptiles,
and small mammals and of being

generous in the distribution of the
findings in a form palatable to and
usable by the intended audience.

Conclusions

An environmental attitude adjust-
ment model must be developed and
promoted that considers where we
want to end up, who we must edu-
cate and influence, and what we
must know and do to achieve the
goal of education in a convincing
manner. The desired end point is a
nationwide attitude among scientists,
managers, politicians, and the public
that amphibians, reptiles, and small
mammals are critical wildlife compo-
nents. Each species population and
community must be identified as
having an intrinsic value in maintain-
ing the integrity of the natural eco-
systems of North America.

Scientists have a responsibility for
collecting extensive and intensive in-
formation on the life history patterns
and habitat requirements of native
amphibians, reptiles, and small
mammals. The required data must be
collected in a rigorous experimental
manner that promotes an under-
standing of these species and com-
munities through strong inferences
and syntheses.

Politicians have a responsibility to
assure that the approval of a govern-
ment project is as contingent on envi-
ronmental consequences as on budg-
etary considerations. Our attitude
must graduate to become one of ac-
ceptance of a proposed project only
after environmental impact determi-
nations have led to an objective deci-
sion that the gain from the project
warrants the loss to the environment.

Managers have a responsibility for
promoting basic research, for apply-
ing the findings to habitat manage-
ment, and for having the patience to
wait for the completion of long-term
studies as required. In situations
where removal of animals or elimina-
tion of habitat is an issue, the burden
of proof should be borne by the har-

vester or developer, and not by the
scientist or manager. The status of a
species should be determined before
the decision to proceed is made, cer-
tainly not after harvesting begins or
during the physical development of a
project. This assessment should be
made and evaluated before the proj-
ect is approved. Each species should
be protected until proven harves-
table.

Both scientists and managers have
a responsibility to inform the public
and political arena that the protec-
tion and ecological understanding of
inconspicuous and non-game species
are vital to proper ecosystem man-
agement and to the preservation and
maintenance of North America's
natural heritage.

Acknowledgments

I thank Justin D. Congdon, Nat B.
Frazer, Trip Lamb, William D.
McCort, Joseph H. K. Pechmann,
David E. Scott and Raymond D.
Semlitsch for commenting on the
original manuscript. I appreciate the
efforts of Marianne Reneau, Marie
Fulmer, Jeff Lovich, Tony Mills, and
Tim Owens in manuscript prepara-
tion. Manuscript preparation was
aided by Contract DE- AC09-
76SR00819 between the U.S. Depart-
ment of Energy and the University of
Georgia's Savannah River Ecology
Laboratory.

Literature Cited

Brisbin, I. Lehr, Charles A. Ross,
M.C. Downes, Mark A. Staton and
Brad R. Gammon. 1986. A bibliog-
raphy of the American alligator
(Alligator mississippiensis). Savan-
nah River National Environmental
Research Park Publ. Aiken, SC.

Brown, J. H., D. W. Davidson, J. C.
Munger, and R. S. Inouye. 1986.
Experimental community ecology:
The desert granivore system,
pp. 41-61. In: Community Ecol-

8

ogy. J. Diamond and T. J. Case,
editors. Harper and Row. NY.

Brown, J. and G. R. Brooks, Jr. 1987.
Box turtles (Terrapene Carolina) as
potential agents for seed dispersal.
Am. Midi. Nat. 117:312-318.

Burton, T.M. and G.E. Likens. 1975.
Salamander populations and bio-
mass in the Hubbard Brook Ex-
perimental Forest, New Hamp-
shire. Copeia 1975:541-546.

Bushley, S. Date of publication un-
known. How to profit from snap-
ping turtles. Reptiles Unlimited,
York, Pennsylvania.

Carolina Biological Supply Com-
pany. 1987. Biology /Science mate-
rials. Catalog 58. Burlington, NC.

Conant, Roger. 1975. A field guide to
reptiles and amphibians of east-
ern/central North America.
Houghton-Mifflin: NY.

Congdon, Justin D. and J. Whitfield
Gibbons. 1988. Biomass productiv-
ity of turtles in freshwater wet-
lands: A geographic comparison.
In Freshwater Wetlands and Wild-
life. Rebecca R. Sharitz and J.
Whitfield Gibbons, editors. Office
of Scientific and Technical Infor-
mation, U.S. Department of En-
ergy. Oak Ridge, TN.

Congdon, Justin D., Gary L. Breiten-
bach and Richard C. van Loben
Sels. 1987. Reproduction and nest-
ing ecology of snapping turtles
(Chelydra serpentina) in southeast-
ern Michigan. Herpetologica
43:39-54.

Congdon, Justin D., Judith L. Greene,
and J. Whitfield Gibbons. 1986.
Biomass of freshwater turtles: A
geographic comparison. Am.
Midi. Nat. 115:165-173.

Constantine, D. G. 1970. Bats in rela-
tion to the health, welfare, and
economy of man. pp. 320-449. In:
W. A. Wimsatt, editor. Biology of
bats. Vol. II. Academic Press. New
York.

Gibbons, J. Whitfield. 1987. Why do
turtles live so long? Bioscience
37:262-269.

Gibbons, J. Whitfield and Judith L.
Greene. 1979. X-ray photography:

A technique to determine repro-
ductive patterns of freshwater
turtles. Herpetologica 35:86-89.

Gibbons, J. Whitfield and Raymond
D. Semlitsch. 1988. Guide to the
reptiles and amphibians of the
Savannah River Plant. SRO-NERP
18. U.S. Department of Energy's
Savannah River National Environ-
mental Research Park. Aiken, SC.

Halls, L. K. 1984. White-tailed deer:
Ecology and management. Stack-
pole Books. Harrisburg, PA.

Hayward, G. F. and J. Phillipson.
1979. Community structure and
functional role of small mammals
in ecosystems, pp. 135-211. In:
Ecology of small mammals. D. M.
Stoddart, editor. John Wiley and
Sons. New York.

Holling, C.S. 1966. The functional re-
sponse of invertebrate predators
to prey density. Mem. Entomol.
Soc. Can. 48:1-86.

Joanen, Ted and Larry McNease.
1987. The management of alliga-
tors in Louisiana, USA. p. 33-42.
In: Wildlife Management. Croco-
diles and Alligators. Grahame J.W.
Webb, S. Charlie Manolis and Pe-
ter J. Whitehead, editors. Surrey
Beatty & Sons Pty. Limited. Chip-
ping Norton, NSW, Australia.

Johns, Paul E. and Michael H. Smith.
1985. Bibliography for the white-
tailed deer on the Savannah River
Plant. University of Georgia, Insti-
tute of Ecology and Savannah
River Ecology Laboratory. Aiken,
SC.

Kushlan, James A. and Frank J. Maz-
zotti. 1986. Population biology and
status of the American crocodile in
South Florida, pp. 188-194. In:
Proceedings of the 7th working
meeting of the crocodile specialist
group. lUCN. Caracas, Venezuela.

Luckenbach, R. A. 1982. Ecology and
management of the desert tortoise
(Gopherus agassizii) in California,
pp. 1-37. In: R. B. Bury (ed.). North
American tortoises: Conservation
and Ecology. USDI, Fish and
Wildlife Service. Wildlife Resource
Department. 12.

Parrish, Henry M. 1980. Poisonous
snakebites in the United States.
Vantage Press, Inc. New York.

Pechmann, Joseph H.K., David E.
Scott, J. Whitfield Gibbons and
Raymond D. Semlitsch. 1988. In-
fluence of wetland hydroperiod on
diversity and abundance of meta-
morphosing juvenile amphibians.
Wetlands Ecology and Manage-
ment 1:1-9.

Peterson, C. 1988. Presentation at
Desert Tortoise Council Meeting,
Laughlin, NV.

Ricker, N.E. 1975. Computation and
interpretation of biological statis-
tics of fish populations. Bull. Fish.
Res. Board Can. 191:1-382.

Rowe, P.P. 1981. Wild house mouse
biology and control. Symposium
of the Zoological Society of Lon-
don. 47:575-589.

Scribner, Kim T., Joseph E. Evans,
Stephen J. Morreale, Michael H.
Smith and J. Whitfield Gibbons.
1986. Genetic divergence among
populations of the yellow-bellied
slider turtle (Pseudemys scripta)
separated by aquatic and terres-
trial habitats. Copeia 1986:691-700.

Semlitsch, Raymond D. 1985. Analy-
sis of climatic factors influencing
migrations of the salamander Am-
bystoma talpoideum. Copeia
1985:477-489.

Semlitsch, Raymond D., David E.
Scott and Joseph H.K. Pechmann.
1988. Time and size at metamor-
phosis related to adult fitness in
Ambystoma talpoideum. Ecology
69:184-192.

Shiffer, Clark N. 1987. The bullfrog.
Pennsylvania Angler. August
1987.

SREL Report. 1980. Savannah River
Ecology Laboratory' Annual Re-
port, FY-1980. A biological inven-
tory of the proposed site of the
Defense Waste Processing Facility
on the Savannah River Plant in
Aiken, South Carolina. NTIS Publ.
SREL-7UC-66e. Aiken, SC.

Taylor, Barbara E., Ruth A. Estes, Jo-
seph H. K. Pechmann and Ray-
mond D. Semlitsch. Trophic rela-

9

tions in a temporary pond: Larval
salamanders and their microin-
vertebrate prey. Can. J. Zool. In
Press.

USDI (United States Department of
the Interior). 1980. Republication
of the lists of endangered and
threatened species and correction
of technical errors in final rules.
Federal Register. 45:33768-33781.

Walker, E. P. 1983. Mammals of the
world. Vol. I. John Hopkins Uni-
versity Press, Baltimore. 568 p.

Young, David P. 1988. Rhadinaea flav-
ilata (Pine woods snake). Herp.
Rev. 19:20.

Douglas-fir Forests in tlie
Oregon and Washington
Cascades: Relation of the
Herpetofauna to Stand Age
and Moisture^

R. Bruce Bury^ and Paul Stephen Corn^

Abstract.— Pitfall traps effectively sampled
amphibians but not reptiles in Douglas-fir
(Pseudofsugo menz/es/O forests. The abundance of
only one amphibian species varied across an age
gradient or a moisture gradient. Salamanders and
frogs that breed in ponds or streams vjere captured
in large numbers in some stands, likely due to the
presence of nearby breeding habitat rather than
forest conditions. Lizards occurred mostly in dry
stands and clearcuts. Time-constrained searches
showed different use of dov^ned v\/oody debris
among terrestrial salamanders. The occurrence and
abundance of species in naturally regenerated
forests markedly differed from clearcut stands.

The value of old-growth forests for
wildlife is highly debated (Fosburg
1986, Harmon et al. 1986, Harris
1984, Kerrick et al. 1984, Ruggiero
and Carey 1984, Salwasser 1987,
Wilcove 1987). Most attention has
been directed toward the spotted owl
(Strix occidentalis), which is one of
several hundred vertebrate species
occurring in the Pacific Northwest
(Bruce et al. 1985). Franklin and Spies
(1984) distinguished old-growth for-
ests of Douglas-fir (Pseudotsuga men-
ziesii) as having a wide range of tree
sizes and ages, a deep mulhlayered
crown canopy, large individual trees,
and accumulations of coarse woody
debris (CWD), including snags and
downed logs of large dimension.
They reported that these forests are
productive, diverse ecosystems, and
highly specialized habitats.

We need to evaluate sampling
techniques continually to better de-
scribe, understand and predict the
species richness, abundance and bio-
mass of herpetological assemblages.
However, few herpetological com-
munities or their habitats have been

^ Roper presented at symposium, Mon-
agement of AmphibioDS. Reptiles, and
Smoll Mammals in North America (Flagstaff,
AZ.July 17-21, 1988).

Bruce Bury is Zoologist (Research)),
USDA Fishi and Wildlife Sen/ice, National
Ecology Research Center, 1300 Blue Spruce
Drive, Fort Collins, CO 80524.

^Paul Stephen Corn is Zoologist, USDA
Fish and Wildlife Service, National Ecology
Research Center, 1300 Blue Spruce Drive,
Fort Collins. CO 80524.

sampled using more than one quanti-
tative technique.

Recently, field techniques for the
study of herpetological communities
have improved (Scott 1982). Some of
the most promising methods employ
pitfall traps and drift fences to cap-
ture amphibians and reptiles. Several
promising pitfall designs have been
developed for varied habitats in Aus-
tralia (Friend 1984, Webb 1985) and
in North America (Bennett et al.
1980, Bury and Corn 1987, Bury and
Raphael 1983, Campbell and Christ-
man 1982, Enge and Marion 1986,
Gibbons and Semlitsch 1981, Jones
1981, 1986, Raphael 1984, Raphael
and Rosenberg 1983, Rosenberg and
Raphael 1986, Vogt and Hine 1982).
Pitfall traps are effective for capture
of commmon terrestrial species and
they are particularly valuable in sam-
pling secretive or rare forms.

Searches by hand (either based on
specific areas or time of collecting) or
observation are used to sample her-
petofaunas (see reviews by Bury and
Raphael 1983, Jones 1986, Rough et
al. 1987). Campbell and Christman
(1982) suggested that time-con-
strained collecting (searching within
a specific period of time by trained
collectors) can sample terrestrial spe-
cies that are under-sampled or not
taken in pitfall traps.

The first year of our old-growth
study (1983) was partly devoted to
refining field techniques. A compari-
son of different pitfall designs is re-
ported elsewhere (Bury and Corn

1987). Here, we employ a standard-
ized pitfall array and time-con-
strained searches to determine the
occurrence and abundance of the ter-
restrial (upland) herpetofauna in the
Cascade Mountains of the Pacific
Northwest.

The current work on small mam-
mals (Anthony et al. 1987, Com et al.
1988, West 1985), birds (Carey 1988,
Manuwal and Huff 1987), and bats
(Thomas in press) are part of an inter-
disciplinary effort to better under-
stand the relationship of nongame
wildlife in old-growth forest stands
(Ruggiero and Carey 1984). Our
study is the first to attempt to iden-
tify which species of the herpe-
tofauna, if any, are associated with
age and moisture gradients in forests
of the Cascade Mountains.

Our specific objectives were (1) to
compare effectiveness and relative
merits of time-constrained collecting
versus pitfall trapping, (2) to com-
pare the species richness and relative
abundance of amphibians and rep-
tiles between different forest stands,
and (3) to examine the association of
the herpetofauna with old-growth
forest conditions.

DESCRIPTION AND
CLASSIFICATION OF STUDY SITES

We sampled 30 sites: 18 in or near
the H. J. Andrews Experimental For-
est in eastern Linn and Lane coun-
ties, Oregon, and 12 stands in the

11

Figure 1 .—Conducting time-constrained searches in an old-growtti stand, Oregon. Note
large amounts of downed woody debris.

Wind River Experimental Forest,
Skamania County, Washington. All
sites are on the western slopes of the
Cascade Mountains. Specific loca-
tions, stand classification, elevations
and other details are provided in
Corn et al. (this volume).

Study sites represent a range of
forest development across a
chronosequence (principally age)
and, for old-growth, a moisture gra-
dient. These stands were independ-
ently selected and assessed by Spies
et al. {in press). They were all in natu-
rally regenerated forest caused by
wildfire. There were three develop-
ment stages in moderate moisture
conditions: young (30-76 years old),
mature (105-150 years) and old-
growth (195-450 years). Clearcut sites
represent recent timber harvest (<10
years old). For old-growth stands
only, there were representative mois-
ture conditions: wet, moderate and
dry sites. Stand classification was
based on age determined by incre-
ment boring of trees or other meth-
ods, characteristic plant species in
the understory, physiography, and
soils. These methods and other para-
meters are described by Corn et al.
(this volume). Franklin et al. (1981)
and Spies et al. {in press).

Following the initial stand selec-
tion, there were minor adjustments
in assignment of stand classification
(Corn et al., this volume). We re-
jected a few sites that were either not
continually accessible for our weekly
checking of pitfall traps or were
being actively logged.

MATERIAL AND METHODS

Time-Constrained Searches (TCS)

Details of this technique are pro-
vided elsewhere (Campbell and
Christman 1982, Bury and Raphael
1983, Raphael and Rosenberg 1983).
A team of 3-8 people intensively
searched each stand for 8 person-hrs
in the spring (8-25 April 1983 in Ore-
gon and 3-12 May 1983 in Washing-

ton). We turned over moveable sur-
face objects (twigs to logs <1 m dia-
mater), dug into decayed wood, and
removed bark from downed wood or
the bases of standing snags by hand
or with potato rakes (fig. 1).

Collectors remained within
boundaries of habitat typical of the
stand, avoiding conspicuous special-
ized habitats such as ponds, creeks
or rock outcrops. Further, we
searched 4 sites in each state again
during warm weather (July- Aug
1983). These surveys were performed
for 4 hrs per plot. We recorded infor-
mation on exact position of capture
for each animal, including vertical
position (e.g., on or under litter; on,
under or in log; etc.), identification of
cover object, length and diameter of
object, time of capture, total length,
and mass of animal.

We determined the decay class of
coarse woody debris occupied by
animals on the forest floor. Large
woody debris or felled trees (logs)
occur in five progressive broad decay
classes (Bartels et al. 1985, Franklin et
al. 1981, Harmon et al. 1986, Maser et
al. 1979, Maser and Trappe 1984): (1)

intact, recently downed trees; (2) logs
with loose bark; (3) loss of bark and
stem partly rotted; (4) invasion of
roots and deep decomposition of
stem; and (5) hummocks of wood
chunks and organic material. Once
fallen, a large tree might require 200
or more years to progress from class
1 to 5 (Spies et al. in press), providing
habitat for many generations of resi-
dent wildlife.

Pitfall Arrays

We installed a pitfall array at each
site in Oregon and Washington (de-
tails in Bury and Corn 1987). Each
array had two triads with their cen-
ters 25 m apart. Each triad was com-
posed of three drift fences 5 m long
and 0.5 m tall; about 0.3 m of fence
was above ground. Fences radiated
at 120° angles, beginning 3 m from
the center point. The compass direc-
tions of the arms depended on open-
ings between trees or large logs on
the forest floor. Pitfall traps were
constructed from two stacked #10 tin
cans (3.2 1 volume) connected with

12

Table 1 .—Numbers of amphibians and reptiles captured during time-constrained searches (TCS) corKiucted 8-25
April 1983 at the H. J. Andrews Experimental Forest In Oregon. Old-growth stands are arranged in order of increasing
dryness.

Old growth

Wet

Moderate

Dry

Mature

Young

Clearcut

Species Stand No.

15

03

24

«02

17

33 25

29

11

35

42

39 47

48

75

55

291

391

Clouded Salamander

3

8

6

9

3

11

17

4

2

1

2

12

2

Oregon Slender Salamander

2

6

4

12

9

n

5

1

9

1

1

Oregon Ensatina

4

3

1

9

5

7 22

2

10

6

4

5 3

9

8

9

4

1

Dunn's Salamander

2

1

Rough-skinned Newt

2

1

1

Pacific Tree Frog

]

4

1

1

1

Western Skink
Norhtern Alligator Lizard
Western Fence Lizard

°Two surveys were conducted in this stand and the results are combined here.

duct tape. A pit trap was placed
fl'jsh with the ground surface at each
end of the fence. Funnel traps were
constructed of aluminum screening,
rolled into a tube 1 m long by 0.1 m
diameter, with inward funnels
stapled at each end of the trap. A
funnel trap was placed midway on
either side of the fence. No water or
preservatives were added to the
traps. A wooden shingle was
propped over each pitfall and
funnnel trap, but water entered pit-

falls during heavy rains. We rou-
tinely removed water from traps
with scoops or a hand-operated aq-
uarium siphon.

We operated pitfall traps conti-
nously for 180 days, from the last
week of May to late November 1983.
Traps were checked 1-2 times each
week. Captures were usually taken
to a field laboratory for identification
and measurements. All retained
specimens are deposited at the Na-
tional Museum of Natural History.

RESULTS

Tinne-Constrained Searches (TCS)
Yield

During spring TCS, we collected 258
amphibians and 4 reptiles (table 1) at
the 18 Oregon sites (1.8 animals per
person-hr) and we took 78 amphibi-
ans and 4 reptiles (table 2) at 12
Washington sites (0.85 per person-
hr). For summer TCS, all Washington
captures included only 4 lizards from
one clearcut, one mature (drier as-
pect) and an old-growth dry stand
(0.25 animals p>er hr) whereas in Ore-
gon we captured 13 salamanders (no
new species) and 2 lizards from 4
sites (0.9 animals per hr).

Although we report the abun-
dance of herpetofauna collected by
TCS (tables 1 and 2), we did not ana-
lyze these results based on the age
and moisture gradients because such
abundance data can be biased.

Habitat Use

TCS provided useful information on
the exact position where individuals
were found (table 3). Oregon ensati-

Table 2.— Numbers of amphibians and reptiles captured during TCS 3-12
May at the Wind River Experimental Forest In Washington. Old-growth
stands are arranged in order of increasing dryness.

Old growth

Wet Moderate Dry Mature Young Clearcut

Species Stand No. 14 12 21 20 31 41 42 50 60 61 70 71

Olympic Salamander 2

Oregon Ensatina 3 7 13 5 5 4 1 1 1 1

Larch Mountain Salamander 14
Western Red-backed

Salamander 6
Rough-skinned Nev4 3 2 1

Red-legged Frog 1

Pacific Tree Frog 1
Rubber Boa 2 1

Common Garter Snake 1

13

Table 3.— Number of salamanders (Oregon data only) captured in different
microhabitats. Percentages are In parentheses.

Oregon

Oregon

Clouded

Slender

Position

Ensotina

Salamander

Salamander

On/Under Litter

3 (2.4)

0 (0)

1 (1.6)

On/Under Rock

3 (2.4)

0 (0)

1 (1.6)

On/Under Log

14(11.5)

8 (10.2)

6 (6.8)

Inside Log

52 (42.6)

27 (34.2)

38 (62.3)

Under Bark on Log

12 (9.8)

37 (46.8)

7 (11.5)

Under Bark on Ground

38 (31.1)

7 (8.9)

8 (13.1)

nas {Ensatim eschscholtzi; fig. 2) oc-
curred more evenly and in more mi-
crohabitats than did the other two
species. Clouded salamanders
(Aneides ferreus) were mostly under
bark on logs and, secondarily, often
were in logs (81% of the sites occu-
pied were related to logs). The Ore-
gon slender salamander (Batrachoseps
wrighti) predominately occurred in
logs (62%) and then under bark on
ground or on logs (87% in or near
logs). Most bark on the ground oc-
curred in piles sloughed from fallen
trees or snags and is essentially an
extension of the log environment.

Terrestrial salamanders that were
captured in or near downed wood
markedly differed in their use of dif-
ferent decay classes of CWD (fig. 3).

We did not include decay class 1
logs, because few of these were
searched and none had salamanders.
These logs are intact material and
offer little cover for salamanders.

We calculated Chi-square statistics
for three species in Oregon. The
clouded salamander was most abun-
dant in younger (class 2) logs (P
<0.001), while Oregon slender sala-
manders were found more often than
expected in the more decayed class 4
and 5 logs (P < 0.05). Numbers of
Oregon ensatina generally followed
the pattern of log abundance (fig. 3),
except that they were found less of-
ten than expected in class 3 logs (P
<0.05). These results are consistent
with microhabitats where the sala-
manders were captured (table 3).

Pitfall Trapping
Total Nunnbers

Pitfall arrays at 18 Oregon sites pro-
vided 1,028 captures (table 4): 685
salamanders, 252 frogs, 64 lizards
and 27 snakes. Pitfalls at 12 Washing-
ton sites yielded 1,152 animals (table
5): 460 salamanders, 663 frogs and 29
snakes. Two Washington sites had
exceptional catches: 253 tailed frogs
(Ascaphus truei) at #21 Old-growth
Moderate and 119 red-legged frogs
(Ram aurora) at #42 Mature.

HtL Alive FREQUCNCV

Figure 3.— Frequency of occurrence of
clouded salannanders, Oregon slender
salannanders, and Oregon ensatinos occu-
pying downed wood in decay classes 2-5.
Density of logs in each decay class are
provided. Data are from 18 sites at the H. J.
Andrews Experimental Forest, Oregon.

Yield

Summer operation of the pitfall ar-
rays added a few reptiles but the
bulk of the catch was amphibians in
the fall months during and after
heavy seasonal rains (Bury and Corn
1987). There was a low catch of rep-
tiles (Oregon, mean = 5 per site;
Washington, mean = 2.4).

Species richness did not differ
across the chronosequence gradient
(table 6, fig. 4). Moderate and dry
old-growth stands had the highest
mean abundance across the moisture
gradient, which was caused by the
capture of large numbers of several
migratory species.

Figure 2.— Adult ensatina (Ensatina eschscholtzi) from Douglas Co., Oregon.

14

r

Table 4.— Abundance of amphibians and reptiles captured by pitfall arrays at ttie H. J. Andrews Experimental Forest in
Oregon. Arrays of pitfall traps withi drift fences were operated continuously for 180 days in 1983. Old-growth stands
are arrariged in order of increasing dryness.

Old growth

Wet

Moderate

Dry

Mature

Young

Clearcut

Species

Stand r4o. 15 03 24 02 17 33 25 29 11 35 42 39 47 48 75 55 291 391

Northwestern Salamander
Pacific Giant Salamander
Clouded Salamander
Oregon Slender Salamander 1
Oregon Ensatina 8
Dunn's Salamander
Rough-skinned Newt 21
Tailed Frog
Red-legged Frog

Pacific Tree Frog 2
Western Skink
Norhtern Alligator Lizard
Western Fence Lizard
Rubber Boa

Northwestern Garter Snake 1
Common Garter Snake 1

1
3
2

28

2
1

10

3
5

18

26
3

1

22

5
4

13
1

17

26

119

3
1

27
7
4

21
1

62
46
23

3
4

1

4

28
3

11
14
5

1

2

10

16

14

20

30

12

10

1

15

13

36

5

14

16

2

6

28

30

3

2

4

1

3

3

2

5

9
8
3

1 11

Table 5.— Abundance of amphibians and reptiles captured by pitfall arrays
at the Wind River Experimental Forest in Washington. Arrays of pitfall traps
with drift fences were operated continuously for 180 days in 1983. Old-
growth stands are arranged in order of increasing dryness.

Old growth

Wet Moderate Dry Mature Young Clearcut

Species Stand No.

14

12

21

20

31

41

42

50

60

61

70

71

Northwestern Salamander

2

5

15

4

1

1

1

9

10

2

Pacific Giant Salamander

1

Olympic Salamander

3

1

1

Oregon Ensatina

7

35

29

18

39

14

13

3

24

25

0

1

Larch Mountain

Salamander

10

Western Red-backed

Salamander

19

Rough-skinned Newt

10

4

5

40

1

10

4

7

38

37

7

4

Tailed Frog

44

22 253

4

27

50

4

2

1

4

Red-legged Frog

8

1

3

15

1

19 119

40

5

23

6

Pacific Tree Frog

3

9

Northern Alligator Lizard

1

1

12

1

Northwestern Garter Snake

2

1

4

Common Garter Snake

Differences in Closed-Canopy
Stands

For Oregon and Washington data
combined, mean abundance of com-
mon species (3 salamanders, 2 frogs)
appeared to differ across either forest
development (age) or moisture gradi-
ent (fig. 5). However, except for the
Oregon ensatina, none of the differ-
ences were statistically significant
(table 6). High numbers of individu-
als at a few stands resulted in large
variances in catch at stand types.

Large numbers of both the rough-
skinned newt (Taricha granulosa) and
Northwestern salamander
(Ambystoma gracile) were captured in
a few stands (tables 4-5). Most of the
tailed frogs taken were juveniles at
one old-growth site in Washington
(table 5), and these were apparently
dispersing away from a nearby
stream. Similarly, most (78%) of the
red-legged frogs were taken at 5 sites
(tables 4-5); the largest number (n =

15

119) were juveniles captured at one
mature stand in Washington.

The only species showing a signifi-
cant difference (table 6) across the
chronosequence of stands was the
Oregon ensatina. Its numbers were
lower in mature stands (fig. 5), per-
haps related to amounts of CWD in
different age classes (fig. 6). Abun-
dance of Oregon ensatinas was most
highly correlated with the number of
decay class 4 and 5 logs per hectare
(Pearson r = 0.48, n = 29, P < 0.01)
and the mean diameter (d.b.h.) of
large-sized canopy trees (r = 0.51, n =
29, P < 0.01). A discussion of the
habitat variables used here is pro-
vided in Corn et al. (1988). Mean
abundance of Oregon ensatina also
differed across the moisture gradient
in old-growth stands with fewer
present in wetter sites than drier.
Paradoxically, most OGW stands
have large amounts of CWD (fig. 6).
Oregon ensatina may be associated
with the amount of CWD, but there
are other components of the habitat
that may be under represented in
OGW stands.

Clearcut Stands

We also trapped 5 clearcut sites (all
<10 years old) to describe herpe-
tofauna occurrence in managed

stands. The relative abundance of the
herpetofauna in these clearcuts
markedly differed from 6 compa-
rable young stands (fig. 7). Reptiles
predominate in clearcuts, most likely
responding to increased ambient
temperature in such areas. The Pa-
cific treefrog (Hyla regilla) also was
most abundant in clearcuts.

DISCUSSION

Comparison and Improvements in
Tectiniques

Time-constrained searches (TCS)
provided insufficent animals for
quantitative analyses in most stands.
The technique might be more worth-
while under optimal environmental
conditions (e.g., after heavy rains for
amphibians) and with increased ef-
fort (16+ person-hr per site). Summer
searches added the occurrence of liz-
ards to some stands, but in general
the effort was not worth the time in-
vestment in forested stands of the
Cascade Mountains.

However, TCS can be effective to
sample terrestrial species of salaman-
ders. Our pitfall trapping (180 days)
caught 257 ensatina, 44 clouded sala-
manders, and 13 Oregon slender
salamanders, whereas TCS yielded
113 ensatina (0.44 times that of pit-

r '. " ^

Table 6.— Analysis of variance of species richness and abundance (log
transformed) categorized by age (old growth, mature, and young) and
moisture (wet, moderate, and dry). Wet and dry old growth stands were
not used in the analysis of stand age, and mature and young stands were
not used in the analysis of stand moisture.

Age (n = 1 7) Moisture <n = 1 3)

F

P

F

P

Species Richness

2.02

0,17

0.30

0.75

Total Abundance

0.92

0.42

2.40

0.14

Northwest Salamander

0.38

0.69

1.90

0.20

Rough-skinned Newt

0.91

0.43

0.26

0.78

Oregon Ensatina

8.09

0.005

11.4

0.003

Tailed Frog

0.92

0.42

0.06

0.94

Red-legged Frog

0.65

0.54

0.12

0.89

falls), 76 clouded salamanders (1.7 X
pitfalls), and 57 slender salamanders
(4-4 X pitfalls). The clouded salaman-
der is a common denizen of Oregon
forests and sometimes the most fre-
quently encountered species, but pit-
fall traps caught few. This species
has large toes and is adept at climb-
ing, and perhaps escaped. Or, they
rarely free-fall into traps on the
ground. The Oregon slender sala-
mander seems to be associated with

SPECIES RICHNESS

MEAN I OF SPECIES

ABUNDANCE

MEAN TOIAL CAPTURES

Figure 4.— Mean species richness and
mean total abundance of annpt»ibians and
reptiles in closed-canopy forest stands.

16

Ambystoma gracila Taricha granulosa

Ensatina aschschaltzi Rana aurara

downed woody debris and the best-
known method to sample such mate-
rial is with TCS, area-constrained
searches (Bury and Raphael 1983,
Raphael and Rosenberg 1983), or
hand-collecting of specific amounts
and types of CWD.

For several reasons, we refrained
from using TCS to compare differ-
ences in herpetofauna across stand
ages and moisture gradients. In 1983,
we did not record the number nor
amount of litter (CWD) searched in
each study site, which could have
affected the results. Unless cover
items are scarce, TCS will result in
equivalent numbers of cover items
searched, e.g., 20 logs per person-hr
of search. However, the type, num-
ber and biomass of logs differs
among stands. Thus, the number of
animals collected is not related to the
availability of cover (Corn and Bury
unpublished data).

On the other hand, sites with large
amounts of CWD may be occupied
by many individuals yet few are re-
vealed because they are dispersed.
Douglas-fir forests can have over
1600 mVha of CWD (Spies et al. in
press). Recently, we found that the
density of salamanders in the Oregon
Coast Range (number/ m^ of CWD)
was inversely related to the amount
of CWD present in the stand (Corn
and Bury unpublished data). TCS
will underestimate abundance in
stands with large amounts of CWD
relative to stands with less CWD.
Underestimation of the numbers of
amphibians and reptiles in ecosys-
tems is often more common than
overestimation. Furthermore, we dis-
covered that some collectors tended
to focus on older decay classes of
CWD (that often yield the highest
catch) rather than uniformly search-
ing all objects.

To estimate abundance of sala-
manders, we suggest recording the
volume of CWD searched, control for
time per object (e.g., 15 minutes
maximum), balance effort (e.g.,
equivalent search between different
decay classes of CWD), and relate

17

catch per volume of objects to sepa-
rate estimates of the total CWD per
hectare. These changes are needed to
improve the value of TCS techniques
for sampling the herpetofauna of for-
est ecosystems.

Pitfall traps catch the large num-
bers of individuals needed for quan-
tified analyses of differences between
forest stand types. They proved to be
particularly important for sampling
migratory species of amphibians,
which we found to be common in
Cascade forests. Also, our recent re-
sults indicate for the first time that
tailed frogs occur in "upland" for-
ested habitats.

Vogt and Hine (1982) pointed out
that pitfall traps were most efficient
during periods of precipitation or
soon thereafter. Our results confirm
these observations and, lately, we
have reduced pitfall operations to SO-
SO days in the fall only. Also, the
triad design used here was highly
effective but required great effort
(900 m of drift fence was installed) in
Pacific Northwest forests, which
have large tree roots and rocky soils.
Drift fences are more cost-effective in
sandy areas where they can be more
readily installed.

We caught few reptiles in the Cas-
cade Mountains and pitfall traps
were ineffective for these animals,
even in the warmer summer months.
Reptiles may be numerous in certain
clearcuts (e.g., tables 4-5), in drier
regions such as interior areas of
northern California (e.g., Raphael
and Barrett 1984, Raphael, this vol-
ume) and, based on our prior experi-
ence, in some young managed stands
(10-30 years old). When present,
these would be worth sampling with
pitfall traps.

Pitfall traps alone are adequate to
capture most amphibians and small
mammals (Bury and Corn 1987) but
overall sample size can be improved
by increasing the number of traps per
site. Thus, we have more recently
employed a 6 by 6 pitfall grid (36
traps; 15-m spacing) and the catch is
large enough for quantitative analy-

ses. These adjustments greatly in-
crease the use and effectiveness of
pitfall trapping in the Pacific North-
west and, likely, in other forested
habitats.

Association of Herpetofauna with
Old-Growth Forests

TCS revealed microhabitat differ-
ences between terrestrial species of
salamanders, confirming general ob-
servations about these species (e.g.,
see Nussbaum et al. 1983, Stebbins
1985). However, the habitat require-
ments of these forms need better in-
vestigation.

The Oregon slender salamander
seems to be associated with coarse
woody debris in older decay classes,
which is a characteristic feature of
old-growth forests. This species is
endemic to the Oregon Cascades, oc-
curring only in Douglas-fir and sub-
alpine forests. Thus, timber harvest
might affect populations of slender

LU
O
Z
<
Q
Z

<
LU

:^

Lli
>

<
_J
Ol

7
6
5
4
3
2
1

1

2
3
4
5
6
7

S

IMATUREI YOUNG laEAR-l
ao GROWTH ^

Figure 6.--Biomass of all (top) and class 4
and 5 (bottom) downed wood at 18 stands
at the H. J. Andrews Experimental Forest,
Oregon.

YOUNG

I — I

TAl£D ENSATMA ROUGH- NORTHWEST RED- PACFIC
FROG 8KJNNED SALA- UE.QGED TREE

KEWT MANDER FROG FROG SNAKES UZARDS

n

CLEARCUT

78

Figure 7.— Relative abundance of herpetofauna in young stands and clearcuts. Above the
horizontal: species more abundant in young stands. Below: species more abundant in
clearcuts. Values are the greater mean adundance divided by the lesser, e.g., lizards were
78 times more abundant in clearcuts than in young forest stands.

18

salamanders, and this species merits
special study.

The Olympic salamander (Rhyacot-
riton olympicus) occurs in or near
small streams, which can be dis-
rupted by timber harvest (Bury 1988,
Bury and Corn 1988, Welsh, this vol-
ume). Our techniques sampled ter-
restrial habitats and we found few of
this species (pitfall traps took only 4
in old-growth and 1 in mature
stands). Many tailed frogs were cap-
tured in pitfall traps in closed-can-
opy forests, but they were absent or
rare in clearcuts (only 1% of the total
catch). Both the Olympic slamander
and the tailed frog seem to be sensi-
tive to timber harvest, and the sur-
vival of these species may depend on
protection of cool, flowing streams
(required for breeding and larval de-
velopment) as well as adjacent for-
ested habitats (for shade and reten-
tion of stream substrate quality, see
Bury and Corn 1988). There is a need
to assess the effects of logging in
streamside and upland forests, which
may directly or indirectly affect am-
phibians in headwaters and small
streams (Cooper et al. 1988, Bury and
Corn 1988).

Adults of the rough-skinned newt
and Northwestern salamander mi-
grate to ponds for breeding and,
later, the adults and juveniles move
back to land, which obfuscates their
relation to forest type. The red-
legged frog breeds in slow-moving
creeks or ponds, and the proximity
of such waters may have influenced
the abundance of the frog in adjacent
stands.

Tailed frogs breed in small
streams and the location of these wa-
ters can greatly influence the occur-
rence of the species in nearby forest
stands. Also, we captured some juve-
nile and adult tailed frogs 100 to
>300 m from the nearest stream
(Bury 1988). Before our study, tailed
frogs were not thought to move far
from water (Metter 1964, Nussbaum
et al. 1983). Proximity of aquatic
breeding sites apparently influenced
the capture of several species in up-

land habitat. At the same time,
aquatic and semi-aquatic species
might depend on the forest habitat
for part of their life history, e.g., dis-
persal. We suggest that future re-
search emphasize the life history re-
quirements and movement patterns
of amphibians, which might help to
resolve which factors are most im-
portant to their continued local oc-
currence and abundance.

Fewer Oregon ensatina were cap-
tured in mature forests than either
young or old-growth stands, and this
salmander might be associated with
large amounts of CWD in the Oregon
Cascades. Mature forests lack input
from large trees and snags (see dis-
cussions by Franklin et al. 1981, Har-
mon et al. 1986, Spies et al. in press).
Disturbance (fire or blow-down) cre-
ates new young stands with appre-
ciable amounts of CWD.

Similar to our results, Raphael and
Barrett (1984) found that the abun-
dance of Oregon ensatina in northern
California was correlated to density
of large Douglas-fir trees. However,
they found few ensatina in the
youngest stands (<150 years) they
studied, and they included ensatina
with species associated with old-
growth stands. In the Oregon Cas-
cades, ensatina were ubiquitous and
there is no apparent correlation with
old-growth stands.

Clouded salamanders were most
abundant under the bark of relatively
young logs. They may prefer class 2
and 3 logs, particularly occupying
logs with loose bark. Also, clouded
salamanders appear to be common in
clearcuts (table 1). This species does
not appear to be associated with old-
growth conditions.

In Washington, we only found the
Larch Mountain salamander (Pletho-
don larselli) at one old-growth stand
(table 2). This species may be associ-
ated with forested stands (Herring-
ton and Larson 1985), but the relation
needs further inquiry and verifica-
tion.

Management Considerations

Current evidence suggests that rich,
abundant populations of herpe-
tofauna occur in naturally regener-
ated forests. Within these stands,
however, we found few differences
in amphibians between wet, moder-
ate, and dry old-growth sites and be-
tween young, mature, and old-
growth stands. These results might
be related to '"old-growth" features
occurring in many or all of these
stands. For example, young and ma-
ture sites retained many characteris-
tics of old-growth forests: complex
structure, snags, and large amounts
of downed woody debris, particu-
larly in older decay classes (fig. 6).
Such material is the result of wildfire
that burns and kills larger trees,
which later fall to the ground.

Wildfire often burns unevenly
through stands, resulting in patches
of lightly burned or unburned vege-
tation surrounded by areas more in-
tensively affected by fire. Some large
trees might not be killed during fires
and these persist into the regenerated
stand. Burned trees become snags
that later fall to the forest floor, creat-
ing huge amounts of CWD. This
heterogeneity and large amounts of
CWD in naturally regenerated forest
likely maintain favorable conditions
for many species of the herpetofauna.

Managed stands (clearcuts) had
little downed CWD in older decay
clases (fig. 6) and, generally, no snags
nor trees (except for a rare spar pole
or small planted trees). Current for-
estry practices usually fell all trees
and snags at sites, eliminating vari-
ability in stand age and structure.
Logging is generally followed by pre-
scribed burning of slash and cull
logs, reducing CWD by 50% or more
(Bartels et al. 1985, Maser et al. 1979).
The large amount of CWD at one of
our Oregon clearcuts reflects light
burning (fig. 6). Also, this site was
surrounded by dense, old-growth
forest, which probably contributed
large amounts of CWD before burn-
ing.

19

Often, the result of current timber
harvest is even-aged stands with
little CWD, especially in larger sizes.
Present logging differs from that per-
formed 30 or more years ago, when
more CWD was left on the forest
floor and smaller trees were left in-
tact or ignored. Also, earlier practices
tended to harvest larger, more valu-
able trees with little or no site prepa-
ration (except tree-planting), particu-
larly on private lands. These were
economic decisions, but the resultant
second-growth stands may differ
markedly from current intensive
management of forests.

In contrast to clearcuts, young
stands (naturally regenerated) we
studied were closed-canopy and had
much downed woody debris. The
predominant species were the tailed
frog and ensatina, and young stands
had more newts. Northwestern sala-
manders and red-legged frogs than
did clearcuts (fig. 7). Thus, there
seem to be major differences in the
herpetofaunas of pre-canopy
clearcuts and naturally regenerated
stands (young to old-growth).

There is a critical need to compare
differences in wildlife in intensively
managed stands and those subjected
to other treatments (e.g., prior log-
ging practices, select-cut). At this
time, there is a lack of information on
herpetofaunas or other wildlife in
managed second-growth forests.
Managed forests soon will be the
predominate forest type in the Pacific
Northwest and the bulk of our wild-
life probably will occur in these
stands. Wise management of these
forests should be of foremost concern
for wildlife managers, and done in
concert with protection of isolated
habitat patches (old-growth forest).

ACKNOWLEDGMENTS

We thank our field crew for their
untiring efforts: S. Boyle, R. Hayes, L.
Hanebury, S. Martin, T. Olson, and S.
Woodis. We thank J. Dragavon, L.
Jones, P. Morrison, R. Pastor, and D.

Smith for checking traps. A. McKee
and J. Moreau at the H. J. Andrews
Experimental Forest, Oregon, and
personnel at Wind River Experimen-
tal Forest and the Carson National
Fish Hatchery, Washington, supplied
logistical support. Data in figure 6
was provided by T. Spies. We also
thank R. E. Beiswenger, A. B. Carey,
and M. G. Raphael for review com-
ments. This is contribution number
67 of the USD A Forest Service Proj-
ect, Wildlife Habitat Relationships in
Western Washington and Oregon.

LITERATURE CITED

Anthony, R. C, E. D. Forsman, G. A.
Green, G. Witmer, and S. K. Nel-
son. 1987. Small mammal popula-
tions in riparian zones of different-
aged coniferous forests. Murrelet
68:94-102.

Bartels, Ronald, John D. Dell, Rich-
ard L. Knight, and Gail Schaefer.
1985. Dead and down woody ma-
terial, p. 171-186. In E. Reade
Brown, tech. ed.. Management of
wildlife and fish habitats in forests
of western Oregon and Washing-
ton. Part 1 - chapter narratives.
USDA Forest Service, Pacific
Northwest Publ. No. R6-F&WL-
192-1985.

Bennett, Stephen H., J. Whitfield Gib-
bons, and Jill Glanville. 1980. Ter-
restrial activity, abundance and
diversity of amphibians in differ-
ently managed forest types.
American Midland Naturalist
103:412-416.

Bruce, Charles, Dan Edwards, Kim
Mellen, Anita McMillan, Tom
Owens, and Harold Sturgis. 1985.
Wildlife relationships to plant
communities and stand condi-
tions, p. 33-55. In E. Reade Brown,
tech. ed.. Management of wildlife
and fish habitats in forests of west-
ern Oregon and Washington. Part
1 - chapter narratives. USDA For-
est Service, Pacific Northwest
Publ. No. R6-F&WL-192-1985.

Bury, R. Bruce. 1983. Differences in

amphibian populations in logged
and old growth redwood forest.
Northwest Science 57:167-178.

Bury, R. Bruce. 1988. Habitat rela-
tionships and ecological impor-
tance of amphibians and reptiles.
[In press] In Kenneth J. Raedeke,
ed., Streamside management: ri-
parian wildlife and forestry inter-
actions. Proceedings of a Sympo-
sium [University of Washington,
Seattle, February 11-13, 1987] Insti-
tute of Forest Resourses, Univer-
sity of Washington.

Bury, R. Bruce, and Paul Stephen
Com. 1987. Evaluation of pitfall
trapping in northwestern forests:
trap arrays with drift fences. Jour-
nal of Wildlife Management
51:112-119.

Bury, R. Bruce, and Paul Stephen
Corn. 1988. Responses of aquatic
and streamside amphibians to tim-
ber harvest. [In press] In Kenneth
J. Raedeke, ed., Streamside man-
agement: riparian wildlife and for-
estry interactions. Proceedings of a
Symposium [University of Wash-
ington, Seattle, February 11-13,
1987] Institute of Forest Resourses,
University of Washington.

Bury, R. Bruce, and Martin G. Ra-
phael. 1983. Inventory methods
for amphibians and reptiles, p.
416-419. In John F. Bell and Toby
Atterbury, eds.. Renewable re-
source inventories for monitoring
changes and trends: Proceedings
of an international conference.
[Corvallis, Oregon, August 15-19,
1983] Society of American Forest-
ers 83-14.

Campbell, Howard W., and Stephen
P. Christman. 1982. Field tech-
niques for herpetofaunal commu-
nity analysis, p. 193-200. In Nor-
man J. Scott, Jr., ed., Herpetologi-
cal Communities. U.S. Fish and
Wildlife Service, Wildlife Research
Report No. 13.

Carey, Andrew B. 1988. The influ-
ence of small streams on the com-
position of upland bird communi-
ties. [In press] In Kenneth J.
Raedeke, ed., Streamside manage-

20

ment: riparian wildlife and for-
estry interactions. Proceedings of a
Symposium [University of Wash-
ington, Seattle, February 11-13,
1987] Institute of Forest Resourses,
University of Washington.

Cooper, Edward J., William J. Trush,
and Allen W. Knight. 1988. Effects
of logging on Pacific giant sala-
manders: influence of age-class
composition and habitat complex-
ity. Bulletin of the Ecological Soci-
ety of America 69(suppl.):104-105.

Enge, Kevin E., and Wayne R. Mar-
ion. 1986. Effects of clearcutting
and site preparation on herpe-
tofauna of a north Rorida flat-
woods. Forest Ecology and Man-
agement 14:177-192.

Fosburg, Whit. 1986. Wildlife issues
in the National Forest system, p.
159-173. In Roger L. Di Silvestro
ed., Audubon Wildlife Report
1986. The National Audubon Soci-
ety, New York. 1094 p.

Franklin, Jerry F., and Thomas A.
Spies. 1984. Characteristics of old-
growth Douglas-fir forests, p. 328-
334. In Proceedings, Society of
American Foresters national con-
vention. [Portland, Oregon, Octo-
ber 16-20, 1983] Society of Ameri-
can Foresters, Washington, D.C.

Franklin, Jerry F., Kermit Cromack,
Jr., William Denison, Arthur
McKee, Chris Maser, James Sedell,
Fred Swanson, and Glen Juday.
1981. Ecological characteristics of
old-growth Douglas-fir forests.
U.S. Forest Service, General Tech-
nical Report PNW-118. 48 p.

Friend, Gordon R. 1984. Relative effi-
ciency of two pitfall-drift fence
systems for sampling vertebrates.
Austalian Zoologist 21:423-433.

Gibbons, J. Whitfield, and Raymond
D. Semlitsch. 1981. Terrestrial drift
fences with pitfall traps: an effec-
tive technique for quantitative
sampling of animal populations.
Brimleyana 7:1-16.
Harmon, M. E., J. F. Franklin, F. J.
Swanson, P. Sollins, S. V. Gregory,
J. D. Lattin, N. H. Anderson, S. P.
Cline, N. G. Aumen, J. R. Sedell,

G. W. Lienkaemper, K. Cromack,
Jr., and K. W. Cummins. 1986.
Ecology of coarse woody debris in
temperate ecosystems. Advances
in Ecological Research 15:133-302.

Harris, Larry D. 1984. The Frag-
mented Forest: Island Biogeogra-
phy Theory and the Preservation
of Biotic Diversity. University of
Chicago Press. 211 p.

Herrington, Robert E., and John H.
Larson. 1985. Current status, habi-
tat requirements and management
of the Larch Mountain salamander
Plethodon larselli Bums. Biological
Conservation 34:169-179.

Jones, K. Bruce. 1981. Effects of graz-
ing on lizard abundance and di-
versity in western Arizona. South-
west Naturalist 26:107-115.

Jones, K. Bruce. 1986. Amphibians
and reptiles, p. 267-290. In Allen Y.
Cooperrider, Raymond J. Boyd,
and Hanson R. Stuart eds.. Inven-
tory and monitoring wildlife habi-
tat. U.S. Dept. Interior, Bureau of
Land Management, Service Cen-
ter, Denver, Colorado. 858 p.

Kerrick, Michael A., Kathy Johnson,
and Richard J. Pedersen. 1984.
What information is necessary for
planning the management of old-
growth forest for wildlife? p. 383-
386. In Proceedings, Society of
American Foresters national con-
vention. [Portland, Oregon, Octo-
ber 16-20, 1983] Society of Ameri-
can Foresters, Washington, D.C.

Manuwal, David A., and Mark H.
Huff. 1987. Spring and winter bird
populations in a Douglas-fir forest
sere. Journal of Wildlife Manage-
ment 51:586-595.

Maser, Chris, Ralph G. Anderson,
Kermit Cromack, Jr., Jerry T. Wil-
liams, and Robert E. Martin. 1979.
Dead and down woody material,
p. 78-95. In Jack Ward Thomas,
tech. ed.. Wildlife habitats in man-
aged forests: the Blue Mountains
of Oregon and Washington. Agri-
cultural Handbook 553. U.S. Dept.
Agriculture, Washington, D.C.

Maser, Chris, and James M. Trappe
(tech. eds.). 1984. The seen and

unseen world of the fallen tree.
USDA Forest Service General
Technical Report PNW-164. 56 p.

Metter, Dean E. 1964. A morphologi-
cal and ecological comparison of
two populations of the tailed frog,
Ascaphus truei Stejneger. Copeia
1964:181-195.

Nussbaum, Ronald A., Edmund D.
Brodie, Jr., and Robert M. Storm.
1983. Amphibians and Reptiles of
the Pacific Northwest. University
Press of Idaho, Moscow. 332 p.

Rough, F. Harvey, Ellen M. Smith,
Donald H. Rhodes, and Andres
Collazo. 1987. The abundance of
salamanders in forest stands with
different histories of disturbance.
Forest Ecology and Management
20:1-9.

Raphael, Martin G. 1984. Wildlife
populations in relation to stand
age and area in Douglas-fir forests
of northwestern California, p. 259-
274 In William R. Meehan, Theo-
dore R. Merrell, Jr., and Thomas
A. Hanley, eds.. Fish and wildlife
relationships in old-growth for-
ests. Proceedings of a symposium.
[Juneau, Alaska, April 12-15, 1982]
425 p. American Institute of Fish-
ery Research Biologists, (available
from John W. Reintjes, Rt. 4, Box
85, Morehead City, North Caro-
lina 28557).

Raphael, Martin G., and Reginald H.
Barrett. 1984. Diversity and abun-
dance of wildlife in late succes-
sional Douglas-fir forests, p. 352-
360. In Proceedings, Society of
American Foresters national con-
vention. [Portland, Oregon, Octo-
ber 16-20, 1983] Society of Ameri-
can Foresters, Washington, D.C.

Raphael, Martin G., and Kenneth V.
Rosenberg. 1983. An integrated
approach to wildlife inventories in
forested habitats, p. 219-222. In
John F. Bell and Toby Atterbury,
eds.. Renewable resource invento-
ries for monitoring changes and
trends: Proceedings of an interna-
tional conference. [Corvallis, Ore-
gon, August 15-19, 1983] Society
of American Foresters 83-14.

21

Rosenberg, Kenneth V., and Martin
G. Raphael. 1986. Effects of forest
fragmentation on vertebrates in
Douglas-fir forests, p. 263-272. In
Jared Verner, Michael L. Morri-
son, and C. John Ralph, eds..
Wildlife 2000. University of Wis-
consin Press. 470 p.

Ruggiero, Leonard F., and Andrew
B. Carey. 1984. A programmatic
approach to the study of old-
growth forest — wildlife relation-
ships, p. 340-345. In Proceedings,
Society of American Foresters na-
tional convention. [Portland, Ore-
gon, October 16-20, 1983] Society
of American Foresters, Washing-
ton, D.C.

Salwasser, Hal. 1987. Spotted owls:
turning a battleground into a blue-
print. Ecology 68:776-779.

Scott, Norman J., Jr. (Ed.). 1982. Her-
petological Communities. U.S.
Fish and Wildlife Service, Wildlife
Research Report 13. 239 p.

Spies, Thomas A., Jerry F. Franklin,
and Ted B. Thomas. In press.
Coarse woody debris in Douglas-
fir forests of western Oregon and
Washington. Ecology.

Stebbins, Robert C. 1985. A Field
Guide to Western Reptiles and
Amphibians. Houghton Mifflin
Co., Boston. 336 p.

Thomas, Don. In press. Forest age as-
sociation of bats in western Ore-
gon and Washington. Journal of
Wildlife Management.

Vogt, Richard C, and Ruth L. Hine.
1982. Evaluation of techniques for
assessment of amphibian and rep-
tile populations in Wisconsin, p.
201-217. In Norman J. Scott, Jr.,
ed., Herpetological Communities.
U.S. Fish and Wildlife Service,
Wildlife Research Report 13.

Webb, G. A. 1985. Habitat use and
activity patterns in some southeast
Australian skinks. p. 23-30. In Gor-
don Grigg, Richard Shine, and
Harry Ehmann, eds.. Biology of
Australasian Frogs and Reptiles.
Surrey Beatty & Sons Pty Ltd,
New South Wales, Australia.

West, Stephen D. 1985. Differential
capture between old and new
models of the Museum Special
snap-trap. Journal of Mammalogy
66:798-800.

Wilcove, David S. 1987. Public lands
management and the fate of the
spotted owl. American Birds
41:361-367.

22

Long-Term Trends in
Abundance of Annphlbians,
Reptiles, and Mammals in
Douglas- Fir Forests of
Northwestern California^

Martin G. RaphaeP

Abstract.— Relative abundance of 55 species of
amphibians, reptiles, and mammals was estimated
at 166 sites representing early clearcut through old-
growth Douglas-fir forest in northwestern California.
Nine species were strongly associated with older
stands and 1 1 species were strongly associated with
younger stands. The remaining species were either
too rare to analyze statistically (22 species) or
exhibited no clear trends of abundance in relation
to stand age (13 species). Estimates of relative
abundance of each species in each stage, coupled
with data on historical, present, and future acreage
of timber in each serai stage, were used to
approximate the long-term impacts of timber
harvest on the fauna of the Douglas-fir region in
northwestern California.

Management of old-growth Douglas-
fir (Pseudotsuga menziesii) forests is
controversial in the Pacific North-
west, primarily because of the pos-
sible value of old-growth as habitat
for certain wildlife species versus the
revenues represented by old-growth
trees (Meslow et al. 1981, Harris et al.
1982). Management to provide wild-
life habitat requires an inventory of
associated wildlife species and an
assessment of their old-growth de-
pendency. An analysis of the size
and distribution of habitat patches
necessary to support viable popula-
tions of those species is also critical
(Burgess and Sharp 1981, Rosenberg
and Raphael 1986, Scott et al. 1987).

This study describes the relative
abundance of amphibians, reptiles,
and mammals in six serai stages rep-
resenting clearcuts, young timber
stands, and mature forest in north-
western California. These estimates
of relative abundance were used to
project probable long-term changes
in population size of amphibians,
reptiles, and mammals as each serai

^ Paper presented at Symposium, Man-
agement of Amphibians, Reptiles and Small
Mammals in North America (Flagstaff, AZ,
July 19-21, 1988).

'Research Ecologist, Forestry Sciences
Laboratory, USDA Forest Service, Rocky
Mountain Forest and Range Experiment
Station, 222 South 22nd Street, Laramie,
Wyoming 82070.

Stage responds to forest management
practices.

METHODS

Stand Selection

Study stands were on the Six Rivers,
Klamath, and Shasta-Trinity National
Forests within a 50-km radius of Wil-
low Creek, Calif. Forest cover was
dominated by Douglas-fir, usually in
association with an understory of
tanoak (Lithocarpus ensiflorus) and Pa-
cific madrone (Arbutus menziesii). Ele-
vations varied from 400 to 1300 m.

Stage

1

2
3

4

5

6

Raphael and Barrett (1984) describe
methods for aging these stands.
Ground surveys were used to verify
stand conditions. Forest Service
stand designations were used to
guide stand selection, but the final
classification of each stand into serai
stages was based on measured vege-
tation characteristics.

The study region is characterized by
warm, dry summers and cool, wet
winters; total precipitation averages
60-170 cm per year.

After selecting potential study
stands using timber maps and aerial
photographs, I then located all stands
that were accessible by road, were
relatively homogeneous with respect
to tree cover, included no large clear-
ings or other anomalous features,
and were free from scheduled timber
harvest for at least the next 3 years.

From this restricted subset of
stands, I randomly chose 10 to 15
stands representing each of six serai
stages:

Vegetation Sampling

The structure and composition of
vegetation on each stand in the three
older serai stages was measured in
three, randomly selected, 0.04-ha cir-
cular subplots within a 90-m radius
of each plot center. Within each sub-
plot, observers recorded species.

Serai state

Age (yrs)

Early

<10

Late

10-20

Pole

20-50

Sawtimber

50-150

Mature

150-250

Old-growth

>250

Classification

Clearcut (brush/sapling)
Young forest (pole/sawtimber)
Mature forest

23

height, diameter at breast height
(d.b.h.) and crown dimensions of
each tree or shrub >2.0 m tall. In ad-
dition, all trees >90-cm d.b.h. were
counted on one 0.50-ha circular sub-
plot centered on the plot. This
sample permitted a better estimate of
the density of large-diameter trees.
Numbers of larger (>8-cm diameter)
logs and volume of other downed
woody debris were estimated along a
30-m transect crossing the center of
each 0.04-ha subplot (Brown 1974).
Marcot (1984) sampled vegetation in
a similar manner on stands in the
three early-seral stages.

Vertebrate Sampling

All field data were collected by a
team of three to six biologists. We
used a variety of techniques to
sample various taxonomic groups.

Pitfall Arrays

We used pitfall arrays to capture
small mammals (especially insecti-
vores), reptiles, and salamanders. An
array was composed of ten 2-gallon
plastic buckets buried flush with the
ground and covered with plywood
lids, arranged in a 2 x 5 grid with 20-
m spacing. We placed one array
within each stand center and checked
traps at weekly to monthly intervals
from December 1981 (sawtimber,
mature, old-growth; n = 27, 56, and
52 sites in each stage, respectively) or
August 1982 (early shrub-sapling,
late shrub-sapling, pole; n = 10 sites
each) until October 1983. All live ani-
mals were marked and released; re-
captures were excluded from analy-
ses. Dead animals were collected and
prepared for permanent deposit in
museum collections. Results for each
species were expressed as captures
per 1000 trapnights on each stand.
Raphael and Rosenberg (1983) dem-
onstrated that abundance estimates
(capture rates) had stabilized after 15
months of continuous trapping.

Drift Fence Arrays

To better sample snakes, we installed
a drift fence array (Campbell and
Christman 1982, Vogt and Hine 1982)
on each of 60 randomly selected
stands (10 of each of the three early
stages and sawtimber, 8 mature, and
12 old-growth). An array consisted of
two 5-gallon buckets placed 7.6 m
apart and connected by an aluminum
fence 7.6 m long and 50 cm tall with
two 20 X 76 cm cylindrical funnel
traps, one on each side of the center
of the fence. These fences were oper-
ated from May through September
1983. All captures were combined
with those from the pitfall arrays
along with the associated trapnights
from each stand.

Track Stations

Tracks of squirrels and other larger
mammals were recorded on each site
on a smoked aluminum plate baited
with tuna pet food (Barrett 1983, Ra-
phael and Barrett 1981, Raphael et al.
1986, Taylor and Raphael 1988).
Based on results of a pilot study (Ra-
phael and Barrett 1981), observers
monitored each station for 8 days in
August or September in 1981-1983,
sampling 20 stations in each of the
three early stages and 81, 168, and
157 stations in the sawtimber, ma-
ture, and old-growth stages, respec-
tively. The proportion of stations in
each serai stage on which a species
occurred was as an index of that spe-
cies' abundance.

Livetrap Grids

To better estimate abundance of
small mammals that were liable to
escape from pitfalls, we established
27 livetrap grids (3 in each of the
three earliest stages and 5, 7, and 6 in
the three later stages), each of which
usually consisted of 100 25-cm Sher-
man livetraps arranged in a 10 x 10
grid with 20-m spacing. Other grid

sizes or shapes were used when the
plot configuration would not contain
the standard grid. Traps were
checked each day for 5 days (based
on pilot studies, Raphael and Barrett
1981) during July in 1981 (late stages
only), 1982, and 1983 (all stages). Re-
sults for each species were expressed
as mean number of captures per 100
trapnights.

Surface Searcti

To better sample certain amphibian
species, we conducted time- and
area-constrained searches (Bury and
Raphael 1983, Raphael 1984) on a
subset of sites in 1981 (late stages),
1982, and 1983 (all stages). A two-
person team searched under all mov-
able objects and within logs on three
randomly located 0.04-ha circular
subplots (fall 1981, 1982) or within a
1-ha area for 4 working hours (spring
1983). We conducted 20 surveys in
each of the three early stages and 29,
39, and 48 surveys in the three late
stages.

Opportunistic Observations

Observers recorded the presence of
vertebrates or identifiable vertebrate
sign incidental to the above proce-
dures. We tallied observations to cal-
culate frequency of occurrence of
rarer species within each stage.

Forest Area Trends

Estimates of historical, current, and
future acreage in each serai stage
were taken from Raphael et al. (in
press). For these analyses, I com-
bined similar pairs of serai stages
into three generalized stages repre-
senting brush/sapling, pole/ sawtim-
ber, and mature timber. I then com-
puted relative abundance of each
vertebrate species in these three
stages using a weighted average
(weights based on sampling effort) of

24

estimates from each of the two stages
forming the pair. Population esti-
mates for historical, present, and fu-
ture time periods were computed
using the formula:

3

where P.^ was the relative population
size of the zth vertebrate species at
time t,D.. was the relative abundance
of the zth vertebrate in the /th serai
stage, and A .^ was the total area of
each of the tkree serai stages at time

RESULTS

Vegetation Structure

Comparisons of vegetation structure
among the serai stages (table 1)
showed that older stands had greater

canopy volume, basal area, litter
depth, and density of Douglas-fir
stems >90 cm d.b.h. Downed wood
mass differed among stages, but the
greatest volume occurred in the
youngest stands, probably in the
form of logging slash, and the lowest
volume occurred in pole and sawtim-
ber stages. Early-seral stands were
higher in elevation than older stands,
probably because of the logistics of
timber harvest in the area (most
clearcuts were located along ridg-
etops). Stands in the two earliest
serai stages, also because of logging,
were smaller in area than stands in
the four older stages.

Vertebrate Abundance and
Diversity

Among all plots and years of study,
we recorded 9,928 captures of all

Table 1 .—Comparisons of vegetation characteristics among serai stages of
Douglas-fir forest, northwestern California, 1981-1983.

Characteristic

Early Late

brush/ brush/ Old-
sapiing sapling Pole SawtimberMature growth

Canopy volume (jTT'lrv?)

10.77

n.26

^3.64

7.15

7,52

7.47

Live stem basal area (m^hc)

^2.6

^52.8

50.5

60.2

65.6

Snog basal area (m^/ha)

2_

4.7

6.1

5.3

Downed wood moss (metric tons/ha)

<8 cm diameter

^9.7

V.9

m.9

12.9

12.3

11.5

>8 cm diameter

^81.4

V4.7

^52.4

32.3

43.6

67.3

Utter depth (cm)

^2.2

M.8

^6.0

6.2

5.1

7.1

Douglas-fir >90 cm d.b.h. (n/ha) —

3.6

19.3

25.7

Elevation (m)

1128

1016

972

660

832

904

Stand area (ha)

12,3

21.9

41.2

47.1

62.0

84.2

Solar radiation index^

0.34

0.41

0.51

0.49

0.49

0.43

Slope (%)

48

30

31

36

41

52

Age (years since clearcut.

or index)

9

14

123

206

294

'Dofo are from Marcot(1984), with permission, and represent a larger number of
sites than v/ere sampled in tt)e present study.

'Dasties indicate no values were available.

^ Index of total yearly solar energy flux (Frank and lee 1966). Larger values indicate
warmer, drier sites.

J

Species during 898,431 trapnights
from pitfalls and drift fences; 1,636
captures of amphibians during sur-
face searches; 3,066 small mammal
captures during 35,070 trapnights
from Hvetrap grids; and 510 detec-
tions of larger mammals from track
stations. Relative abundances of 55
species, based on the most appropri-
ate sampling method for each spe-
cies, are summarized in table 2. Val-
ues are comparable across stages but
not among taxa if different sampling
methods were used. Amphibians
were much more abundant in for-
ested than in clearcut stands,
whereas reptiles were more abun-
dant in clearcuts. None of the am-
phibians and reptiles [except rarer
species such as northwestern sala-
mander (see appendix for scientific
names of vertebrates)] was absent
from any stage.

Mammals exhibited a greater vari-
ety of responses to serai stage. Some
(e.g., Douglas' squirrel, western red-
backed vole) increased in abundance
from earliest to latest serai stages;
others (e.g., deer mouse) decreased
along this gradient. A number of spe-
cies (e.g., Allen's chipmunk, dusky-
footed woodrat, pinyon mouse, Cali-
fornia vole) were most abundant
both in late shrub-sapling and ma-
ture or old-growth stands.

Mean numbers of mammal and
reptile species recorded per stand
differed among serai stages, but
mean numbers of amphibian species
did not differ significantly (fig. 1).
Among mammals, mean numbers of
species were greatest in mature and
old-growth stages. In contrast, mean
numbers of reptile species were
greatest in the two earliest stages.

Long-Term Trends

Estimates of land area in each serai
stage through time (table 3) indicate
more area is occupied by early serai
stages currently than during historic
or future times. Mature and old-
growth stages currently occupy

25

Table 2.~Mean relative abundance of amphibians, reptiles, and mammals among serai stages of Douglas-fir forest,
nortliwestern California, 1981 -1983.

Species'

Sampling .
metliod(s)2

Total
captures

Relative abundance
among serai stages^

6 Significance*

Salamanders

Northwestern salamander^
Pacific giant salamander
Olympic salamander^
Rough-skinned newt
Del Norte salamander
Ensatina

Black salamander
Clouded salamander

Frogs and toads
Tailed frog^
Western toad
Pacific treefrog
Foothill yellow-legged frog^
Bullfrog^

Turtles

Western pond turtle^

PD. TC^
PD
TC
TC^

PD,TC^

6

0

0

0

1

3

4

PD

28

0

0.05

0

0.01

0.02

0.04

PD,TC6

5

0

0

0

1

1

3

PD

68

0.02

0

0

0.05

0.09

0.04

TC

196

0.70

0.60

0.05

0.07

1.92

1.92

TC

1116

2.40

1.85

8.10

6.28

8.15

7.69

TC

32

0.05

0.05

0.05

0.03

0.21

0.42

TC

103

0.35

1.55

0.50

0.10

0.31

0.83

0,114

0.403
0.035
0.001

o.on

0.009

3

0

0

0

0

2

0

54

0.18

0.03

0.02

0.08

0,06

0.01

0.035

51

0.60

0.05

0,10

0.55

0,03

0,06

0.000

6

1

0

0

1

0

0

3

0

0

0

1

0

0

5

0

0

0

4

0

0

Lizards

Western fence lizard PD

Sagebrush lizard PD

Western skink PD

Southern alligator lizard PD

Northern alligator lizard PD

Snakes

Rubber boa^
Rtngneck snake^
Sharp-tailed snake^
Racer^

Gopher snake-^
Common kingsnake^
Common gartersnake5
Western terrestrial gartersnakeS
Western ratt-lesnake5

Mammals

Pacific shrew PD

Trowbridge's shrew PD

Shrev/-mole PD

Coast mole^ PD

Allen's chipmunk LT

Weste rn g ray squi rrel TP^

Douglas' squirrel TP*

Northern flying squirrel TP*

523

1.77

2.38

0.30

0.94

0.54

0.11

0.000

196

2.66

0.76

0.25

0.09

0.11

0.01

0.000

584

3.05

3.47

0.78

0.73

0,42

0.13

0,000

41

0.03

0

0

0.11

0,05

0.03

0.085

586

0.81

1.03

0.90

0.97

0.60

0.44

0,029

CO, PD^

7

0

20

10

0

4

0

OCPD*

6

0

0

0

0

4

0

PD^

22

10

20

30

0

5

4

OCPD^

8

0

20

0

4

5

4

GO, PD*

0

0

10

0

2

0

00,PD*

1

0

0

0

4

0

0

GO, PD6

19

20

10

20

0

5

11

400,PD

11

0

20

10

7

4

6

oo. pd;tc6

5

0

10

0

0

2

6

89

0.02

0.08

0

0.07

0.07

0.17

2384

2.70

4.01

2.83

3.04

3.16

3.80

479

0.04

0.16

0.25

0.76

0.55

0.55

15

0

0.03

0

0.02

0.05

0.06

254

16.7

29.5

0.8

2.8

5.2

5.0

48

0

0

10

12

12

9

104

0

0

20

16

22

30

43

0

0

15

9

18

13

0.004
0.215
0.002

0.003
0.378
0.001
0.046

(continued)

26

Table 2. —(continued).

Species'

Sampling
method(s)2

Total
captures

Relative abundance
among serai stages^

6 Significance*

Deer mouse

PD

1 Ml

5.09

0 C\~l

o.U/

U.OV

u.bo

U.Vo

1.28

0.000

Brush mouse

LT

33

0

0.33

3.67

0.25

0.25

0

0.216

Pinyon mouse

LT

222

1.35

10.34

4.67

10.63

3.86

2.76

0.086

Dusky-footed woodrat

LT

115

1.9

3.5

0.2

1.2

4.4

3.4

0.000

Western red-backed vole

PD

669

0,35

0.36

0.46

0.45

0.82

0.97

0.015

Red tree vole

PD

19

0

0.10

0

0.07

0.11

0.15

0.586

California vole

PD

106

0.89

1.70

0.03

0.02

0.01

0.01

0.000

Creeping vole

PD

22

0.09

0.03

0.05

0.04

0.01

0.01

0,038

Western jumping mouse^

PD

2

0

0

0

0

0.04

0.02

Coyote^

ALL^

7

10

30

0

15

9

15

Gray fox

TP^

63

20

15

10

30

11

8

0.001

Black bear

TP

196

20

25

5

42

45

48

0.028

Ringtail

TP

25

0

0

0

10

6

4

0.249

Raccoon^

TP

3

0

0

0

0

1

1

Fisher

TP

58

0

5

25

6

13

15

0.060

Ermine^

PD

2

0

0

0

0

0,02

0.02

Western spotted skunk

TP

70

10

15

5

10

18

15

0.426

Striped skunk^

TP

17

0

0

0

7

6

1

Bobcat^

TP

3

5

5

0

1

2

0

'A// names follow Laudenslayer and Grenf ell (1983).

^PD = Pitfall plus drift fence, TC = Time- and area-constrained searcti. OO = Opportunistic observations, TP = Track plots. LT= Live
traps, ALL = all observation methiods combined.

^Seral stages (and numbers of stands sampled) are: 1 — early brustt/sapling (n= 10): 2— late brusti/sapling (n= 10): 3— pole (n= 10): 4—
sawtimber (n=27): mature (r]=56); 5—old-growtht (n=63).

"Significance from analysis of variance (means) or cN-square analysis (frequencies) comparing abundances among stages. A dashi
indicates th\at no test was performed.

^Too rare for subsequent analyses.

'^Abundance values based on percent frequencies.

about half of historic acreage, and
these stages will probably occupy
only about 30% of current acreage
under the most likely harvest pat-
terns of the future (table 3).

The implications of these changing
distributions of serai stages for am-
phibians, reptiles, and mammals are
summarized in figure 2. Nearly equal
numbers of species are likely to have
increased or decreased by more than
25% relative to historic abundance at
present and in the future. Three of
the five reptile species are presently
more abundant than in historic times
and all five species will likely be
more abundant in the future. Am-
phibians showed an opposite pattern.

Four of the eight species are pres-
ently less abundant and five of the
eight may be less abundant in the fu-
ture. Among the 20 mammal species,
seven are presently less abundant
than in historic times whereas five
are more abundant. Eight species
will probably be less abundant in the
future and six more abundant.

DISCUSSION

Abundance in Serai Stages

Results suggest late brush/ sapling
and mature/old-growth serai stages
provided more productive wildlife

habitat than early brush/ sapling,
pole, and sawtimber stages. Among
amphibians, only ensatinas were cap-
tured frequently in pole sites.
Clouded salamanders were generally
under bark or inside downed logs
and persisted in clearcut stands as
long as adequate numbers of logs
were retained, especially in late sites
(Raphael 1987, Welsh, this volume).

Lizards were more abundant in
earlier serai stages than in pole and
mature stages. Among snakes, only
sharp-tailed snakes were observed
on early sites; other species occurred
on later sites. However, sampling
was not sufficient for definitive con-
clusions.

27

With the exception of the deer
mouse, small mammals were more
abundant on late brush/ sapling sites.
Dusky-footed woodrats were of spe-
cial interest in this regard as we ob-
served many woodrat nests built
among the stems of tanoak and Pa-
cific madrone in late brush/ sapling
sites. The combination of abundant
mast, good nesting substrate, and
protection from predation (spotted
owls rarely forage in old, brush-
dominated clearcuts) provided by
the dense, brushy cover were proba-
bly the reasons that woodrats and
other small mammals were more
numerous in late clearcut sites (Ra-
phael 1987).

Tree squirrels were most abun-
dant in mature forest sites and
ground squirrels were more abun-
dant in early clearcut sites. Chip-
munks were the only squirrel that
reached peak abundance in early
serai sites. Their abundance was cor-
related with the cover of tanoak in
the understory (Raphael 1987). Man-
agement actions, such as herbicide
treatments, that shorten or delete the
late brush/ sapling stage are probably
detrimental to chipmunks, woodrats,
and certain other rodents.

Several carnivorous mammals
were abundant in the late brush/sap-
ling stage. Greater prey density in
late compared to early and pole sites
may explain this higher frequency of
carnivores although more data will
be necessary to confirm this observa-
tion.

Of the 55 species observed, 20
were strongly associated with either
older (9 species) or younger (11 spe-
cies) stands (table 4). Three salaman-
ders and six mammals were associ-
ated with older stands. One toad,
one frog, five lizards, and four mam-
mals were associated with younger
stands. Five species associated with
old-growth were also abundant in
late (brushy) clearcut stages (table 2).
These species peak in abundance in
old stands and late clearcuts, with
low abundance in intermediate age
classes.

Table 3 — Approximate area (millions of ha) of serai stages in Douglas-fir
forests of northwestern California In historic, present, and future time peri-
ods (after Raphael et al., in press).

Serai stage

Historical

Present

Likely
future

Worst
case
future'

Brush/sapling

0.14

0.49

0.20

0.24

Young forest

0.14

0.20

0.77

0.85

Mature forest

0.81

0.40

0.12

0.00

'Assumes that all mature and old-growth stands ore harvested and all lands man-
aged under short rotatiorts.

CO

PRESOff

12 r

3 4

SERAL STAGE

Figure 1 .—Mean numbers of amphibian,
reptile, and mammal species observed in
serai stages of Douglas-fir forest, northwest-
ern California, 1981-1983. Serai stages (and
numbers of stands sampled) are: 1 - early
brush/sapling (n = 10); 2 - late brush/sap-
ling (n = 1 0); 3 - pole (n = 1 0); 4 - sawtimber
(n = 27); 5 - mature (n = 56); 6 - old-growth
(n = S3). Vertical lines indicate 95% confi-
dence intervals.

UttLYFUIUt

iMMi

jlBTlB

<-75

-75 -50 -TS •

CHANGE h ASUNQANCE (1

Figure 2.— Percent change in population
size of amphibian, reptile, and mammal
species at present and in the future relative
to estimated historical populations. Histo-
grams represent the numbers of species
increasing or decreasing by specified per-
centages.

28

I examined habitat associations
among each of the above 9 species by
computing correlations of their abun-
dance with specific habitat compo-
nents (table 5). Density of large trees
and hardwood volume were corre-
lated with the abundance of most
species. Moisture, as measured by
the presence of surface water, mois-
ture-loving tree species, or north-fac-
ing slopes, was important for most
mammals and one salamander spe-
cies. Four mammal S}:>ecies were sig-
nificantly more abundant on higher
elevation stands. Downed wood vol-
ume also was significantly and posi-
tively correlated with abundance of
four amphibian and mammal spe-
cies. The abundance of hardwoods in
the understory was important for
many species in each group. In con-
trast, snag density was not positively
correlated with the abundance of any
species.

Long-Term Trends

The list of sensitive species (table 4)
is tentative pending results of addi-

r

Table 5.— Habttat components that were correlated with relative abun-
dance of amphibians and mammals associated with late-seral Douglas-fir
forests of northwestem California.

Density of

Hardwood

Downed

conifers

under-

wood

Standing

Species >90-cm d.b.h.

story

mass

snags

Moisture

Elevation

Del Norte salamander

X

Black salamander

X

X

Clouded salamander X

X

Pacific shrew X

X

X

X

Douglas' squirrel X

^(X)

X

X

X

Northern flying squirrel

X

(X)

Dusky footed woodrat X

X

X

X

Western red-backed vole X

X

X

X

Fisher X

X

'Parentheses indicate negative correlations.

tional surveys and more intensive,
species-specific research. The projec-
tions, although based on an intensive
sampling effort, are highly specula-
tive. Three assumptions must be rec-
ognized to interpret these results.
First, I assumed that greater relative
abundance in a serai stage indicates a
species' preference for that stage and
that preferences remain constant
with shifting distribution of acreage

r

Table 4.— Amphibian, reptile, and mammal species most strongly affected
by future harvest of old-growth Douglas-fir forest, northwestern California.'

Decreasers— associated
with late-seral forest

Increasers— associated
with early-seral forest

Species

decline^ Species

% increase^

Del Norte salamander

75

Western toad

45

Black salamander

71

Pacific treefrog

160

Clouded salamander

29

Western fence lizard

60

Pacific shrew

39

Sagebrush lizard

44

Douglas' squirrel

31

Western skink

59

Northern flying squirrel

31

Southern alligator lizard

60

Dusky-footed woodrat

55

Northern alligator lizard

43

Western red-backed vole

37

Pinyon mouse

70

Fisher

26

California vole

44

Creeping vole

102

Gray fox

78

'Species were listed if thieir estimated future abundance differed by more ftian 25%
from estimated historical abundance and if mean abundance differed significantly (P
<0.10) among serai stages (table 2).

'Percent increase or decrease in estimated hjture abundance compared with
estimated historic abundance.

in each stage. Some species have (or
could) adapt to new stages over time.

Second, I assumed total acreage of
each serai stage can be used to esti-
mate responses of vertebrates with-
out regard to size and juxtaposition
of stands comprising each stage.
However, continued fragmentation
of forest habitats may result in dis-
junct patches so small they cannot
support a species that would other-
wise find the habitat suitable. Rosen-
berg and Raphael (1986) found that
at least eight species of amphibians
(2), reptiles (2), and mammals (4)
were significantly less abundant in
stands <10 ha in size than in larger
stands. Some of these (e.g., western
gray squirrel) were not listed in this
study among the sensitive species
(table 4), but the effects of habitat
fragmentation may nonetheless be
cause for concern.

A third assumption is that young
forested stands (pole, sawtimber) in
this study represent young stands of
the future. Naturally occurring pole
and sawtimber stands contain some
large Douglas-fir stems and a sub-
stantial amount of standing and
downed wood (table 1). If future
management activities result in fewer
large live trees, snags, and downed
logs, the abundance of vertebrates
associated with these habitat compo-
nents may also decline. In this case.

29

responses of vertebrates to forest
management may be more extreme
than those projected.

The overall trend is for increased
abundance among species of south-
ern affinity that are associated with
open, drier habitats in other parts of
their ranges, and decreased abun-
dance among species of boreal affin-
ity that are primarily associated with
moist coniferous forest throughout
their ranges. Furthermore, most of
the increasers are widespread species
with large distributions that include
many nonthreatened habitats. In con-
trast, the decreasers are almost all
species with rather restricted total
ranges, most of which are in threat-
ened habitats. Therefore, even
though total numbers of increasers
and decreasers are nearly equal, the
effects of old-growth reduction
should not be viewed as neutral.

Because many of the decreasers
are affected by soil moisture and
other microclimatic conditions, man-
agement to protect stream edges,
moist ravines, and other moist sites
may provide refuges for species that
can later recolonize maturing stands.
Management efforts to retain (or rec-
reate) natural components of regen-
erating stands, such as hardwood
understory, snags, and logs, may
help mitigate against wildlife losses
in future forests. It is not stand age,
per se, but the structural characteris-
tics of forests of various ages that are
important to survival of most spe-
cies.

Finally, results of this study ad-
dress another important forest man-
agement issue in the northwest;
What should managers use as a
baseline for evaluation of impacts:
historic or present conditions? It is
apparent that many species are pres-
ently much less abundant compared
with historic numbers (fig. 2). Addi-
tional reductions because of contin-
ued timber harvest will cause further
declines in some species but most
major declines have already oc-
curred. Therefore, I believe that esti-
mates of historic populations should

be used as baselines for monitoring
biological diversity, rather than pre-
sent populations.

ACKNOWLEDGMENTS

Field studies were funded by the Pa-
cific Southwest Region and the Pa-
cific Southwest Forest and Range
Experiment Station of the USDA For-
est Service and by the University of
California, Agricultural Experiment
Station 3501 MS. I especially thank
my field assistants (Paul Barrett, John
Brack, Cathy Brown, Christopher
Canaday, Lawrence Jones, Ronald
LaValley, Kenneth Rosenberg, and
Cathy Taylor) for their dedication
and blisters; R. H. Barrett, C. J.
Ralph, and J. Verner for their sup-
port; Bruce G. Marcot for freely shar-
ing information from his studies and
for valuable discussions; and Ken-
neth V. Rosenberg, Fred B. Samson,
and Hobart M. Smith for their com-
ments on an earlier draft of this
manuscript.

LITERATURE CITED

Barrett, Reginald H. 1983. Smoked
aluminum track plots for deter-
mining furbearer distribution and
abundance. California Fish and
Game 69:188-190.

Burgess, Robert L., and David M.
Sharpe. 1981. Forest island dy-
namics in man-dominated land-
scpates. Springer- Verlag, New
York. 310 p.

Brown, James K. 1974. Handbook for
inventorying downed woody ma-
terial. USDA Forest Service Gen-
eral Technical Report INT-16. 24 p.

Bury, R. Bruce, and Martin G. Ra-
phael. 1983. Inventory methods
for amphibians and reptiles, p.
426-419. In J. F. Bell and T. Atter-
bury (eds.). Renewable Resource
Inventories for Monitoring
Changes and Trends. College of
Forestry, Oregon State University,
Corvallis, Oregon.

Campbell, H. W., and S. P. Christ-
man. 1982. Field techniques for
herptofaunal community analysis,
p. 193-200. In N. J. Scott (ed.). Her-
petological Communities. USDI
Fish and Wildlife Service Wildlife
Research Paper 13. 239 p.

Frank Ernest C, and Richard Lee.
1966. Potential solar beam irradia-
tion on slopes. USDA Forest Serv-
ice Research Paper RM-18.

Harris, Larry D., Chris Maser, and
Arthur McKee. 1982. Patterns of
old growth harvest and implica-
tions for Cascades wildlife. Trans-
actions of North American Wild-
life and Natural Resource Confer-
ence 47:374-392.

Laudenslayer, William F., Jr., and
William E. Grenfell, Jr. 1983. A list
of amphibians, reptiles, birds and
mammals of California. Outdoor
California 44:5-14.

Marcot, Bruce G. 1984. Habitat rela-
tionships of birds and young-
growth Douglas-fir in northwest-
ern California. Corvallis, OR: Ore-
gon State University; 282 p. Ph.D.
dissertation.

Meslow, E. Charles, Chris Maser,
and Jared Verner. 1981. Old-
growth forests as wildlife habitat.
Transactions of North American
Wildlife and Natural Resource
Conference 46:329-344.

Raphael, Martin G. 1984. Wildlife di-
versity and abundance in relation
to stand age and area in Douglas-
fir forests of northwestern Califor-
nia, p. 259-274. In Meehan, W. R.,
T. T. Merrell, Jr., and T. A. Hanley
(tech. eds.). Fish and Wildlife Rela-
tionships in Old-growth Forests:
proceedings of a symposium (Jun-
eau, Alaska, 12-17 April 1982).
Bookmasters, Ashland, Ohio.

Raphael, Martin G. 1987. Wildlife
tanoak associations in Douglas-fir
forests of northwestern California,
p. 183-189. In Plumb, T. R., N. H.
Pillsbury, (tech. coord.). Proceed-
ings of the Symposium on Mul-
tiple-Use Management in Califor-
nia's Hardwood Resources; No-
vember 12-14, 1986, San Luis

30

Obispo, CA. General Technical
Report PSW-100. Berkeley, CA:
Pacific Southwest Forest and
Range Experiment Station, Forest
Service, U.S. Department of Agri-
culture, 462 p.
Raphael, Martin G., and Reginald H.
Barrett. 1981. Methodologies for a
comprehensive wildlife survey
and habitat analysis in old-growth
Douglas-fir forests. Gal-Neva
Wildlife 1981:106-121.
Raphael, Martin G., and Reginald H.
Barrett. 1984. Diversity and abun-
dance of wildlife in late succes-
sional Douglas-fir forests, p. 352-
360. In New Forests for a Chang-
ing World. Proceedings 1983 Con-
vention of the Society of American
Foresters. 650 p.
Raphael, Martin G., and Kenneth V.
Rosenberg. 1983. An integrated
approach to inventories of wildlife

in forested habitats, p. 219-222. In
J. F. Bell and T. Atterbury (eds.).
Proceedings, conference on renew-
able resource inventories for
monitoring changes in trends.
Corvallis, Oregon, 1983.

Raphael, Martin G., Kenneth V.
Rosenberg, and Bruce G. Marcot.
In press. Large-scale changes in
bird populations of Douglas-fir
forests, northwestern California.
Bird Conservation 3.

Raphael, Martin G., Cathy A. Taylor,
and Reginald H. Barrett. 1986.
Sooted aluminum track stations
record flying squirrel occurrence.
Pacific Southwest Forest and
Range Experiment Station Re-
search Note PSW-384.

Rosenberg, Kenneth V., and Martin
G. Raphael. 1986. Effects of forest
fragmentation on wildlife commu-
nities of Douglas-fir. p. 263-272. In

Appendix

Verner, J., M. L. Morrison, and C.
J. Ralph (eds.). Modeling habitat
relationships of terrestrial verte-
brates. University of Wisconsin
Press, Madison, WI.

Scott, J. Michael, Blair Csuti, James
D. Jacobi, and John E. Estes. 1987.
Species richness — a geographic ap-
proach to protecting future bio-
logical diversity. Bioscience
37:782-788.

Taylor, Cathy A., and Martin G. Ra-
phael. 1988. Identification of mam-
mal tracks from sooted track sta-
tions in the Pacific Northwest.
California Fish and Game 74:4-11.

Vogt, R. C, and R. L. Hine. 1982.
Evaluation of techniques for as-
sessment of amphibian and reptile
populations in Wisconsin, p. 201-
217. In N. J. Scott, (ed.). Herpeto-
logical Communities. USDI Fish
and Wildlife Service Research Re-
port 13, 239 p.

Common and scientific names of vertebrates mentioned in text (nomenclature follows Laudenslayer and

Grenfell (1983)).

Salamanders

Northwestern salamander Amhystoma gracile

Pacific giant salamander Dicamptodon ensatus

Olympic salamander Rhyacotriton olympicus

Rough-skinned newt Taricha granulosa

Del Norte salamander Plethodon elongatus

Ensatina Ensatina eschscholtzi

Black salamander Aneides flavipunctatus

Qouded salamander Aneides ferreus

Frogs and toads

TaUed frog Ascaphus truei

Western toad Bufo boreas

Pacific treefrog Hyla regilla

Foothill yellow-legged frog Rana boylei

Bullfrog Rana catesbeiana

Turtles

Western pond turtle Clemmys marmorala

Lizards

Western fence lizard Sceloporus occidentalis

Sagebrush lizard Sceloporus graciosus

Western skink Eumeces skiltonianus

Southern alligator lizard Gerrhonotus multicarinatus

Northern alligator lizard Gerrhonotus coeruleus

Snakes

Rubber boa Charina botlae

Ringneck snake Diadophis punctatus

Sharp-tailed snake Phyllorhynchus decurtatus

Racer Coluber constrictor

Gopher snake Pituophis melanoleucus

Common kingsnake Lampropeltis zonula

Common gartersnake Thamnophis sirtalis

Western terrestrial gartersnake Thamnophis elegans

Western rattlesnake Crotalis viridis

Mammals

Pacific shrew Sorex pacificus

Trowbridge's shrew Sorex trowbridgii

Shrew-mole Neurotrichus gibbsii

Coast mole Scapanus orarius

AUen's chipmunk Tamias senex

Western gray squirrel Sciurus griseus

Douglas' squirrel Tamiasciurus douglasii

Northern flying squirrel Glaucomys sabrinus

Deer mouse Peromyscus maniculatus

Brush mouse Peromyscus boylii

Pin yon mouse Peromyscus truei

Dusky-footed woodrat Neotoma fuscipes

Western red-backed vole Clethrionomys californicus

Red tree vole Arborimus longicaudus

California vole Microtus californicus

Creeping vole Microtus oregoni

Western jumping mouse Zapus princeps

Coyote Canis latrans

Gray fox Urocyon cinereoargenteus

Black bear Ursus americanus

Ringtail Bassanscus astutus

Raccoon Procyon lotor

Fisher Martes pennanti

Ermine Mustela erminea

Western spotted skunk Spologale gracilis

Striped skunk Mephitis mephitis

Bobcat Lynx rufus

31

Use of Woody Debris by
Plethodontid Salamanders In
Douglas-Fir Forests In
Washington

Keith B. Aubry,^ Lawrence L C. Jones,^ and
Patricia A. Hali^

Abstract.— Ensaf/no eschscholfzii \f^os found most
often under pieces of bark, whereas Plefhodon
vehiculum occurred primarily under logs. Captures
of both species were highest in young stands, but
occurred in all age classes. Our results suggest that
the retention of coarse woody debris in managed
forests would provide for the habitat needs of these
species.

The harvesting of old-growth
Douglas-fir (Pseudotsuga menziesii)
forests in the Pacific Northwest, and
its potential effects on wildlife spe-
cies has been the focus of much con-
cern in recent years (e.g.. Lumen and
Nietro 1980, Franklin et al. 1981,
Meslow et al. 1981, Meehan et al.
1984, Gutierrez and Carey 1985).
Most of this attention has been di-
rected towards birds and mammals
such as the spotted owl (Strix occiden-
talis), Vaux's swift (Chaetura vauxi),
northern flying squirrel (Glaucomys
sahrinus), and red ti'ee vole (Ar-
borimus longicaudus); little concern
has been expressed about amphibi-
ans and reptiles. These groups have
not been studied extensively in the
Pacific Northwest. Only recently has
research been conducted on habitat
associations among different forest
age classes (Raphael 1984, Raphael
and Barrett 1984, Ruggiero and
Carey 1984).

From 1983 to 1986, the USDA For-
est Service and USDI Bureau of Land

'Paper presented at symposium. Man-
agement of Amphibians. Reptiles, and
Small Mammals in Nortti America. (Flag-
staff, AZ, July 19-21, 1988).

^Researcti Wildlife Biologist, USDA Forest
Service, Pacific Nortt)west Research) Station,
3625 93rd Ave. SW, Olympia, WA 98502.

^Biological Technician, USDA Forest Serv-
ice, Pacific Northwest Research Station,
3625 93rd Ave. SW, Olympia, WA 98502.

"Wildlife Biologist, USDA Forest Service,
Pacific Northwest Research Station, 3625
93rd Ave. SW, Olympia, WA 98502.

Management funded a major re-
search effort aimed at identifying
wildlife species that occur in highest
abundances in old-growth Douglas-
fir forests and investigating the eco-
logical basis of observed patterns of
association

Amphibian communities were
sampled using pitfall traps, stream
surveys, and time-constrained
searches (Standard Sampling Proto-
cols on file at the Forestry Sciences
Laboratory, Olympia, WA). Some of
the results of these studies are re-
ported elsewhere in this volume
(Bury and Corn 1988, Welsh 1988).
Here, we report the results of time-
constrained searches conducted in
southern Washington in 1984. Our
objectives are to (1) identify potential
habitat associations, (2) examine pat-
terns of cover object use, and (3)
evaluate the efficacy of this technique
for studying amphibians in this re-
gion.

Study Area

Forty-five forest stands were
sampled in the southern portion of
the Cascade Range in Washington
(fig. 1). Stands ranged in age from 55
to 730 yr and were at least 20 ha in
size. All stands were located within
the western hemlock (Tsuga heter-
phylla) zone and lower elevations of
the Pacific silver fir (Abies amabilis)
zone (Franklin and Dyrness 1973),
which are characterized by a wet and

Figure 1 .—Location of study stands by age
class in the southern Washington Cascade
Range.

mild maritime climate. Snow rarely
accumulates at our sites.

Old-grovv^th stands (210-730 yr)
typically contained high proportions
of Douglas-fir and western hemlock
and, in wet sites, western redcedar
(Thuja plicata). Mature (95-190 yr)
and young (55-80 yr) stands were

32

dominated by Douglas-fir. In all age
classes, other species such as red
alder (Alnus rubra), vine maple (Acer
circinatum), bigleaf maple (A. macro-
phyllum), Pacific silver fir, and west-
ern hemlock occurred in lesser
amounts.

Average age of each stand was
determined through growth ring
counts, either by increment coring or
examination of cut stumps in nearby
stands. Old-growth stands were clas-
sified into wet, moderate, and dry
moisture classes on the basis of flo-
ristic and physiographic characteris-
tics; all young and mature stands
were in the moderate moisture class
(T. A. Spies, unpubl. data). All stands
had resulted from natural regenera-
tion following fires; none had under-
gone silvicultural treatments.

Methods

Surveys for terrestrial amphibians
were conducted from 16 April to 12
June 1984; all but four high-elevation
stands were sampled by 4 May. A
time- constrained search method was
used (Campbell and Christman
1982). A crew of two to four persons
actively searched each stand for am-
phibians for a total of 4 person-
hours. An initial search area was se-
lected at least 50 m within the stand
to avoid edge effects.

In general, woody debris such as
logs, snags, and pieces of bark was
abundant in each stand and consti-
tuted virtually all potential cover ob-
jects. An area was searched for 0.5
person-hours, after which we moved
a minimum of 25 m to search another
suitable area; sampling areas were
not spatially constrained. This was
repeated until the sampling period
was over. All potential cover objects
were searched by hand or with po-
tato rakes, but no single object was
searched for more than 20 min. Logs
of all sizes in advanced stages of de-
composition were pulled apart with
potato rakes. Areas beneath large
undecomposed logs could not be

Table 1.— Amphibian species captured during time-constrained searches
in the southern Washington Cascade Range by stand type.'

Species

Mean Captures ± Standard Error

(N=9) <N=9) (N=6) (N=17) (N=4)
YNG MAT OGW OGM OGD

Caudata

Plethodontidae

Ensafina eschscholfzh

Plefhodon vehiculum
Ambystomatidae

Ambysfoma gracile

A. macrodacfylum
Salamadridae

Taricha granulosa
Dicamptodontidae

Rhyacofrifon olympicus
Anura

Leiopelmatidae

Ascaphus fruei
Ranidae

Rana aurora

R. cascadae

5.9±1.8 2,6±0.8 1.2±1.2 2.8±0.9 2.5+1.0

3.1±2,2 Q.4±0.3 0.2±0.2 0.5+0.3 2.3±2.3

0.3±0.1
0.2+0.2 0.1+0.1

0.1+0.0

0.2+0,1

0.1+0.1

0.1±0.1 0.5+0.5
0.1+0.1

0.2+0.1

0.1+0.1

'YNG^Young. MAT=Mafure. OGW=Wef Old Growth. OGtv1=Moderafe Old Growth.
OGD=Dry Od Growth

searched in most cases. Little effort
was expended searching leaf litter, as
this has been shown to be relatively
ineffective when sampling amphibi-
ans in Douglas-fir forests (Bury and
Raphael 1983). Areas near seeps,
streams, ponds, rock outcrops, and
other areas not representative of the
stand were avoided.

Modifications of methods devel-
oped by Raphael (1984) were used to
describe capture sites. The following
information was recorded for each
individual captured: species, vertical
position in relation to cover object,
snag or log decay class, length and
width of cover object, and slope and
aspect of capture site. All amphibians
were collected, measured, and pre-
served, usually on the same day.
Snout- vent length (to anterior margin
of vent), total length, and weight
were recorded. Specimens were de-
posited in the Museum of Vertebrate
Zoology, University of California,
Berkeley.

Results

Captures

A total of 214 amphibians, including
6 species of salamanders and 3 spe-
cies of frogs, were captured; no rep-
tiles were encountered (table 1). Only
two species of plethodontid salaman-
ders, the ensatina (Ensatina es-
chscholtzii) and western redback
salamander (Plethodon vehiculum),
were captured in sufficient numbers
(141 and 50, respectively) to permit
comparisons of abundance among
stand types or to conduct analyses of
cover object use.

Habitat Occupancy

Ensatinas and redback salamanders
occurred in all forest age and mois-
ture classes. Although both species

33

were most abundant in young forests
(table 1), a one-way AN OVA re-
vealed no significant differences
among stand types. Mean captures
for both species were lowest in wet
old-growth stands. The proportion of
stands containing ensatinas was also
relatively low in wet old growth:
fewer than 20% of wet old -growth
stands sampled contained ensatinas,
whereas all other stand types had a
frequency of occurrence of 65% or
greater (fig. 2). The proportion of
stands containing redback salaman-
ders was generally low in all stand
types (fig. 2), suggesting that at the
time of our sampling, redback sala-
manders were less abundant or more
clumped in distribution than ensati-
nas. We found no amphibians in 67%
of old-growth wet stands, 11% of
young and mature stands, 12% of
moderate old-growth stands, and 0%
of dry old-growth stands.

\ \ \ \

\ \ \ \ \ \
\ \ \

Use of Woody Debris

Cover object selection varied be-
tween the two species. Ensatinas

O.Bt

OGW
(N=6)

STAND TYPE

Figure 2.— Proportion of stands in each stand type with captures of Ensatina eschscholtzii
(ENES) and Plethodon vehiculum (PLVE) in the southern Washington Cascades Range. Stand
type YNG= Young. MAT=Mature, OGW=Wet Old Growth, OGM=Moderate Old Growth,
OGD=Dry Old Growth.

Figure 3.— Use of cover objects by Ensatina eschscholtzii (ENES) and Plethodon vehiculum
(PLVE) in the southern Washington Cascade Range.

were most often found under pieces
of bark (generally within 1 m of a
snag or log) and secondarily under
logs (fig. 3). The pattern was re-
versed for redback salamanders. Nei-
ther species was found under bark
on snags. When found under pieces
of bark, ensatinas most often oc-
curred in bark piles at the base of
moderately decayed snags (see Tho-
mas et al. 1979, p. 64). Seventy-four
percent of these captures occurred
next to snags in which the top had
broken off, the wood was soft, and
most or all of the bark had sloughed
onto the ground. Logs where ensati-
nas and redback salamanders were
captured were most often 10- 30 cm
in diameter (fig. 4). Both species were
captured in low numbers in associa-
tion with very large logs (diameter
>30 cm), but our inability to ade-
quately search this cover type may
account for these results. Virtually all
logs where ensatinas and redback

34

salamanders were found were in in-
termediate stages of decay (fig. 5)
(see Maser et al. 1979, p. 80). Only a
few captures of either species oc-
curred in association with intact or
extensively decomposed logs. Nei-
ther species was commonly found
under rocks, but this cover type is
relatively rare in Douglas-fir forests.
No correlations between slope or as-
pect and amphibian capture sites
could be detected.

Discussion

Old-growth forests do not appear to
provide unique habitat for either en-
satinas or western redback salaman-
ders; both species were well-repre-
sented in all age classes. Our results
suggest that abundance levels of
these salamanders are more likely a
function of the availability of woody
debris for cover than age of the over-
story. Wet old-growth stands in
southern Washington, however, ap-
parently provide low quality habitat
for these plethodontids, especially
ensatinas (table 1, fig. 2). Soils in
these stands were often saturated

INTACT

MODERATELY DECOMPOSED
DECAY CLASS

DECOI^OSED

Figure 5.— Use of logs by Ensatina esc/)scho//z//(E NES) and Plethodon vehiculum (PLVE) by
decay class In the southern Washington Cascade Range.

with water, and such conditions may
reduce the availability of microenvi-
ronments suitable for cover, mainte-
nance of water balance, and success-
ful reproduction. In addition, these

< 10 CM
FINE WOODY DEBRIS

10 -30 CM > 30 CM

COARSE WOODY DEBRIS

Figure 4.— Use of logs by Ensatina eschscholtzii i£N£S) and Plethodon vehiculum (PLVE) by
dianneter class in the southern Washington Cascade Range.

Stands were located in topographi-
cally low sites where cold air accu-
mulates, which may create unfavor-
able microclimatic conditions for ple-
thodonhd salamanders. Our results
also suggest that plethodontid sala-
manders may prefer certain types of
woody debris as cover, especially
those associated with large, moder-
ately to well-decomposed snags and
logs. Captures of ensatinas were
most common under pieces of bark,
especially in bark piles at the base of
well- decayed snags (fig. 3). Snags in
the early stages of decomposition
with shallow or no bark piles at their
bases provide few suitable mi-
crohabitats for salamanders. Depth
of these bark piles increases as
sloughing continues until all bark has
fallen off. Later stages of snag de-
composition provide no additional
bark to the pile and habitable spaces
become compressed as the lower lay-
ers of bark decay and mix with the
underlying substrate.

Bark microhabitats formed by the
deterioration of snags differ in struc-
ture from those formed by the de-

35

composition of logs. As logs decay, a
single layer of bark is deposited on
the forest floor, whereas bark slough-
ing from snags forms multilayered,
structurally complex cover. Such
bark piles could provide microcli-
matic conditions more resistant to
fluctuations in temperature and
moisture than those found under
bark on the ground. Additional for-
aging habitat may also be available.

Redback salamanders, on the
other hand, were most often found
under moderately decayed logs 10-30
cm in diameter (figs. 3-5). In the early
stages of decay, bark has not begun
to slough and branches suspend the
log above the ground. As the bark
begins to slough and branches dete-
riorate, increased cover and moisture
are provided along the length of the
bole where it comes in contact with
the forest floor (Maser and Trappe
1984). The quality of this environ-
ment for salamanders continues to
improve with further decay until the
organic matter becomes incorporated
into the underlying substrate and
habitable interstices become com-
pressed in the advanced stages of
decomposition.

All known nest sites of ensatinas
in the Pacific Northwest have been
found in association with large, mod-
erately decayed logs (Norman and
Norman 1980, Maser and Trappe
1984, Jones and Aubry 1985, Norman
1986, L. L. C. Jones unpubl. data).
This habitat feature may be impor-
tant for the persistence of ensatinas
in these forests. We do not know to
what extent coarse woody debris
may be important for reproduction
of redback salamanders in Douglas-
fir forests; only one nest site has been
found, and this was in moist talus in
the Oregon Coast Range (Hanlin et
al. 1978).

In Douglas-fir stands of the Cas-
cade Range that have regenerated
after catastrophic fires, levels of
coarse woody debris (CWD) (logs
and snags > 10 cm in diameter) are
moderate in young stands, lowest in
mature stands, and highest in old-

growth stands (Spies et al. in press).
In general, this is due to the inheri-
tance of high levels of CWD in young
stands from the preceding old-
growth stands, a low accumulation
of CWD in mature stands as CWD
decays but inputs are low, and high
inputs of CWD in older stands as the
large Douglas-firs die and accumu-
late as snags and logs. Intensive for-
est management results in levels of
CWD substantially lower than that
encountered in unmanaged forests
(Spies and Cline in press). This is be-
cause plantations inherit little CWD
from the preceding stand when it is
clearcut and existing CWD is re-
moved and fragmented. In addition,
thinning operations reduce the input
of CWD from suppression mortality
and short rotations prevent the accu-
mulation of CWD. Maintaining even
moderate amounts of CWD in man-
aged forests will require modifica-
tions of current harvesting and
silvicultural practices (Harmon et al.
1986, Spies et al. in press).

Virtually all available cover ob-
jects we encountered were woody
debris, and both species were found
most often in association with large,
moderately decayed logs and snags.
Our results suggest that the availabil-
ity of coarse woody debris may be
important for maintaining salaman-
der populations in Douglas-fir for-
ests. Additional studies of terrestrial
salamanders in managed vs. unman-
aged forests are necessary to deter-
mine the extent to which they may be
affected by intensive forest manage-
ment.

In general, our study yielded a
relatively low number of captures.
Only two common species (Nuss-
baum et al. 1983) were captured in
high enough numbers to permit
analyses of the data; captures of all
other species were incidental. The
total number of species detected was
also low in relation to known species
richness: pitfall trapping for approxi-
mately 1000 trap nights in each of the
same study sites in the fall of 1984
yielded 916 captures of 13 species (K.

B. Aubry unpubl. data). Research us-
ing time-constrained searches to
study all but the most common spe-
cies in this region would require sub-
stantially more search time. Sam-
pling should also be conducted dur-
ing all seasons of the year to detect
seasonal shifts in habitat selection or
cover object use, and to sample spe-
cies that are active at other times of
the year.

Acknowledgements

We thank R. W. Lundquist, J. B.
Buchanan, B. A. Schrader, A. B.
Humphrey, M. Q. Affolter, M. J.
Reed, B. F. Aubry, and M. J. Crites
for assistance. T. A. Spies at the For-
estry Sciences Laboratory, Corvallis,
OR provided data on stand charac-
teristics. R. W. Lundquist provided
the map used in figure 1 . This study
was funded under USDA Forest
Service Cooperative Agreement
PNW-83- 219 to S. D. West and D. A.
Manuwal at the Univ. of Washing-
ton. We thank personnel of the Gif-
ford Pinchot National Forest and
Mount Rainier National Park for
their cooperation and support. M. G.
Raphael, K. E. Severson, T. A. Spies,
and A. B. Carey provided construc-
tive comments on a previous draft of
the manuscript. This paper is Contri-
bution No. 65 of the Old-growth For-
est Wildlife Habitat Project, USDA
Forest Service, Pacific Northwest Re-
search Station, Olympia, WA.

Literature Cited

Bury, R. B. and P. S. Corn. 1988.
Douglas-fir forests in the Oregon
and Washington Cascades: rela-
tionship of herpetofauna to stand
age and moisture. This volume.

Bury, R. B. and M. G. Raphael. 1983.
Inventory methods for amphibians
and reptiles, p. 416-419. In]. F. Bell
and T. Atterbury, eds. Renewable
resource inventories for monitor-
ing changes and trends: Proceed-

36

ings of an international conference
(Corvallis, Oregon, August 15-19,
1983) Society of American Forest-
ers 83-14). 737 p.
Campbell, H. W. and S. P. Christ-
man. 1982. Field techniques for
herpetofaunal community analy-
sis, p. 193-200. In N. J. Scott, Jr.,
ed. Herpetological communities.
USDI Fish and Wildlife Service
Wildlife Research Report 13.
239 p.

Franklin, J. F., K. Cromack, Jr., W.
Denison, A. McKee, C. Maser, J.
Sedell, F. Swanson, and G. Juday.
1981. Ecological characteristics of
old-growth Douglas-fir forests.
USDA Forest Service General
Technical Report PNW-118, 48 p.

Franklin, J. F. and C. T. Dyrness.
1973. Natural vegetation of Ore-
gon and Washington. USDA For-
est Service General Technical Re-
port PNW-8, 417 p.

Gutierrez, R. J and A. B. Carey, eds.
1985. Ecology and management of
the spotted owl in the Pacific
Northwest. USDA Forest Service
General Technical Report PNW-
184. 119 p.

Hanlin, H. G., J. J. Beatty, and S. W.
Hanlin. 1978. A nest site of the
western red-backed salamander
Plethodon vehiculum. Journal of
Herpetology 13(2):212-214.

Harmon, M. E., J. F. Franklin, F. J.
Swanson, P. Sollins, S. V. Gregory,
J. D. Lattin, N. H. Anderson, S. P.
Cline, N. G. Aumen, J. R. Sedell,
G. W. Lienkaemper, K. Cromack,
Jr., and K. W. Cummins. 1986.
Ecology of coarse woody debris in
temperate ecosystems. In A. Mac-
Fadyen and E. D. Ford, eds. Ad-
vances in Ecological Research
15:133-302.

Jones, L. L. C. and K. B. Aubry. 1985.
Ensatina eschscholtzii (Oregon en-
satina). Reproduction. Herpeto-
logical Review 16(1):26.

Lumen, I. D. and W. A. Neitro. 1980.
Preservation of mature forest serai
stages to provide wildlife habitat
diversity. Transactions of the
North American Wildlife and

Natural Resources Conference
43:78-88.

Maser, C, R. G. Anderson, K.

Cromack, Jr., J. T. Williams, and R.
E. Martin. 1979. Dead and down
woody material, p. 78-95. In J. W.
Thomas, ed. Wildlife habitats in
managed forests- the Blue Moun-
tains of Oregon and Washington.
USDA Forest Service Agricultural
Handbook 553, 512 p.

Maser, C. and J. M. Trappe. 1984.
The seen and unseen world of the
fallen tree. USDA Forest Service
General Technical Report PNW-
164, 56 p.

Meehan, W. R., G. R. Merrell, Jr., and
T. A. Hanley, eds. 1984. Fish and
wildlife relationships in old-
growth forests: Proceedings of a
symposium. (American Institute
of Fishery Research Biologists,
Juneau, Alaska, April 1982). 425 p.

Meslow, E. C, C. Maser, and J.
Vemer. 1981. Old-growth forests
as wildlife habitat. Transactions of
the North American Wildlife and
Natural Resources Conference
46:329-335.

Norman, B. R. 1986. Ensatina es-
chscholtzii oregonensis (Oregon en-
satina). Reproduction. Herpeto-
logical Review 17(4):89.

Norman, C. E. and B. R. Norman.
1980. Notes on the egg clusters
and hatchlings of Ensatina es-
chscholtzi oregonensis. Bulletin of
the Chicago Herpetological Soci-
ety 15(4):99-100.

Nussbaum, R. A., E. D. Brodie, Jr.,
and R. M. Storm. 1983. Amphibi-
ans and reptiles of the Pacific
Northwest. 332 p. University Press
of Idaho, Moscow.

Raphael, M. G. 1984. Wildlife diver-
sity and abundance in relation to
stand age and area in Douglas-fir
forests of northwestern California,
p. 259-274. In W. R. Meehan, T. R.
Merrell, Jr., and T. A. Hanley, eds.
Fish and wildlife relationships in
old-growth forests: Proceedings of
a symposium. (American Institute
of Fishery Research Biologists,
Juneau, Alaska, April 1982). 419 p.

Raphael, M. G. and R. H. Barrett.
1984. Diversity and abundance of
wildlife in late successional
Douglas-fir forests, p. 352-360. In
New Forests for a changing world.
(Society of American Foresters
Convention, Portland, Oregon,
1983).

Ruggiero, L. F. and A. B. Carey. 1984.
A programmatic approach to the
study of old-growth forest - wild-
life relationships, p. 340-345. In
New forests for a changing world.
(Society of American Foresters
Convention, Portland, Oregon,
1983).

Spies, T. A. and S. P. Cline. In Press.
Coarse woody debris in manipu-
lated and unmanipulated coastal
Oregon forests. In C. Maser, R. F.
Tarrant, J. M. Trappe, and J. F.
Franklin, eds. From the forest to
the ocean — a story of fallen trees.
USDA Forest Service General
Technical Rep>ort.

Spies, T. A., J. F. Franklin, and T. B.
Thomas. In Press. Coarse woody
debris in Douglas-fir forests of
western Oregon and Washington.
Ecology. Thomas, J. W., R. G. An-
derson, C. Maser, and E. L. Bull.
1979. Snags. In J. W. Thomas, ed.
Wildlife habitats in managed for-
ests - the Blue Mountains of Ore-
gon and Washington. USDA For-
est Service Agricultural Handbook
553. 512 p.

Welsh, H. H., Jr. 1988. Old growth
forests and the distribution of the
terrestrial herpetofauna. This vol-
ume.

37

Forestry Operations and
Terrestrial Salamanders:
Techniques in a Study of the
Cow Knob Salamander,
Plethodon punctatus^

Kurt A. Buhlmann,2 Christopher A. Pague,^
Joseph C. Mitchell,^ and Robert B. Glasgow^

Abstract.— The status and ecology of Plethodon
punctaf us \^/as investigated in George Washington
National Forest, Virginia to determine potential
effects of logging. Pitfall traps and mark-recapture
supplemented searching by hand. Elevation,
aspect, soil characteristics, and number of cover
objects (rocks) ore the most important features that
identify P. panc^ofus habitat. Intensive logging
operations appear to be detrimental to this species.

Increasing emphasis is being placed
on conservation and preservation of
biological diversity worldwide
(Norse et al., 1986; Wilson, 1988).
U.S. federal and state agencies have
become concerned about the bio-
diversity of their managed lands and
are directing efforts towards preserv-
ing natural biota. From a manage-
ment perspective, research on am-
phibians and reptiles lags behind that
devoted to game animals, such as
some mammals, birds, and fish (Bury
et al., 1980). This is partly due to a
previous lack of interest in these
groups, but also because some spe-
cies can be more difficult to observe
or investigate.

The Cow Knob salamander, Ple-
thodon punctatus, is a dark, moder-
ately large (to 74 mm snout-vent
length), woodland, fossorial amphib-
ian (Martof et al., 1982) found only

'Paper presented at symposium. Man-
agement of Amphibians, Reptiles, and
Small Mammals in Nortti America. (Flag-
staff, AZ, July 19-21, 1988).

^ Kurt A. Buhlmann is a consulting biolo-
gist for tlie U.S. Forest Service. Buhlmann's
current address is 2001 N. Main St.,
Blacksburg, VA 24060.

^Christophier A. Pague is a doctoral can-
didate in ttie Ecological Sciences Program,
Department of Biological Sciences, Old
Dominion University, Norfolk, VA 23508.

^Josept) C. Mitchiell is a Research) Biolo-
gist, Department of Biology, University of
Richmond, Richmond, VA 23173.

^Robert B. Glasgow is the Wildlife Biolo-
gist for the George Washington National
Forest, Harrison Plaza, P.O. Box 233, Harri-
sonburg, VA 22801.

on Shenandoah and North Mountain
of western Virginia and eastern West
Virginia (Highton, 1972, Tobey,
1985). Most of the known range of
this recently described species
(Highton, 1972) is in the George
Washington National Forest. Fraser
(1976) compared some aspects of the
ecology of this species with a sympa-
tric congener, Plethodon hoffmani.
Little else is known of the ecology of
this salamander. Because of its rela-
tively small range and unknown
status, P. punctatus was added to the
U.S. Fish and Wildhfe Service's Cate-
gory 2 list (U.S. Fish and Wildlife
Service, 1985). Potential timber har-
vesting within the range of this spe-
cies (USD A Forest Service, 1986)
prompted us to examine its status in
forest stands of various ages. In this
paper we report the following as-
pects of this study: techniques of cap-
ture and data collection, salamander
habitat characteristics, and potential
effects of logging operations. Our ob-
jective in this paper is to make other
researchers aware of the techniques
we used and the problems we en-
countered in developing useful man-
agement recommendations for the
protection of an apparently rare ter-
restrial salamander.

Materials and Methiods

We conducted this study on Shenan-
doah Mountain, Augusta and Rock-
ingham Counties, George Washing-

ton National Forest, Virginia. Before
its purchase, between 1911 and 1940,
by the U.S. government, this area
was repeatedly logged and burned
(Leichter, 1987; original land deed
documents). Few virgin stands of
forest remain, and regrowth and log-
ging operations has resulted in a mo-
saic of mixed hardwoods of various
ages.

We selected five sites of different
aged forest to determine the relative
abundance of Plethodon punctatus (fig.
1) to see if its presence was affected
by logging. All sites selected had
similar aspects (S-SE) and elevation
(914- 1127 m) (table 1). We used
USDA Forest Service compartment
descriptions and maps to aid in se-
lection of sites and to obtain informa-
tion on the history, physical and bio-
logical descriptions, and future man-
agement goals for each site. A com-
partment is divided into a series of
stands, each of which defines a for-
ested area of similar tree species by
composition, age, and stand condi-
tion. Stand age is defined by the age
of dominant canopy trees. Final
choices of sites were made only after
each was checked in the field and
tree age was verified by tree ring
analysis.

In each site we erected drift fence
arrays (Campbell and Christman,
1982) consisting of four 60 cm x 7.5 m
secHons of aluminum flashing ar-
ranged in a cross pattern. Opposite
arms of the cross were separated by
15 m and all sections were sunk in

38

the ground approximately 10 cm. A
5-gal plastic bucket was placed in the
center of each arm and #10 cans were
placed in the ground on either side of
the ends of each arm so that the tops
of the pitfalls were flush with the
ground surface. Sites A and C con-
tained two drift fence arrays, and the
remaining three sites had one array
each. In each pitfall we put 4-10 cm
of 10% formal-dehyde solution to in-
sure adequate preservation of the
salamanders. We selected this
method to obtain samples of all the
terrestrial fauna for a range of stud-

ies on reproductive cycles and ecol-
ogy. Pitfalls were checked and all
captures (including other vertebrates
and all invertebrates) were collected
weekly May 5 - June 18, 1987, bi-
weekly July 7 - November 22, 1987
and monthly December 1987 and
January 1988. Samples were sorted in
the laboratory and the vertebrates
stored in 10% neutral buffered for-
maldehyde. Invertebrates were
stored in 70% isopropanol.

Hand-collecting supplemented
drift fence collection and was used to
determine the elevational range of P.

Figure 1 .—Plethodon punctatus, the Cow Knob salamander, from Shenandoah Mountain,
Augusta County, Virginia. Photograph by Kurt A. Buhlmann.

punctatus and to obtain information
on range and habitat characteristics.
Results from timed collecting periods
allowed comparison among sites and
dates of collection. Between April 20
and June 2, 1987, we collected data
on eleven microhabitat variables at
67 sites to evaluate those most im-
portant in predicting the presence of
this salamander. These variables
were elevation, aspect, slope, soil
temperature under cover object, soil
moisture, soil pH, soil description,
canopy cover, number of cover ob-
jects available within a 2 m circle,
type of cover object (e.g., rock, log),
and forest type.

One site >1 km away from any of
the collection sites was selected for
estimation of population size and
data on individual movements. We
searched for salamanders in daylight
by turning and replacing all surface
objects and at night while they were
active on the surface (i.e., during
conditions of near 100% relative hu-
midity [sensu Heatwole, 1960; Jaeger
1978]). Each individual was meas-
ured (snout-vent length, tail length to
nearest mm), weighed (nearest 0.1 g),
the sex determined, assigned to adult
or juvenile age-classes, marked by
toe-clipping, and released at its cap-
ture site. We marked each capture
site with survey flags on which the
salamander's number and capture

Table 1.— Descriptions of drift fence study sites for Plethodon punctatus on ShenarKloah Mountain, Virginia. Slope
angle is in degrees and site age is in years since last logging activity.

Site

Timber descrip.

Slope

Manag. type

Site age

Stand condition

Past logging history

A

1 yr old white pine
several hardwood
seed trees

30

white pine

2

seedling/
sapling

90% clearcut
few hardwood trees

B

white oak/
red oak/hickory

. 45

oak/hickory

8

sparse
saw timber

thinned due to ice
damage, 1979

C

white pine/mixed

25

white pine

30

immature

cut in 1956, planted in

hardwoods

pole timber

white pine, some hardwood

seed trees

D

white oak/hickory

30

oak/hickory

60-100

mature saw
timber

no recent management

E

white oak/
red oak/ hickory

5

none

virgin?

low quality
saw timber

none known

39

date were written. We noted all re-
captures and measured movements
in linear fashion (0.1 m) between cap-
ture points.

Results and Discussion

Capture Tectiniques

Nineteen P. punctatus were caught in
the pitfall traps, 2.0% of the total
number of salamanders. Of the 17
recorded, 12 were caught in 5-gal.
buckets and 5 in #10 cans. Eleven P.
punctatus were caught in Site E, six in
Site B, and two Site D. None were
caught in Site A or the Site C. In con-
trast, by hand collecting in areas ad-
jacent to Site E, we found 38 P. punc-
tatus in 7.7 man hours of searching.
The drift fence method appears only
moderately effective in sampling this
salamander. It is feasible that P. punc-
tatus is less likely to fall into the pit-
falls than other salamander species.
We observed several individuals
climbing rocks and tree trunks dur-
ing nocturnal surface activity. This
suggests that this salamander is able
to detect precipices and avoid falling
into pitfalls. Also, this species may be
active on the horizontal surface only
for limited periods of time and under
specific environmental conditions.
Thus, the drift fence technique,
which depends on horizontal activ-
ity, may not be an effective sampling
method for this salamander (R.D.
Semlitsch, pers. comm.).

Data from pitfall traps, combined
with data from hand collecting, can
provide information for management
decisions. For instance, seasonal
trends in surface activity were simi-
larly indicated by both drift fence re-
sults (fig. 2) and captures based on
hand collection (fig. 3). Comparison
of P. punctatus with that of its sympa-
tric congener P. cinereus (fig. 3) re-
veals concordance in seasonal activ-
ity and suggests similar responses to
surface environmental conditions.
This information could be used to
determine the times logging opera-

MAY JUN JUL AUG SEP OCT NOV DEC JAN

Figure 2. —Seasonality of drift fence captures of Plethodon cinereus and P. punctatus at Site E
(Tomat>awl< Mountain), George Washington National Forest. Adults and juveniles are in-
cluded, but not tiatctilings. Sampling period is 5 May 1987 to 24 January 1988.

20

APR 22 MAY 5 MAY 24 JUL 6 JUL 21 AUG 31 SEP 28 OCT 12 NOV 22

Figure 3.— Seasonality of captures per man hiour of Plett)odon cinereus and P. punctatus on
Tomatiawk Mountain, George Wastiington National Forest. Black bars represent P. punctatus
and bars witli diagonal lines represent P. cinereus. Sampling dates are 22 April to 22 Novem-
ber 1987.

40

tions would cause the least impact on
salamanders at or near the surface.

The benefits of the drift fence tech-
nique outweighed the low numbers
of captures of P. punctatus. We
probably would not have otherwise
found this species in Site D because
there were few surface rocks to turn
over. Although the individuals
caught may have been transients, this
species does occasionally occur at
this site. This result would not have
been obtained by hand-collecting
alone.

The drift fence method also pro-
vided estimates of the relative abun-
dance of the salamander fauna and
other species in the community. The
relative numbers of these species and
species groups can generate addi-
tional information on the structure of
the community in which the focal
species lives. Drift fence techniques
have been used for a variety of eco-
logical studies (e.g.. Gibbons, 1970;
Gill, 1978; Pechmann and Semlitsch,
1986) but only recently to answer
questions about vertebrate communi-
ties in relation to forest management
(e.g., Bennett et al., 1980; Gibbons
and Semlitsch, 1981; Enge and Mar-

Table 2.— Seasonal differences in
surface abundance of Plethodon
punctatus at Flagpole Knob and
Skidmore Tract, Shenandoah
f^ountain, George Washington For-
est. These sites are <1 km apart.
Flagpole Knob is a rocky, grassy
ridge habitat containing young
oak (Quercus sp.) and maple
(Acer sp.) pole timber, and Skid-
more is a virgin hemlock (Tsuga
canadensis)/ye\\ow birch (Betula
lutea) forest. Numbers of salaman-
ders are followed by number of
man hours In parentheses. All data
are based on hand-collecting re-
sults.

Date Flagpole Skidmore

June 2 11 (0,5) 10 (1.7)

June 8 0 (0.5) 2 (2.0)

Sept. 28 0 (1.0) 3 (3.0)

V J

ion, 1986; Bury and Com, 1987). Our
results indicate this technique can be
effective in mountainous terrain and
can be used to gain information on
apparently rare terrestrial salaman-
ders.

If an endangered or otherwise
protected species is the focus of
study and cannot be collected, then
slight modifications of the drift
fence-pitfall design must be made.
Traps would need to be checked on a
daily basis, or nearly so, in order to
release the animals unharmed (Gib-
bons and Semlitsch, 1981). Water or
wet leaves can be placed in the pit-
falls for cover and moisture. Poten-
tial problems include killing of the
salamanders in the pitfalls by small
mammals, especially shrews, and
desiccation. The loss of animals by
shrew or raccoon predation in pitfall
containers affects the samples and
may prevent quantitative compari-
sons among sites. Data obtained
from visitation frequencies of every
three days (Bury and Corn, 1987) to
once a week (Enge and Marion, 1986)
probably underestimate actual cap-
tures.

The detection of P. punctatus at a
particular site depends on the time of
year, substrate type, soil depth, soil
moisture, soil temperature, and
weather conditions (see Habitat Re-
quirements). A simple survey of sites
by hand searching and rock turning
in daylight hours without attention
to weather and seasonality will un-
derestimate actual abundance and
fail to detect presence of a species.
Table 2 contains comparative data
for two sites searched the same day
at different times of the year and
demonstrates a strong seasonal ef-
fect. In order to construct effective
management plans, the range and
abundance of a terrestrial salaman-
der must be known. Therefore, re-
searchers conducting distributional
surveys must take seasonal and diet
changes in surface activity into con-
sideration.

Results of our 1987 mark-recap-
ture efforts are preliminary; only

four recaptures were made. One P.
punctatus captured 28 May was re-
captured on 15 October. It had
moved 17.4 m. Three salamanders
were recaptured within ten days of
original capture and had moved < 2
m. Knowledge of movement capabili-
ties by P. punctatus is an important
part of evaluating the consequences
of population fragmentation through
logging operations. Are salamanders
able to move out of a logged area or
repopulate it when suitable habitat
conditions return? We believe mark-
recapture studies can provide useful
information on rare terrestrial sala-
manders, but realize that data may
need to be collected over several
years and under standardized condi-
tions in order to provide direct an-
swers.

Habitat Characteristics

Preliminary evaluation of microhabi-
tat data indicate that four site charac-
teristics are most important in deter-
mining the presence of P. punctatus.
We found P. punctatus at elevations
between 732 m and 1317 m (fig. 4).
Most sites (87%) with this species oc-
curred above 960 m. Plethodon punc-
tatus occurred on all slopes but were
more common on north-facing as-
pects (87% of 11 sites) than east (38%
of 13), south (36% of 8), or west as-
pects (40% of 7). Most of the captures
(67% of 21) were on slopes of 20-45°.
Seven sites were on slopes less than
20° and between 46° and 60°. Sites
without this salamander were on a
similar range of slopes (< 20°, 28.6%;
20-45°, 57.1%; > 45°, 14.3%).

Soil temperatures under cover ob-
jects at sites with P. punctatus (x =
12.3 C, 9.4-16.1, n = 36) were nearly
identical to temperatures at sites
without this species (x = 12.8 C, 9.4-
15.8, n = 15). Soil pH under cover ob-
jects were also similar (with P. punc-
tatus: X = 6.3, 5.4-6.8; without P. punc-
tatus: X = 6.4, 5.8-6.8). Average soil
moisture at sites with P. punctatus
was 37.1% (12- 70%) and 42.8% (24-

41

80%) at sites without this species.
Soils in which P. punctatus were
found are characterized by shallow
black humus intermixed with rocks
(72% of 39 sites). One site where
eleven salamanders were captured
consisted of brown humus and ex-
tensive log cover, but few rocks.
Cover objects under which this sala-
mander was found were rocks < 645
cm2 (13.6%), rocks 645-1290 cm^
(40.0%), rocks > 1290 cm^ (34.8%),
and logs (10.6%). Over 89% of the
captures were found under rock
cover. Number of cover objects
within a 2 m circle of the captured
salamander averaged 15.1 (1-45).
Sites without P. punctatus ranged
from 100% rock cover to 0% rock
cover. Sites with canopy cover equal
to or greater than 50% accounted for
88.2% of the captures (n = 52).

We found P. punctatus in the fol-
lowing forest types: mature oak/
hickory (38.5% of 13), oak/maple/
birch (62.5% of 8), oak/pine (33.3% of
3), young oak/ maple/ hemlock (50%
of 8), virgin hemlock /yellow birch
(100% of 2), hemlock/maple/bass-
wood (62.5% of 8), white pine (0% of
2), and grassy balds (20% of 5). Of
the site characteristics we examined,
the following appear to be most im-
portant in identifying P. punctatus
habitat: elevation, aspect, soil charac-
teristics, and number of cover objects
(rocks).

Habitats of terrestrial salamanders
differ among species and, in some
cases, among geographic areas
within species (e.g., Semlitsch, 1980;
Tilley, 1973). Data derived from the
literature for management studies
and plans must be used with caution.
Baseline habitat and life history stud-
ies should be conducted on the focal
species at the location in question be-
fore developing management plans.

Effects of Logging

Tree removal effects the terrestrial
salamander community in several
ways. Removal of canopy cover

eliminates the moisture-retaining po-
tenUal of the soil and leaf litter, al-
lows an increase in insolation (with a
concomitant increase in soil tempera-
tures), and increases soil erosion
(Bury, 1983).

The use of heavy machinery com-
pacts soil and destroys leaf litter.
Enge and Marion (1986) found that
machine site preparation and
clearcutting had little effect on am-
phibian species richness in a Florida
slash pine forest. However, of the 15
amphibian species they recorded,
none was a terrestrial salamander.
On Shenandoah Mountain, where
most of the terrestrial amphibian
community is comprised of terres-
trial salamanders, logging and
clearcutting are likely to have detri-
mental effects. Salamander abun-
dance in a 60-100 yr-old deciduous
forest in another Virginia site was
more than four times that in 2 yr-old
and 6-7 yr-old clearcuts (Blymer and
McGinnes, 1977). Bury (1983) found
that terrestrial salamanders were
more abundant in old growth com-
pared to logged redwood forest

habitats. Plethodon cinereus was sig-
nificantly less abundant in a clearcut
site compared to an old-growth site
in a deciduous forest in New York
(Pough et al., 1987).

Populations of Plethodon punctatus
inhabiting rocky substrates with a
thin soil cover may be able to with-
stand some logging operations. Our
Site B was logged in a salvage opera-
tion after ice storm damage. Not all
trees were removed and the sub-
strate was not as damaged as that in
Site A, which was clearcut. These fac-
tors, combined with the presence of a
seep near the drift fence array,
probably explain the high numbers of
P. punctatus found at Site B com-
pared to other logged sites.

We found no P. punctatus on Sites
A and C for apparently different rea-
sons. Site A was clearcut, the sub-
strate was greatly disturbed, and the
lack of canopy cover prevented mois-
ture retention. The fact that P. punc-
tatus occurred on the same ridge in a
nearby hardwood stand suggests this
salamander may have occurred on
Site A prior to logging. Site C was

4(37-610 611-762

763-914 915-1067 1068-1219 1220 +■

Elevation (m)

Figure 4.— Elevational distribution of Pleftiodon punctatus on Stienandoati Mountain, George
Washiington National Forest. Solid bars represent sites wtiere P. punctatus was not found and
bars witti diagonal lines represent sites whiere thiis species was found.

42

logged 30 years ago but was re-
planted with white pine (table 1). The
logging operation and change in
vegetation type may have affected
the salamander populations previ-
ously present. However, because of
the lack of rocky substrate, we can-
not disprove the hypothesis that P.
punctatus may not have occurred
there historically.

Plethodon punctatus appears to oc-
cur in greatest abundance on rocky
sites that contain virgin hardwoods
(Site E) and sites that are not heavily
disturbed by logging operations (Site
B). Clearcutting and associated dis-
turbance does appear to eliminate
populations of this salamander. Sala-
mander mortality can be minimized
if logging operations are conducted
outside the seasonal activity period.
If size of the area logged is small, or
if the area is logged in a mosaic, or if
corridors are allowed to remain, rein-
vasion may eventually be possible
from peripheral populations when
suitable conditions return. Fragmen-
tation of the limited range of P. punc-
tatus by a patchwork of clearcuts
could seriously affect its long-term
survival.

Conclusions

Because of budget and time con-
straints, our study attempted to ob-
tain baseline data and evaluate the
effects of logging simultaneously. We
offer the following conclusions to re-
searchers and managers who must
study a salamander whose ecology is
little known.

1 . Multiple capture techniques
should be used when study-
ing an apparently rare terres-
trial salamander.

2. The life history and basic
ecology of the study species
needs to be understood be-
fore the project^ s experimen-
tal design can be erected to
evaluate logging effects.

3. Seasonal and daily activity
patterns of salamander activ-
ity must be taken into con-
sideration when surveys are
conducted to determine
range and population abun-
dance.

4. Project proposals to federal
and state agencies should
contain a two step process, a
field survey phase to obtain
baseline data on ecology and
life history and an experi-
mental phase in which log-
ging or other concerns are
evaluated. The design of the
experimental phase should
be based on the results of the
field survey.

Acknowledgments

We are grateful to the following
people for field assistance: Christian
A. Buhlmann, Kara S. Buhlmann,
Lana C. Buhlmann, Susan J. Fortuna,
Joshua C. Mitchell, and Scott M.
Smith. David A. Young deserves spe-
cial thanks for helping install the
drift fence arrays. The rangers and
foresters at the Dry River District
provided logistical support, equip-
ment, and help with site selection.
This study was supported by a Chal-
lenge Cost Share from the U.S. Forest
Service. Additional support was pro-
vided by the Nongame Wildlife and
Endangered Species Program of the
Virginia Department of Game and
Inland Fisheries.

Literature Cited

Bennett, Stephen H., J. Whitfield Gib-
bons, and Jill Glanville. 1980. Ter-
restrial activity, abundance and
diversity of amphibians in differ-
ently managed forest types.
American Midland Naturalist
103:412-416.

Blymer, Michael J., and Burd S.
McGinnes. 1977. Observations on

possible detrimental effects of
clearcutting on terrestrial amphibi-
ans. Bulletin of the Maryland Her-
petological Society 13:79-83.

Bury, R. Bruce. 1983. Differences in
amphibian populahons in logged
and old growth redwood forest.
Northwest Science 57:167-178.

Bury, R. Bruce, Howard W.

Campbell, and Norman J. Scott, Jr.
1980. Role and importance of
nongame wildlife. Transactions of
the 45th North American Wildlife
and Natural Resources Confer-
ence, 45:197- 207.

Bury, R. Bruce, and Paul Stephen
Corn. 1987. Evaluation of pitfall
trapping in northwestern forests:
trap arrays with drift fences. Jour-
nal of Wildlife Management
51:112-119.

Campbell, Howard W., and Stephen
P. Christman. 1982. Field tech-
niques for herpetofaunal commu-
nity analysis, p.193-200. In Herpe-
tological Communities. N. J. Scott,
Jr., editor. U.S. Fish and Wildlife
Service, Wildlife Research Report
13.

Enge, Kevin M., and Wayne R. Mar-
ion. 1986. Effects of clearcutting
and site preparation on herpe-
tofauna of a north Rorida flat-
woods. Forest Ecology and Man-
agement 14:177- 192.

Eraser, Douglas F. 1976. Coexistence
of salamanders in the genus Ple-
thodon: A variation on the Santa
Rosalia theme. Ecology 57:238-251.

Gibbons, J. Whitfield. 1970. Terres-
trial activity and the population
dynamics of aquatic turtles.
American Midland Naturalist
83:404-414.

Gibbons, J. Whitfield, and Raymond
D. Semlitsch. 1981. Terrestrial drift
fences with pitfall traps: an effec-
tive technique for quantitative
sampling of animal populations.
Brimleyana 7:1-16.

Gill, Douglas E. 1978. The metapopu-
lation ecology of the red-spotted
newt, Notophthalmus viridescens
(Rafinesque). Ecological Mono-
graphs 48:145-166.

43

Heatwole, Harold. 1960. Burrowing
ability and behavioral responses to
desiccation of the salamander, Ple-
thodon cinereus. Ecology 41:661-
668.

Highton, Richard. 1972. Distribu-
tional interactions among eastern
North American salamanders of
the genus Plethodon. p. 139- 188. In
The Distributional History of the
Biota of the Southern Appalachian
Mountains. Part III: Vertebrates.
Virginia Polytechnic Institute and
State University, Research Divi-
sion Monographs 4.

Jaeger, Robert G. 1978. Plant climb-
ing ability by salamanders: peri-
odic availability of plant- dwelling
prey. Copeia 1978:686- 691.

Leichter, Bill. 1987. The land nobody
wanted. Highlands of the Vir-
ginia's Magazine, Autumn 1987/
Winter 1988, pp. 3-7.

Martof, Bernard S., William M.
Palmer, Joseph R. Bailey, and Jul-
ian R. Harrison, III. 1980. Am-
phibians and Reptiles of the Caro-
linas and Virginia. University of
North Carolina Press, Chapel Hill,
264 p.

Norse, Elliott A., Kenneth L. Rosen-
baum, David S. Wilcove, Bruce A.
Wilcox, William H. Romme,
David W. Johnston, and Martha L.
Stout. 1986. Conserving Biological
Diversity in our National Forests.
The Wilderness Society, Washing-
ton, D.C., 116 pp.

Pechmann, Joseph H. K., and Ray-
mond D. Semlitsch. 1986. Diel ac-
tivity patterns in the breeding mi-
grations of winter-breeding anu-
rans. Canadian Journal of Zoology
64:1116-1120.

Pough, P. Harvey, Ellen M. Smith,
Donald H. Rhodes, and Andres
Collazo. 1987. The abundance of
salamanders in forest stands with
different histories of disturbance.
Forest Ecology and Management
20:1-9.

Semlitsch, Raymond D. 1980. Geo-
graphic and local variation in
population parameters of the
slimy salamander Plethodon gluti-
nosus. Herpetologica 36:6-16.

Tilley, Stephen G. 1973. Life histories
and natural selection in popula-
tions of the salamander
Desmognathus ochrophaeus. Ecology
54:3-17.

Tobey, Franklin J. 1985. Virginia's
amphibians and reptiles, a dis-
tributional survey. Privately pub-
lished, Virginia Herpetological So-
ciety, 114 p.

USDA Forest Service. 1986. Final
land and resource management
plan for the George Washington
National Forest. U.S. Department
of Agriculture.

U.S. Fish and Wildlife Service. 1985.
Endangered and threatened wild-
life and plants: review of verte-
brate wildlife. Federal Register
50:37958-37967.

Wilson, Edward 0. 1988. Biodiver-
sity. National Academy Press.
Washington, D.C.

44

Conserving Genetically
Distinctive Populations: The
Case of the Huachuca Tiger
Salamander (Ambystoma
tigrinum stebbinsi Lowey

James P. Collins,^ Thomas R. Jones,^ and
Howard J. Berna^

Abstract.— Huachuca tiger salamanders are a
genetically distinctive race of Ambystoma tigrinum
found only in 1 7 localities in the San Rafael Valley
(SRV) in southeastern AZ. Populations of SRV
salamanders are threatened by introduction of
exotic fishes and disease. Salamanders were largely
eliminated from four habitats after introduction of
sunfish and/or catfish. An unknown fatal disease
killed all aquatic morphs in two other habitats. An
additional threat includes possible hybridization and
introgression of SRV populations resulting from
introduction of exotic salamanders. Introduced
bullfrogs may also prey on salamanders, or act as
vectors for disease.

Technological advances in genetics
now enable characterization of vari-
ation within a species at increasingly
finer levels of description. These de-
velopments are allowing us to begin
the difficult task of identifying which
gene pools should be protected to
preserve genetic attributes significant
for conserving present and future
generations of a species (Echelle
1988, Meffe and Vrijenhoek 1988,
Ryman and Utter 1987). Rather than
considering simply which species to
conserve, we can now ask whether a
conservation effort should be di-
rected at the species, subspecies, or
population levels (Allendorf and
Leary 1988, Behnke 1972, Ryder
1986).

Tiger salamanders, Ambystoma ti-
grinum Green, range throughout
much of North America from south-
ern Canada to the central Mexican
Plateau, and from the east coast of
the United States to California
(Gehlbach 1967). This complex spe-
cies is divided into eight subspecies
(Collins et al. 1980, Gehlbach 1967,

' Paper presented of symposium. Man-
agement of Amphibians, Reptiles, and
Small Mammals in Nortt) America. (Flag-
staff, AZ, July 19-21, 1988).

'James P. Collins is Associate Professor of
Zoology. Department of Zoology, Arizona
State University, Tempe, AZ 85287-1501 .

^Ihomas R. Jones and Howard J. Berna
are Graduate Students, Department of Zo-
ology, Arizona State University, Tempe, AZ
85287-1501.

Jones et al. 1988), several of which
(tigrinum, mavortium, nebulosum, and
melanostictum) are widespread geo-
graphically, locally abundant, and
apparently not in need of protection
at this time. More information is
needed on the Mexican subspecies,
velasci, and a north-central USA race,
diaboli, before conservation needs can
be confidently assessed. Two races
need consideration now.

A. t. californiense occurs only in the
Central Valley and adjacent oak
woodlands of California, placing it
among the more geographically re-
stricted tiger salamanders. Further,
A. t. californiense appears to have
been isolated from the other races of
A. tigrinum for several million years,
and has a level of genetic divergence
equalling species-level differences
among many ambystomatid taxa
(Jones 1988). Two factors suggest this
taxon warrants special conservation
efforts. First, California populations
are as distinct genetically from other
races of tiger salamanders, as other
species of Ambystoma are from each
other. Second, the geographic isola-
tion and apparent spatial subdivision
of A. t. californiense populations
(Gehlbach 1967) likely increases their
probability of extinction (Soule 1987).

A. t. stebbinsi has properties like A.
t. californiense, suggesting it too needs
special conservation efforts despite
being classified as only part of a very
wide-ranging species. Populations of
A. t. stebbinsi occur only in the San

Rafael Valley (SRV) in the border-
lands between Arizona and Sonora,
Mexico. In addition to being geo-
graphically restricted, the race is also
genetically distinctive. Average
heterozygosity among SRV popula-
tions is the lowest in Ambystoma
(Jones et al. 1988). Electrophoretic
analysis, as well as variation in exter-
nal morphology, indicates A. t. steb-
binsi is phylogenetically most closely
related to A. t. mavortium. In contrast,
analysis of the mitochondrial DNA
(mtDNA) in these populations indi-
cates there is a single mitochondrial
clone in the San Rafael Valley. This
clone is derived from A. t. nebulosum,
not A. t. mavortium, suggesting A. t.
stebbinsi actually arose from hybridi-
zation between A. t. nebulosum and A.
t. mavortium (Collins 1988).

A recent paper describes patterns
of variation in external morphology,
allozymes, and geographic isolation
that suggest A. t. stebbinsi is a distinc-
tive race within the A. tigrinum com-
plex (Jones et al. 1988). In a future
paper we will describe mitochondrial
DNA variation in these populations
(Collins et al., in prep.). Our present
goal is to summarize several aspects
of the population biology of A. t. steb-
binsi. In addition to being restricted
geographically, our research indi-
cates salamander populations in SRV
are threatened by several factors in-
cluding disease, and factors sur-
rounding the introduction of exotic
fishes and salamanders.

45

Materials and Methods

SRV is a Plains grassland-Madrean
evergreen woodland habitat extend-
ing from southeastern AZ into north-
eastern Sonora (Brown 1982). A sur-
vey of aquatic habitats in southern
AZ and northern Sonora and Chi-
huahua indicated salamanders refer-
able to A. i. stebbinsi occurred only in
SRV (Jones et al. 1988).

From June 1979 to February 1988,
we sampled seven natural and 23
man-made or man-altered aquatic
habitats in SRV and adjacent slopes
of the Patagonia and Huachuca
mountains. Altered habitats were
primarily livestock watering tanks
constructed where natural water for-
merly existed. Bog Hole tank is a
large, impounded cienega (sensu
Hendrickson and Minckley 1985),
and another may be an impounded
spring. Salamanders occurred in only
17 of the 30 habitats sampled in SRV
(appendix 1, fig. 1). We report life
history variation in A. t. stebbinsi, and
the influence of disease and intro-
duced exotic animals on this taxon.
For describing life history variation
we emphasize four tanks (Parker
Canyon #1, Huachuca, Upper 13, and
Bodie Canyon) sampled routinely.
We also present additional informa-
tion from irregular collections at all
other SRV tanks with salamanders.

We usually collected specimens
using seines and dipnets, but occa-
sionally used gill nets. Depending on
our plans for using a particular col-
lection, we either marked and re-
leased salamanders, returned them
to the laboratory alive, or preserved
them in the field for later analysis.
All preserved specimens are in the
Lower Vertebrate Collections at Ari-
zona State University.

To summarize life history vari-
ation in A. t. stebbinsi, we classified
salamanders by life history stage and
morphology using internal and exter-
nal characters (table 1). Stages 1 or 2
were immature, and 3-5 were ma-
ture. Metamorphosed salamanders
lack gills and a caudal fin, while lar-

vae and mature branchiate salaman-
ders have those structures. All meas-
urements are in mm; snout-vent
length (SVL) is the distance from
snout to posterior margin of the vent.

Results

Life History Variation

Ambystoma tigrinum has the most
complicated pattern of morphologi-
cal and life history variation known
in salamanders. After an egg hatches
a larva begins growing in an aquatic
habitat. At about 30 mm SVL, larvae
of A. t. nebulosum, A. t. mavortium, or
A. t. tigrinum can continue develop-
ment as a typical larva, or develop as
a cannibalistic larval morph. This
dimorphism is unknown in the other
subspecies (Collins et al. 1980). At
about 70 mm SVL, larvae of all sub-

species except A. t. californiense con-
tinue developing in one of two ways.
They may metamorphose, often leave
the aquatic habitat, and must eventu-
ally return to freshwater to breed.
Alternatively, a larva continues
growing beyond 70 mm SVL, ma-
tures, and breeds as a larval-like
form, or paedomorph (Gould 1977).
Thus, depending on the subspecies, a
single population might have two
juvenile morphs, typical or cannibal,
and four adult morphs, typical and
cannibal, mature, branchiate morphs
or metamorphosed morphs of either
type. Relative frequency of each
morph varies among populations in a
subspecies (Collins 1981, Rose and
Armentrout 1976).

In SRV, most populations have
mature, typical, branchiate morphs
as well as mature, typical, metamor-
phosed morphs. Judy Tank is one
population in which we have col-

Figure 1 .—Map of the San Rafael Valley, Arizona. Symbols indicate sampling sites (see ap-
pendix 1). Electrophoretic samples were from sites 1-8; mtDNA samples were from sites
1,2,5,6,9; F=sites with fish; D=sites with diseased salamanders; arrow=J.F. Jones Ranch, type
locality for stebbinsi.

46

lected no mature, branchiate morphs
thus far. Mature, typical, branchiate
morphs dominated the SRV popula-
tions. From July 1979 to August 1985,
we collected more than 1200 mature,
branchiate salamanders and only 64
mature, metamorphosed animals.
We conservatively estimated popula-
tion sizes of mature, branchiate
morphs as varying from 50 (Upper
13 Reservoir, 1984) to several
hundred (Huachuca Tank, 1983,
1984). No population in the SRV had
cannibalistic morphs. Absence of the
cannibal morph is a distinctive fea-
ture of these populations, since the
morph can be common in A. t. nebu-

losum and A. t. mavortium, the nearest
relatives of A. t. stehbinsi.

Salamanders in SRV bred as early
as mid-February and as late as early
May. Most egg laying occurred from
mid-March to late April. Animals
hatched within several weeks and
grew rapidly, so that larvae <40 mm
SVL were often abundant by late
spring (tables 2-5). By mid-July, lar-
vae were usually about 60 mm SVL,
and those that metamorphosed gen-
erally did so from late July to early
September. A relatively small per-
centage of larvae metamorphose an-
nually—about 17% to 40% based on
estimates from Bodie Canyon Tank.

r

Table 1 .—Criteria used to classify salamanders into stages of breeding
readiness. Numerals In parentheses refer to diameter in mm (after Collins
1981).

Oviduct, ovary, peritoneum, and
cloacal characters

Wolffian duct, testes, peritoneum,
and cloacal characters

1 . Gonadal tissue primarily white and flacid; Wolffian ducts or oviducts
narrow, with few folds; cloacal margins not swollen; peritoneum largely
unpigmented.

2. Oviducts enlarged (0.5-1), white,
weal<ly convoluted; ova small (<1),
mostly white-cream colored; dor-
sal third of peritoneum light grey;
cloacal margins not swollen.

3. Oviducts large (3-4), convo-
luted, white; ova small and white,
medium (1- 1 .5) and cream or
cream-tan or black, with some
perhaps large (1 .5-2) and bipolar
cream and tan; at least dorsal
two-thirds of peritoneum grey to
black; cloacal margins swollen,
bulbous with interior margins light
grey to black and rugose.

4. Oviducts large, convoluted,
white, distended in coils; ova small
and white or large and bipolar
cream and tan; peritoneum and
cloaca as in 3.

5. All characters as in 3 except
most ova small and white with a
few darkly pigmented.

2. Duct enlarged (0.5-1), convo-
luted, but not distended in coils;
testes small, flacid; peritoneum
black; cloacal margins swollen,
with grey to grey-black borders,
especially posterior.

3. Duct large(>l),
convoluted, cream colored with
localized black pigment; testes
turgid; cloaca! margins swollen ,
grey to grey-black, rugose borders,
especially posterior; peritoneum
black, especially densely pig-
mented dorsally.

4. Duct large, convoluted, cream
colored with scattered black pig-
ment spots, distended in coils; tes-
tes turgid, enlarged; cloaca and
peritoneum as in 3.

By early autumn, first year ani-
mals that did not metamorphose be-
gan to mature (tables 2-5). From late
autumn through winter most SRV
branchiate salamanders were >100
mm SVL (tables 2-5), and ready to
breed (figs. 2-3). These data indicate
branchiate salamanders in SRV breed
for the first time when one year old.

Disease

During July and August 1985, all
branchiate salamanders in Inez,
Huachuca, and Parker Canyon #1
Tanks were killed by an undiagnosed
disease (fig 1). Salamanders in the
field and laboratory showed little re-
sistance to the disease which was
100% fatal within a few days of the
appearance of symptoms. Attempts
to culture the pathogen(s) were in-
conclusive, but many symptoms re-
sembled those characteristic of Aero-
monas infection [red legl (Fowler
1978), including lethargy, loss of ap-
petite, and the epidermis can become
red from infusion of blood. This type
of epidemic disease in the aquatic
environment is particularly devastat-
ing in A. t. stebbinsi, because popula-
tion structure in SRV is strongly
skewed toward larvae and mature
branchiate animals. In addition to
death of larvae, therefore, most
adults may have been killed in highly
infected populations.

Parker C!anyon #1 and Inez were
recolonized by metamorphosed sala-
manders that presumably escaped
the disease while in terrestrial sites.
We collected two metamorphosed
adults (male and female) and one
larva in Inez Tank in April 1986 and
collected eggs in April 1987. We also
collected eggs in Parker Canyon #1 in
April 1987, and five mature branchi-
ate morphs (3 males, 2 females) in
January 1988. Since all branchiate
morphs in Parker Canyon were killed
in 1985, and none was collected in
1986, these five animals also sup-
ported our conclusion that in SRV
branchiate salamanders can reach

47

sexual maturity when a year old. We
collected no salamanders in
Huachuca Tank as late as spring 1988
(see below).

Introduction of Exotic Animals

Fishes. — A few exceptional species of
salamanders can coexist with fishes,
but most cannot. In SRV exotic
fishes, especially centrarchids and
ictalurids, invariably eliminate sala-
manders. We do not know the effect
of native fish on A. t. stebbinsi, but no
salamanders occur in four natural
SRV habitats (Heron Spring, Sheehy
Spring, Sharp Spring, Santa Cruz
River and tributaries) that have na-
tive fishes (Gila topminnow, Poecili-
opsis 0. occidentalis and Gila chub,
Gila intermedia). We base our general
conclusions concerning exotic fishes
and salamanders in SRV on the fol-
lowing observations (fig.l).

J.F. Jones Ranch Tank. — This is
the type locality for A. t. stebbinsi.
Largemouth bass (Micropterus salmoi-
des) and bluegill (Lepomis macrochirus)
were introduced in the 1950s, and
salamanders no longer occur here
(see photograph of this site in Lowe
1964:106). It is apparently a popular
local fishing spot.

FS 58 Tank.— We first collected
mature, branchiate and larval sala-
manders here in July 1979. There
were only yellow bullheads
(Ameiurus natalis) in June 1980. In
August 1984, we collected 19 mature,
branchiate salamanders, no catfish,
and hundreds of sunfish (Lepomis
sp.).

Huachuca Tank. — First sampled in
May 1982, this tank was a reliable
source of salamanders and natural-
history information for the next two
years. On 22 August 1984 we found
one yellow bullhead, plus many lar-
val and mature branchiate salaman-
ders. On 5 July 1985 we netted >100
salamanders in each of several seine
hauls. Routine sampling on 24 Au-
gust 1985 yielded several thousand
fingerling catfish and no salaman-

>

Table 2.— Seasonal variation in number and size (SVL) of salamanders in
eacti breeding stage collected from Parker Canyon Tank.

Snout/Vent Length (mm)

Date

Stage

0-19

20-39

40-59

60-79

80-89

100-119

120-140

8 Jan

4

5

28-29 Mar

1

13

4

7

2

22-28 Apr

1

4

6

4

39

8

5

4

13-25 Jun

1

3

36

2

1

3

6

9

1

4

1

19

10

5

4

4

14

5-10 Jul

1

1

18

4

7

1

5

1

22 Aug

1

7

2

2

3

7

3

4

9

1

2 Dec

4

10

18

Table 3.— Seasonal variation in number and size (SVl^ of salamanders In
each breeding stage collected from Upper 13 Resen/oir.

Snout/Vent Length (mm)

Date

Stage

0-19

20-39

40-59

60-79

80-89

100-119

8 Jan

4

2

17Mar

4

5

7 May

1

n

43

24 Jun-

1

20

79

24

8

9 Jul

2

13

4

1

7

5

2

5

23-28 Aug

1

1

1

2

3

4

n

4

3

9 Oct

1

2

2

4

2

3

1

1

2

4

4

10 Nov

1

1

2

3

1

2 Dec

4

5

48

ders. We resampled this site several
times through February 1988. Each
time we caught only catfish, although
salamanders were abundant in
nearby tanks.

In this instance disease as well as
predation may have contributed to
decline of the salamander popula-
tion. On 24 August 1985 we found

three dead mature, branchiate
morphs in the tank. We also ob-
served a significant decline in sala-
mander p)opulations on this date at
two other tanks with diseased sala-
manders. Yellow bullheads are
highly carnivorous (Minckley 1973),
and we do not expect salamanders to
successfully recruit at Huachuca

Table 4.— Seasonal variation In number and size (SVL) of salamanders In
each breeding stage collected from Huachuca Tank.

Snout/Vent Length (mm)

Date

Stage

29 Mar

1

4/5

21 Apr

1

4/5

25 Jun

1

4/5

5 Jul

1

3

4

5

22-28 Aug

1

2

3

4

5

2 Dec

1

4

Stage 0-19 20-39 40-59 60-79 80-89 100-119 120-140

13
46

78

21

24

18
1

9
9
1
2
5
13

1

2
14

2

16

20
1

12
17

8
7

11

1
1

4

Table 5.— Seasonal variation In number and size (SVL) of salamanders in
each breeding stage collected from Bodle Canyon Tank.

Snout/Vent Length (mm)

Date

Stage 0-19 20-39 40-59 60-79 80-89 100-119 120-140

28 Mar 1
3

25 Aug 1
2
3
4

26-27 Sep
3
4

11 Nov 3
4

10

74

21

18
3

21
18

8

2
2

10

2
3
2
6
3
1

2
3
1

4
1

Tank as long as the catfish popula-
tion remains high. Catfish will pre-
sumably eat eggs, larvae, and all but
the largest salamanders. We know of
no experiments demonstrating the
minimum number of catfish that will
prohibit salamander reproduction.

Bog Hole Tank. — We collected
salamanders here in 1979, 1980, and
one larva in 1982. Native fishes com-
prised longfin dace (Agosia chrysogas-
ter) and Gila topminnow. Since the
1970s, several exotic fishes including
Gambusia affinis, Cyprinodon macular-
ius eremus, Lepomis spp., and Microp-
terus salmoides (W.L. Jvlinckley, pers.
comm.; Minckley and Brooks 1985)
have become established. Disappear-
ance of A. t. stebbinsi, and the two na-
tive fish species, correlates with es-
tablishment of non-native fish popu-
lations.

Frogs. — During the last decade
bullfrogs (Ram catesbeiana) were in-
troduced in SRV. Their introduction
correlates with reduction in native
frog populations in the valley, but
the impact of bullfrogs on A. t. steb-
binsi is unknown. Bullfrog larvae
may eat salamander eggs, while
adults may prey on larval salaman-
ders. Bullfrogs may also act as vec-
tors for disease, since in the three
tanks where salamanders were heav-
ily affected by disease, bullfrog
populations were large and appar-
ently unaffected. Frogs may be a
natural reservoir for disease, and suf-
fer few negative effects from the
pathogen(s). Since they disperse
readily to colonize surrounding habi-
tats, they may also help spread dis-
ease among amphibian populations.

S alam anders. — Commercial
baitdealers (waterdoggers), fisher-
men, and private landowners intro-
duce native and exotic salamanders
into aquatic habitats in Arizona
(Collins 1981). Salamanders are used
commonly as bait by fishermen in the
American Southwest (table 6), and
Lowe (1955) first noted that salaman-
ders were being introduced into Ari-
zona for this purpose. SRV is closed
to "waterdog" collecting under Ari-

49

zona Game and Fish Commission
order #R1 2-4-311. Enforcement is dif-
ficult, because SRV is large and
sparsely settled. It would be easy to
introduce exotic A. Hgrinum into this
valley. Pre-mating and post-mating
isolating mechanisms in the A. H-
grinum species group within Amby-
stoma are weak (Brandon 1972, Nel-
son and Humphrey 1972).
Introduced A. Hgrinum would be ex-
pected, therefore, to easily interbreed
with native tiger salamanders.

Discussion

In theory, average heterozygosity or
gene diversity of organisms in an
area can be decomposed into gene
diversities within and between any
subpopulations comprising the total
number of organisms in the popula-
tion (Nei 1987). If all organisms in a
population are a panmictic aggre-
gate, then the component describing
variation between subpopulations is
zero. We have no information on dis-
persal between tanks in SRV, so for
this discussion we arbitrarily con-
sider each tank a subpopulation and
together all tanks comprise the total
population of SRV salamanders.
Within this context our results high-
light several factors to consider in
trying to understand the evolution-
ary genetics of SRV tiger salaman-
ders.

Mean heterozygosity (.0015) for A.
t. stebbinsi is the lowest reported for
any salamander (Jones et al. 1988).
Salamanders in SRV went through
one or more bottlenecks at some
point in their history, but cause(s)
and time of reduction in numbers
and associated genetic diversity are
unknown. The effect of a one-time
bottleneck is a drastic decrease in ex-
pected heterozygosity of a popula-
tion, and in theory, repeated bottle-
necks could reduce gene diversity
even more (Motro and Thomson
1982).

Current factors affecting changes
in SRV salamander numbers may

provide some insight into the origin
and /or perhaps maintenance of low
gene diversity in SRV. Increased
heterozygosity generally correlates
positively with traits associated with
high individual vigor and fitness,
plus population stability (Mitton and
Grant 1984). Susceptibility to disease
or apparent reduced ability to over-
come infection may thus be conse-
quences of reduced genetic variation
in SRV salamanders. A historical bot-
tleneck in population size with asso-
ciated loss of gene diversity in SRV
salamanders, therefore, could have
resulted in populations more suscep-
tible to disease. This susceptibility, as
seen in contemporary stocktanks,
could easily cause severe reductions
in numbers of salamanders and re-
tard any expected increase in gene
diversity. O'Brien et al. (1985) pro-
vide a related example. They de-
scribe how extremely low genetic
variation in the South African chee-
tah may derive from a population

bottleneck. Low genetic variation
seen in structural loci also extends to
the major histocompatibility com-
plex. This extreme monomorphism
correlates with a hypersensitivity in
cheetahs to some viral pathogens,
and they feel the sensitivity of this
genetically uniform species to patho-
gens provides an example of the pro-
tection against disease genetic vari-
ation provides to species. The mecha-
nism connecting low genetic vari-
ation revealed by electrophoresis and
susceptibility to disease is unclear.
Hence, for both cheetahs and SRV
salamanders it is uncertain if reduc-
tion of population size and loss of
genetic variation increased suscepti-
bility to disease, or alternatively, sus-
ceptibility increased for some other
reason, and this lead to reductions in
population numbers.

Two additional factors, again
found in present stocktanks, would
reinforce this pattern of change in
numbers of salamanders and reduc-

>

120"

100"

80"

60"

40"

20

2 3 4
Breeding Stage

Figure 2.— Variation in SVL with breeding stage for aninnals from four SRV populations: soiid
llne=Upper 13 Reservoir, dotted line=Parl<er Canyon Tank #1, dots+dasties=Huactiuca Tank,
daslies=Bodie Canyon Tank. (Circles=mean, vertical line=lSE, perpendicular tiorizontal
line=linnits.)

50

tion in heterozygosity. First, in SRV,
most salamanders occur in aquatic
habitats and most, if not all, salaman-
ders in the water at the time of an
epidemic are apparently killed. Since
most SRV salamanders are adult,
branchiate animals, aquatic disease
dramatically reduces effective popu-
lation size. Furthermore, future
population recruitment is reduced

since a larval year class is also lost.
Thus, the preponderance of branchi-
ate morphs in SRV subpopulations
exacerbates any negative effects of
aquatic diseases on population size
and heterozygosity. If disease is a
predictable selection pressure, how-
ever, it is not obvious why relative
frequency of adult morphs in a sub-
population has not shifted from

2 3 4
Breeding Stage

Figure 3.— Variation in SVL with breeding stage for ail SRV populations. Synnbols as In figure 2.

Table 6.— Total bait sales in the Lower Colorado River basin (modified after
Espinoza et al. 1970).

Area Value of sales ($) Volume of sales

Salamanders

Minnows

1 . Las Vegas-Lake Mead

190,000

1,250,000

750,000

2. Mid-river

110,000

570,000

325,000

3. Parker Dam

80,000

400,000

185,000

4, Yuma

53,000

190,000

290,000

5. BIythe-Palo Verde

24,000

30,000

230,000

Total in 1968

457,000

2,440,000

1,780,000

51

branchiate to metamorphosed
morphs. Since disease appears to
equally affect metamorphosed and
branchiate morphs, this may indicate
there is little or negligible difference
in heritable variation for disease re-
sistance betv/een morphs. Selection,
therefore, would have little or no ef-
fect on relative morph frequencies.
Likewise, the genetic basis of paedo-
morphosis versus metamorphosis is
poorly understood. It may be that
genetic differences between morphs
are slight, with environmental condi-
tions largely determining relative fre-
quency of each adult morph in a sub-
population.

Second, exotic predaceous fishes,
like an aquahc-borne disease, will
quickly reduce adult and larval sala-
mander numbers, and coincidently
genetic diversity, in any stocktank in
which they are introduced. Haphaz-
ard introduction of fishes in SRV
habitats may help maintain low lev-
els of genetic diversity.

Other than by mutation, heterozy-
gosity in SRV could be increased by
the introduction of exotic A. tigrinum,
and their interbreeding with native
SRV salamanders. The only report on
salamander introductions in AZ is 20
yrs old, summarizes use of salaman-
ders as bait in only the extreme west-
ern part of AZ, and provides no in-
formation on relative numbers im-
ported into AZ, as opposed to sala-
manders moved within AZ (Espi-
noza et al. 1970). Nonetheless, in
1968, about 2.5 million salamanders
in western AZ were available for po-
tential introduction into aquatic habi-
tats. The increased number of people
living in AZ means these numbers
are probably higher now. Further-
more, salamanders are regularly sold
for bait in all major population cen-
ters in AZ, not just along the Colo-
rado River. Salamanders sold in AZ
come from three primary sources: (1)
seined from AZ populations; (2) col-
lected and imported from popula-
tions in at least NM, OK, CO, TX,
and NE; and (3) adults and/or larvae
collected in AZ or other states, intro-

duced into AZ habitats as "brood
stock," and larvae from these ani-
mals collected in subsequent years
and sold as bait.

We know from discussions with
residents that salamanders are at
least occasionally moved between
tanks in SRV. We have no evidence
salamanders are introduced into SRV
from elsewhere, and two facts sug-
gest such events are rare or non-exis-
tent. First, our electrophoretic data
show heterozygosity is uniformly
low for SRV animals from eight sub-
populations separated by as much as
25 km (fig. 1) (Jones et al. 1988). Alle-
lic diversity should be higher if sala-
manders are regularly being intro-
duced into SRV. Second, there is only
one mitochondrial DNA clone in
SRV. Again, regular introductions
would be expected to result in more
than one mtDNA haplotype in SRV.
Nonetheless, continued active use of
salamanders for bait in AZ means
there is always the possibility exotic
animals might be introduced. This
could lead to introgressive hybridiza-
tion between species or subspecies,
or perhaps interbreeding between
genetically distinctive populations of
the same species. Furthermore, we
cannot completely exclude the possi-
bility that A. t. nebulosum and/or A. t.
mavortium was deliberately or ac-
cidently introduced into SRV, thus
creating the opportunity for hybridi-
zation between these races. How-
ever, several arguments suggest sala-
manders were native in SRV (Jones et
al. 1988).

Among tiger salamanders in SRV,
color pattern of metamorphosed ani-
mals, relative frequency of typical
and cannibal morphs, nuclear gene
frequencies derived from electro-
phoresis, and mitochondrial DNA
genotype each show distinctive vari-
ation relative to the entire A. tigrinum
complex. We conclude, therefore,
that SRV tiger salamander popula-
tions are sufficiently different to war-
rant at least subspecific status as A. t.
stebbinsi (Collins 1988, Jones et al.
1988). Likewise, the small number

and restricted geographic range of
SRV populations increases their like-
lihood of extinction. These facts
coupled with our information con-
cerning life history, incidence of dis-
ease, and potential negative effects of
exotic animals in SRV, argue that
conservation efforts and careful
management of A. t. stebbinsi is
needed. Although A. tigrinum has a
wide distribution, in some races spe-
cial effort needs to be directed at pro-
tecting locally adapted populations
to conserve the diversity of genetic
and life history traits characteristic of
this polytypic species.

Acknowledgments

We thank the following for help in
the field: T. Corbin, B.D. DeMarais,
P.J. Fernandez, W.C. Hunter, A.S.O.
Jones, L.F. Elliott, P. Rosen, C.A.
Schmidt, and C.W. Seyle. We appre-
ciate criticism of any earlier draft of
this paper provided by L. Allison, D.
Begun, W.L. Minckley, and C.A.
Schmidt. G. Goodwin and F. Sharp
generously provided access to their
land in SRV. Arizona Game and Fish
Department provided collecting per-
mits. This research was supported by
funds from U.S. Fish and Wildlife
Service, Office of Endangered Spe-
cies (order #20181-0746-83) awarded
to JPC, a National Science Founda-
tion grant to JPC (BSR-8407930), and
a Tucson Audubon Society grant to
TRJ.

Literature Cited

Allendorf, Fred W. and Robb F.
Leary. 1988. Conservation and dis-
tribution of genetic variation in a
polytypic species, the cutthroat
trout. Conservation Biology 2:170-
184.

Behnke, R.J. 1972. The systematics of
salmonid fishes of recentiy glaci-
ated lakes. Journal of the Fisheries
Research Board of Canada 29:639-
671.

Brandon, Ronald A. 1972. Hybridiza-
tion between Mexican
salamanders Ambystoma dumerilii
and Ambystoma mexicanum under
laboratory conditions. Herpetol-
ogica 28:199-207.

Brown, D.E., ed. 1982. Biotic commu-
nities of the American Southwest-
United States and Mexico. Desert
Plants 4:1-342.

Collins, James P. 1981. Distribution,
habitats, and life history variation
in the tiger salamander, Amby-
stoma tigrinum, in east-central and
southeast Arizona. Copeia
1981(3):666-675.

Collins, James P. 1988. Evolutionary
relationships of Ambystoma ti-
grinum stebbinsi to other South-
western A. tigrinum based on mi-
tochondrial DNA. Final Report,
Nongame Branch, Arizona Game
and Fish Department: 28 p.

Collins, James P., Jeffry B. Mitton,
and Benjamin A. Pierce. 1980. Am-
bystoma tigrinum: a multispecies
conglomerate? Copeia 1980(4):938-
941

Echelle, Anthony A. 1988, in press.
Conservation genetics and genie
diversity in threatened fishes of
western North America. In Battie
against extinction: Twenty years of
native fish management in the
desert West. W. L. Minckley and
James E. Deacon, eds. University
of Arizona Press, Tucson.

Espinoza, F.A., James E. Deacon, and
Andre Simmons. 1970. An eco-
nomic and biostatistical analysis of
the bait fish industry in the lower
Colorado River. Special Publica-
tion of the University of Nevada,
Las Vegas: 1-87.

Fowler, Murray, E.,ed. 1978. Zoo and
wild animal medicine. W.B. Saun-
ders, Philadelphia, PA.

Gehlbach, F.R. 1967. Ambystoma ti-
grinum (Green). Catalogue of
American Amphibians and Rep-
tiles: 52.1-52.4.

Gould, Stephen J. 1977. Ontogeny
and phylogeny. Harvard Univer-
sity Press, Cambridge, MA: 501 p.

52

Hendrickson, Dean A. and W.L.
Minckley. 1985. Cienegas- Vanish-
ing climax communities of the
American Southwest. Desert
Plants 6:131-175.

Jones, Thomas R. 1988. Macrogeogra-
phic and microgeographic pat-
terns of evolutionary differentia-
tion in tiger salamanders (Amby-
stoma tigrinum) in the Southwest-
ern U.S.A., Ph. D. Dissertation,
Arizona State University, Tempe,
AZ.

Jones, Thomas R., James P. Collins,
Thomas D. Kocher, and Jeffry B.
Mitton. 1988. Systematic status
and distribution of Ambystoma ti-
grinum stebbinsi Lowe
(Amphibia:Caudata). Copeia
1988(3):621-635.

Lowe, Charles H. 1955. The salaman-
ders of Arizona. Transactions of
the Kansas Academy of Sciences
58:237-251.

Lowe, Charles H., ed. 1964. The ver-
tebrates of Arizona. University of
Arizona Press, Tucson, Arizona:
270 p.

Meffe, Gary K. and Robert C. Vrijen-
hoek. 1988. Conservation genetics
in the management of desert
fishes. Conservation Biology
2:157-169.

Minckley, W.L. 1973. Fishes of Ari-
zona. Arizona Game and Fish De-
partment, Phoenix, Arizona: 293 p.

Minckley, W.L. and James E. Brooks.
1985. Transplantations of native
Arizona fishes: records through
1980. Journal of the Arizona-Ne-
vada Academy of Sciences 20:73-
89.

Mitton, J.B. and M.C. Grant. 1984.
Associations among protein
heterozygosity, growth rate, and
developmental homeostasis. An-
nual Review of Ecology and Sys-
tematics 15:479-499.

Motro, Uzi and Glenys Thomson.
1982. On heterozygosity and the
effective size of populations sub-
ject to size changes. Evolution
36:1059-1066.

Nei, Masatoshi. 1987. Molecular evo-
lutionary genetics. Columbia Uni-
versity Press, New York, N.Y.:
512 p.

Appendix

Nelson, Craig E. and R.R.

Humphrey. 1972. Artificial inter-
specific hybridization among Am-
bystoma. Herpetologica 28:27-32.

O'Brien, S.J., M.E. Roelke, L. Marker,
A. Newman, C. A. Winkler, D.
Meltzer, L. Colly, J. F. Evermann,
M. Bush, and D. E. Wildt. 1985.
Genetic basis for species vulnera-
bility in the cheetah. Science
227:1428-1434.

Rose, Francis R. and D. Armentrout.
1976. Adaptive strategies of Amby-
stoma tigrinum (Green) inhabiting
the Llano Estacado of West Texas.
Journal of Animal Ecology 45:713-
729.

Ryder, Oliver A. 1986. Species con-
servation and systematics: the di-
lemma of subspecies. Trends in
Ecology and Evolution 1: 9-10.

Ryman, Nils and Fred Utter, eds.
Population genetics and fishery
management. University of Wash-
ington Press, Seattle, WA: 420 p.

Soule, Michael A.,ed. 1987. Viable
populations for conservation.
Cambridge University Press, Cam-
bridge, England: 189 p.

Locality data for all populations of A. t. stebbinsi, taken from ttie following U.S.G.S. 7V2 nnin. quadrangles:
Cannpini Mesa, Canelo Pass, Duquesne, Harstiaw, Huactiuca Peak, Lochiel.

Site Locality Map codes Site Locality Map codes

Bodie Canyon Tank:
Bog Hole Tank:
Campini Mesa Tank #1:
FS 58 Tank:
FS 799 Tank:
Grennan Tank:
Heron Springs Tank:
Huachuca Tank:

NW» SE« sec.2, T.24S, R.18E 8

31*22' 30"N, no •28' 45"W

NW» SE« sec33, T.22S, R17E F

31* 28' 36'H 110° 37" 06"W

SW» E» sec.l9, T.24S, R.19E

31*21' 00", 110*26' 45"W

NE» NE» sec6, T.23S, RITE F

31*27' 03"N, 110*38' 49"W

SW- NE» sec36, T22S, R.17E 7

31*28' 48"N, 110*34' 09"W

S center sec. 14, T.23S, R.16E 6

31*25' 29'H 110°40'47"W

SW- NE- sec.l4, T24S, R17E

31*20' 39"N, 110*34' 54"W

NE- NW» sec.l5, T24S, R18E 3,F,D

31*21' 12"N, 110°30'15"W

Inez Tank:
Judy Tank:

Ki-He-Kah Ranch Tank:

Lower 13 Reservoir:

Meadow Valley Flat
Tank #1

Parker Canyon Tank #1 :
School Canyon Tank #1 :
School Canyon Tank #2:
Upper 13 Reservoir:

SW- NW- sec.2, T.24S, R18E D

31*22' 30"N, 110*29'30"W

SE» SE» sec35, T23S, R.18E 9

31*23' 04"N, 110*29' 19"W

SW» SW- sec.l, T.23S, R17E

31*26' 26"N, 110*35' 22"W

SW» NE» sec.l8,T.24S, R17E

31*20' 49"N, 110*39' 05"W

SW • NE« sec.6, T.22S, R.17E 1

31*27' 49"N, 110*38' 47"W

NE- NE» sec.l9, T.24S, R.18E 2,D

31*20' 16"N, 110*32' 42"W

NE» SE- sec9, T.24S, R.18E 4

31*21' 28"N, 110*24' 04"W

NE» SE» sec.l7, T.24S, R19E

31*21' 14"N, 110*24' 24"W

S center sec.7, T.24S, R.17E. 5

31*21' 18"N, 110*39' 16"W

53

Habitat Requirements of New
IVIexico's Endangered
Salannanders^

Cynthia A. Ramotnik^ and Nornnan J.
Scott, Jr.^

Abstract.— We measured habitat components for
two state-listed endangered salamanders in New
Mexico in 1986 and 1987. Both species ore restricted
to mesic environments within high-elevation, mixed
coniferous forests. Steep slope and high elevation
were the most useful variables for predicting the
occurrence of Jemez Mountains salamanders and
Sacramento Mountain salamanders, respectively.
Although the discriminant models show some
predictive value in detecting salamanders based on
habitat variables, we believe that the best survey
technique is ground-truth surveys in wet weather. A
better fit of the discriminant models might be
obtained by including variables not measured e.g.,
fire and logging history, and soil characteristics. We
offer interim management guidelines as a result of
our analysis.

Two of the three species of salaman-
ders that occur in New Mexico are
restricted to coniferous forests at
high elevations. The Jemez Moun-
tains salamander (Plethodon neomexi-
canus) (fig. 1) is known only from
north-central New Mexico at the
southern terminus of the Rocky
Mountains (Reagan 1972). The Sacra-
mento Mountain salamander (Aneides
hardii) (fig. 2) occurs in the Capitan
and Sacramento Mountains in south-
central New Mexico (Williams 1976).
These lungless salamanders, with
small body sizes and terrestrial juve-
nile development, are restricted to
mesic environments. Lowe (1950)
suggested that both species are rel-
icts of the mid-Tertiary Rocky Moun-
tain fauna.

In 1975, both species were listed
by the state of New Mexico as endan-
gered due to their restricted distribu-
tion (Hubbard et al. 1979). Since
1980, increases in timber harvest by

' Paper presented at symposium. Man-
agement of Amphibians, Reptiles, and
Small Mammals in Northi America. (Flag-
staff, AZ. July 19-21 1988).

^Museum Specialist, U.S. F/sh & Wildlife
Service, National Ecology Researchi Center.
1300 Blue Spruce Drive. Fort Collins, CO
80524.

^Zoologist, U.S. Fisfi & Wildlife Service.
National Ecology Research) Center, Mu-
seum of Southwestern Biology, University of
New Mexico, Albuquerque. NM 87131.

the U.S. Forest Service (USPS) and
changes in timber practices have
prompted concern about the effect of
logging on these salamanders (Scott
et al. 1987, U.S. Fish & Wildlife Serv-
ice 1986). Most of the range of each
species occurs on National Forest
(NF) lands, and the close association
of these salamanders with mixed co-
niferous forests may make them vul-
nerable to some forest-management
practices. In 1985, both species were
placed under review as potentially
threatened or endangered species
under the Federal Endangered Spe-
cies Act (Ramotnik 1986, Staub 1986).
As a result, an interagency commit-
tee was established to identify data
and management needs and develop
strategies to address these needs.

Figure 1 .—Jemez Mountain salannander
(Plethodon neomexicanus). Photo by
Stephen Corn.

Figure 2.— Sacramento Mountain
salamander (Aneides hardii). Photo by
Stephen Corn.

In 1986, the U.S. Fish & Wildlife
Service (USFWS) contracted with the
USFS to study these species on NF
lands. The primary objectives were
to survey for salamanders in plan-
ning units under consideration for
future logging operations and to
characterize salamander habitats us-
ing habitat components that are
meaningful and useful to USFS biolo-
gists and land managers. This infor-
mation would be used to assess po-
tential salamander habitat from maps
or aerial photos, thereby reducing
the need to inventory areas by
ground-truth assessment.

In this paper, we characterize
habitats of Jemez Mountains sala-
manders and Sacramento Mountain
salamanders based on general site
characteristics and surface cover

54

items that could serve as refugia for
salamanders. We use a multivariate
analysis of habitat characteristics that
describes areas with and without
salamanders, and present manage-
ment guidelines as a result of this
analysis.

Study Areas

We studied the Jemez Mountains
salamander within the Santa Fe NF
in the Jemez Mountains (Los Alamos,
Rio Arriba, and Sandoval Counties,
New Mexico), which are located ap-
proximately 100 km north of Al-
buquerque (fig. 3). The Jemez Moun-
tains are volcanic in origin and are
underlain by volcanic rock, ash, and
pumice. The predominant feature in
the area is the volcanic caldera, the
Valle Grande, around which the
mountains lie. Fieldwork on the Sac-
ramento Mountain salamander was
conducted in the Sacramento Moun-
tains, within the Lincoln NF, Otero
County, New Mexico fig. 3). Volcanic
intrusions occur within the Paleozoic
strata of the Sacramento Mountains.
Elevations in the Jemez Mountains
range from 2130-3410 m, and from
2290-3600 m in the Sacramento
Mountains.

Habitat types within these eleva-
tional ranges occur within the Rocky
Mountain upper montane (2290-2900
m) and subalpine (2900-3660 m) for-
est association (Castetter 1956). The
upper montane forest association
(Shelf ord 1963) is characterized by
mixed coniferous forests dominated
by white fir (Abies concolor), Douglas-
fir (Pseudotsuga menziesii), Engelmann
spruce (Picea engelmannii), and blue
spruce (Picea pungens). Deciduous
components include quaking aspen
(Populus tremuloides), Rocky Moun-
tain maple (Acer glabrum), oak {Quer-
cus spp.), New Mexico locust (Robinia
neomexicam), and oceanspray (Holo-
discus dumosus). Ponderosa pine
(Pinus ponderosa) stands predominate
at the lower elevations, particularly
on south-facing slopes. Within the

subalpine forest association, Engel-
mann spruce, Douglas-fir, and white
fir are the most common trees. Aspen
and Rocky Mountain maple are
found to a lesser extent. Aspen
groves, talus fields, and open mead-
ows are present at higher elevations.
Annual precipitation in the Jemez
Mountains ranges from 400-550 mm
(Castetter 1956) and is slightly higher
in the Sacramento Mountains. Much
of the precipitation falls between July
and September (Kunkel 1984).

Methods

We conducted fieldwork in the sum-
mers of 1986 and 1987 (Jemez Moun-
tains: 28 July-14 August 1986, 29
June-11 July 1987, 24 August-5 Sep-
tember 1987; Sacramento Mountains:
22 August-10 September 1986, 8-20
June 1987; 20 July-1 August 1987).
These dates included the surface ac-
tivity periods of Jemez Mountains
salamanders (Reagan 1972) and Sac-
ramento Mountain salamanders
(Williams 1976).

Transects were established in for-
ested areas; most were located in
planning units selected by USPS per-
sonnel. Within these areas, locations
of transects were selected from topo-
graphic maps to sample a variety of
topographic aspects. South-facing
slopes were not searched in the
Jemez Mountains due to the diffi-
culty in locating salamanders on
these slopes (Ramotnik 1988). To en-
sure having sites occupied by sala-
manders, we visited known localities
or areas where salamanders had re-
cently been found. A small number
of sites outside planning units were
chosen from topographic maps.

We established 100-m^ transects (2
m X 50 m) oriented uphill from near
the bottoms of slopes. Our transect is
modified from area-constrained
searches, a technique developed by
others, e.g.. Bury (1983), Bury and
Corn (this volume). Bury and Ra-
phael (1983), Campbell and Christ-
man (1982), Raphael (this volume).

and Raphael and Rosenberg (1983).
The areas of four classes of cover
items (rock, bark, fine woody debris,
and coarse woody debris) were esti-
mated visually. We further divided
coarse woody debris (CWD) into
three decay classes, adapted from a
five-class scheme for rating decom-
position of Douglas-fir logs (Franklin
et al. 1981). To emphasize differences
between decay classes, we combined
classes 1 and 2 (CWDl), and classes 3
and 4 (CWD3), and placed the most
decayed logs, class 5, in a third cate-
gory (CWD5).

Aspect was taken with a magnetic
compass at 10, 30, and 50 m. Com-
pass readings were assigned to one
of four aspect classes where 316-45° =
north-facing; 46-135° = east-facing;
136-225° = south-facing; and 226-315°
= west-facing. Percent slope was de-
termined with a clinometer, and per-
cent canopy cover was estimated
with a spherical densiometer
(Lemmon 1956). Both measurements
were recorded at 10-m intervals. All
readings were made along the
transect and averaged for the

Figure 3.— Distribution of Jemez Mountairw
salamanders (Plethodon neomexicanus)
and Sacramento Mountain salamanders
(Aneides hardii) in New Mexico.

55

transect. Numbers of white fir and
Douglas-fir were pooled in a single
class (TFIR), as were Engelmann and
blue spruce (TSPRUCE), and Pinus
spp. (TPINE). Numbers of trees
within tree classes were counted in a
20-m X 50-m plot centered over the
transect. Twenty-three measured and
derived variables were used in the
analyses (table 1).

We determined numbers of sala-
manders on transects by searching all
cover items manually or with potato
rakes. The locations of salamanders
in other than the four classes of cover
items also were recorded. When a
salamander was found, we recorded
snout-vent length (distance from tip
of snout to anterior edge of vent),
sex, and dimensions and type of
cover item. For coniferous logs, we
also recorded salamander position
relative to the log (in, under, or un-
der bark) and decay class (modified
from Com and Bury, in press, Ra-
phael and Rosenberg 1983). These
data were used to calculate densities
of salamanders on transects and to
determine cover item use by sala-
manders. We acquired additional
data on cover item use by salaman-
ders by locating salamanders in areas
on both sides of the transects.

Statistical Analysis

Data for transects with and without
salamanders were pooled separately.
We calculated descriptive statistics
(mean, standard error, range) for
habitat variables in the two groups
and used a one-way analysis of vari-
ance to compare transformed vari-
ables between groups. Size classes of
fir and spruce were compared be-
tween the two groups with a t-test.

The following transformations
were applied to stabilize the variance
of the habitat variables (Snedecor
and Cochran 1967) and to increase
the probability of a normal distribu-
tion: arcsine (SLOPE CANOPY);
square root + 0.5 (tree densities); and
log + 0.5 (cover items). Elevahon was

not transformed because values were
distributed normally.

A stepwise variable entry proce-
dure (STEPDISC) selected the "best
set" of habitat variables to discrimi-
nate between groups and reduced
the corriplexity of the original vari-
able set. Because the models selected
by STEPDISC are not necessarily the
best possible models (SAS Institute
Inc 1982), cross-validation was ac-
complished by using canonical analy-
sis (CANDISC) or descriptive dis-
criminant analysis (DDA) (Williams
1983). DDA attempts to establish op-

timal separation between groups us-
ing linear transformations of the in-
dependent variables based on vari-
ables selected by the stepwise proce-
dure. The Mahalanobis distance be-
tween group means was tested using
an F-statistic.

Predictive discriminant analysis
(PDA) (Williams 1983) (DISCRIM)
was used to test the discriminatory
power of the variables selected by
DDA. We used chi-square analysis to
compare cover item use (of the four
classes) to availability and to com-
pare aspects of transects with and

Table 1.— Description of measured and derived habitat variables used In
habitat selection analysis of two species of New Mexico salamanders.

Sampling unit
mnemonic

50-m x2-m transect

BARK
CANOPY

CWDl

CWD3

CWD5

CWD
ELEV

FWD

ROCK
SLOPE

Description

50-m X 20-m plot

SFIR

Number of

MFIR

Number of

LFIR

Number of

TFIR

SFIR + MFIR

SSPRUCE

Number of

MSPRUCE

Number of

LSPRUCE

Number of

TSPRUCE

SSPRUCE +

TASPEN

Number of

TNOD

Number of

TOAK

Number of

TPINE

Number of

TSNAGS

Number of

Estimate of amount of bark on ground (m^)
Average percent canopy cover recorded with
G spherical der^iometer
Estimate of amount of poorly decayed coarse
woody debris (m^)

Estimate of amount of moderately decayed
coarse woody debris (m^)
Estimate of amount of well-decayed coarse
woody debris (m^)
CWD1 +CWD3 + CWD5
Estimated from a U.S. Geological Survey topo-
graphic map (m)

Estimate of amount of fine woody debris
(sticks) (m^

Estimate of amount of surface rock (m^)
Average percent slope measured with a cli-
nometer

small fir (<20 cm dbh)
medium fir (20-50 cm dbh)
large fir (>50 cm dbh)
+ LFIR

small spruce (<20 cm dbh)

medium spruce (20-50 cm dbh)

large spruce (>50 cm dbh)

MSPRUCE + LSPRUCE

aspen (all sizes)

non-oak deciduous (all sizes)

oak (all sizes)

pine (all sizes)

snags (all sizes)

56

without salamanders. The Statistical
Analysis System computer package
(SAS, Version 5) was used for all
analyses (SAS Institute Inc 1982). Sig-
nificance levels were set at P < 0.05
unless otherwise indicated.

Results

Jemez Mountains Salanaander

Salamanders (N = 28) were present
on 10 of 43 transects (23%) with a
mean density of 3/100 m^ in occu-
pied areas. One hundred twenty
salamanders were found in areas off
the transects. Transects with sala-
manders occurred on significantly
steeper slopes and at lower eleva-
tions than transects without salaman-
ders (table 2). Analysis of size classes
of fir and spruce showed no signifi-
cant differences between transects
with and without salamanders. Pro-
portions of decay classes of CWD

also did not differ significantly be-
tween the two groups of transects (X^
= 0.28, df = 2, P > 0.90). The amount
of CWDl was similar between
groups but amounts of CWD3 and
CWD5 were higher on transects with
salamanders. Although no south-fac-
ing slopes were searched, propor-
tions of other aspects occupied by
salamanders were not different from
the proportions of total aspects
searched (X^ = 1.3, df = 2, P > 0.50).

Three of the original 20 variables
were selected by the stepwise vari-
able entry procedure for inclusion in
the descriptive discriminant model:
SLOPE, TPINE, and LSPRUCE (table
3). Subsequent analysis by DDA re-
tained these variables. The resultant
discriminant function explained 38%
of the between-group variance; how-
ever, it did not have significant
power in discriminating between
groups (F = 2.34, P = 0.09). This func-
tion describes a multivariate gradient
that ranges from steep slopes with

Table 2,— Comparison of habitat variables measured on transects with and
without Jemez Mountains salamanders, Santa Fe National Forest, 1986-
1987. Significance is based on one-way analysis of variance. Mnemonic
codes for habitat variables are explained in table 1 .

Transects (N = 10)
with salamanders

Transects (N = 33)
without salamanders

Mnemonic

X ± se (range)

X ± se (range) Significance

ELEV

2526 +35.8

(2359-2621)

2635 +22.0

(2332-2886)

•

SLOPE

66 + 2.5

(55-84)

44+ 2.8

(0-82)

CANOPY

62 + 1,8

(56-65)^

64+ 2.1

(21-82)2

NS

TFIR

72 ±10,4

(29-156)

95 + 10.3

(22-292)

NS

TSPRUCE

17 + 6,6

(0-59)

20+ 5,9

(0-163)

NS

TPINE

25 ± 7,8

(0-63)

9 + 2.1

(0-56)

NS

TASPEN

20+ 8.8

(1-96)

17 + 2.5

(0-60)

NS

TOAK

10+ 6,6

(0-59)

7± 2.4

(0-50)

NS

TSNAGS

33 + 6.1

(5-64)

27 + 3.3

(3-82)

NS

TNOD

29 + 10.4

(0-103)

8+ 2.0

(0-51)

NS

ROCK

11 + 2,6

(3-26)

7 + 1.6

(0-37)

NS

FWD

4+ 1.1

(2-12)

4+ 0,5

(0-15)

NS

BARK

1+1.0

(0-10)

1 ± 0.1

(0-3)

NS

CWD

10+ 1.9

(1-20)

9± 0.8

(1-26)

NS

'P<0.05
"P< 0.005

'Data are available for 5 frar\secfs.
'Data are available for 29 transects.

many pine and large spruce trees
containing salamanders, to shallow
slopes with few pine or large spruce
trees without salamanders. SLOPE
had the highest discriminating power
(r^ = 0.73). PDA correctly classified
91% of the 33 transects without sala-
manders and 80% of the 10 transects
with salamanders.

The 10 transects and additional
searches produced 148 Jemez Moun-
tains salamanders; the type of cover
item was known for all but one sala-
mander. Ninety-six percent (141/
147) of salamanders were distributed
among the four major cover classes
as follows: CWD, 100 (68%); ROCK,
40 (27%); FWD, 1 (1%). No salaman-
ders were found under BARK. Three
salamanders (2%) were found on
transects under surface litter and
three salamanders (2%) were found
under aspen logs. The frequency of
salamanders associated with CWD
by decay class was CWDl — 4%;
CWD3— 66%; CWD5— 30%. Of 28
salamanders found on transects, 24
salamanders were associated with
one of the four classes of cover items.
Because of the small sample size, we
were unable to determine a correla-
tion between cover item availability
and use.

Sacramento Mountain
Salanaander

Salamanders (N = 233) were present
on 26 of 80 transects (33%) with a
mean density of 6/100 m^ in occu-
pied areas. We located 387 salaman-
ders in areas off the transects.
Transects with and without salaman-
ders differed in several respects:
transects with salamanders occurred
at significantly higher elevations, on
shallower slopes, and had higher
numbers of spruce and lower num-
bers of pine than transects without
salamanders (table 4). Analysis of
size classes of fir and spruce revealed
that densities of large fir and all size
classes of spruce were significantly
higher on transects with salamanders

57

(LFIR: t = 3.38, P = 0.001; SSPRUCE: t
= 2.85, P = 0.008; MSPRUCE: t = 2.56,
P = 0.016; LSPRUCE: t = 3.04, P =
0.003) (fig. 4). Although the total
amount of CWD on transects with
and without salamanders was not
significantly different, there was sig-
nificantly more CWD5 on transects
with salamanders (X^ = 6.93, df = 2, P
> 0.05). The proportions of transects
by aspect did not differ between the
two groups (X2 = 3.83, df = 3, P >
0.10).

Because numbers of the three size
classes of spruce were significantly
higher on transects with salaman-
ders, we substituted TSPRUCE for
SSPRUCE, MSPRUCE, and
LSPRUCE in subsequent analyses. A
stepwise variable entry procedure se-
lected eight of the original 20 vari-
ables for inclusion in the descriptive
discriminant model (table 5). Subse-
quent DDA kept all but three
(SLOPE, CWDl, and TAPSEN) in the
model. The resultant discriminant
function explained 49% of the be-
tween-group variance and had sig-
nificant power in discriminating be-
tween groups (F = 6.87, P < 0.0001).
This function can be interpreted ecol-
ogically to describe a gradient that
ranges from low elevations with
many pine, few spruce and large fir,
and infrequent CWD5 without sala-
manders, to higher elevations, few
pine, many spruce and large fir, and
abundant CWD5 that contain sala-
manders. ELEV had the highest dis-
criminating power (r^ = 0.64). PDA
correctly classified 96% of the 54
transects without salamanders and
58% of the 26 transects with salaman-
ders.

The 26 occupied transects and ad-
ditional searches produced 620 Sac-
ramento Mountain salamanders.
Ninety-five percent (589) were dis-
tributed among the four major cover
classes as follows: CWD, 377 (64%);
ROCK, 127 (22%); BARK, 58 (10%);
and FWD, 27 (4%). Fourteen sala-
manders (2%) were found under as-
pen logs and 17 salamanders (3%)
were above or below surface litter.

The frequency of salamanders associ-
ated with CWD in the three decay
classes was CWDl— 13%; CWD3—
62%; CWD5— 25%. Of 233 salaman-
ders found on transects, 209 sala-
manders were associated with one of
the four classes of cover items. Ex-
amination of cover item availability
and use for these salamanders re-
vealed that salamanders are associ-
ated with some cover items dispro-
portionate to their availability (X^ =
59.9, df = 3, P < 0.001). In particular,
Aneides was found in association
with FWD proportionately less fre-
quent than expected, and used well-
decayed and moderately decayed
logs to a greater extent than expected
(X2 = 62.1, df =2, P< 0.001).

Discussion

Jemez Mountains Salamander

While canonical analysis did not dis-
criminate between transects with and

without salamanders, it did identify
steep slopes as the most useful vari-
able in determining the occurrence of
Jemez Mountains salamanders. It is
possible that steep slopes contain
more interstitial spaces in the soil
than do shallower slopes. The soils of
steep slopes may be less compacted
than those of more gentle slopes due
to the combined effects of gravity,
and movement of water and soil. As
a consequence of steep slope and the
presence of underlying volcanic rock
characteristic of the Jemez Mountains
(Burton 1982), spaces within this ma-

Tobie 3.— Correlations of habitat
variables with discriminant scores
for transects with and without
Jemez Mountains salamanders.

Mnemonic

DFl

SLOPE

TPINE

LSPRUCE

0.73
0.52
0.35

Table 4.--Comparlson of habitat variables measured on transects with and
without Sacramento Mountain salamanders, Lincoln National Forest, 1986-
1987. Significance Is based on one-way analysis of variance. Mnemonic
codes for habitat variables are explained In Table 1 .

Transects (N = 26)
with salamanders

Transects (N = 54)
without salamanders

Mnemonic

x ± se (Range)

X ± se (Range)

Significa

ELEV

2779

+ 17.6

(2618-2890)

2682

+

8.7

(2450-2792) "

SLOPE

39

+ 2.7

(21-65)

41

+

1.6

(17-70)

♦*

CANOPY

72

+ 1.3

(59-88)

71

+

1.3

(53-90)

NS

TFIR

67

+ 6.3

(8-122)

64

+

4.0

(14-144)

NS

TSPRUCE

17

+ 7.6

(0-186)

1

+

0.6

(0-30)

*•

TPINE

7

+ 2.1

(0-50)

22

+

2.3

(0-71)

♦

TASPEN

14

+ 4.1

(0-74)

17

+

3.3

(0-107)

NS

TOAK

5

+ 2.4

(0-59)

18

3.8

(0-104)

NS

TSNAGS

24

+ 2.8

(6-56)

25

+

2.5

(1-106)

NS

TNOD

33

+ 7.7

(4-180)

34

+

5.6

(0-222)

NS

ROCK

7

+ 1.7

(0-33)

7

+

0.9

(0-29)

NS

FWD

6

+ 0.6

(2-13)

5

+

0.5

(0-14)

NS

BARK

1

+ 0.3

(0-6)

1

+

0.2

(0-10)

NS

CWD

12

+ 1.2

(4-24)

8

+

0.8

(0-26)

NS

"P< 0.005
'P < 0.05

58

130

SPRUCE

50 -I

40

30

20 -
10 -
0

130

WITH SALAMANDERS
WITHOUT SALAMANDERS

120 A
80

FIR

50 4-/-^

40

30

20 -

10 -

<20CM 20-50 CM > 50 CM

D.B.H.

Figure 4.— Comparisons of average size classes (d.b.h.) of spruce and fir on transects witti
and without Sacramento Mountain salamanders. Boxes indicate 95% confidence Inten/als
for ttie mean. Levels of significance indicated by asterisks are 0.05 (*) and 0.005 (**).

trix of rocky soil may provide refugia
for salamanders during inhospitable
times and, thus, may provide a clue
to the survival of this salamander in
the harsh environment of the Rocky
Mountains. The largest concentra-
tions of P. neomexicanus have been
found in association with talus slopes
(Whitford and Ludwig 1975, Clyde
Jones pers. comm.), which are also
important to many other western Ple-
thodon (Brodie 1970). Other pletho-
dontids are virtually restricted to ar-
eas with a loose rocky soil (Aubry et
al. 1987, French and Mount 1978,
Herrington and Larsen 1985, Jaeger
1971).

The variables selected by canoni-
cal analysis showed some predictive
value. Although three transects with-
out salamanders were misclassified
by PDA as transects with
salamanders, Plethodon was found in
areas adjacent to the transects. The
two transects misclassified as
transects without salamanders had
values for TPINE and LSPRUCE
closer to values usually associated
with transects without salamanders.
Because a larger percentage of
transects without salamaders were
correctly classified by PDA, these
three variables may better describe
the conditions under which salaman-
ders are absent from an area, rather
than describing favorable conditions
under which they would occur.

The limited discriminatory and
predictive power of the variables se-

Table 5.— Correlations of habitat
variables v/ith discriminant scores
for transects with and v^lthout Sac-
ramento Mountain salamanders.

Mnemonic

DFl

ELEV

TSPRUCE

TPINE

CWD5

LFIR

CWDl

SLOPE

TASPEN

0.55
0.42
-0.47
0.44
0.34
-0.05
-0.06
-0.02

59

lected by multivariate techniques
may reflect our inability to reliably
and consistently detect the presence
of Plethodon at a site. We believe that
our ability to detect salamanders is
fairly good and repeatable, but we
realize that environmental factors
can influence the relative numbers of
salamanders. During repeated visits
to the same sites, Plethodon was more
abundant when we searched under
wet conditions, and other studies
have reported a significant correla-
tion between movement and activity
of salamanders, and precipitation
(Barbour et al. 1969, Kleeberger and
Werner 1982, MacCullough and
Bider 1975). Low densities and
patchiness of P. neomexicanus popula-
tions also can hinder detection of the
animal. In comparison with densities
of red-backed salamanders, P. cin-
ereus, (0.9-2.2 individuals/ m^; Heat-
wole 1962, Jaeger 1980), our density
estimates for Jemez Mountains sala-
manders are extremely low (0.03 in-
dividuals/m^). Although Williams
(1972) reported estimates of Jemez
Mountains salamanders ten times
greater than ours, he noted that their
distribution was spotty.

A better fit to a discriminant
model might be obtained by includ-
ing variables that we did not meas-
ure, e.g., fire and logging history and
soil characteristics (moisture, pH,
and compaction). Williams (1976)
suggested that logging may have
eliminated Jemez Mountains sala-
manders from part of Peralta Canyon
due to dry conditions resulting from
removal of most of the canopy. How-
ever, there was no documentation
that salamanders occurred at the site
prior to logging. Soil characteristics,
which can be affected by fire and log-
ging practices (Childs and Hint 1987,
DeByle 1981, Krag et al. 1986), also
can influence the distribution of ple-
thodontid salamanders, that occupy
the soil-litter interface. Plethodon cin-
ereus was excluded from 27% of for-
est habitat in eastern deciduous for-
ests because of low soil pH (Wyman
and Hawksley-Lescault 1987), while

the distributions of up to 10 amphibi-
ans in southeastern New York were
significantly influenced by soil pH
and moisture (Wyman 1988).

Salamanders also may be absent
from a given site for reasons other
than unsuitability of habitat. For ex-
ample, access to a particular area by
salamanders may be impossible due
to the unsuitability of the area that
surrounds it, e.g., dry, open field. Or,
a climatic event may have eliminated
salamanders from a given area with-
out sufficient time occurring for them
to recolonize the site.

Sacramento Mountain
Salamander

The variables selected by canonical
analysis were able to discriminate be-
tween transects with and without
salamanders. However, these vari-
ables had limited predictive value.
Although a larger percentage of
transects without salamanders were
correctly classified by PDA, there is
still a one-in-five chance of being
wrong in predicting that salaman-
ders are absent from a site. For most
management decisions, this level of
uncertainty will not be acceptable,
and ground-truth searches will have
to be made.

High elevation was the best pre-
dictor of the presence of Sacramento
Mountain salamanders (table 5).
Weigmann et al. (1980) also found
significantly more Sacramento
Mountain salamanders on transects
at higher elevations. The higher ele-
vations of the Sacramento Mountains
experience greater rainfall, cooler
temperatures, and lower
evapotranspiration rates than the
lower elevations and therefore may
be more hospitable to plethodontid
salamanders. The low critical ther-
mal maximum of Aneides probably
reflects adaptations to the low tem-
peratures characteristic of their mi-
crohabitat (Whitford 1968) and may
restrict salamanders to high eleva-
tions.

Aneides is often present where the
best habitat predictors indicate they
should not occur. While high-eleva-
tion, wet, north-facing slopes with a
mature mixed-conifer forest do har-
bor Aneides, salamanders are also
found less predictably in areas that
may be drier and more exposed than
the model would indicate. With the
exception of elevation, the ranges of
habitat variables on transects occu-
pied by salamanders are not strik-
ingly different from those on plots
without salamanders (table 4). This
overlap may be due to factors not
measured, e.g., fire and logging his-
tory, and it may show an ability of
salamanders to persist after habitats
have been altered.

Management Guidelines

Our data show that, despite some
predictive power of the habitat vari-
ables, the level of uncertainty in pre-
dicting salamander occurrence may
preclude their use by the USFS. At
this time, we feel the best survey
technique for salamanders is ground-
truth surveys in wet weather during
the activity season of each species.
Under proper conditions, both spe-
cies are easy to find and relatively
unskilled persons can be quickly
trained to survey habitats. Our im-
pression was that Plethodon was
more difficult to survey, because it
tended to retreat underground dur-
ing dry periods. Aneides, however,
can usually be found even during ex-
tended dry periods.

Our attempts to explain the ab-
sence of salamanders from a given
area, i.e., potential difficulty of de-
tecting all salamanders present, and
low density or patchy distribution of
populations, may overlook the possi-
bility that absence is not solely due to
unsuitable habitat. Absence does not
necessarily mean avoidance, but may
be due to insufficient time for the
animal to recolonize an area, or inac-
cessibility of a suitable area due to
unsuitable habitat surrounding it.

60

In lieu of specific recommenda-
tions, the USPS needs interim man-
agement guidelines to protect the
salamanders from population de-
clines. We suggest the following
steps:

1. Salamander surveys should
be made on specific sale ar-
eas as early in the planning
process as possible. The
USPS could maintain a team
of seasonal employees for
such surveys and for other
activities related to endan-
gered species.

2. To the extent possible, inten-
sive logging operations (i.e.,
clearcuts, seed-tree cuts, trac-
tor logging) should not be
conducted in areas occupied
by salamanders. Cable log-
ging in winter, when the
ground is frozen and the
salamanders are under-
ground, is probably the least
damaging activity. In com-
parison, tractor logging on
wet soils can compact the
soil to such a degree that
salamanders cannot use it.

3. Modifications of current
practices, such as leaving
slash where it falls or leaving
as much canopy as possible,
help prevent the soil surface
from drying out and will
probably benefit salaman-
ders.

4. Because current timber har-
vest schedules will inevitably
lead to younger-aged stands
with few or only small
downed logs, a mix of young
and old logs should be main-
tained to ensure short-term
and long-term habitat com-
ponents. Old logs provide
cover to Aneides and Pletho-
don, while younger logs are
potential sources of cover in
future years.

Other studies provide some evi-
dence for negative effects of logging
on amphibian populations (Bennet et
al. 1980, Blymer and McGinnes 1977,
Bury 1983, Gordon et al. 1962, Her-
rington and Larsen 1985, Pough et al.
1987, Ramotnik 1988, Staub 1986, and
Williams 1976) and we suspect that
intensive logging, slash removal, and
burning will reduce or eliminate
populations of Plethodon neomexica-
nus and Aneides hardii. Only intensive
observations of salamander popula-
tions throughout the logging cycle
will provide the information needed
to make management recommenda-
tions. These studies are in progress,
but may require years before defini-
tive results are available to assess the
effects of logging on Plethodon and
Aneides.

Acknowledgments

We thank the following U.S. Porest
Service personnel: Santa Pe National
Porest — R. Alvarado, D. Delorenzo,
and M. Morrison; Lincoln National
Porest — R. Dancker, D. Edwards, S.
Lucas, J. Peterson, and D. Zaborske;
and L. Pisher, Regional Office. Much
of the funding was provided by the
U.S. Porest Service (Southwestern
Region).

Pield personnel included M. J. Al-
tenbach, R. R. Beatson, A. Bridegam,
R. B. Bury, C. Campbell, S. Com, T.
H. Pritts, B. E. Smith, and M. C.
Tremble. S. Stefferud (Endangered
Species, U.S. Pish & Wildlife Service)
and C. Painter (Endangered Species
Program, New Mexico Department
of Game & Pish) were welcome field
companions. S. Corn provided pho-
tographs. K. Aubry, K. Buhlmann, S.
Corn, C. K. Dodd, and C. Painter
provided helpful criticism of earlier
drafts.

Literature Cited

Aubry, Keith B., Clyde M. Senger,
and Rodney L. Crawford. 1987.

Discovery of Larch Mountain sala-
manders Plethodon larselli in the
central Cascade Range of Wash-
ington. Biological Conservation
42:147-152.

Barbour, Roger W., James W. Har-
din, James P. Schafer, and Michael
J. Harvey. 1969. Home range,
movements, and activity of the
dusky salamander, Desmognathus
fuscus. Copeia 1969:293-297.

Bennet, Stephen H., J. Whitfield Gib-
bons, and Jill Glanville. 1980. Ter-
restrial activity, abundance and
diversity of amphibians in differ-
ently managed forest types.
American Midland Naturalist
103:412-416.

Blymer, Michael J. and Burd S.
McGinnes. 1977. Observations on
possible detrimental effects of
clearcutting on terrestrial amphibi-
ans. Bulletin of the Maryland Her-
petological Society 13:79-83.

Brodie, Edmund D., Jr. 1970. West-
ern salamanders of the genus Ple-
thodon: systematics and geo-
graphic variation. Herpetologica
26:468-516.

Burton, Barry W. 1982. Geologic evo-
lution of the Jemez Mountains and
their potential for future volcanic
activity. Los Alamos National
Laboratory, LA-8795-GEOL. Los
Alamos, New Mexico. 31 p.

Bury, R. Bruce. 1983. Differences in
amphibian populations in logged
and old growth redwood forest.
Northwest Science 57:167-178.

Bury, R. Bruce and Martin G. Ra-
phael. 1983. Inventory methods
for amphibians and reptiles, p.
416-419. In Renewable resource
inventories for monitoring
changes and trends. J. P. Bell and
T. Atterbury, eds. Oregon State
University, Corvallis.

Campbell, Howard W. and Stephen
P. Christman. 1982. Pield tech-
niques for herpetofaunal commu-
nity analysis, p. 193-200. In Herpe-
tological communities. Norman J.
Scott, ed. U.S Pish & Wildlife
Service, Wildlife Research Rep. 13.

Castetter, Edward, P. 1956. The vege-

61

tation of New Mexico. New Mex-
ico Quarterly 26:256-285.

Childs, S. W. and L. E. Hint. 1987. Ef-
fect of shadecards, shelterwoods,
and clearcuts on temperature and
moisture environments. Forest
Ecol. and Manage. 18:205-217.

Corn, P. Steven and R. Bruce Bury. In
press. Sampling terrestrial am-
phibians and reptiles. In Popula-
tion Monitoring Techniques for
Wildlife in Pacific Northwest For-
ests. A. B. Carey and L. F. Ruggi-
ero, eds. USDA Forest Service
General Technical Report. Pacific
Northwest Research Station, Port-
land, Oregon.

DeByle, Norbert V. 1981. Clearcut-
ting and fire in the larch/ Douglas-
fir forests of western Montana-a
multifaceted research summary.
General Technical Report INT-99.
USDA Forest Service, Intermoun-
tain Forest and Range Experiment
Station. 73 p.

Franklin, J. F., K. Cromack, Jr., W.
Denison, A. McKee, C. Maser, J.
Sedell, F. Swanson, and G. Juday.
1981. Ecological characteristics of
old-growth Douglas-fir forests.
General Technical Report PNW-
118. USDA Forest Service, Port-
land, Oregon. 48 p.

French, Thomas W. and Robert H.
Mount. 1978. Current status of the
Red Hills salamander,
Phaeognathus huhrichti Highton,
and factors affecting its distribu-
tion. Journal of the Alabama
Academy of Science 49:172-179.

Gordon, Robert E., James A. MacMa-
hon, and David B. Wake. 1962.
Relative abundance, microhabitat
and behavior of some Southern
Appalachian salamanders. Zool-
ogica 47:9-14.

Heatwole, Harold. 1962. Environ-
mental factors influencing local
distribution and activity of the
salamander, Plethodon cinereus.
Ecology 43:460-472.

Herrington, Robert E. and John H.
Larsen. 1985. Current status, habi-
tat requirements and management
of the Larch Mountain salamander

Plethodon larselli Bums. Biological
Conservation 34:169-179.

Hubbard, John P., Marshall C. Con-
way, Howard Campbell, Gregory
Schmitt, and Michael D. Hatch.
1979. Handbook of species endan-
gered in New Mexico. New Mex-
ico Department of Game & Fish.

Jaeger, Robert G. 1971. Moisture as a
factor influencing the distributions
of two species of terrestrial sala-
manders. Oecologia 6:191-207.

Jaeger, Robert G. 1980. Microhabitats
of a terrestrial forest salamander.
Copeia 1980:265-268.

Kleeberger, Steven R. and J. Kirwin
Werner. 1982. Home range and
homing behavior of Plethodon cin-
ereus in northern Michigan. Copeia
1982:409-415.

Krag, R., K. Higginbotham, and R.
Rothwell. 1986. Logging and soil
disturbance in southeast British
Columbia. Canadian Journal of
Forest Research 16:1345-1354.

Kunkel, Kenneth E. 1984. Tempera-
ture and precipitation summaries
for selected New Mexico locations.
Climate report. Office of State Cli-
matologist. New Mexico Depart-
ment of Agriculture, Las Cruces,
New Mexico.

Lemmon, Paul E. 1956. A spherical
densiometer for estimating forest
overstory density. Forest Science
2:314-320.

Lowe, Charles H., Jr. 1950. The sys-
tematic status of the salamander
Plethodon hardii, with a discussion
of biogeographic problems in
Aneides. Copeia 1950:92-99.

MacCullough, Ross D. and J. R.
Bider. 1975. Phenology, migra-
tions, circadian rhythm and the ef-
fect of precipitation on the activity
of Eurycea b. bislineata in Quebec.
Herpetologica 31:433-439.

Rough, F. Harvey, Ellen M. Smith,
Donald H. Rhodes, and Andres
Collazo. 1987. The abundance of
salamanders in forest stands with
different histories of disturbance.
Forest Ecol. and Manage. 20:1-9.

Ramotnik, Cynthia A. 1986. Status
report Plethodon neomexicanus

Stebbins and Riemer, Jemez
Mountains salamander. Submitted
to the U.S. Fish & Wildlife Service,
Region 2, Office of Endangered
Species, Albuquerque, NM. 55 p.

Ramotnik, Cynthia A. 1988. Habitat
requirements and movements of
Jemez Mountains salamanders,
Plethodon neomexicanus. M.S. the-
sis, Colorado State University,
Fort Collins. 84 p.

Raphael, Martin G. and K. V. Rosen-
berg. 1983. An integrated ap-
proach to wildlife inventories in
forested habitats, p. 219-222. In
Renewable resource inventories
for monitoring changes and
trends. J. F. Bell and T. Atterbury,
eds. Oregon State University,
Corvallis.

Reagan, Douglas P. 1972. Ecology
and distribution of the Jemez
Mountains salamander, Plethodon
neomexicanus. Copeia 1972:486-492.

SAS Institute Inc. 1982. SAS user's
guide: statistics, 1982 ed. Gary,
North Carolina: SAS Institute Inc.
584 p.

Scott, Norman J. Jr., Cynthia A. Ra-
motnik, Marilyn J. Altenbach, and
Brian E. Smith. 1987. Distribution
and ecological requirements of
endemic salamanders in relation
to forestry management. Summary
of 1987 activities. Part 2: Santa Fe
National Forest. Final report sub-
mitted to the USDA Forest Serv-
ice, Albuquerque, New Mexico.
33 p.

Shelford, Victor E. 1963. The ecology
of North America. University of Il-
linois Press, Urbana. 610 p.

Snedecor, George W. and William G.
Cochran. 1967. Statistical methods.
6th ed. The Iowa State University
Press, Ames, Iowa. 593 p.

Staub, Nancy. 1986. A status survey
of the Sacramento Mountain
salamander, Aneides hardii, with an
assessment of the impact of log-
ging on salamander abundance.
Final report submitted to the U.S.
Fish and Wildlife Service, Region
2, Endangered Species Program,
Albuquerque, New Mexico. 42 p.

62

U.S. Fish & Wildlife Service. 1986.
Distribution and ecological re-
quirements of endemic salaman-
ders in relation to forestry man-
agement. Part 2: Santa Fe National
Forest. Report submitted to the
U.S. Forest Service, Albuquerque,
New Mexico. 40 p.

Weigmann, Diana L., Mark Hakkila,
Karl Whitmore, and Richard A.
Cole. 1980. Survey of Sacramento
Mountain salamander habitat of
the Cloudcroft and Mayhill dis-
tricts of the Lincoln National For-
est. Final report submitted to the
U.S. Forest Service, Lincoln Na-
tional Forest. 45 p.

Whitford, Walter G. 1968. Physiologi-
cal responses to temperature and
desiccation in the endemic New
Mexico plethodontids, Plethodon
neomexicanus and Aneides hardii.
Copeia 1968:247-251.

Whitford, Walter G. and John
Ludwig. 1975. The biota of the
Baca geothermal site. Report for
Union Oil Co. -Geothermal Divi-
sion. 140 p.

Williams, Byron K. 1983. Some obser-
vations on the use of discriminant
analysis in ecology. Ecology
64:1283-1291.

Williams, Stephen R. 1972. Repro-
duction and ecology of the Jemez
Mountains salamander, Plethodon
neomexicanus. M. S. thesis, Univ. of
New Mexico, Albuquerque. 98 p.

Williams, Stephen R. 1976. Compara-
tive ecology and reproduction of
the endemic New Mexico pletho-
dontid salamanders, Plethodon ne-
omexicanus and Aneides hardii. Ph.
D. thesis. University of New Mex-
ico, Albuquerque. 152 p.

Wyman, Richard L. 1988. Soil acidity
and moisture and the distribution
of amphibians in five forests of
southcentral New York. Copeia
1988:394-399.

Wyman, Richard L. and Dianne S.
Hawksley-Lescault. 1987. Soil
acidity affects distribution, behav-
ior, and physiology of the
salamander Plethodon cinereus.
Ecology 68:1819-1827.

utilization Of Abandoned
Mine Drifts and Fracture
Caves By Bats and
Salamanders: Unique
Subterranean Habitat In The
Ouachita Mountains^

Abstract.— Twenty-seven abandoned mine drifts
and four fracture caves constitute one of the most
unique habitats in and adjacent to the Ouachita
National Forest, an area devoid of solutional caves.
Six species of salamanders and nine species of bats
were found to utilize these areas.

David A. Saugey,^ Gary A. Heidt,^ Darrell R.
Heathi^

Caves and mines play an important
role in the ecology of many species,
serving as permanent or temporary
habitats. Culver (1986) stated, "the
variety of species that depends on
caves during some critical time in
their life cycle, such as hibernation in
bats, is impressive and usually
underestimated." To this statement,
we add mines.

Bear Den Caves are located in
Winding Stair Mountain, LeFlore
County, in southeastern Oklahoma.
These four caves occur in an outcrop
belt of a massive sandstone unit and
were formed by a number of factors,
the most important being gravita-
tional sliding and slumpage of sand-
stone. These four caves have more
than 365 meters of mapped passage-
way and represent the only known
caves in the Ouachita National Forest
(Puckette 1974-75).

Additional subterranean habitat
was formed from 1870 to 1890, when
the area extending west from Hot
Springs to Mena, Arkansas was the
scene of a gold, lead, silver and zinc

' Paper presented at symposium. Man-
agement of Amphibians, Reptiles, and
Small Mammals in North America. (Flag-
staff , AZ, July 19-2h 1988).

^David A. Saugey is a Wildlife Biologist,
U.S. Forest Sen/ice, Ouachita National For-
est, Hot Springs, AR. 71902.

^GaryA. Heidt is Professor of Biology,
University of Arkansas at Little Rock, AR.
72204.

"Darrell R. Heath is an Undergraduate
Student, University of Arkansas at Little
Rock, AR. 72204.

rush. During the period of greatest
activity, 1885 to 1888, over a dozen
gold mines were in operation, rang-
ing from shallow test holes to exten-
sive linear and L-shaped drifts ex-
tending up to 150 meters into the
surrounding mountains (Harrington
1986, Hudgins 1971, U.S. Army
Corps of Engineers 1980). The "gold
and silver boom" effectively ended
with the issuance of a report which
in effect stated there were no pre-
cious metals in paying quantities to
be found in the area (Branner 1888).
Soon thereafter, many mines were
abandoned as prospectors moved
West (Harrington 1986, Hudgins
1971). Through the years, other min-
erals, such as manganese and mer-
cury, have been mined from the Ou-
achitas resulting in the excavation of
numerous additional drifts; but for a
variety of reasons, most have been
abandoned (Clardy and Bush 1976,
Stone and Bush 1984). The legacy of
these mining activities has not been
riches and new-found wealth, but the
creation of unusual and unique wild-
life habitat.

The objectives of this study were
to review, compile, and consolidate
existing literature concerning utiliza-
tion of caves and mine drifts by bats
and salamanders in the Ouachita
Mountains. In addition, we provide
new data and propose recommenda-
tions concerning management of
caves and mines in the Ouachita Na-
tional Forest and on other public and
private lands.

METHODS

During the past six years, 27 aban-
doned mines in Garland (8),
Montgomery (3), Pike (4) and Polk
(12) counties, Arkansas (fig. 1) were
located and visited a minimum of
eight times (at least once each sea-
son). In several cases, where endemic
or Category II (U.S. Federal Register
1985) species occurred or breeding
populations were found, mines were
visited much more often. Mist net-
ting of entrances for bats was con-
ducted in spring, summer, and fall.
Bear Den Caves came to our atten-
tion during 1987 and were visited
several times. Collections were mini-
mal (mines only) and voucher speci-

OKLA

Figure 1 .—Location of Ouachita National
Forest (backslastied area) and study area
(crosshiatchied area).

64

mens are located in the Vertebrate
Collections at the University of Ar-
kansas at Little Rock and Arkansas
State University.

Following McDaniel and Smith
(1976), we include the probable eco-
logical position of the species in the
cave and mine environments. This is
followed by comments concerning
the status or life history of each spe-
cies. Following Barr (1963) and
McDaniel and Smith (1976) the terms
"troglophile" (commonly found in
caves), "trogloxene" (may be com-
mon in caves but must leave to com-
plete their life history), and "acciden-
tal" (unable to survive long in the
cave environment) have been em-
ployed in the species accounts.

RESULTS

Nine species of bats and six species
of salamanders were found to utilize
caves and abandoned mine drifts
during some portion of their annual
cycles.

Annotated List of Bats and
Salamanders Utilizing Caves and
Abandoned Mine Drifts

CLASS AMPHIBIA

Order Urodela

Family Plethodontidae

Desmognafhus brimleyorum
(Stefneger). Troglophile.

Means (1974) stated the Ouachita
dusky salamander was confined to
rocky, gravelly, streams in the Ou-
achita Mountains. Rock falls along
the upper portions of streams repre-
sented particularly good adult habi-
tat. This species was most abundant
where water percolated through
rocky substrate in streambeds and
along stream sides. Description of
egg clutch characteristics and
stream/streamside deposition were

given by Means (1974) and Trauth
(1988) provided descriptions of
deposition sites in seepage areas dur-
ing the severe summer drought in
1980. Heath et al. (1986) reported the
occurrence of this endemic salaman-
der in four drifts, with egg clutches
deposited on the underside of rocks
in one mine and the presence of lar-
vae in two others. In those mines
with larvae, pools contained abun-
dant leaf litter and isopods. On one
occasion, larvae were observed feed-
ing on isopods. Since these observa-
tions were made, numerous addi-
tional visits to these four mines re-
vealed the presence of D^smognathus
when epigean conditions would be
considered ideal. The pools within
these and other drifts are the result
of seepage through walls which, in
some instances, provided sufficient
volumes of water to have small
streams flowing from their entrances.
However, unlike the preferred,
gravel-bottomed stream habitat,
pools typically exhibited silted sub-
strates with very little rubble and
few rocks large enough for egg at-
tachment.

Eurycea multiplicata (Cope).
Trogloptiile.

The many-ribbed salamander is pri-
marily an aquatic species endemic to
the Interior Highland region and ad-
jacent areas that contain suitable
habitat. It may be found under
stones, logs, and other debris in
clear, rock or gravel-bottomed
streams (Bishop 1943, Ireland 1971,
Reagan 1974). It inhabits essentially
the same habitat as Desmognathus
brimleyorum (Strecker 1908). Hurter
and Strecker (1909) noted
Desmognathus eating Eurycea indi-
viduals with which they were con-
fined. Heath et al. (1986) reported
both larvae and adults in two mines
and in one, larvae shared the same
pools with Desmognathus larvae. Both
mines contained shallow streams
with a gravel substrate. One addi-

tional mine contained larvae of this
species. A seepage stream in this
mine was approximately five centi-
meters wide, one centimeter deep,
and extended a distance of sixty
centimeters before dropping into a
large pool at the entrance. The pool
connected directly to an epigean
stream.

Plethodon caddoensis Pope and
Pope. Trogloptiile.

Large aggregations of the endemic
Caddo Mountain salamander using
drifts as refugia to escape heat and
dryness during summer and fall
were first reported by Saugey et al.
(1985). Over 100 individuals were
discovered in each of two drifts,
from June through September 1983.
Subsequent visits to these and other
drifts revealed limited use of three
additional drifts and use of one of
the original aggregation sites for egg
deposition and breeding (Heath et al.
1986). Since these observations were
made, summer aggregations of this
salamander have numbered as high
as 383 individuals and additional egg
clutches have been observed and
monitored. Known only from the
Novaculite Uplift area of the Ou-
achita Mountains in Howard,
Montgomery, and Polk counties in
Arkansas (Blair and Lindsey 1965,
Robison and Smith 1982), this sala-
mander and its habitat are of special
concern to the Arkansas Natural
Heritage Commission (ANHC)
(Smith 1984). In 1985, the U.S. Fish
and Wildlife Service (USFWS) desig-
nated it a Category II species. In
1986, the U.S. Forest Service (Ou-
achita National Forest) began infor-
mal consultation with the USFWS
(Jackson, Mississippi, Endangered
Species Field Station) and requested
field assistance from the ANHC con-
cerning preservation of critical mine
aggregation sites and protection of
their vulnerable populations. Place-
ment of a gate at one sensitive site is
planned in 1988 (fig. 2).

65

Plethodon glutinosus glutinosus
(Green). Troglophile.

The slimy salamander, a woodland
species, is widely distributed, ex-
ploiting virtually every available ter-
restrial habitat. This species is com-
monly found under rocks, in and
under well rotted logs and stumps,
and buried deep in moist layers of
leaf litter. During hotter and drier
portions of the year, they usually re-
treat deeper into the substrate. Al-
though primarily epigean, this sala-
mander has been reported to use
caves for aggregation sites, egg depo-
sition and brooding, and escape from
inhospitable surface environmental
conditions (Barnett 1970, Noble and
Marshall 1929). Heath et al. (1986)
reported this salamander from five
mines; two contained breeding popu-
lations and brooding behavior has
been observed several times. Subse-
quent observations have confirmed
another of the five mines as an egg
deposition and brooding site. One of
the mines reported with a breeding
population (Heath et al. 1986) is the
site of an annual aggregation of slimy
salamanders exceeding 600 individu-
als. A gate (fig. 2) has been con-
structed by the U.S. Army Corps of
Engineers to protect this population.
Continuing studies to determine the
effect of gating will allow compari-
son of pre- and post-gating data.

Plethodon ouachitae Dunn and
Heinze. Troglophile.

Endemic to the Ouachita Mountains
of Arkansas and Oklahoma, the Rich
Mountain salamander may be found
living beneath rotting logs and
stumps. However, it lives primarily
under pieces of sandstone on heavily
overgrown talus north slopes (Black
1974, Dunn and Heinze 1933, Pope
and Pope 1951, Sievert 1986). Reagan
(1974) listed this species as "endan-
gered and vulnerable" in Arkansas.
Ashton (1976) and Black (1980) both
considered this salamander "threat-

ened" in Oklahoma. Sievert (1986)
proposed it as a species of "special
concern," conditional on his recom-
mendations concerning silvicultural
practices on National Forest lands.
Black (1974) reported this salaman-
der in Bear Den Caves where they
were found throughout, but most
commonly within the first 19 meters
or twilight zone. A small juvenile
with a snout-vent length (SVL) of < 7
mm was found in an entrance and
the presence of numerous juveniles
with SVLs of > 30mm may indicate
egg deposition and brooding activi-
ties. One of the authors (DAS) visited
these caves in December, 1987 and
observed one adult Rich Mountain
salamander near the entrance of one
cave. An additional visit in June 1988
resulted in the observation of 30+
salamanders of various size classes.
Considerable human refuse and a
well worn path indicated substantial
numbers of visitors. Considering the
uniqueness of this area and the Cate-
gory II status of this salamander,
steps are being taken to exclude ex-
cessive visitation and protect this
population from vandalism and
overcollection. These caves are util-
ized by the small-footed bat, Myotis
leibii, (Caire 1985) also a Category II
species.

Plethodon serratus Grobman.
Troglophile.

The endemic Ouachita Red-backed
salamander is commonly found be-
neath rocks, logs, and in leaf litter at
all elevations throughout the Ou-
achita Mountains. This species has
been observed in one mine on two
separate occasions. In both cases, it
has been in association with large
aggregations of the Caddo Mountain
salamander during extremely dry
epigean conditions. Reagan (1974)
frequently found this species in asso-
ciation with the Caddo Mountain
and Rich Mountain salamanders.

CLASS MAMMALIA
Order Chiroptera

Family Vespertilionidae

Myotis austroriparius (Rhoads).
Trogloxene.

The first Arkansas specimens of the
southeastern bat were collected from
one of several drifts located 12 miles
northwest of Hot Springs, Garland
County, Arkansas (Davis et al. 1955).

66

At the time of collection (November
1952) and during a subsequent visit,
this species was found in association
with the little brown bat, Myotis luci-
fugus, and Keen's bat, Myotis keenii.
This particular drift was inundated
by the filling of Lake Ouachita in
1955 and, since that time, no addi-
tional specimens have been observed
in nearby drifts. The second occur-
rence of this species in the Ouachita
Mountain area was from abandoned
Cinnabar mines located on an penin-
sula in Lake Greeson, Pike County,
Arkansas (Heath et al. 1986). During
a winter visit (January 1984) over 150
individuals of both red and gray
color phases were observed in deep
torpor. A subsequent early spring
visit (March 1986), revealed 15 indi-
viduals. During December, 1986,
only a few scattered individuals were
found. According to personnel famil-
iar with the drift, considerable hu-
man visitation and disturbance may
have been the cause of sharp decline
in use of this excavation. Mumford
and Whitaker (1982) suggested the
southeastern bat does not tolerate
disturbance and is likely to change its
roosting and hibernation sites quite
readily. Caire (1985) did not report
this species, but records exist for the
Little River drainage in southeastern
Oklahoma (Glass and Ward 1959).
The southeastern bat is listed as a
Category II species in the U.S. Fed-
eral Register (1985).

Myotis keenii (Merriam).
Trogloxene.

Utilization of caves and mines by
Keen's bat has been well documented
(Barbour and Davis 1969, Heath et al.
1986, McDaniel and Gardner 1977).
Sealander and Young (1955) first re-
ported the occurrence of Keen's bat
from the Ouachita Mountain area
when three specimens were collected
from the drift located 12 miles north-
west of Hot Springs. Caire (1985)
mist-netted a number of specimens at
Bear Den Caves; the majority were

males with a few postlactating fe-
males. Heath et al. (1986) found this
bat in 12 drifts. The largest hibernat-
ing aggregation consisted of 12 bats,
including both males and females.
Normally, from one to three indi-
viduals (usually males) were found
hibernating in small cracks and crev-
ices near entrances. On occasion, two
have been found together in drill
holes in ceilings and walls and, less
frequently, individuals were ob-
served hanging in the open. The larg-
est non-hibernating cluster was 57
females found in the spring of 1985.
Three were collected and found to be
pregnant (drifts were not used as
maternity roosts). Although utilized
more frequently during winter
months, these drifts contained from
one to several Keen's bats through-
out most of the year.

Myotis ieibii (Audubon and
Bachman). Trogloxene.

The small-footed bat is very common
and widespread in the western
United States where it readily uses
caves and mines for hibernation. In
the eastern United States it is consid-
ered to be rare (Barbour and Davis
1969, Smith 1984). Caire (1985) re-
ported mist-netting four males, three
adults and one subadult, at Bear Den
Caves. Specimens collected in Sep-
tember had descended testes. Heath
et al. (1986) did not record this bat
from drifts in Arkansas. According
to Barbour and Davis (1969), the only
known winter habitats for this spe-
cies are caves and mines. Preferred
hibernation sites are near entrances
where temperatures drop below
freezing and humidity is relatively
low. Abandoned drifts in the Ou-
achitas generally have one, small,
partially collapsed entrance which
ensures relatively warm interiors (18
C) with high humidities, which is un-
suitable hibernating habitat. Mist-
netting of creeks and drift entrances
and subsequent winter visits to drifts
have been unsuccessful in locating

this bat. Caire (1985) indicated this
species is probably restricted to cave
areas. Thus, the few caves in south-
eastern Oklahoma are critical to the
species survival and are in need of
protection. The small-footed bat is a
Category II species (U.S. Federal
Register 1985).

Myotis lucifugus (LeConte).
Trogloxene.

The little brown bat appears to be
extremely rare in the Ouachita
Mountains. It had been reported
from one drift by Sealander and
Young (1955), but an additional
specimen was reported by Heath et
al. (1986) from a drift in Arkansas. In
Oklahoma, the little brown bat has
been collected only from Beavers
Bend State Park in the southeastern
part of the state (Glass and Ward
1959).

Myotis sodalis Miller and Allen.
Trogloxene.

Sealander and Young (1955) reported
a misidentified Indiana bat from a
now inundated drift northwest of
Hot Springs. There is a confirmed
record of the species from a south-
eastern Oklahoma cave (Glass and
Ward 1959). Neither Caire (1985) nor
Heath et al. (1986) found this species
inhabiting mines or caves in the Ou-
achitas.

Pipistrellus subflavus (F. Cuvier).
Trogloxene.

The eastern pipistrelle was described
as fairly abundant in southeastern
Oklahoma (Caire 1985) and as wide-
spread and abundant in the Arkansas
portion of the Ouachitas (Heath et al.
1986). Barbour and Davis (1969) de-
scribed it as the most abundant bat
over much of the eastern United
States. Caves and mines appear to be
important habitats for winter hiber-

67

nation sites and for summer night
roosts (Barbour and Davis 1969,
McDaniel and Gardner 1977). Caire
(1985) reported capturing many indi-
viduals at Bear Den Caves during
summer months. Heath et al. (1986)
reported this species had been ob-
served in every drift at all times of
the year and that, over a three year
period, one drift had an annual
population of between 600-800 hiber-
nating individuals. Visits to this
hibernaculum over the past three
years have revealed the number of
individuals to be fairly constant. Pre-
liminary observations of a drift that
has had a gate in its entrance for two
years have indicated an increase in
numbers of hibernating pipistrelles.

Epfesicus fuscus (Palisot de
Beauvois). Trogloxene.

Heath et al. (1986) reported that, al-
though common in the Ouachita
Mountain area, the big brown bat
was rarely found hibernating in
drifts. The four drifts used during
hibernation had larger, less re-
stricted, openings that created a vari-
able temperature zone. Rarely were
more than two or three observed in
any drift. This species characteristi-
cally chose hibernating sites near the
entrance where temperature and
humidity levels were lower. Similar
hibernating behavior has been docu-
mented in other caves and mines
(Barbour and Davis 1969, Lacki and
Bookhout 1983). Caire (1985) re-
ported this species from Bear Den
Caves.

Lasionycteris noctivagans
(LeConte). Trogloxene.

Typically considered a tree bat, the
silver-haired bat has been found in
numerous caves and mines (Barbour
and Davis 1969, Saugey et al. 1978,
Whitaker and Winter 1977). Heath et,
al. (1986) discovered a single speci-
men hibernating in a breeze way of a

drift near Lake Greeson; the ambient
temperature was 2 C.

The three following species of La-
siurus, normally considered tree bats,
have been captured during swarm-
ing activities at the entrances of, but
not inside drifts (Heath et al. 1983,
1986). Similar behavior in tree bats
has been observed at caves (Barbour
and Davis 1969, Harvey et al. 1981).

Lasiurus borealis (Muller).
Accidental.

The red bat was captured at the en-
trances of three drifts. Caire (1985)
reported capturing this species at
Bear Den Caves. Red bats were re-
ported from inside two Ozark caves
by McDaniel and Gardner (1977).
Saugey et al. (1978) discovered the
remains of 140 red bats in one Ozark
cave.

Lasiurus seminolus (Rhoads).
Accidental.

Heath et al. (1983) reported the cap-
ture of a female Seminole bat at the
entrance to a drift in Polk County,
Arkansas, during September.

Lasiurus cinereus (Palisot de
Beauvois). Accidental.

Previously unreported, a male hoary
bat was captured simultaneously
with the above mentioned Seminole
bat. The occurrence of this species in
mines and caves has been well docu-
mented (Barbour and Davis 1969,
Saugey et al. 1978).

DISCUSSION

Caves are common and widely dis-
tributed in the United States. Caves
are known in every state and, in
some, are very common. It has been
found that most caves contain a bio-
logically interesting fauna (Culver

1986). Where caves are scarse, aban-
doned mineshaf ts occasionally pro-
vide the same specialized habitat as
do natural caves (Barbour and Davis
1969).

Abandoned mine drifts and frac-
ture caves represent important habi-
tat features in the Ouachita Moun-
tains. Six species of salamanders and
nine species of bats utilize these
structures for some purpose. In addi-
tion, four of the six salamanders are
endemic to the Ouachita Mountains,
and a fifth is endemic to the Interior
Highlands. Two of these
salamanders, Plethodon caddoensis
and P. ouachitae, are Category II spe-
cies. For all of these salamanders,
caves and mines may only represent
larger versions of existing subterra-
nean microhabitats, complimenting
existing situations and not replacing
them. However, caves and mines do
provide ''natural laboratories'' where
insights into life histories and species
interactions, otherwise unobservable,
may be studied with the knowledge
gained applied to management of
surface populations.

Six of the nine species of bats
regularly frequent caves or mines
during some portion of their annual
cycles and two of these are listed as
Category II species {Myotis austrori-
parius and M. leibii). Mines provide a
key habitat component for bats
where natural subterranean hiber-
nacula are scarce. Hibernacula can be
viewed as islands of different sizes
and complexities in an ocean of habi-
tat inhospitable for hibernation
(Gates et al. 1984). Most caves and
mines in the Ouachitas are small and
marginal as hibernacula when com-
pared with extensive and complex
cave systems of other regions. How-
ever, minor hibernacula may become
major ones (depending on their size,
configuration, and microclimate), if
the latter are destroyed. Further, they
may function to promote range ex-
pansions (Gates et al. 1984). In addi-
tion, small populations become in-
creasingly important in species man-
agement when large populations are

68

continually threatened (Humphrey
1978).

Fifty- three vertebrate taxa use
Ozark caves (McDaniel and Gardner
1977). Heath et al. (1986) reported the
occurrence of 27 vertebrate taxa util-
izing abandoned mine drifts in the
Ouachita Mountains. Caire (1985)
and Black (1974) reported two spe-
cies from Bear Den Caves. We report
two additional species from aban-
doned mines (Lasiurus cinereus and
Plethodon serratus). Of the 31 re-
corded species that use caves and
mines in the Ouachita Mountains, 22
are common to both the Ouachitas
and Ozarks.

These data further support Maser
et al (1979) when they stated,
"Unique habitats occupy a very
small percent of the total forest land
base, yet they are disproportionately
important as wildlife habitats." From
our measurement, the total area of all
known and inventoried caves and
drifts in the Ouachita Mountains is
approximately one acre in a forest
with nearly 1 .6 million surface acres.
For these reasons, resource managers
should not overlook opportunities to
protect and conserve what may ap-
pear to be marginal sites, especially
in areas where these unique habitats
may be a limiting factor.

MANAGEMENT
RECOMMENDATIONS

While the National Forest Manage-
ment Act (1976) and Endangered
Species Act (1973) specify objectives
and set policy, the Forest Service
Manual provides guidance and di-
rection to realize these objectives re-
lating to species of special concern
and their habitats. These documents
mandate consideration of these
unique and valuable resources in all
phases of planning and project im-
plementation.

Nieland and Thornton (1985), Nie-
land (1985), Hathom and Thornton
(1986), and Chaney (1984) provide
additional information, guidance and

considerations concerning manage-
ment, inventory and evaluation of
caves. Caire (1985) made recommen-
dations about habitat management
for bats, including Bear Den Caves in
southeastern Oklahoma, and Sievert
(1986) proposed guidelines for pres-
ervation of habitat for the endemic
Rich Mountain salamander.

Because management of cave re-
sources are adequately addressed in
these references, the following rec-
ommendations address issues con-
cerning needed management of aban-
doned mine drifts whose importance
to bats and other vertebrates has
been demonstrated by Heath et al.
(1986), Lacki and Bookhout (1983),
Saugey et al. (1985), Whitaker and
Winter (1977) and this study.

In line with these studies, we rec-
ommend the following actions be
taken on National Forests, other pub-
lic lands, and private lands:

1. Address abandoned mine
drifts and shafts as "unique
subterranean habitat" in the
Cave Management section of
the Forest Service Manual.
Most of the language in this
chapter is directly applicable
to these excavations.

2. Incorporate management
prescriptions for abandoned
mine drifts into Forest Land
Management Plans and other
resource management plan-
ning documents, where ap-
plicable.

3. Develop specific supple-
ments, for individual Na-
tional Forests, to the Forest
Service Manual concerning
the inventory, evaluation,
and management of these
excavations.

4. Prepare a chapter in the Ou-
achita National Forest Wild-
life Handbook providing di-
rection and guidance con-
cerning management of

abandoned mine drifts and
coordination with other re-
sources.

5. Use full seasonal or partial
closures to protect species of
special concern during criti-
cal periods of the year.

6. Acquire lands within agency
administrative authority that
contain caves and aban-
doned mine drifts.

7. Prohibit extraction of miner-
als and other materials from
abandoned mine drifts.

8. Identify and designate aban-
doned mine drifts, caves,
and associated above ground
habitat as "key areas" for
wildlife during the silvicultu-
ral prescription process.

9. Set aside and preserve travel
corridors to prevent isolation
and loss of use by terrestrial
vertebrates.

10. Establish monitoring activi-
ties to assess changes in the
drift environment and asso-
ciated wildlife utilization.

11. Continue inventory of spe-
cies utilizing drifts and de-
termine how and what they
are using them for.

12. Cooperate, consult, and coor-
dinate with state and federal
resource management agen-
cies, universities and col-
leges, public and private con-
servation organizations, and
other interested publics to
promote conservation, edu-
cation, and research.

"Ultimately, the survival of most
animal species depends more on
habitat protection than on direct
shielding of the creatures them-
selves" (Smith 1984).

69

ACKNOWLEDGMENTS

We thank District Rangers John M.
Archer and Rex B. Mann, Resource
Assistant Clifford F. Hunt, and Wild-
life Staff Officer Dr. David F. Urb-
ston, all of the Ouachita National
Forest, for their support and encour-
agement during this study. Special
appreciation is extended to Clark
Efaw, Belinda Jonak, Stan Neal, Di-
anne Saugey, and Derrick Sugg for
valuable assistance in the field. Le-
onard Aleshire and David Heath
were most helpful in locating aban-
doned mines in the Polk County
area. The Arkansas Geological Com-
mission provided useful information
concerning the location of mines.
This study was supported, in part, by
the U.S. Forest Service (Ouachita Na-
tional Forest), a University of Arkan-
sas Faculty Research Grant and the
University of Arkansas at Little Rock
College of Science's Office of Re-
search, Science and Technology.

LITERATURE CITED

Ashton, Ray E. 1976. Endangered
and threatened amphibians and
reptiles in the United States. Soci-
ety Study Amphibians Reptiles,
Herpetological Circular 5, 65 pp.

Barbour, Roger W., and Wayne H.
Davis. 1969. Bats of America. Uni-
versity Press of Kentucky, Lexing-
ton. 286 pp.

Barnett, Douglas Eldon. 1970. An
ecological investigation of cavemi-
cole populations in Mansell Cave,
Randolph County, Arkansas. M.S.
Thesis, Northwestern State Uni-
versity, Natchitoches, LA. 51 pp.

Barr, Thomas C. 1963. Ecological
classification of cavernicoles. Cave
Notes 5:9-12.

Bishop, Sherman C. 1943. Handbook
of Salamanders. Comstock Pub-
lishing Company, Ithaca. 555 pp.

Black, Jeffery H. 1974. Notes on Ple-
thodon ouachitae in Oklahoma. Pro-
ceedings Oklahoma Academy Sci-
ence 54:88-89.

Black, Jeffery H. 1980. Amphibians of
Oklahoma — a checklist. Bulletin
Oklahoma Herpetological Society
4:78-80.

Blair, Albert P., and Hague L. Lind-
say, Jr. 1965. Color pattern vari-
ation and distribution of two large
Plethodon salamanders endemic to
the Ouachita Mountains of Okla-
homa and Arkansas. Copeia 3:331-
335.

Branner, John C. 1888. Annual report
of the Geological Survey of Arkan-
sas. Volume 1. Little Rock. 103pp.

Caire, William. 1985. Summer ecol-
ogy of the bats in southeastern
Oklahoma. Final Report. Okla-
homa Department Wildlife Con-
servation, Oklahoma City. 22 pp.

Chaney, Steve W. 1984. Cave Man-
agement Plan. USDI, National
Park Service, Buffalo National
River, Arkansas. 24 pp.

Clardy, Benjamin F., and William V.
Bush. 1976. Mercury District of
southwest Arkansas. Information
Circular 23. Arkansas Geological
Commission, 57 pp.

Culver, David C. 1986. Cave Faunas.
Pp. 427-443. In Conservation Biol-
ogy: The science of scarcity and
diversity. Michael E. Soule, editor.
Sinaeur Assoc. Inc., Publishers,
Sunderland, MA. 584 pp.

Davis, Wayne H., William Z. Lid-
icker, Jr., and John A. Sealander.
1955. Myotis austroriparius in Ar-
kansas. Journal of Mammalogy
36:288.

Dunn, Emmett R., and Albert A.
Heinze. 1933. A new salamander
from the Ouachita Mountains.
Copeia 3:121-122.

Gates, J. Edward, George A. Feld-
hamer, Lizabeth A. Griffith, and
Richard L. Raesly. 1984. Status of
cave-dwelling bats in Maryland:
Importance of marginal habitats.
Wildlife Society Bulletin 12:162-
169.

Glass, Byron P., and Claud M. Ward.
1959. Bats of the genus Myotis
from Oklahoma. Journal of Mam-
malogy 40:194-201.

Harrington, Donald. 1986. Bear City,

Arkansas, p. 381-427. In: Let us
build us a city. Harcourt Brace
Jovanovich, Publishers, New York.

Harvey, Michael J., John J. Cassidy,
and Gary G. O'Hagan. 1981. En-
dangered bats of Arkansas: Distri-
bution, status, ecology and man-
agement. Final Report to Arkansas
Game and Fish Commission, U.S.
Forest Service (Ozark National
Forest) and National Park Service
(Buffalo National River). Memphis
State University, 80 pp.

Hathom, Jim, and Jer Thornton.
1986. The common sense guide to
cave gates. American Cave Con-
servation Association 1:23-44.

Heath, Darrell R., Gary A. Heidt,
David A. Saugey and V. Rick
McDaniel. 1983. Arkansas range
extensions of the seminole bat (La-
siurus seminolus) and eastern big-
eared bat (Plecotus rafinesquii) and
additional county records for the
hoary bat (Lasiurus cinereus), sil-
ver-haired bat (Lasionycteris nocti-
vagans) and evening bat (Nycticeius
humeralis). Proceedings Arkansas
Academy Science 37:90-91 .

Heath, Darrell R., David A. Saugey,
and Gary A. Heidt. 1986. Aban-
doned mine fauna of the Ouachita
Mountains, Arkansas: Vertebrate
taxa. Proceedings Arkansas Acad-
emy Science 40(in press).

Hudgins, Mary D. 1971. Gold and
silver boom town of Bear thrived
for a decade. The Record: Hot
Springs-Garland County Histori-
cal Society 12:98-102.

Humphrey, Stephen R. 1978. Status,
winter habitat, and management
of the endangered Indiana bat,
Myotis sodalis. Florida Science
41:65-76.

Hurter, J. and J.K. Strecker, Jr. 1909.
Amphibians and reptiles of Ar-
kansas. Transactions Academy
Science St. Louis 18:11-27.

Ireland, Patrick H. 1971. Systematics,
reproduction and demography of
the salamander Eurycea multipli-
cata. Ph.D. Dissertation. Univer-
sity of Arkansas, Fayetteville. 127
pp.

70

Lacki, Michael ]., and Theodore A.
Bookhout. 1983. A survey of bats
in Wayne National Forest, Ohio.
Ohio Journal Science 83:45-50.

Maser, Chris, Jon E. Rodiek, and Jack
Ward Thomas. 1979. Cliffs, talus
and caves. Pp. 96-103. In Wildlife
habitats in managed forests: The
Blue Mountains of Oregon and
Washington. Jack Ward Thomas,
editor. Agriculture Handbook
No.553, USDA Forest Service,
Washington D.C. 512 pp.

McDaniel, V. Rick, and Kenneth L.
Smith. 1976. Cave fauna of Arkan-
sas: selected invertebrate taxa.
Proceedings Arkansas Academy
Science 30:57-60.

McDaniel, V. Rick, and James E.
Gardner. 1977. Cave fauna of Ar-
kansas: vertebrate taxa. Proceed-
ings Arkansas Academy Science
31:68-71.

Means, D. Bruce. 1974. The Status of
Desmogmthus brimleyorum
Stejneger and an analysis of the
genus Desmognathus
(Amphibia:Urodela) in Florida.
Bulletin Rorida State Museum,
Biological Sciences 18. 100 pp.
Mumford,

Russell E. and John O. Whitaker.
1982. Mammals of Indiana. Indi-
ana University Press, Blooming-
ton. 537 pp.

Nieland, Jim. 1985. Cave manage-
ment prescriptions. American
Cave Conservation Association
1:36-41.

Nieland, Jim, and Jer Thornton. 1985.
Spelean inventory and evaluation
guide. American Cave Conserva-
tion Association 1:1-35.

Noble, G.K., and Byron C. Marshall.
1929. The breeding habits of two
salamanders. American Museum
Novitates 347:2-12.

Pope, Clifford H., and Sarah H.
Pope. 1951. A study of the
salamander, Plethodon ouachitae,
and the description of an allied
form. Bulletin Chicago Academy
Science 9:129-152.

Puckette, Bill. 1974-75. Bear Caves.
Oklahoma Underground 7:15-17.

Reagan, Douglas P. 1974. Threatened
native amphibians of Arkansas.
Pp. 93-99. In Arkansas natural
area plan. Arkansas Department
of Planning. Little Rock, AR. 248
pp.

Robison, Henry W., and Kenneth L.
Smith. 1982. The endemic flora
and fauna of Arkansas. Proceed-
ings Arkansas Academy Science
36:52-57.

Saugey, David A., Robert H. Baber,
and V. Rick McDaniel. 1978. An
unusual accumulation of bat re-
mains from an Ozark cave. Pro-
ceedings Arkansas Academy Sci-
ence 32:92-93.

Saugey, David A., Darrell R. Heath,
and Gary A. Heidt. 1985. Summer
use of abandoned mines by the
Caddo Mountain salamander, Ple-
thodon caddoensis (Plethodontidae)
in Arkansas. Southwestern Natu-
ralist 30:318-319.

Sealander, John A., and Howard
Young. 1955. Preliminary observa-
tions on the cave bats of Arkansas.
Proceedings Arkansas Academy
Science 7:21-31.

Sievert, Gregory. 1986. An investiga-
tion of the distribution and popu-
lation status of the Rich Mountain
Salamander (Plethodon ouachitae) in
Oklahoma. Final Report. Okla-
homa Department Wildlife Con-
servation, Oklahoma City. 37 pp.

Smith, Kenneth L. 1984. Animal spe-
cies. Pp. 48-70. In Arkansas's natu-
ral heritage. Bill Shepherd, editor.
August House Publishers, Little
Rock. 116 pp.

Stone, Charles G., and William V.
Bush. 1984. General geology and
mineral resources of the Caddo
River Watershed. Information Cir-
cular 29. Arkansas Geological
Commission, Little Rock. 32 pp.

Strecker, John K. Jr. 1908. Notes on
the habits of two Arkansas sala-
manders and a list of batrachians
and reptiles collected at Hot
Springs. Proceedings Biological
Society Washington 21:85-89.

Trauth, Stanley E. 1988. Egg clutches
of the Ouachita dusky

salamander, Desmognathus brimley-
orum (Caudata: Plethodontidae),
collected in Arkansas during a
summer drought. Southwestern
Naturalist (in press).

U.S. Army Corps of Engineers. 1980.
Lake Ouachita design memoran-
dum No. 2A. Re-evaluated and
Updated Master Plan for Develop-
ment and Management of Lake
Ouachita and Blakely Mountain
Dam. Vicksburg, MS. 254 pp.

U.S. Federal Register. 1985. Endan-
gered and threatened wildlife and
plants; review of vertebrate wild-
life; notice of review. Volume 50,
No. 181. September 18. 10 pp.

Wells, Kentwood D., and Roger A.
Wells. 1976. Patterns of movement
in a population of the slimy
salamander, Plethodon glutinosus,
with observations on aggrega-
tions. Herpetologica 32:165-162.

Whitaker, John O., and Francis A.
Winter. 1977. Bats of the caves and
mines of the Shawnee National
Forest, southern Illinois. Transac-
tions Illinois State Academy Sci-
ence. 70:301-313.

71

The Herpetofauna of Long
Pine Key, Everglades
National Park, in Relation to
Vegetation and Hydrology'

George H. Dalrymple^

Abstract.— The amphibians and reptiles of the
Long Pine Key region. Everglades National Park,
were surveyed between 1984 and 1986. This
herpetofauna, with 51 species, is well represented
by habitat generalists and Prairie species, but the
compliment of Upland species, primarily Pineland
species, is low due to the lack of natural soil
development and the isolation of the area.

Many authors have noted a general
reduction in species diversity among
animal groups as latitude decreases
in peninsular Florida (Dinnen 1984,
Loftus and Kushlan 1987, for fishes;
Duellman and Schwartz 1958, Kiester
1971, for amphibians and reptiles;
Cook 1969, Robertson and Kushlan
1984, for birds; Simpson 1964, Layne
1984, for manmials). Simpson (1964)
considered such a "peninsular ef-
fect to be due to a greater rate of
extinction and, or a lower rate of
immigration along peninsulas in
comparison to the mainland.

Species area curves (Preston 1962,
Mac Arthur and Wilson 1967) for liz-
ards and snakes evaluated by Busack
and Hedges (1984) showed that there
was no significant peninsular effect
in Florida. There was, however, a
general trend for reduced species
numbers as one proceeds down the
peninsula of Rorida, most likely
caused by a reduction in habitat
quality. Moreover, Robertson's
(1955) study of breeding land birds
of the Long Pine Key region of Ever-
glades National Park, the southern
most Upland region on the mainland,
revealed both lower species richness
and lower densities within species

'Paper presented at symposium. Man-
agement of Amphibians, Reptiles, and
Small Mammals in Nortti America. (Flag-
staff, AZ, July 19-21, 1988.)

'George H. Dalrymple is Associate Pro-
fessor, Department of Biological Sciences,
Florida International University, Miami, FL
33199.

than in other areas. This reduced
abundance of animals agrees with
the general belief that productivity is
low in southern Florida Pinelands
(oligotrophic, Snyder 1986). When
Duellman and Schwartz (1958) de-
scribed the southern Florida herpe-
tofauna as "depauperate... for a
warm lowland area" they were refer-
ring to the lower number of species
(table 1). It has remained unclear
whether characterization of the her-
petofauna as depauperate applies to
all habitat types in the region, in-

cludes both low species and popula-
tion numbers and applies to all taxa.

The main objectives of this study
are to:

1. develop a species list of am-
phibians and reptiles in Long
Pine Key-Paradise Key area
(abbreviated LPK herein),

2. describe species associations
with vegetation characteris-
tics.

Table 1.— The number of species of amphibians and reptiles found In Flor-
ida, southern Florida and In Long Pine Key'

Taxa

Florida

Southern Florida
# (%)

Long Pine K
# (%)

#

Salamanders

24

4 (17)

3

(13)

Frogs and toads

29

16 (55)

12

(41)

Amphibian

Subtotal

53

20 (38)

15

(28)

Turtles

20

11 (55)

8

(40)

Crocodilians

2

2(100)

1

(50)

Lizards

16

11 (69)

6

(38)

Snakes

41

28 (68)

21

(52)

Reptile

Subtotal

79

52 (66)

36

(46)

Totals

132

72 (55)

51

(39)

'The data for Florida and souttiern Florida are based upon current species lists
(Wilson arud Porras, 1983; Auffenberg, 1982). The numt>ers for Long Pine Key are for the
current study (see text). Since Long Pine Key column includes the exotic species
Eleutherodactylus planirostris. Osteopilusseptentrionalis onaf Anolissagrei they have
been included in the counts for the first two columns also. (Salamander list includes
Stereochilus marginatus; frog list includes the new species Rana okaloosae (Moler,
1986).

72

3. evaluate correlations be-
tween species' phenologies
and rainfall patterns in the
area,

4. estimate abundances of spe-
cies and compare them to
other areas in North Amer-
ica.

Study Area

The Long Pine Key (LPK) region was
chosen for study because this 8000 ha
area is the principal remaining natu-
ral upland region of the original Mi-
I ami (or Atlantic) Rock Ridge physi-
ographic province (Davis 1943) and
jj as part of Everglades National Park it
' has been protected from human
interference for nearly 40 years. The
region includes about 4650 ha of Pi-
nelands (Snyder 1986) with a series
of ''transverse or finger glades," or

seasonally flooded Prairies, inter-
spersed throughout the Pinelands
(fig. 1). Within the Pinelands there is
a series of at least 120 tropical hard-
wood Hammocks (Olmsted et al.
1983, fig. 2) varying in size from .1 ha
to 91 ha (Olmsted, Loope and Hilsen-
beck 1980). Most Hammocks are
completely surrounded by Pineland
and are kept rather small due to the
frequent fires (prescribed burns and
natural fires from lightning) in the
region. The largest Hammock, Royal
Palm, is surrounded by seasonally
flooded Prairies and has almost com-
pletely overgrown the limestone ele-
vation known as Paradise Key (these
names are sometimes used inter-
changeably). Because Paradise Key
figured importantly in the study of
Duellman and Schwartz (1958), I
have included it in the present study
as part of the general area described
herein as LPK.

On the southern border of LPK

about 3600 ha of land were farmed
until 1975 (abandonment was an at-
tenuated process from the 1960's to
1975), when this agricultural area,
known as the "hole-in-the donut,"
was purchased by the Park Service.
Early farming was limited to areas
with deeper soil, and involved little
alteration of the underlying bedrock.
Starting in 1954 (W.B. Robertson, Jr.
pers. comm.) rock-plowing of the
upper 20 cm of the ground surface
created an artificial soil: "deeper,
better drained, better aerated, and
possibly more nutrient-rich than the
pre-farming soil" on 1600 of the 3600
ha (Ewel et al. 1982:1-2). The sub-
strate alteration proved conducive to
the establishment of exotic vegeta-
tion, especially Brazilian Pepper (Sch-
inus terebinthifolius) after the farm-
land was abandoned (Ewel et al.
1982).

Existing detailed surveys of the
region's vegetation in relation to ele-
vation, fire and hydrology (e.g. Olm-
sted et al 1980; Olmsted et al. 1983;
Olmsted and Loope 1984; Taylor and
Herndon 1981) as well as an ex-
tremely detailed vegetation map of
the area (Johnson et al. 1983) have
made it much easier to plan the cur-
rent project. Historical surveys of the
literature in the above cited refer-
ences, among many others, make it
clear that the LPK region has not
been completely free from distur-
bances: logging of the Pinelands dur-
ing the 1930's and 1940's; farming, as
described above; invasion by exotic
vegetation; development of elevated
roadways with marl dug from local
pits and their resulting small canals,
culverts and ponds bordering the
former farmlands (all of which dis-
tort the original associations of eleva-
tion, soil, vegetation and surface wa-
ter); fire roads, to help control pre-
scribed bums; and the inevitable
presence of humans and their build-
ings (both those for visitors and the
complex of staff facilities). All of
these factors play a role in determin-
ing the present herpetofauna. Cur-
rent park management fosters a de-

Figure 1.— Aerial photograph of Pineland and Prairie of Long Pine Key.

73

large enough to ensure lasting pres-
ervation of this unique ecosystem
type.

Materials and Methods

General Collecting and Road
Cruising

For the 3 years of the study reported
on herein many hours were spent
surveying and trapping in areas for
evidence of amphibians and reptiles.
Each time the traps were checked, a
50 km section of unimproved dirt
roads was driven over by van, and
an additional 15 km paved road was
systematically covered by van for a
total of 8 to 16 hours per week, dur-
ing which all animals were captured
and identified. Searches on foot, by
teams of two to four people, were
conducted in all of the major habitats
each week, during which animals
were searched for at the surface and
under rocks and logs. The time spent
collecting and road cruising was di-
vided between day and night to en-
sure that all species in LPK might be
found.

Trapping

I used a system of funnel traps at-
tached to drift fences and transects
(referred to throughout as '"arrays").
Many researchers have used arrays
to study amphibians and reptiles
(Campbell and Christman 1982b,
Clawson and Baskett 1982, Vogt and
Hine 1982, Gibbons and Semlitsch
1981, Clark 1970), however they all
employed arrays that included both
funnel traps and pit traps. Usually
the pit traps are placed at regular in-
tervals by digging holes in the
ground. However, the lack of well
developed soils coupled with an ir-
regular limestone surface made the
use of pit traps impractical to use in
the everglades.

Each array was constructed of
four fifteen meter long sheets of
shade cloth (one meter tall) that

intersected in the middle to form an
"x." The shade cloth was kept up-
right by tieing it to iron rebars that
were hammered into the limestone.
Traps were made of cylinders of one-
eighth inch hardware cloth approxi-
mately 1 m in length and 30 cm in
diameter. Each trap was fitted with
two funnels (one funnel on each side
of the shade cloth fencing) made of
the same material. Funnels were at-
tached to the free ends of the four
arms of the array. Shade cloth had
12-cm flaps sewn onto the bottom
edge to conform to the irregular sur-
faces of the everglades terrain. Flaps
were covered with natural soils and
or leaf litter so that animals would
not crawl under them (figs. 3 and 4).
The square area encompassed by
each array was .10 ha.

Arrays were placed in each of four
main habitat types: seasonally
flooded Prairies, Pinelands, tropical

hardwood Hammocks, and in the
area of secondary succession from
former farming, the "hole-in-the-do-
nut." The latter area is referred to
throughout as "Disturbed." Thirteen
arrays were maintained starting in
May, 1984, and the arrays are still
checked to the current date. Three
arrays were placed in each habitat
type within Long Pine Key and one
extra hammock array was main-
tained in Royal Palm Hammock on
Paradise Key (fig. 5). Arrays were
temporarily taken down during park
service prescribed burns and re-
placed after the bums. Because ar-
rays were in place for different dura-
tions, I assessed yield in terms of rate
of capture, rather than absolute cap-
ture yield, and capture rate was as-
sessed separately for wet and dry
seasons. At each array we main-
tained two 1-m^ pieces of tar-paper,
under which we commonly collected

seasons. At each array we main-
tained two pieces of tar-paper,
under which we commonly collected
animals. All animals caught along the
fences or under the tar paper at an
array were counted as part of the
capture rate at the array in question.

Figure 3.— Aerial photograph of locality known as New Wave Prairie in Long Pine Key with
"x"-shaped trapping array visible at left (each of the four arnr>s of the array is 15 m long).

Symbolic Star Plot Analysis

Symbolic Star Plot Analysis (Cham-
bers at al. 1983) was chosen as a use-
ful multivariate method for graphi-
cally depicting the rates of capture of
species in the major habitats. Only
species for which there were at least
ten captures were chosen, and the
analyses were based on the number
of animals trapped per 1000 array
days because the raw data does not
reflect the fact that arrays were op-
erational for varying time periods.
The data values are used as the
lengths of the rays of the stars for
each habitat. All data values were
rescaled to range from 1 to c, where c
is the length of the smallest ray (set
to 0.1 for these analyses). According
to Chambers et al. (1983:158): "If x.. is
the measurement of the i"" variable

then the scaled variable [x*. ] is

')

x*j = (1 - c)(X|j - min,X|P

/ (maX|X|j - min|X,j) + c."

The scaled variables are arranged
around a circle at equal angles, the
number of angles determined by the
number of variables, and the actual
rays are drawn by connecting points
trigonometrically calculated for an
arbitrarily chosen maximum radius
for the circle.

The lengths of the rays (not the
area adjoining the rays) in the four
habitat stars for a given species rep-
resent the proportion of all captures
for that species in each habitat. The
result is intended to form a simple
yet "dramatic and memorable" im-
pression of the relationships within
species and between habitat types,
for further details see Chambers et al.
(1983:158-163).

Figure 4.— Ground level view of trapping array fencing in Pineland,

75

Population Abundance Estimates

For most species the actual numbers
presented are actual numbers of indi-
viduals captured. All snakes and
turtles were individually marked.
The anurans and lizards were
marked only during 1984, but due to
the lack of recaptures I stopped
marking in 1985. The marking
method used for snakes was that of
Brown and Parker (1976), and even
though snakes were marked for four
consecutive years (1984-1987) the re-
capture rate remained very low
(<0.05, Dalrymple, in prep.).

Concentrations of amphibians and
reptiles around one or more re-
sources, such as water (ponds or
lakes. Carpenter 1952, Reichenbach
and Dalrymple 1986), hibemacula
(caves, pits and dens, Woodbury
1951, Brown and Parker 1982a,
Aleksiuk and Gregory 1974) breeding
sites (Crump 1982, Brown and Parker
1982b, Wiest 1982) and or food (Ha-
milton 1951) lead to recaptures that
allow for density estimates with con-
fidence limits (cf. Turner 1971). These
estimates are dependent on seasonal
fluctuations, and may differ greatly
from estimates of crude density.
However, few concentrations were
found on LPK particularly because
water was readily available in nu-
merous solution holes in every habi-
tat. Moreover, mild winters allowed
most species to be active throughout
the year, and the ability of animals to
readily go underground through the
porous limestone and plentiful solu-
tion holes found in all habitats re-
sulted in the absence of group hiber-
nacula. Further complicating density
estimation were widespread move-
ments in search of mates, and the fact
that major food sources were not
clumped.

All these factors lead to a wide
spread distribution of most species in
the region and most were not habitat
specialists, at least at the major vege-
tation type level. The lack of concen-
trations and the limited number of
recaptures permit only the presenta-

tion of total numbers of captures and
not accurate density estimates at this
time.

Results

Species List

Starting in January, 1984, 51 species
of amphibians and reptiles were ob-
served or collected in LPK (table 2).
Some species were rare because they
are most commonly associated with
more permanently aquatic habitats,
such as the Sloughs (e.g. Acris gryllus,
Ram grylio, Trionyx ferox, Farancia
abacura, Nerodia cyclopion, Nerodia tax-
ispilota, Regim alleni). A few species

that have been recorded in the larger
geographic region were not found in
LPK during this study (Scaphiopus
holbrooki, Pseudobranchus striatus,
Semimtrix pygaea, Masticophis flagel-
lum, Heterodon platyrhinos, Ophisaurus
ventralis, Sternotherus odoratus).

Trapping Results

Between May, 1984 and December,
1986, 1709 amphibians and reptiles
were collected either in the traps,
under associated tar paper, or along
array fences (table 3). These animals
represent 37 of the 51 species (73%)
known from our overall surveys. I
compared the four habitats by re-
cording the number of animals per

Figure 5.— Map of the Long Pine Key-Paradise Key region of Everglades National Park. Array
locations are nunnbered and referred to in \he text as follows: 1 . Pine Block B, 2. New Wave
Prairie, 3. Pine Block E, 4. Junk Hammock, 5. Serenoa Prairie, 6. Wrigfit Hammock, 7. Mud
Prairie, 8. Pine Block H, 9. Palma Vista I Hammock, 10. Royal Palm Hammock, 11. Burnout Dis-
turbed, 12. Sctiinus Disturbed, 13. Grass Disturbed.

76

Table 2.— List of species of amphibians and reptiles observed In the Long Pine Key - Paradise Key region of Ever-
glades National Park during present study, between January, 1984 and December, 1986. The reglonwide natural
habitat associations of Duellman and Schwartz (1958), as they apply In the study area, are given after the scientific
name for each species. Pr = Prairie, Pi = Pine, H = Hammock, A= Permanently Aquatic, I.e. Slough, Canals.

Scientific name

Common name

Scientific name

Common name

Urodela

Amphiumo means -Pr A

Siren lacerfina -Pr A
Nofophfhalmus viridescens -Pr A

Anura

Acris gryllus -Pr

Bufo quercicus -Pr ?\ H

Bufo ferresfris -Pr Pi H

Eleufherodacfylus planirosfris -Pi

Gasfrophryne carolinensis -Pr Pi H

Hyla cinerea -Pr Pi H

Hyla squirella -Pr Pi H

Limnaoedus ocularis -Pr

Osfeopilus sepfenfrionalis -H *

Pseudacris nigrifa -Pr Pi

Rana grylio -Pr A

Rana splienocephala -Pr A

Testudines

Chelydra serpentina -Pr A
Clirysemys floridana -Pr A
Chrysemys nelsoni -Pr A
Deirocheiys reficularia -A
Gopherus polyphemus -Pi
Kinosfernon bauri -Pr A
Terrapene Carolina -Pr Pi H
Trionyx ferox -A

Crocodylia

Alligator mississippiensis -Pr A

two-toed ampl-ii-
uma

greater siren
peninsula newt

Florida cricket frog
oak toad
southern toad
greenhouse frog
eastern narrow-
mouthed toad
green treefrog
squirrel treefrog
little grass frog
Cuban treefrog
Florida chorus frog
pig frog

southern leopard
frog

snapping turtle
peninsula cooter
red-bellied turtle
chicken turtle
gopher tortoise
striped mud turtle
box turtle
Florida soft-shelled
turtle

American alligator

Squamatalacertilia

Anolis carolinensis -Pr Pi H

Anolis sagrei -P\

Eumeces inexpectatus -Pr Pi H

Ophisaurus compressus -Pr Pi
Scincella laterale -Pi
Sptiaerodactylus notatus -Pi

Squamata ,Serpentes

Agkistrodon piscivorus -Pr A
Cemopt)ora coccinea -Pi
Coluber cor^strlctor -Pr Pi H
Crotalus adamanteus -Pi

Diadophis punctatus -Pr Pi H
Drymarct)on corais -Pr Pi H
Eiaphe guttata -Pr Pi H
Elaphe obsoleta -Pr Pi H
Farancia abacura -Pr A
Lampropeltis getulus -Pr Pi H
Lampropeltis triangulum -Pi
Micrurus fulvius -Pi
Nerodia fasciata -Pr A

Nerodia cyclopion -Pr A
Nerodia taxispilota- A
Opheodrys aestivus -Pr Pi H
Regina alleni -Pr

Sistrurus miliarius -Pr Pi
Storeria dekayi -Pr Pi H
Thamnophis sauritus -Pr Pi H
Ttiamnophis sirtalis -Pr Pi H

green anole
brown anole
southeastern five-
lined skink
island glass lizard
ground skink
reef gecko

cottonmouth
scarlet snake
black racer
eastern diamond-
back

ringnecked snake
indigo snake
corn snake
yellow rat snake
mud snake
kingsnake
scarlet kingsnake
coral snake
banded water
snake

green water snake
brown water snake
rough green snake
striped crayfish
snake

pigmy rattlesnake
brown snake
ribbon snake
garter snake

array day. The highest capture rates
were in seasonally flooded Prairie,
which had both the most individuals
and the most species collected, fol-
lowed by Disturbed areas. Hammock
and Pineland (table 3).

Monthly total rainfall for LPK and
maximum water level from well sta-
tion NP-72 in the same area for data
from 1984-1986 were provided from
hydrological stations maintained by
the South Florida Research Center,

Everglades National Park. These data
were correlated with the monthly
values for animals trapped per check
day. There were significant correla-
tions between number of animals
caught per check day and both
monthly rainfall (r = 0.55, p = .001),
and monthly maximum water levels
(r = 0.50, p = .004) for the three year
period (fig. 6). Rates of capture were
significantly greater during the wet
season than the dry season (table 4;

Wilcoxin matched pairs test, T = 3.0,
p < .005). Differences in overall cap-
ture rates between the dry and wet
seasons is greater in Hammock and
Disturbed areas than in the Pinelands
and Prairie.

Relative Abundance

Although 37 species were found at
arrays they were not all equally com-

77

Table 3.— Total numbers of amphibians and reptiles trapped, May 1984-Dec 1986. "Ctieck days" are number of days
on which traps were checked. "Array days" are number of total days arrays were standing. Numbers in parentheses
are animals per 1000 array days. Acronyms at right of table are for species used in figures 7-9.

Taxa

Prairie

Pineland

Hammock

Disturbed

Total

A. means

9

(3.5)

0

(0)

0

(0)

0

(0)

9

A. gryllus

1

(0.4)

0

(0)

0

CO)

0

(0)

\^y

1

B. quercicus

95

(37.2)

7

(2,8)

3

(0 9)

9

(6.2)

114

Bq

B. ferresfris

45

(17.6)

24

(9.4)

50

(15.5)

31

(21.3)

150

Bt

E. plonirosfris

15

(5.9)

17

(6.7)

50

(15.5)

6

(4.1)

88

EP

G. carol inensis

10

(3.9)

1

(0.4)

21

(6 5)

33

(22.6)

65

Go

H. cinerea

on

1
1

7

u-2;

31

He

H. squkella

32

(12.5)

3

(1.2)

6

d 9)

4

(2.7)

45

Hs

OiSepfenfrionalis

2

(0.8)

1

(0.4)

3

CO 9)

6

(4.1)

12

Os

P.nigrifa

5

(2.0)

8

(3.1)

0

CO)

0

(0)

\^y

13

Pn

R. grylio

5

(2,0)

V,*- >^ J

0

(0)

0

CO)

0

(0)

\^y

5

R. sphenocephala

135

(52.8)

10

(3.9)

106

(32 8)

20

(13.7)

271

Rs

A. carolinensis

170

(66,5)

136

(52.3)

19

C5 9)

19

(13.0)

344

Ac

A. sagrei

0

(0)

0

(0)

50

(15.5)

103

(70.7)

153

As

E. inexpecfatus

23

(9.0)

21

(8.2)

42

CI 3.0)

3

(2.1)

89

El

O. compressus

1

(0.4)

1

(0.4)

0

(0)

\^y

1

(0,4)

3

S. late rale

30

(11.7)

9

(3.5)

3

CO 9)

0

(0)

\^y

42

SI

S. not a f us

0

(0)

0

(0)

29

(9.0)

0

(0)

29

Sn

K. bauri

12

(4.7)

2

(0.8)

1

(0.3)

\ V • ^y

1

(0,7)

16

Kb

T. Carolina

n

(4.3)

1

(0.4)

2

(0.6)

3

(2.1)

17

To

A. piscivorus

1

(0.4)

0

(0)

0

(0)

2

(1.4)

3

C. coccinea

2

(0.8)

0

(0)

0

(0)

0

(0)

\^y

2

C. constrictor

8

(3.1)

30

(1 1.8)

14

C4 3)

14

(9.6)

v » • ^y

66

Cc

C. adamanteus

0

(0)

0

(0)

0

(0)

\^y

1

(0,7)

1

D. punctatus

3

(1.0)

3

(1.0)

13

(4 0)

0

(0)

\^y

19

Dp

D. corals

1

(0.4)

2

(0.8)

2

(0,6)

0

(0)

5

E. guttata

0

(0)

1

(0.4)

0

(0)

0

(0)

1

E. obsoleta

0

(0)

0

(0)

4

(1.2)

1

(0.7)

5

L.getulus

0

(0)

0

(0)

0

(0)

1

(0.7)

1

L triangulum

1

(0.4)

0

(0)

0

(0)

0

(0)

1

M. fulvius

0

(0)

0

(0)

4

(1.2)

0

(0)

4

N. fasciata

3

(1.2)

0

(0)

0

(0)

0

(0)

3

R. alleni

1

(0.4)

0

(0)

0

(0)

0

(0)

1

S. millarius

14

(5.5)

8

(3.1)

3

(0.9)

6

(4.1)

31

Sm

S, dekayi

2

(0.8)

0

(0)

4

(1.2)

0

(0)

6

T. sauritus

8

(3.1)

1

(0.4)

10

(3.1)

0

(0)

19

Tsa

T. sirtalis

30

(11.7)

5

(2.0)

2

(0.6)

7

(4.8)

44

Tsi

Totals

695

292

448

274

1709

No. Check days

669

663

789

361

2482

Anis/Check day

1.04

0.44

0.57

0.76

0.70

No. Species

30

22

24

21

37

No. Array days

2555

2550

3229

1458

9792

Anis/ Array day

0.27

0.12

0.14

0.19

0.18

men. The most common species were
anurans and lizards (table 3): Ram
sphenocephala, Bufo terrestris, and Ano-
lis carolinensis. Of the 20 species of
snakes collected during the study, 17
were trapped but only five were cap-

tured in high enough frequency to
allow for more detailed study (Col-
uber constrictor, Thamnophis sirtalis,
Sistrurus miliarius, Diadophis punc-
tatus, and Thamnophis sauritus). As a
preliminary method, abundance can

be minimally estimated as the actual
counts from the 'Total" column of
table 3 as the number per hectare (12
arrays, each one covering approxi-
mately one-tenth of a hectare makes
this a conservative estimate).

78

Habitat Use And Preference

A species' likelihood of being
trapped is more a funcrion of the
number of individuals in the vicinity
of an array than a result of any dif-

Figure 6.— Comparison of number of ani-
mals trapped per check day per month
with monthly rainfall and water table values
from study area between May 1984 and
December 1986.

ference in trap functioning between
habitats. For species with high cap-
ture rates, there were significant dif-
ferences in habitat use for: Coluber
constrictor, more common in Pine-
lands (chi square = 14.59, p = ,0007);
Thamnophis sirtalis, Sistrurus miliarius,
Scincella laterale and Bufo quercicus all
more common in Prairie (chi squares
of 42.9, 9.6, 26.4, 71.8 respectively, all
with p's < .01); while Bufo terrestris is
equally common in all habitats (chi
square = 2.36, p = .51). In most cases,
species were found in more than one
and usually three habitats (cf. Duell-
man and Schwartz 1958). Among
trapped species, 41% were found in
all four habitat types, 27% in two or
three, and 32% in only one habitat
type. Seven of the 13 species from
only 1 habitat type were from Prairie.

Table 4.— Results of 1985 trapping of all individuals of amphibians and rep-
tiles at 13 array sites organized by vegetation type, and season (dry = No-
vember-April; wet = May-October). "Check-days" are the number of days
on which an array was checked for animals. Note that there is no data for
the wet season for "Grass" array (see Materials and Methods). Variation
within habitat types is as great as between habitat types.

Habitat/array

No. No. No. Animals

Individuals species check-days per check day

Season:

Dry

Wet

Dry

Wet

Dry

Wet

Dry

Wet

Prairie

New Wave

65

118

12

20

54

54

1,2

2.2

Mud

38

64

10

18

56

54

0.7

1.2

Serer^oa

12

20

2

10

50

51

0,2

0.4

Pineiand

Pine Block B

26

23

8

8

50

53

0,5

0.4

Pine Block H

25

39

8

11

56

51

0,5

0,8

Pine Block E

16

17

4

8

51

52

0,3

0,3

Hammocks

Royal Palm

18

110

7

17

56

28

0.3

3.9

Pclma Vista 1

11

50

6

12

56

33

0,2

1.5

Wright

15

21

6

8

52

53

0.3

0.4

Junk

17

23

7

7

53

52

0,3

0.4

Disturbed

Schinus

11

76

6

12

55

33

0.2

2.3

Burnout

11

16

6

7

45

17

0.2

0.9

Grass

14

4

18

0.7

Symbolic star plot analyses
(Chambers et al. 1983) were applied
to the 1984-1986 trap data for the
number of animals per 1000 array
days as the data set (table 3), for the
anurans (fig. 7), lizards and turtles
(fig. 8), and snakes (fig. 9). Since the
qualitative general habitat associa-
tions of Duellman and Schwartz
(1958) were corroborated in this
study, I restricted this quantitative
analysis to those species for which
there were at least 10 captures.

It is obvious from the anuran plot
that the majority of individuals and
species are most prevalent in Prairie.
Pseudacris nigrita is strongly repre-
sented in Pineiand, as was noted by
Duellman and Schwartz 1958). In
Hammocks, Eleutherodactylus
planirostris, Bufo terrestris, Gastro-
phryne carolinensis, and Hyla cinerea
were dominant. Rana sphenocephala
was most common in Prairie but was
very abundant in two Hammocks
that are adjacent to wet Prairie and
that retained water in solution holes
throughout most of the year (Royal
Palm and Palma Vista I). Bufo
terrestris, G. carolinensis and the exotic
Cuban tree frog, Osteopilus septentri-
onalis, were dominant in Disturbed
habitat (fig. 7).

For the trap data for turtles, Kinos-
ternon bauri and Terrapene Carolina,
and the lizards. Prairie again had the
greatest abundance; but T. Carolina
was commonly found in the Dis-
turbed habitat. Anolis carolinensis was
well represented in Pineiand and
Prairie, as were the skinks, Eumeces
inexpectatus and Scincella laterale. Ano-
lis sagrei was restricted to Disturbed
sites and Hammocks, especially
those close to roads and parking lots.
Sphaerodactylus notatus is most often
found in leaf litter of Hammocks,
and E. inexpectatus is also well repre-
sented in Hammocks (fig. 8).

For snakes, the star diagram
analysis was restricted to the five
most common species; again the
greatest diversity and abundance is
found in Prairie. Coluber constrictor
was clearly the dominant snake in

79

prairie pineland

Pn Rg

Figure 7.— Star plot diagrams of anuran
data from table 3, comparing th»e frequen-
cies of trapping (anurans per 1000 array
days) of ttie species in \he four tiabitat
types. Genus and species names abbrevi-
ated on key at bottom of figure correspond
to acronyms given in table 3.

Pineland. Sistrurus miliarius was well
represented in all habitats, but is
least common in Hammocks. Tham-
nophis sirtalis was most abundant in
Prairie, while T. sauritus was most
common in Prairie and Hammocks.
Diadophis punctatus is the snake spe-
cies most difficult to keep in traps
(because of their small size they
could more readily escape) but cur-
rent data indicate that they are most
common in the leaf litter environ-
ment of Hammocks (fig. 9).

The most similar habitats with re-
gard to trap data were Prairie and
Pineland, the least similar were Pine-
land and Hammock (table 5). Table 5
includes the only data from the ar-
rays and therefore some species are
excluded from the similarity index
(because the index used, Morisita's
index (Horn 1966; Brower and Zar
1984) requires data on both the num-
ber of species and the number of in-
dividuals per species in the estima-
tion of degree of similarity).

hammock disturbed

prairie pineland

Figure 8.— Star plot diagrams of lizard and
turtle data from table 3, comparing fre-
quencies of trapping (lizards or turtles per
1000 array days) in the four tiabitat types.
Genus and species names abbreviated on
key at bottom of figure correspond to acro-
nyms given in table 3.

Discussion

Species List

Duellman and Schwartz (1958) gave
a complete list of the localities from
which they examined specimens but,
unfortunately this list does not serve
as an effective species list for this
study. Since the intention of their
study was a survey of all of southern
Florida, they did not collect as exten-

Pralrie

Prairie 30

Pine .736

Hammocks .608

Disturbed .314

V

hammock disturbed

prairie pineland

Figure 9.— Star plot diagranr>s of snake data
from table 3, comparing frequencies of
trapping (snakes per 1000 array days) in
ttie four tiabitat types. Genus and species
names abbreviated on key at bottom of
figure correspond to acronynns given in
table 3.

sively in one area as we have been
able to. Nevertheless, the descrip-
tions of habitat preferences they gave
make it clear that a few more species
might be found in the Long Pine Key
region if I continue the study. There
are some noticeable absences from
their list for the Long Pine Key and
Paradise Key areas however: Storerk
dekayi and Diadophis punctatus. It is
possible that these species were
merely overlooked in their surveys

Pine

Hammocks

Disturbed

21

20

17

22

19

16

,303

24

17

.253

.589

21

Table 5.— Measures of similarity among arrays grouped by vegetation type
based on data from table 3 (1984-1986, above). Numbers above the di-
agonal are the numbers of species shared between habitats; numbers
along the diagonal, boldfaced, are numbers of species occurring in each
habitat. Numbers below the diagonal, underlined, are Morisita's indices.

80

and it is extremely unlikely that these
species were not present in the local
area thirty years ago (Duellman and
Schwartz, personal communications).

Salamanders were the taxon most
poorly represented in LPK, only four
of the state's 24 salamanders were
found in southern Florida (table 1),
and only three of these were found in
LPK. The reason for the low count is
obviously the low elevation and poor
soil development of the region.

The majority of Florida's salaman-
ders are members of the family Ple-
thodontidae, and this family is pri-
marily distributed in the Appala-
chian mountains and foothills of the
eastern U.S. Many species are stream
dwellers, others are forest litter in-
habitants that require a moist thick
leaf litter and soil development. The
mole salamanders, family Ambysto-
matidae, also require soils for bur-
rowing. Moreover, salamander lar-
vae are frequently absent from
aquatic settings in which fish are
common.

One notable exception is the newts
(family Salamandridae), but even the
one member of this family from the
region, Notophthalmus viridescens, is
rare. The only successful salaman-
ders in the region are fully aquatic,
neotenic, eel-like animals: Amphiuma
means, Siren lacertina and Pseudobran-
chus striatus. Their cryptic life styles
and easy access to the underground
aquifer through the porous limestone
bedrock may be important reasons
for their success.

The number of anuran, lizard and
turtle species are all rather low in
southern Florida (tables 1 and 2).
Several species of lizards extend
southward past the mainland into
the Rorida Keys, but appear to have
completely by-passed the western
extension of the Miami Rock Ridge
(in particular LPK) e.g. Eumeces
egregius and Cnemidophorous sexlinea-
tus. Two species are endemic to the
sandhills and scrub habitats of Ror-
ida (Sceloporous woodi and Neoseps
reynoldsi) and their absence in the
area is again probably due to the lack

of suitable soils and substrates. The
reason for the absence of the other
two species of Ophisaurus (O. attenu-
atus and O. ventralis) listed by Duell-
man and Schwartz (1958) is not clear,
although they did note that Ophisau-
rus compressus was the "most abun-
dant" of the three species in southern
Florida.

The only notable introduced lizard
was Anolis sagrei. This species is so
common in southern Florida now
that it is no surprise that large popu-
lations are found in some parts of the
current study area (Wilson and Por-
ras 1983). In LPK it was generally
limited to areas where there was a
greater rate of contact with visitors,
and in Disturbed settings. In remote
Hammocks anoles were rarely ob-
served, but Palma Vista I and Royal
Palm Hammocks (both sites that are
popular with visitors and adjoin
roads) Anolis sagrei is extremely com-
mon, as well as throughout the hole-
in-the-donut. At the current time the
park appears to have a limited
"load" of exotic lizards. Hemidactylus
garnoti was observed at the parking
lot at Pahayokee visitors site, and
there are occasional reports of this
species and of Anolis equestris in the
LPK campground area and the "Pine
Island" residential area for park
staff.

Of the few specimens of Gopherus
polyphemus seen during the study, the
only one from the study area was
crossing the road into the hole-in-the-
donut (several others were seen in
the Pine Island residential area and
one shell was near a pond, but no
one is certain of the source of these
animals, and some visitors have been
known to release gopher tortoises
near the entrance to the park).
Whether the sighting within the
study area (the turtle was measured,
and marked) is indicative of a small
population or is a captive released by
a visitor is not at all clear.

The presence of a population of
gopher tortoises on Cape Sable
(Kushlan and Mazzotti 1985) does
not help in explaining the single

specimen, and Duellman and
Schwartz (1958) list only one speci-
men for Dade County. Duellman and
Schwartz (1958:260) described Ster-
notherus odoratus as "the least abun-
dant of the three southern Florida
kinosternids," and I have found it in
the Shark River Slough region but
not LPK. Kinosternon subrubrum is de-
scribed by Duellman and Schwartz
(1958:265) as avoiding "the main part
of the Everglades, an area where K.
bauri reaches its greatest abundance.
When the above three rare species
are noted the turtle list for Long Pine
Key is typical of the southern Rorida
region.

Some of the species listed by Du-
ellman and Schwartz were not com-
mon in the southern everglades, but
were found in other areas of south-
ern Rorida. There were no species of
anurans that I expected to find and
did not. The burrowing nature of
Scaphiopus holbrooki probably pre-
vents it from being common in LPK,
and it was never seen or heard dur-
ing this study.

The crocodilian fauna of LPK is
composed of only one species, the
American alligator (although there
have been rare occurrences of the
American crocodile, Crocodylus
acutus, in the freshwater reaches of
the Taylor Slough drainage in the vi-
cinity of the study area, W.B.
Robertson, Jr. pers. comm). The alli-
gator is found in almost every place
in the everglades where there is wa-
ter. We commonly found evidence of
alligators in the seasonally flooded
Prairie (alligator trails) and in the
willow heads and Hammocks ("ga-
tor holes," a few nests seen, juvenile
and adult alligators observed). The
LPK region is certainly peripheral to
the main distribution of the species
in the park.

The snake fauna is clearly the best
represented fauna in LPK. Of the 26
species listed for southern Florida, 21
were collected during the study. Of
the five not found during this study
only one was expected, Seminatrix
pygaea, and the technique for trap-

81

ping this species described by Lo-
raine (1985) will be tried in the study-
area in the future. Heterodon platy rhi-
nos was described by Duellman and
Schwartz (1958) as not being abun-
dant in southern Florida, and there is
only one report of it from the LPK
area (Roger L. Hammer pers.
comm.).

Masticophis flagellum is still re-
ported from the pineland remnants
of southwest Dade County. Duell-
man and Schwartz (1958) had no rec-
ords of this species from the park,
but since then there has been one rec-
ord from the park.

Pituophis melanoleucus was repre-
sented in the work of Duellman and
Schwartz by a single specimen from
Miami, and a single specimen of this
species was collected in 1984 in
North Miami Beach. The snake was
probably a captive pet released in the
area, since its feces contained white
mouse remains (Robert J. Nodell,
pers. comm.). Tantilla oolitica (T.
coromta wagneri of Duellman and
Schwartz) has never been recorded
from the park, and its range is lim-
ited to isolated Atlantic Coastal
Ridge remnants on the eastern coast
and the Florida Keys (Wilson and
Porras 1983).

Habitat Use and Preferences

Within the LPK region. Prairie habi-
tat has the most diverse and abun-
dant herpetofauna. The Prairie is a
broad transition zone or ecotone be-
tween the longer hydroperiod Slough
habitat and the drier Uplands, and
they are seasonally inhabited by most
species from those two habitats as
well as a semi-aquatic fauna of their
own.

Duellman and Schwartz (1958:206-
213) characterized the habitats of
southern Florida, as they pertain to
Long Pine Key, as: Xeric (including
the rocky Pineland of Long Pine
Key), Mesic (including the tropical
hardwood Hammocks of Long Pine
Key), and Altemohygric (including

Prairie), and their characterization
for each species is given in table 2.

All of the 18 species that Duellman
and Schwartz (1958:211) character-
ized as generalists i.e. "common to
all three" (i.e. Prairie, Pineland, and
Hammock) were found in Long Pine
Key. Seventeen of the 21 species
(81%) they characterized as inhabi-
tants of the Prairie (or Altemohygric
habitat) were found in the study
area.

Only 9 of the 22 species (40%) that
Duellman and Schwartz (1958:210)
characterized as Xeric or Pineland
species are found in the region. Four
of these 9 species were actually more
common in Hammocks (Eleutherodac-
tylus planirostris, Sphaerodactylus
notatus, Anolis sagrei, and Micrurus
fulvius), one (Scincella laterale) was
common in Prairie, three were rare
(Gopherus polyphemus, Lampropeltis
triangulum, and Cetnophora coccinea)
and only one (Crotalus adamanteus)
was actually most common in Pine-
land (see table 2).

Using the species associations of
Duellman and Schwartz (1958), of
the 51 species from Long Pine Key,
35% (18) are generalists, 33% (17) are
Prairie species, 18% (9) are Pineland
or Xeric in habitat association, 6% (3,
Limmoedus ocularis, Pseudacris nigrita
and Ophisaurus compressus) are com-
mon to Prairie and Pineland, 6% (3,
Alligator mississipiensis, Trionyxferox
and Deirochelys reticularia) are pri-
marily Slough or Hygric (Duellman
and Schwartz 1958:212), and 2% (1,
Osteopilus septentrionalis) from Edifi-
carian-Ruderal and Hammock (Me-
sic) habitats.

The limit to the preservation of
overall diversity of the Long Pine
Key region is the extent of rocky Pi-
neland habitat, because it is the ma-
jor habitat type of the area with the
smallest percentage (40%) of its her-
petofauna (as defined by Duellman
and Schwartz 1958) represented. It is
important to note that the common
use of interdigitating finger glades,
i.e. the local Prairie, and Hammocks
by some of the Pineland species

makes it clear that overall diversity
depends upon continued manage-
ment to preserve the current patch-
iness of the area.

Sixty two percent of the species
trapped in the Disturbed habitat are
characterized as generalists by Duell-
man and Schwartz (1958), 14% are
from Pineland and Prairie, 14% are
from Pineland and 10% are from
Prairie.

While the vast majority of am-
phibians and reptiles were either
trapped and, or seen in the Disturbed
habitat, a few were rarely or never
seen in the Disturbed habitat:
Limnaoedus ocularis, Pseudacris nigrita,
Scincella laterale and Sphaerodactylus
notatus. In contrast to these native
species, which were not common to
the Disturbed habitat, the two exotic
species, Osteopilus septentrionalis and
Anolis sagrei were most common
there.

Species composition of the Dis-
turbed habitat primarily depends on
the historical topography of the area.
The vast majority of species there are
generalists, but the area is large
enough that local variations in hy-
droperiod attract a number of species
more commonly associated with
drier or wetter conditions and future
analyses of this very complex area
will involve a more specific separa-
tion of habitat types within the area.
Clearly, most of the species of am-
phibians and reptiles are responding
to basic microhabitat requirements
that have little to do with the actual
species composition of the vegetation
(Campbell and Christman 1 982a: 170-
171).

Abundance

It is impossible to accurately com-
pare the trapping results of this
study to other studies. The methods,
objectives and local circumstances of
each study vary widely. Perhaps
most confounding is the variability in
the number of months per year dur-
ing which species are active, and this

82

makes comparisons based on animals
per check day difficult. There are
also differences in types of arrays
used, the purposes of the trapping
effort, substrate characteristics and
ability to use pit traps, all of which
preclude valid comparisons.
Campbell and Christman (1982b)
summarized their results from north-
ern Florida, in which they operated
30 arrays for 7432 array-days. They
collected 1644 animals of 43 species
from 11 habitats for an average of
0.22 animals per array-day. In LPK,
13 arrays operated a total of 9792 ar-
ray-days and collected 1709 animals
of 37 species in 4 habitats for an aver-
age of 0.18 animals per array day, a
similar catch rate per array day.

Campbell and Christman (1982b)
used both funnel traps and pit traps,
and they estimated that only 36% of
their collection came from funnel
traps. They also state that 69% of the
animals trapped were Eleutherodacty-
lus planirostris, and that 90% of their
trappings were of E. planirostris and
Gastrophryne carolinensis. Both of
these species were readily trapped in
their pit traps. If their pit trap ex-
cluded, and look at the percent from
funnel traps, there was a much trap
yield.

There are so many differences in
the two studies that the only conclu-
sion to be drawn is that the results
compare favorably with that the LPK
region has a moderate diversity and
comparable abundance of animals,
based upon similar trapping effort.

Comparisons to other studies are
even more difficult, since studies in
more temperate climates are done
only during the warmer months of
the year. For example, Clawson and
Baskett (1982), in Missouri, used 13
arrays a total of 3159 array days in
the spring, summer, and fall, and
captured 2545 animals, for an aver-
age of 0.81 animals per array day.
This much higher figure may well be
representative of the greater concen-
tration of both animals and resources
typically found in more temperate
climes.

Species Diversity

Species richness for southern Florida
was described by Duellman and
Schwartz (1958:205) as "depauper-
ate" and "impoverished." They state
that "an impoverished herpetofauna
is what might be expected at the end
of a long peninsula, through the
length of which certain habitats and
their inhabitants disappear."

The difficulty in evaluating this
statement arises from the fact that
there is much more involved in the
biogeography of the peninsula of
Florida than a simple "peninsula ef-
fect" due to reduced area and dis-
tance from centers of distribution
(Robertson and Kushlan 1984). There
is also the recent geological origin of
the land area, the poor development
of soils in the area during the time
since emergence, the lack of variation
in relief of the area (Olmsted and
Loope 1984), and the severe human
disturbance. All of these factors need
to be considered in evaluating the
possible reasons for an "impover-
ished" fauna. Finally there is the is-
sue of deciding whether the fauna
deserves the label of "impoverished"
in the first place.

A reduced species list does not by
itself determine whether the biomass
of the existing species is high or low,
e.g. while the species list for fresh
water fish is considered low for the
area (Loftus and Kushlan 1987) they
are the principal food of an enor-
mous biomass of wading birds.
Robertson and Kushlan (1984:234)
have addressed this point: "...the
nearly unique ability of the South
Florida ecosystem to support such
large numbers of 14 species of super-
ficially similar secondary and tertiary
consumers on a resource base that is
reduced in species diversity by bio-
geographic factors is generally unap-
preciated." and the nesting efforts
(1972 or 1974 numbers) of the White
Ibis and Wood Storks alone are esti-
mated to have required "in excess of
3 billion kilocalories or approxi-
mately 2500 metric tons of food..."

As the impact of the remaining 12
species of wading birds is not known
and the secondary productivity of
South Florida habitats has not yet
been studied, the meaning of this en-
ergy requirement to the total system
is undeterminable."

During this study we have col-
lected data on 51 species of amphibi-
ans and reptiles (table 2). This is not
a low figure for an area the size of
LPK (8000 ha).

Vogt and Hine (1982) list 34 spe-
cies of amphibians and reptiles from
their study area in southern Wiscon-
sin. Clawson and Baskett (1982) list
35 species from their Missouri study
area. Clarke (1958) lists 39 species
from Osage County, Kansas. In trap-
ping studies in the Florida sandhills
of Tampa, Mushinsky (1985) lists 27
species. Campbell and Christman
(1982b) list 60 species from their ex-
tensive study in northern Florida,
and this number comes from a vari-
ety of sampling techniques in, at
least, 11 different habitat types.

Gibbons and Harrison (1981) list
68 species from coastal mainland
South Carolina and Gibbons and Pat-
terson (1978) list 94 species from the
Savannah River Plant in South Caro-
lina. Myers and Rand (1969) list 100
species for Barro Colorado Island,
Panama. Crump (1971) lists 116 spe-
cies for the Belem area of Brazil.

From the temperate to tropic lati-
tudes there is an obvious increase in
overall diversity, but the species rich-
ness for the LPK is not very low for
its latitude. The presence of 51 spe-
cies and the fact that many are abun-
dant makes it clear that the applica-
tion of terms such as impoverished
or depauperate must be used in con-
text. Rather than pondering the ab-
sence of some species (especially
when for the group with the least
representation in the area, the sala-
manders, it is quite clear why they
are not common, see above) I find
myself, like Robertson and Kushlan
(1984, above), more impressed with
the actual abundance of animal life in
this unique area.

83

Conclusions

1. The species list for the LPK
includes at least 51 species,
15 species of amphibians and
36 species of reptiles. The
most poorly represented
group is the salamanders, the
best represented group is the
snakes. The survey of current
species composition is basi-
cally the same as reported 30
years ago for the area by Du-
ellman and Schwartz (1958).
The fact that there has been
no reduction in species rich-
ness of the local area should
be considered a major benefit
of the preservation of the re-
gion inside the national park.

2. Amphibians and reptiles of
LPK are primarily habitat
generalists, usually being
found in three of the four
major habitat types in the
area. The principal separa-
tion by habitat is related to
the characteristics of the sub-
strate, there being a subset of
herptiles most commonly
found in areas with greater
soil development (Ham-
mocks and the Disturbed ar-
eas) and another subset of
herptiles that are more com-
mon in seasonally flooded
Prairie. The most poorly rep-
resented group is that de-
scribed as primarily from
Xeric, Pineland habitat, and
the absence of sandy soils in
the rocky Pineland makes
this the most fragile compo-
nent of the Everglades herpe-
tofauna. The findings of this
study do not differ signifi-
cantly from those of Duell-
man and Schwartz (1958)
from thirty years ago. The
results point out that there is
a significant portion of the
local herpe to fauna that relies
upon the preservation of
large contiguous areas of na-

tive Pineland interspersed
with Hammocks and season-
ally flooded Prairie for its
continued success.

3. Phenologies of amphibians
and reptiles of the LPK can
be described as modified
temperate zone patterns.
While the subtropical charac-
ter of the southern coastal
portion of peninsular Rorida
results in a year long grow-
ing season, with only occa-
sional frosts, the seasonality
of rainfall and the temperate
zone origin of the herpe-
tofauna results in a tradi-
tional spring emergence of
the herptiles, tied to increas-
ing day length, warmer tem-
peratures and the onset of
heavy rainfall.

4. Estimates of density and
relative abundance remain
difficult to give at the current
time. Comparison of current
trapping results with those of
Campbell and Christman
(1982a, 1982b) from 11 habi-
tats in northern Florida indi-
cate a similar level of abun-
dance for the two areas, but
differences in the actual spe-
cies lists, habitat types and
methodologies make such
conclusions tenuous. Com-
parisons of the fauna of the
area with those of a wide va-
riety of other regions indicate
that the herpetofauna of
LPK, with the exception of
the salamanders, has a mod-
erate level of diversity.

Acknowledgnrjents

I wish to thank all the dedicated stu-
dents of ecology and herpetology at
F.I.U. who gave their time so will-
ingly during the study. Doug Barker,
Peter Beck, Laura Brandt, Teresa De-
Francesco, Bob Dunne, Ernesto Her-

nandez, Liz Lewis, Nancy O'Hare,
and Arlene Sackman helped with
trap checking and collecting. To
Frank S. Bernardino, Jr., Bob Nodell,
Todd Steiner, and Joe Wasilewski I
owe a great debt for their dedication
to the field work. I thank David But-
ler and SARLON Industries for the
donation of the shade cloth used to
make the fencing, and David W. Lee,
of F.I.U., for suggesting the use of
shade cloth to us. I thank the staffs of
the South Florida Research Center
and the Division of Resources Man-
agement of the park for their pa-
tience, generosity, perspectives and
spontaneous collection of specimens
for our studies. Most of all I wish to
thank Gary Hendrix and William B.
Robertson, Jr. for their interest and
support.

This research was sponsored by
the U.S. National Park Service-Flor-
ida International University Coop-
erative Agreement (CA-5000-3-8005,
Supplemental Agreement No.2, 1984)
and the Horida International Univer-
sity Foundation.

Literature Cited

Aleksiuk, Michael, and Patrick T.
Gregory. 1974. Regulation of sea-
sonal mating behavior in Thamno-
phis sirtalis parietalis. Copeia
1974:681-689.

Auffenberg, Walter. 1982. Horida
environments and their herpe-
tofaunas. Part III. Herpe togeogra-
phy. Florida Herpetologist.4:l-36.

Brower, James E., and Jerrold H. Zar.
1984. Field and ecological methods
for general ecology. Wm C.
Brown. Dubuque. 226p.

Brown, William S, and William S.
Parker. 1976. A ventral scale clip-
ping system for permanently
marking snakes (Reptilia, Serpen-
tes). Journal of Herpetology
10:247-249.

Brown, William S., and William S.
Parker. 1982a. Niche dimensions
and resource partitioning in a
Great Basin Desert snake commu-

84

nity. p.59-81. In Herpetological
Communities. Norman J. Scott, Jr.,
editor. U.S. Fish and Wildlife Serv-
ice, Wildlife Research Report 13.
Washington D.C.

Brown, William S., and William S.
Parker. 1982b. Growth, reproduc-
tion and demography of the racer.
Coluber constrictor mormon, in
northern Utah. In Contributions to
vertebrate ecology and sys-
tematics: a tribute to Henry S.
Fitch. Richard A. Seigel and oth-
ers, editors. Museum of Natural
History University of Kansas, Spe-
cial Publication No. 10. Lawrence.

Busack, Stephen D. and S.Blair
Hedges. 1984. Is the peninsular
effect a red herring? American
Naturalist 123:266-275.

Campbell, Howard W., and Steven P.
Christman. 1982a. The herpeto-
logical components of Florida
sandhill and sand pine scrub asso-
ciations, p. 163-171. In Herpetologi-
cal Communities. Norman J. Scott,
Jr., editor. U.S. Fish Wildlife Serv-
ice Wildlife Research Report 13.
Washington D.C.

Campbell, Howard W., and Steven P.
Christman. 1982b. Field tech-
niques for herpetofaunal commu-
nity analysis, p.193-200. In Herpe-
tological Communities. Norman J.
Scott, Jr., editor U.S. Fish and
Wildlife Service Wildlife Research
Report 13. Washington D.C.

Carpenter, Charles C. 1952. Com-
parative ecology of the common
garter snake (Thamnophis s. sir-
talis), the ribbon snake (Thamno-
phis s. sauritus) and Butler's garter
snake (Thamnophis butleri) in
mixed populations. Ecological
Monographs 22:235-258.

Chambers, John M., William S.
Cleveland, Beat Kleiner and Paul
A. Tukey. Graphical Methods for
Data Analysis, p. 395. Duxbury.
Boston, Mass.

Clark, Donald R. Jr. 1970. Ecological
study of the worm snake, Car-
phophis vermis (Kenicott). Museum
of Natural History Publication,
University of Kansas 19:85-194.

Clark, Robert F. 1958. An ecological
study of reptiles and amphibians
in Osage County, Kansas. The
Emporia State Research Stud-
ies.7:l-52.

Clawson, Mary E., and Thomas S.
Baskett. 1982. Herpetofauna of the
Ashland Wildlife Area, Boone
County, Missouri. Transactions,
Missouri Acad, of Science 16:5-16.

Cook, R.E. 1969. Variation in species
density in North American birds.
Systematic Zoology 18:63-84.

Crump, Martha L. 1971. Quantitative
analysis of the ecological distribu-
tion of a tropical herpetofauna.
Museum of Natural History, Uni-
versity of Kansas, Occasional Pa-
per 3:1-62.

Crump, Martha L. 1982. Amphibian
reproductive ecology on the com-
munity level, p. 21 -36. In Herpeto-
logical Communities. Norman J.
Scott, Jr., editor. U.S.Fish and
Wildlife Service Wildlife Research
Report 13. Washington D.C.

Davis, J. H. 1943. The natural features
of southern Rorida. Geological
Bulletin 25, Rorida Geological
Survey.

Dinnen, J. Walter. 1984. The fishes of
the everglades, p. 258-268. In Envi-
ronments of South Rorida Present
and Past II. Patrick J. Gleason, edi-
tor. Miami Geol. Soc, Miami.

Duellman, William E., and A.
Schwartz. 1958. Amphibians and
reptiles of southern Rorida. Bulle-
tin of the Rorida State Museum
3:181-324.

Ewel, John J., Dennis S. Ojima, Dori
A. Karl, and William F. DeBusk.
1982. Schinus in successional eco-
systems of Everglades National
Park. Report T-676. Homestead,
Rorida: Everglades National Park.
South Florida Research Center.

Gibbons, J. Whitfield, and Karen K.
Patterson. 1978. The reptiles and
amphibians of the Savannah River
Plant. Savannah River Ecology
Plant. DuPont Press, SRO-NERP-2:
1-24.

Gibbons, J. Whitfield, and Julian R.
Harrison III. 1981. Reptiles and

amphibians of Kiawah and Caper
Islands, South Carolina. Brimley-
ana 5:145-162.

Gibbons, J. Whitfield, and Raymond
D. Semlitsch. 1981. Terrestrial drift
fences with pitfall traps: an effec-
tive technique for quantitative
sampling of animal populations.
Brimleyana. 7:1-16.

Hamilton, W.J. 1951. The food and
feeding behavior of the garter
snake in New York State. Ameri-
can Midland Naturalist 46:385-
390.

Heatwole, Harold. 1982. A review of
structuring in herpetofaunal as-
semblages. p.1-19. In Herpetofau-
nal Communities. Norman J. Scott,
Jr., editor. U.S. Fish and Wildlife
Service Wildlife Research Report
13. Washington D.C.

Horn, Henry S. 1966. Measurement
of "overlap" in comparative eco-
logical studies. American Natural-
ist 100:419-424.

Johnson, J.M., I.C. Olmsted, and O.L.
Bass Jr. 1983. Vegetation map of
Long Pine Key, Everglades Na-
tional Park. South Rorida Re-
search Center. U.S. National Park
Service. Homestead.

Kiester, A. Ross 1971. Species diver-
sity of North American amphibi-
ans and reptiles. Systematic Zool-
ogy 20:127-137.

Kushlan, James A. and Frank J. Maz-
zotti. 1985. Environmental effects
on a coastal population of gopher
tortoises. Journal of Hepetology
18:231-239.

Layne, James N. 1984. The land
mammals of south Florida, p.269-
296. In Environments of South
Rorida Present and Past II. Patrick
J. Gleason, editor. Miami Geologi-
cal Society, Miami.

Loftus, William F. and James A.
Kushlan. 1987. Freshwater fishes
of southern Rorida. Bulletin of the
Rorida State Museum, Biological
Sciences. 31:147-344.

Loraine, Raymond K. 1985. Seasonal
changes in foraging success and
diet composition of Seminatrix
pygaea. Abstract from 1985 com-

85

bined meeting of the Society for
the Study of Amphibians and Rep-
tiles and Herpetologisf s League,
Tampa, Florida. August 4-9, 1985.

MacArthur, R.H., and Wilson. E.O.
1967. The theory of island biogeo-
graphy. Princeton University
Press. Princeton N.J.

Moler, Paul E. 1985. A new species of
frog (Ranidae: Ram) from north-
western Florida. Copeia. 1985:379-
383.

Mushinsky, Henry R. 1985. Fire and
the Rorida sandhill herpetofaunal
community: with special attention
to responses of Cnemidophorus
sexlineatus. Herpetologica 41:333-
342.

Myers, Charles W., and A. Stanley
Rand. 1969. Checklist of amphibi-
ans and reptiles of Barro Colorado
Island, Panama, with comments
on faunal change and sampling.
Smithsonian Contributions to Zo-
ology. 10:1-11. Smithsonian Insti-
tution Press. Washington D.C.

Olmsted, Ingrid C, Lloyd L. Loope,
and Charles E. Hilsenbeck. 1980.
Tropical hardwood Hammocks of
the interior of Everglades National
Park and Big Cypress National
Preserve. National Park Service,
South Florida Research Center
Report T-604. Everglades National
Park, Homestead, FL. 58 p.

Olmsted, Ingrid C, William B.
Robertson Jr., Jill Johnson, and
Oron L. Bass Jr. 1983. The vegeta-
tion of Long Pine Key, Everglades
National Park. National Park Serv-
ice, South Horida Research Center
Rep. Sn<C-83/05. Everglades Na-
tional Park, Homestead, FL. 64 p.

Olmsted, Ingrid C. and Lloyd L.
Loope. 1984. Plant Communities
of Everglades National Park. p.
167-184. In Environments of South
Horida Present and Past II. Patrick
J. Gleason, editor. Miami Geologi-
cal Society, Miami.

Preston, F.W. 1962. The canonical
distribution of commonness and
rarity. Ecology .43:185-215.

Reichenbach, Norman G., and
George H. Dalrymple. 1986. En-

ergy use, life histories, and the
evaluation of potential competi-
tion in two species of garter snake.
Journal of Herpetology. 20:131-
151.

Robertson, William B. Jr. 1955. An
analysis of the breeding-bird
populations of tropical Florida in
relation to the vegetation. Unpub-
lished Ph.D. thesis. University of
Ilhnois. Champa ign-Urbana.
599 p.

Robertson, William B. Jr., and James
A. Kushlan. 1984. The southern
Horida avifauna, p. 219-257. In
Environments of South Horida
Present and Past II. P.J. Gleason,
editor. Miami Geological Society.
Miami.

Rose, Paul W. 1977. Annual hydrol-
ogy review: Everglades National
Park and Big Cypress Preserve
1976. Everglades National Park
Report T-505. U.S. National Park
Service. Homestead. FL. 126 p.

Simpson, G.G. 1964. Species densities
of North American recent mam-
mals. Systematic Zoology 13:57-73.

Snyder, James R. 1986. The impact of
wet season and dry season pre-
scribed fires on Miami Rock Ridge
pineland. Everglades National
Park. National Park Service, South
Horida Research Center Report
SFRC-86/06. Everglades National
Park, Homestead, Fl. 106 p.

Taylor, Dale L., and Alan Hemdon.

1981. Impact of 22 years of fire on
understory hardwood shrubs in
slash pine communities within Ev-
erglades National Park. National
Park Service, South Horida Re-
search Center Report T-640. Ever-
glades National Park, Homestead,
H. 30 p.

Turner, F,B. 1977. The dynamics of
populations of squamates, croco-
dilians and rhyncocephalians. p.
157-264. In Biology of the Reptilia.
Carl Gans and Donald W. Tinkle,
editors. Academic Press. New
York.

Vogt, Richard C, and Ruth L. Hine.

1982. Evaluation of techniques for
assessment of amphibian and rep-

tile populations in Wisconsin.
p.201-217. In Herpetological Com-
munities. Norman J. Scott, Jr., edi-
tor. U.S. Fish and Wildlife Service
Wildlife Research Report 13.
Washington D.C.

Wiest, John A. Jr. 1982. Anuran Suc-
cession at temp)orary ponds in a
post oak-savanna region of Texas.
p.39-47. In Herpetological Com-
munities. Norman J. Scott, Jr., edi-
tor. U.S. Fish and Wildlife Service
Wildlife Research Report 13.
Washington D.C.

Wilson, Larry David, and Louis Por-
ras. 1983. The ecological impact of
man on the south Horida herpe-
tofauna. Museum of Natural His-
tory, University of Kansas, Special
Publication 9. 89+vi p.

Woodbury, A.M. 1951. Introduction —
a ten year study. Herpetologica.
7:4-14.

86

The Herpetofaunal
Community of Temporary
Ponds in North Florida Sand-
hills: Species Composition,
Temporal Use, and
Management Implications^

C. Kenneth Dodd, Jr.^ and Bert G. Ctiarest^

Abstract.— Amphibians and reptiles use an
isolated temporary wetland in a north Florida
sandhills throughout the year despite variation in
environmental conditions. Species composition and
number of individuals varies seasonally and
annually. Temporal variation in habitat use must be
considered in managing small wetlands and
assessing their importance to the herpetofaunal
community.

The sandhills and xeric live oak her-
petofauna of Florida is diverse and
contains a number of endemic spe-
cies. Whereas the terrestrial herpe-
tofauna has been described for a few
sandhills communities (Campbell
and Christman 1982, Mushinsky
1985), there have been no long-term
studies of the ecology of species us-
ing temporary ponds. For breeding
amphibians, sandhills temporary
ponds are often the only sources of
water that are free of predatory fish
and many larger predatory insects,
and such ponds may be extremely
important for amphibian reproduc-
tive success (Ma can 1966, Sexton and
Phillips 1986, Semlitsch 1987, Moler
and Franz 1988). At the same time,
the ephemeral nature of these breed-
ing sites makes reproductive success
uncertain and thus provides an op-
posing selective pressure for their
use (Semlitsch 1987).

Since January 1985, we have been
conducting studies on the herpe-
tofaunal community at a temporary
pond in a north-central Florida long-
leaf pine-turkey oak ("high pine")

'Paper presented at symposiurr). Man-
agement of Amphibians, Reptiles, and
Small Mammals in Nortti America. (Flag-
staff . AZ, July 19-21, 1988).

'C. Kenneth Dodd, Jr. is Zoologist (Re-
search), National Ecology Research Center,
U.S. Fish and Wildlife Service, 412 N.E. 16th
Avenue, Room 250, Gainesville, FL 32601.

^Bert G. Charest is Biological Aid (Wild-
life), National Ecology Research Center,
U.S. Fish and Wildlife Sen/ice, 412 N.E. 16th
Avenue, Room 250, Gainesville, FL 32601.

sandhills. Little is known of the com-
position of such Florida herpetofau-
nal communities, although Moler
and Franz (1988) reported 16 anuran
species breeding in various types of
wetlands surrounded by sandhills on
the 3750 ha Katharine Ordway Pre-
serve-Swisher Memorial Sanctuary in
Putnam County. Nothing is known
about movement patterns and activ-
ity cycles of the herpetofauna, or
about the numbers of individuals
breeding at such ponds and the num-
bers of offspring produced.

The purposes of our study are to
gain insight into the structure of the
herpetofaunal community using a
temporary pond in a sandhills eco-
system, to assess variation in species
composition and temporal use of the
pond, and to gather basic biological
information on the species that com-
prise the community. This paper
presents findings based on two years
of fieldwork of a projected five year
study.

Methods

Breezeway Pond, a 0.16 ha isolated
temporary pond in a shallow 1.3 ha
basin on the Katharine Ordway Pre-
serve-Swisher Memorial Sanctuary,
Putnam County, Florida, was en-
circled with a 230 m drift fence
(mean height = 36 cm above the sub-
strate) following the general proce-
dure of Gibbons and Semlitsch
(1982), reviewed by Jones (1986a).

Buckets were spaced at 10 m inter-
vals and paired on opposite sides of
the fence, making 23 stations of two
buckets each. Sloping covers were
put over the buckets and wet
sponges were placed in them to mini-
mize exposure to direct rays of the
sun and desiccation, respectively. As
a result, mortality among captured
animals was < 1.0% and was caused
primarily by invertebrate predation
(spiders, ants, centipedes, and
beetles).

Breezeway Pond is located at an
ecotone. To the immediate south and
west, the predominant habitat is
"high pine" sandhills dominated by
longleaf pine (Pinus palustris), turkey
oak (Quercus laevis) and wiregrass
(Aristida striata). A xeric hammock
dominated by sand live oak (Q. gemi-
nata) and laurel oak (Q. laurifolia)
faces the north, while a small "Pani-
cum meadow" dominated by
maidencane (Panicum hemitomon),
lies to the east. The distance from the
drift fence to the nearest forested
plant association is no more than
about 50 m in any direction.

Buckets were checked 5 days per
week in the morning (beginning
0700-0900 h depending on season)
from January 16 through April 12,
1985, and from October 1, 1985, until
September 30, 1987. For purposes of
discussion and analysis, a year refers
to a 12-month period from October
through the following September
(e.g. 1986 = October 1985 through
September 1986) because reproduc-

87

tion and metamorphosis generally
cease in early autumn while winter
breeding has yet to commence.

All reptiles and amphibians were
measured in the field (snout-vent
length, carapace and plastron length
[for turtles], tail length [for snakes
and glass lizards]), weighed and
marked for future identification us-
ing a year code (e.g., 0022 identifies
animals marked in 1986) or an indi-
vidual identification number (all
turtles, snakes, gopher frogs [Ram
areolata], red-tailed skinks [Eumeces
egregius], and ground skinks [Scin-
cella lateralis]). Very small animals,
mostly juvenile frogs and lizards,
were not marked because of their ex-
tremely small toes.

Notes were recorded on tail regen-
eration and damage, breeding and
hatchling coloration, and reproduc-
tive status. All animals, except liz-
ards, were released on the opposite
side of the fence from site of capture;
lizards were released on the same
side as captured. Weather condi-
tions, rainfall, pond water level, and
maximum and minimum air and wa-
ter temperatures were recorded.
These data are similar to those re-
corded in other long-term studies
employing drift fences to study am-
phibian communities (Gibbons and
Bennett 1974, Gibbons and Semlitsch
1982) and are vital to the inventory
and management of ecological com-
munities and individual species
aones 1986b).

In this paper we concentrate our
analyses on the two most commonly
captured amphibians, the striped
newt (Notophthalmus perstriatus), a
species listed as of special concern in
Florida (Christman and Means 1978),
and the eastern narrow-mouthed
toad (Gastrophryne carolinensis), a
common Florida frog (Carr 1940).

We also divided the year into bi-
weekly sampling periods and plotted
the cumulative number of species
captured versus sampling period.
The three years were plotted sepa-
rately. Data from October 1985
through September 1987 were treated

two ways: (1) as if sampling began in
October, and (2) as if sampling began
in April. This provided a between
year comparison of how effective
sampling for species numbers would
be if sampling began in the autumn
as opposed to the spring.

Statistical Analysis

Variation in the overall biweekly cap-
ture of amphibians and reptiles be-
tween 1986 and 1987 was compared
using a Chi-square contingency table.
The Spearman Rank Correlation Ma-
trix then was used to compare bucket
capture frequency between first cap-
ture and recaptured individuals, in
both 1986 and 1987, of G. carolinensis
and N, perstriatus. Since there were
no significant differences, captures
and recaptures were combined in
subsequent analyses.

We tested for within-year vari-
ation in capture frequency inside and
outside the fence using a one sample
Chi-square goodness of-fit-test. The
Spearman Rank Correlation Matrix
was again used to make the follow-
ing comparisons: (1) a within year
comparison of animals captured in-
side the fence with those captured
outside the fence for both 1986 and
1987 [both species], (2) a comparison
of juvenile with adult G. carolinensis
in 1986, (3) a comparison of juvenile
G. carolinensis inside and outside the
fence, and (4) a between year com-
parison of animals captured per
bucket inside or outside the fence
[both species].

To determine if N. perstriatus and
G. carolinensis preferentially oriented
to or from one of the three habitat
types surrounding the pond, data
were collapsed and analyzed using a
Kruskal-Wallis 1-way ANOVA.
Buckets 1-3 and 21-23 faced a xeric
hammock, 4-6 faced a small open
field, and 7-20 faced sandhills, thus
producing the three habitat catego-
ries.

Statistical analyses were carried
out using the SAS program for

microcomputers (SAS Institute Inc.
1985) or program ABSTAT version
4.09 (Anderson-Bell 1984). For all
analyses, P < 0.05 was considered in-
dicative of statistical significance.

Results

Environmental Conditions

Severe cold weather and a prolonged
drought characterized the sampling
period from January through April
1985. In Gainesville, 33 km west of
Breezeway Pond, low temperatures
reached -12 C and rainfall was 152.4
mm below normal for the three
month period. Breezeway Pond was
dry throughout this period. Summer
rains filled the pond in mid-July, and
water remained until December 16;
maximum pond depth was 60 cm but
declined steadily after September.
Free water was present from January
10-February 3 and from March 14 to
April 22, 1986. The pond remained
dry throughout the summer of 1986
despite summer thunderstorms and
did not refill until February 24, 1987.
From then until June 20 (115 days),
up to 60 cm of water filled the pond.
On June 20, the pond dried and re-
mained dry through September 30.

Species Composition

Thirty-nine species (7161 individual
captures) used the pond or its pe-
riphery at some point during the 27
months that the traps were moni-
tored (table 1). The amphibians cap-
tured most often were the winter/
spring breeding striped newt, Noto-
phthalmus perstriatus, and the spring/
summer breeding eastern narrow-
mouthed toad, Gastrophryne carolinen-
sis. Only one other salamander was
collected at Breezeway Pond, the
dwarf salamander, Eurycea quad-
ridigitata. Fourteen species of frogs
visited the pond, and six were pres-
ent at virtually any time of the year:
Acris gryllus, Bufo quercicus, B. ter-

88

catesbeiam were caught mainly in the
summer. Adult R. areolata were
caught in the early spring as they

restris, G. carolinensis, Limnaoedus ocu-
laris, and Scaphiopus holbrooki. Adult
Hyla femoralis and juvenile Ram

Table 1.— Species and numbers of individual amphibians and reptiles cap-
tured (first number) and recaptured (second number) at Breezeway Pond,
January 1985 through September 1987. * = very small Individuals not
marked.

1985

1985-1986

1986-1987

Total

(January-

(UCIODGr-

VVJCIODGi-

A nril\

oaiarnana©rs

*Eun/cea quadridigifafa

5/0

10/0

8/0

23/0

Nofophfhalmus persfriafus

29/5

558/309

744/226

1331/540

Frogs

*Acris gryllus

5/0

74/5

64/1

143/6

DU/L/ K^LJCfl ^I^LJO

1 /O

111/31

96/50

208/81

Bufo ferresfris

6/2

65/46

109/109

180/157

Eleufherodaciytus planirosfris

0/0

0/0

2/0

2/0

* Gasfrophryne carolinensis

2/0

1500/226

379/274

1881/500

Hyla chrysoscelis

0/0

1/0

0/0

1/0

Hyla femoralis

0/0

4/0

39/2

43/2

Hvin 'snuirf^lld

0/0

3/0

0/0

3/0

* Limnaoedus ocularis

14/0

20/0

49/0

83/0

Rana areolafa

2/1

9/5

46/23

57/29

Rana cafesbelana

2/4

9/4

0/0

11/8

Rar^a gryllo

1/0

0/0

5/0

6/0

Rana sphenocephala

0/0

5/0

15/2

20/2

Scaphiopus holbrooki

1/1

66/19

165/92

232/112

Turtles

Apalone ferox

0/0

6/4

0/0

6/4

Delrochelys reHcularla

0/0

2/0

0/0

2/0

KInosfernon subrubrum

9/0

7/0

11/4

27/4

Pseudemys Horldana

0/0

17/14

2/2

19/16

Lizards

Cnemidophorus sexllneafus

18/7

140/135

122/115

280/257

*Eumeces egregius

14/2

54/8

30/4

98/14

Eumeces inexpecfatus

0/0

0/0

1/0

1/0

*Ophlsaurus venfralls

0/0

14/2

15/0

29/2

Sceloporus undulafus

4/0

7/2

2/0

13/2

*Scincella lateralis

23/0

217/2

207/2

447/4

Snakes

Cemophora cocclnea

1/1

2/0

2/0

5/1

Coluber constrictor

2/0

7/0

8/8

17/8

DIadophls punctatus

0/0

2/0

2/2

4/2

MIcrurus fulvius

0/0

6/0

8/0

14/0

Nerodia fasciata

3/0

4/1

13/1

20/2

Nerodia floridana

1/0

6/1

4/0

11/1

Reglna alleni

2/1

1/0

2/0

5/1

Seminatrix pygaea

59/14

18/10

13/11

90/35

SIstrurus mlliarlus

0/0

4/0

2/1

6/1

J

moved toward breeding ponds, and
juvenile R. areolata and R. sphenoceph-
ala were caught in late summer and
early autumn presumably as they
emigrated to terrestrial habitats.

The most commonly captured rep-
tiles were the lizards Scincella later-
alis, Cnemidophorus sexlineatus and
Eumeces egregius, and the snake Semi-
natrix pygaea (table 1). Recent hatch-
lings accounted for all individuals of
the lizards Ophisaurus ventralis and
most S. lateralis, as well as the snakes
Coluber constrictor, Nerodia fasciata
and Thamnophis sirtalis, and the
turtles Pseudemys floridana and Kinos-
ternon subrubrum. The only snake
caught in substantial numbers was
the swamp snake, S. pygaea, espe-
cially as they left the pond during the
1985 drought.

Cunnulative Capture Rotes

The rate at which species were cap-
tured varied between 1986 and 1987
(fig. 1). More species were captured
at a faster rate in 1986 than in 1987
for sampling begun in October. How-
ever, the reverse was true for sam-
pling begun in April. In autumn, the
number of new species reached an
asymptote after about six weeks of
sampling in both years but at differ-
ent levels (25 in 1986, 23 in 1987). In
spring, the capture of new species
rose steadily both years; in 1986 it
never leveled off whereas in 1987 it
leveled off (at 31) only after four
months of sampling. In 1985, the rate
at which new species were observed
rose rapidly throughout the period
and was beginning to level off only
when the observations were termi-
nated.

In 1985, three months of sampling
produced 25 of the 39 (64%) species
now known to be present at Breeze-
way Pond. Corresponding percent-
ages for other years and durations of
sampling are as follows: 1986 - 6
months begun in October = 74%, 6
months begun in April = 77%, 12
months = 85%; 1987 - 6 months be-

89

gun in October = 59%, 6 months be-
gun in April = 82%, 12 months =
87%.

Variation in Biweekly Capture

The numbers of amphibians and rep-
tiles captured biweekly varied and
was significantly different between
1986 and 1987 for both amphibians
(X2 = 1366.46, 1 df, P < 0.001) and
reptiles (X^ = 128.08, 1 df, P < 0.001).
For amphibians, very few were
caught from October 1986 through
January 1987 compared with the
same period in 1985-1986. There also
were many fewer individuals caught
during the summer of 1987 com-
pared with 1986. This was due to a
late summer drought which resulted
in the complete drying of the pond
with subsequent reproductive failure
of G. carolinensis. Successful repro-
duction by this species in the sum-
mer of 1985 accounted for the large
numbers of amphibians captured in
1986 (fig. 2). Even if juvenile narrow-
mouthed toads are excluded (N =
690), there were still nearly 1000
more amphibians recorded in 1986
compared with 1987 (3425 in 1986,
2475 in 1987).

The numbers of reptiles recorded
in and around Breezeway Pond were
very similar between years, although
there was enough variation to make
the patterns significantly different.
As might be expected, reptile activity
decreased during the winter from
late October through mid-March al-
though some individuals were active
year round (fig. 3). The peak in num-
bers in mid-July 1986 represents both
a large number of species captured
as well as an influx of hatchling S.
lateralis.

Temporal Capture Variation:
Nofophthalmus perstriatus and
Seminatrix pygaea

An example of annual variation in
numbers of individuals and dates of

Figure 1 .—A comparison of the rate at which species were recorded for sampling from Janu-
ary-April 1985 (1985), October 1985-September 1986 (1986), and October 1986 through Sep-
tember 1987 (1987). For 1986 and 1987, the data were treated as if sampling began either in
October or April.

Figure 2.— Number of amphibiarw captured at Breezeway Pond in 1 986 and 1 987 by 2-week
intervals.

90

120 -1

--'-•--'--oooooooooSSoS-JJ

o o o

8 S S

Figure 3.— Number of reptiles captured at Breezeway Pond in 1986 and 1987 by 2-week Inter-
vals.

NOTOPHTHALMUS PERSTRIATUS - BREEZEWAY POND

capture is illustrated by comparing
collecting data from 1985 through
1987 for striped newts, N. perstriatus
(fig. 4), and swamp snakes, S. pygaea
(fig. 5). From mid-January through
mid-April, the numbers of newts
captured varied from 34 in 1985 to
364 in 1986 and 449 in 1987. Most
captures occurred from the first
week of February through the latter
part of March, and were associated
with rainfall > 10 mm. Movements in
1985 occurred despite bitter cold and
prolonged drought.

In contrast, striped swamp snakes
did not leave the pond during the
cold weather of 1985, but waited un-
til temperatures moderated in early
March (fig. 5). Unlike newts, how-
ever, they did not return in appre-
ciable numbers later in 1986 or 1987
despite favorable habitat and climatic
conditions.

Orientation and Movement
Patterns: Gastrophryne
carolinensis and Notophthalmus
perstriatus

<0 -

M •

108B N-34

f!' !•!« ;!

A

-1016 N-364
■•10B7 N-449

III

in!' i

rrrpr

HP

■

I

101 •»

i
i

TTT

*l 1 I

1.

JAN

FEB

MAR

APR

—I

Figure 4.— Comparison of thie numbers of striped newts (Notophthalmus persfriafus) cap-
tured from January 16 ttiroughi April 16, 1985-1987. The stars indicate days of > 10 mm rain-
fall.

The frequency of bucket capture,
both inside and outside the drift
fence, varied significantly for both
adult G. carolinensis and N. perstria-
tus in 1986 and 1987 (table 2). These
data indicate non-random movement
into and out of the pond. There was
no significant correlation between
inside and outside bucket capture
frequency for G. carolinensis in 1986
(r^ = -0.20, 22 df) or 1987 (r^ = -0.25,
22 df). There was significant correla-
tion between inside bucket captures
between 1986 and 1987 (r = 0.35, 22

s '

df) but not between outside bucket
captures between years (r^ = 0.06, 22
df). These results indicate that nar-
row-mouthed toads left the pond in
similar directions but entered it from
different directions.

Juvenile G. carolinensis entering
and exiting Breezeway Pond showed
distinct differences between capture
frequency at different stations (X^ =
535.73, df = 22, P < 0.001). However,

91

they showed no correlation with
adult capture frequency per station
(r^ = 0.09, 22 df). There also was no
correlation in bucket capture fre-
quencies for juveniles caught inside
and outside the drift fence (r^ = 0.26,
22 df). These data apply only to 1986
because no juveniles were observed
in 1987.

For N, perstriatus, there was like-
wise no significant correlation in in-
side versus outside bucket capture
frequency in 1986 (r^ = 0.23, 22 df) or
1987 (r^ = 0.03, 22 df). Capture fre-
quencies were compared outside the
fence in 1986 versus 1987 (r^ = 0.07,
22 df, P > 0.05) and inside the fence
in 1986 versus 1987 (r^ = 0.55, 22 df, P
< 0.01). As with Gastrophryne, these
results suggest that newts were leav-
ing the pond in similar directions be-
tween years, but that they were en-
tering it from different directions.

Habitat Relationships

Adult Gastrophryne did not m.ove to-
ward specific habitats in either 1986
(X2 = 2.62, 2 df, P = 0.27) or 1987 (X^
= 0.32, 2 df, P = 0.85). On the other
hand, juvenile narrow-mouthed
toads moved toward the sandhills at
a higher frequency than would be
expected if movements were random
(X2 = 13.31, 2 df, P = 0.001), but not
toward the pond from any particular
direction (X^ = 2.26, 2 df, P = 0.32).
Striped newts showed non-random
movement in 1986 (X^ = 7.79, 2 df, P
= 0.02) toward the sandhills but in
1987 moved toward the Panicum
meadow more often than would be
expected by chance alone (X^ = 9.42,
2 df, P = 0.009). Movement in rela-
tion to nearby habitat is illustrated in
figure 6.

Discussion

Was Sampling Effective?

Although we caught 39 species in >
7,000 captures, it is likely that more

14^

II •

>
<

o
c

U I

n
<

> •

o
z

4-

SEMINATRIX PYQAEA - BREEZEWAY POND

t

A

I
t

1 1

I ff

1 1

O' — - " 1985 N-73

1986 N-2

* 1987 N-1

II ■

It

JAN

FEB

a

ill I

ill I

III I

. I

i i

I Mill

I IIII.I

MAR

APR

te

Figure 5— Comparison of the numbers of swamp snal<es (Seminatrix pygaea) captured from
January 16 through April 16, 1985-1987. The stars indicate days of > 10 mm rainfall.

species of amphibians and reptiles
occasionally visit Breezeway Pond.
Some species, such as the eastern
coach whip snake (Masticophis flagel-
lum), Florida pine snake (Pituophis
melanoleucus), and gopher tortoise
(Gopherus polyphemus), are common

in adjacent sandhills but have not
been observed in or near the pond.
Large snakes (e.g., Pituophis, Mastico-
phis) could easily go over the fence
and thus avoid capture. The barking
treefrog (Hyla gratiosa) bred in the
pond before the initiation of our

r

Table 2.— Is the frequency of bucket capture random inside and outside
the drift fence? For ail analyses, there were 23 stations and 22 df. A signifi-
cant value indicates non-random movement.

Species

Year

Orientation

X2

P

Gosfrophryne

1986

Inside

55.68

< 0.001

carolinensis

1986

Outside

81.25

< 0.001

1987

Inside

84.00

< 0.001

1987

Outside

100.69

< 0.001

Nofophfhalmus

1986

Inside

243.56

< 0.001

perstriatus

1986

Outside

93.44

< 0.001

1987

Inside

88.45

< 0.001

1987

Outside

145.48

< 0.001

92

study (R. Franz, pers. comm.), but
we have never captured it or heard it
calling from the pond.

Some species, particularly
treefrogs such as Hyla femoralis,
might be able to climb over the fence
and thus go undetected (Gibbons
and Semlitsch 1982). Newts (N.
viridescens) are known to scale drift
fences (Semlitsch and Pechmann
1985) although we have not observed
N. perstriatus doing so. We have ob-
served a substantial number of un-
marked newts inside the drift fence
even after two years of study, but we
do not know if they were residents
that were moving after remaining in
the pond area for several years, or if
they entered by crawling over or un-
der the drift fence. Harris et al. (1988)
noted that many adult N. viridescens
burrowed into mud at the edge of
North Carolina sandhills ponds as
the ponds dried.

For these reasons, our data proba-
bly underrepresent both the number
of species and individuals using the
pond during the two years of obser-
vation. On the other hand, it is un-
likely that some species (e.g., Bufo,
Scaphiopus) are able to climb the
fence. As such, capture results of
these species may provide a reasona-
bly accurate estimate of pond use.

Activity Patterns

It is difficult to interpret data on ac-
tivity patterns of species with only
two years of data because there are
many variables that influence activity
cycles and the timing of reproduc-
tion. These variables, such as rainfall
amount and distribution, maximum
and minimum temperatures, and
hydroperiod (Wiest 1982, Semlitsch
1985, Pechmann et al. 1988), vary
daily, seasonally and yearly, and
may affect different species in differ-
ent ways. The subtle interaction of
these parameters probably accounts
for the variation in activity patterns
observed between years (Semlitsch
1985, Semlitsch and Pechmann 1985).

HABITAT DISTRIBUTION GC 1986 JUV : LEAVING POND

GO 1986 ADULT GC 1987

NP 1986 NP 1987

Figure 6.— Diagram illustrating the relationship between buckets, emigration from the pond,
and nearby habitat for Gastrophryne corolinensis (GC) and Notophthalmus perstriatus (NP).

93

Amphibians breeding in sandhills
ponds are faced with substantial un-
certainty as to whether or not suit-
able conditions will prevail for repro-
duction. Breezeway Pond was cho-
sen as the site for our study because
it had consistently held water from
the spring of 1983 through January
1985 (R. Franz, pers. comm.). Begin-
ning in January, climatic conditions
changed resulting in two years of
drought with only sporadic free wa-
ter. Temporary ponds may allow re-
production free of certain predators,
but their use comes at the cost of re-
productive uncertainty.

Amphibians are active during or
immediately after periods of rainfall
or high humidities. However, the
interaction of moisture and tempera-
ture and how they affect condensa-
tion probably affects diel activity
(Semlitsch and Pechmann 1985, Du-
ellman and Trueb 1986, Pechmann
and Semlitsch 1986) but also seasonal
activity.

The extremely dry conditions at
Breezeway Pond during the study
makes it difficult to predict whether
patterns observed in early 1985 and
from late 1985 through late 1987 are
"typical" for the amphibian commu-
nity using the pond. Observations
from other long-term studies of her-
petofaunal communities suggest that
there is wide variation in numbers of
individuals at a site and in reproduc-
tive success from year to year (Gill
1978, Semlitsch 1983, 1985, 1987,
Pechmann et al. 1988).

Because of their lack of depend-
ence on standing water, temperature
is probably more important than hy-
droperiod in governing reptile daily
and seasonal activity, at least for spe-
cies in direct spatial proximity to the
pond. However, reptile predators
that opportunistically visit tempo-
rary ponds, such as garter snakes
(Thamnophis sp.), might increase the
number of visits and duration of stay
if a sufficiently long hydroperiod al-
lows amphibian reproduction to take
place. Our data are insufficient as yet
to answer this question.

Some individuals are active even
during unfavorable environmental
conditions of drought and unseason-
ally cold temperatures. Amphibians
and reptiles are generally, but not
always inactive during cold or dry
weather. For instance, Semlitsch
(1983, 1985) noted that mole sala-
manders {Amhystoma sp.) in South
Carolina bred during the coldest but
not necessarily the wettest months.
He felt that most animals moved to
breeding ponds at this time to allow
sufficient time for larval develop-
ment prior to pond drying (Semlitsch
1987). Such may not explain winter/
early spring breeding in N. perstriatus
because the breeding period is ex-
tended (Bishop 1947) and larvae have
been found from April through De-
cember (Christman and Means 1978).
The larval period is unknown, but its
duration is critical to successful re-
production in temporary sandhills
ponds.

Individuals moving at times of
unusually cold and dry weather may
be searching for more favorable re-
treats or escaping adverse condi-
tions. If the onset of migration (sensu
Semlitsch 1985) commenced during
unusually adverse conditions, and
the unfavorable conditions extended
for a long period of time, the popula-
tion could be vulnerable to local ex-
tinction via mortality or emigration.
Prolonged drought brought about
the local extinction, via emigration,
of the resident Semimtrix population.

Movement Patterns and
Orientation

Because of the small size of Breeze-
way Pond, it is difficult to ascribe
directed movements of individuals
as migrating to, or originating from,
a specific habitat type. Because the
pond was located in an ecotone, an
animal captured at buckets facing the
interface between sandhills and xeric
hammock could move in either direc-
tion once beyond the fence. Likewise,
an animal originating from one habi-

tat type could be misclassified if it
moved a relatively short distance
and fell into a bucket facing a differ-
ent habitat type. The open field was
also rather small and, although we
did not feel comfortable assigning
buckets 4-6 to sandhills or xeric ham-
mock, it is likely that animals exiting
or entering the pond through these
buckets came from or went to one or
the other habitat.

Given these qualifications, adult
Gastrophryne did not exhibit habitat
preferences, although juveniles left
the pond primarily toward sandhills.
Gastrophryne are commonly recorded
in sandhills (Carr 1940, Campbell
and Christman 1982, Mushinsky
1985) and have been found in
sandhills > 100 m from the nearest
water source (Franz 1986, Dodd pers.
obs.). Xeric hammock or sandhills
apparently provide narrow-mouthed
toads suitable cover and resources
away from the breeding pond, but
why juvenile Gastrophryne would
move toward sandhills is unknown.

Striped newts are most commonly
found in flatwoods ponds in pine-
palmetto habitats (Christman and
Means 1978) as well as ponds in
sandhills and scrub areas (Campbell
and Christman 1982). To what extent
they use sandhills habitats away
from ponds is unknown. Carr (1940,
reported as N. v. symmetrica) re-
corded efts in high and mesophytic
hammocks in light, porous soil.
However, striped newts at Breeze-
way Pond moved toward sandhills
or meadow rather than hammock.
Migration distances of striped newts
are unknown although displaced N.
viridescens can move 400 m through
deciduous forest to return to a resi-
dent pond (Gill 1979). N. perstriatus
probably can travel similar distances
in its migrations.

Management implications

The Florida sandhills are undergoing
extensive habitat alteration because
of rapid human population growth

94

and associated development. In the
late 1970's, Auffenberg and Franz
(1982) estimated that 70.6% of the
sand pine-scrub oak, 57% of the long-
leaf pine, and 37.7% of the xeric ham-
mock communities had been de-
stroyed by forest plantation agricul-
ture and urbanization. In Putnam
County, the site of our study, > 50%
of the land area originally supporting
such communities no longer does so.
With projected human population
increases of more than 300% between
1972 and 2000 (Auffenberg and Franz
1982), there has been increasing con-
cern for the loss of sandhills habitats
in northern and central Florida. Ex-
tensive loss of habitat is occurring in
other portions of the state and South-
east, such that only 14% of the long-
leaf pine (Pinus palustris) forests re-
main from estimates of over 70 mil-
lion acres that once comprised this
community (Means and Grow 1985).

Because of habitat loss, amphibian
and reptile populations dependent
upon sandhills probably are declin-
ing. Many of the amphibians, such as
the Rorida gopher frog. Ram areolata
aesopus, and the striped newt, N. per-
striatus, are considered endangered,
threatened, or rare (Fogarty 1978,
Christman and Means 1978), yet
there are few data on their life histo-
ries or population dynamics.

The paucity of information on spe-
cies composition and population dy-
namics of amphibians and reptiles
that use temporary ponds in xeric
habitat masks the probable impor-
tance of such habitats. Variation in
annual habitat use, both intraspecifi-
cally and inter-specifically, appears
to be considerable. Long-term eco-
logical studies of the herpetofaunal
community are needed to under-
stand the magnitude of such vari-
ation and its potential significance.

Information on the biology of the
species comprising the sandhills her-
petofaunal community could be im-
portant in planning for the manage-
ment of sandhills ecosystems by
State and Federal agencies. For in-
stance, Florida Statutes Section

373.414 required Water Management
Districts to adopt rules to establish
specific permitting-criteria for small
isolated wetlands, including size
thresholds below which impacts on
fish and wildlife habitats would not
be considered. When these rules
were adopted, almost no data were
available on herpetofaunal communi-
ties on which to make recommenda-
tions for size threshold considera-
tions. Lack of information led, in
part, to variation among regulations
adopted by the different Water Man-
agement Districts.

There is considerable interest
among Rorida biologists, conserva-
tionists, and land use planners in the
concept of wildlife corridors to main-
tain biotic diversity (Harris 1985).
Unfortunately, most discussions
have centered on riparian habitats.
The lack of data on sandhills habitat
use, especially by candidate endan-
gered or threatened species, could
hamper the long-term survival of
such species. Many sandhills species
are likely dependent on small iso-
lated wetlands for at least a portion
of their life cycle. By focusing on ri-
parian habitats, planners may be
overlooking the importance of up-
land habitats and their associated
small wetlands to the maintenance of
biotic diversity.

The following are the most impor-
tant implications of our study for the
conservation and management of
small isolated wetlands and their as-
sociated herpetofaunal communities
in "high pine" xeric habitats in
northern and central Rorida. These
should be kept in mind when evalu-
ating impacts of habitat loss and
planning assessment studies.

1. Many species use these habi-
tats: some are permanent
residents, some are migrants,
and some wander through
the area on an irregular ba-
sis. All pond-breeding spe-
cies live in surrounding ter-
restrial habitats during the
non-breeding season. Thus,

the pond and a portion of the
terrestrial habitat are both
critical to species persistence.

2. Such habitats are used year-
round despite seemingly un-
favorable periods of drought
and cold weather.

3. Species composition varies
within a year: some species
are found only in one season,
some predominate at one
time but are found com-
monly at other times, some
are very rarely observed.

4. Reproductive output among
species varies considerably:
in one year spring breeders
may be successful, in other
years summer breeders may
be successful, in some years
both probably produce
young, in other years neither
may successfully reproduce.
The longer that studies are
conducted, the greater is the
likelihood that multiple pat-
terns will emerge.

5. Activity patterns change sea-
sonally and annually proba-
bly in response to environ-
mental cues, particularly
rainfall, temperature, and
hydroperiod.

6. To determine the total num-
ber of species using such
wetlands, spring and early
summer sampling produces
the best results, but single
season or even yearly sam-
pling will not catch all spe-
cies.

7. Quick surveys underestimate
both numbers of species and
individuals, as well as an-
nual variation, and thus un-
derestimate the importance
of temporary isolated wet-
lands in sandhills.

95

8. To adequately understand
complex communities, long-
term studies are absolutely-
essential for management
and conservation.

Acknowledgments

We thank H. 1. Kochman for advice
on statistical analyses, and R. Ash-
ton, R. Franz, J. Oldemeyer, J. H. K.
Pechmann, R. Seigel and R. D. Sem-
litsch for their comments on the
manuscript. R. L. Burke, K. M. Enge,
and J. N. Stuart assisted with various
phases of fieldwork.

Literature Cited

Anderson-Bell. 1984. ABSTAT. Re-
lease 4. Anderson-Bell, Canon
City, Colorado.

Auffenberg, Walter, and Richard
Franz. 1982. The status and distri-
bution of the gopher tortoise (Go-
pherus polyphemus). p. 95-126. In
North American Tortoises: Con-
servation and ecology. R. Bruce
Bury, editor. Fish and Wildlife
Service, Wildlife Research Report
12.

Bishop, Sherman C. 1947. Handbook
of Salamanders. Cornell Univer-
sity Press, Ithaca, New York.

Campbell, Howard W., and Steven
P.Christman. 1982. The herpeto-
logical components of Horida
sandhill and sand pine scrub asso-
ciations, p. 163-181. In Herpeto-
logical Communities. Norman J.
Scott, editor. Fish and Wildlife
Service, Wildlife Research Report
13.

Carr, Archie Fairly, Jr. 1940. A contri-
bution to the herpetology of Ror-
ida. University of Florida Publica-
tion, Biological Science Series
3(1):1-118.

Christman, Steven P., and D.Bruce
Means. 1978. Striped newt Notoph-
thalmus perstriatus. p. 14-1 5. 7n Rare
and Endangered Biota of Horida,
Vol. 3, Amphibians & Reptiles.

Roy W. McDiarmid, editor. Uni-
versity Presses of Horida,
Gainesville.
Duellman, William E., and Linda
Trueb. 1986. Biology of Amphibi-
ans. McGraw-Hill Book Co., New
York.

Fogarty, Michael J. 1978. Florida go-
pher frog Rana areolata aesopus.
p.5-6. In Rare and Endangered Bi-
ota of Florida, Vol. 3, Amphibians
& Reptiles. Roy W. McDiarmid,
editor. University Presses of Flor-
ida, Gainesville.

Franz, Richard. 1986. Gopherus pol-
yphemus (Gopher Tortoise). Bur-
row commensals. Herpetological
Review 17:64.

Gibbons, J. Whitfield, and David H.
Bennett. 1974. Determination of
anuran terrestrial activity patterns
by a drift fence method. Copeia
1974:236-243.

, and Raymond D. Semlitsch.

1982. Terrestrial drift fences with
pitfall traps: an effective technique
for quantitative sampling of ani-
mal populations. Brimleyana 7:1-
16.

Gill, Douglas E. 1978. The metapopu-
lation ecology of the red-spotted
newt; Notophthalmus viridescens
(Rafinesque). Ecological Mono-
graphs 48:145-166.

. 1979. Density dependence and

homing behavior in adult red-
spotted newts Notophthalmus
viridescens (Rafinesque). Ecology
60:800-813.

Harris, Larry D. 1985. Conservation
corridors, A highway system for
wildlife. ENFO (Florida Conserva-
tion Foundation), November, 10 p.

Harris, Reid N., Ross A. Alford, and
Henry M. Wilbur. 1988. Density
and phenology of Notophthalmus
viridescens dorsalis in a natural
pond. Herpetologica 44:234-242.

Jones, K. Bruce. 1986a. Amphibians
. and reptiles, p.267-290. In Inven-
tory and Monitoring of Wildlife
Habitat. A. Cooperrider, R.J.Boyd
and H.R.Stuart, editors. Bureau of
Land Management Service Center,
Denver, Colorado.

. 1986b. Data types, p. 11-28. In

Inventory and Monitoring of
Wildlife Habitat. A. Cooperrider,
R. J. Boyd and H. R. Stuart, edi-
tors. Bureau of Land Management
Service Center, Denver, Colorado.

Macan, T. T. 1966. The influence of
predation on the fauna of a
moorland fishpond. Archiv fur
Hydrobiologie 61:432-452.

Means, D. Bruce, and Gerald Grow.
1985. The endangered longleaf
pine community. ENFO (Horida
Conservation Foundation), Sep-
tember, 12 p.

Moler, Paul E., and Richard Franz.
1988. Wildlife values of small, iso-
lated wetlands in the southeastern
coastal plain, p. 234-241, In Pro-
ceedings of the Third Southeastern
Nongame and Endangered Wild-
life Symposium. R. R. Odom, K. A.
Riddleberger, and J. C. Ozier, edi-
tors. Georgia Department of Natu-
ral Resources, Athens, Georgia.

Mushinsky, Henry R. 1985. Fire and
the Horida sandhill herpetofaunal
community: with special attention
to responses of Cnemidophorus
sexlineatus. Herpetologica 41:333-
342.

Pechmann, Joseph H. K. and Ray-
mond D. Semlitsch. 1986. Diel ac-
tivity patterns in the breeding mi-
grations of winter-breeding anu-
rans. Canadian Journal of Zoology
64:1116-1120.

, David E. Scott, J. Whitfield Gib-
bons, and Raymond D. Semlitsch.
1988. Influence of wetiand hy-
droperiod on diversity and abun-
dance of metamorphosing juvenile
amphibians. Wetlands Ecology
and Management 1:1-9.

SAS Instihite Inc. 1985. SAS Intro-
ductory Guide for Personal Com-
puters, Version 6 Edition. SAS In-
stitute Inc., Gary, North Carolina.

Semlitsch, Raymond D. 1983. Struc-
ture and dynamics of two breed-
ing populations of the eastern ti-
ger salamander, Amhystoma ti-
grinum. Copeia 1983:608-616.

. 1 985. Analysis of climatic factors

influencing migrations of the

96

salamander Amhy stoma talpoideum.
Copeia 1985: 477-489.

. 1987. Relationship of pond

drying to the reproductive success
of the salamander Amby stoma
talpoideum. Copeia 1987:61-69.

, and Joseph H. K. Pechmann.

1985. Diel pattern of migratory
activity for several species of
pond-breeding salamanders.
Copeia 1985:86-91.

Sexton, Owen J., and Christopher
Phillips. 1986. A qualitative study
of fish-amphibian interactions in 3
Missouri ponds. Transactions,
Missouri Academy of Science
20:25-35.

Wiest, John A., Jr. 1982. Anuran suc-
cession at temporary ponds in a
post oak-savanna region of Texas,
p. 39-47. In Herpetological Com-
munities. Norman J. Scott, Jr., edi-
tor. Fish and Wildlife Service,
Wildlife Research Report 13.

97

Management of Annphlbians,
Reptiles, and Snnall Mammals
In Xeric PInelands of
Peninsular Florida^

I. Jack $tout,2 Donald R. Richardson,^ and
Richard E. Roberts^

Abstract.— The primary xeric pinelands of peninsu-
lar Florida are longleaf pine/turi<ey oak sandhills and
sand pine scrub. Their management on public lands
is largely confined to prescribed burning to maintain
fire climax status of the vegetation. The regulation of
large-scale developments on private land has stimu-
lated interest in preserve design and management.
The suite of techniques used to solve conflict be-
tween natural system preservation and develop-
ment includes: (1) conservation set asides (pre-
serves) on site; (2) habitat restoration; (3) purchase
and dedication of off-site preserves; (4) species relo-
cation; and (5) wildlife resource mitigation fund.

Xeric pinelands seem incongruent
with reference to Florida, a state with
annual rainfall that ranges from 50-65
in (19.6-25.6 cm). Nonetheless, the
Florida peninsula contains thousands
of acres of sandy soil derived from
marine deposits dating to the Pleisto-
cene (White 1970). Two distinct plant
associations, longleaf pine (Pinus
palustris)/ turkey oak (Quercus laevis)
sandhill and sand pine scrub (Pinus
clausa), have developed on these nu-
trient deficient and excessively well-
drained soils. Significant areas of
these plant associations occur at
higher, albeit modest, elevations rela-
tive to the surrounding landscape. In
fact, certain topographic features,
e.g., the Lake Wales Ridge and the
Marion Upland, were likely to have
been true islands during interglacial
periods while the remainder of Flor-
ida was covered by a shallow sea.
Regardless of their exact origin, xeric
pinelands support many relatively
unusual species of amphibians, rep-
tiles, and small mammals.

' Paper presented at symposium. Man-
agement of Amphibians, Reptiles, and
Small Mammals in North America. (Flag-
staff , AZ, July 19-21, 1988).

'I. Jack Stout is Professor of Biological
Sciences, Department of Biological Sci-
ences, University of Central Florida, Box
25000, Orlando, FL 32816-0368.

^Donald R. Richardson is Adjunct Profes-
sor, Department of Biology, University of
South Florida, Tampa, FL 33620.

^Richard E. Roberts is Biologist, Division of
Recreation and Parks, Florida Department
of Natural Resources, Hobe Sound, FL
33455.

Human population growth (3.3%
per year) and development in Florida
continues to encroach on upland
habitats and particularly on xeric pi-
nelands. Most of the habitat loss is to
agricultural uses, principally citrus.
Oddly, the state's excellent wetlands
protection acts have forced develop-
ment into the uplands. Thus, xeric
pinelands and their narrowly
adapted fauna and flora are increas-
ingly threatened by area reduction,
fragmentation and isolation.

It is our intent to discuss the man-
agement of these xeric pinelands in
general and, more specifically, in the
context of small preserves in an oth-
erwise developed landscape. Man-
agement of xeric pinelands as ecosys-
tems is yet in its infancy, and more
detailed prescriptions for designated
species are unproven. However,
progress is being made (Cox et al.
1987) and improvement and revision
of current thinking on management
practices is anticipated. This paper
summarizes selected literature on
xeric pineland and the species associ-
ated with these communities to as-
sess management practices. In addi-
tion, unpublished information has
been used and identified in the text.

Preserve design efforts by us have
been on behalf of developers re-
sponding to development orders pre-
pared by governmental agencies.
These designs are site specific in de-
tail, but nonetheless point to general
problems and solutions. Our ap-
proach has been to focus on provid-

ing the area required to support
minimum viable populations of
"keystone" or otherwise critical ani-
mal species of a given xeric pineland.
Once this area is settled on, manage-
ment should focus on those species
whose minimum area requirements
are met, whereas no special efforts
are expended on species with larger
area requirements.

XERIC PINELAND HABITATS

Longleaf Pine/Turkey Oak
Sandhills

The longleaf pine /turkey oak
sandhill association (LLP/TO) was
about 15% (2, 110, 256 ha) of the
natural landscape of peninsular Hor-
ida in pre-Columbian times (fig. 1)
(Auffenberg and Franz 1982). This
xeric pineland occupies rolling to-
pography of several ridge systems
that run north-south, notably Trail
Ridge, the Lake Wales Ridge, and the
Brooksville Ridge; numerous lesser
ridges and hills are identified by
White (1970). These ridges consist of
deep, well-drained soils of the
Lakeland, Eustis, and Blanton asso-
ciations (Beckenbach and Hammett
1962). Laessle (1942, 1958a) describes
the LLP/TO plant association as a
fire climax system dominated by
longleaf pine; slash pine (Pinus elli-
otti) replaces longleaf pine in the
community in south Florida. Turkey
oak is a minor tree, but can achieve

98

co-dominance when fires are sup-
pressed. The predominant under-
story plant is wiregrass, Aristida
stricta; however, a rich assemblage of
perennial herbs vary in prominence
in concert with seasonal changes.
Monk (1968) recognizes two addi-
tional phases of sandhill vegetation
in north central Rorida: (1) longleaf
pine/sand post oak (Quercus marga-
retta) and (2) longleaf pine /southern
red oak (Quercus falcata). A fourth
phase, longleaf pine/scrub hickory
(Carya floridam), occurs in the south-
ern portion of the Lake Wales Ridge
(Abrahamson et al. 1984). Veno
(1976), Givens et al. (1984), and Abra-
hamson et al. (1984) provide quanti-
tative data on LLP /TO community
structure and dynamics. Myers
(1985) suggests that longleaf pine/
turkey oak and sand pine scrub asso-
ciations are successionally linked in
some portions of their geographic
ranges. Differences in physical/
chemical features of soils of LLP/TO
and SPS communities in the Ocala
National Forest are not considered to
be sufficient to explain the local dis-

Figure 1 .—Potential geographic distribution
of longleaf pine/turkey oak sandhill and
sand pine scrub xeric pinelands in Florida.
Light shading indicates the sandhills and
darker shading indicates the scrub. These
distributions are based on Davis (1980) and
do not reflect nninor sites of either commu-
nity due to the scale of the illustration.

tribution of the communities (Kalisz
and Stone 1984).

Prior to settlement by European
man, ground fires occurred in LLP/
TO sandhills at intervals of 1-5 years.
These relatively "cool" fires favor
regeneration of longleaf pine, flower-
ing by grasses and herbs, and sup-
press growth of woody plants
(Myers 1985).

Sand Pine Scrub

Compared with the LLP/TO sandhill
association, sand pine scrub (SPS)
has less area (250,000 ha) and a far
more limited distribution (fig. 1).
Scrub is associated with old shore-
lines, lake margins, and stream
courses where extremely well
washed, nutrient deficient sands
were deposited during Pleistocene
times (Kurz 1942; Laessle 1958a, b,
1967). The most widespread soils
supporting SPS are the St. Lucie,
Lakewood, and Pomello associations
(Beckenbach and Hammett 1962).

Sand pine scrub is a two-layered
community. Sand pine (Pinus clausa)
normally occurs as a relatively even-
aged overstory species. The under-
story is comprised of 10-20 species of
evergreen shrubs l-5m in height.
Four species of oaks comprise the
bulk of the biomass, Quercus gemi-
nata, Q. myrtifolia, Q. chapmanii, and
Q. inopina. Lesser numbers of other
species including Ceratiola ericoides,
Lyonia ferruginea, and Osmanthus
americanus add to local diversity.
Sand pine scrub is a fire climax com-
munity (Laessle 1958a, Abrahamson
et al. 1984). In contrast with LLP/TO,
SPS burns at intervals of 20-70 years;
a combination of ground and crown
fires destroys all the above-ground
vegetation. Most of the woody
plants, with the notable exception of
the sand pine and Ceratiola, readily
sprout from root crowns following
fires. Laessle (1958a), Veno (1976),
and Richardson (1977) provide data
on plant community structure of
scrubs. Recent quantitative studies

include those of Abrahamson et al.
(1984) and Latham (1985).

Outstanding examples of SPS in-
clude the "Big Scrub," part of the
Ocala National Forest, scrubs of the
Lake Wales Ridge, e.g., the Archbold
Biological Station, and stands along
the Atlantic Coastal Ridge.

SMALL VERTEBRATE SPECIES
ASSEMBLAGES

Longleaf Pine/Turkey Oak
Sandhills

Amphibians and Reptiles

At least 47 species of herptiles, in-
cluding 2 newts, 13 toads and frogs,
3 turtles, 10 lizards, 1 amphis-
baenian, and 18 snakes, are reported
to occur in LLP/TO habitats (table 1).
Campbell and Christman (1982) list 5
categories of reptile and amphibian
species that occur in LLP/TO and
SPS: (1) characteristic (18 species); (2)
associated with tortoise burrows (3
species); (3) frequent (8 species); (4)
occasional (14 species); and (5) asso-
ciated with aquatic habitats (21 spe-
cies). Of the characteristic species, 7
are regarded as adapted to xeric con-
ditions, 3 as sand swimmers, viz.,
Neoseps reynoldsi; Eumeces egregius,
and Tantilla relic ta, and the
remainder (Sceloporus woodi, Mastico-
phis flagellum, Stilosoma extenuatum,
Cnemidophorus sexlineatus) to other
physical features of the habitats.

The gopher tortoise (Gopherus pol-
yphemus) is a terrestrial turtle that
digs deep burrows in the well-
drained sandhill soils (Auffenberg
and Franz 1982). Stout (1981) and
Eisenberg (1983) recognized the go-
pher tortoise was the keystone spe-
cies in xeric pinelands. Some 80 spe-
cies of animals may be classified as
burrow commensals (Cox et al. 1987);
however, the number of obligatory
commensals is much smaller. Herp-
tiles particularly associated with go-
pher tortoise burrows include Rana

99

areolata, Pituophis melanoleucus, and
Drymarchon corais.

The snake fauna of LLP/TO
sandhills is species rich (> 18 spe-
cies). This diversity includes large
forms, e.g., Drymarchon corais couperi
and Cro talus adamanteus, and small,
specialized species like Stilosoma ex-
tenuatum. This latter ophiophagous
species feeds largely on Tantilla
relicta; Tantilla, is in turn specialized
on Tenebrionidae larvae (Mushinsky
1984).

Small Mammals

At least 19 species of small mammals
with body masses less than 6.0 kg

may be anticipated in LLP/TO sand-
hills (table 2). Two are fossorial,
Scalopus aquations and Geomys pinetia,
1 semi-fossorial, P. polionotus, and 2
occur in the surface litter, Blarina
carolinensis and Cryptotis parva.

Arboreal species include Sciurus
carolinensis, S. niger, Glaucomys volans,
P. gossypinus, and Ochrotomys
nuttalli. Podomys floridanus nests in
the burrows of the gopher tortoise
and the pocket gopher (Layne 1969);
it may enlarge other openings in the
soil to establish burrows independ-
ently of the gopher tortoise (R. E.
Roberts, personal observation).
Dasypus novemcinctus is the only ex-
otic species of mammal that is clearly

established in the sandhill commu-
nity.

Sand Pine Scrub

Amphibians and Reptiles

Campbell and Christman (1982)
listed 64 species of reptiles and am-
phibians that may be found in LLP/
TO sandhills and SPS. Pitfall trap-
ping in six different even-aged
stands of SPS on the Ocala National
Forest by Christman et al. (unpub-
lished manuscript and personal com-
munication) revealed 27 species
(table 1). Of 1,624 individuals

Table 1.— Herpetofauno trapped or observed within the xeric pinelonds of peninsula Florida. Standard herp arrays
were used in each study to sample for a period of at least one year.

Species Long leaf pine/ Sand pine Species Long leaf pine/ Sand pine

turkey oak scrub turkey oak scrub

Campbell Mushinsky Stout Christman Campbell Mushinsky Stout Christman

& Christman 1965 etaL etai. & Christman 1985 etal. etal.

1982 unpubl. unpubl. 1982 unpubl. unpubl.

Nofophfhalmus

viridescens
N. persfriafus
ScopNopus holbrookii
Bufo ferresfris
Bufo quercicus
Eleufh ero dac fylus

planirosfris
HyJa fern or alls
Hyla grafiosa
Hyla squirella
Hyla cinerea
Acris gryllus
Rana gryllo
Rana areolata
Rana ufricularia
Gasfrophryne

carolinensis
Kinosfernon bauri
Terrapene Carolina

bauri

Gopherus polyphemus
Anolis carolinensis
Anolis sagrei
Sceloporus undulafus
Sceloporus wood!
Ophisaurus compressus
Cnemldophorus
sexlineatus

X
X

X
X
X

X
X

X
X
X

X

X

X
X

X
X
X
X
X
X

X
X
X
X

X
X

Scincella lateralis x

Eumeces inexpectatus x

E, egregius lividus —

E. egregius onocrepis x

Neoseps reynoldsi x

Rhineura floridgna x

Nerodia fasciafa —

Th amn ophis saurifus —

Riiadinaea flavilafa —

Diadophis punc tafus x

Paranoia abacura —

Coluber constrictor x

Masticophis flagellum x

Opheodrys aestivus x

Drymarclion corais —

Baphe guttata —

Pituophls melanoleucus x
Lampropeltis

triangulum x

Stilosoma extenuatum x

Cemophiora coccinea x

Tantilla relicta x

Heterodon platyrhinos x

Heterodon simus —

Micrurus fulvius fulvius x

Sistrurus millarius x

Crotalus adamanteus —

Totals 29

X
X
X

X

27

X
X
X
X
X
X

X
X
X
X
X

33

27

100

trapped, the common species were
Bufo terrestris (n=332), Cnemidophorus
sexlineatus (n=329), and Sceloporus
woodi (n=216); five species were rep-
resented by single captures. Christ-
man et al. concluded that the herpe-
tofaunal diversity declined with in-
creasing age of SPS stands.

Gopherus polyphemus is the key-
stone species in SPS but is less com-
mon there than in LLP /TO (Auf fen-
berg and Franz 1982). Many, if not
most, of the burrow commensals are
common in SPS (Cox et al. 1987).
Podomys floridanus is an example.

[ Small Mammals

Fourteen species of small mammals
commonly inhabit SPS (table 2). Pod-

omys floridanus is a predictable mem-
ber of the assemblage throughout the
range of scrubs in peninsular Florida
(Layne 1978). Three subspecies of
Peromyscus polionotus occur in scrubs
of the interior and east coast portions
of the peninsula. Common small
mammals in central peninsular Flor-
ida scrubs include Podomys floridanus,
Peromyscus gossypinus, Ochrotomys
nuttalli, and Glaucomys volans (Swin-
dell 1987). Podomys floridanus is the
predominate small mammal in
scrubs of southeast Florida
(Richardson etal. 1986).

Limited data suggest Spilogale
putorius is a major predator on small
mammals in scrubs with lesser roles
played by Mephitis mephitis and
Mustela frenata (Stout and Roberts,
personal observations).

Table 2.— Small mammal community structure In sandhill and sand pine
scrub plant associations of peninsular Florida. The upper limit of body mass
of small mammals was arbitrarily set at 6.0 kg.

Mammal Species Longleaf pine/turkey oak' -^-^ Sand pine scrub*

Didelphis virginiana

X

X

Crypfofis parva

X

Blarina carolinensis

X

X

Scalopus aquaficus

X

Dasypus novemcinctus

X

X

SyMlagus floridanus

X

X

Sciurus carolinensis

X

X

Sciurus niger

X

Glaucomys volans

X

X

Geomys pinefis

X

Peromyscus polionotus

X

X

Peromyscus gossypinus

X

X

Podomys floridanus

X

X

Ochrotomys nuttalli

X

X

Sigmodon hispidus

X

X

Urocyon cinereoargenfeus

X

Procyon lotor

X

X

Mustela frenata

?

X

Spilogale putorius

X

X

Mephitis mephitis

X

No. Species

19

14

^ St out et al., unpublished
^Arata 1959
^Humphrey et al. 1985
"Stout 1982

ENDANGERED AND THREATENED
SPECIES

Ten species of amphibians, reptiles,
and small mammals associated with
xeric pineland are currently listed as
having some level of threatened, en-
dangered, or sensitive status by ei-
ther the state of Florida or the De-
partment of Interior (table 3). The
extensive overlap in species composi-
tion between the two pineland com-
munities results from the high num-
ber of species common to both types.
The Endangered Species Act charges
federal agencies with the responsibil-
ity to manage federally listed species
on federally owned lands. At the
state level, preservation of these
listed species is of major concern
when they occur on parcels of land
scheduled for large-scale develop-
ment. Preserve design and manage-
ment practices for these species have
largely evolved on an ad hoc basis
without adequate time for an evalu-
ation of the management or the long-
term implications for the species.

MANAGEMENT OF XERIC
PINELANDS ON PUBLIC LANDS

Of three national forests in Rorida,
only the Ocala National Forest is lo-
cated in the peninsula. It totals
153,846 ha of which 85,020 ha are SPS
and 18,219 ha LLP/TO. The National
Forest Management Act (1976) and
pursuant regulations (36 CFR 219)
require that each forest be managed
to maintain well-distributed and vi-
able populations of wildlife species,
including species that are endan-
gered or threatened (Norse et al.
1986).

Silvicultural systems differ be-
tween the two pineland communi-
ties. On the Ocala National Forest
sand pine scrub is routinely har-
vested in patchy clearcuts that range
from 16-24 ha in area. Scrub under-
story vegetation is allowed to regen-
erate naturally; however, sand pine
is seeded following site preparation

101

by a single roller chopping. The har-
vest rotation length is about 50 years.
In contrast, LLP /TO is ostensibly
managed on a 80-100 year rotation
and shelterwood cutting favors natu-
ral regeneration of the longleaf pine
(Don Bethancourt, personal commu-
nication). In practice, harvesting of
longleaf pine may occur in 60 years.

Effectiveness of ecosystem man-
agement in the SPS community v;^ill
be judged by the response of desig-
nated indicator species, such as go-
pher tortoises and scrub jays (Aphelo-
coma coerulescens) (table 3). The go-
pher tortoise is also a designated in-
dicator species for the LLP/TO com-
munity. The significance of the go-
pher tortoise as a keystone species
was emphasized in 1986 when har-
vesting of the species on national for-
ests in Florida was made illegal
through an agreement between the
U.S. Forest Service and the Florida
Game and Fresh Water Fish Commis-
sion. Other species-specific manage-
ment practices involving amphibians,
reptiles, or small mammals have not

been deemed necessary to carry out
on the Ocala National Forest (Don
Bethancourt, personal communica-
tion). In fact, the impact of timber
harvesting on small vertebrates of
LLP/TO and SPS communities is
simply not known.

Public lands in Florida supporting
xeric pinelands include, but are not
limited to, state forests and state
parks. State forests with large acre-
ages of LLP/TO, e.g., the Withla-
coochee State Forest, are managed at
the ecosystem level. Prescribed burn-
ing is done every 3-8 years and fu-
ture timber sales will follow a rota-
tion length of 80-120 years; currently
rotation lengths are about 60 years
and are not regarded as favorably for
endemic wildlife. Wildlife manage-
ment areas overlap the state forest
holdings and are managed for sus-
tained yields of wildlife by the Flor-
ida Game and Fresh Water Fish
Commission based on a memoran-
dum of understanding between
agencies (Cathy Ryan, personal com-
munication).

Table 3.— Endangered and potentially endangered amphibians, reptiles,
and small mammals (Wood 1 987) inhabiting xeric pinelands of peninsular
Florida.

Species group

Xeric pineland Designated status'

LLP/TO SPS FGFWFC2 USFWS^

Amphibians and Reptiles

Drymarchon cords couperi

Eumeces egregius lividus

Gopherus polyphemus

Neosepsreynoldsi

Pifuophis melanoieucus mugifus

Rona areolafa

Sfilosoma exfenuafum
Mammals

Geomys pinefis goffi

Podomys floridanus

Sciurus niger shermani

X
X
X
X
X
X
X

X
X
X

X
X
X
X
X
X
X

X

T

T

I

ssc

T

SSC
SSC

T

I

E

SSC
SSC

T
T

UR2

T
UR2
UR2
UR2

UR3
UR2
UR2

'f= Endongeredj T=Jhreatened; SSC= Species of Special Concern: UR2= Under re-
view for listing, but substantial evidence of biological vulnerability and/or threat is
lacking: UR3 = Still formally under review for listing, but no longer being considered for
listing due to existing pervasive evidence of extinction.

^Florida Game and Fresh Water Fish Commission

^United States Fish and Wildlife Service

State parks are managed by the
Division of Recreation and Parks of
the Rorida Department of Natural
Resources (FDNR). An ecosystem
approach is taken in the restoration
and management of xeric pinelands
on state park lands (Jim Stevenson,
personal communication). Prescribed
burning has been used since 1969 to
control hardwood invasion of LLP/
TO stands and to stimulate growth
and flowering of grasses and herbs.
Burning in spring and early summer
appears to best duplicate the historic
timing of lightning initiated fires in
xeric pinelands. The impact of these
management practices on the plant
community has been documented
(Davis 1984); the response of reptiles,
amphibians, and small mammals is
currently under study (Stout et al.
unpublished). Generally, mature
stands of SPS have not been burned
until recently, due to the unpredict-
able behavior of fire in the commu-
nity; however, a prescription for
burning this fuel type has been writ-
ten and tested on private land and
state parks (Doran et al. 1987). Early
recovery stages of SPS appear to sup-
port the greatest diversity of reptiles
and amphibians. However, as can-
opy closure occurs in SPS, ground
cover diminishes and habitat quality
for gopher tortoises declines (Cox et
al. 1987). In contrast, similar numbers
of Podomys have been observed in
early (R. E. Roberts, unpublished
data, J. Dickinson State Park); inter-
mediate (Stout 1982); and old growth
SPS (James N. Layne, unpublished
data, Archbold Biological Station).

State parks, reserves, and pre-
serves appear to be ideal lands to ex-
plore species-specific management
measures for herp tiles and small
mammals. For example, sand swim-
ming herptiles (Smith 1982) require
openings that are relatively root free
in LLP/TO and SPS habitats. The
natural occurrence of such openings
may have been due to "hot" spots
associated with the combustion of
high fuel loads, e.g., fallen trees (Ron
Myers, personal communication).

102

Concentrarion of natural fuels prior
to prescription bums in SPS would
offer a means to create microhabitat
conditions favorable for the sand
swimmers.

MANAGEMENT OF XERIC
PINELAND ON PRIVATE LAND

Development of Regional Impact

Concern with management of am-
phibians, reptiles, and small mam-
mals on private lands in Rorida de-
rives from state and federal protec-
tion of endangered species and the
development guidelines promul-
gated during the Development of
Regional Impact (DRI) process. "The
Rorida Environmental Land and
Water Management Act of 1972"
(Chapter 380, Florida Statutes) de-
fines developments of regional im-
pact in Section 380.06(1), Rorida Stat-
utes, as "...any development which,
because of its character, magnitude,
or location, would have a substantial
effect upon the health, safety, or wel-
fare of citizens of more than one
county (Anonymous 1976)." Large
scale development projects in penin-
sular Florida commonly involve hun-
dreds to several thousand acres of
relatively natural landscape. The DRI
process requires bona fide studies of
wildlife populations and their associ-
ated habitats; emphasis is placed on
listed species. Developers must pre-
pare viable management strategies to
accommodate wildlife resources de-
pendent upon their lands (Cox et al.
1987; Richardson et al. 1986).

Management strategies of devel-
opers with xeric pinelands generally
follow one of two somewhat overlap-
ping approaches to preserve habitat
and/or species values: (1) conserva-
tion set asides or (2) mitigation. Con-
servation set asides are, in principle,
the preferred solution. In practice
some habitat is dedicated in perpetu-
ity as a nature preserve; preserve de-
sign currently is a somewhat ad hoc
process and will be discussed more

completely in a subsequent section of
this paper. Very high land values
may dictate mitigation rather than on
site preservation of habitat.

Mitigation may take many forms
to compensate for development of
xeric pinelands. Restoration of de-
graded land (Humphrey et al. 1985),
not necessarily xeric pinelands, is one
method. Another tactic is to purchase
comparable land or some other type
of land of equivalent natural value
elsewhere and dedicate it to preser-
vation. A formal process for accom-
plishing this option is presently un-
der study by the Florida Game and
Fresh Water Fish Commission.

Preservation of habitat is the basic
purpose of conservation set asides
and mitigations. The value of these
efforts depends on the proximity to
larger, undeveloped tracts of land,
travel corridors, area of preserves,
and future management options.

Another form of mitigation is the
relocation of sensitive species from
tracts of land to be developed to land
dedicated to purposes that are con-
sistent with the long-term survival of
the relocated species. In Rorida, the
gopher tortoise has been the focus of
numerous relocation efforts. Diemer
(1984) discussed the advantages and
disadvantages of relocation of go-
pher tortoises as a species manage-
ment strategy. Formal research on
gopher tortoise relocation was re-
cently reported (Proced. Gopher Tor-
toise Relocation Symp., 27 June 1987,
Gainesville, FL, in press). The Rorida
Game and Fresh Water Fish Commis-
sion regulates relocations by a permit
system based on a standardized relo-
cation protocol.

Preserve Design

Preserve design is an evolving and
controversial area of conservation
biology (Diamond 1975, 1978; Gilbert
1980; Higgs 1981; Margules 1982;
Pickett and Thompson 1978; Pyle
1980; Soule and Simberioff 1986).
Large preserves encompassing a mo-

saic of xeric pinelands, mesic forests,
and seasonal and f>ermanent wet-
lands would perhaps offer the ideal
landscape unit for long-term preser-
vation of amphibians, reptiles, and
small mammals in peninsular Ror-
ida. Because preserves on private
lands must be justified and dedicated
through the DRI process, economics
dictates preserve units of minimal
size. Rarely do we have the opportu-
nity to cluster or juxtapose these
small units to take advantage of the
so called "rescue effect" (Brown and
Kodric-Brown 1977).

In practice, conservation set asides
tend not only to be small in acreage
but also only of one habitat type. The
latter presents a dilemma for species
whose requirements often include
two or more contrasting habitats. For
example, the gopher frog lives in tor-
toise burrows in LLP/TO sandhills
during late spring, summer and early
fall and migrates to temporary wet
season depressions to breed in win-
ter and early spring (Moler and
Franz 1987). Thus a mosaic of up-
land-wetland habitats in close prox-
imity are essential to maintain viable
populations of this species. Other
species such as the indigo snake have
home range requirements that in-
clude 122-202 ha of several upland-
wetland habitat types (Moler 1985;
Moler unpublished data). It is obvi-
ous that large landscape units are
necessary to preserve viable popula-
tions of these animals.

We have prepared a detailed pre-
serve design for a SPS community
within the city of Boca Raton, Rorida
(Richardson et al. 1986; Stout et al.
1987; manuscript in preparation).
The approach taken anticipated
Soule and Simberioff (1986) and rec-
ommended the area of the preserve
be sufficient to support a minimum
viable population (Franklin 1980) of
gopher tortoises because of their
status as the keystone species. Al-
though biologically reasonable, this
basis for determining preserve size is
often economically unrealistic from
the view point of the private land-

103

owner. A consortium of public land-
owners would, however, permit the
purchase and long-term management
of the preserve as recommended.

Cox et al. (1987) offer guidelines
for the design of preserves on private
lands to maintain gopher tortoise
populations. They employed the
computer simulation model
POPDYN (Perez-Trejo and Samson
manuscript) to determine population
viability based on different initial
sizes. Populations of 40-50 individu-
als were found to be likely (>90%) to
persist 200 years. Based on existing
literature on home range require-
ments. Cox et al. (1987) recom-
mended a minimum preserve of 10-

20 ha, depending on habitat quality,
to support 40-50 tortoises.

Another approach to determining
the area of a preserve employs ''inci-
dence functions" (Diamond 1978).
Incidence functions are species spe-
cific and derived from data sets
which reveal the fraction of plots
(discrete habitats) of different areas
that actually support the species. It is
a matter of judgement as to the
probability of occurrence, e.g. 0.5 as
opposed to 0.7, that would set a
lower limit to area for an acceptable
preserve. Data sets useful for evalu-
ating this approach with respect to
amphibians, reptiles, and small
mammals in xeric pinelands are pres-

Table 4.— Incidence of Gopherus polyphemus a keystone species, and
Podomys fioridanus in xeric pinelands of peninsular Florida. Presence (+) or
absence <-) Is indicated. Study sites are ranked according to area within
the xeric pinelands. Quantitative sampling of the 12 LLP/TO study sites con-
sisted of 5 days of live-trapping and observation at invervals of 3 months
over a period of 18 months (1986-1988). Study sites In SPS were sampled by
live -trapping and observation a minimum of 3 consecutive days, often in
the same season of consecutive years (Stout et al. unpublished).

Incidence of species in xeric pineland

Study sites

Area

LLP/TO

SPS

(ha)

Gopherus Podomys Gopherus Podor

Lake Mary

1.2

+ —

Morningside Nature Center

2.0

+ —

San Felasco

4.1

+ —

Spruce Creek

4.1

+ —

Orange City

5.6

+ —

Bok Tower

9.3

+ —

Wekiwa Springs

9.7

+ +

Suwannee River

10.1

+ +

O'Lena

10.5

+ +

J. Butterfield Brooks

15.8

+ +

Starkey Well Field

16.2

+ —

Sandhill Boy Scout Camp

16.2

+ —

Interlachen

21.8

+ —

Yamato Plaza

2.8

+ —

Yamato Scrub, B

3.2

+ +

Quantum Park, A

4.4

+ +

Quantum Park, B

4.4

+ +

Quantum Park, C

4.8

+ +

Yamato Scrub, A

8.5

+ +

Summit Place

10.5

+ —

Potomac Road

17.8

+ +

Cedar Grove

21.5

+ +

J. Dickinson

256.2

+ +

ently lacking. Table 4 provides data
we have gathered on area of discrete
habitats and the presence or absence
of gopher tortoises and Florida mice.
It is apparent that tortoises are less
area sensitive than Florida mice and
that Florida mice are patchy in occur-
rence in LLP /TO, perhaps only sec-
ondarily related to area.

Incidence functions do not neces-
sarily reveal the minimum area re-
quired to support minimum viable
populations (Franklin 1980). We be-
lieve preserve area should be based
on providing this requirement, par-
ticularly when preserves are isolated
relative to average dispersal dis-
tances of keystone species. However,
clusters of preserves within dispersal
distances of keystone species may be
of less area per preserve due to a
high likelihood of reinvasion from
nearby populations following local
population extirpations (Noss and
Harris 1986).

Monagennent of Preserves in Xeric
Pinelands

The future viability of preserves de-
pends largely on their ownership af-
ter development of the surrounding
landscape. It is unlikely that home-
owners associations will assume the
cost of management if preserves re-
main as a part of the overall develop-
ment's "commons." Public owner-
ship is an alternative and might rest
with a city, county, or state. Local
governments seem more appropriate;
however, funds and expertise to
manage may be lacking. One pre-
serve in south Rorida is designed to
border a city park, thus allowing its
maintenance and /or management
costs to be assumed over time as part
of the existing park system
(Richardson, personal observation).
Regardless of the ownership, a com-
mitment to long-term management
must be achieved if a preserve is to
retain natural values.

Management options for nature
preserves range from a decision 1) to

104

do nothing and let nature take its
course; 2) to manage for maintenance
of a viable ecosystem, which implies
the natural biota, including amphibi-
ans, reptiles, and small mammals,
will be present in proportion to their
normal abundance; or 3) to focus
management on the needs of one or
more species. White and Bratton
(1980) have exposed the folly of the
first management option. The deci-
sion to emphasize ecosystem or spe-
cies management depends on the en-
tity responsible for management,
type of preserve, management objec-
tives, area of the preserve, nature of
the surrounding lands, relative over-
all or regional rarity of particular
species, and the resources available
for management.

Management objectives of any
preserve should focus on: 1) mainte-
nance of normal ecosystem proc-
esses; 2) conservation of soil; 3)
maintenance or restoration of normal
hydrologic conditions; 4) prevention
of establishment of exotic species.; 5)
and prevention of human encroach-
ment (e.g., dumping, ATVs, etc.) Be-
yond these generalities, management
of preserves is an idiosyncratic proc-
ess that may concern endemic spe-
cies, genetics of inbred populations,
or restoration of periodic wild fires.

Xeric pinelands of peninsular Flor-
ida depend on periodic fires to main-
tain their structure and function
(Laessle 1958a; Abrahamson 1984).
Thus a burning program is essential
in the management of LLP /TO or
SPS preserves. Spring or early sum-
mer prescribed bums are routinely
used to maintain LLP/ TO communi-
ties on state parks. Doran et al. (1987)
have documented prescribed burns
of SPS preserves in an urban setting
based on rather esoteric fire models
developed by the U.S. Forest Service.
Gopher tortoises respond favorably
to the bums (Stout et al. 1988). A mo-
saic of recovery stages in SPS may
favor beta diversity of herptiles and
small mammals. Mushinsky (1985)
has carefully documented the re-
sponse of the herpetofauna to a vari-

ety of buming schedules in LLP/TO.
Diversity and abundance of amphibi-
ans and reptiles was increased on
experimental plots relative to un-
bumed controls. Re-establishment of
the pine overstory may be necessary
to produce needle cast for carrying
fire (Landers and Speake 1980).

Management of conservation set
asides and /or easements may focus
on particular species or combinations
of species. The smaller the preserve
the more likely that a reduced suite
of species will be present
(Richardson et al. 1986). Given that a
fixed area is available for manage-
ment, major efforts to enhance or
maintain habitat should target those
species that can maintain viable
populations within the preserve
(Shaffer 1986). A species whose mini-
mum area requirements for a mini-
mum viable population exceeds the
preserve area should not be of major
concern (Shaffer and Samson 1985);
nonetheless, such species can benefit
from the preserves if travel corridors
exist (Harris 1984).

DISCUSSION

Xeric pinelands of peninsular Florida
support a species-rich assemblage of
reptiles, amphibians, and small
mammals. (Growth and development
continues to diminish LLP/ TO and
SPS habitats to the detriment of the
associated biota. Land in public own-
ership, e.g. state parks and forests,
national forests, and private hold-
ings, e.g., the Archbold Biological
Station, and institutional lands such
as the Ordway and Swisher Pre-
serves, jointly owned and managed
by the University of Florida and The
Nature Conservancy, will be increas-
ingly valuable as other xeric pine-
lands are converted to land uses not
favorable to the biota. Thus, manage-
ment of these xeric pinelands will
become more important in the fu-
ture. At present management is
largely limited to prescribed bums to
maintain what were historically fire

climax communities. Thus, fire man-
agement is tantamount to small ver-
tebrate management.

In the future as air quality stan-
dards are modified, prescribed bum-
ing, particularly in or near urbanized
areas, will be restricted or eliminated
as a management option. Alternative
means of habitat manipulation need
to be developed, particularly for SPS.

Basic information on the life his-
tory of many amphibians, reptiles,
and small mammals of xeric pine-
lands is lacking. The Nongame Wild-
life Program of the Rorida Game and
Fresh Water Fish Commission has
initiated and funded rather large
scale studies of SPS and LLP/TO
communities. These studies are at the
community level and largely obser-
vational. Management needs of indi-
vidual species may be derived only
secondarily from this research. Stud-
ies that focus on particular species
will ultimately lead to more refined
habitat management guidelines. The
report by Cox et al. (1987) will likely
serve as a model for the preparation
of habitat protection guidelines; man-
agement follows protection (White
and Bratton 1980).

Management alternatives at the
ecosystem and species level are
needed now for xeric pinelands on
private lands undergoing develop-
ment. Regulation of development in
these habitats as currently practiced
will result in a patchwork of small,
isolated nature preserves. Preserva-
tion of natural habitat in a developed
landscape is, of course, desirable.
However, several problems remain:
(1) who will own the preserves, (2)
how will a management plan be pre-
pared, and (3) who will be respon-
sible for management? Even another
decade of rapid growth in peninsular
Florida may result in a few hundred
nature preserves, which will not nec-
essarily be restricted to xeric pine-
land habitat. Ignoring the question of
ownership, no public land manage-
ment agency is currently capable of
assuming the charge of managing
these preserves. Lack of manage-

105

ment, e.g., failure to conduct pre-
scribed burning, will allow succes-
sional changes to occur to the detri-
ment of many small vertebrates nar-
rowly adapted to xeric pinelands.
Loss of habitat and species values
originally used by jurisdictional
agencies to secure preserve set asides
provides a potential basis for private
land owners to request development
rights on the land. This action would
defeat the entire purpose of having
conservation set asides.

An alternative to on site habitat
protection is offered by Cox et al.
(1987) in regard to preserving habitat
for the gopher tortoise. The alterna-
tive, a Wildlife Resource Mitigation
Fund (WRMF), allows a developer to
contribute money to the fund to miti-
gate losses of valuable wildlife habi-
tat on lands being developed. The
collective monies of several develop-
ment projects would allow an inde-
pendent group such as the Trust For
Public Lands to assist in the purchase
of commensurate lands to expand an
existing public park, preserve or for-
est. Management is more likely to be
applied to these lands and ultimately
the resources are better served by the
public agencies.

ACKNOWLEDGMENTS

We thank the authors of the papers
cited herein for their efforts and
dedication to science. Biologists who
contributed to our knowledge of
xeric pineland include but are not
limited to the following individuals:
Dan Austin, Don Bethancourt, Russ
Burke, Steve Christman, David Cook,
David Corey, Jim Cox, Joan Diemer,
Dick Franz, Larry Harris, Randy
Kautz, Jim Layne, Wayne Marion,
Paul Moler, Ron Myers, Reed Noss,
Cathy Ryan, Jim Stevenson, and Don
Wood. Support for research on the
ecology of sandhill communities was
provided by the Nongame Wildlife
Program, RFP86-003, of the Rorida
Game and Fresh Water Fish Commis-
sion. Development groups that

funded work by the authors on xeric
pinelands include Hardy-Lieb Devel-
opment Corporation, The Adler
Group, and Deutsch-Ireland Proper-
ties. The Division of Recreation and
Parks (FDNR), Department of Biol-
ogy, University of South Florida, and
the Department of Biological Sci-
ences, University of Central Florida
assisted in our studies in a variety of
ways. We thank Beverly Bamekow,
Rita Greenwell, Barbara Erwin, and
Nancy Small for typing the manu-
script. Lastly, we thank Paul E.
Moler, James N. Layne and Robert C.
Szaro for providing excellent sugges-
tions to improve the paper.

LITERATURE CITED

Abrahamson, Warren G. 1984. Post-
fire recovery of Florida Lake
Wales ridge vegetation. American
Journal of Botany 71:9-21.

Abrahamson, Warren G., Ann F.
Johnson, James N. Layne, and Pa-
tricia A. Peroni. 1984. Vegetation
of the Archbold Biological Station,
Florida: an example of the South-
ern Lake Wales Ridge. Florida Sci-
entist 47:209-250.

Anonymous. 1976. Developments of
regional impact: guidebook for
preparation of the application for
development approval. 65 p. Divi-
sion of State Planning, Department
of Administration, Tallahassee,
Florida.

Arata, Andrew A. 1959. Effects of
burning on vegetation and rodent
populations in a longleaf pine tur-
key oak association in north Cen-
tral Florida. Quarterly Journal
Florida Academy Sciences 22:94-
104.

Auffenberg, Walter, and Richard
Franz. 1982. The status and distri-
bution of the gopher tortoise (Go-
pherus -polyphemus). p. 95-126. In
North American tortoises: conser-
vation and ecology. R. Bruce Bury,
editor. Wildlife Research Report
12. 126 p. U.S. Fish and Wildlife
Service, Washington, D.C.

Beckenbach, J. R., and J. W. Ham-
mett. 1962. General soil map of
Florida. 1 p. Florida Agricultural
Experiment Stations and the Soil
Conservation Service.

Brown, James H., and Astrid Kodric-
Brown. 1977. Turnover rates in
insular biogeography: effects of
immigration on extinction. Ecol-
ogy 58:445-449.

Campbell, Howard W., and Steven P.
Christman. 1982. The herpetologi-
cal components of Horida sandhill
and sand pine scrub associations,
p. 163-171. In Herpetological com-
munities: a symposium of the So-
ciety for the Study of Amphibians
and Reptiles and the Herpetolo-
gists' League, August 1977. Nor-
man J. Scott, Jr., editor. 239 p. U.S.
Fish and Wildlife Service, Wildlife
Research Report 13.

Christman, Steven P., Howard I.
Kochman, Howard W. Campbell,
Charles R. Smith, and Wallace S.
Lippincott, Jr. Successional
changes in community structures:
amphibians and reptiles in Horida
sand pine scrub. Unpublished
manuscript.

Cox, James, Douglas Inkley, and
Randy Kautz, 1987. Ecology and
habitat protection needs of gopher
tortoise (Gopherus polyphemus)
populations found on lands slated
for large-scale development in
Horida. Horida Game and Fresh
Water Fish Commission Nongame
Wildlife Program Technical Report
no. 4. 75 p. Tallahassee, FL.

Davis, John H. 1980. General map of
natural vegetation of Florida.
Agric. Exp. Sta. Institute of Food
and Agricultural Sciences Circular
S-178. 1 p. University of Horida,
Gainesville.

Davis, Mary M. 1984. Phenology of
flowering plants on experimental
fire plots, St. Marks National
Wildlife Refuge. M. S. Thesis. 54 p.
Horida State University. Tallahas-
see, FL.

Diamond, Jared M. 1975. The island
dilemma: lessons of modem bio-
geographic studies for the design

106

of natural reserves. Biological
Conservation 7:129-146.

Diamond, J. M. 1978. Critical areas
for maintaining viable populations
of species, p. 27-40. In The break-
down and restoration of ecosys-
tems. M. W. Holdgate and M. J.
Woodman, editors. Plenum Press,
New York, N. Y.

Diemer, Joan E. 1984. Tortoise reloca-
tion in Florida: solution or prob-
lem? Proc. 1984 Symposium Des-
ert Tortoise Council: 131-135.

Doran, R. F., Donald R. Richardson,
and Richard E. Roberts. 1987. Pre-
scribed burning of the sand pine
scrub community: Yamato Scrub,
a test case. Florida Scientist 50:184-
192.

Eisenberg, J. F. 1983. The gopher tor-
toise as a keystone species, p. 1-4.
In The gopher tortoise: a keystone
species. R. J. Bryant and R. Franz,
editors. Proceedings of the 4th
Annual Meeting of the Gopher
Tortoise Council. Rorida State
Museum, Gainesville, FL.

Franklin, Ian Robert. 1980. Evolution-
ary change in small populations,
p. 135-149. In Conservation biol-
ogy: an evolutionary-ecological
perspective. Michael E. Soule and
Bruce A. Wilcox, editors. 395 p.
Sinauer Assoc., Sunderland, MA.

Gilbert, F. S. 1980. The equilibrium
theory of island biogeography:
fact or fiction? Journal of Biogeo-
graphy 7:209-235.

Givens, Kerry T., James N. Layne,
Warren G. Abrahamson, and
Susan C. White-Schuler. 1984.
Structural changes and succes-
sional relationships of five Florida
Lake Wales Ridge plant communi-
ties. Bulletin Torrey Botanical
Club 111:8-18.

Harris, Larry D. 1984. The frag-
mented forest. 211 p. The Univer-
sity of Chicago Press, Chicago, IL.

Higgs, A. J. 1981. Island biogeogra-
phy theory and nature reserve de-
sign. Journal of Biogeography
8:117-124.

Humphrey, Stephen R., John F. Eis-
enberg, and Richard Franz. 1985.

Possibilities for restoring wildlife
of a longleaf pine savanna in an
abandoned citrus grove. Wildlife
Society Bulletin 13:487-496.

Kalisz, Paul J., and Earl L. Stone.
1984. The longleaf pine islands of
the Ocala National Forest, Florida:
a soil study. Ecology 65:1743-1754.

Kurz, Herman. 1942. Florida dunes
and scrub vegetation and geology.
Florida Geological Survey Bulletin
23:15-154.

Laessle, Albert M. 1942. The plant
communities of the Welaka Area.
University of Florida Publication,
Biological Science Series 4:1-143.

Laessle, Albert M. 1958a. The origin
and successional relationship of
sandhill vegetation and sand pine
scrub. Ecological Monographs
28:361-387.

Laessle, Albert M. 1958b. A report on
succession studies of selected
plant communities on the Univ. of
Fla. conservation reserve, Welaka,
Florida. Quarterly Journal Florida
Academy Science 21:101-112.

Laessle, A. M. 1967. Relationship of
sand pine scrub to former shore
lines. Quarterly Journal Florida
Academy Science 30:269-286.

Landers, J. Larry, and Dan W.
Speake. 1980. Management needs
of sandhill reptiles in Southern
Georgia. Annual conference
Southeastern Association Fish and
Wildlife Agencies. 34:515-529.

Latham, Pamela J. 1985. Structural
comparisons of sand pine scrubs
of east-central Florida. M. S. The-
sis. 106 p. University of Central
Florida, Orlando.

Layne, James N. 1969. Nest-building
behavior in three species of deer
mice, Peromyscus. Behaviour
35:288-303.

Layne, James N. 1978. Peromyscus flo-
ridanus (Chapman), p. 21-22. In
Rare and endangered biota of
FLorida. Vol. 1: mammals. James
N. Layne, editor. 52 p. University
Presses of Florida, Gainesville, Fl.

Margules, C, A. J. Higgs, and R. W.
Rafe. 1982. Modern biogeographic
theory: are there any lessons for

nature reserve design? Biological
Conservation 24:115-128.

Moler, Paul E. 1985. Distribution of
the eastern indigo snake, Dry-
marchon corais couperi, in Rorida.
Herpetological Review 116:37-38.

Moler, Paul E., and Richard Franz.
1987. Wildlife values of small, iso-
lated wetlands in the southeastern
coastal plain, p. 234-241. In Proc.
3rd S. E. Nongame and Endan-
gered Wildlife Symposium. R. R.
Odom, K. A. Riddleberger and J.
C. Ozier, editors. Georgia Depart-
ment Natural Resources, Game
and Fish Division.

Monk, Carl D. 1968. Successional and
environmental relationships of the
forest vegetation of North Central
Rorida. The American Midland
Naturalist 79:441-457.

Mushinsky, Henry R. 1984. Fire and
the Rorida sandhill herpetofaunal
community: with special attention
to responses of Cnemidophorus
sexlineatus. Herpetologica 41:333-
342.

Myers, Ronald L. 1985. Fire and the
dynamic relationship between
Rorida sandhill and sand pine
scrub vegetation. Bulletin Torrey
Botanical Club 112:241-252.

Norse, Elliott A., Kenneth L. Rosen-
baum, David S. Wilcove, Bruce A.
Wilcox, William H. Romme,
David W. Johnston, and Martha L.
Stout. 1986. Conserving biological
diversity in our national forests.
116 p. The Wilderness Society.

Noss, Reed F., and Larry D. Harris.
1986. Nodes, networks, and
MUMS: preserving diversity at all
scales. Environmental Manage-
ment 10:299-309.

Pickett, S. T. A., and John N. Th-
ompson. 1978. Patch dynamics
and the design of nature reserves.
Biological Conservation 13:27-37.

Pyle, Robert Michael. 1980. Manage-
ment of nature reserves, p. 319-
327. In Conservation Biology: An
evolutionary-ecological perspec-
tive. Michael E. Soule and Bruce
A. Wilcox, editors. 395 p. Sinauer
Associates, Sunderland, Mass.

107

Richardson, Donald Robert. 1977.
Vegetation of the Atlantic coastal
ridge of Palm Beach County, Flor-
ida. Florida Scientist 40:281-330.

Richardson, Donald R., I. Jack Stout,
Richard E. Roberts, Daniel F.
Austin, and Taylor R. Alexander.
1986. Design and management
recommendations for a sand pine
scrub preserve: the Yamato scrub.
124 p. Ecological Consultants,
Tampa, Florida.

Shaffer, Mark L. 1986. The met-
apopulation and species conserva-
tion: the special case of the north-
ern spotted owl. p. 86-99. In Ecol-
ogy and management of the spot-
ted owl in the Pacific Northwest,
Ralph J. Gutierrez, and Andrew B.
Carey, tech editors. USDA Forest
Service. General Technical Report
PNW-185, 119 p. Pacific North-
west Forest and Range Experiment
Station, Portland, Oregon.

Shaffer, Mark L., and F. B. Samson.
1985. Population size and extinc-
tion: a note on determining critical
population sizes. American Natu-
ralist 125:144-152.

Smith, Charles R. 1982. Food re-
source partitioning of fossorial
Rorida reptiles, p. 173-178. In Her-
petological communities: a sympo-
sium of the Society for the Study
of Amphibians and Reptiles and
the Herpetologists' League, Au-
gust 1977. Norman J. Scott, Jr.,
editor. 239 p. U.S. Fish and Wild-
life Service, Wildlife Research Re-
port 13.

Soule, Michael E., and Daniel Sim-
berloff. 1986. What do genetics
and ecology tell us about the de-
sign of nature reserves? Biological
Conservation 35:19-40.

Stout, I. Jack. 1981. Conservation bi-
ology of Peromycus floridanus. Flor-
ida Scientist 44 (Supplement 1):37.

Stout, I. Jack. 1982. Small mammal
community structure in sand pine
scrub. Bulletin Ecological Society
America 63:68-69.

Stout, I. Jack, Donald R. Richardson,
and Richard E. Roberts. 1988. Re-
sponse of resident and relocated

gopher tortoises to a prescribed
bum in a sand pine scrub commu-
nity. Proced. Gopher Tortoise Re-
location Symp., 27 June 1987,
Gainesville, FL.

Stout, I. Jack, Donald R. Richardson,
Richard E. Roberts, and Daniel F.
Austin. 1987. Design of a nature
preserve in a subtropical urbaniz-
ing landscape: application of ecol-
ogic and genetic principles. Bulle-
tin Ecological Society of America
68:423-424.

Swindell, C. E. 1987. Multivariate
analysis of competition and habi-
tat use by small mammals in Flor-
ida Sand Pine Scrub. M. S. Thesis.
59 p. University of Central Florida,
Orlando.

Veno, Patricia A. 1976. Successional
relationships of five Florida plant
communities. Ecology 57:498-508.

White, Peter S., and Susan P. Bratton.
1980. After preservation: philo-
sophical and practical problems of
change. Biological Conservation
18:241-255.

White, W. A. 1970. The geomorphol-
ogy of the Florida peninsula. Bu-
reau of Geology, Tallahassee. 164
p. Geological Bulletin No. 51.

Wood, Don A. 1987. Official lists of
endangered and potentially en-
dangered fauna and flora in Flor-
ida. 19 p. Florida Game and Fresh
Water Fish Commission.

108

Distribution and Habitat
Associations of Herpetofauna
in Arizona: Comparisons by
Habitat Type^

Abstract. —Between 1977 and 1981 , the Bureau of
Land Management conducted extensive surveys of
Arizona's herpetofauna in 16 different habitat types
on approximately 8.5 million acres of public lands.
This paper describes results of one of the most exten-
sive surveys ever conducted on amphibian and rep-
tile communities in North America.

K. Bruce Jones^

With the passage of the Federal Land
PoUcy and Management Act in 1976,
the Bureau of Land Management
(ELM) was mandated to keep an in-
ventory of resources on public lands.
Information collected during inven-
tories or surveys was then to be used
to identify issues for land use plan-
ning and opportunities for land man-
agement. The BLM made a decision
to collect data on all major wildlife
groups and their habitats

Early in the development of its in-
ventory program, the BLM recog-
nized a need to devise a strategy that
would compare animal distributions
and abundance to habitats. This
strategy was important since the
BLM manages wildlife habitats and
not wildlife populations.

In 1977 the BLM initiated invento-
ries of wildlife resources on public
lands. At that time, considerable in-
formation was already available on
game species. However, data on
nongame species were mostly lack-
ing. As a result, priority was given to
collecting data on nongame species
and their habitats.

Amphibians and reptiles are im-
portant members of the nongame
fauna. They use a wide range of habi-

' Paper presented at symposium. Man-
agement of Amphibians. Reptiles, and
Small Mammals in North America. (Flag-
staff , Arizona, July 19-21. 1988).

^K. Bruce Jones is a Research Ecologist
with the Environmental Protection Agency.
Environmental Monitoring Systems Labora-
tory. Las Vegas. Nevada 89193.

tats and are often good indicators of
habitat conditions (Jones 1981a).
Therefore, in order to obtain infor-
mation on these animals, principally
for land-use planning, the BLM con-
ducted extensive inventories of am-
phibians and reptiles by habitat type.
This inventory included a scheme
whereby associations between am-
phibians and reptiles and certain n\i-
crohabitats could be determined. The
inventory, conducted between 1977
and 1981, was one of the most com-
prehensive surveys of herpetological
communities ever conducted in
North America (27,885 array-nights
in 16 habitat types over a five-year
period). It also represents the first
large-scale effort to quantitatively
compare herpetofaunas associated
with ecosystems. This paper reports
the results of these surveys, includ-
ing species distributions and associa-
tions with microhabitats and habitat
types (plant communities).

STUDY AREA

The study area consisted of approxi-
mately 3,441,296 ha (8.5 million
acres) of public lands located in cen-
tral, west-central, southwestern, and
northwestern Arizona (fig. 1). Sixteen
different habitat types were deline-
ated within this area, primarily from
an existing map of vegetation asso-
ciations (Brown et al. 1979). Field re-
connaissance allowed more local as-
sociations to be recognized within

Figure 1.— The study area.

those presented by Brown et al.
(1979). For example, because of the
scale of their map. Brown et al. (1979)
failed to recognize several small, rel-
ict stands of chaparral woodland,
although Brown (1978) had noted the
presence of chaparral woodland
vegetation at several small sites (see
Jones et al. 1985 for the importance of
small woodland stands to certain
herpetofauna). Therefore, the habitat
type map used to allocate samples in
this study drew upon the Brown
(1978) and Brown et al. (1979) maps.

109

and results of field reconnaissance.
For detailed descriptions of these
habitat types see Jones (1981b) and
Buse (1981).

SAMPLING METHODS

Amphibian and reptile distribution
and abundance by habitat type were
determined by on-the-ground sam-
pling efforts between October, 1977,
and July, 1981. Samples were ob-
tained by three methods. The most
extensive sampling was accom-
plished with a pit-fall trapping
method (array) consisting of a series

of 18.3 1 (5 gal) plastic containers bur-
ied in the ground and connected by
0.41 m (8 inches) high aluminum
drift fence; one trap was located in
the center with three evenly dis-
persed (120°) peripheral traps 7.14 m
(25 ft) from the center (Jones 1981a,
Jones 1986). This modified array
method was designed specifically for
sampling amphibians and reptiles in
desert habitats (see Jones 1986 for a
comparison of this procedure with
the original array trapping scheme
designed by Christman and
Campbell 1982). A total of 183 arrays
were used to sample 16 different
habitat types (see table 1 for sum-

Table 1 .—Sampling effort in each habitat type.

#of
arrays

# of trap
nights

# of road
riding road
transects

# of field
searches

Elevation
range (m (ff.))

Ponderosa Pine Woodland (PP)

5 745 10 15 1677-2531 (5500-8300)
Pinyon-Juniper Woodland (PJ)

9 945 14 20 1311-1921 (4300-6300)

Sagebrush (Great Basin Desert) (SB)

3 270 12 12 1311-1830 (4300-6000)

Closed Chaparral (CC)

18 2168 18 20 1250-2287 (4100-7500)

Open Chaparral (OC)

13 1950 22 25 762-1311 (2500-4300)

Desert Grassland (DO)

11 1155 15 14 1006-1525 (3300-5000)

Disclimax Desert Grassland (DD)

3 300 11 10 884-1311 (2900-4300)

Mixed Broadleaf Riparian (MB)

6 784 8 18 884-2287 (2900-7500)
Cottonwood-Wiilow Riparian (CW)

13 3145 23 28 549-1372 (1800-4500)

Juniper Woodland (mixed shrub) (JM)

9 1080 19 22 793-1342 (2600-4400)

Canotia Mixed Shrub (CA)

3 265 11 16 884-1189 (2900-3900)

Mesquite Bosque (floodplain woodland) (ME)

15 3025 18 22 213-915 (700-3000)
Mixed Riparian Scrub (Xeroriparian) (MR)

16 2640 23 18 229-1220 (750-4000)
Mojave Desertscrub (MD)

15 1803 25 24 610-1220 (2000-4000)

Sonoran Desertscrub (Arizona Upland) (SD)

22 3970 33 27 335-1189 (1100-3900)

Creosotebush (Lower Colorado) (CB)

22 3640 32 18 213-915 (700-3000)

mary of sampling effort in each habi-
tat typ>e). Arrays were placed so that
microhabitat variability within each
habitat type was sampled. The num-
ber of arrays used to sample habitat
types was partially influenced by the
size of habitats; generally, more ex-
tensive habitats received prop)ortion-
ally larger samples. However, certain
habitats (e.g., riparian) were known
to be great sources of diversity
within desert regions; therefore, pri-
ority was given to obtaining larger
samples within these habitats. Once
placed into the ground, arrays were
continuously open for a minimum of
60 days. Some arrays (60) were open
for 9 months. Generally, samples
were taken during the spring, sum-
mer, and fall. However, some arrays
(17) were open only during spring
months and others only in the fall
(12). The opening of new arrays at
different locations, and the closing of
other arrays, were often dictated by
BLM's predetermined resource plan-
ning schedule.

Since some amphibians and many
snakes could not be effectively
sampled by pit-fall traps, it was nec-
essary to use two other field tech-
niques. Road riding, consisting of
traveling roads from dusk to ap-
proximately 2300 h throughout de-
lineated habitat types, was used to
determine the occurrence of amphibi-
ans and medium and large snakes
(see table 1 for sampling effort within
each habitat type).

Time-constraint searches (Bury
and Raphael 1983), consisting of
walking along permanent and tem-
porary water sources (natural and
man-made) at night, were used to
verify the presence of frogs and
toads at waters within habitat types
(see table 1 for sampling effort within
each habitat type).

Finally, to get an idea of the
known distribution of amphibians
and reptiles within the study area, I
obtained records from 7 museums
known for their outstanding collec-
tions of amphibians and reptiles from
the Southwest: the University of

110

Michigan, Arizona State University,
the University of New Mexico,
Northern Arizona University, the
University of Arizona, the Los Ange-
les County Museum, and the Univer-
sity of California at Berkeley. In addi-
tion, these data were used to com-
pare the past distribution of amphibi-
ans and reptiles within the study
area with that obtained during the
BLM's inventories.

Microhabitat data were collected
on each array site and along roads by
a modified point-intercept method
consisting of 100 sample points sepa-
rated by 8 m (26 ft) along a randomly
determined compass line; on array
sites, the center of the line crossed
over the array. At each point, the fol-
lowing measurements were taken: (1)
vertical distribution of vegetation be-
tween 0-0.6 m (0-2 ft), 0.6-1.7 m (2-6
ft), 1.7-6.0 m (6-20 ft), and > 6 m (20
ft) (each time vegetation occurred in
a height class above the point, a con-
tact or "hit" was recorded); (2) pene-
tration to the nearest cm into the soil
by a pointed metal rod (1 cm in di-
ameter); (3) depth of leaf litter (if
present); (4) depth of other litter such
as debris heaps (piles of logs, leaves
and other dead vegetative material)
and rotting logs; (5) characterization
of surface rock into size classes of
sand, gravel (< 1 cm or 0.4 inches in
diameter), cobble (1 to 5 cm or 0.4 to
2 inches in diameter), stone (> 5 cm
or 2 inches in diameter), and bed-
rock. Vegetation cover and percent-
age of the surface occupied by each
rock and litter size class was deter-
mined by comparing the number of
"hits" in each category (e.g., litter)
with the total number of sample
points (100). Plant species were also
recorded along each 100 point
transect (see table 1 for the number
of microhabitat samples taken in
each habitat type).

DATA ANALYSIS

I calculated relative abundance of
each amphibian and reptile species as

the total number of any species
caught during a 24-hour period (ar-
ray-night). Relative abundance was
determined for each species on array
sites by taking the greatest number of
individuals of a species trapped dur-
ing a 30-day period and dividing by
the number of days. This calculation
was used because of monthly differ-
ences in species' activity patterns.
The number of arrays in which a spe-
cies was trapped in each habitat type
also was compiled to determine how
widespread a species was within in-
dividual habitat types.

A principal components analysis
(Pimental 1979) was performed to
compress microhabitat data into a
smaller, depictable subset. Mean fac-
tor scores of compressed microhabi-
tat data were computed for each
habitat type and plotted on a 3 vector
(axis) graph. Similarly, mean factor
scores of compressed microhabitat
data were computed for each am-
phibian and reptile species (turtles
were excluded because aquatic mi-
crohabitats were not measured).
These scores were calculated for each
species by averaging mean factor
scores for microhabitats on which a
species occurred.

Species richness (total number of
species) and species diversity were
calculated for each habitat type. Two
calculations of species richness for
habitats were used; one that used
only array data and one that used all
data (array, road-riding, and field-
search data). In addition, the average
number of species collected per array
(30-day period) was calculated and
compared to overall, array-deter-
mined, species richness. Species di-
versity of each habitat was deter-
mined from a Shannon-Weaver di-
versity index (Hair 1980): H' = E p,
logjQ Pi/ where s = the number of spe-
cies and Pj is the proportion of the
total number of individuals consist-
ing of the i**^ species. Average species
diversity per array was calculated for
each habitat type. Because road-rid-
ing and field searches did not yield
estimates of relative abundance simi-

lar to arrays, only array data were
used to calculate species diversity.

Two types of cluster analysis were
used to determine similarities among
habitat types. The first cluster analy-
sis was performed only on array
data, and it was based on euclidean
distances (Pimental 1979). Calcula-
tion of euclidean distances between
habitats were based on a combina-
tion of species' presence or absence
on a site and similarity in species'
dominance (relative abundance) be-
tween habitats. Since medium and
large snakes (> 0.5 m or 1.5 ft) are not
readily caught in pit-fall traps, their
relative abundances could not be cal-
culated accurately. To compare the
overall herpetofaunas of habitat
types, a second cluster analysis was
performed. This procedure involved
calculation of Simpson similarity co-
efficients (Pimental 1979). These coef-
ficients were then submitted to a
cluster analysis. Unlike the analysis
of array data via euclidean distances,
the use of Simpson similarity coeffi-
cients in a cluster analysis did not
consider relative dominance in calcu-
lating distances between habitats.

Several thousand site specific dis-
tributional records were obtained for
amphibians and reptiles within the
study (to 16.2 ha or 40 acre accu-
racy). These individual records were
too numerous to report here; detailed
locality records for each species are
kept at the Bureau of Land Manage-
ment's Phoenix District Office.

RESULTS

Microhabitats

A principal components analysis
(PCA) of microhabitats yielded 3
compressed habitat components
(axes), and the cumulative propor-
tion of eigenvalues was < 1 .0 with
83% of the variability accounted for
by the matrix (p < .05). This analysis
revealed large differences in the mi-
crohabitat among habitat types (fig.
2). Desert grassland, disclimax desert

111

grassland, and creosotebush habitats
had open canopies and low-height
vegetative structure, whereas
pinyon-juniper, mixed riparian
scrub, cottonwood-willow riparian,
mixed broadleaf riparian, and pon-
derosa pine had tree canopies and
large amounts of vegetative debris,
such as leaf litter and logs, on their
surfaces (fig. 2). Closed and open
chaparral habitats consisted of
shrubs with rocky surfaces, and
Sonoran Desert had a combination of
trees and shrubs and rocky surfaces
(fig. 2).

Species Distributions and
Abundances

A total of 28 species of lizards, 30
snakes, 4 turtles, 9 toads, 3 frogs, and
1 salamander were observed or
trapped during the study. Sceloporus

magister, Urosaurus ormtus, Uta
stansburiam, and Cnemidophorus tigris
were the most widely distributed
and abundant lizards throughout the
study area's habitat types (table 2).
These lizards also consistently oc-
curred on a large number of sites
within each habitat type (table 2).
Certain lizards, such as Gambelia wis-
lizeni, Phrynosoma solare, and Dip-
sosaurus dorsalis occurred only on
lower elevation (< 915 m or 3000 ft),
desert habitats, and other lizards,
such as Sceloporus undulatus, Gerrhon-
otus kingi, and Phrynosoma douglassi
occurred only on higher elevation (>
1220 m or 4000 ft) habitats (table 2).
Some species, such as Eumeces gilberti
and Cophosaurus texam, were princi-
pally found on higher elevation habi-
tats, but also inhabited cottonwood-
willow riparian habitats at lower ele-
vations (549-915 m or 1800-3000 ft)
(table 2). Certain lizards, such as

1.5
Trees

Component II

Figure 2.— Mean factor scores of microhabitats for habitat types. (Abbreviations correspond
to thiose listed for hiabitats in table 1 .)

Cnemidophorus burti and Eumeces ob-
soletus, had limited distributions
within the study area (table 2); C.
burti is principally distributed in the
Sonoran Desert and Desert Grass-
land habitats in extreme southern
Arizona and Mexico, and E. obsoletus
only occurs in the chaparral habitat
type in the extreme eastern portion
of the study area. Although re-
stricted to higher elevation and ripar-
ian habitats throughout most of the
study area, C. texana was found in
Sonoran Desert in the extreme east-
ern portion of the study area. Most
lizards occurred throughout the
study area where suitable habitat
was present and were not restricted
by geographic range.

A PCA revealed that lizards dif-
fered in their associations with cer-
tain microhabitats (fig. 3). Some of
the widely distributed species, such
as Cnemidophorus tigris and Uta
stansburiana, showed little association
with any of the principal components
(fig. 3), although the distribution of
other common species, such as Sce-
loporus magister and Urosaurus ornatus
was highly correlated with the pres-
ence of vegetation debris (fig. 3).
More than half of the lizards oc-
curred on sites with relatively open
canopies and shrubs or grasses, and
many also preferred rocky substrates
(fig. 3). Dipsosaurus dorsalis, Callisau-
rus draconoides, and Gambelia wislizeni
occurred on sites with sand
substrate. Gerrhonotus kingi and
Eumeces gilberti occurred on sites
with large amounts of vegetative de-
bris, medium to high canopies, and
rocky substrates, and Xantusia vigilis
on sites with similar substrate but
with a more open canopy (fig. 3).
Crotaphytus collaris and Sauromalus
obesus occurred on sites that were
open, rocky, and shrubby or grassy
(fig. 3).

Snakes showed similar distribu-
tional patterns to lizards. Some
snakes, such as Lampropeltis getulus,
Pituophis melanoleucus, Rhinocheilus
leconti, Crotalus atrox, and Crotalus
molossus, occurred in many habitat

112

Table 2.— Relative abundance of lizards by habitat type. Relative abundance the number of an individual species
caught in an array per 24 h period. * Indicates species verified in a habitat type via road-riding and searches. The
number below the Habitat Type in ( > = the total number of arrays. The number In ( ) to the right of the species' relative
abundance = the number of arrays in which the species was trapped.

PP PJ SB CC

OC

DG

DD

MB

cw

JM

CA

ME

MR

MD

SD CB

(5) (9) (3) (18)

(13)

(11)

(3)

(6)

(13)

(9)

(3)

(15)

(16)

(15)

(22) (22)

Gerrhonotus kingi

- - - .03(1)

.03(1)

Coleonyx vahegafus

03(1) - -

.03(1)

.03(1)

.01(1)

.05(2)

.01(5)

.03(8)

.02(6)

.04(11) .06(11)

Heloderma suspectum

♦

»

.03(1)

♦

♦

.03(1)

.03(1)

.03(1)

.03(2)

Callisaurus draconoides

- - - .06(1) - - - - .10(3) .01(6) 05(2) .03(2) .05(7) .08(4) .06(10) .04(6)
Cophosaurus fexona

.07(1).09(5) - .10(5) .03(1) - - .08(2) .10(4) - - .01(1) .03(2) - .02(2) -
Crofaphyfus collaris

- * * .03(2) • .10(5) .03(2) .03(1) .04(2) -
Dipsosaurus dorsalis

Gambelia wislizenii

------- .01(1) - .03(1) .08(9)

- .07(1) - ♦ _ _ .01(1) .03(2) .01(2) .02(3) .02(3)

Holbrookia maculafa

- - - .08(1) - ,03(1)

Phrynosoma douglassi

.06(3).04(3) .13(1) .04(6) _ _ _
Rirynosoma plafyrhinos

Phrynosoma sol are
Sauromalus obesus

- ' - .07(1) - - .01(1) .11(3) - .03(1) .02(3) .02(3) .05(7)

---- ---- - .03(1) - .03(2) .02(1)

- - - - .03(1) _ _ _ .06(1) .01(7) .02(1) - .03(1) * .02(1) -
Sceloporus clarki

- .03(1) _ _ _ _ . .03(2) .03(2) - - - .03(1) .03(1) .03(1) -
Sceloporus magister

- .05(5) - .05(7) .03(3) .03(2) - .11(4) .23(7) .03(8) .19(3) .13(10) .1 1(16) .10(15) .07(14) .03(6)
Sceloporus undulafus

.13(3). 13(4) .17(3) .07(13) - .10(3) - .02(1) .04(2) - - - - - -

Urosaurus graciosus

---------- - .07(7) .07(11) .01(3) .04(2) .07(13)

Urosaurus omafus

.03(1).04(4) - .04(6) .03(7) .05(3) - .15(4) .20(5) .03(1) - .08(5) .04(5) .03(3) .06(7) ,02(3)
Ufa sfansburiana

- .03(2) - .04(4) ,04(7) .05(1) .10(1) - .11(7) .05(8) .05(2) .08(5) .1 1(13) .05(12) .13(17) .09(15)
Eumeces gilberfi

.03(1).06(3) - .05(9) .11(10) .03(1) - .02(2) .04(4) .03(1) ______

Eumeces obsolefus

Cnemidophorus burfi

Cnemidophorus flagellicaudus

- .05(3) - .04(5) - .07(2) - .08(1) .02(1)
Cnemidophorus inornatus

- .03(2) - * - .03(1)

(continued)

113

c

Table2.— (continued).

PP PJ SB CC OC DG
(5) (9) (3) (18) (13) (11)

DD
(3)

MB CW JM CA
(6) (13) (9) (3)

ME
(15)

MR
(16)

MD
(15)

SD
(22)

CB
(22)

Cnemidophorus figris

nom 10 f/^^ — 07(6) 05(3) 09(4)

23(2)

10(3) 07(7) .14(9) .25(3) .

14(9)

.25(16)

.13(15)

.17(21)

.15(21)

{^noniioopnoius ui ]ii~juit3i lo

_ _ _ .04(1) - .03(1)

—

—

—

—

—

—

Cnemidophorus velox

- * .49(3) .14(5) - .01(1)

.05(2) .02(2) - -

—

—

—

—

Xanfusia vigilis

- .01(1) -

.07(1)

- .02(1) - .05(1) .08(7) -

—

Total Number of Species (includes species verified by road-riding and searclnes)
7 14 4 20 12 14 4 10 16 11 9 10

16

14

17

12

Mean Relative Abundance

.37 .69 .79 .96 .43 .74

.47

.67 1.06 .54 .72

,59

.91

.58

.78

.63

Species Diversity (H')

.56 1.00 .40 1.18 .89 1.07

.54

.91 1.00 .76 .72

.86

1.05

1.00

1.09

.95

J

15
.75

Component I

-.75

Sknitn/
Grasses

Component III

1 - EwMcd obsoWM

2 - Ctleonyi foritfohis

] - Mcdemn ml^mian

4 - CdtMM *ocanoid«l

5 - Cophosounn kenun

6 - OuttphyfaiS cdbnt

7 - D^jMnwva towh

I - GonMio (WiOii

$ - Hotmcldo mooioto
10 - Rnynosoni dougkns

I I - Rfjbosoto plotjrtinos
12 - Rf)iwswno vim

13 - Sann«lut cbsw

U - Scelopann darti
15 - Scikpana najim
IS - Scilcpmf uiduMif

17 - Umwvn fTOCWSus

18 - UoNwvs omotn

19 - Uto ilmlMioM

20 - EiiffltcM ffbvti

21 - DinMcts obsolekis

22 - OMnUai^mi barb'

23 - Ca«nidg|]hona 9aj/kKin

24 - Omidofiham •wmtba

25 - CfwrMoftow ttf*

26 - Cneiitfoftow u^mn

27 - Cn«n4«|iharui lim
21 - bnhso liglil

Opsn

-.75

-.50

-J5

J5

JO

Component II

Figure 3.— Mean factor scores of microhabitats for lizards.

types. Others, such as Chilomeniscus
cinctus, Chiomctis occipitalis, Phyl-
lorhynchus browni, Phyllorhynchus de-
curtatus, and Crotalus cerastes, oc-
curred primarily on lower elevation
(< 915 m or 3000 ft), desert habitats,
and some, such as Lampropeltis py-
romelam and Crotalus viridis cerberus,
occurred only on higher elevation
(>1525 m or 5000 ft) habitats (table 3).
Lichanura trivirgata and P. browni oc-
cur primarily outside the study ar-
eas, and their distributions only
overlap the extreme southern and
southwestern portions of the study
area. Therefore, they were limited to
the small number of sites with suit-
able habitat. Thamnophis cyrtopsis and
Thamnophis marcianus were restricted
to sites with water, with the former
occurring on a large number of habi-
tats and the latter only in a mesquite
bosque habitat along the Gila River
south of Phoenix. Similar to Copho-
saurus texana, Tantilla hobartsmithii
was found on higher elevation
(>1220 m or 4000 ft) and riparian
habitats throughout most of the
study area, but also in Sonoran Des-
ert in the eastern portion of the study
area.

A PC A of microhabitats on which
snakes occurred revealed that, simi-

114

Table 3.— Relative abundance of snakes by habitat type. Relative abundance = the number of an Individual species
caught In an array per 24 h period.* Indicates species verified In a habitat type via road-riding and searches. The
number below the Habitat Type in ( ) = the total number of arrays. The number in () to the right of the species' relative
abundance = the number of arrays in which the species was trapped.

Arizona elegans

* ♦

Chilomeniscus cine f us

Chionacfis occipitalis

Diadophis puncfafus

Hypsiglena for quota

Lampropeltis getulus
« » ♦

Lampropeltis pyromelana

♦

Lichanura trivirgata

- - - - • • - * .02(1) * .03(2)
- - - - .05(3) - - .07(7) .08(7) - .02(2) .03(1)
------- .03(3) .12(2) - .05(4) .06(5)

.02(2) .02(1) -

- - .03(2) * .03(2) .02(1) .02(1) * ,02(1) * -

Masficophis bilineatus

- - - - .03(2) .03(1) - • * * _ _ _ _ _ _

Masticophis flagellum

- ,03(1) - * * - - - .02(1) * .02(1) .02(1) .02(1) ,02(1) ,02(1) .03(2)
Masficophis taeniatus

,03(2) .03(2) - ,02(1) * _ _ _ _ _ _

Pituophis melanoleucus

.02(1) * .02(1) * .02(1) * * • .01(1)

Pliyllorliynclius browni
Phyilorhynclius decurfafus
Rhinoclieilus leconfi

- — - - - - ,02(2) - * .03(2)

- .02(1) .02(1) * * * * * .02(1) .02(1)

Salvadora liexalepis

- - - .02(1) .02(1) * - - .03(2) • .03(1) .03(1) .02(1) .04(2) .02(1)
Sonora semiannulafa

- - - * * .10(2) - - * .03(1) * ♦ _ * .05(3)
Tanfilla hobarf smith ii

- - - .05(5) .08(8) ,07(2) - .02(1) .05(4) .03(1) - .03(2) - -
Thamnophis cyrtopsis

*->»»• »•♦♦♦♦»»♦

Thamnophis marcianus

«

Trimorphodon biscutafus lambda
_ »

Crofalus afrox
Crofalus cerastes
Crofalus mifchelli

.02(1)

.02(2)

(continued)

115

>

Table 3.— (continued).

PP PJ SB CC

OC

DG

DD

MB

CW

JM

CA

ME

MR

MD

SD

CB

(5) (9) (3) (18)

(13)

(11)

(3)

(6)

(13)

(9)

(3)

(15)

(16)

(15)

(22)

(22)

Crofalus molossus

*

«

•

•

•

•

♦

♦

Crofalus scututatus

.02(1)

«

» ■■

♦

Crofalus figris

Crofalus viridis cerberus
• * «

Micruroides euryxanfhus

Leptotyphlops humilis
- .04(3) - -

.05(2) .05(3) .05(3)

.03(2) - - * -
.09(6) .03(2) .02(1) .03(2) .06(6)

Total Number of Species (includes species verified by road-riding and searches)

2

4 11 6 17

22

12

Mean Relative Abundance

- .07 - .07

.18

.23

Species Diversity (H')
- .30 - .26

.63

.54

12

20

18

13

17

18

16

25

16

.18

.24.

16

.07

.36

.28

.12

.22.

.29

.75

.81

.68

.47

.93

.68

,68

.86

.90

15
TrMt

75 _

Component I

Crane*

29 - Arizona ejegans . 4-6 •

30 - Chilomeniscus cinctus 47 •

31 - Chionactis occipitalis 4-8 •

32 - Diodophis punclatus 4-9 •

33 - Hypsiglena torquata 50 ■

34 - Lompropeltis qetuius 51 ■

35 - Lompropeltis pyromelaiM •

36 - Lichanura trivirqata 53 ■

37 - Mosticophis bilineatus 54 •

38 - Mosticophis flagellum 55 •

39 - Masticophis taeniatus 56 •

40 - Pituophis melanoleucu57 ■

41 - Phyllorhynchus browni 58 •

42 - Phyllorhynchus decurtotus

43 - Rhinocheilus leconti

44 - Salvadoro hexolepis

45 - Sonora semiannulqja

Tontilla hobortsmithii
Thomnophis cyrtopsis
Thomnobhis mcrcianus
Trimorpnodon biscutctus lambdc
Crotolus otrox
Crotolus cerastes
Crotalus mitchelli
Crotolus molossus
Crotalus scutulatus
Crotalus tigris
Crotolu? viridis cerberus
Micruroides euryxonthus
Leptotyphlops humilis

" 51 54

Component

50

A2 M

5:i

51)

57

49

58

48

46

56

Open
Canopy

Component

Figure 4.— Mean factor scores of microhabitats for snakes.

lar to those of lizards, microhabitat
associations differed among snakes
(fig. 4). Many of the widely distrib-
uted snakes, such as Hypsiglena
torquata, Lampropeltis getulus, Mastico-
phis flagellum, and Pituophis melano-
leucus, showed no strong relationship
with any of the compressed habitat
components (fig. 4). Conversely,
most species with limited distribu-
tions showed a strong relationship
with certain components (fig. 4).
Chionactis occipitalis, Crotalus cerastes,
Crotalus scutulatus, and Phyllorhyn-
chus browni consistently occurred on
open, sandy sites, and Chilomeniscus
cinctus occurred on sites with sandy
substrate but taller canopy (fig. 4).
Other species, such as Crotalus mitch-
elli and Sonora semiannulata, were
found on sites with open canopies
but rocky substrates (fig. 4). Thamno-
phis marcianus and Tantilla
hobartsmithii occurred on sites with
sandy substrates but closed canopies
and large amounts of vegetative de-
bris, and Lampropeltis pyromelana oc-
curred only on sites with high
amounts of vegetative debris (fig. 4).
Other species, such as Diadophis

116

pundatus, Thamnophis a/rtopsis, and
Crotalus viridis cerherus, occurred on
rocky sites with high amounts of
vegetative debris (fig. 4).

Except for a single Gopherus agas-
sizii captured in an array, all turtle
records came from road-riding and
field searches. Four species of turtles
were recorded within the study area,
three aquatic and one terrestrial
(table 4). Of these, G. agassizii was the
most widely distributed (verified in 9
habitat types, table 4). A more thor-
ough account of this turtle's distribu-
tion is described by Burge (1979,
1980). Pseudemys scripta, an intro-
duced species, was limited to a
stretch of the Gila River from the
99th Street bridge in southwest Phoe-
nix to Gillespie Dam, located ap-
proximately 24 km (15 miles) south
of Buckeye. Trionyx spiniferus oc-
curred at Alamo Lake (confluence of
the Big Sandy and Santa Maria rivers
in western Arizona) and along peren-
nial stretches of the Gila River from
Phoenix to Yuma. Kinosternon sonori-
ense occurred on several permanent
streams and rivers throughout the
study area.

In contrast to the observed distri-
bution patterns among lizards and
snakes, the distribution of amphibi-
ans did not shown an elevational pat-
tern. Although certain species such

as Bufo punctatus and Scaphiopus
couchi occurred in a large number of
habitat types, most species were
found in at least one lower (< 915 m
or 3000 ft) and one higher (> 1220 m
or 4000 ft) elevation site (table 5).
Similar to lizards and snakes, there
are some amphibians whose ranges
are principally outside the study area
and are, therefore, found only on a
few sites (table 5). The ranges of Bufo
debilis, Bufo retiformes, and Gastro-
phyrne olivacea are primarily in north-
ern Mexico, or east and south of the
study area in the Chihuahuan Desert;
within the study areas, their ranges
are limited to desert grassland habi-
tats in the extreme southern portion
(Vekol Valley, 48 km or 30 mi west-
southwest of Casa Grande). All
populations of Ambystoma tigrinum
were located at earthen stock tanks
(dirt tanks). Presumably, all of these
populations were introduced.

A PCA demonstrated correlations
between occurrence of amphibian
species and particular microhabitats
(fig. 5). Bufo debilis, B. retiformes, and
Gastrophyrne olivacea occurred on
sandy, grassy sites, and Bufo cognatus
on sandy, shrubby sites (dg. 5). Bufo
microscaphus and B. punctatus oc-
curred on rocky sites, and Hyla areni-
color on rocky sites generally occu-
pied by trees and large amounts of

vegetation debris (fig. 5). Certain
species, such as Scaphiopus couchi,
Bufo alvarius, and Bufo woodhousei oc-
curred on sites with a wide variety of
substrates (fig. 5).

The occurrence and frequency of
water was not quantitatively meas-
ured at each site; therefore, the influ-
ence of water was not considered in
the development of figure 5. How-
ever, all sites with amphibians had
surface water during some part of
the year, especially during summer
months. AH sites with Bufo mi-
croscaphus, Rana pipiens, R. catesbe-
iana, and Hyla arenicolor had perma-
nent water (e.g., springs, creeks, and
rivers).

At the start of the survey in 1977,
populations of Bufo microscaphus and
B. woodhousei sympatric on major
drainages, such as the Hassayampa,
Santa Maria, Agua Fria, and New
rivers, could be easily distinguished
from one another. By 1981, popula-
tions on all of these drainages were
indistinguishable.

Range Extensions

Thirty-five range extensions were
recorded for amphibians and reptiles
within the study area. Except for the
following discussion, range exten-

Table 4.— Distribution of turtles by habitat type C). Records are entirely from road-riding and searches (except where
otherwise Indicated. All turtles except Gopherus agassizii occurred only at sites with permanent water within habitat
types listed below.

PP PJ

SB

CC OC

DG

DD

MB

CW JM

CA

ME

MR MD

SD

CB

Gopherus agassizii

«

Pseudemys scripta
Trionyx spiniferus
Kinosternon sonoriense

" Trapped in an array
Number of species

117

— - .. , ,,

Table 5 -Relative abundance of amphibians by habitat type.Relative abundance = the number of an individual spe-
cies caught in an array per 24 h period. ' indicates species verified in a habitat type via road-riding and searches.
The number below the Habitat Type in ( ) = the total number of arrays. The number in ( ) to the right of the species
relative abundance = the number of arrays in which the species was trapped.

PP
(5)

PJ
(9)

SB
(3)

CC
(18)

OC
(13)

DG
(11)

DD
(3)

(6)

CW
(13)

JM
(9)

CA
(3)

ME
(15)

MR
(16)

MD
(15)

SD
(22)

CB
(22)

Bufo olvorius

- .07(1)

Bufo cognatus

* — —

Bufo debilis

Bufo mlcroscaphus'

- .05(2) -•
Bufo puncfafus

.03(1).03(1) * .15(8) .11(8)
Bufo refiformis

Bufo woodhouseP

Hyla orenicolor

- .07(1) .03(1)
Gastrophyrne olivacea

.06(2)
.14(3)
.03(1)

.03(2)

.03(4)

.09(3) -

.06(4)

18(2)

.13(3) .06(3) V - -07(4) _ - - -
.20(1) .16(3) .12(2) .23(1) .28(6) .05(2) .06(2) .10(7) -

_ • _ - - .03(1) -

Scaphiopus couchi

Rana pipiens
• ♦ ♦

Rona cafesbeiana
Ambysfoma figrinum

.12(2)

.05(2)
.10(3)

• .20(1)

.15(6) .06(2) .07(6) .06(5) .11(6)

.03(1)

'95% of these were a cross between the two species (B, microscaphus x B. woodhousei)
Total Number of Species (includes species verified by road-riding and searches)
4 4 3 6 6 8 2 4 6 5 5
Mean Relative Abundance

.03 .03 - .34 .26 .56 - .33 .45 .12 .23
Species Diversity (H')

_ - - .56 .42 .71 - .29 .51 -

8

6

5

6

3

.65

.14

.13

.16

.17

.65

.46

.30

.29

.28

sions discovered during this study
have been described elsewhere (Jones
et al. 1981, Jones et al. 1982, Buse
1983, Jones et al. 1983, Jones et al.
1985). The southernmost distribution
of Tantilla hobartsmithii was extended
from the Salt River east of Phoenix,
southwest in the mesquite bosque
habitat along the Gila River to 56 km
(35 miles) east-northeast of Yuma
(fig. 6). A population of T. hobart-
smithii was also discovered in a 10 ha

(25 acres) open chaparral habitat in
the Eagletail Mountains (fig. 6). The
westernmost distribution of Cnemido-
phorus burti was extended from the
Tucson area northwest by discovery
of isolated populations in desert
grassland habitats on summits of the
Tabletop and Estrella mountains (fig.
6).

An isolated population of Mastico-
phis bilineatus lineolatus was discov-
ered on the summit of Tabletop

Mountain in a relict desert grassland
habitat (fig. 6). This population ex-
tends the known distribution of this
subspecies approximately 100 km (62
mi) to the north of the only other
known population (Ajo Mountains).

Finally, an isolated population of
Diadophis punctatus was discovered
in a relict desert grassland commu-
nity on the summit of the Estrella
Mountains southwest of Phoenix (fig.
6).

118

Comparison of Habitat Types

Based on data compiled from pit-fall
trapping, road-riding, and searches,
the Sonoran Desert habitat had the
greatest species richness (49 species,
fig. 7). Closed chaparral and cotton-
wood-willow riparian habitats were
the second richest habitats (44 spe-
cies), and open chaparral and mixed
riparian scrub were third (41 species,
fig. 7).

Disclimax desert grassland had
the fewest species (8), and sagebrush
and ponderosa pine had the second
and third fewest species (13 and 15
species, respectively, fig. 7). All other
habitats had at least 27 species but
not more than 39 (fig. 7). Although
Sonoran Desert had the richest lizard
and snake faunas, mesquite bosque
and desert grassland habitats had the
richest amphibian fauna (fig. 7). The
mesquite bosque habitat type had the

greatest number of turtle species
(four species, fig. 7).

When only array data are com-
piled, disclimax desert grassland,
sagebrush, and ponderosa pine habi-
tats still had by far the lowest num-
ber of species, but Sonoran Desert
and mesquite bosque had the great-
est number of species (fig. 8). As
when all data were taken into ac-
count, mixed riparian scrub, cotton-
wood-willow riparian, closed chap-
arral, and open chaparral had high
species richness (fig. 8). However,
desert grassland was relatively more
diverse using only array data (fig. 8).

The difference between array vs.
all data appears to result from the
inability of arrays to consistently ver-
ify (trap) turtles and medium and
large snakes, although many larger
snake species were verified because
young-of-the-year were easily
trapped.

A more revealing statistic is the
average number of species verified
by an array (fig. 8). This analysis re-
veals which habitats consistently had
the largest number of species at
sample sites. Certain habitats, such
as desert grassland, although high in
overall species richness, had rela-
tively few species verified at each
array site (fig. 8). Other habitats,
such as ponderosa pine, sagebrush,
and disclimax desert grassland, had
the lowest number of total species
and the lowest average number of
species per array site (fig. 8). Many of
the habitats that had high overall
species richness also had high overall
richness at each array site; however,
cotton wood-willow had a higher av-
erage number of species per array
site than did Sonoran Desert (fig. 8).

Species diversity indices (H') cal-
culated from array data reveal pat-
terns similar to those described
above (fig. 9). Disclimax desert grass-
land, sagebrush, and ponderosa pine
continue to exhibit low diversity, and
Sonoran Desert, closed chaparral,
cotton wood-willow riparian, mixed
riparian scrub, and desert grassland
continue to be diverse (fig. 9). How-
ever, as in the previous analysis, the
average diversity per array site is
low when compared to total diver-
sity for individual habitats (fig. 9). Of
the habitats with high overall diver-
sity, mixed broadleaf riparian and
cottonwood-willow riparian had
relatively high average diversity per
array site (fig. 9).

A comparison of herpetofaunas of
each habitat type by cluster analyses
revealed that all desert habitats, such
as creosotebush, Sonoran Desert,
Mohave Desert, and mixed riparian
scrub had very similar herpetofaunas
(figs. 10 and 11). In both cluster
analyses, open and closed chaparral
had similar herpetofaunas, and sage-
brush and disclimax desert grassland
had a herpetofauna different from
any other habitat. However, there
were differences in results of the two
cluster analyses for other habitats.
Whereas the cluster analysis of array

1.5
Tro«

75

Component I

-.75

Grauei

Component

59 - aufo olvoriui

60 - Bufo cognotm

61 - Biifo deUD]

62 - SjTo mlcroicaphui

63 - auto punctatui
- Buto rotltormS

66 - Buf 0 woodhomel

66 - Hyla aronlcolof

67 - Gaitrophyme ollvocea

68 - ScopmopLa coucni

69- nana plplem

70- nono coteiDelana
71 - AmbyitDtno ttgrtnum

iJ 71

Open
Canopy

0 .25

60

VeQerotlva

Component I

Figure 5.— Mean factor scores of microhabitats for amphibians.

119

data revealed large differences be-
tween the herpetofaunas of cotton-
wood-willow and desert habitats,
such as Sonoran and Mohave Des-
erts, these habitats had a relatively
moderate degree of overlap when all
data were analyzed (figs. 10 and 11).
Additionally, ponderosa pine and
pinyon-juniper habitats were similar
when array data were analyzed and
relatively dissimilar when all data
were submitted to cluster analysis
(figs. 10 and 11).

DISCUSSION

Overall, western Arizona has an ex-
tremely diverse herpetofauna, pri-
marily because of its large variety of
habitats zoogeographic location. The
Hualapai Mountains, located in
northwestern Arizona, are adjacent
to three major deserts: the Mohave
Desert to the northwest, the Great
Basin Desert to the northeast, and the
Sonoran Desert to the south. No-
where else on the North American
continent does such a phenomenon
exist. The diversity of habitat in this
area is also enhanced by the occur-
rence of several woodland islands.

Number of Species

50

40

30 _

20 -

10 _

Tuttes

Amphibians

Snakes

Uzofds

PP PJ SB CC OC DG DO MB CW JM CA rvE fv« MD SO CB

Habitat Type

Figure 7.— Number of species by taxonomic group by habitat type. (Abbrev. correspond to
ttiose listed for liabitats In table 1 .)

Number of Species

30

25 —

20

15 -

10 —

6 —

#Of

Species

Ave # of
species/
array

PP PJ SB CC OC DG DD MB CW JM CA ME MR MD SD CB

Figure 6.— Map of range extensions.

Habitat Type

Figure 8.— Total number of species caughit in arrays by habitat type vs. the average number
of species caught per array by habitat type. (Abbrev. correspond to those listed for habitats
In table 1 .)

120

Species Diversity (H')

Patterns of Species Distributions

Habitat Type

Figure 9.— Total species diversity (H') by hiabitat type vs. average species per array by tiabl-
tat type. (Abbrev. correspond to thiose listed for htabitats in table 1 .)

Similarity

MR SD MD CA OC JM CW CC ME CB DG PJ MB PP SB DD

1.0

Figure 10.— Cluster analysis (dendrogram) of array data Illustrating similarities in tiabitat type
tierpetofaunas. (Abbrev. correspond to thiose listed for tiabitats in table 1 .)

This survey reveals that certain spe-
cies are widespread, occurring in
several habitats, but many species
are limited to specific habitat types.
Also, some species occur on most
sample sites within a habitat type
and others on only a few. There ap-
pear to be at least 3 major factors
contributing to distributional pat-
terns of amphibians and reptiles in
the study area.

Geographic Limitations

The ranges of certain species only
peripherally occur in western
Arizona. Cnemidophorus burti, Phyl-
lorhynchus browni, Masticophis bilinea-
tus lineolatus, and Bufo retiformis oc-
cur principally in northern Mexico
whereas others such as Holbrookia
maculata, Eumeces obsoletus, Gastro-
phyrne olivacea, and Bufo debilis are
mostly east and north of the study
area (Stebbins 1985). Bufo retiformis,
Gastrophyrne olivacea, and Bufo debilis
are associated with low elevation
(457-915 m or 1500-3000 ft) desert
grassland (Jones et al. 1983), and
these habitats are mostly absent in
the central and northern portions of
the study area. However, habitat
suitable for other species listed above
appears to be available throughout
most of the study area.

Physical barriers, such as topogra-
phy, elevation, and climate may have
presented these species from coloniz-
ing or immigrating into suitable habi-
tats to the north and west (see Con-
nor and Simberloff 1979, Case 1983,
Jones et al. 1985 for discussion of the
influence of physical barriers on colo-
nization/immigration). In addition,
competition between species may
have limited individual species'
ranges during initial and subsequent
colonization of suitable habitats (e.g.,
during periods of large climatic
changes). Perhaps the best example
of this is the distributional relation-
ship between Eumeces gilberti and E.

121

ohsoletus. E. gilberti belongs to the
skiltonianus group of skinks, whose
evolutionary center is the western
United States (Taylor 1935, Rogers
and Fitch 1947).

Conversely, E. obsoletus evolved in
the Great Plains region (Fitch 1955).
Both of these lizards occupy seem-
ingly identical, but separate, habitats
in central Arizona, and their distribu-
tions come together in chaparral and
desert grassland habitat types near
Cordes Junction; the westernmost
range of E. obsoletus is just east of
Interstate Highway 17 and the east-
ernmost range of E. gilberti is just
west of the highway. These lizards
are similar in appearance, with E. ob-
soletus averaging slightly larger in
size.

Although subtle differences in mi-
crohabitat cannot be ruled out as fac-
tors influencing their ranges, it ap-

Similarity

pears that these lizards are mutual
exclusive (competitive exclusion).

Several remnant stands of chapar-
ral and desert grassland occur in
western and northwestern Arizona at
or near the summits of mountain
ranges. These relict stands or habitat
islands are isolated within creo-
sotebush and Sonoran Desert habi-
tats as a result of the retreat of the
last Ice Age (see Van Devender and
Spaulding 1977). Data collected in
my study show that several reptiles
typically found in "upland" habitats
(e.g., large continuous stands of des-
ert grassland and woodlands associ-
ated with the Colorado Plateau of
central and northern Arizona) inhabit
these isolated mountain stands, al-
though the number and composition
of these upland species vary among
mountains. Habitat island size ap-
pears to be of primary importance in

determining the number of upland
present species (see Jones et al. 1985).

The turtles Pseudemys scripta and
Trionyx spiniferus are present along
the Gila River as a result of
introductions. P. scripta is a popular
pet, and specimens have been re-
leased along the Gila River in south-
west Phoenix. T. spiniferus was intro-
duced along the Colorado River in
the early 1900's (Stebbins 1985); pre-
sumably, these populations ex-
panded into the Gila River at the
confluence of the Gila and Colorado
rivers near Yuma.

Microhabitats and Physical
Characteristics of Habitat

Many studies have shown a strong
relationship between the distribution
and abundance of amphibians and
reptiles and the presence and amount
of certain microhabitats (Norris 1953,
Pianka 1966, Zweifel and Lowe 1966,
Fleharty 1967, Pianka and Parker
1972). The distribution of a number
of species within western Arizona
area appears to be influenced by the
presence of microhabitats on sites,
although most of the widespread
species, such as Cnemidophorus tigris,
Pituophis melanoleucus, and Lam-
propeltis getulus show no strong rela-
tionship with any specific habitat
components, others (e.g., Urosaurus
ornatus and Sceloporus magister) occur
on sites with trees and downed litter.
Many sites in the study area, includ-
ing desert and upland habitat types,
have trees and downed logs, and this
probably accounts for these species'
wide distributions. The habitat analy-
sis revealed that several species are
associated with specific substrate
types (e.g., rock), density or height of
the vegetation canopy, type of vege-
tation (shrubs or grasses vs. trees), or
presence of downed litter.

Species' associations with certain
microhabitats may reflect their physi-
cal or behavioral limitations. For
example, Eumeces gilberti may be re-
stricted to sites with large amounts

MR SD MD CA OC JM CW CC ME CB DG PJ MB PP SB DD

1.0

Figure 1 1 .—Cluster analysis (dendrogram) of all data illustrating similarities in tiabitat type
hterpetofaunos. (Abbrev. correspond to thiose listed for hiabitats in table 1 .)

122

of downed litter (primarily leaves
and logs) because of its low preferred
body temperature and feeding habits
(Jones 1981b, Jones and Glinski 1985).
Large amounts of surface litter on
certain riparian sites may explain the
occurrence of this lizard in cotton-
wood-willow riparian sites within
desert regions (down to 549 m or
1800 ft) (see Jones and Glinski 1985).
Several other species typically found
on upland habitats (e.g., chaparral),
such as Tantilla hobartsmithii, Copho-
saurus texana, Masticophis bilineatus,
and Diadophis punctatus, also may
persist on riparian habitats within
deserts because of the high moisture
regime associated with surface litter,
higher humidity, and surface water
(Jones and Glinski 1985).

A similar relationship appears to
exist in desert habitats occupied by
Xantusia vigilis. This lizard also has a
low preferred body temperature, and
it only occurs on Mojave Desert sites
occupied by agaves (Agave spp.) and
yuccas (Yucca spp. and Nolim spp.);
these plants create cool, moist mi-
crohabitats within desert habitats. In
the southern part of its range, X. vig-
ilis only occupies Sonoran Desert on
steep slopes in mountain canyons, or
on top of mountains (> 1220 m or
4000 ft) in chaparral habitats. This
shift in habitat association may re-
flect increased average temperature
and aridity associated with decreas-
ing latitude; canyons and mountain
summits may be the only sites mod-
erate enough to support this lizard.

A similar moisture or temperature
relationship may also account for dif-
ferences observed in habitat type as-
sociations of Tantilla hobartsmithii,
Cophosaurus texana, and Diadophis
punctatus in the eastern and western
portions of their ranges. In the west-
ern portion of the study area, these
reptiles occur only in chaparral or
riparian habitat types (excluding
mixed riparian scrub habitats). In the
eastern and southeastern portions of
the study area, these species also oc-
cur in the Sonoran Desert habitat
type. Eastern and southeastern Sono-

ran Desert habitats within the study
area are more extensive than those to
the west and northwest, and they are
not interrupted by large creo-
sotebush habitats; western and
northwestern sites are restricted
mostly to mountain slopes, separated
by extensive creosotebush flats. In
addition, eastern and southeastern
sites appear to have more springs
and perennial creeks than western
and northwestern sites, and this ad-
ditional moisture might contribute to
the presence of these species on these
sites.

The presence of surface water also
has a profound affect on the distribu-
tion and abundance of certain species
within the study area. Kinosternon
sonoriense, Trionyx spiniferus, Thamno-
phis cyrtopsis, Bufo alvarius, Bufo mi-
croscaphus, Bufo woodhousei, Rana pipi-
ens, Rana catesbeiana, Hyla arenicolor,
and Ambystoma tigrinum occur only
on sites with permanent water
(springs, creeks, rivers, dirt tanks).
All of these species are restricted to
permanently watered sites because of
a combination of physiological
(Walker and Whitford 1970), mor-
phological (Mayhew 1968), reproduc-
tive (Justus et al. 1977), or behavioral
(Hulse 1974) limitations. In addition
to occurring near permanent water,
Bufo punctatus also occurs in rock-
bound canyons with intermittent wa-
ter, and Bufo cognatus, B. debilis, B.
retiformis, and Gastrophyrne olivacea
occur on sites with clay and clay-
loam soils that accumulate surface
water during summer convectional
rainstorms. All of these species pos-
sess adaptations, such as a rapidly
developing embryo, that are condu-
cive to survival in areas with inter-
mittent surface water (Creusere and
Whitford 1976).

A number of species were verified
on fewer than half of the array sites
within habitat types. These low per-
centages may reflect species' associa-
tion with specific microhabitats and
the abundance and distribution of
microhabitats within habitat types.
For example, Chilomeniscus cinctus

occurred on less than half of the cot-
tonwood-willow and mixed riparian
scrub array sites. The habitat analysis
shows that this species is associated
with sandy and fine gravel soils, but
many of the cotton wood-willow ri-
parian and mixed riparian scrub
sample sites have rocky substrates.
Therefore, the substrate type limits
this species' range within these habi-
tat types.

However, there were other spe-
cies, especially snakes in excess of 0.5
m (1.5 ft), that were not readily
caught in pit-fall traps, although a
small percentage of arrays captured a
few large snakes; these snakes were
feeding on small rodents at the bot-
tom of traps. Therefore, the paucity
of large snakes on samples sites
within habitats probably reflects the
ability of larger snakes to escape
from pit-fall traps rather than the dis-
tribution and abundance of mi-
crohabitats within habitat types. Ad-
ditionally, amphibians and reptiles
with restricted activity patterns (e.g.,
toads) or home ranges (Xantusia vig-
ilis) also were rarely trapped and,
therefore, verified on few sites within
a habitat. The limited number of
mixed broadleaf and chaparral array
sites with Gerrhonotus kingi probably
reflect a low sampling effort in these
habitats during the fall; this lizard's
peak activity is during its breeding
season in the fall (Robert Bowker
personal comm.).

Habitat Conditions

The condition of habitats may play
an important role in determining the
distribution and abundance of am-
phibians and reptiles. In Arizona, the
large variety of land uses within the
area may affects the distribution and
abundance of certain microhabitats
and may account for variation in spe-
cies composition within habitats. A
number of studies have shown the
effects of land uses on amphibians
and reptiles and their habitats. These
include grazing (Bury and Busack

123

1974, Jones 1981a, Szaro et al 1985),
off-road vehicle use (Bury et al. 1977,
Bury 1980), forest management (Ben-
nett et al. 1980), and stream modifi-
cation resulting from water im-
poundments (Jones, this volume).
Generally, these affect habitat struc-
ture. For example, excessive, long-
term livestock grazing reduces the
abundance and diversity of forbs and
perennial grasses. Many former des-
ert grassland habitats are now domi-
nated by shrubs such as creosotebush
(Larrea tridentata) and mesquite
(Prosopis glandulosa) (York and Dick-
Peddie 1969). Jones (1981a) showed
large differences in the presence and
abundance of certain lizards on heav-
ily vs. lightly grazed sites, especially
on riparian, desert grassland, and
woodland habitats, attributable to
differences in lizard ecology and dif-
ferences in habitat structure between
heavily vs. lightly grazed areas. Cer-
tain lizards, such as Cnemidophorus
tigris, prefer open, shrubby sites;
these lizards are more abundant on
heavily grazed sites where shrubs
have replaced grasses and forbs
(Jones 1981a). Conversely, certain
lizards, such as Eumeces gilberti, pre-
fer grassy, moist sites, and are, there-
fore, less abundant on or absent from
sites where grazing has reduced tree
reproduction (e.g., cottonwoods,
Populus fremontii on riparian sites) or
suppressed grasses (e.g., on desert
grassland sites) (Jones 1981a).

The reduction of naturally-occur-
ring water and the modification of
river and stream habitats has been
shown to affect the composition of
amphibians and reptiles within habi-
tats, especially riparian sites (Jones
1988). Platz (1984) attributes the ex-
tinction of Ram onca to modification
of stream habitats along the Virgin
River. Species that prefer lentic or
pool habitats should increase on sites
with water impoundments, whereas
species that prefer lotic or running
water should decrease.

Natural phenomena, such as fire,
also affect species composition
within habitats (Kahn 1960, Simovich

1979). Simovich (1979) showed that
fire set back succession within chap-
arral habitats (grass/forb succes-
sional stage), and that these changes
resulted in increases in certain spe-
cies and decreases in others. As suc-
cession proceeded to shrubs and
trees, reptiles that were abundant in
the grass/forb successional stage
(e.g., Phrynosoma coromtum) became
less abundant, and others that pre-
ferred wooded sites (e.g., Sceloporus
occidentalis) became more abundant.

Historical vs. Present Distributions

Prior to this study, records of am-
phibians and reptiles on the study
area were limited; one of the primary
reasons for which this study was
conducted was to assemble basic dis-
tribution information. Therefore,
range expansions or reductions were
hard to document. This study re-
sulted in range extensions of ap-
proximately 35 species, and clarified
the relationship of Arizona habitats
to habitats in adjacent geographic
regions. Many species, such as Helod-
erma suspectum, Eumeces gilberti, Sce-
loporus clarki, Tantilla hobartsmithii,
and parthenogenic whiptail lizards
(Cnemidophorus flagellicaudus, C. uni-
parens, and C. velox) proved to be
considerably more widespread than
previous records indicated — not sur-
prising since many areas had never
been intensively sampled. The expan-
sion of E. gilberti' s range results from
the discovery of the California
subspecies, E. g. rubricaudatus, in
chaparral and pinyon-juniper habi-
tats; the distribution of E. g. ari-
zonenis is limited to a cottonwood-
willow riparian habitat along an 18
km (11 mi) stretch of the Has-
sayampa River immediately south of
Wickenburg (see Jones et al. 1985,
Jones and Glinski 1985).

Only one species demonstrated a
range reduction. Pure populations of
Bufo microscaphus have apparently
been reduced due to hybridization
with Bufo woodhousei, especially on

major drainages. Water impound-
ment and diversion-associated
changes in aquatic habitats from per-
manent riffles and runs to pools may
have caused the immigration of B.
woodhousei into areas formerly occu-
pied by only B. microscaphus (Brian
Sullivan personal comm.).

There is considerable taxonomic
confusion about a population of
Kinosternon sonoriense on the Big
Sandy River near Wikieup. Because
specimens with raised 9th marginal
scales had been taken from this area,
Stebbins (1966) considered this popu-
lation to be Kinosternon flavescens, but
Iverson (1978) considered it to be K.
sonoriense, based on specimens with-
out 9th marginals. Of the 12 indi-
viduals observed during this study, 6
had raised 9th marginals and 6 did
not. Based on its large separation
from the nearest population of K.
flavescens, Iverson (personal comm.)
considers this population to be an
aberrant form of K. sonoriense.

Similarity of Habitats Types

It is possible to discern definite pat-
terns in the diversity of and similari-
ties between the herpetofaunas of
different habitat types within the
study area. There is an apparent ele-
vational gradient affecting species
diversity. Desert habitats between
610 and 1067 m (2000-3500 ft), ripar-
ian habitats between 549 and 1220 m
(1800-4000 ft), and chaparral habitats
between 1067 and 1525 m (3500-5000
ft) had greater species richness than
higher elevation woodland (> 1677 m
or 5500 ft, e.g., Ponderosa pine) and
desert habitats (> 1220 m or 4000 ft,
e.g., sagebrush). Additionally, low
elevation desert habitats (> 610 m or
2000 ft, e.g., creosotebush), had rela-
tively low species diversity. Higher
species diversity on middle elevation
habitat types may reflect these habi-
tats' moderate environmental and
climatic conditions, whereas higher
and lower elevation habitats possess

124

extreme environmental and climatic
conditions (e.g., temperature). For
example, low elevation creosotebush
habitats have sparse canopies, and
temperatures often exceed 60 C near
the surface in summer (Costing
1956). High elevation sites are cold
and are often snowcovered until late
April so that the growing season is
short. Although possessing relatively
low species richness, low elevation
creosotebush habitats are more di-
verse than high elevation sites. These
differences in diversity may reflect
thermal conditions at these eleva-
tional extremes. Many of the species
that occur within creosotebush are
nocturnal, and, therefore, these ani-
mals avoid exposure to extreme sur-
face heat. On higher elevation habi-
tats, the problem is not avoiding heat
but, rather, gaining heat for activity.
Other than along rock outcrops,
rapid heating is difficult for reptiles
at higher elevations. Differences be-
tween diversity and species composi-
tion on medium elevation habitat
types probably reflect differences in
microhabitat abundance and diver-
sity on habitat types (see earlier dis-
cussion on microhabitats). Lack of
diversity on disclimax desert grass-
land sites probably reflects the lack
of vegetation structure on these sites.

There was similarity in the herpe-
tofaunas of certain habitat types. All
desert habitats, except sagebrush,
had very similar herpetofaunas, as
did most moderate elevation habitats
(e.g., chaparral, pinyon-juniper, and
mixed riparian scrub). This is pre-
dictable because all of these habitats
occur in close proximity and are
structurally similar. There was a
moderate degree of similarity be-
tween cottonwood-willow riparian
and desert habitats, chaparral and
cottonwood-willow riparian, and
chaparral and desert habitats. Be-
cause cottonwood-willow riparian
habitats traverse through both desert
habitats and upland habitats, many
of the species associated with the
surrounding habitats also frequent
riparian sites; riparian sites are im-

portant sources of food and cover
(Ohmart and Anderson 1986). Simi-
larities between chaparral and desert
habitat types, such as Mohave Des-
ert, Sonoran Desert, and mixed ripar-
ian scrub, result from occurrence of
typical desert species (e.g., Callisau-
rus draconoides) on upland sites rather
than the occurrence of upland spe-
cies (e.g., E. gilberti) on desert sites.

The diversity of and similarities
among amphibian and reptile com-
munities of habitat types also may
have been affected by the proximity
of habitat types to evolutionary cen-
ters. Because of the many new rec-
ords for herpetofauna generated by
this study, we now have a better pic-
ture of the sources of diversity for
this area. Many of the amphibians
and reptiles occurring in the Sonoran
and Mohave Deserts evolved in Baja
California and along the western sec-
tion of mainland Mexico; these areas
were linked until their separation 13
million years ago (Murphy 1983).
With the retreat of pleistocene glacia-
tion and spread of xerophyllous and
desert habitats, amphibians and rep-
tiles moved northward into southern
California and southwestern Ari-
zona; hence, Sonoran and Mohave
Desert habitat types have similar her-
petofaunas. Although many species
immigrated into what is today the
Sonoran and Mohave Deserts, only a
few species immigrated as far north
as the Great Basin Desert. Higher ele-
vations may have precluded many of
these species from colonizing the
Great Basin desert habitat types and,
hence, it's herpetofauna is different
from and less rich than those of the
other two deserts.

The discovery of the subspecies
Eumeces gilberti rubricaudatus, for-
merly unknown in Arizona, suggests
that Arizona chaparral was closely
associated with (Zalifomia chaparral
during Pleistocene glaciation; E. g.
rubricaudatus evolved in California
sclerophyll woodland (Taylor 1935).
That parthenogenic whiptail lizards,
such as Cnemidophorus flagellicaudis,
C. uniparens, and C. velox, are absent

from California chaparral suggest
that these species evolved after Pleis-
tocene glaciation.

There were a few inconsistencies
in the results of the two analyses
used to determine similarity between
habitats (the cluster analysis of all
data vs. the cluster analysis of only
array data). These inconsistences par-
tially result from the inconsistency of
arrays to capture turtles and medium
and large-sized snakes, and partially
from the analyses themselves (see the
Methods Section for a more detailed
explanation).

Conclusions and
Recommendations

This survey indicates that most spe-
cies present within western Arizona
are widespread, and that few war-
rant special management considera-
tion. However, it is evident that cer-
tain species are more vulnerable to
range or population reduction than
others. Generally, these species are
those that require microhabitats that
are easily affected by land uses.

It appears that habitat moisture
and moderated surface temperatures
are of primary importance to many
species in western Arizona. Downed
and dead surface litter (debris), such
as logs and leaves, play a major role
in moderating surface temperature
and enhancing moisture (Dauben-
mire 1974). Horizontal and vertical
vegetation structure also help moder-
ate temperatures and increase mois-
ture. In developing management
schemes, priority should be given to
maintaining or enhancing surface lit-
ter and vegetation structure. It is im-
portant to maintain tree reproduc-
tion, and to leave litter on the surface
rather than piling and burning it. The
latter practice is especially important
on cottonwood-willow riparian sites
within deserts, since many species in
riparian sites are totally dependent
on surface litter for their survival
(Jones and Glinski 1985). Many ripar-
ian sites within the study area have

125

reduced amounts of trees and sur-
face litter, principally because live-
stock have greatly reduced the repro-
duction of Cottonwood trees by re-
ducing the survival of seedlings
(Jones 1981a). Management prescrip-
tions are needed on these sites to in-
crease the survivorship of seedling
and young cottonwood trees.

Populations of ''upland" species
(e.g., Eumeces gilberti) on habitat is-
lands are more vulnerable to impacts
associated with certain land uses
than populations occurring on major,
continuous stands. Jones et al. (1985)
described these habitat islands, some
only 10 ha (25 acres) in size. Loss or
fragmentation of any portion of these
islands could result in the local extir-
pation of one or several upland spe-
cies (see Bury and Luckenbach 1983
and Harris 1984 for the effects of
habitat fragmentation and habitat
loss on species occurring on habitat
islands). Because even small modifi-
cations to island habitats can result in
the extirpation of upland species,
proposed projects should be moved
to alternative sites whenever pos-
sible; mitigation strategies should be
used only as a last resort. Top prior-
ity should be given to protecting
these sites in land-use and on-the-
ground activity plans (see Jones et al.
1985 for specific locations of these
sites).

Although all amphibians in the
study area (excluding Bufo mi-
croscaphus) appear to be stable, water
in many habitats continues to be de-
veloped. In addition, new informa-
tion (Bruce Bury personal comm.
Com and Fogleman 1984) suggest
that several populations of ranid
frogs have been extirpated from
western North America, although
there is no apparent cause for their
extirpation. Considering the heavy
use of spring and creek water, and
the reported loss of many ranid
populations in the West, high prior-
ity should be given to monitoring
amphibian populations at springs
and creeks in Arizona. Additionally,
high priority should be given to de-

termining the extent of hybridization
between the toads B. microscaphus
and Bufo woodhousei. Pure popula-
tions of B. microscaphus should be lo-
cated and protected against hybridi-
zation with B. woodhousei. If only a
few pure populations are found, the
Arizona Game and Fish Department
and/or the U.S. Fish and Wildlife
Service should set up a captive
breeding program to reduce this
toad's risk of extinction.

Although I obtained distributional
records of Gopherus agassizii, Burge
(1979, 1980) and Schneider (1980)
provide considerably more detail on
the needs of this species. However,
many biologists consider G. agassizii
to be declining throughout most of
its range. The U.S. Fish and Wildlife
Service (1987) continues to list G.
agassizii as a species that needs fur-
ther study to determine its status,
although it has determined that the
Federal listing of the tortoise
throughout its range is warranted
but precluded by species needing
more immediate listing (e.g., species
in more eminent danger of extinc-
tions). The BLM should continue to
give high priority to the study and
management of this species in Ari-
zona.

If the few measures suggested in
this paper are implemented, western
Arizona should continue to support
one of North America's most diverse
herpetofaunas.

ACKNOWLEDGMENTS

I am indebted to several people for
the completion of this project. Don
Seibert, Bob Furlow, and Ted Cor-
dery were instrumental in obtaining
funding, equipment, and personnel
for this study. Lauren Kepner, Tim
Buse, Dan Abbas, Terry Bergstedt,
Kelly Bothwell, William Kepner,
Dave Shaffer, Bob Hall, Ted Cordery,
Scott Belfit, Ted Allen, Ken Relyea,
Becky Peck, Brian Millsap, Jim Zook,
Jim Harrison, and Greg Watts helped
collect both animal and habitat data.

Special thanks to W.L. Minckley and
M.J. Fouquette for technical contribu-
tions to this project's study design,
and to the Bureau of Land Manage-
menf s line managers and supervi-
sors. Bill Barker, Roger Taylor, Barry
Stallings, Dean Durfee, Gary
McVicker, and Malcolm Schnitkner,
for their continuous support of re-
source inventories on public lands. I
thank John Fay, Scott Belfit, R. Bruce
Bury, and Robert Szaro for review of
this manuscript. Finally, all of us
who strive for the conservation of
nongame wildlife on public lands are
indebted to Gary McVicker, Bill
McMahan, and Don Seibert for their
tireless efforts in getting top-level
management to support nongame
programs.

LITERATURE CITED

Bennett, Stephen H., J. Whitfield Gib-
bons, and Jill Glanville. 1980. Ter-
restrial activity, abundance, and
diversity of amphibians in differ-
ently managed forest types.
American Midland Naturalist
103:412-416.

Brown, David E. 1978. The vegeta-
tion and occurrence of chaparral
and woodland flora on isolated
mountains within the Sonoran and
Mohave Deserts in Arizona. Jour-
nal of Arizona Academy Sciences
13:1-12.

Brown, David E., Charles H. Lowe,
and Charles P. Pase. 1979. A digit-
ized classification system for the
biotic communities of North
America, with community (series)
and association examples from the
Southwest. Journal of Arizona
Academy of Science 14, Suppl. 1,
p. 16.

Burge, Betty L. 1979. A survey of the
present distribution of the desert
tortoise (Gopherus agassizii) in Ari-
zona. U.S. Bureau of Land Man-
agement, Denver, Colorado. Con-
tract No. YA-512-CTB-108.

Burge, Betty L. 1980. Survey of the
present distribution of the desert

126

tortoise, Gopherus agassizii, in Ari-
zona. U.S. Bureau of Land Man-
agement, Denver, Colorado. Con-
tract No. YA-512-CT8-108.

Bury, R. Bruce. 1980. What we know
and do not know about off-road
vehicle impact on wildlife. In R.N.
Andrews and P.F. Nowah (eds.).
Off-road vehicle use: a manage-
ment challenge. U.S.D.A. Office of
Environmental Quality, 748 p.

Bury, R. Bruce and Stephen D.
Busack. 1974. Some effects of off-
road vehicles and sheep grazing
on lizard populations in the
Mojave Desert. Biological Conser-
vation 6:179-183. Bury, R. Bruce,
Roger A. Luckenbach, and
Stephen D. Busack. 1977. Effects of
off-road vehicles on vertebrates in
the California Desert. U.S. Fish
and Wildlife Service Wildlife Re-
search Report No. 8.

Bury, R. Bruce and M.G. Raphael.
1983. Inventory methods for am-
phibians and reptiles. Proceedings
of the International Conference on
Renewable Resources, Inventories
for monitoring changes and
trends. Oregon State University,
Corvallis.

Bury, R. Bruce and Roger A. Lucken-
bach. 1983. Vehicular recreation in
arid lands drives: biotic responses
and management alternatives, p.
217-221. In R.H. Webb and H.G.
Wilshire (eds.). Environmental ef-
fects of off-road vehicles: impacts
and management in arid regions.
Springer- Verlag, New York, New
York.

Buse, Timothy C. 1981. Distribution,
ecology, and habitat management
of the reptiles and amphibians of
the Cerbat planning unit, Mohave
County, Arizona. U.S. Bureau of
Land Management, Kingman, Ari-
zona. Unpubl. Man.

Buse, Timothy C. 1983. Herpetologi-
cal records from northwestern
Arizona. Herpetol. Rev. 14:53-54.

Campbell, Howard W. and Stephen
P. Christman. 1982. Field tech-
niques for herpetofaunal commu-
nity analysis, p. 193-209. In Norm

J. Scott (ed.), Herpetological com-
munities, U.S. Fish and Wildlife
Service, Wildlife Research Report
Number 13.

Case, Thomas J. 1983. The reptiles:
ecology. In T.J. Case and M.L.
Cody (eds.). Island biogeography
in the Sea of Cortez. University of
California Press, Berkeley.

Connor, Edward F. and Daniel Sim-
berloff. 1979. The assembly of spe-
cies communities: chance or com-
petition? Ecology 60:1132-1140.

Com, Paul Stephan and James C.
Fogleman. 1984. Extinction of
montane populations of the north-
ern leopard frog (Rana pipiens) in
Colorado. Journal of Herpetology
18:147-152.

Creusere, F. Michael and Walter G.
Whitford. 1976. Ecological rela-
tionships in a desert anuran com-
munity. Herpetologica 32:7-18.

Daubenmire, Rexford F. 1974. Plants
and environment: a textbook of
autecology. John Wiley and Sons,
New York, New York. 3rd Ed.

Fitch, Henry S. 1955. Habits and ad-
aptations of the Great Plains skink
(Eumeces obsoletus). Ecological
Monographs 25(3):59-83.

Reharty, Eugene D. 1967. Compara-
tive ecology of Thamnophis elegans,
Thamnophis cyrtopsis, and Thamno-
phis rufipunctatus in New Mexico.
Southwestern Naturalist 12(3):207-
230.

Hair, Jay D. 1980. Measurements of
ecological diversity, p. 269-275. In
S.D. Schemnitz (ed.). Wildlife
Management Techniques Manual.
The Wildlife Society, Washington,
D.C.

Harris, Larry D. 1984. Island biogeo-
graphy applied: old growth is-
lands and wildlife conservation in
the western Cascades. University
of Chicago Press, Chicago, Illinois.

Hulse, Arthur C. 1974. An autecol-
ogy study of Kinosternon sonoriense
leconte (Chelonia: Kinosternidae).
Phd. Dissertation, Arizona State
University, Tempe.

Iverson, John. 1978. Distributional
problems of the genus Kinosternon

in the American Southwest.
Copeia 1978:476-479.
Jones, K. Bruce. 1981a. Effects of
grazing on lizard abundance and
diversity in western Arizona.
Southwestern Naturalist 26(2): 107-
115.

Jones, Kenneth Bruce. 1981b. Distri-
bution, ecology, and habitat man-
agement of the reptiles and am-
phibians of the Hualapai- Aquarius
planning areas, Mohave and
Yavapai Counties, Arizona. U.S.
Bureau of Land Manage. Technical
Note No. 353, Denver, Colo.

Jones, K. Bruce. 1987. Amphibians
and reptiles, p. 267-290. In A.Y.
Cooperrider, R.J. Boyd, and H.R.
Stuart (eds.). Inventory and moni-
toring of wildlife habitat. U.S. Bu-
reau of Land Management, Den-
ver, Colorado xviii, 858 p.

Jones, K. Bruce., Dan R. Abbas, and
Terry A. Bergstedt. 1981. Herpeto-
logical records from central and
northwestern Arizona. Herpeto-
logical Review 12(1):16.

Jones, K. Bruce and Patricia C.

Glinski. 1985. Microhabitats of liz-
ards in a southwestern riparian
community, p. 355-358. In R. Roy
Johnson et. al.. Riparian ecosys-
tems and their management: rec-
onciling conflicting uses. First
North American riparian confer-
ence. Rocky Mountain Forest and
Range Experimental Station, Gen-
eral Technical Report Number
RM-120., Fort Collins, Colo.

Jones, K. Bruce, Lauren P. Kepner,
and William G. Kepner. 1983.
Anurans of Vekol Valley, central
Arizona. Southwestern Naturalist
28(4):469-470.

Jones, K. Bruce, Lauren P. Kepner,
and Thomas E. Martin. 1985. Spe-
cies of reptiles occupying habitat
islands in western Arizona: a de-
terministic assemblage. Oecologia
66:595-601.

Jones, K. Bruce, Lauren M. Porzer,
and Kelly J. Bothwell. 1982. Her-
petological records from westcen-
tral Arizona. Herpetological Re-
view. 13(2):54.

127

Justus, J. T., Mark Sandomir, Tom
Urquhart, and Barbara Orgel
Evan. 1977. Developmental rates
of two species of toads from the
desert Southwest. Copeia
1977:592-594.

Kahn, Walter C. 1960. Observations
on the effect of a bum on a popu-
lation of Sceloporus occidentalis.
Ecology 41(2):358-359.

Mayhew, William W. 1968. Biology
of desert amphibians and reptiles,
p. 195-356. In G.W. Brown, Jr.
(ed.). Desert biology. Academic
Press, New York, New York.

Murphy, Robert W. 1983. Paleobio-
geography and genetic differentia-
tion of the Baja California herpe-
tofauna. Occasional Papers Cali-
fornia Academic Sciences 137:1-48.

Norris, Kenneth S. 1953. The ecology
of the desert iguana, Dipsosaurus
dorsalis. Ecology 34:265-287.

Ohmart, Robert D. and Bertin W.
Anderson. 1987. Riparian habitats,
p. 169-200. In A.Y. Cooperrider,
R.J. Boyd, and H.R. Stuart (eds.).
Inventory and monitoring of wild-
life habitat. U.S. Bureau of Land
Management, Denver, Colorado
xviii, 858 p.

Oosting, Henry J. 1956. The study of
plant communities. W.H. Freeman
and Company, San Franciso, Cali-
fornia.

Pianka, Eric R. 1966. Convexity, des-
ert lizards, and spatial heterogene-
ity. Ecology 47:1055-1059.

Pianka, Eric R. and William S.
Parker. 1972. Ecology of the igua-
nid lizard, Callisaurus draconoides.
Copeia 1972:493-508.

Pimental, Richard A. 1979. Mor-
phometries: the multivariate
analysis of biological data. Ken-
dall/Hunt Publishing Company,
Dubuque, Iowa.

Platz, John E. 1984. Status report for
Rana onca Cope. Status report pre-
pared for the U.S. Fish and Wild-
life Service, Albuquerque, New
Mexico. 28 p.

Rogers, Thomas L. and Henry S.
Fitch. 1947. Variation in the skinks
(Reptilia: Lacertilia) of the skil-

tonianus group. University of
California Publications in Zoology
48(4):169-220.

Schneider, Paul B. 1980. A popula-
tion analysis of the desert tortoise
in Arizona during 1980. U.S. Bu-
reau of Land Management, Phoe-
nix, Arizona. Contract No. AZ-
950-279-0014.

Simovich, Marie A. 1979. Post fire
reptile succession. Cal-Neva
1979:104-113.

Stebbins, Robert C. 1966. A field
guide to western reptiles and am-
phibians. Houghton Mifflin Co.,
Boston, 1st Edition.

Stebbins, Robert C. 1985. A field
guide to western reptiles and am-
phibians. Houghton Mifflin Co.,
Boston, 2nd Edition.

Szaro, Robert C, Scott C. Belfit, and
J. Kevin Aitkin. 1985. Impact of
grazing on a riparian garter snake,
p. 359-363. In R. Roy Johnson et.
al.. Riparian ecosystems and their
management: reconciling conflict-
ing uses. First North American
riparian conference. Rocky Moun-
tain Forest and Range Experimen-
tal Station, General Technical Re-
port Number RM-120., Fort
Collins, Colorado.

Taylor, Edward H. 1935. A taxo-
nomic study of the cosmopolitan
scincoid lizards of the genus
Eumeces with an account of the
distributions and relationships of
its species. University of Kansas
Science Bulletin 23:1-643.

U.S. Fish and Wildlife Service. 1987.
Notice of findings on petitions and
initiation of status review. Federal
Register 52(126):24485-24488.

Van Devender, Thomas R. and Wil-
liam G. Spaulding. 1977. Develop-
ment of vegetation and climate in
the southwestern United States.
Science 204:701-710.

Walker, Robert F. and Walter G.
Whitford. 1970. Soil water absorp-
tion capabilities in selected species
of anurans. Herpetologica
26(4):411-416.

York, John C. and William A. Dick-
Peddie. 1969. Vegetation changes

in southern New Mexico during
the past hundred years, p. 157-166.
In W.G. McGinnes and B.J.
Goldman (eds.). Arid lands in per-
spective. University of Arizona
Press, Tucson.
Zweifel, Richard G. and Charles H.
Lowe. 1966. The ecology of a
population of Xantusia vigilis, the
desert night lizard. American Mu-
seum Novitates 2247:1-57.

128

Multivariate Analysis of the
Summer Habitat Structure of
Rana pipiens Schreber, in
Lac Saint Pierre (Quebec,
Canada)^

N. Beauregard^ and R. Leclair Jr^

Abstract.— Thirty stations representing various ripar-
ian habitats typical of the Lac Saint Pierre area were
sampled with a system of drift fences and funnel
traps to characterize the summer habitat structure
of a leopard frog population. A discriminant analysis
indicates that habitats with high frog density (1) ore
close to the marsh line, (2) have a tall herbaceous
stratum with high richness and (3) have a low moss
cover. A stepwise multiple regression model used 5
of the vegetation structure variables, and explains
CO. 70% of the variability associated with frog density
among stations.

The leopard frog, Rana pipiens, is the
most abundant frog species in the
Lac Saint Pierre area (Leclair 1985,
Leclair and Baribeau 1982, Paquin
1982), and also one of the most com-
mon vertebrates in aquatic communi-
ties in North America (Dole 1965a).
Despite this apparent abundance,
many herpetological surveys made in
the last fifteen years have shown dra-
matic reductions in leopard frog den-
sities. Gibbs et al. (1971) estimated a
50% drop in the global population of
leopard frogs in the USA, during the
1960's. Many other workers have re-
ported population reductions and
local extinctions in Canada and the
USA (Collins and Wilbur 1979, Cook
1984, Degraaf and Rudis 1983, Froom
1982, Hayes and Jennings 1986, Hine
et al. 1981).

In area where hypothesis of preda-
tion or competition by introduced
species (Bullfrogs or predatory
fishes) (Hayes and Jennings 1986)
does not apply, two major causes
have been invoked as responsible for

^ Paper presented at symposium. Man-
agement of Amphibians, Reptiles, and
Small t^ammals in North America. (Flag-
staff . AZ, July 19-21, 1988).

^Norman Beauregard is a graduate stu-
dent in Environmental Sciences, Universite
du Quebec a Trois-Rivieres, Department of
chimie-biologie, CP. 500 Trois-Rivieres, Que-
bec, Canada, G9A 51-17.

^Raymond Leclair Jr. is Professor of Her-
pefology and Ecology. Universite du
Quebec a Trois-Rivieres, Department of
chimie-biologie, CP. 500 Trois-Rivieres, Que-
bec, Canada, G9A 5H7.

this situation (1) overexploitation of
natural stocks, and (2) loss or altera-
tion of habitat rendering it unsuitable
for R. pipiens (Cook 1984, Frier and
Zappalorti 1984, Leclair 1985, Mar-
cotte 1981, Rittschof 1975). Riparian
habitats have been especially affected
by human activities (Sarrazin et al.
1983, MLCP 1985). In Canada, 50% of
the wetlands that once supported
wildlife have now been reclaimed for
agricultural, industrial or urban de-
velopment, or have been altered by
pollution (SCF 1980). Even greater
riparian habitat has occurred along
the St.Lawrence river, where 70% of
the riparian habitats have been elimi-
nated.

According to Ministere des
Loisirs, de la Chasse et de la Peche
(MLCP) (1985), essential habitats are
those vital to population or species
survival, whether these habitats are
used temporarily or permanently.
This definition emphasizes several
crucial aspects of amphibian habitat
use, i.e. the use of aquatic as well as
terrestrial habitats, and of migratory
routes between the two. Up to now,
quantitative studies of habitat re-
quirements for anuran species have
focused mostly on the aquatic habi-
tats (Beebee 1977, Clark and Euler

1982, Dale et al. 1984, Gascon and
Planas 1986, Hine et al. 1981). This
situation largely results from the lack
of appropriate quantitative sampling
method for amphibian populations in
terrestrial habitats (Bury and Raphael

1983, Clawson et al. 1984). Recently,

Campbell and Christman (1982), and
Vogt and Hine (1982) have devel-
oped adequate techniques that help
overcome this situation.

The aims of the present study
were (1) to characterize the structural
aspects (biotic and abiotic) of the ter-
restrial habitats of Rana pipiens and
(2) to develop a model relating frog
abundance to habitat descriptors.

Study Area

The study area is a 30 X 0.9 km strip
extending from Trois-Rivieres to Ber-
thierville (Quebec, Canada), on the
north shore of Lac Saint Pierre (73*30'
W X 46°05' N). The Lac Saint Pierre
covers about 300 km^ and is formed
by a widening of the St.-Lawrence
river (fig. 1). The lake flood plain is
extensive (Tessier et al. 1984) and
consequently, spawning sites for
amphibians are abundant in spring.
The habitats most frequently used by
Rana pipiens (based on mating call
frequencies) are flooded fields of
reed phalaris (Phalaris arundinacea)
and of purple loosestrife (Lythrum
salicaria), mixed with willow {Salix
sp.) (Leclair 1983). From these fields,
numerous bays, small rivers, drain-
ing canals and natural or man-made
pools facilitate movement of frog to-
wards adjacent terrestrial habitats.

According to the maps produced
by Denis Jacques (1986) and by
Tessier and Caron (1980) on the ri-
parian vegetation of Lac Saint Pierre,

129

at least ten plant communities may
be recognized on the criterion of
dominant species. These plant com-
munities can be grouped in six differ-
ent physionomic types (table 1).
Thirty stations were selected in order
to sample the diversity of habitats.
From the maps, sampling sites were
located in habitat patches not having
less than 2500 m^ of homogeneous
vegetation. The final choice of sites
was determined by physical and le-
gal accessibility.

Materials and Methods

Sampling Technique

At each station, frogs were sampled
with 12 funnel traps placed on each
side of two 15 m drift fences made of
polyethylene and forming a right
angle (fig. 2). Dirt and/or litter was
brushed into the mouth of each fun-
nel to simulate a natural entrance

(Clawson and Baskett 1982). This de-
sign has been shown to allow for
sampling in various kinds of terres-
trial habitats, and to provide data for
the estimation of demographic para-
meters and for comparison between
various habitats (Campbell and
Christman 1982, Clawson et al. 1984).

Funnel traps were opened for at
least 10 consecutive days in each pe-
riod (10 days in May, 10 in June, 12
in July, 10 in august, 22 in Fall) and
were checked every other day.

Data recorded for each capture
were: date, station number, direction
of capture (N, S, E or W), species, sex
and snout-urostyle length. Captured
frog were marked by clipping the
fourth digit of the hindfoot. Clipped
phalanges were kept for age determi-
nation trough skeletochronological
examination (Leclair and Castanet
1987).

Because of the way the arrays
were used for sampling, captures re-
flected the relative abundance of
frogs among stations, not their abso-
lute density.

Environmental Variables^

Each station was characterized by 6
spatial variables: distance to the
marsh line (DMARSH), to the nearest
permanent pool (DWATERP), to the
nearest temporary pool (DWATERP),

^See appendix 1 for all abbreviations
used in fhe text.

^Variable measured monthly.

^ 7m

h 15 m -I

Figure 2.— Schematic representation of the
trapping arrays.

Figure 1 .—General location of the study area (upper map) and the study area's relationship to
lake St. Pierre (Quebec, Canada) (lower map).

130

to the nearest human aUeration
(road, path, residence, crop) (DHU-
MAN), and to the nearest open habi-
tat without shrub or tree canopy
(DOPEN) or closed habitat with can-
opy (DCLOSE). All distances were
measured in the field with a topofil
marker (lost thread measure appara-
tus), except for some measures of
DMARSH taken from a 1:10 000 to-
pographical map. Elevation from the

marsh ground (ALTREL) was taken
with a Keuffel and Essel altimeter.
Water table level (WTABLE^) was
measured with a piezometer, placed
1 m deep.

Edaphic variables measured were:
soil moisture (MOIST^), from oven-
dried soil samples (80 °C, 24 hrs); soil
fractions (SAND, SILT, CLAY), as
determined by the Bouyoucos
method (Bouyoucos 1936); soil water

Table 1 .—Characteristics of the sampiing stations according to physiog-
nomic type and to major plant species.

Sta.

Physionomic type

Code

1

Open dry field

O

2

Brushy dry field

B

3

Wooded swamp

F

4

Riparian marsh

M

5

Shrub swamp

S

6

Wet prairie

P

7

Open dry field

O

8

Wooded swamp

F

9

Wet prairie

P

10

Wooded swamp

F

11

Wet prairie

P

12

Shrub swamp

S

13

Wet prairie

P

14

Shrub swamp

S

15

Riparian marsh

M

16

Wet prairie

P

17

Brushy dry field

B

18

Wet prairie

P

19

Brushy dry field

B

20

Open dry field

O

21

Open dry field

O

22

Wet prairie

P

23

Shrub swamp

S

24

Riparian marsh

M

25

Wooded swamp

F

26

Shrub swamp

S

27

Wooded swamp

F

28

Riparian marsh

M

29

Wet prairie

P

30

Wet prairie

P

Code Major plant species

Solidago canadensis, Aster umbellafus
Spirea lafifolia, Populus tremuloides
Acer saccharinum, Laporfea canad-
ensis

Sparganium eurycarpum, Scirpus flu-
viafilis

Spirea lafifolia. On ode a sensibilis
Calamagrosfis canadensis, Phalaris
arundinacea

Solidago rugosa. Aster umbellafus
Acersaccharium, Laporfea canaden-
sis

Carex lacustris, Lyft-)rum salicaria
Salix nigra, Laporfea canadensis
Typha lafifolia, Onoclea sensibilis
Salix spp . . Myrica gale
Calamagrosfis canadensis
Salix cordafa, Phalaris arundinacea
Sparganium eurycarpum, Equisetum
fluviafile

Phalaris arundinacea
Spirea lafifolia, Populus tremuloides
Carex lacustris, Lythrum salicaria
Spirea lafifolia, Salix ssp.
Solidago canadensis. Aster umbellafus
Phleum pratense, Agrostis alba
Calamagrosfis canadensis, Phalaris
arundinacea

Salix ssp . , Rorippa amphibia
Sparganium eurycarpum, Sagittaria
lafifolia

Acer saccharinum . Populus deltoides
Salix ssp . , Spirea lafifolia
Acer saccharinum, Onoclea sensibilis
Sparganium eurycarpum, Rorippa am-
phibia

Carex lacustris, Lythrum salicaria
Calamagrosfis canadensis, Lythrum
salicaria

pH (PH), as determined with a Fisher
pH-meter, and soil temperature
(TEMP2). The soil temperature vari-
able used in the statistics is ex-
pressed as the sum (5 reading per
month) of the deviations from the
daily mean taken over all stations.

Percent of ground covered by lit-
ter (LITTER5), dead wood (DEAD-
WOOD), mosses (MOSSCOV^), her-
baceous plants (HERBCOV^), and
percent bare ground (BAREGRND^)
was estimated by two independent
observers in 5 X 5 meters quadrat,
and the mean was recorded. Litter
thickness (LITTHICK^) and height of
the herbaceous stratum
(HERBHGHT^) represented the mean
of 5 measurements taken with a me-
terstick.

Quantitative assessment of vegeta-
tion structure was represented by
Fox's photometric index (Fox 1979)
as:

= In (la/lb)
H (b-a)

where represents the photometric
index for the amount of vegetation
present in a layer between two levels,
when la and lb are the light intensi-
ties immediately above and below
the layer and H(b-a), the layer thick-
ness. Readings of light intensity were
taken with a Sekonic light meter, at 0,
20, 50, and 100 cm above ground,
above the herbaceous canopy and in
the open field adjacent to stahon
having closed canopy. At each site,
measurements were taken at five
points which were then averaged to
provide one value. Five photometric
index were computed: vegetation in-
dex in the 0-20 cm layer (PHOT20^);
from 20 to 50 cm (PHOT505); from 50
to 100 cm (PHOTIOO^); herb layer
above 100 cm (PHOT+s); and shrub
and tree strata (PHOTCAN^).

Vegetation structure was also de-
scribed in 8 growth-form categories:
(TREE) woody plants > 10 cm diame-
ter; (SHRUBHI) woody plants > 2.5
m tall); (SHRUBLO) woody plants <
2.5 m tall; (HGH') high graminoid
herbs > 100 cm tall; (MGH^) medium

131

size graminoid herbs from 20 to 100
cm tall; (HBLH^) high broad-leaf
herbs > 100 cm tall; (MBLH^) me-
dium size broad-leaf herbs from 20 to
100 cm tall; (SMALL^) herbs layer
below 20 cm tall. Basal area (BA-
SARE A) was calculated by measur-
ing tree diameter at breast height
with a caliper. Richness in herbs spe-
cies (NSPHERB), shrubs
(NSPSHRUB), and trees (NSPTREE)
was determined in a 400 m^ quadrat.
Minimal area of homogeneous vege-
tation patch (MINAREA) was esti-
mated according to the graphical
method of Braun-Blanquet (1964).

Statistics

Spearman rank correlations and chi-
square tests were used to test for
non-random distribution of captured
frogs among age class and among
periods of sampling. Chi-square tests
were also used to detect a significant
movement of frogs. Because some
variables were not normally distrib-
uted (Kolmogorov-Smirnov test),
they were square-root transformed
before analysis (indicated on appen-
dix 1).

For final analyses, the number of
variables was reduced by screening
an initial principal component analy-
sis (PCA), and by using Pearson rank
correlations (Green 1979). Because of
heterogeneity in the variables meas-
ured, the correlation matrix was used
to extract the principal components
that explained the greatest propor-
tion of variability. A second PCA
with the 22 extracted variables
served to define the structural differ-
ences among stations, and to reduce
the data set to a few important di-
mensions that could identify most of
the structural variability among
measured habitats. To construct a
classification model for potential
habitats, a discriminant analysis
(DFA) was done on three groupings
of stations based on frog abundance.
The model was validated through a
simulation.

A stepwise multiple regression
was used to identify which habitat
characteristics account for most of
the variability in the analyzed data
(Clawson et al. 1984). An independ-
ent variable was included in the
model when its partial F-value was
significant (a = 0.05). Partial correla-
tion coefficients were used to verify
the statistical relation between the
dependent variables and the inde-
pendent one. This analysis has been
identified as the most appropriate to
study the combined effects of various

habitat variables on wildlife density
(Legendre and Legendre 1984). Inter-
pretation of the models obtained
from such analyses takes into ac-
count combinations of variables, but
not variables taken individually (Sch-
errer 1984). Statistical analyses were
performed with SPSS (Nie et al.
1975).

Results

A total of 798 individuals represent-
ing 4 species of anurans (Ram pipiens,

r

Table 2.— Capture data by sampling period, and by age class.

Number of captures

By month

By age class

Sum

Sum^

WV4I 1 1

Station

M

J

J

A

S-O

Adult Juv. NMY'

adjusted

1

0

1

0

2

2

4

0

1

5

5

2

0

0

3

2

6

5

3

3

11

11

3

1

0

0

1

3

2

2

1

5

5

4

8

n

5

3

7

11

17

6

34

34

5

6

3

4

2

2

15

1

1

17

17

6

7

13

3

3

9

20

11

4

35

35

7

0

1

1

3

1

5

0

1

6

6

8

0

0

1

0

2

0

2

1

3

3

9

8

8

4

2

5

14

7

6

27

27

10

3

9

3

2

2

12

3

4

19

19

11

5

3

1

2

8

7

4

8

19

19

12

11

9

4

4

3

21

3

7

31

31

13

15

8

7

12

5

28

13

4

47

47

14

22

7

10

39

47

16

12

78

107

15

12

15

8

12

39

51

21

11

86

86

16

3

0

2

0

3

1

1

5

7

17

0

0

0

0

0

0

0

0

0

0

18

0

0

0

0

0

0

0

0

0

19

0

0

3

0

0

3

0

0

3

3

20

0

0

0

0

0

0

0

0

0

0

21

0

0

3

1

1

0

0

5

5

5

22

5

2

6

2

0

8

4

0

15

15

23

10

10

2

1

0

9

13

0

23

23

24

26

21

3

2

7

28

23

1

59

59

25

8

3

1

4

3

17

2

0

19

19

26

5

1

1

3

7

12

1

3

17

17

27

0

1

4

9

6

3

5

14

19

28

2

10

12

10

20

3

n

34

47

29

5

2

0

1

1

3

5

0

9

9

30

3

5

6

7

11

3

7

21

29

Total

135

150

86

98

178

362

161

103

647

704

^ Newly metamorphosed young.

'Sum adjusted for stations not inventoried in May.

132

R. catesheiana, R. sylvatica, Bufo ameri-
canus) were captured during the
study. Many small rodents (n = 188)
and a few weasels (Mustela ermim)
were also captured. The results pre-
sented here relate only to R. pipiens.

Table 2 presents the capture data
for the various stations and sampling
periods, along with data on popula-
tion age structure. The mean capture
rate is 0.35 capture/day/ station; sta-
tions range from 0 to 1.77 captures/

day/ station. Preliminary trials on
three stations in fall 1986 had given
4.8 captures/day/station.

Spearman's correlation coefficients
(table 3) from among all possible age
groups and sampling period pairs
were all significant except those be-
tween captures at period 1 and
newly metamorphosed young (R =
0.3471, P = 0.097). We also compute a
contingency table (table 4) to check
for independence of the two vari-

Table 3.— Spearman rank correlations and signifiance level between cap-
tures for all possible age groups and sampling periods pairs.

Periods

Age

Period/Age

M

J

J

A

s-o

Adult

Juv

NMY'

May

«*«

•♦»

NS

June

0,88

»

July

0.59

0.57

««

***

August

0.63

0.56

0.64

♦ ♦•

**•

Sep.-Oct,

0.52

0.47

0.52

0.75

•••

Adult

0.88

0.83

0,74

0.81

0.67

Juvenile

0.82

0.79

0.59

0.49

0.60

0.70

♦

NMYl

0.35

0.44

0.62

0,70

0,81

0.56

0.46

'Newly metamorphosed youngs.

'P<0.05.

"P<0.01

'"P< 0.001.

NS = non significant.

Table 4.--Contingency table for non random distribution of age group cap-
tures among the physlonomic types of habitat.

Physionomic types of habitat

Age

Dry

Shrub

Wooded

Wet

Riparian

groups

habitat'

swamp

swamp

prairie

marsh

NMY

10

23

11

30

29

Count

4.9

26.5

9.9

28.3

33.4

Exp. vol.

Juvenile

3

34

12

48

64

Count

7.7

41.4

15.4

44.2

52.2

Exp. vol.

Adult

17

104

37

94

no

Count

17.3

93.1

34.7

99,5

117.4

Exp. vol.

For all habitats: D.F. = 8. X' = 16.64, 0.025 <P< 0.05.
Without dry habitats: D.F. = 6. = 6.04. 0.10 <P< 0.25.

'Open dry field and Brushy dry field were joined to respect chi-square require-

ments.

ables ''age group" and "physionomic
type of habitat." There was a weak
relationship (D.F. = 8, = 16.64,
0.025 < P < 0.5) created mostly by the
capture of a few young (n = 5) at a
dry open field station (# 21, see table
2). Otherwise, all other habitats
shared proportional distribution for
the different age groups (D.F. = 6,
= 8.04, 0.10 < P < 0.25). Further gen-
eral PCA and DFA models used the
number of total captures per station,
irrespective of sampling periods or
age groups.

Following preliminary screening,
we removed variables that were not
normally distributed (DOPEN,
DCLOSE, DEADWOOD, HBLH,
SMALLH), those correlated with
other variables (DWATERP, LITTER,
SAND, WTABLE), and those related
to the tree and shrub strata
(NSPTREE, NSPSHRUB, TREE,
SHRUBHI, SHRUBLO, BASAREA,
PHOTCAN) which diluted the re-
sults of PCA.

Figure 3 illustrates the distribution
of the remaining 22 variables along
the first two PCA axes, based on data
of table 5. The first axis explains
22.3% of the variation and is corre-
lated to descriptors of vegetation
structure, such as density of grami-
noids (HGH), vegetation height
(HERBHGHT), photometric index

2 «s»)

. M.TRB.

PHOT.

MOSSCOV

H — 1 — ^

' PH071»

your

Ohuuan'

Figure 3.— Projection of the 22 biophysical
variable vectors onto plane defined by the
first two principal components. The circle at
the origin has a radius of 0.30.

133

(PHOT20, PHOT+), litter thickness
(LITTHICK) and moss cover
(MOSSCOV). The second axis ex-
plains 15.2% of the variability and is
correlated to marsh distance
(DMARSH, ALTREL), number of
herb species (NSPHERB), mid-height
graminoids (MGH), short distance to
human alteration (DHUMAN) and
bare ground (BAREGRND). The
third axis explains 10.5% of the data
variability, which is significant ac-
cording to the broken stick model
(Frontier 1976, Legendre et Legendre
1984). It is related to edaphic factors
such as: pH (PH), silt fraction (SILT)
and soil moisture (MOIST) (table 5).
The forth and subsequent axes are
not significant.

Figure 4a gives the relative posi-
tion of stations according to the first
two axes of the PCA. Five groups

may be easily circled at best, accord-
ing to their physionomic type. Dry
habitats (open and brushy fields) are
at the top of the figure and are char-
acterized by a greater distance to the
marsh line, a higher moss cover and
a plant cover which is meager but
has a high species diversity. The dry
open fields with high PH(DT+ are dis-
tinct from the dry brushy fields
which have a lot of bare ground,
those two variables being in opposite
direction (fig. 3). Although the vari-
ables on tree and shrub strata were
removed from PCA, wooded and
shrub swamps appear distinct from
the other habitats. They are clustered
along the BAREGRND and MBLH
vectors (fig. 3) opposed to variables
describing vegetation structure and
positively correlated to the first axe.
Wet prairies and riparian marshes

can be differentiated from the other
three habitats along the first axis by a
more elaborated herbaceous struc-
ture. Stations positioned in the Spar-
ganium eurycarpum community,
which occupies approximately the
first 100 m of the riparian marsh
(Tessier et a. 1984), have a very wet
soil, the water receding only about
the end of May. Wet prairies are dis-
tinguished from the preceding habi-
tat by the conjugated differences of
many variables related to axe 2.

Figure 4b shows the position of
the stations as in figure 4a but are
best circled by classes of frog abun-
dance. This figure emphasizes the
relationship between habitat aridity
and frog density, the lowest frog
densities occurring in the driest habi-
tats (open dry field, brushy dry
field). Higher frog density stations
include those from the marsh line
and those from the wet fields. Inter-
mediate frog densities occur in forest
and shrub sites.

A DFA of frog density classes al-
lowed us to identify a few variables
that were easy to quantify and also to
classify habitats according to their
potential use by leopard frogs. Table
6 presents the standardized coeffi-
cients (computed with z-score) of the
variables for each DFA axis, and
non-standardized coefficients associ-
ated with classification function.
Four such variables were retained
from DFA. DMARSH alone allows
for 60% of the stations to be correctly
classified. Addition of the NSPHERB
variable adds another 13%. When
PHOT+ and MOSSCOV variables
were used, 90% of the stations were
correctly classified.

Figure 5 integrates information
about habitat and density by indicat-
ing the position of each station and
group centroids of frog density
classes along the two canonical axes.
As for PCA, the value of the stan-
dardized coefficients for each vari-
able associated to each DFA axis is
proportional to the length of each ar-
row. The first axis, which represents
the major part of the interclass vari-

r \

Table 5.— Sorted factor loadings for ttie principal component analysis of
habitat variables.

Factor

1

2

3

(22.3%)'

(15.2%)

(10.5%)

HGH

0.845

0.094

0.271

LITTHICK

0.841

0.102

-0.119

MOSSCOV

-0.772

0.180

0.139

HERBHGHT

0.604

0.144

0.178

MINAREA

-0.602

0.158

0.184

PHOT20

0.586

0.266

-0.016

HERBCOV

0.536

0.040

-0.413

PHOT+

0.845

0.400

0.250

ALTREL

-0.347

0.753

0.087

DMARSH

-0.394

0.738

-0.166

NSPHERB

-0.006

0.654

0.464

MGH

-0.067

0.578

-0.125

DHUMAN

-0.146

-0.566

0.050

BAREGRND

-0.444

-0.537

0.143

SILT

-0.098

0.042

0.696

PH

0.097

0.264

-0.640

MOIST

0.481

-0,313

0.509

TEMP

-0.108

0.404

-0.208

PHOTIOO

0.490

-0.090

-0.384

PHOT50

0.452

0.030

0.409

MBLH

-0.337

-0.405

-0.077

CLAY

-0.088

0.214

0.314

'Percentage of total variance explained by each component.

V : : : . . ■ - - ^^^^^^^^^^^^^^^^^ ^^^ ^^ ^ ^

134

Table 6.— Summary statistics for discriminant function analysis of habitat
characteristics according to three classes of frog abundance (as defined
In fig. 4b).

Variable

Wilk's
lambda

% correct Standardized
classification coefficients^
total

Unstandardized
coefficients'*

Axe 1
88.8%

Axe 2

n.2%

Axe 1

Axe 2

DMAf^H

0,505

0.0001

60.0

1.220

0.065

0.00613

0.00033

NSPHERB

0.328

<0.0001

73.3

-1.024

0.505

-0.143

0.071

PHOT+

0,255

<0.0001

73.3

0.537

0.709

0.584

0.771

MOSSCOV^

0,214

<0.0001

90.0

0.500

-0.185

0.480

-0.179

(CONSTANT)

-0.908

-1.522

'DFA useszscore data and gives the relative contribution of each variable to final
discrimination.

^Classification fonctlon uses original data and allows to know to which group sam-
pling stations belong, DMARSH expressed in meter. NSPHERB espressed in number of
herb species, MOSSCOV espressed in % ground cover.

ability (88.8%), is mostly related to
DMARSH, NSPHERB and
MOSSCOV. The second axis (11.2%
of intergroup variation) reflects pri-
marily variation in the photometric
index above 1 m (PHOT-»-) and
NSPHERB.

To validate our discriminant
model, we randomly drew 103
samples of three groups of stations,
and ran a DFA. The distribution of
the 103 samples does not depart sig-
nificantly from normality
(Kolmogorow-Smirnov test = 1.233, P
= 0.096). The results give a mean of
correct classifications of 68.6% with a
maximum of 83.3% and a standard
error of 7.2%. A t-test (T = 2.98, P(_l)
= 0.0025) indicates that the probabil-
ity of obtaining a value equal to 90%
is less than 0.0025.

Finally, using stepwise multiple
regression analyses, we identified
those variables used in models that
best predict frog abundance. For
such modelling, Clawson et al. (1984)
have pointed out the importance of

L. \ 1 U

-2.0 -1.0 0 1.0 2.0

FACTOR 1

Figure 4a.— Ordination of the sampling sta-
tions in the plane defined by the first two
principol components according to station
physiognomy.

incorporating phenological aspects of
habitat utilization. Directions of cap-
tures (table 2) were then analyzed in
order to group the capture data in
different periods of activity based on

■ \ \ — *

•2.0 -1.0 0 1.0 2.0

FACTOR 1

Figure 4b.— Ordination of the sampling sta-
tions in the plane defined by the first two
principle components according to frog
abundance. 1: number of capture < 9; 2: 8
< number of capture < 26; 3: number of
capture > 25.

seasonal patterns of movement (i.e.
movement away from aquatic over-
wintering sites in Spring, movement
within a summer foraging range, and
movement towards aquatic overwin-
tering sites in Fall).

Chi-square values (table 7a)
showed significant movement for
period 1,2 and 5. Individuals cap-
tured in the Fall seem to move back
towards the lake where they pre-
sumably overwinter. A stepwise re-
gression model associated with this
period (model 3) would thus charac-
terize habitat used during Fall migra-
tion. Although we got significant chi-
square in early season (sampling pe-
riods 1 and 2), interpretation is
doubtful whether or not there was a
migration movement from the over-
wintering site (i.e. from south and
east). To test for an actual movement,
we associated the two compass di-
rections in the general direction to-
wards the overwintering site and we
tested them against the two compass
directions in the general direction
away from the overwintering site
(i.e. north and west). No significant
movement was then noted (table 7b).
Consequently, we referred to the
phenology of the leopard frog de-
scribed by Dole (1967) and Rittschof

135

(1975) to decide for grouping of sam-
pling periods.

In May, as leopard frogs remained
at proximity of their reproductive
site and because we had only 24 sam-
pling stations at that time, data from
period 1 were analyzed separately
(model 1). Data from June, July and
August (periods 2, 3 and 4) were
grouped together to construct a
single model (model 2) because in
June individuals normally tend to
disperse in their summer foraging
habitats (Rittschof, 1975), and in July
and August no definite movement
direction was observed (that is typi-
cal when foraging habitat is occu-
pied). We also analyzed the data for
all periods in two general models
(models 4 and 5).

Model 1 (table 8) explains ca. 82%
of the variation in frog density for
the month of May using 6 variables.
The first one is distance to marsh

F
U
N
C
T

Table 7a.— Capture data by sampling period and by direction and chi-
square values for tests of goodness of fit. P values „ 0.05 are considered

significant.

Month

North West

South

East

Exp.vaiue

P

May

45 24

24

41

33.50

9.42

<0.025

June

42 28

27

52

37.25

11.56

<0.010

July

26 T9

15

25

21.25

3.80

>0.25

August

19 24

31

23

24.25

3.08

>0.25

Sep.-Oct.

106 13

26

33

44.50

117.96

<0.001

Table 7b.-

Results of test for nonrandom distribution of captures among the

two general directions of movement from and away overwintering sites.

Month

North + West

South + East

P

May

69

65

0.119

0.067

>0.75

June

70

79

0.272

0.215

>0.50

July

45

40

0.294

0.188

>0.50

August

43

54

1.247

1.031

>0.25

Sep.-Oct.

119

59

20.224

19.556

<0.001

X'. = Chi-square wifh Yates correction for cor^tinuity.

0
N

S

c

0
R
E

2.0--

1 .0--

0 --

-1,0-

NSPHERB

DM ARS

,'22

3.0 - 2.0 - 1,0 0 1.0 2.0 3.0

FUNCTION SCORE 1

Figure 5.— Localization of the sampling stations (represented by tfieir abundance class) in
\he discriminant space according to their function score. The relative contribution of each
variable (NSPHERB, PHOT+, DMARSH AND MOSSCOV) involved in the two discriminant func-
tions is indicated by the length of each vector. Class centroids are represented by *. Mis-
classified stations are circled.

line. Four of the five other variables
are related to soil characteristics:
temperature, moisture, silt fraction
and bare ground. In model 2 (sum-
mer feeding habitats), about 70% of
variation in frog density is explained
by only three variables: distance to
marsh line, number of herb species
and clay fraction. The third model,
for the month of September and Oc-
tober, explains only 34.6% of vari-
ation in Fall captures with two vari-
ables: DMARSH and NSPHERB. It
should be noted that the same two
variables explain 61.5% of the vari-
ation in model 2.

In the next two models (table 8)
the seasonal captures were corrected
to account for the lower number of
stations sampled in May. Model 4
includes five variables: DMARSH
and NSPHERB again, and three vari-
ables related to vegetation structure
(PHOT+, PHOT20, PHOT50). These
last three variables explain an addi-
tional 21.6% of the variation in frog
density in the model.

Hooding of St. Lawrence river
over our study sites is a major mani-
festation in the Lac Saint Pierre area

136

having a strong impact on frog distri-
bution as indicated by the presence
of the variable DMARSH in all previ-
ous models. However, when water
recesses, we get a mosaic of habitats
that can be found elsewhere in North
America but independently of the
presence of such marsh line. That is
the reason why we ran another mul-
tiple regression (model 5) after hav-

ing removed DMARSH. This last
model emphasizes the significance of
vegetation, all 5 variables included
being related to vegetation structure.
This model explains 69.2% of the
variation in total captured frogs.

To facilitate the understanding of
our interpretation, we present in ap-
pendix 2 the significant level of the
Pearson rank correlations between

Table 8.— Multiple regression models for frog captures.

Variable

Coefficient

Probability

Adjusted

(p ±SE)

(a value for F)

Model 1 Capture in May (24 stations)

(Intercept)

-2.51 ± 3.60

0.4950

0,4950

DMARSH

-0.0116 ± 0.0027

0.0005

0,484

TEMP

0.176 ± 0.050

0.0027

0,574

MOIST

0.230 ± 0.053

0.0004

0.636

PHOT50

-1.430 ± 0.429

0.0039

0.691

SILT

0.284 ± 0.074

0.0013

0.770

BAREGRND

-0.999 ± 0.419

0.0290

0,818

Model 2 Captures in June, July and August

(Intercept)

6.05 ± 2.99

0.0532

DMARSH

-0.0355 ± 0.0044

0.0000

0.298

NSPHERB

0.649 ± 0.153

0.0002

0.615

CLAY

0.151 ± 0.051

0.0068

0.700

Model 3 Captures in September and October

(Intercept)

1.21 ± 3.54

0.7358

DMARSH

-0.0196 ± 0.0056

0.0016

0.121

NSPHERB

0.663 ± 0.204

0.0030

0.346

Model 4 Adjusted total captures

(Intercept)

-6.80 ± 8.03

0.4053

DMARSH

-0.0600 ± 0.0099

0.0000

0.351

NSPHERB

1.745 ± 0.362

0.0001

0.553

PHOT+

-14.859 ± 3.001

0.0000

0.607

PHOT20

4.100 ± 1.274

0.0037

0.693

PHOT50

4.804 ± 1.576

0.0055

0.769

Model 5 Adjusted total captures*

(Intercept)

17.64 ± 11.37

0,1339

HGH

20.307 ± 2.732

0.0000

0.196

LITTHICK

-6.275 ± 1 .264

0.0000

0,362

PHOT+

-14.060 ± 3.081

0.0001

0.450

PHOTCAN

-5.234 ± 1.374

0,0009

0.631

MBLH

5.195 ± 2.135

0,0228

0.692

"DMARSH removed from the model 4.

the variables used in the models (1 to
5 and DFA) and all other variables
measured in the field.

Discussion

Model-Related Assumptions

In order to use density (estimated by
captures) as the dependent variable
in multivariate analysis to model sea-
sonal habitat structure selected by
leopard frogs, certain assumptions
must be made. Moreover, we cannot
recommend the use of the models
presented in table 8 to predict den-
sity for leopard frog populations for
which the pattern of seasonal fluctua-
tion and causes of those fluctuations
are unknown (Clawson et al. 1984,
Hine et al. 1981).

1. Density as estimated by cap-
ture reflects density in the
sampled habitats as regards
to immigration or emigration
to or from neighboring habi-
tats (Collins and Wilbur
1979). Ram pipiens is known
to be very mobile (Merrell
1977, Rittschof 1975), and is
capable of nocturnal excur-
sions of 100 m or more (Dole
1965a). Nevertheless, leopard
frogs rarely move more than
10 m away from their home
range, estimated by Dole
(1965b) to vary between 68
and 503 m^.

2. Favorable habitats are char-
acterized by frog densities
that are higher than those in
unfavorable habitats (Par-
tridge 1978). However, if
density is low (as observed
on our study site in 1987
when compared to 1986), all
favorable habitats may not
be occupied (Partridge 1978).

3. Multivariate analyses are
based on matrices of linear
correlation between environ-

137

mental variables and an in-
dex of abundance (Legendre
and Legendre 1984), which
neglects saturation and nega-
tive feedback effects, as well
as non-linear patterns in the
species response to environ-
mental factors.

4. Competition and predation
or the presence of sites for
reproduction may control
frog distribution patterns but

active habitat selection with
respect to vegetation struc-
ture also plays an important
role. Dole (1971) has ob-
served that newly metamor-
phosed young do not neces-
sarily select the first suitable
site during dispersal.

Finally, in models, it is apparently
essential to assume that factors vital
for species survival, i.e. those vari-
ables actively selected by individu-

Figure 6.— Ordination of the stations in relation to marsti distance (DMARSH) and number of
hierb species (NSPHERB). Stations withi thie same abundance class are circled by an ellipsoid

als, and those identified by the analy-
sis do not necessarily coincide. In
fact, apparent cause-and-effect rela-
tionships are not often testable and
require specific study on the func-
tional responses of species to the se-
lected variables. Weller (1978) indi-
cates that the study of habitat stimuli
as attractants for wildlife remains to
be done. The approach used in this
study is valuable when variables de-
scribing favorable habitat are re-
quired (Clark and Euler 1982, Green
1971,Grier 1984).

Classification of Habitats

The PCA analysis facilitated under-
standing of the multidimensional
models, and so allowed for system-
atic description of the various habi-
tats found in the Lac Saint Pierre
floodplain. We found that our pre-
established groupings were not an
analytical artefact but rather con-
firms that there is a structure that can
be defined by environmental vari-
ables not related to species specific
local vegetation.

Our results have shown that dif-
ferent age groups of R. pipiens are not
differently distributed among habi-
tats (tables 3 and 4). This conclusion
have been drawn with recently meta-
morphosed young representing only
16% of total captures but is sup-
ported by others studies describing
the habitats used by young (Dole
1971, Hine et al. 1981, Rittschof 1975,
Whitaker 1961). Our proposed mod-
els are those independent of age or
size groups. This might not be the
same however, for other species as
Clark and Euler (1982) and Roberts
and Lewin (1979) have noted for
Ram clamitans and for R. sylvatica,
respectively.

The models presented in this pa-
per reveal the importance of distance
to marsh line in habitat classification.
This variable has a high degree of
predictive power as to the extent
habitat will be utilized by leopard
frogs, in the Lac Saint Pierre

138

floodplain. However, systematic
sampling in habitats of unknown
value indicates the presence of a sig-
nificant number of leopard frogs in
some wooded and shrub swamps
stations far from the marsh (fig. 6).
The DFA model is then relevant to
show the importance of variables re-
lated to structural components of
habitat such as herbaceous vegeta-
tion (PHOT+, NSPHERB) and moss
cover (MOSSCOV). In a similar
analysis on Missouri herpetofauna,
Clawson et al. (1984) concluded that
proximity to water appeared to over-
ride other variables in determining
the abundance of amphibians.

Other multivariate studies (Beebee
1977, Clark and Euler 1982, Dale et
al. 1985 and Gascon and Planas 1986)
on anuran species habitat have
shown that bio-physico-chemical
variables related only to the body of
water cannot give predictive infor-
mation about the absence or presence
of a respective amphibian species.

Frog Abundance Models

In spring, before the growing season,
frog distribution is related to soil
characteristics, such as temperature.
This variable is not significantly cor-
related with any other variable meas-
ured. It results from the interaction
of many variables and may be a key
element in habitat selection during
that period. The activity of ectoth-
erms is known to be related to ambi-
ent temperatures (Putnam and Ben-
nett 1981), by selecting warmer habi-
tat, ectotherms might improve their
mobility, thus escaping more easily
to predators. Soil moisture is the
third most important variable in the
first model and appears only in this
model. In spring, soil moisture re-
flects the speed of water recess after
snowmelt and obviously is a variable
linked with the proximity of over-
wintering and spawning sites.

The model proposed for the sum-
mer period is the simplest of the
models presented in this paper with

only 3 descriptors (DMARSH,
NSPHERB, CLAY). Soil moisture is
not included into this model al-
though it has been shown to be the
major factor limiting the distribution
of anuran species in terrestrial habi-
tats (Clark and Euler 1982, Dole
1965a, 1971, Rittschof 1975, Roberts
and Lewin 1979). It may be that this
variable contains an information al-
ready carried in DMARSH variable;
its presence in the summer model
would then be a redundancy. Clay,
on the other hand, is a variable
known to play an important role in
soil water retention (Ramade 1984).

Sampling during Fall migration
have shown a significant movement
towards aquatic overwintering sites.
Model 3 however, with two variables
explaining only 34.6 % of frog abun-
dance, did not allow identification of
preferred migratory corridors. It
seems that leopard frogs en route to
overwintering sites do not select any
particular pathway.

The last two models use data from
all sampling periods. Model 4, which
improves on model 2 (summer
model), is interesting because its
photometric variables are signifi-
cantly correlated (appendix 2) with
many of other variables describing
the habitat structure. This suggests
the value of such indices (Fox 1979)
in habitat modeling to quantify vege-
tation structure since they can be
measured with an instrument (light
meter) easy to use.

The last model, with 69.2% vari-
ability explained, is of more general
interest because the local variable
DMARSH has been removed. In
model 5, the importance of vegeta-
tion structure in habitat selection is
obvious, and the model can be ap-
plied to the entire distributional
range of R. pipiens. HGH indicates
the importance of graminoids
(grasses, sedges, etc.) usually abun-
dant in open wetlands. This vegeta-
tion cover provides a refuge from
many predators and may thus con-
tribute to maintaining an abundant
frog population (Whi taker 1961). Lit-

ter thickness has a negative coeffi-
cient in the model, but is positively
correlated with HGH, which sug-
gests the existence of an optimum
foliage density. Dole (1965b, 1967,
1971) mentions that litter may pre-
clude direct contact betv-^een the in-
dividual and the moist substrate and
thus cause higher cutaneous evapo-
ration. The three other descriptors
summarize the information on vege-
tation structure. PHOT+ corresponds
to the presence of broad-leaf herbs >
100 cm tall (Rp = .4066, P = 0.026),
and graminoids (Rp = .3765, P =
0.040); PHOTCAN represents tree
and shrub cover; MBLH indicates
broad-leaf plant obstruction between
20 and 100 cm from the ground.

These results seem to indicate that
vegetation structure, more than spe-
cific species composition, is an im-
portant factor in habitat selection for
Rana pipiens. This finding is similar
to that of MacArthur and MacArthur
(1961) who have demonstrated that
bird species occupying forests and
prairies choose their habitat on the
basis of foliage density at different
levels from the ground, irrespective
of plant species composition.

Conclusion

In summary, we present three types
of complementary analysis dealing
with wet habitats used by the leop-
ard frog during Summer. First, a
PCA gives a qualitative description
of five kinds of habitats typical to the
St. Lawrence river floodplain and
offering potential supports to leop-
ard frog populations. Second, a DFA
model with four easily measured
variables allows classification of
habitats into three groups of frog
abundance. This is a very helpful
way to map potential frog species
habitats for protective purpose. Fi-
nally, five regression models (accord-
ing to each phenological periods or
whole active season) explain frog
abundance variations with only a
few important structural variables.

139

Although the models described in
this paper cannot fully demonstrate
functional relationships between
model variables and frog density,
suitable modifications of some of
these variables (litter thickness, for
instance) may increase frog popula-
tion. Refinement of these models will
require experimental studies on func-
tional responses of leopard frogs to
specific habitat features.

Acknowledgments

Thanks are due to Benoit Levesque,
Sylvain Cote and Jean-Louis Benoit
for field assistance, to Bernard
Robert for graphical art, to Gille
Houle for the English version of the
text and to Marc P. Hayes and Gary
K. Meefe for their very constructive
comments on the first draft of this
paper. Financial support came
through grants to N.B. from National
Research Council of Canada, Cana-
dian Wildlife Federation, and Centre
d'Etude Universitaire (Quebec) and
from direct funding from Ministere
Quebecois du Loisir, de la Chasse et
de la Peche, and Universite du
Quebec a Trois-Rivieres.

Literature Cited

Beebee, Trevor J.C. 1977. Habitats of
the British amphibians (l):Chalk
uplands. Biological Conservation
12:179-193.

Bouyoucos, G.J. 1936. Directions for
making Mechanical Analysis of
Soils by the Hydrometer Method.
Soil Science 32:225-228.

Braun-Blanquet, Josias. 1964. Plant
sociology. McGraw-Hill, New
York. 865 p.

Bury, R. Bruce and Martin G. Ra-
phael. 1983. Inventory methods
for amphibians and reptiles, p.
416-419 in J. F. Bell and T. Atter-
bury, eds. Renewable resource in-
ventories for monitoring changes
and trends. Oregon state Univer-
sity, Corvallis.

Campbell, Howard W., and Steven P.
Christman. 1982. Field techniques
for herpetofaunal community
analysis, p. 193-200; in Scott
N.J.Jr.(ed.), Herpetological com-
munities. United States Depart-
ment of the interior. Fish Wildlife
Service, Wildlife Research Report
13. 239p.

Clark, Karen and David Euler. 1982.
The importance of pH and habitat
disturbance in amphibian distribu-
tions in Central Ontario. Technical
Report Wildlife Habitat compo-
nent, Lakeshore Capacity Study,
Ontario Ministry of Natural Re-
sources, Toronto, Ontario.

Clawson, Mary E. and Thomas S.
Baskett. 1982. Herpetofauna of the
Ashland Wildlife Area, Boone
County, Missouri. Transactions of
the Missouri Academy Science
16:5-16.

Clawson, Mary E., Thomas S. Baskett
and Michael J. Armbruster. 1984.
An approach to habitat modeling
for herpetofauna. Wildlife Society
Bulletin 12:61-69.

Collins, James P. and Henry M.
Wilbur. 1979. Breeding habits and
habitats of the amphibians of the
Edmin S. George Reserve, Michi-
gan with notes on the local distri-
bution of fishes. Occasional Papers
of the Museum of Zoology Uni-
versity of Michigan no 686. 34 p.

Cook, Francis R. 1984. Introduction
aux amphibiens et reptiles du
Canada. Musees Nationaux du
Canada, Ottawa, 211 p.

Denis Jacques, Enr. 1986. Vegetation
des terres humides du lac St-Pi-
erre. Cartographic 1:10 000. Min-
istere du Loisir, de la Chasse et de
la Peche du Quebec, Corporaton
pour la mise en valeur du lac St-
Pierre. Feuillet 7 a 12.

Dale, J.M., B. Freedman and J. Kere-
kes. 1984. Acidity and associated
water chemistry of amphibian
habitats in Nova Scotia. Canadian
Journal of Zoology 63:97-105.

Degraaf, R.M. and D.D. Rudis. 1983.
Amphibians and reptiles of New
England Habitats and natural his-

tory. The University of Massachu-
setts. 85 p.

Dole, Jim W. 1965a. Spatial relations
in natural populations of leopard
frog, Rana pipiens Schreber, in
northern Michigan. American
Midland Naturalist 74:464-478.

Dole, Jim W. 1965b. Sommer move-
ments of adult leopard frogs, Rana
pipiens Schreber, in northern
Michigan. Ecology 46(3):236-255.

Dole, Jim W. 1967. Spring move-
ments of leopard frogs, Rana pipi-
ens Schreber, in northern Michi-
gan. American Midland Naturalist
78(1):167-181.

Dole, Jim W. 1971. Dispersal of re-
cently metamorphosed leopard
frogs, Rana pipiens. Copeia
1971(2):221-228.

Fox, Barry J. 1979. An objective
method of measuring the vegeta-
tion structure of animal habitats.
Australian Wildlife Research
6:297-303.

Frier, Jo Ann and Robert T. Zap-
palorti. 1984. Reptile and amphib-
ian management techniques.
Transactions of the N.E. Section
Wildlife Society 40:142-148.

Frontier, Serge. 1976. Etude de la
decroissance des valeurs propres
dans une analyse en composantes
principales: comparaison avec le
modele du baton brise. Journal of
experimental Marine Biology and
Ecology 25:67-75.

Froom, Barbara 1982. Amphibians of
Canada. McClelland and Stewart,
Toronto. 120 p.

Gascon, Claude and Dolors Planas.
1986. Spring pond water chemis-
try and the reproduction of the
wood frog, Rana sylvatica. Cana-
dian Journal of Zoology 64:543-
550.

Gibbs, E.L., G.W. Nace, and M.B.
Emmons. 1971. The life frog is al-
most dead. Bioscience 21:1027-
1034.

Green, Roger H. 1971. A multivariate
statistical approach to the Hutch-
insonian niche: Bivalve molluscs
of central Canada. Ecology, 52
(4):543-556.

140

Green, Roger H. 1979. Sampling de-
sign and statistical methods for
environmental biologists. Wiley
Inter-science Pub., Wiley and
Sons, Inc. 257 p.

Grier, James.W. 1984. Biology of ani-
mal behavior. Times Mirror/
Mosby College Publishing, St.
Louis.

Hayes, Marc P. and Mark R. Jen-
nings. 1986. Decline of Ranid Frog
Species in Western North Amer-
ica: Are Bullfrog (Ram catesbeiam)
responsible? Journal of Herpetol-
ogy 20(4):490-509.

Hine, Ruth L., Betty L. Les and Bruce
F. Hellmich. 1981. Leopard frog
populations and mortality in Wis-
consin 1974-76. Technical Bulletin
No. 122, Department of Natural
Resources, Wisconsin. 39 p.

Leclair, Raymond Jr. 1983. Utilisation
de differents types d'habitats par
la grenouille leopard pour fin de
reproduction dans la region de
Bainville au lac St-Pierre. Rapport
technique, Ministere du Loisirs, de
la Chasse et de la Peche du
Quebec. 24 p.

Leclair, Raymond Jr. 1985. Les am-
phibiens du Quebec: Biologie des
especes et problematique de con-
servation des habitats. Ministere
du Loisirs, de la Chasse et de la
Peche du Quebec. 119 p.

Leclair, Raymond Jr. and L. Bar-
ibeau. 1982. Problematique des
anoures au Quebec. Ministere du
Loisirs, de la Chasse et de la Peche
du Quebec. 69 p.

Leclair, R. Jr., and J. Castanet. 1987.
A skeletochronological assessment
of age and growth in the Frog
Ram pipiens Schreber (Amphibia,
Anura) form Southwestern Que-
bec. Copeia 1987(2):361-369.

Legendre, Pierre and Louis Legen-
dre. 1984. Ecologie numerique (2^
ed.). Tome 1: Le traitement mul-
tiple des donnees ecologiques;
Tome 2: La structure des donnees
ecologiques. Presses de I'Univer-
site du Quebec et Masson Cie. 260;
335 p.

MacArthur, Robert H., and John W.
MacArthur. 1961. On bird species
diversity. Ecology. 42:594-598.

Marcotte, Alice. 1981. L'exploitation
des grenouilles au Quebec. Rap-
port technique, Ministere du
Loisirs, de la Chasse et de la Peche
du Quebec, Service de I'Amenage-
ment de la Faune, Memoire Uni-
versite du Quebec a Montreal.
75 p.

Merrell, David J. 1977. Life history of
the leopard frog. Ram pipiens, in
Minnesota. University of Minne-
sota Bell Museum Natural History
Occasionnal Paper 15: 23 p.

MLCP. 1985. Des habitats fauniques
a conserver. CoUoque national sur
la faune. Ministere du Loisirs, de
la Chasse et de la Peche du
Quebec. 81 p.

Nie, Norman H., C. Hadlai Hull, Jean
G. Jenkins, Karin Steinbrenner and
Dale H. Bent. 1975. Statistical
package for the social sciences.
McGraw-Hill, New-York. 675 p.

Paquin, Diane. 1982. L'exploitation
de trois expeces d'anoures dans la
region du lac St-Pierre. Rapport
technique, Ministere du Loisirs, de
la Chasse et de la Peche du
Quebec. 64 p.

Partridge, Linda. 1978. Habitat selec-
tion. Pages 351-376 in J.R. Krebs
and N.B. Davies. (ed.). Behavioral
ecology: an evolutionary ap-
proach. Blackwell Scientific Publi-
cations, Oxford.

Putnam, Robert W. and Albert F.
Bennett. 1981. Thermal depend-
ence of behavioral performance of
anuran amphibians. Animal be-
haviour. 29(2):502-509.

Ramade, Francois. 1984. Elements
d'ecologie; ecologie fondamentale.
McGraw-Hill, Paris. 397 p.

Rittschof, Daniel 1975. Some aspects
of the natural history and ecology
of the leopard frog. Ram pipiens.
University of Michigan, Ann Ar-
bor. Ph. D. Thesis. 212 p.

Roberts, Wayne and Victor Lewin.
1979. Habitat utilization and pop-
ulation densities of the amphibi-

ans of northeastern Alberta. Cana-
dian Field-Natur. 93(2):144-154.

Sarrazin, R., M. Cantin, A. Gagnon,
C. Gauthier and G. Lefebvre. 1983.
La protection des habitats
fauniques au quebec. Gouv. du
Quebec. Dir. Gen. Faune. Min.
Loisir Chasse et Peche. 256 p.

Scherrer, B. 1984. Biostatistique.
Gaetan Morin (ed.). Chicoutimi,
Quebec. 850 p.

Service Canadien De La Faune. 1980.
Plan de gestion des oiseaux aqua-
tiques au Canada, Environnement
Canada, 26 p.

Tessier, Claude and Paul Caron.
1980. Cartographie ecologique de
la vegetation de la rive nord du lac
Saint-Pierre. Rapport technique
Service de I'amenagement et de
l'exploitation de la faune, Min-
istere du Loisirs, de la Chasse et
de la Peche du Quebec. 19 p. +
carte.

Tessier, Claude, Antoine Aubin and
Daniele Chenevier. 1984. Les
facteurs elevation et distance dans
la structure et la productivite
d'une communaute riveraine de
macrophytes. Canadian Journal of
Botany 62:2260-2266.

Vogt, Richard C. and Ruth K. Hine.
1982. Evaluation of techniques for
assessment of amphibian and rep-
tile populations in Wisconsin, p.
201-217; in Scott, N.J.Jr. (ed.). Her-
petological communities. United
States Department of the interior.
Fish Wildlife Service, Wildlife Re-
search Report 13. 239p.

Weller, M.W. 1978. Management of
freshwater marshes for wildlife, p.
267-284 in Freswater wetlands;
ecological processes and manage-
ment potential. New York, 365 p.

Whitaker, John 0.,Jr. 1961. Habitat
and food of mousetrapped young
Ram pipiens and Ram clamitans.
Herpetologica 17(3):173-179.

141

Appendix 1.

Abbreviations for variables used in \he text, figures and tables.

Abbreviations

Variables

Abbreviations

Variables

Abbreviations

Variables

DOPEN

Distance to nearest

^ BASAREA

Basal area

PHonoo*

Photometric index,

open habitat

MINAREA,,
PH

Minimal area

between 50 et 100

DCLOSE

Distance to nearest

pH of soil solution

cm

closed habitat

SAND

Sand fraction in soil

PHOT+*

Photometric index.

DMARSH

Distance to marsh

SILT

Silt fraction in soil

for herbs > 100 cm

line

CLAY

Clay fraction in soil

PHOTCAN*

Photometric index

DWATERP

Distance to nearest
permanent pool

DWATER*

Distance to nearest
temporary pool

under shrub and
tree strata

DHUMAIN

Distance to nearest

MOIST*

% soil moisture

HERBHGHT*

Height of herb stra-

human artefact

TEMP*

Temperature at the

tum

ALTREL

Altitude relative to

soil surface

HERBCOV*

% herb cover

shore line

WTABLE*

% bare ground

HGH*

Cover class for high

NSPHERB

Number of herba-

BAREGRNDj^*

Water table level

graminoid herbs ( >

ceous species

LITTER*

% ground covered

100 cm tall)

NSPSHRUB

Number of shrub

with litter

HBLH*

Cover class for high

species

MOSSCOVj^*

% moss cover

broad-leaf herbs

NSPTREE

Number of tree spe-

DEADWOOD*

% ground covered

MGH/

Cover class for me-

cies

by dead wood

dium graminoid

TREE

Cover class for tree

LITTHICK*

Litter thickness

herbs (20 to 100 cm)

stratum

PHOT20*

Photometric index.

MBLH*

Cover class for me-

SHRUBHI

Cover class for
High shrub stratum

between 0 and 20
cm

dium broad-leaf
herbs

SHRUBLO

Cover class for lov^
shrub stratum

PHOT50*

Photometric index,
between 20 and 50

SMALLH*

Cover class for
herbs < 20 cm tall

cm

N: Variable normalized by square-root transformation.
': Variable measured monthly.

142

Appendix 2.

Significance levels of Pearson rank correlations between ttie variables included in ttie nnodels and all vari-
ables measured. (Significance levels 1 : P < 0.05; 2: P < 0.01 ; 3: P < 0.001 ; 4: P < 0.0001 ; +: positive; negative.)

4

Models

1

2

4

5

D

D

T

M

P

s

B

N

C

P

P

P

H

L

P

M

M

M

E

o

H

I

A

s

L

H

H

H

G

I

H

B

o

A

M

I

o

L

R

P

A

o

o

o

H

T

o

L

s

R

P

s

T

T

E

H

Y

T

T

T

T

T

H

s

s

T

5

G

E

+

2

5

H

c

c

H

0

R

R

0

0

I

A

o

N

B

c

N

V

D

K

DOPEN

-1

-1

-2

-1

+1

+1

DCLOSE

+1

DMARSH

-1

-1

DHUMAN

ALTREL

+3

-1

+1

NSPHERB

NSPSHRUB

+3

-1

-1

-1

-1

+1

+3

NSPTREE

+2

-1

-1

-2

-1

-3

-1

+1

+2

+2

BASAREA

+2

-2

-2

-3

-2

+1

+2

+1

MINAREA

-1

-3

+2

+2

PH

SAND

-2

-4

-1

SILT

-2

CLAY

BAREGRND

-2

+1

+1

LITTER

-2

+1

+1

+2

-2

MOSSCOV

-1

-2

-1

DEADWOOD

-1

-1

+1

LITTHICK

-4

+2

+3

-2

-2

-3

TREE

+1

-1

-1

-1

-3

-2

+2

+2

SHRUBHI

-2

-1

-1

-2

+3

+1

SHRUBLO

-1

_i

+2

+2

MOIST

-1

+1

-1

-1

TEMP

PHOT20

+1

+1

+2

+2

-1

PHOT50

+2

+1

+2

-1

PHOTIOO

-1

-1

+1

-2

PHOT+

+1

+2

+1

-2

PHOTCAN

-1

-2

-1

-1

-2

+1

HERBHGHT

+1

+3

+1

-2

WTABLE

+3

D.WATER

HERBCOV

+1

-2

-2

HGH

-1

+1

+2

+2

+3

-1

-2

-2

HBLH

-1

+1

+1

MGH

+2

MBLH

-1

-2

SMALLH

-2

-2

D: DM mode/.

143

Habitat Correlates of
Distribution of the California
Red -Legged Frog (Rana
aurora draytoriii) and the
Foothill Yellow-Legged Frog
(Rana boylii): Implications for
Management^

Marc P. Hayes and Mark R. Jennings^

Abstract.— We examined features of the habitat
for the California red-legged frog and foothill yellow-
legged frog from the Central Valley of California.
Limited overlap exists in habitat use between each
frog species and introduced aquatic macrofaunal
predators. Temporal data implicate aquatic preda-
tors that restrict red-legged frogs to intermittent
stream habitats as explaining limited overlap. Identi-
fication of responsible predators is currently pre-
vented because the alternative of limited overlap
simply due to differential habitat use between frogs
and any one putative predator cannot be rejected.
Until the predators causing the negative effects are
identified, efforts should be made to isolate these
frogs from likely predators and minimize alteration of
key features in frog habitat.

Wright 1920). Despite this history of
exploitation, few attempts have been
made to link species-specific habitat
requirements of ranid frogs to their
management (but see McAuliffe
1978; Treanor 1975a, b; Treanor and
Nicola 1972). Most ''management"
literature has either simply reviewed
the biolog}'' of selected ranid frog
species or indicated vulnerable life
history stages needing study (Baker
1942, Bury and Whelan 1984, Storer
1933, Willis et al. 1956, Wright 1920).

In this report, we examine the
habitat features of two "non-game"
species, the California red-legged
frog (Rana aurora draytonii) and the
foothill yellow-legged frog (Rana
boylii), two ranid frogs found in low-
land California. Each species has dis-
appeared from sizable areas of its
historic range (Hayes and Jennings
1986, Sweet 1983). Although histori-
cal disappearance of red-legged frogs
has been linked to its exploitation as
food (Jennings and Hayes 1985),
causal factors in the continuing de-
cline of both species remain poorly
understood. Insufficient documenta-
tion of the habitat requirements of
each species has especially impeded
identification of the causes of decline
(Hayes and Jennings 1986). In this
report, we reduce this gap by identi-
fying the habitat requirements that
characterize each frog. We then use
these data to suggest the direction
for management of these two species

The application of habitat
analysis to management has a
long, complex history. The Greek
philosopher Aristotle inferred that
seasonal variation in the distribu-
tion of certain commercially ex-
ploited fishes was related to changes
in their food resources and habitat
temperatures (Cresswell 1862). In the
13th century, the Mongol emperor
Kublai Khan encouraged the gather-
ing of data on foraging patterns of
sport-hunted birds to facilitate ma-
nipulating their populations (Leo-
pold 1931). Since these efforts, many
individuals have used diverse habitat
data to help understand factors that
influence the distribution and success
of various species. Most often, such
data have been used to address com-
mercially important or game species,
usually to identify management al-
ternatives intended to enhance exist-
ing populations or avert population
declines (Bailey 1984, Leopold 1933).
This emphasis has resulted in most
studies addressing selected birds,

'Paper presented at symposium. Man-
agement of Amphibians. Reptiles and Small
Mammals in North America. (Flagstaff. AZ.
July 19-21. 1988.)

^Environmental Scientist, Gaby & Gaby,
Inc.. 6832 SW 68th Street, Miami, FL 33 143-
3 1 15 and Department of Biology, P.O. Box
249118, University of Miami, Coral Gables.
FL 33 124-91 18; Research Associate, Depart-
ment of Herpetology, California Academy
of Sciences, Golden Gate Park, San Fran-
cisco. CA 94118-9961.

fishes, and large mammals. In con-
trast, species historically having lim-
ited economic importance (i.e., "non-
game" species) have been largely ne-
glected (Bury 1975; Bury et al. 1980a,
b; Pister 1976). Only over the last 15
years has an appreciation been
broadly realized that non-game spe-
cies are also in need of management.
Non-game species are often linked to
economically important ones, and as
such, provide significant direct and
indirect benefits to humans (Kellert
1985, Neill 1974). Although this ap-
preciation has led to greater empha-
sis in their study (Bury et al. 1980a,
Pister 1976), a broader understand-
ing of the biology of non-game spe-
cies is increasingly urgent because of
widespread habitat modification in-
fluencing declines among ever-great-
er numbers of such species (Dodd
1978, Hayes and Jennings 1986, Hine
et al. 1981, Honegger 1981).

Amphibians are prominent among
groups of organisms given a non-
game label (Bury et al. 1980a). For
ranid frogs, among the most familiar
of amphibian groups, non-game is
really a misnomer (Brocke 1979) be-
cause they have a history of human
exploitation which has its roots in
European and aboriginal cultural tra-
ditions (Honegger 1981, Zahl 1967)
and has included significant com-
mercial enterprises (Abdulali 1985,
Chamberlain 1898, Husain and Rah-
man 1978, Jennings and Hayes 1985,

144

until experiments can identify the
causes of decline.

METHODS

Our analysis draws upon two data
sets, one addressing R. a. draytonii
and the other, R. hoylii. The former is
based on all known occurrences of R.
a. draytonii (n = 143) from the Central
Valley of California, which we define
as the collective drainage area of the
Kaweah, Kern, Sacramento-San
Joaquin (to Carquinez Strait), and
Tule River systems. We assembled
these data from museum records and

field notes or direct observations of
the many investigators listed in the
acknowledgments or whose data are
cited in Childs and Howard (1955),
Cowan (1979), Fitch (1949), Grinnell
and Storer (1924), Grinnell et al.
(1930), Hallowell (1854, 1859), Ingles
(1932a, b; 1933; 1936), Storer (1925),
Walker (1946), Williamson (1855),
and Wright and Wright (1949). We
used records not authenticated by
museum specimens if they were cor-
roborated by at least two sources.
We then determined the subset (n =
131) of records that could be both
mapped (i.e., where we could iden-
tify the aquatic system likely to be

Table 1 .—Habitat variables recorded for the California red-legged frog
(Rana aurora draytor)ll) data set. Subset scored refers to the subset of lo-
calities for which we were able to score each variable. Percent scored re-
fers to the percentage of the entire data set (n = 143) for which we were
able to score each variable. See text regarding further details concerning
the method of data collection for each variable.

Variable

Subset scored % scored

(n=)

DefirtJtion

1.

Habitat type

140

98

As (1) stream or (2) pond

2.

Temporal status

137

96

As (1) perennial or (2) inter-
mittent

3.

Drainage area

129

90

In km^

4.

Local gradient

139

97

In angular degrees C) from
horizontal

5.

Water depth

74

52

As (1) presence or

(2) absence of water
>0.7 m deep

6.

Vegetation matrix
(emergent or shoreline)

44

31

As (1) dense (area >25%
thickly vegetated)

(2) limited (some, but
<25% of area)

(3) absent

7.

Native fishes

56

39

As (1) present or (2) absent

8.

Introduced fishes

32

22

As (1) present or (2) absent

9.

Introduced bullfrogs

115

80

As (1) present or (2) absent

10.

Substrate alteration

113

79

As (1) present or (2) absent

n.

Vegetation reduction

106

74

As (1) present or (2) absent

12.

Stream order

127

89

As defined by Strahler

(1957)

the site of origin of the source popu-
lation upon which the record was
based), and identified as being from
different "point" localities i>OA km
apart). Although our data set was
developed primarily from this sub-
set, we used a few data from the re-
maining 12 localities for the habitat
variables described below. We used
this additional data because they
were either available with the origi-
nal records or could be determined
independent of accurate mapping.

For each locality, we recorded as
many of 12 habitat variables as pos-
sible (table 1). For aquatic habitat
type, we used the term "stream" for
localities with both a well-defined
drainage inflow and outflow,
whereas we used "pond" for locali-
ties lacking a well-defined inflow and
little or no outflow. Temporal . .atus
of the aquatic habitat was scored as
perennial or intermittent based on
7.5'and 15' United States Geological
Survey (USGS) topographic maps,
but the status of some localities was
modified based on field reconnais-
sance or data provided by other in-
vestigators. For many localities, lack
of change in the temporal status of
the aquatic habitat during the time R.
a. draytonii was recorded was veri-
fied by examining USGS topographic
maps bracketing the frog record
date(s). We used the designation
intermittent to describe the interrup-
tion of surface flow in streams or
complete dry-down in ponds, either
occurring at least once seasonally.
Drainage area indicates the size of
the hydrographic basin influencing
the recorded locality. The drainage
area, local gradient, and stream or-
der were largely estimated from 7.5'
USGS topographic maps. We esti-
mated large drainage areas (>130
km^) by extrapolation to the recorded
locality on topographic maps from
either the drainage area for the near-
est upstream gauging station (United
States Geological Survey 1970a, b) or
section counts on United States For-
est Service and county maps. Local
gradient was estimated from map

145

distances of 0.5-1.0 km across the re-
corded locality except in the few
cases where pronounced local relief
required reduction of this distance
for an accurate estimate.

Data for the remaining variables
(water depth, vegetation matrix, na-
tive and introduced fishes, intro-
duced bullfrogs [Ram catesbeiam],
substrate alteration, and vegetation
reduction) were obtained for subsets
of the larger data set from the
sources indicated earlier supple-
mented by Leidy (1984), Moyle and
Nichols (1973), Moyle et al. (1982),
and Rutter (1908). The exact values
used to partition water depth and
vegetation matrix variables are arbi-
trary. However, we chose their gen-
eral dimensions with the intent of
identifying whether the habitat re-
quirements of red-legged frogs sug-
gested by anecdotal data (moderately
deep water associated with dense
vegetation; see Hayes and Jennings
1986) were supported by this data
set. Variation in the collective data
set required scoring the fish and in-
troduced bullfrog data as presence/
absence, but we also used available
data on which fish species were pres-
ent to interpret the habitat require-
ments of red-legged frogs. Substrate
alteration and vegetation reduction
variables indicate alteration of
aquatic habitats that was, directly or
indirectly, human-effected. We
scored substrate alteration as present
if evidence existed that the shoreline
or substrate topography of the
aquatic habitat had been markedly
altered (e.g., dams, rip-rap, bank-
trampling by cattle). Marked altera-
tion meant that at least 25% of the
area of substrate of a locahty ap-
peared altered. We scored vegetation
as being reduced when data indi-
cated that at least 25% of pre-existing
shoreline or emergent vegetation had
been removed.

We also gathered current data on
a subset of the described localities
through field reconnaissance and
some information provided by others
(data gathered during the interval

1980-1987 represented ''current''
data). We used these data to help
identify temporal changes that may
have occurred at sites or in drainage
systems for which we had historical
data. For this analysis, we used
"drainage system" to mean only the
primary and highest-order (fide
Strahler 1957) secondary tributaries
of the Sacramento-San Joaquin drain-
age system. These data were particu-
larly important for indicating where
red-legged frogs were probably ex-
tinct.

The data set addressing R. hoylii
consists of data published by Moyle
(1973) and Moyle and Nichols (1973)
from which we re-examined selected
elements. Collection methods for
these data are thoroughly described
therein. Our reanalysis used most of
the variables described by Moyle
(1973) with some modifications. We
used the original estimates of the
numbers of each fish species rather
than the coded values; the numbers
of yellow-legged frogs and bullfrogs
remained coded because the original
data were recorded as coded.
Moyle' s stream type variable was
reduced to two categories by com-
bining his three intermittent and
three perennial stream categories.
We also added two variables, one
which combines Moyle's cobble and
boulder/bedrock substrate catego-
ries. The other describes the stream
morphology category designated in
Moyle's original data as smooth wa-
ter and fits the definition of a run
(Armour et al. 1983). For correlations
between yellow-legged frogs and
other species, we used only the sub-
set of localities where either or both
of yellow-legged frogs and the spe-
cies being compared was present.

We re-examined these data for
four reasons. First, Moyle (1973)
summarized data from only some of
the sites where yellow-legged frogs
were not found. We were equally
interested in habitat variation among
all sites sampled where yellow-
legged frogs had not been found as
well as sites where they were found.

Second, Moyle (1973) found that the
collective abundance of all fish spe-
cies was inversely correlated with
that of yellow-legged frogs, but also
commented that yellow-legged frogs
were most abundant where native
fishes were present. Because original
estimates of the numbers of each fish
species were available and an inverse
relationship between the abundance
of native frogs and introduced fishes
had already been identified (Hayes
and Jennings 1986), we were espe-
cially interested in relationships be-
tween the abundance of specific na-
tive and introduced fishes and that of
yellow-legged frogs. Third, Moyle
(1973) coded fish abundance when
the data, as originally recorded, per-
mit at least ranking, so, where pos-
sible, we analyzed the original data
directly to minimize bias that can re-
sult from coding (Sokal and Rohlf
1981). Lastly, the fish abundance data
displayed skewed distributions for
several species, so we used non-par-
ametric analyses to avoid having to
make any assumptions about sample
distributions.

Statistical treatments used are de-
scribed in Sokal and Rohlf (1981) and
Zar (1974). All contingency table
comparisons performed had one de-
gree of freedom (df), so all Chi-
square values were calculated with
the correction for continuity (X^^). For
those analyses that required more
than one comparison using some of
the data, alpha (a) was evaluated
based on the number of comparisons
to a level equivalent to 0.05 using Si-
dak's multiplicative inequality (Sokal
and Rohlf 1981).

RESULTS

California Red-Legged Frog

Ram aurora draytonii was recorded
primarily from aquatic habitats that
were intermittent streams which in-
cluded some area with water at least
0.7 meters deep, had a largely intact
emergent or shoreline vegetation.

146

Table S.—Frequency of fish species co-occurrence with Rana aurora dray-
tonii. Percentage Is the number of sites respective fish species were re-
corded as a function of all sites where fishes were recorded as co-occur-
ring with R. a. drayfonii. An asterisk (*) indicates introduced species.

Co-occurrence Percentage

Species

(n =>

<%)

California roach (Lavinia symmefricus)

19

47

Mosquitofish (Gambusia affinis)*

10

25

Hitch (Lavinia exilicauda)

6

15

Green sunfish (Lepomis cyanellus)*

6

15

Threespine stickleback (Gasferosfeus aculeafus)

3

8

Sacramento squowfish (Piychocheiius grandis)

2

5

Sacramento sucker (Cafosfomus occidenfalis)

2

5

Prickly sculpin (Coffus asper)

1

3

Hardhead (Mylopliarodon conoceplialus)

1

3

Rainbow trout (Salmo gairdnerii)

1

3

Brown trout (Salmo fruffa)*

1

3

and lacked introduced bullfrogs
(table 2). We found descriptions ade-
quate to characterize vegetation for
77% (33) of sites where the emergent
or shoreline vegetation variable
could be scored. With three excep-
tions, descriptions indicated that ei-
ther, or both of, an emergent vegeta-
tion of cattails {Typha spp.) or tules
{Scirpus spp.), or a shoreline vegeta-
tion of willows (Salix spp.) were
present. Shrubby willows were re-
corded at 67% (22) of the sites with
vegetative descriptions, and were
identified as arroyo willow (Salix la-
siolepis) in the eight instances where a
species name was provided. Only
juvenile frogs were recorded at five
of the six sites where a limited emer-
gent vegetation was present and at
the only site that lacked a water
depth greater than 0.7 m. We found
no significant difference in the num-
bers of intermittent versus perennial

sites with red-legged frogs that had a
dense vegetation and a water depth
of >0.7 m (X2^ = 0.338, p = 0.561, for

vegetation; X^^ = 0.017, p = 0.897, for
water depth; X^^^^ ^^^^s = 5-024 for
both).

Rana aurora draytonii was also
more frequently recorded at sites
with native fishes and with substrate
alteration, but less frequently re-
corded at sites with introduced
fishes. Fishes were present at 69% (40
of 58) of sites where data as to their
occurrence were recorded; 26 sites
had only native fishes, seven had
only introduced fishes, and seven
had both. Only four fish species,
California roach (Lavinia symmet-
ricus), hitch (Lavinia exilicauda), green
sunfish (Lepomis cyanellus), and
mosquitofish (Gambusia affinis), were
recorded as co-occurring with R. a.
draytonii at more than three sites
(table 3), and only California roach
was recorded at more than 25% (10)
of sites. Sixty of the 70 sites described
as being substrate-altered at the time
R. a. draytonii was recorded were
small impoundments.

California red-legged frogs were
also most frequently recorded at sites
influenced by a small drainage area,
having a low local gradient, and in
streams having a low stream order.
Drainage areas of sites from which R.
a. draytonii was recorded vary from
0.02 km2 to over 9000 km^, but two-

Table 2.— Variation among habitat variables for California red-legged frogs
(Rana aurora draytonii). Number of localities (percentages of localities) in
each category are indicated. See table 1 and text for explanation of vari-
able categories.

Variable

Variable categories

1.

Aquatic habitat type

(a) stream

129

(92%)

(b) pond

10

(8%)

2.

Temporal status of

(a) perennial

49

(36%)

aquatic site

(b) intermittent

88

(64%)

3.

Water depth

(a) > 0.7 meters

73

(99%)

(b) < 0.7 meters

1

(1%)

4.

Emergent and

(a) absent

0

(0%)

shoreline vegetation

(b) limited

9

(20%)

(c) dense

35

(80%)

5.

Native fishes

(a) present

33

(65%)

(b) absent

18

(35%)

6.

Introduced fishes

(a) present

14

(44%)

(b) absent

18

(56%)

7.

Introduced bullfrogs

(a) present

13

(1 1%)

(b) absent

102

(89%)

8.

Significant substrate

(a) present

70

(62%)

alteration

(b) absent

43

(38%)

9.

Significant removal

(a) present

1

(2%)

vegetation (see #4)

(b) absent

44

(98%)

10.

Current status

(a) probably extant

86

(72%)

(among localities)

(b) probably extinct

34

(28%)

11.

Current status

(a) probably extant

18

(42%)

(among drainages)

(b) probably extinct

25

(58%)

147

thirds (n = 83) are from localities
with drainage areas <40 km^ (fig. 1).
Local gradient (slope) at California
red-legged frog localities varies from
0.04° to 12.8° from horizontal, al-
though 87% (n = 100) occur at sites
with slopes <2°. California red-
legged frogs have been recorded in
1st to 6th order streams, but 94% (n =
119) of these localities are 4th- or
lesser-order streams and 42% are 1st-
order streams (fig. 2).

Based on the subset for which cur-
rent data were available (n = 120),
California red-legged frogs are
probably extinct at >25% of the lo-
calities where they were historically
recorded. When clustered into a
sample representing drainage sys-
tems (n = 43; see methods), this sub-
set indicates that California red-
legged frogs are probably extinct in
over 50% of the drainage systems in
the Central Valley area. Three habitat
variables (temporal status of aquatic
habitat, drainage area, and intro-
duced bullfrogs) showed a signifi-
cant relationship to the probability of
survival of local populations of C!ali-
fornia red-legged frogs (table 4). We
found that R. a. draytonii is likely ex-
tant at 82% (n = 70) of localities with
an intermittent aquatic habitat,
whereas it is probably extinct at 71%
(n = 22) of the sites with a perennial
aquatic habitat. Grouping localities
based on drainage area, R. a. dray-
tonii is probably extant at 83% (n =

>«01 1 3

3001-4000 1 1

MOt-3000 |2

ATM ISOI-MOO 2
(H km)

lOOl-ISOO t

SOI-tOOO 1

iKillUMbc)

Figure 1 .—Frequency distribution of locali-
ties wtiere Rana aurora draytonii \nas been
recorded in thie Central Valley, California
based on drainage area. Thie inset details
\he frequency distribution of localities witti
drainage areas < 280 kpn*.

7A\-2tO

201-240

12

161-200

121-160

81-120

13

41-60
0-40

I I I I I
0 20 40 60 60 100

■ \

Table 4.— Contingency analysis relating selected habitat variables to an
estimate of the lilceiihood that historically recorded California red-legged
frog populations are extant. Status of frog populations at recorded locali-
ties are indicated as extant (= probably extant) and extinct <= probably
extinct). A double asterisk (") denotes significant contingency tables,
based a critical ^^^^y^^^^^Q^^j = 7.3, a adjusted for seven comparisons (see
methods).

Locality Status

Variable

Condition

extcnt

extinct

x%

Probability

1.

Temporal status

Perennial

9

22

27.326

o.ooor*

Intermittent

70

15

2.

Drainage area

>300 km2

0

11

31.466

o.ooor*

<300 km^

85

18

3.

Native fishes

■f

13

6

0.276

0.5991

14

11

4

Introduced bullfrogs

+

0

10

27.140

0.0001"

70

16

5.

Substrate alteration^

25

14

0.983

0.3215

47

14

6.

Introduced fishes

+

5

9

0.003

0.9524

7

10

7.

Substrate alteration''

-1-

21

3

<0.001

0.9944

26 5

"Analysis with all localities.

^Analysis with subset of localities having a drainage area <25 km'.

\

affected by the largest drainage areas
(n = 10). Similarly, R. a. draytonii is
probably extant at 81% (n = 70) of
localities lacking introduced bull-
frogs and is probably extinct at all
localities (n = 10) where it has been
recorded with bullfrogs. Remaining
variables either failed to show a sig-
nificant relationship to the probabil-
ity of California red-legged frog sur-
vival (table 4), or one of the variable
categories was so rare that this analy-
sis was not applicable (see table 2).

Foothill Yellow- Legged Frog

Rana boylii was recorded primarily
from shallow, partly shaded stream
sites with riffles and at least a cobble-
sized substrate. All 29 stream sites at
which either post-metamorphic or
larval R. boylii were recorded were
<0.6 m in average water depth (fig. 3)
and had at least some shading (fig.
4). Rana boylii was recorded more

85) of sites influenced by a small
(<300 km^) drainage area, whereas it
is probably extinct at all recorded
localities (n = 11) influenced by a
large (>300 km^) drainage area.
Moreover, available data indicate
that R. a. draytonii is extinct at all re-
corded localities on the Central Val-
ley floor, which includes all localities

0 10 » M 4Q SC M

Figure 2.— Frequency distribution of locali-
ties wt^ere Rana aurora draytonii has been
recorded in the Central Valley, California
based on stream order.

148

frequently at sites with a stream area
that was >20% shaded than at sites
with >20% shading. Only one of 29 R.
hoylii sites lacked riffle habitat and R.
boylii was recorded significantly
more frequently at sites with >40%
riffle area than at sites with a riffle
area of <40% [X^ = 8.680, p = 0.003,
X'df=,a(2)=o.o25 = 5.024; fig. 5]. Only four
of 29 R. boylii sites lacked at least a
cobble-sized substrate and R. boylii
was recorded most frequently (20 of
29) at sites with >40% of the sub-
strate that was at least cobble-sized
(fig. 6). Few other patterns could be
identified from among the environ-
mental variables that we re-analyzed.
Rana boylii was recorded more fre-
quently from perennial streams (n =
19) than from intermittent ones (n =
10), but the difference was not sig-
nificant when compared to the total
number of perennial (n = 71) and
intermittent (n = 59) stream sites
sampled [X^'^ = 1.268, p = 0.260,
X'df=i,a(2)=o.o25 = 5-024]. Of 13 environ-
mental variables that we re-exam-
ined, only the percentage of stream
area in riffles was significantly corre-
lated with the abundance of R. boylii
(table 5).

Rana boylii occurred with 1-5 Cx =
2.5) of the vertebrate members of the
aquatic macrofauna at 26 of the 29
localities where it was recorded.

Figure 3.— Histogram of the proportion of
sites in stream depthi categories whiere
Rana boylii hias been recorded in thie Sierra
Nevada foothiiils, California. Sample sizes as
a function of thie total sample in eachi
stream deptti category are: <0.20 (n=8/24),
0.21=0.40 (n=9/43), 0.41-0.60 (n=12/57), and
>0.60 (n=0/18).

0 I-» Jl-« <!■« 6'-«0

^Ntifi or sirtM Am m (arriH

Figure 5.— Histogram of thie proportion of
sites in riffle categories wtiere Rana boylii
has been recorded in the Sierra Nevada
foothills, California. Sample sizes as a func-
tion of the total sample in each ritfle cate-
gory are: 0% <n=l/36), 1 -20% (n=5/31 ), 21 -
407, <n=4/21). 41-60% (n=l 1/28). 61-80%
(n=7/19), and 81-100% (n=2/6>.

Foothill yellow-legged frogs were
recorded as occurring with 12 differ-
ent species, but co-occurrence, ex-
pressed as the percentage of total
sites at which either R. boylii or the
co-occurring species were recorded,
did not exceed 31% (table 6). Intro-
duced species (n = 6) occurred with
R. boylii less frequently Tx = 2, 1-3)
than native species \x = 9.3, 1-17) and
native species had a significantly
higher percentage of co-occurrence
(3-31%, X = 16.5%) than introduced
species [n = 6; 2-9%, x = 3.7%; Mann-
Whitney test, U' = 32.5, p = 0.0275,
U

criHcala(2)=0.05

= 31]. Only four native

0 1-20 21-« 4l-«0 6l-« SI-100
PrcirUgi if SkiM Slmm Vm

Figure 4.— Histogram of the proportion of
sites in stream shading categories where
Rana boylii has been recorded in the Sierra
Nevada foothills, California. Sample sizes as
a function of the total sample In each
sh'eam shading category are: 0% (n=0/5),
1 -207o (n=3/37>, 2 1 -40% (n=7/38), 41 -607*
(n=8/30), 61 -807. (n=9/23), and 81 -1007,
(n=2/8).

0 1-20 2l-« 41-M i\-m 81-100

Figure 6.— Histogram of the proportion of
sites in substrate categories where Rana
boylii has been recorded In the Sierra Ne-
vada foothills, California. Sample sizes as a
function of the total sample in each sub-
strate category are: 07, (n=4/19), 1-207,
(n=3/32), 21 -407, (n=2/23), 41 -607, (n=7/29),
61-807, (n=9/26). and 81-1007, (n=4/12).

fishes, California roach, Sacramento
sucker (Catostomus occidentalis), Sac-
ramento squawfish (Ptychocheilus
grandis), and rainbow trout (Salmo
gairdnerii), occurred with R. boylii at
more than three of the 29 sites where
the latter was recorded, and of these,
only California roach occurred with
R. boylii at more than 50% of the sites
where R. boylii was recorded. Only
one species assemblage, that consist-
ing of California roach, Sacramento
squawfish, and Sacramento sucker,
occurred with R. boylii more often
than expected by chance alone (table
7). Correlation analysis indicated that
the abundance of 10 of the 12 co-oc-
curring species was significantly in-
versely correlated with the abun-
dance of R. boylii (table 8).

DISCUSSION

Habitat Variation

California Red-Legged Frog

A dense vegetation close to water
level and shading water of moderate
depth are habitat features that ap-
pear especially important to Califor-
nia red-legged frogs. Previous au-
thors have suggested or implied the
occurrence of at least one of these
habitat features. Storer (1925) noted

149

that R. a. draytonii in streams was re-
stricted to large pools, which implies
a moderate water depth. Stebbins
(1966, 1985) emphasized vegetative
cover as important to red-legged
frogs, but his comments confound
habitat characteristics that may be
attributable to northern versus Cali-
fornia (southern) red-legged frogs;
data on these two forms should re-
main partitioned until it is well-es-
tablished that they are not different
species (Hayes and Miyamoto 1984,
Hayes and Krempels 1986). Zweifel
(1955) coupled the water depth and
vegetation features of California red-
legged frog habitat, but he empha-
sizes a herbaceous shoreline vegeta-
tion. Chir data indicate that a more
complex vegetation is a feature of
sites where R. a. draytonii occurs.
Cattails, bulrushes, and shrubby wil-

/fdble 5. —Spearman rankcorrelatlorr
between selected environmental
variables and the coded abun-
dance of R. boylil as measured by
Moyie (1973). Sample size for each
variable is n = 1 30. A double asterisk
(**) Indicates significant correlations;
based on a critical r, = 0,267 at an
aCtwo -tailed) = 0.002, adjusted for 24
comparisons (13 below and 1 1 In |
table 8; see methods).

Variable

Correlation
coefficient <r, =)

Human alteration

-0.160

Vegetation

Aquatic vegetation (%)

-0.157

Floating vegetation (%)

-0.169

Shade (%)

0.219

Stream morphology

Pools (%)

-0.205

Riffles (%)

0.304**

Runs (%)

-0.020

Stream substrate

Mud(%)
Sand (%)
Gravel (%)
i^ubble (%)
Boulder/ Bedrock (%)

-0.035
-0.085
-0.032
0.071
0.192

yRubble/Boulder/Bedrock (%) 0. 1 72|

Table 6.— Occurrences of aquatic macrofaunal species among the 1 30
stream sites sampled by t^oyle (1973) and Moyie and Nichols (1973). Co-
occurrences is the number of sites Rana boy/// was found to co-occur with
each species. Percentage of co-occurrences is co-occurrences as the
percentage of those sites at which either R. boylil or the state species oc-
cur. An asterisk (*) indicates introduced species. Ten other fish species
(Goldfish (Carassius auratus), Prickly scuipln (Coitus asper), Common carp
(Cyprlnus carplo), Threadfln shad /Dorosoma pe/enense;, Threespine stick-
leback (Gasterosteus ocu/ea/us), Yellow bullhead (Ictalurus nebulosus),
Redear sunfish (Lepomis microlophus), Chinook salmon (Onchorhynchus
tshawytscha), Brown trout (Salmo trutta)) were recorded at low numbers of
stations (<8); none were recorded as co-occurring with R, boylil

Species

Occurrences Co-occur- %of
rences co-occur-
(n =) <n =) rences

Bullfrog ('/?ono catesbe/ono/ 68

Green sunfish C/.epom/s c/ane//us)* 61
Sacramento sucker (Cafosfomus occidenfalis) 55
Sacramento squawfish (Ptychocheilus grandis) 48

California roach CLov/n/a symmefncus) 43

largemouth bass (Micropterus salmoides)* 41

Mosquitofish (Gambusia affinisT 37

B\ueQi\\ (Lepomis macrochirus)* 33

Rainbow trout (Salmo galrdnerii) 27

White catfish (7c/o/afus cofus)* 13

Golden shiner (Nofemigonus crysoleucas)* 13

Hitch (Lavinia exilicauda) 12

Hardhead (Mylopharodon conocephalus) 1 1

Smallmouth bass (Micropterus dolomieui)* 9

2
2
13
12
17
0
1
3
11
1
0
1
2
3

2
2
18
18
31
0
2
5
24
2
0
3
5
9

Table 7.— Frequencies of species assemblages of aquatic macrofaunal
vertebrates co-occurring with R. boy/// from data recorded by Moyie
(1973). Assemblages listed include only combinations of species recorded
as co-occurring wlth^. boy/// at least seven localities (see table 6). Listed
species are California roach (RCH), Sacramento sucker (SKR), Sacramento
squawfish (SO), and Rainbow trout (RD. Asterisks (**) identify assemblages
co-occurring at frequencies significantly higher than expected by chance,
based on a critical X^e^,T a-ooos ~ 7.879, adjusted for 1 1 combinations (see
methods). Probabilities (p) are those associated with calculated values.

Species
assemblage

Frequencies

Probability

Observed

Expected

RCH/RT/SKR/SQ

2

1.20

0.077

0.75<p<0.90

RCH/SKR/SQ

9

3.15

9.068**

0.003

RCH/RT/SQ

2

2.67

0.011

0.90<p<0.95

RCH/RT/SKR

2

2.89

0.053

0.75<p<0.90

RT/SQ/SKR

2

2.04

0.104

0.50<p<0.75

RCH/RT

5

6.45

0.139

0.60<p<0.75

RCH/SKR

10

7.62

0.463

0.25<p<0.50

RCH/SQ

9

7.03

0.305

0.50<p<075

RT/SKR

3

4.93

0.415

0.50<p<0.75

RT/SQ

3

4.55

0.243

0.50<p<0.75

SKR/SQ

11

5.38

4.959

0.026

150

lows, the plants comprising emergent
and shoreline vegetation at such
sites, typically shade a substantial
surface area of water with a dense
matrix at or near water level. Califor-
nia red-legged frogs appear sensitive
to the presence of such a vegetation
structure because most sites from
which frogs were recorded lacked
significant alteration of emergent or
shoreline vegetation (see table 2).
Moreover, because only juvenile
frogs were recorded from most sites
with limited shoreline or emergent
vegetation, a minimum amount of
such vegetation appears to be needed
for survival of adults. Parallel argu-
ments apply to water depth. Previ-
ous authors have characterized R. a.
draytonii as a p)Ool- or pond-dwelling
species (Stebbins 1966, 1985; Storer
1925; Zweifel 1955) and descriptions
corresponding to that characteriza-
tion were recorded for this frog at
most sites. Yet, we found that using
minimum water depth was a more
encompassing habitat descriptor be-

cause it included canals and stream
sites where adult frogs were de-
scribed as being conrvmon and that
had the minimum water depth re-
quirement, but could not be de-
scribed as either ponds or stream
pools. Available description of such
sites indicates that they fit the defini-
tion of a run (Armour et al. 1983),
although data upon which part of the
definition is based (the rate of water
flow) are lacking.

We believe that California red-
legged frogs occur primarily in
streams because alternative sites
(ponds) that have suitable water
depth and vegetation characteristics
were historically rare outside of
stream habitats rather than because
red-legged frogs are somehow pre-
adapted for survival in streams. His-
torically, pond habitats below 1500 m
in the Central Valley were mostly
vernal pools, a habitat too shallow
and ephemeral to develop the mac-
rovegetation found associated with
R. a. draytonii (see Holland 1973, Jain

Table 8.— Spearman rank correlation between the numerical <non-coded)
abundance of the vertebrate macrofauna and the abundance (coded) of
R. boylUas recorded by Moyle (1973). Sample size is based on the total
number of sites where either R. boy/// or the species being compared was
present. A single asterisk (*) Indicates introduced species. A double aster-
isk (••) Identifies significant correlations at an _ (two-tailed) = 0.002, ad-
justed for 24 comparisons (1 1 below and 13 In table 5; see methods).
Probability (p) is the probability of obtaining the calculated Spearman cor-
relation coefficient (r^. Common names for the listed species are In
table 6.

Critical

size

coefficient

Probability

S|:>ecles

(n=)

(r,=;

(p=)

Cafosfomus occidenfalis

71

-0.404"

<0.001

-0.363

Gambusia affinis*

62

-0.835"

<0.001

-0.388

Icfalurus cafus*

41

-0.798**

<0.001

-0.473

Lovinia eydlicauda

40

-0.760"

<0.001

-0.479

Lavinla symm e trie us

55

-0.316

0.020

-0.411

Lepomis cyonellus*

88

-0.742"

<0.001

-0.327

Lepomis macrochirus'

59

-0.827"

<0,001

-0.397

Micropferus dolomfeur

35

-0.538"

0.001

-0.510

Mylopharodon conocephalus

38

-0.607"

<0.001

-0.491

PlychocheHus grandis

66

-0.54r*

<0.001

-0.376

Rana cafesbeiana*

90

-0.800"

<0.001

-0.323

Salmo gairdneriJ

44

-0.425

0.005

-0.458

1976). Even the only two exceptions
to R. a. draytonii not occurring in ver-
nal pools support this hypothesis. A
large vernal pool in San (Dbispo
County, California is known to have
a population of California red-legged
frogs (D. C. Holland, pers. comm.).
However, this vernal pool is atypical
because it possesses significant mac-
rovegetation and water depth. These
features appear to be present because
this large (ca. 20 ha) pool does not
dry down each year. The second ex-
ception is a vernal pool in coastal
southern California in which two
frogs with abnormal numbers of legs
were found (Cunningham 1955).
Cunningham thought that the defects
were induced by exposure to high
temperatures during early develop-
ment, a condition facilitated by the
limited vegetative cover that was
present. His speculation may be
valid if California red-legged frog
embryos have a low critical thermal
maximum (Hayes and Jennings
1986). Storer (1925) thought that R. a.
draytonii was excluded from tempo-
rary (vernal) pools because its larval
period is relatively long, but the
more likely mechanism is that frogs
immigrating to such pools were un-
able to establish because suitable
habitat was lacking. The latter hy-
pothesis is supported because Cali-
fornia red-legged frogs are not re-
corded from the many vernal pools
that hold water for intervals longer
than the minimum time required by
R. a. draytonii to complete metamor-
phosis (10 weeks; Hayes, unpubl.
data; see also Jain 1976, Zedler 1987).

Rana a. draytonii also appears to
have responded to the creation of
habitat with the appropriate vegeta-
tion and water depth characteristics.
A significant aspect of the changes in
aquatic habitats that have occurred
in the Central Valley below 1500 m is
an increase in the number of perma-
nent ponds (Moyle 1973). Storer
(1925) reported that R. a. draytonii
occurred in a number of water stor-
age reservoirs and artificial ponds,
but the habitat features of those sites

151

were not described. Thus, it was of
special interest to find that no signifi-
cant difference could be identified
between the probability of extinction
of R. a. draytonii at substrate-altered
sites (mostly small impoundments)
and at sites lacking such alteration.
Moyle (1973) concluded that the de-
cline of R. a. draytonii was related in
part to human-induced alteration,
including creation of impoundments.
Our data suggest that human-in-
duced alteration creating small im-
poundments cannot be related di-
rectly to the disappearance of Cali-
fornia red-legged frogs. We empha-
size that these data do not exclude
the alternative, discussed later,
which indicates that the creation of
small impoundments is likely to have
an indirect negative effect on R. a.
draytonii by facilitating the dispersal
of introduced aquatic predators.

Besides features of habitat struc-
ture associated with R. a. draytonii, its
isolation from one or more aquatic
macrofaunal predators is the other
key element suggested by these data.
No significant variation was found in
the features of habitat structure im-
portant to R. a. draytonii between
intermittent and perennial aquatic
sites, so differences in habitat struc-
ture cannot explain why R. a. dray-
tonii is recorded most frequently
from intermittent aquatic sites. We
believe that California red-legged
frogs were recorded most frequently
from intermittent sites because the
likelihood of extinction at perennial
sites is now higher than at intermit-
tent sites (see table 4) and few his-
torical data are available from when
frogs were often found at perennial
sites.

California red-legged frogs are
now extinct from all sites on the Cen-
tral Valley floor, all of which were
perennial and, except for one, were
recorded prior to 1950. We believe
that the disadvantage associated
with perennial sites and the advan-
tage associated with intermittent
sites is the degree to which the for-
mer allow, and the latter restrict, the

access of aquatic macrofaunal preda-
tors.

The remaining variation in fea-
tures of R. a. draytonii habitat we
have identified can be directly, or
indirectly, linked to a hypothesis in-
voking the influence of one or more
aquatic macrofaunal predators. The
significantly lower likelihood of ex-
tinction at sites with small drainage
areas (table 4) and R. a. draytonii
being recorded from a greater num-
ber of localities with smaller drain-
age areas (fig. 1) and lower stream
orders (fig. 2), are probably unrelated
to either drainage area or stream or-
der effects per se. Rather, they are a
function of both the bias against re-
cording historical data and the fact
that sites with smaller drainages or
lower stream orders have a higher
probability of being intermittent
aquatic habitats, which have a higher
probability of excluding aquatic
predators. Limited co-occurrence
with aquatic predators, namely bull-
frogs and predatory fishes, and a sig-
nificantly higher likelihood of extinc-
tion at sites where bullfrogs were re-
corded (table 4) may indicate a nega-
tive interaction with one or more of
these species. Rana a. draytonii did
not co-occur with any fish species
frequently. It co-occurred most often
with California roach, a small, om-
nivorous native fish that is thought
to have declined, in part, due to pre-
dation by introduced fishes (Moyle
and Nichols 1974, Moyle 1976). We
did not detect a significantly higher
likelihood of extinction at sites with
introduced fishes. However, the
sample was too small to partition to
permit testing individual fish species,
the level at which we believe such an
effect is most likely.

While we are reasonably con-
vinced that the greater restriction of
R. a. draytonii to intermittent aquatic
habitats is an effect due to novel
aquatic predators, we emphasize that
these data cannot identify which are
the aquatic predators producing such
an effect. The inability to identify the
responsible predators is complicated

by the condition of limited overlap
between each potential predator and
R. a. draytonii. That condition pre-
vents excluding the alternative that
different habitat requirements rather
than any predatory interaction may
explain the limited overlap in habitat
use between each putative predator
and California red-legged frogs
(compare Moyle 1973 for bullfrogs
and Moyle and Nichols (1973) for
various fishes, but especially mosqui-
tofish and green sunfish; see also
Hayes and Jennings 1986 for a dis-
cussion). It is this fact and the appar-
ent intolerance of R. a. draytonii to
unshaded habitat that leads us to
suggest that some alteration of ripar-
ian vegetation may be necessary to
create the conditions for a negative
interaction.

Foothill Yellow-Legged Frog

Partly shaded, shallow streams and
riffles with a rocky substrate that is
at least cobble-sized are the habitat
features that appear to be important
to foothill yellow-legged frogs. Previ-
ous authors agree that R. hoylii oc-
curs in streams (Moyle 1973; Stebbins
1966, 1985; Storer 1925; Zweifel
1955), but variation exists in the fea-
tures of streams associated with
these frogs. Of environmental vari-
ables that appear important to R.
boy Hi, the percentage of stream area
in riffles is the only one we were able
to correlate significantly, albeit
weakly, with its abundance. Moyle
(1973) obtained a similar positive
correlation in his original analysis of
the same data, and Stebbins (1966,
1985) also emphasized riffles as one
of the key aspects of R. hoylii habitat.
The reason for the weak correlation
we found is uncertain, but one or
more of three factors probably pro-
duced that result. First, as intermit-
tent streams lose surface flow during
late summer, riffles disappear, and R.
hoylii can then be found associated
with stream pools (Fitch 1938, Slevin
1928, Storer 1925, Zweifel 1955).

152

Moyle's data were collected in late
summer and 10 of the 29 stream sites
at which R. boylii was recorded were
intermittent, so data from these sites
may have diluted the correlation.
Second, riffle area may be correlated
with the abundance of R. boylii only
above or below certain values (see
fig. 5). Lastly, R. boylii has been re-
ported from sites with little or no
riffle habitat unrelated to seasonal
patterns (Fitch 1938, Zweifel 1955).

Apart from riffles, our reanalysis
of environmental variables differs
from that of Moyle (1973), who
found that five of the other variables
that we re-examined were either
positively (i.e., shading and boulder/
bedrock; compare table 1 in Moyle
[1973] and our table 5) or negatively
(i.e., rooted vegetation [= our aquatic
vegetation], pools, man modified [=
our human alteration]) significantly
correlated with the abundance of jR.
boylii. We attribute this difference, in
part, to our analysis being more con-
servative because we adjusted a for
the experimentwise error rate, our
analysis was not restricted to locali-
ties where only frogs were found,
and we used non-parametric tests.
Some of the correlations that Moyle
(1973) observed with R. boylii abun-
dance may have been significant due
to one or more of these differences.
We must emphasize, however, that
several of the variables that Moyle
found correlated with R. boylii abun-
dance vary differentially in their oc-
currence between riffles and pools
(e.g., boulder /bedrock; see Moyle
[1973] and Moyle and Nichols
[1973]). Those variables are also sus-
ceptible to the seasonal correlation-
altering effects discussed for the riffle
variable. Thus, a conservative analy-
sis, like ours, is less likely to detect
variables related to frog abundance
within such a data set.

Nevertheless, variables identified
as important to R. boylii need not be
correlated to its abundance. Stream
depth, shading, and substrate type
may represent such variables. Out
reanalysis of Moyle's data suggests

that sites with a shallow average
stream depth are somehow advanta-
geous (see fig. 3). Moyle (1973) found
no significant correlation between the
abundance of R. boylii and stream
depth, and he did not discuss stream
depth with respect to foothill yellow-
legged frogs in any other context.
Zweifel (1955) noted that streams in
which R. boylii occurred were seldom
more than 0.3 m deep, and Fitch
(1936), Storer (1925), and Wright and
Wright (1949) found that R. boylii
usually lays eggs in shallow water.
Still, overall importance of stream
depth to R. boylii remains unclear.
Our reanalysis also suggests that
some advantage is linked to in-
creased shade up to some intermedi-
ate level (see fig. 4). Zweifel (1955)
described shading in typical R. boylii
habitat as interrupted, whereas
Moyle (1973) reported a positive cor-
relation between frog abundance and
the degree of shading.

Some workers have emphasized
the degree of openness or insolation
in R. boylii habitat, rather than ad-
dressing shading (Fitch 1938; Steb-
bins 1966, 1985). Nevertheless, even
the latter imply that some shading is
present. Fitch's (1938) suggestion that
yellow-legged frogs are excluded by
dense canopy may be supported by
Moyle's data because he recorded no
R. boylii at sites with >90% shading
(see also fig. 4). Our reanalysis also
suggests that some advantage is as-
sociated with sites possessing at least
a cobble-sized substrate (see fig. 6).
Although workers have most fre-
quently emphasized the rocky aspect
of R. boylii habitat (Fitch 1936, 1938;
Moyle 1973; Stebbins 1966, 1985;
Storer 1925), substrate descriptions
of that habitat are probably as varied
as any other single variable. Moyle
(1973) identified a positive correla-
tion between the percentage of
stream area with bedrock and boul-
ders and the abundance of R. boylii,
yet sites with gravely (Gordon 1939),
sandy (Zweifel 1955), or muddy sub-
strates have also been recorded
(Fitch 1938, Storer 1925). Because

Moyle's data do not provide frog
age, we could not determine whether
sites having a substrate that was less
than cobble-sized were simply mar-
ginal habitat with juvenile R. boylii
(see Zweifel 1955), or whether they
represented real variation in habitat
used by established populations.

Fitch (1938) and Zweifel (1955) re-
ported on a few sites with adult frogs
that lacked a substrate that was
cobble-sized or larger and appeared
to have few predators. They sug-
gested that yellow-legged frogs are
rarely recorded from such sites be-
cause their predators may access the
"atypical" habitat more easily. Nev-
ertheless, data on the aforementioned
variables reinforce the conclusion al-
ready arrived at with R. a. draytonii:
Existing data cannot distinguish hy-
potheses explaining the differential
occurrence of R. boylii among habitat
categories due to mechanistic or
physiological restriction (i.e., "habi-
tat preference") from hypotheses in-
voking habitat restriction because of
some novel predator (Hayes and Jen-
nings 1986). The data for R. boylii dif-
fer from that of R. a. draytonii in that
we cannot confidently reject the al-
ternative that no restriction is occur-
ring. For example, it remains unclear
whether earlier reports of "atypical"
habitat use by R. boylii were simply
rare occurrences, or whether those
instances actually reflect a general
pattern of broader habitat use in
years prior to when Moyle (1973) ob-
tained his data, indicating that habi-
tat restriction had occurred.

Management Implications

Both R. a. draytonii and R. boylii need
immediate management considera-
tion if many remaining populations
are to survive into the next century.
Rana a. draytonii is extinct on the
floor of the Central Valley, and is
probably extinct from over half of the
drainage systems in the Central Val-
ley from where it was historically re-
corded. We consider many of the

153

remaining populations at risk since
over half of the localities are within
areas projected to be flooded by res-
ervoirs proposed for the Coast Range
slope of the Central Valley (Wemette
et al. 1980; C. J. Brown, Jr., pers.
comm.). Populations at an additional
10 localities are at an unknown, but
probably high level of risk. Although
these additional localities will not be
flooded by the proposed reservoirs,
flooding will isolate the frogs present
in small (<10 km^) drainage basins
upstream of the reservoirs. We lack
data on how isolation in very small
drainage basins may increase the
probability of extinction (see Fritz
1979), but the only four localities iso-
lated by reservoirs for which data
exist now lack red-legged frogs
(Hayes, unpubl. data). California
red-legged frogs were recorded at
each of the latter sites up to 20 years
ago, between one and five years after
flooding of the adjacent reservoir
had taken place. Comparable data on
the decline of R. boylii in the Central
Valley are lacking, but observations
by experienced workers indicate that
R. boylii no longer occurs at many
localities in the Central Valley drain-
age basin where it was historically
recorded (Moyle 1973; R. Hansen, D.
Holland, S. Sweet, D. Wake, pers.
comm.; Jennings, unpubl. data).

Modal habitat requirements for
both frog species suggested by exist-
ing data should be given special at-
tention in any management attempt.
Since our comments here are based
on data for both species in the Cen-
tral Valley of California, attempts to
apply the management recommenda-
tions we make to other areas within
the geographic range of each species
should be done cautiously. We can-
not overemphasize that preservation
of what appears to be the preferred
(modal) habitat condition for either
species should be stressed where it is
ambiguous whether restriction is due
either to the negative impact of the
introduced aquatic macrofauna, or to
intrinsic mechanical or physiological
limitations. Preservation of non-mo-

dal habitat is not only likely to incur
a greater cost to ensure frog survival,
but more importantly, it may still not
allow survival if the worst-case sce-
nario (restriction of habitat by the
introduced aquatic macrofauna) is
true.

The modal habitat features of R. a.
draytonii and R. boylii are similar in
two ways. First, the aquatic habitat
of each has some shading. Yet, shad-
ing associated with California red-
legged frogs differs because of the
apparently crucial aspect of having
dense vegetation at or near water
level. We lack details on just how the
streams Moyle (1973) sampled were
shaded, but knowledge of some of
the species providing shade suggests
that a higher overstory was typical.
Rana a. draytonii will always be at
greater risk than R. boylii where al-
teration of riparian vegetation is a
problem simply because of its shade
requirement; even altered stream en-
vironments may retain some shad-
ing, but a lesser probability will al-
ways exist that the shading that re-
mains will have the structure needed
by R. a. draytonii. Second, each spe-
cies occurs most frequently in the ab-
sence of any aquatic macrofauna,
and both species have probably expe-
rienced some habitat restriction due
to introduced aquatic predators.
Only one small native minnow co-
occurs at over one-third the sites
where each frog species was re-
corded, and even that species was
not positively correlated with frog
abundance. For R. a. draytonii, the
data are reasonably convincing that
restriction has occurred away from
perennial aquatic sites. For R. boylii,
data do not clearly indicate habitat
restriction. Still, the fact that R. boylii
was found at fewer intermittent sites
leads us to believe that if habitat re-
striction has taken place, it has oc-
curred away from intermittent
aquatic sites. We reason that since
riffles disappear seasonally in inter-
mittent streams, such streams lack
the condition found in perennial
streams that may be an advantage if

riffle habitat is a refuge, i.e., that per-
ennial streams have riffle habitat
year-round.

CXir analysis indicates that at-
tempts at management of these two
frogs should address at least three
other habitat variables: water depth,
stream morphology, and substrate
type. Rana boylii appears to require a
shallow water depth of <0.6 m,
whereas R. a. draytonii seems to re-
quire some water _0.7 m deep. Data
on stream morphology and substrate
type, which were recorded only for
R. boylii, suggest that both of a per-
centage of riffle area and at least
cobble-sized substrate of greater than
40% best suit this species. Parallel
data for R. a. draytonii are lacking,
but since data on other habitat para-
meters measured for R. a. draytonii
are largely "reciprocals" of the corre-
lates of riffle habitat associated with
R. boylii, we anticipate that some re-
lationship to the more lentic water
stream morphology categories (i.e.,
pools and runs) and their associated
finer substrate categories (i.e., silt
and sand) will be demonstrated for
R. a. draytonii.

Experiments may ultimately iden-
tify the introduced aquatic predators
likely responsible for the declines of
these frogs, but management based
on current knowledge should ad-
dress no less than the worst-case sce-
nario; i.e., that any member of the
introduced aquatic macrofauna pres-
ents a risk to the survival of popula-
tions of R. a. draytonii and R. boylii.
Thus, the sound management deci-
sion is to implement measures that
will maximize the degree of isolation
between existing populations of each
frog species and any members of the
introduced aquatic macrofauna. Just
how isolation should be maintained
will vary depending on the site con-
sidered, but some general sugges-
tions can be made. First, passive
measures promoting isolation are
preferable because they are less
costly and are less likely to affect
non-target species. Simply avoiding
habitat modification where the mo-

154

dal habitat features for each frog spe-
cies already exist is a passive meas-
ure that will provide some degree of
within-habitat isolation since mem-
bers of the introduced aquatic
macrofauna show little overlap in
their habitat requirements with each
frog. Yet, populations of either frog
species currently coexisting in a habi-
tat mosaic with members of the in-
troduced aquatic macrofauna may
still be doomed. This possibility
leads us to suggest that most efforts
at management should be spent on
frog populations at sites that cur-
rently lack introduced aquatic preda-
tors. We consider protection of the
entire hydrographic basins of drain-
age systems tributaries (see methods
for definition) an important part of
such management attempts because
intrusion by introduced aquatic
predators is probably most easily
controlled if the only natural access
route is via upstream movement. To
our knowledge, no locality within the
Central Valley drainage area having
an extant California red-legged frog
population has its entire hydro-
graphic basin protected. Moreover,
only two California red-legged frog
populations within this area occur at
sites where the habitat is currently
offered some protection. Second, iso-
lation strategies may differ depend-
ing on whether proximate popula-
tions of introduced aquatic predators
are bullfrogs or fishes or both. Apart
from being physically transported,
fishes are effectively prevented from
moving upstream by a barrier (see
Hayes and Jennings 1986), whereas
bullfrogs, capable of overland move-
ment under wet conditions (Hayes
and Warner 1985), are less likely to
be barrier-limited. We indicated ear-
lier that creation of small impound-
ments may enhance the ability of R.
a. draytonii to establish at certain sites
through the creation of features
found in its habitat, but attention to
the positioning of such impound-
ments is an equally important con-
siderahon. If impoundments are
close enough that bullfrogs reach

them from an adjacent source popu-
lation, such sites can also act as local
refuges at which new bullfrog popu-
lations can become established, and
can serve as new focal points from
which to disp)erse. Moreover, new
impoundments probably favor the
establishment of bullfrogs simply be-
cause their unvegetated condition
more closely matches the habitat re-
corded for bullfrogs (Moyle 1973).
These arguments simply indicate that
particular attention should be given
to avoiding the creation of "step-
ping-stone" pathways, i.e., provision
of access into currently isolated
drainages by the positioning of im-
poundments that permit introduced
predators, like bullfrogs, to encroach
progressively by dispersal.

The limits of our analysis indicate
that significant aspects of habitat
variation for both frog species re-
main to be understood. In particular,
an understanding is needed as to
how key variables influence repro-
duction and refuge sites. Although
available data on oviposition pat-
terns suggest a link between R. a.
draytonii and the presence of emer-
gent vegetation (Hayes and
Miyamoto 1984), and R. boylii and a
rocky substrate (Fitch 1936, 1938;
Storer 1925; Zweifel 1955), it is un-
clear for either species to what de-
gree the substrate can vary before
oviposition may be prevented and
also how aspects of reproduction be-
sides oviposition may be linked to
habitat variation. Perhaps the most
crucial gap is a lack of understanding
of what aspects of habitat variation
are related to frog refuge sites, in-
cluding the often temix)rary refuges
used as an escape from predators as
well as those refuges used during the
season of inactivity. The former type
of refuge site may be related to the
deep-water and dense vegetation
habitat associated with R. a. draytonii,
and the riffle habitat associated with
R. boylii, but what aspects of those
habitat features really comprise the
refuge and to what degree they may
vary before they are no longer a ref-

uge is unknown. A understanding of
the latter is pivotal to the identifica-
tion of predator-induced habitat re-
striction. Most importantly, an
understanding of how reproduction
and refuge sites are related to habitat
variation for these two frogs is essen-
tial if management is to ever be re-
fined to a level where habitat vari-
ables, either individually or in con-
cert, may be manipulated. Finally, if
habitat manipulations are attempted,
they will have to be implemented
with caution in aquatic systems
where both R. a. draytonii and R.
boylii co-occur; differences in habitat
characteristics between each species
suggest that whatever way one or
more of several habitat variables are
manipulated, they will probably re-
sult in a tradeoff between habitat
losses and habitat gains for R. a. dray-
tonii versus R. boylii.

In summary, habitat analysis for
the two ranid frogs, R. a. draytonii
and R. boylii, indicates that each spe-
cies is most frequently associated
with discemibly different aquatic
habitats, the former with densely
vegetated, deep water and the latter
with rocky, shallow-water riffles in
streams. The species are similar in
that they infrequently co-occur with
any aquatic vertebrates, especially
the introduced aquatic macrofauna.
Low levels of co-occurrence between
frogs and the introduced aquatic
macrofauna have two confounded
explanations: 1) preferential use of
different habitats between the intro-
duced aquatic macrofauna and frogs,
and 2) habitat restriction because
frogs and their life stages are preyed
upon by the introduced aquatic
macrofauna. However, even though
it is presently impossible to idenrify
the responsible predator, temporal
data strongly suggest that R. a. dray-
tonii has been restricted by some in-
troduced aquatic predator and the
same possibility cannot be excluded
for R. boylii. For t>oth species, a man-
agement scheme is necessary to avert
existing trends of decline, and ulti-
mately, extinction. A management

155

scheme that minimizes the risk of ex-
tinction based on current data must
address the worst-case scenario
among the ahernatives imphcated in
limiting frog distributions. To ad-
dress anything less increases the risk
of extinction if that alternative is
true. Since that alternative is habitat
restriction by an introduced aquatic
macrofauna, management should
strive to isolate both frog species
from the introduced aquatic macro-
fauna. Moreover, available data indi-
cate that preservation of modal con-
ditions for habitat variables identi-
fied as associated with each species is
a suitable interim strategy, since it is
more likely to promote isolation. Sig-
nificant refinements of this manage-
ment scheme will require a thorough
understanding of how habitat vari-
ables associated with each frog spe-
cies are linked to their refuge re-
quirements and their reproductive
patterns.

ACKNOWLEDGMENTS

Special thanks go to Charles J.
Brown, Jr., Peter B. Moyle, and
David B. Wake for allowing us to use
data in their care.

Sean J. Barry, John M. Brode,
Charles W. Brown, Mark L.
Cay wood, Henry E. Childs, Jr.,
Arthur L. Cohen, Nathan W. Cohen,
Lawrence R. Cory, John B. Cowan,
Robert G. Crippen, Henry S. Fitch,
William J. Hamilton, Jr., George H.
Hanley, George E. Hansen, Robert
W. Hansen, John Hendrickson (Woo-
dleaf, Calif.), John R. Hendrickson
(University of Arizona), Daniel C.
Holland, Samuel B. Horowitz, Alex-
ander K. Johnson, William F.
Johnson, Donald R. Kirk, J. Ralph Li-
chtenfels. Amy R. McCune, Roy W.
McDiarmid, Milton D. Miller, Rich-
ard R. Montanucci, Garth I. Murphy,
Robert T. Orr, Thomas L. Rodgers,
Stephen B. Ruth, Robert C. Stebbins,
the late Ruth R. Storer, Samuel S.
Sweet, Richard Terry, Walter Tor-
doff, Jens V. Vindum, Conrad

Yamamoto, and Richard G. Zweifel
all contributed ancillary data. Addi-
tionally, important data were ex-
tracted from the unpublished field
notes or voucher specimens collected
by the following workers no longer
living: Adrey E. Borell, Harold C.
Bryant, Charles L. Camp, Joseph S.
Dixon, Adolphus L. Heermann,
Henry W. Henshaw, Carl L. Hubbs,
Lloyd G. Ingles, Henry C. Kellers,
William N. Lockington, Donald R.
McLean, Joseph R. Slevin, Tracy I.
Storer, John Van Denburgh, and Al-
bert H. Wright. Phyllis A. Buck, Peter
B. Moyle, C. Mindy Nelson, and
Richard G. Zweifel kindly reviewed
the manuscript.

LITERATURE CITED

Abdulali, Humayun. 1985. On the
export of frog legs from India.
Journal of the Bombay Natural
History Society 82:347-375.

Armour, C. L., K. P. Bumham, and
W. S. Platts. 1983. Field methods
and statistical analyses for moni-
toring small salmonid streams. 200
p. United States Fish and Wildlife
Service, FWS/OBS-83/33.

Bailey, James A. 1984. Principles of
wildlife management. 373 p. John
Wiley and Sons, Inc., New York.

Baker, RoUin H. 1942. The bullfrog: A
Texas wildlife resource. Texas
Game, Fish and Oyster Commis-
sion, Bulletin (23):3-7.

Brocke, R. H. 1979. The name of
nongame. Wildlife Society Bulletin
7:279-282.

Bury, R. Bruce. 1975. Conservation of
non-game wildlife in California: A
model programme. Biological
Conservation 7:199-210.

Bury, R. Bruce, Howard W.

Campbell, and Norman J. Scott, Jr.
1980a. Role and importance of
nongame wildlife. 197-207 p. In
Transactions of the 45th North
American Wildlife and Natural
Resources Conference, Wildlife
Management Institute, Washing-
ton, D.C.

Bury, R. Bruce, C. Kenneth Dodd, Jr.,
and Gary M. Fellers. 1980b. Con-
servation of the Amphibia of the
United States: A review. Depart-
ment of the Interior, United States
Fish and Wildlife Service, Re-
source Publication (134):l-34.

Bury, R. Bruce, and Jill A. Whelan.
1984. Ecology and management of
the bullfrog. Department of the
Interior, United States Fish and
Wildlife Service, Resource Publica-
tion (155):l-23.

Chamberlain, F. M. 1898. Notes on
the edible frogs of the United
States and their artificial propaga-
tion. 249-261 p. In Report of the
United States Commissioner of
Fish and Fisheries for the Year
Ending June 30, 1897. United
States Government Printing Of-
fice, Washington, D.C.

Childs, Henry E., Jr., and Walter E.
Howard. 1955. The vertebrate
fauna of the San Joaquin Experi-
mental Range. United States De-
partment of Agriculture, Califor-
nia Range and Experiment Station,
Miscellaneous Paper (19):l-20.

Cowan, J. B. 1979. Mammals, reptiles
and amphibians of Gray Lodge
Wildlife Area. 12 p. Publication of
the Department of Fish and Game,
Resources Agency, State of Cali-
fornia, Sacramento.

Cresswell, Richard. 1862. Aristotle's
history of animals in ten books.
326 p. Henry G. Bohn, York Street,
Covent Garden, London.

Cunningham, John D. 1955. Notes on
abnormal Rjina aurora draytonii.
Herpetologica 11:149.

Dodd, C. Kenneth, Jr. 1978. Amphibi-
ans and reptiles: the declining spe-
cies. Water Spectrum 10:24-32.

Fitch, Henry S. 1936. Amphibians
and reptiles of the Rogue River
Basin, Oregon. American Midland
Naturalist 17:634-652.

Fitch, Henry S. 1938. Rana hoylii in
Oregon. Copeia 1938:148.

Fitch, Henry S. 1949. A study of
snake populations in central Cali-
fornia. American Midland Natu-
ralist 41:513-579.

156

Fritz, R. S. 1979. Consequences of in-
sular population structure: distri-
bution and extinction of spruce
grouse (Camchites canadensis)
populations. Oecologia 42:57-66.

Gordon, Kenneth. 1939. The amphib-
ian and reptiles of Oregon. Oregon
State College Monographs (l):l-82.

Grinnell, Joseph, Joseph Dixon, and
Jean M. Linsdale. 1930. Vertebrate
natural history of a section of
northern California through the
Lassen Peak region. University of
California Publications in Zoology
35:1-594.

Grinnell, Joseph, and Tracy I. Storer.
1924. Animal life in the Yosemite;
an account of the mammals, birds,
reptiles, and amphibians in a
cross-section of the Sierra Nevada.
752 p. University of California
Press, Berkeley.

Hallowell, Edward. 1854. Descrip-
tions of new reptiles from Califor-
nia. Proceedings of the Academy
of Natural Sciences of Philadel-
phia 7:91-97.

Hallowell, Edward. 1859. Report on
the reptiles collected on the Sur-
vey. 27 p. Pacific Railroad Reports
10 (part 4, Williamson's Report),
zoology report no. 1.

Hayes, Marc P., and Mark R. Jen-
nings. 1986. DecHne of ranid frog
species in western North America:
Are bullfrogs (Rana catesbeiana)
responsible? Journal of Herpetol-
ogy 20:490-509.

Hayes, Marc P., and Dana M. Krem-
pels. 1986. Vocal sac variation
among frogs of the genus Rana
from western North America.
Copeia 1986:927-936.

Hayes, Marc P., and Michael M.
Miyamoto. 1984. Biochemical, be-
havioral and body size differences
between Rana aurora aurora and R.
a. draytoni. Copeia 1984:1018-1022.

Hayes, Marc P., and Joan Warner.
1985. Life history notes: Rana
catesbeiana, food. SSAR Herpeto-
logical Review 16:109.

Hine, Ruth L., Betty L. Les, and
Bruce F. Hellmich. 1981. Leopard
frog populations and mortality in

Wisconsin. Department of Natural
Resources, Madison, Wisconsin,
Technical Bulletin (122):l-39.

Holland, Robert F. 1978. The geo-
graphic and edaphic distribution
of vernal pools in the Great Valley,
California. California Native Plant
Society (4):1-12.

Honegger, Rene E. 1981. Threatened
Amphibians and Reptiles in Eu-
rope. 158 p. Akademie Verlagsges,
Wiesbaden, Germany.

Husain, K. Z., and M. M. Rahman.
1978. The amphibian fauna of
Bangladesh. Bangladesh Journal of
Zoology 6:157-158.

Ingles, Lloyd G. 1932a. Four new
species of Haematoloechus (Trema-
toda) from Rana aurora draytonii
from California. Univ. of Calif.
Publications in Zool. 37(7):189-201.

Ingles, Lloyd G. 1932b. Cephalogoni-
mus brevicirrus, a new species of
Trematode from the intestines of
Rarm aurora from California. Uni-
versity of California Publications
in Zoology 37(8):203-210.

Ingles, Lloyd G. 1933. The specificity
of frog flukes. Science 78:168.

Ingles, Lloyd G. 1936. Worm para-
sites of California amphibia.
Transactions of the American
Microscopical Society 55:73-92.

Jain, Subodh. (editor) 1976. Vernal
pools: their ecology and conserva-
tion. A symposium sponsored by
the Institute of Ecology, Univer-
sity of California, Davis, May 1
and 2, 1976. Institute of Ecology
Publication (9):l-93.

Jennings, Mark R., and Marc P.
Hayes. 1985. Pre-1900 overharvest
of California red-legged frogs
(Rana aurora draytonii): The in-
ducement for bullfrog (Rana
catesbeiana) introduction. Herpetol-
ogica 41:94-103.

Kellert, Stephen R. 1985. Social and
perceptual factors in endangered
sp)ecies management. Journal of
Wildlife Management 49:528-536.

Leidy, R. A. 1984. Distribution and
ecology of stream fishes of the San
Francisco Bay drainage. Hilgardia
52(8):1-175.

Leopold, Aldo S. 1931 . A history of
ideas in game management. Out-
door America (June):22-24,38-
39,47.

Leopold, Aldo S. 1933. Game man-
agement. 481 p. Charles Scribner's
Sons, New York.

McAuliffe, J. R. 1978. Biological sur-
vey and management of sport-
hunted bullfrog populations in
Nebraska. 78 p. Nebraska Game
and Parks Commission, Lincoln,
Nebraska.

Moyle, Peter B. 1973. Effects of intro-
duced bullfrogs, Rana catesbeiana,
on the native frogs of the San
Joaquin Valley, California. Copeia
1973:18-22.

Moyle, Peter B. 1976. Inland fishes of
California. 405 p. University of
California Press, Berkeley.

Moyle, Peter B., and Robert D.
Nichols. 1973. Ecology of some
native and introduced fishes of the
Sierra Nevada Foothills in Central
California. Copeia 1973:478-490.

Moyle, Peter B., and Robert D.
Nichols. 1974. Decline of the fish
fauna of the Sierra Nevada foot-
hills. Central California. American
Midland Naturalist 92:72-83.

Moyle, Peter B., Jerry J. Smith,
Roberts A. Daniels, Thomas L.
Taylor, Donald G. Price, and
Donald M. Baltz. 1982. Distribu-
tion and ecology of stream fishes
of the Sacramento-San Joaquin
drainage system, California. Uni-
versity of California Publications
in Zoology 115:1-256.

Neill, Wilfred T. 1974. Reptiles and
amphibians in the service of man.
248 p. Biological Sciences Curricu-
lum Study Book, Pegasus Press,
The Bobbs-Merrill Company, Inc.,
Indianapolis, Indiana.

Pister, E. P. 1976. A rationale for the
management of nongame fish and
wildlife. Fisheries 1:11-14.

Rutter, Cloudsley. 1908. The fishes of
the Sacramento-San Joaquin Basin,
with a study of their distribution
and variation. Bulletin of the
United States Bureau of Fisheries
27:103-152.

157

Slevin, Joseph R. 1928. The amphibi-
ans of western North America.
Occasional Papers of the Califor-
nia Academy of Sciences (16):1-
144.

Sokal, Robert R., and R. James Rohlf.
1981. Biometry. 687 p. W. H. Free-
man and Co., San Francisco.

Stebbins, Robert C. 1966. A field
guide to western reptiles and am-
phibians. 279 p. Houghton Mifflin
Company, Boston.

Stebbins, Robert C. 1985. A field
guide to western reptiles and am-
phibians. Second edition, revised.
336 p. Houghton Mifflin Com-
pany, Boston.

Storer, Tracy 1. 1925. A synopsis of
the Amphibia of California. Uni-
versity of California Publications
in Zoology 27:1-342.

Storer, Tracy 1. 1933. Frogs and their
commercial use. California Fish
and Game 19:203-213.

Strahler, A. N. 1957. Quantitative
analysis of watershed geomor-
phology. Transactions of the
American Geophysicists Union
38:913-920.

Sweet, Samuel S. 1983. Mechanics of
a natural extinction event: Ram
boylii in southern California. Ab-
stract of a paper presented at the
26th annual meeting of the Society
for the Study of Amphibians and
Reptiles, and the 31st annual meet-
ing of the Herpetologists' League,
University of Utah, August 7-12,
1983.

Treanor, Robert R. 1975a. Manage-
ment of the bullfrog (Ram catesbe-
iam) resource in California. Cali-
fornia Department of Fish and
Game, Inland Fisheries Adminis-
trative Report (75-l):l-30.

Treanor, Robert R. 1975b. Manage-
ment of the bullfrog resource in
California. Cal-Neva Wildlife
Transactions 1975:81-92.

Treanor, Robert R., and Stephen J.
Nicola. 1972. A preliminary study
of the commercial and sporting
utilization of the bullfrog. Ram
catesbeiam Shaw, in California.
California Department of Fish and

Game, Inland Fisheries Adminis-
trative Report (72-4):l-23.

United States Geological Survey.
1970a. Surface water supply of the
United States 1961-1965. Part II.
Pacific slope basin in California.
Volume 3. Southern Central Valley
basins. U. S. Geological Survey,
Water-Supply Paper (1930):1-155.

United States Geological Survey.
1970b. Surface water supply of the
United States 1961-1965. Part II.
Pacific slope basin in California.
Volume 4. Northern Central Val-
ley basin. Uruted States Geological
Survey, Water-Supply Paper
(1931):1-111.

Walker, M. V. 1946. Reptiles and
amphibians of Yosemite National
Park. Yosemite Nature Notes 25:1-
48.

Wemette, F. G., F. A. Hall, Jr., C. J.
Brown, Jr., C. L. Mayer, and N. A.
Villa. 1982. Los Vaqueros project —
fish and wildlife impacts. 263 p.
California Department of Fish and
Game, Status Report.

Williamson, R. S. 1855. Report of
Lieutenant R. S. Williamson. 43 p.
In Report of Exploration in Cali-
fornia for Railroad Routes, to con-
nect with the Routes Near the 35th
and 32nd Parallels of North
Latitude; In Report of Explora-
tions and Surveys to Ascertain the
Most Practicable and Economical
Route for a Railroad from the Mis-
sissippi River to the Pacific Ocean,
Made Under the Direction of the
Secretary of War, in 1853-54. Vol-
ume 5, part I. Beverly Tucker,
Washington, D.C. (1856) [Senate
Executive Document No. 78, 33rd
Congress, 2nd Session].

Willis, Yuell L., Don L. Moyle, and
Thomas S. Baskett. 1956. Emer-
gence, breeding, hibernation,
movements and transformation of
the bullfrog. Ram catesbeiam, in
Missouri. Copeia 1956:30-41.

Wright, A. H. 1920. Frogs: their natu-
ral history and utilization. Depart-
ment of Commerce, United States
Bureau of Fisheries Document
(888):l-44.

Wright, Albert Hazen, and Anna Al-
len Wright. 1949. Handbook of
frogs and toads of the United
States and Canada. Third edition.
640 p. Comstock Publishing Com-
pany, Inc., Ithaca, New York.

Zahl, Paul A. 1967. In quest of the
world's largest frog. National
Geographic Magazine 132:146-152.

Zar, Jerrold H. 1974. Biostatistical
analysis. Prentice-Hall, Inc. 620 p.
Englewood Cliffs, New Jersey.

Zedler, Paul H. 1987. The ecology of
southern California vernal pools: a
community profile. Department of
the Interior, United States Fish
and Wildlife Service, Biological
Report 85(7.11):1-136.

Zweifel, Richard G. 1955. Ecology,
distribution, and systematics of
frogs of the Ram boylei group.
University of California Publica-
tions in Zoology 54(4):207-292.

158

Integrating Anuran
Amphibian Species into
Environmental Assessment
Programs^

Ronald E. Beiswenger^

Abstract.— Anurans are often given minimal
attention in environmental assessments despite their
ecological importance and potential value as
indicator species. Habitat and guild-based models
must be adopted to include all life cycle stages of
anurans. A preliminary habitat suitability model for
the American toad shows how this con be
accomplished.

As a result of our increased under-
standing of the roles of wildlife spe-
cies in ecosystem structure and func-
tion, and legal requirements to de-
velop holistic approaches to environ-
mental management, it has become
increasingly common to include all
species of wildlife in resource inven-
tories and monitoring programs
(Chalk et al. 1984). However, am-
phibians are often ignored or given
minimal attention in such programs,
even though they are important
wildlife resources and should be
given serious consideration in man-
agement evaluations (Bury and Ra-
phael 1983, Bury et al. 1980, Jones
1986). If included in resource evalu-
ations at all, amphibians are usually
lumped with reptiles in a category
called herpetofauna and even then
are often only represented as items in
a species list.

This is unfortunate because, in
addition to their ecological impor-
tance, anurans are potentially valu-
able as a unique form of indicator
species capable of integrating envi-
ronmental changes occurring in both
the terrestrial and aquatic phases of
their habitats. Furthermore, because
they occupy small ponds and the
shallow margins of lakes, anurans

' Paper presented at symposium. Man-
agement of Amphibians, Reptiles and Small
Mammals in North America. (Flagstaff. AZ.
July 19-21. 1988.)

'^Ronald E. Beiswenger is Professor. De-
partment of Geography and Recreation.
The University of Wyoming. Laramie, WY
82071.

are likely to be the first vertebrates to
come in contact with contaminated
run-off or acidified snowmelt. This
could make them useful as elements
of an early warning system for the
detection of environmental contami-
nation. Campbell (1976) found that
the boreal toad, Bufo boreas, would be
an especially effective indicator spe-
cies for monitoring the impact of
cloud seeding in the mountains of
Colorado. It is also significant that
many anurans require specialized
habitats in wetland areas and ripar-
ian zones, and could serve as indica-
tor species for the overall health of
these areas of special ecological im-
portance.

Despite their potential usefulness,
there are several reasons why am-
phibians are not given adequate at-

tention in environmental assess-
ments. The importance of amphibi-
ans in ecosystems is generally unrec-
ognized, particularly by the general
public and the resource managers
who must respond to the desires of
this public as they set management
priorities. Also, the secretive habits
during the non-breeding season, and
complex life cycles of amphibians
make them relatively difficult to
study. Consequently, the natural his-
tory of many amphibian species is
not well known. Another factor is
that current models for monitoring
and assessment have been developed
for either terrestrial or aquatic spe-
cies and have not been adapted to
species with divergent life cycle
stages which depend on both aquatic
and terrestrial habitats (table 1).

Table 1 .— Habitat components and life cycle stages of anurans.

Habitat
component

Eggs/Pre-
feeding
tadpoles

Feeding IVIetamorphosing

tadpoles tadpoles Juveniles

Adults

Aquatic Phase

Spawning sites X

Tadpole habitat X

Aquatic/Terrestrial Interface Phase

Tadpole habitat

Juvenile habitat

Terrestrial Phase

Summer habitat

Hibernation sites

Movement corridors

Interspersion Factors

Distribution of habitat components

Density of habitat components

X

X

X
X

X

X
X
X

X
X
X

X X
X X

159

Approaches for incorporating
wildlife into resource evaluations in-
clude inventories of relative abun-
dance and species richness, develop-
ment of databases, the use of indica-
tor species, and the development of
species diversity indices and models
using guild concepts. However, the
application of these approaches to
species of Amphibia has not kept
pace with applications to other spe-
cies of vertebrates.

The primary purpose of this paper
is to suggest ways to use single spe-
cies models, and models which use
guilds and habitat structure, to more
effectively integrate anuran amphibi-
ans into resource assessments. A
single species model for the Ameri-
can toad, stressing the importance of
tadpole habitat, is presented in some
detail.

Models for Anurons

Guilds and Habitat Structure

Guild-based environmental assess-
ments are especially useful from an
ecological perspective, although they
are most effective when used in com-
bination with other methods (Karr
1987). Unfortunately, when amphibi-
ans are included in guild-based pro-
grams they are usually considered
too simplistically. A common proce-
dure is to categorize them according
to their general spawning and feed-
ing habitat, but to include no further
detail (e.g. see Thomas et al. 1979).

The habitat models developed for
Arizona (Short 1984) represent a
good starting point for producing
effective models for anurans. In these
models wildlife guilds are used to
correlate habitat use with habitat
structure (layers) by associating a
species with a particular plant com-
munity (habitat or cover type), and
then with a habitat layer. Layers of
both terrestrial and aquatic habitat
are included.

This system is as appropriate for
terrestrial adult anurans as it is for
any small, terrestrial vertebrate.
However, the aquatic phases of the
model require further development if
it is to be used with the aquatic larval
stages of amphibians. The adaptive
significance of the tadpole stage has
been established by Wassersug (1975)
and Wilbur (1980), and it is clear that
the habitat requirements of larval
anurans should be an important
component of habitat models. The
selection of a spawning site that will
provide high quality habitat for the
tadpole stage is likely to be critical to
the evolutionary success of an anu-
ran species.

Single Species Models

Habitat models for indicator species
have been developed by the U.S. Fish
and Wildlife Service (1981), the U.S.
Forest Service (Berry 1986) and oth-
ers (e.g. Clawson et al. 1984) for use
in assessing environmental impacts
and in making management deci-
sions. A comprehensive habitat
model for an anuran species must
encompass spawning sites, tadpole
habitat, metamorphic sites, juvenile
and adult feeding habitat, movement
corridors and hibernation sites. For
example, a model developed for the
bullfrog (Rana catesbeiana) illustrates
how the approach can be applied to

Table 2.— Components of habitat for Bufo americanus (measurable attrib-
ute in parentheses).

Spawning Habitat

Shallow, emphemeral ponds (depth range)
m^m:: Emergent or submergent vegetation (% cover)
■ Exposure to direct sunlight (% of area shaded)

Tadpole Habitat

Ponds with access to shallow shoreline areas (< 10 cm) and to

deeper areas (10-100 cm)
Substrates with food W-:W-r

periphyton (% cover)

bottom areas with detritus or microorganisms (% cover)
Microorganisms suspended in water column (density)
Exposure to direct sunlight (% of area shaded)

Metamorphic Habitat

Shallow depth gradient at shoreline (< 1 0 cm)
Exposure to direct sunlight (% of area shaded)
Moist substrate on shore (moisture content)
Vegetative cover on shore (% cover)

Juvenile and Adult Habitat

Availability of insect and other invertebrate prey (prey density)
Access to moist substrates and refugia (moisture content and refu-
gia density)

Access to vegetative cover (distance to cover)

Hibernation Site

Unoccupied animal burrows (burrow density)

Friable soils (soil texture)

Root zones of large trees (large tree density)

Interspersion

Movement corridors between hibernation and spawning sites (distri-
bution of continuous open areas with adequate cover)

Distribution and density of potential spawning sites within the home
range of the population (density of spawning sites)

160

an anuran species that is primarily
aquatic (Graves and Anderson 1987).
While this model is well constructed,
a different modeling approach would
be needed for anurans with terres-
trial adult stages. A limitation of the
bullfrog model is that the habitat re-
quirements of the tadpole stage are
not given in sufficient detail. This is
important because the larval stage
(up to three years in duration) repre-
sents a significant proportion of a
bullfrog's total lifespan.

A different array of habitat com-
ponents for a species that is predomi-

HABITAT VARIABLES

nantly terrestrial is an adult, the
American toad (Bufo americanus) is
outlined in table 2. This outline is
based on extensive field studies in
Michigan (Beiswenger 1975, 1977),
field observations of related toad
species in Oregon and Wyoming
(Beiswenger 1978, 1981, 1986), and
information found in the literature.

Including the terrestrial features of
toad habitat in assessments does not
represent a particularly difficult chal-
lenge because these features can be
described using well-established ap-
proaches developed for other small

COMPONENTS

Percent of water area 1 m or less
in depth (VI)

Percent cover of rooted aquatic
vegetation (V2)

Percent of shoreline v/ith shading
riparian vegetation (V3)

Percent of shoreline v/ith strip of
invegetated shallow water (V4)

Percent of shoreline with terrestrial
vegetative cover or ground debris
within 1 m of water (V5)

Percent tree canopy closure (V5)

Percent of trees that are deciduous
species (V7)

Percent herbaceous canopy cover (V8)

Number of burrows, decaying logs, and
debris objects larger than 20 cm in
diameter on the ground (V9)

Distance along a protected dispersal
corridor to potential spawning
sites (V10)

_ Aquatic cover/
reproduction

Jerrestrial cover/
hibernation

Interspersion

Figure 1 .—Relationships of habitat variables to components of an HSI model for the Ameri-
can toad.

vertebrates that live on and below
the surface of the ground. However,
tadpole habitat is also important and
must be incorporated into habitat as-
sessment procedures. This is some-
what more challenging because less
is known about tadpole ecology and
techniques for describing tadpole
habitat are not well developed.

A Habitat Model for the American
Toad

A preliminary version of a habitat
suitability model for the American
toad is described here to show how
the requirements of all life cycle
stages could be incorporated into
such a model (figs. 1 and 2). The
model includes 10 variables and is
based primarily on the author's expe-
rience and a partial literature review.
Consequently, the model should be
refined through a more extensive
analysis of the literature and a peer
review process before it is field
tested.

The habitat requirements of
spawning adults and tadpoles are
included in the aquatic cover/repro-
ductive component of the model. The
quality of spawning sites selected by
American toads is influenced by
structural features such as depth gra-
dients and vegetation. Adult toads
typically lay their eggs in shallow,
unshaded, vegetated areas (variables
2 and 3), distributing them in strands
on the vegetation. At first the newly
hatched tadpoles do not feed, but
remain at the site where the eggs
were laid.

Older tadpoles are active swim-
mers and display a variety of feeding
modes that arc influenced to a large
measure by structural features of the
habitat (e.g. aquatic vegetation and
depth gradients) (variables 1, 2, and
4). Wassersug (1975) has shown that
tadpoles are essentially non-discrimi-
nant suspension feeders, although
they use a variety of means for ob-
taining food. Tadpoles of the Ameri-
can toad most commonly graze

161

t 0.5-

100!?

PERCEN"0- WATER
AREA I n OR LE55
IN DEPTH
(Variable 1)

m%

PERCENT SHORELINE
WITH SHADING RIPARIAN
VEGETATION
(variaole 3)

PERCFHT f.OVFR OF
ROOTED ACUATIC
VEGETATION
(Variable 2)

TO-i

05-

l(50«

PERCENT SHORELINE
WITH 30-50 cn WIDE
STRIP OF UNVFGFTATFD

SHALLOW WATER
;iOcn OR LESS DEEP)
(Variable 4)

PERCENT SHORELINE
WITH VEGETATIVE
COVER OB GROUND
DEBRIS WITHIN 1 m
0- WATER
(Variable 5)

PERCENT OF TREES
THAT ARE DEC DUOUS
SPECIES
(Variable 7)

100^

PERCENT TREE
CANOPY CLOSURE
(Variable 6)

0.5-

PERCENT HERBACEOUS
CANOPY COVER
(V.¥lable a)

150 m

300 n

NUMBER OF BURROWS,
DECAYING LOGS, AND
DEORISODjECTS
LARGER THAN 20 cm
INDiAMETEROM
THE GROUND
(Va-.able

DISTANCE ALONG A
PROTECTED D 3PER5AL
CORRIDOR TO POTENTIAL
SPAWNING SITES
(VariablelO)

Figure 2.— The assumed relationships
among habitat variables and suitability
index values for the American toad.

periphyton from emergent or sub-
mergent vegetation, or scrape micro-
organisms and detritus from the
pond bottom and other substrates.
However, when blooms of sus-
pended algae are present, the tad-
poles become midwater filter feed-
ers. They also feed on organic mate-
rial supported by the surface film of
the pond. At other times, the tad-
poles are facultatively cannibalistic
or coprophagic. The particular feed-
ing mode employed is usually influ-
enced by a combination of factors
including the type of food available,
depth and temperature gradients,
vegetation structure and the degree
of social behavior exhibited by the
tadpoles (Beiswenger 1975). Most of
the time toad tadpoles feed from
substrates provided by the structural
features of their environment. Diaz-
Paniagua (1987) also found structural
features of aquatic vegetation to be
important in the distribution of the
tadpoles of five anuran species in
Spain.

Habitat use by tadpoles is strongly
influenced by temperature, which in
the shallow ponds they occupy is

highly correlated with depth and so-
lar radiation (variables 1, 3, and 4).
For example, in northern Michigan
ponds were early summer tempera-
tures varied greatly over the diel pe-
riod, toad tadpoles consistently se-
lected the warmest available water in
thermally stratified ponds
(Beiswenger 1977). Thus, they occu-
pied the deepest areas of the pond
(greater than 50 cm in depth) at
night, avoiding the shallow pond
margin where temperatures were 5.5
C cooler. During the day tadpoles
moved to shallow areas near shore
which were 9 C warmer than the
deeper areas of the pond. During
those times when there was no ther-
mal stratification (e.g. cloudy days),
or later in the summer when pond
temperatures were uniformly high,
the tadpoles used all parts of the
pond (Beiswenger 1977). These ob-
servations indicate that tadpole habi-
tat quality is partly determined by
thermal stratification associated with
depth gradients and exposure to di-
rect sunlight.

Habitat quality for mctamorphic
tadpoles is strongly influenced by

their vulnerability to predation (vari-
ables 4 and 5). As Arnold and Was-
sersug (1978, p. 1019) expressed it,
"the transforming anuran is neither a
good larva nor a good frog." The lar-
vae develop forelimbs which impede
swimming, the tail remnant on the
newly emergent juvenile interferes
with its jumping ability. Conse-
quently, the availability of structural
features such as hiding cover and
moist substrates is important for the
successful emergence and dispersal
of metamorphosing tadpoles.

Habitat quality for juvenile and
adult toads is determined by factors
generally associated with deciduous
or mixed coniferous/deciduous for-
ests. These factors include moderate
temperature regimes, invertebrate
prey density, protected microhabitats
with moist substrates, vegetative
cover, and access to hibernation sites.
Some of the variables used as surro-
gate measures of substrate moisture
and other forest floor conditions in
the HSI model for the red-spotted
newt (Sousa 1985) were adapted for
the American toad model (variables
6, 7, and 8). Juvenile and adult toads

162

also need moist cover during hot dry-
periods and for winter hibernacula.
These can be provided by soils which
are suitable for burrowing, existing
small mammal burrow systems, or
decaying logs and other debris ob-
jects on the ground (variable 9).

The American toad model in-
cludes interspersion as a habitat-re-
lated factor. Movement corridors
interconnecting spawning areas,
summer habitat and hibernation sites
are an im|X)rtant component of juve-
nile and adult habitat (variable 10).
Brode and Bury (1984) have pointed
out (cited in Ohmart and Anderson
1986), that such corridors are impor-
tant for dispersal and genetic conti-
nuity, and anurans use riparian
zones as travel lanes. Habitat frag-
mentation by road construction
(Rittschof 1975), or other forms of
habitat destruction can disrupt these
travel lanes and prevent anurans
from reaching spawning ponds or
hibernation sites.

Attention must also be paid to
other aspects of interspersion. For
example, the reproductive success of
toads depends on the continuing
availability of shallow water habitats.
Ponds with optimum spawning con-
ditions in a given year may be dry in
years with low precipitation, or too
deep in years when flooding pre-
vails. At the same time, changing wa-
ter levels may result in the availabil-
ity of new spawning sites, apparently
in response to this kind of variation,
some species of toads do not use the
same spawning site every year
(Kelleher and Tester 1969) and in
some years may not breed at all. Be-
cause of variation like this, it is im-
portant to describe the distribution of
habitat components, such as spawn-
ing sites and movement corridors, in
a broad geographic area and over a
range of environmental conditions.

Relationships among the habitat
variables and habitat components are
expressed by equations in HSI mod-
els. A value for the aquatic cover/
reproduction (SIA) component is ob-
tained by combining the suitability

index values for variables 1 through
4, as shown in the following equa-
tion.

SIA = SIVl X SIV2 x(SI\/3+SIV4)

2

This assumes that the suitability of
aquatic habitats is primarily deter-
mined by the presence of water
depths ranging from less than 10 cm
to 1 m, rooted aquatic vegetation to
provide cover and substrates for
food, and shallow, unshaded shore-
line areas.

It is assumed that terrestrial habi-
tat suitability (SIT) is determined by
the availability of cover with moist
substrates, invertebrate prey and hi-
bernation sites. The following equa-
tion shows how these habitat values
could be evaluated using variable 5
to assess cover for metamorphic
stages, 6, 7, and 8 as surrogate meas-
ures of substrate moisture, and vari-
able 9 for the availability of hibernac-
ula.

SIT= (SIV5+SIV6+SIV74-SIV9)
4

Overall habitat suitability (HSI) is
determined by combining the suita-
bility values for the aquatic (SIA) and
terrestrial (SIT) habitat components
with the suitability value for inter-
spersion (SII) as shown in the follow-
ing equation.

HSI = (SIA x SIT x 511)^^3

This form is used because a value of
zero for the suitability index for any
one of the three components indi-
cates a lack of habitat to maintain vi-
able populations of American toads.

Once it has been fully developed,
a habitat model for the American
toad could be used to assess the ef-
fects of such activities as road build-
ing, housing construction, environ-
mental pollution, landfill operations,
clearing of deciduous forests, drain-
ing or dredging of ponds and wet-
lands, intensive recreational use of
wetlands, floodplains and the shore-
line areas of lakes, and large changes
in water level by removing or intro-
ducing water.

Habitat Models and Endangered
Species Protection

The Wyoming toad (Bufo hemiophrys
baxteri) has recently been listed as
endangered by the U.S. Fish and
Wildlife Service (Baxter et al. 1982).
As of June 1988, there was only one
small breeding population known to
exist. There are no habitat models
available for this subspecies and
there have been few studies of its
natural history. This is unfortunate
because there is an urgent need to
begin a recovery program. Informa-
tion about the related Manitoba toad
(Bufo hemiophrys) which has been
more extensively studied could be
used to infer habitat relationships,
but this is obviously not as valid as
studying the Wyoming toad directly.
This situation illustrates why it is
important to intensify our efforts to
develop databases and habitat mod-
els for all species before they reach
the point of becoming endangered. It
also exemplifies the role a habitat
model can play in identifying infor-
mation gaps and focusing research
efforts.

Discussion

Resource assessments require the
development of models for the quan-
titative assessment of habitat suitabil-
ity. It is essential that such models be
developed in combination with com-
prehensive databases. A long range
goal should be to develop databases
with efficient retrieval systems so
that it is possible to access all of the
site-specific natural history informa-
tion available in the literature, and in
the files of researchers and resource
managers. The databases should also
be constructed so that information
gaps and priority areas for research
can be identified.

This paper has emphasized pro-
ducing habitat models for individual
species as if these species exist in iso-
lation. Hutto et al. (1987) have criti-
cized the overemphasis on species

163

approaches in conservation pro-
grams as too narrow and they point
out that we must not lose sight of the
higher order patterns and processes
which occur among interacting spe-
cies. They suggest supplementing the
species approach with approaches
that consider such things as land-
scape patterns that maintain ecosys-
tem level processes, the use of geo-
graphic information systems, and
other land-based approaches.

Studies emphasizing the role of
anurans in ecosystems should result
in a better understanding of ecologi-
cal process occurring at the terres-
trial-aquatic interface, and could also
contribute to more effective manage-
ment of species which depend on
these edge habitats and ecotones.

Literature Cited

Arnold, Stevan J. and Richard J. Was-
sersug. 1978. Differential preda-
tion on metamorphic anurans by
garter snakes (Thamnophis): social
behavior as a possible defense.
Ecology 59: 1014-1022.

Baxter, George T., Mark K.

Stromberg and C. Kenneth Dodd.
1982. The status of the Wyoming
toad (Bufo hemiophrys baxteri). En-
vironmental Conservation 9: 348,
338.

Beiswenger, Ronald E. 1975. Struc-
ture and function in aggregations
of tadpoles of the American toad,
Bufo americanus. Herpetologica 31:
222-233.

Beiswenger, Ronald E. 1977. Diel pat-
terns of aggregative behavior in
tadpoles of Bufo americanus, in re-
lation to light and temperature.
Ecology 58: 98-108.

Beiswenger, Ronald E. 1978. Re-
sponses of Bufo tadpoles to labora-
tory gradients of temperature.
Journal of Herpetology 12: 499-
504.

Beiswenger, Ronald E. 1981. Preda-
tion by gray jays on aggregating
tadpoles of the boreal toad (Bufo
boreas). Copeia 1981: 459-460.

Beiswenger, Ronald E. 1986. An en-
dangered species, the Wyoming
toad, Bufo hemiophrys baxteri — the
importance of an early warning
system. Biological Conservation
37: 59-71.

Berry, K.H. 1986. Introduction: de-
velopment, testing, and applica-
tion of wildlife-habitat models.
Pages 3-4. In J. Verner, M.L. Mor-
rison, and C.J. Ralph, eds. Wildlife
2000: modeling habitat relation-
ships of terrestrial vertebrates.
Univ. Wise. Press, Madison. 470 p.

Brode, J.M. and R.B. Bury. 1984. The
importance of riparian systems to
amphibians and reptiles. Pages 30-
36. In Warner, R.E. and K.M. Hen-
drix, eds. California riparian sys-
tems: ecology, conservation, and
productive management. Univ.
California Press, Berkeley. 1035 p.

Bury, R. Bruce, Howard W. Camp-
bell and Norman J. Scott, Jr. 1980.
Role and importance of nongame
wildlife. Transactions North
American Wildlife and Natural
Resources Conference 45: 197-207.

Bury, R. Bruce and Martin G. Ra-
phael. 1983. Inventory methods
for amphibians and reptiles. Pages
416-419. In John F. Bell and Toby
Atterbury, eds. Proceedings of the
international conference. Renew-
able resource inventories for
monitoring changes and trends.
Oregon State University, Corval-
lis. 737 p.

Campbell, James B. 1976. Environ-
mental controls on boreal toad
populations in the San Juan Moun-
tains. Pages 289-295. In Harold W.
Steinhoff and Jack D. Ives (eds).
Ecological impacts of snowpack
augmentation in the San Juan
Mountains, Colorado. Final report
San Juan ecology project. Colo-
rado State University Publ., Fort
Collins. 480 p.

Chalk, David E., Stephen A. Miller
and Thomas W. Hoekstra. 1984.
Multispecies inventories: integrat-
ing information on wildlife re-
sources. Wildlife Society Bulletin
12: 357-364.

Clawson, Mary E., Thomas S. Bas-
kett, and Michael J. Armbruster.
1984. An approach to habitat mod-
eling for herpetofauna. Wildlife
Society Bulletin 12: 61-69.

Diaz-Paniagua, Carmen. 1987. Tad-
pole distribution in relation to
vegetal heterogeneity in tempo-
rary ponds. Herpetological Journal
1: 167-169.

Graves, Brent M. and Stanley H. An-
derson. 1987. Habitat suitability
index models: bullfrog. U.S. Fish
Wildl. Serv. Biol. Rep. 82 (10: 138).
22 p.

Hutto, Richard L., Susan Reel and
Peter B. Landres. 1987. A critical
evaluation of the species approach
to biological conservation. Endan-
gered Species UPDATE 4(12): 1-4.

Jones, K. Bruce. 1986. Amphibians
and reptiles. Pages 267-290. In Co-
operrider, A.Y., R.J. Boyd and
H.R. Stuart eds. Inventory and
monitoring of wildlife habitat. U.S.
Department of the Interior, Bureau
of Land Management, Service
Center. Denver, Co. xviii, 858 p.

Karr, James R. 1987. Biological moni-
toring and environmental assess-
ment: a conceptual framework.
Environ. Manage. 11: 249-256.

Kelleher, Kelvin E. and John R.
Tester. 1969. Homing and survival
in the Manitoba toad, Bufo hemio-
phrys, in Minnesota. Ecology 50:
1040-1048.

Ohmart, Robert D. and Bertin W.
Anderson. 1986. Riparian habitat
Pages 169-199. In Cooperrider,
A.Y., R.J. Boyd and H.R. Stuart,
eds. Inventory and monitoring of
wildlife habitat. U.S. Dept. Inter.,
Bur. Land Manage. Service Center.
Denver, Co. xviii, 858 p.

Rittschof, Daniel. 1975. Some aspects
of the natural history and ecology
of the leopard frog, Rana pipiens.
Ph.D. Dissertation, University of
Michigan, Ann Arbor.

Short, H.L. 1984. Habitat suitability
index models: The Arizona guild
and layers of habitat models. U.S.
Fish Wildl. Serv. FWS/OBS-82/
10.70. 37 p.

164

Sousa, P.J. 1985. Habitat suitability
index models: Red-spotted newt.
U.S. Fish Wildl. Serv. Biol. Rep.
82(10.111). 18 p.

Thomas, Jack Ward (ed). 1979. Wild-
life habitats in managed forests.
The Blue Mountains of Oregon
and Washington. Agricultural
Handbook No. 553. USDA Forest
Service, Washington, D.C. 512 p.

U.S. Fish and Wildlife Service. 1981.
Standards for the development of
habitat suitability index models.
ESM103. U.S. Depart. Int. Fish
Wildl. Serv., Division of Ecological
Services. Washington, D.C. 68 p.
and appendices.

Wassersug, Richard J. 1975. The
adaptive significance of the tad-
pole stage with comments on the
maintenance of complex life cycles
in anurans. American Zoologist
15: 405-417.

Wilbur, Henry M. 1980. Complex life
cycles. Annual Review of Ecology
and Systematics 11: 67-93.

165

Preliminary Report on Effect
of Bullfrogs on Wetland
Herpetofaunas in
Southeastern Arizona^

Cecil R. Schwalbe and Philip C. Rosen^

Abstract.— Ranid frogs (Rona cofesbeiano, R.
chiricohuensis, and R. yavopaiensis), garter sr^akes
(Thamnophis eques, T. marcianus) and Sonoran
mud turtles (Kinosfernon sonoriense) were surveyed
in soutt^eastern Arizona. Distribution of the
introduced bullfrog (Rano cofesbeiano) was
negatively correlated witt^ distributions of the two
leopard frogs and garter snakes, The hypothesis that
bullfrog predotion caused decline of a native
wetland herpetofouno is supported by data on
bullfrog diet, on garter snake, leopard frog and mud
turtle population structure, and natural history
observations on the snakes. An experimental
removal of bullfrogs has been initiated at the San
Bernardino National Wildlife Refuge.

The bullfrog (Ram catesbeiam) is
North America's largest frog and one
of the most widely distributed anu-
rans on the continent. Occurring
naturally from Florida to Nova Scotia
and west into central Texas, Okla-
homa, and Kansas, the bullfrog has
been introduced widely into perma-
nent waters throughout the West
(Bury and Whelan 1984, Stebbins
1985, Wright and Wright 1949)
Known to be voracious, opportunis-
tic predators, they have been impli-
cated in declines of native anuran
populations (Bury and Luckenbach

1976, Bury et al. 1980, Conant 1975,

1977, Jameson 1956, Moyle 1973,
Nussbaum et al. 1983, Vitt and
Ohmart 1978 and others). Much less
is known about their impacts on
other vertebrate classes.

A recent investigation of factors
producing decline of Mexican garter
snakes (Thamnophis eques) in Arizona
(Rosen and Schwalbe 1988) sug-
gested that predation by introduced
bullfrogs (see fig. 1) is a present and

'Paper presented of the Symposium, The
Monagemenf of Amphibians, Repfiles, and
Small Mammals in North America. July 18-
22. 1988. Flagstaff. Arizona.

^Cecil R. Schwalbe is Nongame Herpe-
tologist and Philip C. Rosen is Contract Bi-
ologist. Arizona Game and Fish Depart-
ment. 2222 West Greenway Road. Phoenix.
Arizona 65023-4399. Rosen's present ad-
dress is Department of Ecology and Evolu-
tionary Biology. University of Arizona.
Tucson. Arizona 85721.

serious impact on some of the few
remaining snake populations. Obser-
vations during the garter snake sur-
vey suggested a similar effect on
leopard frogs (Rana yavapaiensis, R.
chiricahuensis).

Recently, Hayes and Jennings
(1986) questioned the importance of
bullfrog predation in declines of
western North American ranid frogs.
They include predation by bullfrogs
as one of three major hypotheses to
explain decline of ranid frogs in Cali-
fornia, but suggest that predation by
introduced fish has had greater im-

pact on native frogs. Hayes and Jen-
nings (1986) indicate further that
their hypotheses need to be tested to
determine actual causal factors in
population declines. In this paper we
present distributional and natural
historical data implicating bullfrogs
in population declines of native wet-
land reptiles and amphibians in
southeastern Arizona. We then de-
scribe an experimental program of
bullfrog removal we have initiated to
test the direct and indirect effects of
this introduced predator on wetland
herpetofaunas.

166

Methods and Materials

We report on two phases of our
work. The first phase involves exten-
sive surveys, principally for garter
snakes. The second focuses on inten-
sive surveying and experimental
manipulation at one locality that is
heavily infested with bullfrogs.

Extensive Phase

We sampled over 80 localities
throughout much of central and
southern Arizona during 1985-1987,
searching appropriate aquatic and
semi-aquatic habitats (Rosen and
Schwalbe 1988). Methods and results
are briefly summarized here. Lotic
habitats were surveyed for 2-6 mile
reaches on foot. Lentic habitats were
also examined on foot, in their en-
tirety in most cases. During these

surveys, attempts were made to cap-
ture, measure, mark and release all
garter snakes seen. Detailed observa-
tions were made on distribution and
abundance of other biota on the sites
sampled, with special attention to
anurans, turtles and other snakes.
Intensive mark-recapture studies
were conducted at four sites using
trapping methods described below.

Intensive Phase

San Bernardino National Wildlife
Refuge (SBNWR), one of four sites
where mark-recapture procedures
were initiated during the extensive
phase of our work, was selected for
ongoing observation and experimen-
tation. Beginning in September 1986,
we visited the refuge in September
and May of each year, marking
snakes, observing herpetofaunal dis-

SAN BEffmom NWR

Mexico

Figure 2.— Diagrammatic map of San Bernardino National Wildlife Refuge. Stippled line indi-
cates boundary between upland Chihuahuan desertscrub and riparian scrub and wood-
land vegetation types.

tributions and abundances, and ex-
perimentally removing bullfrogs.

Intensive Site Description

SBNWR (fig. 2) consists of 984 ha in
the San Bernardino Valley on the
Mexican border in Cochise County,
Arizona. Elevations range from 1134
to 1183 m. Higher, rocky slopes and
mesas supporting Chihuahuan de-
sertscrub and lower terraces grading
into desert grassland comprise al-
most two-thirds of the refuge.

The heart of the refuge is a low-
land supporting dense mesquite
(Prosopis velutina) bosques and
sacaton (Sporobolus) grasslands inter-
spersed with four spring-fed ponds
and seven additional springs. In the
center of this low ground is deeply
incised Black Draw, headwater of the
Rio Yaqui, which normally arises at a
natural spring about halfway be-
tween the Mexican border and north
boundary of the refuge. Large, iso-
lated, living and dead cottonwood
trees (Populus fremontii) occur near
almost all aquatic habitats. Broad
swamplikc ciencgas with little open
water occur at the artesian wells that
do not supply ponds.

Vegetation in Black Draw varies
from rank herbaceous plants and tall
grass in the northern one-half,
through open riparian thicket and
cat-tail (Typha domingensis) stands,
into almost impenetrable thickets of
sapling cottonwood and willow
(Salix gooddingi) throughout the
lower 1.2 km to the border. Cienega
pools are cold and reach a depth of
about 2 meters.

North Pond, focal point for the
experimental removal of bullfrogs,
contains 0.1 ha of open water sur-
rounded by earthen levees. Artesian
well flow is piped into the pond and
into a small marshy area north of the
pond. North and west banks are
lined with mesquite. South and west
banks are open or overgrown with
herbaceous vegetation. Cat-tail is
spreading rapidly around the pond

167

margin from foci in northeast and
southwest corners. Open water is
largely choked with submergent
macrophytes.

The wetland herpetofauna of the
refuge includes bullfrogs (Rana
catesbeiana), lowland leopard frogs
(R. yavapaiensis), Mexican garter
snakes (Thamnophis eques), checkered
garter snakes (T. marcianus) , and
Sonoran mud turtles (Kinosternon
sonoriense).

Intensive Field Procedures

Garter snakes were collected by hand
at all times of day and night, and
with minnow traps connected by
aquatic drift fences (see Rosen and
Schwalbe 1988 for details). Four drift
fences, each with a trap at each end,
were set in North Pond during each
visit to the refuge. Two drift fences
with traps were set in Twin Pond in
August 1985 and August-September
1986. Twin Pond was drained during
summer 1987 and remains dry.

The following data were recorded
for each snake captured: date, loca-
tion, sex, snout-vent length (SVL),
tail length, total weight, presence/
absence and number of food items,
and injuries. Females were palped to
determine presence /absence and
number of developing young. For
hand-caught snakes we recorded ac-
tivity at time of first sighting, mi-
crohabitat, time, and cloacal and am-
bient temperatures. Each individual
was uniquely marked by clipping
subcaudal scales.

Bullfrogs were collected mostly
with four-pronged spears at night by
using head lamps to find and blind
them. Additionally, many were col-
lected in turtle hoop nets, which
were set along seine nets rigged as
aquatic drift fences. Some hoop nets
were baited to capture turtles, and
these captured bullfrogs, as well. A
few were collected by hand and with
air guns and light arms. Initial col-
lecting efforts were focused on larger

(>100 mm SVL) bullfrogs. Every
aquatic habitat on the refuge was
checked for frogs by listening for
their calls and searching visually at
night. Captured bullfrogs were kept
on ice overnight and the following
data were recorded the next day:
capture location and date, sex, snout-
vent length, total weight. Most were
dissected to determine stomach con-
tents and reproductive condition.

Results

Distribution and Natural History

Leopard frogs are significantly less
common where bullfrogs abound
(table 1: Spearman rank correlation
r^=-0.434, p<0.025, Rosner 1982, Sokal
and Rohlf 1981). SBNVVR is the only
site where we found both bullfrogs

and leopard frogs. Among the sites
shown in table 1, introduced, non-
native predatory fish were found in
abundance only at Bog Hole and
Babocomari Cienega, where ranid
frogs were absent. Historical records
indicate that leopard frogs once were
abundant in two areas now support-
ing dense bullfrog populations, Ari-
vaca Creek (Wright and Wright 1949)
and SBNWR (Lanning 1981, Lowe
personal communication).

Mexican garter snakes also are sig-
nificantly less abundant in the pres-
ence of bullfrogs (table 1: Spearman
rank correlation r^=-0.420, p<0.03). At
the Potrero Canyon locality, Mexican
garter snakes were known as late as
1970 (Rosen and Schwalbe 1988), but
we found only checkered garter
snakes (N = 24) during 19854987.

At SBNWR, all museum records of
Thamnophis prior to 1970 (N=7) were

Table 1. —Distribution and abundance of ranid frogs and garter snal<es in
wetlands of southeastern Arizona, based upon field work during 1985-1988.
0=absent, l=rare, 2=common, 3=very abundant; P=pond, C=cienega,
M=marsti; NWR=National Wildlife Refuge; SB=San Bernardino. Leopard frogs
may be either Rana chiricahuensis or R. yavapaiensis.

Locality

Ranid abundance Garter snake abundance

Bull-
frog

Leopard
frog

Checkered

Mexican

San Bernardino NWR

P

3

0

1

1

San Bernardino NWR

C

3

1

1

1

Upper SB valley

P

0

2

3

0

Leslie Creek

C

0

3

0

0

Lewis Springs

C

0

3

0

2

San Pedro River

2

0

2

1

San Pedro gravel pit

P

3

0

0

0

Ramsey Canyon

P

0

3

0

0

Parker Canyon Lake

3

0

0

1

Sharp Spring

C

1-2

0

0

1

Bog Hole

P

0

0

0

0

Bog Hole

c

0

0

0

3

Research Ranch

p

0

2

0

3

Research Ranch

c

0

1

0

3

Elgin Cienega

c

0

2

0

2

Babocomari River

p

0

0

0

1

Babocomari River

c

0

0

0

1

Cienega Creek

c

1

0

0

2

Potrero Canyon

M

3

0

3

0

Potrero Canyon

C

3

0

2

0

Sonoita Creek

M

3

0

0

0

Sonoita Creek

C

3

0

0

0

168

eques, while all subsequent (N=5)
were marcianus. Thamnophis eques
comprised 57% of the garter snakes
seen on the refuge during 1985-1988
(table 2). On the refuge, the popula-
tion of Mexican garter snakes was
heavily dominated by large adults, in
significant contrast to populations in
areas lacking bullfrogs, where year-
lings and small adults predominate
(fig. 3, Mann-Whitney U Test,
p<0.001). At SBNWR, most Mexican
garter snakes (61.9%) had damaged
tails which bled between the ventral
scales when handled (fig. 4), suggest-
ing unsuccessful predation attempts
by bullfrogs. This type of injury was
not seen at any other locality.

At SBNWR we found Sonoran
mud turtles (Kinosternon sonoriense)
to be unexpectedly rare. Only four

turtles were captured in 29 trap-
nights on the refuge, a rate of 0.14
captures per trap-night. Elsewhere in
Arizona, 917 trap nights produced
2,092 captures at the 17 other locali-
ties we have sampled (Rosen unpub-
lished data, Rosen 1987). The mean
trap success for those 17 localities
was 4.32 + 0.23 captures per trap-
night (range 0.20-12.23). For the five
habitats in southeastern Arizona
which were comparable to the ref-
uge, and where at least 20 trap-nights
were registered, mean trap success
was 5.42 + 1.03 captures per trap-
night (1.23-12.23). Quitobaquito
Pond, with 0.20 captures per trap-
night was the only area in Arizona
with trapping success approaching
the low level obtained at the refuge.
The Quitobaquito population is

Table 2.— Records of all garter snakes captured on the San Bernardino Na-
tional Wildlife Refuge, Arizona, 1985-1988.

Sampling

Number

Number

Snakes

period

Mexican

checkered

captured

garter snakes

garter snakes

per day

16-18 Aug 85

3

2

1.67

23-27 May 86

4

3

1.40

30 Aug- 1 Sep 86

3

1

1.33

23-25 May 87

5

0

1.67

5-7 Sep 87

4

6

3.33

29-30 May 88

1

3

2,00

Total

20

15

1.84

15-

10-1

w
o

K

III

K

u. O

K

U

■ SJ

3
Z

0-

I I I I " I lll I 1 1 I ll I

200 400 600 800

SNOUT-VENT LENGTH (mm)

Figure 3.— Size-frequency histograms of
Mexican garter snakes in 1985 and 1986
(rTX>clified fronn Rosen and Schwalbe 1988).
Upper histogrann represents snakes fronn
populations where bullfrogs were scarce or
absent. Lower histogrann represents San
Bernardino National Wildlife Refuge sample.

Figure 4.— Bullfrog damage to tail of large
Mexican garter snake, San Bernardino Na-
tional Wildlife Refuge, Cochise County, Ari-
zona, 1986.

known to have been markedly re-
duced by human activities (Rosen
1986).

Including captures obtained by all
methods, only six Sonoran mud
turtles have been found by us on the
refuge. All were large adults, and,
according to growth ring analysis
(see Rosen 1987), all were born prior
to 1981. In all other populations, ju-
veniles comprised over 207o of the
sample (Rosen, unpublished data).

Bullfrog Diet

Stomach contents confirmed the op-
portunistic feeding behavior of bull-
frogs (table 3). Invertebrates consti-
tuted the majority of food items, with
the snail, Planorbella tenuis, and in-
sects of the orders Coleoptera,
Diptera, Hemiptera, Hymenoptera,
Odonata and Orthoptera commonly
eaten. Arthropods consumed in-
cluded adults and larvae of terres-
trial, aquatic and flying forms.

Vertebrates were found in 14.6
percent of the stomachs that con-
tained some food. The most com-
monly consumed vertebrates were
other frogs, including bullfrogs. At
least two species of native fishes,
both endangered, were eaten, the
Yaqui chub (Gila purpurea) and the
Yaqui topminnow (Poeciliopsis oc-
cidentalis sonoriensis). Mammal prey
included Peromyscus, a Sigmodon and
other as yet unidentified small ro-
dents. The two reptile food items
were a neonate checkered garter
snake in a frog from House Pond and
a spiny lizard (genus Sceloporus). Not
shown in table 3 was a nestling bird,
thought to be a red-winged
blackbird, Agelaius tricolor, found in
the stomach of a subadult bullfrog
(100 mm SVL).

Bullfrog Density

Using the numbers of bullfrogs re-
moved from North Pond (table 4),
we can estimate density and bio-

169

mass. After removing 74 adult bull-
frogs in spring 1987, we estimated 5
adults remained. Including the small
area of marsh north of the levee,
there was 0.11 ha of habitat for this
population, giving a minimum den-
sity estimate of 718 adults/ha. Mean
weight for all frogs removed in the
spring 1987 census at North Pond
was 217.1 g, yielding a total biomass
of 23.7 kg, or 215.5 kg/ha. Excluded
from this biomass estimate were re-
maining adults, and numerous juve-
niles that were not hunted. These es-
timates are conservative since we
had already removed 51 adults and
23 juveniles during fall 1986, before
we had determined the most effec-
tive means of removing the frogs.

The fall 1987 census at North Pond
reflects thorough removal the previ-
ous spring, with only about 10 frogs
either maturing into adults or immi-
grating between May 24 and Septem-
ber 5, 1987. We estimated that 4-6
adults remained in North Pond at the
end of our 1987 collecting. Because of
extremely cool, windy weather dur-
ing the spring 1988 trip, we were un-
able to collect bullfrogs effectively
during the last night and left an esti-
mated 15-20 adults.

A total of 552 bullfrogs has been
removed from SBNWR as of June
1988 (tables 4-5), including 358 of
adult size, from a total area of 2.4 ha
of open water. We estimate that take
to represent 55-80% of the adult bull-
frogs on the refuge at that time.

Preliminary Experimer^tal Results

Leopard frogs bred successfully at
the spring source in central Black
Draw in early 1987, a time of unusu-
ally good rainfall. This area was vir-
tually devoid of bullfrogs because it
is open enough for predators and re-
source managers to kill all or almost
all adults. In May 1987, leopard frog
tadpoles and juveniles were moder-
ately abundant from the spring to the
northernmost reach of cienega-
stream and dense sapling thicket.

where they were replaced by bull-
frogs. The first confirmation of leop-
ard frogs in North Pond was five
found in bullfrog stomachs in May
1987. No noticeable further increase
in leopard frog numbers or distribu-
tion was observed in May 1988.

The first juvenile Mexican garter
snake on the refuge during this study
was recorded in fall 1987. The cap-
ture rate of garter snakes on the ref-
uge doubled between May and Sep-
tember 1987 following bullfrog re-
moval (table 2). Extremely cold.

windy weather on the May 1988 trip
greatly depressed reptile activity.
Thus, the 2.0 garter snakes captured
per day (table 2) may reflect a de-
crease in activity rather than a de-
crease in the numbers of garter
snakes on the refuge.

Discussion

Distributional and natural historical
data from southeastern Arizona pro-
vide prima facie evidence that bull-

Table 3 -Stomach contents of aduit (>120 mm snout-vent length) bullfrogs.
San Bernardino National Wildlife Refuge. Arizona.

Sampling date

Prey type

30 Aug-
1 Sep 86

5-6
Sep .87

22-24
May 87

29-30
May 88

Total

Amphibians
Bullfrogs
Tadpoles
Juveniles
Leopard frogs

Juveniles
Unknown anurans
Fishes

Yaqui chub
Yaqui topminnovv
Unidentified
I Mammals
Reptiles
Invertebrates
Detritus

Empty stomachs
Total food items
No. frogs dissected

r

sidered to be adults.

Sampling
period

Adult
males

Fall 1986
Spring 1987
Fall 1987
Spring 1988

^ Totals

33
43
14
17

107

Adult
females

Total
juveniles

Total
removed

18
31
1

15

23
35
13
48

74
109
28
80

65

119

291

170

frogs play a causative role4n popula-
tion decline and disappearance of
native wetland amphibians and rep-
tiles (table 1; Results). For Mexican
garter snakes, this evidence is bol-
stered by data on population struc-
ture (fig. 3) and by observations of
injuries caused by bullfrogs (fig. 4;
Rosen and Schwalbe 1988).

That bullfrogs are predatory gen-
eralists has been thoroughly docu-
mented (see extensive review of bull-
frog foods in Bury and Whelan 1984).
In Arizona alone, bullfrogs have con-
sumed such vertebrate prey as a
nestling bird, young muskrat (On-
datra zibethicus), cotton rat (Sigmo-
don), softshell turtle (Trionyx spinif-
erus), spiny lizard (Sceloporus),
kingsnake (Lampropeltis getulus), sev-
eral species of fish and frogs, garter
snakes, even a rattlesnake (Crotalus
atrox) (fig. 1, table 3; Clarkson and
deVos 1986).

To our knowledge, in southeastern
Arizona, the only place where bull-
frogs abound and where leopard
frogs and Mexican garter snakes also
still occur, albeit rarely, is SBNVVR.
We beheve the native species persist
there because the extent and diver-
sity of aquatic habitats is greater than
elsewhere in the region. Specifically,
the relatively sparse vegetation and
absence of deep pools at the spring
source area in central Black Draw has
remained largely free of adult bull-
frogs. This is where leopard frogs

have bred and where the smallest
Mexican garter snakes have been
found.

We believe the reason only five
leopard frogs and one garter snake
were found in bullfrog stomachs is
due to already severe reduction of
leopard frog and garter snake popu-
lations. The same reasoning may ap-
ply to the absence of hatchling Sono-
ran mud turtles in bullfrog stomachs.

The bullfrog density at North
Pond (SBNWR) was quite high for
Arizona populations, although not
necessarily high for other parts of its
range (Currie and Bellis 1969). Such a
density is equalled and possibly ex-
ceeded at Arivaca, Pima County, Ari-
zona, where both leopard frogs and
Mexican garter snakes have been ex-
tirpated or become extremely rare
(Rosen and Schwalbe 1988). Concen-
trations of bullfrogs similar to that in
lower Black Draw have only been
seen in comparable habitat in por-
tions of one cienega in the San Ra-
phael grasslands of Santa Cruz
County. Abundances comparable to
those in House Pond occur at a
gravel mine south of Arizona High-
way 90 on the San Pedro River, Co-
chise County; at Page Springs,
Yavapai Count}'; and possibly at
Parker Canyon Lake, Cochise and
Santa Cruz counties and Potrero
Canyon marsh, eight kilometers
north of Nogales, Santa Cruz
County.

r

Table 5.— Bullfrog removals from aquatic habitats other than North Pond,
San Bernardino National Wildlife Refuge, Arizona. Individuals > 120 mm
snout-vent length are considered adults.

Adult

Adult

Juveniles

Total

Locality

Date

males

females

removed

Twin Pond

Fall 86

2

2

1

5

Tula Pond

Spring 87

2

3

3

8

House Pond

Spring 87

35

42

34

111

Black Draw

Spring 87

32

25

15

72

Tuie Pond

Spring 88

0

0

3

3

House Pond

Spring 88

9

10

11

30

Black Draw

Spring 88

6

18

8

32

Totals

86

100

75

261

At Potrero Canyon marsh, Mexi-
can garter snakes have disappeared
and checkered garter snakes are
abundant. In the preceding three lo-
calities, checkered garter snakes are
absent, and Mexican garter snakes
persist in low numbers. Both garter
snakes occur along the San Pedro
River but neither utilize the gravel
pit pond (Rosen and Schwalbe 1988,
Rosen personal observations).

Natural cienega-streams, includ-
ing Turkey and O'Donnell Creeks,
where bullfrogs are absent, and
Cienega Creek, where they are rare,
have high densities of Mexican garter
snakes and include many juveniles
and young adults. One spring fed
pond north of Canelo Hills, which is
structurally and vegetatively similar
to North Pond, contained about 95
Mexican garter snakes at a density
near 1055 individuals/ha, and
yielded an average of 5.4 snakes per
trapping day (Rosen and Schwalbe
1988). In contrast, only seven garter
snakes have been trapped on SB-
NWR in fifteen days of similar trap-
ping.

Central Black Draw would ordi-
narily be regarded as relatively poor
habitat for Mexican garter snakes,
because the vegetative cover is too
thin, particularly at the water's edge.
The abundance of Mexican garter
snakes there and the regular occur-
rence of checkered garter snakes at
North Pond display an inversion of
the usual habitat preferences of the
two species in Arizona. In competi-
tion, in a broad sense, with Mexican
garter snakes, checkered garter
snakes may be favored by the pres-
ence of bullfrogs because they are
less aquatic and hence less affected
by the increased predation pressure.

Hayes and Jennings (1986) argued
that predation by introduced bull-
frogs was not a compelling hypothe-
sis to explain population declines of
native ranid frogs in western North
America. They suggest that preda-
tion by introduced fish, mainly cen-
trarchids, is a more promising hy-
pothesis. In southeastern Arizona we

171

found that bullfrogs have invaded a
greater variety of wetland environ-
ments than exotic predatory fish,
and, in some instances, have
achieved population densities suffi-
cient to impact the native herpe-
tofauna. While we do suspect that
introduced fish impact native wet-
land herpetofaunas in Arizona (see
Rosen and Schwalbe 1988), our data
for the southeastern portion of the
state compellingly incriminate the
bullfrog.

Our approach is to attempt to
manage or eliminate bullfrogs from
selected areas. It is principally in-
tended to develop practical manage-
ment techniques for controlling bull-
frogs, but should also provide an ex-
perimental test of the bullfrog preda-
tion hypothesis.

Effective January 1, 1988, the Ari-
zona Game and Fish Commission
opened the season year round and
set an unlimited bag and possession
limit on dead bullfrogs statewide ex-
cept for La Paz, Mohave, and Yuma
counties (Arizona Game and Fish
Commission 1988). The stipulation of
unlimited possession of dead frogs
was to decrease the likelihood of ac-
cidental or intentional release of bull-
frogs into new habitats. The new
regulations will make it easier for
agencies, organizations and individu-
als to put pressure on bullfrog popu-
lations in specific areas in favor of
native species.

No data exist to show impacts of
bullfrogs on native species in the
three western counties, so they have
retained a July 1 to November 30 sea-
son with a bag and possession limit
of 12 per day or in possession live or
dead. Because Arizona's amphibian
and reptile regulations are reviewed
annually, new data can be incorpo-
rated into management decisions.

Conclusions

There is evidence that bullfrogs have
negatively impacted populations of
native amphibians and reptiles in

Arizona. Although some of the
trends are encouraging, preliminary
data from bullfrog removal exp>eri-
ments are inconclusive as to whether
or not bullfrog control measures may
augment recruitment in lowland
leopard frogs, Mexican garter snakes
or Sonoran mud turtles. More inten-
sive efforts will be required to elimi-
nate bullfrogs from even local habi-
tats when such habitats are structur-
ally complex.

Acknowledgments

We are thankful to the U.S. Fish and
Wildlife Service Office of Endan-
gered Species in Albuquerque for
funding parts of this study. We thank
the U.S. Fish and Wildlife Service
and Refuge Manager Ben Robertson
in particular for permission to con-
duct this research on the refuge, and
C.H. Lowe for his information on the
history of the herpetofauna on the
refuge and elsewhere in the South-
west.

For enthusiastic assistance in the
field we gratefully acknowledge the
following: Ron Armstrong, Randy
Babb, Howard Berna, Andrew,
Cindy and Ted Cordery, Mary
Gilbert, Rich Glinski, Dean and Gar-
rett Hendrickson, Julia Hoffman,
Terry Johnson, Charlie Painter, Bruce
Palmer, David Parizek, David
Propst, Cathy Schmidt, Adam and
Ethan Schwalbe, Berney Swinburne,
Ross Timmons and Sabra Tonn.

Literature Cited

Arizona Game and Fish Commission.
1988. Commission Order 41: Am-
phibians. Arizona Game and Fish
Department Publication. 1 p.

Bury, R. Bruce, C. Kenneth Dodd, Jr.,
and Gary M. Fellers. 1980. Conser-
vation of the amphibia of the
United States: a review. U.S. De-
partment of the Interior, Fish and
Wildlife Service, Resource Publica-
tion 134, 34 p. Washington, D.C.

Bury, R. Bruce, and Roger A. Luck-
enbach. 1976. Introduced amphibi-
ans and reptiles in California. Bio-
logical Conservation 10:1-14.

Bury, R. Bruce, and Jill A. Whelan.
1984. Ecology and management of
the bullfrog. U.S. Department of
the Interior, Fish and Wildlife
Service, Resource Publication 155,
23 p. Washington, D.C.

Clarkson, Robert W., and James C.
deVos, Jr. 1986. The bullfrog. Ram
catesbeiam Shaw, in the Lower
Colorado River, Arizona-Califor-
nia. Journal of Herpetology 20:42-
49.

Conant, Roger. 1975. A field guide to
reptiles and amphibians of eastern
and central North America. 429 p.
Houghton Mifflin Company, Bos-
ton.

Conant, Roger. 1977. Semiaquatic
reptiles and amphibians of the
Chihuahuan Desert and their rela-
tionships to drainage patterns of
the region, p. 455-491. In Transac-
tions of the symposium on the bio-
logical resources of the Chihua-
huan Desert region. United States
and Mexico. R.H. Wauer and D.H.
Riskind, editors. U.S. National
Park Service, Transaction Proceed-
ings Series 3, 658 p.

Currie, W., and E. D. Bellis. 1969.
Home range and movements of
the bullfrog. Ram catesbeiam
Shaw, in an Ontario pond. Copeia
1969:688-692.

Dowe, Brian John. 1979. The effect of
time of oviposition and microenvi-
ronment on growth of larval
bullfrogs (Ram catesbeiam) in Ari-
zona. Unpubl. masters thesis, 55 p.
Arizona State University, Tempe.

Fritts, Thomas H., Randy D. Jen-
nings, and Norman J. Scott, Jr.
1984. A review of the leopard
frogs of New Mexico. Unpubl. re-
port to New Mexico Department
of Game and Fish (Santa Fe, New
Mexico). 157 p.

Hayes, Marc P., and Mark R. Jen-
nings. 1986. Decline of ranid frog
species in western North America:
Are bullfrogs (Ram catesbeiam)

172

ogy 20:490-509.

Hendrickson, Dean A., and W.L.
Minckley. 1984. Cienegas — vanish-
ing climax communities of the
American Southwest. Desert
Plants 6:130-175.

Hulse, Arthur C. 1974. An autecol-
ogical study of Kinosternon sonori-
ense Lcconte (Chelonia: Kinoster-
nidae). Unpubl. doctoral disserta-
tion, 105 p. Arizona State Univer-
sity, Tempe.

Jameson, David L. 1956. Growth, dis-
persal and survival of the Pacific
treefrog. Copeia 1956:25-29.

Lanning, Dirk V. 1981. The verte-
brates of San Bernardino Ranch,
Cochise County, Arizona. Unpubl.
report from the Arizona Natural
Heritage Program (Tucson, Ari-
zona) to The Nature Conservancy
and the U.S. Fish and Wildlife
Service, Albuquerque, New Mex-
ico. 88 p.

Moyle, Peter B. 1973. Effects of intro-
duced bullfrogs, Rana catesbeiana,
on the native frogs of the San
Joaquin Valley, California. Copeia
1973:18-22.

Nussbaum, Ronald A., Edmund D.
Brodie, Jr., and Robert M. Storm.
1983. Amphibians and reptiles of
the Pacific Northwest. 332 p.
Northwest Naturalist Books, The
University of Idaho Press,
Moscow.

Rosen, Philip C. 1986. Population de-
cline of Sonoran mud turtles at
Quitobaquito Springs. Final Re-
port to National Park Service, Co-
operative Park Service Research
Unit, Tucson, Arizona. 24 p.

Rosen, Philip C. 1987. Variation of
female reproduction among popu-
lations of Sonoran mud turtles
(Kinosternon sonoriense) in Arizona.
M.S. thesis, Arizona State Univer-
sity, Tempe.

Rosen, Philip C, and Cecil R. Sch-
walbe. 1988. Status of the Mexican
and narrow-headed garter snakes
(Thamnophis eques megalops and
Thamnophis rufipunctatus) in Ari-
zona. Unpubl. report from Ari-
zona Game and Fish Dept. (Phoe-

nix, Arizona) to U.S. Fish and
Wildlife Service, Albuquerque,
New Mexico. 69 p.

Rosner, Bernard. 1982. Fundamentals
of Biostatistics. 547 p. Duxbury
Press, Boston.

Sokal, Robert R. and F. James Rohlf.
1981. Biometry, the Principles and
Practice of Statistics in Biological
Research. 859 p. W.H. Freeman
Co., San Francisco.

Stebbins, Robert C. 1985. A field
guide to western reptiles and am-
phibians. 336 p. Houghton Mifflin
Company, Boston.

Vitt, Laurie J., and Robert D.

Ohmart. 1978. Herpetofauna of the
lower Colorado River: Davis Dam
to the Mexican border. Proceed-
ings of the Western Foundation of
Vertebrate Zoology 2:35-72.

Wright, Albert Hazen, and Anna Al-
len Wright. 1949. Handbook of
frogs and toads of the United
States and Canada. 640 p. Com-
stock Publishing Company, New
York.

173

Developing Management
Guidelines for Snapping
Turtles^

Ronald J. Brooks,^ David A. Galbraith,^ E.
Graham Nanceklvell/ and Christine A.
Bishop^

Abstract.— We examined demographic features
of 2 Ontario populations of snapping turtles
(Chelydra serpentina) io provide an empirical basis
for developing management guidelines. The
northern population matured later (18-20 yr) than
did the southern populations (<10 yr), and displayed
an older age distribution. Long-lived, "bet-hedging"
species have low annual reproductive success and
are unusually susceptible to exploitation. A
preliminary life table is presented for the northern
population. Our results indicate that the northern
population cannot sustain even minimal levels of
exploitation by humans without undergoing a
decline in numbers.

In general, turtles have not been a
major concern of wildlife managers
in North America, and in many juris-
dictions they are given little or no
protection. They are perceived to
have limited ecological, commercial,
aesthetic or recreational value, and
because they are usually cryptic and
slow moving they are uninteresting
to most people. Partly for these rea-
sons, there have been remarkably
few studies of their life history and
ecology. In addition, their great lon-
gevity makes them difficult to study,
except on a long-term basis. Never-
theless, turtles are, or should be, of
interest to wildlife managers for at
least three major reasons.

First, they are major components
of a variety of both terrestrial and
aquatic ecosystems and therefore
play significant, though often unrec-
ognized roles as carnivores, herbi-
vores and scavengers. In both
aquatic and terrestrial habitats, the

'Paper presented at symposium. Man-
agement of Amphibians, Reptiles, and
Small Mammals in North America. (Flag-
staff , AZ, July 19-21, 1988.)

^Professor, Department of Zoology, Uni-
versity of Guelph, Guelph, Ontario, Can-
ada. N1G2W1.

^Graduate Student, Biology Depart-
ment, Queen's University, Kingston, Ontario,
Canada. K7L 3N6.

"Wildlife Technician, Department of Zo-
ology. University of Guelph, Guelph, On-
tario, Canada. N1G2W1.

^Graduate Student, Biology Depart-
ment, York University, Toronto, Ontario,
Canada.

Standing-crop biomass of turtles is
generally much higher than that of
any other reptile (Iverson 1982). In
aquatic systems, turtle biomass often
exceeds that of sympatric endoth-
erms by an order of magnitude and
is similar to levels reported for fishes
(Iverson 1982). Similarly, annual pro-
duction of turtles is comparable to
that reported in most other verte-
brates, although well below levels
found in some fishes (Iverson 1982).
Many turtles that are especially long-
lived may have low annual produc-
tivity. This low productivity may be
overestimated because of the high
standing-crop biomass of turtles.
Their life history is markedly differ-
ent from those of the birds and mam-
mals that typically occupy the atten-
tion of wildlife managers. As such,
these species represent special prob-
lems in conservation and manage-
ment. Therefore, turtles should be of
interest to managers, because they
are important components of a vari-
ety of ecological communities and
because in many cases their longevity
and low annual production relative
to standing crop, characteristic of a
"bet-hedger" (Obbard 1983) is a life-
history strategy that may be highly
susceptible to exploitation or to other
sources of mortality of adult animals
such as unsuitable overwintering
conditions or heavily polluted wa-
ters.

Secondly, managers should have
an interest in turtles because many
species are harvested for commercial

profit, usually as food or for the pet
trade (Bergmann 1983, Congdon et
al. 1987, Lovisek 1982). There is evi-
dence of marked, recent declines in
harvests of most turtle species, but
this evidence is difficult to quantify
because estimates of total stocks do
not exist for any turtle species. For
snapping turtles, the annual commer-
cial catch in Minnesota was esti-
mated at 36000-40800 kg or approxi-
mately 6000-6800 average-sized
adults (Helwig and Hora 1983). In
southern Ontario, Lovisek (1982) esti-
mated the annual catch of C. ser-
pentina to be 30000-50000 kg or 5000-
8300 adults. There is evidence from
trappers (J. Bullard pers. comm.) that
numbers of this species are a fraction
of former numbers over much of
their southern range in Ontario, but
again no quantitative estimates exist.
At present, therefore, it is necessary
to measure the impact of harvesting
turtles on a local basis (Hogg 1975).

Thirdly, snapping turtles may be
of interest to managers because they
are often regarded as pests or as a
danger to human swimmers, or as
destructive predators of waterfowl
and game fish (Hammer 1969, Kiviat
1980, Pell 1941).

In this paper, we review the biol-
ogy of snapping turtles in relation to
these three areas of potential impor-
tance for wildlife managers. We pres-
ent demographic characteristics of 2
populations in Ontario, and in addi-
tion, we develop a life table for the
more northern population of snap-

174

ping turtles which will allow us to
predict the impact of different levels
of harvesting pressure on this popu-
lation.

Snapping Turtles in Aquatic
Ecosystems

Regulation of Population Density

There is at present little understand-
ing of what factors regulate popula-
tions of any turtle species, but it is
known that turtles may reach very
high densities and high biomass den-
sities (Galbraith et al., in press; Iver-
son 1982). It seems likely that pri-
mary productivity would be the best
predictor of variation in numbers of
turtles in a habitat. In snapping
turtles, population density ranges
from 1-75 adult turtles per ha (Gal-
braith et al., in press). Density among
populations correlates positively
with latitude and primary produc-
tion levels and negatively with the
size of the body of water (Galbraith
et al., in press), although data are too
sparse to rely heavily on these corre-
lations. Other possible factors influ-
encing density are predation pres-
sure, especially on nests and hatch-
lings, climatic influences on egg sur-
vival and embryo development, and
availability of suitable nesting sites.
Again, the role of these factors has
not been studied.

Annual Energy Budgets

No complete energy budget has been
determined for any turtle population,
although some efforts have been
made to estimate critical components
of the energy budget (Congdon et al.
1982). Almost all efforts in this area
have concentrated on the energy con-
tent and cost of the eggs (Congdon
and Gibbons 1985, Congdon and
Tinkle 1982, Shine 1980) and on the
rates of digestion, especially in rela-
tion to temperature (Parmenter
1981).

Food-Web Connections

Snapping turtles are widely regarded
as voracious predators, but most
studies of their diet indicate that
plant material is a major com|X)nent
of their food (Alexander 1943, Ham-
mer 1972, Pell 1941). Hammer (1972)
found that plants made up the major-
ity of the diet of snapping turtles in a
North Dakota marsh. In Connecticut,
fish (mostly nongame species) and
aquatic plants were of equal impor-
tance and birds made up only a small
fraction of the diet (Alexander 1943).
In Maine, snapping turtles ate signifi-
cant numbers of ducklings in local
areas where both turtles and water-
fowl were common, but widespread
control of turtles was not recom-
mended (Coulter 1957). Lagler
(1943), working in Michigan, con-
cluded that snapping turtles had
minimal impact on waterfowl and
pan fish and subsisted primarily on
plant material and invertebrates. In
general then, snapping turtle preda-
tion on waterfowl or game and sport
fish poses no serious problem to
these valuable species except perhaps
in local situations where numbers of
turtles may be very high and the
turtles have easy access to young wa-
terfowl.

Adult snapping turtles are largely
immune to predation other than by
humans over most of their range. A
wide diversity of predators prey on
snapping turtle eggs (foxes, skunks,
raccoons) and hatchlings (herons,
large fish), and mortality is very high
during these stages.

Rationale for \he Development of
Life Tables

The demography of populations of
freshwater turtles under exploitation
has not been extensively studied.
Some reports have cited large catches
being removed from specific loca-
tions with apparently little impact on
remaining numbers in the short term
(e.g. Hogg 1975) but no study has

followed an exploited population in
detail for any length of time. It is nec-
essary, therefore, to infer the effect of
harvesting on populations using
demographic parameters of unex-
ploited populations under long-term
study. This paper describes 2 snap-
ping turtle populations in Ontario,
Canada and presents a life table for
one of these populations.

Study Areas

Lake Sasajewun, Algonquin
Provincial Park, Ontario

The Ontario Ministry of Natural Re-
sources Wildlife Research Area
(W.R.A. 45'35' N, 78'30'W, mean an-
nual temperature 4.4 'C), is located in
the central area of Algonquin Provin-
cial Park, in a region of mixed forest
last logged in the 1930s. The snap-
ping turtles inhabiting the lakes and
streams running through the W.R.A.
have been studied since 1972. Each
year, adult female turtles are cap-
tured after nesting and both males
and females are captured using
baited hoop traps. Of the approxi-
mately 185 tagged snapping turtles in
the watershed of the North Mada-
waska River, about 100 are recap-
tured each year. Approximately 70
nests of known females are located
each year.

Snapping turtles are the largest
aquatic vertebrate in the W.R.A.,
with the exception of beavers (Castor
canadensis) and occasional river otters
(Lutra canadensis). The only other spe-
cies of turtle in this watershed is the
midland painted turtle ( Chrysemys
picta marginata), present in very small
numbers (< 10). The density of the
W.R.A. snapping turtle population is
approximately 1.5 adults/ha in lakes
(Galbraith et al., in press). The study
area and the snapping turtle popula-
tion have been described extensively
elsewhere (Galbraith and Brooks
1987; Galbraith et al. 1987, in press;
Obbard 1983).

175

Royal Botanical Gardens,
Hamilton, Ontario

The Royal Botanical Gardens (R.B.G.)
consist of approximately 700 ha of
woodlands and waterways within
the metropolitan Hamilton area
(43'17'N, 79'53'W; mean annual tem-
perature 9.8'C). This study area and
the snapping turtle population in the
R.B.G. have been described previ-
ously (Galbraith et al., in press). We
have captured, tagged, and released
adult and juvenile snapping turtles in
this watershed since 1984. In addi-
tion to snapping turtles, map turtles
(Malaclemys geographica) and painted
turtles are common aquatic cheloni-
ans in this system. The painted turtle
is at least as common as the snapping
turtle.

The turtles inhabit a highly pro-
ductive, eutrophic waterway which
is artificially enriched by effluent
from a sewage treatment plant. West
Pond (9.8 ha), where our trapping
has taken place, also connects with
heavily-polluted Hamilton Harbour.
Despite the contaminants, this popu-
lation exhibits one of the highest den-
sities yet reported for this species,
approximately 60-70 adults/ ha (Gal-
braith et al., in press).

Methods and Results
Life Tables

Two approaches are commonly taken
in preparing life tables. Static or ver-
tical life tables are prepared by deriv-
ing mortality rates from the observed
population age structure. Cohort-
specific, or horizontal life tables are
prepared by following a specific co-
hort and observing age-specific mor-
tality rates throughout life (Deevy
1947). At present, only static life
tables can be prepared for snapping
turtle populations, because individ-
ual cohorts cannot be followed effec-
tively in these animals which may
have a maximum longevity of over a
century (Galbraith and Brooks 1987).

Therefore, we will only consider
static life tables.

Life-Table Parameters for
Algonquin Park (W.R.A.)

Snapping turtles experience large
fluctuations in annual reproductive
success (Obbard 1983). In the W.R.A.
population, for example, most years
do not produce any emergent hatch-
lings (R.J. Brooks, unpubl. data)
whereas occasional years may pro-
duce large numbers of hatchlings.
This highly stochastic survivorship
throws some doubt on the utility of
static life tables, because age curves
could be highly biased by errors due
to irregular recruitment. Therefore,
we will use an average mark-recap-
ture survivorship rate (Galbraith and
Brooks 1987) for all adult females for
the construction of the life table.

Several critical pieces of informa-
tion have never been obtained for
any snapping turtle population. For
example, no estimate of survivorship
of hatchlings or juveniles has ever
been published. A crude estimate of
this rate can be obtained by assuming
that the number of turtles recruited
per year into the population is fairly
represented by the average recruit-
ment rate, and that the number of
eggs being produced per year has not
varied greatly between the years
when recruits were initially pro-
duced (i.e. as eggs) and the present
time. In the W.R.A. population, on
average, one new nesting female is
captured per year on nesting sites
used by approximately 85 other fe-
males. The mean clutch size of 34
eggs once per year gives an annual
egg production of 2890 eggs. Assum-
ing half these eggs produce females,
the net survivorship across all age
classes (including eggs) until age at
first nesting (approximately 19 yr,
(Galbraith 1986)) is therefore 1/1445
(0.000692).

In the W.R.A. population, Obbard
(1983) observed a mean rate of emer-
gence of hatchlings from eggs of

0.0635, averaged over 142 nests in 5
yr. Taking this into account, in addi-
tion to the adult recruitment rate of
one mature female per year, the
probability of mortality between
hatching and maturity for females in
this population is 99.17%. Average
annual juvenile survivorship from
this estimate is therefore 0.7541 from
hatching to 19 yr (table 1).

High rates of statistical errors
within age estimates of individual
turtles (Galbraith 1986) make docu-
mentation of horizontal rates of age-
specific changes in fecundity unreli-
able, and therefore we have con-
structed our life table using mean
clutch size for all age classes. Net fe-
cundity, however, is a function of
both clutch size and clutch fre-
quency. Obbard (1983) estimated that
72.1% of adult females, on average,
lay a clutch each year in this popula-
tion. Mean annual egg production is
therefore 24.514 eggs per female
(mean clutch size is 34 eggs). For the
purposes of a life table, the female
turtles are considered as producing
only female offspring. It is also neces-
sary, therefore, to consider the effects
of biases in hatchling sex ratios.
Snapping turtles experience environ-
mental sex determination, whereby
incubation temperature during the
middle third of the incubation period
determines offspring sex (Yntema
1976). Between 1981 and 1985, the
mean hatchling sex ratio of naturally
incubated nests in the W.R.A. was
66% female (R.J. Brooks, unpubl.
data). Therefore, each female turUe,
on average, produces 16.18 female-
destined embryos per nesting season.
Although snapping turtles are long-
lived, the life table for female snap-
ping turtles in the W.R.A. suggests
that they do not reproduce enough to
sustain the population (table 1).

Life-Table Parameters for the
Royal Botanical Gardens (R.B.G.)

Although data are inadequate to con-
struct a meaningful life table for

176

snapping turtles from the R.B.G.,
some population parameters are
known. For example, females in the
very large snapping turtle population
in the R.B.G. appear to nest for the
first time at 10 yr of age (RJ. Brooks,
unpubl. data), and the mean clutch
size in the R.B.G. population between

1985 and 1987 was 45 eggs. The rate
of mortality in this population is
likely higher than in the Algonquin
population, because numerous dead
turtles are found each year (C.A.
Bishop, unpubl. data). Essential but
currently unavailable information
from the R.B.G. population includes

Table l.—Ufe table for female snapping turtles In Algonquin Park (W.R.A.),
Ontario, Canada.

Year

class

a '

X

1 =

X

^x

X

ml,

X X

Zm.l *

X X

n

u

1007 A

1 nnn

1

1

.uooo

9

01 '^zL'^

0470

O

oo.ooo

.UOO 1

*4

CI o^n

0979

C.

'^0 17ft

090^^

. / 0*4 1

A

O

90 '^i4A

OT^"^
.u I

7^41

7

99 9ft9

01 17

o
o

iA on A

.uuoo

7c; /ll

0
y

19 A7'^

OOAA

i^Ay
, / 0*4 1

in

0 ^^^7

y . oo /

00 '^O

7'^dl
. / 0*4 1

1 1

1 9nft
/ .zuo

00 "^A

l^A^
. 1 1

009 A

7^41

1 o

A noo

0091

. /C^ t

15

3.091

.0316

.7541

16

2.331

,0012

.7541

17

1.758

.0009

.7541

18

1.326

.0007

.7541

19

1.000

.000524

.9660

16.18

.00848

0.00848

20

.000506

.9660

16.18

.00819

0.0167

21

.000489

.9660

16.18

.00791

0.0246

22

.000472

.9660

16.18

.00764

0.0322

23

.000456

.9660

16.18

.00738

0.0396

24

.000441

.9660

16.18

.00714

0.0468

25

.000426

.9660

16.18

.00689

0.0536

30

.000358

.9660

16.18

.03107

0.0847

35

.000301

.9660

16.18

.02615

0.1109

40

.000253

.9660

16.18

.02199

0.1329

50

.000179

.9660

16.18

.03404

0.1633

60

.000127

.9660

16.18

.02409

0.1873

70

.000090

.9660

16.18

.01705

0.1990

80

.000064

.9660

16,18

.01209

0.2111

90

.000045

.9660

16,18

.00853

0.2196

100

.000032

.9660

16.18

.00730

0.2269

' = numbers of individuals.

ax

- probability of survival from year class 0 to year class x.
^n, = probability of survival from year class x to year class x+ 1 .

= net fecundity at year class x (female-destined embryos produced).

*Z//n^ = sum of all reproduction from year class 0 to year class x. equals Ro. total
lifetime reproduction, when xisat its maximum.

long-term estimates of emergence
rates of hatchlings or of adult survi-
vorship, annual nesting frequency,
and primary sex ratio.

Life-Table Implications for
Management Guidelines

Clearly, exploitation of a population
similar to that in Algonquin Park
would quickly reduce numbers be-
low any chance of recovery by repro-
duction within that population. In
formulating our life table for the
W.R.A., we have had to make several
assumptions. The most important
concerns our estimate of the rate of
survival of hatchlings and juveniles.

A comparison between the 2
populations indicates that the advan-
tages in the R.B.G. population of hav-
ing a larger clutch size than the more
northern population and being able
to initiate nesting almost 10 yr before
the W.R.A. population may be tem-
p>ered by overestimating adult survi-
vorship in the R.B.G. population.
Consequently, lifetime reproduction
may not be as high as one might pre-
dict. These comparisons must be im-
proved by direct observation of sur-
vival in the critical juvenile years,
and by following individuals of
known age throughout life, in a vari-
ety of populations.

Considerable variation in popula-
tion characteristics exists between
these 2 populations located about 280
km apart. Trapping guidelines appli-
cable to the R.B.G. population may
not be suitable to the population in
the W.R.A. Regardless, neither could
likely tolerate harvests of more than
10% of the adult population.

Management Practices to
Increase Yields of Snapping
Turtles

It is evident that unregulated har-
vesting of adult snapping turtles will
rapidly decrease population sizes,
because adult turtles are normally

177

subject to very low rates of mortality
(Galbraith and Brooks 1987). Two
strategies are possible to increase
harvestable numbers of turtles.

First, practical experience with sea
turtle farming has shown that large
numbers of eggs can be incubated
under artificial or protected condi-
tions (Mrosovsky and Yntema 1980),
although care must be taken to incu-
bate the eggs at a selection of tem-
peratures which will produce a bal-
anced sex ratio. Similar propagation
of snapping turtles should result in
increased numbers of juveniles in
populations where adult numbers
are not density-dependent.

Secondly, enrichment of the envi-
ronment could provide faster growth
rates for these poikilotherms. In-
creases in available protein will
probably result in an increase in
growth rates of individuals and in-
creases in adult carrying capacities
(MacCulloch and Secoy 1983).

Organochloride Contaminants
and Hunnan Consumption

Long-lived bottom-dwellers can ac-
cumulate high levels of environ-
mental toxins, and snapping turtles
have been found to carry very high
loads of PCBs of various forms
(Bryan et al. 1987a). Several studies
have considered the way in which
PCBs accumulate and in which tis-
sues, and snapping turtles are now
being employed as biomonitors for
organochlorides in some studies
(C.A. Bishop et al., unpubl. data).

Bryan et al. (1987) demonstrated
that local levels of pollutants mark-
edly affected the levels of organo-
chloride toxins in snapping turtle tis-
sues. Tissue-specific accumulation of
PCBs is not random in snapping
turtles, but is a function of lipopro-
tein content of the tissue and the high
lipoprotein solubility of the toxins.
Especially high concentrations (as
high as 1600 ppm PCB in turtles from
polluted locations) are found in fat
bodies, brain, and testes. However,

Bryan et al. (1987) indicated that
toxic PCB congeners did not remain
in the large fat reserves of female
turtles, as some had suggested, but
were passed on in bulk to the egg
yolks.

It is necessary, therefore, to test
tissue or egg samples to ensure that
turtles being harvested for human
consumption are not loaded to a dan-
gerous degree with organochloride
contaminants.

Management of Snapping Turtles
as Predators

Several studies have considered the
impact of snapping turtles on water-
fowl populations (Alexander 1943,
Hammer 1972, Lagler 1943). Highly-
productive bodies of water present
ideal habitat for waterfowl and for
turtles.

Destroying turtle nesting locations
may not reduce local populations of
snapping turtles, because females
may migrate several kilometers be-
tween their usual home range and
their nesting sites (Obbard 1977). In
addition, such habitat interference
will remove nesting opportunities for
other turtle species.

Reduction in numbers of adult
snapping turtles through trapping
will rapidly deplete isolated popula-
tions and should reduce risks to prey
species. However, if turtles can emi-
grate into the management area, then
the expected long-term effect of cull-
ing adults will not be realized be-
cause the population can increase
from these new adult immigrants.

Acknowledgments

We are grateful to the Ontario Minis-
try of Natural Resources and to D.
Strickland for their support and for
granting permission to conduct re-
search in Algonquin Provincial Park,
and to the Royal Botanical Gardens,
Hamilton, Ontario, for permission to
work in Cootes Paradise. We thank

C. Bell, M.L. Bobyn, M. Fruetel, J.
Hughes, K. Lampman, J. Lay field,
and S. Plourde for field and technical
assistance, and K. Kovacs and S. In-
nes for computational help. This
study has been supported by Natural
Sciences and Engineering Research
Council Canada Grant A5990, and an
Ontario Ministry of Natural Re-
sources Renewable Resource Pro-
gram Grant to Ronald J. Brooks.

Literature Cited

Alexander, Maurice M. 1943. Food
habits of the snapping turtle in
Connecticut. Journal of Wildlife
Management 7:278-282.

Bergmann, Brian. 1983. Carapace ca-
pers: uncontrolled trapping threat-
ens Ontario's snapping turtles.
Equinox 2(5):127.

Bryan, A. M., P. G. Olafsson, and W.
B. Stone. 1987. The disposition of
low and high environmental con-
centrations of PCB's in snapping
turtle tissues. Bulletin of Environ-
mental Contamination and Toxi-
cology 38:1000-1005.

Congdon, Justin D., Gary L. Breiten-
bach, Richard C. van Loben Sels,
and Donald W. Tinkle. 1987. Re-
production and nesting ecology of
snapping turtles (Chelydra ser-
pentina) in southeastern Michigan.
Herpetologica 43:39-54.

Congdon, Justin D., A. E. Dunham,
and Donald W. Tinkle. 1982. En-
ergy budgets and life histories of
reptiles, p. 233-271. In C. Gans and
F. H. Rough, editors. Biology of
the Reptilia. 13. Physiology D-
Physiological Ecology. Academic
Press. New York.

Congdon, Justin D., and J. Whitfield
Gibbons. 1985. Egg components
and reproductive characteristics of
turtles: relationships to body size.
Herpetologica 41: 194-204.

Congdon, Justin D., and Donald W.
Tinkle. 1982. Reproductive ener-
getics of the painted turtle
(Chrysemys picta). Herpetologica
38:228-237.

178

Coulter, Malcolm W. 1957. Predation
by snapping turtles upon aquatic
birds in marine marshes. Journal
of Wildlife Manage. 21(1):17-21.

Deevy, Edward S., Jr. 1947. Life
tables for natural populations of
animals. Quarterly Review of Biol-
ogy 22:283-314.

Galbraith, David A. 1986. Age esti-
mates, survivorship, growth, and
maturity of female Chelydra ser-
pentina Linnaeus in Algonquin
Provincial Park, Ontario. M.Sc.
Thesis, University of Guelph,
Guelph, Ontario.

Galbraith, David A., and Ronald J.
Brooks. 1987. Survivorship of
adult females in a northern popu-
lation of common snapping
turtles, Chelydra serpentina. Cana-
dian Journal of Zoology 65:1581-
1586.

Galbraith, David A., Ronald J.

Brooks, and Martyn E. Obbard. In
Press. Female age and body size at
the onset of reproduction in a
northern population of snapping
turtles (Chelydra serpentina). Cana-
dian Journal of Zoology.

Galbraith, David A., Mark W. Chan-
dler, and Ronald J. Brooks. 1987.
The fine structure of home ranges
of male Chelydra serpentina: are
male snapping turtles territorial?
Canadian Journal of Zoology
65:2623-2629.

Hammer, Donald A. 1969. Parame-
ters of a marsh snapping turtle
population: Lacreek Refuge, South
Dakota. Journal of Wildlife Man-
agement 33: 995-1005.

Hammer, Donald A. 1972. Ecological
relations of waterfowl and snap-
ping turtle populations. Ph.D. Dis-
sertation, Utah State University,
Salt Lake City, Utah.

Helwig, D. D., and M. E. Hora. 1983.
Polychlorinated biphenyl, mercury
and cadmium concentrations in
Minnesota snapping turtles. Bulle-
tin of Environmental Contamina-
tion and Toxicology 30:186-190.

Hogg, D. M. 1975. The snapping
turtles of Wye marsh. Ontario Fish
and Wildlife Review 14(2):16-20.

Iverson, John B. 1977. Reproduction
in freshwater and terrestrial
turtles of north Rorida. Herpetol-
ogica 33:205-212.

Iverson, John B. 1982. Biomass in
turtle populations: a neglected
subject. Oecologia 55:69-76.

Kiviat, Erik. 1980. A Hudson River
tidemarsh snapping turtle popula-
tion. Transactions of the Northeast
SecHon, Wildlife Society 37:158-
168.

Lagler, Karl F. 1943. Food habits and
economic relations of the turtles of
Michigan with special reference to
fish management. 29:257-312.

Lovisek, Jim. 1982. An investigation
of the harvesting of turtles in On-
tario. Nongame Program Wildlife
Branch, Ontario Ministry of Natu-
ral Resources.

MacCulloch, Ross D., and D. M.
Secoy. 1983. Demography, growth,
and food of western painted
turtles, Chrysemys picta bellii
(Gray), from southern Sas-
katchewan. Canadian Journal of
Zoology 61:1449-1509.

Mrosovsky, N., and C. L. Yntema.
1980. Temperature dependence of
sexual differentiation in sea
turtles: implications for conserva-
tion practices. Biological Conser-
vation 18:271-280.

Obbard, Martyn E. 1977. A radio-
telemetry and tagging study of
activity in the common snapping
turtle, Chelydra serpentina. M.Sc.
Thesis, University of Guelph,
Guelph, Ontario.

Obbard, Martyn E. 1983. Population
ecology of the common snapping
turtle, Chelydra serpentina, in north-
central Ontario. Ph.D. dissertation.
University of Guelph, Guelph, On-
tario.

Parmenter, Robert R. 1981. Digestive
turnover rates in freshwater
turtles: the influence of tempera-
ture and body size. Comparative
Biochemistry and Physiology
70(A):235-238.

Pell, Stuyvesant Morris. 1941. Notes
on the habits of the common snap-
ping turtle, Chelydra serpentina

(Linn.), in central New York. M.Sc.
Thesis, Cornell University, Ithaca,
New York.

Shine, Richard. 1980. "Costs" of re-
production in reptiles. Oecologia
46:92-100.

Yntema, C. L. 1976. Effects of incuba-
tion temperatures on sexual differ-
entiation in the turtle, Chelydra ser-
pentina. Journal of Morphology
150:453-462.

179

Spatial Distribution of Desert
Tortoises (Gopherus agassizii)
at Twentynine Palms,
California: Implications for
Relocations^

Ronald J. Baxter^

Abstract.— The spatial distribution of desert
tortoises in relation to plant communities was
compared against randomness. Tortoise captures (n
= 120) and tortoise burrows (n = 160) exhibited non-
random distributions across a 1.29 square kilometer
study plot at Twentynine Palms, California. Results
imply high diversity plant ecotones and
communities, and possibly soil characteristics are
important in determining tortoise densities. Non-
randomness in tortoise populations dictates that
relocation sites must include specific vegetational,
topographic and edaphic habitats used by the
parental populations.

The desert tortoise (Gopherus agas-
sizii) is a species whose future is un-
certain. Increased use of the deserts
by man (Luckenbach 1982) has led to
the point where the tortoise was offi-
cially listed as "threatened" in the
state of Utah (Dodd 1980). The U.S.
Fish and Wildlife Service stated in
1985 that ''...listing [of the desert tor-
toise as a threatened or endangered
species] is warranted but precluded
by other pending proposals of higher
priority" (Federal Register. 50(234):
49868-49870, 1985).

In California, the desert tortoise is
the official state reptile, and is fully
protected under law. The tortoise is
also protected in Arizona and Ne-
vada.

As part of a larger population
study (Stewart and Baxter 1987) at
the Twentynine Palnns Marine Corps
Air Ground Combat Center
(MCAGCC), the spatial distributions
of tortoise captures and burrows
were analyzed and compared against
randomly generated distributions.
Questions asked were: (1) Are tor-
toise captures and burrows ran-
domly located across the landscape

'Paper presented at symposium. Man-
agement of Amphibians, Reptiles, and
Small Mammals in Nortt) America. (Flag-
staff, AZ, July 19-21, 1988).

'Ronald J. Baxter received his master's
degree in biology for working on the desert
tortoise while at California State Polytech-
nic University, Pomona. He is currently com-
pleting his doctorate at the Department of
Biological Sciences. Northern Arizona Uni-
versity. Flagstaff. AZ, 8601 1-5640.

and /or are they associated with cer-
tain habitat types or site characteris-
tics, and if so, (2) what implications
do these distributions have for future
management decisions?

Mettiods

Twentynine Palms MCAGCC is lo-
cated approximately 5 kilometers
north of Twentynine Palms, San Ber-
nardino County, California, in the
southwestern extreme of the Mojave
Desert. All fieldwork was performed
in the Sand Hill Training Area which
is in the southwest corner of the
MCAGCC. Elevations ranged from
865 meters atop Sand Hill to about
730 meters in the bottom of Surprise
Springs wash. Data were collected
Monday through Friday, 14 April
through 18 July, 1986.

Systematic searching methods for
tortoises and tortoise burrows were a
derivation of procedures described
by Berry (1984). A 1.29 square kilo-
meter permanent study plot was es-
tablished, with its approximate cen-
ter being the NE 1 /4, SW 1 /4, NE 1 /
4, of S7, T2N, R7E (San Bernardino
Base Meridian) This site offered a
wide variety of habitats including
washes, sandy basins, rolling hills
and alluvial bajadas. The plot was di-
vided into 64 equal sized "grids" of
142 meters on a side, with grid cor-
ners marked by posts. Grids were
searched in parallel belts until the
entire plot had been searched twice;

once with the belts running north-
south, and once with the belts run-
ning east-west. The plot was also
randomly searched.

When an active tortoise was en-
countered, it was marked, weighed,
sexed, measured and photographed.
Each tortoise was assigned a unique
number, and marginal scutes were
notched with a small triangular file
for relatively permanent identifica-
tion. The precise location of the cap-
ture was noted by its distance (meas-
ured by rangefinder) and compass
aspect to the nearest grid post. Data
collected at each capture site in-
cluded plant community, tempera-
tures at the ground, 1 centimeter,
and 1 meter, cloud cover, wind
speed and direction, closest burrow,
closest plant, and any unusual be-
havior.

Precise location of tortoise bur-
rows were similarly determined by
rangefinder and compass. Data col-
lected at each burrow included plant
community, distance and identifica-
tion of nearest ecotone, distance to
nearest wash, distance to nearest Hi-
laria rigida, slope aspect and steep-
ness, opening compass aspect and
position, length, depth, and tunnel
characteristics. In this study area, it
was difficult to determine if a bur-
row high on a slope above a wash
was part of the wash "system."
Therefore, it was arbitrarily decided
to include burrows in the wash plant
community only if they were actually
found the wash bed.

180

Six visually identified plant com-
munities (Latr/ Amdu, Hiri/ Amdu,
Mixed, Wash, Sparse Wash and
Meadow) were mapped within the
study plot, and seven 15-meter line
transects (total of 105 meters) were
measured which included bare
ground as a species. Transects were

i:latr/amdu 5: meadow
2:mixed •'.burrow

3:SPARSE WASH ff-mSH NORTH

4:HIRI/AMDU Soo'

METERS

Figure 1 .—Approximate distribution of piant
communities and tortoise burrows across
the study piot. See text for expianation of
piant community names.

randomly located in each of these six
communities. Standard transect sta-
tistics (density, coverage, frequency,
relative density, relative coverage,
relative frequency and importance
values; Brower and Zar 1984) were
computed for each community.
Simpson's diversity indices (Simpson
1949) were computed and compared
with Student's t-tests (Keefe and
Bergerson 1977). Available annuals
as well as perennials were used to
give the best estimate possible for
diversity. In addition, seven soil
samples were taken in each commu-
nity, and analyzed for soil separates
(Brower and Zar 1984) and soil cal-
cium (Hach 1983). Finally, nine
''sand scats" were collected during
the field work and tested for calcium.

A random model for capture and
burrow locations was formed by
combining a number of statistical
tests. First a master map of the plot
was constructed from actual field
data at a scale of 1:2000. All capture
positions, burrows and plant com-
munity boundaries were plotted on
this map and checked against aerial
photographs. The area covered by
each plant community was then de-
termined by the use of a planimeter.
An X-Y scale ranging from 0 to 8 was
plotted on the sides of the map, and
a list of 328 random numbers was
generated by computer. These num-
bers were paired, and the pairs be-

came the X-Y coordinates of random
positions against which observed
capture and burrow locations were
compared. Distances to the nearest
wash and ecotone were determined
for these random locations by meas-
uring them on the map, and com-
pared against observed by Student's
t-tests (Zar 1974). Observed capture
distances were sometimes combined
with previous data recorded in this
area (Baxter and Stewart 1986).

A lack of habitat preference may
be suggested if burrows and captures
were found in the same relative
abundance as the plant communities.
In addition, if the expected plant
abundance distribution differed sig-
nificantly from random an extrapola-
tion of observed distributional char-
acteristics could be accomplished. An
assumption of this test was that a
distribution of randomly generated
locations (with randomness con-
firmed) produced a random fre-
quency distribution. Expected fre-
quencies for burrows and captures
were generated by multiplying the
total number of actual burrows or
captures by the percent of the plot
encompassed by each plant commu-
nity. These values were compared by
a goodness-of-fit chi-square test (Zar
1974). In addition, the number of
burrows or captures per grid were
compared against expected values as
derived from the Poisson distribu-
tion by a goodness-of-fit test.

Results

Plant Communities and Soils

Vegetation analyses revealed six dis-
tinct plant assemblages (table 1; fig.
1). Plant community distributions
generally reflected the relief of the
plot. The higher, more well-drained
hills were dominated by an associa-
tion of Larrea tridentata and Ambrosia
dumosa, which encompassed plot
area the most ("Latr/ Amdu"; table
1) and exhibited relatively high plant
diversity.

r

Table 1 .—Summary of plant community data from plant transects (total
transect length = 105 meters).

Simpson's

Percent

Plant

No. of

No. of

diversity

of plot

community^

species

individuals

index

area

Sparse Wash

16

733

0.6069

5.6

Hiri/Amdu

8

501

0.6841

4.0

Mixed

12

349

0,7247

37.2

Latr/ Amdu

11

306

0.7688

50.6

Wash

25

292

0.7914

0.2

Meadow

15

662

0.7497

1.7

Bore Areas

0.7

°See text for explanation of community names.

181

Found on 37.2% of the plot area
was the "n\ixed" community that
generally occupied intermediate ar-
eas between the Latr/ Amdu and ei-
ther washes or areas of high Hilaria
rigida density. It was characterized
by the association of L. tridentata, A.
dumosa and H. rigida, and was found
most often on the slopes above, and
narrow linings next to washes. The
edge, or ecotone, of this community
with the Latr/ Amdu community is
extensively discussed below.

A highly diverse plant community
was found in the washes (table 1; ap-
pendix 1). Such areas not only con-
tained these perennial species, but
also a significant number of other
species found only in this commu-
nity, giving it the highest species
richness of any community.

Small uplifts within wash channels
seemed to support a more open type
of wash vegetation, "sparse wash."
Such areas had many species com-
mon to the washes (appendix 1 ), yet
much of this community was essen-
tially pure stands of the opportunis-
tic grass, Schismus barbatus.

A community ("Hiri/Amdu")
consisting primarily of H. rigida and
A. dumosa was located in upland ba-
sins where L. tridentata was not
found. Such areas were low in habit
and diversity, and very sandy.

Finally, near the south boundary
of the plot, a small "meadow" of
mostly Bailey a multiradiata was
found. Since no tortoises or tortoise
burrows were found there, it was
eliminated from further analyses.

Bare ground, when treated as a
species in transect analyses, had
overriding importance values and
dominance in all communities (ap-
pendix 1). This is often the case in
desert environments. Likewise im-
portance values of S. barbatus were
extremely high in all communities,
pointing to the generally disturbed
nature of the site. Comparisons of
Simpson indices for the communities
revealed significant differences (p <
0.05) in diversity for all communities
except two. The Latr/ Amdu and

wash communities were not signifi-
cantly different (p > 0.50) in their di-
versity.

Soils were found to be somewhat
similar in constituency (table 2), each

being composed to a large degree of
sand. Soil calcium levels (table 3)
were shown to differ significantly.
No detectable calcium was found in
any of the sand scats tested.

Tabie 2.— Summary of percent soli separcstes for plant communities.

Plant

Silt

community"

Sand

(%)

Clay

Classification

Sparse mixed

87

2

11

loamy sand

Hiri/Amdu

90

8

2

sand

Mixed

85

12

3

loamy sand

Latr/ Amdu

70

20

10

sandy loam

Wash

81

3

16

sandy loam

Meadow

63

3

33

sandy day loam

°See text for explanation of community names.

r

Table 3.— Summary of soil calcium levels and their significance.

Plant

Mean soil calcium

Significantly

community*'

Cmeq/100 mg soli)

different from

Latr/Amdu

6.43

Hiri/Amdu

<0.006

Wash

<0.05

Mixed

NS

Hiri/Amdu

3.00

Wash

N$

Mixed

<0.05

Mixed

1.48

Wash

NS

Wash

0.57

°Se0 text for explanation of community names.

^2'Sampte t-test; corrected for type t errors: N$ = not significant.

Table 4.— Distributions and significance of tortoise burrows per grid (Pols-
son, n = 64).

°P>0.9d
^P>Q.7S

J

Number of

Number of grids

Expected values

burrows/grid

random

observed

random"

observed^

0

6

6

4.95

5.52

1

11

17

12.67

13.53

2

15

11

16.21

16.58

3

15

14

13.83

13.54

4

10

8

8.85

8.29

5 or more

7

8

4.67

6.30

182

Table 5.— Summary of frequency of tortoise burrows compared to plant
community abundance.

Plant community*'

Sparse
Wasti

HIrl/
Amdu

Mixed

Latr/
AmHij

Wast!

Other

% of Diot

5 6

A n

37.2

50.6

0.2

2.4

Random'^

observed
expected
Observed^

14
9.2

6

6.6

55
61.0

76
83.0

8

0.3

5

3.9

observed
expected

11
8.8

2

6.3

68
58.8

75
80.0

1

0.3

1

3.8

°See text for explanation of community names.

''P<0.00hn=164.

^P>0.25,n= 158.

_J

Table 6.— Comparison of distance to washes between observed and ran-
dom tortoise burrows.

Plant Mean distance (m (SEM))

community*"

random

observed

t

freedom

P

All

96.83

101.21

0.424

318

>0.50

communities

(6.80)

(7.83)

Mixed

79.54

68.66

0.821

120

>0.50

(9.75)

(8.90)

Latr/ Amdu

132.05

145.40

0.845

148

>0.50

(10.07)

(12.17)

°See text for explanation of community names.

J

r

Table 7.— Comparison of distances to ecotone between observed and ran-
dom tortoise burrows.

Plant

community*'

Mean distance (m fSFMD
random observed

t

Degrees of
freedom

P

All

96.83

101.21

0.424

318

>0.50

Latr/ Amdu

33.63

15.21

5.360

137

< 0.0005

(2.65)

(1.80)

Mixed

38.33

12.18

3.650

65

< 0.0005

(9.09)

(1.60)

Combined

34.05

13.99

6.493

203

< 0.0005

(3.00)

(1.26)

""See text for explarxjtion of community names.

/

Tortoise Burrov^s

A total of 164 tortoise burrows was
found on the study plot (fig. 1). Sev-
enty-five percent were found under
bushes, 14% with the opening under
a bush but the tunnel proceeding into
an open area, 8% with entrances in
the open but the tunnels proceeding
under a bush, and 3% entirely in an
open area. Thus, almost all of bur-
rows (97%) were associated with
shrubs. Of these, 71% were associ-
ated with L. tridentata, 13% each with
H. rigida and A. dumosa, and another
3% with other species.

Neither the distribution of ob-
served or random burrows differed
significantly from the Poisson ex-
pected frequencies (table 4). Like-
wise, when the distribution of ob-
served burrows was compared
against the distribution of random
burrows, no significant difference
was found (chi-square = 2.224; DF =
5; p > 0.50). Thus, when the entire
plot area is considered, tortoise bur-
rows exhibited a random pattern
across the landscape. However, this
was a relatively large scale test of
burrows per arbitrary unit area, and
says nothing about the pattern of tor-
toise burrows in relation to plant
communities.

The abundances of tortoise bur-
rows (both observed and random) in
each plant community were com-
pared against expected frequencies
generated by the abundances of the
plant communities (table 5). Burrows
were sparse in the Hiri/Amdu and
wash communities. Observed bur-
row frequency distribution did not
differ significantly (p > 0.25) from the
expected frequency distribution. The
observed frequency distribution dif-
fered significantly from the random
distribution (chi-square = 11.74; DF =
5; p < 0.05), as did the expected dis-
tribution (chi square = 158.9; DF = 5;
p< 0.001).

Mean observed burrow distance
to the closest wash was compared to
the mean distance from the ran-
domly located burrows (table 6).

183

Comparisons for the sparse wash
and Hiri/Amdu communities were
not done because they would be bio-
logically meaningless or had too low
a sample size, respectively. For all
burrows, and for burrows found in
either the Latr/ Amdu or mixed com-
munities, no significant differences
between random and observed wash
distances were detected. Thus, ob-
served tortoise burrows were not lo-
cated closer to washes than a set of
random points predicted. However,
examination of the spatial pattern
(fig. 1) reveals a lack of burrows deep
within Latr/Amdu and Hiri/Amdu
areas which were furthest away from
any possible wash influence.

Past observations seemed to indi-
cate a correlation between burrow
location and the presence of the edge
of the H. rigida distribution (Baxter
and Stewart 1986). The approximate
distribution of observed burrows to
this edge may be seen in figure 1 .
Mean edge (ecotone) distance of ob-
served burrows was compared to
that of random sites (table 7). Highly
significant differences in ecotone dis-
tances were found in both communi-
ties, and also when combined. Thus,
burrows were found closer to the
ecotone than a set of random points.

Tortoise Captures

Similar analyses were performed for
tortoise capture sites. There were a
total of 120 tortoise captures and re-
captures of 41 individual tortoises.
The observed captures per grid,
along with the randomly located cap-
ture frequencies (same points used
for random burrow sites) were com-
pared against expected values de-
rived from the Poisson distribution
(table 8). Observed capture sites
showed a statistically significant de-
parture from Poisson expected fre-
quencies by the goodness-of-fit test
(p < 0.05).

Frequencies of capture sites in
each plant community were com-
pared against expected values gener-

ated by community abundance (table
9). Observed distributions for both
all captures, and for captures of ac-
tive tortoises (those found outside of
burrows) differed significantly from
expected. These two observed distri-
butions did not differ from each
other (chi-square = 0.5385; DF = 5; p
> 0.99), yet differed significantly
from the randomly generated distri-
bution (chi-square = 18.957 and
19.556, respectively; DF = 5; p <
0.005). Thus, tortoise captures were
not found across the plot in a ran-
dom fashion as would be predicted

by a set of randomly generated
points. Habitat preference for washes
was seemingly indicated, as was a
lack of preference for Hiri/Amdu
areas. These results also gave further
support to the non-randomness ex-
hibited in the Poisson analyses.

To further examine this apparent
non-random distribution of capture
locations, the mean observed capture
distance to washes was compared to
that of the randomly located sites
(table 10). When all capture sites, or
captures within the mixed commu-
nity were considered, a significant

Table 8.—Distrlbutlons and significance of tortoise captures per grid (Pois-
son, n= 64).

Number of
captures/grid

Number of grids
random observed

Expected values
random" observed^

0
1
2
3

4 or more

6
VI
15
15
17

19
17
15
4
9

4.95
12.67
16.21
13.83
13.59

9.97
18.54
17.23
10.68

5.54

''P>0.90
'>P< 0,026

Table 9.— Summary of frequt^ncy of captures compared to plant commu-
nity abundance.

Plant community*"

Sparse

HIrl/

IViixed

Latr/

Wash

Other

Wash

Amdu

Amdu

% of plot

5.6

4.0

37.2

50.6

0.2

2.4

Random'^

observed (n=164)

14

6

55

76

8

5

expected

9.2

6.6

61.0

83.0

0.3

3.9

Observed (oll)^

observed (n=120)

14

1

33

48

23

1

expected

6.7

4.8

44.6

60.7

0.3

2.9

Observed (active)^

observed (n=81)

9

1

20

32

18

1

expected

4.5

3.3

30.1

41.0

0.2

1.9

°See text for explanation of community names.
""Pk 0.001

184

difference between random and ob-
served locations was demonstrated.
However, mean distance to washes
within Latr/ Amdu sites was not sig-
nificantly different from the random
set of points, possibly because the
Latr/ Amdu communities were gen-
erally located further away from
washes, as well as the high variation
in observed Latr/Amdu distances.
These results, along with the results
of the community analysis above,
seemed to indicate a high degree of
tortoise activity near the washes.

Distances to the edge of the H.
rigida were compared between ran-
domly generated and observed cap-
ture locations (table 11). Highly sig-
nificant differences in mean distances
were demonstrated for both the
Latr/Amdu community, and for cap-
tures found in the mixed and Latr/
Amdu communities combined. Cap-

tures within the mixed community
alone were not significantly different
from randomly generated locations.
It seems then that captures, like bur-
rows, were generally not found far
within Latr/Amdu areas, but tended
to be near its edge with the H. rigida
distribution (i.e. the mixed commu-
nity). Because there was no differ-
ence within the nnixed community
alone, differences from random for
captures within the mixed and Latr/
Amdu communities combined were
probably significant due to the
higher number of observations
within the Latr/Amdu community
biasing the sample. Thus, it seems
that tortoises tended to stay either
near the washes, the mixed commu-
nity, or its ecotone with the Latr/
Amdu community, and generally
were not going far within the Latr/
Amdu community.

Table 10.— Comparison of distance to washes between observed and ran-
dom capture locations.

Plant

community*'

Mean distance rSfM))

t

Degrees of
freedom

random

observed

P

All

96,83

71.86

2.189

258

<0.05

communities

(6.80)

(9.39)

Mixed

79.54

44.14

2.081

73

<0.05

(9.75)

(10.61)

Latr/Amdu

132.05

133.66

0.917

117

> 0.50

(10.07)

(15.12)

''See text for explanation of community names.

r N

Table 1 1 .—Comparison of distances to ecotone between observed and
random capture locations.

Plant

community*'

Mean distance (m (SEM))
random observed

Degrees of
t freedom P

Latr/Amdu

32.33

18.59

3.389

114

< 0.001

(9.09)

(3.05)

Mixed

38.63

21.05

1.595

42

>0.10

(2.65)

(4.65)

Combined

34.05

13.99

3.485

157

< 0.001

(3.00)

(2.53)

=See text for explanation of community names.

Discussion

Since the establishment in 1975 of the
Desert Tortoise Council, the amount
of literature published on the desert
tortoise has been considerable.
Oddly enough, only a few papers
may be found that attempt to say
what exactly makes good tortoise
habitat.

A paper by Schwartzmann and
Ohmart (1978) quantified the fre-
quency of use by tortoises in a num-
ber of "habitat types." Their study
took place in the Picacho Mountains
of Arizona's Sonoran Desert, where
tortoises are known to frequent
rocky hillsides and are absent from
valley bottoms (Fritts 1985). Habitat
preferences are just the opposite in
the Mojave Desert, and thus their re-
sults may not be applicable. Like-
wise, Walchuck and Devos (1982)
studied tortoise habitat, but this was
also in the Sonoran Desert of Ari-
zona.

In a draft report, Weinstein et al.
(1986) performed several multivari-
ate analyses on the large Bureau of
Land Management tortoise database.
Several attempts were made to corre-
late abundance with habitat charac-
teristics. Not only were many of
these characteristics derived from the
extrapolation of large scale map data,
but the best fit analysis was found by
designating "corrected sign" of the
transects (the dependent variable;
not actual population numbers) into
arbitrary categories. Indeed, one of
the authors (Berry and Nicholson
1984) has shown that roughly one-
third of population estimates (7 out
of 20 and 4 out of 6) based on sign
transects did not agree with intensive
plot censuses. Also, Turner et al.
(1982) stated that sign transects
"...cannot provide the accuracy and
precision needed..." In addition,
Fritts (1985) stated that such
transects are "...subject to error."
Thus the accuracy of sign transects
are open to serious debate, and al-
though the discriminant analysis
showed some promise as a method

185

for accessing regional abundances,
the nature of the analysis and the
underlying assumptions of both the
data acquisition and techniques leave
much to be desired.

When viewed from the larger
scale of regional or even plot area,
these data seem to indicate that bur-
rows were found in a random fash-
ion when predicted by burrows per
unit area. However, different results
may have been obtained by changing
the size and shape of the grids. For
example, 32 larger rectangular grids
may very well have produced differ-
ent results than the 64 smaller square
grids used in this study. In addition,
such an analysis said nothing about
distributions in relation to habitat
characteristics. Therefore, such a test
should be used as a starting point
and /or support for other tests, and
locally is of limited use by itself for
describing ecologically meaningful
patterns which may exist.

With closer examination, these
data also indicate that burrow loca-
tions were assembled in a pattern
similar to the non-random distribu-
tion of plant communities. Within-
community examinations revealed
patterns of burrow site utilization,
and such patterns were strongly non-
random. At Sand Hill then, while a
majority of burrows were not found
in washes, they were often found
within easy walking distance to a
wash. Very often, burrows were on
slopes high above washes, and possi-
bly within its area of influence. They
were not found far within either the
Latr/ Amdu or Hiri/ Amdu commu-
nities, but were tied strongly to the
edge of these communities with the
mixed community.

Washes are sometimes cited as
being of great importance to tortoise
populations (Burge 1978, Hohman
1977, Lowe 1964). However, results
of this study indicated that tortoise
burrows were not significantly closer
to washes than a set of randomly se-
lected sites. Burge (1978) found 207
(26%) of 783 burrows and pallets
were associated with washes. Of

these, 56 (27%) were actually within
a wash bed. However, Burge appar-
ently eliminated some burrows from
the analysis due to their physical
characteristics. The discrepancy may
be due to the definition used. In this
study, wash burrows were defined
as such, only if they were actually
within the sandy wash bottoms. In
this way, burrows which were on
wash banks, were counted as being
in the plant community of the bank.
Burrows located on wash banks, and
even further away, may have been
associated with the wash, and a re-
classification of these burrows may
show washes to have a more impor-
tant influence in burrow analyses.
Examinations of the actual burrow
distribution (fig. 1) seemed to indi-
cate that they were mostly absent
from areas highly isolated from wash
influence.

The significance of capture loca-
tions in relation to the washes also
seemed to refute the burrow /wash
results. Washes clearly supported a
disproportionate amount of activity
in relation to their abundance on the
plot. Preliminary investigations of
tortoise communities near Kramer
Junction, San Bernardino County,
have also shown tortoises are proba-
bly localizing their activities in the
vicinity of washes (Baxter, unpub.
data).

Several things may explain the
disproportionate amount of captures
in the washes. Greater visibility of
tortoises in the washes may be a fac-
tor. Utilization of highly diverse
plant resources there may also con-
tribute to the localization of activity.
Finally, washes may simply serve a
natural highways for tortoise move-
ments. For instance, several relocated
tortoises at Kramer Junction abruptly
turned and followed trails and
washes upon their release (Baxter,
unpub. data). Regardless, these data
seem to support washes as an impor-
tant habitat characteristic for tor-
toises at Sand Hill. If this population
is representative of other Mojave
populations, the importance of

washes in p>otential relocation sites
will be highly significant in assuring
the best chance of survival for the
relocatees. Further, impacts to
washes may have highly significant
impacts on a population if it is local-
izing its activities there.

These data support the impor-
tance of large woody shrubs (i.e., L.
tridentata) for successful burrow con-
struction at this site. Similar results
have been reported by Burge (1978)
who found 72% of "cover sites" as-
sociated with shrubs. Berry and
Turner (1984) found 75% of juvenile
burrows associated with bushes.
Support for the burrow roofs and
added protection from predators are
likely reasons for this association.
Regardless, the absence of L. triden-
tata from the Hiri/Amdu community
is probably a major reason for the
tortoises not utilizing those areas.
Unsuccessful burrow construction by
virtue of the sandier soils is another
possibility. This latter assumption is
supported by the Weinstein et al.
(1986) analysis which showed "soil
diggibility" as a highly significant
regression variable.

However, the lack of burrows
deep within Latr/ Amdu communi-
ties is not explained by the spatial
abundance of L. tridentata. The high
frequency of burrows and captures
point out that something is being
sought there by the tortoises. Yet,
deep ventures within these areas ap-
parently do not provide resources
that are unavailable at their edges.
Perhaps the higher levels of soil cal-
cium found there are being utilized.
Tortoises must support a massive,
ossified shell, as well as lay eggs, and
calcium may be a very important nu-
trient. Tortoises have been observed
eating dirt (geophagy) and then pro-
ducing "sand scats," and calcium
levels have been hypothesized as an
explanation for this behavior (Sokol
1971). The lack of calcium in the sand
scats tested seems to support this
hypothesis.

In contrast, such deep ventures
would take the tortoises away from

186

the distribution of H. rigida, and the
frequented and diverse washes. Al-
though detailed scat analyses were
not performed, field examination of
hundreds of scats seemed to suggest
that H. rigida is a significant dietary
component. Turner and Berry (1986)
found H. rigida as a part of the diet of
tortoises near Goffs, California.

It would seem then that tortoises
in this area are exhibiting some char-
acteristics similar to "edge" species.
That is, tortoise activity is centered
on the two communities with the
highest vegetational diversity that
border extensive areas of H. rigida.
Since burrows are closely associated
with L. tridentata, they in turn are
found primarily along the only
highly diverse ecotone of the H.
rigida distribution where L. tridentata
importance is the highest. This im-
portance of H. rigida and L. tridentata
is further shown in appendix 1 . The
two communities where tortoises
were not found (i.e., deep Latr/
Amdu and Hiri/Amdu) each lack
one of these species. The assumption
that they are focusing on high diver-
sity areas is further supported by
Weinstein et al. (1968) which shows
"food availability" as the single most
significant regression variable. Fi-
nally, Speake (1986) reports that for
the gopher tortoise (G. polyphemus),
"Edge habitats or ecotonal areas ap-
pear important to tortoises. In each
habitat type except oldfields tortoises
tended to cluster near the edges. In
general, the more edge availability in
a given habitat, the higher the tor-
toise density."

In summary, tortoises utilized the
environment at Sand Hill in a mostly
non-random fashion. Tortoise cap-
tures were spread out between two
communities of highly diverse re-
sources, with clustering occurring at
either edge. Tortoises frequented
washes and the ecotonal edge of the
Latr/Amdu community, with many
found in the intermediate mixed
community. Tortoises were not
found deep within Latr/Amdu or
Hiri/Amdu areas. Burrows were

found close to the ecotone of the
mixed and Latr/Amdu connmunities.
Burrows were not found closer to
washes than randomly located bur-
rows, although this point is far from
clear. Burrows were located close to
the one highly diverse edge of tor-
toise activity area where the impor-
tance of L. tridentata and soil calcium
were the greatest, and were not
found in Hiri/Amdu areas where L.
tridentata was absent, and soils were
the most unconsolidated.

Non-randomness in tortoise popu-
lations is especially important for the
management considerations of relo-
cation. Clearly, despite the best ef-
forts of concerned managers, the use
of the deserts will continue to in-
crease and the frequency of tortoise
relocations will also undoubtedly in-
crease. If tortoise distributions are
random, relocation management es-
sentially becomes a search for safe
relocation sites roughly similar to the
"parental" area. No special consid-
erations of unique habitat types are
required. If on the other hand they
are not, then the relocation site(s)
must include such high-use habitats
as those found in the parental site. In
addition, severe disturbance of such
favored habitats will in turn have se-
vere impacts on the populations, par-
ticularly if small.

This study indicates that the non-
randomness exhibited by the Sand
Hill tortoises is probably a function
of the non-randomness of highly di-
verse plant assemblages and edaphic
characteristics. Thus, the presence of
diverse land forms and their associ-
ated plant communities and diverse
edges within future relocation sites
should be of significant importance
to the manager. Areas which "look
good" to the relocation manager may
not supply the needed resources for
the relocatees. These data are in need
of further support however. If such
patterns are exhibited in other popu-
lations, biologists and managers may
use such techniques to successfully
determine possible habitat require-
ments, and help insure the survival

187

of one of the Mojave's most enig-
matic species.

Acknowledgments

The author wishes to express sincere
thanks to Dr. Glenn R. Stewart of Cal
Poly, Pomona for physical help and
moral support during the fieldwork,
and for his abiding friendship. Many
thanks also to the entire staff at the
MCAGCC for logistical support. Fi-
nally, thanks to K. Berry, D. Speake
and R. Szaro for their constructive
reviews of this manuscript.

This work was supported by
United State Navy contract
N6247484RPOOV48, which was ad-
ministered by the Cal Poly Kellogg
Unit Foundation. Additional equip-
ment support was supplied by
graduate research funds of Cal Poly,
and monies received from the Chuck
Bayless and Tim Brown memorial
scholarship funds. Travel funds were
supplied by Sigma Xi, The Scientific
Research Society.

Literature Cited

Baxter, Ronald J., and Glenn R. Ste-
wart. 1986. Report of the continu-
ing fieldwork on the desert
tortoise (Gopherus agassizii) at the
Twentynine Palms marine corps
base. Proceedings of the sympo-
sium, [Palmdale, Calif., March,
1986]. The Desert Tortoise Coun-
cil, Long Beach, Calif [in press].

Berry, Kristin H. 1984. A description
and comparison of field methods
used in studying and censusing
desert tortoises. Appendix II. In
The status of the desert tortoise
(Gopherus agassizii) in the United
States. Kristin Berry, editor. Re-
port to the U.S. Fish and Wildlife
Service, Sacramento, Calif. Order
No. 11310-0083-81.

Berry, KrisHn H., and Lori L.

Nicholson. 1984. The distribution
and density of desert tortoise
populations in California in the

1970's. In Kristin Berry, editor.
The status of the desert tortoise
(Gopherus agassizii) in the United
States. Report to the U.S. Fish and
Wildlife Service, Sacramento, CA.
Order No. 11310-0083-81.

Berry, Kristin H., and Frederick B.
Turner. 1984. Notes on the behav-
ior and habitat preferences of ju-
venile desert tortoises (Gopherus
agassizii) in California. Proceed-
ings of the symposium, [Lake Ha-
vasu City, Ariz., March, 1984]. The
Desert Tortoise Council, Long
Beach, CA.

Brower, James E. and Jerrold Zar.
1984. Field and laboratory meth-
ods for general ecology. William
E. Brown, publishers. Dubuque,
Iowa.

Burge, Betty. 1978. Physical charac-
teristics and patterns of utilization
of cover sites used by Gopherus
agassizii in southern Nevada. Pro-
ceedings of the symposium, [Las
Vegas, Nevada, 1978]. The Desert
Tortoise Council, Long Beach, CA.

Dodd, C. Kenneth. 1980. Endangered
and threatened wildlife and
plants: listing as threatened with
critical habitat for the Beaver Dam
slope population of desert tortoise
in Utah. Federal Register 45(163):
55654-55666.

Fritts, Thomas H. 1985. Ecology and
conservation of North American
Tortoises (genus Gopherus). II.
Evaluation of tortoise abundance
based on tortoise sign detected in
field surveys. Prepared for: U.S.
Fish and Wildlife Service, Denver
Wildlife Research Center, Univer-
sity of New Mexico, Albuquerque.
NM.

Hach Company. 1983. Soil calcium
and magnesium test kit (model
14855) instruction manual. Hach
Company, Inc., Loveland, CO.

Hohman, Judy P. 1977. Preliminary
investigations of the desert tor-
toise on the Beaver Dam slope in
Arizona. Proceedings of the sym-
posium, [Las Vegas, Nev., March
1977]. The Desert Tortoise Coun-
cil, Long Beach, CA.

Keefe, T. J. and E. Bergerson. 1977. A
simple diversity index based on
the theory of runs. Water Re-
sources 11:689-691.

Lowe, Charles. 1964. The vertebrates
of Arizona. The University of Ari-
zona press, Tucson, Ariz.

Luckenbach, Roger A. 1982. Ecology
and management of the desert
tortoise (Gopherus agassizii) in
California. In R. Bruce Bury, edi-
tor. North American tortoises:

conservation and ecology. U.S.
Fish and Wildlife Service, Wildlife
research report No. 12, Washing-
ton, D.C.

Schwartzmann, James L., and Robert
D. Ohmart. 1976. Quantitative
vegetational data of desert tortoise
(Gopherus agassizii) habitat in the
lower Sonoran Desert. Proceed-
ings of the symposium, [Las Ve-
gas, Nev. March, 1978] The Desert
Tortoise Council, Long Beach, CA.

Appendix 1

Summary of Importance Values^ From Plant Transect Data.

Plant Community^

Species

1

2

3

4

5

Dare vjrouna

1 Oft 0

1 IJ.O

1 1 n

1 lU.O

114 7
1 14./

yD.D

Dcnismus oaroatus

Q

Rl ft

AQ 1
^y. 1

OA ft
Zo.o

40 A

Larrea tridentata

16.5

21.6

36.7

6.7

Ambrosia dumosa

lie

lie:
1 1.0

07 Q

Z/.y

0'2 ft
ZO.O

A 1

riuaria rtgiaa

0 Q

l^Q Q

Oy.y

07
L/ .o

A 7
O./

Erodtum texanum

11.9

6.8

10.2

26.4

19.3

Malacothrix spp.

7 A

OQ 1

1C o

oo.z

1 n c

Eriogonum spp.

/.U

o c
Z.o

1 "3 1
lO.l

o.o

Hymenoclea salsola

111
11.1

O 7
Z./

OA A

Zo.4

/iTnsincicui spp.

Q ft
y.o

0 0

z.z

Oenothera deltoides

6.4

4.8

13.6

Baileya multiradiata

2.0

6.8

6.5

Abronia villosa

5.6

2.2

2.1

Bromus rubens

2.7

2.2

Langloisia Matthervsii

4.9

2.6

Langloisia Palmeri

2.3

Oryzopsis hymenoides

2.5

Eriophyllum Wallacei

2.0

Menodora spinescens

2.9

5.7

Lesquerella Palmeri

2.3

3.1

Salazaria mexicana

3.0

Dalea Fremontii

10.6

Cucurbita foetidissima

2.3

Euphorbia polycarpa

2.0

Isomeris arborea

3.4

Prunus fasiculata

6.7

Spheralcea ambigua

2.6

Salvia columbariae

4.0

Phacelia spp.

2.2

Petalonyx Thurberi

2.6

Unknown composite #1

2.9

2.3

7.3

Unknown composite #2

2.3

'Importance value = relative density + rel. domin. + rel. freq.

'Plant community: See text for description of community names: Meadow and bare
areas not listed: I = Sparse Wash): 2 = Hiri/Amdu: 3 = Mixed: 4 = Latr/Amdu: 5 = Washi.

188

Simpson, E. H. 1949. Measurement of
diversity. Nature 163:466-467.

Sokol, O. M. 1971. Lithophagy and
geophagy in reptiles. Journal of
Herpetology 5:69-71.

Speake, Daniel W. 1986. Gopher tor-
toise density in various south Ala-
bama habitats. Alabama Coopera-
tive Fish and Wildlife Research
Unit, Research Information Bulle-
tin 86-105, 1 p. Auburn, Alabama.

Stewart, Glenn R., and Ronald J.
Baxter. 1987. Final report and
habitat management plan for the
desert tortoise (Gopherus agassizii)
in the West and Sand Hill training
areas of the Twentynine Palms
MCAGCC. Report prepared for
the U.S. Department of the Navy,
San Bruno, Calif. Contract number
N6247484RPOOV48.

Turner, Frederick B., and Kristin H.
Berry. 1986. Population ecology of
the desert tortoise at Goffs, CaH-
fornia, in 1985. University of Cali-
fornia, Los Angeles publication
number 12-1544.

Turner, Frederick B., and Carl
Thelander, Daniel Pearson, and
Betty Burge. 1982. An evaluation
of the transect technique for esti-
mating desert tortoise density at a
prospective power plant site in
Ivanpah Valley, California. Pro-
ceedings of the symposium, [Las
Vegas, Nev., March, 1982]. The
Desert Tortoise Council, Long
Beach, CA.

Walchuck, Sandra L., and James C.
devos, Jr. 1982. An inventory of
desert tortoise populations near
Tucson, Arizona. Proceedings of
the symposium, [Las Vegas, Nev.,
March, 1982]. The Desert Tortoise
Council, Long Beach, CA.

Weinstein, Michael and Frederick B.
Turner and Kristin H. Berry. 1986.
An analysis of habitat relation-
ships of the desert tortoise in Cali-
fornia. Draft report prepared for
Southern California Edison Com-
pany, Los Angeles, Calif.

Zar, Jerrold H. Biostatistical analysis.
1974. Prentice-Hall, Inc., Engle-
wood Cliffs, NJ.

Changes in a Desert Tortoise
(Gopherus agassizii)
Population After a Period of
High IVIortality^

David J. Germano^ and Michele A. Joyner^

Abstract.— An apparent high rate of mortality for
desert tortoises at the Piute Valley in southern
Nevada between 1979 and 1983 significantly
decreased mean carapace length and average
age of the population by 1983. but not density. By
1987, average size and age of the population had
increased and density remained stable.

Chelonians, as a group, are charac-
terized by high rates of adult sur-
vival, delayed maturity, and low
rates of juvenile survival (Wilbur and
Morin 1988). Many chelonians live a
long time after reaching adulthood
(Gibbons 1987), potentially leading to
a long period of reproduction offset-
ting low juvenile survival (Wilbur
and Morin 1988). The desert tortoise
(Gopherus agassizii) (fig. 1) is an her-
bivorous chelonian of the desert
Southwest that exhibits these popula-
tion traits (Berry 1986, Luckenbach
1982, Osorio and Bury 1982, Turner
et al. 1984, 1986). In 1983, a large
number of desert tortoise skeletons
were collected from a study plot lo-
cated in southern Nevada and deaths
were believed to have occurred since
the initial census in 1979 (unpub-
lished report, C. Mortimore and P.
Schneider, Nevada Department of
Wildlife, Las Vegas, NV). It was re-
ported that since 1979, mean cara-
pace length of the population de-
creased, sex ratio had become male
biased, and that population density

'Paper presented at symposium. Man-
agement of Amphibians, Reptiles, and
Small Mammals in Nortti America. (Flag-
staff. AZ, July 19-21. 1988).

'David J. Germane is a doctoral candi-
date. Museum of Southwestern Biology.
Department of Biology. University of New
Mexjco, Albuquerque 87131

^Michele A. Joyner is an undergraduate.
Museum of Southwestern Biology. Depart-
ment of Biology. University of New Mexico.
Albuquerque 87131

decreased, and that these changes oc-
curred because long-term grazing of
this plot by cattle weakened tortoises
to such a degree that decreased for-
age production resulting from below-
average rainfall in 1981 killed many
individuals (unpublished report, C.
Mortimore and P. Schneider, Nevada
Department of Wildlife, Las Vegas).

We recensused this population in
1987 in order to determine changes
that might have taken place since
1983 in age distribution, size distri-
bution, sex ratios, and population
density in order to address the fol-
lowing questions: Of what signifi-
cance are such periods of high mor-

tality to the p)opulations' probability
of survival? How do desert tortoise
populations respond to high rates of
mortality? Are changes in population
demographics long-lasting? Can we
predict future changes in desert tor-
toise populations? We also reassess
possible causes of the high rate of
mortality between 1979 and 1983.

METHODS

Study Area

The 2.59 km^ plot is located in the
Piute Valley of southern Nevada in

190

Figure 3.— Creosote bush and white bursage are the most conspicuous plants of much of the
study plot (top) with Mojave yucca abundant In the northwestern portion (bottom). Other
abundant plants at this site are California buckwheat (Eriogonum fasiculatum), rayless
goldenhead (Acamtopappus sphaerocephalus), Opuntia spp,, bush muhly (Muhlenbergia
porteri), gig galleta (Hilaria riglda), six-week fescue (Festuca octoflora), filaree (Erodium
cicutarium), desert dandelion (Malacothrix glabrata), and Chaenactis spp.

the eastern Mojave desert (fig. 2).
Vegetation is Mojave desert scrub
dominated by creosote bush (Larrea
tridentata) and white bursage (Ambro-
sia dumosa) over the southeastern 2/3
of the plot grading into an area with
an overstory of Mojave desert yucca
(Yucca schidigera) in the northwestern
third (fig. 3).

Field Methods

The population was censused be-
tween April and June 1979 by the Bu-
reau of Land Management (unpub-
lished report, A. Karl, BLM, Las Ve-
gas, NV) and again between April
and June 1983 by the Nevada Depart-
ment of Wildlife (unpublished re-
port, C. Mortimore and P. Schneider,
Nevada Department of Wildlife, Las
Vegas, NV). Each tortoise encoun-
tered was measured, weighed,
marked, its sexed determined, and
its location, behavior and general

Figure 2.— The location of the desert tortoise
permanent study plot (PSP) in the Piute Val-
ley of southern Nevada. The dashed and
dotted lines show major washes.

condition noted. Shells were col-
lected and are catalogued in the Mu-
seum of Southwestern Biology, Uni-
versity of New Mexico, Albuquer-
que.

We recensused the plot 13-27 May
and 18-25 August 1987. We collected
similar data on tortoises, but in-
cluded making casts of the second
costal scute using dental casting ma-

191

terial (Galbraith and Brooks 1987).
Measurements of growth rings from
the impressions on the casts were
taken.

Growth rings of desert tortoises
have been found to be valuable for
determining age and growth histo-
ries of many individuals (Germano
1988). Shells were collected and de-
posited in the Museum of Southwest-
ern Biology.

Data Analysis

Density

Densities in 1979 and 1983 were de-
termined by the investigators who
conducted the censuses using the
Schnabel estimator. This method in-
volves making periodic estimates of
density during the census based on
the number of marked and un-
marked animals found (Tanner 1978).
Because of immigration into the plot,
we reestimated density for 1983 us-
ing the Jolly-Seber estimator (Tanner
1978), which does not assume a
closed population.

As a first approximation of den-
sity for 1987, we used a simple mark-
recapture estimator with May as the
period of marking animals and Au-
gust as the recapture period. Only 1 /
2 the plot was recensused in August
because of time constraints. Density
was computed for this half of the
plot.

Carapace Lengtti Distributions

Carapace lengths (CD of individuals
were plotted and mean CLs com-
puted for live tortoises and remains
for each census year. Mean CLs of
the total population, tortoises >180
mm CL, and tortoises <180 mm CL
were compared among years using
anova with comparisons among
means using Scheffe's multiple com-
parisons test.

Age Distributions

Ages of individuals were plotted for
live tortoises and remains and mean
ages compared in a manner similar
to CLs. Ages of skeletons and 1987
live tortoises were determined for
most individuals using scute annuli,
a technique that is accurate up to 20-
25 years (Germano 1988). Several in-
dividuals were considered to be
older than the number of easily seen
annuli based on non-growth since
last capture, or scute edge beveling,
which indicates continued slow
growth. These individuals were cate-
gorized as >25 years old.

Ages were estimated for live tor-
toises found in 1979 and 1983 using
an age-CL regression (Age = 0.106
CL - 3.82). The number of scute an-
nuli is well correlated with CL (r^ =
0.908, n = 150), although the relation-
ship is less accurate in larger indi-
viduals. We corrected for the pres-
ence of older individuals in our esti-
mates by assigning a portion of
adults of various sizes to the >25 age
category based on the percentage of
adults that were into this category
from the 1987 live and 1983 and 1987
shell groups.

Mortality Rates

Age-specific mortality rates were de-
termined for 1979-1983 and 1983-
1987 using the equation q^ = (k [f J)/
g^, where q^ is the mortality rate per
year for age x, k is the per capita
mortality rate of the population, f^ is
the proportion of animals age x that
are known to have died in the past
year, and g^ is the proportion of ani-
mals of age X in the preceding live
population (Fryxell 1986). In order to
compare mortality rates to age distri-
butions, we determined mortality
rates for age groups 0-14 years, 15-27
years, and >25 years. The per capita
mortality rate was divided by 4 to
obtain the yearly mortality rate for
each time period.

Sex Ratios

Sex ratios were compared among live
tortoises and shells. Sex was assigned
to tortoises >180 mm CL based on
secondary sex characteristics or, in
some instances, for males >170 mm
CL when plastron concavity was ob-
vious. Sex can be determined reliably
in desert tortoises based on shell
characters after 180 mm CL (unpub-
lished report, F. Turner and K. Berry,
Southern California Edison Co., CA)
and female tortoises in this part of
the Mojave desert reproduce at 189
mm CL (Turner et al. 1986), indicat-
ing that sexual maturity probably oc-
curs between 180-190 mm CL. Ratios
were tested for deviation from a 1:1
sex ratio with Chi-square analysis (p
< 0.05).

CL/Weigtit Regressions

Carapace length to weight regres-
sions were constructed for 1979 and
1987 tortoises based on the logarith-
mic transformation of both variables.
Data for 1983 were not available.
Slopes were tested against 0 and
against each other using f-tests (Sokal
and Rohlf 1981).

Growtti Rate Comparisons

Individual growth was compared
among 1987 live tortoises and shell
groups in two ways. Growth rings
were compared among groups using
mean annual widths (AW) and mean
percent growth for rings 1-24 (See
Germano 1988 for a description of
growth ring measurements). Percent
growth for a ring is AW/ estimated
CL for the preceding year. CLs were
estimated using the length of growth
rings from the second costal scute,
which are highly correlated to CL (r^
= 0.96, n = 174). Growth estimates
based on annuli have been found to
accurately reflect carapace growth in
gopher tortoises (Landers et al. 1982)
and desert tortoises (Germano In

192

Press). Means of these variables for
each ring were compared among
groups using the nonparametric

Wilcoxon sign test. We also com-
pared the mean AW and mean per-
cent growth of the last two growth
rings for the shells found in 1983 to
the mean AW and mean percent
growth of the 1980 and 1981 growth
rings from live tortoises found in
1987 using f-tests.

'rill i'ri"i"l"rT"("i'T'i"r"i"ri"i'TiTi i i i i i —
I t s 7 I II II B 17 II n u tt n n >za

AGE (yeors)

1 I 7 * II n II IT If ai M tt IT II >ii

AGE (y e a r t )

Figure 4.— Population size distributions for
iive desert tortoises from \he Piute Valley
permanent study plot. Mean carapace
lengths and somple sizes are given in table
1.

Figure 5.— Population size distrit>utions for
desert tortoises found dead in 1983 and
1987 from the Piute Valley permanent study
plot. Mean carapace lengths and sample
sizes are given in table 1 .

Table 1 .—Mean carapace lengths (mm) of tortoises from the Piute Valley
permanent study plot. Standard deviation and sample size are given be-
low \he mean.

All

% of

%of

Group

tortoises

>180 mm CL

total

<180 mm CL

total

1979 live

186.8

217.1

58

144.5

42

(44.0, 84)

(21.0,49)

(30.8, 35)

1983 live

148.2

211.8

37

110.8

63

(59.6,81)

(24.9, 30)

(38.3,51)

1987 live

181.1

213,8

60

125.8

40

(46.6, 48)

(20.0, 29)

(37.8, 19)

1983 shells

197.6

212.9

78

106.4

22

(93.3, 108)

(22.6, 84)

(39.0, 24)

1987 shells

165.4

216.3

49

117.2

51

(58.1,37)

(19.4, 18)

(36.9, 19)

Climate Analysis

Climate was analyzed using weather
information from Searchlight, Ne-
vada. Data were compared for 3 time
periods; 1970-June 1979, July 1979-
1982, and July 1979-July 1987. Means
and variances of rainfall, both annual
and winter, were compared among
time periods. Mean monthly tem-
peratures were compared among
time periods and temperatures below
freezing were analyzed for duration
and relation to unusually warm win-
ter daily highs.

RESULTS
Density

Tortoise density was estimated to be
50/km2 in 1979 and 72/km2 in 1983
by the authors of these censuses.
Eighty-four and 81 tortoises were
found in 1979 and 1983, respectively.
We reestimated the 1983 density to
be 44 tortoises/km^. We estimated
the density in 1987 to be 59 tortoises/
km^ (95% confidence intervals, 19-
173). We found 48 tortoises in 1987,
33 in May and 19 on the southern
half of the plot in August, of which 4
had been marked in May.

Carapace Length Distributions

Distributions of CLs of live tortoise
populations varied significantly for
each census (fig. 4). Mean CL was
significantly smaller in 1983 than in
either 1979 (p<.05) or 1987 (p<.05).
Mean CLs in 1979 and 1987 were not
significantly different, however
(p>.05, table 1). No significant differ-
ences were found among mean CLs
for adults (>180 mm CL). Adults
comprised 58% of the 1979 popula-
tion, 37% of the 1983 population, and
60% of the 1987 population. The '
mean CL of non-adults (<180 mm
CL) was significantly smaller in 1983
than 1979 (p<.05), but was not sig-
nificantly different than 1987 (p>.05.

193

table 1). The mean CL of non-adults
was not significantly different be-
tween 1979 and 1987 (p>.05).

Remains of 37 tortoises were
found in 1987 compared to 109 found
in 1983 (fig. 5). Ten shells were found
in 1979. CLs of remains were not sig-
nificantly different (p>.05), although
mean CL in 1983 was considerably
larger than for 1987 (table 1). Mean
CLs of adult remains in 1983 and
1987 were similar, as were non-adult
CLs, but adults comprised 78% of the
1983 collection and only 49% of the
1987 collection. The mean CL of re-
mains from 1983 was not signifi-
cantly different from the mean CL of
live tortoises in 1979 or 1987, but was
significantly larger than live tortoises
in 1983 (p<.05). Mean CL of remains
from 1987 was not significantly dif-
ferent than any live tortoise means.

Age Distributions

Ages of tortoises varied significantly
among years (table 2). Changes in
age distributions of live tortoises
were similar to the changes seen for
CLs (fig. 6). The estimated mean age
for 1979 was significantly older than
1983 (p<.05) but not 1987 (p>.05).
Mean age for 1987 was not signifi-
cantly different than 1983 (p>.05), but
non-adults were significantly older
(p<.05). Mean age of 1983 remains
was significantly older than 1983 live
tortoises (p<.05), but was not signifi-
cantly different than 1987 live tor-
toises or remains (p>.05, fig. 7).

Mortality Rotes

Death rates for 1983-1987 were lower
than for 1979-1983. Per capita mortal-
ity rate (k) for 1979-1983 was 0.21/
year (N = 130) and was 0.08/year for
1983-1987 (N = 115). Mortality rates
dropped for all age classes after 1983.
For 1979-1983 mortality rates were
0.145/year for 0-14 year olds, 0.247/
year for 15-25 year olds, and 0.195/
year for tortoises >25 years. For 1983-

1987 mortality rates were 0.061 /year
for 0-14 year olds, 0.093/year for 15-
25 year olds, and 0.103 for tortoises
>25 years. Mortality rates for all
adults (15-25 years and >25 years) for
1979-1983 was 0.240/year and for
1983-1987 was 0.103/year.

Sex Ratios

Sex ratios of live tortoises show an
increasing proportion of males (table
3), although only 1987 showed a sig-
nificantly biased sex ratio. When the
1987 sex ratio was analyzed by size,
92% of tortoises >220 mm CL were
males, whereas only 53% of tortoises
180-219 mm CL were males (table 3).
When analyzed by age, 63% of
tortoises >20 years were males, but
71% of tortoises of known sex be-
tween 13-19 years were males, a sig-
nificantly higher proportion than fe-
males. The sex ratios of dead tor-
toises were not significantly different
than 1:1 (table 3).

CL/Weigtit Regressions

The regressions of weight against CL
had significant slopes for 1979 and

CARAPACE LENGTH (mm)

Figure 6.— Population age distributions for
live desert tortoises from the Piute Valley
perrrKinent study plot. The 1979 and 1983
age distributions are estimates based on a
carapace length to annulus number re-
gression. A proportion of adults were
placed in the >25 age category based on
the proportion of adults In this category
from the age distributions for which ages
were assigned by annuli counts. The 1987
age distribution Is based on annuli counts.

r

Table 2.--Mean ages of tortoises from the Piute Valley permanent study
plot in southern Nevada. Standard deviation and sample size are given
below the mean. Ages for 1 979 and 1 983 are estimates based on cara-
pace length (see Methods).

Ages (years)

Group

0-27

0-14

15-27

^>25

1979 live

16.6

10.9

19.5

(5.1,72)

(3.4, 24)

(2.9, 48)

(12)

1983 live

12.1

7.5

18.8

(6.6.74)

(3.7,41)

(3.0, 30)

(7)

1987 live

14.1

11.3

17.0

(3.8, 43)

(3.2, 22)

(2.2,21)

(5)

1983 shells

17.0

7.8

19.9

(6.2, 94)

(3.6, 22)

(3.3, 72)

(14)

1987 sl-»ells

14.0

8.4

19.3

(62., 31)

(2.6, 15)

(3.5, 16)

(6)

'Mean age cannot be determined.

194

1987 (fig. 8). The regression equation
for 1979 is gram weight = 0.000317
CL2.924 ^ 0.952, n = 73) and for
1987 is gram weight = 0.000505
(2L2.826 (jj ^ 0.969, n = 53). Regression
slopes were not significantly differ-
ent from each other (p>.10).

Growth Rate Comparisons

No significant differences were
found in a ring by ring comparison of
growth between 1987 live tortoises
and 1983 remains for either annual
widths (AW) or percent growth.

When 1980 and 1981 rings were com-
pared, no significant difference ex-
isted between the mean AW for the
last two rings of 1983 mortalities (X =
1.98mm, n = 72) and the 1980 and
1981 rings for 1987 live tortoises (X =
1.92nmi, n = 79; p>.10).

CARAPAC E LENGTH (mm)

Figure 7.— Population age distributions for
desert tortoises found dead in 1 983 and
1987 from the Piute Valley pernnanent study
plot. Both the 1983 and the 1987 age distri-
butions are based on counts of annuli.

9

4.0-
l.t-
1.0-
IJ

t.o-
1,1

1.0

I • T 1

I I \ — r— I — l—l — I — I — r— I — I I I — r— I — I — I — I — r

M

1.0

u

l.«
I.I-

1.0-

•.I

HIT

I T I I I I I I I I I I I I I I I I I — I — r
to M no HO NO NO WO tot tW 140 tM

CARAPACE LENGTH (mn)

Figure 8.— Regressions of carapace length
to weight for desert tortoises found in 1979
and 1987. Slopes of both regressions are
significantly different from 0 but not from
each other.

Table 3.— Numbers of males to females for desert tortoises from the Piute
Valley permanent study plot. Significant departures from a 1 :1 sex ratio
were determined by Chl-square analysis. The 1987 live totals were sub-
categorized by size and age.

Year

Males

Females

Ratio

1979

live

24

30

0.88

1

0.667

shells

4

3

1.33

1

0.001

1983

live

22

11

2

1

3.667

shells

35

41

0.85

1

0.474

1987

live (total)

20

9

2.22

1

M.172

size: 180-219 mm CL

9

8

1.13

1

0.059

>220mm CL

11

1

11

1

^8.330

age: 13-19 years

15

6

2.5

1

^3.857

>20 years

5

3

1.67

1

0.500

shells

11

6

1.83:1

1.471

'Significanf departure from 1:1 ratio (p<.05).

Climate Analysis

Average precipitation were higher
between July 1979 and July 1987 than
the previous 10 years (table 4). The
highest average precipitation was
recorded between July 1979 and De-
cember 1982. Winter rainfall (Octo-
ber-March) followed the same pat-
tern, with both 1979-1987 and 1979-
1982 averages higher than 1970-1979
(table 4). The period 1970-1979 was a
drought period with average rainfall
7% below the long-term average of
183.8 mm and 7 of the 10 years were
well below average (table 4). When
1978 and 1979 are excluded, average
precipitation drops to 129.3 mm, 30%
below the long-term average. July
1979-December 1982 averaged 40%
higher rainfall than the long-term av-
erage with only 1981 experiencing
below-average rainfall. Mean
monthly high and low temperatures
were similar among time periods. No
extended periods of freezing tem-
peratures were found for daily read-
ings between 1979 and 1983.

DISCUSSION

Population Parameters

The desert tortoise p>opulation in the
Piute Valley study plot experienced a
high rate of mortality, particularly of
adults, between July 1979 and 1983.
Related to this event was a signifi-
cant decrease in the size and age dis-
tributions of the population in 1983,
although both were returning to 1979
dimensions by 1987. The lower mean
age in 1983 is probably a result of in-
creased survival of hatchlings and
increased immigration. The increased

195

survival of hatchlings, as shown by
the significant increase of tortoises in
the 1-4 age group in 1983, may be
due to more favorable conditions be-
cause of lower densities just after the
high rate of mortality, or to optimal
climatic and habitat conditions.

It is possible that the greater num-
bers of smaller tortoises found in
1983 could have resulted from better
search effort for these sizes (Berry
and Turner 1984), but we censused
the plot carefully in 1987, specifically
looking for small tortoises, yet we
found relatively few. While we do
not doubt that young are missed be-
cause of their inconspicuousness, we
believe that the changes in size and
age distributions between 1979 and
1987 reflect actual population
changes.

The size and estimated age distri-
butions for 1983 indicate that a sig-
nificant number of smaller and
younger tortoises came into the plot
between 1979 and 1983. Judging by
the male-dominated sex ratio after
1979, immigration largely has been
by young males. The biased sex ra-
tios are not due to higher adult male
survival since equal proportions of
males and females died. Most of the
males in the present population are
fairly young, although they are large.
Male turtles are known to disperse
greater distances than females (Gib-
bons 1986).

Although many turtle populations
have biased sex ratios, evolutionary
theory indicates that these ratios
should be under selective pressure to
be relatively even, in most instances
(Fisher 1930, Trivers 1972). However,
desert tortoise age to maturity is ca.
15 years (Germano In Press, Woo-
dbury and Hardy 1948), therefore a
reproductive solution mediated by
selection would require hundreds of
years.

Censuses in other parts of this val-
ley in 1983 indicate that this high rate
of mortality was confined to this plot
and areas close by (unpublished re-
port, C. Mortimore and P. Schneider,
Nevada Department of Wildlife, Las

Vegas, NV). Differences in sex ratios
at this plot may be more a reflection
of higher male movement rates com-
pared to females and not to a real
difference in numbers of males and
females in the population as a whole.
Over time the sex ratios may change
by movement of females into the plot
from outside.

Density may have decreased
slightly since 1979, but it does not ap-
pear to have changed significantly
over the 8 year period, although we
recognize the imprecision of these
density estimates. The number of tor-
toises found has decreased in each
census, but investigators and time
periods in the field have varied, ren-

Table 4.— Annual and winter precipitation (mm) for 1970-1987 and for 3
time periods from the Searchlight, Nevada NOAA Station. Winter precipita-
tion is defined by the months October-March. Means and standard devia-
tions are given for the 3 time periods. Precipitation for 1987 only Includes
the months of January-July.

Year

1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987

Time period

Annual Winter
total total

Jan. 1970- July 1979- July 1979
June 1 979 Dec. 1 982 July 1 987

127.76
68.83
136,65
114.81
184.40
132.08
161.80
107.70
473.71
256.54
313.44
162.81
366.10
376.68
300.48
149.35
166.88
73.66

30,73

17.02
179.02

54,36
100,08

82.79

52.58
183.90
249.43
260.10

67.06
101.09
216.15

61.47
191.52

91.69
126.24

Annual precipitation

170.9
(113.0)

281.2
(86.4)

265.5
(148.8)

Winter precipitation

104.4
(33.0)

139.4
(73.7)

161.0
(71.7)

196

dering this comparison unreliable.
We believe that the lower number of
live tortoises found in 1987 is due to
inexperienced field personnel and
the shorter duration of time in the
field. The most valid of these density
estimates is the Jolly-Seber estimate
of 44 tortoises /km^, because more
assumptions are met with this tech-
nique. Unfortunately, estimates can-
not be made for the first or last cen-
sus with this technique. Density esti-
mates, though, are similar in magni-
tude and we believe this indicates
that density has remained relatively
stable since 1979. The population
must have experienced a decline af-
ter 1979 but we believe that increased
survival of young and immigration
from adjacent non-affected areas has
quickly returned the density to 1979
levels.

Mortality Factors

Causes of the high rate of mortality
have not been demonstrated. The
hypothesis that long-term grazing
confounded by a drought in 1981
was the cause of the high number of
tortoise deaths is not supported by
growth analysis of annuli, CL/
weight data, or climate data. Growth
did not differ significantly between
those that died before 1983 and those
that survived to 1987. In addition,
the weight to size regressions for
1979 and 1987 were the same and
both were almost identical to the re-
gression for tortoises from an un-
grazed plot in Nevada (Medica et al.
1975). As for a drought in 1981, aver-
age rainfall was only 9% below the
long-term average (up to 1987) and
was actually at the average, up to
1981, given the drought in the 1970s.
Preceding 1981 were 3 years of ex-
ceptionally high rainfall. In contrast,
rainfall in 1977 was 41% below aver-
age and followed many drought
years (table 4).

Desert tortoises are known to
store water (Nagy and Medica 1986)
and may be able to store fat. It seems

doubtful that one average year of
rainfall after 3 very good years could
cause starvation or lethal dehydra-
tion. The 2 years preceding our cen-
sus in 1987 were below average in
precipitation, yet mortality rates
dropped. The period 1970-1977 was a
drought, yet only 10 shells were
found in 1979. If these low rainfall
years didn't produce a high rate of
mortality that could be detected in
1979, it is hard to imagine that one
average year after 3 good years
would result in excess mortality. Es-
timates of yearly adult death rates
from 1972-1982 for a population only
42 km south of this site was 1.2%, in
an area that has been grazed by live-
stock for 100 years (Berry and
Nicholson 1984a).

Other possible causes for this mor-
tality could have been disease, pre-
dation, or flooding. Diseases are
known to affect other turtle species
in the wild (Jacobson 1980a,b), but no
evidence exists for disease as a fac-
tor. Many of the shells show signs of
chewing by carnivores, although
whether this indicates predation or
scavenging cannot be determined.
Flooding occurred in or near the plot
in 1980 and 1982 (unpublished re-
port, J. Jamrog and R. Stager, BLM,
Las Vegas, NV). The plot is dissected
by numerous washes that are most
prevalent in this part of the valley
(fig. 2).

The exact cause of the high rate of
mortality may never be known. Star-
vation, disease, flooding, and preda-
tion may have all had an effect. No
singular explanation is supported by
the data. Whatever the causative
agent, the population appears to be
returning to a density and popula-
tion structure as occurred before the
f)eriod of high mortality.

Management Implications

As a long-lived reptile, the desert tor-
toise is more vulnerable to fluctua-
tions in adult mortality than to simi-
lar fluctuations in younger age

groups. Many desert tortoise popula-
tions consist of adult segments that
usually have yearly survivorship
rates of 95-98% (Berry and Nicholson
1984b). High adult survivorship is
often coupled with low juvenile sur-
vivorship (Wilbur and Morin 1988)
and part of the concern for tortoise
populations is that they may not
have the ability to withstand distur-
bance because of low juvenile survi-
vorship. Female desert tortoises in
the eastern Mojave desert have the
ability to lay 2-3 clutches in a season
(Turner at al. 1986). The significant
increase in 1983 of tortoises 1-4 yr of
age suggests more hatchlings have
survived between 1979-1983 than
previously. As with any other popu-
lation parameter, juvenile survivor-
ship can vary, and this may lead to
periodic additions of greater num-
bers of young surviving to adult age.

It appears that desert tortoises
have the ability to recover from dis-
turbance in some instances. This ap-
pears to be what is happening at the
Piute plot. Increased juvenile survi-
vorship and immigration are holding
the population density stable and the
age and size distributions are return-
ing to 1979 dimensions. This kind of
recovery may not occur if a distur-
bance is prolonged or is widespread.
Those managing desert tortoises
must be aware of the dynamics of
each population, but it is apparent
that tortoise populations can recover
from short-term high mortality.

ACKNOWLEDGMENTS

We thank T. Fritts and the National
Ecology Research Center of the U.S.
Fish and Wildlife Service for provid-
ing support during data collection
and analyses. We also thank R. Wil-
ingham, J. Talbert, and C. Isbell for
assistance with the May census. R.
Haley and B. Turner of the Nevada
Department of Wildlife provided re-
|X)rts and shells for this site. T. Fritts,
M. Molles, N. Scott, H. Snell, K. Sev-
erson, and 2 anonymous reviewers

197

read drafts of this manuscript and
greatly improved its content, but any
errors or omissions are our own.

LITERATURE CITED

Berry, Kristin H. 1986. Desert tortoise
(Gopherus agassizii) research in
Cahfornia, 1976-1985. Herpetol-
ogica 42:62-67.

Berry, Kristin H., and Lori L.
Nicholson. 1984a. Attributes of
populations at twenty-seven sites
in California, p. 154-241. In The
Status of the Desert Tortoise (Go-
pherus agassizii) in the United
States, Kristin H. Berry, editor.
Desert Tortoise Council Report to
U.S. Fish and Wildlife Service,
Sacramento, California. Order No.
11310-0083-81.

Berry, Kristin H., and Lori L.

Nicholson. 1984b. A summary of
human activities and their impacts
on desert tortoise populations and
habitat in California, p. 61-117. In
The Status of the Desert Tortoise
(Gopherus agassizii) in the United
States, Kristin H. Berry, editor.
Desert Tortoise Council Report to
U.S. Fish and Wildlife Service,
Sacramento, California. Order No.
11310-0083-81.

Berry, Kristin H., and Frederick B.
Turner. 1984. Notes on the behav-
ior and habitat preference of juve-
nile desert tortoises (Gopherus
agassizii) in California, p. 111-130.
In Proceedings of the 1984 Sympo-
sium of the Desert Tortoise Coun-
cil, Mary W. Trotter, editor. Long
Beach, California.

Fisher, Ronald A. 1930. The genetical
theory of Natural Selection. Ox-
ford Univ. Press, London. 272 p.

Fryxell, J. M. 1986. Age-specific mor-
tality: an alternative approach.
Ecology 67:1687-1672.

Galbraith, David A., and Ronald J.
Brooks. 1987. Photographs and
dental casts as permanent records
for age estimates and growth
studies of turtles. Herpetological
Review 18:69-71.

Germano, David J. 1988. Age and
growth histories of desert tortoises
using scute annuli. Copeia 1988: [In
Press].

Germano, David J. In Press. Prelimi-
nary comparison of growth in des-
ert tortoises and gopher tortoises.
In Proceedings of the 1987 Sympo-
sium of the Gopher Tortoise
Council, Richard Franz and C.
Kenneth Dodd, Jr., editors.

Gibbons, J. Whitfield. 1986. Move-
ment patterns among turtle popu-
lations: applicability to manage-
ment of the desert tortoise. Herpe-
tologica 42:104-113.

Gibbons, J. Whitfield. 1987. Why do
turtles live so long? Bioscience
37:262-269.

Jacobson, Elliott R. 1980a. Viral
agents and viral diseases of rep-
tiles, p. 197-202. In Reproductive
Biology and Diseases of Captive
Reptiles, James B. Murphy and
Joseph T. Collins, editors. Society
for the Study of Amphibians and
Reptiles, Contribution to Herpe-
tology Number 1.

Jacobson, Elliott R. 1980b. Mycotic
diseases of reptiles, p. 235-241. In
Reproductive Biology and Dis-
eases of Captive Reptiles, James B.
Murphy and Joseph T. Collins,
editors. Society for the Study of
Amphibians and Reptiles, Contri-
butions to Herpetology Number 1.

Landers, J. Larry, W. Alan McRae,
and James A. Garner. 1982.
Growth and maturity of the go-
pher tortoise in southwestern
Georgia. Bulletin of the Rorida
State Museum, Biological Sciences,
27:81-110.

Luckenbach, Roger A. 1982. Ecology
and management of the desert tor-
toise (Gopherus agassizii) in Califor-
nia, p. 1-37. In North American
Tortoises: Conservation and Ecol-
ogy, R. Bruce Bury editor. U.S.
Fish and Wildlife Service, Wildlife
Research Report 12. Washington,
D.C

Medica, Philip A., R. Bruce Bury, and
Frederick B. Turner. 1975. Growth
of the desert tortoise (Gopherus

agassizi) in Nevada. Copeia
1975:639-643.

Nagy, Kenneth A., and Philip A.
Medica. 1986. Physiological ecol-
ogy of desert tortoises in southern
Nevada. Herpetologica 42:73-92.

Osorio, Sandalio Reyes, and R. Bruce
Bury. 1982. Ecology and status of
the desert tortoise (Gopherus agas-
sizii) on Tiburon Island, Sonora. p.
39-55. In North American Tor-
toises: Conservation and Ecology,
R. Bruce Bury, editor. U.S. Fish
and Wildlife Service, Wildlife Re-
search Ref)ort 12. Washington,
D.C.

Sokal, Robert F., and F. James Rohlf.
1981. Biometry. Second edition.
W. H. Freeman and Company,
New York. 859 p.

Tanner, James T. 1978. Guide to the
study of animal populations. Uni-
versity of Tennessee Press, Kn-
oxville. 186 p.

Trivers, Robert L. 1972. Parental in-
vestment and sexual selection, p.
136-179. In Sexual Selection and
the Descentof Man— 1871-1971,
Bernard Cambell editor. Aldine
Publishing Co., Chicago.

Turner, Frederick B., Page Hayden,
Betty L. Burge, and Jan B. Rober-
son. 1986. Egg production by the
desert tortoise (Gopherus agassizii)
in California. Herpetologica 42:93-
104.

Turner, Frederick B., Philip A.

Medica, and Craig L. Lyons. 1984.
Reproduction and survival of the
desert tortoise (Gopherus agassizii)
in Ivanpah Valley, California.
Copeia 1984:811-820.

Wilbur, Henry M., and Peter J.

Morin. 1988. Life history evolution
in turtles, p. 387-439. In Biology of
the Reptilia, Volume 16, Carl Gans
and Raymond B. Huey, editors.
Alan R. Liss, Inc., New York.

Woodbury, Angus M., and Ross
Hardy. 1948. Studies of the desert
tortoise, Gopherus agassizii. Eco-
logical Monographs 18:145-200.

198

A Survey Method for
Measuring Gopher Tortoise
Density and Habitat
Distribution^

Daniel M. Spillers and Dan W. Speake^

The only tortoise to occur in the
southeast, the gopher tortoise (Go-
pherus polyphemus) (fig. 1), is limited
to six states. Of these six states, legal
protection is offered by South Caro-
lina, Mississippi, Georgia, Florida
and Alabama; Louisiana does not re-
strict the harvest on gopher tortoises
at present. The gopher tortoise is
now federally listed as threatened in
the portion of its range west of the
Tombigbee river in Alabama.

During the past several years, an
apparent decline of gopher tortoise
p)opulations has been noted. Boze-
man (1971) and Wharton (1978)
noted the rapid loss and alteration of
sand ridge habitat, the habitat in
which most gopher tortoise popula-
tions occur, and argued for the pres-
ervation of these habitats not only for
gopher tortoises but also for other
aspects of their ecological signifi-

'A contribution of the Alabama Coop-
erative Rshi and Wildlife Research Unit: Au-
burn University Agricultural Experiment Sta-
tion and Department of Zoology and Wild-
life Science, Game and Rsh Division of the
Alabama Department of Conservation and
Natural Resources, the U.S. Fish and Wildlife
Service and the Wildlife Management Insti-
tute cooperating. Presented at the Sympo-
sium on Management of Amphibians, Rep-
tiles, and Small Mammals in North America,
July 19, 1988.

'Spillers is a research technician and
Speake is assistant unit leader/wildlife with
the Alabama Cooperative Fish and Wildlife
Research Unit. Auburn University, Alabama
36849-5414.

Abstract.— An underground closed-circuit
television camera and Landsat satellite imagery
were utilized in a 2-year study to examine status of
the gopher tortoise in southern Alabama. Use of this
camera resulted in a complete count of gopher
tortoises in the sample transects. The transects were
located precisely on standard topographic maps
and on Landsat images. An estimation was then
made of the amount of each habitat type in
southern Alabama based on light reflectance of the
vegetation and soil type of the sample transects.
Density measurements were then expanded to
estimate tortoise numbers for the entire area. This
method is effective for estimating gopher tortoise
numbers and for determining quantity and location
of gopher tortoise habitat.

cance. Auffenberg and Franz (1982)
documented a decline of gopher tor-
toise populations on specific sites in
the Southeast. Landers et al. (1980)
found that gopher tortoises have
such a low reproductive rate that
human exploitation of tortoises can
drastically reduce local populations.
Landers and Speake (1980) showed
that population densities of gopher
tortoises can fluctuate widely in re-
sponse to habitat manipulation or
neglect. Other conceivable reasons
for this apparent decline were noted
by Diemer (1986).

Sand ridge habitat is not only im-
portant for gopher tortoises, but also
for many other animals that use go-
pher tortoise burrows for nesting,
feeding, or escape cover. Three sub-
species of the crawfish-gopher frog
complex that are closely associated
with gopher tortoise burrows are the
dusky gopher frog (Ram areolata
sevosa), the Rorida gopher frog, (R. a.
aesypus), and the Carolina gopher
frog (R.a. capito). The threatened
eastern indigo snake (Drymarchon
corais couperi) is dependent on tor-
toise burrows for winter cover in the
northern part of its range (Speake et
al., 1978; Landers and Speake, 1980;
Diemer and Speake, 1981). Several
species of mammals and birds use
gopher tortoise burrows, most often
as escape cover. Several authors have
noted the diversity of animal life
(both vertebrate and invertebrate)

Figure 1 .—A gopher tortoise from southern
Alabama.

inhabiting tortoise burrows and the
dependence of some species on tor-
toise burrows for survival (Allen and
Neill, 1951; Hubbard, 1894; Hutt,
1967; Landers and Speake, 1980;
Speake et al., 1978; Woodruff, 1982).

In view of the apparent decline of
gopher tortoise populations, it is im-
p)ortant to be able to accurately meas-
ure tortoise density in an area and to
determine quantity and distribution
of suitable tortoise habitat. Tortoise
density has been previously esti-
mated by means of a correction fac-
tor applied to counts of burrows
(Auffenberg and Franz, 1982), dig-
ging of burrows, and use of listening
devices. Previous methods do not
ensure accurate determination of tor-
toise density without burrow de-
struction and prohibitive labor. De-
termination of quantity and location
of tortoise habitat is becoming neces-

199

sary due to rapid changes in land use
and increasing relocation and re-
stocking efforts (Diemer, 1984; Lan-
ders, 1981).

The objectives of this study were
to develop and employ a method to:
(1) accurately measure gopher tor-
toise density and (2) locate and quan-
tify tortoise habitat in a 24-county
area of southern Alabama.

We are indebted to James Altiere,
Eugene Carver, Kevin Dodd, Lane
Knight, Sonny Mitchell, Claud
Searcy, and William Sermons, who
assisted in collecting field data. We
are especially indebted to Walter
Stephenson, Chief of the Resource
Development Section of the State
Planning Division, Department of
Economic and Community Affairs,
State of Alabama for his help and co-
operation in giving us access to the
Landsat remote sensing system. Ap-
preciation is extended to Joe Exum,
Raymond Metzler, and Nick Wiley
for their assistance in experimental
design and data analysis. Special ap-
preciation is extended to Dr. Charles
Williams of the Research and Data
Analysis Department, Auburn Uni-
versity, for his advice and aid with
statistical design and analysis. The
project was funded by a grant from
the U.S. Fish and Wildlife Service
and by the Alabama Cooperative
Fish and Wildlife Research Unit.

Methods

Study Area Determination and
Questionnaires

Our study area was determined by
the reported historical range of the
gopher tortoise in Alabama (Mount,
1978; Auffenberg and Franz, 1982).
This included 24 counties in the
coastal plain of Alabama (excluding
the counties west of the Tombigbee
river which were surveyed by other
researchers). Questionnaires were
sent out to wildlife biologists, conser-
vation officers, herpetologists,
county agents, soil conservation

agents and other people who were
likely to have knowledge of gopher
tortoise populations in our 24-county
study area. These questionnaires
asked for locations of areas that sup-
ported or had supported tortoise
populations, and names of landown-
ers or other persons who might have
additional knowledge of tortoise
populations. A map was included
with each questionnaire so that loca-
tions could be marked. A total of 132
questionnaires was mailed out and
58% were returned.

Soil conservation offices were vis-
ited in each surveyed county and fur-
ther inquiries were made concerning
tortoise population occurrence and
habitat availability. Areas in each
county that had soils with sand to a
depth of at least 1 m and that pref-
erably contained a variety of habitat
types were delineated on maps.
These areas were considered poten-
tial tortoise habitat (Garner and Lan-
ders, 1981; Landers, 1981; Landers
and Garner, 1981) and were used to
sample tortoise densities.

After evaluation of the informa-
tion from the questionnaires, per-
sonal interviews, and discussion with
soil conservation agents, the 24-
county study area was divided into
three classes (fig. 2). Class I counties
(n=14) contained widely distributed
gopher tortoise {X)pulations and
habitat. Class II counties (n=4) con-
tained relict or disjunct populations
and scattered, spotty habitat. Class
III counties (n=6) were those in
which no tortoise populations could
be found.

Sampling Sctieme

In Class I counties, regions deline-
ated by the soil conservation agents
(sandy soil > 1 m) were located on
1:24,000 scale topographic maps.
Within these areas, a reference point
for initiation of sampling was chosen
from the map which had a variety of
habitat types (at least 2) within a 1
km radius of the reference point.

These points were chosen before vis-
iting the site. Where necessary, per-
mission was obtained for sampling
on private property.

Upon arrival at the location as
many of the following habitat types
were located as possible: unburned
pine/ scrub oak, burned pine/ scrub
oak, planted pines, clearcuts, old-
fields, agricultural fields, pasture,
and corresponding edges for each
type. The example of each habitat
type nearest to the reference point
was then sampled.

Belt transects measuring 265 x 15
m (0.4 ha) were systematically lo-
cated within the habitat types avail-
able; edge transects were centered on
and followed the edge. If there were
open burrows in the transect, the
burrows were examined using the
MUTVIC (Miniature Underground
Television Inspection Camera)
(Speake and Altiere, 1983). This de-
vice enabled us to insert a closed-

Figure 2.— Distribution of the gophier tortoise
in 24 counties of Alabama.

200

circuit television camera to the bot-
toms of the burrows and determine if
they were occupied (figs. 3-5). Bur-
row width measurements were made
with calipers inserted approximately
70 cm into the burrow. Data gathered
for each transect included habitat

type, number of open burrows, num-
ber of active burrows (burrows with
sign of recent tortoise use), number
of tortoises, and width of burrows.

In Class II counties we searched
each area where tortoise populations
had been reported or where gopher

Figure 3.— Closed-circuit television camera with protective glass globe.

Figure 4.— Crew inserting closed-circuit television camera into gopher tortoise burrow.

tortoise habitat (sandy soil > 1 m) ex-
isted. Observations were made of the
total number of burrows, and total
number of active burrows. Since
these counties lay along the northern
border of the gopher tortoise's range
in Alabama, tortoise populations
were scattered and did not occur as
uniformly in specific habitat types as
those populations in Class I counties.
Therefore we did not sample here
but instead used a correction factor
similar to the one described by
Auffenberg and Franz (1982). The
correction factor (0.67 tortoises/ac-
tive burrow) was obtained from our
sampling of Class I counties by di-
viding the total number of tortoises
by the total number of active bur-
rows. The estimated total number of
tortoises for Class II counties was
very low (56), and did not signifi-
cantly affect our population estimate.

Landsat Satellite Imagery

Having measured tortoise density on
sample areas of the habitat types,
Landsat digital satellite imagery was
used to obtain an estimate of the area
of each habitat type in Class I coun-
ties. Characteristics and usage of this
remote sensing technique are de-
scribed by Anderson, Wentz and
Treadwell (1980), Brabander and
Barclay (1977), Diemer and Speake
(1983), Graham et al. (1981), Taranik
(1978a), and Taranik (1978b). The
system we used makes a scan of the
earth every eighteen days from a
geosynchronous orbit. The multis-
pectral scanner operates in seven dif-
ferent wavelengths of light — four vis-
ible and three infrared. We used near
infrared because it showed vegeta-
tion characteristics more clearly. By
making several passes, the scanner
senses light reflectance based on 0.1
ha pixels. Each 0.1 ha of the earth's
surface is assigned 1 of 256 gray val-
ues based on its reflectance. Using
these gray values we separated the
following habitat types based on
their spectral signature: unburned

201

pine/scrub oak, burned pine/ scrub
oak, planted pine, old-field, agricul-
tural fields, pasture and composite
edge.

Before sampling, we used ground-
truthing to determine if it was fea-
sible to attempt to classify each habi-
tat type using Landsat imagery. On
70-0.4 ha sample plots in Baldwin
County (10 plots in each habitat
type), each plot was correctly classi-
fied. Clearcuts were not included be-
cause they were a rapidly changing
transient stage (1-2 years) leading to
planted pine habitat, and as such
could not be identified on Landsat
images accurately due to their rapid
vegetational change. Habitat was
considered planted pine if pine was a
prominent understory or midstory
component (at least 0.3 m tall). Indi-
vidual edge types were combined
because edge transects had similar
vegetation characteristics and thus a
similar spectral signature. Combined
edge habitat was identifiable.

NASA software used with Land-
sat imagery includes a program for
referencing Landsat digital data to
any scale map. We referenced our
data to standard 1:24,000 topo-
graphic maps using known control
points. This enabled us to use Uni-
versal Trans Mercator coordinates to
locate each transect on the Landsat
image and obtain the correct gray
value for each transect. We then as-
signed a range of gray values to each
habitat type based on the reflectance
of the sample transects. The accuracy
of the habitat classifications was
checked throughout this process.

A polygon was then constructed
enclosing all the Class I counties, and
areas of each gray value within this
polygon were measured. From these
measurements we determined the
total area for each habitat type in
Class I counties.

Data Analysis

We had two concerns relative to data
analysis: (1) to derive a population

estimate based on mean tortoise den-
sity f>er hectare multiplied by the es-
timated area of the respective habitat
type, and (2) to identify and locate
gopher tortoise habitat.

In order to obtain a population
estimate we multiplied the mean
density of gopher tortoises per hec-
tare in a specific habitat type by the
total area of that habitat type in Class
I counties. An allowance was made
for standard error of the mean. The
habitat totals were then summed to
give a final population estimate of
the Class I counties.

In addition to these concerns we
examined age class structure. Lan-
ders et al. (1982) noted that gopher
tortoises pass through two general
life-history stages before they reach
sexual maturity. The juvenile stage
lasts until the carapace is approxi-
mately 100-120 mm. During the juve-
nile stage, the shells are very soft and
carapacial scutes usually have dis-
tinct yellow centers. This stage usu-

ally lasts until about 5 years of age.
Juvenile coloration fades and the
shells begin to harden during the
subadult stage which generally lasts
from 5 to 21 years of age. Carapace
lengths range from about 120-220
mm. At sexual maturity, body vol-
ume has drastically increased and
sexual dimorphism is apparent. This
occurs at approximately 21 years of
age and a carapace length of 230 mm.
Alford (1980) established a mathe-
matical relationship between the
widths of gopher tortoise burrows
and the carapace lengths of their oc-
cupants in northern Florida (this rela-
tionship has not been thoroughly
tested in other states). Using Alford's
equation log^^y = 0.879 log^^x + 0.149,
where y is carapace length and x is
burrow width, we used our burrow
width measurements of occupied
burrows to divide tortoise popula-
tions into juvenile, subadult, and
adult age classes. We considered age
class structure to be an important cri-

Figure 5.— Closed-circuit television monitor displaying picture of a gopher tortoise inside a
burrow.

202

^ .

Table 1 .—Summary of sample variables and derived estimates for Class I counties from 339-0>4 ha transects in south-
em Alabama, 1984-1985.

Standard Area Population
Habitat n Habitat totals Mean densitles/ha error*" (ha) estimate

Open Active Open Active

burrows burrows Tortoises burrows burrows Tortoises

Old-field

21

23

17

13

2.72

2.00

1.53

0.47

35,822

207,808 ±

63,836

Planted Pine

17

7

6

5

1.01

0.87

0.72

0.35

99,855

71,896 ±

34,949

Burned Pine/

Scrub Oak

34

36

13

9

2.62

0.94

0.64

0.27

209,108

133,829 ±

56,459

Edge

129

85

54

34

1.63

1,04

0.64

0.15

102,408

65,541 ±

15,361

Pasture

46

1

1

1

0.05

0.05

0.05

0.05

61,225

3,061

± 3,061

Agriculture

31

0

0

0

0.00

0.00

0.00

0,00

210,386

0

Unburned Pine/

Scrub Oak

10

0

0

0

0.00

0.00

0.00

0.00

133,004

0

Clearcuts

51

1

1

0

0.05

0.05

0.00

0.00

Totals

339

153

92

62

951,808

482,135 ±

173,666

"Standard error o f the tortoise mean dertsity/ha.

teria along with density in evaluating
tortoise population viability. Re-
search has not yet revealed an opti-
mum age class structure. Intuitively,
in a long-lived animal such as the
gopher tortoise, the age class struc-
ture of a healthy population would
be skewed toward the adult class.
The presence of juvenile and sexually
mature adult tortoises does definitely
indicate recent reproduction.

Results

Gopher Tortoise Densities and
Habitat Areas

Tortoise densities and habitat areas
were measured in Class I counties.
These results are summarized in
table 1, which includes sampling
variables by habitat type along with
estimates derived from sampling.

Age Class Structure

Five percent of the sampled popula-
tion (n=100 tortoises) were juvenile
tortoises, 48% were subadult, and
47% were adults. This structure

shows that there has been recent re-
production, and that there is a large
segment of breeding size adults pres-
ent. This suggests that the potential
for successful population mainte-
nance over the estimated 951,808 ha
area of tortoise habitat in Class I
counties is good.

Discussion

Using the referenced Landsat data
and knowing the range of gray val-
ues for each habitat type, we were
able to examine any area in Class I
counties and determine the size and
quantity of gopher tortoise habitat
units. Using a plotter, figures can be
made of all the 0.1 ha pixels that cor-
respond to a given habitat type and
then the figure can be overlaid on a
map. For our purposes we only
needed the area of each habitat type
in Class I counties.

This technique has two distinct
sources of error. First is the variation
of the gopher tortoise densities
within habitat types. These variations
are inherent in sampling biological
populations. In this study the vari-
ance was fairly low. Increased
sample size would likely lower this

J

error. The second source of error is
in estimating total areas of the habi-
tat types over a large region. Al-
though in our preliminary ground-
truthing, Landsat imagery correctly
classified all our habitat types (ex-
cluding clearcuts and individual
edge types), we suspect that when
this technique is applied to a large
diverse region some areas will be
misclassified. Ground-tru thing
should be done after the classifica-
tion to determine what percentage
has been misclassified, which would
allow the researcher to make allow-
ances for this error in final computa-
tions.

Conclusions

We found this technique to be useful
for measuring tortoise density and
for determining quantity and loca-
tion of tortoise habitat. The error in
this technique seems to b>e less than
that for techniques used for census-
ing most other animals. Although it
is difficult to estimate numbers of
animals over a large area, it is helpful
to be able to accurately measure den-
sity in small areas and then extrapo-
late this density on the basis of a

203

quantitative measurement of a desig-
nated area. This method should be
especially valuable for surveys of
animals that are habitat specific.

Literature Cited

Alford, Ross A. 1980. Population
structure of Gopherus polyphemus
in northern Florida. Journal of
Herpetology. 14:177-182.

Allen, Ross, and Wilfred. T. Neill.
1951. The gopher. Florida Wildlife.
5(7):10, 32.

Anderson, William H., W. Alan
Wentz, B. Dean Tread well. 1980. A
guide to remote sensing for wild-
life biologists. Pages 291-303. In S.
D. Schemnitz (ed.). Wildlife man-
agement techniques manual. The
Wildlife Society, Washington, D.C.
viii + 686 p.

Auffenberg, Walter and Richard
Franz. 1982. The status and distri-
bution of Gopherus polyphemus. p.
95-126. In R.B. Bury (ed) North
American tortoises: Conservation
and ecology. U.S. Fish and Wild-
life Service Research Rep. No. 12.

Bozeman, John. R. 1971. A sociologic
and geographic study of the sand
ridge vegetation in the Coastal
Plain of Georgia. Ph.D. disserta-
tion. University of North Carolina,
Chapel Hill. 244 p.

Brabander, Jerry J. and John S.
Barclay. 1977. A practical applica-
tion of satellite imagery to wildlife
habitat evaluation. Proceedings of
the Annual Conference of South-
eastern Fish and Wildlife Agen-
cies. 31:300-306.

Diemer, Joan E. 1984. Tortoise reloca-
tion in Florida: solution or prob-
lem? Proceedings of the 9th An-
nual Symposium of the Desert
Tortoise Council, p. 131-135.

Diemer, Joan E. 1986. The ecology
and management of the gopher
tortoise in the southeastern United
States. Herpetologica 42(1):125-
133.

Diemer, Joan E. and Dan W. Speake.
1981. The status of the eastern in-

digo snake in Georgia. In Ron R.
Odom and J. W. Guthrie (eds.).
Proceedings of the Nongame and
Endangered Wildlife Symposium.
Ga. Dept. Nat. Resources, Game &
Fish. Tech. Bull. WL5. p. 52-61.

Diemer, Joan E. and Dan W. Speake.
1983. The distribution of the east-
ern indigo snake (Drymarchon co-
rais couperi) in Georgia. Journal of
Herpetology 17:256-264.

Garner, James A. and J. Larry Lan-
ders. 1981. Foods and habitat of
the gopher tortoise in southwest-
em Georgia. Proceedings of the
Annual Conference of the South-
eastern Association of Fish and
Wildlife Agencies. 35:120-133.

Graham, Marcellus H., Ronnie W.
Pearson, Benjamin R. Seyfarth,
Bobby G. Junkin, Maria T. Kalcic.
1980. ELAS, earth resources labo-
ratory applications software. Re-
port No. 183 of the NASA, Na-
tional Space Technologies Labora-
tories, Earth Resources Labora-
tory.

Hubbard, Henry. G. 1894. The insect
guests of the Florida land tortoise.
Insect Life 6:302-315.

Hutt, Art. 1967. The gopher tortoise,
a versatile vegetarian. Florida
Wildlife. 21(7):20-24.

Landers, J. Larry. 1981. Techniques
for restocking gopher tortoise
populations. In The Gopher Tor-
toise: Distribution, Ecology, and
Effects of Forest Management. Fi-
nal Report to Georgia Department
of Natural Resources.

Landers, J. Larry and James A. Gar-
ner. 1981. Status and distribution
of the gopher tortoise in Georgia,
p. 45-51. In Ron R. Odom and J.
W. Guthrie (eds.). Proceedings of
the Nongame and Endangered
Wildlife Symposium. Ga. Dept.
Nat. Resources, Game & Fish
Tech. Bull. WL5.

Landers, J. Larry, James A. Garner,
and W. Alan McRae. 1980. Repro-
duction of gopher tortoises (Go-
pherus polyphemus) in southwest-
em Georgia. Herpetologica
36(4):353-361.

Landers, J. Larry and Dan W.

Sf)eake. 1980. Management needs
of sandhill reptiles in southern
Georgia. Proceedings of the An-
nual Conference of the Southeast-
em Association of Fish and Wild-
life Agencies. 34:515-529.

Mount, Robert H. 1975. The Reptiles
and Amphibians of Alabama. Ala-
bama Agricultural Experiment
Station, Auburn. 347 p.

Speake, Dan W. and James A. Alti-
ere. 1983. A device for viewing
and filming the contents of tree
cavities, p. 185-187. In Davis, J. W.,
Gregory, A. F. and R. A. Ocken-
tield (Tech. Coords.) 1983 Snag
Habitat Management: proceedings
of the symposium. USDA Forest
Service General Technical Report
RM-99. 226 pages. Rocky Moun-
tain Forest and Range Experiment
Station, Ft. Collins, CO.

Speake, Dan W., Joe A. McGlincy,
Thaggard A. Colvin. 1978. Ecology
and management of the eastern
indigo snake in Georgia: a prog-
ress report. Pages 64-73. In Ron R.
Odom and Larry Landers (eds.).
Proceedings of the Rare and En-
dangered Wildlife Symposium.
Ga. Dept. Nat. Resources, Game &
Fish Tech. Bull. WL4.

Taranik, James V. 1978a. Characteris-
tics of the Landsat Multispectral
Data System: U.S. Geological Sur-
vey Open-file Report 78-187, 76 p.,
Sioux Falls, South Dakota.

Taranik, James V. 1978b. Principles of
computer processing of Landsat
data for geologic applications: U.S.
Geological Survey Open-file Re-
port 78-117, 50 p., Sioux Falls,
South Dakota.

Wharton, Charles H. 1978. The natu-
ral environments of Georgia.
Georgia Dept. Nat. Res. 277 p.

Woodruff, Robert E. 1982. Arthro-
pods of gopher burrows, p. 24-48.
In R. Franz and R. J. Bryant (Eds.),
The Gopher Tortoise and Its
Sandhill Habitat. Proceedings of
the 3rd Annual Meeting, Gopher
Tortoise Council.

204

Evaluation and Review of
Field Techniques Used to
Study and Manage Gopher
Tortoises^

Russell L. Burke^ and James Cox^

Abstract. —This paper reviews methods used to
census gopher tortoises as well as techniques for
demographic, reproduction, and movement
studies. We also evaluate a refinement for line
transect estimates of gopher tortoise abundance. In
situations where dense vegetation structure may
hinder abilities to locate burrows along transects,
Fourier series estimators of abundance con be used
to overcome the problem, However, our results
indicate that many transects may be needed to
provide precise estimates of gopher tortoise
abundance over large areas. The collection of
vegetation data along transects may also be helpfu
in evaluating habitat preference in this species.

Introduction

Of the approximately 107 genera and
267 species of North American rep-
tiles, two species of tortoises have
received a relatively large amount of
scientific attention. Organizations
dedicated to the conservation and
protection of the gopher tortoise (Go-
pherus polyphemus) (The Gopher Tor-
toise Council) and the desert tortoise
(G. agassizi) (The Desert Tortoise
Council) attest to heightened levels
of amateur and scientific interest in
these species. Past bibliographies
(Diemer 1981, Douglass 1975,
Douglass 1977, Hohman et al. 1980)
together record over 775 different
publications concerning the genus,
and more have been published since
then. Compared to most other reptile
species, an exceptional diversity of
techniques has been employed, and
many field methods have been devel-
oped and used to study their status
and biology.

The gopher tortoise is a large ter-
restrial turtle (15-37 cm carapace
length, 3.6-5.0 kg) that exhibits low
rates of juvenile recruitment, extreme

'Paper presented at syrDposium, Man-
agement of Amphibiarts. Reptiles, and
Small Mammals in Northi America. (Flag-
staff. AZ, July 19-21 1988.)

'Researct) Associate. Tall Timbers Re-
search! Station. Route 1. Box 678. Tallahas-
see. Florida. 32312.

'Biologist. Nongame Wildlife Program.
Florida Game and Fresh Water Fish Com-
mission. 620 S. Meridian Street. Tallahassee.
Rorida. 32399-1600.

adult longevity, and persistent use of
a small number of burrows, often in
a loose aggregation of 10 to 15 indi-
viduals. As a result, tortoises display
a social system that involves indi-
viduals who may have interacted
regularly for decades (Douglass 1976,
Landers et al. 1980, McRae et al.
1980). Tortoises were once a common
feature of the upland habitats of the
southeastern coastal plain (Auffen-
berg and Franz 1982), but the species
is now less common and appears on
several state and federal lists of rare
or endangered species (Lohoefener
and Lohmeier 1984, Wood 1987). The
principal forces driving these popu-
lation declines are rapid urbaniza-
tion, certain forest management prac-
tices, and human predation (Diemer
1986).

Gopher tortoise burrows are im-
portant to a large wildlife commu-
nity, and 332 other species have been
documented to use tortoise burrows
at least occasionally (Jackson and
Milstrey in press). Included among
the several rare species that rely
heavily on tortoise burrows are the
Florida mouse (Podomys floridanus),
Florida and dusky crawfish frogs
(Ram areolata aesopus and R. areolata
sevosa), sand skink (Neoseps reynoldsi),
Florida pine snake (Pituophis melano-
leucus mugitus), and eastern indigo
snake (Drymarchon corais couperi).

In this paper we review tech-
niques used in field research on the
gopher tortoise community. We also
discuss future areas of research and

analyze the use of Fourier series esti-
mators (Burnham et al. 1980) in line
transect censusing techniques. In
doing so we suggest appropriate
methods for future work, standard-
ize some techniques, bring some
lesser known techniques to the fore,
and suggest refinements to com-
monly used methods.

Estinnating Population Size

Burrow Count Transects

Burrow-count transects are currently
the most widely used method for es-
timating the size of local gopher tor-
toise populations, though some tor-
toise populations do not dig burrows
(Auffenberg 1969), while others may
use seven or more burrows per indi-
vidual (McRae et al. 1980). Burrows
are particularly amenable to transect
analysis since they are stationary and
generally visible in many of the open
areas occupied by gopher tortoises.
Transects also require little equip-
ment, can be used to cover relatively
large areas in a short time, and can
be used to estimate abundance over a
large area using random or stratified-
random sampling procedures. A con-
version factor (Auffenberg and Franz
1982) is used to relate the number of
different tortoise burrows to the
number of gopher tortoises in an
area.

The dimensions of reported
transects ranges from 100 to 250 m in

205

length to 7 to 10 m in width (Auffen-
berg and Franz 1982, Cox et al. 1987,
Lohoefener and Lohmeier unpub.
rep.). Lohoefener (in press) points
out that strip transect burrow counts
assume that all burrows are detected
within a strip. Breininger et al. (in
press), however, expressed concern
that dense vegetation could make
strip-transect estimates unreliable
unless the transects were narrow.
The thick oak scrub (Quercus spp.)
vegetation common on many of their
study sites, for example, would have
prohibited surveyors from seeing
burrows more than a few meters
from transect lines.

A possible method of correcting
this problem (Cox et al. 1987,
Lohoefener in press) is to take {per-
pendicular distance measures from
transect lines to observed gopher tor-
toise burrows. Perpendicular dis-
tances can be used in Fourier series
density estimators (or other estima-
tors) (Bumham et al. 1981) to account
for differences in the detectability of
burrows due to vegetation or the size
of the burrow.

To look at this problem in more
depth, we compared strip transects
and line transects by establishing 12
transects (250 m by 20 m) in each of
three areas containing gopher tor-
toise populations. The areas selected
had noticeable differences in vegeta-
tive structure. The first site was a
mixed longleaf pine (Pinus palustris),
turkey oak (Quercus laevis) habitat on
a private ranch; the second site was
an early successional sand pine scrub
(P. clausa) forest on private timber
lands; and the third site was a ma-
ture longleaf pine forest in the
Apalachicola National Forest. The
starting points and directions of
transects within these areas were
randomly selected.

Perpendicular distances from bur-
rows to transect lines were measured
to the nearest 0.25 m, and only bur-
rows detected from the transect line
were recorded (i.e., burrows located
while measuring perpendicular dis-
tances to burrows seen from the

transect line were ignored). Burrow
densities for each of the three areas
were estimated directly using the
number recorded on transects and
Fourier series estimators obtained
from perpendicular distance data
(table 1)". Fourier series estimators
were calculated using the
TRANSECT program developed by
Laake et al. (1979) and are presented
in table 1 for the three sites.

Vegetation structure appeared to
influence the estimate of burrow den-
sity on the early successional site
(Site 2), but the Fourier series esti-
mate of density was no different than
the estimate provided by direct com-
putations on the other sites. The
early successional site had a very
thick shrub component that made it
difficult to locate burrows several
meters from the transect line. Ten
meters was probably too wide a
transect width in this particular set-
ting. The direct computation of bur-
row density from transect data on
Site 2 is only half the density estimate
developed by the Fourier series esti-
mate.

The level of variation observed
among transects (whether they be
strip or line transects) within a site
can be used to estimate the number
of additional transects needed to at-
tain a higher level of accuracy for the
estimate of density (Bumham et al.
1981). To increase the precision of
our estimates by 10%, for example,
an additional 24 transects would be
needed for Site 1, 40 for Site 2, and 78
for Site 3. Such an analysis can help
determine whether additional sur-
veys are needed, given the level of
accuracy desired. For some ques-
tions, levels of accuracy of 20-30%
may be acceptable.

Detecting small burrows of juve-
nile tortoises in transect sampling
can be particularly difficult even in
fairly open habitats (Douglass 1978).
This problem weakens the reliability
of transect data in estimating the
abundance of juveniles. Fourier se-
ries estimators again could be used,
in conjunction with an estimate of

burrow size, to gauge detectability of
small burrows, but extremely large
samples are probably needed to ob-
tain an accurate detectability func-
tion and estimate of abundance for
smaller tortoises.

Point-Center Burrow Counts

Tortoises often form small colonies of
aggregated burrows (McRae et al.
1980), and H. Mushinsky and E.
McCoy (Pers. comm.. University of
South Florida, Tampa, Florida) use a
point-center method (Cottam and
Curtis 1956) to estimate the size of
tortoise colonies. The approximate
center of the aggregation of burrows
is estimated, and the center point of
the census station is placed there.
The distance from the center point to
several tortoise burrows is deter-
mined, and a burrow density esti-
mate is derived using standard
point-center calculations (Cottam
and Curtis 1956). If the abundance of
tortoises over a large area is desired,
all aggregations should be located.

Ottier Indirect Estimates of Density

In some situations (e.g., intensive col-
ony analysis or preparation for popu-
lation relocation), complete burrow
counts are needed. We have used
teams of 6 to 12 inexf)erienced field
assistants, spaced at arm's length, to

Table 1 .—Mean burrow density esti-
mates (burrows per ha) and stan-
dard deviations calculated from
transect data using Fourier series
estimators (D) and direct computa-
tions. Data were collected at three
sites in north Florida.

Location

Fourier series

Direct

estimator <D)

computations

Site 1

5.3 + 0.957

5.5 + 0.932

Site 2 7.9 ±0.464 3.3 + 1.351
Site 3 3.8 + 0799 3.8 + 0.873

206

traverse an area and search inten-
sively for burrows. Later searches by
a more exf)erienced researcher did
not reveal any previously undiscov-
ered burrows, except for a few cryp-
tic hatchling burrows.

Trained dogs and aerial searches
by helicopter (Humphrey et al. 1986)
have also been used to locate gopher
tortoise burrows. Gopher tortoises
often defecate in or near their bur-
rows, and a motivated dog can de-
tect and locate the resulting olfactory
source. Scats and carcasses are also
important field sign used as indices
of desert tortoise populations (Berry
and Nicholson 1984, Woodman and
Berry 1984).

Regularly used burrows often
have several well-defined trails lead-
ing to foraging areas and other bur-
rows (Ernst and Barbour 1972). We
have used these trails to find bur-
rows hidden in extremely dense
vegetation.

Activity Patterns and Correction
Factors for Burrow Counts

Although estimates of gopher tor-
toise burrow abundance are rela-
tively easy to collect, calculating the

number of tortoises associated with
those burrows can be difficult. It
seems logical that the number of tor-
toise burrows would be positively
correlated with the number of go-
pher tortoises in an area, but the pre-
cise nature of this relationship is
poorly understood. Complicating
factors include the level of human
disturbance, soil type, and factors
that influence gopher tortoise activity
patterns (e.g., time of day, season,
and weather conditions).

Most researchers have used a cor-
rection factor of 0.614 times the num-
ber of "active" and "inactive" bur-
rows to estimate tortoises abundance
from burrow counts. This conversion
factor is based on information pre-
sented in Auffenberg and Franz
(1982) that was derived from long-
term data on the occupation rates of
122 burrows. Burrow activity was
defined by Auffenberg and Franz
(1982) in the following manner:

active (burrow) if the soil of
the burrow had been recently
disturbed by the tortoise, inac-
tive if the soil were undis-
turbed but the burrow ap-
peared to be maintained, and
old if the mouth had been

Table 2.-- Examples of reported correction factors.

Tortois©s/active TortoIses/InactlveTortolses/actlve+

burrow burrow inoctiv© Source

•11%"

49/103(48%)"

33/103(32%)"

67/124(54%)

43/44(98%)

4/19(21%)

35/174(20%)

«

'61.5%
9/19(47%)
7/10(70%)
•66,0%

0/30(0%)
3/16(19%)
0/25 (0%)
0/144(0%)

0/225 (0%)
0/47(0%)

n=122,614%

67/154(44%)
45/60 (75%)
4/44 (9%)
35/318(11%)
127/411 (31%)

»

9/244 (4%)
7/57 (12%)

10/89(11%)"*

Auffenberg and Franz (1982)
Breininger et al. (in press)

Burke (pers.obs.)
Doonan(1986)

Fucigna and Nickerson (in press)

Linley(1986)
Lohoefener(1982)
Speake (1983)

Spiders and Speake (1986)
Stout et al. (in press)

'Not reported or additional details not reported.
"Includes 'maybe active' activity classification.
'"Unknown number of tortoises hiad been harvested prior to survey.

washed in or covered with
debris (1982:96) (italics ours).

Little exp)erience is needed to learn
to make these distinctions, but differ-
ent investigators' classifications may
vary, increasing the imprecision of
tortoise abundance estimates. The
precision is also affected by the activ-
ity level of tortoises. During warm
periods tortoises may move among
several burrows during a day; dur-
ing cooler periods a tortoise may stay
in a burrow for several weeks.

R. Stratton (Pers. comm.) suggests
that it is possible to determine
whether a burrow is occupied (i.e.,
active) by the direction of foot tracks
on the burrow apron. Stratton was
able to identify correctly 14 of 15 oc-
cupied burrows using this technique,
but he incorrectly identified 19 unoc-
cupied burrows as being occupied.

I. J. Stout (Pers. connm.. University
of Central Florida, Orlando Florida)
has successfully used a "sewer
snake" to determine if a burrow is
occupied. When extended to the end
of the burrow, the sound of the end
of the wire tapping a tortoise shell is
distinctive. Other methods in-
clude "feeling" for tortoises using
long PVC pipes (Pers. comm., J. Di-
emer, Florida Game and Fresh Water
Fish Connmission Wildlife Research
Laboratory, Gainesville, Florida) and
listening for tortoises using either a
flexible garden hose (Pers. comm.,
D.B. Means, Coastal Plains Institute,
Tallahassee, Florida) or an electronic
"ear" to amplify breathing sounds
(Pers. comm., D. W. Speake, Ala-
bama Cooperative Research Unit,
Auburn, Alabama).

Several small twigs stuck verti-
cally into the soil at the burrow
mouth can also be used to determine
if a burrow is occupied (Hallinan
1923, Beiinger et al. in press). If prop-
erly spaced, one or more twigs will
be knocked over the next time a tor-
toise passes. Direction of travel can
be determined by uniquely marking
the top of each twig (or using a "Y"
shaped stick) and noting which di-

207

rection the twig falls. The twigs can
be resurveyed 1-3 days after place-
ment.

Some recent studies involving to-
tal colony capture (Doonan 1986,
Stout et al. in press, Fucigna and
Nickerson in press, Linley 1986, R.L.
Burke unpublished data), using a
miniature underground television
camera (Burke p>ers. obs., Breininger
et al. in press, Spillers and Speake
1986) or other techniques have pro-
vided reliable determinations of the
number of tortoises per burrow.
These studies (table 2) have reported
a wide variation in the appropriate
correction factor, from 4% of active
and inactive burrows (Speake 1983?)
to 75% (Doonan 1986).

Breininger et al. (in press) suggest
that an appropriate correction factor
must be determined on a case-by-
case basis. They recommended that
at least 20 active and inactive bur-
rows be surveyed by other methods
(e.g., by camera techniques, trapping,
or by stick placement at the mouth of
the burrow) to establish an accurate
correction factor for a site.

Capture Techniques

Gopher tortoises spend most of their
time in burrows (McRae et al. 1980),
which makes it difficult to observe or
capture animals above ground. It is
not known how much time gopher
tortoises spend in above ground ac-
tivities, but the congener desert tor-
toise is inactive for about 98% of its
life (Nagy and Medica 1986).

Once inhabited burrows are lo-
cated, tortoises may be captured and
counted directly by any of several
methods. The methods vary in terms
of time and resource expenditures
required and the degree to which
habitat conditions are disturbed.

Trapping

Many researchers use a version of
bucket trapping similar to that origi-

nally reported by Agassiz (1857).
This fairly non-disruptive technique
involves burying a smooth sided
plastic bucket (usually a five-gallon
size) immediately in front of the bur-
row, and covering the trap loosely
with a cloth or a sheet of heavy pa-
per. The trap is then disguised with a
thin layer of soil.

Drainage holes may be drilled in
the bottom and sides to prevent ac-
cumulation of rainwater, which can
drown a captured tortoise. However,
in extremely hydric soils, traps
should not have holes because water
entering from the ground can cause
the same problem.

In general, traplines should be
closed down during periods of heavy
rains. Traps should be checked at
least daily, and during very hot
weather there is a risk of overheating
and killing captured animals (Burke
1987, Taylor 1982). It may help to
shade exposed traps. Smaller cans
and containers may be used for cap-
turing juvenile and subadult tor-
toises.

Bucket trapping is labor intensive,
but once traps are in place they are
easy to monitor. Up to forty traps
may be installed by an experienced
person per day, and over 100 traps
can be checked and reset if necessary
per person per day. We found that
over 90% of bucket-trapped tortoises
were captured in the first 21 days,
suggesting that three to four weeks is
required to capture nearly all tor-
toises.

These results are very similar to
the results obtained by J. Diemer
(Pers. comm., Florida Game and
Fresh Water Fish Commission Wild-
life Research Laboratory, Gainesville,
Florida). An absence of signs of
above-ground activity after place-
ment of traps helps to indicate
whether all occupied burrows in the
area have been located and trapped.

Martin and Layne (1987) placed
standard live mammal traps at the
entrance of the burrow to capture
tortoises. Snares have also been used
by Novotny (1986) and ourselves

with some success. They may be set
so as to catch the leg of the tortoise
and therefore limit possible injury,
though Taylor (1982) describes the
use of snares to kill pest tortoises.
Although snares are inexpensive and
easy to set, they are easily evaded
and may occasionally injure a noosed
animal.

Auffenberg (in Plummer 1979) and
Recht (1981) described using me-
chanical and electronic burrow-ex-
cluding devices to force tortoises to
remain above ground after leaving
their burrows. Recht (1981) pointed
out that, if such a mechanism was
equipped with transmitting appara-
tus, the tortoise could be captured
immediately.

Deception

"Handbobbing'' (Burke 1987, Linley
1986) may entice tortoises to emerge
from burrows, apparently by eliciting
a territorial response. This technique
involves bobbing a clenched fist in
short, jerky motions at the mouth of
the burrow, which is similar to the
head bobbing that tortoise engage in
as part of social interactions (Auffen-
berg 1969). Once a territorial re-
sponse is initiated, tortoises will at-
tempt to push the intruding hand
from the burrow and can be maneu-
vered into a position to be extracted.
Success may be enhanced by striking
the ground several times before
handbobbing and by tossing a small
amount of soil down the burrow.
Mirrors can also elicit a territorial
response (Legler and Webb 1961).

A somewhat similar technique,
''tapping," has been used to capture
desert tortoises (Medica et al. 1986).
Tapping involves lightly rapping on
the tortoise's shell with a long stick.
This procedure would be difficult to
employ successfully where burrows
are long and curved. We have used
sewer snakes to probe for tortoises at
the end of their burrows, but we
have not elicited a response by shell
tapping.

208

Burrow Excavation and Pulling

Digging up the entire burrow with a
backhoe or hand shovel is both time
consuming and destructive. At one
South Florida site, it took an experi-
enced backhoe operator 2.5 hours to
excavate one burrow that was over
11m long and 6 m deep. Most bur-
rows are excavated in less than 45
minutes using a backhoe, which com-
pares favorably to the approximately
30 days of bucket trapping required
to remove all tortoises from an area
(Diemer et al. in press).

When excavating a burrow, a
sewer snake or garden hose should
be extended to the end of the burrow
to keep track of the tunnel path. The
entire process is complicated by
loose, sandy soils at some sites, and
it is difficult to retain burrow struc-
ture and avoid potentially dangerous
cave-ins. The difficulty of the process
may be reduced by using an elec-
tronic device to locate the burrow
end before digging (see Wolcott
1981). Small commensal species are
likely to be buried when a burrow is
excavated mechanically, but excava-
tion by hand is extremely labor-in-
tensive (Ernst and Barbour 1972).

Taylor (1982) describes the history
of a pulling "hook" first reported by
Fisher (1917). It is the only simple,
quick, and moderately reliable
method for capturing tortoises, used
principally by tortoise hunters. Pull-
ing requires the use of a long flexible
rod attached to a short stout piece of
bent wire. The apparatus is fed into
the burrow, maneuvered behind the
tortoise, and wedged between the
rear of the plastron and the flared
carapace. Success rate is influenced
by a puller's skill and by the length
and curvature of the burrow. In re-
gions that have been heavily
"pulled" in the past, remaining tor-
toises are most often found in wind-
ing burrows that are particularly dif-
ficult to pull (R. Stratton, Pers.
comm.). Taylor (1982) gives details
on the procedure, as well as statistics
on the damage to captured tortoises.

Techniques for Studying Tortoise
Demography and Reproduction

Estimates of Population Structure
Using Burrow Width

Alford (1980) and Martin and Layne
(1987) have demonstrated that a
simple mathematical relationship ex-
ists between the width of a burrow
and the size of the resident tortoise.
Thus, on the basis of a burrow cen-
sus, burrow widths, and a reliable
correction factor, it is possible to esti-
mate population size and evaluate
demographic structure (Alford 1980,
Sauer and Slade 1987). The relation-
ship between burrow width and size
of occupant may be slightly biased,
however, since small tortoises can
occupy large burrows but the ob-
verse is impossible.

Marking Techniques and
Determining Sex and Age

Marking tortoise shells is an easy
way to follow the fate of individuals
over long periods of time. Tech-
niques for marking marginal scutes
of turtles have been reviewed by
Femer (1979) and Plummer (1979).

Based on variation in the shell di-
mensions of 183 adult tortoises of
known sex, McRae et al. (1981) devel-
oped a discriminate equation that
can be used to determine accurately
the sex of adult tortoises from north
Florida and south Georgia. The ap-
plicability of the technique to tor-
toises from other areas, and to
smaller size classes, is untested
(Wester 1986).

Graham (1979) reviews four age-
determination techniques: mark/re-
capture, records of captive speci-
mens, exannination of long bone sec-
tions, and scute ring counts. Of these,
only scute ring counts have been re-
ported for gopher tortoises. W.
Auffenberg (Pers. comm., Florida
State Museum, Gainesville, Rorida)
suggested that a p>encil rubbing of
the plastron was an accurate way

both to record true scute rings and to
avoid counting false rings. This has
been confirmed by L. Landers (un-
pub. data. Tall Timbers Research Sta-
tion, Tallahassee, Florida). Addi-
tional methods of counting and re-
cording scute rings are given by Gal-
braith and Brooks (1987).

Landers et al. (1982) demonstrated
that, in southern Georgia, age can be
accurately estimated by carefully
counting plastron scute rings. Ger-
mano and Fritts (in press) used
mark/ recapture data to show a high
correlation between age and scute
ring counts of 17 known-age desert
tortoises (less than 25 years old) from
Nevada. They propose microscopic
examination of thin scute sections
can help determine age of older tor-
toises. However, Berry (in press)
presents data from 190 desert tor-
toises from 11 study sites in which
scute rings were not annual. Ring
deposition varied from 0 to 3 rings
per year. Berry and Woodman (1984)
discuss the use of shell wear classes
for age determination of adult desert
tortoises.

Studies of Tortoise Reproduction

Indirect indications of reproductive
activity include swelling of the sub-
dentary glands and recent evidence
of gravidity. Auffenberg (1966) and
Rose (1970) suggested that the sub-
dentary glands produce pheromones
important to courtship and mating
behavior, and Landers et al. (1980)
used the swollen condition of these
glands in some captured tortoises as
an index to sexual activity.

Although the clutch size of gravid
tortoises can be determined by radi-
ography (Turner et al. 1986), field
methods are limited to palpation and
weight loss. T. Linley (Pers. comm.)
uses palpation to estimate clutch
sizes for gravid females with well
calcified eggs. Turner et al. (1986)
also regularly weighed transmittered
desert tortoises and used sudden
weight loss to indicate oviposition.

209

Given the fairly predictable nature
of tortoise nest location (Hallinan
1923), it is surprising that so few field
data have been collected on nest pre-
dation, nest microclimate, sex of off-
spring, time of emergence, etc.
Auffenberg and Iverson (1979) in
north Florida, and Landers et al.
(1980) in south Georgia, provide esti-
mates of predation rates and nest
viability, but more information is
needed to construct accurate esti-
mates of nesting success over time,
one of the more critical portions of
tortoise life cycles (Diemer 1984).
Marshall (1987) and Douglass and
Winegamer (1977) also report pre-
liminary studies on nest predation
using sign at a small number of regu-
larly visited nests.

Camera traps may be particularly
useful in egg predation studies, al-
lowing precise identification of tim-
ing and predator. R.L. Burke and M.
Noss (pers. obs.) attempted to detect
soil disturbance due to egg laying by
burying a layer of colored gravel in
46 burrow mounds before oviposi-
tion season. No activity was de-
tected, however. Careful use of an
egg probe (Hallinan 1923) may facili-
tate rapid searching of large numbers
of burrow mounds for egg clutches.

Movement Studies

In addition to studies employing di-
rect observation and capture-recap-
ture techniques (e.g., Auffenberg and
Iverson 1979, Douglass and Layne
1978, McRae et al. 1980, Landers et
al. 1980), various remote sensing de-
vices have been used to monitor tor-
toise movements.

String trailers (see Ferner 1979 and
Plummer 1979) have been used for
daily movement and path length
studies (Pers. comm., W. Auffenberg,
Florida State Museum, Gainesville,
Fl., McRae et al. 1980). Tortoises too
small for radio transmitters may be
tracked using a metal detector to lo-
cate small pieces of different metals
attached to their shells.

Radio telemetry (Legler 1979) of
gopher tortoises has been used by
Burke (1987), Fucigna and Nickerson
(in press), McRae et al. (1980), Stout
et al. (in press), J. Diemer (unpub-
lished data, Florida Game and Fresh
Water Fish Connmission Wildlife Re-
search Laboratory, Gainesville, Flor-
ida) and others. Radios are attached
to anterior of the carapace on females
(to avoid interference with copula-
tion) and either the anterior or poste-
rior of males. Dental acrylic is typi-
cally used to fix the transmitter on
the shell, and the entire device is cov-
ered in silicone sealant for additional
protection. Other researchers (e.g..
Stout et al. in press) have used ma-
chine screws or wire to attach the ra-
dio to the shell. Antennae are usually
glued along the shell or left dragging.

Auffenberg and Iverson (1979)
used a series of microswitches and
sensors buried along, and extending
into, numerous tortoise burrows to
correlate inner-burrow movements
with microhabitat environmental
conditions.

Commensal Studies

General methods for trapping reptile
and amphibian species are reviewed
by Campbell and Christman (1982)
and Vogt and Hine (1982). Crawfish
frogs may be seen at night sitting in
the mouth of the burrow (Hallinan
1923), and are sometimes captured in
bucket traps, small mammal traps,
and funnel traps set for other species
(Franz 1986). General marking tech-
niques for reptiles and amphibians
are reviewed by Ferner (1979).

Day et al. (1980) give a general re-
view of capture and marking tech-
niques for mammals, birds and rep-
tiles, and Mengak and Guynn (1987)
compare different trapping methods
for small mammals and herpe-
tofauna. Eisenberg (1983) describes
successful placement of traps for
Florida mice. As described above,
digging up the burrow by hand is the
only known way reliably to capture

all burrow commensals, especially
invertebrates. W. Auffenberg (Pers.
comm., Florida State Museum,
Gainesville, Rorida) and Milstrey
(1986) have used vacuum systems to
sample invertebrates in burrows.
Milstrey (1986) and Woodruff and
Klein (in prep.) also describe various
small, baited pitfall traps for captur-
ing invertebrates. Butler et al. (1984)
describes a C02 trap that is useful
for collecting ticks and fleas.

Vegetation Analysis

A small number of researchers has
attempted to characterize gopher tor-
toise habitat using quantitative meth-
ods. Breininger et al. (in press),
Marshall (1987), and Wester (1986)
related gopher tortoise densities to
vegetation structure, while Auffen-
berg and Iverson (1979) analyzed the
relationship between tortoise densi-
ties and a single vegetative compo-
nent, herbaceous ground cover.
C^antitative vegetation sampling has
become a standard element in survey
techniques used for other groups
(e.g., breeding bird censuses, James
and Shugart 1970), and these tech-
niques should be more widely ap-
plied to tortoise research.

We collected vegetation data at 50
m points as part of the transect study
described above. Percent canopy
cover (trees > 5 m), percent shrub
cover, percent ground cover, percent
wiregrass (Aristida stricta) cover, and
the relative percent of deciduous
trees to coniferous trees were meas-
ured using methods described in Cox
et al. (1987). These five variables
were selected based on published
information about gopher tortoise
habitat preferences (Campbell and
Christman 1982, Diemer 1986), but
several other variables could also be
considered.

A principal components analysis
was performed on the vegetation
data using a "varimax" rotation pro-
cedure (Wilkinson 1980). The density
(per ha) of active and inactive gopher

210

tortoise burrows along each of the 32
transect segments was then plotted
against the transect's vegetation
score on the first principal compo-
nent axis. This procedure helps
gauge the degree to which variation
in tortoise density along transects

Table 3.— Factor loadings for 6
habitat variables measured along
transects. Weightings and contrasts
were derived from a "varlmax"
principal component (PC) analysis
(Wilkinson 1983).

Variable

PC 1

PC 2

Canopy cover

0.809

-0.278

Shrub cover

-0.896

0.171

Ground cover

-0.832

0.044

Deciduous/conifer-

ous overstory

0.090

0.900

Percent wiregrass

0.607

0.550

Percent variance

explained by axis

50.5%

24.4%

relates to variation in vegetation
structure. The average values for
vegetative samples recorded along
transects was used to compute prin-
cipal component scores. Too few
samples were collected to produce a
very precise evaluation between bur-
row density and vegetation struc-
ture, so the effort should be consid-
ered only as an example of the appli-
cation of vegetation data collected
along transects.

Principal component analysis of
vegetation data accurately projected
the differences we casually observed
among sites. The first principal com-
ponent axis explained 50.5% of the
variation among samples and largely
contrasted decreasing canopy cover
and wiregrass percentages with in-
creasing shrub and ground cover
(table 3). High positive scores along
this axis indicate decreasing {percent-
ages of canopy cover and wiregrass,
increasing amounts of shrub cover

<
X

LU

a.

00

o
o:

D

m

o
>

00
Z
LU
Q

15.0"

10.0"

5.0

0.0

0 (§)

0

CD

0

0© Op. 0

0 0 0 0

0

-1.6

-0.2 1.2
PC 1

Figure 1 .—Gopher tortoise burrow density estirrvates plotted along first principal connponent
axis. High positive scores along PCI have low canopy cover and relatively high levels of
herbaceous ground cover and shrubs.

and ground cover, and increasing
ratios of deciduous to coniferous
trees. The second principal compo-
nent axis explained an additional
24.4% of the sample variance and is
weighted by decreasing amounts of
wiregrass cover and the ratio of de-
ciduous to coniferous trees (table 3).

A plot of burrow densities against
the first principal component shows
a general trend of increasing burrow
density with decreasing principal
comp)onent value (fig. 1). Areas with
greater burrow densities generally
had a lower percentage of canopy
cover, but higher percentages of
shrub and ground cover, than areas
with lower densities. The regression
line drawn through the points has an
adjusted r^ of 0.37 (p<0.05).

Future Directions

Burrow-count transects are efficient
for estimating burrow density, but
they may not produce sufficiently
accurate estimates of gopher tortoise
densities. The relationship between
burrow density and tortoise density
is poorly understood, and studies
analyzing the relationship between
burrow occupancy and burrow activ-
ity class are needed to strengthen
abundance estimates. Whether
transects are appropriate will depend
on the questions being addressed.

The combined effects of variation
in occupancy rates and variation in
burrow counts among transects may
easily produce estimates of tortoise
abundance that span an order of
magnitude. For example, a 95-confi-
dence interval for the density of ac-
tive and inactive burrows on our sec-
ond study area (using the Fourier se-
ries estimate from table 1) is 3.326-
12.55 burrows per ha. If the occu-
pancy rate of 20 active and inactive
burrows was followed for a week on
this site and determined to be 0.60
+0.20 for any one day, then a 95-con-
fidence interval for the estimated
density of tortoises on the site could
range from 0.69 to 12.4 tortoise per

211

ha. Clearly this is too large a range
for some, if not most, ecological
questions. Many more transects and
more precise occupancy rates would
be needed to correct these problems.

Fourier series estimators should
be used when transects are con-
ducted in areas with a dense shrub
component. Some strip-transect esti-
mates of gopher tortoise densities in
thick, scrubby areas may have under-
estimated density. Indeed, Breininger
et al. (in press) found high tortoise
densities on areas with thick shrub
levels that traditionally might not
have been considered appropriate
gopher tortoise habitat.

Repeated samples of burrow activ-
ity over time should be used to esti-
mate site-specific correction factors,
rather than rely on a single general-
ized correction factor. This can be
easily done, requiring only a return
visit to 20 or more randomly chosen
burrows. As such data accumulate,
they may lead to a more appropriate
correction factor.

Additional studies of the commen-
sal community are also needed since
very little is known of the interac-
tions that occur among commensal
species. Certain mutualistic relation-
ships may be critical to the survival
of many of these species and be im-
portant in efforts to relocate compo-
nents of the burrow community (e.g.,
Diemer et al. in press). Video camera
techniques (Breininger et al. in press,
Spillers and Speake 1986) offer a
great potential for investigating bur-
row ecology.

Additional studies of the early life
cycles of gopher tortoises may also
be worth pursuing, particularly in
terms of conducting management for
this species. The critical survival pe-
riod in the gopher tortoise life cycle
occurs during the first few years of
life (Diemer 1984). If nesting success
and hatchling survival can be effec-
tively manipulated through manage-
ment activities, such activities would
need to be conducted fairly infre-
quently to enhance population size
over many years.

Acknowledgments

The authors appreciate the sugges-
tions of W. Auffenberg, D. Breinin-
ger, R. Franz, L. Landers, J. Layne, H.
Mushinsky, I. J. Stout, an anonymous
reviewer, and especially K. Berry and
J. Diemer. D. Bentzein, J. Dudley,
P.K. Harpel-Burke, T. Linley, M.
Noss and R. Stratton provided vital
field assistance and insightful com-
ments. J. Layne, B. Woodruff and D.
Wood provided in press manu-
scripts.

Literature Cited

Agassiz, Louis. 1857. Contributions
to the natural history of the United
States of America. 452 p. Little,
Brown, and Co.

Alford, Ross A. 1980. Population
structure of Gopherus polyphemus
in northern Florida. Journal of
Herpetology 14(2):177-182.

Auffenberg, Walter. 1966. On the
courtship of Gopherus polyphemus.
Herpetologica 22(2):113-117.

Auffenberg, Walter. 1969. Tortoise
behavior and survival. 38 p. Rand
McNally, Chicago.

Auffenberg, Walter, and Richard
Franz. 1982. The status and distri-
bution of the gopher tortoise (Go-
pherus polyphemus). p. 95-126. In: R.
B. Bury (ed.). North American tor-
toises: Conservation and ecology.
U.S. Fish and Wildlife Service,
Wildlife Research Report 12,
126 p. Washington, D.C.

Auffenberg, Walter, and John B. Iver-
son. 1979. Demography of terres-
trial turtles, p. 541-569. In: M. Har-
less and H. Morlock (eds.).
Turtles: Research and perspec-
tives. 718 p. John Wiley and Sons,
New York.

Berry, Kristin, in press. Using growth
rings to age wild juvenile desert
tortoises (Gopherus agassizii). Pro-
ceedings of Desert Tortoise Coun-
cil 1987 Symposium. 5319 Cerritos
Avenue, Long Beach, California
90805.

Berry, Kristin H., and Lori L. Nichol-
son. 1984. The distribution and
density of desert tortoise popula-
tions in California in the 1970's. p.
26-60. In: K. H. Berry, (ed.). The
status of the desert tortoise (Go-
pherus agassizii) in the United
States. Desert Tortoise Council re-
port to the U.S. Fish and Wildlife
Service, Sacramento, CA. 502 p.+
appendixes. Order No. 11310-
0083-81.

Berry, Kristin, and A. Peter

Woodman. 1984. Preliminary in-
vestigations of shell wear in deter-
mining adult age groups in desert
tortoises, p. A4/1-A4/54. In: K. H.
Berry, (ed.). The status of the des-
ert tortoise (Gopherus agassizii) in
the United States. Desert Tortoise
Council report to the U.S. Fish and
Wildlife Service, Sacramento, CA.
502 p.+ appendixes. Order No.
11310-0083-81.

Breininger, D., A. Schmalzer, D. Ry-
dene, and R. Hinkle. 1986. Burrow
and habitat study of the gopher
tortoise in scrub and flatwood
habitat types. Final Project Report,
Project No. GFC-84-016, 313 p.
Nongame Wildlife Program, Flor-
ida Game and Fresh Water Fish
Commission, Tallahassee, Florida.

Burke, Russell L. 1987. An experi-
mental relocation and reintroduc-
tion of a gopher tortoise popula-
tion. M.S. thesis. University of
Rorida, Gainesville, FL. 41 p.

Burnham, K.A., D.R. Anderson, and
J.L. Laake. 1980. Estimations of
density from line-transect sam-
pling of biological communities.
Wildlife Monographs 72:1-202.

Butler, J.F., K.H. Holscher, O.
Adeyeye, and E.P. J. Gibbs. 1984.
Sampling techniques for burrow
dwelling ticks in reference to po-
tential African swine fever virus
vectors, p. 1065-1074. In: D.A.
Griffiths and C.E. Bowman (eds.).
Acarology IV, Volume 2. Ellis
Horwood Ltd., Chichester, Eng-
land. 1345 p.

Campbell, Howard W., and Steven P.
Christman. 1982. Field techniques

212

for herpetofauna community
analysis, p. 193-200. In: N.J. Scott
(ed.). Herpetological Communi-
ties. U.S. Fish and Wildlife Service.
Wildlife Research Report 13.
Washington, D.C. 239 p.

Cottam, G., and J.T. Curtis. 1956. The
use of distance measures in
phytosociological sampling. Ecol-
ogy. 37:451-460.

Cox, James. In press. Demographic
characteristics of small gopher tor-
toise populations and their pos-
sible influence on relocation
efforts. In: J.E. Diemer, D.R.
Jackson, J.L. Landers, J.N. Layne
and D.A. Wood. Proceedings of
the gopher tortoise relocation
symposium. Florida Game and
Fresh Water Fish Commission,
Nongame Wildlife Program Tech-
nical Report. Tallahassee, FL.

Cox, James, Douglas Inkley, and
Randy Kautz. 1987. Ecology and
habitat protection needs of gopher
tortoise (Gopherus polyphemus)
populations found on lands slated
for large-scale development in
Florida. Florida Game and Fresh
Water Fish Commission,
Nongame Wildlife Program Tech-
nical Report No. 4. Tallahassee,
FL. 75 p.

Day, Gerald I., Sanford D. Schem-
nitz. and Richard D. Taber. 1980.
Capturing and marking wild ani-
mals, p. 61-88. In: S.D. Schemnitz
(ed.). Wildlife management tech-
niques manual. 686 p. The Wildlife
Society. Washington, D.C.

Diemer, Joan E. 1981. Addendum to
the bibliography of the North
American land tortoises: Gopherus
polyphemus. p. 75-82. In: L. Lo-
hmeier and G. Johnston (eds.) The
future of gopher tortoise habitats.
Proceedings of the 2nd Annual
Meeting of the Gopher Tortoise
Council. Florida State Museum,
Gainesville, Florida, 32611.

Diemer, Joan E. 1986. The ecology
and management of the gopher
tortoise in the southeastern United
States. Herpetologica 42(1): 125-
133.

Diemer, Joan, Dale Jackson, J. Law-
rence Landers, James Layne, and
Don Wood. In press. Proceedings
of the gopher tortoise relocation
symposium. Florida Game and
Fresh Water Fish Commission,
Nongame Wildlife Program Tech-
nical Report. Tallahassee, FL.

Doonan, T.J. 1986. A demographic
study of an isolated population of
the gopher tortoise, Gopherus pol-
yphemus, and an assessment of a
relocation procedure for tortoises.
M.S. thesis. University of Central
Florida, Orlando, FL. 93 p.

Douglass, John F. 1975. Bibliography
of the North American land tor-
toises (genus Gopherus). U.S. De-
partment of the Interior, Special
Science Report — ^Wildlife No. 190,
60 p. Washington, D.C.

Douglass, John F. 1976. The mating
system of the gopher tortoise, Go-
pherus polyphemus, in southern
Rorida. M. A. thesis. University of
South Florida, Tampa, FL. 79 p.

Douglass, John F. 1977. Supplement
to the bibliography of the North
American land tortoises (genus
Gopherus). Smithsonian Herpeto-
logical Information Service, No.
39. 18 p.

Douglass, John F. 1978. Refugia of
juvenile gopher tortoises, Gopherus
polyphemus (Reptilia, Testudides,
Testudinidae). Journal of Herpe-
tology 12(3):413-415.

Douglass, John F., and James N.
Layne. 1978. Activity and ther-
moregulation of the gopher
tortoise (Gopherus polyphemus) in
southern Florida. Herpetologica
34(4):359-374.

Douglass, John F., and Chester E.
Winegamer. 1977. Predators of
eggs and young of the gopher
tortoise, Gopherus polyphemus
(Reptilia, Testudines, Tes-
tudinidae) in southern Florida.
Journal of Herpetology 11(2):236-
238.

Eisenberg, John F. 1983. Conserva-
tion address: The gopher tortoise
as a keystone species, p. 1-4. In:
R.J. Bryant and R. Franz (eds.).

The gopher tortoise: A keystone
species. Proceedings of the 4th an-
nual meeting of the Gopher Tor-
toise Council. 47 p. Florida State
Museum, Gainesville, Florida,
32611.

Ernst, Carl H., and Roger W. Bar-
bour. 1972. Turtles of the United
States. The University Press of
Kentucky. Lexington, Kentucky.
347 p.

Femer, John W. 1979. A review of
marking techniques for reptiles
and amphibians. Society for the
study of amphibians and reptiles.
Herpetological circular No. 9. De-
partment of Zoology, Miami Uni-
versity, Oxford, OH 45056.

Fisher, G.C. 1917. "Gopher pulling''
in Florida. American Museum
Journal. 17:291-293.

Franz, Richard. 1986. The Florida go-
pher frog and the Florida pine
snake as burrow associates of the
gopher tortoise in northern Ror-
ida. p. 16-20. In: D.R. Jackson and
R. J. Bryant (eds.). The gopher tor-
toise and its community. Proceed-
ings of the 5th Annual Meeting of
the Gopher Tortoise Council. 93 p.
Rorida State Museum,
Gainesville, FL 32611

Fucigna, Thomas F. Jr., and David K.
Nickerson, Jr. In press. Relocations
of two groups of gopher tortoises
(Gopherus polyphemus) from Palm
Beach County to Martin County,
Rorida. In: J.E. Diemer, D.R.
Jackson, J.L. Landers, J.N. Layne
and D.A. Wood. Proceedings of
the gopher tortoise relocation
symposium. Rorida Game and
Fresh Water Fish Commission,
Nongame Wildlife Program Tech-
nical Report. Tallahassee, FL.

Galbraith, David A., and Ronald J.
Brooks. 1987. Photographs and
dental casts as permanent records
for age estimates and growth
studies of turtles. Herpetological
Review 18(4):69-71.

Germano, David J., and Fritts. in
press. Methods for age determina-
tion of the desert tortoise, Go-
pherus agassizii. Proceedings of the

213

Desert Tortoise Council 1987 Sym-
posium. 5319 Cerritos Avenue,
Long Beach, California 90805.

Graham, Terry E. 1979. Life History
Techniques, p. 73-95. In: M. Har-
less and H. Morlock (eds.).
Turtles: Research and perspec-
tives. 718 p. John Wiley and Sons,
New York.

Hallinan, Thomas. 1923. Observa-
tions made in Duval County,
northern Florida, on the gopher
tortoise (Gopherus polyphemus).
Copeia 1923 (115):ll-20.

Hohman, Judy P., R.D. Ohmart, and
James Schwartzmann. 1980. An
annotated bibliography of the des-
ert tortoise (Gopherus agassizi).
Desert Tortoise Council Sp)ecial
Publication No. 1. 121 p.

Humphrey, Stephan R., John F. Eis-
enberg, and Richard Franz. 1985.
Possibilities for restoring wildlife
of a longleaf pine savanna in an
abandoned citrus grove. Wildlife
Society Bulletin. 13:487-496.

Jackson, Dale R., and Eric R. Mil-
strey. In press. The fauna of go-
pher tortoise burrows. In: J.E. Di-
emer, D.R. Jackson, J.L. Landers,
J.N. Layne and D.A. Wood. Pro-
ceedings of the gopher tortoise re-
location symposium. Florida
Game and Fresh Water Fish Com-
mission, Nongame Wildlife Pro-
gram Technical Report. Tallahas-
see, FL.

James, Frances, and Henry H. Shug-
art. 1970. A quantitative method
of habitat description. Audubon
Field Notes 24:727-736.

Laake, Jeffrey L., Kenneth P.
Burnham, and David R. Ander-
son. 1979. User's manual for pro-
gram TRANSECT. Utah State Uni-
versity Press. Logan, UT.

Landers, J. Larry, James A. Garner,
and W. Alan McRae. 1980. Repro-
duction of gopher tortoises (Go-
pherus polyphemus) in southwest-
ern Georgia. Herpetologica
36(4):353-361.

Landers, J. Larry, W. Alan McRae,
and James A. Garner. 1982.
Growth and maturity of the go-

pher tortoise in southwestern
Georgia. Bulletin of the Florida
State Museum, Biological Sciences.
27(2):81-110.

Legler, John M., and Robert G. Webb.
1961. Remarks on a collection of
Bolsbn tortoises, Gopherus fla-
vomarginatus. Herpetologica
17(l):26-37.

Legler, Warren K. 1979. Telemetry, p.
61-72. In: M. Harless and H. Mor-
lock (eds.). Turtles: Research and
perspectives. 718 p. John Wiley
and Sons, New York.

Linley, Terry Reed. 1986. Tortoise
density, age/ size class distribution
and reproductive parameters in a
central Horida population of Go-
pherus polyphemus. p. 21-32. In:
D.R. Jackson and R. J. Bryant
(eds.). The gopher tortoise and its
community. Proceedings of the 5th
Annual Meeting of the Gopher
Tortoise Council. 93 p. Florida
State Museum, Gainesville, FL
32611.

Lohoefener, Ren. In press. Line
transect estimation of gopher tor-
toise burrow density using a
fourier series. Proceedings of the
7th Annual Meeting of the Gopher
Tortoise Council. Horida State
Museum, Gainesville, FL 32611.

Lohoefener, Ren. 1982. Gopher tor-
toise ecology and land-use prac-
tices in southern Desoto National
Forest, Harrisn County, Missis-
sippi, p. 50-74. In: Franz, R. and R.
Bryant (eds.). The gopher tortoise
and its sandhill habitat. Proceed-
ings of the 3rd Annual Meeting of
the Gopher Tortoise Council. 78 p.
Rorida State Museum, Gaines-
ville, FL., 32611.

Lohoefener, Ren, and Lynne Lo-
hmeier. Unpublished report. 1984.
The status of Gopherus polyphemus
(Testudines, Testudinidae) west of
the Tombigbee and Mobile Rivers.
Report to the U.S. Fish and Wild-
life Service to support listing this
population under the Endangered
Species Act of 1973:1-116.

Marshall, John E. 1987. The effects of
nest predation on hatchling go-

pher tortoises (Gopherus pol-
yphemus Daudin 1802). M.S. thesis.
University of South Alabama. Bir-
mingham, AL. 67 p.

Martin, Paige L., and James A.
Layne. 1987. Relationship of go-
pher tortoise body size to burrow
size in a south-central Rorida
population. Florida Scientist
50(4):264-267.

McRae, W. Alan, J. Larry Landers,
and James A. Gamer. 1980. Move-
ment patterns and home range of
the gopher tortoise. American
Midland Naturalist 106(1):165-179.

McRae, W. Alan, J. Larry Landers,
and James A. Gamer. 1981. Sexual
dimorphism in the gopher tortoise
(Gopherus polyphemus). Herpetol-
ogica 37:46-52.

Medica, Philip A., C.L. Lyons, and
Frederick B. Tumer. 1986. Tap-
ping: A technique for capturing
tortoises. Herpetological Review
17(1):15-16.

Mengak, Michael T., and David C.
Guynn, Jr. 1987. Pitfalls and snap-
traps for sampling small mammals
and herptofauna. The American
Midland Naturalist 118(2):284-288.

Milstrey, Eric G. 1986. Ticks and in-
vertebrates in gopher tortoise bur-
rows: Implications and impor-
tance, p. 4-15. In: D.R. Jackson and
R. J. Bryant (eds.). The gopher tor-
toise and its community. Proceed-
ings of the 5th Annual Meeting of
the Gopher Tortoise Council. 93 p.
Rorida State Museum,
Gainesville, FL. 32611.

Nagy, Kenneth A., and Philip A.
Medica. 1986. Physiological ecol-
ogy of desert tortoises in southern
Nevada. Herpetologica 42(1 ):73-
92.

Novotny, Raymond. 1986. The go-
pher tortoise at Momingside Na-
ture Center: A public education
project, p. 33-37. In: D.R. Jackson
and R. J. Bryant (eds.). The gopher
tortoise and its community. Pro-
ceedings of the 5th Annual Meet-
ing of the Gopher Tortoise Coun-
cil. 93 p. Florida State Museum,
Gainesville, FL. 32611.

214

Plummer, Michael V. 1979. Collect-
ing and Marking, p. 45-60. In: M.
Harless and H. Morlock (eds.).
Turtles: Research and perspec-
tives. 718 p. John Wiley and Sons,
New York.

Recht, Michael A. 1981. A burrow-
occluding trap for tortoises. Jour-
nal of Wildlife Management.
45(2):557-559.

Rose, F.L. 1970. Tortoise chin gland
fatty acid composition: behavioral
significance. Comparative bio-
chemistry and physiology. 32:577-
580.

Sauer, J.R., and N.A. Slade. 1987.
Size-based demography of verte-
brates. Annual Review of Ecology
and Systematics. 18:71-90.

Speake, Dan W. 1983. Unpublished
report. Report of the survey of in-
digo snake/ gopher tortoise den-
sity on the Margo Study Area of
Kennedy Space Center, Florida.
Alabama Cooperative Wildlife Re-
search Unit, Auburn, AL. 24 p.

Spillers, D.M., and D.W. Speake.
1986. Status and distribution of the
gopher tortoise (Gopherus pol-
yphemus) in southern Alabama.
Report to U.S. Fish and Wildlife
Service, Jackson, MS. 24 p.

Stout, I. Jack, Terry J. Doonan, Rich-
ard E. Roberts, and Donald R.
Richardson. In press. A compari-
son of results of three gopher tor-
toise relocations in central and
southeast Florida. In: J.E. Diemer,
D.R. Jackson, J.L. Landers, J.N.
Layne and D.A. Wood. Proceed-
ings of the gopher tortoise reloca-
tion symposium. Florida Game
and Fresh Water Fish Commis-
sion, Nongame Wildlife Program
Technical Report. Tallahassee, FL.

Taylor, Robert W. 1982. Human pre-
dation on the gopher tortoise (Go-
pherus polyphemus) in north-central
Florida. Bulletin of the Rorida
State Museum, Biological Sciences.
28(4):79-102.

Turner, Frederick B., Page Hayden,
Betty L. Burge, and Jan B.
Robertson. 1986. Egg production
by the desert tortoise (Gopherus

agassizii) in California. Herpetol-
ogica 42(1):93-104.

Van Home, Bruce. 1983. Density as a
misleading indicator of habitat
quality. Journal of Wildlife Man-
agement. 47:813-901.

Vogt, Richard C, and Ruth L. Hine.
1982. Evaluation of techniques for
assessment of amphibian and repv
tile populations in Wisconsin, p.
201-217. In: N.J. Scott (ed.). Herpe-
tological Communities. U.S. Fish
and Wildlife Service. Wildlife Re-
search Report 13. Washington,
D.C

Wester, Edward Earl. 1986. Demog-
raphy of two insular populations
of the gopher tortoise, Gopherus
polyphemus (Daudin), in a reser-
voir in southwestern Georgia.
M.S. thesis. Auburn University,
Auburn AL. 57 p.

Wilkinson, Leland. 1985. Systat: The
system for statishcs. SYSTAT, Inc.
Evanston, IL. 417 p.

Wolcott, Thomas G. 1981. A radio
probe for mapping burrows with-
out excavation. Journal of Wildlife
Management 45(4):1069-1072.

Woodman, A. Peter, and Kristin H.
Berry. 1984. A description of car-
cass deterioration for the desert
tortoise and a preliminary analysis
of disintegration rates at two sites
in the Mojave Desert, California,
p. A6/1-A6/31. In: K. H. Berry,
(ed.). The status of the desert
tortoise (Gopherus agassizii) in the
United States. Desert Tortoise
Council report to the U.S. Fish and
Wildlife Service, Sacramento, CA.
502 p.+ appendixes. Order No.
11310-0083-81.

Wood, Don. 1987. Official lists of en-
dangered and potentially endan-
gered fauna and flora in Florida.
Florida Game and Fresh Water
Fish Conunission, Tallahassee, FL.
19 p.

Woodruff, Robert E., and Bert C.
Klein, in prep. A new baited pitfall
trap design for beetles. 4 p.

215

Talus Use by Amphibians and
Reptiles in the Pacific
Northwest^

Robert E. Herrington^

Abstract.— Field data and a review of available
literature v/ere used to categorize the extent of talus
usage by individual herpetofaunol species. Five
categories were recognized that ranged from
species essentially restricted to talus slopes to those
that were only occasionally observed there. More
than 60% of the amphibian and reptile species that
occur in the states of Oregon and Washington were
found to utilize talus habitats. In addition to species
essentially restricted to talus slopes, the most
frequent use patterns were to moderate the effects
of adverse seasonal weather conditions and the use
of talus slopes for reproductive activities.

In recent years, biologists have em-
phasized the importance of preserv-
ing habitats with high species diver-
sity (EhrHch and Ehrlich 1981). In
this context, habitats that play a criti-
cal role in the life cycle of a large
number of species should also be
considered for protection. However,
there is little information available
concerning habitat utilization by
many amphibian and reptile species,
and even less on the combined use of
a single habitat by both of these
groups (but see Scott and Campbell
1982).

Obtaining data on habitat use of
amphibians and reptiles is often hin-
dered by the fact that habitat fidelity
is extremely variable for these
groups. Most studies have concerned
eastern species, but some generaliza-
tions have emerged. Small species
may be more or less restricted to a
single habitat (Ashton 1975, Barbour
et al. 1969, Fitch 1958, Gregory et al.
1987, and Rose 1982). Others rou-
tinely occupy two or more distinctly
different habitats over a single sea-
son. The latter group includes species
that migrate to reproduce and those
which use a separate habitat for hi-
bernation and /or aestivation (Brown
and Parker 1976, Duvall et al. 1985).

'Paper presented at symposium. Man-
agement of Amphibiarts. Reptiles, and
Small Mammals in Northi America. (Flag-
staff, AZ, July 19-21. 1988.)

'Robert E. Herringfon is Assistant Profes-
sor of Biology, Georgia Southtwestern Col-
lege, Americus, GA 31709.

Hov^ever, the imjx)rtance of a habitat
to the continued survival of a popu-
lation is not necessarily correlated
with the time that a species spends
within it. Providing reproductive
habitat, refugia from adverse
weather conditions, or protection
from predators can disproportion-
ately influence the role that a particu-
lar habitat plays in the ecology of the
animals that use it.

Talus slopes are "unique habitats"
(Maser et al. 1979), that represent the
gradual accumulation of weathered
rock fragments (mostly basalt and
andesite) from a cliff face (Strahler
1981). Individual slopes are quite
variable in rock size, aspect and in
the amount and type of vegetation
present. These factors interact in
complex ways to provide a broad
range of thermal and moisture re-
gimes that amphibians and reptiles
can select. This study examines the
use of talus slopes by amphibians
and reptiles and compares these
findings with non-talus areas.

Study Area and Mettiods

Herpetofauna associated with talus
slopes and adjacent non-talus areas
was determined by field observation
and a review of the literature
(Campbell et al. 1982). For the pur-
pose of this investigation, talus habi-
tats were those in which the sub-
strate was predominantly weathered
rock fragments (typically with an as-

sociated cliff-face) and included a 10
meter wide band of transitional habi-
tat. Non-talus habitats were those in
which the substrate was not as de-
scribed above and were located a
minimum of 100 meters from a talus
area. Aquatic habitats were not spe-
cifically sampled; however, speci-
mens observed under objects located
above the high water mark were in-
cluded in the analysis.

Field work was conducted be-
tween August, 1981 and August,
1985. During this period, more than
100 days were spent in the Cascade
Mountains of southern Washington
and northern Oregon. Additional
surveys ranging from 2-6 days each,
were conducted in the North Cas-
cades of Washington, the Coast
Range of southern Oregon, and the
Wallowa Mountains of northeastern
Oregon. A total of 183 individual ta-
lus slopes and adjacent non-talus ar-
eas were surveyed. Approximately
equal time was spent searching talus
and adjacent non-talus habitats. Ta-
lus slopes were considered to have
been altered by human activities if
there was evidence of extensive rock
or tree removal.

Searches were conducted by turn-
ing surface debris, raking through
leaf litter, and in the case of talus, by
digging in the upper layers of rock
with a potato rake. Data recorded for
most specimens included habitat
type, the activity the animal was en-
gaged in when first observed (active
or inactive, surface or sub-surface.

216

Table 1— Talus use by amphibians and reptiies obsen^ed during this study.
Use patterns by IndlvlduoJ species were: (1) species generally restricted to
Talus habitats; (2) species which use talus areas for reproductive activities;

(3) species which use talus areas to survive adverse weather conditions;

(4) species frequently associated with talus areas; (5) species occasionally
observed In talus habitats.

Species

Numbers of individuals
obsen/ed In
Taius/non-talus habitats talus use pattern(s)

AMPHIBIANS

Frogs

Hyla regilla
Rana aurora
Rana cascade
Rana prefiosa
Bufo bore as
Ascaphus true}

Salamanders

Ambysfom a gracile
Ambysfoma macrodacfylum
Dicampfodon eDsafus
Rhyacofrifon olympicus
Anejdes ferreus
Plethodon eiongafus
Plefhodon dunni
Plethodon farselli
Plefhodon vehiculum
Plefhodon vandykei
Plefhodon sforrrv
Ensafina eschscholf^
Bafrachoseps wrighfi
Taricha granulosa

REPTILES
Lizards

Elgaria coerulea
Elgaria mulficarinafa
Eum eces skllfonianus

Snakes

Confia tenuis
Coluber constrictor
Charina boftae
Diadophis punctatus
Hypsigfena tor quota
Pituophis melanoleucus
Thamnophis e/egans
Thamnophis ordinoides
Thamnophis sirtalis
Crotalus viridis

NUMBER OF INDIVIDUALS
SPECIES RICHNESS

3/21
2/17

0/6
0/17

2/9
2/18

3/8
5/63
8/5
26/109
3/8
43/3
123/87
383/20
193/146
31/3
19/0
32/26
13/8
5/32

19/9
3/0
5/3

2/0
10/21
4/2
2/2
2/0
8/5
9/11
16/3
13/19
28/5

1017/686
31/29

3
3

3
3
2,3
3
5
1

3,4
1,2,3
2,3,4
1
1

4,5
4,5
3

3,4
2,3,4
4

3,4
3
4
4
3,4
2,3
3
2,3
2,3
2,3

foraging or involved in reproductive
activities), and a subjective evalu-
ation of the individual's approximate
age (hatchling, juvenile, or adult).
The determination that an individual
v^as using talus to avoid unfavorable
weather conditions was based on the
season, prevailing weather condi-
tions, the behavior exhibited by the
animal when uncovered, and the
depth at which the specimen was lo-
cated.

These observations were summa-
rized in an effort to categorize pat-
terns of talus use. Voucher specimens
of most species have been deposited
in the vertebrate collection. Depart-
ment of Zoology, Washington State
University. However, the majority of
specimens were identified in the field
and released at the site of capture.

Results and Discussion
Habitat Use

A total of four species of frogs were
observed in talus habitats (table 1),
with a fifth species reported using
talus areas for feeding (table 2). A
single Hyla regilla and two Rana au-
rora were located under snow cov-
ered talus and were considered to
have been hibernating there. All frog
species were more numerous in non-
talus areas and two species (Rana cas-
cade and R. aurora) observed in non-
talus areas were not recorded from
talus areas.

Salamanders were numerically
and taxonomically the most abun-
dant amphibians encountered during
the study. The number of species re-
corded from talus and non-talus
habitats were 14 and 13, respectively
(table 1). However, species richness
is somewhat misleading, since more
than 90% of the observations of Ple-
thodon elongatus, P. larselli, P. stormi,
and P. vandykei were from talus habi-
tats. I consider these species to be
essentially restricted to forested talus
areas. This observation is supported
by the work of Stebbins and Rey-

217

nolds (1947) with P. elongatus, Nuss-
baum et al. (1983) with P. stormi and
P. vandykei, and Harrington and
Larsen (1985) with P. larselli. Five
additional species (Dicamptodon en-
satus, P. dunni, P. vehiculum, Ensatim
eschscholtzi, and Batrachoseps wrighti)
were observed more frequently in
talus than in other habitats (table 1).

All the salamanders mentioned
above with the exception of Dicamp-
todon ensatus, are capable of complet-
ing their entire life cycle within talus
habitats. I observed portions of the
courtship sequences of Plethodon ve-
hiculum and P. vandykei only on
damp talus. Many of these same spe-
cies probably nest in deep recesses
within the talus. This is based on two
observations. The first is that given
the abundance of some salamander
species, very few nests have ever
been located (Hanlin et al. 1979,
Jones and Aubry 1985). This suggests
that nests are located in places gener-
ally inaccessible to investigators. The
slope and rock size associated with
talus fields generally precludes dig-
ging at depths > 50cm without the
talus caving in. Secondly, I found
small aggregations (1-3 individuals)
of P. larselli, P. vehiculum, and P.
dunni, that approached the size re-
ported for hatchlings (Stebbins 1951,
Peacock and Nussbaum 1973, Her-
rington 1985) only in loose talus ar-
eas, following the first fall rains. This
is the time that recent hatchlings are
likely to to emerge from their nests.

Individuals uncovered from talus
in situations suggesting that they
were in winter dormancy included
Ambystoma gracile, A. macrodactylum,
Dicamptodon ensatus, Rhyacotriton
olympicus, Plethodon dunni, P. larselli,
P. vehiculum, and Taricha granulosa.
Conversely, between June and Au-
gust there was reduced rainfall and
elevated surface temperatures
throughout most of the study areas.
Because of this, surface activity by
salamanders was greatly restricted
and the majority of observations
(83%) were of individuals uncovered
from talus areas.

A total of 5 species of lizards were
observed or reported from talus
habitats (tables 1 and 2). Elgaria coer-
ula was the most frequently observed
species and most individuals were
uncovered from the upper layers of
talus. T\yo behavioral patterns were
apparent. The first involved indi-
viduals uncovered before they had
emerged from nocturnal retreats and
the second was of individuals ther-
moregulating under surface talus.
Elgaria coerula is a live-bearing spe-
cies and this behavior may be impor-
tant to the developmental processes
taking place. Talus habitats have
been identified as oviposition sites
for Sceloporus occidentalis and Uta
stanshuriana (Maser et al. 1979) and
Elgaria multicarinata (Brodie et al.
1969). Elgaria coerula and E. multicari-
nata were uncovered from talus
slopes where they appeared to be hi-
bernating.

Ten species of snakes were ob-
served (table 1) and two additional
species reported from talus habitats
(table 2). Taken as group, snakes
were most frequently observed bask-
ing either on the surface or between
exposed rocks. Species that I consid-
ered to be entering or emerging from

hibemacula located within talus were
Crotalus viridis, Pituophis melano-
leucus. Coluber constrictor, Thamnophis
elegans, T. ordinoides, T. sirtalis, Hyp-
siglena torquata, and Contia tenuis.
Both Hypsiglena torquata and Contia
tenuis were only observed in talus
habitats during the study, but they
are known to occupy a broader range
of habitats elsewhere (Cook 1960;
Diller and Wallace 1981).

Talus slopes play an important
role in the reproductive activities of
snakes. Brodie et al. (1969) reported
several individuals of Coluber con-
strictor, Diadophis punctatus, Contia
tenuis and Pituophis melanoleucus ovi-
positing within an exposed talus
slope in Benton Co., Oregon. I ob-
served gravid females of Thamnophis
sirtalis, T. ordinoides and Crotalus
viridis basking on talus slopes during
late summer. Whether these snakes
delivered their young at the talus
slopes is not known. However,
gravid C. viridis are known to remain
in the vicinity of their hibernacula to
produce young (R. Wallace, Depart-
ment of Biological Sciences, Univer-
sity of Idaho, pers. comm.), and I un-
covered 7 "yearling" T, ordinoides
from an area of talus less than 2 m^.

Table 2.— Amphibian and reptile species not observed during this study,
but which have been reported to utilize talus habitats. The categories of
talus use are described In table 1.

Species

Talus use pattern Reference

AMPHIBIANS
Frogs

Bufo woodhousei

REPTILES

Lizards

Crofaphyfus bicincfores
Sceloporus occidentalis
Ufa sfansburiana

Snakes

Lampropelfls zonafa
Mastlcophis taeniafus

Maser et al. (1979)

4

2
2

1

2

Nussbaum et al. (1983)
Maser et al. (1979)
Maser etal. (1983)

Nussbaum et al. (1983)
Nussbaum et al. (1983)

218

where they appeared to be in hiber-
nation. It was not possible to deter-
mine if these snakes had independ-
ently congregated there, or if they
represented a single litter born at the
talus slope, but the latter explanation
seems more plausible.

The importance of talus slopes in
the feeding ecology of snakes is un-
known. The relative abundance of
garter snakes and salamanders on
talus slopes at certain times of the

year could lead to predator-prey in-
teractions. This is supported by evi-
dence palpated from the stomachs of
two Thamnophis sirtalis and one T.
ordinoides captured on talus slopes.
Each of the T. sirtalis contained a
salamander (1 Plethodon dunni; 1 En-
satina eschscholtzi); the single T. ordi-
noides contained a large slug {Arioli-
max sp.). While other interactions
were not observed, small mammals
often were observed in talus habitats.

Alterations to Talus Slopes

It became apparent after the initia-
tion of this study, that a large num-
ber of the talus slopes being sur-
veyed had been or were being al-
tered by human activities. Habitat
modifications involved two not mu-
tually exclusive alterations. The first
was the removal of rock from the
base of talus slof>es to be used for
road construction raw materials (fig.
1). The second involved tree removal
(clearcutting) from the talus slopes.

I revisited talus slopes surveyed in
the early part of the project to deter-
mine the frequency and type of al-
teration. Of 183 talus slopes sur-
veyed, 106 were altered; 76 had no-
ticeable quantities of talus removed,
13 had been deforested, and 17 had
been altered by both events.

I was able to document few clear
species specific trends between al-
tered and unaltered talus slopes (see
Conclusions). However, there were
differences in the number of indi-
viduals encountered. Unaltered
slopes represented 42% of the habi-
tats surveyed but yielded 73% of the
total number of individuals. Because
there were differences in the amount
of search effort (time) expended sur-
veying altered and unaltered talus
habitats, I did not statistically com-
pare these results.

Conclusions

Talus slopes provide important habi-
tat for a significant segment of the
herpetofauna of the Pacific North-
west. A total 37 of the 58 species of
amphibians and reptiles that occur in
the states of Washington and Oregon
are documented from talus slopes.
Use of this resource by amphibians
and reptiles was quite variable, but
three important patterns emerged.
The first involves species essentially
restricted to talus habitats. Four spe-
cies of plethodontid salamanders fit
this pattern (Plethodon larselli, P. van-
dykei, P. elongatus, and P. stormi).

219

The second category of talus use
consisted of species which use talus
slopes to avoid potentially lethal
temperature extremes. Nineteen spe-
cies (10 reptiles, 9 amphibians) were
included here. Several species of
snakes travel considerable distances
to congregate at communal hibernac-
ula (Duvall et al. 1985, Gregory and
Stewart 1975, and Brown and Parker
1976). This behavior conceivably
could put an entire population at risk
if the hibemacula were irreparably
altered.

A third use pattern of talus slopes
was for reproductive activities. In
addition to an egg-laying aggregation
of 5 species of reptiles reported by
Brodie et al. (1969), live-bearing rep-
tiles were frequently observed in th-
ermoregulatory behaviors on and
along the edge of talus slopes. The
importance in this behavior to com-
pletion of developmental processes
remains to be determined.

Each of these utilization patterns
is important to a particular segment
of the herpetofaunal community.
Whether or not the availability of
suitable talus slopes is a limiting fac-
tor for any of these species remains
unknown. However, talus slopes
typically make up only a small por-
tion of the available habitat. In the
Gifford Pinchot National Forest
(where a large part of this work was
conducted), Scharpf and Dobler
(1985) found talus slopes to occupy
less than 5% of the total land area,
most other areas have less.

The high frequency of altered ta-
lus slopes observed during this study
may pose a significant threat to the
long-term survival of many of the
amphibians and reptiles that use
them. Talus removal for road build-
ing materials and tree removal from
the slopes initiate complex changes
in the structure of the slope. Trees,
through leaf fall, provide a major in-
put of nutrients to the slope, as well
as increasing the moisture retention
capabilities of the sub-surface talus.
Tree removal increases the solar ra-
diation reaching the slope and this

results in the rapid loss of moisture
from the upper layers of talus. In a
study comparing the habitat selection
of P. larselli and P. vehiculum (Her-
rington and Larsen 1985), tree re-
moval was implicated in rendering a
talus slope unsuitable for habitation
by P. larselli, but not for P. vehiculum.

Talus removal results in a major
shift of the slope towards its base.
This results in the extensive move-
ment of both surface and deep layers
of talus. The immediate effect would
be to kill or injure many of the rep-
tiles and amphibians inhabiting the
slope as well as destroy any nests lo-
cated there. A long term consequence
of rock removal is that erosional
processes are increased. This results
in an increase in the amount of soil
present in the talus, and could con-
ceivably close off access and fill in
areas formerly used as hibemacula.

Management Recommendations

Prior to altering a particular talus
slope, a survey should be conducted
to detennine the presence of threat-
ened, endangered, or otherwise sen-
sitive species. Additionally, it should
be determined whether or not the
slope in question serves as a major
snake hibernaculum.

Tree removal from talus slopes
should be restricted and logging
practices should be modified to al-
low for leaving a sufficient border of
trees (20-30 m) along the margin of
talus slopes.

Current practices of removing ta-
lus for road building materials from
each slope encountered should be
discouraged. Selected talus areas
known not to contain threatened, en-
dangered or sensitive species or to be
major snake hibemacula should be
utilized as a source of rock for con-
stmction activities.

One area that needs additional
study is the colonization and use by
amphibians and reptiles of artificially
created talus areas. These would in-
clude areas such as the banks of road

cuts with riprap, and rock piles asso-
ciated mining processes. Those
sampled during the study were
found to have a depauperate fauna
compared to natural talus areas and
the fauna consisted almost entirely of
species known to have broad habitat
tolerances. However, the possibility
remains that with adequate planning,
suitable areas could be constructed in
such a manner to benefit amphibians
and reptile faunas.

Acknowledgments

Portions of this study were funded
by the Washington Department of
Game, the Mazamas, the Society for
the Study of Amphibians and Rep-
tiles, and Washington State Univer-
sity. Brian Miller, Chris Davitt, and
Linda Whittlesey assisted with field
work. Comments and suggestions by
Stephen Corn, Patrick Gregory, and
Kieth Severson substantially im-
proved this manuscript. Shelia Hines
typed the numerous drafts of this
manuscript. For all of this help, I am
exceedingly grateful.

Literature Cited

Ashton, R. E. 1975. A study of the
movements, home range and win-
ter behavior of Desmogmthus fus-
cus. Joumal of Herpetology 9:85-
92.

Barbour, Roger W., M. J. Harvey,
and J. W. Hardin. 1969. Home
range, movements and activity of
the eastern worm snake, Carphohis
amoenus amoenus. Ecology 50:470-
476.

Brodie, Edmund D. Jr., Ronald A.
Nussbaum, and Robert M. Storm.
1969. An egg-laying aggregation of
Oregon Reptiles. Herpetologica
25:223-227.

Brown, William S. and William S.
Parker. 1976. Movement ecology
of Coluber constrictor near commu-
nal hibemacula. Copeia 1976:225-
242.

220

Campbell, R. Wayne, Michael G.
Shepard, Brigitta M. Van Der Raay
and Patrick T. Gregory. 1982. A
Bibliography of Pacific Northwest
Herpetology. Heritage Record No.
14. The British Columbia Provin-
cial Museum, Victoria, British Co-
lumbia, 152 p.

Cook, Sherburne F. 1960. On the oc-
currence and life history of Contia
tenuis. Herpetologica 16:163-173.

Diller, Lowell V. and Richard L. Wal-
lace. 1981. Additional distribution
records and abundance of three
species of snakes in southwestern
Idaho. Great Basin Naturalist
41:154-157.

Duvall, David, M. B. King, and M. J.
Gutz wilier. 1985. Behavioral ecol-
ogy and ethology of the prairie
rattlesnake. National Geographic
Research 1:80-111.

Ehrlich, Paul R. and Ann Ehrlich.
1981. Extinction: The Cause and
Consequence of the Disappear-
ance of Species. Random House,
New York, 305 p.

Fitch, Henry S. 1958. Home range,
territories, and seasonal move-
ments of the vertebrates of the
Natural History Reservation.
Univ. of Kansas Publ., Museum of
Natural History 11:63-326.

Gregory, Patrick T., J. Malcolm
Macartney and Karl W. Larsen.
1987. Spatial patterns and move-
ments, p. 366-395 In Snakes: Ecol-
ogy and Evolutionary Biology.
Richard A. Seigel, Joseph T.
Collins, and Susan S. Novak, edi-
tors. Macmillan Publishing Co.
New York.

Gregory, Patrick T. and K. W. Ste-
wart. 1975. Long-distance disper-
sal and feeding strategy of the red-
sided garter snake (Thamnophis
sirtalis parietalis) in the Interlake of
Manitoba. Canadian Journal of
Zoology 53:238-245.

Hanlin, Hugh G. and Joseph J.
Beatty, and Sue W. Hanlin. 1979.
A nest site of the western red-
backed salamander Plethodon ve-
hiculum (Cooper). Journal of Her-
petology. 13:214-216.

Herrington, Robert E. 1985. The ecol-
ogy, reproductive biology and
management of the Larch Moun-
tain Salamander (Plethodon larselU
Burns) with comparisons to two
other sympatric plethodons. Ph.D.
dissertation Department of Zool-
ogy, Washington State University,
Pullman, WA. 102 p.

Herrington, Robert E. and John H.
Larsen, Jr. 1985. The current
status, habitat requirements and
management of the Larch Moun-
tain salamander. Biological Con-
servation 34:169-179.

Jones, Lawrence L. C. and Keith B.
Aubry. 1985. Ensatina eschscholtzi
oregonensis: Reproduction. Herpe-
tological Review 16:26.

Maser, Chris, Jon E. Rodiek, and Jack
W. Thomas. 1979. Cliffs, Talus and
Caves. In Wildlife Habitats in
managed Forests the Blue Moun-
tains of Oregon and Washington.
Jack W. Thomas, Editor. U.S. De-
partment of Agriculture, Agricul-
ture Handbook No. 553.

Nussbaum, Ronald A., Edmund D.
Brodie, Jr., and Robert M. Storm.
1983. Amphibians and Reptiles of
the Pacific Northwest. The Univer-
sity of Idaho Press, Moscow,
Idaho. 332 p.

Peacock, Robert L., and Ronald A.
Nussbaum. 1973. Reproductive
biology and population structure
of the western red-backed
salamander, Plethodon vehiculum
(Cooper). Journal of Herpetology
7:215-224.

Rose, Barbara 1982. Lizard home
ranges: Methodology and Func-
tions. Journal of Herpetology
16:253-269.

Scott, Norman J. and Howard W.
Campbell. 1982. A chronological
bibliography, the history and
status of studies of herpetological
communities, and suggestions for
future research, p. 221-239. In Her-
petological Communities. Norman
J. Scott, editor. United States De-
partment of the Interior Fish and
Wildlife Service Wildlife Research
Report 13:239 p.

Scharpf, Raymond W. and Fred C.
Dobler. 1985. Caves, Cliffs and Ta-
lus, p. 187-197. In Management of
Wildlife and Fish Habitats in For-
ests of Western Oregon and Wash-
ington. E. Reade Brown, Editor.
U.S. Department of Agriculture,
Forest Service, Pacific Northwest
Region.

Stebbins, Robert C. 1951. Amphibi-
ans of Western North America.
University of California Press,
Berkeley, 539 p.

Stebbins, Robert C. and H. C. Rey-
nolds. 1947. Southern extension of
the range of the Del Norte sala-
mander in California. Herpetol-
ogica 4:41-42.

Strahler, Arthur N. 1981. Physical
Geology. Harper & Row, NY.
612 p.

221

Comparison of
Herpetofaunas of a Natural
and Altered Riparian
Ecosystem^

K. Bruce Jones^

Abstract.— Reptile abundance and diversity were
greater on an unaltered riparian ecosystem tlian on
an altered site; the former l^ad some species
typically found on upland habitats (e.g., chaparral)
and the latter was comprised of species from
adjacent Sonoron Desert. The distribution and
abundance of certain microhobitats appear to
account for differences in reptile abundance and
diversity on the two sites.

Over the past 25 years,
concerns have increased about
the impacts of population growth
and associated development on
wildlife habitats within the
southwestern United States,
especially the impacts of
increased demand for water
resources within arid regions. A
series of long-term studies on the
Colorado River have shown that
dam-induced habitat alternations
have reduced overall bird
abundance and diversity (Ohmart
et. al. 1977). Most of the once wide-
spread riparian woodland along the
Colorado River has been replaced by
non-native salt cedar (Tamarix spp.)
and shrubs typically found in inter-
mittent drainages (Ohmart et. al.
1977). Many of the birds requiring
riparian woodland are no longer
found along the Colorado River.

Many studies demonstrate how
water impoundments impact birds
and fish of riparian and aquatic habi-
tats, but little is known about im-
pacts on amphibians and reptiles in-
habiting these ecosystems. Jones et
al. (1985) and Jones and Glinski
(1985) found that a number of mesic-
adapted or upland amphibians and

'Paper presented at symposium. Man-
agement of Amphibians, Reptiles, and
Small Mammals in Nortt) America. (Flag-
staff, Arizona, July 19-21. 1988.)

'K. Bruce Jones is a Researct) Ecologist
with the Environmental Protection Agency,
Environmental Monitoring Systems Labora-
tory, Las Vegas, Nevada 89193.

reptiles were restricted entirely to
cotton wood- willow riparian habitats
within the Sonoran Desert. Usually
found in habitats of the Upper Sono-
ran Life-zone (e.g.. Chaparral), these
species immigrated into lower eleva-
tions (< 762 m) of the Sonoran Desert
via riparian corridors ([ones et al.
1985). Upland species occur on a few
riparian sites within the Sonoran
Desert that have maintained mesic
habitat conditions (Jones et al. 1985).
These conditions persist on these
sites due to the moderating effects of
leaf litter and logs resulting from Cot-
tonwood trees (Populus fremonti),
perennial waterflow, shading of the
surface by trees, and accumulation of
large debris piles resulting from peri-
odic flooding (fones and Glinski
1985). In California, for example, ri-
parian ecosystems provide habitat
for 83 percent of the amphibians and
40 percent of the reptiles known from
that state (Erode and Bury 1984).

Water impoundment structures
eliminate periodic flooding and sig-
nificantly reduce stands of cotton-
woods and willows (Salix goodingii)
along major drainages (Ohmart et al.
1977). These structures may, there-
fore, significantly reduce mesic con-
ditions in downstream riparian eco-
systems. To determine the possible
impact of impoundment structures
on the herpetofauna of a desert ripar-
ian ecosystem, I studied two low ele-
vation (< 762 m) sites, one with ma-
jor water impoundments and one
without any impoundments.

Figure 1 .—Locations of the study areas.

Methods

To compare herpetofaunas of an un-
altered vs altered desert riparian eco-
system, I chose study sites on the
Hassayampa and Salt Rivers. The
Hassayampa River has no major wa-
ter impoundments. It originates in
the Bradshaw Mountains 160 km
north-northwest of Phoenix, Arizona,
eventually draining into the Gila
River approximately 80 km south-
west of Phoenix. Lower reaches are
mostly intermittent, except for a 15
km perennial section near Wick-

222

enburg, Arizona. The study site was
located approximately 10 km south
of Wickenburg near Palm Lake, a for-
mer resort now owned by The Na-
ture Conservancy, in a mature gal-
lery-type stand of cotton wood (Popu-
lus fremonti) and willow (Salix good-
ingii) (elevation ca. 585 m, fig. 1).

The Salt River originates in east-
central Arizona, flowing southwest
to Granite Reef Dam (approximately
40 km northeast of Phoenix) where
water is diverted for irrigation. Be-
low this point, the floodplain trav-
erses Phoenix, eventually draining
into the Gila River approximately 26
km southwest of Phoenix. Histori-
cally, this river flowed perennially
over its entire course. However, sev-
eral major water impoundments, in-
cluding dams forming Roosevelt,
Apache, d^nyon, and Saguaro lakes,
have significantly altered flows and
consequently physical characteristics.
Flows are regulated by water re-
leases at dams and flooding has
nearly been eliminated; significant
flooding has occurred only when wa-
ter releases from lakes have been
necessary. Before water impound-
ment, riparian vegetation was mostly
Cottonwood and willow, with
mesquite (Prosopis glanulosa) occur-

Tabl© 1 —List of microhabltats measured
at or around each pit-fall trap on eacti
river. Frequency equals tt)e percentage
of quarters around each trap tt^at had a
certain microhabitat.

Soil type (at trap)
Vertical cover

(over trap)
Distance to leaf litter
Leaf litter depth
Leaf litter frequency
Distance to log

Log diameter
Log frequency
Distance to

debris heap
Debris heap width

Rock width
Rock frequency

Distance to tree
Tree height
Tree width (crown)
Tree frequency by

species
Distance to shrub
Shrub height
Shrub width

(crown)
Shrub frequency

Debris heap depth Distance to grass

Patch

Debris heap Grass height

frequency
^Distance to rock Grass frequency

ring primarily on vegas adjacent to
the river(Reference?). Mesquite and
tamarisk now dominate the riparian
community, with only a few small (<
100 m in length) sections of cotton-
wood and willow. The Salt River
sample site was at Blue Point, located
approximately 6 km south of Sa-
guaro Lake (fig. 1). Cottonwood, wil-
low, and mesquite trees were com-
mon at this site, although cotton-
woods and willows were not nearly
as common as on the Hassayampa
River. Blue Point's tree gallery was
poorly developed and I found no evi-
dence of tree reproduction. Substrate
was dominated by sand, with gravel
bars located intermittently through-
out the site. Similar to the Has-
sayampa River site, several small
drainages traversed this site.

The herpetofauna on each site was
sampled by using a pit-fall trapping
grid consisting of 110, double-deep
1.4 kg coffee cans placed 15 m apart
in a 22 X 5 grid trapping configura-
tion (1.9 ha) (see Jones 1987). Covers
were placed approximately 15 cm
above each trap to reduce loss of ani-
mals due to desiccation and expo-
sure. Traps were open continuously
between March and October, 1984.
Traps were checked every three
days, and amphibians and reptiles
captured in traps were measured
(snout- vent length, SVL), weighed,
sexed, uniquely marked, and re-
leased into cover nearest to the cap-
ture site.

While traveling between pit-fall
traps, I recorded observations of all
frogs, toads, lizards, and snakes. I
also flipped rocks and logs to un-
cover hidden herpetofauna.

In order to determine amphibian
and reptile composition in adjacent
Sonoran Desert, a modified array pit-
fall trapping method was used (Jones
1987). Five arrays were placed in
Sonoran Desert habitat adjacent to
each site, and I checked these arrays
for animals whenever I checked the
main grids.

A point-center quarter (plotless)
sampling method (Muller-Dumbois

and Ellenberg 1974) provided data to
characterize microhabitats around
each trap. Each trap was a center
point for quantifying density and fre-
quency of microhabitats within 7m of
each trap. I sampled 110 points or
440 quarters on each site. Microhabi-
tat frequency was determined by di-
viding the number of quarters that a
microhabitat occurred in (7 m or less
from the trap) by the total number of
quarters (440). I also estimated size
(width, height, and depth) of each
microhabitat and frequency of can-
opy cover as the percentage of pit-
fall traps that were covered by vege-
tation (table 1).

Relative abundance equaled the
number of an individual species
trapped during a 24-hour period. I
estimated the diversity of herpe-
tofaunas and microhabitats on each
site using a modified Shannon-
Weaver diversity index (H') (Hair
1980): H' = fiPjdogjoP,), where s =
number of species and pj = the pro-
portion of the total number of indi-
viduals consisting of the i^j^ species. I
used a Student's t-test to determine
differences between herpetofaunas
and microhabitats on the two sites.
Finally, I compared herpetofaunas of
the two riparian sites and adjacent
Sonoran Desert by calculating Jac-
card Similarity Coefficients and then
clustered them using an unweighted
pair group average (Pimental 1979).

Results

Microhabitats

The Hassayampa River had greater
amounts and diversity of microhabi-
tats than the Salt River (table 2). Of
these differences, the frequency of
downed litter on the two sites was
the greatest (table 2). Leaf litter was 3
times more common, debris heaps 10
times more common, and logs and
limbs twice as common on the Has-
sayampa River than on the Salt River
(table 2). Rock substrate and grasses
were more common on the Has-

223

sayampa River and shrubs on the
Salt River (table 2). Trees were com-
mon on the Hassayampa River and
sand substrate on the Salt River, al-
though neither of these differences
were significant (table 2). In addition,
average leaf litter depth was signifi-
cantly greater on the Hassayampa
River than on the Salt River (table 2).

Of the specific types of canopy
covering pit-fall traps, trees were by
far the most common on both rivers,
although the Salt River had more pit-
fall traps with no canopy cover (fig.
2).

Tree composition varied consid-
erably between sites. The Has-
sayampa River had more
cottonwoods (Populus fremonti) and
willows (Salix goodingi) and the Salt
River more salt cedars {Tamarix spp.)
(fig. 3). Mesquite (Prosopis glandulosa)
was the most common tree on both
sites (fig. 3).

The Hassayampa River had more
trees in the 0-1.9, 5.0-9.9, and 10.0-
14.9 m height ranges, but most at the
Salt River were in the 2.0-4.9 m range
(fig. 4). Cottonwood height distribu-
tion was relatively even on the Has-
sayampa River, but most Salt River
cottonwoods were greater than 10 m,
with none less than 5 m, hence no
reproduction (fig. 5).

Herpetofaunas

The abundance and diversity of her-
petofauna was greater on the Has-
sayampa River than on the Salt
River. The Hassayampa River had
nearly twice as many species, more
than twice the number of individu-
als, and a greater species diversity
(1.05 vs. 0.86) than the Salt River (fig.
6).

All but three species (Bufo mi-
croscaphus x woodhousei, B. punctatus,
and Cnemidophorus tigris) were more
abundant on the Hassayampa River,
and this site had five "upland"
species (Cophosaurus texanum, Diado-
phis punctatus, Eumeces gilberti, Masti--
cophis bilineatus, and Tantilla

hobartsmithii) usually found in habi-
tats of the Upper Sonoran Life-zone
(e.g., chaparral). These upland spe-
cies were absent from the Salt River
and adjacent Sonoran Desert (table
3). C. tigris had the same abundance
on both rivers, C. tigris was the most
abundant sp)ecies on the Salt River,
and E. gilberti was the most abundant

species on the Hassayampa River
(table 3). The Hassayampa River also
had 4 species with abundances
greater than 1.0, whereas the Salt
River only had one (table 3).

A cluster analysis of Jaccard Simi-
larity Coefficients using data in table
3 revealed that the Salt River riparian
site had a herpetofauna more similar

labia 2*— Co ^btiot uDundu -^en the Salt und

l^assayampa .v ^..^^ Is ihe m^an rw.HK.^* n^^arters !n which a

tmitcfohablfcsf Was found csrcmd each frcp <v'?thin 7 m) + SD

Mlcrohab8at

$o\i River

Hassoyampa River $lgnlficar>t

Difference
'p<m

Scald substfot

(4ie)

<209)
<326)

lectf litter

Logs/downec tree (fnrtbs

(^65)

<407)

Debris heaps

<123

Trees

1.9±DJ

(213)

<315

Shajbs

(297)

<275,

Grass

0.4±0,3

(48)

<12S)

IVticrohdDitot diversity (H')

J7

CO

Q.

O

a>
o

a>

Q_

Shrub Canopy
Tree Canopy
Shrub/Tree Canopy
Open Conopy

Salt River Hcssayompo River

Canopy Type

Figure 2.— Comparison of canopy types on the Salt and Hassayampa Rivers.
224

100

80 -

-2 70

60
50 -
W h
30
20
10
0

E2L

Cottonwoods

MIon

M«s(|uite

ToTHinsk

Other

Salt FSver Hossoyompa
Tree Types

1^

Figure 3.— Comparison of tree composition on the Salt and Has-
soyampa Rivers.

o

100
90
80
70
60
50
40
X
20
10
0

■ 0.0 - 1.9 m

2.0 - 4.9 m
^ 5.0 - 9.9 m
10.0 - 14.9 m
> 15.0 m

Solt River Hossoyompo
Size Classes

River

Figure 4.— Comparison of tree hteight distribution on ttie Salt and

100

90 -
80 -
70 -
60

50 I-
40

30 h
20
10

0

H 0.0 - 1.9 m

2.0 -4.9 m
\/'^ 5.0 - 9.9 m

I 1 10.0 - 14.9 I

^8881 > 15.0 m

Sott Rivw Hossoyompo

Tree Height Glosses

River

8

1

20

bcrof !

IS

E

10

5

0

3c,

1

Spec

a75

0.5

025

0

SaHRw

Figure 6,— Comporison of ttie total number of amphibians and rep-
tiles, total relative abundance, and species diversity on the Salt
and Hassayampa Rivers.

3,^<wtjpori«)ft of »2ard cft>undance and diversity be-
tween the Salt stir. Hdssoyqmpo R|v^. Abundance is the number
of 8tardsccttight/Qrld/24 Ytotttt. Aitiphlblans and reptiles occu-
pying adjacent Sonofon Desert habllals also are Indicated,

Figure 5.— Comparison of size classes of cottonwoods on the Salt
and Hassayampa Rivers.

Spe<>ies

Salt

Nassayampo

Rivef

Sonoian
Pe$ert

Cafkaurus draconotd^s
Cmmidophoras ifgfis
■X^ofeonyx voftegaius
ieptofypfitcm hm)i{($

ijrpsautu^ orn^tus

Bufo mkifoicuphui x

Bufd puncfatw
ScapNopu$ couchi
Cophosourus t^xonum
Dkfdophf^ punckrfus
fl^^Ss gUberff

Smofn $^mharnJafa
TantSia hobaffsmma
Bufo ofvorius

0,27
147

0.07
D.47

()q7

DM'

&7

XX

0,60^
1,47

0. 03
OW

1. W

0,23

0,40

0<63

0 20^

0<TO*

2,17*

0,07*

0,07*

o,m*

0,37*

X
X
X

X
X

X

X

X

X

*S}gnfifcanf}ygf&:3fef a}:?Undance (ifp< iJ5,

X V^m^ In <3C^Qc:^nf ^ondrati DesGrf habifats viCj ptf-
fdU trapping.

XX V^ifiBtd on th& Satt River $ft& vio field s&atch.

225

to adjacent Sonoran Desert than to
the herpetofauna of the Hassayampa
River riparian site, although the two
riparian herpetofaunas were rela-
tively similar (fig. 7).

Discussion

The distribution, abundance, and di-
versity of herpetofauna on the Salt
River correlate with impoundment-
induced changes in microhabitats.
On the unaltered riparian ecosystem
on the Hassayampa River, many mi-
crohabitats were more abundant and
diverse than on the Salt River, espe-
cially surface litter and trees. These
differences in microhabitats correlate
with differences in species diversity
and abundance on the two rivers.
Species that were most abundant on
the Hassayampa River (Eumeces
gilberti, Sceloporus magister, and Uro-
saurus omatus) prefer sites with
downed vegetative litter and vertical
structure (e.g., trees) (Jones and
Glinski 1985, Jones 1986). These rep-
tiles were not nearly as common on
the Salt River and this may result
from lower surface litter and vegeta-
tion structure (higher percentage of
salt cedar, Tamarix spp., and a lower
percentage of cotton woods, Populus
fremonti, and willows, Salix goodingii)
on this site.

The greatest difference between
herpetofaunas on the two rivers was
presence of five upland species on
the Hassayampa River and the ab-
sence of these species on the Salt
River. Jones and Glinski (1985) sug-
gested these species occur in riparian
habitats within low elevation Sono-
ran Desert because of the moderating
effects of certain microhabitats, espe-
cially surface litter and debris heaps.
Surface litter and debris heaps are
considerably less common on the Salt
River, and this probably accounts for
the lack of any upland species in this
river's herpetofauna. Szaro et al.
(1985) suggest that debris heaps are
the principal source of food and
cover for Thamnophis elegans, and

that grazing-caused reduction in this
n\icrohabitat caused decline of this
snake in a high elevation riparian
community.

The relatively low amounts of sur-
face litter and lack of smaller size
classes of trees (especially cotton-
woods and willows) on the Salt River
appear to result from dam-induced
changes in water flow and flooding.
Periodic flooding is essential in the
long-term maintenance of southwest-
ern U.S. riparian ecosystems (Brady
et al. 1985). Flooding also provides
the physical mechanism by which
large debris piles are built (Jones and
Glinski 1985). Water impoundment
structures on the Salt River appear to
prevent flooding regimes necessary
to maintain cottonwood reproduc-
tion and debris piles.

Over the past 10 years, the major
emphasis in riparian management
has been to manage trees, particu-
larly cottonwoods. Several tech-
niques, such as planting live trees
and tree poles, have been used on
drainages with major water im-

Similarity

poundment structures to improve
reproduction and survival of cotton-
woods (Swenson and Mullins 1985).
Although these techniques generally
increase nesting habitat for birds,
they do not provide enough surface
litter to support litter-dwelling spe-
cies, such as upland herpetofauna.
Szaro and Belfit (1986) studied a arti-
ficially created stand of riparian
vegetation on Queen Creek in south-
central Arizona. This stand of mostly
willows resulted from accumulation
of water behind a dike. Although the
stand emulated vegetation structure
of natural riparian sites, it had a
depauperate herpetofauna, even after
20 years.

This study suggests surface litter
is important in determining abun-
dance and diversity of herpetofaunas
in riparian communities. If we are to
conserve riparian ecosystems, we
must increase our emphasis on pro-
tecting all habitat components, in-
cluding microhabitats such as surface
litter. Like the Salt River site, riparian
areas will loose litter-dwelling and

0 .2 .4 .6 .8 1.0

Sonoran Desert

Salt River

Hassayampa River

Figure 7.— Dendrogram comparing herpetofaunas of the Sonoran Desert and Salt and Has-
sayampa Rivers.

226

mesic-adapted species unless we
consider these other components.

Acknowledgments

I thank Pattie Glinski, Scott Belfit,
Richard GUnski, Chuck Hunter, John
McConnaughey, Dan Abbas, and my
son Justin Jones for helping with data
collection. Special thanks to Dan
James, Mike Bender, James P.
Collins, and David J. Germano for
review of this manuscript.

Literature Cited

Brady, Ward, David R. Patton, and
Jay Paxson. 1985. The develop-
ment of Southwestern riparian
gallery forests. U.S. Forest Service
Gen. Tech. Report No. RM-120. p.
39-43.

Brode, John M. and R. Bruce Bury.
1984. The importance of riparian
systems to amphibians and rep-
tiles, p. 30-36. In: R.E. Warner and
K.M. Hendrix (eds.), California
riparian systems: ecology, conser-
vation, and productive manage-
ment. Univ. California Press,
Berkeley. 1035 p.
Germano, David J. and C. Roger
Hungerford. 1981. Reptile population
changes with manipulation of Sono-
ran Desert shrub. Great Basin Nat.
41(1):129-138.
Hair, Jay D. 1980. Measurements of
ecological diversity, p. 269-275. In:
S.D. Schemnitz (ed.). Wildlife
Management Techniques Manual.
The Wildlife Society, Washington,
D.C.

Jones, K. Bruce. 1986. Amphibians
and reptiles, p. 267-290. In: A.Y.
Cooperrider, R.J. Boyd, and H.R.
Stuart (eds.). Inventory and moni-
toring of wildlife habitat. U.S. Bu-
reau of Land Man., Denver, Colo-
rado xviii. 858 p.

Jones, K. Bruce and Patricia C.

Glinski. 1985. Microhabitats of liz-
ards in a southwestern riparian
community, p. 355-358. In: R. Roy

Johnson et. al.. Riparian ecosys-
tems and their management: rec-
onciling conflicting uses. First
North American riparian confer-
ence. Rocky Mountain Forest and
Range Experimental Station, Gen-
eral Technical Report Number
RM-120., Fort Collins, Colorado.

Jones, K. Bruce, Lauren P. Kepner,
and Thomas E. Martin. 1985. Spe-
cies of reptiles occupying habitat
islands in western Arizona: a de-
terministic assemblage. Oecologia
66:595-601.

Mueller-Dombois, Dieter and Heinz
Ellenberg. 1974. Aims and meth-
ods of vegetation ecology, p. 110-
118. John Wiley and Sons, New
York.

Ohmart, Robert D., Wayne O. Dea-
son, and C. Burke. 1977. A ripar-
ian case history: the Colorado
River, p. 35-46. In: Importance,
preservation and management of
riparian habitat: a symposium.
U.S. Forest Service Gen. Tech. Re-
port RM-43, Fort Collins, CO.

Pimental, Roger A. 1979. Morphom-
etries: the multivariate analysis of
biological data. Kendall/Hunt
Publ. Co., Dubuque, Iowa.

Swenson, E.A. and Charles L.

Mullins. 1985. Revegetating ripar-
ian trees in Southwestern
floodplains. p. 135-138. In: R. Roy
Johnson et. al.. Riparian ecosys-
tems and their management: rec-
onciling conflicting uses. First
North American riparian confer-
ence. Rocky Mountain Forest and
Range Experimental Station, Gen-
eral Technical Report Number
RM-120., Fort Collins, Colorado.

Szaro, Robert C. and Scott C. Belfit.
1986. Herpetofaunal use of a des-
ert riparian island and its adjacent
scrub habitat. J. Wildl. Man.
50(4):752-761.

Szaro, Robert C, Scott C. Belfit, and
J. Kevin Aitkin. 1985. Impact of
grazing on a riparian garter snake,
p. 359-363. In: R. Roy Johnson et.
al.. Riparian ecosystems and their
management: reconciling conflict-
ing uses. First North American

riparian conference. Rocky Moun
tain Forest and Range Exf)erimen
tal Station, General Technical Re-
port Number RM-120., Fort
Collins, Colorado.

227

Critical Habitat, Predator
Pressures, and the
IVlanagement of Epicrates
monensis (Serpentes:
Boidae) on the Puerto Rico
Banl<: A IVIultivariate Analysis^

Abstract.— Ep/crofes monensis is a endangered
boa endemic to the Puerto Rico Bank. Principal
components anaiysis, based on data collected
during five years of study and 200 captures of this
species, was used to identify predator, prey, and
habitat variables critical to survival of the snake.
Management recommendations ore discussed.

Peter J. Tolson^

Epicrates monensis is a small (ca. < 1
m snout-vent length) senni-arboreal
boid snake (fig. 1) that exhibits an ex-
tremely disjunct distribution on the
Puerto Rico Bank. The Mona boa (E.
m. monensis) is endemic to Isla Mona,
a large island in the Mona Passage
between Hispaniola and Puerto Rico
(Schmidt 1926). The other subspecies,
the Virgin Islands boa (E. m. granti),
is found on scattered islands and
cays from La Cordillera eastward
through the Virgin Islands, including
St. Thomas, Tortola, and Virgin
Gorda (Shill 1933; Nellis et al. 1984;
Mayer and Lazell 1988). The boa is
apparently absent from Puerto Rico
and the other large islands on the
bank. Judging from the present dis-
tributions, the historical range of Ep-
icrates monensis encompassed virtu-
ally the whole length of the Puerto
Rico Bank. Today, unfortunately, the
snake is endangered (USFWS 1980)
and absent from far more islands on
the bank than it is resident — doubt-
less the result of a long history of ex-
tirpation. It is improbable that the
decline of the boa can be traced to a
single causative factor; more likely
the survival of the snake at certain
localities is due to a complex series of
biotic, environmental, and stochastic

'Paper presented at symposium. I^an-
agement of Amphibians, Reptiles, and
Small Mammals in Northi America. (Flag-
staff, AZ July 19-21, 1988.)

'Peter J. Tolson is Curator of Amphibians
and Reptiles, Toledo Zoological Society.
2700 Broadway, Toledo, OH 43609.

interactions. The rarity of the snake
has made habitat analysis difficult;
one cannot define critical habitat if
the snake cannot be observed. Prior
to my work, fewer than 13 specimens
of the boa had been encountered, and
habitat descriptions were largely an-
ecdotal with no attempts to quantify
those factors important in determin-
ing population levels (Div. of Fish
and Wildlife, USVI 1983; USFWS
1984, 1986).

The parameters dictating the dis-
tribution and abundance of animal
species within a habitat are often di-
verse. They include not only the
physical structure of habitat, such as
vegetational composition and spatial
heterogeniety (Rotenberry and Wiens
1980), but also species composition
(Matthews 1985; Moulton, 1985) and
other aspects of community structure
which are less easily defined, such as
competition (Cody 1974) or preda-
tion pressure. In the West Indies,
particularly on the Puerto Rico Bank,
utilization of a particular habitat by
the endemic herpetofauna is not only
dependent on the structural attrib-
utes of vegetative cover and the com-
position of the endemic animal com-
munities, but also on the number and
severity of feral and exotic animal
introductions that have occurred.
Colonizations (accidental or other-
wise) of the roof rat, Rattus rattus, the
house cat, Felis catus, and the
mongoose, Herpestes auropunctatus,
have profoundly influenced the sur-
vival and distribution of endemics on

Figure 1 .—Epicrates monensis granti.
Above— adult female, Cayo Diablo, Puerto
Rico. Below—juveniles bom at the Toledo
Zoological Gardens 14 July 87.

the Puerto Rico Bank (Barbour 1917,
1930; USFWS 1986; Div. of Fish and
Wildlife, USVI 1983).

Principal components analysis
(PCA) is a multivariate statistical
technique that has been used by com-
munity ecologists to model distribu-
tions of animal populations in a
multidimensional habitat space de-
fined by a correlation matrix of habi-
tat variables (See Wiens and Roten-
berry 1981 and Matthews 1985). My
current work with Epicrates monensis
utilizes PCA to correlate the abun-

228

A

Cayo fcacos

\

Cayo Lobos

La Cordillera

t8'19'

Cayo Diablo
1

N i 1

1 km

Figure 2.— Location of sampiing plofs in Puerto Rico. Above— plots on Isia Mono. Beiow— plots
on La Cordillera.

dance of the boa with certain critical
elements of habitat structure and in-
dices of p>opulation densities of pre-
ferred prey species and predators.
Compilation of such data is ex-
tremely important in establishing the
critical dimensions of the boa niche,
the identification of suitable release
sites for snakes bom in captivity, and
the selection of likely search localities
for surveys of previously unde-
scribed populations of the snake. By
using PCA, we also hoped to extract
indep)endent patterns of covariation,
such as the degree of niche overlap
with Alsophis, which might explain
certain distributional anomalies of
the boa populations.

Methods

Study Areas

This study is based on habitat analy-
sis of 24 different localities on the fol-
lowing islands and cays of the Puerto
Rico Bank: Buck Is., Cas Cay, Cayo
Diablo, Cayo Icacos, Cayo Lobos,
Congo Cay, Great St. James Cay, Isla
Mona, Outer Brass Cay, Salt Cay,
Saba Cay, and Steven Cay from Feb-
ruary 1986 through April 1988. Some
islands had several plots. Sites were
chosen at random without regards to
presence or absence of boas, but an
attempt was made to select sites so
that sampling included the full spec-
trum of habitat available to the boa.
Figures 2 and 3 illustrate the location
of sampling plots included in the
study.

Vegetational Profiles of Study Sites

Subtropical dry forest is the habitat
where E. monensis is most commonly
observed, particularly on Isla Mona
and St. Thomas. It is characterized by
small (< 5 m) deciduous trees with
small, coriaceous or succulent leaves
and thorns, spines, and secondary
defensive compounds (Ewel and
Whitmore 1973). Examination of the

229

present range of the boa indicates
that it matches the occurrence of dry
subtropical forest on the Puerto Rico
Bank (Ewel and Whitmore 1973).
This is most apparent on St. Thomas,
where E. monensis is restricted to the
dry eastern end of the island despite
presumably suitable habitat else-
where (Nellis et al. 1984). Common
tree species include Burseria
simaruba, Cephalocereus royenii,
Pidetia aculeata, Bucida buceras,
Guaiacum officinale, Leucaena glauca,
Tamarindis indica, Melicoccus bijuga-
tus, Acacia ssp., and Capparis cynoph-
allophora (Little and Wadsworth
1964). In addition, on our dry forest
plots (Cas, Icacos 1, Congo 1, Outer
Brass 1, and Gt. St. James 1), we en-
countered many Byrsonima lucida,
Euphorbia petiolaris, and Metopium
toxiferum. On Buck 1, Diablo 1, Gt. St
James 3, and Mona 2 the vegetation
consisted of tree species with com-
pound trunks, primarily Coccoloba
uvifera, Hippomane mancinella, and
Thespesia populnea. Sabal palm groves
were present on Outer Brass 2 and
Salt 2. Salt-tolerant shrublands pri-
marily composed oiSuriam and
Tournefortia just above the high tide
line was the dominant vegetation on
Diablo 2, while Diablo 3 primarily

consisted of Cassythia/Opuntia
tangles, f icws-dominated forest was
present on Mona 1 and Congo 2.
Guinea grass, Panicum maximum,
dominated the transect on Buck 2
and Acacia macracantha on Buck 3. A
basic summary of the vegetation of
the smaller cays is given in Heatwole
et al. (1981). Figures 4 through 7 il-
lustrate four typical vegetational
types at transect sites: Coccoloba
grove (Buck 1), mixed palm/ shru-
bland (Diablo 2) Opuntia/Cassythia
tangles (Diablo 3) and grassland
(Buck 2).

Geomorphology and Topography
of Study Sites

Geomophology of the various islands
and cays studied varied considera-
bly, from the steep-sided metamor-
phic topography of St. Thomas and
associated cays (Heatwole et al. 1981)
to the cemented dune structure of La
Cordillera (Kaye 1959a). Isla Mona is
composed primarily of a Pleistocene
limestone plateau surrounded by
sheer cliff (Kaye 1959b). In fact, most
islands of the bank have significant
limestone deposits, with varying
amounts of metamorphic rock, in-

Outer Brass Is

St. Thomai"^ Steven Cay

t8»20

Congo Cay

Saba Cay

U. S. Virgin Islands

,65*

Figure A.— Coccoloba uvifera habitat on
Buckl.

Figure 5.— Mixed Cocos and scrubland
hiabitat on Cayo Diabio. Ttie vegetation at
thie center of \he island is primarily Cas-
sythla vine growing over Opuntia cactus.

Figure 6.— Aromatic beactifrontshirubiand,
primarily Surlana and Tournefortia, near
Diablo 2.

Figure 3.— Location of sampling plots, U.S. Virgin Islands.

Figure 7.— Guinea grass, Panicum maxi-
mum, tiabitat on Buck 2.

230

eluding gneiss and basalt, present as
well. The cays of La Cordillera are
exceedingly low, with maximum ele-
vations under 15 m. In the Virgin Is-
lands the cays are of moderate eleva-
tion with eroded limestone hills ap-
proaching 50-300 m in height. An
overview of the geology of the Virgin
Islands is given in Schuchert (1935).

Climate

The climate of the Bank is essentially
subtropical to tropical. Temperatures
of the coastal areas range from over-
night lows of ca. 15° C to daytime
highs approaching 35° C. Rainfall,
especially on Puerto Rico, is geo-
graphically variable (Briscoe 1966).
Areas within the range of E. monensis
typically receive < 750 mm of rainfall
per year.

Sampling Techniques

The presence or probable absence of
the boa on a particular cay was de-
termined by active searching of all
habitat types during surveys (carried
out independently of habitat analy-
sis) from April 1983 to September
1987. Typically 2 weeks or more were

spent searching larger islands and
three to five days for smaller cays.
Only 1 night was spent on Cayo Lo-
bos, as the native vegetation was all
but completely destroyed by human
activity and all densely vegetated ar-
eas could be searched repeatedly in a
single night. Our experience with
multiple recaptures of the same indi-
vidual indicates that the snakes for-
age every night under most circum-
stances. Within each 24-hour period
4 hours per night were spent search-
ing likely foraging sites such as vine
tangles, terminal branches of trees,
palm crowns, and beachfront vegeta-
tion. Ehiring the daylight hours, refu-
gia sites such as debris piles, termite
nests, and palm axils were exanrdned.
After capture, the time, capture
height, habitat description, ambient
temperature, refugium temperature,
and cloacal temperature of each
snake were recorded. Later, sex,
body mass, snout-vent length (SVL),
and caudal length (CL) were re-
corded. The snakes were examined
for reproductive condition, presence
of injuries, and parasite infestation.
Snakes were marked using the tech-
nique of Brown and Parker (1976)
and released at the point of capture.

Habitat variables recorded in-
cluded both physical and biological

parameters (table 1). Predator den-
sity estimates include indices of
abundance for likely predators of E.
monensis: the roof rat, Rattus rattus,
the pearly-eyed thrasher, Margarops
fuscatus, and the Puerto Rican racer,
Alsophis portoricensis. Rattus densities
were estimated using removal trap-
ping over a 3-day span on 100-m
transects with Victor snap traps
spaced every 5 m. Presence of Felis
catus was determined by direct ob-
servation. Because of the extreme
wariness and trap-shy nature of the
Felis on study plots, only their pres-
ence or absence was recorded.

Prey density data includes of
population densities for Anolis cris-
tatellus and Ameiva exsul. Anolis, Also-
phis, Ameiva, and Margarops were
counted by having two observers
slowly walk the transects and count-
ing the individuals of each species
observed within a 5 m distance on
each side of the transect line. On
Cayo Diablo, independent estimates
of Ameiva and Anolis cristatellus
populations were gathered by sur-
veys of 5 m^ quatrats. Anolis cristatel-
lus perch heights were measured
with a metric tape except on Cayo
Lobos and Salt Cay. Canopy height
was estimated for each habitat with
the help of a metric tape. Vegetative
composition was determined by sub-
jective stratified sampling using 10
m^ quadrat plots (Clarke 1986); plant
samples were taken for species iden-
tification from each island. Vegeta-
tion coverage data indicates the per-
centage composition of five different
classes of vegetation: trees (trunk cir-
cumference at shoulder height > 25
cm), palms, Opuntia cactus, shrubs
and small trees (trunk circumfer-
ences < 25 cm), and grasses. Vegeta-
tion structural data includes the
number of dominant plant species,
the height of the canopy, and the
continuity of the vegetation (a meas-
ure of the difficulty for the boa to
crawl from one plant to another
without going to the ground). Plants
were identified by David W. Nellis
and the author.

r

Table L— Factor patterns of the original variables on each of the first six
principal components.

Principal component

Variable

1

11

III

IV

V

VI

Rat density

-0.6795

-0.3582

0.0162

-0.2453

0.4298

-0.0379

Cot presence

0.4976

0.1703

0,0895

-0.4592

-0.3190

-0.2217

Racer density

-0.4702

0,0838

-0,1051

0.5491

-0.1627

-0.0437

Thrasher density

-0.5268

0,2052

0,0410

0,0091

-0.6211

0.4070

Anolis density

-0.1400

0.3391

0.4005

0.4582

0.1560

-0.4850

Ameiva density

-0.0657

0.6091

-0,5302

0.3176

-0.0890

-0.0628

/Anofe perch height

0.7972

0.1830

-0.1991

-0.0386

0.2946

-0.0233

Compound tree density

-0.0810

0,7175

-0,4546

-0.2118

0.2106

0.0407

Singletree density

-0.5156

-0.0546

0,3885

0.1877

0.5079

0.2290

Palm der^ity

0.4478

0.1154

0.3690

-0.5638

-0.2063

0.0771

Shrub der-isily

0.6215

-0.3240

0,4724

0.4028

-0.1720

-0.0718

Grass density

-0,1656

-0.5586

-0.5118

-0.1659

-0,0554

-0.4613

Cactus density

0,3458

-0.3437

-0.1861

0.3130

0.1391

0.6288

Vegetational continuity

0.6882

0.3989

-0.0270

-0,1124

0.3701

0.0988

Canopy height

-0.3978

0.7730

0,4023

-0.1372

0.0139

0.0257

231

I attempted to use continuously
distributed standarized environ-
mental variables whenever possible.
Absence of a particular predator or
prey species on a given a sample plot
did not always indicate its absence
from the island on which the plot
was situated. Only male Anolis perch
heights were used for the statistical
analysis, as female and juvenile A.
cristatellus tend to frequent the
ground under all circumstances (Ki-
ester et al. 1975). Mean male Anolis
perch height data were pooled for
each island for character 16 of the
PCA data matrix, as some plots were
completely devoid of Anolis.

Statistical Analysis

Principal components analysis was
performed using the Statistical
Analysis System '"SAS" release 5.16
(SAS Institute 1985). Significant habi-
tat components, which included both
biotic and structural variables of the
collecting localities (e.g. those which
accounted for > 10% of the total vari-
ance in the data), were clustered on
the basis of their association within
the PCA data matrix. The second
step of the analysis compared the
relative abundance of E. monensis at
each collecting locality with habitats
described by the significant axes of
the principal components. Regression
analysis, ANOVA, and descriptive
statistics (mean, standard deviation,
etc.) were performed using Statview
512^ on an Apple Macintosh Plus.

Results

Multivariate Analysis of Habitats

The PCA indicates that biotic factors,
plant composition, and structural at-

^7he use of trade and company names
is for the benefit of the reader: such use
does not constitute an official endorsement
or approval of any service or product by
the U. S. Department of Agriculture to the
exclusion of others that may be suitable.

tributes of vegetation are all impor-
tant contributors to variance in the
PCA patterns. Factor patterns for the
first six principal components are
given in table 1. Principal component
I accounts for 23.4% of the variance.
This component clusters habitats
with high shrub and palm densities,
low numbers of single trees, vegeta-
tional continuity, and low canopy
heights. Important biotic characteris-
tics of this space include Felis pres-
ence and low Rattus, Margarops, and
Alsophis densities with high Anolis
perch heights. Principal component II
accounts for an additional 17.0% of
the variance observed. This axis de-
scribes sites having low grass den-
sity, high compound tree densities,
canopy height > 3 m , and high
Ameiva densities. Principal compo-
nent III accounted for 11.2% of the
variance and suggested an associa-
tion between low Ameiva density,
low compound tree density, low
grass density, high shrub density,
and canopy height > 3 m. Factor IV
accounted for another 10.7% of the
variance and clustered high Alsophis
and Anolis densities with Felis ab-
sence and low palm density. Compo-
nents V-VI were less significant in
the PCA (e.g. each accounted for < 10
% of the variance) but added some
interesting ecological information to
the habitat analysis. Principal compo-
nent V clustered high Rattus density
with low Margarops density; princi-
pal component VI grouped high Mar-
garops density with low Anolis den-
sity.

Habitat Utilization by Epicrafes
monensis

The vegetational profiles of climax
plant communities (and E. monensis
collection localities) in the dry forest
may differ considerably depending
on island size, geology, geomorphol-
ogy, rainfall, and history of human or
feral mammal disturbance. However,
most dry forest habitats on the Bank
are structurally simple, with usually

only two to five dominant plant spe-
cies (table 2). Captures and sightings
of the Mona boa have been limited to
three distinct localities: dry plateau
forest adjacent to Uvero and Pajaros
(Campbell and Thompson 1978; Riv-
ero et al. 1982) Coccoloba uvifera
groves of Pajaros (M. Frontera, Pers.
Comm.), and Cocos groves and
nearby vegetation adjacent to Playa
Sardinera (G. Rodriguez pers.
comm.). The Virgin Islands boa has
been encountered repeatedly on only
two islands: St. Thomas and Cayo
Diablo. All specimens from St. Tho-
mas were captured on the east end of
the island near Red Hook. Two speci-
mens were found beneath a lime-
stone slab during construction of the
Vessup Bay Estates housing subdivi-
sion, another was taken from a stone
wall, and a third was found as a
roadkill near Smith Bay. R. Thomas
captured a specimen crawling in a
viney tangle ca. 2.4 m high (Sheplan
and Schwartz 1974).

The Red Hook area is dominated
by xeric forest composed primarily
of Burseria, Croton, and Acacia. No
habitat data is available for E. m.
granti on Tortola. I have received re-
ports that the boa was present in the
palm forest of Outer Brass Island (J.
LaPlace pers. comm.) but I was un-
able to find it there even after five
trips to the island. Virgin Islands
residents also report the boa as in-
habiting Great St. James Is. (D. Nellis
pers. comm.). Great Camanoe,
Necker Is., and Virgin Gorda, (Mayer
and Lazell 1988), but these sightings
have not been confirmed by biolo-
gists. Grant (1932) mentioned anec-
dotally (he did not capture the
holotype himself) that "the boa is
found on rocky cliffs on Tortola and
Guana Islands."

On Cayo Diablo, Coccoloba uvifera
is the habitat most commonly associ-
ated with foraging E. monensis. Of the
79 active snakes we captured, 51
were found in Coccoloba, ten in Cae-
salpinea, nine on Cassythia, seven in
Suriana, and two in Opuntia. Twenty-
three percent of the snakes were ac-

232

tive at heights > 2 m. Of these, 67%
had SVLs > 400 mm. Seventy-five
percent of juvenile snakes (under 300
mm SVL) foraged at heights < 1.5 m,
but regression analysis indicated that
these differences were not statisti-
cally significant. Of the 149 inactive
snakes taken from refugia, 43% were
in Cocos or Sahal axils, 36% were in
termite nests, and 21% were under
rocks or debris. Fifty-one percent of
snakes taken from termite nests were
females; over half of these were
gravid. Gravid females use termite
nests or sun-baked debris to ther-
moregulate and may elevate their
body temperatures to over 33° C.

Prey Density and Epicrate$
monensis Distributions

The greatest concentrations of Ep-
icrates monensis are in areas
(particularly Coccoloha groves) with
AnoUs densities > 60 Anolis/ 100 m^.

This Anolis/Epicrates association is
reinforced by PCA (see below). My
field logs indicate that the greatest
success in finding foraging Epicrates
occurs when observations of sleeping
Anolis are > 12 lizards/ h. Numerical
counts of sleeping Anolis and the
times between sightings are regularly
noted in my field book as a rough
guide to potential hunting success in
a study locality.

AnoUs cristatellus is the primary
prey species of E. monensis, and the
mean foraging height of the snake (x
= 1.356, SD = 1.079 N = 54) is close to
the mean perch height of sleeping
Anolis (x males = 1.816 m, SD = 0.993,
N = 17; X females = 1.323 m, SD =
.681, N = 14; X juveniles = 1.417 m,
SD = 0.169, N = 5).

High Ameiva densities are also a
common component of localities
with high boa densities, although I
observed only one instance of a boa
feeding on Ameiva, which are
strongly diurnal.

Feral Mammal Abundance and
Epicrates monensis Distributions

Of the 10 islands surveyed for this
study, only three were completely
devoid of rats: Cayo Diablo, Cayo
Icacos, and Steven Cay. These islands
have high Ameiva and Anolis densi-
ties, but only Diablo Cay harbors a
population of the boa. It also has the
highest densities of Epicrates monensis
found anywhere on the bank, > 100
snakes/hc at some localities. Those
islands with heavy rat densities (ca.
20 rats/ hectare) — Buck Is., Cas Cay,
and Salt Cay — have lower Ameiva and
Anolis densities and apparently no
boa populations, despite suitable
habitat. Rat densities are not always
correlated with low Anolis densities,
however. Some islands, such as
Outer Brass and Congo, have Anolis
densities apparently high enough to
support populations of the boa, but
their perch heights (table 2) are sig-
nificantly different from those Anolis

Table 2.— PCA habitat matrix for the Puerto Rico Bank.

Rat Cat Racer Thrasher Anolis
dens, presence dens, abund, dens.

Ameiva Comp,
dens, tree
dens.

Single
tree
dens.

Palm
dens.

Shrub/
small
tree
dens.

Grass Opuntla Contlg. Plant
dens, dens/ veg. diver,
cover

Canopy Anolis
height perch
height

; Cayo Diablo 1

0.00

0

0.00

0,00

1,48

1,50

0.99

0.00

0.01

0.00

0,00

0,00

1.00

1

1,00

1.70

Cayo Diablo 2

0.00

0

0.00

0.00

0.25

0.05

0,00

0.00

0,03

0,95

0,00

0,00

1.00

2

0,50

1.70

Cayo Diablo 3

0.00

0

0.00

0.00

0,00

0.10

0,00

0.00

0.00

1,00

0.00

0,60

1.00

3

0.26

1.70

Cayo icacos 1

o.oc

1

0.00

0.05

0,36

0.76

0,46

0.06

0.00

0,50

0.00

0,00

0.75

4

1.00

1,57

Cayo icacos 2

0.00

1

0.00

0.00

0.91

0.00

0,14

0.10

0.13

0.63

0.00

0,00

0.60

6

1.00

1.57

i Cayo Icacos 3

0,00

1

0,00

0.00

0,28

0.00

0,02

0.00

0.00

0,98

0.00

0,00

0.75

4

0,60

0.50

; Cayo Lobos

0.08

0

0.00

0,20

0.03

0.02

0,75

0.10

0.10

0.16

0,00

0,00

0.26

3

1.00

0.18

Congo Cay 1

0,08

0

0.50

0.10

2.10

0,00

0,23

0,01

0.00

0.76

0.00

0,00

0.10

1

1.00

0.18

Congo Cay 2

0.09

0

0.60

0.08

2.00

0.00

0.04

0,42

0.13

0.46

0.00

0,00

0.25

4

1.00

0,32

Outer Brass 1

0.04

0

2.60

0.30

1.16

1,00

0.69

0,07

0.00

0.35

0.00

0.00

0.50

2

1,00

0.32

Outer Brass 2

0.02

0

0.50

0.73

0.36

0,60

0.00

0,10

0.86

0.00

0.00

0,00

0.00

2

1.00

1.00

Salt Cay i

o.n

0

0.33

0.10

0.42

0,16

0.90

0.10

0.00

0.00

0.00

0.00

1,00

5

1.00

1.00

Salt Cay 2

0.18

0

0.67

0.05

0,84

0,08

0.00

1.00

1.00

0.00

0.00

0.00

0.60

1

1.00

0.67

Isia Mono 1

0,02

0

0.00

0.21

0,58

0,00

0.14

0.15

0.00

0.71

0.00

0,00

1.00

5

1.00

0.67

Isia Mono 2

0.02

1

0.00

O.n

0.25

0,00

0.58

0.07

0.10

0.26

0.00

0.00

0,76

3

1,00

0.42

Gt, St. James 1

0.06

0

0.50

0,10

0,68

0,00

0,68

0.29

0.00

0.00

0.00

0,13

0,26

1

1.00

0.42

Gt, St. James 2

0.06

0

1.00

0.16

0.53

0,00

0,00

0.27

0.00

0.68

0.00

0.05

0.25

3

1,00

0.42

Gt, St. James 3

0.05

0

0.35

0.15

0.00

0,00

0,27

0.45

0,00

0,24

0.00

0.04

0,50

4

1.00

0.60

Buck Is. 1

0.04

0

0.00

0.00

0.60

0,00

0,28

0.06

0.00

0,66

0.00

0,00

1,00

4

1.00

060

Bucl< Is, 2

0.13

0

0.33

0,00

0.00

0,00

0.00

0.00

0,00

1.00

0.00

0,00

0,25

2

0.50

0.50

Bucl< Is. 3

0.10

0

0.33

0,00

0.05

0,00

000

0.00

0.00

0.00

1.00

0.00

0,00

2

0.26

0.50

Steven Cay

0.00

0

0.00

0,01

2.79

0,00

0.16

0.08

0.00

0.76

0.00

0.00

1,00

3

1.00

1.26

Saba Cay

0.00

0

2.00

0.00

1.06

0,00

0.08

0.08

0.00

0.84

0,00

0.00

1,00

2

1.00

1.02

Cos Cay

0.04

0.00

0.00

0.0476

0.12

0.00

0.78

0.09

0.00

0.13

0,00

0.00

1,00

3

1.00

1.40

'See appendix A for variable descriptioiis.

233

inhabiting rat free islands. ANOVA
performed on the regression line (y =
-5.548x + 1.127) which plots Anolis
perch height vs. rat density on my
study islands (Tolson and Campbell
in prep) shows a negative correlation
(p = .0137) between rat density and
Anolis perch height. This is not
surprising. Anolis cristatellus resident
on rat-infested islands exhibit a typi-
cal escape behavior. Male Puerto
Rican A. cristatellus escape to the can-
opy when threatened (Heatwole
1968), but those on Congo Cay, Outer
Brass, and Salt Cay all run to the
ground when disturbed, even when
suitable cover on the ground is lack-
ing. At night, the Anolis are not usu-
ally found sleeping exposed on vege-
tation, but rather under rocks. This is
extremely unusual behavior for A.
cristatellus (E. Williams pers. comm.).

Although one does not often
discover E. monensis on islands
which are infested with rats, some
sympatry does occur. Isla Mona and
St. Thomas are islands with moder-
ate rat densities and extant (although
apparently dwindling) populations
of E. monensis. Interestingly, at locali-
ties where Epicrates coexists with Rat-
tus, there are also significant num-
bers of introduced mammalian
predators such as Felis and Herpestes
(table 2).

Discussion

PCA and E. monensis Habitat
Utilization

The Puerto Rico Bank encompasses a
total land area in excess of 9,300 km^,
of which 1700+ km^ (or 17.6%) is cov-
ered with subtropical dry forest
(Ewel and Whitmore, 1973). This
xeric forest is widely distributed
throughout the range of Epicrates
monensis, yet the boa, as far as we
know, occupies only seven islands of
the 243 that make up the banks — ef-
fectively exploiting only 0.04% of the
land area available to it. PCA helped
to identify those factors which seem

to define critical boa habitat. Several
vegetative parameters which cluster
together in the PCA are descriptive
of habitat where I or others have
encountered E, monensis repeatedly.
These include areas with high shrub
and palm densities coupled with a
low canopy and vegetational conti-
nuity. These values describe plot
habitat on Diablo 2, Icacos 2, and cer-
tain sites within the Red Hook area
of St. Thomas. Either high shrub or
high palm densities coupled with
vegetational continuity and lower
canopy are found on Diablo 3, Icacos
3, and Mona 1. Of these two subsets
of PC I, boas occur on Diablo 2 and 3,
Mona 1, St. Thomas, and almost cer-
tainly inhabited Icacos 1 and 3 at one
time.

In PC II, habitat correlates include
high compound tree density, high
canopy height, vegetational continu-
ity, and low grass density. This is a
perfect structural and compositional
description of Diablo 1, which has
the highest population of E. monensis
I have ever encountered, and Mona
2 — another locality where E. monensis
has been observed (Campbell and
Thompson 1978). It seems clear from
these data that the unifying variable
which causes an intersection of these
two differing habitat types is vegeta-
tional continuity — an interlocking of
the branches of shrubs or the tree
canopy. I believe this vegetational
characteristic is essential to E. monen-
sis foraging success and survival. It
probably not only decreases the
search time between encounters with
sleeping Anolis while foraging, but it
also potentially limits the encounters
between the boa and Felis and Her-
pestes. Fortunately, at least some
tracts of subtropical dry forest and
Coccoloba have remained relatively
undisturbed on the Virgin Islands,
Isla Mona, and Puerto Rico and its
offshore satellites. Much suitable
habitat does exist — even near popu-
lated areas.

While habitats throughout the
Bank are presumably underutilized
by E. monensis, and suitable areas for

reintroduction apparently exist in a
number of localities, the extant boa
populations are so fragmented and
reduced in numbers that it is crucial
to protect those areas now support-
ing the boa. This may be difficult.
Historically, vegetation on Puerto
Rico and the Virgin Islands has been
severely disrupted, and 17th-18th
century land use patterns on the U.S.
Virgin Islands may partially explain
the limited distribution of the boa on
the east end of St. Thomas and its
absence from St. John. Even now
enormous pressures exist for contin-
ued development on the east end of
St. Thomas. Construction around
Red Hook seems to have accelerated
in recent months, perhaps in re-
sponse to the decline of interest rates
in the United States, and three rela-
tively undeveloped areas on the east
end — Red Hook Mountain, Cabrita
Point, and Water Point — all have proj-
ects in progress that do not involve
federal funding. The management
authority on St. Thomas, U.S. Virgin
Islands — the Division of Fish and
Wildlife — has no control over such
development.

In contrast, Puerto Rican islands
with populations of Epicrates monen-
sis are in no imminent danger of de-
velopment. Cayo Diablo is part of the
Reserva Forestal de La Cordillera,
and Isla Mona is likewise a Forest
Preserve (although it was once pro-
posed to develop the island as a
deep-water oil fX)rt). A problem does
exist, however, with habitat destruc-
tion on isolated cays caused by
campers and fishermen (Heatwole
and Mackenzie 1967). Coccoloba trees
in the larger groves — areas where the
greatest densities of E, monensis are
found — are often used as firewood by
visitors. A survey done in 1987 of
damage to Coccoloba stands on Cayo
Diablo showed that many trees sus-
tained some sort of damage caused
by human activity, primarily ma-
chete cuts and burns from fires
started at the bases of the trees.

234

Effects of Feral Mammals

My analysis shows that Rattus and
Felis are a primary influence on com-
munity composition on the Puerto
Rico Bank. Felis presence is associ-
ated with low Alsophis, Margarops,
and Rattus density (table 1: PC I); Fe-
lis absence is associated high Anolis
and Alsophis densities in PC IV (table
1). Clearly the presence of Felis in E.
monensis habitat is a mixed blessing.
Cats present a great danger to Ep-
icrates because they hunt at night.
Several instances of cat predation of
Epicrates have been reported on St.
Thomas (D. Nellis pers. comm.) In
fact, in April and May of 1988 two E.
monensis were rescued from cats on
St. Thomas and were incorf>orated
into the captive breeding program at
the Toledo Zoological Gardens. In
contrast, however, on islands where
boas and rats coexist — Isla Mona and
St. Thomas — there are also significant
populations of Felis. Cats feed on
Rattus and may keep rat populations
at levels low enough to permit sur-
vival of the boa. Their apparent ad-
verse affect on Alsophis and Marga-
rops density — two potential predators
of E. monensis — may also be of some
small benefit in certain circum-
stances. Weiwandf s (1977) observa-
tion of cat predation of Alsophis on
Isla Mona corroborate the PC I link-
age of cat presence with low Alsophis
density.

I cannot be certain whether Rattus
affect boa populations by acting pri-
marily as a constraint on their re-
source levels or by direct predation.
Although I have been unable to dem-
onstrate that rats forage on boas, I
have every reason to suspect that
they do. Rattus is known to prey on
lizards (Whitaker 1978). While sur-
veying for boa populations on the
Bank I found habitat (Congo Cay,
Outer Brass Cay) which provides op-
timal foraging opportunities for the
boa (e.g. vegetation associated with
population densities of > 60 Anolis/
100 m^ on rat-free islands) but had no
or few boas and were virtually over-

run with rats at night. Rats may also
affect boa populations by preying on
Anolis directly or by influencing their
perching behavior, (indicated by the
negative correlation between rat den-
sity and Anolis perch height (table 1:
PC I) or selection of sleeping sites. If
lizards rarely rest in the canopy at
night but rather seek refuge sites on
the ground, there would be poten-
tially disastrous consequences for
boa foraging success. Rattus also ap-
parently affect Margarops density
(table 1: PC V).

There can be little doubt that the
Indian mongoose, Herpestes auropunc-
tatus, threatens Epicrates monensis di-
rectly as well, but I believe the risk to
Epicrates is sometimes exaggerated.
Herpestes predation on endemic West
Indian snakes is well documented
(Maclean 1982), but the mongoose is
a strictly diurnal, terrestrial predator;
Epicrates monensis is nocturnal and
arboreal. Herpestes poses the greatest
danger to the diurnal West Indian
racers, genus Alsophis, and are di-
rectly responsible for the extinction
of Alsophis sancticrucis on St Croix
and the extirpation of A. portoricensis
from St. Thomas and St. John. In con-
trast, I have found Epicrates monensis
abroad during the daylight hours on
only two occasions over a period of
several years. It seems that Herpestes
would have the greatest chance of
capturing Epicrates when the latter is
resting in some moderately acces-
sible location during the day — in
loose sections of termite nests, for
example. Feral pigs (Sus scrofa) may
also threaten the Mona boa to some
degree, either by eating them or by
destroying vegetation, such as terres-
trial bromeliads, that may act as
snake refugia. I have no data on the
magnitude of this threat.

Natural Predators

The Puerto Rico Bank has no extant
species of native mammalian preda-
tors, but two nocturnal avian preda-
tory species may pose a limited

threat to Epicrates monensis. The yel-
low-crowned night heron, Nyctanassa
violacea, and the Puerto Rican screech
owl, Otus nudipes, are two potential
predators of the boa. While popula-
tions of Otus are declining on the
bank (lUCN 1981) those of the heron
seem quite stable. I have repeatedly
observed herons foraging at night in
boa habitat on both Isla Mona and
Cayo Diablo. Examination of the de-
bris beneath heron rookeries on Cayo
Diablo has revealed numerous frag-
ments of Anolis and Ameiva skin and
skeletal materials, usually ribs, verte-
brae, and jaw elements. No snake
remains have been found, but my co-
workers and I are continuing to in-
vestigate this potential problem. I
also found that Anolis densities and
perch heights are reduced (table 2)
on plots with high pearly-eyed
thrasher densities. In PC I (table 1)
high Anolis perch heights are associ-
ated with low thrasher density.
These birds also prey on Anolis, and
are so common in some areas they
could easily depress Anolis popula-
tion numbers. Principal component
VI (table 1) couples high thrasher
density with low Anolis density.

Two arthropods are potential
predators of E. monensis: the land
crab Gecarcinus and the hermit crab
Caenobita clypeatus. Searches of ter-
restrial refugia for Epicrates have re-
vealed that these snakes are nearly
always absent from areas occupied
by Gecarcinus and Caenobita. This is
especially true in termite nests.
Snakes only occupy areas of the nest
that are inaccessible to crabs. If
weathering or disturbance causes a
section of termite nest to become
habitable for crabs it is abandoned by
Epicrates, despite their prior use of
the refugium for several past field
seasons. In hundreds of examinations
of refugia over the past five field sea-
sons, I found Epicrates in association
with Caenobita on only one occasion: I
found a gravid female thermoregu-
lating under a discarded tarpaulin in
the midst of several Caenobita on 7
September 1987. Evidence for preda-

235

tion by the aforementioned species is
strictly circumstantial, but the fact
remains that over 17% of the Ep-
icrates captured have obvious
wounds, scars, or partially ampu-
tated tails. This is strong evidence
that some form of natural predation
is occurring.

Climatic/Stochastic Events

The apparent extirpation of the snake
from the majority of the islands on
the Bank relate not only to the arrival
of European man on the Bank and
the habitat destruction which fol-
lowed, but also to climatic, eustatic,
and stochastic events, many of which
had profound influences on habitat.
During the late Pleistocene several
climatic and eustatic events occurred
that apparently set the stage for the
decline of E. monensis on the Bank.
Foremost among these was a dra-
matic change in the climate of Puerto
Rico. From a relatively xeric climate,
Puerto Rico became progressively
more mesic during the late Pleisto-
cene. Today, over 81% of Puerto
Rico's vegetation is classified as
moist or wet forest (Ewel and Whit-
more 1973). Pregill (1981) and Pregill
and Olson (1982) describe the effect
this climatic change had on the xeric-
adapted Puerto Rican herpetofauna.
This extreme climatic shift may have
resulted in the extirpation of E.
monensis on Puerto Rico.* In addi-
tion, sea levels rose nearly 100 m
about 8,000-10,000 years ago and
separated the Virgin Islands from
one another and from Puerto Rico,
transforming what was a contiguous
land mass into a scattered series of
islets and cays spread over nearly
400 km. Many of these cays now
have extremely low elevations (Heat-
wole and Mackenzie 1967).

''it is unclear why E. monensis is absent
from the dry forest in southwestern Puerto
Rico. Hab 'ftat in the Guanica forest seems
quite suitable for the boa: perhaps further
survey work will result in its discovery there.

The fragmentation of E. monensis
into several small demes may have
left several populations without the
genetic resources to survive changing
environments, and doubtless allowed
stochastic processes such as disease,
prey fluctuations, or storms to extir-
pate many isolated populations. I as-
sume that the influences of random
events on the present distribution of
the native herpetofauna complicates
the multivariate analysis by introduc-
ing more variance into the correla-
tion matrix. These factors may ex-
plain the absence of snakes from is-
lets with suitable habitat, as some of
these islands may have inadequate
food resources or lower probabilities
of recolonization.

Management Recommendations

The forces threatening Epicrates
monensis are complex. Solutions for
the recovery of the boa will not be
simple, but I am optimistic about the
chances of success. My management
recommendations are summarized
below.

Saving Boa Habitat

This may be impossible on St. Tho-
mas, but with luck the boa may coex-
ist with man (as it now does) at some
relatively developed localities. Con-
tinued protection of Isla Mona and
La Cordillera are absolutely neces-
sary.

Continued protection and man-
agement should be extended to those
cays now protected by the Division
of Fish and Wildlife, U.S. Virgin Is-
lands— particularly Congo Cay, Outer
Brass Cay, Salt Cay, Savana Island
and Steven Cay — as these sites might
eventually be utilized for reintroduc-
tion programs. The smaller islands
should be off limits to casual visitors
to prevent habitat damage and hu-
man persecution of the snakes.

Predator Eradication on Suitable
Offsliore Islets

Rat control programs should be initi-
ated immediately on those islands
with habitat suitable for E. monensis.
Preliminary studies of rat eradication
using anticoagulant poisons on some
small cays near St. Thomas have pro-
duced promising results (Division of
Fish and Wildlife, USVI 1983). It is
critical, however, that time and fund-
ing be committed for follow up stud-
ies on any islands made the subject
for a rat control program. This must
be done to ensure that immunity to
poisons has not evolved or that
populations are being replenished by
recolonization from St. Thomas.

It is unlikely that Felis or Herpestes
will ever be eradicated from larger
islands such as Isla Mona or St. Tho-
mas, but Felis control programs now
in force on Mona should be contin-
ued to further reduce populations
and should be expanded to include
Cayo Icacos. It is important to con-
vince management authorities that
feral mammal control measures on
the Bank must be increased, and
quickly.

It is a credit to the evolutionary re-
silience of this little snake that it has
survived at all. Few endangered spe-
cies have been exposed to such a
wide range of adverse effects and
have still survived. It is my fervent
hope that this, and other endemic
species of the Caribbean, will not be
exterminated in the wake of the liv-
ing human debris, such as Rattus rat-
tus, that we have allowed to pollute
the islands of the West Indies.

Captive Breeding for
Reintroduction Purposes

Captive propagation can figure sig-
nificantly in the recovery of this
snake (USFWS 1986) The current co-
operative breeding plan for E. monen-
sis should be expanded to more
American Association of Zoological
Parks and Aquarium member institu-

236

tions, and Species Survival Plan des-
ignation should be sought for the
snake immediately to facilitate ge-
netic management of the captive
population.

For the present, until genetic
analysis has been completed, the
strategy of deme integrity mainte-
nance should be continued, with St.
Thomas founders and La Cordillera
founders managed as separate popu-
lations. Continuous outcrossing
within demes facilitated by a random
pair mating scheme should be en-
couraged. Fortunately, the first cap-
tive breeding has already taken
place, the proximate factors critical
to reproduction have been identified
(Tolson and Tuebner 1987), and there
is no reason why the captive popula-
tion cannot be expanded quickly for
reintroduction attempts within five
years.

I firmly believe that we are finally
at the point where we can look for-
ward to augmenting boa popula-
tions, rather than helplessly watch
them decline.

Acknowledgments

This research was conducted as part
of USFWS recovery activities under
the support of the Institute of Mu-
seum Services conservation program
(Grant IC-70095-87) and the Toledo
Zoological Society. I am extremely
grateful to Dr. David W. Nellis, Divi-
sion of Fish and Wildlife, U.S Virgin
Islands, Drs. Eduardo R. Cardona
and Jose A. Vivaldi, Departmento de
Recursos Naturales, Commonwealth
of Puerto Rico, and to Hilda Diaz-
Soltero and Robert Pace of the
USFWS Caribbean Field Office for
their counsel and logistical support
during the execution of this project.

I thank Earl W. Campbell III, Jorge
L. Pinero, and Carlos Diez for their
assistance in the field, which was of-
ten given under difficult conditions.
C. Ray Chandler and Earl W.
Campbell III aided in the statistical
analysis.

Literature Cited

Barbour, Thomas 1917. Notes on the
herpetology of the Virgin Islands.
Proceedings of the Biological Soci-
ety of Washington 30:97-104.

Barbour, Thomas 1930. Some faunis-
tic changes in the Lesser Antilles.
Proceedings of the New England
Zoological Club 11:73-85.

Briscoe, C.B. 1966. Weather in the
Luquillo Mountains of Puerto
Rico. U.S.D.A. Forest Service Re-
search Paper ITF-3. 250 p. Institute
of Tropical Forestry, Rio Piedras,
Puerto Rico.

Brown, William S. and William S.
Parker 1976. A ventral scale clip-
ping system for permanently
marking snakes (Reptilia, Serpen-
tes). Journal of Herpetology
10(3)247-249.

Campbell, Howard W. and Frederick
G. Thompson 1978. Observations
on a captive Mona Island boa, Ep-
icrates monensis monensis Zenneck.
Bulletin of the Maryland Herpeto-
logical Society 14(2)98-99.

Clarke, Robert 1986. The handbook
of ecological monitoring. 298 p.
Clarendon Press, Oxford.

Cody, Martin L. 1974. Competition
and the structure of bird commu-
nities. Monographs in Population
Biology 7. Princeton University
Press, Princeton, New Jersey.

Division of Fish and Wildlife, U.S.
Virgin Islands 1983. Rat control on
small tropical cays. 26 p. Unpub-
lished report.

Ewel, Jacob J. and John .L. Whitmore
1973. The ecological life zones of
Puerto Rico and the U.S. Virgin
Islands. U.S.D.A. Forest Service
Res. Paper ITF-18. 71 p. Institute
of Tropical Forestry, Rio Piedras,
Puerto Rico.

Grant, Chapman 1932. Herpetology
of Tortola; notes on Anegada and
Virgin Gorda, British Virgin Is-
lands. Journal of the Department
Agriculture of Puerto Rico 16:327-
329.

Heatwole, Harold, Levins, Robert,
and Michael D. Byer 1981. Biogeo-

graphy of the Puerto Rican Bank.
Atoll Research Bulletin No. 251:1-
55.

Heatwole, Harold, and Frank Mack-
enzie 1967. Herpetogeography of
Puerto Rico. IV. Paleogeography,
faunal similarity, and endemism.
Evolution 21:429-438.

Kaye, Clifford A. 1959a. Shoreline
features and quaternary shoreline
changes in Puerto Rico. U.S. Geo-
logical Survey Professional Paper
317-B:49-140.

Kaye, Clifford A. 1959b. Geology of
Isla Mona, Puerto Rico, and notes
on the age of the Mona Passage.
U.S. Geological Survey Profes-
sional Paper 317-E:141-178.

Kiester, A. Ross, George C. Gorman,
and David Colon Arroyo 1975.
Habitat selection behavior of three
species of Anolis lizards. Ecology
56:220-225.

Little, Elbert L. Jr. and Frank H.
Wads worth 1964. Common trees
of Puerto Rico and the Virgin is-
lands. U.S. Department of Agricul-
ture, Agriculture Handbook 249.
548 p. Washington, D.C.

MacLean, William P. 1982. Reptiles
and amphibians of the Virgin Is-
lands. 54 p. Macmillan Caribbean,
London.

Mayer, Gregory C. and James D,
Lazell 1988. Distributional records
for reptiles and amphibians from
the Puerto Rico Bank. Herpeto-
logical Review 19(l):23-24.

Matthews, William J. 1985. Distribu-
tion of midwestem fishes on
multivariate environmental gradi-
ents, with emphasis on Notropis
lutrensis. American Midland
Naturalist 113(2):225-237.

Moulton, Michael P. 1985. Morpho-
logical similarity and coexistence
of congeners: an experimental test
with introduced Hawaiian birds.
Oikos 44:301-305.

Nellis, David W., Norton, Robert L.,
and William P. MacLean 1984. On
the biogeography of the Virgin Is-
lands boa, Epicrates monensis
granti. Journal of Herpetology,
17(4):413-417.

237

Pregill, Gregory K. 1981. Pleistocene
herpetofaunas from Puerto Rico.
72 p. Miscellaneous Publications
Museum of Zoology, University of
Michigan, No. 71.

Pregill, Gregory K. and Storrs L.
Olson 1981. Zoogeography of
West Indian land vertebrates in
relation to Pleistocene cli-
maticcycles. Annual Review of
Ecology and Systematics 12:75-98.

Rivero, Juan A., Rafael Joglar, and
Idabella Vazquez. 1982. Cinco
nuevos ejemplares del culebron de
La Mona Epicrates m. monensis
(OphidiaiBoidae). Caribbean Jour-
nal of Science 17(l-4):7-13.

Rotenberry, John T. and John A.
Wiens 1980. Habitat structure,
patchiness, and avian communi-
ties in North American shrub-
steppe vegetation: a multivariate
analysis. Ecology 61:1228-1250.

SAS Institute 1985. SAS users guide:
statistics Version 5. 957 p. SAS In-
stitute Inc. Gary, North Carolina.

Schmidt, Karl P. 1926. The amphibi-
ans and reptiles of Mona Island,
West Indies. Field Museum of
Natural History Publications
12(12):149-163.

Sheplan, Bruce R. and Albert

Schwartz. 1974. Hispaniolan boas
of the genus Epicrates and their
Antillean relationships. Annals of
the Carnegie Museum 45(5):57-
143.

Stull, Olive G. 1933. Two new sub-
species of the family boidae. 4 p.
(Occasional Papers of the Museum
of Zoology, University of Michi-
gan, Number 267.

Tolson, Peter J. 1987. Phylogenetics
of the boid snake genus Epicrates
and Caribbean vicariance theory.
68 p. Occasional Papers of the Mu-
seum of Zoology, University of
Michigan, Number 115.

Tolson, Peter J. and Victoria A.
Tuebner 1987. The role of social
manipulation and environmental
cycling in propagation of the boid
genus Epicrates: Lessons from the
field and laboratory. American
Association of Zoological Parks

and Aquariums Regional Confer-
ence Proceedings 1987:606-613.

U.S. Fish and Wildlife Service 1980.
Status of Virgin Islands boa clari-
fied. Endangered Species Techni-
cal Bulletin, Department of Inte-
rior, U.S. Fish and Wildlife Serv-
ice, Endangered Species Program,
Washington, D.C. 5:12.

U.S. Fish and Wildlife Service 1984.
Mona boa recovery plan. 14 p.
U.S. Fish and Wildlife Service, At-
lanta, Ga.

U.S. Fish and Wildlife Service 1986.
Virgin Island tree boa recovery
plan. 23 p. U.S. Fish and Wildlife
Service, Atlanta, Ga.

Weiwandt, Thomas A. 1977. Behav-
ior, ecology, and management of
the Mona ground iguana, Cyclura
stegnegeri. Ph. D. dissertation, Cor-
nell University, Ithaca, New York.

Whitaker, A. H. 1973. Lizard popula-
tions on islands with and without
Polynesian rats, Rattus exulans
(Peale). Proceedings of the New
Zealand Ecological Society 20:121-
130.

Wiens, John A. and John T. Roten-
berry 1981. Habitat associations
and community structure of birds
in shrubsteppe environments. Eco-
logical Monographs 5(1)21-41.

Appendix A.
PCA Variables Measured on island Study Plots.

Variable

Predator

Rattus density
Felis presence
Alsophis density
Margarops density

Prey

Anolis density
Ameiva density
Anolis perch height

Coverage

Percent cover C trees
Percent cover S trees
Percent cover palms
Percent cover Opuntia
Percent cover grasses

Structural

Vegetational continuity

Canopy height
Plant diversity

Description

Rats captured/ trap hour

Present = 1, absent = 0

Mean no. Alsophis observed /day on transect

Mean no. Margarops observed /day on transect

Mean no. Anolis/ 5 m of transect
Mean no. Ameiva /S m of transect
Mean perch height in m of male Anolis

No.compound trees/ no. woody plants
No. single trees/no. woody plants
No. palms/no. woody plants
No. Opuntia/no. woody plants
Grassland area /total area

Contiguous = 1, high = .75, Moderate = .5

low = .25, absent = 0

>3 m = 1, 1-2 m = .5, <1 m = 0

No. of dominant plant species on plot

238

The Use of Timed Fixed-Area
Plots and a Mark-Recapture
Technique In Assessing
Riparian Garter Snake
Populations^

Robert C. Szaro,^ Scott C. Belfit,^ J. Kevin
Aitkin/ and Randall D. Babb^

Abstract.— Wandering garter snake (Thamnophis
elegans vagrans) populations along a thin-leaf alder
(AInus tenuifolia) riparian community in northern
New Mexico were sampled using timed fixed-area
plots and a mark- recapture method. Both methods
served to determine yearly differences and relative
magnitude of snake density between years. But
population estimates determined by timed fixed-
area plots were inconsistent between study plots in
the same year.

Research studies often attempt to de-
termine the effects of disturbance or
management regimes on the abun-
dance of wildlife species (Cooper-
rider et al. 1986, Fitch 1987, Parker
and Plummer 1987, Ralph and Scott
1980). How well the method of data
collection and analyses reflect actual
populations is critically important for
assessing the validity of these stud-
ies. Snakes are difficult subjects for
field studies because of their secre-
tive and cryptic habits (Fitch 1987).

Paper presented at symposium. Man-
agement of Amphibians, Reptile, and
Small Mammais in North America. (Flag-
staff, AZ, July 19-21, 1988.)

^Robert C. Szaro is Research Wildlife Bi-
ologist. USD A Forest Service. Roclcy Moun-
tain Forest and Range Experiment Station.
Arizona State University Campus. Tempe, AZ
85287-1304.

^ Scott C. Belfitis Wildlife Biologist, De-
partment of the Army, Wildlife Manage-
ment Section, Fort Huachuca, AZ 856 13-
6000. Beint's current address is P.O. Box 336,
Fort Belvoir, VA 22060-0336.

^ J. Kevin Aitkin is Wildlife Technician,
USDA Forest Service. Rocky Mountain Forest
and Range Experiment Station. Arizona
State University Campus, Tempe. AZ 85287-
1304.

^Randall D. Babb. formerly Wildlife Tech-
nician. USDA Forest Service. Rocky Moun-
tain Forest and Range Experiment Station.
Arizona State University Campus. Tempe. AZ
85287- 1304 is currently Sportfishing Program
Coordinator, Arizona Game and Fish De-
partment, 2222 West Greenway Road,
Phoenix, AZ 85023.

Many attempts to census snakes have
been inaccurate (Turner 1977, Fitch
1987). Population estimates can be
influenced by sex, reproductive con-
dition, and stage of maturity, all of
which are critical determinants of
activity within species (Gibbons and
Semlitsch 1987). Differences among
juveniles and breeding and non-
breeding females, and males often
lead to much different risks of cap-
ture at various stages of the season
and time of day. Overall population
estimates can be distorted as a result,
requiring separate estimates by sex
and age class (Fitch 1987).

Two methods often used to esti-
mate snake density are direct counts
and mark-recapture analyses. Sys-
tematic searches of defined areas (di-
rect counts) yield species occurrence
data, and usually require less time
and effort than mark-recapture meth-
ods (Jones 1986). Using direct counts.
Bury and Luckenbach (1977) success-
fully censused desert tortoise (Go-
pherus agassizii) populations with a
quartet and grid location system.
Bury (1982) used a removal method
to assess reptile community structure
in the Mohave Desert (Zippin 1956,
1958). Bury and Raphael (1983) refer
to searches conducted per unit effort
of time as time-constraint proce-
dures. Usually it is impossible to find
every snake in an area, making it
necessary to estimate population size

from capture-recapture ratios (Fitch
1987). Yet, when several density esti-
mates become available from the
same area at different times, they of-
ten show such drastic discrepancies
that the basic methods have been
thought invalid (Turner 1977).
Turner (1977) had no confidence in
the density estimates for snakes de-
rived from mark-recapture tech-
niques. However, since his critical
review, estimation techniques have
greatly improved with the develop-
ment of models and computer pro-
grams that test model assumptions
and estimate standard errors (Ar-
nason and Baniuk 1980, White et al.
1978, 1982, Otis et al. 1978, Brownie
et al. 1985).

Although time consuming, deter-
mining accurate pxDpulation estimates
is necessary to develop management
polices not only for abundant spe-
cies, such as the wandering garter
snake (Thamnophis elegans vagrans),
but also for aquatic or semi-aquatic
endangered snake species such as the
Concho water snake (Nerodia harteri
paucimaculata) (Scott and Fitzgerald
1985) and the narrow-headed garter
snake (Thamnophis rufipunctatus)
(Lowe 1985). Flowever, because the
wandering garter snake, is less secre-
tive than most kinds of snakes, and is
concentrated in riparian habitats, it is
probably one of the best adapted to
this sort of investigation (Fitch, per-

239

sonal communication). The results of
this work should be directly appli-
cable to other snake species normally
concentrated in riparian ecosystems
and may be especially useful for cen-
susing endangered species where
large samples to determine the accu-
racy of sampling techniques are not
available. Our previous work
showed the inadequacy of simple
transects and depletion sampling in
determining garter snake popula-
tions along the Rio de las Vacas, New
Mexico (Szaro et al. 1985). The objec-
tive of this study was to compare
timed fixed-area plots and a mark-
recapture technique in assessing the
impacts of management regimes on
riparian ecosystems in the arid
Southwest by sampling wandering
garter snake populations along the
Rio de las Vacas.

Methods and Study Areas

The Rio de las Vacas, is a montane
stream draining the San Pedro Parks
Wilderness Area, Santa Fe National
Forest, New Mexico. Under low flow
conditions, stream width ranges
from 2.8 to 10.5 m and averages 7.6
m. The study area is 17 km southeast
of Cuba, in Sandoval County, at 2600
m. Two cattle exclosures enclosing
stream reaches (each about 1 km long
by 50 m wide) were installed in the
early 1970's (Szaro et al. 1985). Con-
tiguous, downstream areas, privately
owned and grazed by livestock, were
used for comparison. The most ap-
parent difference between the grazed
and exclosed stream segments was
the band of small riparian trees and
shrubs in the exclosures (figs. 1 and
2). Thin-leafed alder (Alnus tenuifolia)
and a mixture of willow species
{Salix spp.) edged the exclosure
streambanks but were widely scat-
tered where the streambanks were
grazed (9.5 ± 1.16, 7.5 ± 1.23, and 0.3
+ 0.14 trees/250 m^ in exclosures 1, 2,
and grazed areas, respectively).

Snake populations were estimated
by timed fixed-area plot sampling.

Figure 1 .—Grazed section of the Rb de las Vacas, New Mexico. Notice the lack of shrub
growth and the unstable stream banks.

and mark-recapture sampling in both
grazed and ungrazed areas. For the
former, 16 plots (10 x 25 m), with the
long edge being defined by the
stream bank, were intensively
sampled for 20 minutes in each of the

two ungrazed exclosures and one
grazed stream segment along the Rio
de las Vacas, for a total of 48 plots
(fig. 3). During sample periods we
turned rocks, logs, debris piles, and
generally searched the area. All plots

Figure 2.— Shrubby growth in Exclosure 2 along the Rio de las Vacas, New Mexico.

240

were sampled once between 0900
and 1300 hours (MST) within a 3-day
period each month. Sampling times
were determined from preliminary
activity period sampling that showed
two distinct periods of activity
(morning and late afternoon). All
snakes captured were placed in a
cloth sack at their point of capture,
until the end of the sampling p>eriod.
Plot sampling began in June 1984 and
was replicated in July, August, and
September of that year and in the
same months in 1985. Total time
spent sampling was approximately
64 hours per year, excluding time be-
tween samples to process snakes.

For mark-recapture estimates, we
searched the entire extent of both ex-
closures and a similarly sized down-
stream grazed stream area. The plots
used for the timed-fixed plot sam-
pling were a subset of the area used
for the mark-recapture sampling. All
captured snakes were marked by
clipping three subcaudal scales (Blan-
chard and Finster 1933, Woodbury
1956). Mark-recapture sampling peri-
ods occurred in the same months as
the plot sampling; but snakes were
captured, marked, and released dur-

ing intensive searches for 6 consecu-
tive days by 3 to 4 collectors. All
snakes were released where cap-
tured. Approximately equal time and
effort was spent searching for snakes
in each of the three areas. Time of
day bias was minimized by alternat-
ing starting areas daily. Sampling
began at 0900 hours (MST) and con-
tinued until dusk. Only captures
within 10 m of the stream were used
in the mark-recapture analyses to al-
low a direct comparison to plot sam-
pling estimates. Thus, the plot sam-
pling represents a sample within the
exclosures and the grazed stream
area, whereas the mark-recapture
sampling represents an ''open"
population estimate of each study
area. Total time sp>ent sampling and
marking snakes was approximately
450 hours per year including time to
process snakes.

The approach to mark-recapture
analysis was to analyze each year
separately using closed population
models calculated by program CAP-
TURE, which allows unequal catch-
ability (Otis et al. 1978, White et al.
1978, 1982) as recommended by Pol-
lock (1981, 1982). Because we were

unable to estimate survival using the
timed fixed-area plots, we do not
present these estimates here for the
mark-recapture analysis. However,
all sampling periods were pooled
and survival estimators between
years estimated using the Jolly-Seber
Model (Seber 1986, Szaro et al., in
preparation).

Inferences about differences be-
tween years and exclosures were
based on Bonferroni's method for
multiple comparisons by fixing the
experimentwise error rate at 0.05
(Milliken and Johnson 1984). Thus,
the overall experimentwise error rate
is less than P (in this case 0.05); but
for each comparison, the compari-
son wise error rate is equal to P/n,
where n is the number of compari-
sons. For example, with 3 compari-
sons the actual P value per compari-
son would be 0.05/3 or 0.017.

Results

We are confident the mark-recapture
estimates accurately reflect popula-
tion densities on the three study ar-
eas and use these as the basis for

Rio de los Voces

%

Iidosure 1

Eidosvt 2 1 n

1 ol i
1 ^10 I
1 bIo I

1 K 1

Grazed

1 KU

1 1

/ Dev. 2600 m

Figure 3.— Study areas and sample plot lay-
out along thie Rio de las Vacas, New Mex-
ico.

Table 1 .—Population estimates of the wandering garter snake (Thamnophis
elegrans vagrans) In 1984 and 1985 within 10 m of the streambank at Rio
de las Vacas, New Mexico.

Mark-recapture'

Times fixed-area plot^

Study area

Year

Mean

S.E. Sig.3

Mean

S.E.

SIg.

Exclosure 1

1984

282

+

23.53 (3.86)^ a

1.28

+

0.18

a

1985

166

15.51 (2.28) b

0.88

0.11

a

Exclosure 2

1984

296

24.42(4.53) a

1.30

0.17

a

1985

146

+

13.92 (2.23) b

0.45

±

0.09

c

Grazed

1984

67

+

10.49(1.00) c

0.28

+

0.07

cd

1985

26

+

5.22 (0.39) d

0.11

±

0.04

d

'Mark-recapture estimates for each stvdy area are for ttie total population using
tt)e best model in CAPTURE for whicti solutions exist. The total area sampled in each
area was 16,240 m^ in Exclosure L 16,340 in Exclosure 2, and 16,760 m^ in the
grazed area.

'Plot samples are mean number of snakes caught per 250 m',
^Population estimates by each method that do not have a letter in common are
significantly different (Bonferroni's method. P<0.05).

^Number in parenthesis is estimated number of snakes per 250 mP using the mark-
recapture population estimate.

241

comparison for the timed fixed-area
plot results. Mark-recapture esti-
mates were based on 118 individuals
and 35 recaptures (118/35) in exclo-
sure 1 in 1984, 72/28 in 1985, 127/30
in exclosure 2 in 1984, 74/26 in 1985,
12/2 in the grazed area in 1984, and
10/1 in 1985.

We asked two questions of the
sampling methods. First, were there
any differences in population esti-
mates between years? Both methods
indicated decreases in populahon
size on all three areas between 1984
and 1985. However, yearly differ-
ences were significant only for mark-
recapture estimates and for the timed
fixed-area plot estimates in exclosure
2 (P < 0.05) (table 1). Mark-recapture
estimates revealed that snake popu-
lations decreased by 41% to 54%
from 1984 to 1985 in all study areas.
Decreases in mean number of snakes
per fixed-area plot were not as uni-
form, varying from 31% on exclosure
1 to 65% on exclosure 2 and the
grazed stream segment.

Second, were there differences be-
tween the study areas? Population
estimates between exclosures and the
grazed stream segment within a
given year were significantly differ-
ent by both census methods and for
both years (P < 0.05) (table 1). Popu-
lation estimates by both methods
were not significantly different be-
tween exclosures, except in 1985
when the estimate determined by
timed fixed-area plots for exclosure 2
was 50% of that on exclosure 1 (P <
0.05) (table 1).

Estimating population size by re-
stricting the mark-recapture esti-
mates to a 10 m band on either side
of the stream served a twofold pur-
pose. First, it allowed us to estimate
the number of snakes per unit area.
Second, it made estimates by both
techniques more readily comparable,
because all plot sampling was con-
fined to the 10-m band next to the
stream where most of the available
down litter, grass clumps, and
shrubby vegetation was concen-
trated. In exclosure 1, there were 3.86

and 2.28 snakes per 250 m^ in 1984
and 1985, respectively. In exclosure
2, there were 4.53 and 2.23 snakes
per 250 m^ in 1984 and 1985, respec-
tively. Along the grazed stream reach
there were 1.00 and 0.38 snakes per
250 m^ in 1984 and 1985, respectively.
Based on these estimates, we caught
between 20.2% (exclosure 2, 1985)
and 38.6% (exclosure 1, 1985) of the
snakes present in the exclosures. On
the grazed area we caught 28% of the
snakes in both 1984 and 1985.

Discussion

Apparent short-term downward
population fluctuations averaging
about 50% have been found in sev-
eral mark-recapture studies (Fukada
1969, Piatt 1969, Fitch 1975, Feaver
1977, Gregory 1977). Many studies of
snakes have related population
changes over several years to succes-
sional changes (Clark 1970, Fitch
1982) or to environmental factors,
such as decreases in annual precipi-
tation (Clark 1974, Clark and Fleet
1976). Another possibility, is that a
study like this actually destroys hid-
ing places (turning rocks, logs, etc.);
and even if each piece is put back
carefully, the site has opened up and
changed (Clark, personal communi-
cation).

We undoubtedly had some impact
on the quality of the available habitat
by our intensive searching tactics;
but we did try to be as careful as pos-
sible to return moved objects back
into their original positions. Parker
and Plummer (1987) suggest that
these apparent fluctuations in den-
sity result from changes in activity
level (which affect recapture proba-
bilities) rather than from actual
changes in density (Lillywhite 1982,
Pough 1983). There are three possible
explanations for these results: (1)
snakes simply moved out of the plot
and exclosure areas; (2) snakes be-
came inactive in burrows or cover
sites because of environmental condi-
tions; or (3) snakes died.

Activity periods of wandering gar-
ter snakes varied between individu-
als from our preliminary sample of
wandering garter snake populations
along the Rio de las Vacas in July
1983. We failed to decrease signifi-
cantly the total numbers of animals
caught per plot even after 3 days of
removal sampling (Szaro et al. 1985);
but at other times snakes were diffi-
cult to find. However, we feel the in-
tensive sampling effort of at least 1
week each month minimized the ef-
fect of changes in snake behavior on
population estimates.

The almost 50% difference in 1985
between exclosures in mean number
of snakes caught while plot sampling
was probably a result of a shift in ar-
eas used by the snakes and not dif-
ferences in mortality between the
two exclosures. Monthly trends in
total number of snakes caught also
showed a dramatic difference in the
number of snakes caught per month
while plot sampling in both exclo-
sures. However, this difference was
not reflected in the overall number of
snakes caught during mark-recapture
sampling (fig. 4). In fact, overall we
caught more snakes in exclosure 2
than in exclosure 1 in all months in
1985.

The difference in plot sampling es-
timates between exclosures in 1985
was not a result of changes in daily
activity patterns, because equal pro-
portions of snake captures in both
exclosures were before 1300 (63% in
exclosure 1 and 59% in exclosure 2,
chi-square, P > 0.05). Furthermore,
differences in captures between years
and methods were not sex-based, be-
cause there were no significant dif-
ferences in sex ratios between years
or method in a given study section
(chi-square, P > 0.05) (fig. 5). How-
ever, there were distributional differ-
ences in snake captures between
years and exclosures.

In 1984, 34.6% and 34.7% of all
captures on exclosures 1 and 2, re-
spectively were made on the plot ar-
eas. In contrast, 42.1% and 20.6% of
all captures on exclosures 1 and 2, re-

242

Plot Sampling

June July Aug. Sept. June July Aug. Sept
1984 1985

Sampling Period

Figure 4.— Total numbers of wandering garter snakes caught in June, July, August, and Sep-
tember 1984 and 1985 along \he Rio de las Vacas, New Mexico during timed fixed-area plot
and mark-recapture sampling.

specrively, were made on the plot
areas in 1985.

We cannot explain this distribu-
tional shift in exclosure 2. Although
we did not plot sample in 1986 and
1987, mark-recapture efforts in those
years showed a similar distributional
pattern (Szaro et al., unpublished). In
exclosure 1, 33.0 % and 37.3% of all
captures in 1986 and 1987, respec-
tively were on the old plot areas,
whereas in exclosure 2, these values
were 10.07o and 9.8%.

We feel that the distributional
changes in exclosure 2 were not an
artifact of plot sampling, because
snakes in exclosure 2 did not return
to plot areas after plot sampling had
stopped. In any case, our sampling
potentially would have been more
destructive in exclosure 1 than in ex-
closure 2 because of the higher inci-
dence of turnable rocks in that exclo-
sure.

Whatever the cause, these changes
in distribution indicate that initial
randomized selection of plots did in-
fluence density estimates for exclo-
sure 2. Although it would increase
substantially the amount of time nec-
essary to adequately sample vegeta-
tion, a better approach would be to
randomly select plots within exclo-
sures each sampling period rather
than repeatedly sampling the same
plots.

In conclusion, the use of timed
fixed-area plots enabled us to quan-
tify dramatic differences in snake
abundance between exclosures and
the grazed area. However, this sam-
pling method is of questionable merit
because of the significant difference
in exclosure population estimates for
1985. Further study incorporating
newly randomized plots for each
sampling period may solve this prob-
lem. Care should be taken to deter-
mine if snakes are distributing them-
selves in a nonrandom pattern. At
this time, we recommend the more
labor-intensive mark-recapture esti-
mators for assessing the impacts of
riparian management regimes on
snake populations.

243

Acknowledgments

We thank D. R. Clark, H. S. Fitch, K.
B. Jones, and N. J. Scott, Jr. for their
constructive reviews of this paper. H.
Berna, M. Cady, C. Engel- Wilson, X.
Hernandez, D. Johnson, M. Lane, W.
Legarde, L. Simon, and D. Smith
aided in the collection of the field
data. Special thanks to Jim and Mary
Bedeaux for their gracious hospital-
ity and allowing us to sample on
their property.

Liferature Cited

Arnason, A. Neil and Leonard
Baniuk. 1980. A computer system
for mark-recapture analysis of
open populations. Journal of Wild-
life Management 44:325-332.

Blanchard, Frank N. and Ethel B. Fin-
ster. 1933. A method of marking
living snakes for future recogni-
tion, with a discussion of some
problems and results. Ecology
14:334-347.

Brownie, Cavell, David R. Anderson,
Kenneth P. Bumham, and Douglas
S. Robson. 1985. Statistical Infer-
ence from band recovery data — a
handbook. USDI Fish and Wildlife
Service Resource Publication
156:1-305.

Bury, R. Bruce. 1982. Structure and
composition of Mojave Desert rep-
tile Communities determined with
a removal method. Pages 135-142.
In Norman J. Scott, Jr., Editor,
Herpetological Communities.
USDI Fish and Wildlife Service
Wildlife Research Report 13.

Bury, R. Bruce and R. A. Lucken-
bach. 1977. Censusing desert tor-
toise populations using a quartet
and grid location system. Desert
Tortoise Council Symposium Pro-
ceedings 1977:169-178.

Bury, R. Bruce and Martin G. Ra-
phael. 1983. Inventory methods
for amphibians and reptiles. Pages
416-419. In J. F. Bell and T. Atter-
bury. Editors, Renewable resource
inventories for monitoring

changes and trends: Proceedings
of an international conference.
[Corvallis, Oregon, August 15-19,
1983] Society of American Forest-
ers 83-14..
Clark, Donald R., Jr. 1970. Ecological
study of the worm snake, Car-
phophis vermis (Kennicott). Univer-
sity of Kansas Publications of the

Museum of Natural History 19:85-
194.

Clark, Donald R., Jr. 1974. The west-
em ribbon snake (Thamnophis
proximus): ecology of a Texas
population. Herpetologica 30:372-
379.

Clark, Donald R., Jr. and R. R. Fleet.
1976. The rough earth snake (Vir-

Exclosure 1

1984

1985

1984

o

E
o

o
>-

1985

1984

1985

40 60
Exclosure 2

Grazed Stream Reach

0 20 40 60

Percent of Total Observations

•.. 1 Femole

100

Figure 5.— Proportion of totai captures by sex during timed fixed-area plot and mark-recap-
ture sampling in two exclosures and a grazed stream reach) along \he Rio de las Vacas,
New Mexico, 1984 and 1985.

244

ginia striatula): ecology of a Texas
population. Southwestern Natu-
ralist 20:467-478.

Cooperrider, Allen Y., Raymond J.
Boyd, and Hanson R. Stuart, Edi-
tors. 1986. Inventory and monitor-
ing of wildlife habitat. U.S. De-
partment of Interior, Bureau of
Land Management Service Center,
Denver, Colorado. 858 p.

Feaver, P. E. 1977. The demography
of a Michigan population of
(Natrix sipedon) with discussions of
ophidian growth and reproduc-
tion. Ph.D. Dissertation. Univer-
sity of Michigan, Ann Arbor.

Fitch, Henry S. 1975. A demographic
study of the ringneck snake (Di-
adophis punctatus) in Kansas. Uni-
versity of Kansas Museum of
Natural History Miscellaneous
Publications 62:1-53.

Fitch, Henry S. 1982. Resources of a
snake community in prairie-wood-
land habitat of northeastern
Kansas. In Norman J. Scott, Jr.,
Editor, Herpetological Communi-
ties. USDI Fish and Wildlife Serv-
ice Wildlife Research Report 13:83-
97.

Fitch, Henry S. 1987. Collecting and
life-history techniques. Pages 143-
164. In Richard A. Seigel, Joseph T.
Collins, and Susan S. Novak, Edi-
tors, Snakes: Ecology and evolu-
tionary biology. MacMillan Pub-
lishing Company, New York.
529 p.

Fukada, H. 1969. Biological studies
on the snakes. XIII. Preliminary
estimate of population size in
Tambabashi study area. Bulletin of
Kyoto University Educational Se-
ries B Mathematics and Natural
Sciences 36:3-9.

Gibbons, J. Whitfield and Raymond
D. Semlitsch. 1987. AcUvity pat-
terns. Pages 396-421. In Richard A.
Seigel, Joseph T. Collins, and
Susan S. Novak, Editors, Snakes:
Ecology and evolutionary biology.
MacMillan Publishing Company,
New York. 529 p.

Gregory, P. T. 1977. Life history
parameters of the red-sided garter

snake (Thamnophis sirtalis parietalis)
in the Interlake of Manitoba. Na-
tional Museum of Canada Publica-
tions in Zoology 13:1-44.

Jones, K. Bruce. 1986. Amphibians
and reptiles. Pages 267-290. In Al-
len Y. Cooperrider, Raymond J.
Boyd, and Hanson R. Stuart, Edi-
tors, Inventory and monitoring of
wildlife habitat. U.S. Department
of Interior, Bureau of Land Man-
agement Service Center, Denver,
Colorado. 858 p.

Lillywhite, Harvey B. 1982. Tracking
as an aid in ecological studies of
snakes. In Norman J. Scott, Jr.,
Editor, Herpetological Communi-
ties. USDI Fish and Wildlife Serv-
ice Wildlife Research Report
13:181-191.

Lowe, Charles H. 1985. Amphibians
and reptiles in southwest riparian
ecosystems. Pages 339-341. In R.
Roy Johnson, Charles D. Ziebel,
David R. Patton, Peter F. Ffolliott,
and Robert H. Hamre, Technical
Coordinators, Riparian ecosys-
tems and their management: rec-
onciling conflicting uses. First
North American Riparian Confer-
ence. USDA Forest Service Gen-
eral Technical Report RM-120.
Rocky Mountain Forest and Range
Experiment Station, Ft. Collins,
Colo. 523 p.

Milliken, George A. and Dallas E.
Johnson. 1984. Analysis of messy
data. Volume 1: designed experi-
ments. Van Nostrand Reinhold
Company, New York. 473 p.

Otis, David L., Kenneth P. Burnham,
Gary C. White, and David R. An-
derson. 1978. Statistical inference
from capture data on closed ani-
mal populations. Wildlife Mono-
graphs 62:1-135.

Parker, William S. and Michael V.
Plummer. 1987. Population ecol-
ogy. Pages 253-301. In Richard A.
Seigel, Joseph T. Collins, and
Susan S. Novak, Editors, Snakes:
Ecology and evolutionary biology.
MacMillan Publishing Company,
New York. 529 p.

Piatt, D. R. 1969. Natural history of

the hognose snakes, Heterodon plat-
yrhinos and Heterodon nasicus. Uni-
versity of Kansas Publications of
the Museum of Natural History
18:253-420.

Pollock, K. H. 1981. Caphire-recap-
ture models: a review of current
methods, assumptions, and ex-
perimental design. Studies in
Avian Biology 6:426-435.

Pollock, Kenneth H. 1982. A capture-
recapture design robust to un-
equal probability of capture. Jour-
nal of Wildlife Management
46:752-757.

Pough, F. H. 1983. Amphibians and
reptiles as low-energy systems.
Pages 141-188. In W. P. Aspey and
S. I. Lustick, Editors, Behavioral
energetics: the cost of survival in
vertebrates. Ohio State University
Press, Columbus.

Ralph, C. John and J. Michael Scott,
Editors. 1980. Estimating numbers
of terrestrial birds. Studies in
Avian Biology 6:1-630.

Scott, Norman J., Jr. and Lee H. Fitz-
gerald. 1985. Status survey of
Nerodia harteri, Brazos and Chon-
cho — Colorado Rivers, Texas.
Status Survey Report. U.S. Fish
and Wildlife Service, National
Ecology Center, Museum of
Southwestern Biology, University
of New Mexico, Albuquerque,
New Mexico. 44 p.

Seber, G. A. F. 1986. A review of esti-
mating animal abundance. Bio-
metrics 42:267-292.

Szaro, Robert C, Scott C. Belfit, J.
Kevin Aitkin, and John N. Rinne.
1985. Impact of grazing on a ripar-
ian garter snake. Pages 359-363. In
R. Roy Johnson, Charles D. Ziebel,
David R. Patton, Peter F. Ffolliott,
and Robert H. Hamre, Technical
Coordinators, Riparian ecosys-
tems and their management: rec-
onciling conflicting uses. First
North A.merican Riparian Confer-
ence. USDA Forest Service Gen-
eral Technical Report RM-120.
Rocky Mountain Forest and Range
Experiment Station, Ft. Collins,
Colo. 523 p.

245

Turner, F. B. 1977. The djmamics of
populations of squamates, croco-
dilians and rhynchocephalians.
Pages 157-264. In C. Cans and D.
W. Tinkle, Editors, Biology of Re|>
tilia, volume 7. Academic Press.

I

Wh
1
1

c

\

J
8
2

Wh
I

d

3

F
L

4

Wo(
n

sU

reptiles. Ecology 37:670-674.

Zippin, C. 1956. An evaluation of the
removal method of estimating ani-
mal populations. Biometrics
12:163-189.

Zippin, C. 1958. The removal method
of population estimation. Journal
of Wildlife Management 22:82-90.

246

Design Considerations for the
Study of Amphibians, Rep-
tiles, and Small Mammals in
California's Oak Woodlands:
Temporal and Spatial
Patterns^

William M. Block, Michael L Morrison, John
C. Slaymaker, and Gwen Jongejan^

Abstract.— We monitored pitfall traps for >50,000
trap nights among three study areas in California's
oak woodlands. Numbers of captures and trap
success varied spatially in comparisons of grids
within and among stand types, as well as among
study areas. Capture numbers also varied
temporally, both within and between the years of
study. Differences in capture rates varied among
taxa (amphibians, reptiles, and small mammals) and
also varied among species within a taxon.
Researchers should design studies to sample
temporal and spatial variations in activity patterns to
provide a more complete understanding of the
habitat associations of the species studied.

position and structure of the vegeta-
tion among the study areas.

Sierra Foothill Range Field Station
(SFRFS), Yuba County, was located
in the foothills of the Sierra Nevada
about 25 km NE of Marysville. Eleva-
tion ranged from 2(X) to 700 m on a
general west-northwest facing slope.
Blue oak (Quercus douglasii), interior
live oak (Q. wislizenii), and digger
pine (Pinus sabiniam) were the major
species of trees with lesser amounts
of California black oak (Q. kelloggii),
California buckeye (Aesculus californi-
cus), and ponderosa pine (Pinus pon-
derosa). Major components of the
shrub layer included buckbrush
(Ceanothus cuneatus), coffeebcrry
(Rhamnus californica), and poison oak
(Toxicodendron diversiloba). Annual
and perennial grasses and forbs
dominated cover within a meter of
the ground, although there were spa-
tial and temporal variations in spe-
cies compositions and also in amount
of ground cover. Further, the compo-
sition and structure of the canopy,
shrub, and ground layers have all
been modified by historic land-use
practices at the Station. Except for 60
ha of fenced areas, the remaining
1800 ha are used for varied research
projects usually entailing cattle graz-
ing and often entailing tree removal.

San Joaquin Experimental Range
(SJER), Madera County, was located
in the foothills of the Sierra Nevada
about 40 km N of Fresno. Elevation
ranged from 200 to 500 m; the aspect
was in a general southwest direction.

The hardwood rangelands of Califor-
nia are coming under increasing
land-use pressures. Cattle grazing,
fuelwood removal, hydro-electric
projects, urban sprawl, and countless
other factors are impacting these
woodlands at local, regional, and
geographical levels (see papers
within Plumb and Pillsbury 1987).
Unfortunately, little is known of the
distributions and ecologies of many
of the vertebrates occurring in these
areas (Vemer 1987). As a conse-
quence, resource managers fre-
quently have too little information
upon which to base land-use deci-
sions. Thus, a research agenda is re-
quired first to obtain baseline infor-
mation on distributions and habitat
associations of these animals, and
then to use these data to predict the
presence or absence of these species,
and ultimately to predict the effects
of habitat change on their popula-
tions. Research should encompass a
hierarchy of spatial scales to account
for variations in patterns of habitat
use, and also to determine if a spe-
cies' habitat exhibits consistent and
measurable features (Allen and Starr
1982, Block, in press). Study must

'Paper presented at symposium. Man-
agement of Amphibians. Reptiles, and
Small t^ommals in Northi America. (Flag-
staff. AZ. July 19-21. 1988.)

'Project Leader. Associate Professor.
Researct) Associate, and Research) Associ-
ate, respectively. Department of Forestry
and Resource Management. University of
California. Berkeley. CA 94720.

also be done year-round to sample
habitat-use by species during differ-
ent stages of their life histories, and it
also should be done over a number
of years to include annual variations
in environmental conditions (Halvor-
son 1984, Morrison, this volume).

As part of an ongoing study to de-
termine habitat relationships of ver-
tebrates in California's oak wood-
lands, we have been using pitfall
traps to sample populations of small
mammals, reptiles, and amphibians
at three distinct areas. To date we
have collected data from greater than
50,000 trap nights distributed among
20 trapping grids. This general de-
sign has allowed us to examine spa-
tial patterns of habitat-use both
within and among areas. Further,
more intensive study has been done
at one area to examine temporal pat-
terns in habitat use both within and
between years. In this paper we pres-
ent these data to examine spatial and
temporal patterns of habitat use and
discuss our results in relation to the
general design of studies of small
mammal, reptile, and amphibian
populations.

STUDY AREAS

The study was done at three areas,
all oak or pine-oak woodlands. Study
areas were distributed along a latitu-
dinal gradient of about 600 km, and
consequently there were notable dif-
ferences in topography and in com-

247

Blue oak, interior live oak, and dig-
ger pine were the major tree species.
These species occurred in mixed-spe-
cies stands, stands of blue oak wood-
land, or as blue oak savannas. An-
nual and perennial forbs and annual
grasses dominated the ground layer.
About 20 ha of SJER have been
fenced to exclude cattle grazing.
Cattle grazing on the remaining 1500
ha has resulted in a sparser shrub
understory at SJER than of that
found at SFRFS (Duncan et al. 1987).
Major shrubs include buckbrush,
whitethorn ceanothus (Ceanothus leu-
codermis), redberry (Rhamnus crocea),
coffeberry, poison oak, and white
lupine (Lupinus alba). The shrub un-
derstory is restricted mostly to
widely scattered stands of mature
shrubs which have grown above the
deer-cattle browse line.

Tejon Ranch (TR), Kern County,
was located about 50 km south of
Bakersfield in the Tehachapi Moun-
tains. Elevation ranged from 1100 to
1700 m; aspects included all cardinal
directions. Major trees found on TR
included blue oak, valley oak (Quer-
cus lobata), California black oak, inte-
rior live oak, canyon live oak ( Q.
chrysolepis), Brewer's oak (Q. garryam
var. breweri), and California buckeye.
At lower elevations, these trees gen-
erally occurred in pure stands of
single species, with mixed-stands of
California black, canyon live, interior
live, and Brewer's oaks occurring at
higher elevations. Buckbrush,
redberry, and mountain mahogany
(Cercocarpus betuloides) were the ma-
jor shrubs with annual and perennial
grasses and forbs comprising the
ground canopy. Cattle grazing and
fuelwood harvest have modified the
composition and structure of the
tree, shrub, and herbaceous layers.

METHODS

Field Methods

At TR we placed three grids in each
of three different stand types — blue

oak, valley oak, and canyon live oak
woodlands — and we placed four
grids in four different stands of
mixed-oak woodlands (California
black, interior live, canyon live, and
Brewer's oaks). At SJER we placed
four grids, one each in a blue oak and
an interior live oak stand, and two in
mixed blue oak-interior live oak-dig-
ger pine stands. The three grids at
SFRFS sampled three stands of
mixed blue oak-interior live oak-dig-
ger pine woodlands. Selection of
stands was not entirely random be-
cause we needed to consider accessi-
bility during inclement weather, and
possible conflicts with other research
projects or with certain management
practices (e.g., excessive cattle graz-
ing, fuelwood harvest, road con-
struction) when selecting stands. The
actual selection of the grid location
within a stand was by a series of ran-
dom procedures to determine dis-
tance of the grid from the stand edge
(>100 m from the stand edge to mini-
mize edge effects) and the direction
of the grid array.

Each grid consisted of 36 2-gal,
plastic buckets arrayed in a 6 x 6
square with 20-m interstation spac-
ings. Buckets were placed within 2 m
of each grid point at a suitable trap-
ping location. Buckets were sunk to
ground level and left closed (a piece
of plywood secured with a rock) for
at least one month prior to being
opened. This period enabled germi-
nation of grasses and forbs to occur
thus making the area near the trap
appear less disturbed and also al-
lowed small mammals and herpe-
tofauna to become accustomed to the
presence of the traps. Traps were
opened by propping a plywood lid 5-
10 cm above the lip of the bucket us-
ing small branches or small rocks
and then placing 3-6 cm of water in
the bottom of the bucket. Traps were
checked once a week and were left
open for 1-2 months at a time. We
noted the species, date, and trap lo-
cation of all captures. Dead animals
were removed from traps; live ani-
mals were removed and relocated to

a similar habitat at least one km from
the nearest trapping grid.

We monitored pitfall traps at TR
from 4 January to 20 May 1987 and
from 10 December 1987 to 20 June
1988. We regarded the first year of
monitoring as a pilot study to evalu-
ate and refine our methods. Traps
were opened and monitored for 30
days using the methods described
above. However, in light of a recent
article by Bury and Corn (1987), we
increased our trapping period from
30 to 60-65 days per grid. Thus, our
design at TR for the second year con-
sisted of opening one grid of each
stand type for 60-65 days, closing
those, and then opening another set
of four grids. We repeated this de-
sign three times. We opened the four
grids at SJER and the three grids at
SFRFS each for 60 days from mid-
January through mid-March 1988.

Data Analyses

We compared standardized capture
numbers among stand types at TR
and among the three study areas (TR,
SJER, and SFRFS) to determine gen-
eral distributional patterns of the ani-
mals caught. Capture numbers were
standardized by pooling all captures
of a species within a stand type or
within a study area and dividing this
number by the total number of trap
nights for each grid within that stand
type or study area. We calculated
Spearman rank-order coefficients
(Marascuilo and McSweeny 1977) to
test for differences in rankings of
captures of species among stand
types at TR and then of captures
among the three areas. We tested for
species-specific differences in capture
rates among stand types and among
study areas using Kruskal-Wallis
analyses (Marascuilo and McSweeny
1977).

We used log-linear analyses (Fien-
berg 1980) to determine the sources
of variation in trap success within
and among years, stand types, and
study areas. We used data only for

248

the presence or absence of a species
at each trapping station, regardless
of the number of individuals of the
species that were captured at the sta-
tion. Because the number of trap
nights varied between grids, we used
this variable as a covariate in all
analyses to factor out the bias this
might have entered in our analyses.

To test for within-year, spatial-
temporal patterns, we restricted our
analyses to data collected in 1988.
Analyses were done for common
species (i.e., those for which we had
adequate numbers of samples) and
taxon variables of mammals, am-
phibians, and reptiles. We used data
from TR to examine seasonal and
stand associations of common spe-
cies of each taxon.

To examine geographic patterns of
captures, we compared trap success
among the three study areas. Be-
tween-year analyses were done by

comparing trap success at TR from
1987 and 1988.

RESULTS
General Patterns
Tejon Rancti

The ranking of species captured in
canyon live oak woodlands was not
significantly correlated with the
rankings of species found in the
other woodland types (all values
were nonsignificant, n = 21, P > 0.05).
These differences were attributable
to a stronger association of amphibi-
ans, particularly Ensatim and Batra-
choseps salamanders, with canyon
live oak stands than with the other
types of woodlands (Kruskal-Wallis
Analyses, df = 2, P < 0.10) (table 1).
Differences among stands were also

Table 1.— Capture numbers of amphibians, reptiles, and small mammals
within four different oak woodlarxl types at Tejon Ranch, Kem County,
California from 1 January 1987 through 20 June 1988.

Species

Bafrachoseps nigrivenfris^
Ensafina eschscholfzii^
Rana boylii

Sceloporus occidentalism
Eumeces gilberfF
Gerrhonofus mulficarinafus
Anniella pulchra
Diadophis pulchellus
Peromyscus maniculafus^
P. boylii
P. frueP

Perognafhus californicus
Microfus californicus
Thomomys botfae
Reifhrodonfomys megalofis
Scapanus lafimanus
Sorex omafus

Total captures

Species richness

Valley
oak
<n=7848)'

Blue
oak
(n=8828)

Canyon
live oak
(n=7848)

f^lxed
oak
(n=8828)

19

1

39
34

38
53

3
13

20
28

1

31
4
3

1

42
33
14

2
8

1

168
10

10
20

1
2
4
1

112
9

22

3
4
2

1

13
138
10

3
24
6

1

6
1
1

6
102
13

'Number of trap nights.

^Significant difference (P <0. 10) of captures among stand types.

noted for captures of Peromyscus
maniculatus, P. truei, Sceloporus oc-
cidentalis, and Eumeces gilberti, which
were captured more frequently in
blue and valley oak stands than in
canyon live or mixed-species oak
stands (table 1). In comparisons of
rankings of taxonomic groups among
stand types, we found a significant
positive correlation between mixed-
species and valley oak stands, but a
significant negative correlation be-
tween blue and canyon live oak
stands (r^ significant, n = 3, P < 0.01)
(fig. 1). All other pair-wise compari-
sons between stand types were non-
significant.

All Study Areas

Rankings of captures of species were
weakly correlated only between TR
and SFRFS (r^ = 0.37, n = 21, P =
0.052); Spearman rank-order correla-
tions were nonsignificant in all other
comparisons. Significant differences
were found among areas in the cap-
ture rates of Sceloporus occidentalis,
Eumeces gilberti, E. sJdltonianus, Batra-
choseps attenuatus, Batrachoseps ni-
griventris, and Ensatina eschscholtzii
(table 2). In contrast, rankings of taxa
were significantly correlated between
SJER and SFRFS (r^ = 1.00, n = 3, P =
1.00), but nonsignificant (P > 0.05) in
all other between-area comparisons.
The differences were primarily be-
cause of differences in capture rates
of reptiles and amphibians (fig. 2).

Log-linear Analyses

Trap success at TR for small mam-
mals, reptiles, and amphibians dif-
fered with stand type and trapping
period (likelihood ratio chi-squares,
P < 0.01). Similar results were found
for the selected common species. In
contrast, fewer differences were
found between years for captures of
amphibians, reptiles, and small
mammals. Only captures of reptiles
in blue oak stands and captures of

249

r .•nil

small mammals within valley oak
stands were significantly different
between years (likelihood ratio chi-
squares, P < 0.01). We noted signifi-
cant differences (P < 0.01) in capture
frequencies of reptiles and amphibi-
ans among study areas, but differ-
ences were nonsignificant (P > 0.05)
for captures of small mammals.

DISCUSSION

Intra-year differences in trap success
at TR were observed for all common
species and taxonomic groups tested.
Much of the intra-year variation in
trap success was probably because of
differences in activity patterns dur-
ing different times of the year (Welsh
1987). Our results further suggested
that activity patterns varied within
and among taxa. For example, few
reptiles were captured from Decem-
ber through March; capture rates
then increased dramatically after
March. In contrast, fewer salaman-
ders were caught in December, Janu-
ary, May, and June than were caught
during March and April. Similar re-
sults emerge when comparing activ-
ity patterns of species within a taxon.
Thus, activity patterns of a species or
of a taxon tend to be somewhat spe-
cific to the animal or group studied.

Differences in trap success were
not as apparent for interyear com-
parisons, however. In fact, the only
differences that we noted were in-
creases from 1987 to 1988 in trap suc-
cess for reptiles in blue oak and for
mammals in valley oak stands. These
results might be interpreted in two
ways. First, species compositions are
fairly consistent from year to year, or
the 2 years of data that we compared
were possibly insufficient to detect
population or habitat shifts (Halvor-
son 1984, Morrison, this volume).
Undoubtedly, a long-term study is
required to determine if these results
remain valid with time or if they are
an artifact of the sampling period.

Species distributions also varied
spatially among the different stand

types at TR and among the three
study areas. For example, canyon
live oak stands contained more am-
phibians and fewer reptiles than
other types of stands, whereas few
amphibians and more reptiles were
captured in blue oak stands. Valley
oak and mixed-species oak stands
contained intermediate numbers of
amphibians and reptiles. We also
noted differences of captures among
grids of the same woodland type.
However, given the short duration of
this study (2 years to date), these dif-
ferences may reflect temporal differ-
ences between sampling periods
more than variation within stand
types. Variation was also noted on a
broader geographical scale of be-
tween study areas.

Pitfall traps are one of many tech-
niques used to sample vertebrate
populations (Day et al. 1980). As
with each technique, however, pitfall
traps are not without limitations
(Bury and Com 1987). Inter- and in-
traspecific differences in motility,
mode of travel, and activity range all

influence the probability of an animal
being captured. Because of probable
species-specific biases in catchability,
a study design should consider alter-
native methods (e.g., live traps for
small mammals, and active searches
for reptiles and amphibians) to
sample the population(s) of the spe-
cies of interest (Halvorson 1984, Ra-
phael and Rosenberg 1983, Welsh
1987).

For example, results from our pit-
fall data do not completely agree
with preliminary results from >6,000
trap nights using live traps or from
20 time-constraint searches, both
done at TR (Block, unpubl. data). In
particular, we captured more Per-
ognathus californicus and Reithrodonto-
mys megalotis using live traps than
we did using pitfall traps, but have
captured no Microtus, Sorex, Tho-
momys, or Scapanus in live traps
whereas we have caught them in the
pitfalls.

Thus, researchers should compare
and evaluate results from alternative
methods to determine the most effec-

8-

1

Valley ook

Bkie ook

Conyon ook

Mixed ook

Amphibians

Reptiles

Mafranols

Figure 1 .—Relative numbers of captures using pitfall traps wittiin four oak woodland types at
Tejon Rancti, Kern County, California from 5 Decemt^er 1987 to 20 June 1988.

250

tive method or combination of meth-
ods to use for the species under
study.

We evaluated our data in two dif-
ferent ways: comparisons of capture
numbers and comparisons of trap
success. Results from both analyses
were generally consistent, although
in some cases we found differences
in comparisons of trap success, but
failed to do so in comparisons of cap-
ture numbers. The discrepancies be-
tween these results may be attribut-
able to both statistical and biological
factors.

Statistical factors stem from the
fact that continuous data were re-
corded for capture numbers whereas
categorical data were recorded for
trap success. Consequently, different
statistical tests were required to ana-
lyze the different types of data. The

lack of concordance between results
may be the result of different as-
sumptions of the different tests and
of different powers of the associated
statistics.

For example, in comparisons of
capture numbers, our use of all cap-
tures from a trap for a given species
may have violated assumptions of in-
dependence of samples; assumptions
underlying most parametric and
nonparametric statistical tests (e.g.,
see Sokal and Rohlf 1969, Marascuilo
and McSweeny 1977). Conversely,
using presence-absence data as we
did in analyses of trap success avoids
the problem of dependency. A short-
coming of using only presence-ab-
sence data, however, is that informa-
tion of the numbers and hence rela-
tive abundance of animals captured
might be lost.

r

Table 2.— Capture numbers of amphibians, reptiles, and small mammals at
three California oak woodlands: Tejon Ranch, Kern County; San Joaquin
Experimental Range, Madera County; and Sierra Foothill Range Reld Sta-
tion, Yuba County, from mid-January through mld-f^arch 1988.

Species

Tejon
Ranch
<n=8828)'

San Joaquin
Exp. Range
{n=8828)

Sierra Foothill
Range Field Stn.
(n=6912)

Bafrachoseps attenuafus^
Bafrachoseps nigrivenfris^
Ensafina eschscholfzP
Taricha forosa
Ran a boy Hi

Scaphiopus hammondii
Sceloporus occidentalism
Bum eces gilb erfF
Eumeces si<ilfonianu$^
Gerrhonotus mulficarinatus
Peromyscus maniculafus
P. boy Hi
P. truei

Perognafhus californicus
P. inomatus
Microfus californicus
Thomomys botfae
Scapanus latimanus
Sorex omafus

Total captures

Species richness

19
3

1

20
9

1

6
13
5
1

1
1
1

82
13

8

1

3
31
46

7
6
9

1

4

113
10

1

96

8
1
3
5
4

6
1

3
128
9

'Number of trap nigtits.

^Significant difference (P <0. 10) of captures among study areas.

CONCLUSIONS

Using pitfall traps to sample amphib-
ian, reptile, and small mammal
populations, we found pronounced
variation within and among study
areas, and within and between years
in capture rates of all taxa and of
many of the species studies. Implica-
tions of these results apply both to
the design of studies for these ani-
mals as well as for their manage-
ment. First, we recognize biases by
using only pitfall traps to sample
populations of free-ranging verte-
brates, and we suggest that research-
ers evaluate all possible methods to
determine the best one or combina-
tion of methods for the study of a
particular organism(s). Second,
within-year variation in capture rates
suggests that researchers should de-
sign a study to sample seasonal vari-
ations in activities and in habitat use.
Similarly, spatial variation, both
within and among stand types and
among distinct geographic locations,
should be studied to better identify
distributional limits of the species
studied and to determine how spe-
cific habitats contribute to the sur-
vival and reproduction of the spe-
cies. From a management perspec-
tive, understanding temporal and
spatial variability in habitat use is
critical when trying to provide suit-
able conditions for the animal to sur-
vive and reproduce. All oak wood-
lands cannot be managed in the same
way for all species. Each oak-wood-
land type contains a unique set of
factors that predispose species to use
the area for some aspect of their life
histories. Management for a species
should be based on information that
considers the spatial and temporal
variability in habitat use to provide
for all life requisites.

ACKNOWLEDGMENTS

We thank J. Bartolome, L. Brennan, J.
Dunne, and T. Pruden for construc-
tive comments on earlier versions of

251

this paper. M. Dixon, S. Kee, W.
Maynard, and T. Tennant assisted in
the field. We thank R. Barrett for use
of field equipment, and S. Lee, L.
Merkle, and I. Timossi for technical
assistance. D. Geivet, D. Duncan, and
J. M. Conner provided logistical as-
sistance at Tejon Ranch, San Joaquin
Experimental Range, and Sierra Foot-
hill Range Field Station, respectively.
University of California, Division of
Agriculture, Integrated Hardwood
Program; U.S. Forest Service, Pacific
Southwest Forest and Range Experi-
ment Station; and California Depart-
ment of Forestry, Forest and Range
Resource Assessment Program pro-
vided funding for this research.

LITERATURE CITED

Allen, T. F. H., and Thomas. B. Starr.
1982. Hierarchy: perspectives for
ecological complexity. 310 p. Uni-
versity of Chicago Press, Chicago,
111.

Block, William M. In press. Geo-
graphical variation in foraging
ecologies: breeding and nonbreed-
ing birds in oak woodlands. Pro-
ceedings of the symposium on
food exploitation by terrestrial
birds [Asilomar, CA, March 19-20,
1988].

Bury, R. Bruce, and Paul Stephen
Corn. 1987. Evaluation of pitfall
trapping in northwestern forests:
trap arrays with drift fences. Jour-
nal of Wildlife Management
51:112-119.

Day, Gerald I., Sanford D. Schem-
nitz, and Richard D. Tabor. 1980.
Capturing and marking wild ani-
mals, p. 61-88. In Wildhfe Manage-
ment Techniques Manual. 686 p.
Sanford D. Schemnitz, editor.
Wildlife Society, Washington, D.C.

Duncan, Don A., Neil K. McDougald,
and Stanley E. Westfall. 1987.
Long-term changes from different
uses of foothill hardwood range-
lands, p. 367-372. In Proceedings
of the symposium on multiple-use
management of California's hard-

wood resources [San Luis Obispo,
CA, November 12-14, 1986].
USDA Forest Service General
Technical Report PSW-100, 462 p.
Pacific Southwest Forest and
Range Experiment Station,
Berkeley, CA.

Fienberg, Stephen, E. 1980. The
analysis of cross-classified cate-
gorical data. 151 p. Second edition.
MIT press, Cambridge.

Halvorson, Curtis, H. 1984. Long-
term monitoring of small verte-
brates: a review with suggestions,
p. 11-25. In Research natural areas:
baseline monitoring and manage-
ment. Proceedings of a sympo-
sium [Missoula, Mont., March 21,
1984). USDA Forest Service Gen-
eral Technical Report INT-173, 84
p. Intermountain Forest and
Range Experiment Station, Ogden,
Utah.

Marascuilo, Leonard A. and Ma-
ryellen McSweeny. 1977. Non-par-
ametric and distribution-free

methods of the social sciences. 580
p. Brooks-Cole, Monterey, CA.

Plumb, Timothy R., and Norman H.
Pillsbury, technical coordinators.
1987. Proceedings of the sympo-
sium on multiple-use management
of California's hardwood re-
sources [San Luis Obispo, CA,
November 12-14, 1986]. USDA
Forest Service General Technical
Report PSW-100, 462 p. Pacific
Southwest Forest and Range Ex-
periment Station, Berkeley, CA.

Raphael, Martin G., and Kenneth V.
Rosenberg. 1983. An integrated
approach to wildlife inventories in
forested habitats, p. 219-222. In
Renewable resources inventories
for monitoring changes and
trends. 731 p. Oregon State Uni-
versity, Corvallis, Oregon.

Sokal, Robert R., and F. James Rohlf.
1969. Biometry: the principles and
practice of statistics in biological
research, 776 p. W. H. Freeman,
San Francisco, Calif.

Tqon Rondi Son Jooquin Sam FooM

Figure 2.— Relative numbers of captures using pitfall traps wittiin thiree oalc woodland study
areas in California: Tejon Rancti, Kern County, San Joaquin Experimental Range, Madera
County; and Sierra Foottiill Range Experimental Field Station, Yuba County. Trapping oc-
curred from 5 December 1987 to 20 Marcti 1988.

252

Vemer, Jared. 1987. The importance
of hardwood habitats for wildlife
in California, p. 162. In Proceed-
ings of the symposium on mul-
tiple-use management of Califor-
nia's hardwood resources [San
Luis Obispo, CA, November 12-
14, 1986]. USDA Forest Service
General Technical Report PSW-
100, 462 p. Pacific Southwest For-
est and Range Experiment Station,
Berkeley, CA.

Welsh, Hartwell, H. Jr.. 1987. Moni-
toring herpetofauna in woodland
habitats of northwestern Califor-

nia and southwestern Oregon: a
comprehensive strategy, p. 203-
213. /n Proceedings of the sympo-
sium on multiple-use management
of California's hardwood re-
sources [San Luis Obispo, CA,
November 12-14, 1986]. USDA
Forest Service General Technical
Report PSW-100, 462 p. Pacific
Southwest Forest and Range Ex-
f>eriment Station, Berkeley, CA.

253

The Importance of Biological
Surveys in Managing Public
Lands in the Western United
States'

Michael A. Bogan,^ Robert B. Finley, Jr.,^ and
Stephen J. Petersburg^

Abstract.— Despite previous studies, incomplete
knowledge of the mammalian fauna of many
national parks hinders our ability to understand the
consequences of either management actions or
natural disasters to such preserves. Fauna! losses
have occurred and con be expected to continue
(Newmark 1986a, 1986b). Our studies in and near
Dinosaur National Monument, one of the parks
studied by Nev^mark (1986a, 1986b), have added 1 1
species to the known fauna. Some species have
increased with human impact; other species hove
either disappeared or are declining. Finally, many
species, which are uncommon and poorly known,
may hove rather specific habitat needs.

era! information is available in only a
few sources (Cary 1911, Warren 1942,
Lechleitner 1969, Armstrong 1972),
each of which treats all Coloradoan
mammals. Detailed studies of this
area are not common and may be dif-
ficult to obtain (Durrant 1963, Bogan
et al. 1983). This paucity of knowl-
edge is frustrating not only to mam-
malogists, but also to land managers
seeking to protect the resources un-
der their care. In the absence of reli-
able information, land stewards may
end up managing for a relatively
small portion of the total fauna, pri-
marily those that are rare or endan-
gered, highly visible or popular,
pests, or those of importance to hunt-
ers and trappers.

Our studies in DNM and adjacent
Browns Park National Wildlife Ref-
uge, conducted since 1980, have pro-
vided new information on the mam-
mals of northwestern Colorado. In
addition, our data can provide a per-
spective on 1) the severity of the
problem of faunal loss as shown by
Newmark (1986a, 1986b) for one area
(DNM); and 2) the continuing need
for a better data base from which to
manage parks and their fauna and
flora. We summarize the gradual ac-
quisition of knowledge about mam-
mals in DNM, the contribution of re-
cent detailed studies to the faunal
data base, and how some species
seem to be responding to human ac-
tivity. Finally, we comment on some
of Newmark's (1986a, 1986b) data
and conclusions for DNM.

The equilibrium model of island bio-
geography (MacArthur and Wilson
1963, 1967) spawned a plethora of
studies that examined ways in which
various kinds of insular faunas be-
have (for mammals see Heaney and
Patterson 1986). Some of the most
interesting applications of the model
have been to animals in islands of
habitat, such as mountains in the
Great Basin (Brown 1971, 1978).
These studies revealed that such fau-
nas often behave in contrast to the
model, which predicts that the num-
ber of species on an island reflects an
equilibrium between processes of ori-
gin, i.e., species emigrating to the is-
land as a function of island size and
distance from the mainland, and
processes of extinction on the island.
Such studies lend support to the con-
tention that montane mammalian
faunas in the Southwest are not in
equilibrium (Brown 1986); rather,
they are relicts derived by extinction

'Paper presented of symposium. Man-
agement of Amphibians, Reptiles, and
Small Mammals in Norfti America. (Flag-
staff, AZ, July 19-21, 1988.)

^Michael A. Bogan is Wildlife Researcti
Biologist, U.S. Fisti & Wildlife Sen/ice. Na-
tional Ecology Research) Center, 1300 Blue
Spruce Drive, Fort Collins, CO 80524-2098.

^Robert B. Finley, Jr., retired from the U.S.
Fish & Wildlife Service, is Research Associ-
ate, The Museum, University of Colorado,
Boulder, CO 80309.

^Stephen J. Petersburg is Resource Man-
agement Specialist, National Park Service,
Dinosaur National Monument, P.O. Box 210,
Dinosaur, CO 81610.

from a set of colonizing species that
reached the mountains when life
zones were lowered during the Pleis-
tocene.

Newmark (1986a, 1986b, 1987) re-
cently examined ways in which west-
ern North American national parks
also behave, biologically, as islands.
Newmark's (1986a, 1986b) analysis of
data for 29 parks (data from only 24
were used in most analyses) in the
United States and Canada showed
that the number of mammalian spe-
cies in these parks is declining.

Newmark (1986a, 1986b) pre-
dicted that western national parks,
under a program of minimal man-
agement, could lose up to 100% of
the extant species of lagomorphs,
carnivores, and artiodactyls in the
next 100 to 200 years. This loss of
species would be dependent upon
the original size of the park (larger
areas have more species and larger
populations that persist better
through time), the degree of insulari-
zation of the parks (although most
parks presently are not completely
isolated, the more isolated they are,
the less likely they will be colonized
from outside), and intensity of man-
agement both within and outside
park boundaries.

One of the mammalian faunas in-
cluded by Newmark (1986a, 1986b)
in his analysis was that of Dinosaur
National Monument (DNM), located
in northwestern Colorado and adja-
cent Utah, where few studies of
mammals have been conducted. Gen-

254

Methods

Data were obtained from our studies
conducted in northwestern Colorado
since 1980. These studies, conducted
in riparian and upland habitats in
and near DNM, involved biological
surveys for mammals and their sign.
Mammals were observed, trapped
and released, and collected. Speci-
mens form a major part of our data
base, confirming the actual presence
of a species at a point in time.

Most habitats were sampled from
one to three nights with 250 to 300
live or snap traps each night. Traps
were set both in linear transects and
opportunistically; mist nets and other
methods were used for some species.
Our study sites included camp-
grounds, subjectively categorized
according to use by humans, as well
as isolated areas rarely visited by
humans. Although data from some

sites are directly comparable and sta-
tistically testable due to standardiz-
ing numbers of traps and techniques,
our purpose here is to present an
overview of the mammals at DNM
using all available information.

Data on distribution and abun-
dance of mammals in this part of
Colorado came from four primary
sources; these are Cary (1911), War-
ren (1942; a slightly revised version
of Warren 1910), Lechleitner (1969),
and Armstrong (1972). Studies of
nearby areas were consulted
(Kirkland 1981, Finley et al. 1984, Fin-
ley et al. 1976). Original surveys of
DNM by Durrant (1963) and Bogan
et al. (1983) were of value, as were
observations and reports by knowl-
edgeable park visitors and specimens
in collections. Historic accounts (e.g.,
Wishart 1979) of fur trappers and ex-
plorers of the nineteenth century
were reviewed for additional infor-

Table 1.— Numbers of species of mammals at Dinosaur Nationai IVlonument
per order as given in various reports on Colorado mammals (see text). Per-
centages in parentheses are the proportion of the total mammal fauna
that a given order represents.

REFERENCE

ORDER

Gary

Warren

Lechleitner Armstrong NewmarkThis pap

1911

1942

1969

1972

1986a

1986

INSECTIVORA

0

0

?

0

0

1

(1.5%)

CHIROPTERA

4

3

7

8

13

14

(21.5%)

LAGOMORPHA

3

3

3

3

4

4

(6.1%)

RODENTIA

18

18

21

20

19

25

(38.5%)

(SCIURIDAE

6

7

7

8

6

9)

(GEOMYIDAE

1

1

1

1

1

1)

(HETEROMYIDAE

1

2

2

2

2

3)

(CRICETIDAE

8

6

9

8

8

10)

(OTHER

2

2

2

1

2

2)

CARNIVORA

13

n

15

12

19

16

(24.6%)

ARTIODACP/LA

4

4

4

4

6

5

(7.6%)

TOTALS

42

39

50

47

62*

65

(%)

65

60

77

72

95

100

'Includes nine species that are not l<nown from DNM.

mation on the occurrence and disap-
pearance of some game species un-
documented by specimens.

Specimens of mammals from
DNM are contained in the University
of Utah Museum of Natural History
(UU), the University of Colorado
Museum (UCM), the Denver Mu-
seum of Natural History (DMNH),
and the Biological Surveys Collection
of the U. S. Fish and Wildlife Service
in Washington, DC (USNM), and
Fort Collins, CO (BS/FC). Original
field notes, photographs, and cata-
logs form an important part of this
data base and are available for in-
spection. Names of mammals follow
Banks etal. (1987).

Results and Discussion

IHistoric Data Acquisition

The growth in knowledge of the
mammals of DNM is shown in table
1. Data in Cary (1911), who worked
just east of the present Monument
and used both specimen data and his
own and others' reports, suggest that
about 42 species (65% of the species
listed in appendix 1) occurred in or
near DNM. Warren (1942), who did
limited work in northwestern Colo-
rado, provides information suggest-
ing that perhaps 39 species occurred
there. Lechleitner's (1969) general
treatise on Coloradoan mammals,
although not intended to provide de-
tailed information on distribution,
supports an expected fauna of about
50 species. Armstrong (1972), in the
first comprehensive study of Colora-
doan mammals, and building upon a
sixty-year data base, relied on speci-
men data to confirm the presence or
absence of mammals in a given area
and recorded 47 species (72% of
those currently known) for DNM or
nearby areas. Although some of these
references perhaps should not be
used to infer the specific occurrence
of species in a given area, we think
they are so used by land managers
and others.

255

During the period covered by
these references little actual work on
the mammals of DNM was con-
ducted. Exceptions were the work of
Hayward et al. (1958), Durrant and
Dean (1959, 1960), and Durrant
(1963) who chronicled the only extant
baseline data for many riparian areas
along the Colorado River and its ma-
jor tributaries (Green, Yampa) prior
to the impoundments at Flaming
Gorge and Glen Canyon.

Durrant (1963) surveyed for mam-
mals in DNM and reported 24 spe-
cies collected or observed, about 37%
of the known fauna. Two later sur-
veys for mammals and other verte-
brates in the Monument produced 29
(Bogan et al. 1983) and 27 (Bogan
unpubl. data) species, 45% and 42%
of the presently known fauna. Many
of the same species were obtained on
both trips.

Contributions of Recent Surveys

The known fauna of DNM includes
65 species (appendix 1) based on
specimens and reliable sight records.
Three species (Canis lupus, Ursus
arctos, and Bison bison) are now extir-
pated; we have omitted one species
of dubious occurrence (Mustela ni-
gripes). The percentage of mammal-
ian species at DNM by order is Insec-
tivora, 1.5%; Chiroptera, 21.5%; La-
gomorpha, 6.1%; Rodentia, 38.5%;
Carnivora, 24.6%; and Artiodactyla,
7.7%. Horses (Equus caballus) and
house mice (Mus musculus) occur at
DNM; we have excluded these intro-
duced species from our list.

What result have enhanced levels
of faunal surveys had on the known
fauna of DNM? Our work has added
11 species to the known fauna. These
include two state records [Per-
ognathus parvus and Euderma macula-
turn (Finley and Creasey 1982) from
Browns Park National Wildlife Ref-
uge, about 8 mi from DNM]; one
county record (Lepus californicus)
from DNM; seven Monument rec-
ords in 1982 (Myotis californicus, M,

thysanodes, Lasionycteris noctivagans,
Pipistrellus hesperus, Perognathus par-
vus, Microtus longicaudus, and M.
montanus); and three records for the
Monument in 1987 (Sorex monticolus,
Euderma maculatum, and Lemmiscus
curtatus).

These 1 1 species represent an in-
crease of 20.3% over the number pre-
viously known from DNM. Much of
this increase (five species) has come
by acquiring a better understanding
of the bats. This has been possible
because of better techniques of sur-
veying for bats, an improved under-
standing of continental and regional
distributions of bats, and an en-
hanced effort in surveying for bats at
DNM. Additional knowledge of
some other groups has come more
slowly, primarily because we are ap-
proaching the asymptote with re-
spect to species occurring in DNM.
The number of cricetid rodents
known or suspected to occur has in-
creased from eight to ten in 75 years;
that for sciurids has increased from
six to nine. Armstrong (1972) re-
ported 20 rodents known from
DNM; our records reveal a rodent
fauna of 25 species. For bats the fig-
ures are 8 in 1972 and 14 in 1987, an
increase of 75 7o.

The extent to which surveys reveal
previously unknown faunal compo-
nents is both fortuitous and regu-
lated by biological phenomena. The
capture of the first records of shrews
and spotted bats from DNM is partly
luck, by being in the right place at the
right time. Yet this ability to "test"
distributions of mammals by examin-
ing (trapping) suitable habitats re-
quires training, skill, and knowledge.
In addition, the ability to find rare
animals often requires removing the
more abundant and common species.

For example, of the 1,469 speci-
mens of small mammals that we
have captured at DNM, 52.6% have
been Peromyscus maniculatus. We
have taken 1,049 Peromyscus {7\A%
of the total trapped) as follows: P.
maniculatus, 772; P. truei, 175; P. crini-
tus, 102. There may be many reasons

why so many Peromyscus are taken;
our techniques may be biased in fa-
vor of them, they are easily trapped,
etc. Still, they are abundant relative
to other species of mammals on the
Monument.

We have no exact density figures
for P. maniculatus in DNM but ex-
trapolations are possible. The area of
DNM is 827 km^ or 82,700 ha; an av-
erage density for P. maniculatus
nnight be 20/ha (French et al. 1975),
or 1,654,000 deer mice. We suspect
that the densities at DNM are higher,
at least seasonally. A higher density
of 50/ha (French et al. 1975) would
yield 4,135,000 deer mice. If the aver-
age deer mouse weighs 20 g (a low
estimate), then the deer mouse bio-
mass at DNM is 33,080 kg to 82,700
kg; the equivalent of 144 to 360 adult
elk (Cervus elaphus) weighing 230 kg
each. The current resident elk popu-
lation of DNM is 150 to 200; up to
600 may be resident seasonally.

This abundance has several impli-
cations. One is that the common spe-
cies can fill the traps, reducing the
possibility of captures of other spe-
cies, and thus biasing the catch. More
interestingly, an accurate under-
standing that there are a few abun-
dant species and many uncommon
ones can provide information of
value in assessing impacts of human
activities and management of the
park, e.g., what species appear to be
increasing, those that are decreasing
or extirpated, those that are adjusting
their ranges, and those for which we
have insufficient information. Ex-
amples for these categories are dis-
cussed below.

Management Implications

Species Increasing in Abundance. —

Peromyscus maniculatus has been sug-
gested (Armstrong 1977, 1979) as one
species that increases in areas dis-
turbed by humans. It is a widespread
and adaptable species; whether it has
actually increased in some situations,
such as in campgrounds, may be de-

256

batable. Armstrong (in litt) has
noted that deer mice are weed spe-
cies and that rather than representing
a moral failure, they represent a suc-
cessful evolutionary strategy. P. man-
iculatus apparently always has been
common in this part of Colorado;
Gary (1911:103) stated that this spe-
cies was "exceedingly numerous de-
spite coyotes, hawks, and owls... in
western Routt [now Moffat] and Rio
Blanco Counties in 1906..." He re-
ports (1911:103) that in one case their
"excessive numbers all but pre-
vented my securing topotypes" of
another species, and that near Lo-
dore they were everywhere a "great
nuisance."

Our data from DNM reveal that
the canyon mouse (P. crinitus) is a
specialist of rocky canyon areas. It
does penetrate to the upper reaches
of some canyons but rarely does it
spread much further. The pinon
mouse (P. truei) is a specialist of pi-
non-juniper forests and occasionally
becomes moderately abundant.
Conversely, P. maniculatus is com-
mon in sagebrush (Artemisia sp.)
flats, a common upland habitat at
DNM. A comparison of relative
abundance of this species in subjec-
tively categorized "natural" and
"campground" situations reveals an
average of 22.4 animals/locality (n =
16) in areas where camping is of low
intensity or absent, versus an average
of 29.6 deer mice/locality in 14 heav-
ily-used areas. Although these num-
bers cannot be tested for significance,
due to non-uniform trapping proce-
dures, there is a difference in relative
abundance of P. maniculatus.

Another species that appears to
show a "campground" effect is the
golden-mantled ground squirrel
(Spermophilus lateralis). We have
taken this species in many areas and
it is widespread. Cary (1911:84) re-
ported that this species was "said to
be abundant" near Lily (just outside
the present Monument), and 7 mi N
of Lily they were reported to be "tol-
erably common," but Cary saw none
there the previous year. They are so

common in campgrounds of the
Monument now that they are a nui-
sance, albeit an attractive one. They
are fed by visitors and thus are en-
couraged to remain near the camp-
grounds. Our data from areas subjec-
tively categorized in terms of human
use reveals an average of 7.1 ground
squirrels from eight areas heavily
used by humans versus 1.2 animals/
locality in six little-used areas. In ar-
eas where golden-mantled ground
squirrels are very common we rou-
tinely close our traps during the day
to prevent being overrun with these
animals.

Species Declining or
Disappearing. — Those elements of a
fauna that disappear over time are
clearly of concern, and may provide
clues to habitat changes or other fac-
tors leading to faunistic changes. At
least three mammalian species are
now extirpated from DNM, and
likely from Colorado. These are the
gray wolf (Canis lupus), the grizzly
bear (Ursus arctos), and the bison (Bz-
son bison). Armstrong (1972) cites a
specimen of C. lupus from Douglas
Spring, near the present-day Monu-
ment. That gray wolves were com-
mon is shown by the fact that about
50 were killed by hired trappers in
Brown's Park in the winter of 1906-07
(Cary 1911). C. lupus was not in-
cluded in the DNM fauna by New-
mark (1986a).

No specimen of 17. arctos from or
near the Monument is known to us,
but there are reports of sightings in
the 1800s. About 60 fur trappers and
800 Indians wintered in Brown's
Park in 1839-40, during which time
they killed six grizzlies and 100 bison
for meat (Dunham and Dunham
1977). Fresh tracks of grizzlies were
seen in 1871 by members of the sec-
ond Powell expedition in Lodore
Canyon, a few miles above Echo Park
(Dellenbaugh 1926); and in 1891 Ann
Willis was rescued from a female
grizzly with two cubs in Zenobia Ba-
sin (Murie and Penfold 1983).

Remains of B. bison were exca-
vated from Hell's Midden, an occu-

pation site of the Fremont Culture in
Castle Park (Lister 1983). In addirion.
Walker (1983) reports the recovery of
remains of bison, as well as black
bear (U. americanus), pronghorn
(Antilocapra americana), mule deer
(Odocoileus hemionus), wapiti (Cervus
elaphus), and bighorn sheep (Oms ca-
nadensis), from Fort Davy Crockett in
Brown's Park. These remains date
from between 1836 and 1842. Ashley
saw several bison in Island Park in
1825 (Murie and Penfold 1983).

The dates of disappearance of
these species are speculative. B. bison,
which wintered in Brown's Park, was
already in decline west of the Conti-
nental Divide in the late 1830s, as ob-
served by concerned fur trappers
(Wishart 1979). According to Wishart
(1979), the Rocky Mountain trapping
system in Wyoming and Colorado
decayed not only because its main
fur-bearer, the beaver, was depleted
but also because the main source of
provisionment, the mountain bison,
was destroyed. Termination of the
fur trade in 1840 allowed mountain
bison to persist for several decades.
The last bison killed in northwestern
Colorado was at Cedar Springs west
of Craig in 1884 (Armstrong 1972).

C. lupus seems to have disap-
peared by 1935-40 (Young 1944,
Lechleitner 1969). The last report of
U. arctos in northern Colorado was
in 1920 in the Medicine Bow Range
(Armstrong 1972). Both species were
victims of increasing human en-
croachment and active predator con-
trol campaigns.

We have chosen to exclude the
black-footed ferret, Mustek nigripes,
from the known fauna of the Monu-
ment, for lack of specimens and
sightings, although it was included
by Newmark. Generally, the ferret
appears to have been a victim of the
active poisoning of its principal prey,
prairie dogs (Cynomys spp.) in addi-
tion to other factors (Clark 1986,
Rath and Clark 1986).

Newmark (1986a) stated that
wapiti (Cervus elaphus) should be
added to the list of mammals extir-

257

pated from DNM. Wapiti did occur
in the Monument in the early nine-
teenth century and are there today,
but their origin is questionable.

The present animals may be de-
scended from remnant populations
from elsewhere in parts of northern
Colorado or Utah, or from later in-
troduced wapiti from Wyoming. We
suspect they may be of mixed de-
scent.

Ovis canadensis occurring on the
Monument today may likewise be of
mixed descent. As noted by Fillmore
(unpubl. ms.) bighorn were common
and highly desired for food by trap-
pers and explorers in northwestern
Colorado in the first half of the 1800s,
but were greatly reduced by the
1880s, when they were protected by
the first game laws. Thereafter the
herds slowly increased until heavy
die-offs were caused by diseases
from domestic sheep. Such losses oc-
curred in Lodore Canyon between
1936 and 1945. By 1947 the superin-
tendent at DNM was ready to "write
them off." In 1954 the Colorado
Game and Fish Department made
two transplants in Lily Fark and
Zenobia Feak, and numbers since
have increased in the Monument
(Murie and Fenfold 1983).

At least two species may be ad-
justing their ranges relative to each
other in reciprocal fashion. We are
aware of no reports of Lepus californi-
cus in Moffat County prior to about
1980, although both specimens and
sightings of L. townsendii exist. In
1972 in western Colorado, the north-
ernmost locality for L. calif ornicus
was Mesa County (Armstrong 1972).
In the summer of 1987, we captured
both species, in close proximity, in
DNM. Based on the pattern of re-
placement seen elsewhere, including
the eastern plains of Colorado (Arm-
strong 1972), it is possible that the
range of L. townsendii is contracting
to the north and that of L. californicus
is expanding to the north. This re-
placement is commonly tied to land
use practices, especially breaking the
ground for cultivation, or over-

grazing, which may lead to increased
amounts of Opuntia (Armstrong
1972). Whether L. californicus is actu-
ally replacing L, townsendii at DNM
is debatable; what is not arguable is
that L. californicus is extending its
range northward in western Colo-
rado.

Species for Which Information is
Inadequate. — There are many species ,
for which scant information exists.
These species include most of the in-
sectivores, bats, and rodents, to-
gether composing 61.5% of the mam-
malian fauna of the Monument. Of
the 40 species in this category, almost
one-third were unknown at DNM
just 15 years ago. Much of this in-
crease comes from a better under-
standing of the bats, but knowledge
of their presence does not tell us if
there are important hibernacula for
bats on DNM, what proportion of the
bats may be migratory, or how best
to manage for this significant compo-
nent (22%) of the fauna. Similar com-
ments can be made for most of the
other small mammals, although few
are as vulnerable to mismanagement
and destruction as are bats (Hill and
Smith 1984).

Cottontails {Sylvilagus spp.) are
commonly seen, even abundant at
times, but it is difficult to identify
animals with certainty as the two
species (S. audubonii and S. nuttallii)
occurring at DNM are externally
similar. The two species overlap in
northwestern Colorado between ap-
proximately 6500 ft and 7000 ft and
specimens of both were collected by
Warren at Douglas Spring. The na-
ture of interactions between the two
species of cottontail at DNM is un-
known and studies based on speci-
mens are needed.

The raccoon was likely absent
from the park and probably the en-
tire upper Colorado River basin prior
to the 1950s (Durrant 1952, Long
1965). Specimens (BS/FC) indicate
that they moved into the upper
Green River and Brown's Fark in the
1960s and 1970s, probably from east-
ern Wyoming.

Newmark's Analysis Applied to
Dinosaur National Monunnent

Newmark's (1986a) analysis is im-
portant because it stimulates us to
consider a problem and assess its
magnitude, and also because he sug-
gests some solutions. He predicts a
depressing picture for some species
in national parks and there is clear
cause for concern. Still, it is useful to
put his analysis in perspective. New-
mark (1986a) lists 62 species of mam-
mals as occurring in DNM, including
E. caballus but not M. musculus. He
(1986a:21) confined his analysis to
only three orders, lagomorphs, carni-
vores, and artiodactyls 'l^ecause
these orders had the most complete
park sighting records. Species of
these orders tend to be more fre-
quently reported because of their
relatively large body size, non-fosso-
rial nature, and popularity." He also
used park sighting records as well as
continental (Hall 1981), statewide
(Armstrong 1972), and local (Ander-
son 1961) reports.

Those orders used by Newmark
(1986a) in his analysis include 39% of
the known mammalian species at
DNM. The most diverse order
(Rodentia) and the third most di-
verse order (Chiroptera) at DNM are
excluded. Furthermore, the 22 spe-
cies he does consider include the
only faunal losses (5) he believes oc-
curred in DNM. We believe that only
three species are extirpated from
DNM, and further suspect that most
of the extinctions occurred prior to
major expansion of the Monument' s
boundaries (1938).

However, the best management
decisions will be derived from the
most accurate data, and we should
try to obtain such data. We also be-
lieve that a holistic approach to ani-
mal management on public lands is
needed. This means including small
and secretive species in our plans, as
well as the large "glamorous" ones.
Newmark recognizes this in his rec-
ommendations; he notes the need to
develop a more extensive monitoring

258

program for vertebrate populations,
including key species of every order.

An examination of Newmark's
(1986a) data reveals that nine of the
62 species he lists for DNM do not
occur there: Plecotus rafinesquii (an
eastern bat perhaps listed due to a
misunderstanding of its taxonomy),
Tadarida brasiliensis (accidental at
best, no records for northern
Colorado), Lepus americanus (perhaps
confused by an observer with L,
townsendii in all-white winter pelage),
Glaucomys sabrinus (may possibly oc-
cur in higher areas of Douglas Moun-
tain at DNM but presently
unknown), Peromyscus boylii (perhaps
mistaken by an observer for the
large-eared P. truei), Vulpes velox (no
specimens north of Mesa County),
Gulo gulo (there is a specimen from
near the Utah-Colorado stateline,
outside the Monument), Mustek er-
minea, and Alces alces (accidental
stragglers only).

Why some of these species were
included by Newmark is unknown,
but in some cases it may have been
because they were listed in park rec-
ords, compiled from observations by
visitors and staff. We reexamined the
records at DNM and also found rec-
ords (mostly sightings) of Sorex cin-
ereus, Tamias umbrinus, Perognathus
flavescens, Ammospermophilus leucurus,
Neotoma lepida (perhaps juveniles of
N. cinerea), and Zapus hudsonius. We
know of no specimens to substantiate
these records and do not include
them in the fauna of the Monument.

These errors are not necessarily
Newmark's, although he may have
been uncritical in some instances, but
likely stem from several sources.
Among these are inadequate or lack-
ing baseline surveys, inaccurate rec-
ord-keeping by park staff, misunder-
standings of current nomenclature by
observers or recorders, unreliable
observations, and human error.
Nonetheless, these errors cloud our
understanding of mammals at DNM
and the management problems they
present. Additionally, although all
data and results age with time. New-

mark did not have the most current
information in many cases and thus
was unaware of recent records of
mammals from DNM.

Conclusion

Lists of species from a given area are
subject to interpretation. We have
taken a conservative approach rely-
ing on specimens (and giving reasons
for inclusions and exclusions where
appropriate) and have added signifi-
cantly to the known mammalian
fauna of DNM. Such lists are not
trivial exercises because they are the
raw materials for making land man-
agement decisions. Incorrect or miss-
ing data will diminish our ability to
manage these lands and their faunas.
We believe that biological surveys,
resulting in verified records (prefera-
bly specimens, but sometimes other
data), are the only reliable means to
determine the presence of a species
and to monitor population trends
over time. We agree with Newmark
(1986a) that such surveys need to be
undertaken immediately, because the
information is needed now; and
where surveys have been initiated
they should be continued on a regu-
lar basis. Monitoring of animal popu-
lations and the incorporation of accu-
rate data into rational management
plans is the only way to ensure that
our public lands continue to support
a diverse fauna that is as complete as
possible.

Acknowledgments

Many people have contributed their
time and efforts to learn more about
the mammals of Dinosaur National
Monument and vicinity. Chief
among them are: R. B. Bury, G. H.
Clemmer, R. D. Fisher, D. Finnic, K.
Hammond, D. Hogan, J. Hogan, M.
L. Killpack, C. A. Langtimm, D. Lan-
ning, B. Lapin, D. Leibman, S. J. Mar-
tin, C. A. Ramotnik, B. R. Riddle, A.
L. Riedel, D. E. Wilson, and D.

Worthington. Our boatpersons, who
also provided research support on
several trips, included R. Buram, H.
DeWitt, C. Frye, J. Rucks, and S.
Walker. Reports from Cloudridge
Naturalists, supplied by D. M. Arm-
strong, added materially to our
understanding of mammals in DNM.
J. Creasey, then-Refuge Manager of
Browns Park NWR, provided much
information and cooperative support
for field work. Early inspirational
support for this work came from A.
R. Weisbrod of the National Park
Service. We appreciate the comments
of D. M. Armstrong, R. B. Bury, C.
Jones, and F. L. Knopf on earlier ver-
sions of this paper.

Literature Cited

Anderson, Sydney. 1961. Mammals
of Mesa Verde National Park,
Colorado. University of Kansas
Publications, Museum of Natural
History, 14:29-67.

Armstrong, David M. 1972. Distribu-
tion of mammals in Colorado.
Monograph, University of Kansas
Museum of Natural History 3:1-
415.

Armstrong, David M. 1977. Ecologi-
cal distribution of small mammals
in the Upp)er Williams Fork Basin,
Grand County, Colorado. South-
western Naturalist 22:289-304.

Armstrong, David M. 1979. Distribu-
tional checklist of rodents in Can-
yonlands National Park, Utah. Oc-
casional Papers, The Museum,
Texas Tech University 59:1-44.

Banks, Richard D., Roy W. McDi-
armid, and Alfred L. Gardner
(eds.). 1987. Checklist of verte-
brates of the United States, the
U.S. Territories, and Canada. U.S.
Fish and Wildlife Service, Re-
source Publication 166:1-79.

Bogan, Michael A., Stephen J. Martin,
and R. Bruce Bury. 1983. Baseline
studies of riparian vertebrates of
the Yampa and Green river corri-
dors within Dinosaur National
Monument, June and July 1982.

259

Final (unpublished) report submit-
ted to Regional Office, Rocky
Mountain Region, National Park
Service, 16 p.

Brown, James H. 1971. Mammals on
mountaintops: nonequilibrium in-
sular biogeography. American
Naturalist 105:467-478.

Brown, James H. 1978. The theory of
insular biogeography and the dis-
tribution of boreal birds and mam-
mals. Great Basin Naturalist Mem-
oirs 2:209-227.

Brown, James H. 1986. Two decades
of interaction between the
MacArthur Wilson model and the
complexities of mammalian distri-
butions. Biological Journal of the
Linnean Society 28:231-251.

Gary, Merritt. 1911. A biological sur-
vey of Golorado. North American
Fauna 33:1-256.

Glark, Tim W. 1986. Technical intro-
duction. The black-footed ferret.
Great Basin Naturalist Memoirs
8:8-10.

Dellenbaugh, Frederick S. 1926. A
canyon voyage. Yale University
Press, New Haven, 277 p.

Dunham, Dick, and Vivian Dunham.
1977. Flaming Gorge country.
Eastwood Printing and Publishing
Company, Denver, 384 p.

Durrant, Stephen D. 1952. Mammals
of Utah. University of Kansas Pub-
lications, Museum of Natural His-
tory 6:1-549.

Durrant, Stephen D. 1963. Mammals
of Dinosaur National Monument,
p. 69-77. In Woodbury, Angus M.
(ed.). Studies of biota in Dinosaur
National Monument Utah and
Colorado. University of Utah, Di-
vision of Biological Sciences Mis-
cellaneous Papers 1:1-77.

Durrant, Stephen D., and Nowlan K.
Dean. 1959. Mammals of Glen
Canyon, p. 73-103. In Woodbury,
Angus M. (ed.). Ecological studies
of the flora and fauna in Glen Can-
yon. University of Utah, Anthro-
pological Papers (Glen Canyon
Series Number 7) 40:1-228.

Durrant, Stephen D., and Nowlan K.
Dean. 1960. Mammals of Flaming

Gorge Reservoir Basin. University
of Utah, Anthropological Papers
48:209-243.

Finley, Robert B., Jr., William Caire,
and Dallas E. Wilhelm. 1983
[1984]. Bats of the Colorado oil
shale region. Great Basin Natural-
ist 43:554-560.

Finley, Robert B., Jr., and James
Creasey. 1982. First specimen of
the spotted bat (Euderma macula-
turn) from Colorado. Great Basin
Naturalist 42:360.

Finley, Robert B., Jr., Dallas E. Wil-
helm, and William Caire. 1976.
Mammals of the Colorado oil
shale region. Annual Progress Re-
port (Unpubl.), National Fish and
Wildlife Laboratory, U.S. Fish and
Wildlife Service, Fort CoUins, xx p.

Rath, Dennis L. and Tim W. Clark.
1986. Historic status of black-
footed ferret habitat in Montana.
Great Basin Naturalist Memoirs
8:63-71.

French, Norm R., Dennis M. Stod-
dart, and B. Bobek. 1975. Patterns
of demography in small mammal
populations, p. 73-102. In Golley,
Frank B., K. Petrusewicz, and L.
Ryszkowski (eds.). Small mam-
mals: their productivity and popu-
lation dynamics. Cambridge Uni-
versity Press, Cambridge. 451 p.

Hall, E. Raymond. 1981. The mam-
mals of North America. Volumes
I, II. Second Edition. John Wiley
and Sons, New York. 1181+90 p.

Hayward, C. Lynn, D. Elden Beck,
and Wilmer W. Tanner. 1958. Zo-
ology of the upper Colorado River
basin. Brigham Young University
Science Bulletin, Biological Series
1:1-74.

Heaney, Larry R., and Bruce D. Pat-
terson. Eds. 1986. Island biogeo-
graphy of mammals. Academic
Press, New York. 271 p.

Hill, John E., and James D. Smith.
1984. Bats, a natural history. Univ.
of Texas Press, Austin. 243 p.

Kirkland, Gordon L., Jr. 1981. The
zoogeography of the mammals of
the Uinta Mountains region. The
Southwest. Naturalist 26:325-339.

Lechleitner, Robert R. 1969. Wild
mammals of Colorado: their ap-
pearance, habits, distribution, and
abundance. Pruett Publishing
Company, Boulder. 254 p.

Lister, Robert. 1983. The ancients of
the canyons, p. 48-57. In Stegner,
Wallace (ed.). This is Dinosaur.
Roberts Rinehart Inc. Boulder,
Colorado, 93 p.

Long, Charles A. 1965. The mammals
of Wyoming. University of Kansas
Publications, Museum of Natural
History 14:493-758.

MacArthur, Robert H., and Edward
O. Wilson. 1963. An equilibrium
theory of insular zoogeography.
Evolution 17:373-387.

MacArthur, Robert H., and Edward
O. Wilson. 1967. The theory of is-
land biogeography. Monographs
in Population Biology, Princeton
University Press 1:1-203.

Murie, Olaus, and Joseph W. Pen-
fold. 1983. The natural world of
Dinosaur, p. 31-47. In Stegner,
Wallace (ed.). This is Dinosaur.
Roberts Rinehart, Inc., Boulder,
Colorado, 93 p.

Newmark, William D. 1986a. Mam-
malian richness, colonization, and
extinction in western North
American national parks. Unpub-
lished Ph.D. dissertation. Univer-
sity of Michigan, Ann Arbor. 172

P-

Newmark, William D. 1986b. Spe-
cies-area relationship and its de-
terminants for mammals in west-
ern North American national
parks. Biological Journal of the
Linnean Society 28:83-98.

Newmark, William D. 1987. A land-
bridge island perspective on mam-
malian extinctions in western
North American parks. Nature
325:430-432.

Walker, Danny N. 1983. Faunal re-
mains and subsistence practices at
5MF605, a fur trading post in
northwestern Colorado. South-
western Lore 49:6-29.

Warren, Edward R. 1910. The mam-
mals of Colorado. G. P. Putnam's
Sons, New York, 300 p.

260

Warren, Edward R. 1942. The mam-
mals of Colorado. University of
Oklahoma Press, Norman. Second
revised edition. 330 p.

Wishart, David J. 1979. The fur trade
of the American west, 1807-1840.
University of Nebraska Press, Lin-
coln, 237 p.

Young, Stanley P. 1944. Their history,
habits, economic status and con-
trol, p. 1-173. In Young, Stanley P.,
and Edward A. Goldman. The
wolves of North America. Ameri-
can Wildlife Institute, Washington.
636 p.

Appendix 1

List of mammalian species from
Dinosaur National Monument.
Species are represented by
specimens in collections unless
otherwise noted in parentheses.
Those specimens not in the U. S.
Fish and Wildlife Service's
Biological Surveys Collections in
Fort Collins (BS/FC), or known only
from near the Monument, are so
noted in parentheses. See text for
species excluded from this list.
Additional information on
specimens or sight records is
available from the authors.

Sorex monticolus (Montane shrew)
Myotis californicus (California myotis)
Myotis ciliolabrum (Western small-
footed myotis)

Myotis evotis (Long-eared myotis)
Myotis lucifugus (Little brown bat; 5
mi SE Elk Springs, UCM)
Myotis thysanodes (Fringed myotis)
Myotis volans (Long-legged myotis)
Myotis yumanensis (Yuma myotis)
Lasiurus cinereus (Hoary bat)
Lasionycteris noctivagans (Silver-
haired bat)

Pipistrellus hesperus (Western pipis-
trelle)

Eptesicus fuscus (Big brown bat)
Euderma maculatum (Spotted bat)
Plecotus townsendii (Townsend's big-
eared bat)

Antrozous pallidus (Pallid bat)

Sylvilagus audubonii (Desert cotton-
tail)

Sylvilagus nuttallii (Nuttall's cotton-
tail)

Lqjus californicus (Black-tailed jack-
rabbit)

Lepus townsendii (White-tailed jack-
rabbit)

Tamias dorsalis (Cliff chipmunk)
Tamias minimus (Least chipmunk)
Tamias quadrivittatus (Colorado chip-
munk)

Marmota flaviventris (Yellow-bellied
marmot; Castle Park, UCM)
Spermophilus lateralis (Golden-
mantled ground squirrel)
Spermophilus elegans (Wyoming
ground squirrel; Two Bar Spring,
DMNH)

Spermophilus tridecemlineatus (Thir-
teen-lined ground squirrel)
Spermophilus variegatus (Rock squir-
rel)

Cynomys leucurus (White-tailed prai-
rie dog)

Thomomys talpoides (Northern pocket
gopher; Pot Creek, DMNH)
Perognathus fasciatus (Olive-backed
pocket mouse)

Perognathus parvus (Great Basin
pocket mouse)

Dipodomys ordii (Ord's kangaroo rat)
Castor canadensis (Beaver)
Reithrodontomys megalotis (Western
harvest mouse)

Peromyscus crinitus (Canyon mouse)
Peromyscus maniculatus (Deer mouse)
Peromyscus truei (Pinon mouse)
Onychomys leucogaster (Northern
grasshopper mouse)
Neotoma cinerea (Bushy- tailed
woodrat)

Microtus longicaudus (Long- tailed
vole)

Microtus montanus (Montane vole)
Lemmiscus curtatus (Sagebrush vole)
Ondatra zibethicus (Muskrat; Castle
Park, UCM)

Erethizon dorsatum (Porcupine; Pot
Creek near Pat's Hole, DMNH)
Canis latrans (Coyote)
Canis lupus (Gray wolf, +; Douglas
Spring, UCM)

Vulpes vulpes (Red fox; ca. Zenobia
Peak, Gary 1911)

Urocyon cinereoargenteus (Gray fox;

Castle Park, UCM)

Ursus americanus (Black bear)

Ursus arctos (Grizzly bear, +)

Bassariscus astutus (Ringtail; Castle

Park, UCM)

Procyon lotor (Raccoon)

Mustela frenata (Long-tailed weasel;

Castle Park, UCM)

Mustela vison (Mink; sightings in Lo-

dore Canyon)

Spilogale gracilis (Western spotted
skunk; Irish Canyon, ca. Lodore)
Mephitis mephitis (Striped skunk)
Taxidea taxus (Badger; Two Bar
Spring, DMNH)

Lutra canadensis (River otter; Yampa
Canyon, Warren 1942)
Felis concolor (Mountain lion; Grey-
stone, UCM)
Felis rufus (Bobcat)
Cervus elaphus (Wapiti)
Odocoileus hemionus (Mule deer; Pot
Creek, USNM)

Antilocapra americana (Pronghorn)

Bison bison (Bison, +)

Ovis canadensis (Bighorn sheep)

C+ = species is extirpated from the Monu-
ment)

261

Sampling Problems in
Estimating Small Mammal
Population Size^

George E. Menkens, Jr.^ and Stanley H.
Anderson^

Abstract.— Estimates of population size are
influenced by four sources of error: measurement,
sampling, missing data, and gross errors.
Measurement error can be reduced by using the
correct estimator, reducing variation in capture
probabilities, and by increasing sample size and trap
period length. Sampling error can be decreased by
increasing the number of grids trapped.

Species conservation and manage-
ment or analysis of environmental
impacts require accurate estimates of
population size. Because censusing
entire populations is difficult, if not
impossible, a sampling program is
generally employed to estimate ani-
mal abundance. In small mammal
studies, sampling is frequently per-
formed using live traps placed in
grids. Numerous approaches have
been used to estimate animal abun-
dance on trapping grids (e.g., catch-
per-unit effort, removal methods) but
capture-mark-recapture techniques
are the most commonly used (Seber
1986).

Four sources of error may influ-
ence an estimator's bias and preci-
sion (Cochran 1977, McDonald 1981).
Two, missing data and gross errors
(e.g., misreading tag numbers) are
"human" errors and can be avoided
by using careful field and laboratory
techniques. The remaining sources,

'Paper presented at symposium. Man-
agement of Amphibians. Reptiles, and
Small Mammals in North America (Flagstaff.
AZ.July 19-21, 1988).

^George E. Menkens, Jr., is a Research
Associate with the Wyoming Cooperative
Fish and Wildlife Research Unit,'' Laramie,
WY 82071.

'Stanley H. Anderson is Leader, Wyo-
ming Cooperative Fish and Wildlife Re-
search Unit," Laramie, WY 82071.

"Cooperators in the Wyoming Coopera-
tive Fish and Wildlife Research Unit include:
the Department of Zoology and Physiology,
University of Wyoming; Wyoming Game
and Fish Department: and the U.S. Fish and
Wildlife Service.

measurement and sampling error,
may, in many cases greatly affect an
estimate (McDonald 1981). Measure-
ment error is the error resulting from
the use of imprecise or biased (or a
combination of these) data collection
methods (McDonald 1981). In mark-
recapture studies, measurement er-
ror influences the bias and precision
of an estimate for any single grid.
Sampling variance is considered to
be a measurement error in mark-re-
capture studies (White et al. 1982).
Sampling error is error introduced
by natural variation between sam-
pling units, i.e., trap grids.

Potentially large sources of meas-
urement error in mark-recapture
studies may result from capture
probability variation and model se-
lection. All mark-recapture estima-
tors make specific assumptions about
capture probability variation within
and among animals and trapping
days. Three factors influencing indi-
vidual capture probability variation
have received attention (Burnham
and Overton 1969, Otis et al. 1978,
Pollock 1981, Seber 1982) and are
time, behavior, and individual
heterogeneity. Models assuming time
variation allow all animals to have
the same capture probability on a
given day, but this probability may
change between days. Models allow-
ing behavioral responses to trapping
assume all animals initially possess
identical capture probabilities, but
these probabilities may change upon
first capture. Capture probabilities

may increase (animals become trap
happy) or decrease (animals become
trap shy) after initial capture. Models
assuming that individual heterogene-
ity is present allow each animal to
have a unique capture probability
that does not change over time. (Com-
binations of these factors may also
occur. For example, an animal's cap-
ture probability may be influenced
by both time and behavioral effects.

Model selection is another source
of measurement error. Selection of an
inappropriate or incorrect model for
data analysis results in estimates
with unknown degrees of bias and
unacceptably large or unrealistically
small standard errors (Otis et al.
1978, White et al. 1982). CAPTURE
(Otis et al. 1978) is a widely used
computer program for estimating
population size using mark-recapture
data that also provides an objective
method for selecting the correct
model when any of the above
sources of capture probability vari-
ation are present.

In this paper, we investigate the
effects that variation in capture
probabilities due to time, behavior,
and individual heterogeneity have on
estimates of animal abundance and
model selection. We also discuss im-
provement of an estimate using data
pooling. We use these results to
show how reducing trap period
length influences estimator bias,
standard error, and confidence inter-
val coverage rate, and discuss how
this may help reduce the number of

262

grids required to detect a given dif-
ference between yearly estimates of
population sizes.

Material and Methods

To investigate effects of both capture
probability variation and trap period
reduction, we used program CAP-
TURE (Otis et al. 1978) to randomly
generate and analyze data sets with
known population characteristics
(see Menkens 1987 for details). CAP-
TURE contains eight models, five
with estimators, for estimating popu-
lation size for closed populations
when capture probabilities do not
vary (model M(o)), or when they
vary with time (model M(t)), behav-

ioral response (model M(b)), individ-
ual heterogeneity (model M(h)) or a
combination of the behavioral and
individual heterogeneity models
(model M(bh)). Using CAPTURE, we
specified the number of trapping pe-
riods, population size, and capture
probabilities, and patterns of vari-
ation. CAPTURE was then used to
analyze each data set.

We analyzed the same data sets
using Chapman's unbiased version
of the Lincoln-Petersen estimator and
its variance estimator (Seber 1982).
Because the Lincoln-Petersen estima-
tor uses data from only two periods,
each data set was split prior to esti-
mation. Thus in a 5 day trapping
study, the first 3 days constituted the
marking period, and the second 2

Table 1 .—Capture probability patterns used in simulations (from Menkens
1987 and Menkens and Anderson in press). Good capture probabilities are
defined as being large (generally > 0.30) with little difference (about 0.15)
between the highest and lowest capture probability. Poor capture proba-
bilities are defined as being low with large differences between the highest
and lowest capture probability. "Model" refers to the CAPTURE model un-
der which the data were generated. See the text for description of the
model abbreviations, p = capture probability, c = recapture probability
(trap shyness = p(0.50), trap happiness = p(1.50)), all simulations were run
for 5 and 1 0 day capture periods (t).

Model

Poor

Good

M(o)
M(h)
M(b)
M(bh)

M(t)

P=.l

p = 0,05, 0.10, 0.25'
p = 0.10, c = 0.50
p = 0.05, 0.20, 0.40
c = 0.50

p = 0.05. 0.20. 0.40
c = 0.50 or 1.502
p = 0.10. 0.15, 0.05,
0.15,0.10
t = 5

p = 0.10, 0.10, 0.15,

0.15,0.05,0.05,

0.15.0.15,0.10,

0.10

t= 10

p = ,5.

p = 0.40, 0.50, 0.60'
p = 0.50, c = 1.50
p = 0.20, 0,30, 0.40
c=: 1.50

p = 0.20. 0,15, 0.25
c = 0.50 or 1.502
p = 0.50, 0,55, 0,40,
0.55, 0.50
t = 5

p = 0.50, 0.50, 0.55,
0.55,0.40, 0.40,
0,55,0,55,0.50,
0.50
t= 10

'Three groups of animals were assumed to be present in the population, the first group
was associated with the first capture probability, the second group with the second
capture probability, the third group with the third capture probability. For N= 50, ani-
mals 1-20 were in group 1, 21-40 in group 2, 41-50 were in group 3. For N= 100, animals
1-40 were in group 1, 4 1-80 were in group 2, 81-100 were in group 3.

'When a heterogeneous recapture probability was assumed, half of the animals
became trap shy, half became trap happy.

days was the recapture period. In
studies 10 days long, the first 5 days
were the marking period, the second
5 days the recapture period.

Data were generated for a wide
range of conditions. We used trap
periods of 5 and 10 days, population
sizes of 50 and 100 and a wide vari-
ety of capture probability patterns
(table 1). One thousand data sets
were generated for each combination
of these conditions. In this paper, we
only generated data meeting the as-
sumptions of one of the five models
with estimators in CAPTURE. For
each data set, CAPTURE was forced
to perform the analysis using the cor-
rect model. For example, if data were
generated under the assumption of
time variation, CAPTURE was forced
to use model M(t) for the analysis.
Simulations were also performed us-
ing the same, and additional, capture
probabilities (table 1), with CAP-
TURE being allowed to select an esti-
mator using its model selection pro-
cedure.

Results

Performance of both the Lincoln-Pe-
tersen estimator and CAPTURE is
dependent upon the size and magni-
tude of the variation in capture
probabilities (table 2). Estimators
have lower degrees of bias, smaller
standard errors, and higher confi-
dence interval coverage rates when
capture probabilities are high and
their variation is low (tables 1 and 2)
over all population sizes. When cap-
ture probability variation is constant,
the estimator's bias tends to decrease
and confidence interval coverage
rates increase with increasing popu-
lation size (table 2). Although this
pattern is evident for standard er-
rors, patterns of change with increas-
ing sample size are not as clear (table
2).

In general, the estimator's bias de-
creases and confidence interval cov-
erage rates increase as trapping pe-
riod length increases (table 2). This

263

pattern is not as obvious for standard
errors, although they do tend to im-
prove with increasing trap period
length (table 2). In most cases the
magnitude of change in bias is
smaller for good capture probabili-
ties than for poor capture probabili-
ties when the trapping period in-
creases (table 2). Although estimated
standard errors tend to decrease
with lengthening trap periods (more
so with good capture probabilities),
the magnitude of this change is gen-
erally smaller than is change in bias
(table 2). As with bias and standard
error, confidence interval coverage
rates improve as trapping period in-
creases; the magnitude of change
tends to be larger when capture
probabilities are poor (table 2).

Except when data were generated
under model M(o), CAPTURE se-
lected the correct model less than
11% of the time (table 3). The Lin-
coln-Petersen estimator failed to pro-
vide an estimate at most 7% of the
time (table 3).

Discussion

In small mammal studies, measure-
ment errors can significantly influ-
ence an estimator's bias and preci-
sion. This study shows the impor-
tance of both reducing capture
probability variation and increasing
the size of those probabilities on
measurement error. Decreasing cap-
ture probability variation reduces the
estimate's bias and coefficient of
variation, and increases its confi-
dence interval coverage rate. This
result has also been stressed by
Burnham and Overton (1969), Menk-
ens (1987), Menkens and Anderson
(in press), Otis et al. (1978), and
White et al. (1982). Of particular sig-
nificance is the need to reduce vari-
ation due to behavioral responses
(i.e., trap-happiness and shyness)
and individual heterogeneity, espe-
cially when these factors act in con-
cert (Menkens and Anderson in
press, Otis et al. 1978, White et al.

Table 2. -r Simulation results for N = 50 and 100 when CAPTURE was forced to use tiie
correct estimator. CAPTURE refers to the appropriate CAPTURE nrtodel for analysis, L-P =
lincoln-Petersen estimate, Model is the model under which the data were generated
by CAPTURE, P = poor capture probabilities, G = good capture probabilities (see table
1 for definitions), t = length of trapping period <in days), PRB = percent relative bias, SE
= empirical standard error, CIC = confidence interval coverage rate.

IModel

L-P

CAPTURE

50

100

SO

100

M(o)
t = 5

M(h)

t = 5
PRB
SE
CIC

-55.4
1.1
16,2

G

PRB

-52.3

0.02

-34.9

-0,5

-9.0

1,4

17.3

-0.2

SE

0.5

0.2

0.9

0,3

1,0

0,2

2.3

0.3

CIC

27.6

87,0

57.1

91,2

80.7

90.7

87.5

92.1

t= 10

PRB

-11.5

0.0

-2,5

0,0

18.3

-0.10

7.1

-0,50

SE

0.6

0.1 ■

1,1

0,1

1,1

0.1

1.2

0.1

CIC

' 73.8 ■

92,3

83,8

93,4

89,1

92.1

91.6

93,2

-1.2
0.2
86.4

-40.9
1.6
40.1

1.6
0.3
89.0

-49.4
0.3
22.8

12.8
0.3
81.1

-43.0
0.4
6.0

16.5
0.5
67.1

PRB

-28,2

-1,0

-24.9

-0.7

-16,1

7.2

-10,0

9.5

SE

0.6

0,1 ,

1.0

0.1

0.5

0,2

0,8

0.2

CIC

48.0

87.9

52.4

91.4

52.8

92.3

59,9

63.2

M(b)

t = 5

PRB

-52.0

-12.2

-30.5

-12.1

-70.0

2.2

-59.1

3.2

SE

1.6

0.2

1.9

0.4

0.3

0,6

0.7

0.9

CIC

9.4

52,4

34.1

40.0

13.3

81.4

36.4

87.3

t= 10

PRB

6.2

-4,0

43,7

-3.7

-33.3

. -1.4

-12.3

-0.6

SE

0.7

0,1

2.3

0.1

0.6

0.1

1.7

0.1

CIC

66,6

54.3

91.3

43.6

497

82.4

64.8

89.9

M(bh) (set 1)

t = 5

PRB

-23.5

26.0

-23.1

45.8

-17.8

-17.8

-1.5

-1.5

SE

0,5

0.9

0.8

2.2

0.6

0.6

1.7

1.7

CIC

49.5

88,0

41.3

99.1

60.7

60.7

70,9

70.9

t= 10

PRB

-0.4

-13.8

-13.7

27.9

-28.8

-4.1

-24.6

-3.5

SE

0.5

0.3

0.5

1.1

0.4

0,4

0.8

0.6

CIC

83.4

43,1

27.5

92.5

40.5

69,6

43.2

75.2

(set 2)

t = 5

PRB

-32.5

-17,8

-31.3

-3.6

-31.3

-40,6

-18.8

-22,1

SE

0.6

0.6

1.1

1.1

0.4

0,4

1.2

1.2

CIC

33.6

64,8

24.4

82.4

46.2

47,6

67.2

64,9

t= 10

PRB

-20.5

0,2

-20.2

-0.2

-28.8

-4,9

-24.6

1,7

SE

0,4

0,4

0.7

0.5

0.4

0,6

0.8

1,4

CiC

32.1

86,7

17.2

91.2

40.5

67,7

43.2

73,5

M(t)

t = 5

PRB

-0.2

0,0

■0.4

0,0

3,8

-1,2

3.5

-0,6

SE

0.6

0,1

0.6

0,1

0.9

0,1

1.0

0.1

CiC

85.6

91,1

90.6

93,7

79.8

78,1

89.0

87.4

t= 10

PRB

-0.2

0,0

-0.1

0,0

-6.4

0,0

-0.8

-0.1

SE

0.3

0,1

0.2

0,1

0.3

0,1

0.3

0.1

CIC

87.4

95.3

92.4

92,7

74.6

95,3

90.7

90.6

264

1982). Reduction of rime variarion,
parricularly if the Lincoln-Petersen
estimator is used, is important, but
not as crirical (Menkens 1987, Menk-
ens and Anderson in press). Again,
reducing variarion in capture proba-
bilities leads to estimates that have
lower bias and increased precision.

Methods for reducing variation in
capture probabilities are numerous
(see Oris et al. 1978, Seber 1986,
White et al. 1982). Behavioral re-
sponses may be reduced by the use
of different capture and recapture
techniques. For example, animals
could be captured using live traps
and marked, and then "recaptured"
visually using spotting scopes.
(Fagerstone and Biggins 1986). In
addirion, use of traps not avoided by
animals, and use of non-intrusive
marking techniques (e.g., ear tags
instead of toe clipping) may also help
reduce behavioral responses. Use of
traps not avoided by animals may
help increase capture probabilities. If
sample sizes are large, heterogeneity
may be reduced by stratifying the
data into sex and age groups with
separate analyses performed on each
group (Oris et al. 1978, White et al.
1982). If data are strarified however,
the effects of small sample size on the
esrima tor's properries must be con-
sidered.

Capture probabiliries may be in-
creased and their variarion reduced
after study completion by fxx)ling
individual trap periods into single
marking and recapture periods as
was done in our simulations (Menk-
ens 1987, Menkens and Anderson in
press). When data are pooled in this

way and the Lincoln-Petersen estima-
tor used, capture probabilities are 20
to 25% higher than those for individ-
ual days (Menkens 1987). In most
cases, data pooling results in esti-
mates with improved properries.

Use of the wrong model for analy-
sis leads to estimates with unknown
degrees of bias and unacceptably
large or unreasonably small standard
errors (Oris et al. 1978, White et al.
1982), thus contributing significantly
to measurement error. In this study,
we forced CAPTURE to use the cor-
rect model for analysis. This, is unre-
alisric however, in that biologist
never know which model is appro-
priate. CAPTURE provides a objec-
tive model selection procedure, how-
ever this procedure works poorly
with the small sample sizes typically
encountered in many field studies
(Menkens 1987, Menkens and Ander-
son in press. Oris et al. 1978, White et
al. 1982). In most cases, the Lincoln-
Petersen estimator is a valid alterna-
tive to CAPTURE when sample sizes
are small, except when capture
probabilities are influenced by severe
behavioral responses or large de-
grees of individual heterogeneity
(Menkens 1987, Menkens and Ander-
son in press). Because use of the most
appropriate model is critical, CAP-
TURE should be used to determine
the type and magnitude of capture
probability variation in a data set,
and if variation is low, the Lincoln-
Petersen estimator should be used in
analysis (Menkens and Anderson in
press).

Many additional factors contribute
to measurement error. Eliminating

these requires detailed knowledge of
species behavior and ecology, and
use of compatible techniques. For ex-
ample, baits identical to, or that
closely approximate natural food
items, should be used (Dobson and
Kjelgaard 1985). Tags that are easily
lost will lead to severe overestimates
of population size and should not be
used. Other factors that could con-
tribute to measurement error include
use of traps or other activities that
decrease survival or increase emigra-
tion or immigration, and use of im-
proper traps for the species.

Sampling error is the error that
results from natural variation be-
tween sampling units; the larger this
variation, the larger the number of
units that must be sampled to detect
a difference in population size. For
example, when environmental im-
pacts are being assessed, sampling
error would be decreased by increas-
ing the number of grids in the control
and experimental groups. Reducing
variation in capture probabilities al-
lows decreasing the number of days
each grid is trapped without large
increases in bias or standard errors
or decreases in confidence interval
coverage rates. By reducing the trap
period, more grids can be sampled in
a shorter period of time, thereby re-
ducing sampling error and improv-
ing the estimate of overall population
size. Trapping in as short a time
interval as possible will also decrease
variation caused by temporal popu-
lation effects.

One approach to reducing sam-
pling error is to reduce intergrid
variation by using a stratified sam-
pling approach. In this case, investi-
gators could stratify the habitat
based on some characteristic that is
correlated with animal density and
trap within these strata. Sample sizes
would be estimated for each strata.

Conclusions

Reduction of capture probability
variation and maximizing their mag-

Table 3.— The percentage of times the Lincoln-Petersen estlnnator (LP) or the
appropriate CAPTURE model (CAPTURE) was selected by CAPTURE'S model
selection routine (from Menkens 1987). Model refers to the CAPTURE model
under which the data were generated.

Model

Estimator

M(o)

M(h)

M(b)

M(bh)

M(t)

LP

CAPTURE

96
100

98
9

93
7

99
11

97
6

265

nitude are critical to obtaining unbi-
ased and precise estimates of popula-
tion size, and also allow the selection
of the proper model for use in analy-
sis. Although we have concentrated
on small mammals, our points con-
cerning reduction of both variation in
capture probabilities and of measure-
ment and sampling error, pertain to
other studies using mark-recapture
techniques (e.g., papers in Ralph et
al.). Our conclusions will hopefully
force investigators to realize that
their techniques, particularly in
poorly designed and carelessly per-
formed studies, may not provide as
detailed and profound conclusions as
they might expect. We reiterate that
care in designing a study can mini-
mize many (but not all) of the
sources of measurement and sam-
pling errors we have discussed.

Acknowledgments

We thank Mark Boyce, Marc Evans,
Richard Greer, Lyman McDonald,
and Brian Miller for comments on
earlier versions of this manuscript.
The Department of Zoology and
Physiology, the University of Wyo-
ming provided the computer time
used for the simulations reported
here.

Literature Cited

Burnham, Kenneth P. and William S.
Overton. 1969. A simulation study
of live-trapping and estimation of
population size. Oregon State Uni-
versity, Department of Statistics,
Technical Report 14. 60 p. and ap-
pendix.

Cochran, William G. 1977. Sampling
Techniques. John Wiley and Sons,
New York.

Dobson, F. Stephen and Julia D.
Kjelgaard. 1985. The influence of
food resources on population dy-
namics in Columbian ground
squirrels. Canadian Journal of Zo-
ology 63:2095-2104.

Fagerstone, Katheleen A. and Dean
E. Biggins. 1986. Comparison of
capture-recapture and visual
count indices of prairie dog densi-
ties in black-footed ferret habitat.
Great Basin Naturalist Memoirs
8:94-98.

McDonald, Lyman L. 1981. "Summa-
rizing remarks: Observer
variability. In Estimating numbers
of birds. Studies in Avian Biology
Number 6, p. 426-435. J.M. Scott
and C.J. Ralph, editors. The Coo-
per Ornithological Society.

Menkens, George E., Jr. 1987. Tempo-
ral and spatial variation in white-
tailed prairie dog (Cynomys
leucurus) populations and life his-
tories in Wyoming. Unpublished
Ph.D. Dissertation, University of
Wyoming, Laramie.

Menkens, George E., Jr. and Stanley
H. Anderson, in press. Estimating
small mammal population sizes.
Ecology.

Otis, David L., Kenneth P. Burnham,
Gary C. White, and David R. An-
derson. 1978. Statistical inference
from capture data on closed ani-
mal populations. Wildlife Mono-
graphs 62:1-135.

Pollock, Kenneth H. 1981. Capture-
recapture models: a review of cur-
rent methods, assumptions and
experimental design. In Estimating
numbers of birds. Studies in Avian
Biology Number 6, p. 426-435. J.M.
Scott and C.J. Ralph, editors. The
Cooper Ornithological Society.

Ralph, C. John and J. Michael Scott
(eds). 1981. Estimating numbers of
birds. Studies in Avian Biology
Number 6, p. 426-435. The Cooper
Ornithological Society.

Seber, G.A.F. 1982. The estimation of
animal abundance. MacMillan
Publishing Company, New York.

Seber, G.A.F. 1986. A review of esti-
mating animal abundance. Bio-
metrics 42:267-292.

White, Gary C, David R. Anderson,
Kenneth P. Burnham, and David
L. Otis. 1982. Capture-recapture
and removal methods for sam-
pling closed p)opulations. Los
Alamos National Laboratory, LA-
8787-NERP.

266

The Design and Importance
of Long-Ternn Ecological
Studies: Analysis of
Vertebrates In the Inyo- White
Mountains, California^

Michael L. Morrison^

Abstract.— This paper reviews the importance of
duration in the design of studies of wildlife-habitat
relationships. Long-term studies are especially suited
to examining slow processes, rare events, subtle
processes, and complex phenomena. Four major
alternatives to long-term studies— retroactive studies,
substitution of space for time, use of systems with fast
dynamics as analogues for systems wrth slow
dynamics, and modeling— are discussed. All studies
should justify their results and (especially)
conclusions-recommendations with regard to study
duration. A suggested design for a long-term study
of small vertebrates is presented, including
preliminary data (as an example) from the Inyo-
White Mountains of eastern California.

A fundamental quesrion that should
arise early in the design process of
any investigation is the duration of
study. Along with questions of sam-
pling methods, sample size, seasons
of study, and the like, is the central
question of how long to collect data:
is 1 week or 1 month ample? Or
should the study extend for 1 or
more years? Naturally, this is a
study-specific question based largely
on the objectives of the research. As I
show in this paper, however, a study
of insufficient length may fail to at-
tain its objectives regardless of the
strength of the design components
(e.g., sample size). Unfortunately, the
researcher and manager may not
even realize that the study gave only
a partial picture of the system under
study; this, then, raises the issue of
study length.

As outlined elsewhere (e.g.. Likens
1983, Wiens 1984, Strayer et al. 1986),
a tradition has developed over the
past several decades — especially
among North American scientists — of
the pursuit of short-term studies.
This situation arose from constraints
imposed by funding duration, the
need to finish graduate programs

'Paper presented at symposium. Man-
agement of Amphibiarxs, Reptiles, and
Small Mammals in North America (Flagstaff.
AZ.July 19-21. 1988).

^Associate Professor of Wildlife Biology.
Department of Forestry and Resource Man-
agement. University of California. Berkeley,
CA 94720.

within short periods of time, the
pressure placed on researchers to
publish, and, of course, human na-
ture. A quote from John A. Wiens
(1984) in his review of long-term
studies in ornithology is appropriate
here: "...an excessive preoccupation
with short-term studies can lead to
short-term insights. By restricting the
duration of investigation, we adopt a
snapshot approach to studying na-
ture. We can only hope that the
glimpses of patterns and processes
that we obtain depict reality accu-
rately and that something critical has
not been missed because we looked
at the system too briefly." These final
thoughts — that the pattern we saw
may not depict reality, and that a
critical factor may have been
missed — have direct implications for
the design of future wildlife-habitat
relationships studies. Such studies
are usually of only 1-3 years in dura-
tion. At best, they give only a partial
view of most ecological systen\s; and,
at worst, lead to false interpretations.

My objectives in this paper are (1)
to compare and contrast short- and
long-term studies, including discus-
sion of when each type of study can
be most useful; I will draw heavily
from the comprehensive review of
long-term ecological studies by
Strayer et al. (1986). (2) Using a shidy
recently implemented in the Inyo-
White mountains of eastern Califor-
nia, I will suggest a design for long-
term studies that seeks to deternrune

trends in abundance and habitat rela-
tionships of small vertebrates.

LONG-TERM STUDIES

Conceptual Framework

As sunrunarized by Strayer et al.
(1986), long-term studies are espe-
cially suited to exploring four major
classes of ecological phenomena:
slow processes, rare events, subtle
processes, and complex phenomena.

Slow Processes

Long-term studies obviously can
contribute to the understanding of
ecological processes that exceeds that
gained from studies of only 1-3 years
in duration. The importance of this
contribution depends on the magni-
tude of the process: results obtained
from any several-year period of the
hypothetical 25-year curve (fig. lA)
could differ substantially from other
periods (e.g., showing an increasing
or decreasing trend). Data obtained
during any short period could be ac-
curate, but only for that period. Al-
though continuous sampling may not
be necessary to identify such a rela-
tionship, certainly regularly-repeated
sampling is. Prominent examples of
such slow processes given by Strayer
et al. (1986) are forest succession, in-
vasion of exotic species, and verte-

267

brate population cycles. Several spe-
cific examples of obvious long-term
relationships or cycles are given in
Halvorson (1984): the 23-fold differ-
ence between peaks and low num-
bers of snowshoe hares (Lepus ameri-
canus) during a 15-year study by
Keith (1983); and it took 12 years for
a relationship between conifer seed-
crop and red squirrel (Tamiasdurus
hudsonicus) abundance to be repeated
(Halvorson, unpubl. data).

Rare Events

Ecological phenomena can occur at
regular intervals (fig. IB); such
events include catastrophes (e.g.,
fires, floods), population eruptions,
and various environmental 'lx)ttle-
necks" or "crunches." Shorter-term
studies are often used to study such
events after their occurrence, focus-
ing on the response or recovery of
the system. Studies of post-fire suc-
cession (e.g.. Bock and Lynch 1970,
Raphael et al. 1987), and changes in
bird populations following oceanic El
Nino conditions (e.g.. Barber and
Chavez 1983, Schreiber and Schreiber
1984), are a few examples. Short-term
studies, cannot, however, be used to
study the frequency and reason (con-
text) for the event.

Subtle Processes

Here Strayer et al. (1986) identified
processes that change over time in a
regular fashion (e.g., monotonic
change, a step-function), but where
the year-to-year variance is large
relative to the magnitude of the
longer-term trend (as depicted in fig.
IC). According to Strayer et al., "A
short-term study will be unable to
discern the long-term trend, or, even
worse, will suggest a completely in-
correct conclusion about the magni-
tude and direction of the change... A
short-term record simply lacks the
statistical power to detect subtle
long-term trends..."

Complex Phenon^no

Evaluation of biological phenomena
are often complicated by the intercor-
related nature of associated environ-
mental factors. Further, relationships
between dependent and independent
variables may be characterized by
both linear and nonlinear responses
(e.g., Meents et al. 1983). Long-term
data are often necessary to sort out
such relationships for several rea-
sons. First, it may simply take many
years for the phenomenon to reveal
enough of its characteristics to allow
meaningful analysis (e.g., to model
the system). Further, it may be neces-
sary to accumulate data for many
years to provide the necessary statis-
tical degrees of freedom to conduct
complex analyses (e.g., multivariate
statistics; Strayer et al. 1986).

Other Considerations. — A myriad
of other, often related, factors indi-
cate the need for long-term studies.
Many of these factors are related to
the basic — albeit complex — biology of
the organism. Vertebrates have long
generation time and long life spans,
which tends to mask a population
response to environmental change.
Site fidelity, another common charac-
teristic of adult vertebrates, may
cause a time-delay in the response of
an animal to perturbation.

How Long is Long-Term?

Strayer et al. (1986) gave two, rather
different, definitions to the concept
of "long-term." The first definition
considers the length of study in
terms of natural processes. Quoting
them, a study is long-term "...if it
continues for as long as the genera-
tion time of the dominant organism
or long enough to include examples
of the important processes that struc-
ture the ecosystem under study.. .the
length of study is measured against
the dynamic speed of the system
being studied."

A different approach is to view
the length of studies relatively, with

long-term studies being those that
have continued for a longer time
than most other such studies. By fol-
lowing this definition, we are accept-
ing human institutions and con-
straints (e.g., human life span, length
of graduate education, pressure to
publish), and not the rate of natural
processes (Strayer et al. 1986).

A

16

YEARS

Figure 1 .—Situations where long-term stud-
ies rTKiy k>e useful. (A) slow processes, (B)
rare events, and (C) subtle changes. The
record in (C) is a long-term trend beginning
at Y = 4 and increasing at 5% per year (dot-
ted line) to which a random error with a
variance equal to the trend line has been
added. Redrawn with permission following
strayer et al. (1986: fig. 3).

268

To illustrate the difficulty in defin-
ing the length of time necessary for a
study to be considered long-term,
Strayer et al. (1986) contrasted the
classic experiment on competitive
exclusion in Paramecium with the for-
est ecosystem studies at the Hubbard
Brook Experimental Forest: Cause
took about 20 days to elucidate the
dynamics of the Paramecium system;
the recovery of a forest ecosystem
from clearcutting has been underway
for 20 years, which is perhaps only
1 / 20 of the time necessary for the
forest to reach steady-state. By the
first definition. Cause's work is long
term, while the 20-year Hubbard
Brook work is not; the latter becomes
"long-term" under the second defini-
tion.

Thus, one cannot establish a for-
mal definition for "long term." Re-
searchers should recognize, however,
that conclusions drawn from any
study should consider the dynamic
speed of the system being studied.
As reviewed by Likens (1983), there
are numerous examples which illus-
trate that 5 to 20 years of baseline
data are required to characterize the
complexity of ecological interactions
and systems.

Length of Study: Advantages and
Disadvantages

Not all studies must be "long term"
to provide reliable results. Descrip-
tive studies of essentially static pat-
terns (e.g., morphology, genetic char-
acteristics of species), of processes at
the individual level (e.g., growth, be-
havior), or evolutionary patterns or
systematic relationships do not nec-
essarily require long-term study.
These phenomena occur on time
frames that are either very short or
very long relative to the normal du-
ration of a short-term study (Wiens
1984). The principal disadvantages of
long-term studies are not ecological,
but practical. The need for continued
support of money, time, staff, and
facilities; the problems associated

with the study falling into unproduc-
tive complacency; and environmental
concerns that often require immedi-
ate, even if incomplete, answers.

As pointed out by Wiens (1984),
long-term studies, because of the in-
tense and continued commitment of
time and money, must focus on just a
few specific situations. Long-term
work, therefore, must sacrifice the
breadth possible with a series of
short-term studies, in exchange for
this increased detail and intensity.
This, of course, reduces the potential
for generalizing from such (long-
term) studies. A degree of compro-
mise between these extremes (short-
term vs. long-term studies) is dis-
cussed below.

Alternatives to Long-Term Studies

There are four classes of short-term
studies that can potentially provide
insight into long-term phenomena:
(1) retrospective studies, (2) substitu-
tion of space for time, (3) use of sys-
tems with fast dynamics as ana-
logues for systems with slow dynam-
ics, and (4) modeling (Strayer et al.
1986). They raise the important point
that such short-term approaches can
be integrated into an overall, longer-
term, study, thus "...extending the
temporal and spatial scales of the in-
vestigation and allowing the ecolo-
gist to explore a wider range of eco-
logical phenomena than nnight be
practical in a direct long-term
study."

Retrospective Studies

The record of past conditions can be
used to help reconstruct a long-term
trend. Obvious examples of such ap-
proaches are tree-rings and pollen
deposition. Unfortunately, conclu-
sions regarding past conditions re-
lated to or even causing the pattern
remaining can only be inferred; fur-
ther, only persistent structures re-
main to be analyzed.

Substitution of Space for Time

This is an often-used substitute for a
long-term study. Here sites with dif-
fering characteristics are used in-
stead of following the course of a
single or a few sites for an extended
period. For example, evaluating suc-
cession by simultaneously using sites
of different age (e.g., 1, 5, 15, 30 years
post-harvest). This approach, how-
ever, requires the assumption that all
important environmental processes
are independent of space and time
(i.e., all sites must have the same en-
vironmental characteristics and his-
tory). To provide valid results
through this approach requires, then,
that many sites with very similar his-
tories and characteristics be used. An
obvious problem, of course, is deter-
mination of how "similar" sites must
be. Although results of such studies
may theoretically approach those of a
long-term study, they can only do so
with a large number of replicates.
Further, such substitutions cannot
capture the historical events that
shaped each site, but can only mask
or "swamp" the effect through a
large sample size (which may yield
adequate results for many applica-
tions).

These problems can best be dealt
with in studies combining direct
long-term studies with space-for-
time substitutions. Long-term studies
done in parallel with carefully
matched, short-term "substitutes,"
can factor out the year-to-year vari-
ation that may mask general trends.

Ottier Mettiods

Applying the results of a simple sys-
tem with rapid generation time can
give insight into how a system with a
slower generation might behave: for
example, applying the results of labo-
ratory studies on rodents to evalu-
ations of population dynamics of
larger mammals. Such extensions of
results have obvious drawbacks, but
can be useful in the development of
general theories used to guide

269

longer-term studies.

Mathematical modeling can, of
course, be used to predict the longer-
term behavior of a system. Such
models are often based on guides
provided by various short-term stud-
ies. Obviously, the predictive ability
of models can only be determined
through long-term studies, and/ or a
series of short-term f)erturbations
that experimentally test them; the
latter will fail unless all likely catas-
trophes and conditions can be ade-
quately simulated. Here again, such
modeling can provide valuable in-
sight into the design and conduct of
parallel long-term studies.

Ecological Monitoring

Monitoring of environmental condi-
tions is a closely aligned aspect of
long-term studies. When a manage-
ment agency such as the USDA For-
est Service discusses the need for
monitoring of wildlife population
numbers, they are essentially de-
scribing a long-term study, the goal
of which is to identify trends. Unfor-
tunately, "monitoring" has a low
status in ecology, being widely re-
garded as possessing little originality
and as unproductive of new scientific
knowledge (Strayer et al. 1986).
Monitoring data can provide, how-
ever, essential support for many re-
search projects and publications aris-
ing from long-term studies. In addi-
tion, monitoring programs can lead
to important and unexpected discov-
eries (e.g., first report of acid rain in
North America; Strayer et al. 1986).

Sutcliffe and Shachak (in Strayer et
al. 1986) outlined several elements
that are essential in the conduct of
monitoring programs: (1) the initial
sampling design, variables to be
measured, and methodology must be
carefully chosen; and (2) a scientist
capable of interpreting the data
should be closely involved with all
aspects of the study, allowing modi-
fication of design to take advantage
of the ever-increasing knowledge

about the system under study. A
critical aspect of any monitoring pro-
gram is to eliminate the unproduc-
tive parts of the program to allow for
more fruitful analyses without de-
stroying some part of the long-term
core data (Strayer et al. 1986).

STUDY DESIGN

Introduction

The design of a long-term study
must be sufficiently simple to persist
over a long period of time. Thus, es-
sential measurements must be simple
enough to be repeatable by workers
with varying degrees of experience
(Strayer et al. 1986). There are also
numerous specific aspects of site pro-
tection and management, manage-
ment of data, quality control, chang-
ing methodologies, and the like that
are all critical to a successful study;
these concerns are discussed by
Strayer et al. (1986) and will not be
repeated here.

The design of a long-term study
must also be sufficiently flexible to
accommodate short-term investiga-
tions. Long-term data often suggest
questions that can be investigated
through short-term exp>erimentation
or observations. A benefit of such an
approach is that overall productivity
can be increased; the longer-term ob-
jectives of a study can also be more
easily funded as a result of such
shorter-term efforts. In summary,
studies of varying lengths can usu-
ally complement one another.

I have designed and implemented
a study to evaluate both short- and
long-term responses of vertebrates to
abiotic and biotic conditions in the
Inyo-White mountains (Inyo and
Mono counties) of eastern California.
The design represents a compromise
among the many different methods
necessary to sample different groups
of small vertebrates on the same site.
Below I briefly describe the sampling
design, and provide data on initial
surveys. I present this design as a

possible template for other studies
that seek to determine wildlife-habi-
tat relationships and responses to
environmental changes (i.e., monitor-
ing) over the short- and long-term.

Rationale

The overall objective of this study is
to determine long-term behavioral
and ecological attributes and interre-
lationships of vertebrates in the Inyo-
White mountains of eastern Califor-
nia. Amphibians, reptiles, small
mammals, and birds will be censused
on a series of sites in the pinyon-juni-
per (Pinus monophylla-Juniperus os-
teosperma) plant community on a
year-round basis. Abiotic factors and
food resources will also be sampled.
Reproductive physiology of small
mammals will be addressed.

Numerous hypotheses can be
evaluated depending upon the taxo-
nomic group(s) (e.g., species level,
class level, guild level) chosen for
analysis; for example:

1. H^l: The p>opulation num-
bers of the group are not re-
lated to (a) food resources,

(b) abiotic conditions, and /or

(c) population numbers of
other groups.

2. H^2: The behavior (e.g., for-
aging behavior) of the group
does not vary with fluctua-
tions in (a) food, (b) abiotic
conditions, and /or (c) popu-
lation numbers of other
groups.

3. H^3: Population numbers of
the group during spring are
not related to (a) food, (b)
abiotic conditions, and/ or (c)
number of other animals
during a previous season.

4. H 4: Guild structure cannot

o

be identified on any tempo-
ral basis.

270

4a. H^4: The guild structure
identified does not vary with
variation in (a) food, (b) abi-
otic conditions, (c) popula-
tion numbers of other
groups, and/ or (d) tempo-
rally.

This study is designed to address
these and numerous other null hy-
p>otheses. The data set necessary to
answer any one hypothesis is very
similar to that required to address
another hypothesis. Thus, the num-
ber of hypotheses generated is, in a
sense, independent of the effort ex-
pended to collect the data.

This study will contribute to our
understanding of the ecology of this
system in several major ways. First,
it will provide data on fluctuations in
population numbers of vertebrates,
thus serving a monitoring role (espe-
cially important to the USDA Forest
Service). Second, it will provide data
which will allow development of
multi-species population models (by
myself and other workers), allow de-
velopment of habitat-relationships
models that incorporate both short-
and long-term responses to biotic
and abiotic factors, and allow devel-
opment and subsequent testing of
models of multi-species interrelation-
ships at various taxonomic levels.
Third, and possibly the most impor-
tant aspect of the study, it will result
in the accumulation of vast amounts
of ecological information on the ver*
tebrate (and invertebrate) commu-
nity. To date, only brief and sporadic
surveys have been conducted in the
Inyo- White mountains. Finally, it is
my goal to use preliminary results to
generate specific hypotheses that can
be tested by my, or others', students.
For example, if initial data indicate
rejection of the null hypothesis of no
relationship between a certain small
mammal and their prey base, then a
student could select additional sites
where food supplementation and /or
removal experiments could be con-
ducted. Additional, study-specific
funding will be sought for such stud-
ies. This study will thus generate

short-term results under the general
framework of its long-term design
and goals.

The Inyo- White mountains were
chosen as the study location for sev-
eral reasons. First, my intent was to
select a type of habitat that offered
structural, especially vertical, diver-
sity intermediate between that of a
grass- or shrubland and that of a ma-
ture, hardwood or coniferous forest.
With a canopy rarely exceeding 10 m,
I will be able to sample arthropod
populations from the upper canopy.
This is not conveniently possible in
mature conifer forest, where the can-
opy extends to 20-30 m or more in
height. Second, I desired an area that
offered only several dominant tree
and shrub species: this allows inten-
sive sampling of all major species,
while allowing some diversity of
plant species beyond that evident in
more monotypic habitats.

The pinyon-juniper woodland was
chosen because of its extensive cov-
erage throughout the intermountain
west. Further, the pinyon-juniper
woodland undergoes few significant
changes in plant species frequency
and density relative to earlier succes-
sional communities. Austin (1987),
for example, showed virtually no
change in a pinyon-juniper commu-
nity in Utah between 1974-84. In con-
trast, seeds and berries undergo of-
ten marked, interyear changes in pro-
duction. Thus, barring some cata-
strophic change, the gross composi-
tion of the plant community used in
this study should remain relatively
stable, helping to control for at least
some of the variance likely to be en-
countered in animal communities.

A definition of what I mean by
"long-term" in this study is not yet
possible, but I have committed my-
self to this study for an indefinite pe-
riod; 15-20 years seems a minimum.
The study is designed to be con-
ducted, at a minimum, by myself and
one assistant. Additional personnel,
primarily undergraduate volunteers
during summer, will also be avail-
able. Thus, the ability to adequately

conduct the study over the long-term
is considered in the design, and will
be possible given my focus (concen-
tration) on this study. The initial de-
sign can accommodate expansion in
size both through the enlargement of
each site (using the original area as a
standard core), and /or the addition
of additional sites (e.g., to sample
from a wider range of ecological con-
ditions). Various ancillary studies
will add to my understanding of this
system, although my primary goal is
to examine the interrelationships
among vertebrates and their environ-
ment.

Sampling intensity will not be in-
creased beyond that discussed herein
(see Methods) to avoid substantial
impact (e.g., trampling) by observers
on the study sites. Thus, an increase
in effort (given adequate time and
funding) will be directed towards an
increase in site size, number of sites,
and/or towards ancillary studies, the
decision based on preliminary data.

I will be intensively involved with
all aspects of this study, including
establishment of the sites (already
accomplished) and collection of data
throughout the duration of the study.
It is essential, in any long-term effort,
that methodology be standardized,
and a high level of quality control be
maintained. My involvement will
serve as the standard upon which
new assistants will be trained. Any
changes in methods, whether this in-
volves modification of sampling in-
tensity or a change in trap type, will
be fully documented. If any proce-
dures must be changed, the old and
new methods will be run simultane-
ously to allow for intercalibration of
methods (as described by Strayer et
al. 1986). All field notes and data will
be duplicated or triplicated and
stored in several locations for safety.

Design Considerations

The study sites — their number, size,
and location — chosen for this study
were selected to restrict samples to

271

convenient and modest-sized popu-
lations. They will be low in cost to
sample and are located in practical
locations for year-round access. With
the few (3) sites chosen, it would be
foolhardy to attempt representation
with probability sampling of entire
populations. The study is designed to
spread effort across important vari-
ables to obtain some measure of con-
formation of results. I follow the
philosophical view of Popper (1959),
as summarized by Kish (1987): "The
choice of the sites should strain to
increase the possibilities for falsifica-
tion." Similar and consistent results
from the replications yield stronger
confirmation than a single site
would. But if the results are discor-
dant, the replications are too few to
yield dependable inference; then fur-
ther research is indicated. Discordant
results yield a healthy skepticism
that naive "success" from a single
site would obscure (taken from Kish
1987). As discussed earlier, the study
is designed to allow an increase in
the number of sites (or their size,
etc.) should early results so indicate.

The general locations of the sites
were not chosen at random. A gen-
eral area was identified based on (1)
ease of access during winter (e.g.,
within 0.5-1.0 km of a maintained,
although usually dirt, road), but also
(2) isolated from access by off-road
vehicles. Using these general guides,
specific sites were chosen to repre-
sent a sampling of slope, aspect, and
longitudinal location in the Inyo-
White mountains. My intent was to
increase the likelihood of "falsifica-
tion," which is better served with
tests obtained in contrasting condi-
tions, as opposed to selecting more
or less "average" sites (see Kish 1987
for a development of this strategy).
The extremes of a relationship are
more informative than either random
or modal or centralized selection.
Three sites were considered a mini-
mum, because two sites might indi-
cate a false, linear relationship in cer-
tain factors. Survey data will be col-
lected on other areas throughout the

Inyo-White mountains. Such data
will provide useful information re-
garding the overall distribution and
habitat associations of vertebrates
throughout the ranges. Further, these
sites will serve as 'l3ack-ups" should
a catastrophic event occur on one of
the three main sites. Because this
study is largely exploratory, such a
strategy was warranted (with the op-
tion of later expansion).

METHODS

Terminology

I have attempted to standardize the
terms used to describe the study ar-
eas described beyond; they are:

"Permanent site": a 400 x 400 m
(16 ha) area that forms the long-term
"study sites" used.

"Point": a trap location within a
site.

"Ancillary site": additional areas
(of various shapes and sizes)
sampled at an intensity less than on
the permanent sites; established to
increase scope of sampling effort.

"Sampling f>eriod": a 5-7-day pe-
riod in which a permanent site is
sampled.

"Transect": the parallel, 400-m
long lines ("transect") forming the
study sites.

"Core trapping area": the central,
200 X 200 m area, location within a
study site where small-mammal and
pitfall traps are placed.

Sampling Schedule

Study sites (described below) were
established during fall 1987. All
methods outlined herein were evalu-
ated during fall 1987, winter 1987-88,
and spring 1988. Each permanent site
will be visited for a 5-7-day period
on two occasions per season. Seasons
are defined as: spring (1 Mar.-31
May); summer (1 June-31 Aug.); fall
(1 Sept.-30 Nov.); and winter (1 Dec-
28 Feb.). The exact length of visit will

be based on trapping results. Initial
order of visit to sites will be random-
ized; this order then followed on the
subsequent visit in that season.

Remaining time available during a
season will be spent sampling the an-
cillary sites. These ancillary sites will
increase my knowledge about the
distribution and relative abundance
of vertebrates and invertebrates in
the Inyo- White mountains. Not all
data from ancillary sites will be di-
rectly comparable — to permanent or
other ancillary sites — because of the
lower sampling intensity. Neverthe-
less, they will supply information
important to the long-term success of
the study.

Permanent (n = 3) Study Sites

Each will be established as a 400 x
400 m (16 ha) site. A 16-ha site was
chosen because: (1) an observer can
travel this distance, even over rough
terrain, in a short period of time; (2)
the utilized area of most small verte-
brates can be sampled within a 16-ha
area; and (3) this area allowed estab-
lishment of sites protected from
roadways, trails, and other human
activity (e.g., fit between cliffs and
gullies that form barriers to illegal
vehicle access). Each site will be
sampled repeatedly within a season
to provide measures on within-site
variability. The effective n per per-
manent site is, of course, one (for
comparison among permanent sites,
n = 3). Each site will have permanent
grid points marked at 25- or 50-m
intervals (using rubber cattle ear
tags).

SAMPLING

Amptiibian, Reptile, and Snnall-
Mammal Trapping

One-hundred-one Sherman live traps
and 41 pit-falls (two 3.2 1 (3 lb) rin
cans taped together) were estab-
lished on each site. Live traps were

272

12.5 m apart in the center 100 x 100
m section of a site, and at 25-m spac-
ings in the remaining trapping area.
The closer trap spacing helps deter-
mine actual population density,
whereas the wider spacing in the sur-
rounding area provides information
on animal movements.

Live traps are baited with seed
mixtures and checked each morning
and late afternoon. Certain rodents
(e.g., Peromyscus) are active through-
out the year. During winter, there-
fore, traps are provided with insulat-
ing material (e.g., wool). All captures
are toe-clipped. Trapping continues
until new captures are minimal (usu-
ally 5 days). Pitfalls are not baited,
but captures are marked and re-
leased. Traps are run "dry": holes
were drilled in each trap, rocks
placed in the bottom of each hole to
provide drainage, and a wooden lid
placed over the trap to reduce expo-
sure.

Bird Activity

The spot-map method (e.g., see
Ralph and Scott 1981) is used to de-
termine bird abundance and territory
(during breeding) size. Following a
census, the observer slowly walks
though the entire site and records
foraging birds as encountered. Data
are recorded on activity and sub-
strate used. (The specific methods
used for birds will not be detailed in
this paper.)

Vegetation Sampling
General site

Trees and shrubs will be sampled
once per year, and grass and herba-
ceous cover will be sampled once per
season. Changes in plant phenology
will be recorded as they occur. Vege-
tation will be sampled using circular
plots and line intercepts centered at
each of the 81, 50-m-transect inter-
cepts. Pinyon and juniper will be

counted and measured (e.g., dbh,
height, vigor, canopy cover) within
20-m-radius plots. Shrubs, grass, and
herbaceous cover will be measured
along 40-m-long line intercepts bi-
secting each circular plot (and run-
ning parallel to the main transect
line).

Trap Locations

Vegetation and soil characteristics
will be measured at each trap site.
The nearest tree in each quarter from
the trap will be measured. Two 5-m-
p)erpendicular transects will be
placed over each trap; shrub and her-
baceous cover will be measured
along each transect. Soil moisture,
compactability, texture, and pH will
be measured on each arm (2.5 m) of
the trap-site transects: at 0.5, 1.0, 1.5,
and 2.0 m (one of these distances per
arm, randomized, for four measure-
ments per trap). These samples will
be gathered once during each season.

Abiotic Factors

A weather station will be established
near the center of each study site.
Temperature, humidity, and rain fall
will be automatically recorded
throughout the year. Snowfall will be
measured by visiting each site fol-
lowing snowstorms.

Ottier Sampling

Data on arthropod abundance
(branch sampling and pan traps) and
cone-seed production (of pinyon, ju-
niper and major shrub species) will
also be collected (but not detailed in
this paper).

RESULTS

Only one site has been sampled with
adequate intensity for presentation of
data at this time. Sampling occurred

during fall (8 days during two trap-
ping sessions), winter (7 days during
two trapping sessions), and spring
(12 days during three trapping ses-
sions). Pitfalls were used only during
the first trapping session in the fall
and the last session in the spring (be-
cause of snow and little or no lizard
activity). Chipmunks and Great Ba-
sin pocket mice were not active from
October-November until early
March.

The sagebrush and western fence
lizards were the most frequently cap-
tured animals in pitfalls (table 1). A
single deer mouse (immature) was
also captured in a pitfall trap. Inten-
sity of pitfall trapping has been in-
adequate to date to make conclusions
on their effectiveness.

Seven small mammal species and
a skink were captured in the live
traps (table 1). The pinyon mouse
was the most abundant species cap-
tured during fall. Relatively few
pinyon mice were captured during
winter, however (a 77.9% decline be-
tween fall and winter). The decline of
pinyon mice continued into spring,
with abundance dropping 64% be-
tween winter and spring. (Dnly a few
deer and pinyon mice were captured
during winter.

The highest overall abundance of
small mammals was found during
spring (table 1). The two species of
chipmunks were the most abundant
animals captured. The Great Basin
pocket mouse and the deer mouse
were also captured frequently during
spring.

DISCUSSION

Live-trapping data indicate the im-
portance of repeated sampling over
time: pinyon mice apparently suf-
fered substantial winter mortality.
Thus, trapping in only fall or spring
would have falsely indicated a rela-
tively high or low population size,
respectively. Although this study can
hardly be considered "long-term,"
initial results do highlight the need

273

for repeated sampling even over the
short term. Only continued sampling
will elicit the frequency and reasons
for such a decline. My initial trap-
ping configuration contained a dense
trap placement (12.5 m trap inter-
vals) in the middle of the grid rela-
tive to the outer traps (25-m spacing).
My intent was to use the outer traps
to determine movement of animals in
and out of the smaller 100 x 100 m
area. Cursory examination of trap-
ping results (unpubl. data) indicate,
however, that even the total 200 x
200 m grid is not sufficiently large to
quantify movements (i.e., animals
moving >200 m). Therefore, I suggest
the following modifications in trap
placement: 10 x 10 trapping grid with
15-m spacing. This placement should
adequately sample the animals pres-
ent. To detect movement (e.g., dis-
persal), trap lines can be established
periodically that run perpendicular
from the edge of the trapping grid.

ACKNOWLEDGMENTS

I thank David Strayer and Robert C.
Szaro for reviewing earlier drafts,
and Lorraine M. Merkle for prepar-
ing the text. C. John Ralph, Redwood
Sciences Laboratory, Areata, Califor-
nia, Jared Verner, Forestry Sciences
Laboratory, Fresno, California (both
Pacific Southwest Forest and Range
Experiment Station, USDA Forest
Service), and Reginald H. Barrett,
Dept. Forestry and Resource Man-
agement, University of California,
Berkeley, provided equipment. The
director ((Tlarence Hall), superinten-
dent (David Trydahl), and staff of
the White Mountain Research Sta-
tion, University of California, Los
Angeles, are thanked for supplying
logistical support. John A. Keane,
Martin L. Morton, and Kimberly A.
With assisted with field work. Kathy
Noland, White Mountain Ranger Sta-
tion, USDA Forest Service, is
thanked for arranging access to
study sites. Harold Klieforth, Atmos-
pheric Sciences Center, Desert Re-

search Institute, University of Ne-
vada, Reno, helped establish weather
stations. This study was funded, in
part, by the Committee on Research
and the Department of Forestry and
Resource Management, University of
California, Berkeley.

LPTERATURE CITED

Austin, Dennis D. 1987. Plant com-
munity changes within a mature
pinyon-juniper woodland. Great
Basin Naturalist 47:96-99.

Barber, Richard T., and Francisco P.
Chavez. 1983. Biological conse-
quences of El Nino. Science
222:1203-1210.

Bock, Carl E., and James F. Lynch.
1970. Breeding bird populations of
burned and unburned conifer for-
est in the Sierra Nevada. Condor
72:182-189.

Halvorson, Curtis H. 1984. Long-
term modeling of small verte-
brates: a review with suggestions,
p. 11-25 In Janet L. Johnson, Jerry
F. Franklin, and Richard G. Krebill
(coord.). Research natural areas:
baseline monitoring and manage-
ment. USDA Forest Service Gen-
eral Technical Rep. INT-173, 84 p.
Intermountain Forest and Range
Experiment Station, Ogden, Utah.

Keith, Lloyd B. 1983. Role of food in
hare populations cycles. Oikos
40:385-395.

Table 1 .—Abundance (no./lOO trap-nights) of reptiles and small mammals,
Inyo-White mountains, California, during fall (Sept.-Nov.) 1987, winter (Dec.
Feb.) 1987-88, and spring (Mar.-Moy) 1988.

Fall=

Winter''

Spring'

Species

Captures Abund. Captures Abund. C^tures AlHjnd.

Western fence lizard

(Sceloporous ocddentalis)
Sagebrush lizard 1^

(S. gradosus)
Gilbert skink

(Eumeces gilberfi)
Golden-mantled ground squirrel

(Spermophifus lateralis)
Least chipmunk 3

(Eufamias minimus)
Panamint chipmunk

(E. panaminfinus)
Great Basin pocket mouse

(Ferognafhus parvus)
Deer mouse 9

(Peromyscus maniculafus)
Pinyon mouse 56

(P. true!)
Desert woodrat 1

(Neofoma lepida)

Total 69

1.3

0.4

1.2
7.7
0.1
9.5

12
12

24

1.7
1.7

3.4

2^
3^
1*

I

39
41
29
26
7

1.4
2.0
0.1
0.1
3.2
3,4
2.4
2.2
0.6

143 11.8

"Trap-nlghfs: 75 for pitfalls, 728 for small-mammal traps.
^Trap-nigt)ts: 0 for pitfalls, 707 for small-mammal traps.
'^Trap-nightts: 147 for pitfalls. 1211 for small-mammal traps.
'^Captured in pitfall.
^Captured in small-mammal trap.

274

Kish, Leslie. 1987. Statistical design
for research. Wiley and Sons, N.Y.

Likens, Gene E. 1983. A priority for
ecological research. Bulletin Eco-
logical Society of America 64:234-
243.

Meents, Julie K., Jake Rice, Bertrin W.
Anderson, and Robert D. Ohmart.
1983. Nonlinear relationships be-
tween birds and vegetation. Ecol-
ogy 64:1022-1027.

Ralph, C. John, and J. Michael Scott
(editors). 1981. Estimating num-
bers of terrestrial birds. Studies in
Avian Biol. No. 6.

Raphael, Martin G., Michael L. Mor-
rison, and Michael P. Yoder-Wil-
liams. 1987. Breeding bird popula-
tions during twenty-five years of
postfire succession in the Sierra
Nevada. Condor 89:614-626.

Schreiber, Ralph W., and Elizabeth
Anne Schreiber. 1984. Central Pa-
cific seabirds and the El Nino
southern oscillation: 1982-1983
perspectives. Science 225-713-716.

Strayer, David, Jeff S. Glitzenstein,
Clive G. Jones, Jerzy Kolasa, Gene
E. Likens, Mark J. McDonnell,
Geoffrey G. Parker, and Steward
T. A. Pickett. 1986. Long-term eco-
logical studies: an illustrated ac-
count of their design, operation,
and importance to ecology. Insti-
tute of Ecosystem Studies, Occa-
sional Paper No. 2. 38 p.

Wiens, John A. 1984. The place of
long-term studies in ornithology.
Auk 101:202-203.

275

An Ecological Problem-
Solving Process for Managing
Special-Interest Species^

Henry L. Short^ and Samuel C. Williamson^

Abstract.— We present a structured problem-
solving process that con help resolve v/ildlife
management issues. Management goals for wildlife
species are expressed in terms of populations to be
attained and maintained. Habitat quantity and
quality necessary to achieve those population goals
can then be determined. Proposed land-use
changes are evaluated in terms of how they will
contribute toward recovery or extinction of the
species of interest.

Land-use problems associated with
the need to protect wildlife habitat
and the desire to develop resources
can sometimes be resolved using an
ecological problem-solving process.
The process requires development of
a management goal for individual
wildlife species, determination of the
quantity of habitat required to
achieve that management goal, and
an appraisal of how development
scenarios will affect the management
goal.

We describe how the process
might work using available data
about the endangered Mount Gra-
ham red squirrel (Tamiasciurus
hudsonicus grahamensis). The exercise
is relevant because the squirrel exists
entirely as a disjunct population in
the high elevation coniferous forest
community of the Pinaleno Moun-
tains of southeastern Arizona, and a
new astrophysics observatory has
been proposed within important
squirrel habitat. Our process was not
applied in the development of the
Environmental Impact Statement
(EIS) prepared for the red squirrel
and its habitat nor in negotiations for
the future management of the squir-
rel. An extensive and current infor-

' Paper presented of symposium, Mon-
agement of Amphibians. Reptiles, and
small Mammals in Northi America. Ragstaff.
Ariz. July 19-21. 1988.

'Ecologist. U.S. Fisti and Wildlife Service.
National Ecology Researct^ Center. 2627
Redwing Road. Fort Collins. Colorado.
60526.

mation base (Spicer et al. 1985; U.S.
Forest Service 1987, 1988) recently
has been developed for the Mount
Graham red squirrel in order to de-
velop the EIS for the proposed astro-
physics observatory. We applied
these data to a cumulative impacts
assessment process being developed
by the U.S. Fish and Wildlife Service.
We assume that species-habitat man-
agement goals can be developed and
that these goal statements can drive
habitat management plans and ac-
tivities. We have not analyzed the
merits of any development scenarios
proposed for the astrophysics obser-
vatory.

The Pinaleno Mountains are an
isolated range that supp)orts one of
the southernmost spruce-fir forests in
North America (Spicer et al. 1985).
The Mount Graham red squirrel is
endemic to the small patches of co-
niferous forests that occur at the
highest elevations of the mountains.
The squirrel has been affected by a
variety of human activities and natu-
ral events that have altered its habi-
tat. Disturbances included comple-
tion of a road to the mountain top in
1933, introduction of the tassel-eared
squirrel (Sciurus aberti) in 1941 to
1943, extensive logging activities in
subalpine coniferous forests from
1946 to 1973, a major fire in 1956, and
extensive windthrows in the 1960's
(Spicer et al. 1985). The squirrel was
first collected from the Pinaleno
Mountains in 1894 and was consid-
ered "connmon" in the spruce-fir

zone above 2,590 m in 1914. Since the
early 1950's it has been considered
"uncommon" throughout the conifer-
ous tree zone of this mountain range
(Spicer et al. 1985).

THE PROCESS

The problem-solving process used in
our analysis contains three principal
steps (fig. 1). Problem description,
the first step, defines the ecological
problem and identifies the spedes,
study area, and time frame of con-
cern.

Problem analysis, the second step,
develops biological information nec-
essary to achieve a solution. An ini-
tial effort is to describe a manage-
ment goal for the species of concern
in terms of a specific population level
to be achieved and maintained. This
numerical target is not a vague state-
ment to "maintain" or "enhance" be-
cause such terms cannot be used to
measure the results of management
actions. The management goal
should be collaboratively develof)ed
so that all interested parties reach a
consensus on the desirability for per-
petuating the spedes and on a popu-
lation level to be achieved by man-
agement. It is understood that mutu-
ally agreed upon goals represent
compromise and that compromises
are rarely satisfactory to all con-
cerned parties.

It is then necessary to determine
the quality and quantity of habitat

276

required to achieve the management
goal. This requires building a model
describing habitat requirements for
the species. An understanding of
how human activities and natural
events impact habitat quality and
quantity is also desirable because the
management of these restricting ac-
tions may help achieve the manage-
ment goal for the species. The identi-
fication of causes contributing to

habitat deficiencies can be made by
interviewing persons familiar with
the sf>ecies and the particular habitat
conditions within the study area.

The third step in the process, solv-
ing the problem (fig. 1), is accom-
plished after: (1) the amount and
quality of habitat necessary to fulfill
the management goal has been deter-
mined, (2) the quantity of suitable
habitat presently available has been

I. Describe the Problon

II. Analyze the Problem

1. Determine the Management Goal for the Species.

2. Describe Inportant Habitat Cticxiitions for the Species.

3. Determine how ftmn Activities Affect Habitat Conditions
Important to the Species.

III. Solve the Problem

1. Determine Acceptable Strategies for Managing Habitats Required
by the Species.

Figure 1 .—Steps of ttie problem-solving process.

Abounds in
mountains of
lh« Graham
range

Common

Infir-spfuce

fcrests

Not

abundant

Uncommon
to rare

Possibly
extirpated

Begun
recovery but
probably has
not acWevod
former
levels

1889

1914

1929

1951-2 1963-7

Present

Figure 2.— Possible trends in abundance of the Mount Graham red squirrel. (Population de-
scriptions are those of Spicer et al. 1 985.)

documented, and (3) the quantity of
suitable habitat that would be avail-
able under different land-use options
has been projected.

Describe the Problem

The Mount Graham red squirrel has
probably declined during this cen-
tury (fig. 2) in part because of the
piecemeal degradation of isolated
forest habitat. The variety of human
activities and natural events causing
this decline might soon be aug-
mented by the development of the
astrophysics observatory on the Pi-
naleno Mountains. Can this and re-
lated developments occur in a man-
ner that does not further jeopardize
the existence of the endangered red
squirrel during the foreseeable fu-
ture?

Analyze the Problem

Determine the Management Goal
for the Species

The management goal is described in
terms of a population to be attained
and maintained. Ideally, population
goals should be based on quantita-
tive historical levels of abundance.
Population goals are more difficult to
establish if historical information
about population levels are fragmen-
tary and descriptive, as for the
Mount Graham red squirrel. In such
cases, criteria for establishing desired
population levels should consider:
(1) estimates of present populations
and trends, (2) threshold values nec-
essary to ensure the survival of the
species, and (3) estimates of the po-
tential p>opulation level that could be
attained if management of an area
was accomplished solely to benefit
the species.

Estimates of population trends for
the Mount Graham red squirrel are
largely qualitative (fig. 2). The results
of field work suggest that the au-
tumn 1987 population of red squir-

277

rels on the Pinaleno Mountains might
be 246 (206-286), (U.S. Forest Service
1988:37). Computer simulations of
population dynamics of the red
squirrel (U.S. Forest Service 1988:74)
are only minimally helpful because
data such as natality and mortality
for the Mount Graham red squirrel
are unknown. The computer simula-
tions suggest probability levels for
extinction under different combina-
tions of mortality and reproduction.
The predicted carrying capacity for
the squirrel under current habitat
conditions has been estimated at 502
squirrels. The potential future carry-
ing capacity, based on the quantity
and present age structure of mixed
conifer and spruce-fir stands, is 725
squirrels (U.S. Forest Service 1988:72-
73). Thus, the current population of
red squirrels might be somewhat
higher than that in the early 1960's
when the species was reported as
possibly extirpated (fig. 2), but less
than one-half the present carrying
capacity for the species. The collabo-
ratively developed management goal
might state, for example, that the
management goal for the species is to
develop and perpetuate a red squir-
rel population equal to the present
carrying capacity of the habitat for
the squirrel which is estimated at 502
squirrels (U.S. Forest Service
1988:73).

Describe Important Habitat
Conditions for \he Species

A species-habitat model for the red
squirrel can be based on the squir-
rel's dependency on seed cones and
trees that produce those cones. Coni-
fer seeds are the primary food of the
red squirrel, which cuts cones in
summer and caches then in middens
in dense needle litter at stumps,
downed timber, and on the base of
snags or live trees in forests with
dense overstory canopies (Spicer et
al. 1985).

We constructed a species-habitat
model for Mount Graham red squir-

rels using data given in the U.S. For-
est Service (1987) report. The struc-
tural stage, tree species, and canopy
density that compose red squirrel
habitats are classified as excellent,
good, fair, poor, very poor, and no
value (fig. 3). These data were devel-
oped by U.S. Department of Agricul- -
ture Forest Service personnel and
others familiar with the habitat re-
quirements of the squirrel and were
based on vegetation type and struc-
tural stage, the number of snags and
downed logs per hectare, aspect, and
slope (U.S. Forest Service 1987:44).
Midden complexes are a focal point
of territories and the number of ac-
tive middens is supposedly associ-
ated with the number of red squirrels
in a stand (Spicer et al. 1985). The
data for middens per hectare have
been adjusted so that a score of 1.0 is
listed for excellent habitats, 0.0 for no
value habitats, and intermediate val-
ues are listed for habitats of interme-
diate quality (fig. 3).

The species-habitat model de-
scribes conditions in habitats of dif-
ferent quality. A simple word model
was then developed to describe a
unit of good or excellent habitat for a
red squirrel (fig. 4). The model devel-
oped from information in figure 3
and U.S. Department of Agriculture
(1987:33-37) defines suitable habitat
for a red squirrel as a 1-ha forested
block that: (1) is contiguous to other
similar forested blocks, (2) provides a
dense overstory canopy of spruce-fir
or mixed conifers, and (3) contains
about 15 "good" seed-bearing trees
per hectare.

Such species-habitat models are
general and approximate. Still, they
provide an estimate of what com-
prises a unit of habitat area and con-
dition that might be required by a
squirrel. If a management goal is to
provide habitat for X red squirrels
then that goal can possibly be
achieved by providing X units of
good to excellent habitat (fig. 4). The
need to provide this quantity of a
specific habitat condition should
drive management plans for the sub-

Trco spccloi,
ftlnirlurnl
fct>(|t.f and
canopy dcosi ty

All titlicr vcgetatiun
coMLiin i I. i es

Mixed conircr and tuicn,
polot,

canopy covar

Spriice-'ir, poles,
'lUl cMnnpy tovyr

Mixiyt conirer and atpen,
poles,

71X canopy ccvar

Sfinico*^ir ard aspon,
iMture,
10S canopy cover
Nixad conlfar. anturo,

7<^ canufjy cover

Spn*C0*rir, acturtt,

41S canopy cover
Spnii:K*rlr uimJ bIwU
coiifcr, old grovUi,

^OX citnupy cover
NIxuJ cuKlrer, BHUiru,

ViX canopy cover

Spnx:e-rir acd ailxud
conircr, old qrowih,
b<it canopy cov«r

TT 1 1 1

Figure 3.— Habitat quality for ttie Mount
Grahiam red squirrel can be described in
terms of tree species, structured stage, and
canopy density wittiin forest tKibitats. Habi-
tat quality is measured in ternrts of middens
per tiectare for different habitats scaled on
ttie basis of 0.0 for no value hiabitats and 1 .0
for excellent tiabitats. Data from U.S. Forest
Service (1988:112).

A habitat block suitable for a Mount Grahan red
squirrel Is at least 1 ha In area aid is
contlgjous to other blocks of suitable habitat
(fragTBitation or isolaftioa totally raroves a
habitat block fran suitability oonsideraticn) .

Yes
i

2. The tree cxnposltitn of the habitat block Is (fran
Figure 3):

a. spcuce-fir, neture, ^40% canopy coror

b. spcuce-fir, old growth

c. mixed ocnlfer, netuie, ^70% caociy cover

d. mixed cxnlfer, old growth

yes

3. The habitat block contains at least 15 "good-
bearing spnioe-fir or mixed conifer seed trees (a
saTB*at arbitrary niiier).

Yes
i

Ihe habitat block provides good to exoeUeot
habitat for one Mount Graham red squirrel.

Figure 4.— A word rTX>del describing good
to excellent tiabitat condition for a red
squirrel.

278

alpine coniferous forests of the Pi-
naleno Mountains.

Determine How HunfKin Activities
Affect Habitat Conditions
Important to ttie Species

Several human activities and natural
events may adversely affect habitats
of the Mount Graham red squirrel
and reduce the opportunity to
achieve the management goal for the
species. A listing of possible impacts
on the Mount Graham red squirrel
and the probable resulting habitat
changes is in figure 5.

The cells in a cause-effect matrix
(table 1) list estimates of the direction
and relative importance of each fac-
tor affecting a habitat criterion. The
cells within the cause-effect matrix
can be completed after synthesizing
information from the literature, from

best professional judgments elicited
from selected personnel or preferably
from analyzing results of appropriate
research. Information within the
cause-effect matrix can indicate the
relative importance of different hu-
man activities on squirrel habitats
and identify actions to be favored or
avoided to help achieve the manage-
ment goal. For example, habitat frag-
mentation, clearcutting, selective har-
vest, and forest management favor-
ing early vegetation successional
stages are important negative factors
to red squirrel habitats whereas man-
agement favoring dense, mature or
old-growth stands of mixed conifer
and spruce-fir forests are important
positive actions, favorable to red
squirrels. Causes of negative and
positive impacts to species or habi-
tats of concern are factors that
should be considered when formulat-
ing and evaluating plans for modify-

cwsts

Habitat fra^MntAtlm (stand size —
reduction, road ccnstiuctlm, otc.)

Sal«ctlv« harvwt (t.MnrUng) of trm (innj

thtOH 1> analogous to s*Lactlv« harvest)

CXasrcuttUig 6oc dw^elopisnt (foc««t firee

are aralcgcxas to clearcutting)

Mnagannt favAsring early SLXxessianal
wtegta

HBDB^erant favociog dense, nature, or -
old grcHtti spcuoe-flr

Moa^anwit favoring dacse, mature, or -
old giouth of mlxsd cxxiifeccus gpeclw

RBwal of anaga and Moody deiarLi

ConpetiUm £ron IntitxlDed tassel-eeivd

VXal ctaogta In tim
gjantity and quality oC
suitable habitat la the
Plnalwp Mountains tvm
adversely affected the Mount
Gattm red agjirr*! populaticn.

-> -

• aiitablllty ce habitat toe squirrel
territociee (large areas with aense
tree canqpy ccwer) is nxliflad

- rjcgi (a food source} pzodLjcticn
in the naedle litt«- is redifled

' Storage effectlvaoBss of middens
is changed

' Maindanca of preferred cones Is
ncdified

Abundance of tot^l cans is

■ QiiBivtity of cavltiee available
as nest sltas Is ciianged

- Quiatity of atrjctural niatArlAls
available for rurways and niddea
sites is changed

Figure 5.— A cause/effect model identifying causes thaf affect the quantity and quaiity of
habitat suitable for the Mount Grahann red squirrel.

Quantity of

suitablfl

habH^

Currtnl habKat corxJ'rtlon

(1a) Habitat condition exceods that
n«cessary for management goal

(2) Required habitat condition for
management goal

(1b3) Future positive impacts
(1b2) No additional future Impacts
(1b1) Future r^egative impacts

Time

Figure 6.— Habitat conditions under a variety of nnanagennent strategies.

279

ing habitats important to selected
wildlife species.

Solve the Problem

Determine Acceptable Strategies
for Managing Habitats Required
by ttie Species

A way to evaluate the diversity of
different land-use scenarios is listed
in figure 6. Threshold values describ-
ing the quantity of suitable habitat
necessary for achieving the manage-
ment goal for the red squirrel can be
represented as habitat condition 2 in
figure 6. If the quantity of suitable
habitat presently available had ex-
ceeded this threshold value (condi-
tion la) then changes to the quantity
of available habitat could be toler-
ated and that fact could be consid-
ered in making a decision about a
potential land use.

The present quantity of good to
excellent habitat for the red squirrel
in the Pinaleno Mountains, however,
is probably more closely approxi-
mated by condition lb in figure 6. A
variety of conditions like those item-
ized in table 1 have reduced habitat
quality and quantity resulting in a
diminished squirrel population with
an endangered species listing. A
land-use plan that continued impacts
(like those listed in table 1) would
further reduce the area and quality of
contiguous blocks of forest habitat
important to the squirrel. Any fur-
ther fragmentation or degradation of
habitat would be exjjected to further
diminish the population (Ibl in fig,
6) and perhaps threaten extinction of
the subspecies. A land-use plan that
neither allowed further degradation
of habitat nor actively improved
habitat conditions for the squirrel
might result in maintaining present
population levels (lb2 in fig. 6). The
most desirable land-use scenarios are
those likely to produce trend lines
such as lb3 (fig. 6). These land-use
plans would minimize fragmentation
of habitats and would actively man-

age habitats to develop large contigu-
ous blocks of old-growth mixed coni-
fers and spruce-fir on the Pinaleno
Mountains to help attain the desired
population level of red squirrels.

CONCLUSIONS

We emphasize that potential land-
use change can be evaluated in a ra-

tional manner if management goals
for wildlife resources have been pre-
viously established and agreed upon.
The merit of this approach is that
planning becomes an active rather
than a reactive exercise. Too often we
evaluate proposed land-use changes
in terms of how they might affect
present habitats and present popula-
tions without considering how pres-
ent conditions compare to desired

populations and necessary habitats.
Without establishing a management
goal and determining the habitat
conditions necessary to achieve that
goal, we could accept the wrong
baseline for developing our manage-
ment strategy (perhaps something
analogous to line lb2 in fig. 6). If this
occurs, we might have little success
in maintaining viable f)opulations
because we frequently strive only to

Table 1 .—A cause-effect matrix that lists the relative importance of causal agents (causes listed in fig. 4) that change
the quantity arKi quality of habitat features (effects listed in fig. 4) for the Mount Graham red squirrel. A {+) value indi-
cates a positive impact and a (-) value indicates a negative impact. Numerical values indicate the magnitude of an
impact: (0) =■ negligible; (1) = minor; (2) = important; and (3) = very important.

- 0

« «)

k

M

I- c

0

c

c

0

c

u

- (.

0

*t

u

L

err

TJ

« *i

u

V

(A

i)

3

3

0)

a

«-)

E

c

A)

0

0

0

M

w n

L

0

0

« w

-0

a.

0

0

c

> 01

0

*» 0

V

c

m **

L

C i-

«^

19

«

0

0

ti ii

0 L

ca

1.

0

«

M l/t

0

C 0

1.

*i

U it

c

«

« C

« 9(-

Z *i
0-

0

a

- 0

A>

£ L 0

>

tt

*»

u

« >

V) -

b

0

«

(9

ii

a

Ai —

a

OwO

V 0

0

n e

E

0-

9

V.

e

E

>»« N

0X3

c

0

0

>

*«»

<«- 0

ft

«

a

- c

4) >>

0

0

0

> «.

- u c

(B C

u

u

«)

g (B

— CB

— 0 (8

0

c

c

0

3 M

n*i 0

0)

K

a

«^>>

n E

«e —

-r

«

T3

0 «

o>*»

L

c

c

3 >

3 M

c

z c

0

3

>

«- C

!.

SOL

(/) 4J «J

3

■A. —

W

<

<

(A (.

Habitat fragmentation (stand size -3
reduction, road construction, etc.)

Selective harvest (thinning) of trees -2
(windthrow is analogous to selective
harvest)

Clearcutting for development (forest -3
fires are analogous to clearcutting)

Management favoring early -2
succession stages

Management favoring dense, mature, +3
or old growth spruce-fir

Management favoring dense, mature, +3
or old growth of mixed coniferous species

Removal of snags and woody debris 0

Presence of tassel-eared squirrels 0

-1

0

-3

0

+3

+3

0
-1

-3

0

+3

+3

-2
0

-1
-2

-3

-2

+3

+1

0
-1

-1

-3

-2-

+3

+3

0
-1

-2

-3

3

+3

+3

-2
0

0
+2

-3

-1

+3

+3

-3
0

-3
-2

-3
-2

-1-3

+2

-1
-1

280

maintain marginal populations in
marginal habitats. A rule for judging
the suitability of a proposed land-use
change might be that land-use
change that can be accomplished
while promoting trend lines like lb3
(with strong positive slopes) or
i which produce conditions like line 2
in figure 6 are environmentally ac-
ceptable and can be accomplished if
they are socially and economically
I desirable.

LITERATURE CITED

Spicer, R. B., J. C. deVos, Jr., and R.
L. Glinski. 1985. Status of the
Mount Graham red squirrel, Tami-
asciurus hudsonicus grahamensis
(Allen), of southeastern Arizona.
Unpublished report by Arizona
Game and Fish Dep. for U.S. Fish
Wildl. Serv., Office of Endangered
Species. 48 p.

U.S. Forest Service. 1987. Mount Gra-
ham red squirrel. A biological as-
sessment of impacts proposed Mt.
Graham Astrophysical Project.
Coronado National Forest,
Tucson, Ariz. 92 p.

U.S. Forest Service. 1988. Mount Gra-
ham red squirrel. An expanded
biological assessment. Coronado
National Forest, Tucson, Ariz.
130 p.

Comparative Effectiveness of
Pitfalls and Live-Traps in
IVIeasuring Small Mammal
Community Structure^

Robert C. Szaro,^ Lee H. Simons,^ and Scott
C. Beifit^

Abstract.— The effectiveness of pitfalls and live-
traps for assessing small mammal community
structure was compared in burned and unburned
upland Sonoran Desert and in an elevational series
of Sycamore riparian and adjacent habitats in
Arizona. Although, live-traps v/ere more effective in
recapturing previously captured small mammals
and usually resulted in more total captures of new
individuals, neither method gave a complete
assessment of small mammal community structure.

Several studies that compared vari-
ous types of pitfalls and live-traps
(also called box- or cage-traps) in the
field (Chelkowska 1967, Boonstra
and Krebs 1978, Peterson 1980, Boon-
stra and Rodd 1984, Mengak and
Guynn 1987) found the sampling effi-
ciency of the two methods varied
considerably (Andrzejewski and
Rajska 1972, Briese and Smith 1974,
Cockbum et al. 1979, Williams and
Braun 1983). Pitfall cone traps were
more effective than live-traps in sam-
pling small mammals, particularly
shrews in southern Finland (Pankak-
oski 1979). In contrast, pitfalls were
less effective than live-traps in cap-
turing small-bodied mice in
Durango, Mexico, although more
shrews (Notiosorex crawfordi) were
taken in pitfalls (Peterson 1976). Pit-
falls of various materials, shapes, and
sizes, with and without drift fences,

'Paper presented of symposium, Mon-
ogement of Amphibions, Repfiles. ortd
Small Mammals in Norfhi America. (Flag-
sf off, AZ. July 19-21 1988.)

'Roberf C. Szaro is Research) Wildlife Bi-
ologist, USDA Forest Service. Roclcy Moun-
tain Forest and Range Experiment Station,
Arizona State University Campus. Tempo, AZ
85287-1304.

^Lee hi. Simorts. formerly a graduate stu-
dent. Arizona State University. Department
of Zoology, Tempo, Arizona is currently a
graduate student. Graduate Group in Ecol-
ogy, University of California, Davis, CA
95616.

"Scott C. Bel fit is Wildlife Biologist. De-
partment of the Army, Wildlife Manage-
ment Section. Fort Huachuca. AZ 856 13-
6000. Belfit's current address is P.O. Box 336.
Fort Belvoir, VA 22060-0336.

have been used for capturing small
mammals (Howard and Brock 1961,
Andrzejewski and Wroclawek 1963,
Pucek 1969, Boonstra and Krebs
1978, Pankakoski 1979). This lack of
standardization makes it difficult to
assess the relative effectiveness of
pitfalls versus live-traps in sampling
small mammals by comparing data
between studies. Conflicting results
from these studies argue for more
comparisons using controls for as
many extraneous factors as possible.

Small mammals respond dramati-
cally to many environmental factors,
thus confounding attempts to assess
species or community relationships.
Sampling biases caused by climate
and differences in activity and loco-
motor adaptations of various species
further compound this problem. Still,
trapping remains the most practical
method for assessing small mammal
populations (Williams and Braun
1983). Because responses to trapping
methods may differ, even within the
same species (Andrzejewski and
Rajska 1972), diverse sampling
schemes might reveal population
dynannics and community structure
more completely than any single
method (Weiner and Smith 1972,
Boonstra and Krebs 1978).

We compared the effectiveness of
live-traps versus pitfalls in riparian
and desert habitats in Arizona to an-
swer the following questions: (1)
Does sampling method influence es-
timates of sf)ecies com{X)sition and
abundance? (2) Are various species

captured or recaptured differen-
tially? (3) Are individuals within a
species captured differentially? (4)
Does habitat structure influence the
effectiveness of these methods?

Study Areas and Methods

Riparian and Adjacent
Communities

The riparian and adjacent communi-
ties (referred to in general as the ri-
parian area) were located at Garden
Canyon, Fort Huachuca Military Res-
ervation, Arizona; elevations ranged
from 1500 to 1630 m. Riparian com-
munities sampled, from lowest to
highest elevation, were sycamore
(Platanus mrightii), sycamore/juniper
(Juniperus monosperma), and syca-
more/juniper (/. deppeam)/odk
(Quercus arizonica, Q. emoryi, and Q.
hypoluecoides) (Szaro 1988). Plant
communities sampled adjacent to the
riparian corridor, from lowest to
highest elevation, were composite
{Heterotheca spp.) /grassland {Poa
spp.), junifjer (/. monosperma) wood-
land, and oak (Quercus emoryi) wood-
land.

Six trap stations were set in each
of six habitats: composite /grassland,
sycamore riparian, juniper wood-
land, sycamore/ juniper riparian, oak
woodland, and sycamore/juniper/
oak riparian forest (figs. 1-6) (36 sta-
tions in all). Trap stations consisted
of two unbaited pitfalls (18.9 L or 5

282

■ ...> -^^ ' ^

.V .■• v'.'-''-" ■-■■' *•'■■■■" ■

■.''"..■.■■.■>.■..■
• ■ ■ . .< ^ < ""

" / " ' - 4 •■ " -

Figure 1 .—Arizona sycarDore (Piafanus wrightil) study site, Garden Figure 4.— Composite (Heterotheca spp.)/grassiand (Poa spp.)

Canyon, Fort Huactiuca Military Reservation, Arizona; eievation ca. study site, Garden Canyon, Fort Huactiuca Military Reservation,
1500 m. Arizona; elevations ca. 1510 m.

Figure 2.— Arizona sycarnore (Platanus wrightii)/one-seed juniper Figure 5.— One-seed juniper (J. monosperma) woodland study site,

(Juniperus monosperma) study site. Garden Canyon, Fort Garden Canyon, Fort Huactiuca Military Reservation, Arizona; ele-

Huact)uca Military Reservation, Arizona; elevation ca. 1565 m. vations ranged from 1570 m.

Figure 3.— Arizona sycamore (Platanus wrighfii)/a\\\ga\of juniper (J.
deppeana)/m\xBd oak (Quercus arizonica, Q. emoryl, and Q. hy-
poluecoldes) study site, Garden Canyon, Fort Huactiuca Military
Reservation, Arizona; elevation ca. 1610 m.

Figure 6.— ErTK>ry oalc (Quercus emoryi) woodland study site, Gar-
den Canyon, Fort Huactiuca Military Reservation, Arizona; eleva-
tions ca. 1590 m.

283

gal.; 29 cm in diameter by 36 cm
deep) with a 7.6-m-long by 20-cm-
high drift fence between buckets.
Covers were propped 2.5-5 cm above
openings mouths. Pitfalls were open
from 16 April through 28 May and
from 20 July through 5 September
1986 (6408 trap-nights) and were
checked three times each week. Sher-
man live-traps (8 by 9 by 23 cm)
baited with rolled oats were set
around each pitfall station in an 8-
trap pattern with at least 5 m be-
tween traps and pitfalls. Live-traps
were set from 12 to 16 May and from
17 to 21 August 1986 (2304 trap-
nights) and were checked each morn-
ing. Most live-trap captures were re-
leased after being ear-tagged. Except
for some Notiosorex, all pitfall cap-
tures were collected. Identification of
all mammals follows Hoffmeister
(1986). Thomomys species include
pure and hybrid T. utnbrinus and T.
bottae.

Desert Community

The desert study area was in the
Tonto National Forest, Maricopa
County, 30 km east of Phoenix, Ari-
zona. The site was rocky desert dis-
sected by sandy washes; elevations
ranged from 450 to 550 m. Vegetation
was typical of the Arizona upland
subdivision of the Sonoran Desert
biome (Brown 1982), with mesquite
(Prosopsis juliflora) along wash banks
and palo verde (Cercidium micro-
phyllum), bursage (Ambrosia deltoides),
and cholla (Opuntia acanthocarpa) on
slopes.

Two grids were established 90 m
apart, each with 100 sampling sta-
tions placed in a 10 by 10 pattern
with 10-m intervals between stations.
Grid 1 was in mature desert and grid
2 had 50% of vegetative cover
burned on 7 June 1985, immediately
before the start of trapping (figs. 7-8).
Interspaced between live-traps (10 by
10 by 25 cm) on each grid, but no
closer than 10-m intervals, were 20
single pitfalls (37.9 L or 10 gal., 34 cm

diameter by 40 cm deep) buried to set and baited with rolled oats for

the rim with a cover propped 5-10 two consecutive nights on 19 occa-

cm over the opening. Live- traps were sions between 10 June 1985 and 3

Figure 7.— Unburned desert study area, Tonto National Forest. Maricopa County, 30 km east
of Phoenix, Arizona; elevation ranged from 450 to 550 m.

Figure 8.— Burned desert study area, Tonto National Forest, Maricopa County, 30 Icm east of
Ptioenix, Arizona; elevation ranged from 450 to 550 m.

284

August 1986 — weekly in spring and
early summer, biweekly from middle
to late summer, and monthly in fall
and winter (Simons 1986). Unbaited
pitfalls were always open during
live-trapping and often in between
when live-trapping occurred weekly
or biweekly (March-September). All
captures except for casualties were
marked and released. Each method
was matched with an approximately
equal sampling effort (about 3800
trap-nights per grid).

Results and Discussion

Species Composition and
Abundance

Live-traps and pitfalls provided dif-
ferent estimates of species composi-
tion and relative abundance at both
study areas. In the riparian area we
observed no consistent pattern be-
tween trapping method and number
of species captured (table 1). Live-
traps caught more species in two

habitats, pitfalls, in three habitats,
and in the sycamore/ juniper/ oak
both methods captured two species.
Neither method captured all species
in a given habitat except in oak
woodland where only two species
were encountered and pitfalls cap-
tured both. However, live-trapping
was significantly more successful
than pitfalls in number of new cap-
tures per trap-night (chi-square, P <
0.05) in all habitats except juniper
woodland, where both methods
yielded equal numbers.

In the desert, live-traps caught
more species than pitfalls (table 2).
Moreover, significantly more new
captures and total captures (chi-
square, P < 0.05) occurred in live-
traps than in bucket-traps in both
burned and unburned plots (table 2).
These results differ from those of
Williams and Braun (1983) who re-
ported that number of species and
total number of captures were
greater in pitfalls than in the com-
bined catch of snap- and live-traps.
They recorded six species in pitfalls

and four in snap- and live-traps.
Their success with pitfalls was no
doubt increased because each trap
was one-third filled with water,
drowning all captures. Trapping suc-
cess for voles (Clethrionomys glareo-
lus) was also reported to be higher in
pitfalls versus live-traps but may
vary with social level, age, and re-
productive period (Andrzejewski
and Rajska 1972, Andrzejewski and
Wroclawek 1963, Chelkowska 1967).

New individuals represented only
31.5% and 26.2 % of total captures in
live-traps on the burned and on the
unburned plots, respectively. In con-
trast, 95.8% and 92.7% of all captures
in pitfalls on the burned and un-
burned areas, respectively, represent
new individuals. The lack of recap-
tures in pitfalls is not explained by
differential mortality between meth-
ods because sampling with both
methods occurred simultaneously,
and most animals were marked and
released. These differences maybe at
least partially due to increased at-
tractiveness of live-traps with bait

Table 1.— Total number of new Individuals captured in riparian and associated habitats using live-traps (384 trap-
nights/habftat) and pitfalls (1068 trap-nlghts/habitat) during spring and late summer 1986.

Composite/ Sycarrtore Juniper Sycarnore/ Oak Sycamore/ Total

grass woodland juniper woodland Juniper/oal<

Species Live- Pit- Live- Pit- Live- Pit- Live- Pit- Live- Pit- Live- Pit- Live- Pit-
trap fall trap fall trap fall trap fall trap fall trap fall trap fall

Neotoma albigula ]

Notiosorex crawfordi 2 31 1 17 22 2

Onychomys forridus 8 2 ] 4 4

Perognafhus fkivus 1 1

Perognathus hispidcs 2 1

Perognafhus pencillatus ^

Peromyscus boylei 12 .1 13 13 3 12

Peromyscus leucopus 3 3

Peromyscus maniculatus 5 2

Reithrodontomys fulvescens 4 2 5 2 1 1

Reithrodontomys megahfis 1
Sigmodon ochrognofhus 1
Sorex arizonae

Jhomomysspp. 2 13

Total captures 21 8

Species richiness 6 4

Overall species richness 8
New coptures/trap-night 6.07 0.75
X 100

All captures/trcp-night 7.26
X 100

1

76

12

-7
1

1

1

3 .

1

51

3

6

5

2

9

6

1
1

2

6

19

37

8

23

16

26

13

5

15

4

92

103

4

5

5

4

2

3

1

2

2

2

12

8

8

7

5

2

4

14

4,95

3.46

2.08

2.15

4.17

2,43

3.38

0.47

3.90

0.37

3.99

1.61

10.38

2.10

9.84

4.92

6.00

6.76

285

and with concentrated odors from
previous captures (Boonstra and
Krebs 1978, Daly and Behrends
1984). Our results show that pitfalls
provide very different estimates of
species composition and abundance
than live-traps. We therefore ques-
tion basic assumptions of the popular
methods of population estimation
that assume either equal catchability
of all members in the population
(Jolly 1965) or nearly complete caf>-
ture and enumeration of a popula-
tion (Krebs 1966, Hilborn et al. 1976).

Differential Trapping Effectiveness
Between Species

In the riparian area, 80 of 81 shrews
(Notiosorex crawfordi and Sorex arizo-
me) and all gophers {Thomomys spp.)
were captured in pitfalls. In contrast,
only 5 of 67 captures of Peromyscus (3
species) were in pitfalls (table 1).
Peromyscus spp. were also recaptured
most frequently (57 of 64 recaptures).
Similar results were found in the Si-
erra Nevada where species such as
shrews (Sorex trowbridgii and S. mon-
Hcolus) and gophers (Thomomys bot-
tae), which tend to travel in burrows
or runways or along obstacles, were
usually captured in pitfalls (Williams
and Braun 1983). Williams and Braun
(1983) reported in their first test that
pitfalls were particularly poor for
capturing white-footed mice (Pero-
myscus). In a subsequent test they
implied these mice might be taken in
pitfalls after losing their caution for
strange objects. This did not happen
in our study because very few Pero-
myscus were captured in pitfalls over
an extended period even though live-
trapping showed them to be com-
mon. More likely Peromyscus may
easily escape pitfalls by jumping out,
but more are recorded after drown-
ing in water-filled pitfalls (Williams
and Braun 1983), especially when
other traps, such as snap- or live-
traps, are missing.

In the desert habitat, a single
shrew (Notiosorex crawfordi) was

caught in a pitfall whereas two
species (Dipodomys merriami and
Peromyscus eremicus) were caught
only in live-traps. Only 1 of 181 cap-
tures (50 different individuals) of
Neotoma albigula was in a pitfall
whereas only 1 of 9 Onychomys tor-
ridus was not captured in a pitfall.
Onychomys was probably unable to
jump out of the buckets used in this
habitat. Noted accumulations of
Neotoma feces overnight in many pit-
falls indicated these rodents had
been present but left. Apparently
larger species either avoid pitfalls or
simply jump out of them (Clockburn
et al. 1979, Williams and Braun 1983).

Differential Trapping Effectiveness
Wittiin Species

Few significant differences in
weights of small mammals caught
with the two methods were ob-
served, but weights tended to be
lower in pitfalls. In the riparian area,
mean weights of Reithrodontomys ful-
vescens were significantly higher in
live-traps (14.3 + 0.65 (S.E.) g versus
5.1 + 0.56, t-test, P < 0.001, N = 12). In

the desert, weight differences be-
tween trap methods were not signifi-
cant for animals less than about 20 g.
However, a significant difference oc-
curred in the mean weight of Per-
ognathus baileyi in live-traps (25.7 +
0.97 g) versus pitfalls (20.8 ± 1.61 g; t-
test, P = 0.014). Similarly, the mean
weight of Neotoma albigula caught in
live-traps was 109.0 + 8.93 g, whereas
the single capture in a pitfall
weighed 31.0 g.

Likewise in Canada and Poland,
voles (Microtus townsendii and Cle-
thrionomys glareolus) captured in pit-
falls were smaller than conspecifics
taken in live-traps (Andrzejewski
and Rajska 1972, Boonstra and Krebs
1978). This apparent relationship be-
tween size and susceptibility to pit-
falls is likely related to jumping abil-
ity which tends to increase with age.
For some sf>ecies, pregnant females
may be more susceptible to pitfalls.

Effects of Habitat on Trapping
Effectiveness

Trapping results for Onychomys tor-
ridus varied substantially between

Table 2.--Total number of new individuals captured in burned and un-
burned desert habitats using live-traps cmd pitfalls (3800trap-nights/habl-
tat/trap type).

Burned Unbumed Total

Species Live-trap Pitfall Uve-trap Pitfall Live-trap Pitfall

AmmospermopNIus harrisii 2 14 1 6 1

Dipodomys memami 11 3 14

Neofomo albigula 11 1 38 49 1

Notiosorex crawfordi 1 1

Onychomys forridus 17. 1 1 8

Peromyscus eremicus 2 3 5

Perognathus amplus 91 76 85 25 176 101

Perognathus baileyi 18 30 21 10 39 40

Total captures 136 115 154 38 290 152

Species richness 7 5 6 5 7 6

Overall species richness 7 8 8

New captures/trap-night 3.58 3,02 4.05 1.00 3.81 2.00
X 100

All captures/trap-night 11.36 3.16 15.45 1.08 13.41 2.11
X 100

286

vegetative communities. On the des-
ert sites, 8 of 9 captures v^ere in pit-
falls whereas in composite/grass
habitat in Garden Canyon, 8 of 10
captures were in live-traps. Four cap-
tures were made with each method
in the juniper woodland. Differences
in trapability of Oncychomys may be
due to different depths of pitfalls in
desert (40 cm) versus riparian (36
cm) habitats.

Except for Perogmthus spp., ro-
dents were about equally susceptible
to pitfalls relative to live-traps in
both burned and unburned desert
habitats. Differences in total number
of individuals captured by both
methods in the desert areas may be
due to (1) difference in abundance of
species on burned and unburned
plots (Simons 1986); or (2) differences
in activity patterns related to the
drastic difference in shrub cover. Per-
ogmthus spp. typically prefer brush
or "cover" microhabitats (Price 1978)
and raised pitfall covers may have
attracted these mice more on the
burned area where natural cover was
scare than on the unburned area
where natural cover was dense (Si-
mons 1986). Whatever the cause, the
results are similar to those found in
desert-shrub and mesquite-grassland
habitats in Durango, Mexico, where
significantly more small-bodied
mammals were captured with live-
traps than with pitfalls (5.4 L tin can
pitfalls with a depth of 25.4 cm) (Pe-
terson 1980). Possibly a greater num-
ber of captures (i.e., sample size) may
be needed to fully reveal the impact
of habitat on trapping methodology.

Conclusions

Neither method alone was able to
fully assess small mammal communi-
ties in the desert-scrub and riparian
communities we investigated. We
recommend the use of both methods,
particularly when it is imjx)rtant to
include species such as shrews that
are not easily caught in live-traps in
investigations of small mammal com-

munity structure and habitat rela-
tionships.

Acknowledgments

We thank T. J. O'Shea, M. G. Ryan,
D. W. Uresk, and D. F. Williams for
their critical reviews of this manu-
script. C. Munns helped check pit-
falls at Garden Canyon. The Depart-
ment of Zoology, Arizona State Uni-
versity provided support for L. Si-
mons.

Literature Cited

Andrzejewski, Roman, and Ewa
Rajska. 1972. Trapability of bank
vole in pitfalls and live traps. Acta
Theriologica 17:41-56.

Andrzejewski, Roman, and Henryka
Wroclawek. 1963. Metal cylinder
as a live trap with a bait. Acta Th-
eriologica 6:297-300.

Boonstra, Rudy, and Charles J.
Krebs. 1978. Pitfall trapping of
Microtus townsendii. Journal of
Mammalogy 59:136-148.

Boonstra, Rudy, and F. Helen Rodd.
1984. Efficiency of pitfalls versus
live traps in enumeration of popu-
lations of Microtus pennsylvanicus.
Canadian Journal of Zoology
62:758-765.

Briese, Linda A., and Michael H.
Smith. 1974. Seasonal abundance
and movement of nine species of
small mammals. Journal of Mam-
malogy 55:615-629.

Brown, David E. 1982. Biotic commu-
nities of the American South-
west— United States and Mexico.
Desert Plants 4:1-342.

Chelkowska, Henryka. 1967. An at-
tempt at comparing two methods
of trapping small rodents (in pit-
falls and live traps). Ekologia
Polska Seria A 15:779-785.

Cockburn, A., M. Fleming, and J.
Wainer. 1979. The comparative
effectiveness of drift fence pitfall
trapping and conventional cage
trapping of vertebrates in the Big

Desert, North-Western Victoria.
Victorian Naturalist 96:92-95.

Daly, Martin, and Philip Behrends.
1984. Effect of moving traps be-
tween trapping stations upon ro-
dent retrapping data. American
Midland Naturalist 112:205-207.

Hilbom, R., J. A. Redfield, and C. J.
Krebs. 1976. On the reliability of
enumeration for mark and recap-
ture census of voles. Canadian
Journal of Zoology 54:1019-1024.

Hoffmeister, Donald F. 1986. Mam-
mals of Arizona. 602 p. University
of Arizona Press and Arizona
Game and Fish Department,
Tucson.

Howard, Walter E., and Elbert M.
Brock. 1961. A drift-fence pit trap
that preserves captured rodents.
Journal of Mammalogy 42:386-391.

Jolly, G. M. 1965. Explicit estimates
from capture-recapture data with
both death and immigration — sto-
chastic model. Biometrica 52:225-
247.

Krebs, Charles J. 1966. Demographic
changes in fluctuating populations
of Microtus californicus. Ecological
Monographs 36:239-273.

Mengak, Michael T., and David C.
Guynn, Jr. 1987. Pitfalls and snap
traps for sampling small mammals
and herpetofauna. American Mid-
land Naturalist 118:284-288.

Pankakoski, Erkki. 1979. The cone
trap — a useful tool for index trap-
ping of small mammals. Annales
Zoologici Fennici 16:144-150.

Peterson, Michael K. 1976. Notewor-
thy range extensions of some
mammals in Durango, Mexico.
Southwestern Naturalist 21:139-
142.

Peterson, Michael K. 1980. A com-
parison of small mammal popula-
tions sampled by pit-fall and live-
traps in Durango, Mexico. South-
western Naturalist 25:122-124.

Price, Mary V. 1978. The role of mi-
crohabitat in structuring desert
rodent communities. Ecology
59:910-921.

Pucek, Z. 1969. Trap response and
estimation of numbers of shrews

287

in removal catches. Acta Theriol-
ogica 14:403-426.

Simons, Lee Hervey. 1986. Rodent
dynamics in burned and unburned
desert habitat with tests of the ef-
fects of fire. M.S. Thesis. 57 p. Ari-
zona State University, Tempe.

Szaro, Robert C. 1988. Riparian forest
and scrubland community types
of Arizona and New Mexico. Des-
ert Plants 9 [In press].

Weiner, James G., and Michael H.
Smith. 1972. Relative efficiencies
of four small mammal traps. Jour-
nal of Mammalogy 53:868-873.

Williams, Daniel F., and Suzanne E.
Braun. 1983. Comparison of pitfall
and conventional traps for sam-
pling small mammal populations.
Journal of Wildlife Management
47:841-845.

The Role of Habitat Structure
in Organizing Smali IVIammal
Populations and
Comnnunities^

Gregory H. Adler^

Abstract.— Microhabitat structure influences
population density more than other demographic
variables such as age and sex composition.
Microhabitat heterogeneity, or quantitative
variation in microhabitat structure, apparently has
little influence on phenomena such as population
stability. Scale mediates effects of habitat structure
and heterogeneity on population and community
organization. I suggest that microhabitat structure
influences density more than other aspects of
demography, whereas mocrohobitat structure and
heterogeneity are more important in influencing
population stability, demography, and community
structure.

Environmental heterogeneity has
maintained a position of prominence
in theoretical population and com-
munity ecology (reviewed by Levin
1976, Wiens 1976, and Wiens et al.
1986). Heterogeneity allows organ-
isms to select different habitats,
which subsequently can have pro-
found consequences for the organiza-
tion of populations and connmunities.
Environmental heterogeneity can be
studied, both theoretically and em-
pirically, at different scales. Conclu-
sions based on the study of habitat
structure may differ widely depend-
ing upon the scale of structure exam-
ined. The scale of environmental sub-
division can be viewed as occurring
along various continua, e.g., from the
area occupied by a single individual
to a biogeographic or continental
area (Wiens et al. 1986), or from mi-
crohabitat to macrohabitat.

In this paper, I concentrate on the
microhabitat to macrohabitat scale. I
define microhabitat as physical habi-
tat characteristics likely to vary over
the home range of a single individual
(e.g., the number of herbaceous
stems within a circumscribed area)
and macrohabitat as the major habi-

' Paper presented at symposium. Man-
agement of Amphibiarts. Reptiles, and
Small Mammals in Nortt) America. (Flag-
staff. AZ. July 19-21. 1988.)

'Gregory H. Adier is Research) Fellow in
Population Sciences. Department of Popu-
lation Sciences. Schiool of Public l-lealthi.
Harvard University. 665 Huntington Avenue.
Boston. MA 021 15.

tat type where an entire population
may be found (e.g., grassy field or
deciduous woodland in the case of
small mammals). Microhabitat struc-
ture therefore can vary substantially
within a single macrohabitat.

I summarize results from a series
of long-term studies on the role of
habitat structure in organizing small
mammal populations and communi-
ties that I conducted in eastern Mas-
sachusetts. These studies were de-
signed to examine (1) habitat associa-
tions and habitat selection and the
roles of intra- and interspecific inter-
actions in affecting habitat utiliza-
tion, and (2) the influence of habitat
structure on density and demogra-
phy. In these studies, I focus primar-
ily on microhabitat structure, and I
develop a conceptual scheme which
shows how microhabitat and mac-
rohabitat structure organize small
mammal populations and communi-
ties.

STUDY SITES AND GENERAL
METHODS

Study Sites

The long-term studies were con-
ducted at three sites in eastern Mas-
sachusetts: Broad moor /Little Pond
Audubon Sanctuary, South Natick;
Great Island, near West Yarmouth;
and the University of Massachusetts
Nantucket Field Station, Nantucket.
Sampling areas within each study

site were confined to a 300-ha area
and were exposed to the same cli-
mate and the same predators, com-
petitors, and parasites.

Broadmoor consists of a mosaic of
grassy fields separated by nnixed de-
ciduous-coniferous woodland. Sam-
pling at Broadmoor was confined to
the fields, which were dominated by
the grasses Agropyron repens and Poa
pratensis. Other herbaceous and
woody plants, including goldenrod
(Solidago spp.), milkweed (Asclepias
syriaca), poison ivy (Rhus radicans),
and several species of deciduous tree
saplings, were much less prevalent.

Great Island is a 240-ha island
connected to mainland Cape Cod by
a causeway. The island is dominated
by deciduous and coniferous wood-
land but has structurally simpler
habitat along the shore. This shore-
line habitat consists primarily of
beach grass (Ammophila breviligulata),
with patches of poison ivy, Virginia
creeper (Parthenocissus quinquefolia),
bayberry (Myrica pensylvanica), rose
(Rosa Carolina), and juniper (Juniperus
virginiana).

Nantucket Island (ca. 12,300 ha
and lying approximately 30 km off
the coast of Cape Cod) has large ar-
eas of low, dense woody growth
(heath) where small mammals were
sampled. Heath at the study site was
composed primarily of rose and bay-
berry, with patches of goldenrod and
other herbaceous plants and grasses
interspersed within the brush. Scat-
tered juniper trees also were present.

289

Sampling Procedures

I sampled small mammals at each
study site by monthly live-trapping
with Longworth live-traps for ap-
proximately 4 to 5 years. At each
study site, I monitored two 0.4-ha
grids located in grassy or brushy
habitat. One grid served as a control
in which all small mammals were
individually marked by ear-tags (ro-
dents) or toe-clips (insectivores). The
other grid, located 30.4 m from the
control and situated in contiguous
habitat, served as an experimental
grid from which all small mammals
were removed permanently upon
first capture (Adler 1985). All small
mammals captured on this grid after
the initial removal period were con-
sidered colonists. I also sampled
small mammals on 4 nearby trapping
plots which also were located in
similar macrohabitat but covered a
range of microhabitats. Each plot
consisted of two parallel traplines
located 30.4 m apart. Each trapline
was 15 stations long at Broadmoor
(except on one plot where both tra-
plines were 12 stations long) and on
Nantucket and 20 stations long on
Great Island. These plots were
trapped on a rotation basis (Adler
1987). On Great Island, an additional
4 control grids were monitored
monthly from April through Septem-
ber for five years (Adler and Wilson
1987). These grids were not confined
to structurally simple macrohabitats
but ranged from grassland to mature
woodland habitats.

Grid 1 was located at the edge of a
stand of pitch pine. (Pinus rigida),
white oak (Quercus alba), and black
oak (Q. velutim). Dense brushy
understory covering a large portion
of the grid consisted of bayberry,
huckleberry (Gaylussacia baccata), and
inkberry holly (Ilex glabra). Low-lying
areas of the grid were damp and har-
bored large cranberry (Vaccinium
macrocarpon) and sundew {Drosera
spp.). Very little herbaceous vegeta-
tion was present. Grid 2 also was lo-
cated at the edge of a pitch pine.

white oak, and black oak woodland
but was more elevated and conse-
quently drier. A dense brushy under-
story consisted of bayberry, poison
ivy, and common greenbrier (Smilax
rotundifolia). Grass was present in the

brushy, treeless portions of the grid.
Grid 3 was located within a white
oak and black oak woodland. A
dense shrub cover of blueberry,
bearberry (Arctostaphylos uva-ursi),
common greenbrier, and bullbrier

Table 1 .—Description of the habitat variables measured at each trap sta-
tion. All variables measured as proporHons was arcsin square root trans-
formed.

Name

Description

WOOD Density of woody stems within a 1 -m^ circle at ground

level.

HERB Number of Inerbaceous stems (excluding grasses and

sedges) within a 1-m^ circle at ground level.

WDSPEC Number of woody species within a 1 -m^ circle at ground

level.

HBSPEC Number of herbaceous species (excluding grasses and

sedges) within a 1-m^ circle at ground level.
HB50 Number of herbaceous sterris (excluding grasses and

sedges) within a 1-m^ circle at 50 cm above ground level.
HBICX) Number of herbaceous stems (excluding grasses and

sedges) within a 1-m^ circle at 1 m above ground level.
VHBDEN Mean of HERB, HB50, and HBIOO.

WD50 Number of woody stems within a 1 -m^ circle at 50 cm

above ground level,
WDIOO Number of woody stems within a 1-m^ circle at 1 m

above ground level.
VWDEN Mean of WOOD, WD50, and WDIOO.

OVER Number of overstory species within a 15-m^ circle.

UNDER Number of shrub level species within a 15-m^ circle.

FORB Number of f orb species within a 15-m^ circle.

GRASPEC Number of grass and sedge species within a 1 5-m^ circle.
GRNDSPEC Number of woody ground-dwelling vine species within a

IS-m^circie,

SPECIES Total number of angiosperm and gymnosperm species

within a 15-m^ cirde.
TPSHRUB Trar^formed proportion of a 15-m^ circle dominated by

woody shrub-level vegetation.
TPHERB Transformed proportion of a ]S-m'^ circle dominated by

herbaceous vegetation (excluding grasses and sedges).
TPGRND Trar^sformed proportion of ]5-m^ circle dominated by

woody ground-dwelling vines.
TPGRASS Trar>sformed proportion of IS-m^ circle dominated by

grasses and sedges.
TPVEG Trortsformed proportion of a 15-m^ circle covered by

vegetation.

HBGRND An index of herbaceous ground cover (excluding grasses

and sedges), calculated as HERB-HB50,
WDGRND An index of woodv ground cover calculated as WOOD-

WD50.

TPCANOPY Trar^sformed proportion of a canopy cover measured only

on the five Great Island control grids.
TPGREEN Trar^sformed proportion of evergreen canopy cover,

measured only on the five Great Island control grids.

290

greenbrier (S. bom-nox) was present,
along with bracken fern (Pteridium
aquilinium). Grid 4 was located on
Pine Island, a 7-ha islet 37 m from
Great Island and connected to the
latter by a narrow sandy spit. White
oak and black oak formed a canopy
over much of the grid, and a dense
woody understory of bayberry and
other shrubs also was present. Dense
beach grass was present in the tree-
less portions of the grid. Grid 5 was

the companion control for the experi-
mental grid and was located in dense
beach grass containing scattered
patches of bayberry, juniper, and poi-
son ivy.

I sampled vegetation structure at
every trap station on all grids and
plots by measuring 23 habitat vari-
ables related to plant structure and
species richness (table 1). Two addi-
tional habitat variables describing
canopy structure were included in

Tcrt>le 2.— Summary of the sampling design and statistical approach em-
ployed in this study.

Sampling area

Control grid

Experimental
grid

Trapping plots

Additional con-
trol grids (4 at
Great Island)

Topic

Memods

Habitat associotions.

Temporal dynamics of
habitat use.

Habitat selection.

Relationship between
demography and mi-
crohabitat structure
within a macrohabitat.

Relationship between
demography and mi-
crohabitat structure of
a habitat generalist
across macrohabitat
boundaries.

Multiple linear regression
of numbers of captures at
a trap station on mi
crohabitat variables de-
rived from PCA,

DFA to derive a quantita-
tive measure of seasonal
habitat use (the distinction
between favorable and
unfavorable microhabi-
tafs, or habitat discrimina-
tion). Regression of dis-
crimination values on
population densities to de-
termine the relatior^ship
between microhabitat use
and intra- and interspeci-
fic population der^ities.

Multiple linear regression
of numbers of captures at
a trap station (in a per-
turbed area) on mi-
crohabitat variables de-
rived from PCA/ compared
with control grid.

Regression (and residual
analysis) of demographic
variables on plot means of
microhabitat gradients
and heterogeneity.

Regression (and residua!
analysis) of demographic
variables on grid means of
microhabitat gradients
and heterogeneity.

the analysis on Great Island control
grids (table 1). Measurement proce-
dures were given by Adler (1985)
and Adler and Wilson ( 1987).

Data Analysis

I relied extensively upon principal
components analysis (PCA) and dis-
criminant function analysis (DFA) in
order to uncover the structure of
complex and temporally variable
small mammal populations and their
relationships to habitat structure.
Specifically, my aims were to (1) re-
duce the number of habitat dimen-
sions, (2) derive a quantitative meas-
ure of habitat heterogeneity, (3)
quantify patterns of habitat utiliza-
tion, (4) combine covarying demo-
graphic traits into single variables,
and (5) derive indices of demo-
graphic variability.

In these studies, I recognized two
related descriptors of microhabitat
structure. I defined a microhabitat
structure-diversity variable or gradi-
ent as a characteristic that described
the physical structure of the mi-
crohabitat and that varied in magni-
tude along a continuum. I defined
microhabitat heterogeneity as a
quantitative measure of horizontal
variation in microhabitat characteris-
tics (August 1983, Adler 1987).

I subjected the habitat data meas-
ured at each trap station to PCA to
reduce the number of habitat vari-
ables. At each site, I conducted two
PC As of the 23 variables, one with
control and experimental grids com-
bined and one with the 4 trapping
plots combined. I also conducted a
PCA of 24 habitat variables for all
five control grids on Great Island
combined. HBIOO was eliminated
from this analysis because only one
nonzero value was recorded on the
five grids.

Each principal component (PC)
with an eigenvalue greater than 1.0
was retained for further analysis as a
new habitat variable. Principal com-
ponents derived from PCAs of grid

291

and plot data were quite similar
within each site, based upon factor
loadings on the original habitat vari-
ables (Adler 1985, 1987). At Broad-
moor, five PCs were retained for
analysis from both grid and plot
data, whereas six were retained from
analysis of grid data; four PCs were
interpreted similarly in both data
sets. PCAs of Nantucket grid and
plot data both yielded seven retain-
able PCs, three of which could be in-
terpreted similarly between the two
data sets. The PCA of habitat data
from the five control grids on Great
Island yielded seven PCs.

I computed a microhabitat hetero-
geneity index for each of the four
trapping plots at the three study sites
and for each of the five control grids
on Great Island (Adler 1987, Adler
and Wilson 1987). This index was
based on the supposition that the
standard deviation of the within-plot
or within-grid mean vector of a PC
described the variability of a mi-
crohabitat gradient on a given plot or
grid. Since each successive PC con-
tributed less to the total variance in

habitat data, I adjusted for each PC's
contribution to the total variance by
multiplying the factor scores by the
square root of that PC's eigenvalue.

I examined capture data in rela-
tion to habitat structure at both the
level of individual trap stations
(habitat association and selection)
and at the level of a grid or plot (de-
mography). I used multiple linear
regression and residuals analysis to
relate these small nnammal (depend-
ent) variables to habitat (independ-
ent) variables. More complete de-
scriptions of analytical techniques are
given in each section below, and a
brief outline of the sampling design
is given in table 2.

SPECIES COMPOSITION

I recorded 9,170 captures of 10 small
mammal species in 42,773 trapnights
at the 3 study sites (table 3). Each
study site generally had an abundant
herbivore (Microtus pennsylvanicus),
an abundant granivore (Peromyscus
leucopus, except at Broadmoor where

it was rare in the grassland trapping
areas), a common insectivore (Blarina
brevicauda or Sorex cinereus), and any
of several rarer granivores, omni-
vores, or insectivores.

HABITAT STRUCTURE AND
POPULATION STATISTICS

Habif at Associations and
Selection

Study Purpose

I exan\ined both small mammal mi-
crohabitat associations and selection
at all three study sites (Adler 1985).
Density-dependent effects of con-
specifics and other species may re-
strict access to certain habitat types,
thereby resulting in different patterns
of habitat utilization. I therefore re-
served the term habitat selection for
situations where individuals had
more or less unrestricted access to a
variety of habitat types.

Table 3.— Trapping effort and numbers of captures of small mammals at three study sites In eastem Massachusetts.
Species designations are MP (Microtus pennsytvanhus), PL (Peromyscus leucopus), (Bfarina brevicauda), SC
(Sorex cinereus), ZH (Zapus hudsonlus), RN (Rattus norvegicus), SA (Scalopus aquaticus), TS (Tamias striatus), CG
(Clethrionomys gapperl), and CC (Condylura crisfata).

Site

Trap
periods

Trap-
nights

MP

PL

BB

SC

ZH

RN

SA

TS

CG

cc

TOTAL

Broadmoor

Control

32

3332

416

1

76

7

6

0

0

0

0

0

507

Exptl.

32

3136

225

11

61

5

23

0

0

2

0

1

327

Plots

32

1824

212

19

48

7

8

1

0

0

0

0

295

Great Islar^d

Ctl(l)

43

4157

86

400

0

22

18

0

0

15

7

0

548

Cti(2)

31

2940

146

359

25

12

7

0

0

39

0

0

588

Ctl (3)

31

2989

12

358

13

10

0

0

0

71

0

0

464

Ctl (4)

30

2934

551

404

0

41

0

0

0

0

0

0

996

Ctl (5)

43

4193

603

349

11

61

21

0

0

3

0

0

1048

Exptl.

35

3381

219

111

15

74

18

0

0

1

0

0

438

Plots

35

2686

336

221

3

54

12

0

0

0

0

0

626

Nantucket

Control

35

5782

1364

255

83

0

3

2

2

0

0

0

1709

Exptl.

35

3621

420

270

75

2

0

3

0

0

0

0

770

Plots

35

1798

400

420

28

0

0

6

0

0

0

0

854

TOTAL

449

42773

4990

3178

438

295

116

12

2

131

7

1

9170

292

Analytical Approach

I defined an association as a staristi-
cal relationship between the numbers
of captures of a species at trap sta-
tions and a quantitative measure of
microhabitat structure. To determine
these relationships, I regressed the
total number of captures of a species
at each control grid trap station on
factor scores of each PC. The experi-
mental grid represented an area
where densities were continually
being reduced and vacant microhabi-
tats were more often available to
colonizing individuals.

To determine differences in mi-
crohabitat associations between con-
trol and experimental grids, I in-
cluded a dummy variable coding for
grid (control or exjjerimental) and
habitat variable x grid interaction
tern\s (Adler 1985).

Inferences

Most small mammals (8 of 11 fx)pu-
lations examined) demonstrated af-
finities for specific microhabitat types
on either control or experimental
grids (table 4). These affinities gener-
ally were consistent with other pub-
lished reports of habitat associations
of these species. For instance, P. leu-
copus generally were associated posi-
tively with woody microhabitats or
negatively associated with herba-
ceous microhabitats. M. pennsylvani-
cus generally showed the opposite
associations. Microhabitats selected
by small mammals, as determined
from capture data on experimental
removal grids, sometimes differed
from associations determined from
capture data on the adjacent control
grids (table 4). Differences in habitat
selection and association were attrib-

Table 4.— Habitat associations of small mammals In eastern Massachusetts
determined from regressions of numbers of captures at trap stations on
habitat variables derived from principal components analysis. The direc-
tion of Vne regression slope Is given by + or - ond tti« strength of the rela-
tionship Is given by the number of signs (1 , P<0.05; 2, P<0.01 ; 3, P<0.001).
Species designations are as In table 3.

Site and
Species

Description of habitat variable

Control
grid

Experimental

Broadmoor
BB
MP

ZH

Great Island
SC
BB
PL

MP

ZH

Nantucket
BB
PL

MP

Alt variables

Herbaceous density and height
Shrubby vegetation
Ground-level herbaceous density
Herbaceous density and height

All variables

Vertical woody vegetation density
Total vegetation cover, primarily
herbaceous
Habitat complexity
Habitat structure, reflecting
increasing herbaceousness
and decreasing woodiness
Ground-level woody vine density
All variables

Herbaceous species richness
Number of overstory trees
Vertical vegetation structure
Herb species richness

N$
+++

NS

NS
+

NS

NS

+

+

NS

NS

NS
+

+

NS

NS
NS

NS

NS
NS
+

NS

utable to opportunistic responses of
small mammals to between-grid dif-
ferences in microhabitat structure
and to differences in the level of in-
traspecific interactions brought about
through density reductions on the
experimental grids (Adler 1985).

Tennporol Patterns of Habitat Use
Study Purpose

I examined temporal patterns of mi-
crohabitat use by M. pennsylvanicus
at the three study sites and by P. leu-
copus on Great Island and Nantucket.

Analytical Approach

Monthly trapping periods were
grouped into winter (Dec.-Feb.),
spring (Mar.-May), summer (Jun.-
Aug.), and fall (Sep.-Nov.) seasons
each year. I divided trap stations on
control grids into favorable and unfa-
vorable nnicrohabitats each season
depending upon whether the total
number of captures in a season was
above (favorable) or below (unfavor-
able) the seasonal mean (Van Home
1982; Adler 1985). I then used a two-
group DFA, with favorable and unfa-
vorable trap stations defining the
two groups, to develop a discrimina-
tion index of habitat use (Rice et al.
1983; Adler 1985). This index was the
percentage of trap stations classified
correctly as either favorable or unfa-
vorable. High discrimination values
indicated a sharp distinction between
favorable and unfavorable micro-
habitats; low values indicated little
difference between favorable and un-
favorable areas.

To determine the importance of
intra- and interspecific population
densities on temporal patterns of
habitat discrimination by P. leucopus
and M. pennsylvanicus, I regressed
the seasonal discrimination values on
the mean seasonal densities of each
of the major small mammal species
present at each study site.

293

Inferences

In the case of M. pennsylvanicus, den-
sity and discrimination were nega-
tively related at Broadnnoor and
positively related on Great Island.
The unexpected positive relationship
on Great Island could be explained
by the distribution of captures over
the grid; 17 capture stations had less
than two captures during the entire
study and were in a sparsely vege-
tated area. As density increased, the
reniaining 32 trap stations became
increasingly utilized. The distinction
between favorable and unfavorable
microhabitats increasingly became a
distinction between unoccupied,
sparsely vegetated stations and occu-
pied, densely vegetated stations.

On Nantucket, discrimination fol-
lowed a pattern similar to density
but was not linearly related to the
latter. For P. leucopus on both Great
Island and Nantucket, habitat dis-
crimination was related negatively to
density (fig. 1), indicating that the
distinction between favorable and
unfavorable microhabitats decreased
with increasing density. Densities of
other species were not related to
temporal variation in habitat use
(Adler 1985).

Therefore, intraspecific competi-
tion appeared to be more important
than interspecific interactions in de-
termining microhabitat use by the
species I examined. As intraspecific
density increased, the range of mi-
crohabitat types utilized also in-
creased, as predicted by early theo-
ries of habitat selection (e.g.,
Svardson 1949).

Microhabitat Structure and
Dennography

Study Purpose

I examined the relationship between
demography of M, pennsylvanicus
and microhabitat structure from data
collected on the four trapping plots
at each study site (Adler 1987).

Analytical Approacti

I calculated density (log^^ number
per 100 trapnights), sex composition
(proportion males, arcsin square root
transformed), age structure (propor-
tion of adults captured during sam-
pling periods from April through
September, arcsin square root trans-
formed), and breeding intensity (pro-
p>ortion of adults in breeding condi-
tion captured in sampling periods
from April through September,
arcsin square root transformed) each
trapping period.

I also computed variability meas-
ures for each of these demographic

variables as squared distances from
plot means. I divided the estimates
for density variability on each plot by
the mean density of the respective
plot in order to adjust for population
size.

I regressed the estimates for each
of the eight demographic variables
separately on plot means for each
microhabitat variable derived from
PC A and the index of heterogeneity.
I then regressed the unstandardized
residuals from each of these regres-
sions separately on each habitat PC
and the heterogeneity index to search
for nonlinearities and missing vari-
ables (Framstad et al. 1985).

GREAT ISLAND

80-1

4 6

DENSITY
NANTUCKET

o
I—
<

100

90-

80-

o

00

^ 70-

60-1

10

DENSITY

Figure 1.— Relationships between seasonal habitat discrimination and population density In
Peromyscus leucopus at two study sites In eastern Massachusetts.

294

Inferences

Densities of M. pennsylvanicus and P.
leucopus were ordered linearly along
microhabitat gradients (Adler 1987),
consistent with patterns of mi-
crohabitat associations and selection
in these two species (table 5). In
general, M. pennsylvanicus densities
were higher on plots with more her-
baceous and grassy cover or less
woody cover. Nantucket was excep-
tional, however, with M. pennsylvani-
cus densities increasing along gradi-
ents of increasing woody growth and
shrub species richness. I captured
large numbers of this vole in dense
heath with little or no herbaceous
vegetation.

Peromyscus leucopus densities on
Great Island could not be related to
microhabitat structure, probably be-
cause of the generalist nature of this
mouse relative to the breadth of mi-
crohabitats sampled. Indeed, when
sampling areas included other mi-
crohabitats, density could be related
to overall microhabitat structure (see
below). P. leucopus densities on Nan-
tucket increased with increasing
shrub species richness (table 5). Den-
sities of both species were more vari-

able in poorer habitats. Microhabitat
structure was a poor predictor of
other aspects of demography such as
age and sex composition. However,
variability in demographic structure
often was greater in low-density
habitats. While some of the variabil-
ity in density and demography may
have been due to statistical depend-
ence on population size (i.e., greater
sampling error at small population
sizes), biological effects
(e.g.,response to environmental fluc-
tuations) also must have been impor-
tant. More favorable microhabitats
should have maintained a more
stable composition over time due to
greater intraspecific interactions,
whereas poorer microhabitats should
have contained a more unstable as-
semblage of predominantly transient
and subordinate individuals due to
spillover during periods of high den-
sity (Adler 1987). In contrast to the
importance of microhabitat gradi-
ents, the quantitative measure of mi-
crohabitat heterogeneity generally
was unrelated to demographic phe-
nomena. In only one case did mi-
crohabitat heterogeneity explain vari-
ation in demography better than any
structure-diversity variable.

Macrohabrtdt Structure and
Dennography

Study Purpose

I further examined the relationship
between demography of P. leucopus
and microhabitat structure across
macrohabitats. P. leucopus is a habitat
generalist which occurs in habitats
ranging from grassland to mature
deciduous and coniferous forests in
southeastern Massachusetts.

Analytical Approacti

For this purpose, data from the five
control grids on Great Island were
analyzed (Adler and Wilson 1987).
Monthly trapping data were ana-
lyzed with respect to 10 demo-
graphic variables. Grid means of
density (log^^ minimum number
known alive), adult male body mass,
and observed range length (ORL, the
maximum linear distance between
capture points of an individual,
Stickel 1954) were compared using
Tukey's multiple comparisons test.
Mean male and female ORLs were
compared on each grid using t-tests.

Contingency table analysis was
used to compare age structure (pro-
portion adult), adult survival (stan-
dardized 14-day rates), sex composi-
tion (proportion of mice tagged that
were males), adult residence rates
(proportions of adults captured in at
least two trapping periods), overwin-
ter residence (proportions of mice
present during Sep. and surviving to
the subsequent Apr.), the propor-
tions of adults that were reproduc-
tively active, and the proportions of
young mice (mice with some grey
pelage remaining) that were repro-
ductively active. These 10 variables
were examdned for intersex differ-
ences within a grid (except sex com-
position) and for intergrid differ-
ences.

To examine temporal dynamics of
demography, monthly trapping data
were grouped into early summer

r

ldb\e 5.— Relationships between density of Peromyscus leucopus (PL) and
Microtus pennsylvanicus (MP) and microhabitat variables (derived from
principal components analysis) at three study sites In eastern Massachu-
setts. Signs of correlation are as Indicated In table 4.

Site

Species

Habitat variable

Correlation

Broadmoor
Great Island

Nantucket

MP Decreasing vertical woody stem +

der^ities and shrub cover.
PL All gradients. NS
MP Increasing woody and herbaceous —

stem der^ties, cover and species

richness; decreasing grassiness

!ncreasir>g woody ground vine —

species richness and cover,

increasing total vegetation cover ++

decreasing overstory species richness.
PL Increasing shrub species richness, +
MP Increasing herbaceous growth;

decreasing woody growth.

Increasing plant species richness. +

Increasing shrub species richness. +

295

(Apr.-Jun.) and late summer (Jul.-
Sep.) seasons. The following demo-
graphic variables were estimated on
each grid during each season: density
(mean log^g minimum number
known alive), proportions of males
and of females that were adults, pro-
portion of males, mean adult male
body mass, proportions of adult
males and of adult females breeding,
and survival rates of adult males and
of adult females (weighted mean 14-
day rates). Variables expressed as
proportions were arcsin square root
transformed.

Many rodent population parame-
ters are known to covary (e.g., Schaf-
fer and Tamarin 1973). Accordingly,
a PCA of the eight variables was exe-
cuted in order to include covarying
parameters as single demographic
variables; four PCs with eigenvalues
greater than 1.0 were retained for
further analysis.

These PCs were correlated with

(1) density and adult survival,

(2) adult female breeding activity,

(3) adult male breeding activity, and

(4) the proportion of males. Variabil-
ity indices of each of these PCs were
calculated each season for each grid
as squared distances from grid
means (Adler and Wilson 1987). A
measure of overall demographic
variability was calculated for each
grid each season as squared dis-
tances of the factor scores from the
mean factor score, summed over the
four PCs.

Factor scores within each PC were
multiplied by the square root of that
PC's eigenvalue in order to account
for the unequal contributions to
overall variance of each PC (Adler
and Wilson 1987). This method al-
lowed variables with different scales
of measurement to be included to-
gether without further scaling or
weighting. Seasonal estimates of each
of the PCA-derived demographic
variables and their variability esti-
mates were regressed separately on
each of the PCA-derived microhabi-
tat variables and the index of hetero-
geneity.

inferences

Statistical tests which were signifi-
cant at P<0.05 are qualitatively sum-
marized in table 6. Grid means of the
first three demographic PCs revealed
three demographic groups. Grids 1
and 5 were located farthest from any
adjacent grid in three-dimensional
space, whereas grids 2, 3, and 4 were
clustered more tightly together with
respect to demographic structure
(table 7). Grid 1 was characterized by
low density and survival, a low pro-
portion of females, low breeding in-
tensity, and high demographic vari-
ability. Grid 5 was characterized by
low density and survival, a high pro-
portion of females, moderate breed-
ing intensity, and high demographic
variability. Grids 2, 3, and 4 were
characterized by high density and
survival, low to moderate proportion
of females, moderate to high breed-
ing intensity, and low demographic
variability. Two low-density groups
(represented by grid 1 and grid 5)
and one high-density group (repre-
sented by grids 1, 2, and 3) therefore

Variable Comment

Density

Adu't male body mass
Observed range length

Proporfion male

were evident. The low-density
groups were more variable in terms
of each of the demographic PCs and
in overall demographic structure. In
general, density, survival, and breed-
ing activity increased along gradients
of increasing woodiness or decreas-
ing herbaceousness, whereas demo-
graphic variability decreased along
these gradients (table 8).

SYNTHESIS

I found microhabitat structure to be a
potentially important force in organ-
izing small mammal fX)pulations,
particularly in relation to associa-
tions and densities. Small mammals
generally were associated with par- !
ticular microhabitats, as revealed by \
analysis of single trap stations. How- j
ever, associations often differed be- |
tween control and experimental \
grids. I suggest that the small mam- j
mals I studied selected specific mi- i
crohabitats and were opportunistic in }
their responses to habitat not occu-
pied by other individuals (as on the |

i
i

Adult breeding activity

Young breeding activity
Adult residence

Overwinter residence
Adult survival

Grids 1 and 5 had lower densities than grids 2.

3, and 4.

No differences.

Males had a greater ORL than females on grid

2.

Grids 2. 3. and 4 had a higher proportion of
adult males. Grid 3 had a higher proportion of
adult males than females.
Grids 1 and 5 had a lower proportion of males
breeding than grids 2, 3.and 4. Grid 1 had a
lower proportion of females breeding than did
the other grids. A higher proportion of females
was breeding on grids 1,3, and 4 than were
males.

No differences.

Grids 1 and 5 had lower residence rates of
adult males than grids 2, 3, and 4,
No differences.

Males on grids 1 and 5 had poorer survival
rates than on grids 2,3, and 4.

Table 6.— Summary of differences In demography of Peromyscus leucopus
on Greal Island, determined from monthly trapping data on five grids.

296

experimental grids). Since most small
mammals that I studied were mi-
crohabitat selectors, microhabitat
structure therefore was a crucial de-
terminant of local community com-
position. Furthermore, microhabitat
structure also should have affected
temporal variability of community
structure since populations in low-
density areas were more variable.

Affinities of each small mammal
species for particular microhabitats
resulted in density-habitat relation-
ships when averaged over a larger
sampling area (grids or plots). Thus,
small mammal densities generally
could be related to nnicrohabitat
structure. Survival and breeding ac-
tivity, which generally covary with
density, also could be related to mi-
crohabitat structure when sampling
areas spanned macrohabitat bounda-
ries. The importance of microhabitat
structure in affecting other demo-

graphic characteristics such as sex
composition and age structure was
not as pronounced. Gradients of n\i-
crohabitat structure can be envi-
sioned as comprising an environ-
mental suitability gradient, with the
endpoints being uninhabitable and
optimal (where individual fitness is
highest). Demographic characteris-
tics then vary along this gradient of
suitability and along other gradients.
The gradient of suitability is com-
posed of factors related not only to
habitat structure but also to food re-
sources and release from predation,
competition, and parasitism. Density
alone may not be a strong correlate
of suitability (Van Home 1983), but
density in concert with survival and
breeding activity should increase
along the gradient of suitability. By
contrast, demographic variability
should decrease along this gradient.
Several habitat types may represent

Table 7.— Distances between grid means of the first three principal compo-
nents derived from an analysis of demographic data of Peromyscus leu-
copus.

Grid

1
2
3
4

1.14

L38
0.46

1.36
0.45
0.22

1.09
1.00
1.26

Table 8.— Relationships between Peromyscus /et/copt/s demographic vari-
alDles arKl habitat variables derived from PCA on Great Island. Signs of re-
lationships are as Indicated in table 4.

Demographic variables Habitat variables

Correlation

Derisity and survival
Adult male breeding
activity

Variability of der^ity
and survival
Variability of male
breeding activity
Variability in sex
composition

Herbaceous ground-level vegetation
Herbaceous ground-level vegetation
Woody ground vine species richness
Herbaceous cover and species richness +

Herbaceous ground-level vegetation

Woody vegetation der>sity and richness —
Herbaceous cover and species richness -»-++
Canopy cover —

similar conditions of environmental
suitability, particularly for habitat
generalists such as Peromyscus leu-
copus. Therefore, it may be difficult to
relate demography to microhabitat
structure because similar demo-
graphic structure may be found in
different habitats (Adler and Wilson
1987).

Quantitative measures of habitat
heterogeneity generally were unre-
lated to demographic variables, in
contrast to the mass of theory pre-
dicting that heterogeneity promotes
population stability (e.g., den Boer
1968, Levins 1969, Smith 1972, Mayn-
ard Smith 1974, Steele 1974, Tanner
1975, Stenseth 1977, 1980, Lomnicki
1978, 1980, de Jong 1979, Hassell
1980). The contrast between my re-
sults and theoretical predictions may
be reconciled by introducing scale.
My measures of heterogeneity were
at the microhabitat level, whereas
many models have implied mac-
rohabitat heterogeneity so that or-
ganisms may disperse into a patch
and establish a resident population
(e.g.. Levins 1969). Increasing the
numb>er of such patches increases the
spatial heterogeneity of an area,
which then promotes population sta-
bility. I suggest that microhabitat
structure will affect density more
than it will other demographic char-
acteristics, whereas macrohabitat
structure and heterogeneity will be
more important in stabilizing popu-
lations and in influencing demo-
graphic structure (e.g., sex composi-
tion and age structure).

My conclusions concerning the
importance of habitat structure in
organizing small mammal popula-
tions and communities can be shown
schematically (fig. 2). According to
this scheme, microhabitat structure
primarily affects habitat selection,
density, and density variability (since
density generally is related inversely
to variability). Macrohabitat struc-
ture primarily affects population sta-
bility (stability being enhanced by
macrohabitat heterogeneity) and
demographic structure. Habitat se-

297

lection and demography then deter-
mine local community composition
and variability, respectively. While
habitat structure ultimately deter-
mines community composition, it
does so at the population or individ-
ual level. Therefore, I added no di-
rect links between habitat structure
and the community variables.

Additional links may be added;
factors such as random events, com-
petition, predation, parasitism, and
infection manifest their effects at
various levels. For instance, a preda-
tor may selectively feed on a particu-
lar species, thereby depressing its
density and affecting community
composition and structure. Competi-
tion between species also may affect
species densities in certain small
mammal communities. The structur-
ally simple habitats that I have stud-
ied generally contain an abundant
herbivore, an abundant granivore, a
common insectivore, and any of sev-
eral rarer omnivores. These poorly
diversified communities are quite
different from other systems such as
deserts or tropical forests where
communities are comprised of regu-
larly structured guilds containing
several species. Competition, which
apparently is important in structur-
ing communities in other areas (e.g..
Brown and Bowers 1984), should not
be very important. The opportunity
for competition between guilds
would be expected to be quite low.
The occurrence of several easily stud-
ied genera with interesting life-his-
tory traits (e.g., Microtus, Peromyscus,
and Tamias) made these sites ideal
for population-level studies, but be-
cause of poorly diversified guilds or
even guild singularity (only one spe-
cies per guild) at my study sites,
these same areas were far less suit-
able for community-level studies.

Habitats with which different spe-
cies of small mammals are associated
are well known, but the effects of
relevant scales of habitat structure
are only now becoming apparent. I
suggest that future studies shift from
repetitious descriptions of habitats

with which well-studied species as-
sociate to innovative experimental
approaches that test hypothesized
effects of habitat structure on popu-
lation and community organization
and that identify relevant scales of
such structure.

ACKNOWLEDGMENTS

Bruce Lund, Wesley N. Tiffney, Jr.,
and the Chace family granted per-
mission to trap. John W. Classer,
Thomas W. Schoener, and Robert H.
Tamarin read an earlier draft of the
manuscript. Mark L. Wilson shared
in much of the field work on Great
Island, and Marita Sheridan shared
in field work on Nantucket. I thank
Robert H. Tamarin for providing re-
search supplies and facilities at Bos-
ton University while I conducted the
field research. I also thank Thomas
W. Schoener for providing facilities
and resources at the University of
California (Davis), where I wrote the
manuscript. This research was sup-
ported by NSF grants DEB-8103483
to Robert H. Tamarin and BSR-
8700130 to G.H.A. and from Sigma Xi
and the Boston University chapter of
Sigma Xi to GHA. This study is a

contribution of the University of
Massachusetts Nantucket Field
Station.

LITERATURE CITED

Adler, Gregory H. 1985. Habitat se-
lection and species interactions: an
experimental analysis with small
mammal populations. Oikos
45:380-390.

Adler, Gregory H. 1987. Influence of
habitat structure on demography
of two rodent species in eastern
Massachusetts. Canadian Journal
of Zoology 65:903-912.

Adler, Gregory H., and Mark L.
Wilson. 1987. Demography of a
habitat generalist, the white-
footed mouse, in a heterogeneous
environment. Ecology 68:1785-
1796.

August, Peter V. 1983. The role of
habitat complexity and heteroge-
neity in structuring tropical mam-
mal communities. Ecology
66:1495-1507.

Brown, James H., and Michael A.
Bowers. 1984. Patterns and proc-
esses in three guilds of terrestrial
vertebrates, p. 282-296. In Ecologi-
cal communities: conceptual issues

Density

Density
Variability

Macrohabitat
Structure

Microhabitat
Structure

Habitat
Selection

Population
Performance

z

Demographic
Structure

Commimity
Composition

Community
Structure

Temporal Variability
of Commvmities

Figure 2.— Conceptual scheme showing the effects of habitat structure on population and
community processes.

298

and the evidence. Donald R.
Strong, Jr., Daniel Simberloff,
Lawrence G. Abele, and Anne B.
Thistle, editors. Princeton Univer-
sity Press, Princeton, NJ.

de Jong, G. 1979. The influence of the
distribution of juveniles over
patches of food on the dynamics
of a population. Netherlands Jour-
nal of Zoology 29:33-51.

den Boer, P. J. 1968. Spreading of risk
and stabilization of animal num-
bers and the heterogeneity of the
environment: the problem of the
persistence of sparse populations,
p. 77-97. In Dynamics of Popula-
tions. Proceedings of the Ad-
vanced Study Institute on Dynam-
ics and Numbers of Populations,
Oosterbeek, 7-18 September 1970.
P.J. den Boer and G. R. Gradwell,
editors. Centre for Agricultural
Publishing and Documentation
(PUDOC), Wageningen, The Neth-
erlands.

Framstad, E., S. Engen, and N.
Stenseth. 1985. Regression analy-
sis, residual analysis, and missing
variables in regression models.
Oikos 44:319-323.

Hassell, Michael P. 1980. Some con-
sequences of habitat heterogeneity
for population dynamics. Oikos
35:150-160.

Levin, Simon A. 1976. Population
dynamic models in heterogeneous
environments. Annual Review of
Ecology and Systematics 7:287-
310.

Levins, Richard. 1969. Some demo-
graphic and genetic consequences
of environmental heterogeneity for
biological control. Bulletin of the
Entomological Society of America
15:237-240.

Lomnicki, Adam. 1978. Individual
differences between animals and
the natural regulation of their
numbers. Journal of Animal Ecol-
ogy 47:461-475.

Lomnicki, Adam. 1980. Regulation of
population density due to individ-
ual differences and patchy envi-
ronment. Oikos 35:185-193.

Maynard Smith, John. 1974. Models
in ecology. Cambridge University
Press, Cambridge, United King-
dom.

Schaffer, William M., and Robert H.
Tamarin. 1973. Changing repro-
ductive rates and population
cycles in lemmings and voles.
EvoluHon 27:111-124.

Smith, Frank E. 1972. Spatial hetero-
geneity, stability, and diversity in
ecosystems. Transactions of the
Connecticut Academy of Arts and
Sciences 44:309-335.

Steele, John H. 1974. Spatial hetero-
geneity and population stability.
Nature 248:83.

Stenseth, N. C. 1977. On the impor-
tance of spatio-temporal heteroge-
neity for the population dynamics
of rodents: towards a theoretical
foundation of rodent control.
Oikos 29:545-552.

Stenseth, N. C. 1980. Spatial hetero-
geneity and population stability:
some evolutionary consequences.
Oikos 35:165-184.

Stickel, Lucille F. 1954. A comparison
of certain methods of measuring
ranges of small mammals. Journal
of Mammalogy 35:1-15.

Svardson, Gunnar. 1949. Competi-
tion and habitat selection in birds.
Oikos 1:157-174.

Tanner, James T. 1975. The stability
and the intrinsic growth rates of
prey and predator populations.
Ecology 56:855-867.

Van Horne, Beatrice. 1982. Niches of
adult and juvenile deer mice (Pero-
myscus maniculatus) in serai stages
of coniferous forest. Ecology
63:992-1003.

Van Home, Beatrice. 1983. Density as
a misleading indicator of habitat
quality. Journal of Wildlife Man-
agement 47:893-901.

Wiens, John A. 1976. Population re-
sponses to patchy environments.
Annual Review of Ecology and
Systematics 7:81-120.

Wiens, John A, John F. Addicott, Ted
J. Case, and Jared Diamond. 1986.
Overview: the importance of spa-
tial and temporal scale in ecology
investigations, p. 145-153. In Com-
munity ecology. Jared Diamond
and Ted. J. Case, editors. Harper
and Row Publishers, New York,
NY.

299

Microhabitat as a Template
tor the Organization of a
Desert Rodent Community^

Michael A. Bowers^ and Christine A.
Flanagan^

Abstract.— We used 20 0.25-ha fenced plots to
experimentally study microhabitat use by 1 1 desert
rodent species in southeastern Arizona. Removal of
the largest granivore, Dipodomys spectabilis.
produced the most pervasive shifts in the use of
microhobitats while adding food or removing ants
produced few responses. These results support the
idea that this community is organized around
competitive interactions involving aggression,
preemption, and relegation.

It is generally believed that species
have different fitnesses in different
habitats, that most communities are
comprised of sufficient habitat vari-
ation over which fitness differentials
can be expressed, and that species
select habitats that maximize their
fitness (e.g.. Levins 1962, Schoener
1971). The manner and degree to
which species respond to the habitat
template involves elements of selec-
tion in its purest form (i.e., choice),
relegation, and correlation.

At the community level rarely do
species occupy habitats in an ideal or
cost-free fashion. By occupying space
or using resources in a habitat spe-
cific manner organisms alter habitat
suitability and thereby change the
basis over which habitats are selected
(Fretwell and Lucas 1970). Species
that use limited resources in an effi-
cient manner or are behaviorally
dominant can monof)olize the choic-
est habitats and relegate, directly or
indirectly, subordinate or competi-
tively inferior species to secondary

'Paper presented at symposium. Man-
agement of Amphibians. Reptiles, and
Small Mammals in Northi America. (Flag-
staff. AZ, July 19-21. 1988.)

'Michiael A. Bowers is Assistant Professor
in ttie Department of Environmental Sci-
ences and Researchi Coordinator at the
Blandy Experimental Farm. University of Vir-
ginia. Clark Hall. Charlottesville, VA 22903.

^Christine A. Flanagan is Assistant Cura-
tor of the Orland E. White Arboretum. Uni-
versity of Virginia. P.O. Box 175. Boyce. VA
22620.

habitats (Col well and Fuentes 1973,
Bowers et al. 1987). If the capture
success rates of predatory species
varies among habitats this can also
affect the absolute and relative fit-
ness of prey species and their distri-
bution among habitats (Kotler 1984,
Bowers 1988).

Marked patterns of habitat occu-
pancy and segregation are often cited
as evidence that ecological communi-
ties are structured. The general pat-
tern is that some (if not most) species
in a community utilize habitats dif-
ferently from random and differently
than if each species occurred by it-
self. Observational and manipulative
experiments have shown that dy-
namical properties of populations
(including patterns of growth, demo-
graphics, and interaction) often be-
come expressed as spatial phenom-
ena, thereby establishing a connec-
tion between habitat occupancy and
population dynamics (see Connor
and Bowers 1987).

Many communities are comprised
of an array of microhabitats which
represent discrete, exploitable re-
sources which occur with suffident
variability so as to be parti tionable
among spedes. The availability and
distribution of microhabitats have
been shown to limit the growth and
density of many populations and,
thereby provide an ecologically rele-
vant and readily identifiable context
over which species interactions and
population growth can be studied
(Price 1978, Rosenzweig 1981).

Desert rodents have long pro-
vided ecologists with a model system
for examining the role of microhabi-
tat in structuring communities. The
basic pattern throughout the major
North American deserts is that lo-
cally co-occurring species character-
istically forage in microhabitats that
are structurally distinctive with re-
spect to perennial vegetation and soil
type (Rosenzweig and Winakur 1969,
Price 1978; for reviews see Brown et
al. 1979, Munger et al. 1983, Price
and Brown 1983).

Three mechanisms, alone or in
combination, apparently account for
the general pattern. First, because of
differences in body size, mode of lo-
comotion and behavior, rodents dif-
fer in their abilities to exploit particu-
lar distributions of food (i.e., seed)
resources that are created by struc-
tural features of the microhabitat
(Bowers 1982, Harris 1984, Price
1983, Reichman 1981). Second, ro-
dents may differ in their ability to
escape visually oriented predators so
that the most susceptible rodents are
limited to the safest microhabitats
(i.e., under vegetative cover) while
more vagile rodents show more un-
restricted use of alternate microsites
(Kotler 1984). Third, the ability of
some species to aggressively defend
areas from other rodents may be
high in some habitats and low in oth-
ers resulting in habitat dependent
segregation involving domination/
relegarion (Hutto 1978, Frye 1983,
Bowers et al. 1987).

300

Desert rodent populations are re-
markable in their ability to respond
to short-term changes in the abun-
dance and distribution of food re-
sources; primarily seeds. Some of the
more marked responses involve
changes in use of microhabitats. For
example, enriching microhabitats
with supplemental seeds increases
the use of these by desert rodents
(Harris 1984, Kotler 1984, Price and
Waser 1985). Such shifts are particu-
larly noteworthy for microhabitats
where the risk of being preyed upon
is high, and suggests that both ener-
getic profits and predatory risk play
a role in determining which mi-
crosites are used (Hay and Fuller
1981, Price and Waser 1985, Bowers
1988). Food availability also can
change the manner in which some
rodent species interact: from com-
1 petitive exploitative interactions un-
I der low levels of food to aggressive
interference interactions under high
levels of food (e.g., Congdon 1974).
i In complex communities mi-
' crohabitat use originates with prefer-
ences of individual species for certain
microhabitats, but these basic re-
|| sponses may become altered, directly

or indirectly, by interactions with
I other species. Moreover, at the com-
munity level it is not clear how
changes in the resource base are
manifest in patterns of spatial usage.
Some important questions are: Does
interspecific competition become
more or less important with increas-
ing food availability? Does the mode
of competition change? How does
food availability change the relative
roles of preference and relegation in
determining habitat occupancy?
Thus, detailing the interplay between
population and community-level re-
sponses to changes in resource
availability should reveal much
about the processes influencing mi-
crohabitat use and, thereby, the fac-
tors responsible for the organization
of these communities.

In this paper we describe patterns
of microhabitat use of 11 Chihua-
huan Desert rodents over a span of

more than eight years. We exp)eri-
mentally manipulated both species
composition and food supply and
measured resulting shifts in mi-
crohabitat use. By detailing shifts in
microhabitat use in response to our
manipulations we were able to iden-
tify the most important interactions
among species, estimate their relative
strengths, and say something about
the mode of interaction promoting
the shifts.

Our results suggest that the or-
ganization of this community re-
volves more around differences in
the ability of species to occupy and
defend certain key microhabitats
than changes in food availability.

Study Site and Methods

The present paper details changes in
microhabitat use in response to long-
term experimental manipulation of
rodent composition and food supply.
Our study site was located at an ele-
vation of 1330 m in a relatively
homogeneous desert shrub habitat
on the Cave Creek Bajada 6.5 km east
and 2 km north of Portal, in Arizona,
USA. Manipulations were performed
in twenty 0.25-ha plots. Each plot
was fenced with 0.64-cm mesh hard-
ware cloth, extending 0.7-m above
and buried 0.2-m below ground. In
addition to an unmanipulated fenced
control (see below), the remaining
treatments consisted of two general
classes: treatments where one or
more rodent species were removed,
potentially changing both food
availability and the ix)tential for di-
rect behavioral interactions; and food
alteration treatments where supple-
mental millet seeds were added at a
rate of 96 kg per year or seed-eating
ants were removed. Experimental
treatments were assigned to plots at
random.

Fourteen rodent species of which
11 were commonly captured, inhab-
ited the study site, all except those
mentioned above had equal access to
all plots (fig. 1). Because of problems

in consistently identifying the two
Onychomys species (as either O. tor-
ridus or O. leucogaster) we group
these together under the designation,
Onychomys spp.

Sixteen equally-spaced gates in
each plot allowed the selective exclu-
sion of rodent species above a thresh-
old body size while allowing all
other species access. Access gates
varied in size among the treatments.
Large gates (3.7 x 5.7-cm) allowed all
rodent species free access to control
(2 plots), ant removal (4 plots), and
the seed addition plots (8 plots; see
below); medium-sized gates (2.6 x
3.0-cm) were used to exclude only
the largest granivore, Dipodomys
spectabilis (2 plots); and small gates
(1.9 x 1.9-cm) were used to exclude
all Dipodomys species (4 plots). The
seed addition treatments included
six plots where supplemental seeds
were applied in 12 monthly applica-
tions (hereafter referred to as "con-
stant seed additions"); two plots re-
ceived the total allotment of seeds in
three applications during the fall
(September-November; referred to as
"pulsed seed additions"). Seeds were
uniformly scattered by hand over
each plot.

It was estimated from productiv-
ity measurements at the site that the
addition of 96 kg of seeds per year
should have approximately doubled
the total biomass of seeds produced
annually (our estimate of seed pro-
duction was ca. 400 kg/ha/yr). The
constant seed additions included two
plots where whole millet (Panicum
miliaceam) was added (mean seed
mass = 6 mg); two plots where
cracked millet was added (mean
mass = 1 mg); and two plots where
an equal mixture of whole and
cracked millet was added. The
pulsed seed treatment was designed
to represent a doubling of the seed
production of summer annual plants,
a particularly important food source
for the rodents in this community
(Davidson et al. 1985). Brown and
Munger (1985) found no differences
in responses of rodents to addition of

301

seeds of different size, so the four
constant seed addition treatments
will be lumped together here (6
plots).

Rodents were censused monthly
during the week of the new moon
(moonlight has been shown to effect
the microhabitats used by desert ro-
dents; Bowers 1988) using live traps
placed in each plot in 7 x 7 grids with
6.5-m between trap stations. Traps
were baited with millet and opened
for one night per month with plot
gates closed so that only plot resi-
dents would be captured. For more
details concerning the experimental
design, see Bowers et al. 1987, Brown
and Munger (1985), and Brown et al.
(1986).

Following the lead of many previ-
ous studies on desert rodent commu-
nities we used the percent cover of
perennial plants to characterize the
microhabitat at each of the 980 trap
stations. Percent cover within a 2-m

radius of each trap station was meas-
ured by ocular estimation using ref-
erence disks of known percent cover-
age. Cover was measured in 1978,
1981, and 1983. There was no signifi-
cant changes in perennial cover over
this five year period (Mann-Whitney
U-test; P > 0.05), so we used data
from 1983 to characterize microhabi-
tats. Table 1 summarizes vegetation
cover data over the entire study site.

Fence installation was completed
in June, 1977; premanipulative trap-
ping was conducted from July-Sep-
tember, 1977; and the manipulations
were initiated in October, 1977. We
restrict our analyses to include post-
manipulation data compiled from
October 1977 to December 1984 and
to only those 20 plots to which ro-
dents had access.

Analyses were designed to answer
two questions: first, what are the pat-
terns of species associations occur-
ring at the community level; and sec-

ond, what role does microhabitat
play in the distribution of individual
species. In this study patterns of as-
sociation (including the association
of rodent species with each other and
with structural microhabitats) are
analyzed at the level of individual
trap stations (980 total). Hence, we

^able l,— Mean percent cover of
the seven most common perennial
plant species over the study site
(standard deviations In parenthe-
ses).

Species

% cover

Acacia constricfa 1.5(4.4)

Ephedra forreyana 2.7 (3.6)

Florensia cemva 2 7 (4.7)

Gufierrezla lucida 3.3(3.9)

Lycium andersonii 0.4 (2,2)

Mimosa biuncifera 0.2 (1 .4)

Prosopis Juliflora 0.6(4.5)

i^otal cover (all species) 127 (3.7) ^

Re i thro-

Ueotoma

Olnedomyt Dipodomft Dipodomys Perognathus PtrognalhuM PeromyMcus Peiomyscut donto»,yt

' penicillatus mtnlculmtut •ramleua megalotis albigula

Mpaclablllt mmrrlami ordU

120g

Control 184 529 51

(2 plots)

Ant removal 539 787 248

(4 plots)

Seed-pulse 278 390 116

(2 plots)

Seed-constant 821 1005 217
(6 plots)

Dipodomys- 0 0 0

removal

(4 plots)

D. spectabills- 0 465 50

removal

(2 plots)

^ ^ ^ ^

45g 52g 17g

25

33

27

42

49

7g

25

82

22

121

243

13

8

9

10
74

21

Figure 1 .—Rodent species on study site (Including their body sizes) along with their capture
frequencies In each of ttie experimental treatments. Included are the number of plots in
each treatment.

10

22

75

65

42

24g 25g 11g 156g

24
48
21
32
175

33

60

94

56

145

61

66

Onyehomyt
tpp.

29g

49

106

47

132

104

53

302

were interested in measuring re-
sponses of rodents to microhabitat
variation occurring at a scale of a me-
ter or two. However, we acknowl-
edge that habitats may also be se-
lected at larger spatial scales (Morris
1987). For example, rodents may also
select areas on the basis of mi-
crohabitat composites (e.g., at the
level of the home range) which might
be best examined by considering
structural microhabitats over trap
station aggregates. However, there is
reason to believe that even if selec-
tion does occur at these larger scales
it is still oriented towards excluding
or including certain key microhabi-
tats. Hence, we were confident our
analyses would detect patterns at
both scales.

Indices of species association were
calculated by using the frequency
that species were captured at the
same trap station using trap data for
the eight year period. This involved
several steps: (i) tabulating the pro-
portion of trap stations where each
species was captured over the eight
year study; (ii) tallying the number of
trap stations where each pair of spe-
cies co-occurred; and (iii) comparing
the observed frequency of co-cap-
tures to that expected if species cap-
tures were distributed independently
and randomly among trap stations.
The expected frequency of species
co-capture was calculated by multi-
plying together the proportion of sta-
tions capturing species individually
to generate a probability of joint oc-
currence. A modified chi-square sta-
tistic, including the sign of associa-
tion, was then used as an index of
association: i.e., a measure of the dif-
ference between the observed and
expected values. The null hypothesis
was that there would be an equal
number of positive and negative as-
sociations with less than 5% of the
association values being statistically
significant at a P = 0.05.

The analysis described above can
also be used to examine the associa-
tion of all species in the community
at individual trap stations. Specifi-

cally, instead of asking how fre-
quently species pairs associate we
can use the maximum likelihood esti-
mation technique to estimate how
many trap stations should have cap-
tured 0, 1, 2, . . n species (where n is
the number of species in the commu-
nity) over the eight year period. As
in the above analysis, this uses the
proportion of stations capturing each
species, multiplies these together in
all possible combinations that might
produce co-captures of from 0 to n
species, and sums these probabilities
for each number of possible co-cap-
tures to give an expected distribution
over the population of trap stations.
The null expectation here is that spe-
cies captures are independently and
randomly distributed among trap
stations.

Analyses were also performed to
examine the individualistic responses
of species to variation in microhabi-
tat and, particularly, how these
change when manipulations are ap-
plied at the level of the entire com-
munity. We used percent cover by
perennial plants at trap stations as a
general descriptor of microhabitat
type. Our goal was not to use a series
of variables to explain the largest
amount of variation in nnicrohabitats
where species were captured but
rather we were interested in identify-
ing a major resource axis over which
both species distributions and com-
munity-level responses could be ana-
lyzed. Past work justified using cover
as such a variable (Brown et al. 1979,
Munger et al. 1983, Price and Brown
1983). Our scheme of categorizing
microhabitats was simple: trap sta-
tions were grouped into those with
greater-than-median and those with
less-than-median cover. This was
performed separately for stations in
each of the six treatments. Hence,
each microhabitat category was rep-
resented by an equal number of trap
stations in each treatment type. The
null hypothesis for analyzing the trap
data was if rodents use microhabitats
randomly, and without regard to
vegetative cover, they should be

trapped in equal frequencies at sta-
tions in the two microhabitat catego-
ries. Avoidance or preference for mi-
crohabitats would be indicated by a
disproportionate number of captures
in one or the other category.

We were also interested in exam-
ining (1) the microhabitat affinities of
species in the different treatments,
and (2) shifts in types of microhabi-
tats used by the same species over
the different seasons of the year and
over the six experimental treatments.
In the first case we used the Fisher
Exact Probability procedure in a two-
tailed test of the null hypothesis that
captures in the two microhabitats did
not differ from a 1:1 ratio (Siegel
1956); in the second we subjected the
proportion of species' captures in the
two microhabitats to a 2-way
ANOVA where season and treatment
represented treatment factors.

Results

Results are based on 8,019 captures
of the 11 most common rodent spe-
cies. Figure 1 lists the frequency of
capture for each species in the six
treatments summed over the eight
year study period.

Community-Wide Patterns of
Microtiabitot Use

What are the patterns of species asso-
ciation at the level of the entire com-
munity? In answering this we consid-
ered the frequency that species were
captured at the same trap station. We
performed two tests. We first calcu-
lated species associations for all pos-
sible pairings of the 11 species occur-
ring in plots with intact rodent as-
semblages (i.e., those 14 plots with
large gates) resulting in a total of 45
values of species association. Plotting
all association values show that most
species in this community are cap-
tured at the same station much less
frequently than predicted by chance
(fig. 2; the null hypothesis is that

303

there would be an equal number of
positive and negative associations
and that only 5% of these would be
statistically significant at P < 0.05).
The deviation from what is expected
is particularly striking considering
that 27 of the association values ex-
ceeded the cutoff value for signifi-
cance (3.84 for p <0.05 and d.f.=l)
and all of these were in the direction
of negative species associations; there
was not a single significant positive
association. This suggests a high
level of organization revolves around
the spatial segregation of species.

Among those factors that could be
responsible for this marked segrega-
tion are unique habitat preferences of
species. These could work alone or in
conjunction with habitat segregation
that is mediated through interactions
with other rodent species. The design
of our experiment allows a further
examination of the role of species in-
teractions in producing the pattern.
Specifically, our experiment includes
treatments with an intact rodent as-
semblage (14 plots; 686 stations) as
well as treatments where either D,
spectabilis (2 plots; 98 stations) or all
Dipodomys (4 plots; 196 stations)
were selectively removed and ex-
cluded. Because previous studies
have shown Dipodomys (and
especially D. spectabilis) to be behav-
iorally dominant over many of the
species they co-occur with (Blaustein
and Riser 1974, Frye 1983, Bowers et
al. 1987) there is reason to think that
by their removal the patterns of asso-
ciation of the remaining species may
change. To evaluate this possibility
we restricted the analyses to include
just those eight non-Dipodomys spe-
cies that occurred in all three treat-
ments (number of pairwise associa-
tion values for this group = 21). The
degree to which these species were
associated with each other at trap
stations in each of the three treat-
ments was calculated as before, and
then compared across the three treat-
ments (fig. 3). The results show that
removing either all Dipodomys or just
D. spectabilis significantly alters the

degree to which the remaining spe-
cies are spatially segregated (X^ =
17.33, df = 2; P < 0.000). While the
trend is clearly towards more posi-
tive and fewer negative associations
when competitors are removed, most
of the species are still negatively as-
sociated with each other.

The previous analysis can be ex-
tended from the two-species case to
one considering the association of all
11 species. Specifically, instead of
asking how frequently species pairs
associate we can use the maximum

likelihood estimation technique to
estimate how many trap stations
should have captured 0, 1, 2 ... 11
species over the eight year period.
Comparing the actual number of spe-
cies captured per station with that
expected (fig. 4) shows that the ob-
served distribution is shifted to the
left of that expected (significantly dif-
ferent at P < 0.05 using Kolmogorov-
Smimov one sample test), that there
are significant differences in the
mode of species co-captured per sta-
tion (expected=4; observed=3), and

18 T

Chi Square

Figure 2.— Estimates of species associations for plots witti intact rodent assemblages (i.e.,
thiose with large gates). Association values represent modified ctii-squares (witti ttie sign of
association) and wt^ere calculated according to wtiettier species were captured at ttie
sanr»e trap station more or less frequently tt>an expected by chiance. See text for rrjore de-
tail.

Figure 3.— Histogram of ttie number of positive and negative species associations for non-
Dipodomys species broiten into thiree treatrT>ent categories: (i) treatments witti intact rodent
assemblages; (ii) D. spectabilis renrwval plots; and (iii) Dipodomys removal plots.

304

that there are large differences in the
proportion of stations capturing two
species (ca. 5% for the expected com-
pared to 23% for the observed). The
main result is that trap stations cap-
tured fewer species than expected if
species captures were random, which
further evidence that species in this
community are spatially segregated.

Use of Space by Individual
Species: The Role of Cover

In this section we are interested in
the individualistic responses of spe-
cies to microhabitat variation and,
particularly, how these change when
manipulations are applied at the
level of the entire community.

Observed
Expected

# species per Station

Figure 4.— Histogram of expected and observed number of species captured at individuai
trap stations.

r

Table 2,— Results of analyses testing for <l) microhabitat associations of
species In control plots and (II) for shifts In microhabltats between control
and experimental plots. Microhabitat associations of species In the unma-
nipulated community are indicated under the "control" treatment: "c" If
they were trapped significantly more often in grecrter-than-median cover;
and "o" If more often In lesser-than-median cover. Significant shifts In mi-
crotxibitat use relative to that on "controls" are Indicated by a "+" if the
shift was towards high cover and It towards low cover (more open)
sites. "R" Is used to Indicate which species were removed from treatments;

Indicates the level of statistical significance C for P <0.05; " for P
<0.01).

Treatments

Seed

Seed

Ant

D.s.

Species

Control pulsed constant removal

removal

removal

D, specfabilis

o

R

R

D. merriami

c

R

D. ordii

R

P. pencillafus

•

•

+•

P. flavus

o

R. megalofis

c

+•

P. maniculafus

+•

P. eremicus

c

+*

N. albigula

c

•»

O. spp.

There was marked variability both
within and between species in the
usage of microhabitats (table 2 and
figs. 5 and 6). On control plots Pero-
myscus eremicus, Neotorm albigula,
Reithrodontomys megalotis, and Dipod-
omys merriami (in all treatments but
the D. spectabilis removals) all
showed positive associations for trap
stations with greater-than-median
cover.

Treatment

CO
3

0
CC

Figure 5.— Distribution of captures in
greater-ttian and less-than rrtedian cover
for the five heteromyid species listed ac-
cording to treatnr»ent and season. Capture
data is graphed relative to what the null
hypothesis predicts (i.e., an equal number
of captures in both microhabitat types; the
zero line). Preference for higher-than -me-
dian sites is represented by positive values;
less-than -nr»edian cover by negative val-
ues. Bars within treatment categories indi-
cate season: from left to right Spring
(March-May), Summer (June-August), Fall
(September-November), Winter (Decem-
ber-February). Treatment designation Is as
follows: "-DS", Dipodomys spectabilis re-
rTX>val; "C", control; "SC", corwtant seed
addition; "-A", ant removal; "-D", Dipod-
omys rerrxjval; "SP", pulsed seed additions.

305

i

Those species associated with
more open microhabitats included
the large kangaroo rat, Dipodomys
spedabilis, and the smallest species,
Perogmthus flavus. The remaining
species used the two microhabitats
more indiscriminantly with the ex-
ception that Peromyscus maniculatus
was captured more frequently in
high-cover microsites in the D. specta-
bilis removal treatment.

Figures 5 and 6 and table 2 show
our experiments were of the kind
and were of sufficient intensity to
promote community-wide changes in
the use of microhabitats by all spe-
cies; only the Onychomys showed sig-
nificant seasonal shifts in microhabi-
tat use (captured more frequently in
higher-cover areas during the fall
than in the other seasons). Using the
control treatment as a reference point
showed that the majority of species
shifted their use of microhabitats on
plots where D. spectabilis was experi-
mentally removed. These shifts, in-
volving eight of the nine species
present, included an increase in the
use of microsites with less-than-me-
dian cover by D. merriami, P. pencilla-
tus, P. flavus, and N. albigula, and an
increase in the use of high-cover sites
by P. maniculatus, P. eremicus, R.
megalotis, and D, ordii.

The remaining manipulations reg-
istered fewer and less dramatic
shifts: i.e., increased use of open mi-
crohabitats by P. pencillatus and P,
maniculatus on constant seed addi-
tion plots; and shifts towards higher-
cover microsites by R. megalotis and
P. pencillatus in ant removal and Di-
podomys removal treatments, respec-
tively.

The role of microhabitat in the or-
ganization of this community can be
further evaluated by comparing the
distribution of trap captures for all
species with what is available at trap
stations (fig. 7). The objective was to
determine whether certain types of
microhabitats are used by the rodent
community more frequently than
others. This analysis shows that the
distribution of captures in control, D.

spectabilis removal, and Dipodomys
removal plots all differ significantly
from that expected if the use of mi-
crohabitats was random with respect
to vegetative cover (Kolmogorov-
Smimov two sample test; P < 0.05).
However, there are characteristic
ways these differ from expected. On
control plots there were fewer than
expected rodent captures in traps
having < 5% cover; on D. spectabilis
removal plots there were a greater-
than-expected number of captures
for this same cover category; and on
Dipodomys removal plots most ro-
dents were captured at trap stations
with > 10 % cover.

Discussion

Out results identify species interac-
tions as the principal factor produc-
ing structure in this community. It is
significant that, by adding supple-
mental seeds or removing ants, we
were able to change microhabitats
used by only a few of the species but
removing a large, potentially domi-
nant competitor produced many
shifts. This suggests that the primary
mode of interaction, as it effects the
patterns of microhabitat use in this
community, involves the direct re-
sponses of rodent species to each
other rather than interactions medi-
ated through the exploitation of food
resources, or the individualistic re-
sponses of rodents to particular mi-
crohabitat types.

The results point to the impor-
tance of one dominant species, D.
spectabilis, whose presence in the
community plays a disproportionate
role in determining which microhabi-
tats are utilized by the other species,
and thus the organization of the com-
munity as a whole. Whenever it is
present, regardless of how much
food is available, it appears to rele-
gate the majority of other rodent spe-
cies to higher-than-median cover
habitats, thereby reducing the den-
sity of potential competitors in the
open habitats it prefers. A notable

exception is Perogmthus flavus which
was captured in open sites along
with D. spectabilis. Because of its
small size (ca. 7 g) and low popula-
tion density, P. flavus may have only
a negligible impact on the food re-
sources that can be harvested by D,
spectablilis and, therefore, may not
compete directly with or be subjected
to its aggressive behavior. The im-
portance of such size-ratio thresholds
in allowing species to coexist has
been discussed (Bowers and Brown
1982). Defending open areas from
other rodents may be a mechanism
by which D. spectabilis is able to
preempt food resources for its exclu-
sive use. Supporting evidence for this
comes from other research at our
study site where it was found that

Treatment

CO
=3

tr

Figure 6.— Distribution of captures in thie two
microtiabitat categories for the six Cricetid
rodents listed by treatment and season.
See legend to figure 5 for nriore details.

306

experimental seeds placed in open
microhabitats remained largely un-
harvested when D. spectabilis was
present but quickly disappeared in
plots where it was removed (see
Bowers et al. 1987).

Our results also infer something
about the mechanism by which D.

spectabilis affects the use of space by
other rodent species in the commu-
nity. Competition can be mediated
through two processes: (i) exploita-
tive interactions where species inter-
act through a shared resource base;
or (ii) contest interactions involving
aggressive dominance and relegation

to suboptimal areas and resources.
For exploitation alone to account for
the patterns of microhabitat use, D,
spectabilis, through its foraging,
would have to significantly alter the
distribution of food (seed) resources
among the microhabitats in ways
that are ecologically significant for
the other species. This is unlikely for
several reasons. First, many of the
seeds utilized by the smaller species
appear to be too small to be economi-
cally harvestable by D. spectabilis (see
Bowers et al. 1987). Second, many of
the species showing significant mi-
crohabitat shifts were non-granivores
(i.e., Neotoma), and hence, should be
relatively insensitive to changes in
the resource base attributable to the
foraging of D. spectabilis. Third, add-
ing seeds should have made food
more available to all species and re-
duced the degree to which D. specta-
bilis was able to alter the distribution
of food resources, so that shifts by
the other species would have been
expected in response to this treat-
ment. Moreover, significant changes
in the distribution of food resources
were more likely to have been caused
by D. merriami that occurs at higher
densities than D. spectabilis. Our re-
sults show that adding supplemental
seeds or removing D, merriami pro-
duced fewer shifts than removing
just D. spectabilis.

As an alternative to exploitation,
competitors of large body size may
directly restrict the foraging activities
of smaller species through interfer-
ence. Under an interference mode of
competition adding seeds may not
alter the intensity or outcome of the
interaction. Because most significant
shifts in microhabitat use occurred in
the D. spectabilis removal treatment —
coupled with the fact that adding
seeds had litUe effect on the patterns
of microhabitat — leads us to the con-
clusion that aggressive interference
by D. spectabilis is the mechanism
most consistent with our results.

Our study also indicates that the
majority of shifts in microhabitat use
originate with the D, spectabilis-D.

UJ
>

<

-I
UJ

cr

50-1

25 -

CONTROL

-1
I

>-

y 50

UJ

o

UJ

cr

U-

25 -

D. spectabilis
REMOVAL

i

I.

50 n

25 -

Dipodomys
REMOVAL

I
I
I

I

0.0 2.5

— I J 1 1 1 —

5.0 10.0 20.0 40.0 80.0

PERCENT COVER

Figure 7.— Distribution of trap captures (broken line; all species combined) and available
trap sites (sold line) relative to vegetative cover on (i) control; (Ii) D. spectobiWs removal;
and (iii) Dipodomys rerrtoval plots.

307

merriami interaction and, at the com-
n\unity level, this one interaction af-
fects the microhabitat utilization of
the majority of rodent species
through a complex network of direct
and indirect interactions. Perhaps the
most striking shift (not in the magni-
tude of response but in the number
of individuals involved) was the in-
creased use of open areas by the nu-
merically dominant D. merriami
when D. spedabilis, which had for-
mally used these sites was removed.
Most other shifts by the smaller ro-
dents, including the increased use of
open microhabitats by Perognathus
flavus, Peromyscus maniculatus and
Reithrodontomys megaloHs when all
Dipodomys were removed, suggest
that these species responded directly
to D. merriami and only indirectly to
D. spectabilis. Hence, there appears to
be a hierarchy of interactions. The
primary one is between the
behavioral (D. spectabilis) and
numerical CD. merriami) dominants
and it is this interaction around
which the community is organized.
Other studies have noted the poten-
tial for interference betv/een desert
rodents (Blaustein and Riser 1974,
Hutto 1978, Rebar and Conley 1983),
especially between D. spectabilis and
D. merriami (Frye 1983), and our
study shows how this one interaction
can resound throughout the commu-
nity to affect many other species.

A primary motivation for our
study — and most studies focusing on
the role of habitat — is that microhabi-
tats represent a limited and exploit-
able resource and the manner in
which they are used directly im-
pinges on population growth and
density. Many of the experimentally
induced microhabitat shifts we have
reported were accompanied by
changes in local species density
(Brown and Munger 1985, Brown et
al. 1986) that support the contention
that D. spectabilis controls the dynam-
ics of this community through a com-
bination of direct and indirect effects.
For example, increasing food levels
by adding seeds resulted in an in-

crease of D. spectabilis and a decrease
in D. merriami densities. Removal of
D. spectabilis resulted in positive den-
sity compensation of D. merriami but
no changes in densities of the smaller
seed-eaters; removal of all Dipod-
omys, however, resulted in large den-
sity increases in several of the
smaller rodents. Taken together, the
microhabitat and density responses
to our manipulations indicate that
interference competition for certain
foraging sites not only determines
the spatial organization of this com-
munity but that it is directly involved
in the regulation of rodent densities.

There are several aspects that war-
rant further comment. First, our re-
sults show that when D. spectabilis is
present open sites are underutilized
by the community as a whole; when
D. spectabilis is removed the
remaining Dipodomys shift to use
these open sites; but when all Dipod-
omys are removed the remaining sp)e-
cies are unable to fully utilize the va-
cated microhabitats (fig. 7). Hence,
there appears to be a limit to how far
the community can compensate for
the absence of certain species.
Among the possible explanations for
this might be that assemblages of
desert rodents have been associating
together for a sufficient time to have
lost the flexibility to respond to situ-
ations where one or more of the spe-
cies are absent (Schroder and
Rosenzweig 1975). Another is that
quadrupedal species may have a lim-
ited ability to avoid predators in
open microhabitats and this limits
the degree to which they can com-
pensate when the bipeds are re-
moved. In either case the relaxation
of one factor (in this case the removal
of dominant competitors) appears to
be accompanied by the increased im-
portance of others.

Second, the effects of interference
competition by D. spectabilis appear
to be effective in excluding inter-
specifics primarily in open areas al-
though this dominant does occur in
greater-than-median cover habitats.
It may be that aggression is of lim-

ited value in bushy microsites where
subdominant species may readily
find refugia. As a result, D, spectabilis
may be involved in two kinds of
interactions with each of its competi-
tors; exploitatively for seeds in bushy
sites and through interference in
open microhabitats. As a result, the
highly asymmetrical interactions be-
tween the dominant/subordinates in
open sites may become more nearly
symmetrical in bushy sites where
premiums are on foraging efficiency.

Third, the existence of strong, ag-
gressive interactions among sp>ecies
increases the potential for indirect
and high-order interactions that in-
volve species that overlap very little
in resource utilization. For example,
the large herbivore, Neotoma albigula
was as likely to shift its microhabitat
use as the granivorous species. How-
ever, it is interesting to note that al-
though the non-granivores shifted
microhabitat use when granivorous
species were removed, significant
density changes were limited to just
other granivores (Brown and Munger
1985). Hence, while interference may
play a role in determining use of mi-
crohabitats by rodents in several for-
aging guilds, its effects appear to be
most significant for ecologically simi-
lar species.

The goal of experimental pro-
grams is to hold most variables con-
stant while manipulating others, and
then to measure for shifts in response
variables. In this paper we have used
patterns of microhabitat use in con-
trol plots as a reference point for
interpreting our experimental results.
The assumption in doing this is that
the degree to which the community
responds to a particular manipula-
tion provides an estimate of its im-
portance in producing the basic pat-
tern. In our particular case we
wanted to know how the baseline
patterns of microhabitat use (i.e.,
those in control plots) change when
supplemental food is added or spe-
cies are removed. While some of our
patterns are easy to interpret, others
are very complex and appear to in-

308

volve a hierarchy of responses that
operate over different scales in time
and space. The existence of such a
dynamic and diverse set of responses
shows the limitations of most two-
species models of interspecific inter-
actions upon which past theories of
community organization have largely
been based; they also call into ques-
tion the value of studies seeking to
understand the mechanistic proc-
esses that determine community
composition through comparative,
nonexperimental methods.

Implications for Management

While the spatial association of small
mammals with particular microhabi-
tats has been rigorously and repeat-
edly documented, and the patterns
suggest almost a universal role of
microhabitat in ''structuring" small
mammal communities, the processes
responsible for producing these asso-
ciations are poorly understood (Price
and Brown 1983, Bowers 1986). To

(successfully manage/ manipulate
such communities there is a clear
need to better understand the proc-
esses that determine which mi-
crohabitats are used and which are
I not. Towards this end we identify
two particularly relevant areas for
our discussion: (1) the scales in time
j and space over which microhabitat
' use occurs; and (2) the roles of corre-
lation, and selection/ relegation in the
occupancy of microhabitats.

Vagile organisms, e.g. small mam-
mals, can potentially respond to fea-
tures of the habitat at several differ-
ent scales. At the macro-end of the
habitat spectrum animals choose ar-
eas in which to establish home
ranges. Microhabitat selection, in
contrast, usually involves the use/
disuse of small areas within the
home range. There are also temporal
differences in schedules of usage:
macrohabitat selection occurs over a
much longer timescale (weeks-
I months) while microhabitat use oc-
curs more immediately (seconds-

minutes). While it was assumed for
years that macrohabitat selection oc-
curred through the selection of com-
posite microhabitats, recent work on
small mammals suggests that the two
may be largely separate (Morris
1987).

Most factors that are demonstra-
bly important to the structure of
small mammal communities, i.e., pri-
mary productivity, plant species and
foliage height diversity, vegetation
cover, substrate type, competitor di-
versity and abundance, and preda-
tory pressure, vary more between
macrohabitats than among mi-
crohabitats within particular locales.
For example, primary productivity
and plant cover are determined by
plant species composition and gen-
eral conditions for growth that vary
over large environmental gradients
at the macrohabitat scale. These large
scale gradients influence patterns of
microhabitat use by determining
which rodent species are present,
their densities, the distribution and
abundance of food resources, and the
types of microhabitats that are avail-
able for selection. As a consequence,
the composition, densities and demo-
graphical behavior of small mammal
populations and communities may
more closely reflect habitat variabil-
ity at the macro — rather than the mi-
cro— scale. On the other hand, mi-
crohabitat usage is a phenomena in-
volving choices of individuals. Mi-
crohabitats that, by definition, vary
over scales smaller than individual
home ranges, have significance for
the survivorship or reproduction of
foraging individuals, but may have
little relevance when integrated over
the population as a whole.

Most experimental studies exam-
ining the role of microhabitat in
structuring small mammal communi-
ties tend to confound micro- and
macrohabitat effects. Typically, ma-
nipulations (e.g., food addition, spe-
cies removal, tailoring of vegetation)
are applied at the level of the mac-
rohabitat with microhabitat usage by
individuals measured as a response

variable. The research reported here
suffers from such a confounding.
Other field experiments that examine
the allocation of foraging time among
patches restrict manipulations to the
level of microhabitats (Kotler 1984,
Price and Waser 1985), and are not
confused by responses of entire
populations. Clearly, the time has
come to utilize the information we
now have to design comprehensive
studies that distinguish between mi-
cro- and macrohabitat selection: i.e,
studies that manipulate certain mi-
crohabitats on a scale over which
populations might respond.

Correctly gauging the scale over
which species respond to the envi-
ronmental mosaic is critical to the
successful management of that spe-
cies. Programs aimed at managing
species by manipulating microhabi-
tats may or may not be successful
depending on the scale at which the
manipulation is applied. If the goal is
to manage pK)pulations then mac-
rohabitat may be the correct context
for the program. This is not to sug-
gest that microhabitat is an inappro-
priate context for management pro-
grams. What it does suggest is that
management oriented programs
should be directed towards popula-
tions rather than the behavior of indi-
viduals. In many cases this may in-
volve changing the focus from the
micro to macro level.

Our second point for discussion
involves habitat correlation versus
selection/ relegation. Habitat usage is
determined by the habitats available,
the tolerances/preferences of organ-
isms for these habitats, and the
among-habitat variability in fitness.
Clearly, there must be some variabil-
ity in the structure of the habitat in
order for selection to occur. Habitats
that are relatively homogeneous at
the smaller scales may not exhibit
habitat associations even by highly
selective species. Conversely, show-
ing that a habitat has a significant
degree of microhabitat variability
does not imply that organisms have
the ability or inclination to respond

309

to that variability. In order to apply
the patterns of microhabitat use from
one site to predict what is occurring
at another requires an understanding
of the biological factors underlying
microhabitat use. Achieving this has
proved difficult because of several
problems. First, it is clear from a
growing body of experimental work
(including the present study) that
habitat association does not necessar-
ily imply habitat selection. Because
microhabitats are rarely discrete,
usually grade from one type to an-
other, and involve a suite of factors
that either characterize or are corre-
lated with specific microhabitats, it is
rare that habitat occupancy can be
tied to a single factor. As a result it is
difficult to conclude that an animal is
selecting a habitat per se, some fea-
ture of that habitat, or some factor
that is only correlated with that mi-
crohabitat. As a complicating factor
habitat selection probably reflects
integrated responses of organisms to
maximize fitness relative to several
largely independent processes. For
example, animals might select mi-
crohabitats so as to minimize preda-
tory risk, or food encounter rates, or
to jointly maximize food intake while
minimizing predatory risk (Bowers
1987).

Second, the present results and
those of others (Price 1978,
M'Closkey 1978, Wondolleck 1978,
Bowers et al. 1987) show that mi-
crohabitat provides a template over
which species interactions and com-
petitive hierarchies become ex-
pressed. The pattern is one of selec-
tion/relegation— the competitive
dominant selecting its preferred mi-
crohabitat and through exploitative
or interference competition relegat-
ing other species to less preferred
sites. The more ecologically similar
two species — and hence, the greater
the intensity of competition between
them — the greater the potential role
of interspecific competition in deter-
mining microhabitat usage.

Competitive interactions represent
dynamical processes impinging on

microhabitat association and usage.
Seasonal or year-to-year fluxes in re-
source availability or changes in the
distribution of resources among mi-
crohabitats can alter the economical
basis underlying competitive interac-
tions, and thereby promote shifts in
microhabitat usage. For example,
Congdon (1974) found during peri-
ods of low resource availability that
the large D. deserti and the smaller,
D. merriami, coexisted in the same
microhabitats but that the former be-
came aggressive and excluded the
latter from these sites when food lev-
els increased. Similarly, Frye (1983)
found that D. spectabilis excluded D,
merriami from areas around its bur-
rows just in the fall when seeds from
summer annuals were abundant.

Competitively based selection/
relegation has the effect of increasing
usage of secondary habitats while
decreasing usage of the most pre-
ferred ones. The result is that compe-
tition promotes the segregation of
species among microhabitats and the
degree to which the community is
spatially organized. Thus it is no ac-
cident that the most striking patterns
of microhabitat use and segregation
are in communities that are highly
competitive (Connor and Bowers
1987). As the present study has dem-
onstrated even one strong interaction
involving just two species (in this
case the behavioral and numerical
dominants) can affect microhabitat
usage by all species in the commu-
nity through direct and indirect path-
ways of interaction.

Care must be taken when examin-
ing the spatial organization of com-
munities where competition might be
occurring. Efforts to understand mi-
crohabitat utilization through recon-
stitution studies that measure indi-
vidual species preferences for mi-
crohabitats, then combines these in a
general model of microhabitat asso-
ciation, will miss higher-order com-
petitive effects that may be the main
determinants of microhabitat use.
Further, since competition can be in-
determinate, work over complex

pathways, and operate over widely
varying scales in time and space it is
doubtful that any one model can be
used to predict microhabitat use over
all communities. As a first step to-
wards using microhabitat utilization
as a tool for management programs
we need to know which communities
are interactive (i.e., structured
around selection/ relegation
schemes), which are non-interactive,
and something about ecological at-
tributes of each. It may be that in
some communities microhabitat is
the correct context for management
programs while in other communi-
ties the focus should be on species
interactions. Species removal experi-
ments such as the one described here
provide a straightforward test of
these models.

What we are suggesting here is
that microhabitat use be viewed as a
manifestation of process and that
these processes provide the basis for
management. We feel that the most
important question is not which
habitats are being used by a particu-
lar species but why it is using that
microhabitat and not others. Recent
work has shown that the pathways
by which species interact at the level
of ecological communities can be
very complex and that similar pat-
terns of microhabitat usage need not
share a common sequence of causa-
tion (see papers in Diamond and
Case 1986).

Without knowing something
about which processes are locally im-
portant it is risky to extrapolate find-
ings from one site in managing an-
other. For example. Bowers (1986)
found in rarefaction studies of the
same three species rodent commu-
nity that microhabitat use at one site
was affected by intersp)ecific compe-
tition but not at two others. Such re-
sults underscore the fact that mi-
crohabitat use involves multidimen-
sional responses of organisms to
their environment. Understanding
the basics of such relationships
should be the goal of community
ecologists and managers alike.

310

Acknowledgments

We thank J.H. Brown and D.B. Th-
ompson for help in the field and for
discussion. J.H. Brown and R.T.
M'Closkey provided critical reviews
of the n[\anuscript.

Literature Cited

Blaustein, Andrew R., and Arthur C.
Riser. 1974. Dominance relation-
ships of the dark kangaroo mouse
(Microdipodops megacephalus) and
the little pocket mouse (Per-
ogmthus longimembris) in captivity.
Great Basin Naturalist 34:312-314.

Bowers, Michael A. 1982. Foraging
behavior of heteromyid rodents:
field evidence of resource parti-
tioning. Journal of Mammalogy
63:361-367.

Bowers, Michael A. 1986. Geographic
comparison of microhabitats used
by three heteromyid in response to
rarefaction. Journal of Mam-
malogy 67:46-52.

Bowers, Michael A. 1987. Seed re-
moval experiments on desert ro-
dents: the microhabitat by moon-
light effect. Journal of Mam-
malogy 69:201-204.

Bowers, Michael A., and James H.
Brown. 1982. Coexistence of desert
rodents: Chance or community
struchire? Ecology 63:391-400.

Brown, James H., Diane W.

Davidson, James C. Munger, and
Richard S. Inouye. 1986. Experi-
mental community ecology: the
desert granivore system, p. 41-61.
In (J. Diamond and T.J. Case, eds.)
Community ecology. Harper and
Row, New York.

Brown, James H., and James C.
Munger. 1985. Experimental ma-
nipulation of a desert rodent com-
munity: food addition and species
removal. Ecology 66:1545-1563.

Brown, James H., O.J. Reichman, and
Diane W. Davidson. 1979. Gra-
nivory in desert ecosystems. An-
nual Review of Ecology and Sys-
tematics 10:201-227.

Bowers, Michael A., Daniel B. Th-
ompson, and James H. Brown.
1987. Spatial organization of a des-
ert rodent community: food addi-
tion and species removal. Oecolo-
gia 72:77-82.

Colwell, Robert K., and Eduardo R.
Fuentes. 1975. Experimental stud-
ies of the niche. Annual Review of
Ecology and Systematics 6:281-
310.

Congdon, Justin. 1974. Effect of habi-
tat quality on distributions of
three sympatric species of desert
rodents. Journal of Mammalogy
55:659-662.

Connor, Edward F., and Michael A.
Bowers. 1987. The spatial conse-
quences of interspecific competi-
tion. Annales Zoologici Fennici
24:213-226.

Davidson, Diane W., Douglas A.
Sampson, and Richard S. Inouye.
1985. Granivory in the Chihua-
huan Desert: interactions within
and between trophic levels. Ecol-
ogy 66:486-502.

Diamond, Jared, and T.J. Case. 1986.
Community Ecology. Harper and
Row, New York, New York. 663 p.

Fretwell, Steven D., and H.L. Lucas,
Jr. 1970. On territorial behavior
and other factors influencing habi-
tat distribution in birds. I. Theo-
retical development. Acta Biotheo-
redca 19:16-36.

Frye, Robert J. 1983. Experimental
field evidence of interspecific ag-
gression between two species of
kangaroo rats (Dipodomys).
Oecologia 59:74-78.

Harris, John H. 1984. An experimen-
tal analysis of desert rodent forag-
ing ecology. Ecology 65:1579-1584.

Hay, Mark E., and P.J. Fuller. 1981.
Seed escape from heteromyid ro-
dents: the importance of mi-
crohabitat and seed preference.
Ecology 62:1395-1399.

Hutto, Richard L. 1978. A mecha-
nism for resource allocation
among sympatric heteromyid ro-
dent species. Oecologia 33:115-126.

Kotler, Burt P. 1984. Risk of preda-
tion and the structure of desert

rodent communities. Ecology
65:689-701.

Lemen, Cliff A., and Patricia W. Free-
man. 1986. Interference competi-
tion in a heteromyid community
in the Great Basin of Nevada,
USA. Okios 46:390-396.

Levins, Richard. 1962. Theory of fit-
ness in a heterogeneous environ-
ment. I. The fitness set and adap-
tive function. American Naturalist
96:361-373.

M'Closkey, Robert T. 1978. Niche
separation and assembly in four
species of Sonoran Desert rodents.
American Naturalist 112:683-694.

Morris, Douglas W. 1987. Ecological
scale and habitat use. Ecology
68:362-369.

Munger, James C, Michael A. Bow-
ers, and W. Thomas Jones. 1983.
Desert rodent populations: factors
affecting abundance, distribution,
and genetic structure. Great Basin
Naturalist Memiors 7:91-116.

Price, Mary V. 1978. The role of mi-
crohabitat in structuring desert
rodent communities. Ecology
59:910-921.

Price, Mary V. 1983. Ecological con-
sequences of body size: a model
for patch choice in desert rodents.
Oecologia 59:384-392.

Price, Mary V., and James H. Brown.
1983. Patterns of morphology and
resource use in North American
desert rodent communities. Great
Basin Natur. Memiors 7:117-134.

Price, Mary V., and Nickolas M.
Waser. 1985. Microhabitat use by
heteromyid rodents: effects of arti-
ficial seed patches. Ecology 66:211-
219.

Rebar, Cindy, and Walt Conley. 1983.
Interactions in microhabitat use
between Dipodomys ordii and
Onychomys leucogaster. Ecology
64:894-899.

Reichman, O.J. 1981. Factors influ-
encing foraging in desert rodents,
p. 195-213. In (A.C. Kamil and
T.D. Sargent, eds.) Foraging be-
havior: ecological, ethological and
psychological approaches. Gar-
land STPM, New York.

311

Rosenzweig, Michael L. 1981. A the-
ory of habitat selection. Ecology
62:327-335.

Rosenzweig, Michael L., and Jerald
Winakur. 1969. Population ecology
of desert rodent communities:
habitats and environmental com-
plexity. Ecology 50:558-572.

Schoener, Thomas W. 1971. The the-
ory of foraging strategies. Annual
Review of Ecology and Sys-
tematics 2:369-404.

Schroder, Gene, and Michael L.
Rosenzweig. 1975. Perturbation
analysis of competition and over-
lap in habitat utilization between
Dipodomys ordii and Dipodomys
merriami. Oecologia 19:9-28.

Siegel, Sidney 1956. Nonparametric
statistics for the behavioral sci-
ences. McGraw-Hill, New York.

Wondolleck, John T. 1978. Forage-
area separation and overlap in
heteromyid rodents. Journal of
Mammalogy 59:510-518.

Response of Small Mammal
Communities to Silvicultural
Treatments in Eastern
Hardwood Forests of West
Virginia and Massachusetts^

Robert T. Brooks and William M. Healy^

Abstract.— We studied small mammal
communities and associated habitats in West
Virginia and Massachusetts hardwood forests with
different silvicultural treatments. In Massachusetts,
white-tailed deer (Odocoileus virginianus) density
was a second interactive treatment. Total capture
rates were relatively stable across all treatment
classes. Small mammal community composition and
individual species capture rates varied according to
treatment. White-toiled deer density had a greater
effect on the small mammal community than did
silvicultural practices.

Small mammals (i.e. New World
mice, voles and jumping mice [Crice-
tidae and Zapodidae], shrews [Sori-
cidae], and squirrels [Sciuridae]) are
an important component of north-
eastern forest ecosystems. Their posi-
tions in the food web are broad,
functioning as foragers on plant and
faunal biomass and as prey to nu-
merous predators. Small mammals
play an important role in forest dy-
namics by dispersing seeds and my-
corrhizal fungal sjx)res and by en-
hancing organic matter decomposi-
tion and mineral cycling (Spurr and
Barnes 1980).

Relatively little is known of the
response of small mammals, by spe-
cies and as a community, to silvicul-
tural treatments of northeastern
hardwood forests. Several studies
have shown that the response varies
by species but that the small mam-
mal community is generally resilient
to forest harvesting (Healy and
Brooks 1988, Kirkland 1977, Lovejoy
1975, Clough 1987, Monthey and
Soutiere 1985). These studies report
the predominant effect of silvicultu-
ral treatments on small mammal
habitat is the enhancement of the

'Paper presented at symposium. Man-
agement of Amphibians, Reptiles, and
Small Mammals in Nortti America (Flagstaff,
AZ.July 19-21, 1988).

'Robert T. Brooks and William M. Healy
are Researcti Wildlife Biologists, USDA Forest
Service, Norttieastern Forest Experiment
Station, University of Massact^usetfs,
Holdswortt) Hall, Amherst, MA 01003.

ground cover and lesser woody
vegetation. Stenotopic species sensi-
tive to understory plant cover and its
influence on microclimate seem to be
encouraged, at least temporarily, by
most forest harvesting, while eu-
ry topic sp)ecies seem unaffected.

The study began in West Virginia
(WV), where one field season was
completed, and was continued in
Massachusetts (MA). Our objective
was to investigate the response of the
small mammal community, as char-
acterized by live-trapping statistics,
to standard eastern hardwood
silvicultural treatment (Marquis et al.
1975, Hibbs and Bentley 1983). In
WV, we studied the effects of even-
aged regeneration clearcutting and
subsequent succession on small
mammal trapping data. In MA, the
silvicultural treatment was interme-
diate thinnings, with a second inter-
active treatment of differential white-
tailed deer density.

STUDY AREAS

The WV study sites were on the
Cheat Ranger District, Monongahela
National Forest. Three randomly lo-
cated stands in each of four stand-
age classes were studied to evaluate
the small mammal community over a
silvicultural rotation for even-aged
management of a northern hard-
wood forest. The four age classes
were seedling (8-9 years), sapling
(12-14 years), sawtimber (61-76

years), and mature (>100 years). The
12 stands averaged 19.4 ha and
ranged in area from 6.1 to 38.8 ha.
The study area is described in Healy
and Brooks (1988).

The MA study sites were on the
Quabbin Reservation in Franklin
county. This watershed is managed
by Boston's Metropolitan District
Commission for water production.
Four randomly selected stands in
each of four treatment classes were
studied to evaluate the interactive
effects of intermediate thinning and
white-tailed deer density on a south-
em New England oak forest's flora
and fauna. The treatments were com-
binations of thinned vs. unthinned
and low (6-8/mi2) vs. high (34-59/
mi^) deer density. The 16 stands av-
eraged 19.1 ha and ranged in area
from 4.9 to 57.5 ha. The MA study
site is described in Healy et al. (1987).

METHODS

SnrKili Mammals

Small mammals were live-trapped at
10 systematically located stations
along a transect in each of the 28
stands. Transects were located along
the long axis of each stand. Trap sta-
tions were no less than 80 m apart in
any stand. At each station three Sher-
man-type box traps (7.6 X 7.6 X 30.5
cm) were baited with a mixture of
peanut butter, rolled oats, and bacon
fat and set within 1 m of each station.

313

Traps were set for three successive
nights, left closed for one (WV) or
four (MA) nights, and then set for
three additional successive nights.
Sprung traps were noted and their
numbers subtracted from the total
number of trap nights per station (18)
to calculate the number of effective
trap nights. Each forest stand was
trapped once per year. Mammals
were trapped from mid-September to
early October 1981 in WV. Mammals
were trapped during June and July of
1985-87 in MA. Captured mammals
were marked for individual identifi-
cation and released.

Vegetation

Vegetation sampling techniques var-
ied between states. Vegetation plots
were systematically located along the
same transects as were the small
mammal trapping stations. In WV,
trees (> 2.5 cm) were sampled using
point-centered-quarter method (Cot-
tam and Curtis 1956), while in MA,
trees were sampled using fixed-ra-
dius plots. Herbaceous and woody-
stemmed understory, including trees
< 2.5 cm, were sampled in WV using
the line intercept method (Eberhardt
1978) and in MA, these flora were
sampled using fixed-radius plots.
Tree and understory sampling oc-
curred at the same locations along

the transects. These data were used
to estimate tree density, dominance,
and average diameter and under-
story cover by major plant life form
(i.e. forb, fern, graminoid, and
woody-stem species).

Analysis

Small mammal trapping results and
vegetation samples were summa-
rized by treatment class and forest
stand. Treatment effects on small
mammal capture rates, standardized
as captures per 100 trap nights (TN),
were analyzed by one-way (WV) or
two-way (MA) analysis of variance
in a balanced, nested design with
stand sum-of-squares the error term
for treatment effect. Treatment ef-
fects on species composition of small
mammal capture rates were analyzed
using multivariate analysis of vari-
ance. Testing of treatment effects was
done using the SPSS MANOVA pro-
cedure (Hull and Nie 1981).

RESULTS
Vegetation Structure
West Virginia

Tree density declined and both basal
area and average tree diameter in-

creased as the forest stands matured
from an even-aged regeneration har-
vest (table 1).

The understory changed more in
life form composition than in total
cover. Forb cover increased in per-
centage of cover with stand-age as
did ferns while shrub cover declined
(table 1). All forest stands supported
a luxuriant understory regardless of
age.

Massachusetts

Tree density and basal area de-
creased with thinning while average
tree diameter changed little (table 1).
The effect of deer density is under-
standable if one considers the low
deer-unthinned treatment to be a
"control" condition.

From this perspective, high deer-
density stands had lower tree density
and basal area, and a larger average
diameter because of poor regenera-
tion resulting from browse damage
(table 1).

Forb cover declined with higher
deer densities, while graminoid
cover increased (table 1). Shrub and
fern cover responded irregularly to
the treatments except for a dramatic
increase in fern cover in high deer-
thinned stands, an effect reported
elsewhere (Marquis 1987).

Table 1 .—Average structural characteristics of scmnpied forest stands by state and treatment class.

West Virginia

Massachusetts

Low Deer

High Deer

Characteristic

Seedling

Sapling

Sdwtlmber

Mature

Unthinned

Thinned

Unthinned

Thinned

Tree stems > 2.5 cm

Stems/ha

1970

2482

969

772

1334

876

974

645

Basal area (mVha)

5.3

12.3

41.7

35.9

24.5

15.7

22.8

15.7

Average diameter (cm)

5.2

6.8

17.9

17.6

12.2

10.7

13.5

13.5

Percent understory cover

Forb species

17

18

18

36

18

16

7

14

Fern species

5

11

21

14

15

13

12

32

Graminolds

2

2

<1

<1

1

2

5

17

Shrubs and trees < 2.5 cm

32

17

9

9

15

31

26

26

314

SrTKill Mammals
West Virginia

In the one trapping season, 662 indi-
viduals of 15 sf)edes were captured.
Total capture rate averaged 33.2 indi-
viduals/100 TN. Average total cap-
ture rate declined with stand-age,
from 42.4 individuals/100 TN in
seedling stands to 27.4/100 TN in
sawtimber stands, and then in-
creased to 31.0/100 TN in mature
stands (table 2). The effect of stand-

age class on total capture rate was
not statistically significant (F = 3.16,
P = 0.086, d.f. = 3,8).

Six species were captured in all
four forest age classes, eight addi-
tional species were captured in three
or fewer treatment classes (table 2).
Species richness was greatest in the
sawtimber stands, intermediate in
the younger stands, and least in the
mature stands.

The southern red-backed vole (see
table 2 for small mammal scientific
nomenclature) was the most com-

mon species, averaging 12.7 indi-
viduals/100 TN. Capture rate for this
species declined with stand-age
through sawtimber stands (table 2),
but treatment effect was not signifi-
cant (F = 2.37, P = 0.146, d.f. = 3,8).
Deer mice were the second most
common species, with an average
capture rate of 10.0 individuals/ 100
TN. Capture rates for this species
were similar across treatment class
except for a lower rate in the seedling
stands. No significant differences
were found between stand-age class

Table 2.— Average number of individual small mammals captured per 100 trap nights by species, state, and treat-
ment class.

West Virginia

Massachusetts

Low deer

High deer

Characteristic

Seedling Sapling SawtimlDer Mature Unthinned Thinned Unthlnned Thinned

S. red-backed vole

19.8

12.7

7,3

n.i

15.0

12,3

2.8

3.8

(y^i&iiiiiufiofTiys yufjf^&fi/

Short-tailed shrew

9 J

3.5

3.6

2.1

1.1

1.4

0.2

0,9

(Blarina brevicauda)

E. chipmunk

0,6

0.6

1.2

1.8

0.7

0.4

0.4

1.2

(Tamias sfriafus)

White-footed mouse

0.6

17,2

17.8

30.9

23.4

(Peromyscus leucopus)

Deer mouse

7.1

ID.O

11.7

11.3

(P. maniculafus)

Woodland jumping mouse

2.1

2.3

0.6

3.0

0.1

(Napaeozapus insignis)

Rock vole

1.6

1.4

0.4

1.2

(Microfus chroforrhinus)

S. flying squirrel

1.1

1.4

0.4

(Glaucomys volans)

Smoky shrew

0.2

0.2

0.4

(Sorex fumeus)

Meadow vole

1.0

0.1

(M. pennsyivanicus)

Red squirrel

0.7

<0.1

(Tamiasciurus hudsonicus)

Masked shrew

0,2

<0.1

(S, cinereus)

Long-tailed shrew

0.2

(S. dispar)

Woodland vole

0.2

0.1

0.1

(M. pineforum)

Total all species

42.4

32.1

27.4

31.0

34.2

32.2

34.3

29.4

Total number trap nights^

497

485

510

608

2036

2055

2021

2008

'Scientific names from Jones et al. 1975.

^Total number of possible trap nighits (WV=540; MA=2160) minus sprung traps.

315

(F = 0.29, P = 0.766). Short-tailed
shrews were the only other species
frequently caught in all stand-age
classes. Shrews were most common
in the seedling stands but no signifi-
cant treatment effect was found (F =
0.96, P = 0.459). No significant treat-
ment effect was found for eastern
chipmunks (F = 1.26, P = 0.351),
woodland jumping mice (F = 0.21, P
= 0.885), and rock voles (F = 0.41, P =
0.749), which were caught infre-
quently in all stand-age classes (table
2). The remaining eight species were
caught with less regularity. No fur-
ther analysis was completed for these
species. No significant treatment ef-
fect was found in the simultaneous
capture rates of the six most com-
monly trapped species (i.e., red-
backed and rock voles, short-tailed
shrews, chipmunks, and deer and
jumping mice) (Wilks lambda =
0.046, Rao's F = 0.979, P = 0.54).

Massachusetts

Over 3 years, 2,630 individual small
mammals of nine species were cap-
tured. Average total capture rate was
32.6 individuals/ 100 TN. There was
a significant decline in capture rate
across the years (F = 30.02, P < 0.001,
d.f. = 2,24). The capture rate of 43.7
individuals/100 TN in 1985 declined
to 33.7 in 1986 and 20.3 in 1987. The
decline was observed across all treat-
ments and all stands.

We found no significant full model
treatment effect on total capture rate
(F = 1.78, P = 0.204, d.f. = 3,12). Total
capture rate for all species was high-
est in the unthinned stands and low-
est in the thinned stands, especially
in the high deer-density stands (table
2). Neither thinning (F = 3.99, P =
0.069, d.f. = 1,12) nor deer density (F
= 0.60, P = 0.453) had a significant
effect on total capture rates.

Species richness was highest in the
high deer-thinned treatment class,
intermediate in the two low deer-
density classes, and lowest in the
high deer-unthinned treatment (table

1). White-footed mouse was the most
commonly captured species, fol-
lowed by southern red -backed voles
(table 2). Capture rates for both spe-
cies differed by treatment class (F =
9.01, P = 0.002, d.f. = 3,12 for mice; F
= 6.06, P = 0.009 for voles), with deer
density a significant effect (F = 20.7,
P = 0.0007. d.f. = 1.12 for mice; F =
17.5, P - 0.01 for voles), and thinning
effect nonsignificant (F = 2.72, P =
0.125 for mice; F = 0.11, P = 0.74 for
voles). Voles were most commonly
captured in stands of low deer-den-
sity, and mice most commonly cap-
tured in stands of high deer-density.

Short-tailed shrews and eastern
chipmunks were the only other spe-
cies captured in each of the four
treatments. Shrew captures, like
those for red-backed voles, declined
with increasing deer density (F = 6.2,
P = 0.028) but showed no significant
response to thinning (F = 3.1, P =
0.1). Chipmunk captures showed no
significant response to either deer
density (F = 0.95, P = 0.35) or thin-
ning (F = 1.52, P = 0.24). The remain-
ing five species were infrequently
caught in three or fewer treatment
classes, and no further analysis was
performed.

Relative capture abundance of the
four most commonly captured spe-
cies (i.e., white-footed mice, red-
backed voles, short-tailed shrews,
and chipmunks) differed between the
two levels of deer density (Wilks
lambda = 0.237, Rao's F = 7.25, P =
0.007). No difference in relative cap-
ture abundance was found between
the two thinning classes (Wilks lamb-
da = 0.496, Rao's F = 2.28, P = 0.14).

DISCUSSION

Silvicultural treatments had no sig-
nificant effect on total small mammal
captures. Total capture rates were
stable across the range of treatments
in both WV (clear-cutting and subse-
quent regrowth) and MA (intermedi-
ate thinning) with the exception of
WV seedling stands (table 2). In

those stands, where regenerating
trees, shrubs, and herbaceous plants
flourish in the sunlight afforded by
the removal of the overstory, total
capture rates increased. Otherwise,
treatment effects on habitat structure
were insufficient to alter total cap-
ture rates, as changes in the species
composition of small mammal cap-
tures were compensatory.

Six of the 14 small mammal spe-
cies captured in WV were captured
in all four stand-age classes. Of the
other species: red squirrels were ob-
served in all stands but poorly cap-
tured in our traps; white-footed
mice, woodland and meadow voles,
and masked and long-tailed shrews
were each captured in one stand;
four smoky shrews were captured in
three stands; and southern flying
squirrels were captured in sapling
and older stands. McKeever (1955)
generally concurs that these species
are uncommon in WV (woodland
vole, masked and long-tailed shrew),
or are common in forests not
sampled in this study (white-footed
mouse in lower elevation forests) or
other habitats (meadow vole). Smoky
shrews and southern flying squirrels
are more common WV small mam-
mals but were poorly represented in
our sample. Capture data for these
species are insufficient for drawing
any conclusions regarding species
response to clearcutting.

West Virginia red-backed vole and
short-tailed shrew captures increased
concurrent with a decline in deer
mouse captures (table 2). Vole and
shrew capture rates were highest in
seedling stands. Kirkland (1977) and
Lovejoy (1975) reported a similar re-
sponse in vole captures but not for
shrew captures. The increase in vole
captures could be a response to the
flush in vegetation associated with
overstory removal and to the volume
of slash occurring immediately sub-
sequent to harvest. These factors al-
ter ground level microclimate, in-
creasing humidity and improving
conditions for red-backed voles
(Lovejoy 1975, Merrit 1981).

316

Vole and shrew captures declined
and deer mice captures increased as
the forest stands matured. Forb cover
remained stable with increasing
stand-age while fern cover increased
and shrub cover declined (table 1).
These changes presumably altered
microhabitat conditions to the detri-
ment of red-backed voles. In mature
forest stands, vole and mouse cap-
tures were equal. In these stands,
forb cover increased dramatically
from conditions observed in sawtim-
ber stands, fern cover declined, and
shrub cover remained stable (table 1).
These habitat conditions resulted in
an increase in red-backed vole cap-
tures in mature stands over capture
rates for the species in sawtimber
stands.

Less frequently trapped rock voles
were captured in stands with rock
outcrops. Eastern chipmunks cap-
tures increased with stand age, and
woodland jumping mice captures
showed no clear response to stand
age. Capture rates for these two spe-
cies were not related to measured
habitat variables (Healy and Brooks
1988).

Species composition of WV small
mammal captures and individual
species capture rates were not signifi-
cantly different between treatment
classes. No major small mammal spe-
cies was eliminated by clearcutting
and the subsequent maturing of the
regeneration of the hardwood
stands. These species either survived
within clearcut stands or recolonized
harvested stands from adjacent un-
cut stands. Within maturing stands,
habitat conditions were sufficiently
diverse to support all major species.

These results demonstrate that
clearcutting of WV northern hard-
wood forests allowed for the contin-
ued maintenance of the small mam-
mal community. Out data showed
the small mammal community to be
relatively stable across a silviculhiral
rotation, with no major changes in
composition or capture rates that
could alter forest ecosystem function-
ing or character.

Total capture rates were stable
across treatment classes in MA.
Treatment effects upon habitat struc-
ture in these stands were insufficient
to alter total capture rates. However,
capture rates for individual small
mammal species varied among forest
treatments. Deer-density had a
greater influence on both individual
species capture rates and species
composition than did silvicultural
treatment. There was a reciprocal
change in the relative abundance of
red-backed voles and white-footed
mice with changes in deer density
(table 2).

During the 3 years of this study,
fall deer density averaged 18/km^ in
the high deer-density stands and 3/
km^ in the low deer-density stands
(Healy et al. 1987). Red-backed voles
were scarce in high deer-density
stands. Ferns and ericaceous shrubs
dominated the understory of these
stands while the understory of low
deer-density stands contained a
greater overall number of plant spe-
cies and forb species were more
abundant (Healy et al. 1987). Red-
backed voles prefer mesic to hydric
sites, especially in the southern por-
tion of their New England range
(Miller and Getz 1972, 1973). It seems
that foraging by deer may have suffi-
ciently altered the understory vegeta-
tion to depress vole populations.

The response of white-footed mice
to deer density in MA was less clear.
Although capture rates in low deer-
density stands were fewer than in
high deer-density stands, they never-
theless exceeded capture rates for
red-backed voles in all treatment
classes (table 2). White-footed mice
are ubiquitous in habitat preference
within the forest ecosystem (King
1968, Godin 1977, Hamilton and
Whitaker 1979). Whereas Wolff and
Dueser (1986) suggest that the these
two species can coexist noncompeti-
tively through microhabitat and food
habit differences, our data suggest
that mice capture rates are sup-
pressed in stands with high vole cap-
ture rates. Our stand data are at too

coarse a scale to address microhabi-
tat separation. One would need to
manipulate vole jx)pulations experi-
mentally to evaluate whether the
abundance of voles is competitively
suppressing mice populations in low
deer-density stands with better qual-
ity vole habitat.

Short-tailed shrews captures were
more common in low deer-density
stands, a possible response to the
greater forb cover observed in these
stands and probable increase in
ground level humidity. Eastern chip-
munk captures offer no ready inter-
pretation in regard to response to
treatment effect or habitat structure.
The remaining five species were cap-
tured so infrequently that it is impos-
sible to draw any conclusions as to
the effects of either thinning or deer
density on capture rates.

Thinning MA oak forests had no
significant effect on capture rates of
the four major small mammal species
or species composition of the cap-
tures. From a management perspec-
tive, intermediate thinning of these
forests did not alter the continuation
of the pretreatment small mammal
community. For those situations
where white-tailed deer have been
allowed to reach population levels
where vegetation is altered, signifi-
cant changes in the small mammal
community are found. Silvicultural
treatment effects on small mammal
habitat are temporary and ecosystem
resources (i.e. nutrients, energy) re-
main available to small mammals.
Long-term, high populations of deer,
a large, possibly competing herbi-
vore, alter the structure and compo-
sition of small mammal habitat to the
detriment of some species.

CONCLUSIONS

The small mammal community is an
important component of northeast-
em forested ecosystems, functioning
both as a consumer of plant and ani-
mal biomass and as prey to numer-
ous predators. Intermediate thinning

317

and clearcutting treatments, which
are common silvicultural practices,
have minimal or ephemeral effects
on the numbers of small manunals
and the composition of the small
mammal community found in these
forests. Long-term, high deer popula-
tions may permanently alter habitat
structure to the extent that changes
occur in small mammal community
composition.

LITERATURE CITED

Clough, Garrett C. 1987. Relations of
small mammals to forest manage-
ment in northern Maine. Canadian
Field-Naturalist 101:40-48.

Cottam, Grant and J. T. Curtis. 1956.
The use of distance measures in
phytosociological sampling. Ecol-
ogy 37:451-460.

Eberhardt, L. L. 1978. Transect meth-
ods for population studies. Journal
of Wildlife Management 42:1-31.

Godin, Alfred J. 1977. Wild mammals
of New England. 304 p. The John
Hopkins Press, Baltimore, Md.

Hamilton, William J. and John O.
Whitaker. 1979. Mammals of the
eastern United States. 346 p. Cor-
nell University Press, Ithaca, N.Y.

Healy, William M. and Robert T.
Brooks. 1988. Small mammal
abundance in northern hardwood
stands of different ages in West
Virginia. Journal of Wildlife Man-
agement 52(3) :[in press].

Healy, William M., Robert T. Brooks,
and Paul J. Lyons. 1987. Deer and
forests on Boston's municipal wa-
tershed after 50 years as a wildlife
sanctuary, p. 3-21. In Deer, for-
estry, and agriculture: Interactions
and strategies for management.
David Marquis, editor. Sympo-
sium Proceedings, Plateau and
Northern Hardwood Chapter, Al-
legheny Society American Forest-
ers, Warren, Pa.

Hibbs, David E. and William R.
Bentley. 1983. A Management
Guide for Oak in New England. 12
p. The University of Connecticut,

Cooperative Extension Service,
Storrs.

Hull, C. Hadlai and Norman H. Nie.
1981. SPSS Update 7-9: New pro-
cedures and facilities for releases
7-9. 402 p. McGraw-Hill Book
Company, NY.

Jones, J. Knox, Dilford C. Carter, and
Hugh H. Genoways. 1975. Revised
checklist of North American mam-
mals north of Mexico. 14 p. The
Museum, Texas Tech. Univ., Lub-
bock. Occasional Paper No. 28.

King, John A. 1968. Biology of Pero-
myscus (Rodentia). 593 p. The
American Society of Mammalo-
gists. Special Publication No. 2.

Kirkland, Gordon L. 1977. Responses
of small mammals to the clearcut-
ting of northern Appalachian for-
ests. Journal of Mammalogy
58:600-609.

Lovejoy, David A. 1975. The effect of
logging on small mammal popula-
tions in New England northern
hardwoods. University of Con-
necticut Occasional Papers, Bio-
logical Science Series 2:269-287.

Marquis, David A., editor. 1987.
Deer, forestry, and agriculture:
Interactions and strategies for
management. 183 p. Symposium
Proceedings, Plateau and North-
ern Hardwood Chapter, Al-
legheny Society American Forest-
ers. Warren, Pa.

Marquis, David A., Ted J. Grisez,
John C. Bjorkbom, and Benjamin
A. Roach. 1975. Interim Guide to
Regeneration of Allegheny Hard-
woods. USDA Forest Service Gen-
eral Technical Report NE-19. 14 p.
Northeastern Forest Experiment
Station, Broomall, PA.

McKeever, Sturgis. 1955. Ecology
and distribution of the mammals
of West Virginia. Ph.D. Disserta-
tion. 335 p. North Carolina State
College, Raleigh.

Merrit, Joseph F. 1981. Clethrionomys
gapperi. 9 p. The American Society
Mammalogists Mammalian Spe-
cies No. 146.

Miller, Donald H. and Lowell L.
Getz. 1972. Factors influencing the

local distribution of the redback
vole, Clethrionomys gapperi, in New
England. University of Connecti-
cut Occasional Papers, Biological
Science Series 2:115-138.

Miller, Donald H. and Lowell L.
Getz. 1973. Factors influencing the
local distribution of the redback
vole, Clethrionomys gapperi, in New
England. II. Vegetation cover, soil
moisture, and debris cover. Uni-
versity of Connecticut Occasional
Papers, Biological Science Series
2:159-180.

Monthey, Roger W. and E. C. Souti-
ere. 1985. Responses of small
mammals to forest harvesting in
northern Maine. Canadian Field-
Naturalist 99:13-18.

Spurr, Stephen H. and Burton V. Bar-
nes. 1980. Forest Ecology. 687 p.
3rd edition. John Wiley and Sons,
N.Y.

Wolff, Jerry O. and Raymond D.
Dueser. 1986. Noncompetitive co-
existence between Peromyscus spe-
cies and Clethrionomys gapperi. Ca-
nadian Field-Naturalist 100:186-
191.

318

Habitat Structure and the
Distribution of Small
Mamnnals in a Northern
Hardwoods Forest^

Jeffery A. Gore^

Abstract.— The influence of habitat structure on

the distribution of small mammals was studied in an
old-growth northern hardwoods forest in New
Hampshire. Logistic regression equations developed
with data from three live-trapping grids were able to
classify locations of just three of eight small mammal
species better than expected by chance. For all
species the regression models failed to correctly
predict presence in an independent grid. At the
scale tested, habitat structure hod little effect on the
distribution of small mammals within this forest type.

In northern temperate forests small
mammals are distributed unevenly
across available habitat, even within
a single forest type or age class
(Dueser and Shugart 1978, Vickery
1981, Parren 1981, Seagle 1985a). Dif-
ferential use of certain segments or
microhabitats within a broader habi-
tat type has most often been reported
for sympatric species of small mam-
mals, but intraspecific variation in
microhabitat use has also been noted
(Kitchings and Levy 1981, Vickery
1981, Seagle 1985a).

Differential use of microhabitats
by small mammals may be a conse-
quence of the ecological require-
ments of each species (i.e. habitat se-
lection) or it may be the result of par-
titioning of habitat by competing sf)e-
cies (Crowell and Pimm 1976, Porter
and Dueser 1982). Another hypothe-
sis is that the observed use of mi-
crohabitats by small mammals is pri-
marily a function of the density of
small mammal populations.

Under this hypothesis, use of a
certain microhabitat is determined
more by the availability of animals to
occupy the area than by structural

'Paper presented at symposium. Man-
agement of Amphibians, reptiles, and Small
Mammals in Nortti America. (Flagstaff, AZ,
July 19-21, 1988.)

^Jeffery A. Gore, formerly a graduate
student. University of Massachiusetts, De-
partment of Forestry and Wildlife Manage-
ment is currently Nongame Wildlife Biolo-
gist, Florida Game and Fresh Water Fish
Commission, f?t. 4. Box 759. Panama City.
Florida 32405.

characteristics of the microhabitat.
Extrinsic factors, such as food availa-
bility, disease, or predation, could
alter population levels and thus indi-
rectly influence the distribution of
small mammals among microhabi-
tats. Observed microhabitat use
might also be a function of some
combination of habitat selection,
competitive partitioning, and factors
affecting population density.

The question of which mechanism
most influences the distribution of
small mammals is of more than aca-
demic impK)rtance. If small mammals
select among microhabitats based on
structural features, then disturbance
such as timber harvesting may have
a considerable impact on population
density or species composition. Con-
versely, if the distribution of small
mammals is primarily a function of
population density, then habitat dis-
turbance is likely to'have less effect,
or at least a less direct effect, on local
populations. Furthermore, if mi-
crohabitat requirements are known,
it might be possible to manipulate
population levels by altering struc-
tural components of the habitat.

I measured use of microhabitat by
small mammals in an old-growth
northern hardwoods (Acer-Fagus-Bet-
ula) forest, a habitat that contains a
variety of microhabitats (Bormann
and Likens 1979) and supports sev-
eral species of small mammals (Love-
joy 1970). In this paper I identify the
small mammal-microhabitat associa-
tions observed, compare them to re-

sults of previous studies, and suggest
that the distribution of small mam-
mals among microhabitats in the
northern hardwoods forest is influ-
enced little by structural features of
the forest.

Methods

The study area was located in the
White Mountain National Forest,
New Hampshire in a topographically
isolated site known as the Bowl
(Martin 1977). All fieldwork was con-
fined to the uncut, old-growth north-
em hardwoods forest that comprises
about 210 ha in the lower (600-750 m)
elevations of the Bowl. The old-
growth forest is structurally hetero-
geneous; numerous treefalls and
gaps in the canopy are present and
the portion of the forest floor cov-
ered by rock, soil, water, or vegeta-
tion varies greatly across the stand
(Gore 1986).

Trapping

In 1983, small mammals were live-
trapped on three 60 x 105-m grids,
each consisting of 40 trapping sta-
tions spaced at 15-m intervals along
five rows. Two stations were added
to each grid in 1984 to increase sam-
pling at seeps and along streanns. A
fourth grid of 42 trapping stations
was also established in 1984. This
grid was used to evaluate the robust-

319

ness of models of microhabitat use
that were developed with data from
the initial three grids.

One Sherman (5 x 6 x 17 cm),
Pymatuning (Tyron and Snyder
1973), and pitfall trap were located at
each trapping station (Gore 1986).
Traps were baited with sunflower
seeds and set simultaneously for four
consecutive days each month from
August to October, 1983 and June to
October, 1984. Each trapping period
began 3-5 days after a new moon (to
minimize variation in ambient light)
and continued regardless of weather
conditions. Captured animals were
marked and released at the capture
site. In 1983, 4,320 trap-nights (12
nights X 120 stations x 3 traps) were
recorded on three grids. In 1984 four
grids provided 10,080 trap-nights (20
nights X 168 stations x 3 traps).

combinations of variables to be incor-
porated, but at the risk of complicat-
ing the interpretation of results.

Data Analysis

Chi-square tests (Snedecor and Co-
chran 1980) were used to identify sta-
tistically significant associations be-
tween capture locations of all pos-
sible pairs of species. Within species,
differences between capture locations
in different years and seasons were
analyzed. The relative strength of as-
sociations was measured via the con-
tingency correlation coefficient. Phi
(Brown 1983).

Habitat variables from plots with
and without captures in 1984 were
compared for each small mammal
Sf)ecies. Capture locations from the

entire trapping period were grouped
even though some values, such as for
vegetative cover, varied between sea-
sons. If the habitat values within the
two groups were normally distrib-
uted and had equal variances, signifi-
cant differences between groups
w6re determined via t-tests; if not,
Mann-Whitney tests were used
(Snedecor and Cochran 1980).

Logistic regression (Bishop et al.
1975, Engelman 1983) was used to
identify, for each mammal species,
the microhabitat variables that ac-
counted for statistically significant
portions of the variation in capture
success. The product of the analysis
is a set of regression equations for
predicting presence of each species at
a station based upon quantitative
measures of the station's habitat
characteristics. Logistic regression

Microhabitat

At each trapping station microhabitat
was quantified by measures from
within the 15 x 15-m plot in which
the station was centered. The vari-
ables used to quantify the habitat are
defined in table 1 and the methods
for measuring them are described in
detail by Gore (1986).

The 26 habitat variables selected
for analysis were a subset of a larger
group of variables on which meas-
ures were made. The number was
reduced in order to facilitate inter-
pretation of results and to increase
the ratio of sample cases to variables
(Morrison 1985). The initial group
was condensed by deleting one of
each pair of highly correlated
(r>0.50) variables that were similar in
ecological form or function. Some
highly correlated variables, such as
the number of dead trees and the
number of logs, were retained be-
cause I felt they represented distinct
ecological features. Variables were
included regardless of whether their
means or distributions varied be-
tween capture and no-capture sta-
tions. This allowed significant linear

Table 1.— Names and definitions of 26 habitat variables measured at each
trapping station in 1984.

Name

Definition

SLOPE angle of ground from horizontal (% of 90 )

NTREES number of trees (stems > 1 0 cm diameter)

BATREE basal area (m^ of trees

DEDTREE number of dead trees

NSTUMPS number of stumps (dead trees < 1.5 m high)

TREEDIS mean distance (m) from trees to station center

N$HRBL3 number of shrubs/saplings (i.e. stems <10 cm diameter)

with diameter <3 cm (measured 10 cm above ground)

NSHRB36 number of shrubs/saplings with diameter 3-6 cm

NSHRBG6 number of shrubs/saplings with diameter 6-1 0 cm

NLOGS number of logs (stems >10 cm diameter)

LOGDIST mean distance (m) from logs to station center

AVGVOLA mean volume (m^) of logs in class of least decay

AVGVOLB mean volume of logs in class of moderate decay

AVGVOLC mean volume of logs in class of advanced decay

DVEG relative cover (%) by vegetation <0.5 m above ground

CVEG relative cover by coniferous vegetation <0.5 m above

ground

ROCK relative cover by rocks >0.5 m^

SOIL relative cover by exposed soil

WATER relative cover by water

VEG52 relative cover by vegetation 0.5-2 m above ground

CVEG52 relative cover by coniferous vegetation 0.5-2 m

VEGT2 relative cover by vegetation >2 m above ground

CVEGT2 relative cover by coniferous vegetation >2 m

AVGLTR mean depth (cm) of leaf litter

AVGHUMS mean depth (cm) organic soil (humus)

AVGSHR mean horizontal sheer strength (kg/m) of soil

320

was used instead of discriminant
analysis because it does not assume
that independent or explanatory
variables are normally distributed or
have homogeneous variances (Press
and Wilson 1978).

I used the BMDP-LR computer
program (Engelman 1983) to con-
struct regression equations, or mod-
els, for predicting microhabitat use
as defined by captures. The program
selected, in a stepwise manner, habi-
tat variables that distinguished sta-
tions where animals were captured
from those where they were not.
Variables were entered into an equa-
tion if their F value was significant at
P<0.10 and removed if P subse-
quently exceeded 0.15.

Initial regression models were
formed using data obtained in 1984
from Grids 1-3. The regression mod-
els for each mammal species were
used to classify stations within the
three grids as locations where the
species was either present or absent.
The models were then used to pre-
dict the presence of each sp)ecies at
stations in Grid 4. Finally, new re-
gression models were developed for
each species using data from all four
grids. These were compared to the
models from the initial three grids in
order to assess the effect of different
sites on the models. The Kappa sta-

tisHc (Reiss 1973, Engelman 1983)
determined the significance of agree-
ment between locations where a spe-
cies was observed and the species-
present locations predicted by the
regression models.

Results

Trapping

Thirteen species of small mammals
were captured during the study, but
only those captured more than ten
times are considered. The number of
captures varied widely among spe-
cies and, for some species, between
years (table 2). The smokey shrew
(Sorex fumeus), pygmy shrew (S.
hoyi), and eastern chipmunk (Tamias
striatus) were captured infrequently
in both 1983 and 1984, while the
masked shrew (S. cinereus) and the
southern red-backed vole (Clethriono-
mys gapperi) were common in both
years. The northern short-tailed
shrew (Blarina brevicauda), deer
mouse (Peromyscus maniculatus) and
woodland jumping mouse (Na-
paeozapus insignis) were captured
more often in 1984 than in 1983.

Captures from August through
October at the 120 stations trapped in
both years were compared for each

Table 2.— Number of individuals captured, total captures, and individuals
captured per TOO trap-nights (TN) for eight small mammal species in 1983
and 1984.

Number of captures

_ 1984

Indivi- /TOO Indlvl- /1 00

Species duals Total TN duals Total TN

Short-tailed Shrew

5

5

0.2

160

170

2.4

Masked Shrew

32

32

1.1

69

73

1.0

Smokey shrew

3

3

0.1

8

9

0.1

Pygmy shrew

6

6

0.2

4

4

<0.1

Eastern chipmunk

6

7

0.2

22

26

0.3

Deer mouse

41

61

1.4

303

671

4.5

Red-backed vole

25

30

0.9

64

87

1.0

Jumping mouse

71

92

2.5

305

494

4.5

species to determine whether 1983
and 1984 capture locations were as-
sociated. For all but two species, in-
dividuals were captured in both
years at few (0-8 percent) of the sta-
tions with captures. No species
showed a significant association in
capture locations between years (X^
tests, P>0.05). Even for deer mice and
jumping mice, which were abundant
in both years, stations with captures
in both years comprised only 32-41
percent of all stations with captures
of these species.

A similar comparison was made
between the capture locations of each
species in summer (June- August) and
fall (September-October) of 1984 on
all four grids. No species exhibited a
significant (P>0.05) association be-
tween summer and fall locations.

For all trapping periods and spe-
cies combined, each trapping station
had at least two captures. In 1984, all
168 stations recorded at least one
capture and 85 percent had more
than four captures. The maximum
number of captures at a single station
was 22 for all species combined and
15 for a single species, the jumping
mouse. Only for deer mice and jump-
ing mice did stations with multiple
captures outnumber stations with
single captures. All other species
were taken only once or not at all at
the majority of trapping stations.

The only pairs of species captured
at the same locations more often than
expected by chance (X^ tests, P<0.05)
were jumping mouseired-backed
vole, short-tailed shrew:masked
shrew, and short-tailed
shrew:smokey shrew. The associa-
tion between each pair was positive
and weak (0.15<Phi<0.30).

Some animals died after capture,
but the effect upon local populations
of each species was unknown. For
abundant species that experienced
low mortality, such as the deer
mouse and jumping mouse, the effect
was probably negligible. For the
shrews, which had high mortality
rates during capture, the effect may
have been substantial. However, cap-

321

hire rates indicated no adverse ef-
fects on shrew abundance. In 1984
more shrews of each species were
captured in September than in the
previous three months. Furthermore,
of the four shrew species, only cap-
tures of pygmy shrews decUned be-
tween years (table 2).

Microhabitat Use

Differences between habitat values
for the capture and no-capture sta-
tions of Grids 1-3 from 1984 were
compared. For each species, at least
one habitat variable differed signifi-
cantly between stations with and
without captures (table 3).

Logistic regression is not useful if
the number of cases of either of the
dependent variable values (species
presence or absence) is less than five
percent of the total number of cases
(D. Hosmer, University of Massachu-
setts, personal communication). Be-
cause the pygmy shrew was found at
only two percent of the stations, it
was deleted from the analysis. Deer
mice and smokey shrews each also
had widely disparate group sizes, 95
percent present and 95 percent ab-
sent respectively; therefore, results
for these species should be consid-
ered cautiously.

The first set of logistic regression
models of microhabitat use were
based on captures in 1984 from the
126 stations in Grids 1-3. The number
of significant variables included in
each model ranged from one, when
presence of red-backed voles was the
dependent variable, to 12, when
presence of eastern chipmunks was
used (table 4). For most species the
variables included in the regression
models were not the same as those
whose means differed between cap-
ture and no-capture stations (table 3).
This suggests that some linear combi-
nation of habitat variables was im-
portant in defining the microhabitat
where a species was captured, even
though individual variables alone
were not.

The habitat variables included in
the logistic regression models (table
4) were selected because each was
associated with a significant (P<0. 10)
portion of the variance in the capture
data of a species. However, if these
variables and their regression coeffi-
cients cannot be used to correctly
predict the capture success of a spe-
cies at a station, they are of limited
practical value regardless of their sta-
tistical significance. To assess the
utility of regression models as de-
scriptors of microhabitat, they were
independently used to classify each
trapping station, based on habitat
pararneters, as one with the species
present or absent (table 5).

Tests of the agreement between
predicted and observed capture suc-
cess were not possible for the
smokey shrew because no sites were
classified as having the species pres-
ent. The regression model was not
able to identify, based upon the habi-
tat variables measured, the eight sta-

tions that captured smokey shrews.
Conversely, nearly all stations were
predicted to capture deer mice and
jumping mice (table 5). The regres-
sion models for those two species
were unable to distinguish those sta-
tions where the animals were not
captured.

For the other four species the
numbers of stations with and with-
out captures were more similar and,
consequently, so were the number of
species-present and species-absent
classifications. For red-backed voles,
however, only 47 percent of the clas-
sifications of present were correct
and this was not significantly better
than chance (table 5). (Dnly for the
eastern chipmunk, short-tailed
shrew, and masked shrew were the
regression models able to classify
capture success at a level better than
chance agreement. For these three
species, the logistic regression mod-
els may be useful descriptors of the
microhabitat used within Grids 1-3.

Table 3.— Means (SE) of habitat variables that differed signlficqntly be-
tween stations with and without captures of each species in Grids 1-3 In
1984.

Habitat

Stations with

Stations without

Species

Variable'

captures

captures

Short-tailed shrew

NSHRBL3

188(10)

158(15)

<0.05

CVEGT2

0.2(0.1)

0.0

<0.05

VEG52

30.9 (2.0)

24.7 (2.2)

<0.05

Masked shrew

LOGDIST

5.1 (0.1)

5.5 (0.1)

<0.05

AVGVOLA

0.20 (0.06)

0.06 (0.02)

<0.001

DVEG

38.2 (2.4)

27.6(1.3)

<0.001

AVGLTR

3.4(0,1)

3.0 (0.1)

<0.05

DEDTREE

0.63 (0.12)

1.0 (0.1)

<0.05

Smokey shrew

AVGLTR

3.8 (0.3)

3.1 (0.1)

<0.05

Eastern chipmunk

BATREE

0.93 (0.08)

0.77 (0.02)

<0.05

AVGVOLA

0.23(0.10)

0.08 (0.02)

<0.05

VEG52

37.1 (4.5)

26.9(1.6)

<0.05

NSHRBL3

220 (28)

168(9)

<0.05

Deer mouse

DEDTREE

0.85 (0.0)

1.6 (0.42)

<0.05

NSHRBG6

7.8 (0.4)

10.8(1.2)

<0.05

AVGVOLB

0,23 (0.03)

0.51 (0.14)

<0.05

Red-backed vole

DVEG

36.9 (2.3)

27.7(1.4)

<0.001

Jumping mouse

VEGT2

75.4 (0.9)

70.2 (2.6)

<0.05

'Definitions ofhobifaf variables are given in table h
'Probability that capture groups have equal means.

322

Because of the large number of
variables included in the regression
model for the eastern chipmunk
(table 4), it was difficult to concisely
describe the microhabitat of this spe-
cies. Briefly, the eastern chipmunk
was associated with large trees
(+BATREE), downed wood
(+NLOGS, -LOGDIST, +AVGVOLB,
+AVGVOLC), and dense vegetation
taller than 0.5 m (+VEG52, +VEGT2,
+NSHRB36).

The microhabitat of the short-
tailed shrew has fewer variables but
is also difficult to characterize. Cap-
tures were negatively associated with

numbers of medium-sized shrubs
and with vegetative cover between
0.5 and 2 m above ground. The
masked shrew was found at stations
with numerous logs and dense vege-
tation <0.5 m tall. The pxDsitive asso-
ciation with slightly decayed logs
and dense ground cover and the
negative association with standing
dead trees suggest that recent
treefalls may provide good habitat
for masked shrews.

To be useful predictors of species
microhabitat, regression models
should be successful with data that
are indef>endent of those from which

the models were formed. To test site-
specificity of regression models, they
were applied to data from 42 stations
in Grid 4. Unlike the other three
grids. Grid 4 had a perennial stream
running through it, two extensive
canopy gaps from recent treefalls,
and highly variable soil conditions.

Regression models from each of
the seven species were used to clas-
sify the stations in Grid 4 according
to capture success. None of the clas-
sifications, even those of the eastern
chipmunk, short-tailed shrew, and
masked shrew were correct more of-
ten than expected due to chance (for
all tests Kappa <0.48, P>0.05).

Because none of the regression
models were useful in predicting
capture locations in Grid 4, data
from all four grids were combined
and new logistic regression models
for predicting species presence were
developed to determine the influence
of data from Grid 4 (table 7). Deer
nnice, jumping mice, and smokey
shrews again had widely disparate
group sizes and the models could not
correctly classify the stations in the
less common group (table 7).

Agreement between observed and
predicted locations was significant
for red-backed voles as well as east-
em chipmunks, short-tailed shrews,
and masked shrews (table 7), which
had significant models earlier. Some
of the variables included in the mod-
els of each species (table 6) were dif-
ferent from those included when
only data from Grids 1-3 were used
(table 4). For the masked shrew, east-
ern chipmunk, and red-backed vole
the regression models created with
and without the data from Grid 4
were similar, even though the predic-
tions from Grids 1-3 for the red-
backed vole were not better than ex-
pected by chance. The coefficients
changed but most variables were the
same. This suggests that for these
three species the microhabitats in
Grid 4 were similar to those identi-
fied in the other three grids. The rela-
tionship of species to microhabitat
parameters may not be as sensitive

Table 4.— Logistic regression models for predicting presence of small
mommcd species based on data collected In Grids 1 -3 in 1 984.

independent

Regression

Coefficient/

Species

Variable'

Coefficient

Standard Error

Short-tailed shrew

N$HRB36

-0.062

-1J84

VEG52

-0.063

2.059

CONSTANT

0.487

0.907

Masked shrew

DEDTREE

-0.704

-2.538

LOGDIST

-O.660

-2.577

AVGVOLA

2.621

2.212

DVEG

0J37

3.273

CONSTANT

0.784

0.566

Smokey shrew

AVGLTR

1.514

2.631

CONSTANT

-7.869

-3.651

Eastern chipmunk

NTREES

-0.210

-1.431

BATREE

2.523

1.957

NSHRB36

0.123

1.457

NSHRBG6

-0.447

-2.831

NLOGS

0.188

2.171

LOGDIST

-0,702

-1.651

NSTUMPS

-1,706

-1.272

AVGVOLB

3.031

2.584

AVGVOLC

2.305

2.560

VEG52

0.052

1.187

VEGT2

0.086

1.772

AVGSHR

-0.102

-2.369

CONSTANT

-3.238

-0.965

Deer mouse

NLOGS

-0.132

-2.033

AVGVOLB

-1.155

-1.586

CONSTANT

4.543

4.804

Red-backed vole

DVEG

0.112

3.230

CONSTANT

-2.450

-4.415

Jumping mouse

CVEG52

-0.512

-1.833

VEGT2

0.070

2.120

CONSTANT

-2.544

-1.205

'See table 1 for definition oftiabifat variables.

323

as the regression models suggest.
This would account for the poor per-
formance of the models from Grids
1-3 in predicting captures on Grid 4.

For short-tailed shrews the regres-
sion model changed greatly when
data from Grid 4 were included.
Four new variables were added, and
the sign of the coefficient was re-
versed on the only variable, vegeta-
tion between 0.5 and 2 m, that was
retained. This suggests that captures
of short-tailed shrews or the meas-
ured habitat parameters poorly re-
flect the microhabitat requirements
of the species, or that short- tailed
shrews are not restricted by mi-
crohabitat within this forest.

Discussion

In an environment of limited re-
sources, sympatric species are ex-
pected to partition resources as a
means of coexisting, i.e. avoiding
competitive exclusion (Schoener
1974). Since Brown (1973) first sug-
gested that temperate forest rodents
would be likely to partition habitat
rather than seasonally variable food
supplies, numerous studies in north-
ern temperate forests have identified
statistically significant associations
between habitat structure and small
mammal distributions (Dueser and
Shugart 1978, Kitchings and Levy
1981, Parren 1981, Vickery 1981,
Schloyer 1983, Seagle 1985a).

Statistical significance, however,
does not necessarily impart biologi-
cal meaning to observed patterns of
species distributions. Few authors
have tested the biological relevance
of their models of microhabitat use
by using them to predict microhabi-
tat use at independent locations or
times. Parren and Capen (1985)
found that capture locations of deer
mice could not be accurately pre-
dicted using discriminant functions
of microhabitat use developed with
data from similar habitats the previ-
ous year. Similarly, none of the logis-
tic regression models I developed

were useful in predicting capture lo-
cations at stations other than those
from which the models were devel-
oped.

One reason for the poor predictive
capabilities of the multivariate mod-
els may be that trapping does not ac-
curately portray the relationship be-
tween species presence and habitat
requirements. In addition, the way in
which habitat features are measured
may not depict the variability per-
ceived by small mammals or the
variation in microhabitat structure
may be small relative to the niche
breadth of each species. Unfortu-
nately, these problems are not easily
identified or solved. Ideally, the ac-
tivity of many individual animals
would be intensively monitored, but
that is very difficult to accomplish.

Another reason for the poor per-
formance of the models is that prob-
lems in applying the multivariate
analyses, such as disparate sizes of
presence and absence groups and
multicollinearity of variables, make it
difficult to interpret the results of
habitat models (Noon 1984). The
scale at which habitat and small
mammals are sampled also greatly

affects the relationship that can be
defined (Morris 1984).

Despite these potential limitations,
I believe the inability of my models
to predict species presence on a inde-
pendent grid in the same forest stand
suggests that structural features
alone, at least at the microhabitat
level, are not important to the distri-
bution of small mammals. Compari-
sons of capture locations and review
of habitat requirements for each spe-
cies supports my argument.

The locations where species were
captured suggest that no interspecific
segregation of microhabitats oc-
curred. Overlap in capture sites was
high among species and no inverse
relationships were observed, even
when data were examined by season.
This suggests that habitat partition-
ing or microhabitat selection is ab-
sent or operating at a finer scale than
my trap stations.

The weak association I found
among capture locations of each spe-
cies between years and seasons sug-
gests that individual species were not
selecting particular trapping stations.
It is possible that subtle shifts in the
microhabitat used would not be per-

Table 5.— Classification of 126 trapping stations in Grids 1 -3 as locations
where each of eight small mammal species Is present or absent based on
logistic regression models, and agreement between predicted and ob-
sen^ed classifications.

No. of stations

classified^

% Correct^

Agreement^

Present

Absent

Present

Absent

K

ASE

P

Short-tailed shrew

104

22

64

64

0.185

0.079

<0.025

Masked shrew

23

103

65

81

0.381

0.093

<0.001

Smokey shrew

0

126

0

94

Eastern chipmunk

12

114

75

92

0.648

0.114

<0.001

Deer mouse

125

1

94

0

0.014

0.013

NS

Red-backed vole

15

111

47

72

0J13

0.084

NS

Jumping mouse

124

2

87

50

0.079

0.090

NS

'Prior probability of presence = 0.05.

'Percent ofstatioris wt)ere present/absent classificotion agreed withi observations
^om trapping in 1984,

= Kappa statistic (Fleiss 1973). ASE = asymptotic standard error. P = ProbabiHty
tttat agreement is due to ctiance. i.e. K=0. NS = not significant >0.05.

324

ceived by examination of trapping
locations alone. The niicrohabitat oc-
cupied by sn-iall mammals has been
reported to shift with season (Kitch-
ings and Levy 1981, Vickery 1981),
population density (M'Closkey 1981,
Adler 1985), and species composition
(Seagle 1985b). This suggests that mi-
crohabitat use is dynamic, regardless
of whether the shifting is determinis-
tic or stochastic.

Another argument against differ-
ential use of microhabitats by the
small mammals I observed is the va-
riety of habitats they occupy. The

eight species I found in the old-
growth northern hardwoods forest
have been found in other age- and
size-classes of northern hardwoods
forests as well as in other forest types
(Lovejoy 1970, Richens 1974,
Kirkland 1977, Miller and Getz 1977,
Hill 1981, and others). Except for
smokey shrews and pygmy shrews,
the sf)ecies are common in a variety
of habitats comprising a wide range
of structural characteristics. In fact,
descriptions of the important habitat
features associated with each species
do not always agree [e.g. see Hamil-

Table 6.— Logistic regression models for predicting presence of small
mammal species based on data collected In Grids 1 -4 In 1 984.

Species

Independent
Variable'

Regression
Coefficient

Coefficient/
Standard Error

Short-tailed shrew

Masked shrew

Eastern chipmunk

Deer mouse

Red-backed vole

Jumping mouse

DEDTREE

DVEG

ROCK

VEG52

AVGLTR

CONSTANT

DEDTREE

LOGDIST

DVEG

AVGHUMS

CONSTANT

NSHRBG6

NLOGS

NSTUMPS

AVGVOLB

AVGVOLC

VEGT2

AVGLTR

CONSTANT

DEDTREE

NSHRB36

AVGVOLB

AVGHUMS

CONSTANT

DEDTREE

DVEG

CONSTANT

LOGDIST

AVGVOLB

SOIL

CONSTANT

-0.355
0.053
0,207
0,043
0.809
-3,463
-0.401
-0.810
0.159
0.499
-0.475
-0.280
0,125
-1.761
2.085
1.042
0.072
0.675
-9.267
-0.934
-0.170
-1.064
-0.794
9763
-0.321
0.083
-1.841
0.475
-0.869
-0.192
-0.140

-2.157
2.275
2,721
1.826
-3.198
-3.392
-2.029
-3.415
5.348
2.467
-0,336
-2.825
1.979
-1.639
2.762
1.874
2.088
1.867
-3.506
-2.511
-2.366
-1.599
-2.165
4.248
-1.849
3.778
-4.365
1.742
-1,815
-2.351
-0.098

'See table 1 for deftiifion of habitat variables.

ton (1941), Brower and Cade (1966),
Lovejoy (1970), Vickery (1981), and
Parren (1981) for descriptions of
jumping mouse habitat]. If each spe-
cies is common under a wide range
of habitat conditions, it seems un-
likely that they would partition or
select habitat based on the advan-
tages of structural features alone.

Fine discrimination of the forest
habitat seems more improbable when
the temjx)ral variability of mi-
crohabitats is considered. Within the
northern hardwoods forest of New
Hampshire microhabitats are greatly
modified in winter by deep snow
cover, in summer by closed canopies
and sparse ground cover, and in fall
by deep leaf litter. Therefore, resi-
dent species must accommodate sea-
sonally variable microhabitats as well
as seasonally variable food supplies.
The reasoning Brown (1973) used to
suggest that temperate forest rodents
could not specialize on seasonally
variable food resources seems appli-
cable also to seasonally variable mi-
crohabitats.

In the forest I sampled, presence
of most species at individual trap-
ping stations could not be accurately
predicted based on structural fea-
tures of the habitat. If microhabitat
structure does not greatly influence
the distribution of small mammals
within this forest type, disturbance of
the habitat should not directly affect
population levels. However, the scale
at which the disturbance occurs may
determine to what extent local popu-
lations are affected. Small scale dis-
turbance of the habitat, such as har-
vesting by single-tree or small-group
selection, would likely not affect Sf)e-
cies composition or density of the
resident small mammals. More wide
scale disturbance, such as clear-cut-
ting of the entire forest stand, might
alter the habitat so greatly that spe-
cies abundance and distribution is
affected (Kirkland 1977). Given the
apparent wide range of habitat con-
ditions in which these small mam-
mals occur, even a large scale distur-
bance of the northern hardwoods

325

forest would likely cause only tem-
porary changes in species composi-
tion or population levels of small
mammals.

The relationship between small
mammals and habitat structure
within the northern hardwoods for-
est remains poorly understood.
However, the data presented here, as
well as comparisons at different
scales (Morris 1984, 1987), suggest
that microhabitat features play only a
minor role in the distribution of
small mammals within the forest. A
more important determinant of small
mammal distribution may be popula-
tion size and the factors that affect it,
such as food, weather, and preda-
tors. Consequently, models for pre-
dicting the distribution of small
mammals within the northern hard-
woods forest will likely remain un-
successful until factors that affect
population size are included. The
temporal and spatial scales at which
these factors influence distribution
must also be addressed.

Acknowledgments

I thank C. D. Warren, J. M. Prince,
and S. L. Crane for assistance with
fieldwork. Staff of the U.S. Depart-
ment of Agriculture, Forest Service
(USPS) facilitated work in the White
Mountain National Forest. W. A. Pat-
terson III, R. M. DeGraaf, S. L. Gar-
man, S. W. Seagle, M. W. Sayre, D. P.
Snyder, and an anonymous reviewer
provided helpful comments on ear-
lier drafts of the manuscript. This re-
search was supported by grants from
the USPS and the Mclntire-Stennis
Cooperative Forestry Research Pro-
gram (Grant No. MS-47).

Literature Cited

Adler, Gregory H. 1985. Habitat se-
lection and species interactions: an
experimental analysis with small
mammal populations. Oikos
45:380-390.

Bishop, Yvonne M. M., Stephen E.
Feinberg, and Paul W. Holland.

1975. Discrete multivariate analy-
sis: theory and practice. 557 p.
MIT Press, Cambridge, MA.

Bormann, F. Herbert, and Gene E.
Likens. 1979. Pattern and process
in a forested ecosystem. 253 p.
Springer- Verlag, New York, New
York.

Brower, J. E., and T. J. Cade. 1966.
Ecology and physiology of Na-
peozapus insignis (Miller) and other
woodland mice. Ecology 47:46-63.

Brown, James H. 1973. Species diver-
sity of seed-eating desert rodents
in sand dune habitats. Ecology
54:788-797.

Brown, M. B. 1983. Frequency tables,
p. 143-206. In BMDP Statistical
Software 1983. W. L. Dixon, edi-
tor. University of California Press,
Berkeley.

Crowell, Kenneth L., and S. L. Pimm.

1976. Competition and niche shifts
of mice introduced on small is-
lands. Oikos 27:251-258.

Dueser, Raymond D., and H.H.
Shugart, Jr. 1978. Microhabitats in
a forest-floor small mammal
fauna. Ecology 59:89-98.

Engelman, Laszlo. 1983. Stepwise lo-
gistic regression, p. 330-344. In
BMDP statistical software 1983.
W.L. Dixon, editor. University of
California Press, Berkeley.

Heiss, J. L. 1973. Statistical methods
for rates and proportions. 223 p.
John Wiley and Sons, New York,
New York.

Gore, Jeffery A. 1986. Small mam-
mals and habitat structure in an
old-growth northern hardwoods
forest. Ph.D. Dissertation. 135 p.
University of Massachusetts,
Amherst.

Hamilton, W.J., Jr., and David B.
Cooke. 1941. Small mammals and
the forest. Journal of Forestry
38:468-473.

Hill, Barbara J. 1981. Small mammal
habitat associations in selected
timber community types on the
White Mountain National Forest,
New Hampshire. M.S. Thesis. 140
p. University of Massachusetts,
Amherst.

Kirkland, Gordon L., Jr. 1977. Re-
sponses of small mammals to the
clearcutting of northern Appala-
chian forests. Journal of Mam-
malogy 58:600-609.

Table 7.— Classification of 1 68 trapping stations in Grids ] -4 as locations
where each of seven small mammal species Is present or absent based on
logistic regression models, and agreement between predicted and ob-
sen^ed classifications.

No. of stations

Species

classified^

% Correct*

Agreement'

Present Absent

Present Absent

K

ASE P

Short-tailed shrew

106

62

71

68

0.371

0.072 <0.001

Masked shrew

45

123

67

78

0.412

0.075 <0.001

Smokey shrew

2

166

100

96

0.351

0.183 NS

Eastern chipmunk

7

161

71

90

0.314

0.116 <0.01

Deer mouse

167

1

95

0

0.011

0.010 NS

Red-backed vole

27

141

57

72

0.223

0.076 <0.01

Jumping mouse

165

3

88

62

0.133

0.093 NS

'Prior probability of presence = 0.05.

'Percent of stations wtiere present/ absent classificotion agreed withi observations
frorr) trapping in 1984.

= Kappa statistic (Fleiss 1973). ASE = asymptotic standard error. P = Probability
tt>at agreement is due to ctiance, i.e. K=0. NS = not significant

326

Kitchings, J. Thomas, and Douglas J.
Levy. 1981. Habitat patterns in a
small mammal community. Jour-
nal of Mammalogy 62:814-820.

Lovejoy, David A. 1970. The effect of
logging on small mammal popula-
tions in the White Mountain Na-
tional Forest of New Hampshire.
Ph.D. Dissertation. 173 p. Univer-
sity of Connecticut, Storrs.

Martin, C. Wayne. 1977. Distribution
of tree species in an undisturbed
northern hardwood-spruce-fir for-
est, the Bowl, New Hampshire.
USDA Forest Service Research
Note NE-244, Northeastern Forest
Experiment Station, Broomall,
Pennsylvania.

M'Closkey, Robert T. 1981. Mi-
crohabitat use in coexisting desert
rodents — the role of population
density. Oecologia 50:310-315.

Miller, Donald H., and Lowell L.
Getz. 1977. Factors influencing lo-
cal distribution and species diver-
sity of forest small mammals in
New England. Canadian Journal
of Zoology 55:806-814.

Morris, Douglas W. 1984. Patterns
and scale of habitat use in two
temperate-zone, small mammal
faunas. Canadian Journal of Zool-
ogy 62:1540-1547.

Morris, Douglas W. 1987. Tests of
density-dependent habitat selec-
tion in a patchy environment. Eco-
logical Monographs 57:269-281.

Morrison, Michael L. 1985. Influence
of sample size on discrinninant
function analysis of habitat use by
birds. Journal of Field Ornithology
55:330-335.

Noon, Barry R. 1984. Summary: Bi-
ometric approaches to modeling —
the researcher's viewpoint, p. 197-
201. In Wildlife 2000: Modeling
habitat relationships of terrestrial
vertebrates. Jared Verner, Michael
L. Morrison, and C. John Ralph,
editors. The University of Wiscon-
sin Press, Madison.

Parren, Steven G. 1981. Habitat selec-
tion by small mammals in a north-
ern hardwoods forest. M.S. Thesis.
77 p. University of Vermont, Burl-
ington.

Parren, Steven G., and David E.
Capen. 1985. Local distribution
and coexistence of Peromyscus in
Vermont. Journal of Mammalogy
66:36-44.

Porter, John H., and Raymond D.
Dueser. 1982. Niche overlap and
competition in an insular mammal
fauna: a test of the niche overlap
hypothesis. Oikos 39:228-236.

Press, S. James, and Sandra Wilson.
1978. Choosing between logistic
regression and discriminant analy-
sis. Journal of the American Statis-
tical Association 73:699-705.

Richens, Voit B. 1974. Numbers and
habitat affinities of small mam-
mals in northwestern Maine. Ca-
nadian Field-Naturalist 88:191-196.

Schloyer, Carroll R. 1983. Ecological
relationships of sympatric Pero-
myscus in western Massachusetts.
Ph.D. Dissertation. 149 p. Univer-
sity of Massachusetts, Amherst.

Schoener, Thomas W. 1974. Resource
partitioning in ecological commu-
nities. Science 185:27-39.

Seagle, Steven W. 1985a. Patterns of
small mammal microhabitat utili-
zation in cedar glade and decidu-
ous forest habitats. Journal of
Mammalogy 66:22-35.

Seagle, Steven W. 1985b. Competi-
tion and coexistence of small
mammals in an east Tennessee
pine plantation. American Mid-
land Naturalist 114:272-282.

Snedecor, George W., and William G.
Cochran. 1980. Statistical methods.
507 p. Iowa State University Press,
Ames.

Tyron, C. A., and D. P. Snyder. 1973.
Biology of the eastern chipmunk,
Tamias striatus: life tables, age dis-
tributions, and trends in popula-
tion numbers. Journal of Mam-
malogy 54:145-168.

Vickery, W. L. 1981. Habitat use by
northeastern forest rodents.
American Midland Naturalist
106:111-118.

327

The Value of Rocky Mountain
Juniper (Juniperus
scopulorum) Woodlands in
South Dakota as Small
Mammal Habitat'

Abstract.— Small mammals and vegetation were
sampled over two years in Rocky Mountain juniper
woodlands and adjacent grasslands in South
Dakota. Juniper woodlands provided specialized
habitat for two woodland species, white-footed
mice and bushy-tailed woodrats, and attracted a
number of species generally associated with
grasslands.

Carolyn Hull Sieg^

Native woodlands constitute only a
small percentage of the total land
area in the Northern Great Plains, yet
they provide critical habitat for many
wildlife species. Isolated woodlands
provide a sharp contrast with adja-
cent grasslands, increasing available
cover, vertical structure, and habitat
interspersion, and, hence, the num-
ber of potential niches available for
wildlife. Research on the value of na-
tive woodlands as wildlife habitat
has focused mainly on wildlife use of
deciduous woodlands (Faanes 1984,
Gaines and Kohn 1982, Hopkins et al.
1986, Uresk 1982), although the im-
portance of Rocky Mountain juniper
woodlands for mule deer (Odocoileus
hemionus) has been documented (Sev-
erson 1981, Severson and Carter
1978). Information on small mam-
mals associated with Rocky Moun-
tain juniper stands is limited to brief
studies conducted in North Dakota
(Hansen et al. 1980, Hopkins 1983,
Seabloom et al. 1978).

Native woodlands in the Northern
Great Plains are limited to areas of
increased moisture, such as along
streams and rivers, and to areas with

'Paper presented af symposium. Man-
agement of Amphibians, Reptiles, and
Small Mammals in Nortti America. (Flag-
staff, AZ, July 19-21. 1988.)

'Carolyn Hull Sieg is Research! Wildlife
Biologist, USDA Forest Service, Rocky Moun-
tain Forest and Range Experiment Station,
located at Rapid City, Southi Dakota.
Headquarters is in Fort Collins, in associa-
tion with) Colorado State University.

increased topographic variation.
Rocky Mountain juniper is restricted
to areas of steep topography, such as
the "Badlands" of North and South
Dakota, the Black Hills, areas along
drainageways of major rivers, and
areas on high limestone plateaus in
South Dakota and Wyoming. It is
more likely to occur on steep, north-
facing slopes, and is often associated
with soils that are calcareous, poorly
developed, and shallow (Powells
1965).

The purpose of this study was to
characterize small mammal species
composition and distribution in
Rocky Mountain juniper woodlands
and in adjacent mixed-grass range-
lands in the Badlands National Park,
southwestern South Dakota. The ob-
jectives were to determine if the pres-
ence of isolated juniper woodlands
increased mammal species richness
of the area, and to form preliminary
hypotheses as to how these wood-
lands function as small mammal
habitat.

Study Area and Methods

The study area is in Pennington
County, South Dakota, approxi-
mately 15 km south of the town of
Wall, in Sage Creek Basin, Badlands
National Park. Elevation ranges from
950 to 1000 m above sea level. An-
nual precipitation averages 36 cm,
most of which is received in May,
June, and July. The terrain in Bad-

lands National Park is typically
rough and irregular, with steep
bluffs rising above floodplains onto
upland grasslands. Dense stands of
Rocky Mountain juniper occur on
steep, north-facing slopes and in
draws. Upland grasslands are domi-
nated by western wheatgrass
(Agropyron smithii), green
needlegrass (Stipa viridula),
buffalograss (Buchloe dactyloides), and
blue grama (Bouteloua gracilis).

Eight study sites were established,
four in Rocky Mountain junijjer
woodlands on north-facing slopes in
draws, and four on adjacent grass-
lands. Vegetation and relative abun-
dance of small mammals were
sampled on a regular basis for 2
years. Plant canopy cover on grass-
lands and understory cover in the
juniper woodlands were sampled in
June and August of both sampling
years. Plant canopy cover, by species,
was estimated in 150, 0.1-m^ quad-
rats spaced at 1-m intervals along
three permanent 50-m transects on
each site (Daubenmire 1959). Over-
story vegetation in Rocky Mountain
juniper study sites was sampled in
eight, 7- by 7-m macroplots spaced at
30-m intervals on each site. Tree den-
sities, heights, diameters (d.b.h.), and
crown heights of all trees were meas-
ured.

Small mammal abundance was
sampled monthly from June through
October in both years. Forty Sherman
live traps, spaced at 10-m intervals
along two permanent 200-m

328

transects, were set on each study site
for four consecutive nights, after one
night of prebaiting. Total trap effort
was 6400 trap nights per vegetation
type per year. Rolled oats mixed

-with peanut butter were used as bait.
Captured animals were identified by
species and assigned a unique 4-digit
number by toe clipping (Taber and
Cowan 1971).

Differences in small mammal
numbers and vegetation between the
two vegetation types were tested
with ref>eated measures analyses of
variance (SPSS 1986). Both years
were combined for analyses. Total
unique small mammals and numbers
of each species were analyzed sepa-
rately: trap session and year were
within-subject factors; vegetation
type was the between-subject factor.
Total plant canopy cover was the
vegetation parameter analyzed: sam-
pling session and year were within-
subject factors; vegetation type was
the between-subject factor. Homoge-
neity of variances was tested with
Bartlett's Box F test; variables with
heterogeneous variances were log-
transformed.

Results

Vegetation

Overstory vegetation in juniper
woodlands was nearly a monocul-
ture of Rocky Mountain junip>er, al-
though an occasional green ash (Frax-
inus pennsylvanica) tree was ob-
served. Tree density averaged 260
trees/ha (+117 SD), and ranged from
an average of 160 to 380 trees/ha on
the four sites. Tree heights ranged
from a mean of 2.8 to 3.1 m, and the
crowns extended nearly to the
ground, averaging approximately 2.3
m in height. The diameters of the ju-
niper trees were small, ranging from
a mean of 4.8 cm to 7.6 cm.

Total plant canopy cover of under-
story vegetation in the juniper wood-
lands was lower (P < 0.01) than on
grasslands. Total cover in the juniper
woodlands averaged 25% (table 1).
Yellow sweetclover (Melilotus offici-
nales) was the most common under-
story plant, then stonyhills muhly
(Muhlenbergia cuspidata) and littleseed
ricegrass (Oiyzopsis micrantha).
Shrubs were uncommon in juniper
woodlands; chokecherry (Prunus vir-
giniam), western wild rose (Rosa
luoodsii), western snowberry (Sym-

Table 1.— Two-year average percent (± SD) canopy cover of dominant
species in four Rocky Mountain juniper woodlands and four grassland
sites, Badlands National Park, South Dakota.

Category

Juniper

Grassland

Total cover

24.5

5.5

52.8 + 4.9

Litter cover

44.6

± 11.3

39.5+ 10.2

Bareground

36.7

±11.8

15.8 + 8.3

Forbs

Yellow sweetclover (Melilotus officinalis)

8.7

+ 4.9

1.2+ 1.0

Russian thistle (Salsola kali)

< 1

±< 1

2.4 + 2.8

Scarlet globemallow (Sphaeralcea coccinea)

< 1

+ < 1

2.9+1.8

Grasses

Western wheatgrass (Agropyron smifhii)

1.4

+ 1.4

9.7 + 7.5

Blue grama (Boufeloua gracilis)

< 1

+ < 1

5.4 + 4.0

Cheoigross (Bromus fecforum)

< 1

±< 1

2.5+1.5

Buffalogross (Buchloe dacfyloides)

< 1

+ < 1

3.4 + 3,5

Threadleof sedge (Carex filifolia)

1.3

+ 2.1

3.3 + 4,2

Sun sedge (Carex heliophila)

< 1

+ < 1

3.0+1.5

Stonyhills muhly (Muhlenbergia cuspidata)

3.4

+ 3.2

0

Littleseed ricegrass (Oryzopsis micrantha)

1.5

+ 1.4

0

Needleondthread grass (Sfipa comafa)

< 1

+ < 1

6.5 + 9.0

Table 2— Two-year average number (+ SD) of small mammals captured
per site in four Rocky Mountain juniper woodlands and four adjacent
grassland sites, BadlarKis National Park, South Dakota.

Species

Juniper

Grassland

Meadow vole (Microtus pennsylvanicus)

<l±l.la

14.8 ±17.2^

House mouse (Mus musculus)

< 1 + 0.4°

0°

Bushy- tailed woodrat (Neotoma cinerea)

1.0± ].2°

0°

Northern grasshopper mouse

(Onychomys leucogaster)

1.0+1.1°

4.9 + 3.3'^

Plains pocket mouse (Perognafhus flavescens)

1.0+ 1.8°

1.0+ 1.8°

Hispid pocket mouse (Perognafhus h'ispidus)

1.0 ±0.9°

2.1 ±2.0°

White-footed mouse (Peromyscus leucopus)

20.4+12.7'^

1.6 + 2.2°

Deer mouse (Peromyscus maniculafus)

44.5 + 24.9°

35.9 ±27,4°

Western harvest mouse

(Reithrodontomys megalotis)

< 1±1°

2.9 ±3.5*^

Thirteen-lined ground squirrel

(Spermophilus tridecemlineafus)

< 1 + 1°

11.0±6.9^

Total

68.5 + 27.4°

74.1+21.9°

'Means in a row followed by the same superscript were not significantly (P > 0. 1)
different

329

phoricarpos occidentalis), and
skunkbush sumac (Rhus aromatica)
each comprised less than 1% of the
total canopy cover. Litter cover in the
juniper woodlands averaged 45%
and bare ground 30%.

Total plant canopy cover on grass-
lands averaged 53% (table 1). West-
ern wheatgrass was the most com-
mon plant species, then
needleandthread grass (Stipa comata),
blue grama, and buffalograss. Scarlet
globemallow (Sphaeralcea coccinea)
was the most common forb. Shrub
species were limited to fringed sage
(Artemisia frigida) and dwarf
sagebrush (A. cam), each comprising
a small percentage of the total cover
on grasslands. Mean litter cover was
40% and bare ground 16% over the
two sampling years.

Small Mammals

Average numbers of small mammals
were similar (P = 0.4) on the two
vegetation types; however, species
composition differed between juni-
per woodlands and adjacent grass-
lands (table 2). Deer mice (Pero-
myscus maniculatus) were the most
common species captured in both
juniper woodlands and on grass-
lands, constituting 66% of the total
capture in juniper woodlands and
48% on grasslands. Number of deer
mice captured was similar (P = 0.4)
in both vegetation types, averaging
42 and 36 individuals per site in juni-
per woodlands and grasslands, re-
spectively. White-footed mice (P. leu-
copus) were the next most abundant
small mammal species captured in
juniper woodlands, constituting ap-
proximately 29% of the total cap-
tures; their numbers were much
lower (P = 0.04) on grassland sites.
Bushy-tailed woodrats (Neotoma cin-
erea) were captured in small numbers
in the juniper woodlands but were
absent from grasslands. Average
numbers of meadow voles (Microtus
pennsylvanicus) (P = 0.03), thirteen-
lined ground squirrels (Spermophilus

tridecemlineatus) (P = 0.03), northern
grasshopper mice (Onychomys leu-
cogaster) (P = 0.06), and western har-
vest mice (Reithrodontomys megalotis)
(P = 0.08) were higher on grasslands
than in juniper woodlands. Small
numbers of plains pocket mice (Per-
ognathus flavescens) and hispid pocket
mice (P. hispidus) were captured in
both vegetation types. One house
mouse (Mus musculus) was captured
in a juniper woodland.

Discussion

Rocky Mountain juniper stands did
not support significantly higher num-
bers of small mammals than did ad-
jacent grasslands, but enhanced
small mammal diversity by provid-
ing specialized habitat for white-
footed mice and bushy-tailed
woodrats. White-footed mice prefer
and are commonly restricted to ri-
parian forests and shrubby habitats
in this region (Armstrong 1972, Sea-
bloom et al. 1978), and were a com-
mon species in Rocky Mountain juni-
per woodlands in North Dakota
(Hopkins 1983). White-footed mice
forage (M'Closkey 1975) and nest
(Wolff and Hurlbutt 1982) in trees
and show a tendency to use woody
vegetation as escape routes (Barry
and Francq 1980). Their preferred
habitat is often characterized by
dense woody understory (Yahner
1982). Rocky Mountain juniper
woodlands lack vertical layering pro-
vided by shrubs, but the dense tree
canopy and presence of branches
nearly to the ground may substitute
for shrub layers found in other
woodlands. Further, juniper wood-
lands may function as dispersal path-
ways for woodland species such as
white-footed mice. Turner (1974)
postulated that riparian habitats
along major drainageways allowed
the western expansion of the white-
footed mouse.

Bushy-tailed woodrats are often
restricted to rocky areas in this re-
gion (Jones et al. 1983), and their

presence has been documented in
deciduous woodlands in northwest-
ern South Dakota (Hodorff et al. In
Press). Bushy-tailed woodrats were
captured in ponderosa pine (Pinus
ponderosa) stands, toe slopes, hilly
scoria, and upland breaks in western
North Dakota (Seabloom et al. 1978).
Juniper stands likely provide den
sites, which grasslands lacked. Mid-
dens constructed of juniper branches
were observed in three of four Rocky
Mountain juniper sites in this study.

Three species — deer mice, plains
pocket mice, and hispid pxKket
mice — apparently showed no prefer-
ence between grasslands or juniper
woodlands. The high proportion of
deer mice in the total capture on both
grasslands and in juniper woodlands
is not uncommon on the Northern
Great Plains. Deer mice are a ubiqui-
tous species, occurring in nearly ev-
ery habitat in this region (Jones et at.
1983). Deer mice were the most com-
monly captured species in green ash
woodlands in northwestern South
Dakota (Hodorff et al. In Press), and
were abundant in both green ash and
Rocky Mountain juniper woodlands
in western North Dakota (Hopkins
1983). Rocky Mountain junif)er
woodlands in South Dakota are
probably not critical habitat for deer
mice, but when available, will be ex-
ploited by this adaptive species.

Hispid pocket mice apparentiy
prefer rocky areas, where a variety of
shrubs, forbs, and yucca (Yucca spp.)
grow (Jones et al. 1983). Plains pocket
mice are considered rare mammals in
South Dakota (Houtcooper et al.
1985); hence little is known about the
distribution and habitat preferences
of this species in the state. Hodorff et
al. (In Press) captured low numbers
of both plains and hispid pocket mice
in green ash woodlands in north-
western South Dakota. Haufler and
Nagy (1984) captured plains pocket
mice in pinyon pine (Pinus edulis)-
Utah juniper (/. osteosperma) wood-
lands in Colorado, and reported that
juniper comprised 17% of the pocket
mouse's diet. The small captures of

330

both species of pocket mice make
generalizations about habitat prefer-

' ence suspect, but Rocky Mountain
juniper woodlands likely provided
habitat interspersion and food re-

! sources for these species.

; Juniper woodlands, with sparse
understory cover, are atypical habitat
for grassland inhabitants such as
meadow voles, thirteen-lined ground
squirrels, northern grasshopper mice,
and western harvest mice. Meadow
voles, in particular, are generally as-
sociated with dense stands of grass
(Birney et al. 1976). However, Rocky
Mountain juniper woodlands in
southwestern North Dakota sup-
ported meadow voles in some areas
(Seabloom et al. 1978) and prairie
voles (M. ochrogaster) on other sites
(Hopkins 1983).

The ability of North Dakota juni-

1 per woodlands to support microtines
was attributed to differences in plant
community attributes. Littleseed
ricegrass and mosses dominated the
understory and total plant cover av-
eraged over 60% (vs. 25% in South
Dakota) in most juniper stands
sampled by Hopkins (1983) (Hansen
et al.l984). The more dense under-
story of the North Dakota wood-

I lands, which South Dakota wood-
lands lacked, apparently provided

(adequate cover for microtines.
Thirteen-lined ground squirrels
were most frequently captured in
northwestern South Dakota in road-
ways and fencerows in shortgrass
prairies (Andersen and Jones 1971).
Northern grasshopper mice are gen-
erally restricted to shortgrass and
desert sites (McCarty 1978), in areas
with adequate dust-bathing sites
I) (Egoscue 1960). Western harvest
mice were occasionally captured in
pinyon-juniper woodlands in south-
eastern Colorado, but were associ-
ated with dense herbaceous cover
lacking tree canopy cover (Ribble
and Samson 1987). Rocky Mountain
juniper woodlands may provide sup>-
plemental food resources for small
mammals generally restricted to
grasslands.

Conclusion

Rocky Mountain juniper woodlands
enhance small mammal richness of
the generally treeless Northern Great
Plains by providing specialized habi-
tat for at least two species, bushy-
tailed woodrats and white-footed
mice. Juniper woodlands lack well-
developed shrub layers, but the
dense canopy of the juniper trees and
crowns that extend nearly to the
ground may provide foraging and
nesting substrates for woodland
mammals. Further, Rocky Mountain
juniper woodlands may function as
dispersal pathways for these two
species. Juniper woodlands lack
dense herbaceous understories neces-
sary to support microtines such as
meadow voles, but likely serve as
food resource supplemental areas for
a variety of mammals associated
with grasslands. Adaptable species
such as the deer mouse may not re-
quire juniper woodlands, but will
exploit this habitat when available.
Finally, Rocky Mountain juniper
woodlands may figure into the habi-
tat needs of pocket mice, but low
captures of two species preclude
clear definition of preferred habitat.

Acknowledgments

Critical reviews by Dan Uresk, Deb
Paulson, Dick Hansen, and Bill Clark
were helpful in improving this
manuscript. Personnel at Badlands
National Park were most cooperative
during this study. Deb Paulson and
Bob Hodorff helped with field work.

Literature Cited

Andersen, Kenneth W., and J. Knox
Jones, Jr. 1971. Mammals of north-
western South Dakota. University
of Kansas Museum of Natural His-
tory 19:361-393.

Armstrong, David M. 1972. Distribu-
tion of mammals in Colorado.
415 p. Monograph No. 3, Museum

of Natural History, University of
Kansas, Lawrence, KS.

Barry, Ronald E., Jr., and Edward N.
Francq. 1980. Orientation to land-
marks within the preferred habitat
by Peromyscus leucopus. Journal of
Mammalogy 61:292-303.

Birney, Elmer C, W.E. Grant, and
Donna Day Baird. 1976. Impor-
tance of vegetation cover to cycles
of Microtus populations. Ecology
57:1043-1051.

Daubenmire, R. 1959. A canopy-cov-
erage method of vegetation analy-
sis. Northwest Science 33:43-64.

Egoscue, H. J. 1960. Laboratory and
field studies of the northern grass-
hopper mouse. Journal of Mam-
malogy 51:622-623.

Faanes, Craig A. 1984. Wooded is-
lands in a sea of prairie. American
Birds 38:3-6.

Powells, H.A. 1965. Rocky Mountain
Juniper {Juniperus scopulorum
Sarg.). In silvics of Forest trees of
the United States. Agricultural
Handbook No. 271. p. 217-222.
Washington, D.C.

Gaines, Ronald C, and Stanley C.
Kohn. 1982. Raptor use of hard-
wood draws in central North Da-
kota. Prairie Naturalist 14:55-58.

Hansen, Paul L., George R. Hoffman,
and Ardell J. Bjugstad. 1984. The
vegetation of Theodore Roosevelt
National Park, North Dakota: a
habitat type classification. General
Technical Report RM-113. 35 p.
Fort Collins, CO: U.S. Department
of Agriculture, Forest Service,
Rocky Mountain Forest and Range
Experiment Station.

Hansen, Paul L., Ricky B. Hopkins,
and George R. Hoffman. 1980. An
ecological study of the habitat
types and their animal compx)-
nents at Theodore Roosevelt Na-
tional Park, North Dakota. 182 p.
Department of Biology, Vermil-
lion, University of South Dakota.

Haufler, Jonathan B., and Julius G.
Nagy. 1984. Summer food habits
of a small mammal community in
the pinyon-juniper ecosystem.
Great Basin Naturalist 44:145-150.

331

Hodorff, Robert A., Carolyn Hull
Sieg, and Raymond L. Linder. In
Press. Wildlife response to stand
structure of deciduous wood-
lands. Journal of Wildlife Manage-
ment.

Hopkins, Rick B. 1983. Small mam-
mal component of habitat types in
Theodore Roosevelt National
Park. Prairie Naturalist 15:45-46.

Hopkins, Rick B., Frank J. Cassel,
and Ardell J. Bjugstad. 1986. Rela-
tionships between breeding birds
and vegetation in four woodland
types of the Little Missouri Grass-
lands. Research Paper RM-270. 12
p. Fort Collins, CO: U.S. Depart-
ment of Agriculture, Forest Serv-
ice, Rocky Mountain Forest and
Range Experiment Station.

Houtcooper, Wayne C, David J.
Ode, John A. Pearson, and George
M. Vandel, III. 1985. Rare animals
and plants of South Dakota. Prai-
rie Naturalist 17:143-165.

Jones, J. Knox, Jr., David M. Arm-
strong, Robert S. Hoffman, and
Clyde Jones. 1983. Mammals of
the Northern Great Plains. 379 p.
Lincoln, NE: University Nebraska
Press.

McCarty, Richard. 1978. Onychomys
leucogaster. Mammal Species 87:1-
6. American Society of Mammalo-
gists.

M'Closkey, Robert T. 1975. Habitat
dimensions of white-footed mice,
Peromyscus leucopus. American
Midland Naturalist 93:158-167.

Ribble, David O., and Fred B. Sam-
son. 1987. Microhabitat associa-
tions of small mammals in south-
eastern Colorado, with special
emphasis on Peromyscus (Roden-
tia). Southwestern Naturalist
32:291-303.

Seabloom, Robert W., Richard D.
Crawford, and Michael G. McK-
enna. 1978. Vertebrates of south-
western North Dakota: amphibi-
ans, reptiles, birds, mammals. Re-
search Report 24, 549 p. Grand
Forks, ND: University of North
Dakota; Institute of Ecological
Studies.

Severson, Kieth E. 1981. Plains habi-
tat. In Wallmo O.C. Mule and
black-tailed deer in North Amer-
ica, p. 459-468. Lincoln, NE: Uni-
versity of Nebraska Press and
Washington, D.C., Wildlife Man-
agement Institute.

Severson, Kieth E., and Arthur V.
Carter. 1978. Movements and
habitat use by mule deer in the
Northern Great Plains, South Da-
kota, p. 466-468. In D.N. Hyder,
editor. Proceedings of the First
International Rangeland Congress.
[Denver, CO, August 14-18, 1978]
Society for Range Management.
Peerless Printing, Denver, CO.

SPSS. 1986. SPSSX user's guide. 988
p. New York, N.Y. McGraw-Hill
Book Company.

Taber, Richard D., and McT. I.
Cowan. 1971. Capturing and
marking wild animals, p. 277-317.
In Wildlife Management Tech-
niques, 3rd edition. R.H. Giles Jr.,
editor. Washington, D.C., Wildlife
Society.

Turner, Ronald W. 1974. Mammals of
the Black Hills of South Dakota
and Wyoming. Miscellaneous Pub-
lication No. 60, 178 p. Lawrence,
KS: University of Kansas, Museum
of Natural History.

Uresk, Daniel W. 1982. Importance of
woodlands to wildlife and live-
stock use on the Northern High
Plains, p. 7-12. In North Platte,
NE: Proceedings of the Great
Plains Agricultural Council.

Wolff, Jerry O., and Bethia, Hurlbutt.
1982. Day refuges of Peromyscus
leucopus and Peromyscus manicula-
tus. Journal of Mammalogy 63:666-
668.

Yahner, Richard H. 1982. Microhabi-
tat use by small mammals in farm-
stead shelterbelts. Journal of
Mammalogy 63:440-445.

332

Postfire Rodent Succession
Following Prescribed Fire In
Southern California
ChaparraP

William O. Wirtz, 11,^ David Hoekman,^ John
R. Muhm/ and Sherrie L. Souza^

Abstract.— This paper describes species
composition and density changes in rodent
populations during postfire succession following
prescribed fire in the chaparral community of the
San Gabriel Mountains. Conclusions are drawn from
a 4-year, live-trap, mark and release study of postfire
succession in two watersheds receiving "hot" burns
and two receiving "normal" burns.

The chaparral community of south-
ern California is associated with
nearly two million years of fire his-
tory (Hanes 1971). In recent centuries
major fires have occurred at intervals
of 20 to 40 years (Byrne et al. 1977;
Philpot 1977). Postfire plant succes-
sion (Patric and Hanes 1964, Hanes
and Jones 1967, Hanes 1971) and the
fire itself have varying short term
effects on the birds and small mam-
mals found in the chaparral (Law-
rence 1966, Quinn 1979, Wirtz 1977,
1979). Wirtz (1977) summarized the
work of earlier authors concerning
conditions in small vertebrate mi-
crohabitats during fire, vertebrate be-
havior during fire, and survival of
small vertebrates exposed to fire.
Both Lawrence (1966) and Quinn
(1979) studied rodent populations
before and after a bum, in addition
to documenting microhabitat condi-
tions during the fire. Wirtz (1977,

'Paper presented at symposium. Man-
agement of Amphibians, Reptiles, and
Small Mammals in North America. (Flag-
staff, AZ, July 19-21. 1988.)

'William O. Wirtz. II is Professor of Biology.
Department of Biology. Pomona College.
Claremont. CA 91711.

^David Hoekman is a research assistant.
Department of Biology, Pomona College,
Claremont. CA9171I.

"John R. Muhm is a research assistant.
Department of Biology. Pomona College.
Claremont. C A 91711.

^Sherrie L. Souza is a research assistant.
Department of Biology. Pomona College.
Claremont. CA 91711.

1982, 1984) presents preliminary
analyses of data collected on postfire
rodent succession following wildfire
in the chaparral community of south-
ern California.

Because of the recently recognized
significance of the use of prescribed
fire in the management of chaparral
ecosystems, the Pacific Southwest
Forest and Range Experiment Sta-
tion, USD A Forest Ser vice, began for-
mulating plans in 1983 for a series of
prescribed fires in the San Dimas Ex-
perimental Forest, located in the San
Gabriel Mountains of southern Cali-
fornia, that might be utilized for long
range studies of the effects of pre-
scribed fire in chaparral. In October,
1984, the Forest Service burned four
chaparral watersheds of approxi-
mately 40 ha each in the San Dimas
Experimental Forest. This paper de-
scribes the changes in rodent com-
munity structure for the 4-year pe-
riod following prescribed burning.

Methods

In October, 1984, four chaparral wa-
tersheds of approximately 40 ha each
were subjected to prescribed bums in
the San Dimas Experimental Forest.
The vegetation of the two of these
watersheds (874 and 775) had been
hand cut in the spring of 1984 to pro-
duce the dried fuel for an exception-
ally hot fire. Two adjacent water-
sheds (804 and 776) burned normally

for climatic conditions at the time. A
fifth watershed (803), which has been
extensively studied since 1976 (see
Wirtz 1977, 1979, 1982, 1984), serves
as a control for studies on the pre-
scribed bum areas.

Rodent live-trap, mark and re-
lease, studies were conducted on all
experimental areas prior to the bums
to document the size and species
composition of the prefire rodent
community on all watersheds, and
175 individuals were permanently
marked by toe-clipping to provide a
prefire pool of marked rodents from
which to determine survival rates
following the bum. Following the fire
grids of 50 stations at 15 m intervals
were established in each of the four
watersheds on the sites of the prefire
censusing, and a live-trap, mark and
release, program was initiated to de-
termine fire survival and postfire ro-
dent succession patterns.

For this paper, population esti-
mates were done by the Hayne (1949)
equation. Area sampled, for each
species, for each month, was esti-
mated by determining the mean dis-
tance traveled for each species be-
tween captures, for each month, and
then adding a zone equal to the mean
distance travelled to the perimeter of
the grid. Biomass was determined by
the product of the estimated popula-
tion times the mean weight for each
species for the month, and these val-
ues are then summed for all species
taken on the grid for the month.

333

Results

Postfire trapping was initiated in
February 1985, and both experimen-
tal and control plots were sampled
bi-monthly. Hayne equation popula-
tion estimates for rodent populations
on each study plot are presented in
figures 1-5, The absence of data
fX)ints from February through April
or May means that no rodents were
trapped, except for watershed 803 in
which trapping was not begun until
June 1985.

Mice of the genus Peromyscus
(deer mouse, P. maniculatus; brush
mouse, P. boylii; California mouse, P.
calif ornicus), and California pocket
mice, Perogmthus californicus, consti-
tute the bulk of the postfire rodent
population. Pacific kangaroo rat, Di-
podomys agilis, dusky-footed wood
rat, Neotoma fuscipes, and California
vole, Microtus californicus, are present
in low numbers, and a few Botta's
pocket gopher, Thomomys bottae, and

1985 1066 1987 1088

Figure 1 .—Hayne equation estimates of
population size of rodent species in Bell
803, ttie 28-year-old ctiaparral control plot.

.or

" L

20
N 10
0

■e

70
«0

so

N 40

30

20
10
0

p. boylii

p. maniculslus

D. •gilit

•T"i"

Perognathus i

A
\ V

o jumjsnjmmjsnjmmjsnjmmj
1985 . 1986 . 1987 • 1988

Figure 2.— Hayne equation estimates of
population size of rodent species in Bell
804, normal prescribed burn. Note ttiat
points at 0 on tt>e x-axis against ttie y-axis
are populations estimates prefire.

'r 0. agilit

I -r-i-T

O JMMJSHJMMJ8NJMMJSNJWU

198S • 1986 • 1987 • 1988

Figure 4.— Hayne equation estirrtates of
population size of rodent species in San
Dimas 776, nornnal prescribed burn. Note
that points at 0 on the x-axis against the y-
axis are populations estimates prefire.

N 30

P.cali'ornic jg

P. boylii

\ / p\maniculatu» v.

O JMMJ9NJMMJSNJMUJSNJMMJ

1985 • 1986 • 1987 • 1988

Figure 3.— Hayne equation estimates of
populations size of rodent species in Bell
874, hot prescribed burn. Note that points at
0 on the x-axis against the y-axis are popu-
lations estimates prefire.

N 30

0 » ■ —l—t'

h II

A Parognathua > Ml
!/ V •- •'^ D. aglris ^

Naotoma

10 r

I MIcrolua

O JMMJSNJMMJSNJMMJSHJMM
1986 • 19B« • 1987 • 1988

Figure 5.— Hayne equation estimates of
population size of rodent species in San
Dimas 775, hot prescribed bum. Note thiot
points at 0 on the x-axis against the y-axis
are populations estirrKites prefire.

334

western harvest mouse, Reithrodonto-
mys megalotis, have also been taken.
Larger mammals observed in burned
watersheds, for which no quantita-
tive data are available, included
Beechey ground squirrel, Spermophi-
lus beecheyi, Audubon's cottontail,
Sylvilagus auduboni, brush rabbit, S.
bachmani, coyote, Canis latrans, black
bear, Ursus americanus, badger. Tax-
idea taxus, and mule deer, Odocoileus
hemionus.

Fire Survival

No marked wood rats survived the
fires. Nine (12.5%) Peromyscus sur-
vived normal fires, and one (1.4%)
survived hot fires. Tow (12.5%)
Pocket mice survived normal fires,
and two (12.5%) Survived hot fires.
These data support the currently
held opinion that some rodents do
survive fires, and help provide the
nucleus, along with immigration
from unburned areas, for rodent
postfire succession.

Larger mammals seen in the
burned watersheds in the first month
postfire included coyote, black bear,
badger, and mule deer.

Early Postfire Succession

Pocket mice and all three Peromyscus
species were present on one hot burn
(874) by April 1985, six months
postfire, but no rodents were present
on the other hot burn (775). Pocket
mice moved into this hot burn (775)
by May, and two Peromyscus species,
(P. californicus, P. maniculatus) were
present by July.

Pacific kangaroo rats appeared on
some burned areas by June or July
1985 (they are rare in mature chapar-
ral). Woodrats appeared on one nor-
mal bum (804) by June 1985, and an-
other (776) by September 1985, and
on one hot burn (874) by August
1985. Single pocket gophers and har-
vest mice have been taken on one hot
bum (775).

Demography

Sampling was not begun on the con-
trol plot (803) until June 1985. The
rodent population on this plot con-
sists chiefly of wood rats, California
mice, and pocket mice (fig. 1). The
California mouse population peaked
during the fall, winter, and spring of
1985-86, and again in the winter and
spring of 1986-1987. Pocket mice
were rare on the control until the fall
of 1986 and remained common until
the summer of 1987 (fig. 1). The
wood rat population has peaked in
each summer studied to date.

The prefire rodent population on
the normal bum in Bell (804) was
composed primarily of woodrats,
with smaller numbers of other spe-
cies (fig. 2) (note that symbols at 0 on
the X-axis against the y-axis represent
prefire density estimates). The
postfire rodent population on this
grid has been composed primarily of
brush mice and pocket mice, with
population peaks of the latter in each
winter (1985, 1986, and 1987). Wood
rat populations did not show signifi-
cant increases on this grid until the
spring of 1987, about 30 months after
the burn, and they have yet (June
1988) to reach prefire densities (fig.
2). Pacific kangaroo rats have oc-
curred on this bumed area in num-
bers above prefire densities since the
summer of 1985. Brush and Califor-
nia mouse populations have oc-
curred in numbers above prefire den-
sities since the winter of 1985-86 (fig.
2).

The prefire rodent population on
the hot bum in Bell (874) was com-
posed largely of wood rats, Califor-
nia mice, and pocket mice (fig. 3). All
species, except kangaroo rats, were
present again on this grid by August
1985, 10 months postfire. The postfire
rodent community on this hot burn
has been dominated by brush mice
and pocket mice (fig. 3), with both
species reaching, or exceeding, pre-
fire densities by the winter of 1985,
approximately a year after the burn.
Califomia mouse and wood rat

populations have yet (June 1988) to
reach prefire densities (fig. 3).

The prefire rodent population on
the normal bum in San Dimas (776)
was composed primarily of Califor-
nia mice and wood rats, with smaller
numbers of pxxket mice and no
brush mice (fig. 4). The postfire ro-
dent community has been dominated
by Califomia mice and pocket mice,
with both species exceeding prefire
densities by the winter of 1985, ap-
proximately one year postfire. Wood
rats have yet (June 1988) to reach
prefire densities, brush mice have not
appeared on this grid, and Califomia
voles were common in the summer
of 1987 and the spring of 1988 (fig. 4).

The prefire rodent population on
the hot burn in San Dimas (775) was
very similar to that on the normal
bum here (fig. 5). And, like the nor-
mal bum, the postfire community
has been dominated by Califomia
mice and p)ocket mice, with pocket
mice exceeding prefire densities by
the summer of 1985 and Califomia
mice exceeding prefire densities by
the fall of 1986 (fig. 5). Pacific kanga-
roo rats also exceeded prefire densi-
ties within one year postfire on this
grid.

Comment should be made about
the presence of deer mice (P. manicu-
latus) and Olifornia voles (Microtus)
on these grids. Neither species was
present on any grid prefire, and nei-
ther has been taken on the control
(figs. 1-5). P. maniculatus has been
taken on all burned grids, with peaks
of abundance by the second year
postfire and declining abundance by
the fourth year postfire (figs. 4 and
5).

Effects of Hot and Normal Fires

The effects of hot and normal fires on
rodent demography were examined
by (1) comparing pre and post fire
populations in areas exposed to these
two fire regimes (figs. 6 and 7), (2)
comparing the number of captures of
each species postfire under each fire

335

regime (fig. 8), and (3) comparing
total postfire biomass on areas ex-
posed to different fire regimes (fig. 9)
(note again that points at 0 on the x-
axis against the y-axis are prefire
populations estimates). Only species
with relatively high abundances are
considered in this paper.

Prefire populations of brush mice
were essentially the same on both
areas to be burned in Bell, while den-
sities of pocket mice and California
mice were greater on the area to re-
ceive the hot bum, and deer mice
were not present on either grid (fig.
6). All prefire populations were se-
verely impacted by fire, dropping in
most instances to near zero for sev-
eral months postfire. Pocket mice in-
creased to twice their prefire density
on the hot bum and 25 times prefire
density on the normal burn (fig. 6).
Bmsh mice increased to 14 times
their prefire density on the hot burn
and six times prefire density on the
normal bum (fig. 6). Califomia mice
returned to prefire density by one
year postfire on the normal burn, and
numbers have remained relatively
constant since then. Deer mice were
present on both burned areas
postfire, but have been more abun-
dant on the hot bum (fig. 6).

Prefire populations of Califomia
mice and pocket mice were similar
on both areas to be burned in San
Dimas (fig. 7). Some individuals sur-
vived the normal burn. Pocket mouse
populations exceeded prefire densi-
ties on both normal and hot bums by
eight months postfire (fig. 7). Califor-
nia mouse populations exceeded pre-
fire densities by one year postfire on
the normal burn, but took two years
to reach prefire densities on the hot
bum (fig. 7). Two species not present
prefire. Pacific kangaroo rats and
deer mice, colonized both burned
areas by eight months postfire; kan-
garoo rats have remained numerous
on the hot burn, and deer mice are
more numerous on the hot bum than
on the normal bum (fig. 9).

Captures of Califomia mice
postfire are greater on normal burns

than on hot burns, and exceed cap-
tures on the control on one normal
bum (776) (fig. 8). Captures of bmsh
mice postfire are greater on both hot
bums and one normal burn than on
the control, and captures on hot
bums are greater than on normal
bums for each pair of watersheds
bumed (fig. 8). Deer mice have not
been captured on the control; cap-
tures are greater postfire on hot
bums than on normal burns for each
pair of watersheds burned (fig. 8).
Captures of wood rats are less on all
bumed areas than on the control, and
they are less on hot bums than on
normal burns for each pair of water-
sheds bumed (fig. 8).

California voles have not been
taken f)ostfire on the control nor on
one normal burn, and are greater on
the other normal burn than on either
hot bum (fig. 8). Captures of Pacific
kangaroo rats p>ostfire are greater on

" I ■ I I • ■ -^r I" -- ■ -

O JMMJtNJMMJSNJMMJtNJMMJ

1685 • 1»ae • 1987 • 1888

Figure 6.— Comparison of roder^t postfire
population growthi on normal (804) and hot
(874 prescribed fire plots in Bell. Note thiat
points at 0 on \he x-axis against ttie y-axis
are populations estimates prefire.

336

both normal and one hot bum than
on the control, while captures of
pocket mice postfire are greater on
all bumed areas than on the control
(fig. 8).

Total biomass on the control, not
28 years old, has fluctuated during
the period of study, but shows a
slight increasing trend (fig. 9). Total
biomass on both bumed plots in Bell,
the location of the control, has also
fluctuated, with a slight increasing
trend, in a fashion similar to that of
the control (fig. 9). Total biomass on
the bumed plots in San Dimas has
also fluctuated, with slight increasing
trend, but with two dramatic bio-
mass increases, one in the Spring of

1987 and the other in the spring of

1988 (fig. 9). The pattem of fluctua-
tion, and increase, on the normal
bum in San Dimas is similar to that
observed for the control, and the pat-
tern of fluctuation, and increase, if

1986 • 1988 • 1987 • 1986

Figure 7.— Comparison of rodent postfire
population growtti on normal (776) and ho\
(775) prescribed fire plots in San Dioxis.
Note ttiat points at 0 on \he x-axis against

the two sharp peaks are not consid-
ered, is also similar to the control
(fig. 9).

Discussion

General patterns of rodent postfire
succession following these prescribed
bums are similar to those reported
by Wirtz (1977, 1982, 1984) for suc-
cession following wildfire in the
chaparral of the San Gabriel Moun-
tains, but lack the dramatic increases
in density, and therefore biomass,
observed in these earlier studies. He
notes (1984) that rodent succession
following wildfire takes about four
years before populations stabilize at
essentially prefire conditions found
in older chaparral stands. The re-
sponse of species to these prescribed
fires varied, with some species reach-
ing prefire densities in less than four
years and others having not yet
reached prefire densities at essen-
tially four years postfire.

Only slight differences are noted
between rodent postfire succession

on normal and hot bums, and these
may probably be attributed to differ-
ences in the biology of individual
species. In Bell, both normal and hot
bums were dominated postfire by
pocket mice and brush mice, though
pocket mice had the highest density
on the normal burn (804) and bmsh
mice had the highest density on the
hot burn (874) (fig. 6). Califomia
mice recovered to prefire density on
the normal bum, but have not yet
(June 1988) recovered on the hot
burn, and wood rats have not recov-
ered to prefire densities on either
bumed area (fig. 6). Deer mice have
been more prevalent on the hot bum
than on the normal burn during the
period of the study. By the second
year postfire, populations of all spe-
cies, except wood rats, exceeded pre-
fire densities on the normal burn (fig.
2), and populations of brush mice
and p>ocket mice had exceeded pre-
fire densities on the hot bum (fig. 3).

In San Dimas, where considerable
brush was left alive on the normal
bum (776), both normal and hot
bums were dominated postfire by

{XKket mice and Califomia mice (fig.
7). Both of these species recovered to
prefire densities on the normal burn
by one year postfire (fig. 4), as did
pocket mice on the hot bum (fig. 5),
but Califomia mice did not reach
prefire densities on the hot burn until
the second year postfire (fig. 5). For
reasons not immediately apparent,
but probably because of the presence
of some grass prefire, California
voles were found only in these two
watersheds postfire. The greater rela-
tive abundance of Pacific kangaroo
rats on the hot burn is most likely
due to the fact that more open space,
necessary for kangaroo rat saltitorial
locomotion, was left by the hot fire
here.

Pocket mice increase rapidly on
bumed areas, there being essentially
no difference between normal and
hot burns (figs. 6 and 7). Bmsh mice,
if present prefire, recover more rap-
idly p)ostfire than California mice,
and the latter recover more rapidly
on normal burns than on hot burns
(figs. 6 and 7). Deer mice, virtually
nonexistent in mature chaparral,
colonize both normal and hot bums,
and increase more rapidly on hot
bums (figs. 6 and 7).

Data on captures (fig. 8) indicate
that increase of deer mice on hot
bums. The sf)ecies is known to colo-
nize disturbed areas, whether they be
caused by fire, logging, or over-

Figure 9.— Total postfire biorrxjss (grams) for
control and burned plots.

800

700

eoo

CO

UJ 500

QC

3

Q.

< 400

u

300

200

100

Peromyscus
cahlornicus

CONTROL
NORMAL ' »
HOT HW

CNHNH CNHNH CNHNH

Ptro-nyscut

Pvromy acu*
maniculaluS

200

100

jQlL

0 0

CNHNH CNHNH

luselptt

300

200

100

OtpodoMyt
■ gills

Uicrotut
oUlornleut

CNHNH CNHNH

Parognathua

ealliornicut

Figure 8.— Connparison of postfire captures of all rodent species on control and prescribed
burn plots.

337

grazing (Williams 1955). These data
also illustrate the decline of Califor-
nia mice on hot bums and its in-
crease in normal burns, and the in-
crease of brush mice, where present
prefire, on both normal and hot
bums. Buming favors density in-
creases of pocket mice, with essen-
tially no difference between normal
and hot burns. Kangaroo rats exhibit
variable increases in response to fire,
and wood rats are severely impacted
by fire.

Biomass increases in response to
fire are variable, and in this study,
were similar in variability to those
occurring on the control (fig. 9). The
sharp peaks in biomass observed on
one hot burn (775) are due to large
density increases in pocket mice dur-
ing these periods.

It is important to note, when com-
paring data for normal and hot
bums, that in one normal burn (776)
a lot of unburned brush remained,
perhaps more accurately simulating
an "island" in a bum rather than a
burn per se. So, for this study, the
data for 776 are somewhat atypical,
and 804 represents more accurately
the situation following a normal
bum. But it is also important to note
that "islands" of unburned vegeta-
tion are frequently left by wildfire,
providing refugia for both plants and
animals from fire.

Several general conclusions may
be drawn from the rodent data: (1)
fire may impact rodent species se-
verely, probably chiefly through loss
of habitat resources, especially shel-
ter and food; (2) some individuals
survive fire; (3) colonization from
adjacent habitats may be rapid; (4)
postfire succession is somewhat de-
pendent on prefire species composi-
tion of the area; (5) in southem Cali-
fornia chaparral, at least two species,
deer mouse and California vole, are
fire specialists, entering the system
only for relatively short periods of
the postfire succession; (6) species
requiring brush for cover and /or
food, like wood rats and California
mice, are most severely impacted by

fire, and require the longest time to
recover to prefire densities; (7) there
is no clear-cut difference in rodent
postfire succession following normal
and hot fires; (8) rodent postfire suc-
cession is characterized by increases
in successionally-adapted species,
with declines in those species for
which essential habitat features are
lacking; and (9) recovery of the ro-
dent community to its prefire condi-
tion probably takes four to six years,
with the exact pattern of recovery
being dependent on prefire species
composition and features of the pre-
fire plan community and postfire
plant succession that have not been
delineated.

Acknowledgments

This research was supported by
USDA Forest Service, Pacific South-
west Forest and Range Experiment
Station Grant Number PSW-85-
0004CA to WOW and a summer re-
search assistantship from Pomona
College to JRM.

We are indebted to Susan Conard,
Project Leader, Pacific Southwest
Forest and Range Experiment Sta-
tion, Forest Fire Laboratory, River-
side, CA, for her support and coop-
eration during this study. Many biol-
ogy students at Pomona College have
assisted with field work. The senior
author wants to acknowledge 5 years
of field work by Sherrie Souza and
David Hoekman, all computer pro-
gramming by David Hoekman, and
all data analysis by John R. Muhm.
We are grateful to Helen Wirtz for
our figures. Preparation of this paper
was greatly assisted by a summer
research assistantship to JRM.

Literature Cited

Byrne, Roger, Joel Michaelsen and
Andrew Soutar. 1977. Fossil char-
coal as measure of wildfire fre-
quency in southern Califomia: A
preliminary analysis, p. 361-367. In

338

Harold A. Mooney, Eugene C.
Conrad, eds. Proceedings of the
Symposium on the Environmental
Consequences of Fire and Fuel
Management in Mediterranean
Ecosystems. USDA Forest Service,
General Technical RepK3rt WO-3.

Hanes, Ted L. 1971. Succession after
fire in the chaparral of southern
California. Ecological Monographs
41(l):27-52.

Hanes, Ted L. and Harold W. Jones.
1967. Postfire chaparral succession
in Southem Califomia. Ecology
48:259-263.

Hayne, Don W. 1949. Two methods
of estimating population from
trapping records. Journal of Mam-
malogy 30(4):399-411.

Lawrence, George E. 1966. Ecology
of vertebrate animals in relation to
chaparral fire in the Sierra Nevada
foothills. Ecology 47(2):278-291.

Patric, James H. and Ted L. Hanes.
1964. Chaparral succession in a
San Gabriel Mountain Area of
Califomia. Ecology 45:353-360.

Philpot, Charles W. 1977. Vegetative
features as determinants of fire
frequency and intensity, p. 12-16.
In Harold A. Mooney, Eugene C.
Conrad, eds. Proceedings of the
Symposium on the Environmental
Consequences of Fire and Fuel
Management in Mediterranean
Ecosystems. USDA Forest Service,
General Technical Report WO-3.

Quinn, Ronald D. 1979. Effects of fire
on small mammals in the chapar-
ral. Cal-Neva Wildlife Transac-
tions, 125-133.

Williams, Olwen. 1955. Distribution
of mice and shrews in a Colorado
montane forest. Journal of Mam-
malogy 36(2):221 -231.

Wirtz, William O., II. 1977. Verte-
brate postfire succession, p. 46-57.
In Harold A. Mooney, Eugene C.
Conrad, eds.. Proceedings of the
Symposium on the Environmental
Consequences of Fire and Fuel
Management in Mediterranean
Ecosystems. USDA Forest Service,
General Technical Report WO-3,
Washington, D.C., 498 p.

Wirtz, William O., II. 1979. Effects of
fire on birds in chaparral. Cal-
Neva Wildlife Transactions, 114-
124.

Wirtz, William O., II. 1982. Postfire
community structure of birds and
rodents in southern California
chaparral, p. 241-246. In Eugene C.
Conrad, Walter C. Oechel, techni-
cal coordinators. Proceedings of
the Symposium on Dynamics and
Management of Mediterranean-
type Ecosystems. USDA Forest
Service, General Technical Report
PSW-58.

Wirtz, William O., II. 1984. Postfire
rodent and bird communities in
the chaparral of southern Califor-
nia, p. 167-168. In MEDECOS IV.
Proceedings of the 4th Interna-
tional Conference on Mediterra-
nean Ecosystems. University of
Western Australia, Perth, Western
Australia.

Douglas- Fir Forests in the
Cascade IVIountains of
Oregon and Washington: Is
the Abundance of Small
Mammals Related to Stand
Age and Moisture?^

Paul Stephen Corn,^ R. Bruce Bury,^ and
Thomas A. Sples^

Abstract.— Red tree voles (Arborimus longicaudus)
were the only small mammal strongly associated
with old-growth forests, whereas vagrant shrews
(Sorex vagrans) were most abundant in young
forests. Pacific marsh shrews (S. bendirii) were most
abundant in wet old-growth forests, but abundance
of this species in young (wet) forests needs further
study. Clearcuts had a mammalian fauna distinct
from young forest stands. Abundance of several
species was correlated to habitat features unique to
naturally regenerated forests, indicating an urgent
need to study the long-term effects of forest
management on nongome wildlife.

Management of old-growth Douglas-
fir (Pseudotsuga menziesii) forests west
of the Cascade Mountains in the Pa-
cific Northwest is an increasingly
controversial topic, arising from a
fundamental conflict. These forests
are extremely valuable sources of
timber; 40 ha of old growth is valued
at about $1.6 million (Meslow et al.
1981). At the same time, conserva-
tionists view old growth as a unique
ecosystem that is nonrenewable un-
der current management practices
(Cutler 1984, Schoen et al. 1981). Old-
growth forests are disappearing; dur-
ing the past 30 years, removal of
Douglas-fir saw timber from western
Oregon and Washington has ex-
ceeded annual growth by a factor of
three (Harris 1984). Now, less than
20% of the original old-growth forest
in the Pacific Northwest remains
(Spies and Franklin in press).

Historically, old-growth forests
were viewed as decadent stands of

'Paper presented at symposium. Man-
agement of Amphibiar^. Reptiles, and
Small Mammals in North America. (Flag-
staff. AZ. July 19-21. 1988.)

'Paul Stephien Corn is Zoologist. USDI Rshi
and Wildlife Service. National Ecology Re-
search) Center. 1300 Blue Spruce Drive. Fort
Collins. CO 80524.

^R. Bruce Bury is Zoologist (Research)).
USDI Fish and Wildlife Service. National Ecol-
ogy Research Center. 1300 Blue Spruce
Drive. Fort Collins. CO 80524.

^Thomas A. Spies is Research Forester.
USDA Forest Service. Pacific Northwest Re-
search Station. Corvallis. OR 9733 1.

wasted timber that provided little
wildlife habitat. For example, Tevis
(1956) stated:

Virgin forest in the Douglas-fir
(Pseudotsuga taxifolia [menzi-
esii]) region of northwestern
California is sterile habitat for
wildlife. Dense shade and
competition from large old
trees prevent the growth of
nearly all bushy and herba-
ceous vegetation except a
weak understory of tan oak
(Lithocarpus densiflora). Food
for animals is scarce.

The value of old growth has been
rehabilitated. Currently, old-growth
Douglas-fir forests are considered
excellent wildlife habitats, dominated
by large trees, but possessing a com-
plex and varied structure (Franklin et
al. 1981, Franklin and Spies 1984),
including some of the highest
amounts of coarse woody debris
(CWD) reported for any forest eco-
system (Spies et al. in press).

Most remaining old growth in the
Pacific Northwest is on Federal land
managed by the Forest Service and
Bureau of Land Management (Harris
1984). The policy of the U.S. Depart-
ment of Agriculture is to "...maintain
viable populations of all existing na-
tive vertebrate populations..." (Cut-
ler 1980) but, until recently, the infor-
mation needed to achieve this goal
did not exist. Most lists of species
with some degree of dependence on

or association with old growth are
incomplete or inferential (e.g., Harris
and Maser 1984, Meslow et al. 1981).
Recent research has improved this
situation, but little of it is directed
toward nongame species. A recent
symposium on wildlife and old-
growth relations (Meehan et al. 1984)
included 27 papers. Two-thirds (17)
of the papers concerned game spe-
cies, and only four papers discussed
ecology of nongame wildlife. Re-
maining pap>ers discussed either
characteristics of old-growth forests
(three papers) or management objec-
tives (three papers).

In 1981, to provide the informa-
tion necessary for managing wildlife
in the national forests of the Pacific
Northwest, the U.S. Forest Service
chartered the Old-Growth Wildlife
Habitat Program^ (OGWHP). Its
goals (Ruggiero and Carey 1984)
were to: (1) identify old-growth for-
ests were unique components of co-
niferous forest ecosystems, (2) iden-
tify the ecological characteristics of
old growth, (3) identify any wildlife
species dependent on old growth for
survival or optimal habitat, and (4)
determine the amount and distribu-
tion of old growth necessary to meet
the needs of dependent species.

Vegetation and vertebrate commu-
nity studies were performed on a
matrix of forest conditions in natu-
rally regenerated stands. Forest de-

^Now the Wildlife Habitat Relationships in
Western Oregon and Washington Project.

340

Figure 1 .—Maps of study areas where pitfall trapping was conducted in 1983. HJAEF = H. J.
Andrews Experirr>ental Forest; WREF = Wind River Experimental Forest. Note ttiat \he scale for
eachi nr>ap is different.

velopment was examined across a
chronosequence, and a moisture gra-
dient was examined for the old-
growth stands.

Field work began in 1983 with
vegetation and vertebrate commu-
nity pilot studies at 30 stands spread
between two sites in the Oregon and
Washington Cascade Mountains. The
primary goal of the first year was to
evaluate and recommend sampling
techniques. The pilot studies were
successful in developing and refining
sampling methods (e.g.. Bury and
Corn 1987, Thomas and West 1984,
West 1985). In 1984 and 1985, com-
munity studies expanded to more
than 180 stands in the Washington
Cascades, the Oregon Cascades, the
Oregon Coast Range, and the
Siskiyou and Klamath mountains of
southern Oregon and northern Cali-
fornia. Since 1985, species-specific
studies have been emphasized,
largely concerning the ecology and
management of the spotted owl
(Strix occidentalis) and its prey base.

Our paper concerns the commu-
nity ecology of small mammals as
revealed by pitfall trapping in 1983.
The data collected in 1983 are useful
for other than evaluating techniques,
but these data are difficult to inte-
grate into 1984 and 1985 results, be-
cause the sampling methods were
changed (Bury and Corn 1987).
Therefore, we report these results
with the caveat that variation be-
tween years is not examined.

Our specific objectives are to ex-
amine the relations of the abundance
of small mammal sf)ecies to the
chronosequence and the moisture
gradient and to identify specific habi-
tat features that contribute to abun-
dance. The effects of forest manage-
ment are also discussed.

METHODS

Study Areas

Forest stands were studied in two
areas on the western slop)es of the

341

Cascade Mountains (fig. 1). Twelve
stands were in the Wind River Ex-
perimental Forest (WREF) or the sur-
rounding Gifford Pinchot National
Forest, Skamania County, Washing-
ton, and 18 stands were in the H. J.
Andrews Experimental Forest
(HJAEF) or Willamette National For-
est, Lane and Linn counties, Oregon.
Appendix A lists ages, elevations,
and locations of all stands.

Stand Selection and Classification

Initial stand selections were made by
OGWHP investigators studying the
structure of old growth (Franklin and
Spies 1984, Spies and Franklin in
press). Age was the primary criterion
for establishing a stand's position on
the chronosequence. Topographic
position and understory vegetation
provided a first approximation of a
moisture gradient (south- or west-
facing ridges were generally dry,
whereas stands on north-facing
slopes were usually moist to wet).
Most stand boundaries were not
highly distinct (e.g., forest islands
surrounded by clear cuts) but were
determined by several factors, in-
cluding age, disturbance history,
vegetational composition, physiogra-
phy, and soils. Stands were first cho-
sen from aerial photographs and for-
est type maps, but an on-site inspec-
tion was completed before any of the
vertebrate sampling plots were estab-
lished. Stand sizes varied from about
10 to 20 ha.

Coarse woody debris (CWD),
vegetation, and site characteristics
were sampled in five nested, circular
plots in each stand (Spies et al. in
press). Classification of downed
CWD (=logs) followed Franklin et al.
(1981) and Maser and Trappe (1984):
from class 1 logs (essentially unde-
cayed) to class 5 logs (well decayed,
appearing as raised hummocks in the
forest floor).

The chronosequence consisted of
four categories beginning with
clearcuts (< 10 years old), closed-can-

opy young stands (30-80 years), ma-
ture stands (80-195 years), and old
growth (195-450 years). The latter
three categories were all composed
of naturally regenerated forests.
Ages of young and mature stands
were estimated by increment coring
of at least five dominant Douglas-fir
trees per stand (Spies et al. in press).
Ages of old-growth stands were esti-
mated from increment cores and by
examining stumps in adjacent
clearcuts and roadsides.

In an ideal chronosequence analy-
sis, age classes should have similar
means and ranges of site characteris-
tics. We were only partly successful
in achieving this goal, because the
age classes were not equally distrib-
uted over the landscajje, and other
criteria such as stand size, accessibil-
ity, and absence of logging activity
took precedence over site uniformity.
Consequently, young and mature
stands spanned a wider range of en-
vironments than originally planned
and for some variables (such as ele-
vation at the HJAEF), the younger
age classes differed from old growth.

We conducted analysis of mois-
ture effects across the old-growth

stands. Adjustments were made to
the preliminary field classification of
dry (OGD), moderate (OGM), and

C3RY Moer
DECORANA AXIS 1

Figure 2.— Detrended correspondence
analysis (DECORANA) of percent occur-
rence of understory plant species in old-
growthi stands in Oregon and Washington.
Stands were placed in nnoisture categories
(wet, moderate, or dry) based on ttieir rela-
tive positions on ttie two gradients.

Figure 3.— A pitfall array in clearcut stand #291 in Oregon, Ptioto by L. Hanebury.

342

wet (OGW), after conducting ordina-
tions of old-growth stands using de-
trended correspondence analysis
(DECORANA). DECORANA is a
weighted average technique that is
computationally related to principal
comp>onents analysis (Gauch 1982).
The percent occurrence of understory
plant species in five 1,000-m^ plots in
each stand was used in separate
analyses of each study area (fig. 2).
The first axis in both areas separated
stands along a moisture gradient cor-
related with indicators of topo-
graphic moisture, such as aspect and
slope. The second axis in both analy-
ses separated stands along a complex
gradient of temperature and mois-
ture and was correlated with eleva-
tion.

Pitfall Trapping

We installed a pitfall trap array (fig.
3) in each stand. An array included
two triads, 25 m apart, each consist-
ing of three 5-m long aluminum drift
fences with screen wire funnel traps
on each side and pitfall traps at each
end. Thus, each array had six fences
and twelve pitfall and twelve funnel
traps. Bury and Com (1987, this vol-
ume) provide more complete de-
scriptions and illustrations.

The traps were opened the last
week in May 1983 and were operated
continuously for 180 days. No water
was put in traps, because this has a
deleterious effect on the preservation
of amphibians, which were a major
target of the traps (Bury and Corn,
this volume ). In practice, most traps
accumulated some water and most
mammals drowned. Traps were
checked initially every three days,
but as trap rate declined over time,
the interval between checks in-
creased to about seven days.

Mammals taken from traps were
identified, sexed, measured, and pre-
served as skulls, skeletons, or skins
and skulls. All specimens from Ore-
gon and most from Washington were
deposited in the National Museum of

Natural History (USNM), where all
identifications were verified. Com-
mon and scientific names used in this
paper follow Banks et al. (1987).

We encountered one problem that
significantly affected the data analy-
sis. The high trap success at the
WREF stands exceeded the field
crew's ability to process sp>ecimens,
and approximately 25% of the mam-
mals were discarded in the field.
When the remaining specimens were
examined later at the USNM, about
10% of the field identifications of
Trowbridge's shrews (Sorex trowbr-
idgii), montane shrews (S. montico-
lus), and vagrant shrews (S. vagrans)
were inaccurate. Thus, the exact
numbers of these shrews captured at
WR are in doubt (Bury and Com
1987), and analyses of overall species
richness and individual abundance
of these sp>ecies were only reported
for Oregon data.

Statistical Analyses

We analyzed the mean abundance
(total number of captures) of each
species, mean total abundance, and
mean species richness across each
gradient with one-way analysis of
variance (ANOVA). No traps were
missing or damaged during the 180-
day trapping period, so it was unnec-
essary to adjust raw abundance for
trap nights. Scavengers may remove
animals from traps when there are
long intervals between checks (M. G.
Raphael, personal communication),
and traps with water may be more
effective than dry pitfalls at captur-
ing rodents with good leaping abil-
ity. Because 70% of all mammals
were captured in the first 60 days of
trapping (Bury and Corn 1987), when
traps were checked frequently, we
feel these considerations are minor
and we made no adjustments to the
data.

Abundances were log transformed
before the ANOVAs were run.
Clearcuts, OGW and OGD stands
were not included in the ANOVA of

the chronosequence. Clearcuts,
young, and mature stands were not
included in the ANOVA on moisture
(Spies et al. in press). A comparison
of species' abundances in clearcuts
versus young stands is presented
separately. Pearson correlation coef-
ficients were calculated between
abundance (transformed as
ln[abundance + 1]) and 24 of the
habitat variables (app>endix B). Per-
centage variables (e.g., % cover of
grasses) were arcsin transformed,
other variables were log trans-
formed. We also performed a princi-
pal components analysis using the
habitat variables, but because the
first three factors explained only 52%
of the variation among stands, we
report only the significant (P < 0.05)
bivariate correlations between abun-
dance and individual habitat vari-
ables. All analyses were performed
using the statistical program SYS-
TAT^ (Wilkinson 1988).

RESULTS

The pitfall arrays were highly effec-
tive at capturing small mammals,
producing 3,877 captures of 27 spe-
cies. Insectivores and microtine ro-
dents were best caught by pitfalls,
while deer mice (Peromyscus manicu-
latus) were under-sampled (Bury and
Com 1987). Captures of each species
in each stand are listed in tables 1
(HJAEF) and 2 (WREF).

Mean species richness (number of
species) varied from about nine in
mature stands to 12 in OGW stands
(fig. 4). There was no significant dif-
ference across either the chronose-
quence or the moisture gradient. To-
tal abundance was highest in young
and mature stands and lowest in
OGM stands, but the difference was
not significant. There was no appar-
ent trend in small mammal abun-
dance across the moisture gradient.

^Trade names are provided for the
benefit of the reader: such use does not
constitute an official endorsement by the
Fish and Wildlife Service.

343

Table 1 —Abundance of small mammals captured at the H. J. Andrews Experimental Forest In Oregon. Arrays of pit-
fall traps with drift fences were operated continuously for 1 80 days in 1 983,

Species Stand no.

Trowbridge's Shrew
Montane Shrew
Vagrant Shrew
Pacific Marsh Shrev/
Northern Water Shrew
Pacific Shrew
Unidentified shrews
Shrew Mole
Coast Mole
Western

Red-backed Vole
Creeping Vole
Red Tree Vole
Water Vole
Heather Vole
Townsend's Vole
Deer Mouse
Pacific Jumping Mouse
Western Pocket Gopher
Others"

Old growth

Wet

Moderate

Dry

Mature

Young

Clearcut

15

03

24

02

17

33

25

29

11

35

42

39

**/

*KJ

75

291 391

33

48

48

76

35

60

75

70

51

56

78

70

139

71

83

18

39

17

lO

1 o

9R
zo

23

13

g

7

19

15

16

15

26

17

14

22

3

8

13

lo

Z

1

9

4

0

7

2

17

3

5

6

1

1

74

7

14

o

1

1

1

7

9

2

2

4

13

1

4

2

1
1

1

1

1

3

5

3

1

1

4

■ . 2

9

4

2

5

1
1

1

4

5

4

6

2

2

1

7

1

2

3

2

14

4

1

2

9

6

3

10

6

15

18

52

13

7

4

1

1

1

2

1

1

2

3

1

5

28

1

3

3

4

1

1

2

1

1

1

1

2

1

1

5

1

2

1

3

3

3

1

3

2

3

3

3

1

2

5

2

3

24

1

1

2

1

3

1

14

1

3

]

2

1

16

1

2

1

2

2

2

1

1

1

1

°Townsend's Chipmunk (8). Northern Rying Squirrel (3), Errriine (2), Spotted Sl(unk (1), Sr)owshoe Hare (1).

r

Table 2.-Abundance of small mammals captured at the Wind River Experimental Forest In Washington, Arrays of pit-
fall traps with drift fences were operated continuously for 180 days in 1983.

Old growth

Species Stand No.

Wet
14

Moderate

12

21

20

Dry
31

Mature

Young

41

42

60

60

61

Clearcut

70

71

Pacific Marsh Shrew 10 3
Other shrews^ 86 73

Shrew Mole 6
Coast Mole 3
Southern

Red-backed Vole 15 10
Creeping Vole 2
Townsend's Vole 1
Other Microtines*^

Deer Mouse 8 16

Pacific Jumping Mouse 2
Northern Pocket Gopher
Others^ 1

2
93
9
3

21
4

40
2

3
46
1

20
2

11

115
2
3

40
3
5

28

6
192
9
4

13
6
2

23
2
2

3
127
2
1

16
9
1

11

3

158
6
1

3
2
1

16
1

2
117
1

2

41
1
1
3
9
1

1

7
97
1

31
4

1

7

1

3
50

4
9

32
1

1

86
3

1

11
5
2
7

4

3

°unider)tified (701 Trowbridge's Shrew (6967). Mor)tane Shrew (35 1 7), Vagrant Shrew (12071 Masked Shrew (7), and Northern Water
Shrew (3).

^unidentified (6). Heather Vole (1).

^Ermine (6). Townsend's CNpmunk (3), Yellow-pine Chipmunk (2). Snowshoe Hare (2), Northern Hying Squirrel (1), Pika (1).

344

species- Habitat Associations

Trowbridge's Shirew

These shrews were the most abun-
dant small mammal (about 46% of all
captures). At HJAEF, this species
was most abundant in young stands
(fig. 5), but the variation across the
chronosequence was not statistically
significant. Most of the high mean
abundance in young stands was due
to one stand (#47) at HJAEF (table 1).
Abundance on the moisture gradient
increased from OGW to (X^D, but
the differences were not significant.

Habitat variables that were posi-
tively correlated with abundance of
Trowbridge's shrews included the
total basal area and mean diameter at
breast height (d.b.h.) of live trees, the
number of decay class 4 and 5 (most
decayed) downed logs, and litter
depth (table 3). Variables negatively

MEAN § OF SPECIES

TOTAL^ /KBUtsO/KNCEI

MEAN TOTAL CAPTURES

Figure 4.— Mean species richness (HJAEF
only) and total abundance (all stands) of
snnall PDommals in closed-canopy stands.

correlated were percent cover by
herbs and grasses and the biomass of
least decayed logs (class 1 and 2).
Montane Stirew

This was the second most abundant
species, occurring in similar numbers
in stands of different ages (fig. 5).
There is a trend on the moisture gra-
dient of decreasing abundance from
OGW to OGD, but the differences
are not significant. Abundance of
montane shrews was positively cor-
related with tree size (MDBH) and
negatively correlated with percent
cover by grasses and number of de-
cay class 1 and 2 logs (table 3).

Vagrant Shirew

Vagrant shrews were significantly
less abundant in older forest stands
(fig. 5, P = 0.02), and variation across
the moisture gradient was not signifi-
cant. This species reached its greatest
abundance in one clearcut (see be-
low). Abundance of vagrant shrews
was negatively correlated with sev-
eral characters associated with old-
growth forests: number of decay
class 4 and 5 logs, percent cover by
mosses, litter depth, and slope (table
3).

Pacific Marsti Shrew

The Pacific marsh shrew (Sorex
bendirii) is a large shrew generally
associated with small streams and
swamps (Maser et al. 1981, Whitaker
and Maser 1976). Our results agree.
The greatest abundance was in (DGW
stands (fig. 5), and the difference
across the moisture gradient was sig-
nificant (P < 0.001). Marsh shrews
were captured (albeit in low num-
bers) in moderate and dry old-
growth stands where the pitfall ar-
rays were some distance from flow-
ing or standing water, but many of
the younger stands in which this spe-
cies occurred (e.g., stands 11, 35, and
75 at the HJAEF) contained streams

or ponds. Variation across the
chronosequence was not significant,
but this may be misleading given the
high abundance in OGYJ stands. Our
study design precluded us from de-
termining whether Pacific marsh
shrews would be abundant in
younger wet stands.

Several habitat variables were as-
sociated with abundance of Pacific
marsh shrews. Positive correlations
reflected older, wet forests and in-
cluded litter depth, total density of
live trees, mean d.b.h., and biomass
of class 4 and 5 logs. The number of
decay class 1 and 2 logs and slope
were negatively correlated with
abundance (table 3).

Stirew Mole

Shrew moles (Neurotrichus gibbsii) are
small moles but are more like shrews
in appearance and habits. Patterns of
their abundance were similar to the
Pacific marsh shrew (fig. 5). Shrew
moles were most abundant in (DGW,
but there were no significant differ-
ences across the moisture gradient or
the age gradient. Unlike the marsh
shrew, none of the habitat variables
were correlated with abundance.

Coast Mole

We captured 59 coast moles (Scapa-
nus orarius), a form rarely taken by
conventional snap- or live-trapping
techniques. This species might be
more active on the surface than other
moles (Maser et al. 1981), or our drift
fences (which were sunk 20-30 cm
into the ground) might have inter-
rupted their burrowing (Williams
and Braun 1983). There was no sig-
nificant variation on the chronose-
quence, but there was on the mois-
ture gradient (P = 0.05). Coast moles
were most abundant in OGM and
OGD stands and were virtually ab-
sent from (DGW stands (fig. 5).

Coast moles might prefer well-
drained soils (Maser et al. 1981). This

345

is supported by their low abundance
in OGW stands where soils are satu-
rated for long periods. Abundance of
coast moles was positively correlated
with percent cover by deciduous
trees. Habitat variables negatively
correlated were the number of decay
class 3 logs and the number of large-
diameter logs.

Red- Backed Voles

We captured two species of red-
backed voles: the southern red-
backed vole (Clethrionomys gap-peri) at
WREF, and the western red-backed
vole (C. californicus) at HJAEF. We
caught more southern than western
red-backed voles (fig. 6), but the pat-
terns of abundance were similar.
Both species were combined in the
ANOVAs to maximize the sample
size. No differences were detected on
either the age or moisture gradients.

Habitat variables were tested
separately for each species, but the
results were similar (table 4). Abun-
dance of western red-backed voles
was positively correlated with total
basal area of live trees, mean d.b.h.,
and percent cover by evergreen
shrubs (mainly Oregon grape, Ber-
beris spp., and salal, Gaultheria shal-
lon).

Negative correlations were with
grass cover, biomass of decay class 1
and 2 logs, and aspect (abundance
was greatest on southern exposures).
Southern red-backed voles were
positively correlated with density
and basal area of live trees, and mean
d.b.h., and were negatively corre-
lated with grass cover.

Red Tree Vole

This species has been identified as an
old growth associate (Meslow et al.
1981) and is a major food item of the
spotted owl (Forsman et al. 1984).
We captured only 17 red tree voles
(Arborimus longicaudus) in the stan-
dard arrays, too few to run the

ANOVA. But, 12 voles were cap-
tured in the eight old-growth stands
at HJAEF, compared to only five
voles in the 10 younger stands (G =
4.73, P < 0.05). Corn and Bury (1986)
provide a more detailed account of
these results.

Creeping Vole

Creeping voles (Microtus oregoni)
were uncommon in closed-canopy
stands (fig. 6), and there was no dif-
ference in abundance on either gradi-
ent. As with vagrant shrews, this
species was more abundant in

Figure 5.— Mean abundance of insectivores in closed-canopy forest stands. Data for Trowbr-
idge's, montane, and vagrant stirews are from HJAEF only. Pacific nrrarsti stirews, stirew
moles, and coast moles use data from all stands.

346

I

Table 3.— Significant (P < 0.05) Pearson correlations of insectivore abun-
dance and stand structure and vegetation variables. See appendix B for
descriptions of ttie variables.

Positive

Negative

Species

Variable

r

Variable

TrnwhriHop'*; Shrpw

TOTBA

HFRR

-0 7S

(n= 17-18)

LNDC45

0.59

GRASS

-0.70

MDBH

0,57

L6DC12

-0.53

LfTTER

0 49

Montane Shrew

MDBH

0.51

GRASS

-0.52

(n= 17-18)

LNDC12

-0.47

Vagrant Shrew

LNDC45

-0.55

(n= 17-18)

MOSS
SLOPE
UTTER

-0.50
-0.51
-0.50

Pacific Marsh Shrew

LITTER

0.41

LNDC12

-0,50

(n = 28-30)

TOTDEN

0.41

SLOPE

-0.37

MDBH

LBDC45

DECTR

0.44
0.40
0,52

Coast Mole

-0.43

LNDC3

(n = 28-30)

LNDM3

-0.43

Figure 6.— Mean abundance of rodents in closed-canopy forest stands. Data from all stands
were used.

clearcuts. Reflecting this, creeping
vole abundance was positively corre-
lated with percent cover by grasses
and negatively correlated with sev-
eral "forest" variables: number and
biomass of decayed logs, density, ba-
sal area and d.b.h. of live trees, and
litter depth.

Deer Mouse

Although pitfall traps are not as ef-
fective for catching deer mice as snap
traps (Williams and Braun 1983, Bury
and Corn 1987), we caught moderate
numbers of this species, particularly
at WREF (table 2). Deer mice were
most abundant in OGM stands and
least abundant in OGW and young
stands. Differences were not signifi-
cant on either the chronosequence or
the moisture gradient. Deer mouse
abundance was negatively correlated
with percent of coarse fragments in
the soil.

Clearcuts Versus Forests

Pitfall arrays were installed in five
clearcuts, three at HJAEF and two at
WREF. We compared the relative
abundance of several of the common
small mammals in clearcuts and
young stands (fig. 7). Trowbridge's,
montane, and vagrant shrews were
compared only for the three clearcuts
and four young stands at HJAEF.

Southern and western red-backed
voles were virtually absent from
clearcuts, while creeping voles were
more than six times more abundant
in clearcuts than in young stands.
Most insectivores were two to six
times more abundant in young
stands, but vagrant shrews were
most abundant in clearcuts. Much of
the difference in the relative abun-
dance of vagrant shrews is due to
their great abundance in clearcut
#391 at HJAEF (table 1). Only one
vagrant shrew was captured at each
of the other clearcuts at HJAEF. Al-
though roughly eshmated, vagrant

347

shrews were the most common small
mammal at both of the clearcuts at
WREF. Deer mice were about three
times more abundant in clearcuts
than in young stands. A few pocket
gophers {Thomomys mazama at
HJAEF, T. talpoides at WREF) were
captured and are not depicted in fig-
ure 7. Most pocket gophers (20/28)
were captured in clearcuts; none
were captured in old growth.

DISCUSSION

Old-Growth Species

Answering the question of if a spe-
cies is dependent on old-growth for-
est for critical habitat is complex, in-
corporating several aspects of ecol-
ogy and needs to account for tempo-
ral and random variation (Carey
1984). Also, abundance of individual
species within a specific region de-
pend not only on the multidimen-
sional niche, but on the geographic
distribution of each species (Brown,
1984). The community ecology stud-
ies of the Old-Growth Program were
not intended to provide definite an-
swers on old-growth dependencies,
but rather the results were to be used
as guides for designing species-spe-
cific research (Ruggiero and Carey
1984). Our results are based on one
season's data and must be inter-
preted cautiously, but they are useful
for comparison with other studies
and for suggesting new research.

Only one small mammal, the red
tree vole, displayed a significant as-
sociation with old-growth stands,
and the sample size for it was small.
Additional captures of this species in
the Oregon Coast Range in 1984-1985
were almost exclusively in old-
growth forests (Com and Bury, un-
published data). Recent studies of
vertebrates across a similar chronose-
quence of Douglas-fir forests in
northern California (Raphael 1984,
this volume, Raphael and Barrett
1984) found significant positive cor-
relations between abundance of sev-

eral species and stand age: Trowbr-
idge's shrews. Pacific shrews (Sorex
pacificus), coast moles, shrew moles,
Allen's chipmunks (Tamias senex),

Townsend's chipmunks (T. town-
sendii), Douglas' squirrels (Tamias-
ciurus douglasii), dusky-footed
woodrats (Neotoma fuscipes), deer

Table 4.— Significant (P < 0.05) Pearson correlations of rodent abundance
and stand structure and vegetation variables. See appendix B for descrip-
tions of the variables.

Positive

(n = 28-30)

Negative

Species

Variable

r

Variable

r

Western Red-backed Vole

TOTBA

0.66

GRASS

-0.54

(n= 17-18)

MDBH

0.56

TRASPECT

-0.51

EGSHR

0.48

LBDC12

-0.53

Southern red-backed Vole

TOTDEN

0.78

GRASS

-0.81

(n= 11-12)

TOTBA

0.70

MDBH

0.71

Creeping Vole

GRASS

0.51

LNDC45

-0.58

(n = 28-30)

LBDC45

-0.43

MDBH

-0.52

TOTDEN

-0.40

TOTBA

-0.49

irrrER

-0.62

Deer Mouse

TOTOF

-0.36

15.4

YOUNG

Z RB5-aACKED PACFC SHREW TROW- UCNXM^ COAST

^ VOLES MARSH MOLE BHDQE3 SHREW MOLE DBW VAQRAMT

LU . SHREW SHCW MOUSE ShFEW VOLE

Figure 7.— Relative mean abundance of snnall mamrrxjls In young stands and clearcuts.
Species more abundant in young stands are above \he horizontal, species more abundant
In clearcuts below. Values are ttie greater mean abundance divided by the lesser, so, for
example, red-backed voles were 15.4 times more abundant In young stands than in
clearcuts.

348

mice, western red-backed voles, and
fishers (Martes penmnti). Many of
these correlations were not strong,
however, with most species repre-
sented in the youngest forest stages.
Mean species richness was about 10
in all forest age classes. Analysis of
the similarity of species composition
showed little difference on the
chronosequence (Raphael 1984). This
is very similar to our results and sug-
gests that old-growth forests do not
harbor unique communities of small
mammals.

Anthony et al. (1987) snaptrapped
small mammals in riparian zones of
old-growth, mature, and young
stands at HJAEF in 1983. They found
greater abundance of deer mice in
old-growth stands, but Pacific
shrews (S. pacificus) were evenly dis-
tributed. They trapped 14 other spe-
cies, though none in sufficient num-
bers to analyze. Although both An-
thony et al. (1987) and Raphael (1984)
found more deer mice in older for-
ests, this species is ubiquitous and
reaches its highest densities in the
Pacific Northwest in clearcuts (see
below).

Smell Mammals in Managed
Forests

Most studies of habitat relations of
small mammals in the Pacific North-
west have compared clearcuts to for-
ested stands. Although there is con-
siderable variation among studies,
general trends are similar, likely re-
lated to the variety of factors exam-
ined (time since logging, burned, un-
bumed, herbicides applied, etc.).
Populations of deer mice, creeping
voles, and Townsend's chipmunks
increase after logging, while red-
backed voles and Trowbridge's
shrews decline (Anthony and Morri-
son 1985, Gashwiler 1959 1970,
Hooven and Black 1976, Sullivan and
Krebs 1980, Raphael, this volume,
Tevis 1956). Red-backed voles are
probably most affected by clearcut-
ting. Western red-backed voles are

obligate fungivores, and their food
supply disappears after clearcutting
(Maser et al. 1978, Ure and Maser
1982). Gunther et al. (1983) found
southern red-backed voles to be the
most common animals on the
clearcuts they trapped, but they
trapped only three months after log-
ging and probably were sampling a
residual p)opulation. Also, this spe-
cies is less dependent on fungi (Ure
and Maser 1982) and might be able to
persist for a time after logging.

Other studies have not noted the
increase of vagrant shrews in
clearcuts that we observed. Several
factors might be involved, including
random variation. Although mean
abundance was high, vagrant shrews
were rare (one capture each) on two
of our five clearcuts. Other studies
probably underestimated shrew
abundance, because they used either
snap or live traps. Also, some inves-
tigators might have followed Findley
(1955) and considered montane and
vagrant shrews to be the same spe-
cies.

Changes in small mammal com-
munities after logging can be dra-
matic, but clearcuts per se might not
be the main factor influencing species
diversity in managed forest land-
scajjes in the Pacific Northwest. In a
managed forest with a 90-year rota-
tion, about 30% of the area will be in
clearcuts and young plantations lack-
ing canopy closure. The remaining
70% of the landscape will be in
stands 30-90 years old that have
closed forest canopies. The habitat
characteristics of these forest planta-
tions will be a major determinant of
biological diversity in managed
lands. For example, the extensive
logging of low-elevation old-growth
forests in Oregon has probably elimi-
nated much of the habitat of red tree
voles. The giant Douglas-fir trees,
which seem to be preferred as nest
sites, will not occur in managed for-
ests. Meanwhile, the heather vole
(Phemcomys intermedius), a species of
alpine meadows, might be benefit-
ting from increased logging of high-

elevation forests (Com and Bury
1988).

Although we have found few dif-
ferences between old-growth and
younger naturally regenerated for-
ests for small mammals or the herpe-
tofauna (Bury and Com, this vol-
ume), the same probably cannot be
said for comparisons of old-growth
to managed forests. Our analysis of
habitat variables revealed that abun-
dance of several species was corre-
lated with habitat features that
would be absent or greatly reduced
in managed forests. Aside from large
trees, CWD is the primary compo-
nent of old growth that is eliminated
by current forestry practices (Harris
et al. 1982, Spies et al. in press). CWD
is correlated to abundance of shrews
(this study), salamanders (Bury and
Com, this volume, Raphael 1984),
and probably is required habitat for
red-backed voles (Maser and Trappe
1984). Bury and Corn (this volume)
provide further discussion of the role
of CWD as wildlife habitat.

Research Needs

These types of community ecology
studies provided baseline data on
nongame wildlife in naturally regen-
erated forests of the Pacific North-
west. For example, we can use the
data on abundance and the correla-
tions with habitat variables to begin
classifying species as to their degree
of rarity (Rabinowitz et al. 1986).
Species with small geographic distri-
butions, restricted habitat specificity,
and small local populations (e.g., red
tree voles. Pacific marsh shrews) are
likely to be affected by habitat altera-
tion. Species with large populations,
broad habitat specificity, and either
large (deer mice) or small (Trowbr-
idge's shrews) geographic distribu-
tions, are less likely to be affected by
forest management.

Our study does not address
changes in habitats in managed for-
ests stands or the effects of forest
fragmentation as remaining old

349

growth is harvested. Further studies
of small mammals should emphasize
managed stands and managed land-
scapes.

Even with the creation of old-
growth habitat areas on National
Forests, most of the landscape will
probably be in plantations less than
100 years old. Research needs to be
focused on the degree of loss of di-
versity in these managed forests and
evaluate silvicultural options for
maintaining or enhancing habitat
structure.

Thus far, there is little evidence
that small mammal populations in
Douglas-fir forests are strongly influ-
enced by stand size or amount of in-
sularization (Raphael 1984, Rosen-
berg and Raphael 1986). As these au-
thors point out, however, forest frag-
mentation in western coniferous for-
ests might not have advanced far
enough or existed long enough for
effects to be observed. Conversely,
forest fragmentation in the Pacific
Northwest is not usually conversion
of forest to farmland or urban areas
as is the case in other temperate re-
gions (e.g., Wilcove et al. 1986,
Askins et al. 1987, Dickman 1987).
Rather, it results in the replacement
of one forest habitat with another.
Patches of old growth in a managed
forest are not strict analogs of oce-
anic islands or isolated mountain
tops (Harris 1984), so the ability of
forest-floor small mammals to main-
tain populations in managed forests
is dependent on habitat availability
after logging.

Our results indicate that some
"old-growth species" are found in
younger stands, but the proximity of
old growth to younger forest might
partly explain their occurrence. The
effect of stand size, shape, edge, and
juxtaposition on small mammal
populations needs attention. Where
old growth and other habitat areas
are set aside to maintain biological
diversity in intensively managed
landscapes, the long-term viability of
these habitats and their vertebrate
populations needs to be monitored.

ACKNOWLEDGMENTS

We thank S. Boyle, L. Hanebury, D.
Hayes, S. Martin, T. Olson, and S.
Woodis for helping to install pitfall
arrays. J. Dragavon, L. Jones, P. Mor-
rison, R. Pastor, and D. Smith
checked traps and processed ani-
mals. R. Fisher of the USNM verified
all identifications. A. McKee and J.
Moreau, H. J. Andrews Experimental
Forest, and staff at the Wind River
Experimental Forest and Carson Na-
tional Fish Hatchery assisted with
housing and logistics. We appreciate
the critical review of this manuscript
by M. Bogan, A. Carey, M. Raphael,
and F. Samson. This is contribution
number 66 of the Wildlife Habitat
Relationships in Western Washington
and Oregon Project.

LITERATURE CITED

Anthony, R. G., E. D. Forsman, G. A.
Green, G. Witmer, and S. K. Nel-
son. 1987. Small-mammal popula-
tions in riparian zones of conifer-
ous forests in western Oregon.
Murrelet 68:94-102.

Anthony, Robert G., and Michael L.
Morrison. 1985. Influence of
glyphosphate herbicide on small
mammal populations in western
Oregon. Northwest Science 59:159-
168.

Askins, Robert A., Margarett J. Phil-
brick, and David S. Sugeno. 1987.
Relationship between the regional
abundance of forest and the com-
position of forest bird communi-
ties. Biological Conservation
39:129-152.

Banks, Richard C, Roy W. McDi-
armid, and Alfred L. Gardner.
1987. Checklist of vertebrates of
the United States, the U.S. Territo-
ries, and Canada. USDI Fish and
Wildlife Service, Resource Publica-
tion 166. 79 p. Washington, D.C.

Brown, James H. 1984. On the rela-
tionship between abundance and
distribution of species. American
Naturalist 124:255-279.

Bury, R. Bruce, and Paul Stephen
Corn. 1987. Evaluation of pitfall
trapping in northwestern forests:
trap arrays with drift fences. Jour-
nal of Wildlife Management
51:112-119.

Carey, Andrew B. 1984. A critical
look at the issue of species-habitat
dependency, p. 346-351. In Pro-
ceedings, Society of American For-
esters national convention. [Port-
land, Or., October 16-20, 1983] So-
ciety of American Foresters,
Washington, D.C.

Com, Paul Stephen, and R. Bruce
Bury. 1986. Habitat use and terres-
trial activity by red tree voles (Ar-
borimus longicaudus) in Oregon.
Journal of Mammalogy 67:404-406.

Corn, Paul Stephen, and R. Bruce
Bury. 1988. Distribution of the
voles Arborimus longicaudus and
Phenacomys intermedius in the Cen-
tral Oregon Cascades. Journal of
Manunalogy 69:427-429.

Cutler, M. Rupert. 1980. A wildlife
policy for the U.S. Department of
Agriculture. Transactions of the
North American Wildlife and
Natural Resources Conference
45:56-66.

Cutler, M. Rupert. 1984. Keynote ad-
dress. Old-growth habitat: an en-
dangered species, p. 205-211. In
William R. Meehan, Theodore R.
Merrell, Jr., and Thomas A.
Hanley, eds.. Fish and wildlife re-
lationships in old-growth forests.
Proceedings of a symposium. [Jun-
eau, Alaska, April 12-15, 1982] 425
p. American Institute of Fishery
Research Biologists, (available
from John W. Reintjes, Rt. 4, Box
85, Morehead City, NC 28557).

Dickman, C. R. 1987. Habitat frag-
mentation and vertebrate species
richness in an urban environment.
Journal of Applied Ecology 24:337-
351.

Findley, James S. 1955. Speciation of
the wandering shrew. Publication
of the University of Kansas Mu-
seum of Natural History 9:1-68.

Forsman, Eric D., E. Charles Meslow,
and Howard M. Wight. 1984. Dis-

350

tribution and biology of the spot-
ted owl in Oregon. Wildlife Mono-
graphs 87:1-64.

Franklin, J. F., K. Cromack, Jr., W.
Denison, A. McKee, C. Maser, J.
Sedell, F. Swanson, and G. Juday.
1981. Ecological characteristics of
old-growth Douglas-fir forests.
USDA Forest Service General
Technical Report PN[W-118, 48 p.
Pacific Northwest Forest and
Range Experiment Station, Port-
land, Or.

Franklin, Jerry F., and Thomas A.
Speis. 1984. Characteristics of old-
growth Douglas-fir forests, p. 328-
334. In Proceedings, Society of
American Foresters national con-
vention. [Portland, Or., October
16-20, 1983] Society of American
Foresters, Washington, D.C.

Gashwiler, Jay S. 1959. Small mam-
mal study in west-central Oregon.
Journal of Mammalogy 40:128-139.

Gashwiler, Jay S. 1970. Plant and
mammal changes on a clearcut in
west-central Oregon. Ecology
51:1018-1026.

Gauch, Hugh G., Jr. 1982. Multivari-
ate analysis in community ecol-
ogy. Cambridge University Press,
Cambridge, England. 298 p.

Gunther Pamela M., Brenda S. Horn,
and Geoffrey D. Babb. 1983. Small
mammal populations and food
selection in relation to timber har-
vest practices in the western Cas-
cade Mountains. Northwest Sci-
ence 57:32-44.

Harris, Larry D. 1984. The frag-
mented forest. Island biogeogra-
phy theory and the preservation of
biotic diversity. University of Chi-
cago Press, Chicago, 111. 211 p.

Harris, Larry D., and Chris Maser.
1984. Animal community charac-
teristics, p. 44-68. In The frag-
mented forest. Island biogeogra-
phy theory and the preservation of
biotic diversity. University of Chi-
cago Press, Chicago, 111. 211 p.

Harris, Larry D., Chris Maser, and
Arthur McKee. 1982. Patterns of
old growth harvest and implica-
tions for Cascades wildlife. Trans-

actions of the North American
Wildlife and Natural Resources
Conference 47:374-392.
Hooven, Edward F., and Hugh C.
Black. 1976. Effects of some
clearcutting practices on small-
mammal populations in western
Oregon. Northwest Science 50:189-
208.

Maser, Chris, Bruce R. Mate, Jerry F.
Franklin, and C. T. Dymess. 1981.
Natural history of Oregon coast
mammals. USDA Forest Service
General Technical Report PNW-
133. 496 p. Pacific Northwest For-
est and Range Experiment Station.
Portland, Or.

Maser, Chris and James M. Trappe.
1984. The seen and unseen world
of the fallen tree. USDA Forest
Service General Technical Report
PNW-164. 56 p. Pacific Northwest
Forest and Range Experiment Sta-
tion. Portland, Or.

Maser, Chris, James M. Trappe, and
Ronald A. Nussbaum. 1978. Fun-
gal-small mammal interrelation-
ships with emphasis on Oregon
coniferous forests. Ecology 59:799-
809.

Meehan, William R., Theodore R.
Merrell, Jr., and Thomas A.
Hanley, eds. 1984. Fish and wild-
life relationships in old-growth
forests. Proceedings of a sympo-
sium. [Juneau, Alaska, April 12-15,
1982] 425 p. American Institute of
Fishery Research Biologists, (avail-
able from John W. Reintjes, Rt. 4,
Box 85, Morehead City, N.C.
28557).

Meslow, E. Charles, Chris Maser,
and Jared Verner. 1981. Old-
growth forests as wildlife habitat.
Transactions of the North Ameri-
can Wildlife and Natural Re-
sources Conference 46:329-335.

Rabinowitz, Deborah, Sara Cairns,
and Theresa Dillon. 1986. Seven
forms of rarity and their frequency
in the British Isles, p. 182-204. In
Michael E. Soule, ed.. Conserva-
tion biology. The science of scar-
city and diversity. 584 p. Sinauer
Associates, Sunderland, Mass.

Raphael, Martin G. 1984. Wildlife
populations in relation to stand
age and area in Douglas-fir forests
of northwestern California, p. 259-
274. In William R. Meehan, Theo-
dore R. Merrell, Jr., and Thomas

A. Hanley, eds.. Fish and wildlife
relationships in old-growth for-
ests. Proceedings of a symposium.
[Juneau, Alaska, April 12-15, 1982]
425 p. American Institute of Fish-
ery Research Biologists, (available
from John W. Reintjes, Rt. 4, Box
85, Morehead City, N.C. 28557).

Raphael, Martin G., and Reginald H.
Barrett. 1984. Diversity and abun-
dance of wildlife in late succes-
sional Douglas-fir forests, p. 352-
360. In Proceedings, Society of
American Foresters national con-
vention. [Portland, Or., October
16-20, 1983] Society of American
Foresters, Washington, D.C.

Rosenberg, Kenneth V., and Martin
G. Raphael. 1986. Effects of forest
fragmentation on vertebrates in
Douglas-fir forests, p. 263-272. In
Jared Verner, Michael L. Morri-
son, and C. John Ralph, eds..
Wildlife 2000: modeling habitat
relationships of terrestrial verte-
brates, [symposium held October
7-11, 1984] 470 p. University of
Wisconsin Press, Madison.

Ruggiero, Leonard F., and Andrew

B. Carey. 1984. A programmatic
approach to the study of old-
growth forest — wildlife relation-
ships, p. 340-345. In Proceedings,
Society of American Foresters na-
tional convention. [Portland, Or.,
October 16-20, 1983] Soc. of Amer-
ican Foresters, Washington, D.C.

Schoen, John W., Olof C. Wallmo,
and Matthew D. Kirchoff. 1981.
Wildlife-forest relationships: is a
reevaluation of old growth neces-
sary? Transactions of the North
American Wildlife and Natural
Resources Conference 46:531-544.

Spies, Thomas A. 1983. Study plan.
Characterization of old-growth
Douglas-fir forests of western Ore-
gon and Washington. 22 p. USDA
Forest Service, Corvallis, Or.

351

Spies, Thomas A., and Jerry F. Fran-
klin. In press. Old growth and for-
est dynamics in the Douglas-fir
region of western Oregon and
Washington. Natural Areas Jour-
nal.

Spies, Thomas A., Jerry F. Franklin,
and Ted B. Thomas. In press.
Coarse woody debris in Douglas-
fir forests of western Oregon and
Washington. Ecology.

Sullivan, Thomas P., and Charles J.
Krebs. 1980. Comparative demog-
raphy of Peromyscus maniculatus
and Microtus oregoni populations
after logging and burning of
coastal forest habitats. Canadian
Journal of Zoology 58:2252-2259.

Tevis, Lloyd, Jr. 1956. Responses of
small mammal populations to log-
ging of Douglas-fir. Journal of
Mammalogy 37:189-196.

Thomas, Donald W., and Stephen D.
West. 1984. On the use of ultra-
sonic detectors for bat species
identification and the calibration
of QMC Mini Bat Detectors. Cana-
dian Journal of Zoology 62:2677-
2679.

Ure, Douglas C, and Chris Maser.
1982. Mycophagy of red-backed
voles in Oregon and Washington.
Canadian Journal of Zoology
60:3307-3315.

West, Stephen D. 1985. Differential
capture between old and new
models of the Museum Special
snap trap. Journal of Mammalogy
66:798-800.

Whitaker, John O., Jr., and Chris Ma-
ser. 1976. Food habits of five west-
ern Oregon shrews. Northwest
Science 50:102-107.

Wilcove, David

system for statistics. Version 4.0.
Systat Inc., Evanston, 111. 822 p.
Williams, Daniel F., and Suzanne E.
Braun. 1983. Comparison of pitfall
and conventional traps for sam-
pling small mammal populations.
Journal of Wildlife Management
47:841-845.

S., Charles
H. McLellan,
and Andrew
P. Dobson.
1986. Habitat
fragmenta-
tion in the
temperate
zone. p. 237-
256. In Mi-
chael E.
Soule, ed..
Conservation
biology. The
science of
scarcity and
diversity.
584 p. Sin-
auer Associ-
ates, Sunder-
land, Mass.
Wilkinson, Le-
land. 1988.
SYSTAT. The

Table Al .—Ages, elevations, and locations of 18 stands in
Oregon. Locations are distances (km) from McKenzie Bridge.

Stand

Type

Age (yr)

Elev. (m)

Location

15

OGW

450

795

Unn Co., 9.2 N, 1.2 W

03

OGW

450

815

Lane Co., 6.4 N, 0.6 E

24

OGW

450

860

Lane Co.. 6.0 N, 0,6 E

02

OGM

450

560

Lane Co., 4.6 N. 6.5 W

17

OGM

450

790

Lane Co., 6.6 N, 0.3 W

33

OGD

200

670

Lane Co., 6.0 N, 7.5 E

25

OGD

195

500

Lane Co.. 2.4 N, 7.5 W

29

OGD

200

700

Lane Co.. 2.6 S

11

Mature

140

670

Lane Co.. 5.4 S, 8.8 E

35

Mature

130

900

Linn Co., 10.5 N, LOW

42

Mature

150

1030

Lane Co.. 3.1 N,3.0W

39

Young

76

1050

Lane Co., 4.4 S, 14,3 E

47

Young

50

1110

Lane Co., 3.3 N,2.4W

48

Young

69

1075

Unn Co,, 13.2 N,0,8E

75

Young

30

560

Lane Co., 1.6 S, 5,2 W

55

Clearcut

9

830

Lane Co., 2.8 N,6.6W

291

Clearcut

5

690

Lane Co., 2.6 S, 1.4E

391

Clearcut

5

1100

Lane Co., 3.8 S, 14.8 E

r

Table A2.— Ages, elevations, and locations of 12 stands in
Wastiington. Locations are distances (km) from Carson, Ska-

mania Co.

Stand

Type

Age (yr)

□ev. (m)

Location

14

OGW

375

520

17.7N, 16.9W

12

OGM

450

485

6.4 N, 11,3W

21

OGM

375

440

17.2 N, 14.0 W

20

OGM

375

480

11.3 N, 11.9 W

31

OGD

375

670

18.5 N, 16,5 W

41

Mature

105

485

19.3 N, 13,7 W

42

Mature

140

500

13.7N,2.4W

50

Mature

130

610

16,0 N, 2.1 W

60

Young

65

475

13.6 N, 12.1 W

61

Young

65

640

8.1 N,6.3W

70

Clearcut

5

535

1 1.3 N, 13.4 W

71

Clearcut

5

730

16,9 N. 7.2 W

Table Bl .—Stand structural and vegetation variables.

Variable name Description

SLOPE

TRASPECT

LNDC12

LNDC3
LNDC45

LNDM1
LNDM2

LNDM3
LBDC12

LBCD3
LBDC45

MDBH
TOTDEN
TOTBA
LiTFER
TOTCF
MOSS
FERN
GRASS
HERB
EGSHR
DESHR
EVGTR
DECTR

Percent slope

Transformed aspect

Number of logs per tia, decay class 1

and 2

Number of logs per ho, decay class 3
Number of logs per t^a, decay class 4
and 5

Number of logs per ha. <30cm diameter
Number of logs per ha. >30cm and <60
cm

Number of logs per ha. >60 cm
Biomass (1 ,000 kg per ha) of logs, class 1
and 2

Biomass (1 .000 kg per ha) of logs, class 3
Biomass (1 ,000 kg per ha) of logs, class 4
and 5

Mean d.b.h. (cm) in stand

Density of live trees (number per ha)

Basal area of live trees (m' per ha)

Litter depth (01 + 02 horizons; cm)

Volume (%) of coarse fragments in soil

% cover by mosses

% cover by fems

% cover by grasses

% cover by herbaceous vegetation

% cover by evergreen shrubs

% cover by deciduous shrubs

% cover by evergreen trees

% cover by deciduous trees

352

Evaluation of Small Mammals
as Ecological Indicators of
Old-Growth Conditions^

Kirk A. Nordyke^ and Steven W. Buskirk^

Abstract.— The use of small mammals as
ecological indicators of old-growth conditions was
evaluated from trapping studies conducted in forest
stands reflecting a range of old-growth conditions in
southeastern Wyoming. The relationship between
abundance of Clethrionomys gapperi an6 old-
growth conditions was expressed in a quadratic
function. Tamias minimus and Peromyscus
maniculatus were negatively correlated with old-
growth conditions. C. gapperi '\s the most likely
candidate for a small mammal ecological indicator
of old-growth conditions in spruce-fir stands.

Recent emphasis in forest manage-
ment has been placed on an inte-
grated multiple-benefit approach to
land and resource planning and man-
agement (Salwasser et al. 1982). The
National Forest Management Act
(NFMA) was enacted in 1976 to es-
tablish revised goals for the USDA
Forest Service. NFMA regulations
require that detailed plans be devel-
oped and implemented in each na-
tional forest. A specific goal is to
manage wildlife and fish habitats to
maintain viable populations of all
existing native vertebrate species in
the planning area and to maintain
and improve habitats of management
indicator species (MIS) (36 CFR
219.19). In addition, population
trends of MIS are to be monitored
and relationships of those trends to
habitat changes must be determined
(36 CFR219.19[a][6]).

Ecological indicator species com-
prise one category of MIS and were
defined for management purposes as
"...plant or animal species selected
because their population changes are

'Paper presented at symposium. Man-
agement of Amphibians. Reptiles, and
Small Mammals in Nortt^ America. (Flag-
staff. AZ. July 19-21. 1988.)

'Kirk A. Nordyke is a graduate student.
The University of Wyoming. Department of
Zoology and Physiology. University Station
Box 3 166. Laramie. W/ 82071.

^Steven W. Buskirk is Assistant Professor.
The University of Wyoming. Department of
Zoology and Physiology. University Station
Box 3166. Laramie. WY 82071.

believed to indicate the effects of
management activities on other spe-
cies of selected major biological com-
munities or on water quality" (36
CFR 219.19[a][l]). Ecological indica-
tors should have a high degree of
sensitivity to perturbation and be
representative of habitat needs of
other species (Patton 1987). Thus,
population responses of an ecological
indicator species to habitat perturba-
tions should reflect similar, yet less
severe, responses in more tolerant
species (Graul and Miller 1984). Eco-
logical indicators should be easily
monitored to achieve realistic goals.

Compliance with the monitoring
requirements of NFMA presents a
major challenge to national forest
management because costs may be
high and because methods are still
being developed (Verner 1983). The
challenge is most pressing in old-
growth forests: this important habitat
is disappearing at an alarming rate
and vertebrate populations depend-
ent on old-growth features are de-
clining (Harris 1984). NFMA guide-
lines mandate that old -growth be a
significant element in the diversity of
forest conditions. To accomplish this,
old-growth and associated fauna
must be characterized and monitored
to determine that management prac-
tices will not impair their productiv-
ity (Juday 1978).

Of 22 species selected as ecological
indicator species for the Medicine
Bow National Forest (MBNF), Gap-
Iyer's red-backed vole (Clethrionomys

gapperi) is the only small mammal
ecological indicator for old-growth
conditions (USDA Forest Service
1985). Old-growth forests represent
optimal habitat for C. gapperi (Jerry
1984). Limiting factors to habitat use
by C. gapperi may include require-
ments for water (Getz 1968, Merritt
1981) and log cover (Tevis 1956).
Old-growth generally exhibits more
mesic conditions than other forest
habitats. Logs provide cover from
predators and weather (Maser et al.
1979), pathways into new habitats
(Franklin et al. 1981), and mesic sites
for fungal growth (Maser and Trappe
1984). The importance of fungi as a
food for C. gapperi has only recently
been recognized (Martell 1981, Maser
et al. 1978a). Mesic conditions of old-
growth stands favor the occurrence
of fungi (Maser et al. 1978b).

The indicator species concept was
adopted by the Forest Service in the
late 1970s, but its viability as a moni-
toring approach has not been investi-
gated. Certain parameters of C. gap-
peri |X)pulations were assumed to
reflect changes in old-growth condi-
tions that result from management
activity. This paper describes a study
investigating the application of the
indicator species concept to old-
growth management. Oir objective
was to evaluate the responses of
small mammal populations to a
range of old-growth conditions. Spe-
cifically, we investigated whether
abundance of C. gapperi was related
to old-growth condition.

353

study Area and Methods

Our study area in the Snowy Range
included upper montane (2300-2750
m) and subalpine (2750 m-timberline)
zones. Lodgepole pine (Pinus con-
torta) was the dominant overstory
species in the montane zone; it also
dominated south slopes and ridge
tops at higher elevations (Romme
and Knight 1981). Engelmann spruce
(Picea engelmannii) and subalpine fir
(Abies lasiocarpa) were generally co-
dominant in the subalpine zone (Al-
exander 1974). Engelmann spruce
and subalpine fir are climax species
(Romme and Knight 1981) and often
develop old-growth conditions. Old-
growth conditions in lodgepole-
dominated stands are less common
due to a shorter fire interval and a
slower rate of succession (Romme
and Knight 1981). Understory vege-
tation was sparse and generally con-
sisted of common juniper (Juniperus
communis) and broom huckleberry
(Vaccinium scoparium).

Field studies were conducted in
the MBNF (fig. 1) from June to Sep-
tember of 1986 and 1987. We estab-
lished eight study plots in spruce-fir
stands reflecting a range of old-
growth conditions. Because plots
were located on both the east and
west slopes, a paired design was
used to control for the effects of ma-
jor relief. A 1.42-ha trap grid with 80
Museum Special snap-traps (8 by 10
pattern with 15-m intervals) was lo-
cated on each plot. In 1987, four ad-

ditional grids were located on plots
dominated by lodgepole pine. Snap-
traps were baited with peanut butter
and oatmeal. Beginning in July 1986,
we trapped each grid for three con-
secutive nights and checked traps
daily in early morning. If rainfall
caused the release of snap-trap
mechanisms, trapping effort was ex-
tended by as many nights as it rained
(table 1).

In 1987, we rated the old-growth
condition of our study plots with the
old-growth scorecard developed spe-
cifically for the MBNF (Marquardt

1984) (table 1). The scorecard is com-
pleted subjectively by Forest Service
personnel and is based on structural
characteristics of stands. Structural
characteristics that define old-growth
stands in the MBNF include trees
with large diameters, long-lived
dominant species (i.e., Engelmann
spruce and subalpine fir), a multi-
storied stand structure, dense cano-
pies, multiple species, woody debris
on the forest floor, and standing
snags (Marquardt 1984). The score-
card incorporates sub-scores for each
of these structural characteristics to

Figure 1 .—Map of the study area in the Medicine Bow National Forest of southeastern Wyo-
ming, about 65 km west of Laramie. Twelve study plots <A-L) were establlshied In 1986 and
1987 for Intensive trapping and habitat characterizations.

^— ■ — — ~>i

Table 1.— Information collected from 12 study plots In the l^edicine Bow National Forest of southeastern Wyoming In
1986 and 1987.

Study plot

Characteristic A B C D E F G H I J K L

USFSIocation 201307 201303 201304 205403 205404 201303 201303 208103 201810 201809 204907 205503
site 03 07 05 34 20 08 10 23 15 10 07 26
Dominant overstory spruce spruce spruce spruce spruce spruce spruce spruce lodge- lodge- lodge- lodge-
fir fir fir fir 'fir fir fir pole pole pole pole
Old-growth rating 51 35 48 44 50 37 40 41 25 22 22 19
Trapping effort 1986 240 240 240 320 240 400 240 240 0 0 0 0
(# trap nights) 1987 320 320 320 320 240 240 240 240 240 240 240 240

: : . y

354

achieve an overall old-growth rating,
ranging from 0 to 60.

We quantitatively determined
sub-scores for characteristics we be-
lieved were most important to meet-
ing habitat needs of C. gapperi. Log
density was estimated with the
point-quarter distance method (25
sampling points) and the diameter of
each log sampled (100 logs were
sampled) was measured to deter-
mine the mean log diameter.

Data analyses were performed
with the SPSS computer package
(Nie et aL 1975). Analyses involved
linear and quadratic correlation tests
between small mammal abundance

40.0

^ 35.0
o

2 30.0

B 25.0

w

IT

S 20.0

s

O 5.0
0.0

□ 1986

□ 1987

n / /

/

mrfl

7

^ 14

/

D E F G
Study PM

I J KL

Figure 2.— Capture success of Clethriono-
mys gapperi in 12 study plots In \he Medi-
cine Bow National Forest of southieastem
Wyoming in 1986 and 1987. Temporal vari-
ation in abundance was extreme In five of
ttie eigtit spruce-fir plots sampled botti
years. Plots l-L were dominated by lodge-
pole pine and were sampled only in 1987.

35.0

z

o 30.0
o

r 25.0

ir

« 20.0

§ 15 0

10
e

5 10.0

" 5.0

0.0

1967
P . 0.007
r«0.81

0.0 100 200 30.0 40.0

Stand Rating

50.0

Figure 3.— Capture success of Clethriono-
mys gapperi \n \he Medicine Bow National
Forest of souttieastem Wyoming in 1987 as
a function of old-growth ratings. Ttiis rela-
tionshiip is best explained by a quadratic
correlation. Dahcened data p>oints repre-
sent lodgepole-dominated study plots;
open data points represent spruce-fir-
dominated study plots.

(as inferred from capture success)
and old-growth ratings. While inter-
ested primarily in the responses of C.
gapperi populations, we also evalu-
ated the responses of other small
mammal species that were captured.

Results and Discussion

A total of 695 small mammals were
captured in 5,360 trap nights (TN). In
decreasing abundance, these were C.
gapperi, Tamias minimus, Sorex spp.,
Peromyscus maniculatus, Phenacomys
intermedins, Sorex cinereus, S. montico-
lus, and Microtus longicaudus. Only
captures of C. gapperi and T. minimus
were frequent enough to provide
data for analysis both years; captures
of P. maniculatus were adequate only
in 1987. Other species were rarely
captured.

Temporal Fluctuations in
Abundance

Mean capture success increased
three-fold from 1986 (5.6/lOOTN) to
1987 (18.0/lOOTN). Capture success
of C. gapperi is representative of this
variation (fig. 2). Natural fluctuations
in small mammal abundance are well
documented (Krebs and Myers 1974,
Vaughan 1969). Such fluctuations are
a major source of confounding vari-
ation and hinder the ability of man-
agers to monitor p)opulations for
changes that result from human-in-
duced disturbance. Because of this
temporal variation in abundance, we
separated the data for analysis.

Association of C. gapperi witti
Old-Growth Conditions

In 1986, the abundance of C. gapperi
was weakly correlated linearly with
old-growth ratings (r = 0.62, P =
0.097). However, this result repre-
sented only the range of old-growth
conditions found in spruce-fir stands
(scores ranged from 35 to 51). Four

lodgepole pine study plots, which
rated lowest on the old-growth
scorecard and provided a greater
range of ratings (19-51), were added
in 1987. A more complete pattern
emerged: C. gapperi was most abun-
dant in the lowest-scoring lodgepole
study plot, decreased in the remain-
ing lodgepole plots, further de-
creased to a minimum in the mid-
range spruce-fir plots, and then in-
creased in abundance with increasing
old-growth condition in the remain-
ing spruce-fir plots. A quadratic cor-
relation model best explained the re-
lationship between abundance of C.
gapperi and old-growth ratings in
1987 (r = 0.81, P = 0.007; fig. 3).

The highly significant quadratic
function that described the relation-
ship between abundance of C. gapperi
and old-growth rating in 1987 should
be interpreted separately for the
lodgepole pine and spruce-fir seg-
ments. In spruce-fir plots, the rela-
tionship was positive (r = 0.89, P =
0.003), as it was (suggestively) in
1986. However, a comparison of C
gapperi abundance in spruce-fir plots
between 1986 and 1987 was not sig-
nificant (r = 0.43, P = 0.290). This in-
dicated that the spruce-fir plots sup-
porting high densities of C. gapperi in
1986 were not the same plots sup-
porting high densities in 1987. In
lodgepole plots (1987 only), abun-
dance of C. gapperi was not signifi-
cantly correlated with old-growth
raHng (r = -0.88, P = 0.116). There-
fore, we are not confident in the re-
sults from the lodgepole plots, but an
interpretation is warranted. The
abundance of C. gapperi in both serai
phases (lodgepole and spruce-fir)
was strongly influenced by the abun-
dance of woody debris (particularly
logs) on the forest floor. However,
these two stand types differ mark-
edly in terms of the source, size and
likely persistence of logs.

In spruce-fir plots, logs were large
(mean diameter was 31.0 cm) and
were recruited through the natural
processes of wind throw and snag
decay. Log size and biomass are

355

greater in older forests than in
younger forests (Franklin et al. 1981).
Thus, availability and size of logs in-
crease with time in young spruce-fir
stands, and we believe that this in-
crease was primarily responsible for
the relationship we found between
abundance of C. gapperi and old-
growth rating of spruce-fir plots. In
lodgepole plots, logs were smaller
than in spruce-fir plots (mean diame-
ter was 22.7 cm; t = 7.93, P = 0.004)
and were recruited almost entirely
by thinning. One lodgepole plot (plot
L, in site 205503-26, table 1) had been
thinned 13 months before we
sampled it and had a high density of
logs and the greatest abundance of C.
gapperi. This single plot overwhelm-
ingly influenced the lodgepole phase
of the quadratic function.

Lodgepole stands do not thin well
naturally (Alexander 1974), so log
recruitment rates and densities are
generally low. We predict that, be-
cause they are larger and are re-
cruited at a less variable rate, logs in
spruce-fir stands are more persistent
over time than are logs in lodgepole
stands. Kirkland (1977) and Martell
and Radvanyi (1977) found high den-
sities of C. gapperi in clearcuts one
year after logging spruce forests.
Three years after logging, Martell
and Radvanyi found that C. gapperi
had become rare. Gunther et al.
(1983) attributed the abundance of C.
gapperi in clearcuts to high ground
cover created by felled trees and
slash and to an abundant food sup)-
ply of lichens.

Interpretation of C. gapperi abun-
dance as an indication of old-growth
condition must be undertaken with
caution. C. gapperi appears to re-
spond to natural processes of log ac-
cumulation; however, C. gapperi
populations also appear to respond
to accumulation of woody debris re-
sulting from management actions.
Stand thinning is more common in
lodgepole than in spruce-fir stands in
the MBNF (T. Cartwright, pers.
comm.), so use of C. gapperi as an in-
dicator of old-growth conditions of

spruce-fir stands appears less likely
to be confounded by this factor.

Association of T. minimus and P.
maniculatus with Old-Growtti
Conditions

The broad habitat affinities of these
two species are well documented
(Armstrong 1977). In forested habi-
tats, they are associated with early
successional stages (Martell 1984). In
our study, T. minimus abundance de-
creased with increasing old-growth
rating in 1986 (r = -0.71, P = 0.046;
fig. 4), but the correlation was based
on a narrow range of ratings so that
its reliability is questionable.
Vaughan (1974) noted this species'
dependence on stumps and rocks for
lookout points. Certain structural
features that characterize old-growth
conditions (e.g., restricted average
sight distance) are inconsistent with
the open habitat requirements of T.
minimus. There was no significant
correlation in 1987. Given the high
population levels that year, limited
resources in preferred habitat may
have caused T. minimus to disperse
into less preferred habitat.

Abundance of P. maniculatus de-
creased with increasing old-growth
raring in 1987 (r = -0.60, P = 0.039;
fig. 5), but the correlation was driven
by one data point (study plot L, table
1). The abundance of P. maniculatus
has been shown to increase with
understory vegetation (Tevis 1956). If
this is due to an affinity for cover,
then the conditions present in study
plot L may explain the high numbers
of P. maniculatus found there. If the
data point is excluded from the
analysis, the result supp>orts the
broad habitat distributions P. manicu-
latus is known to exhibit.

Conclusions

We found that abundance of C. gap-
peri was correlated with old-growth
ratings in spruce-fir stands, and at-

tribute that correlation primarily to
the log component of the old-growth
rating. C. gapperi was strongly corre-
lated with old-growth conditions in
spruce-fir and may be predictive of
old-growth condition in that stand
type. However, C. gapperi appears to
respond to logs recruited from man-
agement activities, and caution
should be used in interpreting abun-
dance data.

Our results neither support nor
refute the assumption that C. gapperi
represents the habitat needs of other
species. Alternative monitoring ap-
proaches may have utility in forest
management. These include guild-
indicator sf)ecies, whole-guild, and
community-based monitoring
schemes.

4.0

I

° 3.0

2.0

1.0

0.0

1966
P = 0.046
r = -0.71

0.0 10.0 20.0 30.0 40.0

Stand FUting

50.0

Figure 4.— Capture success of Tamias mini-
mus in the Medicirie Bow National Forest of
southeastern Wyoming in 1 986 as a function
of old -growth ratings in spruce-fir-domi-
nated study plots.

6.0

K 5.0

s

o

3.0

% 2.0

3

O 1.0

0.0

1967
P . 0.039
r.-0.60

0.0 10.0 20.0 30.0 40.0

Stand Rating

50.0

Figure 5.— Capture success of Peromyscus
maniculatus in the Medicine Bow National
Forest of soutt>eastem Wyoming in 1987 os
a function of old-growth ratings. Darkened
data points represent lodgepole-domi-
nated study plots; open data points repre-
sent spruce -fir-dominated study plots.

356

Tamias minimus and P. maniculatus
populations responded in a manner
consistent with their habitat affini-
ties. Thus, C. gapperi may be the only
choice for consideration as a small
mammal ecological indicator of old-
growth conditions in the MBNF.

Acknowledgments

We wish to thank the USDA Forest
Service, Medicine Bow National For-
est, and the Wyoming Game and Fish
Department for funding this project.
We appreciate very constructive re-
views by M. Raphael and W. Block.

Literature Cited

Alexander, Robert R. 1974. Silvicul-
ture of subalpine forests in the
central and southern Rocky Moun-
tains: the status of our knowledge.
USDA Forest Service Research
Paper RM-121, 88 p. Rocky Moun-
tain Forest and Range Experiment
Station, Fort Collins, CO.

Armstrong, David M. 1977. Ecologi-
cal distribution of small mammals
in the Upper Williams Fork Basin,
Grand County, Colorado. The
Southwestern Naturalist 22:289-
304.

Franklin, Jerry F., Kermit Cromack,
Jr., William Denison, Arthur
McKee, Chris Maser, James Sedall,
Fred Swanson, and Glen Juday.
1981. Ecological characteristics of
old-growth Douglas-fir forests.
USDA Forest Service General
Technical Report PNW-118, 48 p.
Pacific Northwest Forest and
Range Experiment Station, Port-
land, OR.

Getz, Lowell L. 1968. Influence of
water balance and microclimate
on the local distribution of the red-
back vole and white-footed
mouse. Ecology 49:276-286.

Graul, Walter D., and Gary C. Miller.
1984. Strengthening ecosystem
management approaches. Wildlife
Society Bulletin 12:282-289.

Gunther, Pamela M., Brenda S. Horn,
and Geoffrey D. Babb. 1983. Small
mammal populations and food
selection in relation to timber har-
vest practices in the western Cas-
cade Mountains. Northwest Sci-
ence 57:32-44.

Harris, Larry D. 1984. The frag-
mented forest: island biogeogra-
phy theory and the preservation of
biotic diversity. 211 p. University
of Chicago Press, Chicago, IL.

Jerry, Danielle G. 1984. Old-growth
management in the Idaho Pan-
handle National Forests, p. 373-
379. In Fish and wildlife relation-
ships in old-growth forests: Pro-
ceedings of the symposium. Wil-
liam R. Meehan, Theodore R. Mer-
rell, Jr., and Thomas A. Hanley,
editors. [Juneau, AK, April 12-15,
1982] 425 p. American Institute of
Fishery Research Biologists.

Juday, Glenn Patrick. 1978. Old-
growth forests: a necessary ele-
ment of multiple use and sus-
tained yield national forest man-
agement. Environ. Law 8:497-522.

Kirkland, Gordon L., Jr. 1977. Re-
sponses of small mammals to the
clearcutting of northern Appala-
chian forests. Journal of Mam-
malogy 58:600-609.

Krebs, Charles J., and Judith H.
Myers. 1974. Population cycles in
small mammals. Advances in Eco-
logical Research 8:267-399.

Marquardt, Steve. 1984. Old-growth
scorecard user's guide. USDA For-
est Service. Unpublished report.

Martell, Arthur M. 1981. Food habits
of southern red-backed voles (Cle-
thrionomys gapperi) in northern On-
tario. Canadian Field-Naturalist
95:325-328.

Martell, Arthur M. 1984. Changes in
small mammal communities after
fire in north-central Ontario. Ca-
nadian Field-Naturalist 98:223-226.

Martell, Arthur M., and Andrew
Radvanyi. 1977. Changes in small
mammal populations after
clearcutting of northern Ontario
black spruce forest. Canadian
Field-Naturalist 91:41-46.

Maser, Chris, Ralph G. Anderson,
Kermit Cromack, Jr., Jerry T. Wil-
liams, and Robert E. Martin. 1979.
Dead and down woody material,
p. 78-95. In Wildlife habitats in
managed forests: the Blue Moun-
tains of Oregon and Washington.
Jack Ward Thomas, editor. Agri-
cultural Handbook 553, 512 p.
USDA Forest Service.

Maser, Chris and James M. Trappe.
1984. The fallen tree — a source of
diversity, p. 335-339. In New for-
ests for a changing world: Pro-
ceedings of the Society of Ameri-
can Foresters National Confer-
ence. [Portland, OR, October 16-
20, 1983].

Maser, Chris, James M. Trappe, and
Ronald A. Nussbaum. 1978a. Fun-
gal-small mammal interrelation-
ships with emphasis on Oregon
coniferous forests. Ecology 59:799-
809.

Maser, Chris, James M. Trappe, and
Douglas C. Ure. 1978b. Implica-
tions of small mammal mycoph-
agy to the management of western
coniferous forests. Transactions of
the North American Wildlife and
Natural Resources Conference
43:78-88.

Merritt, Joseph F. 1981. Qethriono-
mys gapperi. Mammalian Species
146:1-9. The American Society of
Mammalogists.

Nie, Norman H., C. Hadlai Hull, Jean
G. Jenkens, Karin Steinbrenner,
and Dale H. Bent. 1975. SPSS: sta-
tistical package for the social sci-
ences, 2nd edition. 675 p.
McGraw-Hill, Inc.

Patton, David R. 1987. Is the use of
"management indicator species"
feasible? Western Journal of Ap-
plied Forestry 2:33-34.

Romme, William H., and Dennis H.
Knight. 1981. Fire frequency and
subalpine forest succession along a
topographic gradient in Wyoming.
Ecology 62:319-326.

Salwasser, Hal, Ira D. Luman, and
Don Duff. 1982. Integrating wild-
life and fish into public land forest
management. Western Association

357

of Fish and Wildlife Agencies: Pro-
ceedings 62:293-299.

Tevis, Lloyd, Jr. 1956. Responses of
small mammal populations to log-
ging of Douglas-fir. Journal of
Mammalogy 37:189-196.

USD A Forest Service. 1985. Land and
resource management plan for the
Medicine Bow National Forest and
Thunder Basin National Grass-
land. USD A Forest Service, Rocky
Mountain Region. Laramie, Wyo-
ming.

Vaughan, Terry A. 1969. Reproduc-
tion and population densities in a
montane small mammal fauna, p.
51-74. In Contributions in Mam-
malogy. J. Knox Jones, Jr., editor.
Miscellaneous Publications of the
Museum of Natural History 51:1-
428. University of Kansas.

Vaughan, Terry A. 1974. Resource
allocation in some sympatric, sub-
alpine rodents. Journal of Mam-
malogy 55:764-795.

Vemer, Jared. 1983. An integrated
system for monitoring wildlife on
the Sierra National Forest. Trans-
actions of the North American
Wildlife and Natural Resources
Conference 48:355-366.

Habitat Associations of Small
Mammals in a Subalpine
Forest, Souttieastern
Wyoming^

Martin G. Raphael

Abstract.— Mammal capture rates were greatest
at sites with mature timber and other old-growth
attributes. Shrews (both dusky (Sorex monticolus)
and masked (S. cinereus)) and southern red-backed
voles (Clefhrionomys gapperi) were much more
abundant at sites dominated by spruce or fir
compared to drier sites dominated by lodgepole
pine. Deer mice (Peromyscus maniculofus). in
contrast, were most abundant on drier, pine-
dominated sites. The southern red-backed vole,
because of its high abundance and strong
association with mature forest, is a good ecological
indicator of late serai conditions for forest planning
purposes.

Figure 1.— Map of study area showing iocation of study area and distribution of trapping
stations.

Subalpine forests of spruce, fir, and
lodgepole pine cover about 5 million
ha, or 38% of forested land in the
central Rocky Mountain region —
more than any other forest type (Al-
exander 1974, USD A Forest Service
1980). Subalpine forest is harvested
heavily, accounting for over 90% of
total sawtimber volume in this region
(USDA Forest Service 1980). These
forests also are managed to produce
water, and timber harvest practices
have been developed that can sub-
stantially increase water yield (Tro-
endle 1983, Swanson 1987). The Coon
Creek Water Yield Augmentation
Pilot Project (Bevenger and Troendle
1984, 1987) is a large-scale demon-
stration of the feasibility and costs/
benefits of increasing water yield
through specially designed clearcuts.
To evaluate the response of wildlife
species to such harvests, studies
were initiated to describe the pre-
treatment structure and compxDsition
of the vertebrate community (Ra-
phael 1987b) and, ultimately, to com-
pare responses of vertebrates on the
treated watershed and on the unhar-
vested control.

The present study summarizes the
structure of the small mammal com-

' Paper presented at Symposium, Man-
agement of AmphibioDs. Reptiles, and
Small Mammals in Nortti America (Ragstaff,
AZ.July 19-21. 1988).

'Research! Ecdogist, USDA Forest Serv-
ice. Rocky Mountain Forest and Range Ex-
periment Station. Forestry Sciences Labora-
tory. 222 Southi 22nd Street. Laramie. Wyo-
ming 82070.

munity, describes habitat associa-
tions of the dominant species during
the pretreatment phase of the longer
term project, evaluates the efficacy of
an old-growth scorecard to rate old-
growth characteristics of stands, and
assesses designation of mammals as
ecological indicators of old-growth
conditions.

STUDY AREA

Studies were conducted within two
watersheds, the Upper East Fork of

the Encampment River (911 ha) and
Coon Creek (1,615 ha). These adja-
cent watersheds are part of the Sierra
Mad re range of southern Wyoming,
located about 25 km south of the
town of Encampment (fig. 1). Eleva-
tions vary from 2,600 to 3,300 m.
Soils are 50-150 cm deep and are well
drained.

Mean annual precipitation is about
100 cm, 70% falling as snow that usu-
ally covers the site from late Septem-
ber through late June at depths of 2-4
m in winter. Forest cover is domi-
nated by lodgepole pine (~ 60% of

359

land area), and a nuxture of Engel-
mann spruce and subalpine fir. Pole
stands with trees <23 cm d.b.h. occur
on 24% of the two watersheds, ma-
ture stands occur on 72%, and mead-
ows or rock outcrops cover 4%.

METHODS

Vegetation Sampling

In each watershed, 90 sampling sta-
tions were established at 200-m inter-
vals along N-S lines that were 400 m
apart (fig. 1). At each of the 180 sta-
tions, an observer measured basal
area of each tree species using a 1-
factor metric reloskop. Canopy cover
was estimated from the average of
four readings taken at cardinal direc-
tions with a spherical densiometer.
Slope was measured with a clinome-
ter and aspect was measured with a
hand-held compass. All snags >20
cm d.b.h. and 1.8 m tall were
counted within a 0.04-ha circular plot
centered at the station; cover per-
centages of shrubs, forbs, grasses,
rocks, litter, and bare ground were
visually estimated over the same
0.04-ha plot. Hard (class 1,2) and soft
[class 3,4,5 (Maser et al. 1979)] logs
also were counted on each plot.
Height and d.b.h. of one representa-
tive tree were measured at each sta-
tion with a clinometer and metric
d.b.h. tape.

All stands on each watershed
were assessed by personnel of the
Medicine Bow National Forest and
assigned an old-growth rating based
on canopy structure, d.b.h., tree
height, snag size and density, and log
size and density (app)endix). Possible
scorecard values range from 0 (no
old-growth characteristics) to 60
(maximum).

Stand maps were used to associate
a sampling station with the old-
growth scorecard value for the stand
in which the station was located.
Habitat types were also assigned to
each station based on classifications
used by Medicine Bow National For-

est personnel. Also recorded was the
presence or absence of permanent
streams within 100 m of each sam-
pling station.

Red Squirrel Counts

Three observers visited each sample
station twice each year (totaling six
visits/station/yr) from 13 June to 25
July 1985, 18 June to 23 July 1986,
and 15 June to 17 July 1987. At each
visit, the observer recorded all red
squirrels seen or heard within a 100-
m radius of the station center. All
counts were begun within 30 minutes
after sunrise; each observer visited 15

stations per day and most counts
were completed before noon.

SnKili Mammal Trapping

To sample shrews, six pitfall traps
were installed in a 2 x 3 grid (15-m
spacing) centered on each station.
Each pitfall trap was a 3-gal plastic
bucket buried flush with the ground
surface and covered by logs or bark.
To capture other small mammals,
two 50-cm Sherman livetraps were
placed within 2 m of each pitfall sta-
tion.

Mammals were trapped during
late summer from 1985 to 1987 (20

Table 1.— Vegetation and stand attributes on small mammal trapping sta-
tions, estimated or measured on 0.04-ha circular plots, among tidbltat
types' on a Sierra Madre forest, Wyoming.

Characteristic

Lodgepole pine Spruce/fir

Unclossified Pole Mature Mature Signifi-
(n=9) <n=36) (n=76) (n=:59> cance^

Basal area (mVha)

Lodgepole pine

Engelmonn spruce

Subalpine fir
Tree height (m)
D.B.H. (cm)
Snags/0.04 ha
Percent cover

Shrubs and trees >2 m tall

Forbs

Grasses

Rocks >1 5 cm

Utter

Bare soil

Hard logs >20 cm diameter
Soft logs >20 cm diameter
Overstory canopy

Old-growth scorecard index

Stream presence^

Solar radiation index^

Elevation (10^ m)

12.P

19.2^

21.3^

10.68

0.01^

5.2^

6.3^

11.9^

0.01^

10.7

8.6

7.4

10.2

0.18

-19 9AC

18.2^

20.3^

21.2^

0.00

29.8AB

27.0^

32.9^

36.6^

0.00

2.1

1.2

2.0

2.4

0.29^

45.3^c

54.r

50.8^

38.4«^

0.01

14.0

6.4

7.6

15.0

0.12^

14.8

7.4

7.6

16.4

0.18^

0.8^

4.4^^

3.9^

3.2^^

0.00^

82.0^

85.6^<^

82.7^

73.6«

0.07^

0.4

1.0

1.2

1.8

0.30^

0.6^

1.9^B

2.3^

2.7^

0.00^

9.6

10.7

11.7

10.9

0.63^

65.8

69.2

68.2

62.4

0.32^

190ABC

29.4^

34.9^

41.

0.00

22.0

25.0

38.2

37.3

0.45

0.45^

0,50^

0.50^

0.48^

0.02

9.0^

9.4^

9.5»

9.48

0.00^

'Letter superscripts denote results ofmuttiple compartsais (Tukey-Kramer or Dun-
nett's simultaneous procedures); means witt) same letter did not differ. Experiment-
wise error rate maintairted ata = 0,05.

'Significance of analysis of variance P-tests among habitats: W indicates fhiat
Welch's test was performed when variances were unequal.

'Percent of stations within 100 m of stream.

"Index of yearly solar radiation input (Frank and Lee 1966).

360

August through 26 September 1985, 5
August to 11 September 1986, and 4
August to 10 September 1987). Ob-
servers checked traps once daily dur-
ing each of three, 10-day sampling
sessions each year. Sampling sessions
were separated by four days, encom-
passing six weeks each year. All cap-
tured specimens were identified, toe
clipped, sex determined, aged,
weighed, and checked for reproduc-
tive status (currently breeding or
not).

Dead animals were assigned a
permanent catalog number. Shrews
were preserved in 70% ethanol and
all other species were frozen for later
identification.

Data Analysis

Total rumbers of detections (red
squirrels) or first captures (all other
species) were calculated at each sta-
tion over the 3 years. Thus, the total
numbers of captures represented the
results of 450 trapnights of effort at
each of the 180 stations (81,000 total
trapnights). Despite efforts to close

pitfall traps between sessions, some
mammals were captured before the
start of each 10-day session. These
specimens were retained, but num-
bers were not included in analyses.

To assess habitat associations of
the more abundant mammals, I per-
formed a principal components
analyses (with varimax rotation) us-
ing the SPSS/PC+ program package
(Norusis 1988). Principal components
analysis derives linear combinations
of attributes (in this case vegetation
characteristics as listed in table 1).
All components with eigenvalues
>1.0 were retained for subsequent
analyses. The equations were then
''solved" for each station, resulting in
a set of scores that were interpreted
as habitat gradients. I identified these
gradients from those original habitat
variables most highly correlated with
the principal components scores. To
relate abundance of the more abun-
dant mammals to habitat features at
each station, I performed multiple re-
gressions of capture rates at each sta-
tion (dependent variable) with the
habitat gradients or principal compo-
nents scores (independent variables).

r

Table 2.— Habitat gradients derived from principal components analysis of
19 variables (table 1) describing vegetation structure and composition at
each small mammal sampling station, Sierra Madre, Wyoming.

Percent of Cumulative
Gradient vcffiance^ percent

Interpretation of tiabltat gradienF

1

26.2

26.2

2

16.1

42.3

3

8.7

51.0

4

7.4

58.4

5

6.6

65.1

6

5.6

70.7

Greater cover of shrubs and litter; greater
basal area of lodgepole pine; lower cover
of herbs, grasses,

Greater expression of old-growth attrib-
utes; greater basal area of Engelmann
spruce.

Upland sites with greater cover of soft logs;
greater basal area of subalpine fir.
Lower cover of bare ground; greater can-
opy cover.

Greater cover of rocks.
Higher elevation sites with greater solar ra-
diation (southerly slopes).

'Amount of total variance (among all anginal variabies) accounted for by eact)
principal component.

'Interpretation based on magnitude of correlations of original variables with de-
rived components. Descriptions indicate positive extreme of each gradient.

To summarize patterns of co-oc-
currence of the more common mam-
mal species, I performed an average-
linkage-between-groups cluster
analysis [UPGMA (Norusis 1988)]
based upon Pearson correlations be-
tween abundances of all pairs of spe-
cies among the 180 stations. Results
of the cluster analysis were displayed
using a dendrogram showing the
relative similarities of all species. The
similarity measure, for this display,
was rescaled to values ranging from
0 (no similarity) to 25 (maximum
similarity).

RESULTS AND DISCUSSION

Vegetation

Structure and composition of vegeta-
tion (table 1) were typical of sub-
alpine forest in the central Rocky
Mountains (Alexander 1974; Raphael
1987a, 1987b). VegetaUon characteris-
tics have been shown to be similar
between the two watersheds (Ra-
phael 1987b); therefore, no distinc-
tion was made between the two wa-
tersheds for this study.

Principal components analysis re-
sulted in the creation of six synthetic
habitat gradients that, together, con-
tained 68% of the total variance from
the 19 original habitat variables (table
2). I used the variables that were
most highly correlated with values of
each gradient to interpret the biologi-
cal meanings of the gradients (table
2).

MannnKils

Over the 3 years of study and over
all sampling stations, observers cap-
tured 4,553 individuals of 17 small
mammal species and recorded 987
detections of red squirrels (table 3).
The most abundant species was the
southern red-backed vole, account-
ing for over 50% of all captures.
Other dominant species included
masked shrew (15%), deer mouse

361

(15%), red squirrel, dusky shrew
(6%), and chipmunks (2 species, 6%).

Specific Habitat Associations
Masked Shrew

Masked shrews were more abundant
than other shrews and were captured
more frequently in mature lodgepole
and spruce/fir sites (table 3) with
higher cover of herbs and grasses;
they were less abundant on dry,
south-facing sites (table 4). Their
abundance at each station was mod-
eled (R^ = 0.42) by a regression that
included gradients 2,1,6, and 4 (in
order of their statistical significance)
(table 4). Other studies (Negus and
Find ley 1959, Spencer and Pettus
1966, Brown 1967a, Armstrong 1977)
also report this species' preference
for moist sites. However, I did not
find a strong association with bogs,
as reported by Brown (1967a) and
Spencer and Pettus (1966).

Dusky Shrew

Dusky shrews were captured in
greater numbers in more moist, ma-
ture spruce/fir sites (table 3). They
were most strongly associated with
dense herbaceous cover and (to a
lesser degree) with old-growth attrib-
utes. Unlike the masked shrew, their
abundance was positively and sig-
nificantly correlated with gradient 3
(moist, streamside sites; tables 2,4).
Like masked shrews, they were less
abundant on southerly, steeper sites.
The regression model explained 41%
of variance in abundance (table 4).
Brown (1967a) captured this shrew in
a greater variety of habitats and in
drier sites than the masked shrew.
Negus and Findley (1959) also re-
jx)rted use of a greater variety of
habitats; Spencer and Pettus (1966)
found dusky shrews in association
with marshy habitats.

The association of this shrew with
old-growth conditions has not, to my

knowledge, been previously re-
ported.

Least Chipmunk

The abundance of least chipmunk
was significantly and negatively cor-
related with gradient 4 (bareground)
and positively correlated with gradi-
ent 6 (southerly exposure). Although

the regression was statistically sig-
nificant, it explained only 5% of vari-
ance in abundance (table 4); thus, the
regression model was not statistically
meaningful.

Nonetheless, the associations sug-
gested by the model, particularly the
preference for open, drier slopes, are
in accordance with results of other
studies (e.g., Telleen 1978, Clark and
Stromberg 1987).

r

Table S.—Smali mammal capture rates amor>g generalized habitat types^
in the Sierra Madre, Wyoming, 1 985-1 987.

Species

Toh3l no. Lodgepole pine Spruce/fir
individuals Pole Mature Mahjre
captured (r>=36) (f»=76) (n=59)

Masked shrew 700
(Sorex cinereus)

Dusky shrew 253
(S. monficolus)

Dwarf ^rew 2
(S. nanus)

Water shrew 7
($. palustris)

Pygmy shrew 11
(S.hoyO

Least chipmunk 101
(Tamias minimus)

Uinta chipmunk 150
(Tamias umbrinus)

Golden-mantled ground sq. 11

(Spermophiius lateralis)

Red squirrel 3987

(Tamiasciurus hudsonicus)

Northern pocket gopher 1

(Thomomys falpoides)

Deer mouse 696

(Peromyscus maniculafus)

Southern red-backed vole 2^75

(Clefhrionomys gapperi)

Heather vole 17

(Phenacomys intermedius)

Montane vole 32

(Microfus montonus)

Long-tailed vole 1 1

(M. longicandus)

House mouse : 1

(Mus musculus)

Western jumping mouse 80
(Zapus princeps)

Ermine 6
(Musfela erminea)

2.6^
0.6^
0
0

0.03
0.8*
1.1*
0,06
5.8*
0

3.3*
10.8*
0.08
0.03*
0,03
0

0.1*
0

3.7*8

0.8*

0

0.08
0.09
0.4*
0.9*
0.11
4.7*
0.01
4.2*
11.4^
0.09
0.09*
0.08
0

0,4»
0.06

5.08
2.78
0,03
0.02
0,03
0,7*
0,6*
0.02
6.0*
0

3.5*
18.08
0,10
0.37*
0.07
0.02
0.78
0.02

0.01
0,01
NT
NT
NT
0,50
0,51
NT
0.23
NT
0,31
0,02
NT
0,08
NT
NT
0,00
NT

'Values are mean capture rates (captures/450 trapnigtits) or mean numbers of
detections (red squirreO among tiabitat types for all years combined. Letter super-
scripts indicate results of multiple comparisons; means witti same letter did not differ
significantly.

^Significance from one-way analy^s of varlartce: NT = not tested because of small
sample size.

'Results are expressed as numbers of detections during call counts.

362

Uinta Chipmunk

Uinta chipmunks were most abun-
dant on rocky slopes (gradient 5), as
also reported by Clark and
Stromberg (1987). They were rela-
tively more abundant in younger
stands (gradient 2). The regression
model explained 17% of the variation
in abundance of this species (table 4).
Compared with the least chipmunk,
this species is reported to be more
restricted to subalpine forest habitats
(Negus and Findley 1959). Telleen
(1978) found an association with
closed canopy, open understory
habitats.

Red Squirrel

Red squirrel abundance was some-
what greater on dry, gently sloping
sites (gradients 3, 5), but only 16% of
variation in abundance was ex-
plained by the regression model.
These squirrels were abundant
throughout the study area, which
seemed to be comprised of excellent
red squirrel habitat. Therefore, vari-
ation in vegetation among sites was
probably minor in relation to the po-

tential variation that would distin-
guish suitable from unsuitable habi-
tat. Clark and Stromberg (1987) de-
scribe red squirrels as widespread
throughout coniferous forest habitats
of Wyoming.

i:>eer Mouse

Deer mice were associated with
streamside sites having lower basal
area of subalpine fir (gradient 3). Al-
though widespread on the study
area, they tended to be more abun-
dant on open, lodgepole-dominated
sites and meadows than on spruce/
fir sites. The regression model ex-
plained 15% of the variance in deer
mouse abundance (table 4). Contrary,
to these results, other studies (Brown
1967b, Campbell and Clark 1980,
Ramirez and Homocker 1981) re-
ported associations of deer mice with
xeric sites away from streams. The
species is known to be abundant on
cutover sites (Ramirez and Hor-
nocker 1981, Scrivner and Smith
1984), tolerant of a wide range of eco-
logical conditions (Clark and
Stromberg 1987), and omnivorous
(Clark 1975).

Table 4.— Results of stepwise multiple regressloru of small mammal abun-
dance with habitat gradients (principcd components from table 2) Sierra
Mddre, Wyoming.

Species

Habitat gradient'
3 4 5

Explained
variance^

Masked shrew

(2)

1

(3)

0.42

Dusky shrew

(1)

2

(4)

(3)

0.41

Least chipmunk

(1)

2

0.05

Uinta chipmunk

(2)

1

0.17

Red squirrel

(4)

1

(3)

(2)

0.16

Deer mouse

(1)

0,15

S. red-backed vole

(2)

1

4

5

(6)

(3)

0.46

W. jumping mouse

(1)

2

(3)

0.20

'Numbers below each gradient indicate /he order of entry of that gradient into the
stepwise regression (using F-to-enter significance ofP< 0.05). Parentheses indicate
negative associations.

^Adjusted values indicating the proportion of variance in capture (or detection)
rates accounted for by gradients included in regression model. All regressions were
significant atP < 0.001.

Southern Red- Backed Vole

This vole, the most abundant species
on the study area, was most abun-
dant in mature spruce/ fir stands
(table 3). Its abundance was also
greater in stands that had more herb
and grass cover (gradient 1), on
northerly slopes (gradient 6), and on
sites with greater basal area of sub-
alpine fir and greater log cover (gra-
dient 3). Its abundance was modeled
well by the regression, which ac-
counted for 46% of variation in red-
backed vole abundance among sites
(table 4).

The association of red-backed
voles and mature spruce/ fir forest is
well documented (Ramirez and Hor-
nocker 1981, Allen 1983, Scrivner and
Smith 1984). This association may be
due, at least in part, to the greater
cover of logs and other woody debris
that provides protection during criti-
cal periods of freezing and thawing
(Merritt 1976, 1985; Merritt and Mer-
ritt 1978, Sleeper 1979) and supports
fungi used as food (Williams 1955,
Clark and Stromberg 1987, Wywia-
lowski and Smith 1988).

Western Jumping Mouse

Jumping mice were most abundant
in spruce/ fir and mature lodgepole
habitats (table 3). As reported in
other studies (Negus and Findley
1959, Brown 1967b, Clark 1971,
Scrivner and Smith 1984), these mice
were associated with dense herba-
ceous or grassy vegetation (gradient
1) along moist streamsides (gradient
3) in more mature stands (gradient 2)
(table 4). The regression model ac-
counted for 20% of variation in abun-
dance across all stations. These mice
feed primarily on grass seeds and
fungi (Jones et al. 1978, Vaughan and
Weil 1980), which may account for
their close association with grassy
streamside habitats.

363

General Relationships

Moisture and stand maturity were
two habitat features that separated
patterns of abundance of the various
species. This is illustrated most effec-
tively through the cluster analysis
based on interspecific correlations of
relative abundance (fig. 2). The den-
drogram shows two groups: one
comprised of the two shrews, two
voles, and jumping mouse; and one
comprised of the red squirrel, two
chipmunks, and deer mouse. The for-
mer group is associated with more
moist, old-growth conditions (table
4). The latter group is associated with
drier, less mature conditions.

The association of species with
old-growth conditions is of special
interest because of concern over
identifying species that are ecological
indicators of old-growth (USDA For-
est Service 1985; Nordyke and
Buskirk, these proceedings). The
Medicine Bow National Forest, the
site of this study, lists the southern
red-backed vole as an ecological indi-
cator representing late successional
stages in conifer forests. Because the
forest uses the old-growth scorecard
to rate old-growth conditions,
whether or not red-backed vole
abundance is related to old-growth
index values is of interest. Raphael
(1987b) confirmed such a trend based
on analyses of the first 2 years of the
present study.

The trend is even more pro-
nounced when all 3 years are in-
cluded in the analyses (fig. 3). South-
ern red-backed voles are increasingly
abundant as old-growth scorecard
index values increase. Similar trends
are evident for masked and dusky
shrews (fig. 3).

CONCLUSIONS

The small mammal community, as
sampled in this study, was similar in
composition to that described in
other studies in subalpine forests of
the Rocky Mountain region (cf. Ra-

phael 1987a). The southern red-
backed vole was the most abundant
species and can be considered the
species most representative of ma-
ture spruce/ fir forest stands. Stand
age and moisture conditions were
the two most important generalized
gradients that were predictors of
summer abundance of the various
species. The southern red -backed
vole was confirmed as a suitable eco-
logical indicator of old-growth forest;
but, two other species, the masked
shrew and the dusky shrew, are
good candidates as well.

ACKNOWLEDGMENTS

I am indebted to Christopher Cana-
day, Anita Kang, Jeffery Waters,
Gary Rosenberg, Scott Stoleson, San-
dra Spon, Sandra Pletschet, Steven
Larson, Lindsay Hall, Thomas
Batchelor, Daniel Maltese, and Lisa
Smith for their help in the field. I also

thank personnel of the Wyoming
Game and Fish Department, the
Medicine Bow National Forest, espe-
cially the Hayden Ranger District, for
their cooperation; Carron Meaney of
the Denver Museum of Natural His-
tory and Steven Buskirk of the Uni-
versity of Wyoming for technical as-
sistance and use of museum facilities;
and Gregory D. Hayward, Graham
W. Smith, Mark R. Stromberg, and
Richard H. Yahner for comments on
the manuscript.

LITERATURE CITED

Alexander, Robert R. 1974. Silvicul-
ture of subalpine forests in the
central and southern Rocky Moun-
tains: the status of our knowledge.
Research Paper RM-12. Fort
Collins, CO: U.S. Department of
Agriculture, Forest Service, Rocky
Mountain Forest and Range Ex-
periment Station; 88 p.

r

SOMO

MIMO

SOCI

CLGA

ZAPR

TAHU

TAMI

PEMA

TAUM

0

5 10 15

RELATIVE SIMILARITY

20

25

Figure 2.— Dendrogram showing relative similarity (Pearson correlations) of abundances of
snrxill nrKimmal species across sampling stations. Species are: Sorex monticolus (SOMO),
Microtus montanus (MIMO), Sorex cinereus (SOCI), Clethrionomys gapperi (CLGA), Zapus
princeps (ZAPR), Tamiasciurus hudsonicus (TAHU), Tamias minimus (TAMI). Peromyscus
maniculatus (PEMA), and Tamias umbrinus (TAUM).

364

Allen, Arthur W. 1983. Habitat suita-
bility index models: Southern red-
backed vole (Western United
States). U.S. Dept. Int., Fish Wildl.
Serv. FWS/OBS-82/10.42. 14 p.

30

20

10

CLGA

10

C3

-a;

TAMi

1

0

PEMA

1

1

lOf

0

SOCI

Anonymous. 1985. Medicine Bow
National Forest old growth habitat
scorecard. Southwest Habitater
6:2-7.

Armstrong, David M. 1977. Ecologi-

SOMO

10

TAUM

2.0

1.5

1.0

0.5

0

ZAPR

1

0 17 27 37 45 60 0 17 27 37 45

OLD-GROWTH SCORECARD VALUE

60

10

0.5

0.4

0.3

0.2

0.1

0

TAHU

.It T

MIMO

0 17 27 37 45 60

if

Figure 3.— Mean abundance of selected snrxall mammal species In relation to old-growtti
scorecard values. Larger scorecard values Indicate greater expression of old-growtfi condi-
tions. Vertical lines wittiin bars indicate 95% CI of means. See figure 3 for species codes.

365

cal distribution of small mammals
in the Upp)er Williams Fork Basin,
Grand County, Colorado. The
Southwestern Naturalist 22:289-
304.

Bevenger, Greg S.; Troendle, Charles
A. 1984. The Coon Creek water
yield augmentation project, p. 240-
251. In Water for the twenty-first
century: will it be there? proceed-
ings of the symposium, 1984 April
3-5, Dallas, TX. Dallas, TX, South-
em Methodist University.

Bevenger, Greg S.; Troendle, Charles
A. 1987. The Coon Creek water
yield augmentation pilot project,
p. 145-149. In Water for the
twenty-first century: will it be
there? proceedings of the sympo-
sium, 1984 April 3-5, Dallas, TX.
Dallas, TX, Southern Methodist
University.

Brown, Larry N. 1967a. Ecological
distribution of six species of
shrews and comparison of sam-
pling methods in the central Rocky
Mountains. Journal of Mam-
malogy 48:617-623.

Brown, Larry N. 1967b. Ecological
distribution of mice in the Medi-
cine Bow Mountains of Wyoming.
Ecology 48:677-680.

Campbell, Thomas M. Ill; Clark, Tim
W. 1980. Short-term effects of log-
ging on red-backed voles and deer
mice. Great Basin Naturalist
40:183-189.

Clark, Tim W. 1971. Ecology of the
western jumping mouse in Grand
Teton National Park, Wyoming.
Northwest Science 45:229-238.

Clark, Tim W. 1975. Ecological notes
on deer mice in Grand Teton Na-
tional Park, Wyoming. Northwest
Science 49:14-16.

Clark, Tim W.; Stromberg, Mark R.
1987. Mammals in Wyoming. Uni-
versity of Kansas, Museum of
Natural History. Public Education
Series No. 10. 314 p.

Frank, Ernest C; Richard Lee. 1966.
Potential solar beam irradiation on
slopes. Research Paper RM-18.
Fort Collins, CO: U.S. Department
of Agriculture, Forest Service,

Rocky Mountain Forest and Range
Experiment Station. 166 p.
Jones, Gwilym S.; John O. Whitaker,
Jr.; Chris Maser. 1978. Food habits
of jumping mice (Zapus trinotatus
and Z. princeps) in western North
America. Northwest Science 52:57-
60.

Maser, Chris; Ralph G. Anderson;
Kermit Cromack, Jr.; Jerry T. Wil-
liams; Robert E. Martin. 1979.
Dead and down wood material, p.
78-95. In Thomas, J. W., ed. U.S.
Department of Agriculture, Forest
Service Agriculture Handbook
553. Washington, DC, U.S. Gov-
ernment Printing Office.

Merritt, Joseph F. 1976. Population
ecology and energy relationships
of small mammals of a Colorado
subalpine forest. Ph.D. disserta-
tion, Univ. Colorado, Boulder. 146
p. From Diss. Abstr. Int. B Sci.
Eng. 37:3756B.

Merritt, Joseph F. 1985. Influence of
snowcover on survival of Clethri-
onomys gctpperi inhabiting the Ap-
palachian and Rocky Mountains of
North America. Acta Zoological
Fennica 173:73-74.

Merritt, Joseph F.; Joanie M. Merritt.
1978. Population ecology and en-
ergy relationships of Clethrionotnys
gapperi in a Colorado subalpine
forest. Journal of Mammalogy
59:576-598.

Negus, Norman C; James S. Findley.
1959. Mammals of Jackson Hole,
Wyoming. Journal of Mammalogy
40:371-381.

Norusis, Marija J. 1988. SPSS/PC+:
Advanced statistics. SPSS, Inc.,
Chicago, IL.

Ramirez, Pedro Jr.; Maurice Hor-
nocker. 1981. Small mammal
populations in different-aged
clearcuts in northwestern Mon-
tana. Journal of Mammalogy
62:400-403.

Raphael, Martin G. 1987a. Nongame
wildlife research in subalpine for-
ests of the central Rocky Moun-
tains, p. 113-122. In Troendle, C.
A., M. R. Kaufmann, R. H. Hamre,
and R. P. Winokur, editors. Pro-

ceedings, Management of sub-
alpine forests: building on 50 years
of research. USDA Forest Service,
Rocky Mountain Forest and Range
Experiment Station, General Tech-
nical Report RM-149.

Raphael, Martin G. 1987b. The Coon
Creek wildlife project: effects of
water yield augmentation on wild-
life, p. 173-179. In Troendle, C. A.,
M. R, Kaufmann, R. H. Hamre,
and R. P. Winokur, editors. Pro-
ceedings, Management of sub-
alpine forests: building on 50 years
of research. USDA Forest Service,
Rocky Mountain Forest and Range
Experiment Station General Tech-
nical Report RM-149.

Scrivner, Jerry H.; H. Duane Smith.
1984. Relative abundance of small
manrmnals in four successional
stages of spruce-fir forest in Idaho.
Northwest Science 58:171-176.

Sleeper, Roger A. 1979. Small mam-
mal population fluctuations in re-
lation to snowpack. Ph.D. Disser-
tation, Colorado State Univ., Fort
Collins. 134 p. From Diss. Abstr.
Int. B Sci. Eng. 40:3047.

Spencer, Albert W. and David Pettus.
1966. Habitat preferences of five
sympatric species of long-tailed
shrews. Ecology 47:677-683.

Swanson, Robert H. 1987. Applying
hydrologic principles to the man-
agement of subalpine forests for
water supply, p. 79-85. In Tro-
endle, C. A.; M. R. Kaufmann; R.
H. Hamre; R. P. Winokur, editors.
Proceedings, Management of sub-
alpine forests: building on 50 years
of research. USDA Forest Service,
Rocky Mountain Forest and Range
Experiment Station General Tech-
nical Report RM-149.

Telleen, Steven L. 1978. Structural
niches of Eutamias minimus and E.
umbrinus in Rocky Mountain Na-
tional Park. Ph.D. Dissertation,
Univ. Colorado, Boulder. 152 p.
From Diss. Abstr. Int. B Sci. Eng.
39:2115.

Troendle, Charles A. 1983. The po-
tential for water yield augmenta-
tion from forest management in

the Rocky Mountains. Water Re-
sources Bulletin 19:359-373.

USDA Forest Service. 1980. An as-
sessment of the forest and range
situation in the United States. FS-
345, Washington, DC, U.S. Depart-
ment of Agriculture, Forest Serv-
ice, 631 p.

USDA Forest Service. 1985. Final En-
vironmental Impact Statement,
Medicine Bow National Forest,
Laramie, WY. U.S. Department of
Agriculture, Forest Service.

Vaughan, Terry A.; Wendy Pearse
Weil. 1980. The importance of ar-
thropods in the diet of Zapus prin-
ceps in a subalpine habitat. Journal
of Mammalogy 61:122-124.

Williams, Olwen. 1955. Distribution
of mice and shrews in a Colorado
montane forest. Journal of Mam-
malogy 36:221-231.

Wy wialowski, Alice P.; Graham W.
Smith. 1988. Selection of mi-
crohabitats by the red-backed
vole, Clethrionomys gctpperi. Great
Basin Naturalist 48:216-223.

366

Appendix.— Old growth habitat scorecard (Anonymous 1 985) used to rate stands in the Roclcy Mountain Region. Point val
ues from 1 to 5 are assigned to each category A-L Values are summed over all rows and the grand total is used as the in
dex value.

Point value

5 4 3 2 1

A. Overstory

3 or more species 3 or more species 2 species 2 species 1 species

Spruce and/or Fir >50% Spruce and/ or Fir <50% Spruce and/or Fir >50% Spruce and/or Fir <50% 100%

B. Midstory

3 or more species

Spruce ond/or Fir >50%

3 or more species
Spruce ond/or Fir <50%

2 species

Spruce and/or Fir >50%

C. Understory
3 or more species
Spruce ond/or Fir >50%

3 or more species
Spruce ond/or Fir <50%

2 species

Spruce ond/or Fir >50%

D. Total Canopy Cover
70%+

70-50%

E. Overstory. Canopy Cover

50-30% 70-50% or 30-10%

F. Midstory Canopy Cover

40-20% 70-40% or 20-10%

G. Overstory Ave. DBH (Live)

16'+

H. Midstory Ave. DBH (Live)

9'+

15--13-

8"-6"

50-30%
100-70% or 10-1%
100-70% or 10-1%

12--10"
5"-3-

I. Standing Snogs Ave. DBH (Record only those snags above 6" in height.)

16"+ 15"-13" 12"-10"

J. Standing Snogs #/Acre (Record only those snags above 6' in height and 7" DBH.)
6+ 6-4 3-1

K. Dead. Down Logs Ave. DBH
16"+

15"- 13"

L. Dead. Down Logs #/Acre (Record only those above 7" DBH.)
12+ 12-6

Column

Totals

12--10-
6-2

2 species

Spruce and/or Fir <5CD%

2 species

Spruce and/or Fir <50%

1 species
100%

1 species
100%

30-10%

9'-7"

<10%

9'-7" <7"

<3-

9"-7"

367

Differences in tlie Ability of
Vegetation IVIodels to Predict
Small Mammal Abundance
in Different Aged Douglas- Fir
Forests^

Cathy A. Taylor,^ C.John Ralph,^ and Ariene
T. Doyle^

Abstract.— Three trapping techniques for small
mammals were used in 47 study stands in northern
California and southern Oregon and resulted in
different capture frequencies by the different
techniques. In addition, the abundances of
mammals derived from the different techniques
produced vegetation association models which
were often quite different. Only the California red-
backed vole (Clefhrionomys californicus) showed
any association with stand age, and no species had
any marked associations with the moisture regime of
the stands or the geographical region.

Habitat association patterns have
been presented for many small mam-
mal species (e.g. Rosenzweig 1973,
M'Closkey 1975, Dueser and Shugart
1978, MacGracken, et al. 1985). In
most instances, models representing
habitat use have been derived for a
single species using a single trapping
technique. Most community based
studies have also used a single trap-
ping technique. Individual species,
however, have different sensitivities
to capture, making it difficult to com-
pare capture rates across species (Se-
ber 1981).

To better understand the habitat
associations across a sequence of for-
est ages in the Pacific Northwest, we
studied the population status in se-
lected forest stands in northern Cali-
fornia and southern Oregon during
summer and fall of 1984 and 1985.
This study was part of a U.S. Forest
Service research project extending
from northern California through
Oregon and north into Washington
(e.g. Ruggiero and Carey 1984,
Manuwal and Huff 1987). The im-
pacts of the harvesting of old-growth
forests on vertebrate populations in

'Paper presented at symposium. Man-
agement of Amphibians. Reptiles, and
Small Mammals in Nortt) America (Ragstaff.
AZ.July 19-21. 1988).

'USDA Forest Service, Pacific Southwest
Forest and Range Experiment Station, 1700
Bayy^ew Drive, Areata, California 95521.

^USDA. Tongass-Chatham Area National
Forest, Juneau Ranger District, 8465 Old
Dairy Road. Juneau, Alaska 99801.

this area are uncertain (Hagar 1960,
Raphael and Barrett 1984, Raphael et
al. in press).

We trapped mammals over a gra-
dient of different-aged forest stands
using three techniques. Our primary
objectives were: (1) to determine if
the relative abundance of each spe-
cies differed between the stands; (2)
to determine which habitat variables
were associated with the relative
abundances of each species; and (3)
to study the efficiency of different
trapping techniques. In this paper we
discuss differences in habitat models
derived from different techniques for
the five most abundant species of
small mammals.

Methods and Materials

Study Area

We selected 47 study stands in three
regions of northwestern California
and southwestern Oregon. These
stands represented a successional
gradient typical of the Douglas-fir
communities of the region. Stands
ranged in elevation from 414 m to
1,556 m and were generally domi-
nated by Douglas-fir in association
with tanoak (Lithocarpus densiflora)
and madrone (Arbutus menziesii).
Three low elevation stands had a
redwood (Sequoia sempervirens) com-
ponent; four high elevation stands in
the Cave Junction region were domi-

nated by white fir (Abies concolor).

Fifteen stands were located at each
of three regions (in the vicinities of
Branscomb and Willow Creek, Cali-
fornia, and Cave Junction, Oregon),
with an additional 2 stands at Butte
Creek, near Dinsmore, California.
These stands were divided into three
age classes based on core samples of
2 to 10 of the dominant Douglas-firs
in each young and mature stand (up
to approximately 180 years of age)
(B. Bingham, USFS Pacific Southwest
Station, pers. commun.). In old-
growth stands, tree cores could not
always be taken because of large tree
size and rotten tree cores, thus some
stand ages were based on rings
counted on stumps in adjacent
clearcuts, along roads, or on core
samples provided by local Forest
Service offices. Each forest stand was
assigned to one of three age classes:
young forest < 100 years; mature for-
est 100-180 years; and old-growth
forest, > 180 years. Those that were
classified as old-growth forest were
further classified as to moisture re-
gime: dry, mesic, or wet, based on
species composition and percent
cover of the herb and shrub layers of
the stand (B. Bingham, pers. com-
mun.). All young and mature stands
represented the modal, or mesic
moisture class.

An index to the yearly solar radia-
tion was derived by the method of
Frank and Lee (1966), which is based
on slope, aspect, and latitude. Values
are largest on south-facing, moderate

368

slopes, and lowest on north-facing,
steep slopes.

Mammal Trapping

A single trapping grid for snap/liv-
etrapping was established in each
stand in 12 rows, with 12 trap sta-
tions per row. Trap stations were
placed at 15-meter intervals resulting
in a grid 165 m x 165 m. Each stand
was relatively homogeneous, and
grids located in each stand were, in
most cases, separated from different
habitat types by at least 100-m of the
same habitat.

In 1984, two snaptraps ("Museum
Special") were placed in 1984 at each
trap station within 1.5 meters of the
grid coordinate on all 47 stands. Six
stands were trapped simultaneously
(two in each region) for five days
(four nights) until all stands were
sampled (July 3 to August 31). In
1985, a single Sherman livetrap (7.6 x
8.9 x 22.0 cm) was used at the same
stations on 43 of the stands; six
stands again were trapped during
each five-day session from July 9 to
August 30. We did not livetrap four
stands (two in Branscomb area and
two in Cave Junction area). In both
years, each trap was placed along-

side downed logs, brushy vegetation,
or rodent runways. Baited with pea-
nut butter and oat groats in 1984, and
oat groats and sunflower seeds (3:1
ratio) in 1985, the traps were left in
place for four nights. We feel that the
four nights of trapping did not sig-
nificantly alter populations between
years. In the analyses below, we stan-
dardized captures to the number per
1000 trap-nights.

We used pitfall traps to sample
small mammal populations on all 47
stands during both 1984 and 1985. A
pitfall grid consisted of six rows of
six pitfall traps per row at 15 m spac-
ing in each stand. Grids were located
usually more than 100 m from snap
and pitfall grids. Traps were two No.
10 cans taped together and sunk until
the top was flush with the ground. A
funnel collar to prevent animals from
escaping was made from a margarine
container with the bottom cut out.
We propped a cedar shake 3-4 cm
above the opening to the pitfall trap
to act as a cover.

Traps were examined at 5-day in-
tervals for 50 days in October and
November 1984, and for 30 days in
October 1985. In the analyses that
follows, we used the number of
mammals captured unstandardized
for effort.

r

Table 1.— Vegetation variables measured for each cluster of trapping sta-
tions in a study of small mamma! abundance in Douglas-fir forests of
southem Oregon and northem California, 1984-85.

Ground cover

Vegetation variables

Rock
Soil

Small Litter

Moss

Lichen

First litter layer
Second litter layer
Solar index

Decoy class°1 and 2 logs
Decay class 3. 4. 5 logs

Herb

Grass

Fern

Douglas-fir
Tanoak

Pacific madrone
Live oaks
Oregon grape
Redwood
Poison oak
True fir
Alders

Dogwood
California hazel
Pines

White and black oaks
Said

Manzanita
Rosa spp.
/?ut»L/s spp.
California laurel
Huckleberry
Big-leaf maples
False cedars

°Jhomas 0979:80).

The complication that not all cap-
ture methods were used in both
years of the study, resulted in an un-
known year-effect that may influence
capture frequency. Despite this prob-
lem, we feel that the data are instruc-
tive as to the variety of models pro-
duced, and the implications for in-
vestigators.

Vegetation Sampling

Vegetation for each snap /livetrap
grid was measured on 16 plots over-
laying the 144 trap stations. Nine
vegetation plots were uniformly dis-
tributed among the 36 pitfall stations.
Vegetation and structure were meas-
ured in 5.6 m and 15 m radii circular
plots. On each plot, we recorded:
percent cover of ground cover vari-
ables (i.e. rocks, woody debris); per-
cent cover of vegetation at five height
strata; and counts of trees and snags
in varying size classes.

We averaged the percent cover
values for 25 vegetation stations (16
in the snap/ livetrap grids plus nine
stations in the pitfall grids), to obtain
mean values of percent cover for 11
ground cover variables and 24 spe-
cies of plants (or groups of species)
in each of the 47 stands (table 1). We
combined some taxa into genera
prior to calculating means: the true
firs {Ahies spp.), alders {Alnus spp.),
huckleberries {Vaccinium spp.), live
oaks iQuercus spp.), manzanita (Arte-
mesia spp.), various roses {Rosa spp.),
and Rubus spp. The vegetation data
were vertically stratified into five
levels: ground (0-0.5 m), shrub level
(>0.5-2.0 m), mid-canopy (>2.0 m-
midlevel), canopy (those trees at the
average height of the stand), and
supercanopy (those trees substan-
tially above the canopy). Mean val-
ues for cover by stand were com-
bined into two strata: ''understory"
included ground and shrub layers,
while ''canop/' incorporated mid-
canopy, canopy, and supercanopy.

The small and medium trees (<50
cm dbh) were counted on a 5.6 m cir-

369

cular plot, while large trees (>50 cm
dbh) were counted on a 15 m circular
plot. The counts of 18 species of trees
were averaged over the stations for
each grid and were used in an all-
subsets regression.

Analyses

We used one-way analysis of vari-
ance (ANOVA) to evaluate differ-
ences in mammal abundances rela-
tive to three stand age classes, three
moisture classes of the old-growth
stands in each of the three regions
(Branscomb and Butte Creek area.
Willow Creek area, and Cave Junc-
tion area).

These analyses were done on the
three separate sets of data, without
reference to the each other. Interac-
tion among the factors was ignored
in these analyses. When significant
differences were found among cap-
ture frequencies of individual species
by classes of: age, moisture, or study
area, a multiple comparison test was
used to determine which of the
groups were significantly different. A
Tukey test (Zar 1984:186) was per-
formed if variances were found to be
equal, while a Games and Howell
modification was used in the case of
unequal variances (Keselman and
Rogan 1978).

Pearson product moment correla-
tion coefficients were calculated be-
tween capture frequencies for each
combination of trapping techniques
and between capture frequencies and
vegetation means over all stands.
Variables from ground cover, herb
and shrub cover, and canopy trees
were included in all-possible-subsets
regression analyses for each small
mammal species when a significant
correlation existed with capture fre-
quency from any capture technique.
Five-variable models were selected
for each species when greater than
100 individuals were captured by a
particular technique. Vegetation vari-
ables were excluded when found on
less than 25% of the stands.

Results and Discussion

Twenty-three species of small mam-
mals were captured during the

study, though several were repre-
sented by only a few individuals
(table 2). The three techniques dif-
fered in their effectiveness in captur-

r

Table 2.— Number of captures by species and trapping technique, from a
study of small mammals in northern Califomia and southern Oregon, 1984
and 1985.

Species

Pitfalls Snaptraps Uvetraps

Trowbridge's Shrew 892
(Sorex frowbridgii)
Pacific Shrew 33
(Sorex pocificus)

Vagrant Shrew 1
(Sorex vagrons)

Pacific Water Shrew 1
(Sorex bendidi)

Shrew-Moie 40

(Neurofrichus gibbsii)

Coast Mole 1

(Scapanus orahus)

Chipmunks 2

(Tomias spp.)

Golden-mantled Ground SquirrelO
(Sperm ophilus lateralis)
Northern Flying Squirrel
(GlaucorDys sabrinus)
Botta's Pocket Gopher
(Thomomys boffae)
Deer Mouse

(Peromyscus manicufatus)
Pinyon Mouse
(Peromyscus fruei)
Dusky-footed Woodrat
(Neofoma fuscipes)
Bushy-tailed Woodrat
(Neofoma cinerea)
California Red-backed Vole 572
(Clethrionomys califomicus)
Red Tree Vole
(Arborimus longicaudus)
California Vole
(Microfus califomicus)
Long-tailed Vole
(Microfus longicaudus)
Creeping Vole
(t^icrofus oregoni)
Black Rat
(Ratfus raffus)
Pacific Jumping Mouse
(Zapus frinofafus)
Short- tailed weasel
(Musfela erminea)
Number of Trapnights^

6
5
115
16
2
0

1
14
2
6
1

3
0

135,360

357
70
1

0
27
0
33
0
1
2

524
205
4
0
161
0
14
0
5
0

n

0

55,284

101

n

0

0

0

282

Total

1350

114

72

317

8

15

0

7

404

1043

213

434

28

34

5

5

101

834

0

1

5

33

0

2

10

21

0

1

0

14

6

6

23,367

214.011

°Totals were adjusted for traps damaged by bears, etc.

370

ing different species of mammals.
Five species had sufficient captures
(> 100 individuals or more, by one or
more of the trapping techniques) to
undergo intensive analyses. These
were the California red-backed vole,
deer mouse, pinyon mouse, Trowbr-
idge's shrew, and the combined chip-
munk species.

Associations with Area, Age, and
Moisture Class

Most mammals were found in all
three areas, with the exception of
three species with only 1-2 captured.
The California red-backed vole had

significantly fewer captures in the
more southerly Branscomb region
than in the central and northern re-
gions (table 3). The vole's abundance
was significantly correlated with true
firs (r = 0.46, P < 0.05), which were
found on 11 stands in the north and
no stands in the south. The two mice
(Peromyscus) species exhibited the
opposite trend with captures signifi-
cantly greater in the south than in the
north. The pinyon mouse was corre-
lated with solar index which is gen-
erally greater in the southern area.
The shrews and chipmunks were
found equally in all areas.

The red-backed vole was the only
species to have a significant associa-

tion with age of the forest stand (P <
0.01). This confirms the study of Ra-
phael and Barrett (1984) and Raphael
(this volume) in the Willow Creek
area. Our capture frequency was
fairly low in stands aged at less than
150 years, while greater densities
were evident in many older stands
(fig. 1). No such relationship was
found for the deer mouse, although
Raphael and Barrett (1984) earlier
showed a significant association with
age in the Willow Creek area.

We tested the abundance of small
mammals in the three moisture
classes of old-growth forests: dry,
mesic, an wet. Among the five mam-
mal species with large sample sizes,
there were no differences in capture
frequency according to the various
moisture classes.

Therefore, we found that within
our study areas in the Douglas-fir
type, there were few significant or
strong associations between five
small mammals and age of the forest
stand. The stands chosen to represent
the different age and moisture
classes in this study were all natu-
rally occurring. The young stands
originated from fire or other cata-
strophic events, rather than by tim-
ber harvest, and therefore often were
heterogeneous in character with
structural and floristic components
similar to old-growth stands. Scat-
tered old trees and abundant dead
and down material were sometimes
present in young stands, characteris-
tics which are absent from stands
that originated from clearcuts; results
in even-aged stands may be very dif-
ferent.

Effectiveness of Capture

Captures of small mammals varied
greatly by trapping technique (table
1). The two mice were most effec-
tively captured by baited snap and
livetraps. Very few individuals were
collected in unbaited pitfalls. Microt-
ine voles, shrews, and moles were
trapped most efficiently by the pitfall

t/i 30 -
25-

3:
o

I

a.

2

o
o
o

UJ

OL.

to

UJ

=>
I—

Q-
-<

o

20-
15-
10-

0-

o = BR
• = WC

A = CJ

T

50

100

150 200
FOREST AGE

Figure 1.— Captures of Cdlfomia red-backed voles per 1000 trapnights in a study of snrxill
mamnrxji atHjndance relative to stand age, 1984-1985. BR = Branscomb stands, WC = Willow
Creek stands, CJ = Cave Junction stands.

Table 3.— Significance of differences in capture frequency by area for five
species of small mammals. The areas are CJ = Cave Junction, WC = Willow
Creek, and BR = Branscomb and Butte Creek. Methods with no significant
differences in capture frequencies at the various areas are Indicated by
NS; dashed lines indicate inadequate sample size.

Species

Snaptrap

Uvetrap

Pitfalls

California red-backed vole NS NS BR<WC+CJ

Deer mouse CJ<WC+BR CJ<WC+BR NS

Plnyor^ mouse CJ<BR CJ<BR —

Trowbridge's shrew NS NS NS

Chipmunks — NS — •

371

traps and somewhat by snaptraps.
Sciurids and woodrats were cap-
tured almost exclusively by livetraps.

We correlated the captures of each
species by the different techniques.
We found significant correlation be-
tween capture frequencies only in
those techniques effective at sam-
pling large numbers of a particular
species (table 4). Demonstrating the
closest agreement between tech-
niques were the pinyon mouse (r =
0.88 between snap and livetraps) and
the vole (r = 0.73 between the two
years of pitfall traps). The Trowbr-
idge's shrew, on the other hand,
showed no relationship between cap-
tures by pitfalls and snaptraps (r =
0.14), or pitfalls and livetraps (r =
0.09). Biological interpretation of
such varied results may be very diffi-
cult, as discussed in the following.

Vegetation Models

Depending on which method was
used to predict the dependent vari-
able, we obtained very different
vegetation models, potentially result-
ing in very different biological inter-
pretations. Models from snap and
livetrapping show that areas with
high captures of pinyon mice were
characterized by high densities of
pacific madrone and tanoak, high
solar index, and bare soils (r^ = 0.64
and 0.65) (table 5). Four of the five
habitat variables were identical in
both models suggesting that within
our study area, the pinyon mouse
used dryer, southern exposures with
exposed soils and large amounts of
hardwoods.

Models developed for the Trowbr-
idge's shrew from snaptrap and pit-
fall methods were quite different
(table 6). Only one variable was in-
cluded in both models, and the asso-
ciation with dogwood trees switched
from negative to positive. Both mod-
els included some indication of
greater use of older stands, i.e., the
model using snaptrap data included
well decayed logs and the livetrap

model incorporated the decomposed
litter layer, representing a well devel-
oped layer of organic soil. The incon-
sistency in these vegetation models
was predicted by the lack of correla-
tion between capture frequencies by
the two techniques. It appears that in
our Douglas-fir habitat type, the
shrew may be broadly distributed,
independent of finer vegetation com-
position.

Models for the red-backed vole
developed from capture frequencies
associated with different trapping
techniques (table 7) were more simi-
lar than those for the shrew, but less

similar than those for the pinyon
mouse. In models developed from
snap and livetrap captures, three of
the five variables were selected by
both models. Models from pitfall and
snaptrap data shared two of the five
variables selected. Models from pit-
fall and livetrap capture data also
shared two of the five variables se-
lected, but one of these variables
switched from a positive to a nega-
tive association. Only the response to
an abundant herbaceous layer was
consistent in models from all three
trapping techniques. Interpretation
of the snaptrap model suggests that

r

Table 4.— Correlation between years and methods of the capture fre-
quency of four small mammal species in snaptraps (Snap), livetraps (Live),
or pitfall traps (Pits). (Chipmunks were only caught in significant numbers In
livetraps and could not be compared).

Between years

Within yecffs

Snap84/live85 Pits:84/85 84:pits/snap 85:pits/live

California red-backed vole 0.540" 0.727** 0.459** 0.162

Deer mouse 0.392** 0.015 -0.092 0.320*

Pinyon mouse 0.884** 0.124 0.250 0.320*

Trowbridge's shrew 0.102 0.332* 0.141 0.088

'=P<Q.Qb.

"=p<om.

Tcrf>le 5,— Habitat association mod-
els for the pinyon mouse deter-
mined from capture frequencies
by two different trapping tech-
niques used. NS indicates the vari-
able was not selected. + or - Indi-
cates a positive or negative asso-
ciation with capture frequency.

Selected predictor
variables

Snaptrap Livetrap

Exposed rock NS +

Bar© soil + +

Solar index + +

Poison oak - NS

Tanoak + ■»■

Pacific madrone + +

0,64 0.66

Correlotion between capture fre-
quencies ofttie two tectiniques = 0.88.

Table 6.— Habitat association mod-
els for the Trowbridge's shrew de-
termined from ccqDture frequencies
by two different trapping tech-
niques used. Symbols as In table 5.

Selected predictor
variobles

Snaptrap Livetrap

Highly decayed logs

+

NS

Fern

-»■

NS

Dogwood shrub

NS

Dogwood tree

+

Deciduous oaks

+

NS

True firs

NS

+

Tanoak

NS

California hazel

NS

+

Deep titter layer

NS

+

0.59

0.55

Correlation between capture fre-
quencies by two techniques = 0.14,

372

the vole is associated with a fairly
moist habitat (abundant herbs and
presence of huckleberr}^. The pitfall
nnodel also suggests an association
with a moist habitat (more herbs and
lichens and less solar index). The liv-
etrap model includes some indication
of moist habitats (herbs, Rosa spp.,
and huckleberry) but also a sugges-
tion of a dryer habitat (solar index).

The deer mouse, despite its abun-
dance, had large differences between
variables selected in habitat models
(table 8). Its relative abundance did
not appear to be associated with the
same habitat variables in the same
way for the three different trapping
techniques. Only two of the 12 vari-
ables selected in these models were
included in more than one model

r

Jab\e 7.— Habitat association models for the Calitomia red-backed vole
determined from capture frequencies by three different trapping tech-
niques. Symbols as in table 5.

Selected predictor
variables

Herbs
Rose

Huckleberry
False cedar
Douglas-fir
solar index
Live oaks
Lichen
Grass

R2

Correlation between capture frequencies: snaptrap and llvetrap = 0.64 (P < 0.0?A'
\^ snaptrap and pitfall = 0.50(P < 0.01)j pitfall andJivetrap = 0. 16 (NS).

Table 8. ~ Habitat association models for the deer mouse determined from
capture frequencies by three different trapping techniques. Symbols as In
table 5.

Snaptrap

Llvetrap

Pitfall

+

+

+

+

NS

+

+

NS

+

NS

NS

+

NS

+

NS

+

NS

NS

NS

NS

+

NS

NS

+

0.58

0.55

0.63

Selected precfictor
variables

Snaptrap

Llvetrap

Pitfall

Lichen

True firs

Douglas-fir

California laurel

Pacific madrone

Manzanita

Rose

Dogwood
Deciduous oaks
Lower litter layer
Herbs

False cedars

R2

NS
NS
NS
NS
NS
NS
NS

0.33

NS
NS
NS
+

NS

+
+

NS
NS
NS

0.38

NS
NS
+

NS
NS
NS

NS
NS

+

+

0.63

Correlation between capture frequencies; snaptrap and livetrap = 0.39(9 < 0.01):
snaptrap and pitfall = -0.09 (NS); pitfall arid livetrap = 0.32 (P < 0.05).

with the same sign (avoidance of
Rosa spp. and preference for areas
with California laurel). Model dis-
parity may, of course, simply indi-
cate that one or more of the tech-
niques estimated the dependent vari-
able with considerable bias, thus pro-
ducing an erroneous model.

The chipmunks were captured pri-
marily by livetrapping. The resulting
5-variable model suggests that chip>-
munks were more common in the
true fir stands at high elevation that
had an understory of live oaks and
huckleberries (table 9).

While we are sure that there
would be some seasonal differences
in the habitat association patterns
from autumn captures in pitfalls and
summer captures in snap and liv-
etraps, we suggest that this seasonal
effect would be much less than the
differences that we noted, because of
the relatively low vagility of the
small mammals involved.

All capture methods are assumed
to sample individuals of a given spe-
cies at some unknown proportion of
their true abundance. These propor-
tions, within a species, likely differ
by capture method. If the capture ef-
ficiency of all methods were consis-
tent across sampled areas, then the
rank correlation of abundance be-
tween methods should be close to
1.0. However, for most species that
we studied, correlations of capture
frequencies between methods were
low and the ranking of stands based

Table 9.— Habitat association mod-
els for the chipmunks determined
from capture frequencies by liv-
etrapping. Symbols as in table 5.

Selected predictor

variables Llvetrap

True fir +
Douglas-fir

Lichen +

Vaccinium +

Live Oaks +

R2 0.59

373

on capture frequencies varied con-
siderably depending on technique
used. This suggests that the assump-
tion of a constant proportion of cap-
tures, within a given method, across
sampled areas was violated.

Acknowledgments

We thank the many members of the
old-growth field crew for their hours
of work under difficult conditions.
Linda Doerflinger and Howard Sakai
were helpful in many ways, keeping
the field crews supplied, organizing
data, and efficiently expediting other
aspects of complex field work. We
are also grateful to Andrew B. Carey,
Richard Golightly, Barry R. Noon,
Nancy A. Tilghman, and Martin G.
Raphael who all made helpful com-
ments on the manuscript.

Literature Cited

Dueser, R.D. and H.H. Shugart. 1978.
Microhabitats in a forest-floor
small mammal fauna. Ecology
59:89-98.

Frank, E. C. and R. Lee. 1966. Poten-
tial solar beam irradiation on
slopes. U.S. Dept. Agric. For. Serv.
Res. Pap. RM-18.

Hagar, D.C. 1960. The interrelation-
ships of logging, birds, and timber
regeneration in the Douglas-fir re-
gion of northwestern California.
Ecology 41:116-125.

Keselman, H.J. and J.C. Rogan. 1978.
A comparison of the Modified-
Tukey and Scheffe methods of
multiple comparisons for pairwise
contrasts. Journal of the American
Statistical Association 73:47-52.

MacGracken, J.G., D.W. Uresk, and
R.M. Hansen. 1985. Habitat used
by shrews in southeastern Mon-
tana. Northwest Science 59:24-27.

M'Closkey, R.T. 1975. Habitat Di-
mensions of white-footed mice,
Peromyscus leucopus. American
Midland Naturalist 93:158-167.

Manuwal, D.A. and M.H. Huff. 1987.
Spring and winter bird popula-
tions in a Douglas-fir forest sere.
Journal of Wildlife Management
51:586-595.

Raphael, M.G. and R.H. Barrett.
1984. Diversity and abundance of
wildlife in late succession
Douglas-fir forests, p. 352-360. In
New Forests for a Changing
World. Proceedings 1983 Conven-
tion of the Society of American
Foresters. 650 pp.

Raphael, M.G., K.V. Rosenberg, and
B.G. Marco t. In press. Large-scale
changes in bird populations of
Douglas-fir forests. Northwestern
California. Bird Conservation 3.

Rosenzweig, M.L. 1973. Habitat se-
lection experiments with a pair of
coexisting heteromyid rodent spe-
cies. Ecology 54:111-117.

Ruggiero, L. F. and A. B. Carey. 1984.
A programmatic approach to the
study of old-growth forest-wildlife
relationships, p. 340-345. In Pro-
ceedings of the 1983 convention of
the Society of American Foresters:
New forests for a changing world.
Society of American Foresters, Be-
thesda, MD.

Seber, G.A.F. 1981. The estimation of
animal abundance and related
parameters. Second Edition. Ox-
ford University Press, New York.

Thomas, J. W. (Ed.). 1979. Wildlife
habitats in managed forests.
USDA Forest Service Agriculture
Handbook No. 553.

Zar, J.H. 1984. Biostatistical Analysis.
Prentice-Hall, Inc. Englewood
Cliffs, N.J. 718 pp.

374

Small Mammals in
Streamside Management
Zones in Pine Plantations^

James G. Dickson^ and J. Howard
Williamson^

Abstract.— Small mammals were captured in live
traps in 6 mature-forested streamside management
zones of 3 widths, narrow (< 25 m), medium (30-40
m), and wide (50-90 m). which traversed young,
brushy pine plantations. More small mammals were
captured in the narrow zones (165) than in the me-
dium (82), or wide zones (65).

Many second-growth pine-hardwood
stands in southern forests are being
cut and replaced by pine plantations,
especially on industrial land. From
1971 to 1986, the amount of
Mid south timberland in pine planta-
tions increased from 6 to 8% (Birdsey
and McWilliams 1986). White-tailed
deer adapt well to young brushy
clearcuts with ample forage and soft
mast. Also, many species of birds are
abundant in this diverse brushy habi-
tat (Dickson and Segelquist 1979).
But the effects of clearcutting and
planting on all vertebrate species are
not well assessed or defined.

Various environmental conces-
sions are being implemented along
with stand conversion. One practice
used to protect water quality and
enhance wildlife habitat is to retain
mature forest stands along intermit-
tent and permanent streams when
adjacent stands are cut and planted
to pines (Dickson and Huntley 1986,
Seehorn 1986). These areas of mature
pine or pine-hardwoods are called
riparian zones, filter strips, stringers,
streamers, or streamside manage-
ment zones (SMZ). These areas en-

' Paper presented at symposium, Man-
ogement of Amphibians. Reptiles, and
Small Mammals in Northi America. (Flag-
staff. AZ. July 19-21, 1988.)

'James G. Dickson. Supervisory Research)
Wildlife Biologist, Wildlife Habitat Labora-
tory. Southern Forest Experiment Station,
USDA Forest Service, Nacogdoches. Texas.

^J. Howard Williamson. Forestry Techni-
cian. Wildlife Habitat Laboratory. Southern
Forest Experiment Station. USDA Forest Serv-
ice. Nacogdoches. Texas.

hance habitat diversity and "edge,"
offer suitable habitat for wildlife spe-
cies associated with mature stands,
serve as travel corridors for animals,
and may permit genetic interchange
between otherwise isolated popula-
tions of animals. Retention of SMZ
for reduction of non-point pollution
and for wildlife has been widely rec-
ommended.

These mature hardwood strips can
be good squirrel habitat. In Missis-
sippi (Warren and Hurst 1980) and in
eastern Texas (McElfresh et al. 1980),
gray (Sciurus carolinensis) and fox (S.
niger) squirrel numbers were higher
in riparian areas than in adjacent up-
land stands. In another facet of the
present investigation, gray and fox
squirrels were abundant in SMZ
wider than 50 m but virtually absent
from zones less than 40 m wide
(Dickson and Huntley 1986). A wide
variety of reptiles and amphibians
were abundant in zones greater than
30 m wide, where a closed canopy
offered shaded understory, but were
scarce in SMZ less than 25 m wide,
which were dominated by low,
brushy vegetation (Rudolph and
Dickson In Press). The relationships
of SMZ and other wildlife species are
largely unknown.

The objective of this study was to
determine the relationship of SMZ
width to small mammal communi-
ties. We assessed the effects of nar-
row (<25 m), medium (30-40 m), and
wide (>50 m) SMZ widths on small
mammal captures in 6 SMZ in east-
ern Texas.

Study Areas and Mettiods

Study areas consisted of 6 pine plan-
tations on the western edge of the
southern coastal plains in eastern
Texas. Mature pine and hardwood
trees on the areas had previously
been harvested. The plantations had
been planted to loblolly pine (Pinus
taeda) seedlings 5 to 6 years before
this study was begun and were vege-
tated by diverse flora, dominated by
hardwood and other woody brush.
Oaks {Quercus spp.) and sweetgum
(Liquidambar styraciflua) sprouts,
American beautyberry (Callicarpa
americam), blackberry and dewberry
(Rubus spp.), and sumac {Rhus spp.)
were abundant.

Each of the 6 study areas was trav-
ersed by a SMZ of mature vegeta-
tion. Dominant trees (> 13 cm dbh) in
decreasing order of abundance and
stem density (No. /ha) were as fol-
lows: sweetgum, 63; white oak (Q.
alba), 36; southern red oak (Q. fal-
cata), 28; red maple (Acer rubrum), 19;
black gum (Nyssa sylvatica), 14;
shortleaf pine (P. echinata), 14; and
eastern hophornbeam (Ostrya virgini-
ana), 14. Dominant understory vege-
tation (5-13 cm dbh) and stem den-
sity (No./ha) included sweetgum,
140; eastern hophornbeam, 71; black
gum, 40; flowering dogwood (Cornus
florida), 40; loblolly pine, 21; and red
maple, 19.

Assigned treatments were 3 SMZ
widths: narrow (<25 m), medium
(30-40 m), and wide (>50 m). Two
replications of each treatment were

375

sampled at 2 locations. In each of the
6 study areas two 200-nn transects
were established along each of the 6
streamside zones. Distance from
points along the transects to the SMZ
edge was variable because each zone
orientation changed somewhat with
stream meanders. Thirteen Sherman
live traps were placed 12.5 m apart
on each of the 12 transects. Trapping
was conducted 4 consecutive nights
in each of 2 consecutive weeks (8
nights) during February and March
in 1986 and again in 1987 (52 traps/
treatment X 8 nights X 2 years = 832
trap nights). Traps were baited with
oatmeal each morning and checked
the following morning.

Captures per treatment were ap-
proximately normally distributed ac-
cording to the Kolmogorov-Smirnov
Goodness of Fit Test. Each of the 3
treatments was tested for differences
between years with the T-Test. There
were no significant differences be-
tween years (P > .10); therefore, cap-
ture data were combined for both
years. Treatment effects (captures/
treatment) were tested for differ-
ences by ANOVA and the Duncan's
Multiple Range Test at the 0.05 level
of confidence.

White-footed mice (Peromyscus leu-
copus) and cotton mice (P. gossypinus)
were grouped together because of
difficulty in positive field identifica-
tion. Davis (1974) determined that
white-footed mouse adults were
smaller (15 to 25 g, as opposed to >
30 g for the cotton mouse) and had
brighter colors. Also, adult hind-foot
length was shorter (21 mm, as op-
posed to 23 mm for cotton mice).
However, numerous sub-adults were
captured during the trapping period,
making identification extremely diffi-
cult.

Results and Discussion

Significantly more small mammals
were captured in the narrow SMZ
(165) than were captured in the me-
dium (82) or wide (65) SMZ (table 1).

The absence of tree canopy in the
narrow zones permitted dense,
brushy vegetation growth, abundant
seeds, and dense logging slash cover,
but medium and wide zones were
characterized by shaded sparse
understories under closed canopies.
Other studies have shown higher
densities of small mammals in young
brushy stands than in mature stands.
In an earlier study in eastern Texas,
64 small mammals were captured in
a 6-year-old clearcut, but only 24 in a
pine-hardwood stand more than 35
years old. Small mammal species di-
versity was also higher in the young
stand (Fleet and Dickson 1984). In
pine plantations in Georgia, small
mammal abundance was higher in 1-
to 5-year-old pine plantations than in
older stands with closed canopies
(Atkeson and Johnson 1979). Seed-
eaters were abundant in the 1 -year-
old plantation, but herbivores were
abundant in older young brushy
stands.

In Pennsylvania, relative abun-
dance of small mammals was greater
in recent clearcuts of both northern
hardwood and oak forests than in
adjacent mature stands (Kirkiand
1978). A similar pattern was noted in

deciduous and boreal forests in West
Virginia (Kirkiand 1977). After
clearcutting, small mammal abun-
dance and diversity increased and
remained relatively high until stands
returned to forest. In Arizona, rodent
populations were higher in thinned
ponderosa pine (P. ponderosa) stands
with slash than in unthinned stands
(Goodwin and Hungerford 1979).

The most abundant species, the
fulvous harvest mouse (Reithrodonto-
mys fulvescens) and the white footed
mouse/cotton mouse complex, were
much more abundant in the narrow
zone. For the fulvous harvest mouse,
there were 73 captures in the narrow,
4 in the medium, and 3 in the wide
zones.

Apparently, the dense brushy
vegetation with ample down logging
slash provided ideal habitat for this
species. There was abundant vegeta-
tive forage, seeds, and dense log and
brush cover. Schmidly (1983) de-
scribed the best habitats for fulvous
harvest mice in the pineywoods as
grassland, pine-grass ecotone, and
grass-brush. In an earlier study in
eastern Texas (Fleet and Dickson
1984), fulvous harvest mice were
captured regularly in a young pine

Table Number of small mammals captured In streamside management
zones in pine plantations (832 trap nights) per treatment.

SMZ width

Narrow

Medium

Wide

Hispid Cotton Rat

(Sigmodon hispidus)

9

3

Fulvous Harvest Mouse

(Reithrodontomys fulvescens)

73

4

3

Eastern Harvest Mouse

(Reithrodontomys humulis)

1

1

White-footed and Cotton Mouse

(Peromyscus leucopus and gossypinus)

76

67

50

Golden Mouse

(Perom yscus muttalii)

3

4

Florida Wood Rat

(Neotoma floridana)

3

5

4

Short-tailed Shrew

(Blarina brevicauda)

2

4

Totals

165

82

65

376

plantation, but were not captured in
the adjacent mature pine-hardwood
stand. In a study of small mammal
populations in 5 pine stands in Lou-
isiana, fulvous harvest mice were
captured most frequently in a pine
seed-tree harvest cut having dense
hardwood brush (Hatchell 1964).

Differences among treatments
were less pronounced for the Pero-
myscus complex, with captures of 76
in the narrow, 67 in the medium, and
50 in the wide SMZ. In a 1 -year-old
pine plantation in Georgia, the white-
footed mouse was the dominant sp)e-
cies (Atkeson and Johnson 1979). It
also was the most abundant species
in the mature oak-hickory forest type
in eastern Tennessee (Dueser and
Shugart 1978). Cotton mice were cap-
tured regularly in 5 mature pine
stands in Louisiana (Hatchell 1964)
and in a pine-hardwood stand in
eastern Texas (Fleet and Dickson
1984). Neither species was captured
in a pine plantation in the Texas
study. Schmidly (1983) describes pre-
ferred habitat of the cotton mouse as
flatland hardwood, flatland hard-
wood-pine, and lower slope hard-
wood-pine. McCarley (1954) associ-
ated the white-footed mouse with
upland forest habitat.

Six other species were not cap-
tured frequently enough for conclu-
sions concerning habitat preference.
Habitat preferences have been docu-
mented to some degree in other stud-
ies. The hispid cotton rat is often
very abundant and normally is asso-
ciated with low, dense vegetation
(Atkeson and Johnson 1979, Reet and
Dickson 1984, Goertz and Long 1973,
Schmidly 1983). It has occasionally
been found in habitats dominated by
early successional grasses and forbs.

The golden mouse is associated
with forested stands having low,
dense vegetation (Fleet and Dickson
1984, Hatchell 1964, McCarley 1958).
The Florida wood rat occupies for-
ested upland and streamside habitat
and thrives in bottomland hardwood
stands with low brushy understories
(Schmidly 1983). Short-tailed shrews

were captured in the medium (2) and
wide zones (4). Other investigations
have found them inhabiting a variety
of mature stands (Fleet and Dickson
1984, Hatchell 1964, Schmidly 1983).

In conclusion, more small mam-
mals, especially fulvous harvest
mice, were captured in narrow SMZ
than in medium and wide SMZ. Ap-
parently, this is related to the abun-
dance of low, dense vegetation, with
ample forage, fruits, and seeds; and
down logs and logging slash. But
medium and wide SMZ with closed
tree canopies provide limited mature
habitat for some species associated
with mature stands, such as the
short-tailed shrew, and are positive
for a variety of other wildlife.

Acknowledgments

We thank Jimmy C. Huntley for trap-
ping assistance and James A. Neal
and W. V. Robertson for reviewing
an earlier draft of this manuscript.

Literature Cited

Atkeson, Thomas D. and A. Sydney
Johnson. 1979. Succession of small
mamimals on pine plantations in
the Georgia piedmont. American
Midland Naturalist 101:385-392.

Birdsey, R. A. and W. H. McWil-
liams. 1986. Midsouth forest area
trends. USDA Forest Service Re-
source Bulletin SO-107. 17 p.
Southern Forest Experiment Sta-
tion, New Orleans, La.

Davis, William B. 1974. The mam-
mals of Texas. Texas Parks and
Wildlife Department Bulletin No.
41, Ausrin, TX. 294 p.

Dickson, James G. and Jimmy C.
Huntley. 1986. Streamside man-
agement zones and wildlife in
southern forests: the problem and
squirrel relationships, p. 37-39. In
Managing southern forests for
wildlife and fish — a proceedings.
J.G. Dickson and O.E. Maughan,
eds. USDA Forest Service General

Technical Report SO-65. 85 p.
Southern Forest Exf)eriment Sta-
tion, New Orleans, La.

Dickson, James G. and C. A.
Segelquist. 1979. Breeding bird
populations in pine and pine-
hardwood stands in East Texas.
Journal of Wildlife Management
43:549-555.

Dueser, Raymond D. and H. H.
Shugart, Jr. 1978. Microhabitats in
a forest-floor small mammal
fauna. Ecology 59:89-98.

Fleet, Robert R. and James G.

Dickson. 1984. Small mammals in
two adjacent forest stands in East
Texas, p. 264-269. In Proceedings
of workshop on management of
nongame species and ecological
communities. [Lexington, Ky.,
June 11-12, 1984] W. C. McComb,
editor. 404 p. University of Ken-
tucky, Lexington.

Goertz, John W. and Roland C. Long.
1973. Habitats of five species of rat
in Louisiana. American Midland
Naturalist 90:460-465.

Goodwin, John G., Jr., and C. Roger
Hungerford. 1979. Rodent popula-
tion densities and food habits in
Arizona ponderosa pine forest.
U.S. Department of Agriculture,
Forest Service Research Paper RM-
214. 12 p. Fort Collins, Co.

Hatchell, Glyndon E. 1964. Small-
mammal species and populations
in the loblolly-shortleaf pine forest
type of Louisiana. U.S. Forest
Service Research Paper SO-10, 12
p. Southern Forest Exp)eriment
Station, New Orleans, La.

Kirkland, Gordon L., Jr. 1977. Re-
sponses of small mammals to the
clearcutting of northern Appala-
chian forests. Journal of Mam-
malogy 58:600-609.

Kirkland, Gordon L., Jr. 1978. Initial
responses of small mammals to
clearcutting of Pennsylvania hard-
wood forests. Proceedings of the
Pennsylvania Academy of Science
52:21-23.

McCarley, Howard. 1958. Ecology,
behavior and population dynam-
ics of Peromyscus nuttalli in eastern

377

Texas. Texas Journal of Science
10:147-171.

McCarley, W. H. 1954. The ecological
distriburion of Peromyscus leucopus
species group in eastern Texas.
Ecology 35:375-379.

McElfresh, Robert W., Jack M. Inglis,
and Bennett A. Brown. 1980. Gray
squirrel usage of hardwood ra-
vines within pine plantations.
Louisiana State University Annual
Forestry SymfX)sium 19:79-89.

Rudolph, D. Craig and James G.
Dickson. Reptiles and amphibians
in streamside management zones
of clearcuts. (In press).

Schmidly, David J. 1983. Texas mam-
mals east of the Balcones Fault
zone. 399 p. Texas A & M Univer-
sity Press. College Station.

Seehorn, Monte E. 1986. The influ-
ence of silvicultural practices on
fisheries management: effects and
mitigation measures, p. 54-63. In
Managing southern forests for
wildlife and fish — a proceedings.
J. G. Dickson and O. E. Maughan,
eds. USDA Forest Service General
Technical Report SO-65. 85 p.
Southern Forest Experiment Sta-
tion, New Orleans, La.

Warren, R. C. and G. A. Hurst. 1980.
Squirrel densities in pine-hard-
wood forests and streamside man-
agement zones. Proceedings of the
Annual Conference of the South-
east Association, Fish and Wildlife
Agencies 34:492-498.

378

Patterns of Relative Diversity
Within Riparian Snnall
Mammal Communities, Platte
River Watershed, Colorado^

Thomas E. Olson^ and Fritz L. KnopP

Abstract.— Relative diversity within and between
small mammal assemblages of riparian and upland
vegetation was evaluated at 6 study areas across
an elevational gradient. In contrast to avian diversity
analyses conducted at the same sites, species rich-
ness, relative diversity, and faunal similarity of small
mammals were greater among upland rather than
riparian communities across the dine. Beta diversity
between riparian and upland small mammal com-
munities is greater at higher elevations within the wa-
tershed. These higher elevation portions of water-
sheds must be emphasized in management strate-
gies to conserve regional integrity of native small
mammal faunas.

Figure 1 .—Location of study areas wittiin the Platte River drainage, norttiern Colorado, 1981 .

Riparian communities in the western
states are mesic vegetative associa-
tions occurring along ephemeral, in-
termittent, and perennial streams.
Although relatively limited in area,
these communities contribute more
biotic diversity within a region than
upland vegetation communities
(Thomas etal. 1979).

Riparian communities have been
substantially affected by land-use
changes such as conversion to agri-
culture, grazing, and water manage-
ment (Knopf et al. 1988). Further al-
terations in the western United States
have been caused by the widespread
naturalization of salt cedar {Tamarix
spp.) (Horton 1977) and Russian-
olive (Elaeagnus angustifolia) (Olson
and Knopf 1986). Because of the bio-
logical significance and potential for
perturbations caused by conflicting
land uses, riparian communities have
been the focus of numerous technical
conferences during the past 10 years
(Knopf et al. 1988).

An earlier study of the pattern of
avian species diversity in riparian
and upland study areas within a wa-
tershed (Knopf 1985) showed that

'Paper presented at symposium. Man-
agement of Amphibians. Reptiles, and
Small Mammals in Nortii America. (Flag-
staff . AZ. July 19-21. 1988.)

'Wildlife Biologist. Dames & Moore. 175
Cremona Drive. Suite A-E. Goleta. Califor-
nia 931 17.

^Leader, Riparian Studies. U.S. Fist) and
Wildlife Service. 1300 Blue Spruce Drive. Fort
Collins, Colorado 80524-2098.

although more species of birds occur
in riparian vegetation, upland sites
contribute more to avifaunal diver-
sity between habitats (beta diversity)
and within a region (gamma diver-
sity). Those findings were attributed
to greater similarity among riparian
avifaunas across the altitudinal cline
due to the riparian vegetation pro-
viding a corridor for movement of
birds within a region. Beta diversity
between upland and riparian avian
assemblages was greatest at the up-
per and lower ends of a watershed,
and the study concluded that avifau-

nal conservation efforts should be
concentrated at those sites.

Implications of the earlier avian
study to conservation of small mam-
mal assemblages are unclear. Numer-
ous studies (Anderson et al. 1980;
Honeycutt et al. 1981; Kirkland 1981)
have examined small mammal distri-
bution along environmental gradi-
ents, but with a focus on upland
rather than riparian species assem-
blages. The objectives of this study
were to evaluate diversity within,
and between, small mammal assem-
blages of riparian and upland vegeta-

379

Figure 2.— Study areas: South Platte River (SPR), 1200 m; Lone Pir>e Creek (LPC), 1909m; Sheep
Creek (SC), 2341 m; Illinois River (IR), 2S00m; Laramie River (LR). 2631 m; South Fork of the Cache
la Poudre River (SFPR), 2747 m.

380

tion across an elevational gradient.
We believed that the results would
indicate relative elevations within
watersheds at which small mammal
conservation efforts should be fo-
cused. Such efforts could include
policies of state and federal agencies
concerning type of land use within
portions of watersheds.

Study Areas

Six study areas ranged in elevation
from 1200 to 2747 m within the Platte
River drainage of northern Colorado
(fig. 1). With the exception of an al-
pine area, riparian communities
within each major life zone of upland
vegetation along the Front Range
were represented (fig. 2). Within each
upland, we located a riparian site
that contained a permanent stream.
Cattle grazing had not occurred on
any of the study areas for at least
three years prior to 1981.

The South Platte River (SPR) study
area was on the South Platte Wildlife
Management Area, 2 km south of
Crook, Logan County (elevation 1200
m). This community was dominated
by sand sagebrush mixed-prairie.
Several species of grass and 1 woody
species, sand sagebrush (Artemisia
filifolia), occurred on the upland
sandhills. Dominant riparian species
were plains cottonwood (Populus
sargentii), western snowberry (Sym-
phoricarpos ocddentalis), and willows
(Salix spp.).

The Lone Pine Creek (LPC) study
area was 11 km west of Livermore,
Larimer County, at 1909 m elevation.
This area of mountain shrub transi-
tion vegetation was dominated by
true mountain mahogany (Cercocar-
piis montanus), antelope bitterbrush
(Purshia tridentata), and gooseberry
(Ribes spp.) in the upland site. The
riparian site was dominated by
plains cottonwood, willows, and
common chokecherry (Prunus virgini-
am). Rocky Mountain junipers
(Juniperus scopulorum) were scattered
throughout both sites.

The Sheep Creek (SO study areas
was 21 km north of Rustic, Larimer
County, at an elevation of 2341 m
(fig. 2). Ponderosa pine (Pinus ponder-
osa) forest, along with scattered big
sagebrush (Artemisia tridentata) domi-
nated the upland site. Riparian vege-
tation was dominated by narrowleaf
cottonwood (Populus angustifolia),
willows, and alders (Alnus spp.).

The Illinois River (IR) study area
contained sagebrush steppe vegeta-
tion and was within the Arapaho Na-
tional Wildlife Refuge, 10 km south
of Walden, Jackson County (eleva-
tion 2500 m). Upland vegetation was
predominantly big sagebrush. The
riparian site included eight species of
shrub willows dominated by S. gey-
eriana (Cannon and Knopf 1984).

The Laramie River (LR) study area
was 6.5 km north of Chambers Lake,
Larimer County (elevation 2631 m).
Aspen (Populus tremuloides) domi-
nated the upland site, along with
Douglas fir (Pseudotsuga menziesii)
and lodgepole pine (Pinus contorta).
The riparian site was comprised of
shrub willows.

The highest study area (elevation
2747 m) was along the South Fork of
the Cache la Poudre River (SFPR), at
the Pingree Park Campus of Colo-
rado State University, Larimer
County. Upland vegetation was com-
px)sed of lodgepole pine, limber pine
(Pinus flexilis), Engelmann spruce
(Picea engelmannii), Douglas-fir, sub-
alpine fir (Abies lasiocarpa), and a
sparse understory of aspen. The ri-
parian site was exclusively shrub wil-
lows.

Methods

Small mammal trapping was con-
ducted in 1981 to determine the rela-
tive abundances of small mammal
species at the 6 study areas. In each
study area, two 400-m survey lines
were established, including one ri-
parian and one upland site. Riparian
survey lines were within riparian
vegetation and generally paralleled

the stream course. Upland survey
lines began 500 m from the stream
and were oriented perpendicular to
the direction of the stream.

Trap surveys were conducted be-
tween 30 July and 26 August 1981 .
Survey lines included 20 trap stations
spaced 20 m apart. Each trap station
contained 1 rat trap and 2 museum
special snap traps located within a
1.8-m radius of the measured point.
Three traps were used at each station
to minimize any bias in the data to-
ward more aggressive species, such
as Peromyscus maniculatus. Traps
were baited with a mixture of
ground raisins, carrots, and chipped
beef, blended in a peanut butter base,
and set for 3 consecutive nights in
the riparian and upland sites of a
study area sim.ultaneously. Traps
were checked in the morning and
evening during the 72 hours. Thus,
trap effort p)er study area was 360
trap-nights, including 180 trap-nights
each in the upland and riparian sites.
Total number of trap-nights for all
study areas was 2160.

Diversity indices were calculated
to compare species diversity within
(alpha) and between (beta) riparian
and upland sites across the altitudi-
nal cline. Because preference for type
of index varies, we selected two each
of the most commonly used indices
to measure alpha (Simpson Index,
Shannon-Weiner Index) and beta (co-
efficient of community, percentage
similarity) diversity (Whittaker 1975:
95,118).

The former two differ in the gen-
eral relationship between output
value and species diversity. Shan-
non-Wiener Index (H') varies directly
with number of species trapped,
while the Simpson Index (C) varies
inversely. Coefficient of community
(CO values are ratios of the number
of species common to both riparian
and upland sites to the total number
of species occurring in the two sites
combined. Those values are based
only on presence or absence and vary
directly with diversity. Although
percentage similarity values are

381

based on the differences in impor-
tance values between the two sites,
they also vary directly with diversity.

Results

A total of 471 small mammals of 22
species was trapped in all study ar-
eas in 1981 (table 1). Three species
(14% of all species captured) were
trapped in riparian sites only, 9 spe-
cies (41%) were trapped in upland
sites only, and 10 species (47%) were
trapped in both. Nine species (41%)
were rare, being represented by 2 or
fewer captures.

Within-Habitat Comparisons

Species composition within riparian
sites differed among the study areas.
Deer mice (Peromyscus maniculatus),
voles (Microtus spp.), and jumping
mice (Zapus princeps) accounted for
182 of 189 (96%) total captures at the
3 lower study areas, although jump-
ing mice did not occur at SPR. In
contrast, shrews (Sorex spp.) ac-
counted for 69% of all captures at the
remaining, higher areas. Of 68 small
mammals trapped at the higher sites,
only 14 (20%) were either voles or
jumping mice. No deer mice were
trapped in riparian sites at elevations
higher than 2293 m.

Changes in species composition of
small mammals in upland sites were
not distinct. Deer mice were the most
frequently trapped of all species at
the 4 intermediate study areas. Over-
all, 112 of 214 (52%) small mammals
trapped in the uplands were deer
mice. The next 3 species in abun-
dance (least chipmunk [Tamias mini-
mus], northern grasshopper mouse,
[Onychomys leucogasterj and prairie
vole [Microtus ochrogaster]) accounted
for only 67 of 214 (31%) total cap-
tures. Of these 4 species, only the
deer mouse was trapped at all 6 sites.

Species richness varied among ri-
parian and upland sites. The number
of small mammal species trapped in

riparian sites was least at the lowest
elevation study area (SPR) and great-
est at the second highest study area
(LR) (table 2). All other riparian sites
were intermediate in species richness
with no apparent altitudinal trend.
Values for Simpson's Index (C) (a
measure of the concentration of
dominance) and Shannon-Wiener
Index (HO (Whittaker 1975:95)
yielded similar results.

The highest diversity among ripar-
ian sites occurred at LR, which had
the lowest dominance. The SPR
study area, which had a high C
value, also contained very low spe-
cies diversity.

The number of small mammal spe-
cies trapped in upland sites was
comparatively high at 2 of 3 study

areas under 2500 m (LPC and SO
and at the highest elevation study
area (SFPR) (table 2). Simpson's In-
dex values varied from a high at IR
(2500 m elevation) to a low at SFPR
(2747 m). Shannon-Wiener values in
upland sites ranged from a low of
0.22 at IR to a high of 0.74 at SFPR.

A matrix of percentage similarity
values (Whittaker 1975:118) revealed
a mean similarity of 0.29 + 0.06
among upland sites and 0.18 ± 0.05
among riparian sites. These results
suggest that small mammal commu-
nities in upland sites were more simi-
lar across the cline than were those in
riparian sites. Overall, beta diversity
along the altitudinal gradient was
greater (less faunal mixing) in ripar-
ian sites.

r

Table 1 .—Species of small mcmimals trapped at 6 study areas across an
altltudirKil cline, northern Colorado, 1981.

Study arecP

SPR

[PC

SC

IR

LR SFPR

Rip" Upl^ Rip Up! Rip Up! Rip Upl Rip Up! Rip Up!

Sorex cinereus
S. monficdus
S. spp.

Sylvilagus nutfallii
Lepus americanus
Tamias minimus
T, quadriviifafus
T. umbrinus

Spermophilus spilosoma
S. lateralis

Thomomys talpoides
Dipodomys ordii
Reittvodontomys megafotis
Peromyscus moDiculatus
Onychomys leucogaster
Neotoma mexicana
Clefhrionomys gapperi
Microtus longicaudus
M. ochtrogaster
M. spp.

Lagurus curfatus
Zapus princeps

Totals

1

15
2

20
10
1

17

2
10

2
66

A
2
1

10

1

42
1

33
2

1

68

1

n

7 16

18
2

23 1

1

67

18 83 84 39 38 17 21 20 21 31 32

°Study areas: SPR = Soutti Platte River; LPC = Lone Pine Creek; SC = Stieep Creek; IR
Illinois River; LR = Laramie River; SFPR = Soutti Fork ofCactie la Poudre River.

''Rip = Riparian site.
'=Uj^ = Upland site.

382

Between-Habrtot Comparisons

Species richness was substantially
higher in upland sites than in adja-
cent riparian sites at the lowest and
highest study areas (table 2). The val-
ues were similar at 3 study areas of
intermediate elevation. Only at LR
(the second-highest area) was species
richness higher in the riparian site.
At that study area, number of species
trapped in riparian was greater than
the upland even when captures of

Lepus americanus and Thomomys
talpoides were excluded.

Coefficient of community (CC)
values (Whittaker 1975:118) suggest
that small mammal communities in
riparian and adjacent upland sites
were relatively similar at lower ele-
vations, and became more dissimilar
at 2500 m and higher (table 3). More
species (3) were common to both ri-
parian and upland sites at the 3
lower shidy areas than at the higher
areas. Percentage similarity (PS) val-

]^^LlnZ^^''^\'^l^''^'' ^^'""P'^" '"^^^ = ^ relative alpha diversity

Riparian

Upland

Study Area

South Ratte River (SPR)
Lone Pine Creek (LPC)
Sheep Creek (SC)
Illinois River (IR)
Laramie River (LR)
South Fork of Cache la
Poudre River (SFPR)

Number of
species (0«

2
6
5
3
7
3

0.94

0.42

0.40

0.45

0.21

0,52

0.06
0.46
0.50
0.38
0.74
0.33

Number of
species <0

5
7
7
3
3
7

0.38
0.67
0.34
0.75
0.68
0,21

m

0.53
0.30
0.58
0.22
0.26
0.74

s s
I p2 = X (n,/N)2
i=l i=l
s

- I p, log p
=1

In^riJ^ria";; n^^^ ^.^^^^ ^^^''^ °' communities

study area

South Ratte River (SPR)
Lone Pine Creek (LPC)
Sheep Creek (SC)
Illinois River (IR)
Laramie River (LR)
South Fork of Cache la
Poudre River (SFPR)

No. species Species common
(riparian/upland) to both sites

2/5
6/7
5/7
3/3
7/3
3/7

2
4
3
0
1

2

0.57

0.08

0.62

0.66

0.50

0.26

0.00

0.00

0.20

0,25

0.40

0.12

°CC (Coefficient of community) = 2S ^(S^ +
''PS (Percentage simiiarity) = min (p^ or

ues indicate the same trend, with the
exception of the lowest study area.
The low value at that study area is
due primarily to the abundance of
Peromyscus maniculatus dominating
this calculation (table 1).

Discussion

To date, shidies of small mammal
distribution along environmental
gradients (Anderson et al. 1980;
Armstrong et al. 1973; Honeyciitt et
al. 1981; Kirkland 1981) have been
conducted in upland sites. Knopf
(1985) compared distribution of
breeding birds in riparian and adja-
cent upland sites within the 6 areas
used in this study. The focus of this
study was to analyze patterns of
small mammal faunal similarity
within and between riparian and ad-
jacent upland sites in the same water-
shed. Such patterns, although based
on relatively small sample sizes, may
indicate elevations along the gradient
at which management should be em-
phasized to conserve regional diver-
sity.

A pronounced change in species
composition occurred within riparian
sites at 2500 m elevation. The study
areas below that elevation, represent-
ing foothills and plains, were domi-
nated by deer mice and voles. At
2500 m and above, dominance
shifted primarily to shrews. The
means for PS values comparing the 3
lower study areas (0.31 ± 0.10) and 3
higher study areas (0.43 ± 0.02) were
both considerably higher than the
mean for all study areas (0.18 + 0.05).
Faunal similarity changed as riparian
sites shifted from cotton wood- wil-
low to willow shrub systems. This
shift in small mammal community
composition could have reflected a
shift from xeric site willows (S.
amygdaloides, S. exigua) to mesic site
willows as described in Cannon and
Knopf (1984). Other factors may have
influenced composition of small
mammal communities. Among those
suggested in previous research are

383

soil type, nutrient availability, and
vegetation structure (Huntley and
Inouye 1984, Moulton et al. 1981).
Others have found specific mi-
crohabitat components to be impor-
tant (cf. M'Closkey 1981, Szaro and
Belfit 1987).

Dominance by deer mice was par-
ticularly obvious at the lowest site,
SPR, where 65 of 67 captures were of
this species. The remaining 2 small
mammals trapped were western har-
vest mice (Reithrodontomys megalotis).
These findings were supported by an
earlier study of total small mammal
richness conducted in the same study
area. During the 1982 and 1983 field
seasons of that study, 98.3% of all
small mammals captured in 25,000
trap-nights were deer mice and west-
ern harvest mice (Bennett 1984).

High numbers of deer mice
trapped could indicate behavioral
differences (deer mice being more
aggressive), rather than a dominance
in absolute numbers. We believe,
however, that the number trapped
reflected higher relative abundances
of Peromyscus maniculatus for several
reasons. First, although this species
was the most frequently trapped spe-
cies, it dominated only 4 of 12 total
sites, and was infrequent to absent at
7 sites (table 1). Total captures in 180
trap-nights at each of those 4 sites
(riparian at SPR, upland at IR, both
sites at LPC), ranged from 18 to 68.
That is, deer mice captures ac-
counted for no more than 38 percent
of all available traps at any site.
Moreover, in the riparian site at SPR
(where deer mice were most com-
monly caught), the percentages of all
captures that were deer mice were
similar for this study (97%) and that
of Bennett (1984) (95%).

Dominance by ecological general-
ists at the lowest site, SPR, likely is
explained by periodic catastrophic
events, specifically flooding. In con-
trast to periodic severe flooding ob-
served in floodplains of the western
Great Plains, riparian systems at
higher elevations are not subject to
severe overbank flooding. During a

study of riparian avifauna at SPR,
annual spring flooding varied tre-
mendously (Knopf and Sedgwick
1988). Maximum mean daily flow in
1982 was 44 m^/ sec, compared to 405
m^/ sec in 1983, when all of the ripar-
ian zone, as well as portions of adja-
cent upland habitat were flooded. No
overbank flooding occurred in 1982.
Habitats of small mammals in lower
riparian systems are periodically
subjected to total inundation for vari-
able amounts of time. Those habitats
appear to be too unstable to assure
prolonged survival by species popu-
lations, and are recolonized by indi-
viduals from the uplands following
each perturbation.

Changes in small mammal com-
munities among upland sites were
less pronounced. Faunal similarity
was greatest at the intermediate sites,
especially LPC (1909 m), SC (2293 m)
and IR (2500 m). The mean of PS val-
ues comparing those sites was 0.57 +
0.12, compared to the overall mean
of 0.29 + 0.06. Deer mice were a
dominant species at all sites but SPR
(sand sagebrush mixed-prairie) and
LR (aspen). The distribution of other
species appeared to be influenced by
changes in upland vegetation types
along the altitudinal gradient. For
example, northern grasshopper mice
were relatively abundant at the low-
est site, which contained grassland
areas. Boreal redback voles (Clethri-
onomys gapperi) were similarly abun-
dant at the highest site in spruce-fir.
Neither species was trapped else-
where. Honeycutt et al. (1981) also
reported that the distribution of
some species along an altitudinal
gradient in Utah was strongly influ-
enced by type of vegetation. We
(Knopf and Olson 1984) have noticed
regional differences in small mam-
mal communities in riparian zones of
similar woody communities but dif-
ferent herbaceous composition that
can be attributed to variations in site
dryness.

Beta diversity was low (high CC
values) at elevations of less than 2500
m (SPR, LPC, and SC), indicating

that small mammal communities in
riparian and adjacent upland sites
were quite similar. At 2500 m (IR),
the CC value declined to 0 (no spe-
cies common to both sites), then re-
mained low at the higher study areas
that contained aspen and spruce-fir
uplands. With the exception of an
extremely low value at SPR (caused
by the overwhelming dominance of
deer mice in the riparian site), PS val-
ues followed the same pattern. Thus,
within the Platte River watershed,
beta diversity between riparian and
upland small mammal communities
is much greater at the upper end of
the altitudinal cline.

These results differ from the
avifaunal studies of Knopf (1985)
who found beta diversity between
riparian and upland sites to be great-
est at the higher and lower ends of
the watershed, and upland/ riparian
assemblages to be similar at interme-
diate study areas. Also in contrast to
Knopf's (1985) findings were greater
relative diversity in, and faunal simi-
larity among, upland communities.
In support of the avian study conclu-
sions, however, riparian sites at the
higher elevations contributed sub-
stantially to small mammal beta and
gamma (regional) diversity.

Implications to Conservation

Historically, management of riparian
zones has occurred primarily on ar-
eas at lower elevations. Management
that is concentrated in a limited num-
ber of habitats or at selected eleva-
tions may result in higher local (al-
pha) diversity at the expense of beta
and gamma (regional) diversity
(Sannson and Knopf 1982). Despite
different beta diversity patterns, our
findings support the conclusion by
Knopf (1985) that greater emphasis
needs to be placed upon conserva-
tion of riparian communities at
higher elevations regionally.

Knopf et al. (1988) have recom-
mended that agencies develop guide-
lines for region wide rather than local

384

management of riparian systems. Re-
spective agencies should realize that
small mammal communities at
higher elevations contribute more to
regional diversity than those at lovs^er
elevations. In order to conserve re-
gional integrity in native small mam-
mal faunas, land uses allowed in,
and adjacent to, high elevation ripar-
ian zones should be critiqued as care-
fully as those in lowland floodplains.
For example, livestock grazing can
affect structure of small mannmal as-
sociations by reducing understory
vegetation (Moulton et al. 1981).
Grazing and other activities that po-
tentially reduce understory vegeta-
tion in higher elevation riparian
zones can seriously affect abun-
dances of certain species such as
shrews that are not present at lower
sites. The consequences to regional
diversity of small mammals would
be greater than livestock grazing at
lower elevations because our find-
ings suggest that: (1) higher elevation
(above 2500 m) sites contribute more
to regional diversity of small mam-
mals; and (2) small mammal commu-
nities in some lower elevation ripar-
ian zones are composed mostly of
species populations of ecological gen-
eralists that are regulated by cata-
strophic, natural perturbations.

Acknowledgments

We thank Richard W. Cannon, John
F. Ellis, and Elizabeth A. Ernst for
field and analytical assistance. Ron
Desilet assisted in locating field sites.
Eugene C. Patten of Arapaho Na-
tional Wildlife Refuge and Marvin
Gardner of the South Platte Wildlife
Management Area granted access to
the IR and SPR sites, respectively.
The U.S. Forest Service and Colorado
State University allowed us to work
within their holdings. This research
is a product of Cooperative Agree-
ment 2463-4 between the Colorado
Division of Wildlife and U.S. Fish
and Wildlife Service. This manu-
script was improved by the com-

ments of Steven J. Bissell, Michael A.
Bogan, Lawrence E. Hunt, Mel
Schamberger, Robert C. Szaro, Don
E. Wilson, and an anonymous
reviewer.

Literature Cited

Anderson, Douglas C, James A.
MacMahon, and Michael L. Wolfe.
1980. Herbivorous mammals
along a montane sere: community
structure and energetics. Journal
of Mammalogy 61:500-519.

Armstrong, David M., Benjamin H.
Banta, and Edward J. Pokropus.
1973. AlHtudinal distribution of
small mammals along a cross-sec-
tional transect through the Arkan-
sas River Watershed, Colorado.
Southwestern Naturalist 17:315-
326.

Bennett, Lisa. 1984. The initial effects
of cattle introduction on a riparian
small mammal community, north-
eastern Colorado. M.S. Thesis,
Colorado State University, Fort
Collins.

Cannon, Richard W. and Fritz L.
Knopf. 1984. Sp)ecies composition
of a willow community relative to
seasonal grazing histories in Colo-
rado. Southwestern Naturalist
29:234-237.

Honeycutt, Rodney L., Michael P.
Moulton, Jerry R. Roppe, and
Lynn Fifield. 1981. The influence
of topography and vegetation on
the distribution of small mammals
in southwestern Utah. Southwest-
ern Naturalist 26:295-300.

Horton, Jerome S. 1977. The develop-
ment and perpetuation of the per-
manent tamarix type in the phrea-
tophyte zone of the Southwest, p.
124-127. In Importance, preserva-
tion and management of riparian
habitat: a symposium. USDA For-
est Service General Technical Re-
port RM-43. Rocky Mountain For-
est and Range Experiment Station,
Fort Collins, Colo.

Huntley, Nancy, and Richard S.
Inouye. 1987. Small mammal

populations of an old-field
chronosequence: successional pat-
terns and associations with vege-
tation. Journal of Mammalogy
68:739-745.

Kirkland, Gordon L. 1981. The zoo-
geography of the mammals of the
Uinta Mountains regions. South-
western NaturaHst 26:325-339.

Knopf, Fritz L. 1985. Significance of
riparian vegetation to breeding
birds across an altitudinal cline. p.
105-111. In Riparian ecosystems
and their management: reconciling
conflicting uses. First North
American Riparian Conference.
USDA Forest Service General
Technical Report RM-120, 523 p.
Rocky Mountain Forest and Range
Experiment Station, Fort Collins,
Colo.

Knopf, Fritz L., R. Roy Johnson, Ter-
rell Rich, Fred B. Samson, and
Robert C. Szaro. 1988. Conserva-
tion of riparian ecosystems in the
United States. Wilson Bulletin
100:278-284.

Knopf, Fritz L., and Thomas E.
Olson. 1984. Naturalization of
Russian-olive: implications to
Rocky Mountain wildlife. Wildlife
Society Bulletin 12:289-298.

Knopf, Fritz L., and James A.

Sedgwick. 1988. Latent population
responses of summer birds to a
catastrophic climatological event.
Condor 89:869-873.

M'Closkey, Robert T. 1981. Mi-
crohabitat use in coexisting desert
rodents — the role of population
density. Oecologia 50:310-315.

Moulton, Michael P., Jerry R. Choate,
and Steven J. Bissell. 1987. Small
mammals on revegetated agricul-
tural land in eastern Colorado.
Prairie Naturalist 13:99-104.

Moulton, Michael P., Jerry R. Choate,
Steven J. Bissell, and Robert A.
Nicholson. 1981. Associations of
small mammals on the central
high plains of eastern Colorado.
Southwestern Naturalist 26:53-57.

Olson, Thomas E., and Fritz L.
Knopf. 1986. Naturalization of
Russian-olive in the western

385

United States. Western Journal of
Applied Forestry 1:65-69.

Samson, Fred B., and Fritz L. Knopf.
1982. In search of a diversity ethic
for wildlife management. Transac-
tions of the North American Wild-
life and Natural Resources Confer-
ence 47:421-431.

Szaro, Robert C, and Scott C. Belfit.
1987. Small mammal use of a des-
ert riparian island and its adjacent
scrub habitat. USD A Forest Serv-
ice Research Note RM-473, 5 p.
Rocky Mountain Forest and Range
Experiment Station, Fort Collins,
Colo.

Thomas, Jack Ward, Chris Maser,
and Jon E. Rodiek. 1979. Wildlife
habitats in managed rangelands —
the Great Basin in southeastern
Oregon. Riparian zones. USDA
Forest Service General Technical
Report PNW-80, 18 p. Pacific
Northwest Forest and Range Ex-
periment Station, Portland, Ore.

Whittaker, Robert H. 1975. Commu-
nities and ecosystems. MacMillan
Co., New York, 385 p.

Estimated Carrying Capacity
for Cattle Competing with
Prairie Dogs and Forage
Utilization In Western South
Dakota^

Daniel W. Uresk and Deborah D. Paulson^

On the Great Plains, black-tailed
prairie dogs (Cynomys ludovidanus)
compete with livestock for forage
and have been a major concern
among livestock producers since the
late 1800's (Merriam 1902). For live-
stock producers, increased cattle-car-
rying capacity on range land is the
primary objective of large-scale prai-
rie dog control programs (Collins et
al. 1984). However, carrying capaci-
ties for cattle have not been fully
evaluated comparing effects in the
presence versus the absence of prai-
rie dogs. Carrying capacities for
cattle competing with prairie dogs
for forage have historically been de-
termined by estimating standing
crop of herbage and then arriving at
range condition and estimated carry-
ing capacity. Information on diets of
cattle and prairie dogs, consumption
rates, production of forage, and prai-
rie dog densities has never been col-
lectively evaluated to determine car-
rying capacities on rangelands sup-
porting both cattle and prairie dogs.

'Paper presented at symposium. Man-
agement of Amphibians, Reptiles, and
Small Mammals in Northi America. (Flag-
staff , AZ, July 19-21, 1988.)

'Daniel W. Uresk and Debar att D. Paul-
son are Research) Biologist and Wildlife Bi-
ologist, respectively, at the Rocky Mountain
Forest and Range Experiment Station's Re-
search Work Unit in Rapid City, SD 57701, in
cooperation with the South Dakota School
of Mines. Station headquarters is in Fort
Collins, in cooperation with Colorado State
University.

Abstract.— Carrying capacities for cattle compet-
ir^g with black-tailed prairie dogs (Cynomys ludov'h
cianus)vjeTe estimated by a linear programming
technique for management of cool-season grasses
in western South Dakota. Forage utilization was al-
lowed to range from 20% to 80%. Under manage-
ment for cool-season grasses (western wheatgrass
(Agropyron smifhii): needlegrasses iStipa spp.)),
stocking rotes of cows ranged from 43 to 214 per
hectare over a 6-month grazing season, and cow-
calf stocking rotes ranged from 43 to 214 per hec-
tare over a 6-month grazing season, and cow-calf
stocking rates ranged from 23 to 161 . Needlegrasses
and needleleaf sedge (Corex eleocharis)v/eTe key
forage species.

This study utilized a linear pro-
gramming approach (GOAL) to de-
termine carrying capacities of cattle
as limited by prairie dog town sizes
and forage utilization while still
maintaining pastures in a near climax
stage of mixed perennial cool-season
grasses. Cool-season grasses in-
cluded western wheatgrass (Agropy-
ron smithii) and needlegrasses (Stipa
spp.).

Study Area and Methods

The study was conducted in Conata
Basin, approximately 29 km south of
Wall, S. Dak. Average annual pre-
cipitation at the Cedar Pass Visitor
Center, Badlands National Park, ap-
proximately 21 km east of the study
area, is 39.7 cm, of which 79% falls
from April through September. Av-
erage annual temperature is 10°C.
Effective forage-year (October 1 to
September 30) precipitation for plant
growth was 46.3 cm.

Major graminoids of the study
area included blue grama (Bouteloua
gracilis), buffalograss (Buchloe dacty-
loides), western wheatgrass,
needleleaf sedge (Carex eleocharis),
and red threeawn (Aristida longiseta).
Common forbs were scarlet
globemallow (Sphaeralcea coccinea),
Patagonia Indianwheat (Plantago pat-
agonica), and prairie dogweed (Dysso-
dia papposa). Shrubs were snakeweed
(Xanthocephalum sarothrae) and silver
sagebrush (Artemisia carta).

The area is grazed by cattle and
black-tailed prairie dogs. Prairie dogs
graze within towns and were active
throughout most of the year. Cattle
grazed the entire area from approxi-
mately mid-May to the last of Octo-
ber. Stocking levels of cattle varied
from year to year depending upon
moisture levels and available forage.

We applied the GOAL computer
program to a resource decision prob-
lem using data from a 2,100-ha pas-
ture following similar procedures by
Bartlett et al. (1976), Bottoms and
Bartlett (1975), and Connolly (1974).
Basic data collected on or near the
pasture included cattle diet composi-
tion (Uresk 1986), black-tailed prairie
dog diet composition (Uresk 1984),
prairie dog densities (Cincotta 1985),
and forage production (Uresk 1985).
Forage consumption of a cow and
cow-calf unit was estimated as 355
kg/month [1 AUM (Animal Unit
Month)], and 485 kg/month (1.32
AUM), respecrively (USDA 1968).
Forage consumption of a black-tailed
prairie dog over a 12-month period
was estimated at 10.95 kg (Hansen
and Ca vender 1973). Prairie dog den-
sities were estimated as 44 animals/
ha (Cincotta 1985).

Serai stages (table 1) were esti-
mated for the entire pasture, based
on discriminant functions developed
for canopy cover and frequency of
occurrence of major plants. Climax
or near-climax (serai stage A) was
dominated by western wheatgrass;
serai stage B was high in blue grama;

387

while serai stage C was high in buffa-
lograss. Range serai stage D con-
sisted of approximately equal but
smaller amounts of all three plant
species. Estimates of forage produc-
tion and area occupied by prairie dog
towns were specified separately for
each range serai stage in the analysis.

In the analysis, forage utilization
was varied for the entire pasture at
four levels (20%, 40%, 60% and 80%)
when both cattle and prairie dogs
were grazing. Prairie dog towns
were allocated to serai stages B, C,
and D; but not to range condition
class A because prairie dogs do not
occur in or near climax vegetation.
Prairie dogs were confined to areas
that totalled from 20 to 40 ha for the
entire pasture. Forage utilization on
these areas was adjusted to 100%.

Major forage plants of both herbi-
vores included western wheatgrass
blue grama, buffalograss, needleleaf
sedge, sand dropseed ( Sporobolus
cryptandrus), needlegrasses, scarlet
globemallow and categories of other
graminoids, and other forbs (Uresk
1984, Uresk 1986). Shrubs were ex-
cluded because they were minor
components of the diets and range-
land. Average herbivore diets for the
season were used in this linear pro-
gramming analysis.

With linear programming, man-
agement options for amounts of for-
age utilization and area occupied by
prairie dog towns were analyzed un-
der management for cool-season
grasses. Under management for cool-
season grasses, no forage species was
utilized over the selected percent-
ages.

For the GOAL programming
analysis, the following assumptions
were made:

1. Adequate forage of major
plant species were available
within limits of prescribed
utilization so that herbivores
did not adjust their normal
diets and consumption in re-
sponse to a decrease in for-
age.

2. Common use of the range by
the two herbivores did not
alter the preference for for-
age within established utili-
zation limits.

3. Forage consumption was
proportional to population
densities of the herbivore
species.

Cattle stocking numbers were esti-
mated as follows. Diet composition
and forage consumption rates of both
herbivores were specified and held
constant. Forage availability was
specified for each species by serai
stage and held constant. Prairie dog
density per hectare of town was
specified and held constant. The
management variables — percent for-
age utilization and hectares in dog
towns — were varied within specified
limits.

Finally, the GOAL program solved
cattle-stocking numbers that could be
supported by the available forage for
a given forage utilization {percentage
and hectares in prairie dog towns.
When present, prairie dogs were
given first priority for forage.

Results

Plant Production

Forage production for individual
species was greatest for western
wheatgrass, followed by buffalograss
and blue grama (table 1). The pasture
at or near climax serai stage (A) had
the lowest plant production (1970
kg/ha); serai stage C had the greatest
overall production (2267 kg/ha).
Most of the pasture was at or near
cHmax serai stage A (58%) and did
not have prairie dogs, a factor that
results in a relatively low impact by
prairie dogs. Serai stages B, C, and D
made up 3%, 7%, and 32%, respec-
tively, of the pasture. All had prairie
dogs residing.

Carrying Capacity

Carrying capacity for mature cows
without calves (6-month grazing pe-
riod) on range with no prairie dogs
competing ranged from 55 to 221
cows/2100 ha when forage utiliza-
tion levels were from 20% to 80%

Table 1.— Estimated peak plant production (kg/ha) (Uresk 1985) by range
class on a 2, 1 00-ha pasture.

Range serai stages' (ha)

A

B

c

D

Plant taxa

(1226)

(55)

(144)

(675)

Western wheatgrass (Agropyron smifhii)

1354

514

72

301

Blue grama (Boufeloua gracilis)

204

441

396

88

Buffalograss (BucNoe dacfyloides)

47

601

1172

192

Needleleaf sedge (Carex eleocharis)

9

12

43

38

Needlegrasses (Sf/pa spp.)

32

44

0

55

Sand dropseed (Sporobolus cryptandrus)

0

1

5

48

Other graminoids

138

180

253

372

Total graminoids

1784

1793

1941

1094

Scarlet globemallow (Sphaera/ceo cocclnea)

36

36

47

96

Total forbs

150

388

279

1046

Total production^

1970

2217

2267

2236

'A = climax; D = low serai stage. Uresk, D. W. submitted. A quantitative methiod for
estimating ecological stages in a mixed-grass prairie witt) multivariate tectiniques. J.
Range t^anage.

'Shrubs ore not included in production estimates.

388

(table 2). In estimating stocking rates,
no single forage species was allowed
to be utilized at levels greater than
the set levels from 20% to 80%. Thus,
a range of 1.6 to 6.4 ha/AUM was
required. Numbers of cows de-
creased as hectares of prairie dog
towns increased; stocking rates de-
creased by approximately 3 for every
additional 20 ha of prairie dogs (880
prairie dogs or 293 prairie dogs/
cow) up to 40 ha on the pasture.

Cow-calf stocking rates ranged
from 40/2,100 ha to 161 (1 cow-calf
unit = 7.92 AUMs for 6-months)
when utilization levels varied from
20% to 80% without prairie dogs
(table 2). At these stocking rates, ap-
proximately 2.1 to 8.7 ha were re-
quired for each AUM. Stocking rates
decreased by approximately 2 cow-
calf units for every additional 20 ha
of prairie dogs.

Discussion

Needlegrasses and needleleaf sedge
limited carrying capacity for cattle on
pastures managed for cool-season
grasses. Western wheatgrass was
never a limiting species; that is, con-
sumption of western wheatgrass by
both herbivores never exceeded the
amount available. The 80% level of

utilization of some cool-season
grasses is too high to maintain the
viability of these plants, and lower
utilization levels (30-45%) are recom-
mended (Lewis et al. 1956). With
fewer cattle grazing under manage-
ment for cool-season grasses, cattle
gain more weight per day, but fewer
kilograms per hectare (Black et al.
1937, Lewis et al. 1956, Bement 1969).

Prairie dog expansion can be re-
duced under management for cool-
season grasses because vertical cover
and grass heights increase (Cincotta
1985). Prairie dogs did not signifi-
cantly expand over a 4-year period
on areas where cattle were excluded
(Uresk et al. 1982). Furthermore, a
lower stocking rate (management for
cool-season grasses) would increase
vertical grass cover on the range and
would thereby further reduce prairie
dog expansion. Snell and Hlavachick
(1980) and Snell (1985) reported re-
duced expansion rates and elinuna-
tion of prairie dog colonies by using
a summer-deferred grazing system.
Prairie dogs prefer habitat managed
for warm-season grasses [blue grama
(Bouteloua gracilis), buffalograss
(Buchloe dactyloides)]. Increased stock-
ing rates of cattle and shortgrass stat-
ure with low vertical cover allows for
prairie dog expansion (Uresk et al.
1982, Cincotta 1985).

r

Table 2. —Estimated 6-month carrying capacity for mature cows with and
without calves with management for cool-season grasses. Stocking rates
are related to tiectares of prairie dogs and allowable forage utilization on a
2,100-ha pasture In westem South Dakota. Consumption of needlegrasses
and needleleaf sedge Is 1 00% on prairie dog occupied areas.

Forage

Prairie dogs occupied areas (ha)

utilization %

0

20

40

0

20 40

Cow numbers^

Cow-calf numbers'

20

55

53

50

40

39 37

40

110

108

105

81

79 77

60

166

163

160

121

119 117

80

221

218

214

161

159 157

'355 kg of forage consumed/cow/monfh (1 AUM).

'485 kg of forage cor\sumed/cow-calf/monfh (1.32 AUM).

Cattle stocking rates estimated in
this study were conservative, be-
cause upper limits of forage con-
sumption and prairie dog densities
(44 animals/ ha) were used in the
analyses. The guidelines reported
here for cow or cow-calf stocking
rates for cool-season grasses repre-
sent viable options for management.
Key forage species used to estimate
cattle numbers and monitor utiliza-
tion for management of cool-season
grasses included needlegrasses and
needleleaf sedge. Generally, stocking
rates were limited by production and
use of needlegrasses, although
needleleaf sedge and sand dropseed
also influenced cow numbers. When
hectares of prairie dogs are high,
needleleaf sedge can become the ma-
jor limiting factor in determining
cow numbers. Needlegrasses were
generally the limiting plant compo-
nent in determining cow-calf units.
Sand dropseed can be limiting when
the area with prairie dogs is greater
than or equal to 200 ha.

This study only presents estimates
for up to 40 ha of prairie dog colonies
(approximate current levels of prairie
dogs) on a 2,100-ha pasture, and lim-
ited extrapolation is suggested be-
yond data in table 2. An additional
constraint is availability of needle-
grasses and needleleaf sedge. Ex-
trapolation of results to pastures
with lower availability of these spe-
cies should be done cautiously. In
fact, where forage availability and
composition are much different from
the pasture studies, extreme care
should be used in extrapolating re-
sults to other areas. The assumptions
and required constraints for GOAL
linear program analysis imposes
some limitations on biological sensi-
tivity.

Acknowledgments

Appreciation is extended to Ne-
braska National Forest for providing
study areas. Partial funding of this
study was provided by National Ag-

389

ricultural Pesticide Impact Assess-
ment Program (NAPIAP) and Ne-
braska National Forest.

Literature Cited

Bartlett, E. Thomas, Kenneth E. Bot-
toms, and Randal P. Pope. 1976.
GOAL-multiple objective pro-
gramming. Colorado State Univer-
sity, Ft. Collins, CO. 157 p.

Bement, R. E. 1969. A stocking-rate
guide for beef production on blue
grama range. J. Range Manage.
22:83-86.

Black, W. H., A. L. Baker, V. I. Clark,
and O. R. Mathews. 1937. Effect of
different methods of grazing on
native vegetation and gains of
steers in northern Great Plains.
U.S. Dept. Agr. Forest Service,
Rapid City, SD Tech. Bull. 547. 18

P-

Bottoms, Kenneth E., and E. T.
Bartlett. 1975. Resource allocation
through GOAL programming. J.
Range Manage. 28:442-447.

Cincotta, Richard P. 1985. Habitat
and dispersal of black-tailed prai-
rie dogs in Badlands National
Park. Ph.D. Diss. Colo. State
Univ., Ft. Collins, CO.

Collins, Alan R., John P. Workman,
and Daniel W. Uresk. 1984. An
economic analysis of black-tailed
prairie dog (Cynomys ludovicianus)
control. J. Range Manage. 35:358-
361.

Connolly, J. 1974. Linear program-
ming and the optimum carrying
capacity of range under connmon
use. J. Agric. Sci. 83:259-265.

Hansen, Richard M., and Barbara R.
Ca vender. 1973. Food intake and
digestion by black-tailed prairie
dogs under laboratory conditions.
Acta Theologica 18:191-200.

Lewis, James K., George M. Van
Dyne, Leslie R. Albee, and Frank
W. Whetzal. 1956. Intensity of
grazing: its effect on livestock and
forage production. South Dakota
Agr. Expt. Sta. Brookings Bull.
459. 44 p.

Merriam, C. Hart. 1902. The prairie
dog of the Great Plains, p. 257-270.
In Yearbook of the United States
Department of Agriculture, Wash-
ington, DC. U.S. Gov. Print. Off.

Snell, Glen P. 1985. Results of control
of prairie dogs. Rangelands 7:30.

Snell, Glen P., and Bill D. Hlavachick.
1980. Control of prairie dogs — the
easy way. Rangelands 2:239-240.

Uresk, Daniel W. 1984. Black-tailed
prairie dog food habits and forage
relationships in western South
Dakota. J. Range Manage. 37:325-
329.

Uresk, Daniel W. 1985. Effects of con-
trolling black-tailed prairie dogs
on plant production. J. Range
Manage. 38:466-468.

Uresk, Daniel W. 1986. Food habits
of cattle on the South Dakota High
Plains. Prairie Nat. 18:211-218.

Uresk, Daniel W., James G. Mac-
Cracken, and Ardell J. Bjugstad.
1982. Prairie dog density and
cattle grazing relationships, p. 199-
201. In R. M. Timm and R. J.
Johnson, eds. Fifth Great Plains
Wildlife Damage Control Work-
shop Proc. October 13-15, 1981.
Univ. Neb., Lincoln.

USDA Forest Service. 1968. Range
Environmental Analysis Hand-
book. U.S. Dept. of Agr. Forest
Service Handbook 2209.21. 244 p.

390

Cattle Grazing and Small
Mammals on the Sheldon
National Wildlife Refuge,
Nevada^

John L. Oldemeyer^ and Lydia R. Allen-
Johnson^

Abstract.— We studied effects of cattle grazing on
small mamnnal microhabitat and abundance in
northwestern Nevada. Abundance, diversity, and
microhabitat were compared between a 375-ha
cattle exclosure and a deferred-rotation grazing al-
lotment which hod a three-year history of light to
moderate use. No consistent differences were found
in abundance, diversity, or microhabitat between
the two areas.

Grazing by livestock is a common
and economically important practice
throughout much of the western
United States. Because grazing alters
wildlife habitat, much attention has
centered on its impact on wildlife
abundance, diversity, and habitat
use. However, relatively little infor-
mation exists on effects of grazing on
small mammal communities. Such
information would aid development
of effective grazing progran\s where
small mammals are a management
concern.

Several authors have demon-
strated that removal or alteration of
cover can cause changes in small
mammal communities (Bimey et al.
1976, Geier and Best 1980, Grant et
al. 1982, LoBue and Darnell 1959).
More specifically, grazing altered ro-
dent species diversity through
changes in plant species diversity on
several habitats in northeastern Cali-
fornia (Hanley and Page 1982). Simi-
larly, Grant et al. (1982) found differ-
ential changes in several small mam-
mal community parameters between

'Paper presented at symposium. Man-
agement of Amphibians, Reptiles, and
Small Mammals in Nortt) America. (Flag-
staff, AZ, July 19-21. 1988.)

'Leader, Ecology and Sysfematics Sec-
tion, National Ecology Research) Center.
U.S. Rsh and Wildlife Sen/ice, 1300 Blue
Spruce Drive, Fort Collins, CO 80524.

^Biological technician. Ecology and Sys-
fematics Section, National Ecology Re-
search Center. U.S. Hsh and Wildlife Sen/ice,
1300 Blue Spruce Drive. Fort Collins, CO
80524.

grazed and ungrazed sites in four
western grassland communities; tall-
grass and montane grasslands ap-
peared to be most affected by graz-
ing.

In assessing grazing impacts on
small mammal communities, Hanley
and Page (1982) stressed the impor-
tance of evaluating effects on a habi-
tat-type basis. Grant et al. (1982) con-
cluded that the response of a small
mammal community to grazing de-
pended on the site and the original
mammal species composition.

In 1980, the Sheldon National
Wildlife Refuge (SNWR) initiated a
deferred-rotation grazing system on
the 6,954-ha Badger Mountain graz-
ing allotment to improve soil and
range conditions. The management
plan was designed to graze 1,444 ani-
mal-unit-months (AUMs) with the
grazing period alternating between
mid-June through early August dur-
ing one year and early August
through late October the next (five-
year average, David Franzen, Range
Conservationist, SNWR, pers.
comm.). Prior to 1979, the allotment
had been on a season-long grazing
system from early April through Sep-
tember with an estimated 1,700
AUMs being removed from the unit
(U.S. Fish and Wildlife Service 1980).

In Spring 1981, we constructed a
375-ha cattle exclosure on the Badger
Mountain allotment to evaluate the
effects of cattle grazing on wildlife
and their habitat (Oldemeyer et al.
1983). The purpose of this element of

the study was to evaluate the effect
of the grazing system on small mam-
mals. Specifically, we wanted to de-
termine the following: (1) is there a
difference in small mammal abun-
dance and diversity between the ar-
eas over time, (2) is there a difference
in the available small mammal habi-
tat between areas, and (3) what mi-
crohabitat characteristics are indica-
tive of capture sites by individual
small mammal species for the two
ecosites? We tested the null hypothe-
sis of no significant difference be-
tween the exclosure and the allot-
ment.

Study Area and Mettiods

The Badger Mountain allotment
ranges from 1,890-2,152 m elevation
and is composed of two dominant
range ecosites (Anderson 1978). The
shrubby rolling hills (SRH) ecosite
occurs on moderate to deep soils and
is dominated by big sagebrush (Ar-
temisia tridentata) and antelope
bitterbrush (Purshia tridentata) with
grass understory dominated by
Idaho fescue (Festuca idahoensis). The
mahogany rockland (MR) ecosite oc-
curs on rocky ridges and slopes with
bedrock outcrops. Curlleaf mountain
mahogany (Cercocarpus ledifolius) is
predominate in this ecosite with a
grass understory dominated by west-
ern needlegrass (Stipa occidentalis)
(fig. 1). Precipitation on Badger
Mountain ranges from 27-33 cm an-

391

nually with most coming as snow
and as spring and autumn rains (U.S.
Fish and Wildlife Service 1980).

We conducted the study during
the summers of 1983 and 1984, four
and five years, respectively, after ini-
tiation of the deferred-rotation graz-
ing system. Grazing intensities were
1,650 AUMs from 10 July to 10 Au-
gust, 1980, 1,770 AUMs from 7 Au-
gust to 30 September, 1981, and 1,036
AUMs from 24 June to 22 August,
1982. In 1983, cattle were grazed on
the allotment from 1 August through
15 October at a rate of 980 AUMs.
The following year, the unit sup-
ported 1,337 AUMs during a 28 June
to 18 August grazing period (David
Franzen, Range Conservationist,
SNWR, pers. comm.).

In 1983, eight live trap grids were
established with trap stations 15 m
apart. Four grids were located inside
the exclosure and four were located
in the allotment. We arranged each 7
X 7 grid so that approximately half of
the traps were in the SRH ecosite and
half were in the MR ecosite. We
sampled only four grids (two in the
exclosure and two in the allotment)
in 1984, but we increased the size of
the grids to 64 (8 X 8) trap stations.

We trapped from 1 July through
11 August in 1983, and 19 June
through 1 July in 1984. Only one pair
of grids were trapp>ed at a time (one
grid in the exclosure, one in the allot-
ment), for a total of four trap sessions
in 1983, and two sessions in 1984. A
Sherman live trap containing a hand-
ful of cotton wool and baited with
rolled oats was placed at each sta-
tion. Trapping began in the afternoon
and continued for five consecutive
days. Traps were opened each day
between 1600-1730 hrs and closed the
following morning between 0730-
11(X) hrs to prevent daytime trap
mortality. Species, trap number, age
(adult or juvenile), sex, weight and
tag number were recorded. We used
toe clips or aluminum ear tags to
identify individuals.

We estimated relative abundance
of small mammals as the total num-

ber of individuals captured per trap
night (catch /effort) for each ecosite
type, area and grid. Abundance was
calculated for all small mammals as
well as for each individual sp>ecies.

Small mammal diversity was de-
rived for each area using Patil and
Taillie's (1979) diversity profiles. This
is a graphic ordering of the diversity
of two or more communities. The y-
axis represents the percent of small
mammals remaining in the sampled
population when a species is re-
moved. This is plotted against the
number of species that have been
removed from the sampled popula-
tion, with species removal being cu-
mulative.

The profile of an intrinsically more
diverse community will plot above
that of a less diverse community. If
profile lines intersect, then the com-
munities do not differ in diversity.

Vegetation measurements describ-
ing microhabitat structure were
taken at each station prior to trap-
ping. The characteristics we meas-
ured are similar to those reported in
other small mammal studies (e.g.
Geier and Best 1980, Hallett 1982).
These included:

1. Percent canopy cover of
grass, forbs, and litter (all
downed dead material; e.g.
twigs, dead grass, leaves) in
a 1.0 X 0.5 m quadrat having
the trap station stake as its
center;

2. Height (cm) of the nearest
shrub (crown foliage >2 dm
in diameter) in each quarter
around the trap station stake;

3. Line intercept distance (cm)
of living and dead shrubs (in
the 25 to 50 cm layer above
the ground) occurring within
two perpendicularly oriented
2-m transects centered at the
trap station stake.

Five microhabitat variables were
derived from these measurements
for analysis. These included: (1) %
forb cover, (2) % grass cover, (3) %
litter cover, (4) total shrub intercep-
tion distance (cm), and (5) mean
height (cm) of the live shrubs around
each stake.

Small mammal abundance data
were analyzed using a three-way
analysis of variance to determine if

Figure 1 .—View from the study site on Badger Mountain, Sheldon National Wildlife Refuge.
Nevada.

392

small mammal abundance differed
between areas, years, and ecosites.
We used a one-way analysis of vari-
ance test to detect differences be-
tween areas for individual years and
ecosites. To determine the microhabi-
tat preferences of individual species
we coded trap locations as being ei-
ther capture or non-capture stations.
We employed a nested two-way
analysis of variance to test these pref-
erences among areas and codes, the
interaction of areas by codes, and the
nested interaction of grids within ar-
eas. We considered P<=0.1 to be sig-
nificant. Subsequent discussion of
small mammal microhabitat selection
concerns only the two most abun-
dant species, the deer mouse (Pero-
myscus maniculatus) and the least
chipmunk (Tamias minimus).

Results and Discussion

Species Composition

Species of small mammals occurring
in the two ecosites of our study area
are widely distributed throughout

the Great Basin (Hall 1946). These
species and their percentage of the
total catch were: deer mouse (46.7%),
least chipmunk (29.8%), Great Basin
pocket mouse (Perognathus parvus)
(12.3%), sagebrush vole (Lagurus cur-
tatus) (7.8%), Townsend's ground
squirrel (Spermophilus townsendii)
(1.2%), golden-mantled ground
squirrel (Spermophilus lateralis)
(1.2%), and long-tailed vole (Microtus
longicaudis) (0.6%).

Abundance

Total relative abundance of small
mammals did not differ between
year or area (table 1). However, more
animals were captured in the SRH
ecosite than in the MR ecosite
(P=0.05).

There was a general decline in
deer mouse (P=0.08) and least chip-
munk (P=0.06) abundance from 1983
to 1984, although this probably re-
flects the difference in season and
length of trapping between the two
years. We found no significant differ-
ence in abundance for these two spe-

cies between areas or ecosites. This is
not surprising given the opportunis-
tic, adaptable, nature of these small
mammals. Others have found that
heavy grazing in big sagebrush habi-
tat appears to promote an increase in
deer mice (Black and Frischknecht
1971, Larrison and Johnson 1973),
and least chipmunk numbers (Larri-
son and Johnson 1973). Hanley and
Page (1982) observed a different re-
sponse for the two species on their
big sagebrush-Idaho fescue site 60-80
km west of Badger Mountain. In that
study, deer mice were captured in
the same numbers in both grazed
and ungrazed sites, while least chip-
munks were four times more abun-
dant in the grazed site than in the
ungrazed site.

Great Basin pocket mice were
more abundant (P<0.01) in 1983 than
1984, and they were more commonly
captured in the SRH ecosite than in
the MR ecosite (P=0.08). However,
there was no significant difference in
abundance between the areas. Others
have found Great Basin pocket mice
to be more abundant on ungrazed
big sagebrush sites (Black and Fris-

Table 1 .—Abundance of small mammals (number caught per trap night) by year, area and ecosite on the Sheldon
National Wildlife Refuge, 1 983-84.

Species

Area

Shrubby-Rolling Hills

Mohogany Rocklands

1983
#/trapnite(S.E.)

1984
#Arapnlte(S.E.)

1983
#/trapnite(S.E.)

1984
#/trapnlte(S.E)

Deer mouse

Excl.

0.081(0.012)

0.047(0.014)

0.061(0.019)

0.050(0.011)

Allot.

0.063(0.005)

0.052(0.017)

0.064(0.027)

0.016(0.002)

Least chipmunk

Excl.

0.029(0.007)

0.020(0.005)

0.067(0.017)

0.037(0.006)

Allot.

0.046(0.025)

0.031(0.010)

0.049(0.011)

0.005(0.005)

Great Basin

Excl.

0.013(0.005)

0.028(0.003)

0.003(0.003)

0.029(0.014)

pocket mouse

Allot.

0.011(0.007)

0.031(0.004)

0.005(0.003)

0.014(0,006)

Sagebrush vole

Excl.

0.019(0.010)

0.038(0.020)

0.002(0.002)

0.004(0.004)

Allot.

0.004(0.003)

0.019(0.004)

0

0

Long-tailed vole

Excl.

0

0.004(0.004)

0

0

Allot.

0

0.009(0.002)

0

0

Townsend's ground

Excl.

0

0

0

0

squirrel

Allot.

0.005(0.005)

0

0

0.005(0.005)

Golden-mantled

Excl.

0.006(0.006)

0

0.007(0.007)

0

ground squirrel

Allot.

0

0

0

0

Total Catch

Excl.

0.149(0.013)

0.135(0.027)

0.141(0.024)

0.120(0.028)

Allot.

0.154(0.030)

0.143(0.005)

0.119(0.026)

0.042(0.018)

393

chknecht 1971), or more abundant on
grazed sagebrush sites (Hanley and
Page (1982).

Relative abundance of the sage-
brush voles and long-tailed voles
could not be compared statistically
because of the small number of voles
captured. There was, however, a
general trend for microtine rodents
to be more abundant in the SRH
ecosite even though grass and forb
cover in the MR ecosite were higher.
Birney et al. (1976) and Grant et al.
(1982) have discussed the importance
of cover for microtine rodents in
grasslands. Although grass cover
was lower in the SRH ecosite, the
combination of higher litter cover
and shrub intercept in that ecosite
may provide better habitat for these
rodents. The sagebrush vole was
more abundant in the exclosure than
in the allotment. Although we were
unable to test this trend, it is possible
that the sagebrush vole found the ex-
closure, with its slightly greater grass
and shrub cover, to be more inhabit-
able. It is apparent from other studies
that grass and shrub cover are im-
portant components of sagebrush
vole habitat (MacCracken et al. 1985,
Maser et al. 1974, Maser and Strickler
1978, O'Farrell 1972).

Diversity

In 1983, diversity of small mammals
in the exclosure was greater than in
the grazing allotment (fig. 2). Rela-
tive abundance of deer mice, the
most common species (table 1), was
similar in both areas; however, we
caught one more species in the exclo-
sure. In 1984, small mammal diver-
sity was greater in the allotment than
in the exclosure. During that year,
deer mice made up a somewhat
smaller relative proportion of the
small mammal total in the allotment
(table 1); thus the line for the allot-
ment starts higher on figure 2 indi-
cating greater evenness in the per-
centage each species contributed to
the population. We captured one

more species in the allotment than in
the exclosure which extended the tail
of the profile further to the right. Be-
cause of this change from one year to
the next, we were unable to conclude
what impact the grazing system had
on small mammal diversity. Hanley
and Page (1982) observed a higher
diversity index on their ungrazed

sagebrush-Idaho fescue site 60-80 km
west of Badger Mountain.

Vegetation on the Small MamnrKil
Study Area

Generally, the SRH ecosite had lower
grass and litter cover and a greater

I
I

I
I

NUMBER OFSPEC/ES

Figure 2.— Small mammal diversity profiles for the cattle exclosure and ttie allotment, Stiel-
don National Wildlife Refuge, 1983-84. If profile lines intersect, tt>en diversity does not differ
betv/een areas (Patil and Taillie 1 979).

394

I

shrub intercept value than did the
MR ecosite (fig. 3). In the SRH
ecosite, microhabitat characteristics
did not differ between the exclosure
and allotment, except for 1983 when
shrub height in the allotment was
lower (P<0.05) than that in the exclo-

sure. In the MR ecosite, shrub inter-
cept was lower (P<0.03) in the allot-
ment than in the exclosure both years
and grass cover was higher (P<0.10)
in the exclosure in 1983. In both
ecosites, there was a general trend
for cover of both grasses and forbs to

be lower in the allotment than in the
exclosure.

This trend is probably due to the
cattle grazing. However, the fact that
the means are relatively similar (es-
pecially in the SRH ecosite) and do
not differ significantly between areas
indicates that the grazing effect is
within goals established by the ref-
uge.

Microhabitat Characteristics of
Deer Mice Catch Sites

In the SRH ecosite, traps where deer
mice were caught had significantly
greater litter cover (P=0.07 in 1984),
shorter shrubs (P=0.09 in 1984), and
greater shrub intercept (P=0.10 in
1983) than traps where deer mice
were not caught (fig. 4). These pat-
terns tended to hold for both years.

In the MR ecosite, litter cover,
which is greater than in the SRH
ecosite, did not appear to be a signifi-
cant vegetative characteristic (fig. 4).
Grass cover in 1984 was lower
(P=0.06) and shrub height (P=0.02)
and shrub intercept (P=0.08) were
greater at traps where deer mice
were caught than where they were
not caught.

In both the SRH and MR ecosites,
deer mice appeared to use mi-
crohabitat that had greater shrub
intercept. This corresponds with the
findings of Feldhamer (1979) who
noted an increase in deer mouse den-
sity with increased foliage in the
shrub layer. Other studies have
found that deer mice were associated
with light cover in heavily grazed
sites (Black and Frischknecht 1971),
with increasing forb cover (Geier and
Best 1980), or with no measured
habitat variable (Hallett 1982).

70 -1

SRHFCOS/TE

MR FC OS/ ft

i

I

t

m

fW
100
90
80
70
60
50
40
JO
20
W
0

8J 84 8J 84 8J 84
OmSS FO/^ff LIUER

8J 84 8J 84 8J 84
cms FORB LIFTER

8J 84

SHRUB
HF/GHT

8J 84

SHRUB
/NTFRCFRT

SJ 84

SHRUB
HF/GHT

85 84

SHRUB
/HTFRCFPr

M/CmAB/TAT mmBLES

FXCLOSURF

ALLOWFHT

Figure 3.— Microhabitat characteristics arourxJ trap stations in tt>e shrubby-roliing hiils and
nrxjhogany rockiands, Sheidon National Wildlife Refuge. Variables with an "a" denote a P
value of <0.1 between the two areas.

Microhabitat Characteristics of
Chipmunic Catch Sites

In the SRH ecosite, shrub height was
lower (P<0.08 in 1984) in catch loca-
tions in the exclosure and the allot-

395

ment than in non-catch locations.
This pattern held in 1983 (fig. 5).

In the MR ecosite there were no
consistent patterns of chipmunk mi-
crohabitat use (fig. 5). Shrub inter-
ception, in 1984, was greater (P<0.05)
in chipmunk catch locations than
non-catch locations; however this
pattern was not evident in 1983.

Microhabitat selection by the least
chipmunk lacked a consistent pattern
for either ecosite or year. However,
the fact that the least chipmunk is an
opportunistic forager and is the most
widespread of all North American
chipmunks (Hall 1981), suggests that
this rodent adapts rapidly to a vari-
ety of habitat types. Sullivan (1985)
found that the least chipmunk was
associated with a wide variety of
ecological situations in the southwest
and suggested that this species may
be predisposed to exploiting mar-
ginal environments.

Conclusions

These results indicate that the graz-
ing regime initiated on the Badger
Mountain allotment had no discern-
ible impact on the relative abundance
and diversity of small mammals,
four and five years after its implem-
entation. The dominance of two op-
portunistic species on the study area
probably contributed to this lack of
difference. We suggest future moni-
toring of the study area to determine
the long-term response of small
mammals to the grazing program.
Particular attention should be given
to the two vole species which are the
most sensitive to changes in cover.

Acknowledgments

We thank the Sheldon National Wild-
life Refuge personnel for their sup-
port during this project. We appreci-
ate V. Reid's assistance with the de-
sign of this study. J. Sedgwick pro-
vided expertise in the statistical
analysis of the data. And we thank

the following people for their assis-
tance in the field: B. Allen-Johnson, S.
Boyle, C. Halvorson, B. Oldemeyer,
E. Rominger, M. Woodis, and S.

Woodis. This manuscript benefited
by reviews from M. Bogan, D. Fran-
zen, W. Grant, M. Kaschke, B. Keatt,
J. Sedgwick, and K. Severson.

SRHECOS/TE

MRECOS/TE

I

I

8J 84 8J 84 8J 84
cms FO/?B L/TTER

8J 84 8J 84 84 84
GRASS FORB L/J7FR

8J 84
SHRUB
HE/GHT

85 84

SHRUB
/NTFRCEPr

8J 84
SHRUB
Hf/GHr

8J 84

SHRUB
INIFRGEPr

M/CROHAB/TAT MR/ABLES

CAUGHT

HOrCAUGHT

Figure 4.— Microhabitat characteristics around traps where deer mice were captured and
not captured by year and ecosite, Sheldon National Wildlife Refuge. Variables with an "o"
denote a P value of <0.1 between the two areas.

396

Literature Cited

Anderson, W. E. 1978. Range site
handbook for the high desert re-

>,n . SRHfCOS/Tf

source province, Oregon. Un-
paged. Lake Oswego, Oregon.
Birney, Elmer C, William E. Grant,
and Donna Day Baird. 1976. Im-

MR ECO SITE

portance of vegetative cover to
cycles of Microtus populations.
Ecology 57:1043-1051.

Black, Hal L., and Neil C. Fris-
chknecht. 1971. Relative abun-
dance of mice on seeded sage-
brush-grass range in relation to
grazing. U.S. Department of Agri-
culture Forest Service, Resource
Note INT-147, 8p.

Feldhamer, George A. 1979. Vegeta-
tive and edaphic factors affecting
abundance and distribution of
small mammals in southeast Ore-
gon. Great Basin Naturalist 39:207-
218.

Geier, Anthony R., and Louis B. Best.
1980. Habitat selection by small
mammals of riparian communi-
ties: evaluating effects of habitat
alterations. Journal of Wildlife
Management 44:16-24.

Grant, William E., Elmer C. Bimey,
Norman R. French, and David M,
Swift. 1982. Structure and produc-
tivity of grassland small mammal
communities related to grazing
changes in vegetative cover. Jour-
nal of Mammalogy 63:248-260.

Hall, E. Raymond. 1946. Mammals of
Nevada. University of California
Press, Berkeley and Los Angeles.
710 p.

Hall, E. Raymond. 1981. The mam-
mals of North America. Volumes 1
and 2. John Wiley & Sons, New
York. 1181 p.

Hallett, James G. 1982. Habitat selec-
tion and the community matrix of
a desert small mammal fauna.
Ecology 63:1400-1410.

Hanley, Thomas A., and Jerry L.
Page. 1982. Differential effects of
livestock use on habitat structure
and rodent populations in Great
Basin communities. California Fish
and Game 68:160-174.

Larrison, Earl J., and Donald R.
Johnson. 1973. Density changes
and habitat affinities of rodents of
shadscale and sagebrush associa-
tions. Great Basin Naturalist
33:255-264.

LoBue, Joseph, and Rezneat M. Dar-
nell. 1959. Effect of habitat distur-

I

84 8J 84 8J 84 8J 84 8J 84 8J 84

GRASS FORB UTTE/? GRASS FORB L/JTFR

t
I

8J 84
SHRUB
HF/GHr

8J 84
SHRUB
MFRCFPr

8J 84
SHRUB
HF/GHF

85 84
SHRUB
MFRGFPr

M/CmAB/TAT m/ABLES

CAUGHf

HOrCAUGHT

Figure 5.— Microhabitat characteristics around traps where least chipmunks were captured
and not captured by year and ecosite, Sheldon National Wildlife Refuge. Variables with an
"a" denote a P value of <0.1 between the two areas.

397

bance on a small mammal popula-
tion. Journal of Mammalogy
40:425-437.

MacCracken, James G., Daniel W.
Uresk, Richard M. Hansen. 1985.
Rodent vegetation relationships in
southeastern Montana. Northwest
Science 59:272-278.

Maser, Chris, E. Wayne Hammer,
Cheri Bzown, Robert E. Lewis,
Robert 1. Rausch, and Murray L.
Johnson. 1974. The sage vole,
Lagurus curtatus (Cope, 1968), in
the Crooked River National Grass-
land, Jefferson County, Oregon: a
contribution to its life history and
ecology. Saugetierkundliche Mit-
teilungen 22:193-222.

Maser, Chris, and Gerald S. Strickler.
1978. The sage vole, Lagurus cur-
tatus, as an inhabitant of subalpine
sheep fescue, Festuca ovina, com-
munities on Steens Mountain — an
observation and interpretation.
Northwest Science 52:276-284.

O'Farrell, Thomas P. 1972. Ecological
distribution of sagebrush voles,
Lagurus curtatus, in south-central
Washington. Journal of Mam-
malogy 53:632-636.

Oldemeyer, J. L, S. J. Martin, and S.
G. Woodis. 1983. A preliminary
report on the effects of a deferred-
rotation grazing system on wild-
life at the Sheldon National Wild-
life refuge. Cal-Neva Wildlife
Transactions 1983:26-42.

Patil, G. P., and C. Taillie. 1979. A
study of diversity profiles and or-
derings for a bird community in
the vicinity of Colstrip, Montana,
p. 23-48. In Contemporary quanti-
tative ecology and related ecomet-
rics. International Cooperative
Publishing House, Fairland, Mary-
land.

Sullivan, Robert M. 1985. Phyletic,
biogeographic, and ecologic rela-
tionships among montane popula-
tions of least chipmunks (Eutamias
minimus) in the southwest. Sys-
tematic Zoology 34:419-448.

U.S. Fish and Wildhfe Service. 1980.
Sheldon National Wildlife Refuge
renewable resources management
plan-Final EIS. Department of the
Interior, Fish and Wildlife Service,
Region 1, Portland, Oregon.

398

Effect of Seed Size on
Removal by Rodents^

William G. Standley^

Abstract.— Plots located in southeastern Arizona
were seeded with small and large grass seeds. After
3 days, virtually all large seeds were removed by ro-
dents, while small seeds were still present 36 days af-
ter planting. Thus, managers may increase seed sur-
vival in this area, without removing rodents, by seed-
ing with small seeds rather than large seeds.

Seeding is commonly used for restor-
ing depleted vegetation. Many seed-
ing projects fail because rodents eat
the seeds (Bramble and Sharp 1949,
Spencer 1954, Nelson et al. 1970). A
variety of techniques for reducing
the impact of rodents have been
tested, but few have been successful.
Most often resource managers poison
rodent populations before seeding,
but this method is largely unsuccess-
ful because of rapid immigration of
new individuals (Sullivan 1979, Sulli-
van and Sullivan 1984). New meth-
ods of biological management could
be developed that use information
gained from diet and behavior stud-
ies to reduce destruction of seeds by
rodents. Many studies show that cer-
tain rodents prefer particular species
or sizes of seeds (Reynolds and Has-
kell 1949, Reynolds 1950, Abbott
1962, Gashwiler 1967, Smith 1970,
Lockard and Lockard 1971, Smigel
and Rosenzweig 1974, Everett et al.
1978, Price 1983). Thus, whenever
alternative plant species are available
that both meet the resource man-
ager's objectives and have seeds not
preferentially foraged by local seed-

' Paper presented at symposium. Man-
agement of Amphibians. Reptiles, and
Small Mammals in North America. (Flag-
staff . Al. July 19-21. 1988.)

'William G. Standley. formerly a gradu-
ate student. University of Arizona. Arizona
Cooperative Fish and Wildlife Research
Unit, is currently Animal Ecologist. EG&G
Energy Measurements. Inc.. c/o NPR-1. P.O.
Box 127. Tupman. CA. 93276.

eating rodents, seeding could be suc-
cessful even with rodents present.

In southwestern deserts of North
America, where range managers are
attempting to restore rangelands de-
pleted by overgrazing (Cox et al.
1982), kangaroo rats (Dipodomys sp.)
and pocket mice (Perognathus sp. and
Chaetognathus sp.) are some of the
primary seed eaters (Brown et al.
1979b). As early as 1950, Reynolds
suggested that the influence of Mer-
riam's kangaroo rats (Dipodomys mer-
riami) on seeding success depends on
the size of seeds used. Brown et al.
(1979b), Inouye et al. (1980), and
Price (1983) all found that hetero-
myids preyed selectively on large
seeds. In this study, I investigated
the prediction that fewer small seeds
than large seeds would be removed
by rodents in a seeded area in south-
eastern Arizona.

Study Area and Mettiods

The study area was on the USDA
Forest Service Santa Rita Experimen-
tal Range located 45 km south of
Tucson, Pima County AZ, which is
thoroughly described by Martin and
Reynolds (1973). The vegetation was
typical Sonoran desert-scrub, domi-
nated by mesquite (Prosopis juliflora),
burroweed (Haplopappus tenuisectus)
and cholla {Opuntia spp.). Annual
precipitation averages 36 to 43 cm
and is bimodal, with p>eaks in winter
and summer. Plots were seeded fol-

lowing all recommended procedures
(Jordan 1981), using large and small
seeds, both separately and together.
Seeded plots were located within a
slightly sloped 1-hectare area with a
Comoro soil type, at an elevation of
1300 m. I compared the number of
seeds surviving on 4 experimental
plots to the number of seeds surviv-
ing on a control plot which was pro-
tected from rodents.

The study area was prepared by
removing large shrubs by hand and
plowing small plants with a disk. The
control plot was protected from ro-
dents with a rodent-proof fence simi-
lar to that used by Brown et al.
(1979a). All rodents within the exclo-
sure were removed by trapping be-
fore seeding. Each of the 5-15 x 17 m
plots was seeded with 3 evenly
placed pairs of 15 m rows, one pair
for each of 3 treatments which were:
(1) small seeds planted at a rate of
175/m (0.15 g/m), (2) large seeds
planted at a rate of 100/m (4.7 g/m),
and (3) 5 small and large seeds
planted together at 88/m and 50/ m
(0.07g/m and 2.35 g/m), respec-
tively. Seeding rates were chosen ac-
cording to recommended rates for
similar sized seeds (Jordan 1981).
The treatm.ent assigned to each pair
of rows was randomly selected. The
small seeds were blue panicgrass
(Panicum antidotale) which weighed
an average of 0.85 mg each. The large
seeds were barley (Hordeum vulgar e)
which weighed an average of 47.0
mg each. All seeds were planted with

399

a cone seeder at a depth of 1 to 2 cm
on 21 June 1984, just before expected
summer rains. Because blue pan-
icgrass seeds are very small and dif-
ficult to recover from the soil, they
were dyed with water soluble green
food coloring before planting. Barley
seeds were also dyed to avoid a pos-
sible bias.

The species of rodents on the ex-
perimental plots were monitored by
placing 100 live traps at 10 m inter-
vals on and around the plots on the
5th and 6th nights after planting.
Traps were baited with a mixture of
both sizes of seeds and checked at
midnight and sunrise.

The number of seeds surviving on
plots was monitored by collecting
soil samples from the rows immedi-
ately after planting and at 3, 9, 18,
and 36 days after planting. One ran-
dom sample was taken from each
quarter of every row each time.
Samples were not taken from the
outer meter of any row because the
cone seeder applied seeds at a more
variable rate at the beginning and
end of each row.

Soil samples, 2.5 to 3.5 cm deep
and 15 X 25 cm in area, were taken
lengthwise along each row with the
aid of a two-sided, fixed-area sam-
pler and a trowel. The samples were
placed in paper bags, and oven-dried
at 50 C for 24 hours. Seeds were re-
covered by shaking soil samples
through a series of Tyler sieves
(#5,#10,#14,#18,#20,and #25) for 3
minutes. The number of seeds re-
maining were counted by examining
the contents of each sieve, both dry
and immersed in a salt water solu-
tion, through a lOX viewing scope.

The average number of seeds re-
covered in the soil samples taken
from the 4 experimental plots di-
vided by the total number found in
the control plot times 100 was used
as a seed survival index (SSI). This
dimensionless index permits com-
parison of the removal of different
sized seeds by rodents even though
they were planted at different rates.
It also standardizes for the experi-

mental error contributed by the diffi-
culty of recovering small seeds. The
granivorous arthropods and birds
present on the study area had equal
access to control and experimental
plots, so should not have biased SSIs.

Results

Eleven of 17 individual rodents cap-
tured on or around the plots were
heteromyids: 9 were Merriam's kan-
garoo rats, and 2 were bannertail

Seeds sown separately

0 3 9 18 36

DAYS AFTER PLANTING

Figure 1 .—Seed survival index (average number of seeds recovered in 4 experinr»ental plots
divided by total nunnber of seeds recovered in ttie control plot tinnes 1 00) for snrxali and large
seeds. (A) Seeds sown separately and (B) Seeds sown togetlier.

400

kangaroo rats (D. spectabilis). Two
white-throated woodrats (Neotoma
albigula), 2 southern grasshopper
mice (Onychomys torridus), 1 deer
mouse (Peromyscus maniculatus) and
1 cotton rat (Sigmodon hispidus) were
also captured.

Whether large and small seeds
were planted separately or together,
the SSIs were higher for small seeds
than for large seeds starting with 3
days after planting (fig. 1). After 36
days, the large seeds planted either
separately or with small seeds were
virtually gone from experimental
plots (SSI = 0.4 and 2.1, res|:>ectively).
The SSI for small seeds planted sepa-
rately was 76.5 after 36 days, while
the small seeds planted with large
seeds had an SSI of 43.6. The SSIs of
large seeds planted separately de-
creased at a faster rate than the SSIs
of large seeds sown with small seeds.
The SSIs of small seeds planted sepa-
rately decreased at a slower rate than
the SSIs of small seeds sown with
large seeds, however. Complete data
are presented in Standley (1985).

Discussion

I do not present inferential statistics
to test for significant differences be-
tween large and small seed survival
because the experimental plots were
actually sub-plots rather than true
replicates (Hurlbert 1984). For this
study site, however, striking differ-
ences between the SSIs of large and
small seeds whether planted sepa-
rately or together are certainly evi-
dence that smaller seeds have a
much higher survival rate than large
seeds due to differential predation by
rodents.

The higher rate of removal of large
seeds planted separately compared
to large seeds planted with small
seeds may have occurred because the
lower density of large seeds in the
mixed rows made them less attrac-
tive to rodents. The relatively higher
rate of removal of small seeds
planted with large seeds, compared

to small seeds planted separately,
likely occurred because large seeds
attracted rodents to the rows, where
the rodents then ate both sizes of
seeds. Sullivan and Sullivan (1982)
observed the opposite effect when
seeding lodgepole pine (Pinus con-
torta). Lodgepole seed consumption
by rodents was reduced by planting
the relatively small lodgepole seeds
with sunflower seeds, which were
larger and more preferred by gra-
nivorous rodents present. The oppos-
ing results may be due to differences
in method of seeding (Sullivan and
Sullivan broadcast their seeds) or the
size of plots (Sullivan and Sullivan's
plots were larger). Another possibil-
ity is that the main granivorous ro-
dents in their study area, deer mice,
are more selective than the hetero-
myids present in this study. Nine
days after planting there was a lower
SSI for small seeds planted sepa-
rately (fig. la) than on 18 or 36 days,
which can only be attributed to vari-
ability in seeding rate and sampling
error.

It is possible that not all seeds re-
moved by rodents, small or large,
were destroyed. Reynolds and Glen-
dening (1949) found that the seed
caching behavior of Merriam's kan-
garoo rats actually increased spread
of some plant species.

Factors other than seed size affect
selection by rodents for particular
seed species, such as percent soluble
carbohydrates (Kelrick and MacMa-
hon 1985, Kelrick et al. 1986; but also
see Jenkins 1988), moisture content
(Frank 1988a), and moldiness (Frank
1988b). For most seeds, however, re-
source managers have only the infor-
mation on size available. This study
only compared the effect of size by
using grass seeds of similar composi-
tion that differ most in their linear
dimensions. The results of this study
support other studies which showed
that heteromyid rodents selected
large seeds and reduced standing
stocks of large seeds in the soil to a
greater extent than small seeds
(Brown et al. 1979a, Inouye et al.

1980, Price 1983). Therefore, when
site conditions and management
needs allow a choice of which species
to seed, resource managers should
consider the size of seeds when plan-
ning seeding in areas inhabited by
heteromyids.

Acknowledgments

Appreciation is extended to Drs. J. H.
Brown, H. L. Morton, and N. S.
Smith for guidance, and to S. Collins,
Dr. J. Cox, S. Horton, M. Podborny,
B. Travis, and D. Youkey for field
assistance.

The research was funded by the
Arid Land Ecosystems Improvement
Unit of the USD A, Agricultural Re-
search Service.

Literature Cited

Abbott, Herschel G. 1962. Tree seed
preferences of mice and voles in
the northeast. Journal of Forestry
60:97-99.

Bramble, William C. and Ward M.
Sharp. 1949. Rodents as a factor in
direct seeding on s]X)il banks in
central Pennsylvania. Journal of
Forestry 47:477-478.

Brown, James, H., Diane W.
Davidson, and O. J. Reichman.
1979a. An experimental study of
competition between seed-eating
desert rodents and ants. American
Zoologist 19:1129-1143.

Brown, James, H., O. J. Reichman,
and Diane W. Davidson. 1979b.
Granivory in desert ecosystems.
Annual Review of Ecology and
Systemarics 10:201-227.

Cox, Jerry, R., Howard L. Morton,
Thomas. N. Johnson, Jr., Gilbert L.
Jordan, S. Clark Martin, and Louis
C. Fierro. 1982. Vegetation restora-
tion in the Chihuahuan and Sono-
ran deserts of North America.
United States Department of Agri-
culture, Agricultural Research
Service, Western Series Number
28, August 1982.

401

Everett, Richard L., Richard O.
Meeuwig, and Richard Stevens.
1978. Deer mouse preference for
seed of commonly planted species,
indigenous weed seed, and sacri-
fice foods. Journal of Range Man-
agement 31:70-73.

Frank, Craig L. 1988a. The influence
of moisture content on seed selec-
tion by kangaroo rats. Journal of
Mammalogy 69:353-357.

Frank, Craig L. 1988b. The effects of
moldiness level on seed selection
by Dipodomys spectabilis. Journal of
Mammalogy 69:358-362.

Gashv^iler, Jay S. 1967. Conifer seed
survival in a western Oregon
clearcut. Ecology 48:431-438.

Hurlbert, Stuart H. 1984. Pseu-
doreplication and the design of
ecological field experiments. Eco-
logical Monographs 54:187-211.

Inouye, Richard S., Gregory S. Byers,
and James H. Brown. 1980. Effects
of predation and competition on
survivorship, fecundity, and com-
munity structure of desert annu-
als. Ecology 61:1344-1351.

Jenkins, Stephen H. 1988. Comments
on relationships between native
seed preferences of shrub-steppe
granivores and seed nutritional
characteristics. Oecologia 75:481-
482.

Jordan, Gilbert L. 1981. Range seed-
ing and brush management on
Arizona rangelands. Cooperative
Extension Service, Agricultural
Experiment Station, University of
Arizona, College of Agriculture,
T81121.

Kelrick, Michael Ira, and James A.
MacMahon. 1985. Nutritional and
physical attributes of seeds of
some common sagebrush-steppe
plants: Some implications for eco-
logical theory and management. J.
Range Manage. 38:65-69.

Kelrick, M. I., J. A. MacMahon, R. R.
Parmenter, and D. V. Sisson. 1986.
Native seed preferences of shrub-
steppe rodents, birds and ants: the
relationships of seed attributes
and seed use. Oecologia 68:327-
337.

Lockard, Robert, B., and Joan S.
Lockard. 1971. Seed preference
and buried seed retrieval of Dipod-
omys deserti. Journal of Mam-
malogy 52:219-221.

Martin, S. Clark and Hudson G. Rey-
nolds. 1973. The Santa Rita Experi-
mental Range: your facility for re-
search on semidesert ecosystems.
Journal of the Arizona Academy
of Science 8:56-67.

Nelson, Jack R., A. M. Wilson, and
Carl J. Goebel. 1970. Factors influ-
encing broadcast seeding in
bunchgrass range. Journal of
Range Management 23:163-170.

Price, Mary V. 1983. Laboratory
studies of seed size and seed spe-
cies selection by heteromyid ro-
dents. Oecologia 60:259-263.

Reynolds, Hudson G. 1950. Relation
of Merriam kangaroo rats to range
vegetation in southern Arizona.
Ecology 31:456-463.

Reynolds, H. G. and G. E. Glenden-
ing. 1949. Merriam kangaroo rat as
a factor in mesquite propagation
on southern Arizona range lands.
Journal of Range Management
2:193-197.

Reynolds, Hudson, G. and Horace S.
Haskell. 1949. Life history notes
on Price and Bailey pocket mice of
southern Arizona. Journal of
Mammalogy 30:150-156.

Smigel, Barbara W., and Michael L.
Rosenzweig. 1974. Seed selection
in Dipodomys merriami and Per-
ogmthus penicillatus. Ecology
55:329-339.

Smith, Christopher C. 1970. The co-
evolution of pine squirrels (Tami-
asciurus) and conifers. Ecological
Monographs 40:349-371.

Spencer, Donald A. 1954. Rodents
and direct seeding. Journal of For-
estry 52:824-826.

Standley, William G. 1985. The effect
of seed size on reseeding in the
presence of heteromyids. M.S. the-
sis. University of Arizona, Tucson.
15 p.

Sullivan, Thomas P. 1979. Repopula-
tion of clear-cut habitat and coni-
fer seed predation by deer mice.

Journal of Wildlife Management
43:861-871.

Sullivan, Thomas, P. and Druscilla S.
Sullivan. 1982. The use of alterna-
tive foods to reduce lodgepole
pine seed predation by small
mammals. Journal of Applied
Ecology 19:33-45.

Sullivan, Thomas P. and Druscilla S.
Sullivan. 1984. Influence of range
seeding on rodent populations in
the interior of British Columbia.
Journal of Range Management
37:163-165.

402

Habitat Use by Gunnison's
Prairie Dogs^

C. N. Slobodchikoff,^ Anthony Robinson,^
and Clark Schaack*

Abstract.— Gunnison's prairie dogs (Cynomys gun-
nisoni) ore social, colonial mammals found in Colo-
rado, New Mexico, and Arizona. Colony location
depends to a great extent on the distribution and
abundance of plants used as food, Colonies with
the highest densities of prairie dogs occur in habitats
where there is a high abundance of native species
of plants. From a management standpoint, prairie
dog populations can be conserved by maintaining
habitats that offer such resources.

Prairie dogs often have been consid-
ered "weedy" species that thrive in
disturbed habitats. However, uncer-
tainty remains about the impact of
prairie dogs on their habitat, and
about their economic impact as com-
petitors of domesticated herbivores.
Some studies of primarily black-
tailed prairie dogs (Cynomys ludovi-
cianus) show that they have a nega-
tive effect on their habitat, while
other studies show a positive effect.
Negative effects include decreased
forb and grass cover in prairie dog
towns (Knowles 1982, Archer et al.
1984), higher silicon concentrations
in grasses found in areas grazed by
prairie dogs (Brizuela et al. 1984),
and removal of plant biomass that
could be utilized by cattle (Crocker-
Bedford 1976, Hansen and Gold
1977, Crocker-Bedford and Spillett
1981). Positive effects include in-
creased plant species diversity in
prairie dog towns (Lerwick 1974,
Boddicker and Lerwick 1976, Gold
1976, Severe 1977, Beckstead and

'Paper presented at symposium, Man-
ogemer^t of Amphibians. Reptiles, and
Small Mammals in Northi America. (Flag-
staff. Al, July 19-21. 1988.)

^C. N. Slobodchikoff is Professor of Biol-
ogy. Norttiern Arizona University. Flagstaff.
AZ86011.

^Anttiony Robinson is a graduate stu-
dent in ttie Department of Biology. Norttiern
Arizona University. Flagstaff. AZ 8601 1.

^Clark Sctiaack is Assistant Scientist, De-
partment of Botany, University of Wisconsin,
Madison. Wl 53706.

Schitoskey 1980, Fagerstone 1981,
Archer et al. 1984); greater produc-
tion of forbs and grasses (Uresk and
Bjugstad 1980, Agnew 1983); and bet-
ter quality food and growing condi-
tions inside prairie dog towns
(Hassien 1976, Beckstead and Schi-
toskey 1980, Fagerstone 1981, Cop-
pock et al. 1980, 1983a, 1983b, Det-
ling and Painter 1983). Prairie dog
colonies have also been shown to
provide habitat for many different
species of vertebrates other than
prairie dogs (Campbell and Clark
1981, O'Meilia et al. 1982, Agnew

1983, Clark et al. 1982).

The economic effects of prairie
dogs are also currently unclear. Al-
though they are considered pests
(Uresk 1985), a series of studies has
shown that controlling or eradicating
prairie dogs has little effect on in-
creasing the amount of food available
for cattle (Crocker-Bedford 1976,
Klatt and Hein 1978, Collins et al.

1984, Uresk 1985), and experimental
studies of competition between prai-
rie dogs and steers failed to show
that the prairie dogs had any signifi-
cant negative impact on the weight of
the steers (O'Meilia et al. 1982).

Prairie dogs have been character-
ized as being oriented to disturbed
sites that are overgrazed by cattle or
buffalo (Osbom and Allan 1949). The
relationship between prairie dog oc-
currence and overgrazing, however,
is a correlational one: prairie dogs
can be found at sites that are over-
grazed by large herbivores, but this

does not necessarily imply that the
prairie dogs specialize in colonizing
sites that are overgrazed. Over-
grazing might be occurring subse-
quent to colonization. For example,
bison are attracted to prairie dog
towns as grazing sites, because the
vegetation associated with such
towns may be more digestible,and
have a higher nitrogen content than
the vegetation at sites not colonized
by prairie dogs (Coppock et al.
1983a, 1983b).

Disturbance of a habitat can be
provided by the activities of the prai-
rie dogs themselves. By digging ex-
tensive burrow systems (King 1984),
prairie dogs disturb soil, promoting
the growth of disturbance-oriented
vegetation and increasing plant di-
versity (Gold 1976; Hansen and Gold
1977). Because prairie dogs have a
system of vigilance that depends on
being able to see terrestrial predators
from some distance away (Slobod-
chikoff and Coast 1980), they clip
shrubs and other tall vegetation that
impede visual detection. This in turn
alters the habitat into one that has
predominantly short grasses and an-
nual forbs, rather than the taller
grasses and shrubs that are more
characteristic of climax communities
(Koford 1958).

The goal of this paper is to evalu-
ate habitat use by Gunnison's prairie
dogs (Cynomys gunnisoni), and to
consider this habitat use in the con-
text of managing existing popula-
tions of this species. Many previous

403

ecological studies of prairie dogs
have focused on the blacktailed prai-
rie dogs (Cynomys ludovicianus)
found in the midwestem states. Gun-
nison's prairie dogs offer a better
opportunity to evaluate habitat re-
quirements, because this species is
associated with habitats that have
been modified less by man than habi-
tats where blacktailed prairie dogs
are currently found.

In an attempt to establish some
common habitat conditions that are
preferred by Gunnison's prairie
dogs, we have examined the follow-
ing factors at several prairie dog
sites: (1) burrow density as an indica-
tor of prairie dog population density;
and (2) plant diversity, evenness,
cover, and proportions of native and
introduced species.

Study Areas

Seven colonies in the vicinity of Flag-
staff, Arizona, were investigated.
These were: (1) Humane Society
(HS), within the city limits at an ele-
vation of 2250 m, in a meadow sur-
rounded by Ponderosa pine (Finns
ponderosa) trees on three sides and a
heavily-utilized dirt road on the re-
maining side; (2) Denny's (D), also
within the city limits at an elevation
of 2250 m, in a small meadow en-
circled by a traffic loop that serves as
an approach to the 1-17 freeway; (3)
Snow Bowl (SB), 10 km north of Flag-
staff in an old-field pasture at an ele-
vation of 2400 m; (4) Upper Michel-
bach (UM), on a privately owned
ranch 20 km north of Flagstaff at an
elevation of 2650 m; (5) Lower Mich-
elbach (LM), also at 2650 m and lo-
cated within 1 km east of UM; (6) Po-
tato Lake (PL), in an alpine meadow
surrounded by forested slopes, 25
km northeast of Flagstaff at an eleva-
tion of 2850 m; and (7) Bismark Lake
(BL), another alpine meadow 20 km
northeast of Ragstaff at 2900 m.

Grazing pressure on these sites
varied. The most heavily grazed site
was Upper Michelbach, with grazing

levels of 0.8 ha per AUM. The Hu-
mane Society site was heavily grazed
(1.2 ha per AUM) until 1978, after
which there was no grazing. Both
Lower Michelbach and Snow Bowl
had the same level of grazing (6 ha
per AUM). The Potato Lake and Bis-
mark Lake sites had relatively light
levels of grazing (12 ha per AUM at
PL; 14 ha per AUM at BL). The
Denny's site was not grazed at all in
the last 20 years (all grazing informa-
tion from J, Mundell, pers. comm.).

Methods

To estimate relative densities of prai-
rie dog populations, we sampled
burrow densities at six of the sites
(HS, SB, UM, LM, PL, and BL). Bur-
rows were estimated by laying out
twelve 50 m transects, and counting
all the burrows that were within 0.5
m of each side of the transect line.
Based on the counts of burrows per
transect, mean numbers of burrows /
0.005 ha (mean number of burrows
per 50 m-sq) were calculated for each
colony. Because of the small size of
the colony at BL, only six transects
were used there. Although this
method did not provide a total num-
ber of burrows per site (a number
constantly changing depending on
prairie dog construction activity), it
did provide a measure that allowed
comparison of the six sites.

As an estimate of habitat composi-
tion, vegetation at five sites (HS, SB,
D, PL, and BL) was sampled from
May-October, 1986-87. All plant spe-
cies found at each site were identi-
fied to species and classified as na-
tive non-weedy, native-weedy, or
introduced-weedy. Reference speci-
mens for each species from each site
have been deposited in the Herbar-
ium at Northern Arizona University.

For estimates of plant diversity
and percent cover, we sampled
plants every month along transects at
two sites (HS and SB) from May-Oc-
tober, 1986 and 1987. Each site had
six 100 m parallel transects spaced 20

m apart. Presence or absence of
plants by species were recorded ev-
ery 2 m along each transect.

Similarity indices (SI) were calcu-
lated for plant species composition
between sites, as follows:

Number of Species Common
to Both Site A and B
S|=

Total Number of Species in
Site A + Site B

This is an index that allows compari-
sons of sites based on the percentage
of species common to the two.

Prairie dog densities were deter-
mined at two sites, HS and SB, by
actual counts of all the animals at
each site. The prairie dogs were
trapped weekly in squirrel-sized
Tomahawk live traps and marked
with hair dye. Movements of marked
prairie dogs were observed and plot-
ted with respect to a 100 x 120 m grid
of stakes set up 10 m apart. Territo-
ries were determined behaviorally,
on the basis of aggressive behaviors
such as chases between interterritory
members, and cooperative behaviors
such as greet-kisses between intrater-
ritory members. At these two sites,
HS and SB, the number of burrows in
each territory was counted.

All statistical analyses were done
on a Honeywell Sigma 6 mainframe
computer, using SPSS statistical
packages (Nie et al. 1975). Analyses
included regression, correlation,
analysis of variance, and least signifi-
cant difference. Additionally, eco-
logical indices were calculated: even-
ness, percent cover, Simpson's domi-
nance, Shannon-Weaver diversity,
and H max (Poole 1974).

Results

Plant Species Composition

Similarity indices show that some
sites were quite dissimilar from other
sites (table 1). The HS and D sites
were most similar (63.7 percent simi-
larity), and SB was fairly similar to

404

Table 1.— Similarity Indices for five Gunnison's prairie dog colonies, based
on plant species composition at each site. A similarity of 100 Implies that
all the plant species at both sites are the same. A similarity of 0 implies ttiat
no plant species are common to the two sites. Sites are: BL = Bismark Lake;
D = Denny's; PL = Potato Lake; SB = Snow Bowl; HS = Humane Society.

Site

BL

D

PL

SB

Humane Society (HS)

6.7

63.7

8.6

54.1

Snow Bowl (SB)

10.2

44.1

12.9

Potato Lake (PL)

23.4

4.8

Denny's (D)

8.2

100 -I

N NW I N NW I N NW I N NW I N NW I

HS SB BL PL D

SITES

Figure 1 .—Composition of plant species at five Gunnison's prairie dog colonies near Rag-
staff, Arizona. Percentages shown are for Native-nonweedy species (N), Native-weedy spe-
cies (NW), and Introduced-weedy species (I). Sites are: HS = Humane Society; SB = Snow
Bowl; BL = Bismark Lake; PL = Potato Lake; D = Denny's.

Table 2— Mean burrow densities and standard deviations at 6 Gunnison's
prairie dog colony sites. Means that are not significantly different (LSD Test)
are associated by the same letter.

Site Mean + SD LSD Test

Upper Michelbach (UM) 5.42 + 2. 15 a

Humane Society (HS) 4.17+1.90 a b

Snow Bowl (SB) 3.17+1.69 be

Lower Michelbach (LM) 2.92 + 2.47 b c

Potato Lake (PL) 2.83 + 1 .64 be

Bismark Lake (BL) 2.17+1.72 c

V J

405

the HS site (54.1 percent similarity)
and to the D site (44.1 percent simi-
larity). The HS, D, and SB sites were
quite dissimilar from the other two
sites, PL and BL,and the two latter
sites had a low level of similarity
(23.4 percent) to each other.

The five sites differed in plant spe-
cies composition based on the pro-
portion of native-nonweedy, native-
weedy, and introduced-weedy plant
species (fig. 1). The PL site had the
greatest proportion of native-non-
weedy species (93.1 percent), and the
D site had the lowest (27.2 percent).
Conversely, the PL site had no (0
p>ercent) native-weedy species, while
the D site had the highest proportion
(45.7 percent) of native-weedy spe-
cies. The BL site has the greatest pro-
portion (33.3 percent) of introduced-
weedy species found at any site.

Prairie Dog Burrow Density

The mean numbers of burrows per
0.005 ha found at sites HS, SB, UM,
LM, PL, and BL are shown in table 2.
The highest burrow density was at
UM, and the lowest density was at
BL. These differences between sites
were significant (LSD = 1 .62, P =
0.05). The two sites from the Michel-
bach colonies (UM and LM) had sig-
nificantly different burrow densities,
even though these two sites were
within 1 km of one another.

Burrow density was positively
correlated with prairie dog density at
both sites (HS and SB) where prairie
dog densities were determined and
all burrows were counted. Burrow
density significantly correlated with
prairie dog density at r = 0.665, ac-
counting for 44.2 percent of the vari-
ance (F = 10.32, df = 1, 13, P < 0.01).

For a pooled 15 territories at the
two sites, the mean burrow density
was 13.73 burrows per territory (s =
8.3), and the mean number of prairie
dogs per territory was 6.4 (s = 6.7).
Consequently, on the average, there
were twice as many burrows as prai-
rie dogs per territory.

Burrow Density, Evenness, Plant
Cover, and Plant Species Diversity

Plant cover and plant species diver-
sity were negatively correlated with
burrow density. Multiple regression
analysis with burrow density as the
dependent variable and plant even-
ness, percent cover, Simpson's domi-
nance, Shannon-Weaver diversity,
and H max as independent variables
was significant (F = 5.25, df = 5, 7, P
< 0.05), accounting for 88.8 percent of
the total variance in burrow density.
Of these, evenness (F = 7.47), percent
cover (F = 10.37), and Shannon-
Weaver diversity (F = 7.39) were sig-
nificant to the regression. Evenness
had an r = -0.416, percent cover had
an r = -0.349, and Shannon- Weaver
diversity had an r = -0.427.

Burrow Density, Native Species,
and Introduced Species

Burrow density was negatively cor-
related with the number of intro-
duced-weedy plant species (F =
18.14, df = 1, 10, P < 0.01). Regression
analysis showed that burrow density
was correlated with introduced-
weedy plant species at r = -0.673, ac-
counting for 45.3 percent of the vari-
ance in burrow density.

Burrow density was not signifi-
cantly correlated with either native-
non weedy species or native-weedy
species when each of these was con-
sidered as an independent variable.
However, when these two were com-
bined into a single variable, native
species, this produced a highly sig-
nificant positive correlation of r =
0.803 (F = 18.14, df = 1, 10, p < 0.01),
accounting for 64.5 percent of the
variance in burrow density.

Burrow Density, Plant Species, and
Levels of Grazing

Burrow density was significantly cor-
related with the level of grazing (r =

0.903, F = 17.8, df = 1, 4, P < 0.05).
The more a site was grazed, the
higher was the burrow density. Re-
gression analysis showed that graz-
ing levels were not significantly cor-
related with either the number of in-
troduced species or the number of
native nonweedy species at a site.
Grazing was significantly correlated
with the number of native weedy
species (r = 0.975, F = 37.9, df = 1,2, P
< 0.05), and weakly correlated with
the total number of plant species (r =
0.947, F = 17.4, df = 1,2, P = 0.06).
Multiple regression with burrow
density as the dependent variable
and native species, introduced spe-
cies, and grazing level as independ-
ent variables showed that native spe-
cies (number of native weedy and
native nonweedy species combined)
explained 97.9 percent of the vari-
ance in burrow density, while graz-
ing level explained an additional 1 .8
percent and introduced species ex-
plained 0.2 percent.

Discussion

Our results show that Gunnison's
prairie dogs thrive at sites with na-
tive-nonweedy and native-weedy
Sf)ecies of plants. Gunnison's prairie
dogs apparently do not prefer sites
that have a high proportion of intro-
duced-weedy species. This is not sur-
prising when one considers the die-
tary requirements of these animals.
Shalaway and Slobodchikoff (1988)
found that the diet of Gunnison's
prairie dogs at three sites in the Flag-
staff area consisted primarily of na-
tive plants: native-weedy and native-
nonweedy species made up 60-80
p)ercent of the animals' food. Intro-
duced-weedy species made up a rela-
tively low proportion of the diet of
Gunnison's prairie dogs in that
study.

Contrary to the findings of studies
with blacktailed prairie dogs (Ler-
wick 1974, Boddicker and Lerwick

1976, Gold 1976, Hansen and Gold

1977, Beckstead and Schitoskey 1980,

Archer et al. 1984), Gunnison's prai-
rie dogs did not increase plant spe-
cies diversity, but instead decreased
it. This effect can be produced by the
clipping action of prairie dogs on
plants that tend to grow tall and ob-
scure the animals' view of terrestrial
predators. Such clipping action can
lower the competitive ability of
shrubs and other tall plants, eventu-
ally eliminating them from prairie
dog towns. Many of these species are
introduced weedy plants. A similar
effect was described by Clements
and Clements (1940) with Gunnison's
prairie dogs.

The effects of Gunnison's prairie
dogs on plant cover were consistent
with those found by other studies
(Knowles 1982, Archer et al. 1984). In
each case, prairie dogs decreased
plant cover. This is to be expected,
since all species of prairie dogs graze
on vegetation and can eat up some
24-90 percent of the primary produc-
tion of a site (Osbom and Allan 1949,
Hansen and Gold 1977, Crocker-
Bedford and Spillett 1981). To the
extent that blacktailed prairie dogs
and cattle have a dietary overlap of
76 percent (Kelso 1939), prairie dogs
have been construed as competitors
of large herbivores such as cattle.
However, because prairie dogs feed
very selectively on plants, 80 percent
of the biomass they ingest may come
from plant parts not utilized by cattle
(Crocker-Bedford 1976). Also, any
potential competitive effect might be
minimized by the relatively small
size of most extant prairie dog colo-
nies (King 1955; Koford 1958; Smith
1955), and the beneficial effects that
large herbivores may obtain from
plants that grow in prairie dog colo-
nies (Coppock et al. 1983a).

The positive correlation between
grazing level and density of prairie
dog burrows suggests that prairie
dogs are found more in habitats that
are highly grazed. However, merely
addressing prairie dog management
in terms of possible comp)etition with
cattle misses a much more funda-
mental issue: that of the prairie dog' s

406

place in a natural ecosystem. While
our study has found a positive corre-
lation between prairie dog densities
and grazing, the presence of these
animals at ungrazed sites indicates
that they can establish themselves in
ungrazed areas that have the right
configuration of habitat characteris-
tics.

A much more important point
than grazing is the strong link be-
tween the presence of prairie dogs
and the success of native species of
plants. Introduced weeds are not fa-
vored in prairie dog colonies, even
though the soil is disturbed through
the burrowing actions of these ani-
mals. Rather than being "weed/'
pests who come into overgrazed
lands, prairie dogs might actually
have the function of repairing over-
grazed land, and driving the plant
community toward a more natural
one.

The mechanism for how prairie
dogs might drive the ecosystem to-
ward more native plant species is
still unclear. We have found that
Gunnison's prairie dogs decrease
both species diversity and plant
cover. The decrease in species diver-
sity apparently comes from a de-
crease in the component represented
by the introduced weedy plant spe-
cies, and not from the native plant
species. The decrease in plant cover
comes from herbivory on the plants
growing in the colonies. Some native
plant species produce more flower-
ing stalks and more seeds when they
are grazed by herbivores (Paige and
Whitham 1987). Experimental evi-
dence for black-tailed prairie dogs
shows that both forbs and grasses
increased in plots that contained both
prairie dogs and cows (Uresk and
Bjugstad 1980). In the arid conditions
of the Southwest, native plants might
be better adapted to climatic condi-
tions than introduced weedy species,
and might respond to herbivory by
increasing their numerical abun-
dance. The relationship that we
found between levels of grazing and
prairie dog burrow densities may be

the result of herbivory stimulating
the growth of plants necessary to the
diet of Gunnison's prairie dogs.

Our results suggest that Gunni-
son's prairie dogs must be conserved
by maintaining habitats with a large
component of native vegetation.
Gunnison's prairie dogs are a natural
part of native ecosystems, and have
evolved alongside large herbivores
such as elk, deer, and buffalo, all of
which feed to some extent on native
species of grasses and forbs. Native
plant species have evolved to com-
pensate for these effects of herbivory,
and possibly for this reason prairie
dogs might have a beneficial function
of restoring rangeland that has been
damaged by grazing; this is a man-
agement question that must be ad-
dressed experimentally in the future.
In addition to the positive association
between prairie dogs and native
plant species, prairie dog towns are
habitat sites that are integral to the
existence of large numbers of other
vertebrates and invertebrates, and
eradication of prairie dogs can have
detrimental consequences to natural
ecosystems. Experimental and eco-
nomic evidence currently indicates
that eradication of prairie dogs is nei-
ther economically feasible nor par-
ticularly beneficial to cattle. We sug-
gest that prairie dogs should be
looked at in a more positive role that
reflects their impact on the mainte-
nance of natural ecosystems.

Acknowledgments

We thank Jim Benedix, Ed Creef, Ch-
eryl Fischer, Kitty Gehring, Gene
Hickman, and Steve Travis for assis-
tance in data collection and summa-
rization. We also thank Judith Ki-
riazis for her helpful and insightful
comments on this manuscript.

Literature Cited

Agnew, W. 1983. Flora and fauna as-
sociated with prairie dog ecosys-

tems. M. S. Thesis, Colorado State
University, Fort Collins, 47 p.

Archer, S. R., I. D. Lebreton, M. G.
Garrett, and J. K. Detling. 1984.
Structural changes in a mixed-
grass prairie dog community as a
function of prairie dog coloniza-
tion history. Bulletin Ecological
Society of America 65:162.

Beckstead, M. and F. Schitoskey, Jr.

1980. Assimilation efficiency of the
black-tailed prairie dog. Proceed-
ings South Dakota Academy of
Science 59:184-194.

Boddicker, M. J. and A. Lerwick.
1976, Vegetation changes induced
by prairie dogs on shortgrass
range. Journal of Range Manage-
ment 29:221-225.

Brizuela, M. A., J. K. Detling, and M.
S. Cid. 1984. Seasonal silicon con-
tent of grasses from sites receiving
different grazing pressure by
grassland herbivores. Bulletin Eco-
logical Soc. of Amer. 65:161-162.

Campbell, T. M., Ill, and T. W. Clark.

1981. Colony characteristics and
vertebrate associates of white-
tailed and black-tailed prairie
dogs in Wyoming. American Mid-
land Naturalist 105:269-275.

Clark, T. W., T. M. Campbell, III, D.
C. Socha, and D. E. Casey. 1982.
Prairie dog colony attributes and
associated vertebrate species.
Great Basin Naturalist 42:572-582.

Clements, F. E. and E. S. Clements.
1940. The biotic significance of dis-
turbance. Climate, Climax, and
Conservation: Carnegie Institute
of Washington Yearbook 39:174-
175.

Collins, A. R., J. P. Workman, and D.
W. Uresk. 1984. An economic
analysis of black-tailed prairie dog
(Cynomys ludovicianus) control.
Journal of Range Management
37:358-361.

Coppock, D. L., J. K. Detling, and M.
I. Dyer. 1980. Interactions among
bison, prairie dogs and vegetation
in Wind Cave National Park. Final
Report to National Park Service,
Wind Cave National Park, Hot
Springs, South Dakota. 177 p.

407

Coppock, D. L., J. K. Detling, J. E.
Ellis, and M. I. Dyer. 1983a. Plant-
herbivore interactions in a North
American mixed-grass prairie. I.
Effects of black-tailed prairie dogs
on interseasonal aboveground
plant biomass and nutrient dy-
namics and plant species diver-
sity. Oecologia 56:1-9.

Coppock, D. L., J. E. Ellis, J. K. Det-
ling, and M. I. Dyer. 1983b. Plant-
herbivore interactions in a North
American mixed-grass prairie. II.
Responses of bison to modification
of vegetation by prairie dogs.
Oecologia 56:10-15.

Crocker-Bedford, D. C. 1976. Food
interactions between Utah prairie
dogs and cattle. M. S. Thesis. Utah
State University, Logan. 131 p.

Crocker-Bedford, D. C. and J. J.
Spillett. 1977. Home ranges of
Utah prairie dogs. Journal of
Mammalogy 58:672-673.

Detling, J. K. and E. L. Painter. 1983.
Defoliation responses of western
wheatgrass populations with di-
verse histories of prairie dog graz-
ing. Oecologia 57:65-71.

Fagerstone, K. A. 1981. A review of
prairie dog diet and its variability
among animals and colonies, p.
178-184. In R. M. Timm and R. J.
Johnson, eds. Proceedings Fifth
Great Plains Wildlife Damage
Control Workshop. Institute of
Agriculture and Natural Re-
sources, University of Nebraska,
Lincoln.

Gold, I. K. 1976. Effects of blacktail
prairie dog mounds on shortgrass
vegetation. MS Thesis, Colorado
State University, Fort Collins,
Colorado.

Hansen, R. M. and I. K. Gold. 1977.
Blacktail prairie dogs, desert cot-
tontails and cattle trophic relations
on shortgrass range. Journal of
Wildlife Management 30:210-214.

Hassien, F. 1976. A search for black-
footed ferrets in the Oklahoma
panhandle and adjacent area and
an ecological study of black-tailed
prairie dogs in Texas County,
Oklahoma. M. S. Thesis, Okla-

homa State University, Stillwater.
Ill p.

Kelso, L. H. 1939. Food habits of
prairie dogs. USDA Circular
Number 529.

King, J. A. 1955. Social behavior, so-
cial organization, and population
dynamics in a black-tailed prairie
dog town in the Black Hills of
South Dakota. Contributions of
the Laboratory of Vertebrate Biol-
ogy, University of Michigan 67:1-
123.

King J. A. 1984. Historical ventila-
tions on a prairie dog town. p.
447-456. In J. O. Murie and G. R.
Michener, eds. The biology of
ground-dwelling squirrels. Uni-
versity of Nebraska Press, Lincoln,
Nebraska.

Klatt, L. E. and D. Hein. 1978. Vege-
tative differences among active
and abandoned towns of black-
tailed prairie dogs (Cynomys ludo-
vicianus). Journal of Rangeland
Management 31:315-317.

Koford, C. B. 1958. Prairie dogs,
whitefaces, and blue grama. Wild-
life Monographs 3:1-78.

Knowles, C. J. 1982. Habitat affinity,
populations, and control of black-
tailed prairie dogs on the Charles
M. Russell National Wildlife Ref-
uge. Ph. D. Dissertation, Univer-
sity of Montana, Missoula. 171 p.

Lerwick, A. C. 1974. The effects of
the black-tailed prairie dog on
vegetative com}X)sition and their
diet in relation to cattle. M.S. The-
sis, Colorado State University,
Fort Collins. 106 p.

Nie, N. H., C. H. Hull, J. G. Jenkins,
K. Steinbrenner, and D. H. Bent.
1975. SPSS: Statistical package for
the social sciences, ed. 2, McGraw-
Hill, New York.

O'Meilia, M. E. 1980. Competition
between prairie dogs and beef
cattle for range forage. M.S. The-
sis, Oklahoma State University,
Stillwater. 32 p.

Osbom, B. and P. F. Allan. 1949.
Vegetation of an abandoned prai-
rie dog town in tallgrass prairie.
Ecology 30:322-332.

Paige, K. N. and T. G. Whitham.
1987. Overcompensation in re-
sponse to mammalian herbivory:
the advantage of being eaten.
American Naturalist 129:407-416.

Poole, R. W. 1974. An introduction to
quantitative ecology. McGraw-
Hill, N. Y. 532 p.

Severe, D. S. 1977. Re vegetation of
black-tailed prairie dog mounds
on shortgrass prairie in Colorado.
M.S. Thesis, Colorado State Uni-
versity, Fort Collins. 92 p.

Shalaway, S. and C. N. Slobod-
chikoff. 1988. Seasonal change in
the diet of the Gunnison's prairie
dog. Journal of Mammalogy, in
press.

Slobodchikoff, C. N. and R. Coast.
1980. Dialects in the alarm calls of
prairie dogs. Behavioral Ecology
and Sociobiology 7:49-53.

Smith, R. E. 1955. Natural history of
the prairie dog in Kansas. Univer-
sity of Kansas Museum Natural
History and State Biological Sur-
vey Publication, Number 16.

Uresk, D. W. 1985. Effects of control-
ling black-tailed prairie dogs on
plant production. Journal of
Rangeland Management 38:466-
468.

Uresk, D. W. and A. J. Bjugstad.
1980. Prairie dogs as ecosystem
regulators on the Northern High
Plains, p. 91-94. In C. L. Kucera,
ed. Proceedings 7th North Amer-
ica Prairie Conference. Southwest-
ern Missouri State University,
Springfield.

408

Environmental Contanninants
and the Management of Bat
Populations in the United
States^

Donald R. Clark, Jr.^

Abstract. —Food-chain Residues of orgono-
chlorine pesticides probably have been involved in
declines of some U.S. Bat populations; examples in-
clude free-toiled bats at Carlsbad Cavern. New
Mexico, and the endangered gray bat at sites in
Missouri and Alabama. If a long-lived contaminant
has not been dispersed in large amounts over large
areas, its impact may be controlled by administra-
tive action that stops its use or other environmental
discharge, or that results in physical isolation of local-
ized contamination so that it no longer enters food
chains.

Several species of bats in the U.S.
Form large aggregations in caves, old
mines, or other shelters, and many of
these colonies are of management
concern to biologists working for the
states or federal government (e.g.
Prichard 1987). Four taxa, the gray
bat (Myotis grisescens), Indiana bat
(M. sodalis), Ozark big-eared bat (Ple-
cotus townsendii ingens), and Virginia
big-eared bat (P. t. virginianus), are of
particular concern because they are
endangered (USDI, FWS 1987).

Habitat destruction such as defor-
estation, water pollution, stream
channelization, and stream sedimen-
tation (Tuttle 1979, Prichard 1987) or
direct human disturbance and de-
struction of bats (Tuttle 1979, for a
recent example see Anon. 1987) are
primary known threats to bat colo-
nies. However, environmental con-
taminants, such as organochlorine
pesticide residues and heavy metals,
probably have been involved in some
declines of bat populations. In this
paper I discuss the management im-
plications of these contaminants.
(Note: for purposes of this discus-
sion, "management' refers broadly
to human activities undertaken in the
interest of a bat colony with the goal

' Paper presented at symposium Man-
agement of Amphibians. Reptiles, and
Small Mammals in Nortf^ America. (Flag-
staff . AZ. July 19-21, 1988.)

^Donald R. Clark, Jr., is Researcti Wildlife
Biologist, U.S. fish and Wildlife Service,
Patuxent Wildlife Researcti Center, Laurel,
MD 20708.

that colony size will remain at a
steady, sustainable level or will in-
crease to such a level.)

Examples of Possible Food-Ctioin
Contaminant Impacts on Bat
Populations

Free-Tailed Bats at Carlsbad
Cavern, New Mexico

The Carlsbad population of Mexican
free-tailed bats (Tadarida brasiliensis
mexicam) was estimated at 8.7 mil-
Hon bats in 1936 (Allison 1937) but
only 2(X),(X)0 bats remained in 1973
(Altenbach et al. 1979). Several die-
offs occurred during this interval
(Altenbach et al. 1979), and none was
linked directly to pesticide poison-
ing; however, routine testing of tis-
sues was not available. The question
of pesticide involvement was ad-
dressed by simulating migratory
flight in young bats taken from the
colony in 1974 (Geluso et al. 1976).
Some of these bats died of DDE (1,1'-
(dichloroethylidene)bis[4-chloroben-
zene]) poisoning (DDE is the princi-
pal metabolite of DDT; 1,1 '-2,2,2-
(tricholoroethylidene)bis[4-chloro-
benzene]) due to mobilization of
DDE received in their mother's milk
and stored in their fat (Geluso et al.
1976). This result suggests that DDT
has contributed to the decline of this
population.

High DDE concentrations in the
Carlsbad colony probably resulted

from heavy DDT use in New Mexico
before its ban in 1972; however, other
more-recent inputs have been postu-
lated to explain high DDE levels in
wildlife in parts of Texas, New Mex-
ico, and Arizona (Clark and Krynit-
sky 1983, Hunt et al. 1986, White and
Krynitsky 1986).

Gray Bats in Missouri

Dieldrin (3,4,5,6,9,9-hexachloro-
la,2,2a,3,6,6a,7,7a-octahydro-2,7:3,6-
dimethanonaphth[23-b]oxirene)
killed gray bats in 1976, 1977, and
1978 in two maternity colonies in
Franklin County, Missouri (Clark et
al. 1978b, 1983a). Residues of hep-
tachlor-related chemicals
(1,4,5,6,7,8,8 -heptachloro-3a,4,7,7a-
tetrahydro-4,7-methanoindene) in
bats from both colonies increased to
potentially dangerous concentrations
in 1977 and remained elevated in
1978 (Clark et al. 1983a). Populadon
size at one colony was estimated at
1,800 bats in 1976 and 1978, but no
bats were present from 1979-82
(Clark et al. 1983a,b). Dieldrin, per-
haps in conjunction with heptachlor,
may have caused the decline and dis-
appearance of this colony. Dieldrin
also killed gray bats at three Boone
County, Missouri, caves in 1980,
1981, and 1982 (Clark et al. 1983b,
Clawson and Clark in manuscript).

Death of gray bats were attributed
to dieldrin because this chemical was
measured in the bats' brains at con-

409

centrations known to be lethal in
other species (Clark et al. 1978b). Di-
eldrin and heptachlor-related resi-
dues came from the use of aldrin
(Dieldrin's parent compound) and,
subsequently, heptachlor, to control
cutworms (moth larvae. Family Noc-
tuidae) in corn.

Gray Bats at Cave Springs Cave,
Alabama

DDT was manufactured at Redstone
Arsenal near Huntsville, Alabama,
from 1947 to 1970, and massive
amounts of DDT and its metabolites
(DDD; l,l'-(2,2-dichloroethylidene)-
bis[4-chlorobenzene] and DDE) were
discharged into the Tennessee River
via Huntsville Spring Branch-Indian
Creek (Fleming and Atkeson 1980).
Local biota remains heavily contami-
nated (O'Shea et al. 1980, Heming
and Cromartie 1981, Fleming et al.
1984, Reich et al. 1986.).

Samples of dead or dying bats and
bat guano collected between 1976
and 1986 from four gray bat colonies
as far as 140 km downriver contained
residues from this former discharge
(Clark et al. 1988). Residues were
identifiable by their high DDD to
DDE ratio, which resulted from their
breakdown under anaerobic condi-
tions. Cave Springs Cave at Wheeler
National Wildlife Refuge houses the
colony nearest the contaminant
source — about 20 km. Biologists
judged that bat mortality at Cave
Springs Cave was far above normal
in 1978, 1985, and 1986. Residues of
DDT, DDD, and DDE in brains of
dead or dying bats from this cave,
although elevated in comparison
with residues from colonies up-
stream from Redstone Arsenal, were
well below concentrations believed
to be lethal (Clark et al. 1988). The
single exception was a bat collected
in 1978 with sufficient DDD in its
brain (29 ppm wet weight) to have
been poisoned (Clark et al. 1988). The
measured residues, therefore, did not
explain the observed mortalities.

Although there is no explanation
for this mortality yet, another con-
taminant may by involved. A guano
sample collected from Cave Springs
Cave in 1987 was analyzed for heavy-
metals and cadmium measured 8.5
Ppm (dry weight). This amount may
be compared with 2.2 Ppm cadmium
in guano (mixed gray and southeast-
ern bats, M. austroriparius) from a
Florida cave where the bats were ex-
posed to contaminations from a bat-
tery salvage plant. Kidneys of south-
eastern bats from this Florida cave
averaged 0.89 Ppm (wet weight) cad-
mium with a maximum of 2.9 Ppm.
Concentrations of cadmium as low as
3.4 Ppm in kidneys of voles (Microtus
pennsylvanicus) were associated with
reduced survivorship in enclosed
populations. Also, six gray bats
found dead in Cave Springs Cave in
June 1986 were examined by the U.S.
Fish and Wildlife Service's National
Wildlife Health Research Center,
Madison, Wisconsin. There was no
evidence of injury or infectious dis-
ease, but all bats showed mild renal
tubular degeneration. Because cad-
mium caused kidney damage (Nomi-
yama 1981), this metal, perhaps in
combination with DDD and DDE,
may have caused the recent die-off of
gray bats at Cave Springs Cave. The
cadmium source is unknown. Addi-
tional samples for chemical analysis
will be collected in 1988.

Managerrient of Contaminant
Impacts on Bat Populations

Screening for Possible
Contaminant Problems in
Apparently Healttiy Colonies

Contaminants that biomagnify or
bioaccumulate in ecosystems include
organochlorine pesticides such as
DDT (and its metabolites DDE and
DDD), dieldrin, heptachlor-related
chemicals, and the industrial poly-
chlorinated biphenyls (PCBs). Also
included are heavy metals such as
lead, cadmium, chromium, zinc, and

mercury. For chemicals that biomag-
nify or bioaccumulate, analyses of
guano samples collected from the
surface of a guano deposit can indi-
cate body burdens in bats during
their most recent activity season.
Samples from greater depths may
indicate contaminant concentrations
in previous years.

Relationships between concentra-
tions in guano and carcasses of bats
from the same colony have been de-
scribed for dieldrin, heptachlor epox-
ide, and DDE (Clark et al. 1982).
Limited data are available on concen-
trations of lead, cadmium, chro-
mium, zinc, and mercury in guano
from contaminated colonies (Petit
and Altenbach 1973, Clark 1979,
Clark et al. 1986, this paper). About
20 grams of guano, dry weight, are
necessary for analyses.

Sublethal exposure of bats to the
newer organophosphorus and car-
bamate insecticides is demonstrated
by depressed brain cholinesterase
(ChE) activity in exposed individu-
als. Depression is determined by
comparison to normal ChE activity
for a sample of control bats of the
same species. Measurement of ChE
activity (for methods, see Ellman et
al. 1961, Hill and Fleming 1982) in-
volves removal of the brain, hence
death of the bat.

Recognizing Organoctilorine
Pesticide- Induced Mortality in Bat
Colonies

Managed colonies are usually cen-
sused annually so that any significant
decline will be recognized. By also
estimating numbers of dead and
dying bats at these censuses, manag-
ers can differentiate between "nor-
mal" mortality and increased mortal-
ity, which may be the first sign of a
contaminant problem.

May of the colonies considered
most important are maternity colo-
nies, and in maternity colonies, or-
ganochlorine chemicals kill mostly
young bats. There are two reasons

410

for this. First, organochlorines be-
come concentrated in the fat of
mother's milk and these chemicals
continually and rapidly accumulate
in the young as they nurse.

For example, insects collected in
foraging areas of Missouri gray bats
contained a maximum of 3.1 Ppm
(wet weight) dieldrin, but milk taken
from the stomach of a young dead
gray bat contained 89 ppm (wet
weight) dieldrin (Clark and Prouty
1984). Second, young bats are 1.9
Times more sensitive than adults to
dieldrin and 1.5 Times more sensitive
to DDT (Clark et al. 1978a, 1983a).
Young bats dying of organochlorine
poisoning may still have milk in their
stomachs unlike young dying of star-
vation. Therefore, increased infant
mortality in a maternity colony with
some young having milk in their
stomachs may indicate poisoning by
an organochlorine chemical.

Diagnosing Chemical Poisoning in
Bats

Diagnosis for organochlorine chemi-
cals requires analyses of brains and
interpretation of the resulting meas-
urements. However, because concen-
trations in brains are closely corre-
lated with concentrations in carcass
fat (Clark 1981a), analyses of car-
casses may serve if brains are un-
available. For example, analysis of
carcasses may be the only option
when bats are partly decomposed.
Correlations between brain and car-
cass fat concentrations only have
been quantified for DDE, DDT, and
dieldrin (Clark 1981a).

Lethal brain concentrations for
DDE, DDT, dieldrin, and PCB (Aro-
clor 1260) have been determined for
at least one sp>ecies of bat (Clark
1981b). Because lethal brain levels are
fairly similar among mammals and
birds, comparisons can provide clues
about the effect on a populations,
even though the lethal level for the
species under investigation has not
been determined yet.

Diagnosis of death in bats from
heavy-metal poisoning is less certain,
but interpretations often can be made
based on other species of mammals
(Clark 1979, this paper). Diagnosis
for heavy metals involves analyzing
liver and kidneys along with histo-
logical examination for damage.

Death in bats caused by the anti-
cholinesterase insecticides could be
diagnosed by measurement of de-
pressed brain ChE in combination
with detection of an anticholinester-
ase chemical in the contents of the
gastrointestinal tracts or other tissues
of the affected bats. Lethal depres-
sion of brain ChE has been measured
in little brown bats (M. lucifugus) in
the laboratory for methyl parathion
(phosphorothioic acid 0,0-dimethyl
(3-(4-nitrophenyl)ester) and
Orthene® (acephate; acetylphosphor-
amidothioic acid 0,S-dimethyl ester)
(Clark 1986, Clark and Rattner 1987).

Even though a firm diagnosis of
contaminant-induced mortality re-
quires tissue analyses, analysis of a
guano sample, as a first step, may
indicate whether organochlorines or
metals are involved.

Chemical analyses of tissues or
guano are not something that manag-
ers usually can perform themselves.
However, an Environmental Con-
taminant Field Specialist from the
U.S. Fish and Wildlife Service can be
contacted (there are 1-3 in each
state); if he or she determines that the
situation warrants, analyses can be
done. The Specialist also may send
specimens to the National Wildlife
Flealth Research Center if disease is
suspected.

Bat specimens for diagnostic study
generally should be frozen immedi-
ately. However, examinations for
diseases and histopathology require
that specimens be kept refrigerated
but not frozen until organs can be
removed and preserved in fluid.
Control specimens of the same spe-
cies are necessary for diagnosis of
depressed brain ChE activity. Guano
does not require freezing or refrig-
eration. The Contaminant Field Spe-

cialist can provide detailed instruc-
tions for specimen collection and
handling.

Possible Impacts of New
Generation Pesticides on Bat
Colonies

Most organochlorine p)esticides have
been banned or their use otherwise
reduced in the U.S., And some wild-
life-related problems have improved.
Organochlorines largely have been
replaced by organophosphorus (e.g.,
Acephate, diazinon [phosphorothioic
acid 0,0-diethyl 0-[6-methyl-2-(l-
methylethyl)-4-pyrimidinyl]ester],
and methyl parathion) and car-
bamate (e.g., Aldicarb [2-methyl-2-
(methylthio)propanal 0-[(methyl-
amino)carbonyl]oxime], carbaryl [1-
naphthalenol methylcarbamate], and
carbofuran [2,3-dihydro-2,2-di-
methyl-7-benzofuranol methylcar-
bamate]) insecticides. These chemi-
cals are relatively short-lived and
generally do not accumulate in food
chains. Exposure in bats probably
occurs when they feed over fields or
orchards that are being, or have just
been, sprayed. In these cases, bats
might be sprayed directly and re-
ceive the chemical through their skin
and lungs. Pesticides are frequently
sprayed in the evening, at night, or
early in the morning to avoid killing
honey bees, to kill adult mosquitoes,
or to take advantage of quiet wind
conditions and thereby avoid drift.
Bats also may be exposed by eating
insects that have just been sprayed
but are still alive.

New-generation pesticides have
not yet been linked to bat die-offs,
but, in 1968, ranchers and farmers in
a cotton-growing area of Arizona re-
ported "...unusual Numbers of dead
or dying (free-tailed) bats in their
fields.. .Many Were found convuls-
ing, incapable of flight" (Reidinger
and Cockrum 1978). This mortality
was attributed to DDT; however,
chemical analyses indicated that nei-
ther lethal residues of DDT nor its

411

metabolites had been present in these
bats (Clark 1981b). Because methyl
parathion also was commonly used
on cotton in this region, mortality
may have been caused by this or-
ganophosphorus pesticide. The mor-
tality pattern described by ranchers
and farmers where bats were scat-
tered on the ground in an incapaci-
tated condition suggests quick intoxi-
cation after direct contact with a
chemical of high acute toxicity such
as the organophosphate methyl para-
thion (see Clark 1986).

Reducing Contaminant Impacts in
Bat Colonies

What can be done once it is deter-
mined that bats have died from a
food-chain contaminant? The answer
will depend on the contaminant, its
source, and on the ability or author-
ity of the manager to change local
practices or obtain cleanup proce-
dures.

When large quantities of a long-
lived chemical have been incorpo-
rated into soils over vast areas, such
as DDE in New Mexico or dieldrin in
Missouri, the chemical will continue
to enter food chains for many years.
The manager of an affected bat col-
ony can only protect the colony form
other sources of damage and hope
that it survives until the contamina-
tion dissipates. If the colony is extir-
pated, the manager can protect the
site so that it might be recolonized
from outside the contaminated area
in the future.

After a colony is known to be
heavily contaminated with an or-
ganochlorine or metal, annual analy-
ses of guano can determine whether
contamination is decreasing, increas-
ing, or remaining stable, and also can
alert the manager to potential prob-
lems. For example, in Missouri, hep-
tachlor epoxide increased from mi-
nor amounts in bats in 1976 to near
lethal levels in 1977 (Clark et al.
1983a). Such information promptly
passed to the state authorities might

persuade them to recommend a dif-
ferent pesticide to farmers before the
problem chemical becomes heavily
dispersed over wide areas.

The Alabama example given pre-
viously shows that large cleanup ef-
forts are possible if the contamina-
tion is, in total or in part, localized.
State and federal agencies represent
routes open to managers. In this in-
stance, the U.S. Environmental Pro-
tection Agency exercised its author-
ity. Whether a large cleanup effort
would be undertaken if only bats
were affected is not known; however,
if organochlorine contamination is
heavy enough to cause mortality in
bat colonies, it probably affects other
wildlife as well. Bat colonies are
good places to look for food-chain
contaminant problems because bats
feed over wide areas but congregate
in only a few roosts. Thus, problems
from many potential areas are
brought to a single site where symp-
toms may be seen as dead or dying
bats. The disadvantage is that it may
be difficult to locate the source area,
or areas, unless the feeding locations
of the bats are known.

Heavy metals in the environment
often have industrial point sources
that are subject to existing emission
regulations. Therefore, such contami-
nation may be easier to stop.

Acknowledgments

I thank R.L. Clawson, E.L. Flickinger,
K.N. Geluso, C.E. Grue, and T.H.
Kunz for critical reviews of the
manuscript.

Literature Cited

Allison, Vernon C. 1937. Evening bat
flight from Carlsbad Caverns.
Journal of Mammalogy 18:80-82.

Altenbach, J.Scott, Kenneth N.
Geluso, and Don E. Wilson. 1979.
Populations size of Tadarida brasil-
iensis at Carlsbad Caverns in 1973.
p. 341-348. In H.H. Genoways and

R.J. Baker, eds. Biological investi-
gations in the Guadalupe Moun-
tains National Park, Texas. Pro-
ceedings and Transactions Series
Number 4, National Park Service.
Anonym.ous. 1987. Thornhill cave bat
killer convicted. Bats, December
5:2,8.

Clark, Donald R., Jr. 1979. Lead con-
centrations: bats vs. Terrestrial
small mammals collected near a
major highway. Environmental
Science and Technology 13:338-
341.

Clark, Donald R., Jr. 1981a. Death in
bats from DDE, DDT or dieldrin:
diagnosis via residues in carcass
fat. Bulletin of Environmental
Contamination and Toxicology
26:367-374.

Clark, Donald R., Jr. 1981b. Bats and
environmental contaminants: a
review. United States Fish and
Wildlife Service Special Scientific
Report— Wildlife, Number 235.

Clark, Donald R., Jr. 1986. Toxicity of
methyl parathion to bats: mortal-
ity and coordination loss. Environ-
mental Toxicology and Chemistry
5:191-195.

Clark, Donald R., Jr., Fred M. Bagley,
and W. Wynon Johnson. 1988.
Northern Alabama colonies of the
endangered gray bat Myotis gris-
esens: organochlorine contamina-
tion and mortality. Biological Con-
servation 43:213-225.

Clark, Donald R., Jr., Christine M.
Bunck, and Eugene Cromartie.
1983a. Year and age effects on resi-
dues of dieldrin and heptachlor in
dead gray bats. Franklin County,
Missouri— 1976, 1977, and 1978.
Environmental Toxicology and
Chemistry 2:387-393.

Clark, Donald R., Jr., Richard L.
Clawson, and Charles J. Stafford.
1983b. Gray bats killed by dieldrin
at two additional Missouri caves:
aquatic macroinvertebrates found
dead. Bulletin of Environmental
Contamination and Toxicology
30:214-218.

Clark, Donald R., Jr., and Alexander
J. Krynitsky. 1983. DDT: recent

412

contamination in New Mexico and
Arizona? Environment 25:27-31.

Clark, Donald R., Jr., Thomas H.
Kunz, and T. Earl Kaiser. 1978a.
Insecticides applied to a nursery
colony of little brown bats (Myotis
lucifugus): lethal concentrations in
brain tissues. Journal Mammalogy
59:84-91.

Clark, Donald R., Jr., Richard K.
LaVal, and Douglas M. Swineford.
1978b. Dieldrin-induced mortality
in an endangered species, the gray
bat (Myotis grisescens). Science
199:1357-1359.

Clark, Donald R., Jr., Richard K.
LaVal, and Merlin D. Tuttle. 1982.
Estimating pesticide burdens of
bats from guano analyses. Bulletin
of Environmental Contamination
and Toxicology 29:214-220.

Clark, Donald R., Jr., and Richard M.
Prouty. 1984. Disposition of die-
tary dieldrin in the little brown bat
and correlation of skin levels with
body burden. Bulletin of Environ-
mental Contamination and Toxi-
cology 33:177-183.

Clark, Donald R., Jr., and Barnett A.
Rattner. 1987. Orthene® toxicity to
little brown bats (Myotis lucifugus):
acetylcholinesterase inhibition, co-
ordination loss, and mortality.
Environmental Toxicology and
Chemistry 6:705-708.

Clark, Donald R., Jr., Anne Shapiro
Wenner, and John F. Moore. 1986.
Metal residues in bat colonies,
Jackson County, Florida, 1981-
1983. Florida Field Nat. 14:38-45.

Ellman, George L., K. Diane Court-
ney, Valentino Andres, Jr., and
Robert M. Featherstone. 1961. A
new rapid colorimetric determina-
tion of acetylcholinesterase activ-
ity. Biochemistry and Pharmacol-
ogy 7:88-95.

Fleming, W. James, and Thomas Z.
Atkeson. 1980. Situation report:
heavy DDT contamination at
Wheeler National Wildlife Refuge.
Proceedings of the Annual Confer-
ence of the Southeast Association
of Fish and Wildlife Agencies
34:453-461.

Fleming, W. James, and Eugene Cro-
martie. 1981. DDE residues in
young wood ducks (Aix sponsa)
near a former DDT manufacturing
plant. Pesticides Monitoring Jour-
nal 14:115-118.

Fleming, W. James, Burline P. Pullin,
and D.M. Swineford. 1984. Popu-
lation trends and environmental
contaminants in herons in the Ten-
nessee Valley, 1980-81. Colonial
Waterbirds 7:63-73.

Geluso, Kenneth N., J. Scott Alten-
bach, and Don E. Wilson. 1976. Bat
mortality: pesticide poisoning and
migratory stress. Science 194:184-
186.

Hill, Elwood F., and W. James Flem-
ing. 1982. Anticholinesterase poi-
soning of birds: field monitoring
and diagnosis of acute poisoning.
Environmental Toxicology and
Chemistry 1:27-38.

Hunt, W. Grainger, Brenda S.

Johnson, Carl G. Thelander, Brian
J. Walton, Robert W. Risebrough,
Walter M. Jarman, Alan M. Sprin-
ger, J. Geoffrey Monk, and Way-
man Walker II. 1986. Environ-
mental levels of p,p'-DDE indicate
multiple sources. Environmental
Toxicology and Chemistry 5:21-27.

Maly, Mark S. 1984. Survivorship of
meadow voles, Microtus pennsyl-
vanicus, from sewage sludge-
treated fields. Bulletin of Environ-
mental Contamination and Toxi-
cology 32:724-731.

Nomiyama, Kazuo. 1981. Renal ef-
fects of cadmium, p. 643-689. In
J.O. Nriagu, ed. Cadmium in the
environment; part 2 health effects.
Environmental Science and Tech-
nology: A Wiley Series of Texts
and Monographs. John Wiley and
Sons, Inc., New York.

O'Shea, Thomas, J., W. James Flem-
ing III, and Eugene Cromartie.
1980. DDT contamination at
Wheeler National Wildlife Refuge.
Science 209:509-510.

Petit, Michael G., and J. Scott Alten-
bach. 1973. A chronological record
of environmental chemicals from
analysis of stratified vertebrate

excretion deposited in a sheltered
environment. Environmental Re-
search 6:339-343.

Prichard, Dennis E. 1987. Managing a
national wildlife refuge for bats.
Bats, December 5:6,8.

Reich, Andrew R., Jimmy L. Perkins,
and Gary Cutter. 1986. DDT con-
tamination of a north Alabama
aquatic ecosystem. Environmental
Toxicology and Chemistry 5:725-
736.

Reidinger, Russell F., Jr., and E. Len-
dell Cockrum. 1978. Organo-
chlorine residues in free-tailed
bats (Tadarida brasiliensis) at Eagle
Creek Cave, Greenlee County,
Arizona, p. 85-96. In R.J. Olembo,
J.B. Castelino, and F.A. Mutere,
eds. Proceedings of the Fourth
International Bat Research Confer-
ence, Kenya Literature Bureau,
Nairobi.

Tuttle, Merlin D. 1979. Status, causes
of decline, and management of
endangered gray bats. Journal of
Wildlife Management 43:1-17.

United States Department of the Inte-
rior, Fish and Wildlife Service.
1987. Endangered and threatened
wildlife and plants, April 10, 1987,
50 CFR 17.11 and 17.12. United
States Government Printing Of-
fice.

White, Donald H., and Alexander J.
Krynitsky. 1986. Wildlife in some
areas of New Mexico and Texas
accumulate elevated DDE resi-
dues, 1983. Archives of Environ-
mental Contamination and Toxi-
cology 15:149-157.

413

Habitat Structure, Forest
Composition and Landscape
Dimensions as Components
of Habitat Suitability for the
Delmarva Fox Squirrel^

Raymond D. Dueser,^ James L. Dooley, Jr.,^
and Gary J. Taylor^

Abstract.— Discriminant function analysis compar-
ing 36 occupied and 18 unoccupied sites revealed
ti^ot structural variables discriminated betv\/een
sample groups better than compositional variables,
and the latter discriminated better than landscape
variables. These results ore encouraging that habitat
structure will provide a reliable basis for a predictive
classification model of habitat suitability. Such a
model would be useful both for pre-screening the
biological suitability of potential release sites and for
planning, implementing and monitoring prescriptive
habitat management.

The Delmarva fox squirrel ( Sciurus
niger cinereus) was placed on the fed-
eral endangered species list in 1967
(32 FR 4001; U.S. Department of Inte-
rior 1970). Remnant populations
were restricted to four counties in
eastern Maryland (Taylor and Flyger
1973), representing less than 10% of
the historic range of the subspecies
on the Delmarva Peninsula. Forest
clearing and habitat fragmentation
throughout the range undoubtedly
contributed significantly to the pres-
ent endangerment (Taylor 1973).

The U.S. Fish and Wildlife Service
Recovery Plan for the restoration of
the Delmarva fox squirrel to secure
status emphasizes both the reintro-
duction of this subspecies to suitable
habitats throughout the former range
and prescriptive habitat management
for established populations (Taylor et
al. 1983). A thorough understanding

'Paper presented at symposium. Man-
agement of Amphibians, Reptiles and Small
Mammals in Nortt) America. (Ragstaff. AZ,
July 19-21, 1988.)

'Raymond D. Dueser is an Associate
Professor in ttie Department of Environ-
mental Sciences, University of Virginia,
Chiarlotfesville, VA 22903.

^ James L Dooley, Jr., is a graduate re-
search) assistant in the Department of Envi-
ronmental Sciences, University of Virginia,
Charlottesville. VA 22903.

''Gary J. Taylor is Associate Director of
Wildlife. Department of Natural Resources.
Maryland Forest, Park and Wildlife Service.
Tawes State Office Building. Annapolis, MD
21401.

of habitat requirements will be essen-
tial for both initiatives (Dueser and
Terwilliger 1988).

Habitat requirements might be
expressed through any of three sepa-
rate but related components of habi-
tat suitability: forest habitat struc-
ture, forest tree species composition,
and surrounding landscape struc-
ture. Both habitat structure and for-
est composition have been shown to
influence the distribution and abun-
dance of fox squirrels in heterogene-
ous landscapes (Nixon and Hansen
1987).

Recent research has demonstrated
the potential influence of landscape
composition and structure on popu-
lations of woodland mammals occu-
pying farmland mosaics (Wegner
and Merriam 1979, Middleton and
Merriam 1983, Fahrig and Merriam

1985) . Furthermore, changes in the
landscape of the Delmarva Peninsula
almost certainly played a major role
on the decline of the fox squirrel
(Taylor 1973).

Given this background, the objec-
tive of this study was to compare the
apparent effects of habitat structure,
forest composition and landscape di-
mensions on the presence and ab-
sence of the Delmarva fox squirrel on
54 study sites in eastern Maryland.
This analysis is the first step in the
development of a predictive classifi-
cation model of habitat suitability for
this subspecies (cf. Houston et al.

1986) .

414

Methods

Data Base

During a 12-mo search for remnant
populations of the Delmarva fox
squirrel on the Maryland Eastern
Shore, Taylor (1976) located 36 "fox
squirrel present'' (Present) sites with
extant populations and 18 "fox squir-
rel absent" (Absent) sites. The gray
squirrel (Sciurus caroUnensis) was
present on all 54 sites. Taylor then
sampled the forest habitat of each
site, to compare Present forest stands
with Absent stands. He established a
representative 4 m x 200 m belt
transect on each site, on which he re-
corded the number of trees by spe-
cies per diameter-breast-height
(DBH) size class (5-20 cm, 20.1-30
cm, 30.1-50 cm, and 50.1+ cm), per-
cent crown cover, percent understory
cover, understory density, and
understory species composition. All
habitat measurements were taken
from June through September 1972
and 1973. These data formed the ini-
tial data base for this study.

Taylor (1976) reported the number j|
of trees measured in each of two size
classes: "small" trees (5-30 cm DBH)
and "large" trees (> 30 cm DBH). We
assigned each tree to one of five taxo-
nomic groups: loblolly pine (Pinus
taeda), combined oak species (Quercus
spp.), American beech (Fagus gran-
difolia), combined hickory sp>ecies
{Carya spp.), and combined mixed |

hardwoods. We estimated the ap-
proximate total basal area for each
size-taxonomic class by assuming an
average DBH of 17.5 cm for small
trees and 40.0 cm for large trees. We
then estimated total basal area for all
trees > 5 cm DBH and the fraction of
that total basal area represented by-
each taxonomic group. These basal
area estimates provide a basis for
comparing forest "composition" in-
dependently of forest "structure" as
reflected, for example, in the raw
percentage of trees counted in each
taxonomic group.

Original data were collected on
land use and cover composition of
the landscape surrounding a random
subset of Taylor's (1976) study areas
(fig. 1). Landscape variables included
area of open fields, percentage of
area forested, internal forest perime-
ter ("edge") within the sample unit,
forest shape (Blouin and Connor
1985), and distance to next nearest
woodland. These variables are re-
ferred to below as landscape "dimen-
sions." They were measured by

planimetry of 1:10(X) black-and-white
photographs (dated 1978) obtained
from Eastern Shore offices of the
USDA Agricultural Stabilization and
Conservation Service. We initially
measured each landscap)e variable
for a 2-km^ circular sample unit cen-
tered on the sample woodland. This
unit was chosen as a first-approxima-
tion of "minimal population area" on
the basis of home range size and ac-
tivity (Flyger and Smith 1980). Based
on the results of analyses for the 2-
km^ unit, both smaller (1-km^) and
larger (4-km^) sample units subse-
quently were described in the same
way.

Statistical Analyses

This comparison of habitat compx)-
nents is based on multivariate statis-
tical analyses of three separate but
related components of forest habitat
suitability: (1) habitat structure
("What does the forest 'look like' to
an observer passing through on the

Figure 1 .—Schematic diagram of 1-, 2- and 4-krTP sample units for rrjeasuring landscape
dimensions of "fox squirrel present" and "fox squirrel absent" study areas on the Eastern
Shore of Maryland.. Each sample unit was centered on one of Taylor's (1976) study areas.

ground?"), (2) tree species composi-
tion ("Which tree species predomi-
nate in this forest and give it its char-
acter?"), and (3) landscape dimen-
sions ("What are the land use and
cover dimensions of the landscape
mosaic in which this forest is embed-
ded?"). Conceptually, these compo-
nents represent a gradient of scales
from "microscopic" habitat structure
to "macroscopic" composition to
"megascopic" context.

Two-group discriminant function
analysis was used to compare the
Present and Absent forest stands
identified by Taylor (1976). Each
analysis (1) computed the univariate
F-ratio comparing Present and Ab-
sent sites for each habitat variable,

(2) tested the centroids of Present
and Absent sites for equality on the
basis of a linear combination of the
habitat variables (i.e., a linear dis-
criminant function), using multivari-
ate analysis of variance (MANOVA),

(3) indicated the relative contribution
of each habitat variable to any ob-
served difference between centroids,
based on the correlation between the
variable and the discriminant func-
tion, (4) tested the sample variance-
covariance matrices of Present and
Absent sites for homogeneity using a
Box's M test statistic, and (5) indi-
cated the percentage of the variation
in group membership (Present or
Absent) explained by the discrimi-
nant function, based on the correla-
tion between the membership vari-
able and the discriminant function.
(Dueser and Shugart 1978). All analy-
ses were computed both with and
without arcsin-square root transfor-
mations of percentage variables. Re-
sults of the parallel analyses were
qualitatively similar in each case. For
purposes of interpretability, only the
results for untransformed variables
are presented here. All analyses used
the MANOVA and DISCRIMINANT
routines of the Statistical Package for
the Social Sciences (SPSS, Nie et al.
1975).

As an unbiased test of the ability
of each set of habitat variables to

415

classify the group membership of the
study sites (i.e.. Present or Absent), a
jackknife procedure (Efron 1979) was
used to classify each of Taylor's
study areas. Each site was deleted in
sequence, DISCRIMINANT was run
for data from the remaining 53 sites,
and a classification function was
computed from these data. The de-
leted site was then classified on the
basis of this independent classifica-
tion function. The probabilistic ("pre-
dicted") classification was then com-
pared with the actual ("observed")
classification for each site.

Brennan et al. (1986) have pro-
posed an alternative solution to the
problem of habitat analysis. Logistic
regression analysis is superior to
multivariate analysis of variance
when one or more of the predictor
(e.g., habitat) variables is categorical
(i.e., non-continuous), when the vari-
ance-covariance matrices are non-
homogeneous and /or when the data
violate the assumption of multivari-
ate normality (Press and Wilson
1978). Parallel analyses demonstrate
that logistic regression analysis offers
no inherent advantage over discrimi-
nant function analysis in the present
case (Dooley, unpublished).

Results

Habitat Structure

Present sites had a greater percent-
age of trees larger than 30 cm DBH,
lower percentage shrub-ground
cover, and slightly lower understory
vegetation density than Absent sites
(table l,p< 0.05). Present and Absent
sites differ structurally on the aver-
age (MANOVA Chi-square (5) =
14.825, p < 0.011). The linear combi-
nation of structure variables ac-
counted for 26% of the variation in
group membership. The variance-
covariance matrices were marginally
homogeneous (Box's M = 20.056, p >
0.06). Percentage of trees greater than
30 cm DBH (r = -0.735), understory
vegetation density (0.564), and per-

centage shrub-ground cover (0.564)
are particularly important in dis-
criminating between sites. Conceptu-
ally, Present sites have larger trees,
less shrub-ground cover vegetation,
and less understory than Absent sites
(fig. 2). Present sites were correctly
classified 79% of the time in the jack-
knifing procedure, and Absent sites
were correctly classified 48% of the
time.

Forest Tree Species Composition

All 54 study areas supported a mix
of hard- and soft-mast tree species.
Although Present sites had some-
what greater basal areas for Ameri-
can beech {p > 0.07) and mixed hard-
woods (p > 0.05), there were no clear-
cut univariate differences between
sites in forest composition (table 2).
There also was no difference in total

Table 1.— Comparison of average forest habitat structure for "fox squirrel
present" and "fox squirrel absent" study areas on the Eastern Shore of
Maryland, based on data of Taylor (1976). Tabled values are means and
(standard deviations).

Habitat variables

Present

Absent

(N

= 36)

(N =

18)

% Trees > 30 cm DBH

32.3

(12.14)

22.1

(9.26)

<0.01

% Crown cover

75.6

(17.72)

70.6

(16.08)

>0.30

% Shrub-ground cover

51.1

(26.60)

67.5

(21.85)

<0.05

Understory "der^ity"

2.6

(1.38)

3.4

(1.04)

<0.05

% Pine composition

10.5

(10.63)

17.1

(22.23)

>0.10

Present
ceritroid

Absent
centroid

, ."Absent". .

1 1 1

1 1 1 1

1 1

-2

1 0 +1

Discriminant Score

+2

Larger trees
Sparse understory
Sparse grourxicover

SmaJler trees
Dense understory
Dense groundcover

Figure 2.— Interpretation of discrimirxition between average "fox squirrel present" and "fox
squirrel absent" study areas on ttie Eastern Stiore of Maryland, based on analysis of forest
habitat structure. Ttie tiorizontal dastied lines indicate ttie range of observations for a
sample group (Present or Absent).

416

basal area (F (1,52) = 2.300, p > 0.13).
The two types of sites were similar in
composition for both small and large
trees (fig. 3). Reflecting this similar-
ity, there was only a marginally sig-
nificant multivariate difference in
forest composition (MANOVA Chi-
square (5) = 10.584, p > 0.06).

The linear combination of compo-
sition variables accounted for 19% of
the variation in group membership.
The variance-co variance matrices
were conspicuously non-homogene-
ous (Box's M = 61.549, p < 0.001).
Present sites were correctly classified
79% of the time, and Absent sites
were correctly classified 48% of the
time.

Although the correct classification
rates were the same as for structural
variables, the two sets of variables
misclassified different sites.

Landscape Dimensions

Five landscape variables were meas-
ured for the 2-km^ circular sample
unit centered on the target woodland
of 38 of the Taylor's (1976) study ar-
eas. Present sites were somewhat
closer to the next nearest forest tract
than Absent sites (table 3,p< 0.03).
Despite this modest difference, there
was no significant multivariate dif-
ference in landscape dimensions be-
tween sites (MANOVA Chi-square
(5) = 8.574, p> 0.127).

Present and Absent woodlands
also were similar in area, averaging
9.4 and 10.0 ha, respectively, as pho-
tographed in 1978. The linear combi-
nation of landscape variables ac-
counted for 23% of the variation in
group membership. The variance-co-
variance matrices were homogene-

r

Table 2.— Comparison of average tree species composition for "fox squirrel
present" and 'lox squirrel absent" study areas on the Eastern Shore of
Maryland, based on estimated basal area (cm^ per SOO-m^ sample
transect) per taxonomlc groip. Data from Taylor (1976). Tabled values are
means and (standard deviations).

Taxonomic group

Present
(N = 36)

Absent
(N = 18)

Loblolly pine

5359

(1099,84)

7339

(2278.43)

>0.35

Oak species

9547

(1061.24)

9628

(1378.18)

>0.95

American beech

3293

(679.62)

1400

(546.56)

>0.07

Hickory

1683

(611,77)

1050

(263.26)

>0.50

Mixed hardwoods

9498

(1032.96)

6514

(690.13)

>0.05

Table 3.— Comparison of average landscape dimensions for "fox squirrel
present" and "fox squirrel absent" study areas on the Eastern Shore of
Maryland. Variables measured for 2-km^ circular sample unit centered on
study woodlarKl. Tabled vcdues are means and (standard deviations).

Landscape variables

Present

Absent

(N

= 27)

(N =

11)

Area open fields (ha)

99.3

(6.4)

96.3

(11.9)

>0.81

% Forested area

56.4

(3.6)

50.1

(6.0)

>0.35

Internal perim. (km)

5.3

(2.0)

6.2

(1.9)

>0.21

Forest "shape"

136.4

(54.5)

153.0

(44.9)

>0.38

Dist. next woodlot (km)

0.4

(0.1)

0.8

(0.2)

<0.03

ous (Box's M = 19.926, p > 0.39). As
with forest composition, there was
no consistent difference in landscape
dimensions between Present and Ab-
sent sites. Present sites were correctly
classified 787o of the time, and Ab-
sent sites were correctly classified
40% of the time.

To evaluate the possibility that the
negative result in the test for equality
of group centroids came about be-
cause we were measuring landscape
variables on an "incorrect" spatial
scale, we repeated the landscape
analysis for both smaller (1-km^) and
larger (4-km^) circular sample units,
still centered on the woodland of
interest. Again, there were no consis-
tent group differences on either scale
ip > 0.40, table 4).

Either the landscapes surrounding
the sample Present and Absent
woodlands do not differ consistently,
or they differ on a scale of measure-
ment or in a way not revealed by the
present analyses.

<

lu 10

IT
<

DBH 5-30cm

^ 8PECC8Aae£NT
■i 8PECCS CreBEKT

JLM_ii ilIl

DBH>30CM

Figure 3.— Average forest tree species
comF>ositlon of "fox squirrel present" and
"fox squirrel absent" study areas on tt»e
Eastern Shore of Maryland. "Other" cate-
gory includes a variety of snr»ail trees such
as cherry (Prunus spp.) and flowering
dogwood (Cornus florida).

417

Discussion
Present Habitat

The present habitat of the Delmarva
fox squirrel consists primarily of
relatively small stands of mature
mixed hardwoods and pines having
relatively closed canopies, relatively
open understory, and a high propor-
tion of forest edge. Occupied tracts
include both groves of trees along
streams and bays and small woo-
dlots located near agricultural fields.
In some areas, particularly in south-
ern Dorchester County, Maryland,
occupied habitat includes tracts
dominated by mature loblolly pine
located adjacent to marshes and tidal
streams. The woodland habitats now
occupied by the Delmarva fox squir-
rel are consistent with those occu-
pied by other subspecies of fox squir-
rel (Bakken 1952; Brown and Yeager
1945; Weigl et al., in press).

The picture of the Delmarva fox
squirrel that emerges from the litera-
ture is one of a species relatively
adept at utilizing a dissected, hetero-
geneous landscape dominated by ag-
riculture and woodlot forestry. Fox
squirrels are more cursorial than
gray squirrels, and often are found
on the ground several hundred me-
ters from the nearest woodlot. They
occupy larger home ranges than gray
squirrels (30 ha vs. 3 ha), travel far-
ther between captures (307 m vs. 119
m), and thus are generally more mo-
bile (Flyger and Smith 1980). Fox
squirrels more readily exploit agri-
cultural crops such com, oats, soy-
beans and fruit. They more fre-
quently utilize forest edges. Fox
squirrels would thus appear to be
relatively well-adapted to exploit the
landscape created by settlement of
the coastal plain.

One might conclude that man's
activities on the Delmarva Peninsula
should have been to the benefit of the
fox squirrel. Land clearing has cre-
ated woodlots. Grazing and burning
have opened up the understory. Ag-
riculture has increased the availabil-

ity of alternative food sources and
perhaps stabilized the food supply.
Indeed, Allen (1943) and Nixon and
Hansen (1987) indicate that settle-
ment and agriculture have worked to
the advantage of the fox squirrel
throughout the midwestern United
States, resulting in increased abun-
dance and an expanded geographic
range.

Why has this not occurred with
the Delmarva subspecies? Why has
the abundance of this fox squirrel
continued to decline throughout the
period of the recorded literature
(since approximately 1850)?

Taylor (1976) attributes the contin-
ued decline of the Delmarva fox
squirrel to habitat destruction. While
many of the landscape changes re-
sulting from settlement might have
benefited the fox squirrel, others
have been detrimental. Taylor be-
lieves that extensive timber harvest
has been particularly detrimental.
The removal of mature hardwoods
has reduced the availability of suit-
able den trees, removed reliable
sources of concentrated hard mast,
promoted the luxuriant growth of
understory vegetation, and perhaps
altered the competitive relationship
between fox and gray squirrels to
favor grays. Furthermore, coastal
plain woodland management typi-
cally has involved both short timber
rotations (i.e., frequent harvests) and
reforestation with pure stands of lob-
lolly pine. Finally, gradual urbaniza-
tion has added yet another detrimen-
tal land-use practice.

Habitat Suitability

It is assumed that Present (i.e., occu-
pied) sites are more "suitable" on the
average than are Absent (i.e., unoc-
cupied) sites. Present sites are re-
garded here as the "standard of ex-
cellence" by which to judge the habi-
tat requirements of the Delmarva fox
squirrel. Given that a number of un-
known (and unknowable) ecological,
biogeographical, and /or historical
factors may actually be responsible
for the absence of this subspecies
from any particular site within the
historic range, this assumption is cor-
rect only as a first approximation
(ref. Hanski 1982). It clearly would be
unwarranted if the distribution of
squirrels among these 54 study sites
were highly variable through time.
Nevertheless, the chance presence of
the squirrel on "unsuitable" sites and
its absence from "suitable" sites be-
cause of factors other than habitat
suitability per se can only make it
more difficult to distinguish between
Present and Absent sites. These
analyses based on presence-or-ab-
sence population information thus
circumvent many of the potential pit-
falls associated with the use of popu-
lation density as an indicator of habi-
tat suitability (Van Home 1983).

Given its present habitat, it
seemed reasonable to propose that
the capacity of a woodland to sup-
port a population of the Delmarva
fox squirrel could be determined by
habitat structure, forest composition
and/or the land use and cover com-

Table 4.— Comparison of average "fox squirrel present" and "fox squirrel
absent" study areas at the 1-, 2- and 4-square kilometer scales of obser-
vation. Testing for similarity of landscape dimensions listed In table 3.

Statistic

l-km2

2-km2

4-km2

7

27

7

9

n

9

4.791

8.574

2.750

5

5

5

0.442

0.127

0.738

Number "Present" areas
Number "Absent" areas
Chi-square
df
P

V

418

position of the surrounding land-
scape. We anticipated originally that
each of these components of habitat
suitability would prove to be impor-
tant in its own way, and that each
would have a perceptible influence
on fox squirrel presence or absence
in Maryland woodlots today. Our
results indicate, however, that habi-
tat structure is the component most
likely to contribute meaningfully to
the formulation of a predictive model
of habitat suitability. Only the struc-
ture variables discriminate strongly
between Present and Absent sites:
Present sites have larger trees, less
ground cover and less understory
(fig. 2). These variables account for
the greatest fraction of the explained
variation, their dispersion matrices
are effectively homogeneous, and
they classify sites to the correct
group (i.e.. Present or Absent) at
least as well as any of the variable
sets examined.

Forest composition is highly vari-
able among locations in eastern
Maryland, but this variation seems to
exert only a marginal influence on
the likelihood of occurrence of fox
squirrels on any given site. The com-
position variables classify sites as re-
liably as the structural variables, and
they account for only a slightly lower
fraction of the explained variation.
They do not, however, discriminate
strongly between Present and Absent
sites and their dispersion matrices
are strongly non-homogeneous. Of
course, this conclusion is based on a
comparison of two groups of sites, all
of which are known to be "squirrel
woods." Had there been a "tree
squirrel absent" category of study
area, forest composition might well
have appeared to be more significant
(cf. Nixon et al. 1978, Sanderson et al.
1976).

Landscape composition also varies
among locations, but this variation
seems not to be important on the av-
erage in discriminating between oc-
cupied and unoccupied sites today.
The landscape variables account for a
comparable fraction of the explained

variation, they classify sites almost as
reliably as the structural variables,
and their disp>ersion matrices are
homogeneous. They do not, how-
ever, discriminate meaningfully be-
tween Present and Absent sites.
Given the suggested importance of
landscape changes in bringing about
the decline of the fox squirrel, this
result was somewhat unexpected.
The correct interpretation probably
requires recognition that most of the
Eastern Shore landscape has been
altered, fragmented and homoge-
nized. Most of the remaining wood-
lands are mere remnants of forest in
a mosaic of agricultural fields, wet-
lands and suburban development.
There may simply be little important
variation remaining among these for-
est patches. At the same time, it must
be recognized that a number of po-
tentially important landscape vari-
ables— e.g., proximity to streams and
ponds (McComb and Noble 1981)
and proximity to roadways (Flyger
and Lustig 1976) — were not consid-
ered in this analysis.

Management Implications

The Recovery Plan for the Delmarva
fox squirrel calls for both the translo-
cation of squirrels to suitable habitats
throughout the historic range and the
maintenance of occupied habitat
(Taylor et al. 1983). Will objective,
quantitative habitat analysis be help-
ful in evaluating potential release
sites and planning prescriptive habi-
tat management? Results of the
analyses presented here provide
some basis for optimism. A number
of management implications follow
from these results:

1. Of the variable sets exam-
ined, habitat structure is the
best indicator of biological
habitat suitability for the Del-
marva fox squirrel at the
present time. Even this mini-
mal list of structure variables
(table 1) has the power to

discriminate meaningfully
between occupied and unoc-
cupied forest stands. Present
sites have larger trees, less
ground cover, and less
understory than Absent sites.
Significantly, these results
corroborate the general habi-
tat descriptions rep>orted by
Flyger and Lustig (1976).

2. In addition to this clear-cut
discrimination, the structure
variables exhibit the most
desirable combination of sta-
tistical properties, including
the highest variance explana-
tion, homogeneity of disper-
sions, and high correct classi-
fication rates. These proper-
ties will simplify the formu-
lation of a predictive classifi-
cation model of habitat suita-
bility.

3. Although the absence of
meaningful discriminating
information in forest compo-
sition and landscape dimen-
sions was somewhat surpris-
ing, these results have the
effect of simplifying the ef-
fort to quantify habitat suita-
bility for the Delmarva fox
squirrel. It would be impru-
dent to disregard forest com-
position and landscape at-
tributes in the evaluation of
{X)tential release sites; these
components of habitat suita-
bility must be imjx)rtant at
some level (Ryger and Lus-
tig 1976). There appears to be
little potential, however, for
the variables analyzed here
to contribute to a predictive
model of habitat suitability.

4. The discriminating structure
variables are easy and rela-
tively inexpensive to meas-
ure. Including site reconnais-
sance, approximately one-
half day of field time is re-
quired for a team of two ex-

419

perienced observers to col-
lect a Taylor-type data set.

5. It should therefore be practi-
cal to pre-screen potential
release sites for habitat suita-
bility relative to Present sites.
Objective pre-screening has
not always been possible be-
cause no "standard of excel-
lence" has been available.

6. It also should be practical to
plan, implement and evalu-
ate prescriptive habitat man-
agement for the benefit of the
Delmarva fox squirrel on oc-
cupied sites or potential re-
lease sites. The important
measures of habitat structure
(e.g., understory vegetation
density) tend to be variables
which are amenable to
silvicultural manipulation
(Nixon etal. 1980).

Conclusions

We anticipated at the outset that each
of three potentially important com-
ponents of habitat suitability — forest
habitat structure, forest tree species
composition, and surrounding land-
scape dimensions — would influence
the present occurrence of the Del-
marva fox squirrel in forest stands on
the Eastern Shore of Maryland. The
analyses reported here produced a
number of surprises.

Habitat structure is the only com-
ponent that both discriminates be-
tween occupied and unoccupied sites
in a meaningful way and exhibits a
combination of statistical properties
favorable for the formulation of a
predictive classification model of
habitat suitability.

The analysis of habitat structure
provides a basis for optimism that
such a model would prove useful
both for pre-screening potential re-
lease sites and for planning, imple-
menting and monitoring prescriptive
habitat management.

Acknowledgments

This study would not have been pos-
sible without the cooperation of
those who assisted G. J. Taylor in the
identification of remnant populations
of the Delmarva fox squirrel on the
Maryland Eastern Shore. G. D. Ther-
res and G. W. Willey, Sr., of the
Maryland Department of Natural
Resources assisted with locating sites
for the landscape analysis. J. H. Por-
ter assisted with air photo interpreta-
tion and with the statistical analyses.
J. Peatross prepared the figures, and
L. M. McCain assisted with prepar-
ing the manuscript. V. Flyger and K.
E. Severson provided thoughtful re-
views of the manuscript. Finally, the
owners of the Maryland study sites
generously provided access to their
property for purposes of habitat
characterization. This study was sup-
ported by funding from the
Nongame and Endangered Species
Program of the Virginia Department
of Game and Inland Fisheries and by
the Virginia Coast Reserve Long-
Term Ecological Research Program
(NSF Grant BSR-8702333).

Literature Cited

Allen, D. L. 1943. Michigan Fox
Squirrel Management. Game Divi-
sion Publication 100. Lansing,
Michigan. 404 p.

Bakken, Arnold. 1952. Interrelation-
ships of Sciurus caroUnensis (Gme-
lin) and Sciurus niger (Linnaeus) in
nnixed populations. Ph.D Thesis,
University of Wisconsin. 212 p.

Blouin, Michael S., and Edward F.
Connor. 1985. Is there a best shape
for nature reserves? Biological
Conservation 32: 277-288.

Brennan, Leonard A., William M.
Block, and R. J. Gutierrez. 1986.
The use of multivariate statistics
for developing habitat suitability
index models, p. 177-182. In J.
Verner, M. I. Morrison and C. J.
Ralph, eds. Wildlife 2000: Model-
ing Habitat Relationships of Ter-

restrial Vertebrates. U. Wisconsin
Press, Madison. 470 p.

Brown, Louis G., and Lee E. Yeager.
1945. Fox squirrels and gray squir-
rels in Illinois. Illinois Natural His-
tory Survey Bulletin 23:449-536.

Dueser, Raymond D., and H. H.
Shugart, Jr. 1978. Microhabitats in
a forest-floor small mammal
fauna. Ecology 59:89-98.

Dueser, Raymond D., and Karen Ter-
williger. 1988. Status of the Del-
marva fox squirrel (Sciurus niger
cinereus) in Virginia. Virginia Jour-
nal of Science 38(4):380-388.

Efron, Bradley. 1979. Computers and
the theory of statistics: thinking
the unthinkable. SI AM Review
21:460-480.

Fahrig, Lenore, and G. Merriam.
1985. Habitat patch connectivity
and population survival. Ecology
66:1762-1768.

Hyger, Vagn, and Loren W. Lustig.
1976. The potential for re-estab-
lishing fox squirrels in portions of
their former range in the north-
eastern states. Transactions of the
Northeast Section of the Wildlife
Society 33:13-17.

Flyger, Vagn, and Deborah A. Smith.
1980. A comparison of Delmarva
fox squirrel and gray squirrel
habitats and home range. Transac-
tions of the Northeast Section of
the Wildlife Society 37:19-22.

Hanski, 1. 1982. Dynamics of regional
distribution: the core and satellite
species hypothesis. Oikos 38:210-
221.

Houston, B. R., Tim W. Clark, and S.
C. Minta. 1986. Habitat suitability
index model for the black-footed
ferret: a method to locate trans-
plant sites. Great Basin Naturalist
Memoirs 8:99-114.

McComb, William C, and Robert E.
Noble. 1981. Nest-box and natu-
ral-cavity use in three mid-South
forest habitats. Journal of Wildlife
Management 45:93-101.

Middleton, J., and G. Merriam. 1981.
Woodland mice in a farmland mo-
saic. Journal of Applied Ecology
18:703-710.

420

Nie, Norman H., C. Hadlai Hull, Jean
G. Jenkins, Karen Steinbrenner,
and Dale H. Brent. 1975. SPSS: Sta-
tistical Package for the Social Sci-
ences, 2nd ed., McGraw-Hill, N.Y.
675 p.

Nixon, C. M., and L. P. Hansen. 1987.
Managing forests to maintain
populations of gray and fox squir-
rels. Illinois Dept. of Conservation
Technical Bulletin 5. 35 p.

Nixon, C. M., S. P. Havera, and R. E.
Greenberg. 1978. Distribution and
abundance of the grey squirrel in
Illinois. Illinois Natural History
Survey Biological Notes 105. 55 p.

Nixon, C. M., S. P. Havera, and L. P.
Hansen. 1980. Initial response of
squirrels to forest changes associ-
ated with selection cutting. Wild-
life Society Bulletin 8:298-306.

Press, S. James, and Sandra Wilson.
1978. Choosing between logistic
regression and discriminant analy-
sis. Journal of the American Statis-
tical Association 73:699-705.

Sanderson, H. Reed, William M.
Healy, James C. Pack, John D. Gill,
and Jack Ward Thomas. 1976.
Gray squirrel habitat and nest-tree
preference. Proceedings of the
Annual Conference of the South-
eastern Association of Game and
Fisheries Commissions 29:609-616.

Taylor, Gary J. 1973. Present status
and habitat survey of the Del-
marva fox squirrel (Sciurus niger
cinereus) with a discussion of rea-
sons for its decline. Proceedings of
the Annual Conference of the
Southeastern Association of Game
and Fisheries Commissions 27:278-
289.

Taylor, Gary J. 1976. Range determi-
nation and habitat description of
the Delmarva fox squirrel in
Maryland. M.S. Thesis, University
of Maryland, College Park. 76 p.

Taylor, Gary J., and Vagn Flyger.
1973, Distribution of the Delmarva
fox squirrel (Sciurus niger cinereus)
in Maryland. Chesapeake Science
14:59-60.

Taylor, Gary J., Vagn Flyger, N.
Swink, and G. Willey. 1983. Del-

marva fox squirrel recovery plan.
First revision. U.S. Fish and Wild-
life Service. Washington, D.C.
49 p.

U.S. Department of Interior, Bureau
of Sport Fisheries and Wildlife.
1970. Endangered Species of the
United States. Washington, D.C.
4 p.

Van Horne, B. 1983. Density as a mis-
leading indicator of habitat qual-
ity. Journal of Wildlife Manage-
ment 47:893-901.

Wegner, J.F., and G. Merriam. 1979.
Movements by birds and small
mammals between a wood and
adjoining farmland habitats. Jour-
nal of Applied Ecology 16:349-357.

Weigl, P. D., M. A. Steele, L. J. Sher-
man, J. C. Ha, and T. S. Sharpe. In
press. The ecology of the fox
squirrel (Sciurus niger) in North
Carolina: Implications for survival
in the southeast. Tall Timbers Re-
search Series.

421

Effects of Treating
Creosotebush With
Tebuthiuron on Rodents^

William G. Standley^ and Norman S. Smith^

Abstract.— Three years after creosotebush (Lauea
tridentata) v/os treated with tebuthiuron, rodent
abundance was 71% higher on treated plots than
on controi plots in southeastern Arizona. Arizona cot-
ton rats (Sigwodon arizonae) and Western harvest
mice (Reifhrodontomys megalotis)v^ere more abun-
dant while the abundance of Merriam's kangaroo
rots (Dipodomys merriami) similar. We conclude
that tebuthiuron may be safely used to control creo-
sotebush in semidesert grasslands unless the pres-
ence of rare or endangered species precludes any
alterations to the community.

Herbicides are often used to control
shrubs such as mesquite (Prosopis ju-
Uflora) and creosotebush (Larrea
tridentata), which have invaded mil-
lions of hectares of semidesert grass-
lands (Cox et al. 1982). The reduction
of shrub cover usually results in an
increase in forage production (Box
1964).

The herbicide 2,4-D has been used
for more than two decades and its
effects on rodent communities have
been extensively studied (Keith et al.
1959, Johnson and Hansen 1969,
Spencer and Barrett 1980). 2,4-D has
varied effects on rodent communi-
ties, increasing the abundance of
some species, while decreasing the
abundance of others (Johnson and
Hansen 1969, Spencer and Barrett
1980).

Tebuthiuron is a thiadiazolyl-urea
herbicide (Walker et al. 1973) used to
control shrubs in the southwest (Her-
bel et al. 1985). No studies have been
conducted to determine effects of
tebuthiuron treatments on rodent

'Paper presented at symposium. Man-
agement of Amphibians. Reptiles, and
Small Mammals in Northi America. (Flag-
staff. AZ. July 19-21. 1988.)

'William G. Standley. formerly a gradu-
ate student. University of Arizona. Arizona
Cooperative Rshi and Wildlife Research»
Unit, is currently Animal Ecologist. EG&G
Energy Measurements. Inc.. c/o NPR-1. P.O.
Box 127. Tupman. CA. 93276.

^Norman S. Smitti is Assistant Leader. Ari-
zona Cooperative Fist^ and Wildlife Re-
search! Unit. University of Arizona. Tucson. AZ
85721.

communities. We studied grasslands
invaded by creosotebush in south-
eastern Arizona in order to deter-
mine changes that take place in a ro-
dent community due to treatment
with tebuthiuron. We compared
vegetation and nocturnal rodents
present on control and treated plots.
Because tebuthiuron is nontoxic to
laboratory mice, rats, and rabbits
(Morton and Hoffman 1976) we as-
sumed that any changes in the rodent
community would be in response to

changes in food supply, ground
cover, or both.

Methods

Two adjacent 150 x 600 m plots were
fenced from cattle, and one was aeri-
ally treated with tebuthiuron (1.0 kg/
ha) in May 1981 as part of an ongo-
ing experiment on the USD A Forest
Service Santa Rita Exp)erimental
Range, 45 km south of Tucson Ari-

r

Table L— Mean (%) vegetative cover on tebuthiuron treated and control
plots (N=6).

Treated

Control

Species

SE

SE

Grasses
Threeawn

(Aristida sp.)
Bush muhly

(MuNenbergia porter!)
Fluff grass

(Tridens pulchellus)
Other

Total

Shrubs

Creosotebush

(Larrea iridentafa)
Mesquite

(Prosopis juliflora)
Desert zinnia

(Zinnia pumila)
Desertbroom

(Baccharis sarofhroides)

Total

18.5

3.2

0.1

0.1

10.0

2.5

11.0

3.4

3.0

1.3

0.7

0.7

0.3

0.3

0.0

0.0

31.8

1.9

11.8

3.2

0.2

0.1

33.9

4.6

0.0

0.0

1.4

0.9

0.0

0.0

3.3

1.4

0.5

0.5

0.0

0.0

0.7

0.5

38.6

4.5

422

zona. Vegetation on the plots is
dominated by creosotebush, with
sparse grasses such as threeawn
{Aristida sp.) and bush muhly
(Muhlenbergia porteri) (Martin and
Reynolds 1973).

We sampled vegetation and ro-
dent communities in June 1984, three
years after herbicide treatment.
Vegetation was sampled using the
line intercept method (Canfield
1941). Six 30 m parallel lines were
systematically located on each plot.
Total intercepts of each species were
averaged and transformed into per-
cent ground and canopy cover. Ro-
dent communities were surveyed us-
ing the removal method. Sherman
live-traps (7.5 x 7.5 x 25 cm) were
used so that rodents could be used
for other studies. Three 8 by 8 grids
with traps spaced at 10 m intervals,
were placed on each plot. Grids were
placed as far from each other and
from plot boundaries as possible, re-
sulting in a uniform distribution.
Traps, opened at sunset and closed at

sunrise, were baited with p>eanut
butter and oats. We prebaited traps
for one night then removed all ro-
dents captured during the following
four nights. The total number of each
species captured on the three grids
on each plot were averaged.

Results

Average grass cover on the tebuth-
iuron-treated plot was almost three
times that on the control plot (table
1), with threeawn contributing most
of the difference. Average shrub
cover on the treated plot was 98%
lower than on the control plot, with
creosotebush accounting for the big-
gest difference.

On tebuthiuron-treated grids we
captured 162 rodents of eight species,
and on control grids 95 rodents of
eight sf>ecies (table 2). Higher num-
bers of Arizona cotton rats (Sigmodon
arizonae) and western harvest mice
(Reithrodontomys megalotis) on the

r

Table 2.— Mean number of rodents captured on tebuthiuron treated and
control plots (N=3).

Treated

Control

Species

SE

SE

Merriam's kangaroo rat

(DIpodomys m erriami)
Arizona pocket mouse

(Perognafhus amplus)
White-throated wood rat

(Neotoma olbiguta)
Western harvest mouse

(Reifhrodonfomys megalotis)
Arizona cotton rat

(Sigmodon arizonae)
Desert pocket mouse

(Perognafhus penicillatus)
Southern grasshopper mouse

(Onychomys torridus)
Bailey's pocket mouse

(Perognatfius baileyi)
Deer mouse

(Peromyscus maniculatus)
House mouse

(Mus musculus)

Total

15.7
5.3
8.0

10.0
8.0
3.0
3.7
0.0
0.0
0.3

54.0

4.3
2.6
1.5
1.5
3.2
0.0
0.3
0.0
0.0
0.3
5.5

13.0
7.0
4.3
0.3
0.0
4,0
1.7
1.0
0.3
0.0

31.6

2.1
0.6
1.5
0.3
0.0
1.0
0.7
1.0
0.3
0.0
2.7

treated grids accounted for most of
the difference in abundance. Cotton
rats and house mice (Mus musculus)
were captured only on the treated
grids, while Bailey's pocket mice
(Perogmthus baileyi) and deer mice
(Peromyscus maniculatus) were caught
only on control grids.

Discussion

The dramatically greater grass cover
and lesser shrub cover on the treated
plot are consistent with results of
other experiments with tebuthiuron
(Herbel et al. 1985), as well as with
2,4-D (Spencer and Barrett 1980).
This difference in vegetative struc-
ture app>ears to account for most of
the differences in the rodent commu-
nity. Studies of cotton rats and har-
vest mice have shown that both spe-
cies are strongly associated with
dense stands of grass (Goertz 1964,
Ford 1977). The similarity in abun-
dance of Merriam's kangaroo rats on
control and treated plots was unex-
f)ected since heteromyids are gener-
ally more abundant in areas with
sparse ground cover (Stamp and
Ohmart 1978).

We do not present inferential sta-
tistics to test differences in ground
cover or rodent numbers because
both the line intercept transects and
trap grids were actually subsamples
rather than true replicates (Hurlbert
1984). We are convinced, however,
that differences between plots in
numbers of cotton rats and harvest
mice, are the result of habitat
changes following treatment with
tebuthiuron.

Because of the low numbers of
deer mice and Bailey's pocket mice
captured on the control plots, we do
not feel their absence on the treated
plots is significant. Because the re-
sponses were either neutral or posi-
tive, we feel that tebuthiuron can be
safely used by managers to control
shrubs in semidesert grasslands
without fear of endangering rodents
directly. However, the impact of

423

habitat changes on rare or endan-
gered species should not be ignored.

Acknowledgments

This study was funded by the USDA
Arid Land Ecosystems Improvement
Unit. We thank J. Ard, J. Brown, B.
Kotler, and B. Zoellick for field assis-
tance.

Literature Cited

Box, Thadis W. 1964. Changes in
wildlife habitat composition fol-
lowing brush control practices in
south Texas. Transactions of the
North American Wildlife Confer-
ence 29:432-438.

Canfield, R. H. 1941. Application of
the line interception method in
sampling range vegetation. Jour-
nal of Forestry 39:388-394.

Cox, Jerry R., Howard L. Morton,
Thomas N. Johnsen Jr., Gilbert L.
Jordan, S. Clark Martin, and Louis
C. Fierro. 1982. Vegetation restora-
tion in the Chihuahuan and Sono-
ran deserts of North America.
United States Department of Agri-
culture, Agricultural Research
Service, Agricultural Reviews and
Manuals, Western Series Number
28, August 1982.

Ford, Steven D. 1977. Range, distri-
bution and habitat of the western
harvest mouse, Reithrodontomys
megalotis, in Indiana. American
Midland Naturalist 98:422-432.

Goertz, John W. 1964. The influence
of habitat quality upon density of
cotton rat populations. Ecological
Monographs 34:359-381.

Herbel, Carlton H., Howard L. Mor-
ton, and Robert P. Gibbens. 1985.
Controlling shrubs in the arid
southwest with tebuthiuron. Jour-
nal of Range Management 38:391-
394.

Hurlbert, Stuart H. 1984. Pseu-
doreplication and the design of
ecological field experiments. Eco-
logical Monographs 54:187-211.

Johnson, Donald R. and Richard M.
Hansen. 1969. Effects of range
treatment with 2,4-D on rodent
populations. Journal of Wildlife
Management 33:125-132.

Keith, James O., Richard M. Hansen,
and A. Lorin Ward. 1959. Effect of
2,4-D on abundance and food of
pocket gophers. Journal of Wild-
life Management 23:137-145.

Martin, S. Clark, and Hudson G.
Reynolds. 1973. The Santa Rita
Experimental Range: Your facility
for research on semidesert ecosys-
tems. Journal of the Arizona Acad-
emy of Science 8:56-67.

Morton, D. M., and D. G. Hoffman.
1976. Metabolism of new herbicide
tebuthiuron (l(5-(l,l-dimethyl-
l,3,4-thiadiazol-2-yl)-l,3-dimethyl-
urea) in mouse, rat, rabbit, dog,
duck, and fish. Journal of Toxicol-
ogy and Environmental Health
1:757-768.

Spencer, Stephen R. and Gary W.
Barrett. 1980. Meadow vole popu-
lation response to vegetational
changes resulting from 2,4-D ap-
plication. American Midland
Naturalist 103:32-46.

Stamp, Nancy E., and Robert D.
Ohmart. 1978. Resource utilization
by desert rodents in the lower
Sonoran desert. Ecology 59:700-
707.

Walker, J. C, M. L. Jones, and J. E.
Shaw. 1973. Total vegetation con-
trol with tebuthiuron — a new
broad spectrum herbicide. Pro-
ceedings of the North Central
Weed Control Conference 28:39.

424

Foraging Patterns of Tassel-
Eared Squirrels In Selected
Ponderosa Pine Stands^

Jack S. States,^ William S. Gaud,^ W.
Sylvester Allred/ and William J. Austin^

Abstract.— Pine seed, primarily available in the fall,
and hypogeous fungi, potentially available in all
seasons, were major food items whose consumption
was associated with an increase in biweekly body
weights of marked squirrels. Use of alternative foods
such OS twigs (inner bark) and apical buds occurred
when these food items were unavailable. Consump-
tion of inner bark and buds was highest in winter
(93%) and spring (86%). Although feed tree prefer-
ence was noted, widespread feeding occurred in
more than half of the trees in both study sites. The
resulting variability in physical evidence of foraging
suggests caution in its use for indexing squirrel popu-
lation densities.

bark in the absence of supplemental
foods could potentially threaten
squirrel survival during adverse
weather conditions (Patton 1974).
The obvious selection of certain trees
by squirrels for inner bark consump-
tion sup]X)rts the assumption that
there are differences in quality of
trees in the same stand. In a food
preference study using captive Abert
squirrels, Farentinos et al. (1981)
were able to show a significant rela-
tionship between selective consump-
tion of inner bark and low oleoresin
content. However, Pederson and
Welch (1987) noted a strong feeding
preference for trees with inner bark
that was easily peeled with no appar-
ent relationship between inner bark
oleoresin content and "feed tree" se-
lection.

Studies on the impacts of squirrel
herbivory on ponderosa pine have
lead to mixed conclusions. Hall
(1981) and Ffolliott and Patton (1978)
found that heavy utilization of pine
twigs had negligible effect on stand
productivity, although Hall demon-
strated significant growth decreases
of individual feed trees. Soderquist
(1987) reported twig clipping to de-
crease tree growth in ecotonal stands
of ponderosa pine. Pearson (1950)
and Larson and Schubert (1970)
noted extensive, but seasonally vari-
able, damage to cone crops. They
were unable to determine the causes
of the highly variable pattern of her-
bivory during several years of obser-
vation.

The tassel-eared tree squirrel (Sciurus
aberti) and its several subspecies, has
a unique and apparent obligatory as-
sociation with Southwestern ponder-
osa pine (Pinus ponderosa). The die-
tary dependence of these squirrels on
pine, including inner bark and buds
of terminal twigs and both staminate
and ovulate cones, identifies the
squirrel as an herbivore having a po-
tentially major influence on the
growth and reproduction of ponder-
osa pine. Conversely, extensive har-
vest of pine for wood products has
resulted in deterioration of the squir-
rel's habitat since the turn of the cen-
tury (Keith 1965).

A number of studies have at-
tempted to explain the '"boom and
bust" p>opulation fluctuations that
seem to be characteristic of tassel-
eared squirrels. In his observations

'Paper presented at symposium, f\/1an-
agement of Amphibians, Reptiles, and
Small Mammals in Nortti America. (Flag-
staff, AZ, July 19-21, 1988.)

^Jack S. States is Professor of Biology,
Department of Biological Sciences, Noriti-
em Arizona University, Flagstaff, AZ 8601 1.

^William S. Gaud is Associate Professor of
Biology, Department of Biological Sciences,
Northern Arizona University. Ragstaff, AZ
86011.

^W. Sylvester AJIred is a doctoral candi-
date in the Department of Biological Sci-
ences, Northern Arizona University, Flagstaff,
AZ86011.

^William J. Austin is a doctoral candi-
date in the Department of Biological Sci-
ences, Northern Arizona University. Flagstaff,
AZ86011.

on the ecology of Abert squirrels (S.
a. aberti), Keith (1965) attributed short
term fluctuations to changes in quan-
tity and quality of major food items
assumed to be provided by pine.
However, high mortality in some
years apparently resulted from some
factors other than food. Hall (1981),
in a study of Kaibab squirrels (S. a.
kaibabensis) also observed population
fluctuations. He suggested that sea-
sonal differences in food resources
and snowfall were potential causes of
declines and recovery. Availability
and use of various food items have
not been adequately studied.

Stephenson (1974) in a study of
Abert diets discovered that fungi
were a major part of the total food
consumed. The fungi in Stephenson's
samples were identified as belonging
to a subterranean group of mush-
rooms popularly called truffles (].
States, unpublished data), which are
known to form mycorrhizal associa-
tions with pine roots. Some of these
fungi were found to be new records
for the Southwest (States 1983, 1984)
and they were found to be primarily
associated with blackjack age-class
ponderosa pine stands with high can-
opy densities (States 1985).

A telltale sign of the activity of the
tassel-eared squirrel is the presence
of clipped twigs on the ground under
a tree after the squirrel has removed
the terminal shoot from a branch.
The nutritional value of the inner
bark consumed by the squirrel is
low. A diet comprised solely of inner

425

The tassel-eared squirrel's variety
of diet and use of the forest has lead
to differences of opinion regarding
the best management plan for both
squirrel and forest. Patton et al.
(1985) considered tree density, size,
and patchy distribution to be major
factors constituting habitat quality
since squirrels use pine for cover and
nesting as well as for food.

The purpose of this study was to
determine the seasonal patterns of
food resource utilization by Abert
squirrels in selected ponderosa pine
stands and to relate the results to
squirrel population levels within the
stands. The use of fungi and inner
bark as major food items is discussed
as it pertains to stand characteristics
and the potential impact of herbivory
on ponderosa pine.

Study Areas and Methods

Two sites in clumped, uneven-aged
ponderosa pine stands were studied
in areas that had not been disturbed
by fire and timber harvest for the
past 35 years. A 9.3 ha site was lo-
cated on the property of the Lowell
Observatory and adjacent to the Co-
conino National Forest. The other site
of 2.5 ha was located in the Mount
Elden Environmental Study area of
the Coconino National Forest. The
elevation of both sites was 2150 m,
and they were within 10 km of the
Flagstaff airport where weather data
used in the study was collected
(NOAA 1987).

Squirrels were captured, marked,
and released at the observatory site.
At this location there were 90 plots,
625 m^ each, in a nested trapping
grid similar to the one described by
Patton et al.(1985). The grid con-
tained 42 systematically spaced
Tomahawk live-traps baited and set
for eight daylight hours once each
week. Trapping was conducted from
September 1985 through June 1987
and squirrel body weight was re-
corded. Fecal pellets deposited in
traps were collected and analyzed to

{ \

Table 1 .—Estimated food availability for Abert squirrels.

Truffles Seed cones/tree Acorns/tree Mushroom
Year kg/ha xn=25 xn=20 abundarice

1983 2.88 144 552 high

1984 0.86 10 83 low

1985 0.39 10 123 very low

1986 0.72 137^ 20^- very high

1987 0,65 10 10»^ moderate

°Cone crop 2}% aborted due to insect damage; 31% of total cone crop harvested
by September.

^Acorn production low due to early frost.

summer {a\\

BA -branch inn«rbark Ml-mistletoe SC -staminate cones

BU -terminal buds MU-mushrooms TR -truffles

TW-twig innerbaric

co-pine cones OP -open pine cones

1

3 winter spring

Figure 1 .—Percentage of feeding time by Abert squirrels for each diet item in each season
(from eighiteen two-hour periods per secjson, Coconino National Forest, AZ).

426

determine the ratio of dietary fungus
to plant matter and to identify the
fungi through spore characteristics.

Observational data on foraging
behavior was collected using focal
animal sampling (Altmann 1974). We
observed four individually marked
males from July 1986 through No-
vember 1987 as they foraged in the
study site. Data were collected in 18
two-hour observation periods in each
of four "ecological" seasons. These
seasons were established by combin-
ing months with similar temperature
and precipitation means. The seasons
correspond to periods of truffle pro-
duction: Season I, December-March
(winter); Season II, April-June

(spring); Season III, July- August
(summer); Season IV, September-
November (fall).

Resource availability and physical
signs of foraging acrivity were re-
corded over a 20-month period (June
1986 through January 1988). Cone
consumption, twig clipping, and dig-
ging activity were documented on 26
contiguous 625 m^ plots (1.6 ha) on
the observatory site. During each of
the biweekly censuses, all twig
remnants, cone cores, and digs
(truffle excavations) were recorded
with a notation of the nearest tree.
Trees were characterized by age-class
(blackjack, or yellow pine) and di-
ameter at breast height (DBH). The

entire 2.5 ha Mount Elden site was
sampled for seasonal foraging pat-
terns. Permanently marked pines and
oaks in 20 plots were censused yearly
in September for acorn and cone pro-
duction. Cones and acorns were
counted on a quarter of the tree and
multiplied by four to obtain a pro-
duction estimate. Truffle production
estimates were made according to
States (1985). Relative mushroom
abundance was determined by
counting numbers of mushrooms
present within 10 randomly placed 50
m^ quadrats sampled in the fall.

Results

Resources available as food for tas-
sel-eared squirrels showed consider-
able annual variability (table 1). The
four food items were all relatively
abundant in 1983, but availability
subsequently dropped. Production of
cones and mushrooms was relatively
high again in 1986, but there were
considerably fewer truffles and
acorns present than in 1983. In gen-
eral, the quantity of truffle produc-
tion was more consistent that it was
for the other foods.

Seasonal foraging behavior of the
squirrels (fig. 1) reflected changes in
availability of food items. The ani-
mals heavily utilized a large cone
crop in 1986 before the seeds were
mature, and continued to utilize it
into November when the remaining
seeds were released. Cone and truffle
use dropped abruptly from a fall
high of 80% feeding time when the
squirrels switched to intensive feed-
ing on buds and inner bark of twigs.
This behavior comprised 85% of the
spring feeding time. Collectively,
pine products constituted the largest
portion of the diet through winter
and spring. Seasonal patterns were
apparent in the use of different parts
of the tree.

The physical evidence left in the
forest by the squirrels verified a sea-
sonal progression of food item
availability (fig. 2). Numbers of digs

2400 n
2200-
2000-
3 1600
^ 1600-

o

i

TWIGS CUPPED CONES HARVESTED KUICBEK OF DIGS

BO

70
00
50
40
30

g 20

10
0

m

mm

JJASONDJFMAUJJASONDi
1066 1067 1066

MONTHS

Figure 2.— Monthly resource usage by Abert squirrels compared to total precipitation (solid
bar) and snow deptti (steaded bar), Coconino Natiorral Forest, AZ.

427

peaked in late fall and dropped in
the winter, a pattern which corre-
sponds to foraging time percentages
for truffles (fig. 1). Subsequent digs
during winter and early spring repre-
sented retrieval of cones buried the
previous fall. Numbers of cone cores
left after seed removal increased as
the cones matured. Numbers of cone
cores susequently decreased in the
winter months. As the use of cones
and truffles declined, numbers of
clipped twigs increased. Twig clips
decreased to a moderate level in
summer but again reached a high
level the following winter. This peak
was coincident with increased snow
depth, a decrease in availability of
truffles, and the absence of seed
cones. By the end of January 1988, a
majority of the 1114 trees (67.9%) in
the observatory site had been clipped
at least once with an average of al-
most 10 clips per tree.

In spite of relatively low food
availability in the year from Septem-
ber 1985 through August 1986 (table
1), resident squirrels maintained a
fairly constant weight throughout the
winter (table 2). The average weight
of four male squirrels dropped 6%
during spring and early summer
from the previous fall's high. Subse-
quent weight gains, 5%, occurred
concurrently with maturation of the
1986 fall cone crop. Winter weight
loss paralleled the decrease in availa-
bility of fungi, as evidenced by their
presence in fecal contents (fig. 3).

The number of terminal shoots
removed by squirrels in the Mt.
Elden site from 1984-1987 was appar-
ently related to snowfall (table 3).
Squirrel densities remained relatively
constant, but the number of trees
clipped increased by 30% and the av-
erage number of clips per tree de-
creased by 81%. Total snowfall was
greater in 1985 than in either 1986
and 1987. The map of clipping behav-
ior shows marked shifts in areas of
heaviest clipping in the 2.5 ha site
(fig. 4). The smallest area of clipping
intensity (46% of the site) occurred in
1985, and it also had the highest

Table 2.— Mean body weight (grams) and standard deviation of Individual
male squirrels In each season from September 1985 through November
1986. The numbers of captures per season are in parentheses.

Squirrel

IV
Fall

Winter

!l

Spring

III

Summer

IV
Fall

1

2
3
4

614+ 11.7

(11)
748 + 37.0

(5)
655 + 20.4

(13)
737 + 24.7

(6)

689

632 + 30.1
(12)

680 + 34.3
(7)

681 + 15.2
(12)

688 + 31.5
(5)
670

610+ 12.7
(3)
723 +

(1)
671 + 18.0
(7)
669

(1)
668

587 + 27.9

(7)
703+ 18.6

(5)
648 + 29.8

(5)
662 + 33.6

(7)

650

610 + 27.6

(9)
769 + 20.9

(5)
674 + 16.9

(8)
674 + 33.8

(11)
682

number of clips per tree. The area of
clipping expanded in 1986 and 1987
to 67% and 64%, respectively. Of the
604 trees clipped in the three years,
45% were clipped once while 23%

were clipped every year. Yellow pine
constituted 10% of the stand and 82%
of these were clipped. Most yellow
pines not clipped were isolated in
open areas. Sixty-one percent of all

FALL WINTER SPRING SUMMER

Figure 3.— Weight loss (solid line) by Abert squirrels as compared to fungal content of feces
(dotted line) during the period September 1985 to August 1986. Coconino National Forest,
AZ.

Table 3.— Twig clipping data for Yellowpine (VP) and Blackjack (BJ) age-
class trees over ttiree successive winter seasons In the Mount FIden study
area.

Snow

SquirrelNumber Trees

Clipped

Clips/

Total

Dry wt.

Year

cm.

number YP

BJ

total

tree

clips

kg

1984-85

345.4

6 38

155

193

124.7

24,061

295

1985-86

266.7

8 52

268

320

65.4

20,640

253

1986-87

217.7

6 58

533

591

24.2

14,288

175

428

blackjack pine with a DBH greater
than 10 cm were cHpped. Eleven of
these and one yellow pine died fol-
lowing virtually total canopy re-
moval by squirrels. The average
number of twigs clipped in yellow
pine was greater than in blackjack
pine but their mean dry weight, 11.2
• 8.7 g, was less than that of blackjack
pine, 13.2 • 4.8 g. Five of the yellow
pines in 1985 had more than 1000
twigs removed from each. A majority
of the trees clipped only once before
1987 were also clipped in 1987 and
were located in an area with little
previous squirrel use (fig. 4).

Discussion

The tassel-eared squirrel is a whole
forest species in the sense that essen-
tially all age classes of trees are util-
ized. Although pine provides much
of the squirrel's food, the various
items are taken by the squirrel from
different age classes of trees (table 3).
The largest number of cones is pro-

Figure 4.— Map of the 2.4 ha Mount Elden
study site illustrating shifts in clipping inten*
sity for each of three years (spring to
spring), Coconino National Forest, AZ. Clip-
ping data corresponding to these areas is
presented in table 3.

duced by mature yellow pines (Lar-
son and Schubert 1970), while
truffles tend to be associated with
pole sized blackjack pines (States
1985). Thus, prime squirrel habitat
provides optimal food in stands con-
taining a combination of tree age
classes whose size, density, and
grouping provides cover and nesting
sites as well (Patton 1984).

Major shifts in foraging by the tas-
sel-eared squirrel are apparently as-
sociated with variations in the availa-
bility of food resources in the forest.
In 1986 squirrels relied heavily on
pine seeds with moderate utilization
of truffles and the inner bark of twigs
(fig. 2). Cone and acorn failure in
1987 resulted in a reversal of the rela-
tive emphasis on seeds and twigs.
Observation of squirrel foraging re-
vealed a corresponding opportunistic
shift from such ephemeral foods as
staminate cones and developing pine
buds in the spring to mushrooms in
summer to truffles in the fall. Similar
opportunism in food utilization has
also been reported for other tree
squirrels: the European tassel-eared
squirrel, Sciurus vulgaris (Wauters
and Dhondt 1987), the American red
squirrel, Tamiasciurus hudsonicus (Per-
ron et al. 1986), and the western gray
squirrel, S. griseus (Stienecker 1977).

In spite of a seasonal emphasis on
temporary supplies of certain food
items, the squirrel removed terminal
shoots of ponderosa pine throughout
the year. When cones and truffles be-
came scarce, as in winter and early
spring, squirrels increased consump-
tion of inner bark (figs. 1 & 2). There
seemed to be a clear preference for
the inner bark of certain trees to the
extent that some individual trees
were nearly defoliated. However, a
decreasing amount of snowfall in
three years (table 3) was associated
with an increasing number of trees
from which inner bark was taken and
a decreasing average number of clips
per tree. Thus, the identification of a
particular tree as a favorite "feed
tree" (Pfolliott and Patton 1978)
seemed to depend to some extent on

other relevant environmental condi-
tions influencing access to food sup-
plies, e.g., mobility over snow-cov-
ered ground.

The repeated use of individual
trees for inner bark was surprisingly
high. We found that 23% of all
clipped trees had shoots removed in
each of three years, while Ffolliott
and Patton (1978) reported only 2%
over four years of potential use. This
difference between the two studies
may have resulted from differences
in squirrel population densities and/
or from differences in the availability
of alternative foods. Nevertheless, it
is important to note that a resident
squirrel population may not rotate
feed trees to the extent previously
reported. In addition, more than half
a stand's individuals may become
feed trees. We expect that continued
observation will increase that per-
centage.

The quantity of hypogeous fungi
remained a fairly consistent food
supply, aside from its unusual abun-
dance in 1983 (table 1). Truffles ap-
pear to be a common component of
the diet of tree squirrels (Gronwall
and Pehrson 1984, Moller 1983), if
not of most small herbivorous mam-
mals (Maser et al. 1978). Judging
from the analysis of gut contents in
this (Vireday 1984) and other squir-
rels (Gronwall and Pehrson 1984,
Grachev and Fedosenko 1974,
McKeever 1964), hypogeous fungi
constitute one of the primary food
resources. The drop in winter squir-
rel weight as inner bark replaced
truffles in the diet (table 2 and fig. 1)
is also suggestive of the importance
of this fungal diet component. Ken-
ward (1983) showed similar weight
losses for gray squirrels, S. carolinen-
sis, feeding heavily on inner bark.

Truffle production has been re-
ported to be correlated with high
canopy cover (States 1985), which is
more characteristic of blackjack
stands than of stands with a high
proportion of yellow pines. This rela-
tionship between truffles and canopy
cover may explain the preponder-

429

ance of squirrel foraging activity ob-
served in the blackjack stands.

Observations of food supply (table
3) and squirrel food use (fig. 2)
showed considerable variability,
much of it related to precipitation
patterns. Consequently each year
presented a different pattern of food
combinations, which nnay take 5 to 10
years to repeat. Nevertheless, it is
clear that squirrels clipped twigs to
some extent every year, but greatly
increased clipping when cones were
scarce. Moreover, hypogeous fungi
were a regularly used resource.

Management impiications

The number of twig clips found has
been suggested as an index of squir-
rel population density (Brown 1982,
Keith 1965). However, the complex
pattern of clipping observed in these
three years suggests some limita-
tions. We advocate restricting such
an index to comparisons of relative
population densities between differ-
ent sites within the same year, when
one can reasonably presume that
weather conditions, pine seed abun-
dance, and availability of alternative
foods to be similar over a large area.

Maintenance of clustered stands is
essential to provide the canopy cover
needed for truffle production as well
as cover and nesting sites for squir-
rels. Reduction of stand heterogene-
ity and removal of trees in large dis-
junct blocks will likely have a nega-
tive impact on Abert squirrel habitat
(see also Pederson et al. 1987). Over
time, squirrels utilize a majority of
blackjack and yellow pine within the
stands. Forest management practices
that provide corridors for squirrel
movement among stands will poten-
tially reduce localized herbivory and
avoid severe tree damage.

Acknowledgments

We thank the Lowell Observatory for
providing a site for this study. This

research was supported by a faculty
grant from Northern Arizona Uni-
versity.

Literature Cited

Altmann, J. 1974. Observational
study of behavior: sampling meth-
ods. Behavior 49:227-265.

Brown, David E. 1982. The use of
"clippings" to index tassel-eared
squirrel population levels. Journal
of Wildlife Management 46:520-
525.

Farentinos, Robert C, P. J. Capretta,
R. E. Kepner, and Victoria M. Lit-
tlefield. 1981. Selective herbivory
in tassel-eared squirrels: role of
monoterpenes in ponderosa pine
chosen as feeding trees. Science
213:1273-1275.

Ffolliott, Peter F. and David R. Pat-
ton. 1978. Abert squirrel use of
ponderosa pine as feed trees.
USDA Forest Service Research
Note RM-362. 4 p. Rocky Moun-
tain Forest and Range Experiment
Station, Fort Collins, CO.

Grachev, Uri A., and A. K. Fe-
dosenko. 1974. On the ecology of
the Teleutka squirrel {Sciurus vul-
garis exalhidus Pall.) in the Dzun-
garskii Alatau. Proceedings of the
Academy of Sciences of the
Kazachstan SSR, Series Biology
(6):32-39.

Gronwall, Olavi, and A. Pehrson.
1984. Nutrient content in fungi as
a primary food of the red squirrel
Sciurus vulgaris L. Oecologia (Ber-
lin) 64:230-231.

Hall, Joseph G. 1981. A field study of
the Kaibab squirrel in Grand Can-
yon National Park. Wildlife Mono-
graphs No. 75. 54 p.

Keith, James 0. 1965. The Abert
squirrel and its dependence on
ponderosa pine. Ecology 46(1 &2):
150-163.

Kenward, Roger E. 1983. The causes
of damage by red and gray squir-
rels. Mammal Review 13:159-166.

Larson, M.M. and Gilbert H.

Schubert. 1970. Cone crops of pon-

derosa pine in central Arizona in-
cluding the influence of Abert
squirrels. USDA Forest Service
Research Paper RM-58. 15 p.
Rocky Mountain Forest and Range
Experiment Station, Fort Collins,
CO.

Maser, Chris, James M. Trappe, and
Ronald A. Nussbaum. 1978. Fun-
gal-small mammal interrelation-
ships with emphasis on Oregon
coniferous forests. Ecology 59:799-
809.

McKeever, Sturgis. 1964. Food habits
of the pine squirrel in northeastern
California. Journal of Wildlife
Management 28:402-404.

Moller, H. 1983. Foods and foraging
behavior of Red (Sciurus vulgaris)
and Grey (Sciurus carolinensis)
squirrels. Mammal Review 13:81-
98.

National Oceanic and Atmospheric
Administration. 1987. Local clima-
tological data. Annual summaries.
Flagstaff, Arizona. National Cli-
matic Data Center, Asheville, N.C.

Patton, David R. 1974. Estimating
food consumption from twigs
clipped by the Abert squirrel.
USDA Forest Service Note RM-
272. 4 p. Rocky Mountain Forest
and Range Experiment Station,
Fort Collins, CO.

Patton, David R. 1984. A model to
evaluate Abert squirrel habitat in
uneven-aged ponderosa pine.
Wildlife Society BulleHn 12(4):408-
414.

Patton, David R., Richard L.

Wadleigh, and Howard G. Hudak.
1985. The effects of timber har-
vesting on the Kaibab squirrel.
Journal of Wildlife Management
49:14-19.

Pearson, G.A. 1950. Management of
ponderosa pine in the Southwest.
USDA Agriculture Monograph
No. 6:1-218.

Pederson, Jordan C, R.C. Farentinos,
and Victoria M. Littlefield. 1987.
Effects of logging on habitat qual-
ity and feeding patterns of Abert
squirrels. Great Basin Naturalist
47:252-258.

430

Pederson, Jordan C. and Bruce L.
Welch. 1987. Comparison of Pon-
derosa pines as feed and nonfeed
trees for Abert squirrels. Journal
of Chemical Ecology 11:149-157.

Soderquist, Todd R. 1987. The impact
of tassel-eared squirrel defoliation
on ecotonal ponderosa pine. Jour-
nal of Mammalogy 68:398-401.

States, Jack S. 1983. New records of
hypogeous ascomycetes in Ari-
zona. Mycotaxon 16:396-402.

States, Jack S. 1984. New records of
false truffles in pine forests of Ari-
zona. Mycotaxon 19:352-367.

States, Jack. 1985. Hypogeous, my-
corrhizal fungi associated with
p)onderosa pine: sporocarp
phenology. In Proceedings of the
6th North American Conference
on Mycorrhizae. 271 p. Bend, Ore-
gon. R. Molina (ed.)

Stephenson, Richard L. 1974. Sea-
sonal food habits of Abert' s
squirrels, Sciurus aberti. Journal of
Arizona Academy of Sciences 9,
1974. Proceedings Supplement.

Vireday, Carol C. 1982. Mycophagy
in tassel-eared squirrels (Sciurus
aberti aberti and S. a. kaibabensis) in
northern Arizona. Masters thesis.
Northern Arizona University.
Ragstaff, AZ.

Wauters, L. A., and A. A. Dhondt.
1987. Activity budget and foraging
behavior of the red squirrel
{Sciurus vulgaris L.) in a coniferous
habitat. Zeitschrift Saugetierkunde
52:341-353.

431

Small Mammal Response to
the Introduction of Cattle into
a Cottonwood Floodplaln^

Fred B. Samson,^ Fritz L. Knopf,^ and Lisa B.
Hass^

Abstract.— Few differences between pastures in
small mammal communities were evident prior to
grazing, 1 month following grazing, and no differ-
ences in numbers or distribution of small mammals
were observed 5 months following grazing. Each
small mammal species exhibited different habitat
use compared to availability and few habitat vari-
ables differed on grazed versus ungrazed pastures.
Grazing at SCS recommendations in winter did not
appear to have an initial effect on small mammal
populations or their habitats in a Colorado
floodploin.

Grazing by cattle in upland areas can
affect vegetation and wildlife popula-
tions (Geier and Best 1980, Moulton
et al. 1981, Madany and West 1983),
but there is little understanding of
how grazing influences wildlife
populations and habitats in western
riparian areas (Kaufman et al. 1982).
Riparian areas of the western United
States provide habitats for greater
diversities and densities of wildlife
than adjoining upland communities
(Thomas et al. 1979, Knopf 1985), and
livestock grazing is one of many uses
that impacts riparian ecosystems.

Grazing of riparian zones gener-
ally occurs in winter along the South
Platte River and similar stream or
river systems in northeastern Colo-
rado. Overgrazing is reported, and in
some cases all ground cover includ-
ing shrubs is removed (Beidleman
1954). The purpose of this study was
to determine if small mammal com-
munities and vegetation structure
were similar in grazed and ungrazed

'Paper presented at symposium. Man-
agement of Amphibians, Reptiles, and
Small Mammals in Nortti America. (Flag-
staff, AZ. July 19-21, 1988.)

'Fred B. Samson is Regional Wildlife Bi-
ologist. USDA Forest Service, Alaska Region,
Juneau, Alaska 99802-1628.

^Fritz L. Knopf is Leader of Avian Studies,
U.S. Fisti and Wildlife Service, National Ecol-
ogy Research) Center, 1300 Blue Spruce
Drive. Fort Collins, CO 80524-2098.

''Lisa B. Mass was a graduate student.
Department of Fisheries and Wildlife Biol-
ogy, Colorado State University. Fort Collins.

riparian areas in northeastern Colo-
rado. The approach was to alter a
riparian area experimentally by in-
troducing cattle into an area that had
not been grazed for 30 years. The
specific objective was to contrast
small mammal communities and
vegetation structure before, during,
and after grazing and between
grazed and ungrazed communities.

Study Area and Methods

The study was conducted on the
Colorado Division of Wildlife's
Tamarack Ranch Unit, South Platte
State Wildlife Area, in Logan County
near Crook, Colorado, from March
1982 to August 1983. The climate is
semi-arid. Mean annual precipitation
is 47.4 cm and average monthly tem-
perature is 22.1 CO. Shallow clay-
gravel soils in highly stratified allu-
vial deposits supported an overstory
of mature plains cottonwood (Popu-
lus sargentii) and understories of
shrubs (Salix exigua, S. interior, Sym-
phoricarpos occidcntalis, Toxicodendron
radicans, Vitis vulpina, and Rhus radi-
cans), forbs (Phragmitcs communis,
Spartina pectinatus, Chenopodium al-
bum, Conium maculatum, Rumexcris-
pus, and Melilotus alba), and grasses
(Elymus canadensis and Spartina pecti-
natus).

The riparian zone adjoining the
South Platte River was last grazed in
the early 1950's (M. Gardner, pers.
comm.). Ten 16-ha pastures were es-

tablished within the riparian zone
and spaced at least 0.4 km apart to
eliminate interactive effects among
pastures. Five pastures selected at
random were grazed from mid -No-
vember 1982 to mid-March 1983 at
levels recommended by the U.S. Soil
Conservation Service, with 35.5, 30.8,
9.0, 37.2, and 36.8 AUMs allocated.
Pre-treatment data were collected on
all pastures in March, June, and Au-
gust 1982. Posttreatment data were
collected on all pastures in March
and August 1983.

A lOO-trap grid of Sherman live
traps with 15-m spacing between
rows and columns (135 x 135 m, 2.25
ha) was established in each pasture
to sample small mammal communi-
ties. Three, five-night trap sessions
were scheduled per year: prior
(middle March), during (late June),
and after (late August) the peak
small mammal breeding season. The
total number of trap nights for the
study was 25,000: 15,000 trap nights
pretreatment and 10,000 trap nights
post-treatment. Individuals were
marked with a numbered aluminum
ear tag, and species, sex, age, breed-
ing condition, trap number, and
weight were recorded. Density esti-
mates were made using the com-
puter program CAPTURE (Otis et al.
1978, White et al. 1982). CAPTURE
examines capture-recapture data,
gives population and density esti-
mates for five different models, and
indicates the model most appropriate
for estimation. Model M (H) was de-

432

termined to be the most robust of the
five estimators.

For each of the five trap sessions,
trap sites were categorized according
to trap success (no-capture vs. cap-
ture) and to the species captured at
that site. In March 1982, five no-cap-
ture sites and five sites for each spe-
cies were selected at random for
vegetation sampHng within each pas-
ture. Beginning in June 1982 and
thereafter, the sample size per pas-
ture was increased to ten no-capture
sites and ten capture sites for each
species.

Habitat variables were measured
using two line intercept transects 5-m
in length at each selected trap site.
Variables included percentage cover
of sand, litter, grass, forb, and shrub
along the 5-m transect. Transects
were centered on the trap site and
oriented toward randomly chosen
cardinal compass directions (north,
south, east, or west). The linear inter-
cept of each variable with the

transect was measured with an incre-
mental tape. Two additional meas-
urements at each trap site were dis-
tance-to-ncarest-undcrstory (<10 m)
and distance-to-ncarest-overstory
(>10 m). Vegetation sampling oc-
curred concurrently or immediately
following each trap session.

Chi-square tests were used to test
for pretreatment differences in sp>e-
cies composition among those pas-
tures chosen for grazing and those
chosen for controls. Chi-square tests
were also used to evaluate posttreat-
ment data. A f-tcst was performed to
examine differences in mean body
weight between treatment groups.

T-tests were used to compare
habitat variables between species and
between species-specific capture sites
from all other trap locations. In each
season, the vegetation variables asso-
ciated with the capture sites of a spe-
cies were compared to the pooled
sample of vegetation variables con-
sisting of no-capture sites in addition

to sites for all other species (Dueser
and Shugart 1978). The degree of
habitat specificity was indicated by
the number of variables for which
the species sample differed from the
pooled sample. Following the sp>e-
cies-sp)ecific and pooled sample two-
group comparison, mean vegetation
values associated with each species
were compared on grazed and con-
trol pastures using f-tests. These pro-
cedures determined whether habitat
used by a specific species differed
from the average habitat available
and compared a species habitat use
on control and grazed areas regard-
less of habitat availability. Some
overlap in use of trap sites was ob-
served, thus the pooled sample is not
expected to be completely distinct
from the species specific sample
(Dueser and Shugart 1978).

All statistical tests and density es-
timates were performed using the
Statistical Package for the Social Sci-
ences (Nie et al. 1975).

Table 1.— Total numbers of small mammals captured in grazed vs. un-
grazed pastures, March 1982 to August 1983, South Platte River Wildlife
Management Area, near Crook Colorado.

Species/

Pretreatment

Posttreatment

Treatment

March

June

August

March

August

Deer Mouse

Control

297

^372

M27

^268

104

Grazed

498

609

575

344

155

Western Harvest Mouse

Control

19

24

27

45

9

Grazed

39

27

22

40

3

Prairie Vole

Control

5

5

12

11

3

Grazed

4

7

2

4

6

Kangaroo Rat

Control

3

12

9

10

0

Grazed

3

3

0

0

0

Others

Control

6

7

17

2

5

Grazed

7

5

6

6

9

'Significantly different than other treatment (P <.05).
^Significantly different than other treatment (P <.00l).

^Includes house mouse, hispid pocket mouse, northern grasshopper mouse,
masked shrew, and spotted skurtk.

Results

Species Composition

Nine species of small mammals were
captured in 1982 and 1983 (table 1).
The deer mouse (Peromyscus manicu-
latus) was the most abundant species,
with the western harvest mouse (Rei-
throdontomys tnegalotis), kangaroo rat
(Dipodomys ordii), prairie vole (Micro-
tus ochogaster), house mouse (Mus
musculus), hispid pocket mouse (Per-
ognathus hisidus), northern grasshop-
per mouse (Onychomys leucogaster),
masked shrew (Sorex cinereus), and
spotted skunk (Spilogale putorius)
comprising less than 2 % of the 9,304
captures.

Pretreatment Sf)ecies richness did
not differ among grazed versus un-
grazed in March 1982 (X^ = 2.47, P =
0.650) but significant were evident in
June (X2 = 15.39, P = 0.017) and Au-
gust (X2 = 33.18, P = 0.001) (table 1).
The differences in June and August
were caused by the abundance of

433

kangaroo rats, prairie voles, and
house mice on control pastures.
While three species — the hispid
pocket mouse, masked shrew, and
spotted skunk — were found only on
pastures to be grazed. Number of
captures of the two common species,
the deer mouse and western harvest
mouse, were not different in June (X^
= 1.71, P = 0.187) or August (X^ =
2.97, P =0.091) between pastures to
be grazed and control pastures.

Following nearly 4 months of
grazing, the composition of small
mammal communities in control ver-
sus grazed pastures differed in
March 1983 (X^ = 15.9, P = 0.001)
(table 1) but not in August (X^ = 6.05,
P = 0.109). The kangaroo rat was not
captured on treated pastures in
March or August 1983 although pres-
ent in two of five pastures prior to
treatment in 1982. The number of
harvest mice captured in grazed pas-
tures increased markedly from
March 1982 to March 1983 (19 vs. 45)
in contrast to control pastures (39 vs.
40).

Inundation of all pastures in May-
July 1983 (see Knopf and Sedgwick
1987) appeared to influence species
distributions and abundances in Au-
gust. From March to August cap-
tures of deer mice on all pastures de-
clined from 61 1 to 259, western har-
vest mouse from 85 to 12, and kanga-
roo rats and mask shrews were no
longer captured.

Densities and Population
Structures

Only the deer mouse was captured in
sufficient numbers to calculate densi-
ties accurately. Deer mice densities
were consistently higher on grazed
pastures before and after treatment
(table 2). However, the density of
deer mice decreased 18.7% from pre-
to posttreatment on the five control
pastures (x= 33.6/ha vs. x=27.3/ha)
versus 42.9% on the five treated pas-
tures (63.2/ha vs. 36.1 /ha) for the
same interval.

Age ratios appear unaffected by
grazing (table 2). In contrast, sex ra-
tios in deer mice shifted significantly
following grazing (X^ = 4.90, P =
0.049) with three of five grazed pas-
tures having substantially more
males. Western harvest mice sex ra-
tios also changed following grazing,
with a higher percentage of females
captured, but sample sizes were in-
sufficient for separate tests on each of
the 10 pastures.

The percentage of female deer
mice in breeding condition was simi-
lar on all pastures prior to grazing
except in June 1982, when a higher
percentage of females (X^ = 3.84, P =
0.049) were in breeding condition on
control pastures. Following grazing,
the percentage of breeding females
was higher in March (X^ = 5.53, P =
0.019) on control pastures yet grazed
pastures had a higher percentage of
breeding females (X^ = 5.44, P =
0.020) in August 1983. No significant

differences in the percentage of
breeding males or females between
treatment groups was observed for
the other species.

Deer mice body weights were
similar across pastures prior to graz-
ing, except in June (t = 3.18, P =
0.002). After treatment, mean body
weights for mature (subadult plus
adult) deer mice were significantly
less {t = 2.66, P = 0.008) on grazed
pastures (18.56 + 0.18g) than on un-
grazed pastures (19.3 + 0.21 g) when
data from all replicates were com-
bined. The divergence in mean deer
mouse body weight between control
and grazed pastures continued into
August 1983 (t =3.02, P = 0.003).

Species Habitat Use

Only sample sizes for the deer
mouse, western harvest mouse, prai-
rie vole, and kangaroo rat were suffi-

Toble 2.— Selected population characteristics including population density
(mean no. per ha), age ratio (% juveniles), sex ratio <% females), and
breeding condition (% breeding females), March 1982 to August 1983,
South Platte River Wildlife Area, near Crook Colorado,

Characteristic/

Species/

Treatment

Pretreatment
Mar 1982 Jun 1982 Aug 1982

Posttreatment
Mar 1983 Aug 1983

Density
Deer Mouse

Control
■: Grazed
Age Ratios
Deer IVIouse
Control
Grazed
Sex Ratios
Deer Mouse

33.6
63.2

1.3
0.7

36.3
55.3

5.3
4.4

27.3
36.1

1.9
4.3^

18.7 24.8
42.7 24.7

Control

46.0

48.8

48.7

44.0

Grazed

53.0

51.3

46.2

43.8

Western Harvest Mouse

Control

60.9

52,0

31.6

0.0

Grazed

37.0

31.8

42.5

0.0

Breeding Condition

Deer Mouse

Control

65.8

67.8

16.2

64.6

Grazed

49.0'

69.5

6.61

85.1

'Significanfly different than other treatment (? <.06).

434

cient for subsequent analysis. Habitat
use by deer mice differed from that
available in 34% (12/35) of the t tests
on control pastures and 12% (4/35)
of the tests on grazed pastures over
all seasons (table 3). Deer mice were
most frequently associated with a
lower percentage of grass cover and
litter as well as presence of shrubs.
Although habitat near deer mouse
capture sites differed from that avail-
able, habitat use was similar on con-
trol and grazed pastures. Among
those habitat variables associated
with the deer mouse, 66.7% (2/3) in
March 1982, 100% (2/2) in June 1982,
66.7% (2/3) in August 1982, 80% (4/
5) in March 1983, and 0% (0/5) in
August 1983 were similar on control
and grazed pastures.

Like deer mice, the harvest mouse
used habitats differing from those
available and preferred similar sites
on control and grazed pastures (table
4). Thirty-four percent (12/35) of the
tests on control pastures and 37%
(13/25) of the tests on grazed pas-
tures were significantly different
whereas the majority (68%, 13/19)
had similar values on control and
grazed pastures. The occurrence of
harvest mice was most strongly asso-
ciated with a high percentage of litter
and grass cover and a low percent-
age of sand around the capture site.

Prairie vole capture sites differed
from the average available site for
only 11% (4/35) of the habitat com-
parisons on control pastures and 17%
(6/35) of the habitat comparisons on

Table 3.— Comparison of mean vegetation values between deer mouse
capture sites and the pooled sample on grazed and ungrazed pastures,
March 1982 to August ! 983, South Platte River Wildlife Management Area,
near Crook Colorado.

Variable/

Pretreotment

Posttreatment

Treatment

Mar 1982 Jun 1982 Aug 1982

Mar 1983

Aug 1983

Sand (%)

Control

16.9

4.4^

8.1

3.3

20.1

Grazed

7.5

6.1

A.V

2,1

10.72

Litter (%)

Control

74.41

71.4

86.2^

89.8

24.9

Grazed
Grass (%)

83.5

79.4

88.7

87.3^

21.8

Control

20.6^

32.9

52.9

38, V

23.51

Grazed

37.32

46.62

68.1

53.52

43.22

Forb (7o)

Control

16.4

48.7

55.2

30,9

I8.51

Grazed

19.4

38.22

41.52

23.42

25.72

Shrub (%)

Control

6.9

10.3

17,2

20.91

31.41

Grazed

12.0

15,6

25,4

23.8

I6.32

Disto3

Control

12.31

n.8

10.6

13.51

10.0

Grazed

9.4

10.3

12.2

12.0

21.11^

Distu'^

Control

5,5

3.V

3.5

2.8

1.21

Grazed

7.1

3.0

1.91^

2.04.42

'Significanf (P < 0.05) difference between deer mouse capture sites and pooled
sample.

^Significant (P < 0.05) difference between grazed and control pastures.
^Distance to nearest overstory (> 10m).
''Distance to nearest understory (<10m).

grazed pastures (table 5). Prairie vole
habitat was similar to habitat used by
western harvest mouse, as both ex-
hibited a preference for sites with a
high percentage of litter. For vegeta-
tion variables which were signifi-
cantly different on prairie vole cap-
ture sites compared to the pooled
sample of sites, 88% (7/9) had simi-
lar values on control and grazed pas-
tures.

Kangaroo rats exhibited the high-
est habitat specificity among the four
major mammal species (table 6).
Habitat variables from kangaroo rat
capture sites differed from the
p)ooled sample of sites for 64% (18/
28) of the habitat comparisons on
control pastures and 50% (7/14) of
the comparisons on pastures to be
grazed. The factors which appeared
most critical in determining the dis-
tribution of kangaroo rats was the
high percentage of sand, moderately
high percentage of forbs, and low
percentages of litter and grass.

Discussion and Conclusions

Kaufman et al. (1982) in Oregon
noted that small mammal densities
decreased just following grazing only
to increase to pre-grazing levels
within a year. Riparian grazing in
Oregon, as in most western range-
lands, is often in late spring to early
fall. A similar pattern, however, is
evident following winter grazing in a
riparian area in northeastern Colo-
rado with few detectable differences
observed in small mammal commu-
nity 5 months following grazing.

The elimination of kangaroo rats
from grazed areas appears to be a
consequence of grazing although
they were never really abundant on
pastures to be grazed (table 1). In
sandhill rangeland of eastern Colo-
rado, Green (1969) found the density
of kangaroo rats approximately the
same on ungrazed and grazed pas-
tures. Kangaroo rats may not have
colonized riparian grazed pastures
because of a change in microhabitat

435

prior to, or unrelated to, cattle intro-
duction. Regardless, the riparian
zone appeared to be a marginal habi-
tat for this upland species.

Differences in age ratios appear
unrelated to grazing. Abramsky
(1976) found that juvenile deer mice
do not readily enter traps and, thus,
may be under represented in age-
class ratios. The Trivers-Willard hy-
pothesis suggests that a population
under stress will produce an in-
creased proportion of females (Myers
1978). The imbalance in deer mouse
sex ratios observed in this study on
grazed, but not control pastures,
does not appear to be related to
change in primary sex ratio or sur-
vival of young as suggested by the
above hypotheses. Rather, most ani-
mals captured in March 1983 trap
session were adults, 70% of which
were tagged in 1982. The mean body
weight of deer mice on grazed pas-
tures following treatment was lower
than on control pastures. A more
parsimonius hypothesis for the ob-
served shift in sex ratio is emigration
of females. Bowers and Smith (1979)
found that female deer mice inhabit
more mesic microhabitats than
males. Grazing by cattle may have
altered microhabitats preferred by
females and or other resources, par-
ticularly seeds, may have been more
abundant on control areas. There is
substantial evidence in other studies
that deer mouse populations are lim-
ited by seasonal food availability
(Gashwiller 1979), specifically in win-
ter (Taitt 1981).

Small mammal habitat use and
seasonal habitat shifts were similar
on grazed and control pastures. Each
species illustrated differential habitat
use compared to availability, and
patterns in habitat use were little af-
fected by grazing. Deer mice habitat,
largely areas with little grass cover,
was consistently distinguishable
from that of other species as reported
elsewhere (Bowers and Smith 1979,
Kantak 1983, Lovcll 1983). Habitat
use and number of captures of the
western harvest mouse, prairie vole.

and kangaroo rat reported in this
study are also consistent with that
previously documented. The western
harvest mouse is reported to be
closely associated with grassy sites
(Hill and Hubbard 1943, Lovell 1983)
and use of sandy sites by kangaroo
rats was noted by Green (1969). The
importance of vegetative cover to the
prairie vole has been well docu-
mented (Birney et al. 1976, Green
1969).

In summary, research reported in
this paper was conducted in an ex-
perimental framework, with five rep-
lications, to evaluate the initial effects
of cattle grazing in winter on small
mammal community in a riparian
area. Winter grazing of riparian areas
based on Soil Conservation Service

recommended levels appears to have
little initial effect on small mammal
populations and their habitats. The
study further indicates that pretreat-
ment assessment of habitat and small
mammal populations in studies to
evaluate effects of grazing in riparian
areas is important. Significant differ-
ences in small mammal numbers and
species-specific habitat use observed
following grazing could have been
attributed to treatment without
knowledge of pretreatment popula-
tion and habitat conditions.

Acknowledgments

The study was partially funded by
the U.S. Fish and Wildlife Service

Table 4.— Comparison of mean vegetation values (%) between western
harvest mouse capture sites and the pooled sample on grazed and un-
grazed pastures, March 1982 to August 1983, South Platte River Wildlife
f^anagement Area, near Crook Colorado.

Pretreatment

Posttreatment

Treatment

Mar 1982 Jun 1982 Aug 1982

Mar 1983

Aug 1983

Sand (%)
Control
Grazed
Litter (%)
Control

0.0^
5.9

0,V
1.81

5,5
0.0

0.31
0.2

12.9
0.02

95.41

73.8

89.91

93.41

42.6

Grazed

93.1

58.4'

89.2

95.01

16.0

Grass (%)

Control

61. r

40.3

64.4

78.2

43.2

Grazed

78.61-2

53.6

78.0

79.71

55.0

Forb (%)

Control

19.9

49.6

53.0

20.21

28.2

Grazed

14.8

35.52

38.6

12.21-2

38.7

Shrub (%)

Control

11.6

20.51

22.8

8.81

22.2

Grazed

4.81

14.3

30.1

25.2

26,7

Disto^

Control

9.1

12.2

8,9

11.7

6.9^

Grazed

8.6

11.4

14.11

12.91

65.0

Distu^

Control

7.2

5.1

2.7

3.61

1.4

Grazed

7.6

4.71

1.31

1.52

4.8

'Significant (P < 0.06) difference between western harvest mouse capture sites and
pooled sample.

'Significant (P < 0.05) difference between grazed and control pastures.

^Distance to nearest overstory (> Wm).
"Distance to nearest understory (<10m).

436

through the Colorado Cooperative
Wildlife Research Unit, and is a
product of Cooperative Agreement
No. 2463-4 between the Colorado
Division of Wildlife and the U.S. Fish
and Wildlife Service's National Ecol-
ogy Research Center.

Literature Cited

Abramsky, Z. Z. 1976. Small mam-
mal studies in natural and ma-
nipulated shortgrass prairie. Ph.
D. dissertation, Colorado State
University, Fort Collins.

Beidleman, Richard G. 1954. The Cot-
tonwood river-bottom community
as a vertebrate habitat. Ph.D. dis-
sertation. University of Colorado,
Boulder, Colorado.

Birney, Elmer D., William E. Grant,
and Duane D. Baird. 1976. Impor-
tance of vegetative cover to cycles
of Microtus populations. Ecology
57: 1043-1051.

Bowers, Michael A., and Howard D.
Smith. 1979. Differential habitat
utilization by sexes of the
deermouse, Peromyscus manicula-
tus. Ecology 60: 869-875.

Dueser, Raymond D. and Herman H.
Shugart, Jr. 1978. Microhabitats in
a forest floor small mammal
fauna. Ecology 59: 89-98.

Gash wilier. Jay S. 1979. Deer mouse
reproduction and its relationship
to the tree seed crop. American
Midland Naturalist 102: 95-102.

Geier, Arnold R., and Louis B. Best.
1980. Habitat selection by small
mammals of riparian comnmni-

Table 6,— Comparison of mean vegetation values between prairie vole
capture sites and the pooled sample on grazed and ungrazed pastures,
March 1982 to August 1983, South Platte River Wildlife Management Area,
near Crook Colorado.

Pretreatment

Posttreatment

Treatment

Mar 1982 Jun 1982

Aug 1982

Mar 1983

Aug 1983

Sand (%)

Control

0.0

20.0

13.9

0.0

0.0

Grazed

0,0

0.0

0.0'

0,0

0.0

Litter (%)

Control

88.0

80.0

85.4

71.0

100,0

Grazed

99.7^^

85.7

84.5

99.5'

33.5

Grass (%)

Control

52.2

52.6

36.6

80.0

100.0

Grazed

78.7^

43.4

56.0

65.0

63.7

Forb (%)

Control

25.0

42.8

49.0

2.0

100.0

Grazed

22.8

29,92

21.0

38,0

40.72

Shrub (%)

Control

0.0

4.8

48.6'

22,0

0.0

Grazed

5.3

17,3

22.5

20,6

23.0

Disto^

Control

5.8^

14.1

10.9

2.4

15.0

Grazed

15.6^2

10.6

4.0'^

12.4

95.7'

Distu^

Control

5.8

2.4

0.2'

3,2

7.5

Grazed

14.0

1.8

5.62

2,7

12.8

'Significont (P

< 0.05) difference between the

prairie vole

and pooled sample.

^Significant (P < 0.05) difference between grazed and control pastures.
^Distance to nearest overstory (> 10m).
^Distance to nearest understory (<10m).

ties: Evaluating effects of habitat
alterations. Journal of Wildlife
Management 44: 16-24.

Green, N. E. 1969. Occurrence of
small mammals on sandhill range-
lands in eastern Colorado. M.S.
thesis, Colorado State University,
Fort Collins.

Hill, John E. and Claude W. Hub-
bard. 1943. Ecological differentia-
tion between two harvest mice
(Reithrodontomx/s) in western Kan-
sas. Journal of Mammalogy 24: 22-
25.

Kantak, Gail E. 1983. Behavioral,
seed preference and habitat selec-
tion experiments with two
sympatric Peromyscus species.
American Midland Naturalist 109:
246-252.

Kaufman, J. Boone, William C.
Krueger, and M. Vavra. 1982. Im-
pacts of a late seasonal grazing
scheme on nongame wildlife habi-
tat in a Wallowa Mountain ripar-
ian ecosystem, p. 208-220. In Wild-
life-livestock Relationships Sym-
posium. Coeur d'Alene, Id., [April
21-22, 1981] University of Idaho
Forest, Wildlife, and Range Ex-
p)eriment Station, Moscow, Idaho.

Knopf, Fritz L. 1985. Significance of
riparian vegetation to breeding
birds across an altitudinal cline. p.
105-111. In Riparian ecosystems
and their management: Reconcil-
ing conflicting uses. [Tucson,
Ariz., April 16-18, 1985] USDA
Forest Service General Technical
Report RM-120, Rocky Mountain
Forest and Range Experiment Sta-
tion, Fort Collins, Colo.

Knopf, Fritz L. and James. A.

Sedgwick. 1987. Latent Population
responses to a catastrophic, clima-
tological event. Condor 89: 869-
873.

Lovell, David C. 1983. Succession of
mammals, and associations of
small mammals, in disturbed habi-
tats at Barr Lake State Park,
Adams County, Colorado. M.S.
thesis. For Hays State University,
Fort Hays, Kan.

437

Madany, Michael H., and Neil E.
West. 1983. Livestock grazing-fire
regime interactions within mon-
tane forests of Zion National Park,
Utah. Ecology 64: 661-667.

Moulton, Michael P., Jerry R. Choate,
Stephen J. Bissell, and Robert A.
Nicholson. 1981. Associations of
small mammals on the central
high plains of eastern Colorado.
Southwestern Naturalist 26: 53-57.

Myers, Judith H. 1978. Sex ration ad-
justment under food stress: Maxi-
mization of quality of numbers of
offspring? American Naturalist
112: 381-388.

Nie, Norman H., C. Hadlei Hull, Jean
G. Jenkins, Karin Steinbrenner,
and Dale H. Bent. 1975. Statistical
Package for the Social Sciences,
McGraw-Hill Book Company,
New York, NY.

Otis, David L., Kenneth P. Burnham,
Gary C. White, and David R. An-
derson. 1978. Statistical inference
from capture data on closed ani-
mal populations. Wildlife Mono-
graph 62: 1-135.

Taitt, M. J. 1981. The effect of extra
food in small mammal popula-
tions. I. Deermice (Peromyscus
maniculatus). Journal of animal
ecology 50: 111-124.

Thomas, Jack W., Chris Maser, and
Jon E. Rodick. 1979. Riparian
zones. In Wildlife habitats in man-
aged forests, the Blue Mountains
of Oregon and Washington, USDA
Handbook 553: 40-47.

Table 6.— Comparison of mean vegetation values between kangaroo rat
capture sites and the pooled sample on grazed arxi ungrazed pastures,
March 1982 to August 1983, South Platte River Wildl'ife Management Area,
near Crook Colorado.

Treatment

Pretreatment
Mar 1982 Jun 1982 Aug 1982

PosttrGatment
Mar 1983 Aug 1983

Sand (%)

Control

Grazed

Litter (%)

Control

Grazed

Grass (%)

Control

Grazed

Forb (%)

Control

Grazed

Shrub (%)

Control

Grazed

Disto2

Control

Grazed

Dlstu^

Control

Grazed

53.01
53.0^

69.0
38.01

12.0
33.7

18.0
25.3

21.5
10.7

3.61
7,5

9.6
3,21

67.41
51.01

44.4'
70.0

10.71
11.01

71.11
60.5

i.r

26.0

11.2
6.01

7.51
0.41

38.61

66.5^

29.41

65.6

lo.r

6.71

5.1

34.91
4971
15.11
44.81
4.21
9.8
4.3

'Significant (P < 0.05) difference between l<ongaroo rat capture sites and pooled
sample.

'Distance to nearest overstory (> 10m).
^Distance to nearest understory (< 10m),

White, Gary C, David R. Anderson,
Kenneth P. Burnham, and David
L. Otis. 1982. Capture-recapture
and removal methods for sam-
pling closed populations. Los
Alamos National Laboratories,
Los Alamos, NM.

Zimmerman, Earl G. 1965. A com-
parison of habitat and food of two
species of Microtus. Journal Mam-
malogy 46: 605-612.

438

Old Growth Forests and thie
Distribution of the Terrestrial
Herpetofauna^

Hartwell H. Welsh, Jr.^ and Amy J. Lind^

Abstract.— Terrestrial herpetofauna were sampled
by pitfall traps and time-constrained searches on 42
stands of Douglas-fir/hardwood forest in southwest-
ern Oregon and northwestern California. Stands
ranged in age from 40 to 450 years. We found 25
species of herpetofauna. Species diversity was
greater in older forest stands than in young stands.
Amphibians were significantly more abundant in old
than in young stands and significantly less abundant
in dry than in moist stands. Our research indicates
that changes in forest structure due to forest prac-
tices results in reduced species diversity and abun-
dance among the herpetofauna.

The coniferous forests of the Pacific
Northwest are currently the focus of
a national conflict between compet-
ing interests. These ancient forests,
previously more species rich and
continuous across the continental
United States, have undergone a
natural decline since the Mesozoic in
conjunction with broad climatic and
geologic changes (Axelrod 1976).
This process eliminated most of the
wooded areas of the Midwest, but
left expansive tracts of forest in the
eastern and western United States. In
the last hundred years, many of these
remaining ancient forests have been
harvested for wood products, chang-
ing the species composition, struc-
ture, and forest age (Harris 1984).
These natural forest ecosystems have
been altered so rapidly that we are
only now recognizing the loss of
some plant and animal species and
the threat to others [e.g., the spotted
owl (Strix occidentalis)] (Simberloff
1987). Recent concern for the health
and well-being of these forest ecosys-
tems, and the need for more knowl-

' Roper presented at tlie Symposium on
Management of Amphibians. Reptiles, and
Small Mammals in Nortii America (July 19-
21. 1988. Flagstaff Arizona).

^Wildlife Biologist, Pacific Soutt)west For-
est and Range and Experiment Station.
Forest Service. U.S. Department of Agricul-
ture. Areata. California 95521.

^Biological Technician, Pacific Southwest
Forest and Range and Experiment Station,
Forest Service, U.S. Department of Agricul-
ture, Areata. California 95521.

edge to meet management goals and
the requirements of the National For-
est Management Act 1976 and the
Endangered Species Act 1973 has
prompted research into the structure
and composition of the vertebrate
communities of these forests
(Meslow et al. 1981, Raphael 1984,
Ruggiero and Carey 1984).

From 1981 through 1983, Raphael
(1984, 1987, this volume) used a vari-
ety of sampling methods to collect
data on the forest age, moisture, and
habitat associations of birds, mam-
mals, reptiles, and amphibians in for-
ests of northwestern California. From
1984 through 1986, researchers from
the Forest Service's Pacific Southwest
Forest and Range Experiment Station
extended these studies to include
southwestern Oregon. By measuring
differences in the species composi-
tion and relative abundance of the
herpetofaunal community in altered
versus unaltered habitats it is pos-
sible to indicate biologically mean-
ingful differences in habitat quality
(e.g.. Bury et al. 1977, Busack and
Bury 1974, Jones 1981, Luckenbach
and Bury 1983, Ortega etal. 1982).
Such information on differences in
the composition of the herpetofauna,
relative to forest age and moisture,
have scientific value as well as practi-
cal value, as indicators of habitat
change, useful to natural resource
managers.

This paper reports on a study to
determine the occurrence and abun-
dance of the forest herpetofauna rela-

Oregon

California

# Coastal Stand
▲ Inland Stand

Figure 1.— Study stands In Douglas-fir forests
were located in norttiwestern California
and southwestern Oregon. Triangles =
stands in the inland area, circles = stands In
the coostal area.

tive to forest age and moisture, and
to compare two methods (time-con-
strained searches and pit-fall trap-
ping) used to sample this herpe-
tofauna in northwestern California
and southwestern Oregon.

439

STUDY AREA

This study was conducted in
Douglas-fir (Pseudotsuga menziesii)/
hardwood forests at low to mid-ele-
vations in the Klamath Mountains
and Coast Range. We sampled 54
stands, but we use data from only 42
stands, omitting nine higher eleva-
tion, white-fir dominated, stands and
three stands on serpentine soils be-
cause they differed so greatly from
our remaining stands. Even-aged
stands in the above forest type were
selected in three areas within the Kla-
math Mountains and Coast Range
(fig. 1) in accordance with proce-
dures outlined by Spies et al, (in
press). Using stand characteristics
(Franklin et al. 1986) and tree age, we
assigned stands to one of three age
classes: young, mature, and old-
growth forests. Stands ranged in age
from 40 to 450 years. Stands in old-
growth were further categorized into
three moisture classes: dry, mesic,
and wet (fig. 2). Stands ranged in size
from 21 to 150 hectares, and in eleva-
tion from 53 m to 1205 m. One-half of
the stands occurred within the Coast
Range, an area formed primarily of
Franciscan parent materials and
dominated by the maritime climatic
influences of the Pacific Ocean. These
stands were classified as coastal for-
est stands (fig. 1). All stands were
dominated by Douglas-fir and con-

Old Crowlh

Dry;

Old Growth
Mcsic:

Old Growth
Wet:

1 coastal
3 inland

4 coastal
6 inland

3 coastal
3 inland

MhIuth:

5 coastal

6 inland

Young;

8 coastal
3 inland

Age

Class

Moisture Class

tained a significant hardwood ele-
ment, primarily tanoak (Lithocarpus
densiflora) and madrone (Arbutus
menziesii); about half also contained
coast redwood (Sequoia semperznrens).

The other sites were designated
inland stands (fig. 1), occurring
within the Klamath Mountains, pri-
marily on granitic and metamorphic
parent materials. This area is subject
to colder winters and drier, hotter
summers than the Coast Range. The

inlands stands were dominated by
Douglas-fir in association with
tanoak, madrone, and to a lesser ex-
tent, canyon live oak (Quercus
chrysolepis), black oak (Quercus kellog-
gii), ponderosa pine (Pinus ponder-
osa), sugar pine (Pinus lamhertiana),
and incense cedar (Calocedrus decur-
rens). For a more complete descrip-
tion of the vegetations of these two
provinces see Raphael (in press) and
Sawyer and Thomburgh (1977).

r~ ^ —

Table 1 .—Structural features' of Douglas-fir stands on which herpetofauna
were sampled in northwestern California and southwestern Oregon.'

Forest age class

Figure 2.— Distribution of study stands by
forest age and moisture class, and by
coastal and Inland area.

Young (9)^

Mature (11)

Old (19)

< 100 yrs

100-200 yrs

> 200 yrs

Structural feature

(mean')

(mean)

(mean)

Live trees

Age of dominant size class

44.3

129.0

264.6

of Douglas-fir

+ 18.3

+38.9

+74.4

Diameter at breast ineight

38.4

85.1

111.5

^r)Rl-l^ of Hominont <?i7fs

+ 14 ft

+99 9

+23 8

class of Douglas-fir (cm)

Ig. conifers— trees/ha

1.5

17.1

34.7

(> 80 cm DBH)

+2.1

+12.0

±14,1

Ig. hardwoods— trees/ha

9.1

10.7

13.0

(> 50 cm DBH)

±10.8

+ 12.0

+ 10.1

sm. trees— trees/ha

1430.0

764.1

630.0

(conf.-5-80 cm DBH

±594.0

±331.0

±240.0

+ hdwds-5-50 cm DBH)

Snags

(conifers and hardwoods)

large— snags/ha

3.4

2.2

6.1

(> 50 cm DBH and

±6.3

±2.2

±5.9

> 4.5 m in height)

Logs

Ig. conifers— logs/ha

1.3

0.7

4.0

(> 50 cm DBH and

±4.0

±1.8

±4.0

> 15 m long)

Ig. conifers— mt/ha^

1.2

1.0

6.9

+3.5

+2.5

+6.9

sm. conifers— logs/ha

334.4

151.4

192.2

±179.5

+ 129.9

+ 107.8

hardwoods— logs/ha

95.6

146.4

113.7

±64.8

+123.1

+59.0

'Bruce Bingham, unpublished data on file with Pacific Southwest Forest and Range
Experiment Station, 1700 Bayview Drive. Areata, CA 95621.

^Sampling occurred from 1984-1986.

'Number of stands.

^f\/lean ± 1 standard deviation.

^ mt = metric tons.

440

METHODS

Herpetofauna Sampling

A herpetofaunal sampling design
was developed for the USDA Forest
Service's old-growth wildlife habitats
project in Oregon and Washington
by Corn and Bury (in prep.). Their
design used two methods to sample
species composition and relative
abundance of the herpetofauna: pit-
fall traps (PF) and time-constrained
searches (TCS) (Bury and Corn, this
volume; Welsh 1987). The TCS
method employed in our study dif-
fered from that described by Corn
and Bury (in prep.) in that headwater
habitats (springs, seeps, and first or-
der streams) were included in the
sampling. Pitfall trap grids consisted
of 36 cans buried at ground level and
spaced 15 m apart. Traps were cov-
ered with bark or cedar shakes. We
sampled 40 stands in the fall of 1984
and 1985, for 50 and 30 nights, re-
spectively. Our total pitfall trapping
effort amounted to 115,200 trap-
nights. Time-constrained searches
consisted of intensively searching all
terrestrial microhabitats in the forest
environment for a fixed amount of
time. Only actual search time was
counted, when an animal was en-
countered the timer was stopped
while data were collected. A 4-per-
son-hour TCS was conducted on
each of the 42 stands in 1984 and

1985. An additional 4-person-hour
TCS was conducted on 30 stands in

1986. Our total effort for TCS
amounted to 456 person-hours.

Forest Age

Forest age was determined for each
stand by increment borer, or by
counting rings on stumps in adjacent
logged areas. Dominant or co-domi-
nant size class Douglas-fir trees were
selected for aging and trees were
cored at breast height. Two to 10
trees (average 3) were cored on each
stand and the sample mean was used

to estimate forest age for the stand.
On the basis of tree coring, ring
counts, and structural characteristics
(Franklin et al. 1986), we grouped
stands into three age classes: young
forest, <100 years; mature forest, 100-
200 years; and old-growth forest,
200+ years (table 1).

Moisture Class

Stands that were classified as old-
growth were also assigned a mois-
ture classification (dry, mesic, or
wet), depending on plant species
composition and percent cover of the
herb and shrub layers within the
stand. The data were independently
recorded from three to five 0.1 ha
circular plots selected at random
within each stand. Moisture class as-
signment was based on mean percent
cover values and the absolute con-
stancy of particular shrub and herb
species within each stand.

Faunal Comparisons

We tested the null hypotheses (H^)
that mean capture frequencies for
herpetofauna did not differ between
either forest age or moisture classes
(1) within the coastal and inland ar-
eas, (2) between the coastal and in-
land areas, and (3) among all stands
(coastal and inland areas combined).
Only the mesic old-growth stands
were used in the age analysis (fig. 2).
One coastal old-growth dry stand
prevented testing for differences in
means among moisture classes
within the coastal area, and between
the coastal and inland dry stands.

We emphasize that our inferences
are drawn from observations and not
experimental manipulations. Though
our results are described in the con-
text of hypothesis testing, our study
is primarily exploratory. In addition,
the power of our tests was low be-
cause our sample sizes were rela-
tively small. Our approach yields
preliminary results about forest age

and moisture relationships among
the herpetofauna, but we caution
against making broad inferences.

Combining Data Across Years

Data from pitfall trapping were to-
taled, by stand, for each species, di-
vided by 50 (1984 data) or 30 (1985
data) nights x 36 traps and multi-
plied by 1000 to yield captures per
1000 trap-nights. Data from time-
constrained searches were adjusted
for unequal sampling effort by ex-
pressing abundance of each species
in captures per person-hour.

We performed paired t-tests be-
tween years (total captures per
stand) for each data set. TCS samples
were not significantly different be-
tween years: 1984 vs. 1985, t = 1.16, P
= 0.25; 1984 vs. 1986, t = 1.24, P =
0.22; 1985 vs. 1986, t = 1.85, P = 0.075.
PF samples were also not signifi-
cantly different between years: 1984
vs. 1985, t = 1.85, P = 0.072. Conse-
quently, we combined years for each
sampling method for all analyses.

Statistical Comparisons

For each method, we tested for statis-
tical differences in mean capture fre-
quencies among age and moisture
classes, across, within, and between
inland and coastal areas. These tests
were performed on the total herpe-
tofauna, taxa at the level of class, or-
der, and sub-order, and on those spe-
cies captured on at least one third of
our stands in either area.

Mean capture frequencies of each
faunal grouping were tested for sta-
tistical differences among three forest
age classes and three moisture
classes. In cases where group vari-
ances were equal among classes, we
used one-way analysis of variance
(ANOVA). We used Hartley's F max
test (Milliken and Johnson 1984:18)
with P < 0.01 to determine the equal-
ity of variances for all three-group
tests. We used P < 0.01 because

441

ANOVA is robust under moderate
violations of the assumption of equal
variances (Zar 1984:170). If a signifi-
cant F-statistic resulted from the
ANOVA test, we tested further for
significant differences between pairs
in order to isolate the source of the
differences by using the Tukey test
(TU) for multiple comparisons (Zar
1984:186). Where group variances
were not equal or where one of the
three age or moisture classes had no
captures, we performed all pairwise
tests (multiple comparisons) using
the Games and Howell modification
of the Tukey test (GHMC) (Keselman
and Rogan 1978).

To test for statistical differences in
capture frequencies in age and mois-
ture classes between coastal and in-
land areas, we used two sample t-
tests (Zar 1984:131). We followed the
more conservative approach of not
pooling variances. Between-area
comparisons consisted of two fami-
lies of tests: (Da single paired com-
parison based on all stands, and (2)
five pairwise comparisons defined by
the different forest age and moisture
classes. Tests in the first family were
considered statistically significant at
the P < 0.05 level. A Bonferroni ad-
justment (Miller 1981:67) was used
for tests done within the second fam-
ily to maintain an overall significance
level of P < 0.05.

For the species richness analyses,
stand records from the TCS and PF
data were combined. The means of
the total number of species for each
forest age and moisture class were
tested for differences described.

Also, the similarity of species'
composition among equal numbers
of stands (selected randomly) in each
forest age class were determined by
using Jaccard's similarity coefficient
(Sneath and Sokal 1973:131):

a

S =

o + b + c

only, and c = number of species in
the second class only.

RESULTS AND DISCUSSION

We sampled 25 species. Amphibians
accounted for 97.8% (salamanders,
96.3%) of all captures, and reptiles
2.2%. The TCS method yielded more
than 66% of all captures (table 2),
sampling 22 species (table 3) and ac-
counting for 67% of the amphibians
and 85% of the reptiles. The PF
method sampled 19 species (table 4)
and accounted for slightly less than
1/3 of all captures (table 2).

Species Composition, Richness

Similarity Indices

Based on species presence-absence
data, an analysis of faunal similari-
ties between forest age classes

(coastal and inland areas combined)
indicated that greatest similarity in
species composition occurred be-
tween the mature and old-growth
stands (table 5). Jacard's Similarity
Index (JSI) values, for comparisons
between young and old-growth
stands and young and mature
stands, indicated that young stands
were different in species composition
from both classes of older forest
stands. These differences were great-
est between young and old-growth
stands (table 5).

Species Rictiness

The number of species per stand for
all 42 stands ranged from 3 to 13 (fig.
3). The coastal mature stands yielded
the highest mean number of species
overall, while the lowest mean num-
ber of species occurred on the inland
mature stands (table 6, fig. 3). The

in which, for any two classes, a =
number of species in common, b =
number of species in the first class

Table 2.— Captures of herpetofauna by time-constrolned searches (TCS)
and pitfall traps (PF) in Douglas-fir forests of northwestern California and
southwestern Oregon from 1 984 to 1 986.

Salamanders Frogs Lizards Snakes All species

Method

(mean^)

(mean)

(mean)

(mean)

(mean)\o\a\ captures

PF 19842

13.72

0.38

0.07

0.00

14.18

1021

(40 stands)

+ 14.18

±1.06

+0.18

+ 14.27

PF 19853

11.20

0.32

0.21

0.02

11.76

508

(40 stands)

±10.23

±0,83

±0.61

±0.15

±10.19

PF Totals

1529

(32.6%)«

TCS^ 1984

6.46

0.04

0.11

0.

04

6.66

1118

(42 stands)

±3.63

±0.14

+0.27

±0.

13

+3.61

TCS 1985

5.80

0.01

0.15

0.06

6.11

1027

(42 stands)

±3.80

±0.20

±0.30

±0.16

+3.81

TCS 1986

8,10

0.18

0.15

0.08

8.51

1021

(30 stands)

±4,10

±0.41

±0.31

±0.23

±4.09

TCS totals

3166

(67.4%)

Totals, both methods

4695

'Mean for pitfall trapping = per 1000 trap-nigtits; X for tirDe-constrolned searches =
per person-hour of search time: both ore ± 1 standard deviation.

'PF 1984 = 50 trap-nights per stand.

^PF 1985 = 30 trap-nights per stand.

"All TCS = 4 person-hours per stand per year.

^Percentage of total captures.

442

Table 3.— Mean number of captures per person-hour' captured by time-constrained searches (TCS) In different age
and moisture classes of Douglas-fir forests of northwestern California and southwestern Oregon In the springs of 1984,
1985, and 1986.^

Young

Mature

Old -wet

Old-mesic

Old-dry

Total old

Total

Species

(11)^

(11)

(6)

(10)

(4)

(20)

Captures

Frogs

Tailed frog

0.000

0.008

0.000

0.013

0.000

0.006

2

(Ascaphus fruei)

±0.025

±0.040

±0.028

Pacific treefrog

0.049

0.166

0.028

0.117

0.000

0.067

44

(Hyla regilla)

±0.128

+0.263

±0.068

±0.153

±0.123

Total

0.049

0.174

0.028

0.129

0.000

0.073

46

±0.128

±0.259

±0.068

±0.148

±0.122

Salamanders

Northwestern

0.000

0.008

0.000

0.000

0.000

0.000

1

Salamander

_.

±0.025

, , , ,

(Ambysfoma gracile)

Clouded salamander

0.496

0.390

0.361

0.725

0.146

0.500

227

(Aneides ferrous)

+0.914

±0.457

±0.215

±0.451

±0.172

±0.415

Black salamander

0.099

0.121

0.000

0.050

0.000

0.025

35

(A. fiavipuncfafus)

+0.178

±0.272

±0.070

±0.065

Calif, slender^

2.718

5.533

4.470

5.542

0.417

4.500

972

salamander

±1.958

±1.065

±1.320

±1.738

±2.190

(Bafrachoseps

affenuafus)

Pacific giant

0.091

0.008

0.000

0.008

0.021

0.008

12

salamander

±0.183

±0.026

±0.026

±0.042

±0.026

(Dicampfodon

ensatus)

Ensatina

2.265

2.595

2.625

4.508

2.938

3.629

1447

(Ensafina

±1.653

±1.391

±2.321

±2.816

±1.332

±2.506

eschscholfzii)

Del Norte*

0.278

0.396

1.722

2.278

0.208

1.622

258

salamander

±0.411

±0.970

±2.237

±3.349

±0.191

±2.607

(Plefhodon

elongafus)

Olympic salamander

0.000

0.038

0.070

0.192

0.000

0.116

31

(Rhyacofrifon

±0.086

±0.111

±0.258

±0.203

olympicus)

Rough-skinned newt

0.038

0.140

0.028

0.192

0.021

0.108

49

(Taricha qranulosa)

+0.101

+0.183

+0.068

+0.399

+0.042

+0.290

Total

5.041

6.030

6.180

9.260

3.385

7.160

3032

±1.917

+2.991

±1.490

±4,900

±1.485

+4.234

Total

5.090

6.204

6.208

9.390

3.385

7.233

3078

amphibians

±1.969

±3.078

±1.532

±4,900

±1.485

±4.270

Lizards

Western skink

0.008

0,045

0.000

0,008

0.063

0.017

10

(Eumeces

±0.025

±0.101

±0,026

±0.125

±0.058

skilfonianus)

Northern

0.095

0.167

0.014

0.042

0.084

0.042

42

Alligator lizard

±0.160

±0.230

±0.034

±0.044

±0.096

±0.057

(Elgaria coeruleus)

Southern

0.000

0.008

0.000

0.000

0.063

0.013

3

Alligator lizard

±0.025

±0.125

±0.056

(E. mulficarinafus)

(Continued)

443

Table Z — (continued)

Species

Young
(11)^

Mature
01)

Western fence lizard
(Sceloporus
occldentalis)
Total

Snakes
Rubber boa
(Charina botfae)
Sharp-tailed snake
(Contia tenuis)
Ringneck snake
(Diadophis punctatus)
Western aquatic
garter snake
(Ttiamnophis couchii)
Terrestrial
garter snake
(T. elegans)
Nortti western
garter snake
(T. ordinoides)
Common garter snake
(T, sirtalis)

Total

Total reptiles
All

herpetofauna

0.000

0.102
±0.170

0.008
±0.025

0,008
±0.025

0.000

0.000

0,015
±0.050

0.023
+0.075

0.023
±0.054

0.242
±0.313

0.000

0.008
±0.025

0.057
±0.109

0.008
±0.025

0.000

0.011
+0.038

Old-wet

Old-mesic

Old-dry
(4)

Total old
(20)

0,000

0,000

0.083
±0.118

0.017
±0.058

0,014
±0.034

0.058
±0.068

0.292
±0.323

0.092
±0.173

0,000

0.000

0.000

0.000

0.014

+0 n'^4

0.000

0.083

±U. 10/

0.021
±U,u/6

0.000
0.000

0.008
±0.026
0.000

0.073
±0.086
0,000

0.019

+0 048
0.000

0.000

0.000

0,000

0.000

0,000

0.000

0.031

0.006

Total
Captures

+0,063

'Mean ± 1 standard deviation.

^Data are from inland and coastal stands combined.

^Number of stands.

"Absent from inland stands.

^Absent from coastal stands.

+0.028

0.000

0.000

0.014

0,000

0.000

0.000

±0.034

0.053

0.083

0.028

0.008

0.187

0.050

±0.086

±0.138

±0.043

±0.026

±0.239

±0.122

0.155

0.326

0.042

0,066

0.479

0.142

±0.210

±0.418

±0.069

±0.086

±0,473

±0.265

5.246

6.530

6.250

9.450

3.865

7.371

±2.004

±3.205

±1.559

±4.900

±1.543

+4.203

61

1
7
11
1

1

27
88
3166

coastal stands had significantly more
species per stand than the inland
stands (fig. 3, table Al).

With coastal and inland areas
combined, our mean species values
indicated that species richness was
greatest on mature stands (table 6),
but was not statistically different.

In the inland area, the old-growth
dry stands had the greatest mean
number of species (table 6) but no
comparisons yielded significant dif-
ferences (fig. 3). Within the coastal

area, mean numbers of species were
significantly different between forest
age classes. Multiple comparisons
(TU) indicated that the greatest dif-
ferences occurred between young
and mature stands (fig. 3).

The significantly higher number of
species in the coastal vs. the inland
area (fig. 3) is attributable to the
salamander Aneides lugubris and four
snakes (Thamnophis couchii, T. sirtalis,
T. elegans, and Charina bottae), which
were all sampled in very low num-

bers and only in the coastal area
(tables 3-4). We believe this is an arti-
fact of the difficulty of sampling for
snakes in forested habitats (Bury and
Corn 1987, Raphael and Marcot 1986,
Welsh 1987). Most snake species exist
in low densities, and available sam-
pling methods only establish pres-
ence. All of these snake species occur
in the ii\land area. The arboreal
salamander, Aneides lugubris, is ab-
sent inland at the northern latitudes
we sampled (Stebbins 1985).

444

Table 4.— Mean number of captures per 1000 trap-nights' captured by pitfall traps (PF) in different age and moisture
classes of Douglas-fir forests of northwestern California and southwestern Oregon. Sampling occurred In the falls of.
1984 and 1985.2

Young Mature Old-wet Old-mesic Old-dry Total old Total
Species (10)^ (11) (6) (9) (4) (19) Captures

Prone
n uyo

Tnilpd froa

0.000

V V 1 ^ III 1 V-/ \-A

0.000

Pacific treefroa

0.139

^Hvln rf^nilln)

+0.293

Yellow-l©aaed froa

0.000

fQnnci Hnv/lii)

Total

0 139

±0.293

Snlnmnnciers

North wpstern

1 '1 \y 1 II 1 VV O 1 1 1 1

0.035

<;alannand©r

±0.110

(Ambysfomo gracile)

Clouded salamar^der

0.035

±o.no

Black salamander

0.035

(A. flavipuncfofus)

±0.1 10

Arhorenl ^nlnmnnder

0.035

('/A, lugubris)

±0.1 10

Calif, slender^

0.298

salamander

±0.422

(Bafrach oseps

affenuafus)

Pacific giant

0.104

salamander

±0.168

(Dicompfodon

ensnfijs)

Ensatina

W 1 1 \J III 1

8.646

CEnsofino

±7.107

eschscholfzii)

Del Norte^

0,810

salamander

±1.120

(Pie f hod on

elongafus)

Olympic salamander

0.000

(Rhyacofrifon

olympicus)

Rough-skinned newt

0.174

(Toricha granulosa)

+0.245

Total

9.620

±7.340

Total

9.760

Amphibians

±7.480

Lizards

Western skink

0.000

(Eumeces

skilfonionus)

V

n 06'^

n nnn

0 n'^9

+0 909

+0 116

0 06'^

0 oon

0 000

+0 140

0.315

0,058

0,077

±0.105

±0.142

±0.159

0 '^lA

0 058

0 579

±0.667

±0.142

±1,493

0 473

0 116

0 694

±0.948

±0.179

±1.483

0.032

0,116

0.000

+0 105

+0 284

0 063

0 058

0 039

+0.140

±0.142

±0,116

0.410

0.000

0.193

+ 1 9S0

+0 352

0.000

0,000

0,000

2.153

0.463

2,517

+ 1 162

±0,401

±1.037

0.126

0.578

0.154

±0.234

±0.474

±0.252

10.164

6.539

9,375

±8.996

±4.328

±7,209

0 120

1.500

13.060

±0.280

±2.310

±25,370

0.000

0.058

0.000

±0.142

0.442

0.174

0.579

+0.493

+0.290

+ 1.264

12.280

8.510

18.750

+8.590

±4.680

±21.040

12.750

8.620

19.440

±8.340

±4.670

±20.940

0.095

0.058

0.000

±0.225

±0.142

0 nnn

n niR

+n nfto

0 087

0 018

3

+0 174

+0 080

0.087

0.073

9

±0.174

±0.145

0 087

0 31 1

±0.174

±1.035

27

n 96n

n 49n

±0.332

±1.039

0.000

0.037

4

+0 159

0 nnn

0 037

5

+0.109

0.087

0.110

20

+0 174

+0 260

0.000

0.000

1

0.347

1.476

72

+ 1 322

0.000

0.256

21

±0.381

14.757

9.613

1097

±11.260

±7.648

0 000

6.340

213

±17.330

0.000

0.018

1

±0,080

0.087

0.347

38

+0.174

+0.889

15.020

14.730

1472

+ 11.380

±15.710

15.280

15.150

1514

±11.610

±15.710

0.087

0.037

5

±0,174

±0.109

(Continued)
J

445

Table A.— (continued).

Species

Young

Mature Old-wet Old-mesic Old-dry Total old Total
(11) (6) (9) (4) <19) Captures

Northern
alligator lizard
(Elgoria coeruleus)
Southern
alligator lizard
(E. mulficarinafus)
Western fence lizard
(Sceloporus
occidenfalis)
Total

Snakes
Northwestern
garter snake
(T. ordinoides)

0.035
±0.110

0.000

0.000

0.037
±0.110

0.000

0.000

0.000

0.032
±0.105

0.126
±0.321

0.000

0.000

0.000

0.000

0.058
±0.142

0.000

0.039
±0.116

0.039
0.116

0.000

0.077
±0.153

0.000

0.260
±0.521

0.000

0.174
±0.347

0.521
±0.601

0.087
+0.174

0.073
±0.248

0.018
±0.080

0.037
±0.159

0.164
±0.335

0.018
+0.080

14

Total

0.000

0.000

0.000

0.000

0.087

0.018

1

±0.174

±0.080

Total reptiles

0.035

0.126

0.058

0.077

0.608

0,183

15

±0.110

±0.321

±0.142

±0.153

±0.716

±0.390

All

9.791

12.877

8.680

19.527

15.888

15.333

1529

herpetofauna

±7.572

±8.408

±4.742

±20.888

±11.555

±15.694

'Mean ± 1 standard deviation.

^Data are from inland and coastal stands combined.

^Number of stands.

^Absent from inland stands.

^Absent from coastal stands.

The fact that we generally found
more species on older stands and
that we found a greater similarity
between mature and old-growth
stands than between either of these
older classes and young stands (see
also Raphael, this volume) suggests
that both the mature and old forest
age classes provide more suitable
habitat and a more diverse herpe-
tofauna than young forests.

Relative Abundance Analysis

Differences Between TCS and PF

A notable aspect of our data is the
differences between the TCS and PF
methods — both in terms of kinds of

species and numbers of individuals
captured. These differences follow
from the different natures of these
sampling methods. TCS is an active
search method that permits the in-
vestigator to seek out animals where
they hide. PF is a passive method
that relies on animal surface move-
ment or the seeking of shelter under
trap covers (Welsh 1987.)

The results of our comparisons of
salamander captures between coastal
and inland areas using TCS and PF
data, which appear contradictory,
serve to illustrate the pronounced
differences between the two meth-
ods. With TCS data, in all compari-
sons except the old-growth wet cate-
gory, the coastal area had higher
mean captures than the inland area.

This result was due to high captures
(over 900 individuals) of a single spe-
cies of salamander, Batrachoseps at-
tenuatus, a species that occurred in all
age and moisture classes. This spe-
cies is absent inland. However, sev-
eral factors unique to the inland area
acted to counter the effects of the
high captures of B. attenuatus. Those
factors were the high captures of Ple-
thodon elongatus (more than 250 cap-
tures), a species found almost exclu-
sively on the inland stands, and
higher relative captures of Ensatim
eschscholtzii inland (865 inland vs. 580
coastal).

In contrast, results from PF, indi-
cated significantly higher captures on
inland stands than on coastal stands,
for all stands combined (table Al).

446

PF captured few (n=72) of the highly
sedentary Batrachoseps attenuatus
relative to TCS (n=972). Captures of
the relatively more vagile salaman-
der species, P. elongatus and E. es-
chscholtzii, were greater on the inland
stands than the coastal stands, for the
PF data.

TCS provided a more complete
data set, sampled more species (par-
ticularly reptiles) and had twice as
many individuals as did PF (tables 2-
4). The active nature of TCS accounts

for the disparities in capture num-
bers, and in the lack of consistency of
statistically significant differences
among forest age and moisture
classes between these data sets, even
for the same species (table Al). Most
significant results from our analyses
derived from the larger TCS data set.
Subsequent discussion of results will
refer to these data unless they are
identified as PF data. Mean captures
(+ one standard deviation) for all
taxa analyzed are found in tables 3

Table 5.— Jaccard similarity index (JSI) values for species of herpetofauna
in 3 age classes of Douglas-fir forests of rK>rthwestem California and south-
western Oregon. Values were calculated using 10 randomly selected
stands from each forest age class, including coastal and inland areas.
Greater JSI values indicate greater similarity in species composition.

All stands

(Areas combined)

Young

Mature

Old-growth

Mature
Old-growth

Total number of species

.542
.467
16

.846
21

15

Table 6.— Mean (± 1 standard deviation) numbers of species of herpe-
tofauna among three age and three moisture classes of Douglas-fir forests

of northwestern California and southwestern Oregon.

Inland stands

Young Mature Old-dry Old-mesic

Old-wet

Number of

stands

3

6

3

6

3

Mean number

5.67

4.67

6.67

5.17

6.33

of species

±3.06

±1,51

±3.51

±1.33

±2.08

Total number

of species

10

13

14

12

10

Coastal stands

Number of

stands

8

5

1

4

3

Mean number

5.50

10.00

6.00

9,25

5.00

of species

±2.56

±2.92

±2.63

±2.00

Total number

of species

16

21

6

15

10

All stands

Number of

stands

11

n

4

10

6

Mean number

5.55

7.10

6.50

6.80

5.67

of species

±2.54

±3.51

±2.89

±2.78

±1.97

Total number

of species

17

23

17

17

14

V

and 4. Results of all tests on both
data sets, and test statistics for those
tests with significant differences, are
found in table Al.

Salannanders

Almost all captures (96.3%) were
salamanders (table 2), consequently,
the results of our analyses were es-
sentially the same for all herpe-
tofauna, amphibia, and salamanders
(species combined) (table Al). Sala-
manders were not equally distrib-
uted among forest age classes. Test-
ing the equality of mean captures
among age classes, with coastal and
inland areas combined, yielded sig-
nificant differences. Multiple com-
parisons (TU) indicated these differ-
ences were between the young and
old stands, with more captures on
the old stands (fig. 4).

Salamanders were not equally dis-
tributed among forest moisture
classes. Multiple comparisons
(GHMC), with areas combined, indi-
cated a significant difference in mean
captures between the old-growth
mesic and old-growth dry stands,
with more captures in the mesic
stands (fig. 4). These differences are
probably a result of the fact that drier
sites offer less equable habitat for
amphibians. We also captured fewer

at

I ♦ %

urn

I - X

wvuc (uati umc oavt ai sukb
STAND TTPt

Figure 3.— Numbers of species of herpe-
tofauna captured \n the coastal arxl inland
areas in three forest age and three forest
moisture classes of Douglas-fir dominated
forests from 1984-1986. Captures were by
time-constrained search (TCS) and pitfall
traps (PF).

447

amphibians on old-wet stands than
old-mesic stands, although the differ-
ence is not statistically significant.

Within the coastal area, multiple
comparisons (TU) indicated that both
mature and old-growth mesic stands
were significantly different from
young stands, but not from each
other, with the lowest mean captures
occurring on the young stands (fig.
5a). Between-area comparisons for
salamanders indicated a significant
difference in means between coastal
and inland mature stands (fig. 5a).

The PF data yielded no significant
differences between mean captures
in age or moisture classes with
coastal and inland areas combined or
within either area (table Al). How-
ever, comparisons between these ar-
eas indicated a significant difference
with all stands combined (fig. 5b).
The greatest differences occurred be-
tween the old-growth wet stands;
however the results were not signifi-
cant (fig. 5b).

The greater number of individuals
in older stands parallel our findings
of greater numbers of species in
older forest age classes (table 6). As
with the species richness analysis, the
number of individuals was greater in
older forests of the coastal area than
in the inland area. These differences
suggest that older forests support
both a richer and more abundant
salamander fauna.

The lower capture rates on old-
wet stands compared to old-mesic
was an unexpected result. We offer
two possible explanations for these
lower sample values. One possibility
is that the habitat structure is more
complex on these wet forest stands,
with more and larger downed
woody material, a thicker duff layer,
and denser undcrstory vegetation
requiring more time to search and
making it more difficult to find ani-
mals (TCS method) and making them
less likely to be moving about on the
surface and encountering our traps
(PF method). A second possibility is
that the wet stands actually contain
fewer salamanders.

Salamanders play an important
functional role in forest ecosystems
because of several unique aspects of
their ecology. Though they are small,
with 90% of species having adult
body masses less than those of small
birds and mammals (Pough 1980),
they are often a major portion of the
vertebrate biomass in a forest. At the
Hubbard Brook Experimental Forest
in New Hampshire, a single species
of salamander accounted for a
greater portion of biomass and sec-
ondary productivity than any other
vertebrate group (Burton and Likens
1975a,b). Their small size enables
them to exploit prey too small to be
used by birds and mammals and
subsequently to convert these prey
into biomass that is available to
larger vertebrates (Pough 1983).
Pough et al. (1987) cites both direct
observations of predation and the
ubiquity of defensive mechanisms
among salamanders as evidence of
their importance as a food source for
both avian and mammalian preda-
tors. Because salamanders are ectoth-
erms and have the lowest metabolic
rates of any terrestrial vertebrates
(Feder 1983), this biomass conversion
process is extremely efficient, with
40-80% of the energy invested being
used to produce new biomass
(Pough et al. 1987). As a consequence
of these characteristics, salamanders
are quantitatively and qualitatively
important components of food webs

of many forest ecosystems. The fact
that their numbers appear to be re-
duced by certain forest practices
could potentially affect energy flow
and biomass production at all bio-
logical levels.

Frogs

Testing the equality of mean captures
yielded significant differences in cap-
tures of frogs in coastal age and
moisture classes, with significantly
higher mean captures in old vs.
young stands and mesic vs. wet
stands (table Al). These results are
attributable to a single species, the
Pacific treefrog. No other significant
differences were found (table Al).

17 1

osMwc looav ~j~
4 ■ siK I - xa

IS ' KM KM

a: 'J-

fOJM WTME OU KT OU KSK «fi OT

STAWD TYPE

Figure 4.— Captures of salarrtanders per
person-hour (TCS) in tt^ree forest age and
three moisture classes. Data are from the
coastal and inland areas combined, and
sampling occurred from 1984-1986.

f < mi

misM wrna > cnei*. loac

« ■ \nt
emeu. OU <tx > Cttsm nuc
« - loi
Km

0COHSTAI.
HJttl

ICM t SC

new
HUN - X

u 90
z
I

HXJNC IHROT 010 HC OU) 1£SC OLD DBT MX SIAMDS

STAND TYPE

L sua ><ii s

I - 2JI

» . »II

T

mm « z

ION

YTKMO iw\j*i (uwi aamx aeon usunm
STAHDTYPE

Figure 5.— Captures of salamanders per person-hour (A:TCS) and per 1000 trap-nights (B:PF)
in the coastal and inland areas. Data are from 1984-1986 aCS) and 1984-1985 (PF).

448

Reptiles

The reptile fauna in the forests of the
Pacific Northwest is depauperate
(Nussbaum et al. 1983, Stebbins 1985)
with most species occurring in rela-
tively low abundance (tables 3-4).
Distribution of reptile species, by age
and moisture class, indicated about
equal numbers of species in the
young, mature, and old-growth age
classes, with lower numbers of spe-
cies in old-growth wet forests.

Based on TCS and PF data, our
mean captures of reptiles (species
combined) were higher on both drier
and older stands, but the differences
were not statistically significant. Our
sample sizes were not sufficient to
analyze for differences among age
and moisture classes at the species
level, except for the northern alliga-
tor lizard for which our data indi-
cated no statistically significant asso-
ciation with a particular forest age or
moisture class (table Al).

We did not sample in any recently
harvested areas, but given their pref-
erences for open areas and their re-
lated heliothermic natures, reptiles,
particularly lizards, probably in-
crease following logging, and
through the early serai stages of re-
generating forests (see Raphael, this
volume). Raphael and Marcot (1986)
indicated that the sagebrush lizard
(Sceloporus graciosus) was four times

a» IOC > lOK

a* icac> M 1CT

42}

I

I

jTuoc CLD w <u neve cm I
STAND TYPE

Figure 6.— Captures per person-hour (TCS)
of the Pacific treefrog (Hyla regilla), in three
forest age and three moisture classes. Data
are fronn the coastal area from 1 984- 'i 986.

as abundant in early vs. late shrub
stages.

Relative Abundance of Connmon
Species

Common species (captured on at
least one third of our stands in either
area by either sampling method)
were analvzed for differences in
mean captures in age and moisture
classes, across, within and between
coastal and inland areas (table Al).
Besides the northern alligator lizard,
these species consisted of amphibi-
ans— 2 frogs and 7 salamanders.
Other amphibians whose distribu-
tions relative to forest age were con-
sidered noteworthy are also dis-
cussed.

Yellow-Legged Frog (Rana
boylii). — This species was absent
from all young stands (table 4), but
they were also captured at such low
frequencies on our inland stands as
to preclude analyses within this area.
Within the coastal area, no significant
differences were found for capture
frequencies of this species in forest
age or moisture classes (table Al).

The yellow-legged frog is a highly
aquatic species (Stebbins 1985) and
therefore our PF captures (table 4)
must be considered incidental. These
captures may have been frogs seek-
ing terrestrial overwintering cover
above high water levels (PF sampling
was done in the fall). However, this
frog was absent from young stands.
Three facts need be considered: (1)
all but a single capture occurred in
the coastal area; (2) in general, the
coastal stands were closer to peren-
nial streams and creeks than were
the inland stands; (3) within the
coastal area, only two out of eight
young stands had PF grids near suit-
able aquatic habitat, whereas all the
mature and old-growth stands had
PF grids near such habitat. Thus we
can not rule out the possibility that
this frog's absence from young
stands in our sam.ples is an artifact of
our stand locations relative to avail-

able and suitable aquatic habitat
(Bury and Corn, this volume).
Twenty-one records from area-con-
strained aquatic surveys (H. Welsh,
unpubl. data) were almost equally
divided between creeks in young and
mature forests. On the other hand, it
is possible that older forests provide
some particulars of microhabitat re-
quired by overwintering yellow-
legged frogs not present in young
forests.

Pacific Treefrog (Hyla regilla). —
The Pacific treefrog is the only frog
for which our data indicated signifi-
cant differences in captures between
both forest age and moisture classes
(fig. 6). Within the coastal area, this
frog was captured at significantly
different frequencies in both forest
age and moisture classes. However,
these differences were not observed
within the inland area, probably due
to lower captures and higher vari-
ances on these stands (table Al).

Because the Pacific treefrog is not
restricted to forested habitat (Steb-
bins 1985), we are suspicious of our
data indicating greater abundance in
older forests (fig. 6). Conceivably
older forests provide more cover and
foraging areas for this species than
do young forests and thus support
higher relative abundances. Most of
our captures of treefrogs occurred in
association with large downed
woody material. However, we can-
not rule out the ]X)ssible influence of
proximity of breeding sites on these
results (Bury and Com, this volume).
The older forest stands were gener-
ally closer to standing water than the
young stands (as with Ram boylii) in
the coastal area.

The difference in captures of
treefrogs between the mesic and wet
moisture classes (fig. 6) may be an
artifact of unequal detectability. Most
treefrogs were captured by TCS and
they are more easily exposed and
seen by investigators in the more
open understory of the mesic stands.
The alternate possibility, that there
are actually more treefrogs on mesic
stands, is consistent with the in-

449

creased incident radiation in the me-
sic stands which would promote
higher productivity of invertebrate
prey, and thus possibly support
more treefrogs.

The Tailed Frog (Ascaphus
truei), — This frog was captured only
on mature and old-growth stands
(tables 2-3); however, the total num-
ber of captures (5) was too low for
statistical tests. This species is of
interest, nonetheless, because of its
absence from young stands. The
tailed frog, like the yellow-legged
frog, is highly aquatic (Bury 1968,
Stebbins 1985). Therefore these rec-
ords based on terrestrial sampling
are considered incidental. However,
results from another study employ-
ing an area-constrained aquatic sam-
pling method yielded more than 400
captures of tailed frogs (Welsh, in
prep.). These data were consistent
with the incidental records reported
here; there were significant increases
in tailed frog abundance with in-
creased forest age.

Olympic Salamander (Rhyacotri-
ton olympicus). — This species was
absent from all young stands (tables
3-4). Low captures prompted us to
combine moisture classes for the age
analysis. Multiple comparisons
(GHMC), coastal and inland areas
combined, indicated that older
stands had significantly greater num-
bers of Olympic salamander than
young stands (fig. 7).

This species is restricted to head-
water habitats, such as seeps,
springs, and small creeks in forests
where it prefers cold water flowing
over rocky substrates (Anderson
1968, Nussbaum et al. 1983). Because
of the relative scarcity of this mi-
crohabitat in the areas of our study,
Rhyacotriton occurs in a patchy distri-
bution. It can be abundant where
conditions are suitable, but we found
appropriate microhabitat islands for
this species to be few, small, and
widely scattered on our stands. This
resulted in relatively few captures
(tables 3-4). We found Rhyacotriton
absent in younger forests (fig. 7),

which is consistent with results from
other studies (Bury 1983; Bury and
Corn 1988; Welsh, in prep.). This spe-
cies appears to be sensitive to forest
harvest practices because of its par-
ticular habitat requirements (Bury
and Corn 1988; Welsh, in prep.). Cur-
rent harvest practices do not protect
headwater habitats. Such habitats are
often radically altered by harvest
practices, which can change water
flow and temperature, increases sedi-
ment loads, and change the structure
and composition of the riparian
vegetation (Bury and Corn 1988). The
result of these changes is often the
extirpation of local populations of
this species.

Clouded Salamander (Aneidesfer-
reus). — Multiple comparisons
(GHMC) indicated significant differ-
ences in mean captures of clouded
salamanders between young and old
stands in the inland area but not in
the coastal area (fig. 8a). Testing for
differences with coastal and inland
areas combined revealed significant
differences in mean captures among
moisture classes; multiple compari-
sons (TU) indicated that the mesic
stands had significantly higher mean
captures than did dry stands (fig.
8b).

This species, a habitat specialist,
occurs most often under exfoliating
bark on downed conifer logs (Steb-
bins 1985, Nussbaum et al 1983). At
several coastal redwood localities.

z

(3 i-o

c

3

I - SO
P < GB

I

I

UIM * SE
UCM

VlUi - St

rOREST ACE CUSS

Of ncsc)

Bury (1983) and Bury and Martin
(1973) found it to be more abundant
in young stands than older stands.
They attributed the differences to an
increase in bark on downed woody
material from logging. Our data from
the coastal area (fig. 8a) indicated
slightly more A. ferreus in younger
than older forests, but the differences
were not significant. However, in the
inland area the clouded salamander
was found in significantly higher
numbers on old vs. young stands
(fig. 8a). We suspect that these differ-
ences are due to the differences in
moisture regimes between the two
areas. This idea is supported by our
findings of significant differences in
capture means between mesic and

OJ > KM

Km

WW « a

mm
ia« - s

roREsr AC£ cuss

Figure 7.— Captures per person-hour (TCS)
of the Olympic salamander (Rhyacotriton
olympicus), in three forest age classes.
Data are from the inland and coastal areas
combined, from 1984-86.

T

■UHM

MM 4 a

ISM
ICJM - X

rOtEST MOtSTURt CLASS

Figure 8.— (A) Captures per person-hour (TCS) of the clouded salamander (Aneldes ferreus)
in the coastal and inland areas, in three forest age classes. (B) Captures per person-hour
(TCS) in three forest moisture classes; data are from coastal and inland areas combined
Sampling occurred from 1984-86.

450

dry old-growth sites (fig. 8b). We
suggest that logs on inland young
stands are subjected to higher
evapotranspiration rates than are
logs on old-growth stands because of
greater incident radiation. Possible
increases in clouded salamanders on
young stands from an increase in
slash and logs after harvesting may
be outweighed by the loss of suitable
microclimatic conditions due to in-
creased exposure.

Black Salamander (Aneides
flavipunctatus). — We found signifi-
cantly greater numbers of this spe-
cies in the coastal area than in the in-
land area (fig. 9). Lynch (1981)
pointed out that inland populations
occur in a patchy distribution charac-

0a»sm.

ICW

•ntii mniK tUKS as tea: cuar «i stmos
STAND TYPE

Figure 9.— Captures per person-hour (ICS)
of the black salamander (Aneides flavip-
unctatus) in coastal and inland areas, in
three forest age and three nnoisture classes.
Data are fronn 1984-86.

-r I

I

I

MCNI - 3
I—

ST*KD TYPt

teristic of a species on the decline.
Further, he attributed the inland
patchiness to climatic constraints and
noted that the black salamander is
restricted to low-lying suitable areas
receiving at least 75 cm of annual
precipitation. Its restriction to rocky
habitats and its low relative abun-
dance in northwestern California
preclude drawing any conclusions
from our forest age and moisture
class analysis (table Al).

California Slender Salamander
(Batrachoseps attenuatus). — The
slender salamander, like the black
salamander, appears to be restricted
to low-lying suitable areas with rela-
tively high annual precipitation
(Maiorana 1976a). This species was
absent from our inland sites, but ac-
counted for the highest captures of
any species within the coastal area.
This was one of the few species we
captured in sufficient numbers with
both sampling methods to test both
data sets for differences between for-
est age and moisture classes (see
table Al). Within the coastal area,
both TCS and PF data indicated sig-
nificant differences in mean captures
among forest age classes (figs. lOa-b).
Multiple comparisons (TU) indicated
that these differences were between
both young and mature and young
and old-growth stands (figs. lOa-b).
Our findings here were consistent
with trends found by others (Bury
1983, Bury and Marrin 1973).

<l

1.0

u
u
u

\s

1.0

u

&0

I

h

1

now • St

low
IC» - 3

HNIUI

STAHO TYPt

Figure 10.— Captures per person-hour (A:TCS), and captures per 1000 trap-nights (B:PF). of
the California slender salamander (Batrachoseps atfenuatus), in three forest age and three
moisture classes. Data are from the coastal area from 1984-86 (TCS) and 1984-85 (PF).

The PF data showed a significant
difference between captures in mois-
ture classes, with a higher mean cap-
tures on mesic than on old-growth
wet stands (fig. 10b), but the TCS
data did not (fig. 10a). For a salaman-
der species whose presence and rela-
tive abundance is correlated with
relatively high and predictable mois-
ture (Maiorana 1974, 1976a), this re-
sult is unexpected and may be an ar-
tifact of different sampling efficien-
cies between forest moisture classes.
The old-growth wet stands appear to
contain habitat with relatively great
structural complexity: a thick and
complex layer of understory, decom-
p>osing woody material, and mossy
duff. Such habitat provides abundant
microhabitat for a ground dwelling
and semi-fossorial species like the
slender salamander.

Slender salamanders may not fre-
quent the surface as much to forage
as they would on drier stands. Forag-
ing in more protected areas would
reduce exposure to predation and
thus incur a selective advantage.
Maiorana (1976b) termed this sub-
mergent behavior (our concept is a
slight variation of her idea; she hy-
pothesized that a species might actu-
ally forage less at times to avoid ex-
posure to predation). As a result of
less surface activity, fewer slender
salamanders are captured in the pit-
fall traps. The same logic can also be
applied to the TCS method, in which
lower captures would be expected in
the structurally more complex habi-
tat per unit of search time. With TCS,
we did get slightly lower captures on
old-growth wet stands for this spe-
cies (table 3), but the active nature of
TCS allowed us to detect enough
slender salamanders that the capture
rates between moisture classes were
not significantly different.

Ensatina (Ensatina es-
chscholtzii). — Ensatina has broad
ecological tolerances, occurring from
relatively dry woodland habitats to
moister forests at high elevations
(Stcbbins 1954). This species has the
most extensive geographic distribu-

451

tion of all the western woodland
salamanders, ranging from British
Columbia to Baja California (Stebbins
1985). Ensatina were captured in the
highest numbers of any species we
sampled (table 3-4). There were sig-
nificant differences in mean captures
among forest age classes, with
coastal and inland areas combined
(fig. 11). Multiple comparisons (TU)
indicated that old stands had signifi-
cantly higher captures than young
stands (fig. 11).

Both PF and TCS data indicated
significant differences in mean cap-
ture frequencies between the coastal
and inland areas (figs. 12a-b). Greater
numbers were found on the inland
stands. These differences between
areas indicate that this species may
be more abundant in the drier inland
area than along the coast.

Del Norte Salamander (Plethodon
elongatus). — Except for three cap-
tures from our most northern coastal
stand, this species was sampled only
on our inland stands. These salaman-
ders are found primarily on or in
rocky substrates (Stebbins 1985,
Nussbaum et al. 1983), and reach
high densities in talus and outcrops
of fractured metamorphic rock. Such
habitats were not present on some of
our stands. Also, our study region
encompassed the geographic range
of this species, and all of our south-
ern and some of our easternmost
stands were beyond its geographic
limits. Despite the patchy distribu-
tion of this species due to habitat re-
strictions, and absence from sites be-
yond its range, both methods indi-
cated a higher relative abundance on
older forest stands and a lower rela-
tive abundance on drier stands (figs.
13a-b, tables 3-4). These differences
were not statistically significant;
something we attribute to high vari-
ances within forest age classes result-
ing from this lack of appropriate mi-
crohabitat and the inclusion of stands
beyond the range (table Al). A sepa-
rate analysis of only stands from
within the geographic range of the
Del Norte salamander indicated that

the abundance of this species is sig-
nificantly correlated with increased
forest age (Welsh, in prep.).

Rough-Skinned Newt (Taricha
granulosa). — Both TCS and PF
showed a marked increase in cap-
tures of this species in older forests
(figs. 14a-b, tables 3-4). Lack of statis-
tically significant differences in cap-
tures between forest age classes
(table Al) is probably related to spe-
cific habitat requirements of this spe-
cies. We suspect that the critical habi-
tat component was proximity to
creeks or ponds, a breeding require-
ment for this species (Stebbins 1985).
Many of our stands, particularly
within the inland area, were a con-
siderable distance from suitable
breeding habitat for this newt. We
had no TCS captures of this species
on old-growth stands in our inland
area, yet the rough-skinned newt is
common there (Stebbins 1985, pers.
observ.).

CONCLUSIONS

Our research indicates that salaman-
ders comprise the majority of both
sf)ecies and individuals among the
herpetofauna of the Douglas-fir/
hardwood forests of northwestern
California and southwestern Oregon.
We found species diversity of the to-
tal herpetofauna to be greater in
older forest age classes. Amphibians,

•UM cnsui.
u am > «i siucs

1 . 213

]CtMSM
J fUM

h

«« < St

ISW - 5E

rate IMUt CU KT OH) UCSK OLD Om AU SUMS

STAMO lYPe

particularly salamanders, were sig-
nificantly more abundant in older
forests and significantly less abun-
dant in drier forests.

We found the TCS method, ac-
tively searching for animals in their
preferred microhabitats (usually as-
sociated with downed woody mate-
rials in these forest habitats), yielded
more useful data on herpetofaunal
diversity and abundance relative to
forest age and moisture class than
did PF. The TCS method sampled
more individuals and species in ad-
dition to taking less time and ex-
pense than PF (see Welsh 1987).

Recent research in forested habi-
tats (Bury and Corn 1988, Pough et
al. 1987, Enge and Marion 1986, Bury

ICAM 4 9E
ICM

K3MS WIUC an 0C9C)

roRCST Acc cuss

Figure 11.— Captures per person-hour (ICS)
of Ensatina (Ensatina eschschotltzii). In
three forest age classes. Data are from
coastal and inlartd areas combined, from
1984-86.

T »

CO»SI»L

mi > ai sua

I • IM

t' aa

MM « a

KAN
ICM - X

yoMC MTum. 010 vt ou icac <u on ta sumr
STWe TYPE

Figure 12.— Captures per person-hour (A:TCS), and captures per 1000 trap-nights (B:PF), of
Ensatina (Ensatina eschschotltzii) in coastal and inland areas, in three forest age and three
moisture classes. Data are from 1984-86(TCS) and 1984-85 (PF).

452

1983, Bennett et al. 1980, Bury and
Martin 1973) has indicated a pattern
of fewer species and reduced abun-
dance of herpetofauna after logging.
We also found lower numbers of
both species and individuals on
younger stands.

Greater species diversity and
greater relative abundance, for most
species, on mature and old-growth
stands may be related to greater
structural complexity in older forests
(Franklin and Spies 1984, Franklin et
al. 1981). Older forests also have a
narrower and more stable range of
moisture and temperature than pre-
canopy and young forests (Bury
1983, Harris 1984). Bury (1983)
sampled amphibians on four paired

plots in coastal redwood forest, each
pair consisting of a logged and an
old-growth forest stand. He attrib-
uted the lower diversity and relative
abundance of amphibians on the
logged sites to microclimatic differ-
ences. Bury (1983) also found higher
numbers of amphibians associated
with a greater volume of downed
woody material, but he considered
these differences in cover habitat to
be of secondary importance. Re-
cently, Bury and Corn (this volume)
found that coarse woody debris is
related to salamander occurrence
and abundance in the Oregon and
Washington Cascades.

We believe that structural com-
plexity or spatial heterogeneity (Pi-

B

X

I

MCUI - SE

iMno tuTuK OLD «n as not OU) MY

Sim TYP£

5 30
!£!

in

HEW « X

WW - SE

TMNC lunnc OLD «rr out Mac ou orr
STAND TYPE

Figure 13.— Captures per person-hour (A:TCS), and captures per 1000 trap-nights (B:PF), of
the Del Norte salamander (Plethodon elongatus), in three forest age and three moisture
classes. Data are from the inland area, from 1984-86 (\CS) and 1984-85 (PF).

X

KM 4 Z

UNftJLW

route lur.JM OLD WT 01* MISIC OLD c

SIAHO TYPt

8

o 1:6

X

itMt - 9

IUT1JI 0U> ITT OLD IC9C

SUNO TYPC

Figure 14.— Captures per person-hour (A:TCS), and captures per 1000 trap-nights (B:PF), of
the rough-skinned newt (Taricha granulosa), in three forest age and three moisture classes.
Data are from coastal and inland areas combined, from 1984-86 (ICS) and 1984-85 (PF).

anka 1966) plays an important role in
promoting the addition of species
and numbers of individuals in older
forests. Downed woody material, be-
sides affording cover, creates micro-
climatic pockets that can act to buffer
the moisture and temperature fluc-
tuations in the forest at large, and it
provides protection from predation
as well. Maiorana (1978) reported
that space (small cavities and bur-
rows) was more important in regu-
lating relative abundance between
two sympatric salamanders (Aneides
luguhris and Batrachoseps attenuatus)
than competition for food resources.
Therefore, more salamander species
and individuals should be expected
in more structurally complex habi-
tats. In fact, both microclimate and
cover are probably interrelated, ulti-
mate factors (Baker 1938) determin-
ing habitat suitability for temperate
forest herpetofauna. Both are clearly
affected by forest harvest practices
and probably jointly account for
most of the differences in diversity
and abundance observed in the her-
petofauna between young, mature,
and old-growth forests in northwest-
ern California and southwestern Ore-
gon.

ACKNOWLEDGMENTS

We thank the members of the field
crews of the Pacific Southwest Forest
and Range Experiment Station's Tim-
ber/Wildlife Research Unit for their
help in collecting data; James A.
Baldwin and Barry R. Noon for ad-
vising on statistical methods; C. John
Ralph, R. B. Bury, and M. G. Raphael
for their reviews of the manuscript;
and Dana L. Waters for his help with
the figures and tables in the manu-
script.

LITERATURE CITED

Anderson, James D. 1968. Rhyacotri-
ton and R. olympicus. American
Society of Ichthyologists and Her-

453

petologists. Catalog of American
Amphibians and Reptiles 68.1-
68.2.

Axelrod, Daniel 1. 1976. History of
coniferous forests, California and
Nevada. University of California
Publications in Botany 70:1-62.

Baker, J.R. 1938. The evolution of
breeding seasons, p. 161-177. In
Evolution, essays presented to E.S.
Goodrich. G.R. de Beer, editor.
Oxford University Press, Oxford.

Bennett, Stephen H., J. Whiti^ield Gib-
bons, and Jill G. Glansville. 1980.
Terrestrial activity, abundance,
and diversity of amphibians in dif-
ferent managed forest types.
American Midland Naturalist
103(2):412-416.

Burton, Thomas M., and Gene E. Lik-
ens. 1975a. Energy flow and nutri-
ent cycling in salamander popula-
tions in the Hubbard Brook Ex-
perimental Forest, New Hamp-
shire. Ecology 56:1068-1080.

Burton, Thomas M., and Gene E. Lik-
ens. 1975b. Salamander popula-
tions and biomass in the Hubbard
Brook Experimental Forest, New
Hampshire. Copeia 1975:541-546.

Bury, R. Bruce. 1968. The distribution
of Ascaphus truei in California.
Herpetologica 24(l):39-46.

Bury, R. Bruce. 1983. Differences in
amphibian populations in logged
and old-growth redwood forest.
Northwest Science 57(3):167-178.

Bury, R. Bruce and Paul S. Corn.

1987. Evaluation of pitfall trapping
in northwestern forests: trap ar-
rays with drift fences. Journal of
Wildlife Management 51:112-119.

Bury, R. Bruce and Paul S. Corn.

1988. Responses of aquatic and
streamside amphibians to timber
harvest, (in press). In Streamside
management: riparian wildlife and
forestry interactions. K.J. Raedeke,
editor. College of Forestry Re-
sources, University of Washing-
ton.

Bury, R. Bruce, Roger A. Lucken-
bach, and Stephen D. Busack.
1977. Effects of off-road vehicles
on vertebrates in the California

desert. U.S. Department of the In-
terior, Fish and Wildlife Service,
Report 8, 23 p.

Bury, R. Bruce and Michael Martin.
1973. Comparative studies on the
distribution and foods of Pletho-
dontid salamanders in the red-
wood region of Northern Califor-
nia. Journal of Herpetology
7(4):331-335.

Busack, Stephen D., and R. Bruce
Bury. 1974. Some effects of off-
road vehicles and sheep grazing
on lizard populations in the
Mojave Desert. Biological Conser-
vation 6(3):179-183.

Corn, Paul S., and R. Bruce Bury. (In
prep.). Sampling terrestrial am-
phibians and reptiles. In Monitor-
ing techniques for wildlife in Pa-
cific Northwest forests. Andrew B.
Carey and Leonard F. Ruggiero,
editors. U.S. Forest Service, Gen-
eral Technical Report.

Enge, Kevin M., Wayne R. Marion,
1986. Effects of clearcutting and
site preparation on herpetofauna
of a north Florida flatwoods. For-
est Ecology and Management
14:177-192.

Feder, Martin E. 1983. Integrating the
ecology and physiology of pletho-
dontid salamanders. Herpetol-
ogica 39:291-310.

Franklin, Jerry F., Kermit Cromack,
Jr., William Denison, Arthur
McKee, Chris Maser, James Sedell,
Fred Swanson, and Glen Juday.
1981. Ecological characteristics of
old-growth Douglas-tir forests.
U.S. Department of Agriculture,
Forest Service, General Technical
Report PNW-118, 48 p. Pacitic
Northwest Forest and Range Ex-
periment Station, Portiand, Oreg.

Franklin, Jerry F., F. Hall, W.
Laudenslayer, Chris Maser, J.
Nunan, J. Poppino, C. John Ralph,
and Thomas Spies. 1986. Interim
definitions of old-growth Douglas-
fir and mixed conifer forests in the
Pacific Northwest and California.
U.S. Department of Agriculture,
Forest Service, PNW Research
Note PNW-447, 7 p.

Franklin, Jerry F. and Thomas Spies.
1984. Characteristics of old-
growth Douglas-fir forests, p. 328-
334. In Proceedings, Society of
American Foresters national con-
vention. [October 16-20, 1983],
Portland, Oregon.

Harris, Larry D. 1984. The Frag-
mented Forest. 211 p. University
of Chicago Press.

Jones, K. Bruce. 1981. Effects of graz-
ing on lizard abundance and di-
versity in western Arizona. South-
western Naturalist 26(2):107-115.

Keselman, H.J. and Joanne C. Rogan.
1978. A comparison of the modi-
fied-Tukey and Scheffe methods of
multiple comparisons for pairwise
contrasts. Journal of the American
Statistical Association 73(361 ):47-
52.

Luckenbach, Roger A. and R.Bruce
Bury. 1983. Effects of off-road ve-
hicles on the biota of the Al-
godones Dunes, Imperial County,
California. Journal of Applied
Ecology 20:265-286.

Lynch, James F. 1981. Patterns of on-
togenetic and geographic variation
in the black salamander. Amides
flavipunctatus (Caudata: Pletho-
dontidae). Smithsonian Contribu-
tions to Zoology 324:1-53.

Maiorana, Virginia C. 1974. Studies
in the behavioral ecology of the
plethodontid salamander Batra-
choseps attenuatus. Ph.D. disserta-
tion. University of California,
Berkeley.

Maiorana, Virginia C. 1976a. Size and
environmental predictability for
salamanders. Evolution 30(3):599-
613.

Maiorana, Virginia C. 1976b. Preda-
tion, submergent behavior, and
tropical diversity. Evolutionary
Theory 1:157-177.

Maiorana, Virginia C. 1978. Differ-
ences in diet as an epiphenom-
enon: space regulates salaman-
ders. Canadian Journal of Zoology
56(5):1017-1025

Meslow, E. Charles, Chris Maser,
and Jared Verner. 1981. Old-
growth forests as wildlife habitat.

454

Transactions of the North Ameri-
can Wildlife and Natural Resource
Conference 46:329-335.

Milliken, G.A. and D.E. Johnson.
1984. Analysis of Messy Data.
Wadsworth Inc. Belmont, Califor-
nia. 473 p.

Miller, Rupert G., Jr. 1981. Simultane-
ous Statistical Inference, 2nd. ed.
Springer-Verlag, New York. 299 p.

z- Nussbaum, Ronald A., Edmund D.

Brodie, Jr., and Robert M. Storm,
h- 1983. Amphibians and Reptiles of

the Pacific Northwest. University
n, of Idaho Press, Moscow Idaho. 332

P-

of Ortega, A., M.E. Maury, and R. Bar-
se bault. 1982. Spatial organization

n and habitat partitioning in a

mountain lizard community of
Mexico. Ecol. Gen. 3(3):323-330.

Pianka, Eric R. 1966. Convexity, des-
ert lizards, and spatial heterogene-
ity. Ecology 47:1055-1059.

Pough, Harvey F. 1980. The advan-
tages of ectothermy for tetrapods.
American Naturalist 115:92-112.

Pough, Harvey F. 1983. Amphibians
m and reptiles as low-energy sys-

tems, p. 141-188. In Behavioral En-
ergetics the Cost of Survival in
Vertebrates. W.P. Aspey and S.I.
Lustick, editors., Ohio State Univ.
Press, Columbus, Ohio.

Pough, Harvey F., Ellen M. Smith,
Donald H. Rhodes, and Andres
Collazo. 1987. The abundance of
salamanders in forest stands with
different histories of disturbance.
] Forest Ecology and Management

20:1-9.

Raphael, Martin G. 1984. Wildlife
populations in relation to stand
age and area in Douglas-fir forests
of northwestern California, p. 259-
274. In Fish and Wildlife Relation-
ships in Old-Growth Forests. [Jun-
eau, Alaska, April 12-15, 1982]
W.R. Meehan, T.R. Merrelly Jr.,
and T.A. Hanley, editors.. Pro-
ceedings of the American Institute
of Fishery Research Biologists.
1 Raphael, Martin G. 1987. Wildlife-
' tanoak associations in Douglas-fir

forests of northwestern California.

In Multiple-use management of
California's hardwood resources.
Plumb, T.R. and N.H. Pillsbury,
editors.. General Technical Report.
PSW-100. Pacific Southwest Forest
and Range Experiment Stations,
U.S. Department of Agriculture,
Forest Service. Berkeley, Califor-
nia.

Raphael, Martin G. In press.

Douglas-fir (DFR). In A Guide to
Wildlife Habitats of California. K.
Mayer and W.F. Laudenslayer, Jr.,
editors.

Raphael, Martin G., and Bruce G.
Marcot. 1986. Validation of a
Wildlife-Habitat-Relationships
model: vertebrates in a Douglas-fir
sere. p. 129-138. In Wildlife 2000:
modeling habitat relationships of
terrestrial vertebrates. Jared
Vemer, Michael L. Morrison, and
C. John Ralph, editors. University
of Wisconsin Press, Madison.

Ruggiero, Lenard F. and Andrew B.
Carey. 1984. A programmatic ap-
proach to the old-growth forest:
wildlife relationships, p. 340-345.
In Proceedings of 1983 convention
of the Society of American Forest-
ers, Bethesda, MD.

Sawyer, John O. and Dale A. Thorn-
burgh. 1977. Mixed evergreen for-
est, p. 59-382. In Terrestrial Vege-
tation of California. M.G. Barbour
and J. Major, editors. J. Wiley and
Sons, New York.

Simberloff, Daniel 1987. The spotted
owl fracas: mixing academic, ap-
plied and political ecology. Ecol-
ogy 68(4):766-779.

Sneath, P.H.A. and Robert R. Sokal.
1973. Numerical Taxonomy. W.H.
Freeman Co. San Francisco, Calif.
573 p.

Spies, Thomas A., Jerry F. Franklin,
and Ted B. Thomas. Coarse
woody debris in Douglas-fir for-
ests of western Oregon and Wash-
ington. Ecology. (In Press.)

Stebbins, Robert C. 1954. Natural his-
tory of the salamanders of the ple-
thodontid genus Ensatina. Univer-
sity of California Publications in
Zoolology 54(2):46-123.

Stebbins, Robert C. 1985. A Field
Guide to Western Reptiles and
Amphibians. 336 p. Houghton
Mifflin Company, Boston, MA.

Welsh, Hartwell H., Jr. 1987. Moni-
toring herpetofauna in woodland
habitats of Northwestern Califor-
nia and southwestern Oregon: a
comprehensive strategy, p. 203-
213. In Multiple-use management
of California's hardwood re-
sources. Timothy R. Plumb and
Norman H. Pillsbury, editors.
USDA Forest Service General
Technical Report PSW-100. Pacific
Southwest Forest and Range Ex-
periment Station, Berkeley, CA.

Welsh, Hartwell H., Jr. (In Prep.). Re-
lictual amphibians and old-growth
forests.

Zar, Jerrold H. 1984. BiostaHstical
Analysis. Prentice-Hall, Inc. Engle-
wood Cliffs, N.J. 718 p.

455

Table Al .-Comparisons of mean capture frequencies of herpetofauna, captured by two sampling methods time-
clo^aZ, Nand ^^eT' '"^^ c^sas":nd^|rin

Spedes richness
TCSandPF
data combined

multiple
comparisons:
All herpetofauna
TCS data

multiple
comparisons:

PF data

multiple

comparisons:
Reptiles

TCS data

multiple

comparisons:
All lizards

TCS data

multiple

comparisons:
Elgaria
coeruleus

TCS data

multiple

comparisons:
All snakes

TCS data

multiple

comparisons:
Amphibia

TCS data

multiple
comparisons:

PF data

multiple
comparisons:
All frogs
TCS data

multiple
comparisons:
PF data
multiple
comparisons:

Inland stands Coastal stands Comparisons between coastal and Inland stands Stands combined

Moisture Age Moisture

All All Old

je stands old wet

F=5.20 C>l

.025>P>.01 t=2.13
P=.04

mature>young
q=4.160, P<.05

F=10.75 * +
.0025>P>.001

old>young
q=6.035, P<.05
mature>young
q=4.558, P<.05

l>C
t=2.44
P=.023

F=9.74
.005>P>.0025

old>young, q=5.091
mature>young, q=4.004
P<.05

l>C
t=2.43
P=.024

mesiowet
t=4.34
P=.023

old>young
t=3.97, P<.05

Old

mesic Mature Young Moisture Age

C>l

C>l
t=3.70
P=.01

+

t=5.51
P<.0001

C>l
1=5.061
P<.0001

F=3.87

.05>P>.02

mesiodry old>young
t=3.22 q=3.860
P<.05 P<.05

F=4.27
.025>P>.01

mesiodry old>young
t=3.49 q=3.998
P<.05 P<.05

(Continued) ^

456

Table A 1 . — (continued).

Inland stands Coastal stands Comparisons between coastal and Inland stands Stands combined

All All

Moisture Age Moisture Age stands old

Old Old

wet mesic Mature Young Moisture Age

Hyla regilla
TCS data

multiple

comparisons:
Rana boylii

PF data

multiple

comparisons;
All Salamanders

TCS data

multiple
comparisons:

PF data

multiple

comparisons:
Rhyacotriton olympicus

TCS data

multiple

comparisons:
Aneides ferreus
TCS data

multiple
comparisons:

Aneides flavipunctatus
TCS data ^

multiple
comparisons:
PF data
multiple
comparisons:
Batrachoseps attenuatus
TCS data 4

multiple

comparisons:

PF data *

multiple
comparisons:
Ensatina eschscholtzii
TCS data

multiple
comparisons:

PF data

meslowet
t=4.34,P=.023

okJ>young
t=3.97,P<.05

old>young
t=3.42, P<.05

F-8.67
.005>P>.0025

old>young, q=5.601
mature>young, q=3.706
P<.05

l>C
t=2.59
P=.017

C>l +
t=2.12,P=.04

F=5.82 *
.05>P>.025
old>young, q=3.836, P<.05
mature>young, q=4.108, P<.05
mesiowet F=10.94 *
t=3.62 .0025>P>.001
P=.022

old>young,q=5.799, P<.05
mature>young, q=5.188, P<.05

l>C
t=2.13.P=.04

l>C

t=3.34,P=.003

C>l
t=5.091
P<.0001

F-4.26
.025>P>.01

meslodry old>young
t=3.42 q=3.970
P<.05 P<.05

old>young
t=2.57,P<.05

+ F=4.45
.05>P>.025

meslodry
q=3.903, P<.05

4 4

F=3.72
.05>P>.025

old>young
q=3.60,P<.05

(Continued) J

457

r

Table A 1 . ~ (continued).

Inland stands

Coastal stands

Comparisons between coastal and Inland stands

Stands connbined

Moisture Age

Moisture Age

All
stands

All Old Old

old wet mesic Mature Young

moisiure Age

Plethodon elongatus

ICS and PF

4 4

■ .4 ■

4 4 4

4 4

4 4

multiple

comparisons:

• ♦'

Taricha granulosa

ICS data

• . ■.

+ +

multiple

comparisons:

■ * ■

» ■ ♦ .

PFdata

*

+ + +

multiple

comparisons:

*

♦ ■ ■ ■

' * = not significont at P<.05.
'+ = not significant at P < .01.

^Capture frequency in designated category was too low for analysis.
^Species absent from inland or coastal area.
^Absent from young stands.

<rU.S. GOVERNMENT PRINTING OFFICE: 1989-674-391/5024

458

1022423217