BLM LIBRARY
88055769
2004 Annual Report
SUBLETTE MULE DEER STUDY (PHASE II):
Long-term monitoring plan to assess potential
impacts of energy development on mule deer
in the Pinedale Anticline Project Area.
Prepared for:
Questar Exploration and Production Company
Independence Plaza
1050 17'" St., Suite 500
Denver, CO 80265
TRC IVIariah Associates Inc.
605 Skyline Road
Laramie, WY 82070
Bureau of Land IVIanagement
Pinedale Field Office
432 E. Mill Street
PO Box 768
Pinedale, WY 82941
Wyoming Game and Fish Department
Pinedale Regional Office
PO Box 850
Pinedale, WY 82941
•
Prepared by:
Hall Sawyer
Ryan Nielson
Lyman McDonald
Dale Strickland
Western EcoSystems Technology, Inc.
2003 Central Avenue
Ctieyenne, WY 82001
QL November 2004
737
.U55
S399
2004
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
ACKNOWLEDGEMENTS
Phase II of the Sublette Mule Deer Study has been a cooperative effort among agencies,
industry, and private organizations. This project was largely funded by Questar Exploration and
Production Company (QEP) and subcontracted through TRC Mariah Associates Inc. (TRC).
The Pinedale Field Office of the Bureau of Land Management (BLM) provided additional
funding. The Jackson/Pinedale Region of the Wyoming Game and Fish Department (WGFD)
provided logistical support, project assistance, and was responsible for portions of the data
collection. Many thanks to Ron Hogan (QEP), Jane Seller (QEP), Steve Belinda (BLM), Keith
Andrews (BLM), Karen Rogers (BLM), Pete Guernsey (TRC), Scott Smith (WGFD), Herb Haley
(WGFD), Bernie Hoiz (WGFD), Scott Edberg (WGFD), Scott Werbelow (WGFD), Dan Stroud
(WGFD), Dean Clause (WGFD), Dennis Almquist (WGFD), Brad Hovinga (WGFD), Fred
Lindzey (University of Wyoming), Hawkins & Powers Aviation, and Gary Lust (Mountain Air).
Thanks to John Amos (SkyTruth) for image processing.
LIST OF ACRONYMS
BACI Before-After Control-Impact
BLM Bureau of Land Management
CR County Road
DAU Data Analysis Unit
EIS Environmental Impact Statement
GIS Geographic Information System
GPS Global Positioning System
HA Hunt Area
JCR Job Completion Report
MWRC Mesa Winter Range Complex
NEPA National Environmental Policy Act
NGO Non-Government Organization
PAPA Pinedale Anticline Project Area
PFWRC Pinedale Front Winter Range Complex
QEP Questar Exploration and Production Company
ROD Record of Decision
RSPF Resource Selection Probability Function
TPB Trapper's Point Bottleneck
TRC TRC Mariah Associates, Inc.
UD Utilization Distribution
USGS United States Geological Survey
UW University of Wyoming
VHF Very High Frequency
WEST Western EcoSystem Technology, Inc.
WGFD Wyoming Game and Fish Department
-fA
Sublette Mule Deer study: 2004 Annual Report WEST, Inc.
TABLE OF CONTENTS
Page S'-^^l'l
1.0 OVERVIEW 1
2.0 SUBLETTE MULE DEER STUDY (PHASE II).. 3
2.1 INTRODUCTION.......... 3
2.2 STUDY AREA 6
2.3 METHODS.. 6
2.3.1 Deer Capture 6
2.3.2 Winter Movement and Distribution Patterns 7
2.3.3 Population Characteristics 7
2.3.3.1 Abundance and Density Estimates 7
2.3.3.2 Reproduction 9
2.3.3.3 Adultfemale winter survival 9
2.3.3.4 Over-winter fawn survival 9
2.3.4 Direct Habitat Loss 9
2.3.5 Resource Selection 10
2.3.5.1 Study Area Delineation 10
2.3.5.2 Predictor Variables 10
2.3.5.3 Modeling Procedures 10
2.4 RESULTS 11
2.4.1 Deer Capture 11
2.4.2 GPS Data Collection 13
2.4.3 Winter Movement and Distribution Patterns 13
2.4.3.1 Treatment Area (PAPA) 13
2.4.3.2 Control Area (PFWRC) 14
2.4.4 Population Characteristics 28
2.4.4.1 Abundance and Density Estimates 28
2.4.4.2 Reproduction 30
2.4.4.3 Over-winter fawn survival..... 31
2.4.4.4 Adult winter survival 32
2.4.5 Direct Habitat Loss 36
2.4.51 Pre-Development 36
2.4.5.2 Yearl of Development 36
2.4.5.3 Year 2 of Development 37
2.4.5.4 Year3 of Development....... 38
2.4.5.5 Year 4 of Development 39
2.4.6 Resource Selection 40
2.4.6.1 Pre-Development: Winters 1998-99 and 1999-00 42
2.4.6.2 Yearl of Development: Winter 2000-01 43
2.4.6.3 Year2of Development: Winter 2001 -02 44
2.4.6.4 Year 3 of Development: Winter 2002-03 45
2.5 DISCUSSION AND FUTURE DIRECTION.... 46
2.6 PROJECT TIMELINE 50
3.0 LITERATURE CITED 51
APPENDIX A: Equations Used to Calculate Abundance and Density
Estimates 55
APPENDIX B: Resource Selection Modeling Procedures...... 56
APPENDIX C: GPS locations and 60-mile seasonal migration of deer #863 59
iii pO.^''
OenvG'
:^m^
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
LIST OF TABLES
Page
Table 2.1 Number and type of radio-collars functioning in treatment and control
areas during the 2003-04 winter. 12
Table 2.2 Summary statistics for abundance and density estimates in tine
treatment area during February 2002, 2003, and 2004 28
Table 2.3 Summary statistics for abundance and density estimates in the control
area during February 2002, 2003, and 2004 29
Table 2.4 Mule deer fawn:doe ratios measured for treatment (Mesa) and control
(Pinedale Front) areas by Wyoming Game and Fish Department,
1992-2004 30
Table 2.5 Mule deer count data and calculations for over-winter fawn survival in
the control area (Rnedale Front), 1999-2004 31
Table 2.6 Mule deer count data and calculations for over-winter fawn survival in
the treatment area (Mesa), 1999-2004 31
Table 2.7 Winter (2003-04) survival rates and summary statistics for adult female
deer in treatment and control areas 32
Table 2.8 Summary of annual and cumulative direct habitat loss (i.e., surface
disturbance) associated with road networks and well pads on the Mesa,
2000-2003 38
Table 2.9 Habitat variables and estimated coefficients for winter mule deer
resource selection probability functions (RSPF), 1998-2003 40
Table 2.10 Optimal values for predictor variables with quadratric terms 41
LIST OF FIGURES
Figure 1 .1 Location of Pinedale Anticline Project Area in western Wyoming
(Fig.1 from BLM 2000)..... 2
Figure 2.1 Location of the Sublette Mule Deer Data Analysis Unit, Wyoming Game
and Fish Department Hunt Areas, and the Mesa and Pinedale Front
Winter Range Complexes 4
Figure 2.2 Location of 38 quadrats used in control area during 2003 helicopter
surveys 8
Figure 2.3 Location of 70 quadrats used in control area during 2004 helicopter
surveys 8
Figure 2.4 GPS locations (r = 923) of deer #860 in the Pinedale Anticline Project
Area(PAPA), December 22, 2003 -April 15, 2004 15
IV
Sublette Mule Deer Study: 2004 Annual Report WES T, Inc.
LIST OF FIGURES (continued)
Page
Figure 2.5 GPS locations (r = 1 ,367) of deer #872 in tine Pinedale Anticline Project
Area (PAPA), December 22, 2003- April 15, 2004 16
Figure 2.6 GPS locations (r = 1 ,367) of deer #878 in the Pinedale Anticline Project
Area (PAPA), December 22, 2003- April 15,2004 17
Figure 2.7 GPS locations of all deer (n = 3) and land ownership of the Pinedale
Anticline Project Area (PAPA), December 22, 2003 -April 15, 2004.. 18
Figure 2.8 GPS locations (r = 924) of deer #861 in the Pinedale Front Winter Range
Complex, December 22, 2003-April 15, 2004.. 19
Figure 2.9 GPS locations (r = 91 9) of deer #869 in the Pinedale Front Winter
RangeComplex, December 22, 2003-April 15,2004 20
Figure 2.1 0 GPS locations (r = 1 ,978) of deer #871 in the Pinedale Front Winter
RangeComplex, November 1, 2003- April 15, 2004 21
Figure 2.1 1 GPS locations {r = 1 ,969) of deer #873 in the Pinedale Front Winter
RangeComplex, November 1, 2003- April 15,2004 22
Figure 2.1 2 GPS locations (r = 1 ,942) of deer #874 in the Pinedale Front Winter
RangeComplex, November 1,2003 -ApriM 5, 2004...... 23
Figure 2.1 3 GPS locations (r = 1 ,378) of deer #876 in the Pinedale Front Winter
RangeComplex, December 22, 2003 -April 15, 2004 24
Figure 2.1 4 GPS locations (r = 3,354) of deer #871 in the Pinedale Front Winter
Range Complex, December 20, 2002 - April 15, 2003 (green) and
November 1, 2003-April 15, 2004 (blue) 25
Figure 2.1 5 GPS locations (r = 3,344) of deer #873 in the Pinedale Front Winter
Range Complex, December 20, 2002 - April 1 5, 2003 (green) and
November 1,2003- April 15, 2004 (blue) 26
Figure 2.1 6 GPS locations (r = 3,328) of deer #874 in the Pinedale Front Winter
Range Complex, December 20, 2002- April 15, 2003 (green) and
November 1, 2003-April 15, 2004 (blue) 27
Figure 2.17 December fawnidoe ratios in treatment and control areas, 1999-2004 30
Figure 2.1 8 Estimated over-winter fawn survival in treatment and control areas,
1999-2004 31
Figure 2.1 9 Winter survival rates of adult female radio-collared deer in treatment and
control areas, 1998-2004 32
Figure 2.20 Satellite image of the Mesa on October 1 999, prior to development of the
Pinedale Anticline Project Area (PAPA) 36
Sublette Mule Deer Study: 2004 Annual Report WES T, Inc.
LIST OF FIGURES (continued)
Page
Figure 2.21 Satellite image of the Mesa taken in August 2001 , following 1 full year
of gas development in the Pinedale Anticline Project Area (PAPA) 37
Figure 2.22 Satellite image of the Mesa taken in October 2002, following 2.3 years
of gas development in the Pinedale Anticline Project Area (PAPA) 38
Figure 2.23 Satellite image of the Mesa taken in September 2003, following 3.25
years of gas development in the Pinedale Anticline Project Area (PAPA).. 39
Figure 2.24 Relationship between probability of habitat use and distance to well pad,
for development years 1-3 • 41
Figure 2.25 Surface map depicting probability of habitat use for mule deer prior to
development (1998-99 winter). Color-coded based on percentile of
predictions (i.e., 0-25%, 25-50%, 50-75%, and 75-100%) 42
Figure 2.26 Surface map depicting probability of mule deer habitat use during Year 1
of development (2000-01 winter). Color-coded based on percentile of
predictions (i.e., 0-25%, 25-50%, 50-75%, and 75-100%) 43
Figure 2.27 Surface map depicting probability of mule deer habitat use during Year
2 of development (2001-02 winter). Color-coded based on percentile of
predictions (i.e., 0-25%, 25-50%, 50-75%, and 75-100%) 44
Figure 2.28 Surface map depicting probability of mule deer habitat use during Year
3 of development (2002-03 winter). Color-coded based on percentile of
predictions (i.e., 0-25%, 25-50%, 50-75%, and 75-100%) 45
LIST OF PHOTOS
Photo 2.1 Winter deer mortality on south end of Mesa 33
Photo 2.2 Cross-section of mule deer femur. Bone marrow color (reddish) and
consistency (gelatinous) suggests cause of death was chronic starvation. . . 33
Photo 2.3 Snow conditions in PFWRC along the Big Sandy River (view north),
February 2004 ■■■■■■■ 34
Photo 2.4 Snow conditions in PFWRC along the Big Sandy River (view south),
February 2004 ■ • 34
Photo 2.5 Snow conditions in MWRC near Mount Airy (view north), February 2004.. 35
Photo 2.6 Snow conditions in MWRC near Two Buttes (view south), February 2004.. . . 35
VI
Sublette Mule Deer Study: 2004 Annual Report WEST, Inc.
1.0 OVERVIEW
In 1998 the Wyoming Cooperative Fish and Wildlife Research Unit began the Sublette Mule Deer Study,
a collaborative effort with industry, agencies, and private organizations intended to examine movement
patterns and population characteristics of the Sublette mule deer {Odocoileus hemionus) herd in western
Wyoming. Although a variety of agencies and non-government organizations (NGOs) contributed to the
study, it was funded largely by industry (Ultra Petroleum). Concurrently, the Bureau of Land
Management (BLM), in compliance with the National Environmental Policy Act (NEPA), initiated an
Environmental Impact Statement (EIS) to assess natural gas development in the SOQ-mi^ Pinedale
Anticline Project Area (PAPA) (BLM 2000) (Figure 1.1). Because the PAPA provides important winter
range to a large segment of the Sublette mule deer herd, there were concerns about the potential effects
gas field development may have on the deer population.
The Sublette Mule Deer Study was originally designed to have two phases. The first phase of the study
was intended to gather information needed by agencies to improve management of the Sublette deer
herd, including the identification of seasonal ranges, determination of migration routes, and estimation of
survival rates (Sawyer and Lindzey 2001). Additionally, these data were collected so that pre-
development information on the mule deer population would be available if Phase II of the study
materialized. Phase II was envisioned as a long-term study that would examine the potential impacts of
energy development on mule deer, using treatment and control areas, with energy development as the
treatment. The BLM completed the PAPA EIS and released their record of decision (ROD) in July of
2000 (BLM 2000). Phase I of the Sublette Mule Deer Study was completed in March of 2001 (Sawyer
and Lindzey 2001). Following a 1-year pilot study (Sawyer et al. 2002) funded by QEP, Phase II was
initiated in December of 2002, as a Before-After/Control-Impact (BACI) study design (Green 1979,
Morrison et al. 2001) that uses the PAPA as a treatment area and a portion of the Pinedale Front as the
control area. Mule deer population characteristics (survival, reproduction, density) and habitat use in
relation to development features will be measured in both areas, and over time, performance of mule
deer in the PAPA will be compared to those in the control area, both before and after the treatment. This
report summarizes the results from the 2004 study period.
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
Rgure 1
General Location
of the
Pinedole Antlcfc© Froject Area
rrc.inir; Anltc'r^ iif^^*lrc«rTtentaJ JmpocJ Staie^i?".**
IS JO' ^hfS
Figure 1.1 Location of Pinedale Anticline Project Area in western Wyoming (from BLIVl 2000).
Sublette Mule Deer Study: 2004 Annual Report WEST, Inc.
2.0 SUBLETTE MULE DEER STUDY
2.1 INTRODUCTION
Western Wyoming is home to the largest, most diverse ungulate populations in the Rocky Mountain
region. Maintenance of these populations and protection of their habitats are primary concerns among
the public and state and federal agencies. Because of their large numbers and economic importance,
mule deer continue to be a top priority for the Wyoming Game and Fish Department (WGFD). The
Sublette mule deer herd unit includes 15 hunt areas (HA) (130, 138-142, 146, 150-156, and 162) (Figure
2 1) and has a post-season population objective of 32,000 (WGFD 2002). Results from the Sublette
Mule Deer Study (Sav^/yer and Lindzey 2001) indicate that these mule deer seasonally migrate 60-100 mi
from winter range near Pinedale, Wyoming to summer in portions of the Salt River Range, Wyoming
Range, Wind River Range, Gros Ventre Range, and Snake River Range. During the lengthy spring and
fall migrations, mule deer spend a substantial amount of time, often 4-5 months out of the year, on mid-
elevation transition ranges that connect summer and wintering areas. By late-fall, most mule deer
annually converge in the Green River Basin to winter in one of two major complexes; the Mesa Winter
Range Complex (the Mesa) and the Pinedale Front Winter Range Complex (the Pinedale Front) (Figure
2.1). Generally, the Mesa includes the PAPA and those wintering areas west of US 191, while the
Pinedale Front includes those areas east of US 191 to the base of the Wind River Mountains.
Population parameters measured during the 3-year (1998-2000) Phase 1 study (WGFD 2002, Sawyer
and Lindzey 2001) suggested the Sublette deer herd was a healthy and productive population prior to
development of energy resources on the PAPA. Annual survival rates of radio-collared adult females
(n=149) averaged 85% and were consistent with populations studied in other western states (Unsworth
et al. 1999). Fawn:doe ratios, an indicator of reproductive success, were among the highest in the state,
averaging >75 fawns per 100 does for the study period and approximately 70 fawns per 100 does over
the last decade (WGFD 2002). Although the Sublette deer herd has been very productive in the past and
recent studies have improved management, this deer herd is similar to others in the region in that habitat
loss due to urban expansion and energy development continue to create major management concerns.
Natural gas production in Wyoming has steadily increased since the mid-1980s, particularly in the five
counties that form the southwest quarter of the state: Sublette, Fremont, Lincoln, Uinta, and
Sweetwater. This area of the state contains some of the largest and most productive gas fields in the
nation, including the Jonah, Continental Divide/Wamsutter, Fontenelle, Big Piney-LaBarge, Moxa
Arch, Riley Ridge, Desolation Flats, and the Pinedale Anticline. While coalbed methane development
in the Powder River Basin has received much attention as of late, Sublette County continues to
produce approximately 2.5 times more gas than Campbell County. Sublette County is the state's
largest producer and accounts for >30% of all Wyoming gas production (BLM 2002). Because of
renewed political and economic support for developing domestic energy reserves, natural gas
exploration, development, and production are at an all time high in Wyoming and expected to
increase.
Because the PAPA encompasses the Mesa, which is used by thousands of mule deer, pronghorn
{Antilocapra americana), and sage grouse {Centrocercus urophasianus), development of this area may
have adverse impacts on wildlife. Impacts to wildlife may be direct or indirect. Direct impacts include the
loss of habitat to well pads, access roads, and pipelines. Indirect impacts may include changes in
distribution, stress, or activity caused by increased human activity associated with energy development
(i.e., traffic, noise, human presence). The best way to evaluate the impact(s) of energy development on
wildlife populations is through long-term studies where pre-development data, such as, estimates
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
Location of Sublette Mule Deer Data Analysis
Unit(DAU) in Wyoming
Hunt Areas in Sublette DAU
j^v^r^l]
Beaver
Ridge
RyegrasX O^ \ ^1
Cottonwood cre^H % 4 / The Mesa
\ *K\ \ / springs
I Half Moon ;■ ' 'f ^N V. X ^ i* »f.
^inedale --^ ' — * • ' '
,_,_ J' 4 . ' ' :\^0C \ '•- ".^''
139 \%\ , "
143
HWY
(189)
138 \%\
^ S Ross Ridge
■-^^ \
Reardon\ \ /^ \
Draw n \ Vyi \
Sandy\0\ ,
Elk \ O. \ -^
Mountain X'Q^ '••■ "m
130 WV
Figure 2.1 Location of the Sublette Mule Deer Data Analysis Unit, Wyoming Game and Fish
Department Hunt Areas, and the Mesa and Pinedale Front Winter Range Complexes.
of survival and reproduction are available. Because these studies are by necessity observational,
determining cause and effect relationships is very difficult. Simply documenting a behavioral response
(e.g., avoidance, acclimation, dispersal) to a disturbance adds very little to our knowledge of the impact, if
it cannot be linked to the survival or reproductive success of the species involved. And conversely,
documenting a change in reproduction or survival does not add significantly to our understanding of the
impact if the cause (e.g., weather, habitat loss, disease) of the change cannot be determined. Because
•
Sublette Mule Deer Study: 2004 Annual Report WES T, Inc.
of the difficulty with designing and funding long-term studies, impacts of energy development on free-
ranging ungulate populations are poorly understood and often debated. However, both direct and
indirect impacts associated with energy development have the potential to affect ungulate population
dynamics, particularly when disturbances are concentrated on winter ranges, where energetic costs are
great and animals occur at high densities.
The major shortcoming of efforts to evaluate the impact(s) of disturbances on wildlife populations is that
they seldom are addressed in an experimental framework, but rather tend to be short-term and are
almost always observational. Brief, post-development monitoring plans associated with regulatory work
generally result in little or no information that allow agencies and industry to assess impacts on wildlife or
identify new, and potentially more effective, mitigation measures. On the other hand, long-term studies
are difficult to implement because they are expensive and require interagency and industry cooperation
and commitment. Additionally, the acquisition of pre-development data on movement patterns and
population characteristics, and identification of suitable control and treatment areas is extremely
uncommon. The situation in the PAPA and upper Green River Basin is unique because all the
necessary information is available to conduct a BACl study to suggest if, and if so, how natural gas
development affects the PAPA mule deer population.
The basic idea with a BACI study design is that the potentially impacted (treatment) site is sampled both
before and after the time of the disturbance (e.g., energy development), and one or more control sites
that do not receive any disturbance are sampled at the same time (Manly 2001). The assumption is that
any naturally occurring changes will be similar at the control and treatment sites, and in the absence of
the treatment the parameters of interest will be similar for both areas, or at least the magnitude of the
differences will be relatively constant from year to year. Thus, potential changes at the treatment site
may be attributed to the disturbance. It is not critical that the control and treatment populations be
identical, only that the subpopulations are independent and that both respond to the same environmental
factors.
For this study, energy development on the Mesa is considered the treatment and a portion of the
Pinedale Front serves as the control area. The Rnedale Front consists mostly of federal lands located
along the southwest portion of the Wind River Range, where no energy development is anticipated. The
Pinedale Front is a suitable control site because: 1 ) there is little or no exchange of deer between the
Mesa and Pinedale Front, 2) the two deer subpopulations use separate winter ranges, but share
common transition and summer ranges, so they have comparable foods available during parturition and
arrive on winter ranges in similar condition, 3) although the two deer subpopulations occupy distinct
winter ranges, they are in close proximity to one another (15-30 mi), so both are exposed to similar
weather patterns and environmental conditions, 4) habitat characteristics on both winter ranges are
similar and dominated by sagebrush communities, and 5) population characteristics of the two
subpopulations have consistently tracked one another prior to development of the PAPA.
We believe four population parameters should be monitored to detect the potential impacts of energy
development on mule deer, including: 1) adult doe survival, 2) over-winter fawn survival, 3) reproduction,
and 4) density. As these parameters are measured in treatment and control areas, comparisons can be
made, and over time, the potential impacts of energy development on mule deer may be assessed. If
mule deer in the PAPA continue to function as well as before development and as well as those in the
control area it vyould suggest that energy development has no adverse impacts on mule deer in the
region. If however, mule deer survival or reproduction in the PAPA decreases, while the same
parameters in the control area remain unchanged or increase, then energy development may be the
cause of those declines. Again, this does not demonstrate a cause-effect relationship; rather it is simply
one piece in a weight of evidence approach, where the BACI study design examines several direct (e.g,
survival, reproduction) and indirect (e.g., habitat use, displacement) parameters that are statistically
analyzed and carefully interpreted.
Sublette Mule Deer Study: 2004 Annual Report WES T, Inc.
Results from Phase I identified seasonal migration routes and distribution of deer in the Mesa and
Pinedale Front (Sawyer and Lindzey 2001). Although mule deer migrations of >60 mi have been
reported in parts of Idaho (Thomas and Irby 1990) and Montana (Mackie et al. 1998), mule deer on and
adjacent to the PAPA are likely the most migratory deer in the western states, annually migrating 60-100
mi between winter and summer ranges. Because these deer are highly mobile and demonstrate strong
fidelity to seasonal ranges, the potential for energy development, or other human disturbances, to disrupt
migratory routes and/or winter distribution patterns exists. While changes in distribution or migratory
patterns may not necessarily result in decreased deer survival or reproduction, it is useful to include
within the monitoring plan to: 1) document if migration routes remain intact, 2) document if deer continue
using pre-development winter ranges, 3) provide industry and agencies with accurate, precise movement
data for site-specific analyses (e.g., seasonal range designation or comparison of effects of multiple well
pads versus single well pad), 4) identify mitigation opportunities on and off-site treatment and control
areas (e.g., migration corridors, habitat improvements), and 5) allow for analyses that estimate and
describe indirect habitat loss (e.g., avoidance of roads or well pads) or changes in habitat use.
Properly designed long-term monitoring and examination of adult survival, over-winter fawn survival,
reproduction, density, and seasonal distribution/movement patterns will allow for population-level
inferences concerning the potential impacts of energy development on mule deer.
2.2 STUDY AREA
The PAPA is located in west-central Wyoming in Sublette County, near the town of Pinedale (Figure 1.1).
The PAPA is characterized by sagebrush communities and riparian habitats associated with the Green
and New Fork Rivers. Elevations range from 6,800 to 7,800 ft. The PAPA consists primarily of federal
lands (80%) and minerals (83%) administered by the BLM. The state of Wyoming owns 5% (15.2 mi^) of
the surface and another 15% (46.7 mi^) is private. Aside from the abundant energy resources, the PAPA
is an important area for agriculture and provides winter range for 4,000-6,000 mule deer, 2,000-3,000
pronghorn, and 3,000-4,000 sage grouse. While the project area is fairly large, most deer occur in the
northern portion of the PAPA, an area locally known as "The Mesa", which includes approximately 100-
mi^. In July of 2000, the BLM approved the development of 700 producing well pads in the PAPA and
recognized that this may require as many as 900 well pads to be constructed and drilled (BLM 2000).
Additionally, 401 mi of pipeline and 276 mi of access roads were approved for development of energy
resources on the PAPA.
2.3 METHODS
2.3.1 Deer Capture
Helicopter net-gunning was used to capture deer across winter ranges in treatment (Mesa) and
control (Pinedale Front) areas. Captured deer were fitted with collars supporting either a GPS or VHF
radio transmitter. Both types of collars were equipped with mortality sensors that change pulse rate if
the collar remains stationary for more than 8 hours. The VHF collars (Advanced Telemetry Systems,
Isanti, MN) were duty-cycled to transmit signals October 1 through May 31. The GPS collars
(Telonics, Mesa, AZ) were store-on-board units capable of storing approximately 3,000 locations and
programmed to obtain fixes every 2 hours during winter months (November-April). Additionally, each
GPS collar was equipped with a remote release mechanism programmed to activate at a specified
time, so that collars could be retrieved and data downloaded.
•
Sublette Mule Deer Study: 2004 Annual Report WEST, Inc.
2.3.1 Winter Movement and Distribution Patterns
Data collected from GPS-collared deer accurately identified winter distribution, movement patterns
and migration routes of the marked deer on and adjacent to winter ranges. Over time, these data will
be used in conjunction with pre-development data to identify changes in mule deer movement
patterns habitat use and distribution relative to energy development activities. Because a portion
(n=11) of GPS collars are to remain on the same deer for consecutive winters (2003-04 and 2004-05),
some data for the 2003-04 winter will not be available until 2005.
2.3.3 Population Characteristics
2.3.3.1 Abundance and Density Estimates
Deer abundance and density were estimated in treatment (Mesa) and control (Pinedale Front) areas
using aerial counts of deer in systematically sampled l-mi^ quadrat units. Winter distribution data
collected from radic^collared deer in the study area between 1998 and 2003 was used to delineate 68-
mi^ and /O-mi^ sampling frames for the treatment and control areas, respectively. Sampling frames were
expected to contain high-densities of deer so stratification was unnecessary. We sampled 34 quadrats
from each sampling frame, covering approximately 50% of the geographic area. Equations used to
calculate abundance and density estimates were taken from Thompson et al. (1998) and are reproduced
in Appendix A. Standard 90% confidence intervals were calculated using a Z statistic.
The size of the sampling frame in the control area has changed over the course of the study (See
Section 2 4 41) During the first year of surveys (2002) v\e identified a 35-mi sampling frame that we
believed represented the core winter range in the Pinedale Front. During 2003 we made some slight
modifications to improve our sampling and used a similar SS-mi^ sampling frame (Figure 2.2). However,
during the 2003 surveys many of our marked deer moved out of the sampled area. At this time it
became apparent that these deer utilize a much larger area than we originally thought. To accurately
adjust the size and extent of our sampling frame we conducted a telemetr/ flight pnor to the 2004 survey
to adjust the size of our sampling frame based on locations of marked deer. The new sampling frame
was 70-mi2 (Figure 2.3), nearly double the size of the 2002 and 2003 frames and approximately the same
size as the sampling frame for the treatment area.
Group size and vegetative cover may significantly influence visibility bias in ungulate helicopter surveys
(Samuel et al 1987) However, the treatment and control areas for this study consist of homogenous
sagebrush stands with no tree cover. Additionally, telemetry data from Phase 1 indicated male and
female deer did not winter in areas with different habitat characteristics, so potential group size vanation
resulting from sexual segregation should not influence counts. Further, when survey areas contain large
concentrations of animals that are widely distributed, recognition of individual groups may be near
impossible Attempting to determine visibility correction factors for groups is likely not feasible in these
situations (Samuel et al. 1987). Counts of animals within the sampled quadrats are assumed to provide
valid indices on density and abundance. That is, if not all animals present were detected, we assume the
same visibility bias in both treatment and control areas overtime.
Counts were conducted from a piston-powered Bell helicopter flown approximately 100-150 ft above
around and at speeds of 20-40 knots. The northeast UTM coordinates for each quadrat were
programmed into a GPS unit on the helicopter. Quadrat perimeters were then flown clockwise, such that
the observer was positioned on the inside, while the pilot navigated. A real-time flight path was traced
into the on-board GPS and once the perimeter was established the quadrat interiors were systematically
searched Observer and navigator collectively detected deer groups and determined whether groups
were inside or outside quadrat boundaries. Deer detected inside and moving out were considered in the
quadrat while deer detected outside and moving in the quadrat were considered out. Half of the deer
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
detected on perimeter boundaries were considered in tine quadrat. For each quadrat, tine observer
recorded total number of deer, number of deer groups, and total search time.
,'yT-^'^~ \ ,y'
W^^9%
igure 2.2 Location of 38 quadrats used in control area dunng 2003 helicopter surveys.
:X
■1- x^'^-y^
^ ^ 4^ — v^
l-c:^-/"-i-k£>w--i J- r >vij^iv44wv.--rt
tiic4-fLv /rt-*' "
■igure 2.3 Location of 70 quadrats used in control area during 2004 helicopter surveys.
Sublette Mule Deer Study: 2004 Annual Report WES T, Inc.
2.3.3.2 Reproduction
Doe:fawn ratios are commonly used as an index to herd productivity or reproduction. Doe:fawn ratios
were calculated from composition data collected during the WGFD's annual helicopter surveys in
December, consistent with the previous 10 years of WGFD data collection (WGFD 2002). Sample
sizes were adequate to obtain desired levels of precision in ratio estimates (Czaplewski et al. 1983).
2.3.3.3 Adult Female Winter Survival
Adult doe survival was estimated from telemetry records using the Kaplan-Meier procedure (Kaplan
and Meier 1985). We attempted to maintain a sample of 30 marked deer in both control and
treatment areas. Marked deer were located at least once per month, December through May.
2.3.3.4 Over-winter Fawn Survival
Deer from both the Mesa and Pinedale Front congregate on the northern ends of their respective
winter ranges every spring which allows large numbers (>1,000) of animals to be counted and
classified. Ground-based composition surveys conducted in April were used to calculate post-winter
adult:fawn ratios. These data were used in conjunction with adult survival rates and December
adult:fawn ratios to estimate over-winter fawn survival, using the change-in-ratio estimator from White
etal. (1996):
. B
S r =S„x—, where A - count of December fawns/count of December adults
^ A
B = count of April fawns/count of April adults
S = estimate of adult survival
a
Adult survival rates were estimated from telemetry records, rather than carcass counts. The delta
method (Sober 1982) was used to estimate variance.
2.3.4 Direct Habitat Loss
Satellite imagery and geographic information system (GIS) software were used to digitize road
networks and well pads associated with natural gas development in the northern portion of the PAPA
(i.e.. The Mesa), from 2000 through 2003. Areas within the PAPA, but outside the Mesa were not
considered. Landsat images were purchased from the USGS and processed by SkyTruth
(Sheperdstown, WV). Images were generally obtained in early fall (i.e., September-October), after
most annual construction activities (e.g., well pad and road building) were complete, but prior to snow
accumulation. Pipelines and seismic tracks were not included in this analysis. Roads and well pads
were digitized in ArcViev\P(ESRI, Redlands, California, USA). Length of road segments and size of
well pads were calculated in ArcViev\P(ESRI, Redlands, California, USA). Acreage estimates
associated with road networks were based on an average road width of 30 ft. We recognize there is
some error associated with the digitizing process, however it is expected to be minimal and the
resulting digital GIS coverages are considered the best available data. During the digitizing process
we assumed full reclamation of well pads had not occurred, since the gas field is only 3 years old and
successful reclamation (i.e., re-establishment of native plant species) of native shrub communities in
arid environments is extremely difficult and unlikely to occur during a short time period.
Sublette Mule Deer Study: 2004 Annual Report WES T, Inc.
2.3.5 Resource Selection
2.3.5.1 Study Area Delineation
The treatment study area was defined by mapping 30,626 GPS locations from 34 mule deer over a
five-year period (1998 to 2002) and creating a minimum convex polygon (MCP), consistent with
McClean et al. (1998)'s recommendation that study-area level of habitat availability should be based
on the distribution of radio-marked animals. Additionally, the MCP generated from GPS data was
consistent with winter distribution patterns documented for this deer population using >60 VHP radio-
collars, between 1 998 and 2000 (Sawyer and Lindzey 2001 ).
2.3.5.2 Predictor Variables
We identified five habitat variables as important landscape predictors of winter mule deer distribution:
elevation (Vi), slope (V2), aspect (V3), road density (V4), and distance to well pad (V5). We did not
include vegetation as a variable because the sagebrush-grassland vegetation is generally
homogeneous and difficult to divide into finer vegetation classes across the study area. We believed
differences in sagebrush characteristics could be largely explained by elevation, slope, and aspect. A
GIS was used to measure variables Vi to V5 for sampled habitat units. The Spatial Analyst Extension
for ArcView® (ESRI, Redlands, California, USA) vvas used to calculate slope and aspect from a 26 x
26 m digital elevation model (DEM) (USGS 1999). Grid cells with slopes >2 degrees were assigned
an aspect value (i.e., NE = 2-< aspect = 90', SE = 90' < aspect = 180', SW = 180-< aspect = 270', NW
= 2700-< aspect = 360^, while grid cells with slopes of = 2 degrees were considered flat and had no
aspect value. Road and well pads were annually digitized from Landsat satellite images acquired from
the U.S. Geological Survey and processed by SkyTruth (Sheperdstown, WV). Images were obtained
from late-summer or early fall, prior to snow accumulation, but after most annual natural gas
development activities were complete. Road density was calculated using three buffer sizes; 0.5 km
radius, 1.0 km radius, and 1.5 km radius, and the modeling process was used to determine which had
the most predictive power. A pairwise correlation analysis (PROC CORR; SAS Institute 2000) was
conducted on the linear forms of the variables to identify possible multicollinearity (Neter et al. 1996)
and determine if any variables should be dropped from the analysis prior to modeling.
No distinction was made between producing and developing well pads. The rationale for treating all
well pads equally was based on the assumption that there are two potential factors that may affect
deer habitat selection near well pads: 1) direct loss of habitat, usually in the form of a 3-4 acre surface
disturbance and 2) human activity (e.g., traffic, noise) associated with the well pad. Because the
PAPA occurs in a dry desert shrub community, re-establishment of native shrub communities in
disturbed areas is difficult and unlikely to occur in < 10 years, following current reclamation standards
and practices in the project area. Based on our experience in the PAPA and given development is in
its early stages (i.e., < 5 years), we felt it was appropriate to assume similar habitat loss among all
well pads. Additionally, there is no evidence from the PAPA that suggests the type of well pad (e.g.,
developing, producing) is an accurate indicator of the amount of human activity that occurs at the site.
Without an accurate measure of human activity (e.g., traffic), we believed it was inappropriate to
distinguish between producing and developing well pads in this analysis.
2.3.5.3 Modeling Procedures
Our modeling approach consisted of four basic steps: 1) estimate the utilization distribution's (UD) for
GPS-collared deer during each winter, 2) use the height of the UD as a response variable in a multiple
regression setting to model the probability of habitat selection for each deer, 3) develop a population-
level model from the individual deer models for each winter, and 4) map predictions of population-
10
Sublette Mule Deer Study: 2004 Annual Report WEST, Inc.
level models from each winter. Refer to Appendix B for modeling procedure details.
2.4 RESULTS
2.4.1 Deer Capture
We captured and radio-collared 21 adult female deer on December 15, 2003. Deer capture in the
PAPA was restricted to those areas where deer congregate across the northern end of the PAPA in
early winter, as they move onto the Mesa from Trapper's Point and/or the Ryegrass/Grindstone area.
We assumed this represented a random sample of deer in the subpopuiation because the deer were
congregated on the north end, before they moved south to their respective winter ranges. For the
same reason, deer capture in the Pinedale Front was restricted to the Big Sandy area; bounded to the
north and west by the Big Sandy River, east to the Prospects, and south to Elk Mountain. Of the 21
deer captured, 18 were equipped with GPS radio-collars and 3 equipped with traditional VHF radio-
collars. All GPS collars were store-on-board units equipped with VHF transmitters on 24-hour duty
cycles, 8-hour mortality sensors, and remote-release mechanisms programmed to drop collars at
0800 hours on April 15, 2004 or April 15, 2005. The programming schedule for collars was as follows:
Generation II GPS collars:
• obtain 1 location every 3 hours December 20, 2002-April 15, 2003
• obtain 1 location every 3 hours December 20, 2003-April 1 5, 2004
• for a possible ~1 ,800 locations
Generation III GPS collars:
• obtain 1 location every 2 hours December 20, 2002-April 15, 2003
• obtain 1 location every 2 hours November 01 , 2003-April 1 5, 2004
• for a possible -3,300 locations
Consistent with previous years, our goal was to maintain a sample size of 30 deer in each area, including
10 GPS and 20 VHF radio-collars (Table 2.1).
Additionally, we captured deer #863 and recovered the GPS collar that had failed to drop off in 2003.
This collar collected locations every 3 hours for 22 months (January 2002 - October 2003), including
the spring and summer of 2003. Appendix C includes a map that illustrates the 60-mi seasonal
migration of deer #863.
11
Sublette Mule Deer Study: 2004 Annual Report
Table 2.1 Number and type of radio-collars functioning in treatment a
2003-04 winter.
WEST, Inc.
nd control areas during the ^^
Treatme
(Thef
(^f\nfrf\l Aro9
l/lesa)
(Pineda
e Front)
Deer ID
Collar Type
Deer ID
Collar Type
810*
VHF
804
VHF
801
VHF
807
VHF
803
VHF
811
VHF
805*
VHF
812
VHF
806
VHF
814
VHF
809
VHF
816
VHF
813
VHF
820
VHF
814*
VHF
821
VHF
815
VHF
823
VHF
817
VHF
824
VHF
818
VHF
825
VHF
822
VHF
826
VHF
827
VHF
829
VHF
830
VHF
833
VHF
842
VHF
835*
VHF
844
GPS
836
VHF
845
VHF
848
VHF
849
VHF
850
VHF
852
VHF
851
VHF
853
VHF
861
GPS
854
VHF
864
GPS
855
GPS
867
GPS
8553
VHF
869
GPS
860
GPS
870
GPS
862
GPS
871
GPS
865
GPS
873
GPS
866
GPS
874
GPS
868
GPS
876
GPS
870*
VHF
958*
VHF
872
GPS
VHF = 20
GPS = 9
Total = 29
878
GPS
884
GPS
886
VHF
887
GPS
889
GPS
898*
VHF
905*
VHF
989
VHF
946*
VHF
976*
VHF
VHF = 28
GPS = 12
Total = 40
* Radio-collars
left over from P
lase 1 (Sawye
r and Lindzey 2
001).
12
•
Sublette Mule Deer Study: 2004 Annual Report WES T, Inc.
2.4.2 GPS Data Collection
We intended to collect data from 9 GPS collars following the 2003-04 winter. All nine release
mechanisms successfully dropped collars on April 15, 2004. However, because of several deaths
due to the severity of the winter, we recovered 1 1 GPS collars, including 5 from the Mesa and 6 from
the Pinedale Front. Four of the collars (#865, #871 , #873, and #874) contained data for consecutive
winters (2002-03 and 2003-04).
Of the 11 collars that were retrieved, six were Generation III and five were Generation II. All collars
functioned properly and collected the expected number of locations. Generation II and Generation 111
collars averaged 923 and 1,521 successful fix attempts, respectively. Consistent with GPS
performance in previous years (Sawyer et al. 2002), success rates for GPS fix attempts were very
high (99%) and locations precise (-93% 3-D locations for GEN II and 73% 3-D for GEN 111). A
minimum of four satellites are needed to generate 3-D locations, which typically have less than 20-
meter error (Di Orio et al. 2003).
2.4.3 Winter Movement and Distribution Patterns
2.4.3.1 Treatment Area (Mesa):
Individual maps were compiled for three GPS-collared deer (#860, #872, and #878) captured in the
treatment area (Figures 2.4-2.6). Data from the eight remaining collars were not mapped because
they will not be recovered until April 15, 2005. Data from deer #865 was not mapped because it died
early in the winter (-January 15).
Deer locations were color-coded by month so that timing of winter movements and distribution
patterns could be easily deciphered (Figures 2.4-2.6). Distribution and movement patterns were
variable among deer. Despite unusually harsh winter conditions deer shifted areas of use through the
winter and utilized a large portion of the Mesa. Locations from GPS-collared deer were similar with
observations and locations of VHF-collared deer.
Most deer did not begin their northerly migration off the Mesa until late-March and early-April,
presumably because of the above average snow conditions. Consistent with previous years, all deer
traveled to the Cora Butte area via the Trapper's Point Bottleneck (TPB) (Sawyer and Lindzey 2001,
Sawyer et al. 2003). With the exception of deer #872, rate of movement in and near the TPB
increased, as evidenced by distance between bcations (Figures 2.2-2.6). Deer #872 spent ten days
(April 5-14, 2004) in the TPB.
Figure 2.7 includes locations {r - 3,657) from all three deer and illustrates the importance of BLM
lands to this mule deer population. Boundaries between private and BLM lands generally correspond
with habitat type and topography; with private lands consisting of flat river bottoms and agricultural
areas, whereas BLM lands contain sagebrush hills in drier, more rugged terrain. Mule deer
demonstrated a strong affinity to the sagebrush-dominated BLM lands.
Figure 2.7 clearly defines the TPB, located 7 mi west of Pinedale, near the junction of US 1 91 , WYO 352,
and CR 110. Sawyer and Lindzey (2001) defined bottlenecks as "those areas along migration routes
wliere topography, vegetation, development and/or other landscape features restrict animal movements
to narrow or limited regions." Bottlenecks create management concerns because the potential to disrupt
or threaten established migratory routes are much greater in these areas. This naturally-occurring
bottleneck is approximately 1 mi in width and length, restricted to the southwest by the Green River
riparian complex and to the northeast by the New Fork/Duck Creek riparian complex. Sagebrush habitats
13
Sublette Mule Deer Study: 2004 Annual Report WES T, Inc.
north and south of US 191 are used extensively by mule deer during certain times of the year (Sawyer
and Lindzey 2001 ). IVlule deer use the narrow strip of sagebrush connecting the 2 areas to cross US
1 91 . Development of small, fenced house lots adjacent to BLM lands has narrowed the effective width of
the TPB to < 0.5 mi.
2.4.3.2 Control Area (Pinedale Front):
Individual maps were compiled for six GPS-collared deer (#861, #869, #871, #873, #874, and #876)
captured in the control area (Figures 2.8-2.13). Data from three collars (#867, #864, and #870) were not
mapped because they will not be recovered until April 15, 2005. Again, deer locations were color-coded
by month so that timing of winter movements and distribution patterns could be easily deciphered
(Figures 2.8 - 2.1 3). Consistent among deer was their mobility and tendency to shift areas of use through
the winter, utilizing areas that exceeded 100-mi^. Similar to last year (Sawyer et al. 2003), deer moved
outside the core winter range area around Buckskin Crossing to peripheral areas, such as Elk Mountain,
Squaw Teat, and a large section of Little Sandy Creek east of the Prospects. While distribution patterns
of deer were variable across the winter range, the migratory routes to and from the winter range were
nearly identical among deer. The strong affinity deer demonstrate to this migration route is clearly
illustrated by comparing locations from deer #871, #873, and #874. These three deer collected data for
two consecutive years (2002-03 and 2003-04) and used the same migration route both years, during
spring and fall movements (Figures 2.1 4-2.1 6).
Most deer began migrating north along the Pinedale Front in mid-March. Deer that winter along the
Pinedale Front were known to migrate northerly along the Wind River Range to the New Fork Lake
area before shifting their migration in a westerly route towards the Hoback Basin and adjacent
mountain ranges (Sawyer and Lindzey 2001). Details of this migration route, in terms of size, width,
specific location, and deer fidelity were unknown prior to GPS data collected over the last two years.
Consistent with last year, all six GPS-collared deer captured in the control area migrated along what
appeared to be a distinct movement corridor located at the base of the Wind River Range. While deer
sometimes remained in one area for a number of days, they appeared to follow a well-defined route
that rarely exceeded 1 -mile in width and covered a distance of 50 mi.
The migration route deer followed from the Buckskin Crossing area took them north across the Big
Sandy River, then northerly across the sagebrush flats below Sheep Creek and Muddy Creek. Deer
then moved into slightly rougher terrain among the boulders and sagebrush draws east of CR 353,
south of the East Fork, and west of Irish Canyon. Deer then moved northerly, crossing the East Fork
and Pocket Creek approximately 2-3 mi east of CR 353. Once across Pocket Creek, deer contoured
through the sagebrush slopes and aspen pockets, northerly through Cottonwood Creek and Silver
Creek. From Silver Creek, deer continued northwesterly across Lovett and Scab Creek. Deer
continued to contour across the sagebrush slopes below Soda Lake, towards the outlet of Boulder
Lake. Deer crossed Boulder Creek near the outlet of Boulder Lake, and then moved north to Fall
Creek, apparently to avoid an agncultural area between Fall Creek and Pole Creek. Deer crossed
Fall Creek just below the confluence of Meadow Creek, and then moved northwesterty toward the
outlet of Fremont Lake. Deer crossed Pine Creek at the Fremont Lake Bottleneck, as described by
Sawyer and Lindzey (2001). GPS collars released on April 15, 2004 and no movements were
recorded beyond the outlet of Fremont Lake. Last year (2003), however, spring conditions allowed
deer to continue movements through the Fremont Lake Bottleneck, and north along the Willow Creek
Road and Fremont Ridge before collars were released (Sawyer et al. 2003: Figure 2.11). Deer
moved within ;^mi either side of the Willow Lake Road from Soda Lake to the outlet of Willow Lake.
No movements were recorded beyond the outlet of Willow Lake for GPS collars released on April 15,
2003.
•
14
Sublette Mule Deer Study: 2004 Annual Report
GPS 2003-04 #B60
0 DecombG'r
• January
Q February
o M^rdi
® April
I I PAPA Boundary
4 luitles
Figure 2.4 GPS locations (r = 923) of deer #860 in the Pinedale Anticline Project Area (PAPA),
December 22, 2003 -April 15, 2Q04.
15
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
Figure 2.5 GPS locations (r = 1 ,367) of deer #872 in the Pinedale Anticline Project Area (PAPA),
December 22, 2003 - April 1 5, 2004.
16
Sublette Mule Deer Study: 2004 Annual Report
WEST. Inc.
Figure 2.6 GPS locations (r = 1 ,367) of deer #878 in the Pinedale Anticline Project Area (PAPA),
December 22, 2003 -April 15, 2004.
17
ty^ttaasj
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
"p^/ -"7 v--^^^^=r4-^^:;%f^^
■igure 2.7. GPS locations of all deer (n = 3) and land ownership of the Pinedaie Anticline Project Area
(PAPA), December 22, 2003- April 15, 2004.
18
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
•
,0 , - ;
^':^^^' Boutder Lakfepitr- -fiyi n
- l^-^-frg^g^^tt-^ Fremont Buttet ----%- -|—-^^^
^ ^d^^"
It^--
v.;
- — '^ i -^ r \
;Sandy|-^
^i
GfS 2003-04 #861
0 December
• January
c February
c March
• April
r~l PAPA Boundary
. V-
^ r-,
Buckskin n-
>«C!;^^L
Crossjtig I 1*^ - ^ ^.y'
:n^5?^-- ^ O PAPA Boundary ^^^^^'^^S^fc o^ ^ -'P
^irt-^r-^ ' ^^^^^ V^lF'- Elk Mountalni * \^*V''^'#'% TT!
Figure 2.8. GPS locations (r= 924) of deer #861 in the Pinedale Front Winter Range Complex,
19
•
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
f]iJ-J I "s
^yc '.^f^!?'"^!** Scab Creek
■;iv ^
\K^
Muddy
4
ip^-x-ki GPS 2003-04 #869
. „ ^_' o December
[
l>
~ — 'J/* - **< i?
e January
o February
Q March
• Aprii
piPAPABoundaty
5
MM
+ K-T
'^-•^ jBudcskin
•1
^t_j
~t — t-rf^
10 Miles
7V'
">.>;-*-»*-'— "1
■'-<¥^-
' Elk Mountatn
:^s^!]v^
N
Elkhom Jet
igure 2.9. GPS locations (r = 919) of deer #869 in the Pinedale Front Winter Range Complex,
December 22, 2003 -April 15, 2004.
20
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
t^'v ;!ElkMourHalnl
t.1- ' " -I... ......I. ■"iL, I . i
imm
Elkhom Jet
■igure 2.1 0 GPS locations (r = 1 ,978) of deer #871 in tiie Pinedale Front Winter Range Complex,
November 1, 2003- April 15, 2004.
21
•
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
/Sr'JScabCreek
.\ ^(Boulder Lake Is '^
i--\ .>riiLI''flJ'>| Ill' .'
X"='ir'-f tr'^-::rT t'\ Ffemonl Buitej— ^r—^^. -r^"— "^^^r' ^
*^ J, ./
l-V
_ — jji^.jf .,■>,: — ^
\.
-i
BigSandyn
6PS 2003-04 #873
a November
M| • December
• January
o Februaiy
0 March
• April
PPAPA Boundary
5
Buckskin I "ir-i;^*'— i^'^^Bjf-y*:<^'-— ^,
'v.
•s 1
-_ ._ , ;-4^ . 1': 'pTi
i BkhomJctJ
■igure 2.1 1 GPS locations (r = 1 ,969) of deer #873 in the Pinedale Front Winter Range Complex,
November 1, 2003- April 15, 2004.
22
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
^B'^Sa-^ Boulder Lake^^|gF '
.'--'1) Vi N. ' „ ->"? -'i 3.
';^4f 1 "'V,-' fc • ■viV'^ 1 't-^ ■
•
/ V VI 4'"' ■ '^M^^;^
GPS 2003-04 #874
• November
8 December
• January
o Febricary
b Mardi
• Apdl
n PAPA Boundary
i) 5 10 Miles
. ._.,^^:.,;:i-X-*~T*'>:
_JBuckskIn
i- ] Crossing
I „iii— ii.,.i.. |i
iBkhom Jet
Figure 2.1 2 GPS locations (r = 1 ,942) of deer #874 in the Pinedale Front Winter Range Complex,
November 1, 2003- April 15, 2004.
23
•
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
-:^ i J
IhL ■':/>• ■-'T<J^ I I I P^A Boundary
10 Miles
Bk Mountain
;TA^^ -=-i5r /.^-
ti-4Vfc
' 'i- «ii JElktiomJct
i^.l
■igure 2.1 3 GPS locations (r= 1,378) of deer #876 in the Pinedale Front Winter Range Complex,
December 22, 2003 - April 1 5, 2004.
24
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
v^-m-F^—
-— ^
. tfc\?
-/-
iS!<:-
VW ■'
T"^
ite
N
• GPS 2003-041^71 A.
« GPS 2002-03 #871 ft qg^' e
"1 PAPA Boundary |
{:2^ "^jBodffiWn
^•- s Crossing
.♦ V
~,„..^.^.-^> , 10 Miles i^ 'n '•"■■AT^£llkrf^}*-''-^
Fiaure 2.1 4 GPS lonations ^r = 3 354^ of rifier #871 in the Pinfirialfi Fr
I«i4
Bk Mountain I
Elkhom .
"igure 2.1 4 GPS locations (r = 3,354) of deer #871 in the Pinedale Front Winter Range Complex,
December 20, 2002 - April 1 5, 2003 (green) and November 1 , 2003 - April 1 5, 2004 (blue).
25
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
""•ryrl 'j • GPS 2003-04 #873
L, ■ ■ i| m GPS 2002-03 #873
V^'^i H □ PAPA Boundajy
■igure2.15 GPS locations (r= 3,344) of deer #873 in the Pinedale Front Winter Range Complex,
December 20, 2002 - April 1 5, 2003 (green) and November 1 , 2003 - April 1 5, 2004 (blue).
26
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
^jtfegulC'^ Boulder Laket?]"-^" T I *^ ' W
^"t^yfr-^^irr^'i^i Fremont Butta|- '- -—4-
0.
? *5>^«»vfC!r' -■ .''-^): ■■■■ :■< Y /i
• GPS 200344 #S74
« GPS 200243 #874
r^: PAPA Boundijry
10 MIJeB iu^
'''.h-Y'
:y:/<r-ir-;4-J..-rW|
-^•*'
■igure 2.16 GPS locations (r= 3,328) of deer #874 in the Pinedale Front Winter Range Connplex,
December 20, 2002- April 15, 2003 (green) and November 1, 2003 -April 15, 2004 (blue).
27
•
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
2.4.4 Population Characteristics
2.4.4.1 Abundance and Density Estimates
Helicopter flights were conducted on February 16-17, 2004 to count deer in selected l-mi' quadrats of
both treatment and control areas. Average flight time per quadrat was 10 minutes. Estimated deer
abundance {n ) in the treatment area was 3,564 ± 650 and deer density (d) was 52 ± 10 deer/mi (Table
2 2) Deer abundance and density were lower than previous years. Since 2002, the standard errors and
coefficients of variation have steadily decreased as the number of sampled quadrats {u) has increased.
Table 2.2 Summary statistics for abundance and density estimates in the treatment area during
Summary Statistics
Treatment Area
(The Mesa)
Year
2002
2003
2004
Total Quadrats {U)
68
66
68
Quadrats Sampled (w)
18
32
34
Deer Counted {N)
1,384
2,267
1,782
Density Estimate ( D )
77
71
52
Variance ( VariD) )
146
87
34
Standard Error (5^(D))
12.07
9.30
5.82
90% Confidence Interval
(57, 97)
(56, 86)
(42, 62)
Abundance Estimate(iV )
5,228
4,676
3,564
Variance ( Var{N) )
673863
377132
156318
Standard Error (5£(A^))
821
614
395
90% Confidence Inten/al
(3878, 6578)
(3666, 5686)
(2914, 4214)
Coefficient of Variation ( CV{N) )
16%
13%
11%
28
Sublette Mule Deer Study: 2004 Annual Report WEST, Inc.
Estimated deer abundance (w) in tine control area was 2,411 ± 541 and deer density (o) was 34 ± 8
deer/mi^ (Table 2.3). Abundance and density estimates for tiie control area inave been variable since
2002. Because the sampling frame in the control area did not reflect the area utilized by our marked
population, abundance and density estimates are expected to be biased high during 2002 and biased low
during 2003.
Table 2.3 Summary statistics for abundance and density estimates in the control area during
February 2002, 2003, and 2004.
Summary Statistics
Control Area
(Pinedale Front)
Year
2002^
2003f
2004
Total Quadrats {U)
35
38
70
Quadrats Sampled (u)
7
18
34
Deer Counted (A/)
810
849
1,171
Density Estimate ( D )
116
47
34
Variance (FarCD))
406
64
22
Standard Error (5£0))
20.14
8.01
4.70
90% Confidence Interval
(83, 149)
(31,63)
(26, 42)
Abundance Estimate( TV^ )
4,050
1,792
2,411
Variance (Fa7-(iV'))
496752
92661
108347
Standard Error (5£(7\0)
705
304
329
90% Confidence Interval
(2891 , 5209)
(1291,2293)
(1870,2952)
Coefficient of Variation { CV{N) )
17%
17%
14%
Abundance and density estimates expected to be high.
" Abundance and density estimates expected to be low.
29
Sublette Mule Deer Study: 2004 Annual Report
2.4.4.2 Reproduction
WEST, Inc.
Year
Treatment
Control
fawnidoe
December
fawn:doe
December
Pre-Development
1992-93
62
61
1993-94
47
51
1994-95
61
72
1995-96
56
63
1996-97
73
75
1997-98
92
81
1998-99
67
76
1999-00
85
76
Average
68
69
Development
2000-01
85
81
2001-02
69
71
2002-03
64
65
2003-04
78
78
Average
74
74
The WGFD conducted helicopter composition (buck:doe:fawn)
surveys to collect pre-winter (December) information on the sex (i.e.,
buck or doe) and age (i.e., fawn or adult) structure of the population.
A total of 9,704 deer were classified in December, 2003, including
3,264 on the Mesa and 6,440 on the Pinedale Front (S. Smith,
WGFD, unpublished data). Estimated fawnidoe ratios were 78:100
in both areas (Table 2.4), well above the 12-year averages.
Treatment and control areas have displayed similar trends in
reproduction (fawn:doe ratios) prior to and since the PAPA ROD in
2000 (Figure 2.1 7).
Table 2.4 (Left) Mule deer fawn:doe ratios measured for
treatment (Mesa) and control (Pinedale Front)
areas by Wyoming Game and Fish Department,
1992-2004.
90 n
1999-00 2000-01 2001-02 2002-03
Winter Year
2003-04
Figure 2.17 December fawn:doe ratios in treatment and control areas, 1999-2004.
30
Sublette Mule Deer Study: 2004 Annual Report
2.4.4.3 Over-winter Fawn Survival
WEST, Inc.
The WGFD conducted ground-based composition (adultifawn) surveys to estimate post-winter
fawn:adult ratios during April of 2004 (S. Smith, WGFD, unpublished data). A total of 1,534 and 2,212
deer were counted in the Mesa and Pinedaie Front, respectively (Tables 2.5-2.6). Estimates of over-
winter fawn survival were 0.36 and 0.33 in the Mesa and Rnedale Front, respectively (Tables 2.5-
2.6). Until this year, over-winter fawn survival has generally been lower in the treatment area
compared to the control, since the PAPA ROD in 2000 (Figure 2.18). Low fawn survival rates were
indicative of the severe winter conditions during 2003-04 that appeared to affect both winter ranges.
Table 2.5 Mule deer count data and calculations for over-winter fawn survival in the control (Pinedaie
Front), 1999-2004.
Year
December
Adults
December
Fawns
April
Adults
April
Fawns
A*
e**
1999-00
2698
1517
959
494
0.56
0.52
0.83
0,76
2000-01
2853
1769
955
478
0.62
0.50
0.85
0.69
2001-02
4593
2455
790
300
0.53
0.38
0.85
0.60
2002-03
3565
1813
704
254
0.51
0.36
0.96
0.68
2003-04
3977
2463
1771
441
0.62
0.25
0.82
0.33
A = count of December fawns/count of December adults
** B- count of April fawns/count of April adults
Table 2.6
Mule deer count data and calculations for over-winter fawn survival in the treatment
(Mesa), 1999-2004.
Year
December
Adults
December
Fawns
April
Adults
April
Fawns
A
B
1999-00
2550
1547
1390
764
0.61
0.55
0.87
0.74
2000-01
2420
1458
1685
707
0.60
0.42
0.82
0.59
2001-02
2546
1275
1366
460
0.50
0.34
0.85
0,57
2002-03
1864
914
1489
470
0.49
0.32
0.88
0.57
2003-04
2063
1201
1215
319
0.58
0.26
0.79
0,36
0.90
I 0.80
I 0.70
OT 0.60
I 0.50
(0
^ 0.40
0)
-£ 0.30
1 0.20
a>
<5 0.10
0.00
-El — Control
-A— Treatment
1 1 1 1
1999-00 2000-01 2001-02 2002-03 2003-04
-igure 2.1 8 Estimated over-winter fawn survival in treatment and control areas, 1999-2004
31
m
Sublette Mule Deer Study: 2004 Annual Report
2.4.4.4 Adult Winter Survival
WEST, Inc.
Winter (December 15 - May 1) survival rates were estimated using the Kaplan-Meier technique
(Kaplan and Meier 1985) and telemetry records of 68 radio-collared adult female deer, including 40 in
the treatment and 28 in the control area (Table 2.1). Winter survival rates were 0.79 and 0.82 for the
treatment and control areas, respectively (Table 2.7). Field necropsy of 10 animals indicated that
starvation was the cause of death (Photos 1-2). Survival rates were lower in both areas than previous
years (Figure 2.19) and likely a result of the severe winter conditions during the 2003-04 winter.
Snow conditions during the February flights were the heaviest and most extensive since the inception
of this study in 1998 (Photos 3-6). Most adult mortality occurred in March and April.
Table 2.7 Winter (2
treatment
D03-04) survival rates and summary statistics for adult female radio-collared deer in
and control areas.
Study Area
Time Period
Ni
A/2
S
90% CI
SE
Pinedale Anticline
(Treatment)
December 15, 2003 -
April 15, 2004
40
8
0.79
0.71 to 0.86
0.06
Pinedale Front
(Control)
December15, 2003-
April15, 2004
29
5
0.82
0.73 to 0.91
0.07
Ni= number of available collars, N2= number of deaths, S = survival estimate, Cl= confidence interval, SE= standard error
•
1.00
0.90
0.80
15 0.70
>
(0
3
■u
<
0.60
0.50
^ 0.40
0.30
0.20
0.10
0.00
ssBsssmmsimsm
m
-m— Control
-A— Treatment
1 — 1 1 1 1
1998-99 1999-00 2000-01 2001-02 2002-03 2003-04
Winter Year
•
Figure 2.1 9 Winter survival rates of adult female radio-collared deer in treatment and control areas,
1998-2004.
32
•
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
Photo 2.1 Winter mule deer mortality on south end of Mesa.
^^'^ ite
Photo 2.2 Cross-section of femur from radio-collared mule deer on the Mesa, March 2004. Bone marrow
color (reddish) and consistency (gelatinous) suggests chronic starvation.
33
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
Photo 2.3 Snow conditions in the Pinedale Front along the Big Sandy River (view north),
February 2004.
Photo 2.4 Snow conditions in Pinedale Front along the Big Sandy River (view south),
February 2004.
34
Sublette Mule Deer Study: 2004 Annual Report
WEST. Inc.
Photo 2.5 Snow conditions on the IVlesa near IVlount Airy (view north), February 2004
Photo 2.6 Snow conditions on the Mesa near Two Buttes (view south), February 2004.
35
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
2.4.5 Direct Habitat Loss
2.4.5.1 Pre-Development
Prior to development, The Mesa portion of tiie PAPA was relatively undisturbed, with very few
improved roads and approximately a dozen existing well pads (Figure 2.20).
•
•
Figure 2.20 Satellite image of the Mesa on October 1999, prior to development of the Pinedale
Anticline Project Area (PAPA).
2.4.5.2 Year 1 of Development
The BLM's ROD for the PAPA was released in July, 2000. Accordingly, natural gas development was
minimal during this year. Approximately 11 mi of new roads and 39 acres of well pads were
constructed on the Mesa during 2000 (Table 2.8). Approximately 51% of total surface disturbance
was associated with road building, while the other 49% was attributed to well pad construction (Table
2.8).
36
•
Sublette Mule Deer Study: 2004 Annual Report
2.4.5.3 Year 2 of Development
WEST, Inc.
2001 marked the first full calendar year of gas field development in the PAPA. Most development
occurred along the central portion of the Mesa, adjacent to Lovatt Draw (Figure 2.21). Based on
satellite imagery, approximately 13 mi of new roads and 113 acres of v\/ell pads were constructed on
the Mesa during the first nine months 2001 (Table 2.8). Approximately 30% of total surface
disturbance was associated with road building, while the other 70% was attributed to well pad
construction (Table 2.8).
Figure 2.21 Satellite image of the Mesa taken in August 2001 , following 1 full year of gas
development in the Pinedale Anticline Project Area (PAPA).
37
Sublette Mule Deer Study: 2004 Annual Report
2.4.5.4 Year 3 of Development
WEST, Inc.
Similar to 2001, most development in 2002 occurred along the central portion of the Mesa, adjacent to
Lovatt Draw, from the Paradise Road northwest to Stewart Point (Figure 2.22). Drilling activity was
also evident on the northern Mesa, east of Stewart Point. Based on satellite imagery, approximately
18 mi of new roads and 201 acres of well pads were constructed on the Mesa between August 2001
and October 2002 (Table 2.8). Approximately 25% of total surface disturbance was associated with
road building, while the other 75% was attributed to well pad construction (Table 2.8).
Figure 2.22 Satellite image of the Mesa taken in October 2002, following 2.3 years of gas
development in the Pinedale Anticline Project Area (PAPA).
38
•
Sublette Mule Deer Study: 2004 Annual Report
2.4.5.5 Year 4 of Development
WEST. Inc.
Similar to 2001 and 2002, most gas development in 2003 occurred along the central portion of the
Mesa, adjacent to Lovatt Draw, from the Paradise Road northwest to Stewart Point (Figure 2.23).
Drilling activity was also evident on the northern Mesa, east of Stewart Point. Based on satellite
imagery, approximately 14 mi of new roads and 237 acres of well pads were constructed on the Mesa
between October 2002 and September 2003 (Table 2.8). Approximately 18% of total surface
disturbance was associated with road building, while the other 82% was attributed to well pad
construction (Table 2.8)
pw^i .-,„V'*-1».* ^-
^y
'u' ;
>
i?..
IverflHJe. 1
^m
k.'?'
1 ' /■ '"'€:■
J Green R
■|3:
t'- :•
t'
l^^w
1
W^
II ^
1 ^
2 3 Milesf^l
1. ^H
i m
-''^^9||P^HpI^^ New Fork River '
Figure 2.23 Sateilite image of the Mesa taken in September 2003, following 3.25 years of gas
development in the Pinedale Anticline Project Area (PAPA).
Table 2.8 Summary of annual and cumulative direct habitat loss (i.e., surface disturbance) associated
with road
networks and well pads on the Mesa
, 2000-2003.
Year
Roads
(mi)
Roads
(acres)^
Well Pads
(acres)
Total
(acres)
% Roads
% Well
Pads
2000
11.4
41
39
80
51%
49%
2001
13.1
48
113
161
30%
70%
2002
18.1
66
201
267
25%
75%
2003
13.9
51
237
288
18%
82%
Total
a „
56.5
206
590
796
26%
74%
Based on an average road width of 30
39
•
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
2.4.6 Resource Selection
Note: Because some GPS-collars operate on the same deer for consecutive years, there is a lag time
between when the data are collected and when they are analyzed. Resource selection analyses
have been completed for data collected through the 2002-03 winter.
Population-level models (Table 2.9) and predictive maps (Figures 2.25-2.28) were estimated for four
winter periods: Pre-Development (Winters 1998-99 and 1999-00), Year 1 of Development (Winter
2000-01), Year 2 of Development (Winter 2001-02), and Year 3 of Development (Winter 2002-03).
Stepwise' model building resulted in slightly different models for the three years of development.
However, consistent predictor variables across years included slope, elevation, and distance to well
pad. Road density and aspect were variable across years, in terms of their ability to predict deer use.
Probability of mule deer habitat use increased as elevation increased. Slope, distance to well pad,
and road density had non-linear (quadratic) effects in the models. When plotted against probability of
use, the optimal values of the variables: slope, distance to well pad, and road density, occurs at the
highest point of the curve, as illustrated in Figure 2.24. The optimal values can be approximated from
the figures, but specific estimates were calculated using calculus (i.e. '1'' derivative test') to find the
value of the variable corresponding to the maximum probability of use. Table 2.10 lists the estimated
optimal values for all predictor variables with quadratic terms in each model. These values were
helpful in quantifying and interpreting how mule deer responded to predictor variables through time.
Using the distance to well pad variable as an example, the optimal values increased as development
progressed, indicating that mule deer were selecting for habitats farther away from well pads as
development progressed from Year 1 to Year 3. This non-linear (quadratic) relationship suggests
mule deer selected areas away from well pads, up to a certain distance. We believe this reflects the
presence of less suitable habitats near the study area boundary, compared to the more suitable
habitats in the central portion of the Mesa where most gas development has occurred. Additionally,
habitat use near the study area boundary is likely influenced by factors outside the boundary, such as
less desirable agricultural and riparian habitats (e.g.. Green and New Fork Rivers), and other gas
development infrastructure and activity (e.g.. Sand Springs Draw).
Table 2.9 Habitat variables and estimated coefficients for winter mule deer resource selection
probability functions (RSPF), 1 998-2003.
Variable
Intercept
Elevation
Slope
Slope^
Well
distance
Well
distance^
Road
density
Road
density^
Aspect =
NE
Aspect =
NW
Pre-Development
(i
-29.736
0.009
0.088
-0.003
0.116
-0.007
-0.185
-0.063
Aspect ■
SE
Aspect
SW
SE
Yearl
R.
-46.219
0.014
0.304
-0.018
4.327
-0.939
-1.650
0.513
ns
SE
9.836
0.004
0.074
0.005
1.903
0.372
0.606
0.212
ns
0.001
0.009
0.003
0.004
0.967
0.039
0.023
0.032
Year 2
fi
-44.398
0,013
0.260
-0.015
2.834
-0.408
0.838
-0.442
-0.481
SE
6.655
0.003
0.075
0.004
1.283
0.174
0.430
0.161
0.151
Year 3
(i
<0.001
<0.001
<0.001
<0.001
0.017
0.012
0.032
0.005
0.007
0.305
0.051
-0.321
0.167
0.154
0.165
0.09
-76.067
0.023
0.250
-0.012
7.593
-0.883
ns"=
ns
ns
ns
0.743
0.072
SE
8.681
0.003
0.095
0.006
2.430
0.292
<0.001
0.001
0.069
0.132
0.04
0.045
ns
VV \ _1 ! I 1 I ' ■ "—
Standard errors and P-values were not presented for tine Pre-Development RSPF coefficients because stepwise model
building was not performed
40
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
' P-values for Year 1 based on student's t distribution with 9 degrees of freedom.
' P-values for Year 2 based on student's t distribution with 14 degrees of freedom.
' P-values for Year 3 based on student's t distribution with 6 degrees of freedom.
' Not significant.
I \
/ \
0 2 4
Distance to Nearest
Year1
□ 2 4
Distance to Nearest Well Pad
Year 2
N
/
/
\
\
/
\
0 2 4 6
Distance to Nearest Well Pad (km)
Years
Figure 2.24 Relationship between probability of habitat use and distance to well pad, for development
years 1-3.
Table 2.10 Optimal values 1
for predictor variables with quadratric terms.
Year
Variable
Optimum Value '
Year 1 of Development
2000-01
Distance to Well Pad"
2.3 km
Road Density"^
0 km/knf
Slope^
8°
Year 2 of Development
2001-02
Distance to Well Pad"
3.5 km
Road Density^
0.95 km/km''
Slope^
9°
Year 3 of Development
2002-03
Distance to Well Pad"
4.3 km
Road Density^
Ns
Slope^
10°
41
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
•
2.4.6.1 Pre-Development: Winters 1998-99 and 1999-00
The population- level model was estimated from a random sample of 4,469 habitat units and 953 VHF
deer locations collected from 76 adult female mule deer during the winters (December 1- April 15) of
1998-99 and 1999-00. Because pre-development data consisted of many deer with relatively few
locations, we could not treat the deer as the experimental init and conduct the same model building
process we did with the GPS-collared deer in subsequent years. Rather, we fit a model based on
habitat variables that were consistently included in models from subsequent years (i.e., elevation,
slope, distance to vvell, and road density). Consequently, we did not estimate standard errors or p-
values for the coefficients (Table 2.9). Predicted probabilities indicate deer selected for areas with
high elevation and moderate slopes.
Figure 2.25 Surface map depicting probability of mule deer habitat use prior to development (1 998-
99, 1999-00 winters). Color-coded based on percentile of predictions (i.e., 0-25%, 25-
50%, 50-75%, and 75-100%).
42
•
•
Sublette Mule Deer Study: 2004 Annual Report
WEST, Inc.
2.4.6.2 Year 1 of Development: Winter 2000-01
The population-level model was estimated from a random sample of 4,489 habitat units and 18,706
GPS locations collected from 10 adult female mule deer during the winter (January 1- April 15) of
2000-01. The model included elevation, slope, distance to well pad, and road density (Table 2.9).
Deer selected for areas with higher elevations, moderate slopes, low road densities, and away from
well pads. Areas with the highest probability of use had elevations of 2,325 - 2,400 m, slopes of 15
degrees, road densities of 0.0 km/km^, and were 2.3 km away from the nearest well pad. Predicted
probabilities indicate deer use was lowest in those areas of the PAPA where well pads and associated
road networks were developed (Figure 2.26). Portions of the study area that were classified as "high
deer use" prior to development were classified as "low deer use" during year 1 of development. The
model suggests these changes in deer distribution were due to well pad and road development.
Figure 2.26 Surface map depicting probability of mule deer habitat use during Year 1 of development
(2000-01 winter). Color-coded based on percentile of predictions (i.e., 0-25%, 25-50%,
50-75%, and 75-100%).
43
Sublette Mule Deer Study: 2004 Annual Report
2.4.6.3 Year 2 of Development: Winter 2001-02
WEST, Inc.
The population-level model was estimated from a random sample of 4,464 habitat units and 14,851
GPS locations collected from 15 adult female mule deer during the winter (January 4 - April 15) of
2001-02. The model included elevation, slope, aspect, distance to well pad, and road density (Table
2.9). Deer selected for areas with northwest aspect (positive coefficient, smallest p-value), high
elevations, moderate slopes, low road densities, and away from well pads. Areas with the highest
probability of use had northwest aspects, elevations of 2,325 - 2,400 m, slopes of 8 degrees, road
densities <0.95 km/knf , and were 3.5 km away from the nearest well pad. Predicted probabilities
indicate deer use was lowest in those areas of the PAPA where well pads and associated road
networks were developed (Figure 2.27). Portions of the study area that were classified as "high deer
use" prior to development were classified as "low deer use" during year 2 of development. The model
suggests these changes in deer distribution were due to well pad and road development.
Figure 2.27 Surface map depicting probability of mule deer habitat use during Year 2 of development
(2001-02 winter). Color-coded based on percentile of predictions (i.e., 0-25%, 25-50%,
50-75%, and 75-100%).
44
Sublette Mule Deer Study: 2004 Annual Report
2.4.6.4 Year 3 of Development: Winter 2002-03
WEST, Inc.
The model was estimated from a random sample of 4,462 habitat units and 5,131 GPS locations
collected from 7 adult female mule deer during the winter (December 20 - April 15) of 2002-03. Our
target sample of 10 marked animals was not met because several deer died early in the season
and/or spent most of their time outside the study area. Deer were distributed farther north than most
years (Sawyer et al. 2003). The RSPF included elevation, slope, and distance to well pad (Table 2.9).
Deer selected areas with higher elevations, gentle slopes, and away from well pads. Areas with the
highest probability of use had elevations of 2,325 - 2,400 m, slopes of 9 degrees, and 4.3 km away
from the nearest well pad. Predicted probabilities indicate deer use was lowest in those areas of the
PAPA where well pads were developed (Figure 2.28). Portions of the study area that were classified
as "high deer use" prior to development were classified as "low deer use" during year 3 of
development. The model suggests these changes in deer distribution were due to well pad
development.
Figure 2.28 Surface map depicting probability of mule deer habitat use during Year 3 of
development (2002-03 winter). Color-coded based on percentile of predictions
(i.e., 0-25%, 25-50%, 50-75%, and 75-100%).
45
Sublette Mule Deer Study: 2004 Annual Report WES T, Inc.
2.5 DISCUSSION AND FUTURE DIRECTION
Approximately 15,000 GPS locations were collected from 10 deer during the 2003-04 winter. Another
12 GPS collars should be recovered in 2005 and should provide an additional 24,000 bcations from
both the 2003-04 and 2004-05 winters. Although data collection is delayed a full year when GPS
collars operate on the same deer for two consecutive winters, acquiring movement and distribution
information from the same animal over a period of years provides useful year-to-year comparisons.
Last year, for example, we were able to demonstrate that deer on the Mesa wintered farther north
than in previous years (Saw/yer et ai. 2003). If we had not had GPS collars running for consecutive
winters on the same deer, it could have been argued that we captured deer that only utilized the
northern Mesa. And, this year we were able to demonstrate the strong affinity mule deer have for a
50-mile migration route along the Pinedale Front.
Basic distribution maps generated from GPS data illustrated winter use of the control and treatment
areas, demonstrated the importance of BLM lands, and refined information on migration routes and
seasonal ranges. Deer in the treatment area (Mesa) continued to utilize the Trapper's Point Bottleneck
(TPB) as a migratory route between winter and spring/fall transition ranges. Deer movements through
the TPB were quick, as evident by the distance between locations, but the TPB continued to function
effectively during 2004. Agencies, industry, NGO's, and the public recognize the value of maintaining
this movement corridor for the Sublette deer herd. Land-use decisions in and adjacent to the TPB should
consider the migration routes and seasonal ranges of the Sublette deer herd.
Consistent with the 2002-03 winter, deer distribution and movement patterns in the control area (Pinedale
Front) were variable during the 2003-04 winter. While deer spent some time in the original sampling
frame (Figure 2.2), they also moved (10 - 15 mi) in all directions; south to Elk Mountain, southeast along
the Big Sandy, easterly to the Little Sandy and Prospects, and northerly to Muddy Mountain. The ability
to alter their rates of movements, to change their pathways, and occupy a variety of winter habitats as
needed are behaviors that likely allow these deer to best exploit winter ranges. However, the
unpredictable movement patterns have made calculation of abundance and density estimates difficult, as
the size of the sampling frame has progressively increased over the study to encompass the range of our
marked deer. Because the sampling frame did not reflect the area utilized by our marked population,
abundance and density estimates are expected to be biased high during 2002, and biased low during
2003. Assuming the 2004 sampling frame accurately reflects the winter range extent used by marked
deer, the abundance and density estimates should be more reliable than those calculated during 2002
and 2003.
Although winter distribution patterns varied among deer in the Pinedale Front, the migration route to
northerly transition and summer ranges was surprisingly consistent. All GPS-collared deer captured in
the control area migrated along a distinct movement corridor located at the base of the Wind River
Range. Deer followed a well-defined route that rarely exceeded 1-mi in width and covered a distance of
50 mi. GPS data collected from individual deer for consecutive winters showed a strong affinity for this
migration corridor during both spring and fall migrations. Deer that winter in the Rnedale Front were
known to migrate northerly along the Wind River Range to the New Fork Lake area, before shifting
their migration in a westerly direction towards the Hoback Basin and adjacent mountain ranges
(Sawyer and Lindzey 2001). However, details of this migration route, in terms of size, width, specific
location, and deer fidelity were unknown prior to GPS data collected over the last two years. Although
these deer may migrate 100 mi between winter and summer ranges (Sawyer and Lindzey 2001), our
GPS collars do not collect locations May - October, and therefore do not record the entire migration
route(s). Deer management in the Sublette DAU is complicated by the long-distance migrations that
occur through a variety of habitats and across a mix of land ownership. Knowledge of this migration
route should provide agencies with the necessary information to maintain deer movements through the
Pinedale Front, adjust harvest strategies accordingly, and prioritize habitat enhancement projects.
46
•
Sublette Mule Deer Study: 2004 Annual Report WES T, Inc.
Because several thousand mule deer rely on this migration corridor to access their seasonal ranges,
maintenance of the corridor should be a priority for agencies and conservation groups alike.
In addition to basic distribution and movement maps, GPS data can be used to conduct more
rigorous scientific analyses, such as estimation of resource selection models (Manly et al. 2002) and
site-specific analyses for industry (TRC Mariah 2003, Sawyer et al. 2004), agencies, or NGOs
regarding permitting, stipulations, or other wildlife-related issues. Resource selection, as described by
Manly et al. (2002), is a rapidly advancing methodology for analyzing, modeling, and interpreting
wildlife field studies. Resource selection analyses have broad applications, and in the case of this
study, were used to determine how mule deer use their habitats in relation to various habitat features,
including well pads and road networks associated with energy development. Our basic approach to
resource selection treats the GPS-collared deer as the experimental unit and estimates a population-
level model (i.e., RSPF) so inference can be made to the entire population of mule deer on the Mesa.
Sample size, in this case the number of collared mule deer, is an important consideration for statistical
procedures that rely on simple random sampling to obtain population-level inference.
Numerous authors have calculated minimum sample sizes for home range analysis, survival
estimates, and resource selection, using both simulations (Alldredge and Ratti 1986) and empirical
data (Powell et al. 2000). We recognize the number of marked animals in our analysis was smaller
than the sample size of 25 recommended by Powell et al. (2000). However, we believe our sample
adequately represents the Mesa deer population because of our a priori knowledge of deer movement
and distribution patterns. Sawyer and Lindzey (2001) studied the movement and distribution patterns
of this mule deer population for three years (1998-2000) prior to gas field development. Their results
indicated that most deer congregate in the northern portion of the study area during eariy winter,
before moving on to their respective winter ranges. Our sampling design incorporated this
knowledge, in that we assumed obtaining a random sample was most likely to occur when deer were
congregated on the north end. And, in fact when we compared the distribution (i.e., spatial extent) of
2,563 locations from 82 radio-collared deer for a 3-year period (1998-2000), to the distribution of
18,711 locations from 10 GPS-collared deer for a 1-year time period, the spatial extents were cleariy
similar (H. Sawyer, WEST, Inc., unpublished data). Thus, while our sample sizes may be less than
those suggested by Powell et al. (2000), we believe our sampling strategy and model-building process
adequately represents the PAPA deer population. Larger sample sizes should reveal the same
relationships between the probability of mule deer habitat use and environmental conditions, but with
higher precision.
Recent advances in GPS technology allow for the collection of thousands of locations per animal and
thereby make it necessary to reassess the performance of resource selection methods in relation to
the number of animals versus the number of observations per animal (Leban et al. 2001). When a
large number (e.g., >1,000) of animal locations evenly spread over a study period are available, as is
often the case with GPS-collar studies, the utilization distribution (UD) will provide an accurate
estimate of the probability of use of habitat units for a particular animal. Accurate estimation of the
UD is most likely to occur when precise, frequent, and unbiased animal locations are collected from a
representative sample of study animals. Given these requirements, our GPS data appears well-suited
for UD estimation, as they lack the inherent bias often associated with traditional telemetry methods
because they are systematically collected, irrespective of human error, poor weather conditions, or time
of day. Forest density and canopy cover may reduce the fix rates and positional accuracy of GPS
collars (Rempel et al. 1995, Moen et al. 1996, D'Eon 2002, Di Orio 2003). However, the open
shrublands and gentle topography of the PAPA allow for optimal satellite coverage and GPS
performance. Fix-rate GPS bias associated with different habitat types has not been an issue in this
study, as evidenced by high (99%) fix rate success and precision (80% 3-D locations). And, since the
discontinuation of selective availability in May 2000, 3-D GPS locations typically have less than 20 m
error (Di Orio 2003).
47
APPENDIX B: Resource Selection Modeling Procedures
We fit the same RSPF to each of the n individuals from each winter and estimated population-level
RSPF coefficients by
where p.. is the estimate of coefficient / for individual / The variance of the estimated population-level
RSPF coefficients was estimated by
var(A.) = -^Z()8,-A)'- (4)
A stepwise model building procedure (Neter et al. 1996) was used to estimate population-level models
for winters 2000-01, 2001-02, and 2002-03. Significance levels for determining variable entry (a =
0.15) and exit (a = 0.2) from the RSPF were based on Hosmer and Lemshow (2000). The stepwise
model building involved seven steps:
1) Each variable {Xj) was used to fit a simple UD model containing only that variable for each
deer («,) within a winter.
2) The average and standard error of the «, estimated coefficients were calculated for each
variable.
3) A student's t-test for the statistical significance of each variable determined which variable
entered the model first. The variable with the lowest p-vaiue, provided it was < 0.15 , was
the first to enter the model.
4) The new probability model was fit to data from each deer; each model contained the most
significant variable identified in step (3) and each of the other variables previously
selected. Again, all remaining variables were considered.
5) The average and standard error of the «, estimated coefficients were computed for those
variables not identified in step (3) as most significant. A student's t-test for the statistical
significance of those variables determined the next variable to enter the model. Again, the
variable with the lowest p-value less than or equal to a = 0.15 was chosen.
6) This new model containing two predictor variables was fit for each deer and the statistical
significance of the variable chosen in step (3) was re-assessed. If the p-value from the t-
test for this variable still met the criteria to be in the model, it was l<ept. If the p-value
changed and exceeded the a = 0.2 cutoff, it was dropped from the model.
7) This process was repeated, and variables were added or dropped from the model until no
further variables entered or exited the model.
Our analysis correctly identified the deer as the experimental unit, and because it was necessary to
have the same UD model (i.e., same predictor variables) for each deer within a winter, we used we
used a stepwise (Neter et al. 1996) model building process to produce both individual and population
level models for each winter, as recommended by Erickson et al. (2001). It is not possible to use
Akaike's Information Criterion (AlC) and similar alternatives such as Akaike s Information Criterion
second order variant (AlCc) and Bayesian Information Criterion (BIC) (Burnham and Anderson 2002)
for model selection in this context, since this could result in a different model for each deer within a
winter and prevent estimation of population level models.
Stepwise model building was conducted for all winters except pre-development. Prior to development
of the PAPA, the deer monitoring consisted of many VHF-collared deer with relatively few locations
(Sawyer and Lindzey 2001), compared to the GPS collars used from 2000 to present. Because of the
limited number of locations recorded for the VHF-collared deer we could not fit individual RSPF's to
these deer. Therefore, we pooled data across deer (n=76) and winters (1998-99, 1999-00) to
57
APPENDIX B: Resource Selection Modeling Procedures
~ " ~ """ •
estimate a population-level model (RSPF) for the pre-development time period. Simple random
samples of 30 locations were taken from deer with > 30 locations to ensure approximately equal
weight was given to each deer in the analysis. All variables except aspect were included in the
population level model for pre-development.
Predictions (i.e., probability of mule deer habitat use) of population-level models were mapped for
each winter on 104 x 104 m grids that covered the study area. Predictions were first scaled so that
the total probability corresponding to all grid cells in the study area totaled 1.0. To aid in mapping the
estimated probability of use, each grid cell was assigned a value of 1 to 4. Grid cells with the largest
25% of values for predicted probability of use were assigned 1 (= high use), cells in the 51 to 75
percentiles were assigned 2 (= medium high use), cells in the 26 to 50 percentiles were assigned 3 (=
medium low use), and cells in the 0 to 25 percentiles were assigned 4 (= low use).
,^«
58 \>-'Ae-"<'^-'^.v->>'' e^
-•A
■<!••
v^
^^El
APPENDIX C: GPS locations and 60-mile migration route of deer #863.
^ ^ ^ '- irm-m;] ■ '- — '
•
59
•
&\^- ^'"^ ..^...1 Center
Denver re^;-^"^
P'^-"^'-^omdo 80225
Denver /--"^^^"