| SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

BULLETIN

UTS 83 | : | LI RB R A RY | Number 3
DEC 28 1084

NEW YORK
BOTANICAL GARDEN

a

Saar rarer i 5, 5 nT 25) ys EL
et ale \
f “Po 1

aK

yay

_ BCAS-A83(3) 113-168 (1984) DECEMBER 1984

Southern California Academy of Sciences

Founded 6 November 1891, incorporated 17 May 1907
© Southern California Academy of Sciences, 1984

OFFICERS

Richard E. Pieper, President
Peter L. Haaker, Vice President
Camm C. Swift, Secretary
Robert G. Zahary, Treasurer
Robert J. Lavenberg, Technical Editor
Gretchen Sibley, Managing Editor

BOARD OF DIRECTORS

1982-1984 1983-1985 © 1984-1986
Jules M. Crane William A. Akersten Charles P. Galt
Michael H. Horn Takashi Hoshizaki Peter L. Haaker
Richard E. Pieper Steven N. Murray Harlan Lewis
John D. Soule Camm C. Swift June Lindstedt-Siva
Gloria Takahashi Robert G. Zahary Martin L. Morton

Membership is open to scholars in the fields of natural and social sciences, and to any person interested
in the advancement of science. Dues for membership, changes of address, and requests for missing
numbers lost in shipment should be addressed to: Southern California Academy of Sciences, the Natural
History Museum of Los Angeles County, Exposition Park, Los Angeles, California 90007.

Annual Members .°3 < 6-2°i.50-. 3 3. be ew ew we ee ws oe Kaw 64 a Oe ee $ 15.00
Student Members.... . a ar en ee a errr ee ee 10.00
Life Members............. ete eee ee Be ee ee Be 300.00

Fellows: Elected by the Board of Directors for meritorious services.

The Bulletin is published three times each year by the Academy. Manuscripts for publication should
be sent to the appropriate editor as explained in “Instructions for Authors” on the inside back cover
of each number. All other communications should be addressed to the Southern California Academy
of Sciences in care of the Natural History Museum of Los Angeles County, Exposition Park, Los
Angeles, California 90007.

Date of this issue 11 December 1984

Bull. Southern California Acad. Sci.
83(3), 1984, pp. 113-120
© Southern California Academy of Sciences, 1984

Factors Affecting Germination of Chaparral Seeds

Jon E. Keeley

Abstract.— Factors affecting germination of chaparral seeds by Jon E. Keeley.
Bull. Southern California Acad. Sci., 83(3):113-120, 1984. Seedling establishment
is uncommon under mature chaparral but abundant after fire. Germination from
soils collected beneath chaparral and in an adjacent burned site were compared
after various treatments. Application of an aqueous leachate from the dominant
overstory shrub (Adenostoma fasciculatum) failed to produce any inhibition of
seedling emergence from the mature chaparral soil but rather increased germi-
nation. Soil heating and powdered charred wood (charate) stimulated germination
of dicots but not grasses. Tests of specific species showed Phacelia cicutaria and
Salvia columbariae markedly stimulated by charate whereas Cryptantha muricata
and Lotus salsuginosus were not. Grasses were abundant in the burn soil and
germinated readily without treatment; their residence time in the soil, however,
may be limited as they were uncommon in the mature chaparral soil. Light was
a significant factor in germination of both monocots and dicots. Continuous
darkness significantly reduced germination over a 12 hour photoperiod at ~230
HE m~?s_! or ~20 wE m-’sS"! or periodic (several hours/week) light of variable
intensity.

Chaparral is an evergreen scrub vegetation which dominates much of the un-
developed landscape in southern California. Widespread wildfires are relatively
frequent in chaparral due to the dry fuels (resulting from the mediterranean-
climate summer drought) and the dense contiguous nature of the vegetation (re-
sulting from the winter and spring rains which are coincident with mild temper-
atures). Today, man supplies the ignition source for the vast majority of wildfires,
though lightning-ignited fires are not uncommon (Keeley 1982). Wildfires are
thought to have played a major role in the evolution of chaparral, as the shrubs
exhibit a number of adaptations which have been interpreted as evolutionary
responses to fire; e.g., seedling recruitment largely restricted to the first postfire
year and sprouting from specialized basal burls.

Herbaceous species are generally uncommon in mature chaparral, however,
they dominate the post-fire environment for one to several years (Hanes 1977;
Keeley et al. 1981). The presence of many of these herb species on recently burned
sites is due to seeds dormant under the chaparral and dating back to the previous
fire.

Two classes of theories have been proposed to account for relatively depauperate
shrub and herb seedling establishment under mature chaparral, but abundant
recruitment after fire. Seeds of these species are either 1) inhibited from germi-
nating by chemical components of the environment (e.g., allelopathic substances
produced by the overstory shrubs or microbial by-products produced in the litter),
or 2) the seeds are refractory and require a stimulus from fire (cues that may
stimulate germination include heat, which may crack hard seed coats or melt
waxy coverings, and chemical stimulants released from charred wood).

114 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

The evidence for chemical inhibition of seed germination under the shrub
canopy is not compelling. Sweeney (1956) found that a leachate made from chap-
arral litter had no inhibitory effect on the germination of various native seeds.
McPherson and Muller (1969) could only demonstrate inhibition on four native
seeds and only after scarifying them and applying a highly concentrated (10x)
leachate from the Adenostoma fasciculatum shrub canopy (names according to
Munz 1974). Christensen and Muller (1975) found that a leachate from Adenos-
toma did inhibit the germination of two native and two non-native chaparral
herbs but had no effect on six of the more typical “‘pyrophyte endemics.”’ Ka-
minsky (1981) suggested microbial toxins produced in chaparral soils could inhibit
germination of native herbs and demonstrated such an effect on lettuce seeds.
Also, he found that soils from a second year burn showed no inhibitory effect.
Recently Pack (1985), using native herb seeds and 6 mature and 6 burn soils, was
not able to duplicate Kaminsky’s results.

Attempts to demonstrate that fire-related cues play a role in stimulating ger-
mination have likewise produced conflicting results. Sweeney (1956) found that
excelsior burned on top of soil did stimulate germination of native herb seeds
but attempts to attribute this to either a chemical product from burned wood or
heat were unsuccessful. Likewise McPherson and Muller (1969) and Christensen
and Muller (1975) were unable to demonstrate heat stimulated germination of
herbs although the latter study as well as others (e.g., Stone and Juhren 1951;
Hadley 1961; Quick and Quick 1961) have shown heat stimulated germination
of shrub seeds. Although Sweeney (1956) could not demonstrate any effect of
ashed wood products, Wicklow (1977), and later Jones and Schlesinger (1980),
demonstrated that partially charred wood stimulated germination of Emmenanthe
penduliflora.

It is clear that no single study is likely to completely clarify the situation. The
purpose of this study is to extend these previous investigations to examine how
“‘allelopathic leachate,”’ charred wood and heat affect germination of seeds from
natural chaparral soils. In addition, the role of light, which has previously been
ignored, was investigated.

Methods

The top 2—4 cm of soil was collected (in early fall) either from beneath mature
chaparral (~75 yrs as determined by ring counts) or an adjacent burn (after the
end of the second year of regrowth). These sites were in the San Gabriel Mtns. at
900 m, | km north of the Monte Cristo Ranger Station. Soils were sieved with a
3 mm screen and the few seeds that did not pass through were added back. The
soil was spread to ~3 mm depth on a tray and heated in a forced convection
oven at 80°C for 2 hrs or 120°C for 5 min. Petri dishes (100 x 15 mm) were filled
with 50 cc of this soil and leachate or charate (see below) treatments were applied.

Leachate from Adenostoma foliage was prepared as described by McPherson
and Muller (1969). Branches from the upper canopy were cut and placed in a 1
m? funnel and 3 liters of deionized water were applied in a fine mist over a period
of 2 hrs. The concentrate was prepared by warming and stirring leachate on a hot
plate, maintained below 50°C as suggested by McPherson and Muller (1969).

Charate was prepared from Adenostoma wood by charring stem segments with

CHAPARRAL SEED GERMINATION 115

a torch and grinding in a Wiley mill to pass through a #20 mesh screen. Ap-
proximately 2.5 g of powdered charate were applied per dish.

For dishes receiving the leachate treatment, 25 ml of leachate were added.
Charate treatments received 27 ml of deionized H,O, and 25 ml dH,O were added
to all other dishes. Dishes were stacked on trays and covered with black plastic
bags to reduce evaporation. The material was then incubated at 5°C for 1 wk,
followed by 23°C for 1 wk; and this cycle was repeated two times for a total of 6
weeks. The seeds of four herb species were then added and incubated in the same
manner. .

In these experiments seeds were incubated in the dark but were exposed to
several hours of light each week when germinated seeds were scored. An additional
experiment, using some of the treatments described above, was done with seeds
in continuous darkness for 6 wks (scoring was done under a green light) or under
a 12 hr light regime. The latter dishes were covered with clear plastic bags and
placed 150 mm beneath fluorescent lamps where they received ~230 wE m~?s“!.

The two soils, as well as a commercial potting soil (for comparison) were
analyzed for major inorganic nutrients by an outside commercial soil testing lab.

Results

Fertility of the soils from mature and burned chaparral are compared with a
commercial potting soil in Appendix I. Relative to the potting soil, both chaparral
soils were markedly depauperate in several important elements. The major soil
changes after fire included an increase in pH, Na*, PO,” and CO; ~? but a decrease
in NO,~.

The buffering capacity of all soils was similar with respect to Adenostoma
leachate. For example the pH of this leachate varied from 3.5 to 3.8, dependent
on concentration. When added to either soil, leachate lowered the pH of the soil
solution by 0.1 pH unit. Charate had a pH of 7.4 and when added to burn soil
it did not change the pH of the soil solution, but when added to the mature soil
the pH increased from 5.6 to 6.0. Since the organic matter was largely broken
down in the burn soil it tended to absorb much less water and thus these soils
tended to be wetter. Because of this difference it is possible that seeds were exposed
to different osmotic pressures in the two soils. Circumstantial evidence suggests

Table 1. Germination from soil collected beneath mature Adenostoma chaparral and in an adjacent
second year burn (n = 84).

Seedlings/50 cc soil

1x 4~x
Ade- Ade- Ade-
nostoma nostoma nostoma 80°C 120°C

Soil Control leachate leachate charate (2 hrs) (5 min) P LSD

Dicots Mature Chaparral 0.67 1.16 0.90 leo? Niel * 040
2nd yr Burn 4.76 4.70 4.48 Bo) GS Sali} BAP).

Monocots? Mature Chaparral 0.23 0.15 0.05 0.17 0.04 0.11 FA OLN
2nd yr Burn 10:82) WS) NOS OWT 908) 42460 2592232

IP SOW, 12 < Oe
2 All but one of the monocots were grasses.

116 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

MATURE BURNED

MONOCOTS ama Memece
DItGOus

19

Zz
O
ke
<x
{10
=
[ac
Ww
O
4
: :
eS \\ -
25 oN &
ia) . \ s
a s
fe) \eeegettennn
ox

7) UA scx Genel 28
ee
5°C 23°C 5°¢ I3°6

Fig. 1. Rate of germination of monocots and dicots from soils collected from mature chaparral
and an adjacent two year burn.

this was not a complication. As will be seen below, germination of Cryptantha
muricata did not differ on the two soils yet a separate test showed it was sensitive
to elevated osmotic pressure; germination was reduced from 77% (S.D. = 12, N =
3) in distilled water controls to 24% (S.D. = 7) in 0.2 M mannitol.

Dicot seedling emergence from mature soils was significantly increased by all
treatments, including the supposedly “‘allelopathic” leachate treatments (Table 1).
Charate produced significantly greater germination than any other treatment;
sixteen dicot species emerged from these plates: 51% were herbs, largely Descu-
rainia pinnata, Camissonia hirtella, Gnaphalium californicum and the remaining
49% were nearly all Adenostoma fasciculatum.

From the second year burn soil, dicot germination was much greater than from
the mature chaparral soil and no treatment increased germination over controls.
No shrub seedlings emerged from these soils though many of the herb species
were the same as those from the mature soil. Addition of charate to this soil
significantly reduced germination.

Monocot seeds were rare in mature chaparral soil but abundant in burned
chaparral soil. These seeds were essentially all from grasses, largely Bromus tec-
torum and Festuca megalura. Heat reduced germination of these grasses in both
soils.

CHAPARRAL SEED GERMINATION 117

Table 2. Percentage germination of species planted into soils after the end of the experiment in
Table 1. C.m. and P.c. were sown together and L.s. and S.c. were sown together thus n = 42 dishes
for each species. Due to the large sample size, seeds were added with a small measuring spoon; mean
number of seeds per spoon, +S.E. of the mean for n = 14, was: 128 + 6 (C.m.), 92 + 6 (P.c.), 108
+ 5 (L.s.) and 140 + 9 (S.c.). Note, for the heat treatment, the soils were heat treated but not the
seeds of the four species added.

4x
Con- Leach- Char- 120°C
Soil trol ate ate (5 min) P LSD
Cryptantha muricata Mature Chaparral 62 WD 78 74 NS
2nd yr Burn 64 58 V2 63 NS
Phacelia cicutaria Mature Chaparral 17 17 26 19 oS 6
2nd yr Burn 20 20 14 18 = 5
Mature Chaparral 21 19 21 19 NS
Lotus salsuginosus aA oe aid eae
2nd yr Burn 12 12 14 12 NS
Salvia columbariae Mature Chaparral 9 10 52 9 a 7
**K eK kK OK
2nd yr Burn 25 22 27 23 NS

NS P > 0.05, * P < 0.05, ** P < 0.01; P > 0.05 for mature vs burn comparisons not starred.

The rate of germination from these soils varied markedly (Fig. 1). For both
monocots and dicots germination was much more rapid from the burn soil than
from the mature chaparral soil.

After the indigenous seed pool had germinated, these same soil samples were
sown with four native herb species (Table 2). There were marked differences in
species responses. Cryptantha muricata germinated well under control conditions
and was not affected by any of the soil treatments. Phacelia cicutaria germination
was significantly increased with charate on mature soil and decreased with charate
on burn soil. Germination of Lotus salsuginosus was significantly lower on burn
soil regardless of treatment whereas Salvia columbariae germination was higher
on burn soil unless charate was added to control soil, where germination was high.

The observation that charate enhanced germination in mature chaparral soil
but inhibited germination (of some species) in burn soil (Tables 1 and 2) was
surprising. In the case of P. cicutaria and S. columbariae (Table 2) the soils had

Table 3. Seedling emergence from soils incubated in the light (12 hr photoperiod at ~230 hE m~
s-!) and total darkness, with and without charate (n = 30).

Seedlings/50 cc soil

Control Charate
Soil Light Dark Light Dark
Dicots Mature Chaparral 1.40 -* 0.37 3.80 Eee 1.70
2nd yr Burn 6.99 = 3°38 4.87 bah 2.80
Monocots Mature Chaparral 0.13 0.17 0.03 0.03
2nd yr Burn 10.30 on 7.30 10.67 + 8.53

* P < 0.05, ** P < 0.01; P > 0.05 for light vs dark comparisons not starred.

118 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

Table 4. Percentage germination of species sown into soils after the end of the experiment in Table
3 (n= 15).

Control Charate
Soil Light Dark Light Dark
Cryptantha muricata Mature Chaparral 73 x 2, 82 aa 12
2nd yr Burn 62 ce 10 56 haba 9
Phacelia cicutaria Mature Chaparral 10 x* 7 38 on 8
2nd yr Burn 6 9 3 11
Lotus salsuginosus Mature Chaparral 18 17 24 2
2nd yr Burn 14 14 13 15
Salvia columbariae Mature Chaparral 44 ice 1 74 i! 33
2nd yr Burn 84 ris 48 59 a 49

*P < 0.05, ** P < 0.01; P > 0.05 for light vs dark comparisons not starred.

been wetted and incubated for a month (to germinate the indigenous seed pool)
before sowing. I hypothesized that this long pre-incubation of the soil (and in
particular with the addition of charate) allowed growing microbial populations to
inhibit germination. I predicted that germination would be much greater if seeds
were sown into the burn soil without prior wet incubation of the soil. This was
tested with P. cicutaria and it was found that burn soils treated as before (with
prior incubation) produced 8% (S.D. = 8, N = 15) germination whereas if P.
cicutaria were sown into burn soil without prior incubation of that soil, germi-
nation was 39% (S.D. = 5, N = 15).

Effect of Light

In previous experiments light was not considered; seeds were incubated in the
dark although they were exposed to the light when scored. In this experiment
germination under a twelve hour photoperiod was compared with germination
in total darkness (Table 3). For both mature and burn soils, dicot and monocot
seed germination was significantly inhibited by total darkness. The four species
sown into these soils illustrate different responses to total darkness (Table 4); L.
salsuginosus was insensitive to darkness; C. muricata and S. columbariae were
markedly inhibited by total darkness with or without charate; P. cicutaria was
inhibited by darkness only in the presence of charate.

Apparently the previous experimental conditions of dark incubation followed
by scoring in the light approximated the “‘light stimulating”’ conditions of the 12

Table 5. Effect of light intensity on germination of seeds from mature chaparral soil and Cryptantha
muricata planted in the same soil (n = 12). For C.m. seed number is as in Table 2.

wE ms!
~230 ~20 ~0 Ie LSD
Dicots (NO/50 cc) 1.92 1.92 0.7 a 1.39
Monocots (NO/50 cc) 0.00 0.17 0.17 NS
Cryptantha muricata (%) 75 57 10 nabs 16

NS P > 0.05, ** P= 0:01:

CHAPARRAL SEED GERMINATION 119

hr photoperiod incubation. This is suggested by a comparison of the charate
response in the previous experiments (Tables 1 and 2) with the charate effect
under the light treatment (Tables 3 and 4). Although the absolute level of ger-
mination was higher under 12 hr photoperiod (Tables 3 and 4) than under periodic
light exposure every week (Tables | and 2), the relative effect of charate was the
same under both conditions. Table 5 shows that the light intensity may be quite
low (~20 wE m~?’ s“') and still produce a stimulatory effect on germination.

Discussion and Conclusions

This study does not support the theory that native seeds are inhibited from
germinating under the mature chaparral canopy by chemical inhibitors. Indeed
so-called “‘allelopathic”’ leachate from Adenostoma foliage stimulated germination
of seeds from the mature chaparral soil. This stimulatory effect has in fact been
observed for a variety of species (Keeley et al. 1985) where it has been attributed
to the known stimulatory effect of NO;~ which is abundant in the leachate.
Circumstantial evidence of microbial inhibition was observed in the present study
but only on burn soil.

Both heat and charred wood produce a highly stimulatory effect on germination
of seeds from mature chaparral soil. However, charate stimulated the germination
of Phacelia cicutaria and Salvia columbariae but not Cryptantha muricata or
Lotus salsuginosus. E

In comparing the mature chaparral soil and the second year burn soil it is clear
that 1) the seed pools are quite different, and 2) the soils affect germination of
the same species quite differently. Annual grass seeds were uncommon in the
mature chaparral soil whereas they were abundant in the second year burn soil.
This suggests that these species invade recently burned areas but their residence
time in the soil is low in the absence of fire. Although ignored in previous studies
it appears that light plays a major role in the germination of certain chaparral
seeds. Certainly the evidence presented here should be sufficient warning that
future studies need to examine the effect of light on chaparral seed germination.

Highest germination from mature chaparral soil was with charate under a 12
hr photoperiod. Estimates based on the range of depths soil was sampled suggest
this treatment produced ca. 1500 seedlings/m? + 500. Such numbers are clearly
large enough to produce seedling densities typical of burned chaparral stands.

In conclusion it appears from these data that the bulk of the dormant seed pool
in mature chaparral requires the stimulating effect of heat, charred wood, and
light for germination. Seeds in soil from a 2nd year burned site (still largely devoid
of shrub cover) germinated readily without heat or charred wood although a large
portion of this seed pool required light for germination.

Literature Cited

Christensen, N. L., and C. H. Muller. 1975. Effect of fire on factors controlling plant growth in
Adenostoma chaparral. Ecol. Monogr., 45:29-55.

Hadley, E. B. 1961. Influence of temperature and other factors on Ceanothus megacarpus seed
germination. Madrono, 16:132-128.

Hanes, T. L. 1977. California chaparral. Pp. 417-470 in Terrestrial vegetation of California. (M. G.
Barbour and J. Major, eds.), John Wiley and Sons, New York, 1002 p.

Jones, C. S., and W. H. Schlesinger. 1980. Emmenanthe penduliflora (Hydrophyllaceae): Further
consideration of germination response. Madrono, 27:122-125.

120 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

Kaminsky, R. 1981. The microbial origin of the allelopathic potential of Adenostoma fasciculatum
H & A. Ecol. Monogr., 51:365-382.
Keeley, J. E. 1982. Distribution of lightning and man-caused wildfires in California. Pp. 431-437
in Proceedings of the symposium on dynamics and management of mediterranean type eco-
systems. (C. E. Conrad and W. C. Oechel, eds.), USDA Forest Service, General Technical
Report PSW-58. Berkeley. 637 p.
, B. A. Morton, A. Pedrosa, and P. Trotter, 1985. The role of allelopathy, heat, and charred
wood in the germination of chaparral herbs and suffrutescents. J. Ecol. Jn press.
Keeley, S. C., J. E. Keeley, S. M. Hutchinson, and A. W. Johnson. 1981. Postfire succession of the
herbaceous flora in California in southern California chaparral. Ecology, 62:1608-1621.
McPherson, J. K., and C. H. Muller. 1969. Allelopathic effects of Adenostoma fasciculatum, ‘“‘cha-
mise,” in the California chaparral. Ecol. Monogr., 39:177-198.

Munz, P. A. 1974. A flora of southern California. Univ. Calif. Press, Berkeley. 1086 p.

Pack, P. 1985. The role of microbial toxins and charate in the germination of chaparral herb seeds.
M.A. Thesis, Occidental College, Los Angeles.

Quick, C. R., and A. S. Quick. 1961. Germination of Ceanothus seeds. Madrono, 16:23-30.

Stone, E. C., and G. Juhren. 1951. The effect of fire on the germination of the seed of Rhus ovata
Wats. Amer. J. Bot., 38:368-372.

Sweeney, J.R. 1956. Responses of vegetation to fire. A study of the herbaceous vegetation following
chaparral fires. Univ. Calif. Publ. Bot., 28:143-216.

Wicklow, D. T. 1977. Germination response in Emmenanthe penduliflora (Hydrophyllaceae). Ecol-
ogy, 58:201-205.

Accepted for publication 6 April 1984.

Department of Biology, Occidental College, Los Angeles, California 90041.

Appendix I. Soil characteristics for media used in this study compared to commercial (Gro-Lite)
potting soil. The mature soil pH in parenthesis is for the soil alone, i.e., without the fine litter (mostly
Adenostoma leaves).

Potting soil Mature chaparral soil 2nd yr burn soil

pH 6.7 5.6 (6.6) 133

Na (ppm) 230 1S) 130

Mg (ppm) 250 21 20

Ca (ppm) 300 200 210

Fe (ppm) 0.6 1.4 1.6

Zn (ppm) 0.2 0.4 0.4

PO, (mg/l) Wes) 0.04 0.60

SO, (mg/l) 96.0 0.00 0.00
NO, (mg/1) 7.9 eS 0.7

CO, (mg/1) 0.1 0.1 0.6

Bull. Southern California Acad. Sci.
83(3), 1984, pp. 121-132
© Southern California Academy of Sciences, 1984

Seasonal Abundance of Pinnipeds at San Nicolas Island,
California, 1980-1982

Brent S. Stewart and Pamela K. Yochem

Abstract. —Seasonal abundance of pinnipeds at San Nicolas Island, California,
1980-1982 by Brent S. Stewart and Pamela K. Yochem. Bull. Southern California
Acad. Sci., 83(3):121-132, 1984. Seasonal cycles in abundance of northern ele-
phant seals, California sea lions, harbor seals, and Guadalupe fur seals at San
Nicolas Island, California, were monitored by frequent ground and aerial surveys
from February 1980 through September 1982. Northern elephant seals were most
abundant in late January and early February during the height of their breeding
season and again in late April to early May when juveniles and adult females
hauled out to molt. Sea lions were most abundant in late June to early July during
the height of their breeding season and were least abundant in winter and early
spring. Harbor seals were in greatest abundance in late May to early June when
they were molting and numbers were lowest in winter. Guadalupe fur seals were
present from June through September during their breeding season. Seasonal
populations and pup production of elephant seals, sea lions, and -harbor seals
increased each year.

Three species of pinnipeds, the northern elephant seal (Mirounga angustirostris),
California sea lion (Zalophus californianus) and Pacific harbor seal (Phoca vitulina
richardsi, Shaughnessy and Fay 1977), breed at San Nicolas Island (SNI), Cali-
fornia (33°15'N, 119°30’W). Guadalupe fur seals (Arctocephalus townsendi) may
have been abundant in the Southern California Bight (SCB) prior to commercial
exploitation by sealers in the 19th century (Scammon 1874; Lyon 1937; Walker
and Craig 1979), but it is not known if they ever bred at SNI. Steller sea lions
(Eumetopias jubatus) have been sighted occasionally (Bartholomew 1951; Bar-
tholomew and Boolootian 1960), but they have apparently never bred at SNI.
The earliest published data for the SNI sea lion population are from a 1946 census
(Ripley et al. 1962) and it has been censused occasionally since then (Bonnot and
Ripley 1948; Bartholomew 1951; Bartholomew and Boolootian 1960; Peterson
and Bartholomew 1967; Frey and Aplin 1970; Carlisle and Aplin 1971; Odell
1971, 1975a, 1975b; Bonnell et al. 1980). The harbor seal population at SNI has
rarely been censused (Bartholomew and Boolootian 1960; Odell 1971; Bonnell et
al. 1980). Elephant seals have been censused sporadically since reestablishment
of a breeding colony in the early 1950’s (Bartholomew 1951; Bartholomew and
Boolootian 1960; Odell 1971, 1974, 1977: Bonnell et al. 1980; Antonelis et al.
1981: Cooper and Stewart 1983). The sizes and distributions of populations of
each species have changed substantially during the last several years. We report
here on the seasonal abundance and distribution of northern elephant seals, Cal-
ifornia sea lions, harbor seals, and Guadalupe fur seals at SNI from February
1980 through September 1982.

122 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

Methods

Forty-three ground and three aerial surveys were made of pinniped breeding
and hauling areas at SNI from 3 February 1980 through 1 October 1982. In 1980,
we identified census areas by placing permanent wooden markers at the boundaries
of each area. These areas and the numerical designations used here (Fig. 1) are
in most cases identical to those used by Bonnell et al. (1980) and generally com-
parable to those that have been used by other authors in the past (Bartholomew
1951; Peterson and Bartholomew 1967; Odell 1974, 1975b, 1977). Ground counts
were made on foot using Bushnell 6-18 <x 35 mm zoom binoculars and a Bushnell
20-45 mm zoom spotting scope; seals were classified by species, sex, and age.
Aerial surveys were made in a Cessna 337 Skymaster; hauling and breeding areas
were photographed obliquely with a Nikon 35 mm camera equipped with a 43-
86 mm zoom or 200 mm telephoto lens using Ektachrome (400 ASA) and Fu-
jicolor 400 film. Mounted slides were projected and numbers of animals of each
species were counted and classified by age and sex.

Age and sex classification of seals differed for each species. This was necessary
because of differences among species in the ease and reliability of consistently
identifying large numbers of individuals, often in very dense aggregations, as
belonging to specific age and sex categories.

Methods of assessment of pup production (an index of natality; usually the
number of pups surviving to some age) and pup mortality also varied among
species. We define pup production as the number of live pups that are counted
on a particular date; determination of this date is based on knowledge of the
breeding biology of each species. We determined pup production for each species
by censusing all breeding areas in late February (elephant seals), early April
(harbor seals) or early July (California sea lions).

We define pup mortality generally as the number of pups that were observed
dead during a specific time period divided by an estimate of the number of pups
born during that same time period. Pup mortality was estimated from early
December through early March for elephant seals, from late February through
late April for harbor seals, and from late May through early July for California
sea lions. These estimates are therefore of pre-weaning pup mortality for elephant
seals and harbor seals and of whelping season pup mortality for sea lions.

In 1982 we counted essentially all elephant seal pups that died prior to weaning
by conducting daily censuses of some rookeries and weekly censuses of all rook-
eries. In 1980 and 1981 we conducted bi-weekly or monthly censuses of all
rookeries. We tied red surveyor’s tape around the ankles and necks of most dead
pups after we measured and sexed them. To avoid disturbance to females and
nursing pups we did not examine a few pups that died in the middle of groups
of females. However, we noted the location and state of decomposition of each
unmarked carcass on grid maps during each census. Marking and mapping the
locations of dead pups prevented double-counting of carcasses.

We recorded all dead harbor seal pups observed from late February through
late April. These counts are, necessarily, minimal estimates of mortality for several
reasons. Harbor seals can swim at birth (many births are apparently aquatic) and
pups move daily between terrestrial hauling areas (beaches, intertidal reefs and
ledges) and the sea with their mothers. Pups that die (or are still-born) while ashore

PINNIPEDS AT SAN NICOLAS ISLAND 123

SAN NICOLAS ISLAND

|

| 20 a2 neg I |

N (4A) capycacy 2 262
(6A)

266

Nauticol Miles

Fig. 1. San Nicolas Island showing census areas and numerical codes.

may therefore not be detected unless the carcasses initially appear well above the
high tide line. Pups that die at sea are also unlikely to be detected. Therefore,
without specific mark-recapture studies of pups or of parous females, the rela-
tionship (and its variability) of observed dead pups to estimates of pup production
is unknown.

We estimated sea lion pup mortality by recording the number of new pup
carcasses observed on each biweekly census during the whelping season (25 May
through 4 July each year). Dead pups found before 25 May were considered to
be premature and were not considered as part of the whelping season mortality.

Because most dead sea lion pups could not be marked without major disturbance
to rookeries, we simply tallied the number of fresh pup carcasses from observation
points above rookeries during biweekly censuses between 25 May and 4 July each
year. Sea lion pups are smaller than elephant seal pups and faster rates of decay
(both passive and by gull scavenging) and burial in the sand prevented detailed
monitoring of the fate of sea lion pup carcasses during the biweekly censuses. We
summed counts of dead pups from all censuses during the whelping season to
determine a minimal estimate of whelping season mortality. This estimate was
derived by dividing the cumulative total of dead pups by an estimate of natality
(pup production plus dead pup total).

During the course of other studies, we placed numbered plastic tags (Dalton
Jumbo Roto) in the interdigit webbing of the rear flippers of harbor seals and
elephant seals at San Nicolas and San Miguel Islands. During the breeding season
we also marked adult and subadult male, adult female and newborn elephant seals
with bleach and dye, which permitted identification of individuals for several
months. Resightings of these tagged and marked animals provided data on sea-
sonal age and sex composition of the populations, which we summarize below.

124 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

Results

Northern Elephant Seal.— Adult males began arriving at SNI in late November
to early December. The first pregnant females arrived in early to mid-December
and some arrived as late as mid-February. The breeding population increased
rapidly in January and peak numbers of seals were ashore the last week in January,
by which time approximately 85% to 90% of the season’s pups had been born
(Stewart 1983). Females gave birth approximately 7 days (X= 6.8, standard
error = .05) after coming ashore and remained ashore suckling their pups an
average of 27 (standard error = .5) additional days before returning to sea. Some
females weaned their pups and departed the rookery as early as 2 January, but
mosi did so during the first two weeks of February (Stewart 1983). All females
that gave birth departed the rookery by the first week of March. Females remained
at sea for an average of 75 days (standard error = 3.5) before returning to molt,
although females that lost pups during the breeding season spent less time at sea
(x = 56.3 days, standard error = 5.21 days) than did females that successfully
weaned pups (X = 75.5 days, standard error = 2.7 days, t = 3.3 P < .01). Using
15 February as a mean date of departure for females, the adult female molt peak
would be expected to occur the first week of May. Our observations of peak
numbers of females ashore in late April to early May support this. Some molting
females arrived in early March and most had left the island by early June.

Pups were weaned about 27 days (standard error = .5) after birth but remained
on the rookery for an additional four to six weeks before going to sea. Weaned
pups began leaving the rookery in early March although most departed between
mid-March and late April (Table 1). A few were present through late May. A few
pups-of-the-year hauled out briefly in September and October and although most
had departed by December, a few (now yearlings) remained through January.

The adult and subadult male populations remained fairly constant from early
January through late February, although there was some movement of both sub-
adult and adult males between beaches in January. Males began leaving in late
February; few were present in mid-March and they were absent from April through
early June (Table 1). Subadult males returned to molt in early May and numbers
ashore peaked in late May to early June. Numbers of molting adult males increased
from June to July peaking in mid-July. Most departed by late August and none
were ashore from September through mid-November. Juvenile males and females
were absent from the island during the breeding season but numbers ashore
increased in late March and early April and peaked in late April when they returned
to molt. Most juveniles had returned to sea by late May although a few were
present in summer and fali months. Molting seals hauled out almost exclusively
at areas 221, 231, 232, and 233 from March through August of each year.

Approximately 1188, 1546, and 2312 pups were born at SNI in 1980, 1981,
and 1982 respectively. Annual pup mortality on the island remained fairly con-
stant at 3% to 4%, although beach topography affected pup mortality at specific
sites. Highest mortality occurred on narrow, steep-backed beaches when high tides
(primarily at peak season in late January) combined with storm-generated surf to
wash those beaches, drowning some pups and separating others from their moth-
ers. Numbers of pups born in each area increased during the three-year study
(Table 2). Colonization of new beaches on the southern coastline also continued

PINNIPEDS AT SAN NICOLAS ISLAND 125

Table 1. Censuses of northern elephant seals (Mirounga angustirostris) at San Nicolas Island, 1980-
1982.

Sub-adult
Date Adult males males Females Juveniles Live pups Yearlings Totals
1980
3 Feb 97 30 1109 D 1075 0 2313
9 Mar 46 2 25 7 884 0 964
21 Apr 0 2 275 965 279 0 1521
17 May 0 54 637 1305 154 0 2150
8 Jun 51 273 146 162 0 0 632
12 Dec V3 32 39 13 1 93 251
27 Dec 106 70 166 39 54 112 547
1981
10 Jan 106 38 750 1 379 24 1298
23 Jan 130 39 1430 0 1176 4 2779
6 Feb 151 41 1151 0 1401 4 2748
19 Feb 170 24 237 0 1399 0 1831
21 Mar 21 1 19 322 952 0 1315
3 Apr 1 0 198 1944 911 0 3054
*18 Apr 0 12 428 2618 603 0 3661
25 Apr 0 11 692 2310 460 0 3473
8 May 0 149 1458 2004 118 0 3729
*16 May 0 357 1512 1101 0 0 2970
23 May 0 590 1209 504 0 0- 2303
12 Jun 0 338 212 18 0 0 568
28 Jun 70 208 0 21 0 0 299
5 Jul 69 188 2D; 7 0 0 266
11 Jul 96 142 24 12 0 0 274
24 Jul 87 112 1 18 0 0 218
7 Aug 81 37 2 24 g** 0 153
22 Aug DD 20 0 82 13a 0 137
31 Oct 0 8 6 371 218** 0 603
1982
2 Jan 147 112 468 90 197 10 1024
9 Jan 129 HONE, 849 69 459 8 1631
15 Jan 127 124 1567 23 836 1 2678
22 Jan 151 136 2071 16 1455 0 3829
29 Jan 142 163 2168 19 2064 0 4556
6 Feb 154 100 1747 1 2196 0 4198
12 Feb 126 137 1225 0 2224 0 3712
19 Feb 126 138 506 5 2218 0 2993
26 Feb 106 132 115 7 2238 0 2600
9 Mar 71 50 25 51 1817 0 2014
13 Mar 20 39 68 89 1981 0 2197
10 Apr 0 5 453 1410 1194 0 3062
21 Apr 0 27 473 1943 747 0 3190
11 May 0 130 867 1305 207 0 2509
18 May 0 252 1462 525 36 0 2275
29 May 0 446 639 174 1 0 1259
12 June 9 284 150 253 0 0 696
4 July 131 82 18 78 0 0 309
1 Oct 0 0 184 48 D3 eee 0 505

* Aerial survey; all others are ground counts.
** Pups-of-the-year.

126 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

Table 2. Natality and pup mortality of northern elephant seals (Mirounga angustirostris) at San
Nicolas Island, 1980-1982.

Pups born Dead pups % Mortality

Site 1980 1981 1982 1980 1981 1982 1980 1981 1982
266 0 14 Dil 0 3 2 — 21% 7.4%
265 0 61 175 0 1 8 — 1.6% 4.6%
264 29 30 98 0 1 2 0) 3.3% 2.0%
261, 262, 263 258 298 361 15 12 7 5.8% 4.0% 1.9%
243 62 30 153 5 0) 5 8.1% 0 3.3%

242 39 37 53 1 0) 0 2.6% 0) 0
234, 241 195 214 232 4 6 7 2.1% 2.8% 3.0%
233 234 406 478 2 10 17 9% 2.5% 3.6%
232 69 98 DAG, 3 13 5 4.3% 13.3% 2.3%
231 58 87 149 1 2 6 1.7% 2.3% 4.0%
223 96 107 121. 9 2 8 9.3% 1.9% 6.6%
DPA 147 164 246 4 2 7 2.7% 1.2% 2.8%

201 0 0 2 0 0 0 — — 0
Total 1187 1546 2312 44 52 74 3.7% 3.4% 3.2%

and pups were born on northern shores for the first time in 1982 (Table 2,
Fig. “f).

Harbor Seal.—Harbor seals hauled out and gave birth at seven sites and spo-
radically used 13 others seasonally (Stewart and Yochem 1983). Areas 231 and
266 were the most consistently used hauling areas year-round, although significant
numbers were found at area 270 during the whelping and molt seasons. Hauling
patterns of harbor seals at most sites on SNI appear to be similar to those of
harbor seals at San Miguel Island (Stewart 198la, 1984; Stewart and Yochem
1983), with peak numbers hauled out in early afternoon.

Few harbor seals were ashore from September through February. Numbers
began to increase in late February when females began giving birth. Pups were
born from late February through late April and pups were most abundant in early
April. In 1982 one premature pup, still in lanugo, was observed on 9 February.
Most pups were weaned by mid-May. At least 26 pups were born in 1980 and at
least 65 and 85 pups were born in 1981 and 1982, respectively.

The number of seals ashore increased through spring; maximum numbers were
present in late May to early June (Table 3). This peak in abundance was related
to the annual molt; some seals began shedding in early May and most had com-
pleted the molt by late July. Numbers ashore declined steadily from late June
through December (Table 3). Live pups represented approximately 14% of the
whelping season population in 1980, 28% in 1981, and 24% in 1982; but 10%,
11%, and 12% of peak (late May) populations in 1980, 1981, and 1982 respec-
tively.

Based on the maximum number of pups observed and the total number of pup
carcasses encountered, minimal pup mortality estimates were 7% in 1980, 12%
in 1981, and 7% in 1982. Of eight dead pups found in 1981, six were found at
sites where disturbance by humans is frequent. Disturbance at beaches when young
harbor seal pups are present often causes separation of mothers and pups and

PINNIPEDS AT SAN NICOLAS ISLAND 127

Table 3. Censuses of harbor seals (Phoca vitulina richardsi) at San Nicolas Island, California, 1980-
1982.

Date Pups Total Date Pups Total

1980 1982
3 Feb 62 2 Jan 81
9 Mar 3 45 15 Jan 46
21 Apr 26 207 22 Jan 80
4 May 5 226 29 Jan 37
17 May 9 257 6 Feb 1 131
8 Jun 312 12 Feb 76
6 Jul 283 19 Feb 41
12 Dec 17 26 Feb 44
27 Dec 29 9 Mar 9 82
13 Mar 38 346
1981 3 Apr 84 304
10 Jan 7 21 Apr 48 229
23 Jan PA 11 May 39 605
6 Feb 39 18 May 26 609
19 Feb 122 29 May 4 566
21 Mar 47 184 2 June 13 694
3 Apr 63 227 12 June 7 661
*18 Apr 20 242 4 July 4 399
25 Apr 65 296
8 May 38 371
*16 May 59 566
23 May 40 530
12 Jun 13 452
*29 Jun 31 444
5 Jul 25 350
11 Jul 62 333
24 Jul 28 177
7 Aug 3 121
22 Aug 3 192
10 Oct 194

* Aerial survey; all others are ground counts.

subsequent starvation and death of pups (Pitcher and Calkins 1979; Johnson
1977; Stewart and Yochem, pers. obs.).

California Sea Lion.—The California sea lion population peaked in early July
coincident with the height of the breeding season. The July peak increased 14%
from 1980 to 1981, and 24% from 1981 to 1982 (Table 4).

Births occurred from late May through early July. The greatest number of births
occurred in mid-June. Premature pups (6 in 1980, 7 in 1981, 16 in 1982) were
observed from January through mid-May. Pup production was approximately
6096 in 1980, 6704 in 1981, and 7738 in 1982 (Table 5). During the whelping
season 186 dead pups were observed in 1980, 113 dead pups were observed in
1981, and 123 dead pups were observed in 1982. Minimal mortality during the
whelping season at specific breeding sites ranged from 1.4% to 20% in 1980, 2.1%
to 12.5% in 1981, and 1.0% to 4.9% in 1982. Absolute pup mortality is presumably
somewhat greater but its measurement is complex. Detection of dead sea lion
pups is influenced by weather and tide conditions, beach topography, variable

128 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

Table 4. Censuses of California sea lions (Zalophus californianus) at San Nicolas Island, California,
1980-1982.

Subadult males/

Date Males females/juveniles Live pups Total
1980
3 Feb 186 4649 0 4835
9 Mar 195 5822 0 6017
19 Apr 149 6260 4 6413
17 May 140 4096 34 4271
8 Jun 360 5205 1276 6841
6 Jul 284 7927 6096 14,307
12 Dec 232 3545 _ 3777
27 Dec 60 3615 _ 3675
1981
10 Jan 62 : 3576 _ 3638
23 Jan 75 3404 — 3479
6 Feb 74 3721 — 3795
19 Feb 148 4650 — 4798
21 Mar 119 5029 — 5148
3 Apr 201 6460 D, 6663
*18 Apr 248 7703 0) 7951
25 Apr 216 9244 5 9465
8 May 187 5572 0 5759
*16 May 153 5224 0 5377
22 May 157 5015 121 5293
12 Jun 925 10,204 3336 14,465
5 Jul 637 8668 6704 16,006
11 Jul 1437 8290 6626 16,353
24 Jul 458 7541 6676 14,654
7 Aug 33 4275 6370 10,678
22 Aug 19 5246 4851 10,116
Subadult Females/’
males juveniles
1982
2 Jan 25 100 3990 — 4115
9 Jan 29 182 3573 _ 3784
15 Jan 43 209 6728 _ 6980
22 Jan 43 285 5033 _ 5361.
29 Jan 62 196 4839 — 5097
6 Feb 69 276 3315 — 3660
12 Feb 122 342 5473 - 5937
19 Feb 131 280 5655 — 6066
26 Feb 105 577 6696 — 7378
9 Mar 70 431 6365 — 6966
13 Mar 43 407 6757 _ 7207
10 Apr 35 245 6610 — 6890
11 May 43 231 4124 _ 4424
18 May 42 205 3839 — 4109
29 May 90 194 5793 274 6381
12 June 261 10,845 3396 14,502
4 July 519 12,035 7738 20,292
1 Oct 0 43 3296 3790 7129

* Aerial survey; all others are ground counts.

PINNIPEDS AT SAN NICOLAS ISLAND 129

Table 5. Pup production of California sea lions at specific sites at San Nicolas Island, 1980-1982.

Pup production

Site 1980 1981 1982
266 0 0 0
264 0 0 0
263 0 0 0
262 483 527 474
261 281 293 355
243 859 864 1030
242 501 560 786
241 795 867 358
234 359 368 366
233 1078 1140 1422
232 469 503 1207
231 351 387 497
223 465 533 806
221 72 101 53
213 12 22 25
212 182 254 164
211 189 285 195
Total 6096 6704 7738

rates of carcass decomposition and disappearance, and frequency of censuses
(Heath and Francis 1983; Stewart and Yochem, unpubl. data).

The number of adult males ashore fluctuated at low levels throughout the winter.
Breeding males first arrived and established territories in mid-May. The male
population then increased rapidly in June and peaked in early July during the
height of the breeding season. The adult male population declined rapidly in late
July as territories disintegrated and males returned to sea; few were present in
August and September (Table 4). Subadult males increased in number briefly in
February and March and then declined before increasing again in late May.

Adult females and juveniles of both sexes were included in a single census
category (‘others’); their numbers were lowest in January and February, peaked
briefly in April and then declined through late May. Numbers then increased
sharply to a peak in mid-late June, corresponding to a peak in whelping activity.

The number of sea lions declined from October through December, during
which time the population was composed primarily of pups-of-the-year, attending
females and, to a lesser extent, juveniles. Most areas used by breeding sea lions
were also used as hauling areas during the non-breeding season. The east end of
221 was used exclusively by bachelor and subadult males during the breeding
season. Although a few pups were born at the west end of 221, this area was not
used by sea lions during the non-breeding season. Relatively few sea lions used
the rocky shoreline in areas 211, 212, and 213 for hauling during the non-breeding
season. During the breeding season, most juveniles and non-breeding females
aggregated on beaches in areas 231, 241, and 262. Large numbers of pups were,
however, born on reefs and bluffs at the west ends of areas 262 and 233. Area
241 was used considerably less for pupping in 1982 than in the previous two
years. The reasons for this are unclear but repeated human disturbance of hauled

130 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

seals (resulting in relocation of some seals to other sites) may be among them.
Although pup production increased 25% from 1980 to 1982 the relative proportion
of pups born in each area remained fairly constant. In 1982 approximately 34%
of all pups were born within a 2 to %4 km stretch of coastline in areas 232 and
233. Areas 223, 242, and 243 were other major whelping areas (Table 5).

Guadalupe Fur Seal.— An adult male Guadalupe fur seal maintained a territory
in area 242 from about 26 June through 2 August 1981; a juvenile fur seal was
seen in the same area on 4 July 1981 but was not seen again (Stewart 1981b).

In 1982 the same adult male (identified by scars) maintained a territory in the
same location from at least 26 May through | October. Lone immature seals were
observed at other sites (232, 223) in July 1982 (P. Yochem, unpubl. data; J.
Francis and C. Heath, pers. comm.). We also saw an immature fur seal within
100 meters of the adult male on 1 October 1982.

Discussion

Elephant seals were first reported at San Nicolas Island in 1949 (Bartholomew
1951) and although published records of pups do not appear until 1959 (Bartho-
lomew and Boolootian 1960) pups may have been born there as early as 1951
(Odell 1974). The initial breeding population was apparently restricted to areas
232 and 233. Pup production increased from 1959 through 1982 at an average
rate of about 16% per year (Cooper and Stewart 1983). Coincident with the
population growth has been continued expansion of breeding to new areas along
the south side of SNI; in 1981 pups were born from area 221 eastward into the
western portion of 266. In 1982 pups were born even farther east in area 266 and
two pups were born on the island’s north side in area 201. We assume that food
availability is not presently a limiting factor to population growth and we expect
the SNI population to increase until beaches become saturated and growth is
stabilized by density dependent influences on female breeding success and pup
survival. Adequate beach space is available in areas 266 to 268; increased use of
some northern beaches may also occur depending on the effects of weather on
breeding success there.

Few historical data are available on the harbor seal population at SNI. Bar-
tholomew (1951) reported 141 and 177 seals on 18-20 June 1949 and 9-11 July
1949, respectively. Bartholomew and Boolootian (1960) reported 21 harbor seals
on 16 April 1958 and 34 on 9 June 1958. Bartholomew (1967) believed .the
population of harbor seals at the Southern California Channel Islands in 1965 to
be about 500. Odell (1971) estimated the minimum Channel Islands population
at 645 in June 1964; on 20 June 1964, 95 harbor seals were counted at SNI.
Bonnell et al. (1980) surveyed the Southern California Channel Islands from 1975
to 1978; the greatest number of harbor seals observed at SNI was 311 in late May
1977. The largest number of pups observed by these authors was 29 in March of
1977. The peak 1982 population count reported here (694) is the largest yet
reported for San Nicolas, suggesting the population is increasing. The consistent
increase in number of pups observed from 1980 through 1982 also suggests that
the population is increasing.

The first data on sea lion pup production at SNI were presented by Bartholomew
(1951); all 895 pups seen in 1949 were in areas 211 to 213. Subsequent pup counts
reported by Peterson and Bartholomew (1967) and Bartholomew (1967) indicated

PINNIPEDS AT SAN NICOLAS ISLAND 131

that approximately 62% (2244) of the pups born in 1965 were found in areas
211-213. Since then there has been a drastic decline in use of these areas for
whelping, presumably due to repeated human disturbance. Bonnell et al. (1980)
reported that 55 and 73 pups were born in areas 211 to 213 in 1975 and 1976
respectively; no pups were observed there in 1977. The Point Mugu Naval Air
Station and SNI Commands subsequently closed these areas to non-authorized
personnel during the whelping and breeding season. By 1980 sea lions had re-
colonized these areas and at least 400 pups were born there that year. A continued
increase in pup production is expected if human disturbance is minimized.

Although data on the Guadalupe fur seal population at Isla de Guadalupe,
Mexico, are few, they indicate a slow recovery from nineteenth century overhar-
vesting (Peterson et al. 1968; Fleischer 1978). Increased frequency of sightings of
young and mature animals on the Southern California Channel Islands (Bartho-
lomew 1950; Antonelis and Fiscus 1980; Stewart 1981b and this report; DeLong
1982) in recent years seem indicative of a growing population on the species’ sole
rookery at Isla de Guadalupe. The health and trends in growth of that population
will likely be the major factors influencing successful (re)establishment of a breed-
ing colony on the Southern California Channel Islands in the near future.

Acknowledgments

J. Jehl, W. Evans, J. Thomas, D. DeMaster, D. Odell, C. Heath, J. Francis, M.
Bonnell, and two anonymous reviewers commented on the manuscript. We thank
LCDR E. Giffen and Mr. R. Dow and his staff for arranging access to San Nicolas
Island and for their interest in our research. We also thank Clairol Research
Laboratories, Stamford, Connecticut, for providing bleach and dye products. This
work was conducted as part of a study to examine the potential effects of Space
Shuttle sonic booms on the vertebrate fauna of the Southern California Channel
Islands under USAF contract No. FO4701-81-C-0018. The research was author-
ized under the Marine Mammal Protection Act of 1972 by Marine Mammal
Permits No. 341 and No. 367.

Literature Cited

Antonelis, G. A. Jr., and C. H. Fiscus. 1980. The pinnipeds of the California Current. CalCOFI
Rep., 21:68-78.

——, S. Leatherwood, and D. K. Odell. 1981. Population growth and censuses of the northern
elephant seal, Mirounga angustirostris, on the California Channel Islands, 1958-78. Fish. Bull.,
79:562-567.

Bartholomew, G. A. 1950. A male Guadalupe fur seal on San Nicolas Island. J. Mamm., 31:175-

80.

1951. Spring, summer and fall censuses of the pinnipeds on San Nicolas Island, California.

J. Mamm., 32:15-21.

1967. Seal and sea lion populations of the California Islands. Pp. 229-244 in Proc. Symp.

Biol. Calif. Islands (R. Philbrick, Ed.), Santa Barbara Botanic Gardens. 363 pp.

, and R. A. Boolootian. 1960. Numbers and population structure of the pinnipeds on the

California Channel Islands. J. Mamm., 41:366-375.

Bonnell, M. L., B. J. LeBoeuf, M. O. Pierson, D. H. Dettman, G. D. Farrens, C. B. Heath, R. F.
Gantt, and D. J. Larsen. 1980. Pinnipeds of the Southern California Bight. Jn Marine mammal
and seabird surveys of the Southern California Bight area 1975—1978. Vol. 3—Investigator’s
Reports, Part 1. NTIS PB81-248-71. 535 pp.

Bonnot, D., and W. E. Ripley. 1948. The California sea lion census for 1947. Calif. Fish and Game,
34:89-92.

132 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

Carlisle, J. G., and J. A. Aplin. 1971. Sea lion census for 1970, including counts of other California
pinnipeds. Calif. Fish and Game, 57:124—-126.
Cooper, C. F., and B. S. Stewart. 1983. Demography of northern elephant seals, 1911-1982. Science,
219:969-971.
DeLong, R. L. 1982. Population biology of northern fur seals at San Miguel Island, California. Ph.D.
Dissertation. Univ. of California, Berkeley, 203 pp.
Fleischer, L.A. 1978. Distribution, abundance, and population characteristics of the Guadalupe fur
seal, Arctocephalus townsendi. MS Thesis. Univ. Washington. 64 pp.
Frey, H. W., and J. A. Aplin. 1970. Sea lion census for 1969, including counts of other California
pinnipeds. Calif. Fish and Game, 56:130-133.
Heath, C. B., and J. M. Francis. 1983. Population dynamics and feeding ecology of the California
sea lion with applications for management. SWFC Admin. Rept. No. LJ-83-04C. 47 pp.
Johnson, B. W. 1977. The effects of human disturbance on a population of harbor seals. Appendix
I, in Environmental assessment of the Alaskan Continental Shelf. Annual Reports of Principal
Investigators, Volume I.
Lyon, G. M. 1937. Pinnipeds and a sea otter from the Point Mugu shell mound of California. Univ.
Calif. Los Angeles, Publ. Biol. Sci., 1:133-168.
Odell, 1971. Censuses of pinnipeds breeding on the California Channel Islands. J. Mamm., 52:187-
190.
—. 1974. Seasonal occurrence of the northern elephant seal, Mirounga angustirostris, on San
Nicolas Island, California. J. Mamm., 55:81-95.
1975a. Abundance of sea lions on San Nicolas Island, California. J. Wild. Management,
39:729-736.
. 1975b. Breeding biology of the California sea lion, Zalophus californianus. Rapp. P-V. Reun.
Cons. Int. Explor. Mer., 169:374-378.
1977. Structure of northern elephant seal population breeding on San Nicolas Island, Cal-
ifornia, in 1971. Anim. Behav., 25:208-214.
Peterson, R. S., and G. A. Bartholomew. 1967. The natural history and behavior of the California
sea lion. Amer. Soc. Mamm. Spec. Publ., 1:1-79.
, C. L. Hubbs, R. L. Gentry, and R. L. DeLong. 1968. The Guadalupe fur seal; habitat,
behavior, population size and field identification. J. Mamm., 49:665-675.
Pitcher, K. W., and D. G. Calkins. 1979. Biology of the harbor seal, Phoca vitulina richardsi, in the
Gulf of Alaska. Final Report, Contract 03-5-002-69, OCSEAP, BLM. 72 pp.
Ripley, W. E., K. W. Cox, and J. L. Baxter. 1962. California sea lion census for 1958, 1960 and
1961. Calif. Fish and Game, 48:228-231.
Scammon, C. M. 1874. The marine mammals of the northwestern coast of North America. John
Carmany and Co., San Francisco, Calif. 319 p.
Shaughnessy, P. D., and F. H. Fay. 1977. A review of the taxonomy and nomenclature of North
Pacific harbor seals. J. Zool., Lond., 182:385—419.
Stewart, B.S. 1981la. Seasonal abundance, distribution and ecology of the harbor seal (Phoca vitulina
richardsi) on San Miguel Island, California. MS Thesis. San Diego State University. 66 pp.
—. 1981b. The Guadalupe fur seal on San Nicolas Island, California. Bull. S. Calif. Acad.,Sci.,
80:100-102.
. 1983. Studies on pinnipeds on the Southern California Channel Islands, 1982-1983. HSWRI
Tech. Rept. 83-154. 125 pp.
1984. Diurnal hauling patterns of harbor seals at San Miguel Island, California. J. Wildl.
Management, 48(3).
,and P. K. Yochem. 1983. Hauling patterns, movements, and site fidelity of harbor seals on
San Nicolas and San Miguel Islands, California. HSWRI Tech. Rept., 82-152. 25 pp.
Walker, P. L., and S. Craig. 1979. Archaeological evidence concerning the prehistoric occurrence of
sea Mammals at Point Bennett, San Miguel Island. Calif. Fish and Game, 65:50-54.

Accepted for publication 3 April 1984.

Hubbs-Sea World Research Institute, 1700 South Shores Road, San Diego, Cal-
ifornia 92109.

Bull. Southern California Acad. Sci.
83(3), 1984, pp. 133-147
© Southern California Academy of Sciences, 1984

The Use of Jackknife Confidence Intervals with the Richards
Curve for Describing Avian Growth Patterns

David W. Bradley,' Ross E. Landry,” and Charles T. Collins?

Abstract.—The use of jackknife confidence intervals with the Richards Curve
for describing avian growth patterns by David W. Bradley, Ross E. Landry, and
Charles T. Collins. Bull. Southern California Acad. Sci., 83(3):133-147, 1984. A
method is presented for measuring growth patterns in birds in which the Richards
curve is used as a smoothing device for extracting convenient summary statistics
from raw growth data. Jackknife confidence intervals for these statistics are de-
veloped and evaluated with Monte Carlo simulation. Strategies for making in-
terspecific and intraspecific comparisons among populations are discussed for
both longitudinal and cross-sectional sampling schemes with emphasis on the
appropriate scope of inference from different levels of sampling.

Growth patterns have been a focal point in studies of avian reproduction (Rick-
lefs 1968a, 1973, 1974). However, difficulties arise in such analyses because the
pattern of growth for any individual organism is the result of a continuous process
of change through time which is observed at a finite number of points. Depen-
dencies among these data complicate the process of estimation and inference. One
method which has been widely used in avian studies was developed by Ricklefs
(1967) and employs graphical techniques to fit three common growth curves. A
more general approach (requiring an iterative curve fitting algorithm) adds flex-
ibility by making use of Richards’ (1959) growth function.

Sandland and McGilchrist (1979) reviewed past work and described desirable
features for models of growth data. They emphasized differences between linear,
nonlinear, deterministic, and stochastic statistical models. Following Ricklefs
(1983), we stress the importance of levels sampling in determining the scope of
inference within longitudinal, mixed longitudinal, and cross-sectional sampling
designs. Our primary concern is development of techniques applicable with cross-
sectional data for comparing growth patterns among populations.

Our analysis techniques are illustrated with two sets of data which use body
weight as the size criterion. The first dataset contains 2002 measurements from
76 banded, nestling Burrowing Owls (Athene cunicularia) of known age. These
data were recorded during 1974 and 1975 at Seal Beach Naval Weapons Station,
Orange County, California (Landry 1979). They form a longitudinal sample which
can be used to compare methods designed for longitudinal data with those ap-
propriate for cross-sectional sampling schemes. The second data set includes 423
weighings of Least Terns (Sterna antillarum). Some terns were measured more
than once, but the separation between successive measurements, about one week,
was long enough so that the data could safely be treated as strictly cross-sectional
throughout this study. The terns were captured and banded between 1979 and
1981 in three coastal nesting sites in Los Angeles and Orange counties, California
(Collins and Atwood, unpublished data).

134 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

LEVELS Fixed set of Random Selection Systematic or random
OF : populations of individuals from selection of points
SAMPLING chosen for study the populations per individual

@oe fe}
= eee °o
= eee fe}

FORM OF . Ny | Between Between
INFERENCE we populations individuals

Fig. 1. Levels of sampling for growth data.

Figure | illustrates the three levels of sampling which are employed during
acquisition of growth data and indicates which levels may serve as proper bases
for different forms of inference from the data. In particular, inferences concerning
differences between populations must be based upon random sampling of indi-
viduals from these populations. With longitudinal data, this is easily accomplished
by fitting growth curves for each individual and then building a data matrix of
summary measurements per individual based upon the parameters of these curves.
This matrix can then be used with standard statistical techniques such as analysis
of variance.

This cannot be done with cross-sectional data since there are not enough data
on any one individual to fit a curve. A growth curve can be fitted to all of the
data per population, but care must be taken in testing for differences between
populations. With purely cross-sectional data, residual differences between data
points and the fitted curve confound the second and third levels of sampling (Fig.
1) and thus can be used as a basis for inter-population inference but with less
power than in the longitudinal case. Mixed longitudinal data contain dependencies
among the residuals which undermine this approach. However, construction of
confidence intervals for growth summary statistics which are founded upon vari-
ation between individuals would permit valid comparisons between populations.
We employ the jackknife procedure (Quenouille 1956; Mosteller and Tukey 1977)
for this purpose.

Selection of Summary Statistics

Raw growth data may be expressed as the sum of two components: a smooth
curve which portrays the major trend, plus residuals which incorporate short term
deviations from the general trend. The smoothed curve provides the foundation
for measurement of growth patterns. It is generally accepted that nonlinear curves
derived from differential equations yield the most readily interpretable parameters
(Sandland and McGilchrist 1979). We employed the flexible curve developed by
Richards (1959):

JACKKNIFE CONFIDENCE INTERVALS FOR GROWTH CURVES 135

1

W = A(1 + (M — I)e-KO-4),0 ™ 1)

T = the time of measurement (age of the organism)
W = the size of the organism at time T
A = the asymptotic size after growth is completed
K = a growth constant (controlling the rate of growth)
I = the time to reach the inflection point in the curve
M = ashape constant (controlling the shape of the curve).

This differs slightly from Richard’s original parameterization of the curve by
expressing the multiplier for the exponential term as the time to reach the inflection
point in the curve. It also has the advantage that the ambiguity of sign in the
exponential term is removed while the other parameters are left unchanged in a
manner generally comparable to but not quite identical to that proposed by Fletch-
er (1975). A derivation of our parameterization from Richards’ original formu-
lation is presented in the appendix.

Four parameters are used to fit Richards’ curve although only three are of
importance in describing the growth pattern. The time scale parameter (I) merely
positions the curve along the x-axis, and is not directly comparable for separate
curves since it depends upon the values for K and M. An alternative fitting strategy
which is more appropriate in the context of stochastic models eliminates the need
for this fourth parameter by setting the starting weight as a fixed initial condition
(Fletcher 1975; White and Brisbin 1980). That approach simplifies the fitting
process by reducing the number of parameters but has the disadvantage (for our
purposes) that the curve is forced to pass exactly through the first data point. We
choose to retain the time scale parameter (1) thus giving the initial data point
equal standing with the rest of the points.

The parameters lead :to convenient summary statistics describing the general
growth pattern. We follow Richards (1959) in selecting asymptotic size (A), weight-
ed mean growth rate (R), and percentage of asymptotic size acheived at inflection
(P) as effective statistics. We also include the time to pass from ten percent to
ninety percent of asymptotic size (G) since it has been used in prior studies
(Ricklefs 1967, 1968a, 1973; Ricklefs and White 1981) as an indicator of the
duration of the primary growth period. These statistics are defined as follows:

A = the raw parameter value from the curve

K
R=— 2)
M
1
P= Mog? 3)
G = In(l — .10'-)/(1 — .90'-))/K. 4)

Figure 2 illustrates how these summary Statistics arise from the growth curve.
The time to reach the inflection point (I) merely sets the time scale and conse-

136 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

(R) Weighted mean growth rate
(slope at inflection)

Percentage of
asymptotic size
at inflection

Asymptotic
Size

Time to grow from ten to ninety percent of asymptotic size

Fig. 2. Graphical representation of summary statistics from growth parameters. Note that the
weighted mean growth rate corresponds to the slope at inflection only when the scale of the ordinate
is set so that the inflection point occurs at 1.0.

quently is not reflected in the summary statistics. The formula for G is derived
in the appendix.

The formulas for Richards’ curve degenerate when the shape parameter (M)
reaches exactly one. However, the curve asymptotically approaches the Gompertz
curve as M approaches 1. Thus the following formulas for the Gompertz curve
replace the above equations in this limiting case:

—K(T—I)

W = Ae®< la)
P = l/e = 0.36788 3a)
G = In(In(.10)/In(.90))/K = 3.0844/K. 4a)

The curve can be fitted to raw data with careful application of nonlinear regression
methodology. Davies and Ku (1977) point out that Richards’ curve is difficult to
fit because it tends to produce a long narrow valley in the response surface. In
particular, they demonstrated that Causton’s (1969) Newton-Raphson approach
fails. We used a modified Gauss-Newton algorithm which resembles Marquardt’s
(1963) method in its use of ridge regression steps to recover from singularities
but differs because pure Gauss-Newton steps are taken whenever possible. Thus
in our method, the ridge parameter remains at zero until singularities arise.

Since variability in size is often essentially proportional to overall size, we
employ a log transformation of the size measures when fitting the curve. This
equalizes the variance of the residuals across time and substantially improves the
curve fit at early ages over results obtained with raw residuals.

The fact that the Richards’ curve is specified by four parameters does not imply
that four independent aspects of the growth pattern are being quantified. These

JACKKNIFE CONFIDENCE INTERVALS FOR GROWTH CURVES 137

Table 1. Correlations among parameters and summary statistics.

Raw parameters Summary statistics
A 1.00
K —.57 1.00
I 16 26 1.00
M —.50 95 oo 1.00
A 1.00 —.56 16 —.50 1.00
R —.26 .08 —.80 = allts} —.26 1.00
P —.47 .89 54 95 —.47 = 2) 1.00
G ) —.65 .O7 —.56 so —.31 —.68 1.00
A K I M A R P G

parameters have substantial interaction in the fitting process, most notable being
parallel adjustments of K and M which can result in very similar curves. The
summary statistics themselves are interdependent as well.

Table 1 illustrates this effect for the complete set of Burrowing Owl data. There
are very high correlations among the raw parameters and moderately high cor-
relations among the final transformed summary statistics. The strong correlations
between the raw parameters and the summary statistics are, of course, induced
by the definition of the summary statistics. -

Confidence Intervals for Summary Statistics

In addition to providing measures which summarize growth patterns, it is
desirable to have some indication of the accuracy of these measurements. We
explore three ways of achieving this by computing confidence intervals for the
summary statistics. One approach fits separate curves to each individual bird in
the sample and then uses the ordinary confidence interval around the mean for
each summary statistic. A second method, widely used in nonlinear regression,
builds a confidence interval for function parameters from the asymptotic variance
of their estimates (Venus and Causton 1979). A third method, which we discuss
in detail in this section, is the Jackknife, developed by Quenoille (1956) and
Mosteller and Tukey (1977).

The jackknife simulates the sampling distribution of the targeted summary
Statistic by estimating how the value might have changed if a different sample
had been drawn. This is done by repeatedly calculating the statistic on different
subsamples of the data, where each subsample includes all but one of the original
cases. These resampled estimates are then converted by the jackknife transform
into pseudo values which approximate the sampling distribution of the statistic:

Ve Ne (N = 1) ee: 5)

= The statistic’s pseudo value for the ith individual
U = Parameter estimate for the whole population

U_, = The parameter estimate excluding the i individual

= The number of individuals in the sample.

Z
|

138 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

Previous studies of the effectiveness of the jackknife in nonlinear regression
had reasonable success in obtaining univariate confidence intervals tempered by
poor performance in the multivariate case (Duncan 1978; Fox, Hinkley, and
Larntz 1980). They computed separate regressions omitting one case at a time
and transformed the resulting parameter estimates with the jackknife formula.
Approximate ninety-five percent confidence intervals were then obtained from
the mean (m) and standard error (s) of the pseudo values:

Clam "S(t gos — 1) 6)

where N is the number of independent pseudo values (cases).

We computed pseudo values for three of the summary statistics (A, R, G) after
conversion from the raw parameters (A, K, I, M). A logarithmic transformation
prior to applying the jackknife transform gave them a reasonably symmetric
distribution. Thus logarithms of A, R, G, and M were substituted for U in equation
5 instead of the raw values. Confidence intervals were obtained by inserting
logarithmic scale values into equation 6 and then converting back to the original
scale by taking anti-logarithms. Values for M were further transformed with equa-
tion 3 into percentage of asymptotic size.

Fortunately, it is not really necessary to perform separate nonlinear fits with
each case deleted. The pseudo values can be accurately approximated with linear
regression near the estimate obtained for the entire sample (Fox et al. 1980). This
is achieved by approximating the nonlinear response surface by a linear manifold
near the optimal solution point. We verified that this method performs well for
the growth curve application using a subset of the Burrowing Owl data. Corre-
lations between true least squares estimates and the linear approximations ranged
from .998 to .999 for the four parameters and none of the linear estimates differed
by more than one percent from their corresponding true values.

The entire set of data for each individual should be considered as a single case
for resampling. Thus with mixed or longitudinal data omission of a single case
involves deletion of more than one data point whenever multiple points are taken
for one case. Pseudo values are then obtained for each case rather than each
separate data point. This assures that the obtained pseudo values are independent,
and avoids overstating the degrees of freedom present in the data.

A Monte Carlo experiment was used to assess the accuracy of the estimates
and the confidence interval coverage rates. The idea was to sample individuals
from known populations and then compare the sample-based point estimates and
confidence intervals to the known population values. Bias, relative efficiency, and
confidence interval coverage rates were used to compare the effectiveness of es-
timates based upon the mean pseudo value with those centered at the summary
statistics directly obtained from the curves.

Ideally, this would be done with infinite populations. In order to obtain realistic
growth data, we used the actual data on owls and terns (described earlier), with
the hope that this restriction to finite populations would not seriously bias the
results. Figure 3 reports the results for six trials (each based upon 500 random
samples) with varying sample sizes. All of the Monte Carlo work done in this
project was accomplished with programs written in APL using the internal random
number generator (Roll: “*?”?) with APLUM 2.1 on a Control Data Corporation
Cyber 174 computer.

JACKKNIFE CONFIDENCE INTERVALS FOR GROWTH CURVES 139
Estimates Coverage Rates
Relative for 95% Confidence
viele Efficiency Intervals
BIRDS IN MEAN DIRECT MEAN DIRECT
RANDOMLY PSEUDO CURVE PSEUDO CURVE
DRAWN VALUE FIT VALUE FIT

Population samptes A R P G A RP G AR P G6 AR P &G
a3 4 BHO BHBH eC0oV5@e eo0ee
Burrowing
Owls 8 S888 SEES ee050e e0cge

(longitudinal)

2002 Points 16 BEES 88868 eeSe ee eo
423 32 OMB@O0 S8oOGBs 00@e0 eee
Least
Terns 64 GESEGB OagO eSee6e e800

(cross-sectional)
423 128 HOGS OHAG e0ee eeoe
@ 97-100% of best O 99-100% coverage
@ 90-96% © 97-98%
& 80-89% @ 94-96%
O 53-79% © 92-93%
i O 88-91%
Fig. 3. Relative efficiency of estimators and confidence interval coverage rates. The direct fit refers

to estimates obtained directly from the least squares curve parameters and the corresponding confidence
intervals centered about these estimates. The other estimates (and confidence intervals) were based
upon the mean of the pseudo values from the jackknife.

There are two common reasons for application of the jackknife: 1) confidence
interval estimation, and 2) bias reduction. Our simulations were designed to assess
its value in the growth curve application for each of these purposes. The results
indicated that bias in the estimates of the summary statistics should be of little
concern since the squared bias averaged only three percent of the mean squared
error in these estimates. Furthermore, Figure 3 shows that the direct estimates
performed slightly better than those based upon the mean of pseudo values with
respect to both relative efficiency and confidence interval coverage rates. We
conclude that the primary value of the jackknife in this situation is for confidence
interval estimation and that estimates of the summary statistics should be com-
puted directly from the curve parameters. The coverage rates displayed in Figure
3 are sufficiently close to the nominal level to recommend the use of jackknife
confidence intervals in this application.

Scope of Inference

The three methods described above for building confidence intervals around
the summary statistics vary in the scope of inference which they support. Intervals

140 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

obtained as standard confidence intervals for the mean from separate fits of curves
to each individual allow direct inference to the population from which these
individuals were sampled (Fig. 1). Ordinary statistical analysis techniques for
comparing populations, such as multivariate analysis of variance are also appli-
cable in this situation. In particular, nested designs may be appropriate when
several broods are measured for each population.

In contrast, use of the asymptotic covariance matrix as the basis for confidence
intervals yields a more narrow scope for inference. This approach would provide
confidence intervals for the parameter values for each bird individually if separate
curves were fitted per bird. This corresponds to the third level of sampling (Fig.
1) where data points are selected for a single individual. It indicates how the results
might have changed if different time points had been measured for the same birds.
These intervals could give tests for differences in parameters between selected
birds. This approach also provides.a basis for computing confidence intervals for
a bird’s weight at any point in time (Venus and Causton 1979).

If cross-sectional sampling was employed, these intervals could provide tests
for differences between populations, since in that case deviations from a curve
correspond to random variation among individuals in the population. However,
the asymptotic variance approach would not be appropriate for mixed longitudinal
sampling schemes since the independence of residuals would not be satisfied and
the degrees of freedom would be overstated (Ricklefs 1983).

Fortunately, the jackknife confidence intervals (when built by treating all points
measured on any single individual as a single case) are appropriate for inference
about populations even when used with mixed sampling schemes. They have the
additional advantage that (unlike asymptotic variance intervals) they can easily
be obtained for summary statistics which are nonlinear transformations of more
than one original parameter (e.g., G obtained from K and M).

When these jackknife confidence intervals are used with longitudinal sampling,
they provide estimates of the accuracy of fit for a single curve through all of the
data for a sample. This would permit comparison to other similar intervals ob-
tained with cross-sectional or mixed sampling schemes. Intervals obtained from
separate fits to each individual in the longitudinal design would not be comparable
to these jackknife intervals, since they estimate a different population parameter.

Table 2 shows the parameter estimates and confidence interval widths obtained
by all three methods for asymptotic weight with the complete set of 78 Burrowing
Owls, and also for a selected subset of six owls. The estimates for the asymptotic
and jackknife approaches are of course identical, and differ from the separate
curve estimate. The asymptotic theory intervals are narrower than the others.
However, they cannot be used as interval estimates of the asymptotic weight for
the population of Burrowing Owls from which this sample was taken. Instead,
the asymptotic theory intervals properly only estimate how the asymptote for
these particular owls might vary had they been measured at different time points.
Since the former scope of inference (to population parameters) is of more general
interest, the separate curve or jackknife intervals would usually be preferred to
the asymptotic theory intervals. The difference is most pronounced with the
smaller sample.

In contrast, when the jackknife (width 3.6) and asymptotic (width 4.3) confi-
dence intervals are compared for the complete data set on Least Terns, there is

JACKKNIFE CONFIDENCE INTERVALS FOR GROWTH CURVES 141

Table 2. Estimates and confidence interval widths for asymptotic weight.

All 78 owls Subset of 6 owls
Asymptote Width Asymptote Width
Asymptotic theory 140 (3.7) 134 (17)
Jackknife 140 (6.6) 134 (40)
Separate fit 143 (6.2) 138 (53)

much less difference in the results. This reflects the fact that both approaches have
the same scope of inference (to population parameters) with cross-sectional sam-
pling. They differ only in the form of approximation used to build the intervals.

Comparison of Sampling Strategies

Many factors influence the choice of sampling scheme. Often non-statistical
considerations take priority. For example, if subjects must be sacrificed in order
to make the measurements, only cross sectional data can be obtained (Brisbin
and Tally 1973). In field studies, gathering longitudinal data requires frequent
handling and possible disturbance of the subjects. This may not be acceptable
with some species, necessitating restricted visitation and limited collection of data
per individual. In contrast, the difficulty of making repeated measurements in
laboratory work is often minor compared with the cost of preparation for the
total experimental setting. In such situations observers normally collect longitu-
dinal data.

Statistical considerations generally favor longitudinal sampling schemes. A wide
variety of analytical techniques may be used to explore longitudinal data once
they have been transformed into summary statistics. Longitudinal designs have
one additional advantage. Since the summary statistics do not depend upon the
time scale parameter (I), they can be calculated even if the exact age of a subject
is unknown. The main disadvantage is the difficulty of data collection, particularly
under field conditions.

Proper choice of sampling design also depends upon the relative power of
different strategies. Is it better to add more cases or more points per case? We
performed a Monte Carlo simulation to answer this question. This experiment
involved only the Burrowing Owl data and was designed to assess the relationships
among 1) the power of the sampling design (as expressed by confidence interval
width), 2) the number of points per bird, and 3) the number of birds measured.
Each trial proceeded with a two stage sampling scheme. First n (ranging from 2
to 32) owls were randomly selected from the 78 available. Then p (from 2 to 16)
points were chosed from each bird’s data. Median confidence interval widths
based upon 100 trials per cell appear in Table 3.

The results confirm our a priori expectation that it is better to sacrifice points
per case (if necessary) in order to measure as many individuals as possible. One
potential exception to this recommendation is the case where sufficient points
may be obtained per case to treat the data as a longitudinal data set. Here the
advantages of longitudinal designs may outweigh the loss of power which accom-
panied reduction in sample size. In any case, a reasonable number of data must
be obtained if any generalizations are to be drawn from the sample. Overall, the

142 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

Table 3. Confidence interval widths based upon 100 trials for each cell with the Burrowing Owl
data.

Number of points per case

Number of cases 2 4 8 16
2 _ — 93 .90

4 _ 73 Sl .26

8 .63 34 Dit 14

16 34 pip) 13 .09

32 eal 13 .09 .06

Cells in the upper left corner included too few total data points and consequently were omitted.

intervals in Table 3 were reasonably narrow only when at least eight cases and
128 total data points were included.

Example

To illustrate how these techniques work with actual data, we arbitrarily selected
a pair of broods of Burrowing Owls for presentation as an example of comparison
between groups. Since the owl data are longitudinal in format, it was possible to
run the standard analysis obtained by fitting separate curves to each individual.
Thus confidence intervals for the mean of each summary statistic were computed
using the traditional technique based upon the t-distribution (Sokal and Rohlf
1969). In addition, jackknife confidence intervals were developed for summary
statistics based upon fitting a single curve through each brood’s data.

In addition to the confidence intervals, we computed comparison intervals
which were inspired by the work of Gabriel (1978), and in this application cover
exactly half of the range covered by the confidence intervals. The intent is sim-
plification of comparison between groups, with significant differences occurring
whenever comparison intervals fail to overlap each other. When several samples
are to be compared, the overall experimentwise error rate could be controlled by
adapting multiple comparison techniques. For example, the Bonferroni inequality
could be used (Green 1978). The method of Gabriel (1978) or the more recent
refinements by Andrews, Snee, and Sarner (1980) or Hochberg, Weiss, and Hart
(1982) provide alternatives.

The resulting confidence and comparison intervals are plotted in Figure 4. Here
the first brood grew more quickly than the second since it had a higher weighted
mean relative growth rate (R) and a shorter growth period (G). This brood also
reached the inflection point at a higher percentage of asymptotic weight (P).
However, it stabilized at a lower asymptote (A) than the second brood. These
patterns are essentially the same with both methods of computing confidence
intervals. In fact, the relative widths and positions of these intervals are quite
similar for the two methods. The main difference is that, for this example, the
estimates are lower when a single curve is fitted through pooled data than those
obtained as means from separate fits. This is not surprising, since the two methods
estimate different population parameters.

The brood that had a higher weight at inflection was the one which grew most
rapidly. This seems reasonable, since a larger proportion of its growth period

JACKKNIFE CCNFIDENCE INTERVALS FOR GROWTH CURVES 143

Summary Statistic
Strategy Brood A

Single Curve 1 +
Jackknife

+

Separate 1 +
Curve Fit

+

100 175 09 17 .25 60 10 30

Fig. 4. Comparison of the summary statistics for two arbitrarily selected broods of Burrowing
Owls. The upper panel shows results from fitting a single curve to each brood’s data and then using
the jackknife to compute confidence intervals. The lower panel fits separate curves for each bird and
uses the ordinary confidence intervals for a mean (based upon the t-statistic). The plot symbols identify
the estimate with a vertical bar. Comparison intervals correspond to the broad horizontal bars, while
the narrower horizontal bars mark the confidence intervals. The four summary statistics are asymptotic
size (A), weighted mean growth rate (R), percentage of asymptotic size at inflection (P), time to pass
from 10% to 90% of asymptotic size (G).

occurred in the exponential gain phase. This brood also reached a lower asymptote,
perhaps resulting from its restricted growth period. These observations reflect the
fact that the four summary statistics are highly interdependent.

If the two groups of birds being compared here had been randomly sampled
from two different populations, or randomly assigned two different treatments,
then the proper scope of inference in this analysis would be to differences between
populations or treatments based upon the second level of sampling presented in
Figure 1. In reality, the data incorporate two entire broods, with the individuals
completely enumerated rather than randomly sampled. Thus in this case the scope
of inference is to differences between broods based upon the third level of sampling
(Fig. 1). Here shorter confidence intervals could likely have been obtained by
jackknifing individual data points rather than entire birds, although the observed
differences were so substantial that this extra power was not needed in order to
detect differences between the groups.

Discussion

Problems in application of the Richards curve may arise from three sources:
1) autocorrelation among the residuals from the curve, 2) intercorrelation among
the parameters for the curve, and 3) truncation of the data prior to stabilization
at the asymptote. Each will be considered in turn.

Successive measurements for any particular bird will tend to show autocorre-
lation among the residuals from the Richards curve if they are separated by a
short enough time interval. If these residuals are then used as the basis for statistical
inference about the data, bias caused by the autocorrelation may undermine the
results. The effect could cause confidence intervals based upon the asymptotic
covariance matrix to be too narrow. White and Ratti (1977) discussed the au-
tocorrelation problem and proposed ways to correct for the bias. Fortunately, the

144 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

problem is relevant only at the third level of sampling (Fig. 1). Thus the separate
fit and jackknife approaches presented here evade the problem of autocorrelation
by treating each sampling unit as a separate entity.

The second problem (correlation among parameters) has caused difficulties
throughout this study. First, it exacerbates the nonlinear regression effort, and in
the process, destabilizes the parameter estimates. Interpretation is complicated
by the fact that the parameters are not really measuring separate aspects of the
growth patterns. The extremely high correlation (.95) between the growth constant
(K) and the shape of the curve (M) displayed in Table | indicates that these two
factors work together to estimate the rate of growth of the birds. They are not
separable into independent components. Simultaneous adjustment of these two
parameters can yield alternative but very similar curves which fit data nearly
equally well. This is at least partly due to the fact that the slope at inflection (R)
can remain the same despite changes in shape (M) if corresponding adjustments
are made to K.

The close tie between K and M reduces the utility of these values in comparing
growth patterns for separate populations since K is not comparable between curves
of different shape. The instability will often cause M and its transform (P) to have
broad confidence intervals and indicate differences between populations only when
their growth curves have a substantially different inflection point. One proposed
growth rate measure, R (the slope at inflection) is not affected by indeterminacy
in K and M and seems to be a potentially useful rate measure since it indicates
the maximal growth rate achieved during the entire growth period. Unfortunately,
it depends on the location of the inflection point and thus has limited compara-
bility across curves of different shape. In fact, the only strong association revealed
in Table 1 for R is a positive correlation with the time to reach inflection which
measures a different percentage of the total growth for curves that differ in shape.
The best overall growth measure appears to be the time to pass from 10% to 90%
of asymptotic size (G). This has two advantages: 1) it is less affected by instability
of the nonlinear fit than K or M, and 2) it produces values on an easily interpreted
scale (days or hours).

The asymptote is an alternative measure which seemingly should not be related
to the growth rate. However, with the Burrowing Ow! data, it showed a substantial
negative correlation with K, M, P, and G. This does not appear to be an artifact
of the fitting process for the curves, since for the owls, there was plenty of data
(30 days) beyond the primary growth period which provided an accurate basis
for estimating the true asymptotes.

The techniques for computing confidence intervals described earlier all produce
separate univariate intervals for each summary statistic. The extension to multi-
variate confidence regions is possible, but has shown little success in past work
with nonlinear regression (Duncan 1978). It is important to remember that the
high intercorrelation among the summary statistics guarantees that corresponding
confidence intervals are not independent of each other. For example, if two sam-
ples differ substantially on the basis of both asymptotic size and length of the
primary growth period, it is incorrect to argue that these samples differ in two
independent aspects of their growth pattern. Instead, the two observed differences
should be viewed as separate manifestations of a single underlying pattern.

The third problem in fitting Richards’ curve arises when data cannot be obtained

JACKKNIFE CONFIDENCE INTERVALS FOR GROWTH CURVES 145

beyond the time when the birds reach their asymptotic size. For example, if the
birds fledge and disappear as they reach their asymptotes, or if there is weight
recession (Ricklefs 1968b) and truncated data must be used to fit the curves, then
the asymptote will not be accurately specified by the data (Fletcher 1975). In these
cases, it may be necessary to tightly restrict the estimated asymptote to a reasonable
range in order to produce acceptable growth curve fits. Perhaps a more effective
strategy would be to fix the asymptote a priori at an appropriate value such as
the adult mean. In any case, with truncated data sets, it will not be generally
possible to separate the effects of asymptote and growth rate in determining the
curves.

In summary, the general strategy which we have pursued is to provide statistics
which measure certain aspects of the observed growth patterns. The Richards
curve serves in this context merely as a measuring device (an extension of the
spring balance) which extracts these summary statistics from the raw data. Other
researchers have explored a different area of application for the Richards growth
function: direct modeling of the growth process. Sandland and McGilchrist (1979)
and White and Brisbin (1980) have employed stochastic modeling techniques
with the Richards curve. In contrast, we do not consider our technique as an
appropriate model of the true growth process. Rather, it is merely a way of
summarizing growth patterns and comparing separate populations. For this pur-
pose, we recommend use of the duration of the primary growth period (G) and
(with adequate data) the asymptotic size (A) as appropriate statistics to describe
the growth patterns.

An APL program which performs the growth analyses presented in this paper
is available from the lead author upon request. It employs the previously discussed
modified Gauss-Newton method for nonlinear regression.

Acknowledgments

We thank Doug Bonett, I. Lehr Brisbin, Robert Ricklefs, Terry Root, and Gary
White for providing helpful criticism of an earlier draft of this manuscript. In
particular, we thank Robert Ricklefs for originally raising the critical question of
appropriate degrees of freedom. We thank Larry Fellow, Harry Metcalf and the
U.S. Navy for providing access to the Seal Beach Naval Weapons Station for our
research. We also thank Mary Allan, Jon Atwood, John Eckhoff, Andy Hughes,
Virginia Landry, Barbara Massey, and Dennis Minsky for assistance in the field.
Computer time was provided by the State University Data Center of the California
State Universities, and the Data Processing Department of California State Uni-
versity, Long Beach.

Literature Cited

Andrews, H. P., R. D. Snee, and M. H. Sarner. 1980. Graphical display of means. The American
Statistician, 34:195-199.

Brisbin, I. L. Jr., and L. J. Tally. 1973. Age-specific changes in the major body weight components
and caloric value of growing Japanese quail. The Auk, 90:624-635.

Causton, D. R. 1969. A computer program for fitting the Richards function. Biometrics, 25:401-
409.

Davies, O. L., and J. Y. Ku. 1977. Re-examination of the fitting of the Richards growth function.
Biometrics, 33:546—-547.

146 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

Duncan, G. T. 1978. An empirical study of jackknife-constructed confidence regions in nonlinear
regression. Technometrics, 20:123-129.

Fletcher, R. I. 1975. A general solution for the complete Richards function. Mathematical Biosci-
ences, 27:349-360.

Fox, T., D. Hinkley, and K. Larntz. 1980. Jackknifing in nonlinear regression. Technometrics, 22:
29-33.

Gabriel, K. R. 1978. A simple method of multiple comparison of means. Journal of the American
Statistical Association, 73:724-729.

Green, P. E. 1978. Analyzing multivariate data. The Dryden Press. xvi + 519 pp.

Hochberg, Y., G. Weiss, and S. Hart. 1982. On graphical procedures for multiple comparisons.
Journal of the American Statistical Association, 77:767—772.

Landry, R.E. 1979. Growth and development of the Burrowing Owl, Athene cuniculara, unpublished
masters thesis, California State University, Long Beach.

Marquardt, D. W. 1963. An algorithm for least squares estimation of nonlinear parameters. S.I.A.M.
Journal of Applied Mathematics, 11:431-441.

Mosteller, F., and J. W. Tukey. 1977. Data analysis and regression, Addison-Wesley Publishing
Company, Reading, Massachusetts. xi1 + 588 pp.

Quenouille, M. H. 1956. Notes on bias in estimation. Biometrika, 43:353-360.

Richards, F. J. 1959. A flexible growth function for empirical use. Journal of Experimental Botany,

10:290-300.
Ricklefs, R. E. 1967. A graphical method of fitting equations to growth curves. Ecology, 48:978-
983.

1968a. Patterns of growth in birds. The Ibis, 110:419-451.
—. 1968b. Weight recession in nestling birds. The Auk, 85:30-35.
1973. Patterns of growth in birds II, growth rate and mode of development. The Ibis, 115:
177-201.
1974. Energetics of reproduction in birds. Pp. 152—297 in Avian energetics. (R. A. Paynter,
Jr., ed.), Publications of the Nuttall Ornithological Society, 15:1-334.
1983. Avian postnatal development. Pp. 1-83 im Avian Biology, Vol. VII (D. S. Farner, J.
R. King, and K. C. Parkes, eds.), Academic Press, xxiv + 542 pp.
, and S. C. White. 1981. Growth and energetics of chicks of the sooty tern (Sterna fuscata)
and common tern (S. hirundo). The Auk, 98:361-378.
Sandland, R. L., and C. A. McGilchrist. 1979. Stochastic growth curve analysis. Biometrics, 35:
255-271.
Sokal, R. R., and F. J. Rohlf. 1969. Biometry, W. H. Freeman and Company. xxi + 776 pp.
Venus, J. C., and D. R. Causton. 1979. Confidence limits for Richards functions. Journal of Applied
Ecology, 16:939-947.
White, G. C., and I. L. Brisbin Jr. 1980. Estimation and comparison of parameters in stochastic
growth models for barn owls. Growth, 44:97-111.
, and J. T. Ratti. 1977. Estimation and testing of parameters in Richards growth model for
western grebes. Growth, 41:315-323.

Accepted for publication 25 January 1984. S

! Department of Data Processing and Information Systems, and ? Department of
Biology, California State University, Long Beach, California 90840.

Appendix
Richards original parameterization took the form:

1
W = A(1 + Be-KT)(1—M) Al)

where B represented the time scale adjustment. The sign flips from negative when M is less than 1| to
positive for M greater than 1. Richards also provided an expression for the weight at the time of
inflection:

1

W, = AM(-™) A2)

JACKKNIFE CONFIDENCE INTERVALS FOR GROWTH CURVES 147

Substituting the weight at inflection (from equation A2) for W in equation A1, dividing both sides by
A and raising both to the 1 — M power yields:

M=1+ Be*! A3)
where I (the time of inflection) is inserted for T. Solving for B gives:

B=(M — I)e™! A4)
Note that in this formulation B is automatically negative when M is less than one and positive otherwise,
eliminating the sign ambiguity when B (from equation A4) is substituted for +B in equation Al. This

substitution results in our form of Richards’ curve as expressed by equation | in the text.
The calculation of the time from 10% to 90% of asymptotic size begins with the formula for size

as a percentage (E) of the asymptote:

1
Ww
Eig Ct (OM Ne, cea) A5)

Solving for T in terms of E gives:
T =I — In(1 — M)/K — In(1 — E'-™)/K. A6)
The time differential from T ,9 to T 99 arises by subtracting equation A6 for E = .10 from the formula
with E = .90:
G = [I — Ind — M)/K — In(1 — .10!~-™)/K] — [I — Ini — M)/K — Ini — .90!~-™)/K]
Cancellation and composition yields the final formula for G:

G = In(l — .10'-“)/(1 — .90'-™))/K,

Bull. Southern California Acad. Sci.
83(3), 1984, pp. 148-151
© Southern California Academy of Sciences, 1984

A Recent Specimen of Collisella edmitchelli from San Pedro,
California (Mollusca: Acmaeidae)

David R. Lindberg

Abstract.—A recent specimen of Collisella edmitchelli from San Pedro, Cali-
fornia (Mollusca: Acmaeidae) by David R. Lindberg. Bull. Southern California
Acad. Sci., 83(3):148-151, 1984. The first known Recent specimen of Coilisella
edmitchelli (Lipps, 1966) is reported. The Recent specimen suggests that some of
the fossil specimens’ characters are preservational. Fossil specimens also may be
present on San Miguel Island. This species may have lived lower in the intertidal
region than previously thought.

Previously, Collisella edmitchelli (Lipps 1966) was the only acmaeid limpet
present in the Pleistocene of southern California that was not extant today; it was
known only from Pleistocene deposits on San Nicolas Island, Ventura County,
California (33°16'N, 119°30’W).

During the summer of 1983, I found a specimen lot identified as Acmaea
spectrum [=Collisella scabra (Gould 1846)] in the collections of the Museum of
Paleontology, University of California, Berkeley (UCMP). The specimen lot
(UCMP Loc. #2388) had been collected at San Pedro, California by J. G. Cooper
and contained three shells, two of which were typical C. scabra, but the third
specimen was quite different. Further examination of the odd specimen revealed
that it is a specimen of C. edmitchelli. In addition to the discovery of this Recent
specimen of C. edmitchelli, study of fossil specimens from several localities on
San Nicolas Island has produced additional information on this enigmatic species
that is included herein.

Examination of the shell layers, as they appear on the interior surface, revealed
that the Recent specimen belongs to MacClintock’s (1967) shell structure group
no. 16. Collisella scabra and C. edmitchelli are the only species known to have
this shell structure (Lindberg 1978).

The shell morphology of the Recent specimen (Figs. 1, 4) is clearly within.the
range of variation of C. edmitchelli (Figs. 2, 3, 5) and not C. scabra (Figs. 6, 7).
In C. scabra, the radial ribs number between 10 and 20, whereas in C. edmitchelli
there are between 20 and 30 radial ribs (Lindberg 1978). The Recent specimen
has 25 ribs. Moreover, the size of the ribs differs substantially between the two
species. In C. scabra, the primary ribs are about twice as wide as the secondary
ribs. In C. edmitchelli the secondary ribs are only slightly narrower than the
primary ribs. Overall, the ribbing of the Recent specimen (Fig. 4) is most similar
to that of C. edmitchelli and not that of C. scabra (Figs. 6, 7). Other characters
such as the shell profile, the crenulation of the aperture, and the lack of a central
stain, all are characteristic of C. edmitchelli and not C. scabra.

The San Pedro specimen is the only known Recent record of C. edmitchelli.
Dried mantle tissue along the posterior edge of the muscle scar confirms the Recent

RECENT COLLISELLA EDMITCHELLI 149

Figs. 1-5. Collisella edmitchelli 1, Recent, San Pedro, California; UCMP 37424. 2, Pleistocene,
San Nicolas Island, California, UCMP 37425. 3, Holotype, San Nicolas Island, LACMIP 1126. 4,
UCMP 37424 coated with ammonium cloride. 5, UCMP 37425 coated with ammonium cloride.

Figs. 6-7. Collisella scabra 6, Recent, Santa Catalina Island, California; UCMP 37426. 7, Pleis-
tocene, San Nicolas Island; UCMP 37427. Scales = 10 mm.

150 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

nature of the specimen. It also is the first record of C. edmitchelli outside of San
Nicolas Island.

The locality and collection date are accurate, Coan (1981) places J. G. Cooper
at San Pedro, California in 1861 and again in 1863. Collections were made on
both occasions. Evidence that Cooper collected the specimen comes from several
labels with the specimen and the entry in the UCMP locality register. Moreover,
on the interior of the specimen are the words “‘San Pedro”’ (Fig. 1). Similar locality
names, written in Cooper’s hand, are on other specimens collected by him from
Southeast Farallon Island and Santa Barbara.

The Recent specimen of C. edmitchelli significantly increases our knowledge of
this species. Lipps (1963) considered the dark interior margin of C. edmitchelli
(Figs. 2, 3) to be a diagnostic character of the species. However, it now appears
that this character is a product of diagenesis. The outer modified foliated shell
layer is glossy white in the Recent specimen but brown in Pleistocene specimens.
The same change occurs in the outer layer of C. scabra (Figs. 6, 7).

Another important character, the lack of the central stain, is a feature of the
Recent specimen. C. scabra has an irregular, dark brown central stain (Fig. 6) but
C. edmitchelli lacks a stain (Figs. 2, 3). The lack of a central stain is a conservative
character and not preservational. The Recent and fossil specimens of C. edmitch-
elli (Figs. 1, 3), lack a central stain. In C. scabra, an irregularly-shaped stain is
present on Recent and fossil specimens (Figs. 6, 7).

The restriction of fossil specimens of C. edmitchelli to San Nicolas Island and
the unique occurrence of a Recent specimen at San Pedro is puzzling. On San
Nicolas Island, C. edmitchelli is abundant on the higher terraces (terraces 11
through 5 of Vedder and Norris 1963). These terraces are between 770,000 +
100,000 yrs BP and 400,000 + 100,000 yrs BP (Lajoie, Wehmiller, and Kennedy
1980; G. L. Kennedy, pers. comm.). Other acmaeid limpets occurring with C.
edmitchelli include Collisella limatula, C. digitalis, C. pelta, and Acmaea mitra
(Vedder and Norris 1963). At most localities C. /imatula is the most abundant
followed by A. mitra and C. edmitchelli (D. R. Lindberg, unpublished data). C.
scabra is absent from the higher terraces and is abundant only on terraces 1-4.

C. edmitchelli may also be present in the higher terraces on San Miguel (34°3’N,
120°21'W). Cockerell (1938:9) reported limpets from the high terraces of the island
that were an “‘extreme variant”? of C. scabra. Specimens of C. edmitchelli could
certainly be mistaken for an “extreme variant” of C. scabra. Unfortunately these
specimens could not be located.

Lipps (1963), using shell morphology, postulated that C. edmitchelli was a high
intertidal species. However, one specimen (Fig. 2) has been drilled by a predatory
gastropod. The borehole in the posterior region of the shell is similar to those
produced by the muricid gastropod Nucella in the Recent San Nicolas fauna.
Nucella also occurs in some of the same terrace deposits as C. edmitchelli. Dis-
counting the possibility of physiological changes in Nucel/la spp., and the possi-
bility that the snail drilled an empty limpet shell, and the habitat suggested by
the Recent intertidal range of Nucella spp. would suggest that C. edmitchelli
occurred in the mid-intertidal region rather than the high intertidal. C. scabra
reaches its highest densities in the mid-intertidal region on San Nicolas Island
today (Estes and Lindberg, in preparation).

No other Recent specimen of C. edmitchelli was found in the collections of

RECENT COLLISELLA EDMITCHELLI 151

UCMP, the California Academy of Sciences, or the Natural History Museum of
Los Angeles County. However, the presence of this species during the 1860’s at
San Pedro, California suggests that additional specimens may be extant, possibly
in older collections.

Acknowledgments

I thank J. H. McLean, B. Roth, and E. C. Wilson for their helpful comments
and suggestions, E. C. Wilson for the loan of the type material, and J. W. Durham
for assistance with the photography. I am grateful to the U.S. Naval Pacific Missile
Test Center and Ron Dow for access to San Nicolas Island.

Literature Cited

Coan, E. V. 1981. James Graham Cooper, Pioneer Western Naturalist. University Press of Idaho,
225 pp.

Cockerell, T. D. A. 1938. Studies of island life. Univ. Colorado Stud., 26:3-20.

Lajoie, K. R., J. F. Wehmiller, and G. L. Kennedy. 1980. Inter- and Intrageneric trends in apparent
racemization kinetics of amino acids in Quaternary mollusks. Pp. 305-340 in Biogeochemistry
of amino acids (P. E. Hare, T. C. Hoering, and K. King, Jr., eds.), John Wiley and Sons,
Xvili + 558 pp.

Lindberg, D. R. 1978. On the taxonomic affinities of Collisella edmitchelli (Lipps) (Gastropoda:
Acmaeidae) a late Pleistocene limpet from San Nicolas Island, California. Bull. Southern Cal-
ifornia Acad. Sci., 77:65—-70.

Lipps, J. H. 1963. Anew species of Acmaea (Archaeogastropoda) from the Pleistocene of San Nicolas

Island, California. Los Angeles Co. Mus. Nat. Hist. Contrb., Sci., 75:1-15.

1966. A-new name for Acmaea mitchelli Lipps. The Veliger, 9:250.

MacClintock, C. 1967. Shell structure of patelloid and bellerophontoid gastropods (Mollusca). Pea-
body Mus. Nat. Hist. Yale Univ. Bull., 22:1-140.

Vedder, J. G., and R. M. Norris. 1963. Geology of San Nicolas Island California. U.S. Geol. Surv.
Prof. Pap., 369:1-65.

Accepted for publication 6 April 1984.

Museum of Paleontology, University of California, Berkeley, California 94720.

Bull. Southern California Acad. Sci.
83(3), 1984, pp. 152-153
© Southern California Academy of Sciences, 1984

Research Notes

Observations of Harbor Seal, Phoca vitulina richardsi
Feeding in Southern California Waters

Although the diet of the harbor seal, Phoca vitulina richardsi, has been de-
scribed for many areas, little is known of the diet in southern California waters.
Harbor seals from northern California to Alaska are generally opportunistic feed-
ers. Their diet varies with season and location and includes pelagic, benthic, and
anadromous fishes, as well as molluscs and crustaceans (Spalding 1964; Kenyon
1965; Pitcher 1980a,b; Antonelis and Fiscus 1980; Jones 1981).

Additionally, little is known of the feeding behavior of harbor seals. Hobson
(1966) noted that night feeding harbor seals swam on their backs and approached
pelagic prey from below. This would be advantageous in that the prey is silhouetted
against the water surface by the light, even starlight, coming from above. This
note describes feeding behavior of harbor seals observed off the southern Cali-
fornia Channel Islands and reports a new prey species.

On 20 October 1982, between 2050 to 2110 hours, we made a night SCUBA
dive at Eagle Reef, a rocky pinnacle offshore of Lion Head Point, near Isthmus
Harbor, Santa Catalina Island. Dive depths were 7.6 to 10.7 m and water visibility
was estimated at greater than 10 m. Each of us carried a bright underwater
flashlight (Underwater Kinetics SE 200).

Eagle Reef includes several pinnacles bordered by shelves and flat rocky reef
areas with many crevices and shallow undercuts. The blacksmith, Chromis punc-
tipinnis, a nocturnally inactive pomacentrid fish, uses such cracks and crevices
for night-time shelter (Hobson et al. 1981).

A harbor seal first appeared early in the dive, swimming under us in a supine
attitude. During the first encounters, the seal circled, then brushed two of us
(D.O.P. and K.C.H.). Still in an inverted position, the seal proceeded to a shallow
crevice and began probing along it with its muzzle. Whenever it approached a
blacksmith, the seal would stop momentarily, then dart at the fish, often catching
it. We observed this behavior repeatedly during the dive. Apparently the seal
swallowed the fish underwater as it once captured several blacksmith on a single
dive.

On the night of 15 December 1982, two of us (P.L.H. and D.O.P.) returned to
Eagle Reef. Another, smaller harbor seal was observed feeding as described above.
During this encounter, the seal released the almost separated halves of a black-
smith, bit the head, and swallowed the fish head first. During feeding dives this
seal swam in both a supine and a prone attitude.

On 8 March 1977, Jack Engle and Tim Tricas (pers. comm.) of the University
of Southern California Catalina Marine Science Center made a night dive at Ship
Rock, also near Isthmus Harbor, Santa Catalina Island. The dive began at 2245
hours and covered waters to 10 m deep. The divers observed “Fish Hook,” a
somewhat tame harbor seal foraging in “every conceivable crevice.” Engle noted
that “away from the range of our lights, he seemed to be catching blacksmith by

RESEARCH NOTES 153

touch. With the aid of our lights he could see to dart out at fleeing blacksmith.
Usually he bit the fish immediately upon contact, then pulled back and readjusted
it for swallowing head first.”’ This behavior was observed over a 45 minute period,
during which “Fish Hook” ate 20 or more blacksmith. Tim Tricas also saw the
seal eat a blackperch, Embiotoca jacksoni, which was lying in low-growing algae.

On 10 October 1982, at about 2100 hours, Todd Tognazzini (pers. comm.), a
National Park Service Ranger, observed a harbor seal pursuing fishes beneath
rock ledges in 6.5 to 7.6 m deep water off Middle Canyon at Santa Barbara Island.
The seal forced its body under ledges and churned up substrate as it attempted
to swim deeper into the crevice. The seal removed itself from the crevice, turned
from the light, and swam away. It could not be determined what prey, if any,
had been taken.

Blacksmith are commonly seen during the day in large schools around kelp
forests. They appear to be sufficiently abundant to be an important prey item in
the diet of P. vitulina. At night, when taking shelter in crevices, blacksmith are
probably more susceptible to seal predation. However, we cannot say whether
the behavior observed above is normal or just an artifact of the presence of artificial
illumination. If such behavior is normal for the species, other fishes having similar
nocturnal habits as that of the blacksmith, may also be susceptible to predation
by harbor seals.

Acknowledgments

We thank Jack Engle, Tim Tricas, and Todd Tognazzini for allowing the use
of their observations; and Doug DeMaster, Brent Stewart, and Harold Clemens
for their helpful comments on the manuscript. This work was accomplished under
the California Department of Fish and Game Nearshore Invertebrate Evaluation
Project.

Literature Cited

Antonelis, G. A., Jr., and C. H. Fiscus. 1980. The pinnipeds of the California current. Rept. California
Coop. Ocean. Fish. Invest., 21:68-78.

Hobson, E. S. 1966. Visual orientation and feeding in seals and sea lions. Nature, 210:326-327.

, W.N. McFarland, and J. R. Chess. 1981. Crepuscular and nocturnal activities of Californian

nearshore fishes, with consideration of their scotopic visual pigments and the photic environ-

ment. Fish. Bull. U.S., 79:1-30.

Jones, R.E. 1981. Food habits of smaller marine mammals from northern California. Proc. California
Acad. Sci., 42:409-433.

Kenyon, K. W. 1965. Food of harbor seals at Amchitka Island, Alaska. J. Mammal, 46:103-104.

Pitcher, K. W. 1980a. Food of the harbor seal, Phoca vitulina richardsi, in the Gulf of Alaska. Fish.

Bull. U.S., 78:544-549.

. 1980b. Stomach contents and feces as indicators of harbor seal, Phoca vitulina, foods in the

Gulf of Alaska. Fish. Bull. U.S., 78:797-798.

Spalding, D. J. 1964. Comparative feeding habits of the fur seal, sea lion and harbor seal on the
British Columbia coast. Bull. Fish. Res. Bd. Canada, 146:1-52.

Accepted for publication 4 October 1983.
Peter L. Haaker, David O. Parker, and Kristine C. Henderson, California De-

partment of Fish and Game, 1301 West 12th Street, Long Beach, California
90813.

Bull. Southern California Acad. Sci.
83(3), 1984, pp. 154-157
© Southern California Academy of Sciences, 1984

Two Species of Lytechinus (Toxopneustidae: Echinoidea:
Echinodermata) Are Completely Cross-fertile

The sympatric sea urchins, Lytechinus anamesus Clark 1912 and L. pictus
(Verrill, 1867) are found in different microhabitats. L. anamesus primarily resides
in open coast areas, while L. pictus is found in bays and lagoons (Ricketts and
Calvin 1968; T. A. Ebert, S. Schroeter and others, pers comm.). Although Mor-
tensen (1943) considered the two forms to be valid species, Mayr (1954) omitted
L. pictus from his zoogeographical discussion of the genus because he thought it
was possibly a synonym of L. anamesus. A key to the species of the genus dis-
tinguishes between the two forms on the morphology of the spicules found in the
stalk of the globiferous pedicellariae (Chesher 1968). Word and Charwat (1975)
separate the two species in their key on the basis of spine morphology and color.
Furthermore, the two forms are readily cross-fertilizable and embryonic cells will
reaggregate to produce viable early embryos (Vacquier, quoted in Durham et al.
1980).

The observations above could be accounted for by one of at least two hypotheses.
The two forms could be ecotypic variants within a single species or truly isolated
populations between which no genetic exchange occurs. In order to exclude the
isolation hypothesis, it must be shown that: (1) the hybrids can successfully de-
velop and survive to the adult form and (2) exchange of gametes or larvae can
occur in nature. The tests for development and survival can be easily accomplished
since techniques for larval rearing are readily available (Hinegardner 1969; Cam-
eron and Hinegardner 1974).

Materials and Methods

Adults obtained from Mission Bay, San Diego, California, and off San Onofre,
California were maintained in running seawater at the Long Marine Laboratory,
University of California, Santa Cruz. Observations were made on these animals
as well as laboratory reared populations of unknown pedigree. é

To test for larval survival, reciprocal crosses between the L. pictus and L.
anamesus were made. Gametes were obtained by injection of 0.55 M KCl through
the peristomial membrane. Eggs from each female were separately collected in
filtered seawater and semen was collected undiluted on ice. Fertilization was
accomplished by mixing 10 ml of a 10% egg suspension with a few drops of a 1%
suspension of semen. After thorough washing to remove sperms, fertilized eggs
were diluted and maintained at 18 degrees C.

At two days post-fertilization, the embryos were diluted to one per 15 ml and
placed in stirred culture vessels (Cameron and Hinegardner 1974). The cultures
were fed a unicellular cryptophyte, Rhodomonas lens, which was kept at approx-
imately 3000 cells/ml by daily inspection. Mature larvae were challenged to settle
by placing them in plastic petri dishes which had acquired a bacterial film while
lying in aquariums containing adult urchins.

RESEARCH NOTES 155

Random aliquots of seawater from vigorously stirred cultures containing em-
bryos or larvae were transferred at each succeeding culture step in order to prevent
unwitting selection of particularly healthy individuals.

Embryos and larvae were examined using a Baush & Lomb dissecting micro-
scope and a Zeiss compound microscope with phase and Nomarski optics. Ob-
servations were made on the following schedule:

Time after
fertilization Event scored Total observed
2 hrs cleavage 800
2d gut development 4000
general appearance
49d metamorphosis 5153
Results

Using the criteria of pedicillarial stalk spicule morphology described by Chesher
(1968), I was unable to separate the two species. The forms of the spicules graded
from club-shaped to hooked regardless of the type observed. Based on the external
features of spine length and test color used by Word and Charwat (1975), the two
species are easily separated.

Because the two forms of Lytechinus freely interbreed and produce embryos
under laboratory conditions, developmental biologists have used them extensively
without considering the species differences. Offspring from inadvertent mixing of
the two forms have been reared to sexual maturity without regard to pedigree.
They assume the form of L. anamesus in the wild. They have longer light-colored
spines and a whitish test blotched with purple. A gradation in form from dumbell-
to C-shaped pedicellarial spicules is also evident.

After two hours of development, 80% or more had completed at least one
cleavage (Table 1). Some had progressed to the four-cell stage.

Table 1. Percent of zygotes exhibiting cleavage furrows at two hours after fertilization. Estimates
based on state of 200 eggs for each cross.

2 anamesus Q pictus
8 anamesus 89% 99%
6 pictus 92% 94%

The appearance of the early plutei which were transferred to feeding cultures
was also the same among the four groups. At each weekly water change, no
difference in growth or development was observed between any of the four groups
of larvae. After 49 days in culture the larvae appeared mature and competent to
undergo metamorphosis. Of the 2153 individuals scored in all the crosses, 87%
or more had undergone the rapid events of metamorphosis and begun to develop
adult structures (Table 2).

Discussion

These experiments indicate that the integrity of these species is not maintained
by a barrier to successful larval development. In all crosses, the eggs cleaved

156 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

Table 2. Percent of larvae from each cross which successfully completed metamorphosis.

Percent
ou x49 metamorphosis Number
anamesus X anamesus 97 526
pictus X pictus 93 364
anamesus X pictus 93 475
pictus X anamesus 87 788

successfully and the embryos developed normally within the limits of the sea
urchin culture techniques (Hinegardner 1969; Cameron and Hinegardner 1974).
The percents of metamorphosis were within a normal range for all the crosses
also. From these investigations it is clear that no barrier to cross breeding or
intertypic larval development exists. Therefore, isolation of these populations
could result from difference in settlement, juvenile survival or larval and gametic
transport. Since the intratypic matings settled successfully in the laboratory, it is
reasonable to assume that substrate selection is the least likely isolation mecha-
nism. It is also probable that the larvae of each form would occasionally find their
way into the habitat of the congener. Perhaps after settlement juveniles cannot
survive in the contratypic habitat.

Morphological variation of the same magnitude as seen in Lytechinus has been
documented in the latitudinally broad ranging sea urchin, Arbacia punctulata. In
addition, the genetic variation between populations of this species has been in-
ferred from protein variation (Marcus 1977). Although the gene frequencies were
significantly different between populations, the overall values calculated for gene
distance were within the range observed for geographical populations of other
species. Both migration accomplished by larval transport and selection have been
proposed to account for the differences in gene frequency. While the selection
factors were implied to be physical ones, no data is available on this question.

Interspecific hybrids of sea urchins have been reported many times (see Harvey
1956 and Giudice 1973 for review). Hybrids between the purple sea urchin,
Strongylocentrotus purpuratus, and the red sea urchin, S. franciscanus are of
particular interest because they are a co-occurring pair of congeneric species found
within the range of the Lytechinus forms. Crosses between untreated purple sea
urchin eggs and red sea urchin sperms yielded as high as 18% fertilization while
reciprocal crosses yielded 9% (Chaffee and Mazia 1963). These levels were ac-
complished by adding large amounts of sperms to a concentrated egg suspension.
They can be regarded as unusually high yields, generally the percent of fertilization
is much less (Hinegardner 1967). In these experiments development was only
observed until the pluteus stage (72 hr). It is not known if the hybrids would
develop normally to metamorphosis. Because the levels of fertilization in recip-
rocal crosses described here have always exceeded 90%, it is more likely that these
forms are not separate species, unlike the Strongylocentrotus case. However, before
these species can be combined, other critical tests must be made. For example,
do real barriers to migration or gamete exchange exist; 1.e., is hybridization pos-
sible in the field. The Lytechinus forms may simply be ecotypic variants. This
hypothesis has not been examined here. The best test of the hypothesis would be
transplantation of young adults from one habitat to the other with observations
on changes in spine length and test color over a period of time. Recently it has

RESEARCH NOTES 157

been shown that Strongylocentrotus drobachiensis and S. pallidus can produce
viable hybrids in the laboratory. In nature, the species probably remain distinct
because of sufficient selection against hybrids (Strathmann 1981).

Literature Cited

Cameron, R. A.,and R. T. Hinegardner. 1974. Initiation of metamorphosis in the laboratory-cultured
sea urchin. Biol. Bull., 146:335-342.

Chaffee, R. R., and D. Mazia. 1963. Echinochrome synthesis in hybrid sea urchin embryos. Dev.
Biol., 7:502.

Chesher, R. H. 1968. Lytechinus williamsi, a new urchin from Panama. Breviora, No. 305, 13 pp.

Durham, W. J., C. Wagner, and D. P. Abbott. 1980. Echinoidea: The sea urchins. Jn Intertidal
Invertebrates of California. (R. Morris, D. P. Abbott and E. C. Haderlie, eds.), Stanford Uni-
versity Press, Stanford, California.

Giudice, G. 1973. Developmental Biology of the Sea Urchin Embryo. Academic Press, New York.

Harvey, E. B. 1956. The American Arbacia and other sea urchins. Princeton University Press,
Princeton, New Jersey.

Hinegardner, R. T. 1967. Echinoderms. /n Methods in Developmental Biology. (F. Wilt and N.

Wessells, eds.), Thomas Crowell Company, New York.

1969. Growth and development of the laboratory-cultured sea urchin. Biol. Bull., 137:465-

475.

Marcus, N. H. 1977. Genetic variation within and between geographically separated populations
of the sea urchin, Arbacia punctulata. Biol. Bull., 153:560-576.

Mayr, E. 1954. Geographic speciation in tropical echinoids. Evolution, 8:1-18.

Mortensen, T. 1943. A Monograph of the Echinoidea 3 (2): Camarodonta. 1. Orthopsidae, Glyco-
cyphidae, Temnopleuridae and Toxopneustidae. C. A. Reitzel, Copenhagen, 553 pp., 321 figs.,
56 pls.

Ricketts, E. F., and J. Calvin. 1968. Between Pacific Tides. Stanford University Press, Stanford,
California.

Strathmann, R. R. 1981. On barriers to hybridization between Strongylocentrotus drobachiensis
(O. F. Miiller) and S. pallidus (G. O. Sars), J. Exp. Mar. Biol. Ecol., 55:39-47.

Word, J., and D. Charwat. 1975. Invertebrates of Southern California Coastal Waters. I. Select
Groups of Annelids, Arthropods, Echinoderms, and Molluscs. Southern California Coastal
Water Research Project, El Segundo, CA.

Accepted for publication 20 June 1982.

Dr. R. Andrew Cameron, Department of Marine Sciences, University of Puerto
Rico, Mayaguez, Puerto Rico 00708.

Bull. Southern California Acad. Sci.
83(3), 1984, pp. 157-162
© Southern California Academy of Sciences, 1984

Humpback Whale (Megaptera novaeangliae) Sighting off
Los Angeles Harbor, Southern California

Humpback whales, Megaptera novaeangliae, are found in all oceans. In the
Northern Hemisphere they migrate from southern winter grounds to northern
summer feeding grounds. They follow no clear-cut migration routes, and the
boundaries of their summer and winter grounds are uncertain. No sightings of
this species in the San Pedro Channel have been reported to date. Data are
presented here on a nearshore sighting of 7. novaeangliae off the San Pedro area

158 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

on 5 May 1982. The whale’s distinctive scars and disfigurations may enable its
migration movements to be traced.

Severe whaling pressures have reduced the world humpback population to
6000-7000 animals (Anonymous 1982). The North Pacific population, about
1000 whales (Leatherwood et al. 1982), migrates from mostly Arctic waters to
three main wintering sites. Over 650 humpbacks migrate to the Hawaiian Islands,
100-200 to Mexico, and a small, undetermined number, to island waters off Asia.
Information concerning migration routes and site fidelity has been largely deter-
mined from the tracking or capture of whales with implanted tags. Another method
utilizes photographic identification based on natural tags such as color patterns
and shapes of flukes and dorsal fins. Photo-identification systems have been com-
puterized for both North Pacific and North Atlantic humpback whale populations
(Anonymous 1982; Katona et al. 1981). Until a few years ago, most researchers
believed that ““Mexican’’ humpback whales summered off southern Alaska and
‘‘Hawaiian’’ humpbacks migrated to the Aleutian Islands. Photo-documentation
has established that at least part of these two “stocks” share feeding grounds in
southeastern Alaska (Lawton et al. 1979; Payne 1982). Photographs also depict
some whales migrating to Hawaii one summer and to Mexico the next, thus
merging the breeding stocks (Payne 1982).

121° 120° 119° 118° 117°
] = ii aD T at T + 7
| —
i CONSE nnIONE === Se Santa _Barbara
| es
| oes
a
5 SANTA BARBERA CHanye, Y=
Las aoe Se it
o| ee ==S2,_ SRA As SSS °Los Angeles
aap % 2 _3 s=Ie 3 J
| 4
°
ta
a
=
R e 6h a _ 1
ey ", —
fa) & =.
° aa a
e, SSs o =
3 Ne
| A =
a ¥ ' S
in | - ° yy c E: “|
oN e =
= - San
| Diego
| - 5
Aa
e
*
2h Ale aN _ \ t | '

LEGEND

SIGHTINGS BY MONTH: JANUARY-MARCH CAA); APRIL-JUNE €*%); MAY, 1982 CA);
JULY-SEPTEMBER (@); OCTOBER-DECEMBER ([)).

ISLANDS: SAN -MIGUEL C1); SANTA ROSA (2); SANTA CRUZ (3); ANACAPA C4);
SAN NICOLAS (5); SANTA BARBARA (6); SANTA CATALINA (7);
SAN CLEMENTE (8); LOS CORONADOS (9).

Fig. 1. Megaptera novaeangliae sightings from 1949-1978, and 1982 sighting, in the southern
California Bight.

RESEARCH NOTES 159

a —

e Los Angeles Breakwater. Rostrum is at A, and nares

Fig. 2. M. novaeangliae surfacing ne r
at B.

Humpback whales are regularly sighted in their eastern North Pacific winter
grounds October through May. Mexican grounds extend south along the west
coast of Baja California and around the peninsula into the Gulf of California
north to at least San José Island. According to Leatherwood et al. (1982), the
summer grounds extend from the Bering Sea south to Point Conception. Rice
(1974) maintained that many sightings occur throughout the summer off central
California. For the last five years, Woodhouse (pers. comm.) has recorded annual
sightings of 20-30 whales in the Santa Barbara Channel between May and Sep-
tember. A Bureau of Land Management (BLM) survey (Norris et al. 1976, 1981)
summarized 28 humpback whale sightings reported from Point Conception to
the Mexican border between 1949 and 1978. Most whales were sighted from June
to November in water 90 to 1500 m deep. The study concluded that the Santa
Barbara Channel appears to be a humpback whale seasonal gathering ground, and
the Santa Rosa-Cortes Ridge may be a major migratory pathway. No sightings
have been reported from the San Pedro Channel. The 1982 sighting documented
below is incorporated with BLM sighting data in Figure 1.

While conducting marine biology classes for Los Angeles school children, the
vessel DAIWA sighted a single whale at 1330 h on 5 May 1982. Three other
vessels engaged in similar activities joined the first: the MAVERICK, the AT-
LANTIS, and the GOLDEN DOUBLOON. The following observations took place
from the GOLDEN DOUBLOON. Calm sea conditions and clear photographs
aided verification of the whale’s identity as a humpback.

160 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

Fig. 3. M. novaeangliae surfacing in kelp, with scars at A and B, and deformation posterior to
dorsal fin at C.

The whale was initially sighted at latitude 33°42.4’N and longitude 118°15.1'W.
At this time it was swimming near the Los Angeles Harbor entrance (Angel’s
Gate), less than 50 m from the breakwater (Fig. 2). The vessel stayed with the
whale for 5.6 km, last sighting it at 33°41.8’N and 118°19.4’W at 1445 h. The
whale’s size was estimated at 12.5—13.5 m. Water depth varied from 18 to 22 m
during most of the encounter, but had increased to 50 m when contact broke off.
The whale’s proximity to shore ranged from 0.5 to 3.5 km. It surfaced as close
as 15 m from the vessel, enabling detailed observations of scars and prominent
deformation posterior to the dorsal fin. Traveling 6.5 to 7.5 km/h, it typically
blew three to five times, and then made a short dive lasting between 2 and 4 min.

The whale usually came up blowhole first, slowly arching with apparent torquing
of its trunk to the right, and dove without raising its flukes. The dark dorsal surface
of the flukes was glimpsed only once. Either the relatively shallow water traversed
by the whale during the observation period or the presence of a back injury could
have accounted for the lack of characteristic fluking behavior. The pectoral fins
were never observed. The watex’s high plankton content, combined with the boat’s
low viewing angle, limited the underwater visibility to about 3 m.

=

Figs. 4a and 4b. Dive sequence showing view of whale’s right side, with scars at A and B, and
deformation posterior to dorsal fin at C.

162 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

The rostrum showed clearly twice. The first time followed a relatively long dive
(Fig. 2). The second display occurred when the whale raised its head through the
sole patch of giant kelp (Macrocystis pyrifera) in its path. The humpback paused,
draped in seaweed, and then moved through the fronds (Fig. 3). Rostral tubercles
showed distinctly, as did an apparently healed injury (A, B, and C) illustrated in
Figures 3 and 4. A small white scar (A) is visible at the posterior base of the
dorsal fin. These figures also indicate the presence of at least six fairly parallel
scars (B) below the dorsal fin on the right side. The dorsal ridge posterior to the
fin is s-shaped, a feature which produced laterally distorted ridge bulges (C). The
parallel nature of the scars, and their relative position to the spinal deformation,
suggest that this injury may have been the result of a collision with a boat.

In addition to identifying a distinctly marked individual in an area from which
no previous encounters have been reported, this shallow water sighting is signif-
icant because of the paucity of published data concerning the humpback whale’s
migration movements. The unusual individual documented here can be readily
identified by features other than its flukes, which will ease tracking of its move-
ments and behavior among other whales. If it regularly winters in Mexican waters,
the small number of neighboring whales increase its chances of being resighted.
Evidence provided by resightings will add to knowledge concerning migration
routes, seasonal range parameters, and perhaps even population crossovers.

Acknowledgments

I thank Mr. William F. Samaras and Dr. Donald R. Patten for their critical
reviews of this manuscript. The helpful suggestions of Dr. Oscar Janiger, M.D.
are also appreciated.

Literature Cited

Anonymous. 1982. Marine mammal protection act of 1972:-annual report. U.S. Dept. Commerce,
NOAA, NMEFS. 77 pp.

Katona, S., P. Harcourt, J. Perkins, and S. Kraus, eds. 1981. Humpback whales: a catalogue of
whales identified in the western North Atlantic by means of fluke photographs. Bar Harbor,
Maine: College of the Atlantic. 169 pp.

Lawton, W., D. Rice, A. Wolman, and H. Winn. 1979. Unpubl. abstract, Proc. Third Conf. Biol.
Marine Mammals, Seattle, Washington, 7-11 Oct. 1979, p. 35.

Leatherwood, S., R. R. Reeves, W. F. Perrin, and W.E. Evans. 1982. Whales, dolphins, and porpoises
of eastern North Pacific and western Arctic waters: a guide to their identification. U.S. Dept.
Commerce, NOAA Tech. Rept., NMFS Circ. #444. 245 pp.

Norris, K. S., et al. Center for Coastal Marine Studies. 1976. Marine mammal and seabird survey
of the southern California bight area: cetacea. Prepared for U.S. Bureau of Land Management.
Nat. Tech. Info. Service, Washington, D.C. Vol. II, part H, pp. 294-306.

———. 1981. Summary of marine mammal and seabird surveys of ihe southern California bight
area, 1975-1978. Nat. Tech. Info. Service, Washington, D.C. Vol. III, part II, pp. 348-358.

Payne, R. 1982. New light on the singing whales. National Geographic Magazine, 161(4):464-466.

Rice, D. W. 1974. Whales and Whale Research in the Eastern North Pacific. Pp. 185-187 im The
whale problem: a status report. (W. E. Schevill, ed.), Harvard Univ. Press, Cambridge, 419 pp.

Woodhouse, C. D. Personal communication. 15 Nov. 1982. Vertebrate zoologist, Santa Barbara
Museum of Natural History, 2559 Puesta del Sol Road, Santa Barbara, California 9310S.

Accepted for publication 1 September 1983.

Alisa Schulman, Sea Education Afloat (SEA) Instructor, San Pedro Science Cen-
ter, 2201 Barrywood Ave., San Pedro, California 90731.

Bull. Southern California Acad. Sci.
83(3), 1984, pp. 163-166
© Southern California Academy of Sciences, 1984

A New Genus and Species of Iphitimid Parasitic in an
Aphroditid (Polychaeta), with an Emendation of the
Family Iphitimidae

Endoparasitic polychaetes taken from the coelomic cavity of Aphrodita longi-
palpa Essenberg are herein described as a new genus and species of the Iphitimidae.
The familial definition of the Iphitimidae is therefore emended to encompass the
new genus. Type materials are deposited at the Allan Hancock Foundation (AHF),
University of Southern California, Los Angeles, California.

Family Iphitimidae Fauchald, 1970, emended
[Type genus: [phitime Marenzeller, 1902]
Iphitimidae Fauchald, 1970:118 (By monotypy)
Iphitimidae: Pilger, 1971:84 (emended)

Familial diagnosis.—Prostomium short, rounded or truncate; 1 pair of frontal
antennae present or absent. Peristomium up to 2 segments long. Parapodia uni-
ramous; 1-8 acicula present per parapodium. Pharyngeal apparatus with 1-3 pairs
of maxillae; 1 pair of short maxillary carriers fused to falcate maxillae I; third
maxillary carrier absent. Mandibles present; partially to completely fused medi-
ally. Setae including simple or compound falcigers, all without distal hoods.

Generic diagnosis.— Body long and slender, ventrally flattened; branchiae ab-
sent. Prostomium without antennae. Two peristomial segments, each apodous,
lacking cirri. Parapodial lobes simple, uniramous. Setae of one kind, entirely
simple. Pharyngeal apparatus with one pair of maxillae, these fused basally with
carriers, these fused together medially. Mandibles completely fused medially into
a single oblong structure.

Remarks.— This new genus is included in the Iphitimidae since the maxillary
carriers are basally fused to maxillae I. This character is atypical of other families
of the Eunicida, although it may be present in dorvilleids (Fauchald 1970). Ve-
neriserva resembles [phitime only in body form, lacking both branchiae and pro-
stomial appendages. Veneriserva has only | pair of maxillae while [phitime has
2-3 pairs. The fusion of the mandibles into a single oblong structure is unique to
Veneriserva.

Etymology.—The generic name (venerius, Latin = of or pertaining to Venus;
serva, Latin = servant) refers to the close association with an Aphrodita.

Veneriserva pygoclava, n. sp.
Fig. la-f

Material examined.—A total of 8 specimens as follows: Holotype (AHF 1275)
and 3 paratypes (AHF 1276) taken from an 4. Jongipalpa collected at Southern
California Coastal Water Research Project (SCCWRP) trawl station C-10
(33°31.5’N, 118°13.7'W; 457 m; July 1976); 4 specimens from another A. /on-

164 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

Fig. 1. Veneriserva pygoclava n. sp.—a, anterior end, dorsal view; b, maxillae I and carriers; c,
mandibles; d, median parapodium; e, seta; f, pygidium, dorsal view. (Scale a, f = 0.25 mm; e = 0.04
mm; b = 0.05 mm; d = 0.1 mm; c = 0.01 mm.)

gipalpa collected at SCCWRP trawl station L-1500 (33°30.5’N, 117°47.7'W; 457
m; Nov. 1976).

Description.— Length of holotype 18 mm, width maximally 1 mm, with 205
segments; largest paratype (2) 35 mm long with 400 segments; colorless in alcohol.
Prostomium (Fig. la) irregularly rounded, ventrally flattened; eyes absent. Peri-
stomial segment | short, fused dorsally with prostomium; second peristomial
segment a complete ring, three times longer than the first. Maxillae I (Fig. 1b)
smooth, thick and distally falcate. Maxillary carriers sigmoidally curved (Fig. 1b).
Fused mandibles small, longer than wide (Fig. 1c). Parapodia (Fig. 1d) each with
2 acicula. Setae with slightly curved shafts and a subdistally inflated, minutely
spinous boss terminating in a digitiform tip (Fig. le); number 4—5 per parapodium.
Body tapering posteriorly, terminating in a small club shaped pygidium (Fig. 1f).

Remarks.—This record is the first report of an iphitimid as a polychaete en-
doparasite. It is also the first record of an aphroditid parasitized by another
polychaete. The Iphitimidae, as originally defined, were exclusively associated

RESEARCH NOTES 165

either in the branchial cavity (5 species) or under the tail (1 species) of decapod
custaceans (Pilger 1971; Kirkegaard 1977). Previous records of polychaete en-
doparasitism show that only 13 species of arabellids parasitize other polychaetes.
The known host families include the cirratulids, eunicids, nereids, onuphids,
terebellids, trichobranchiids, serpulids, spionids, and syllids (Clark 1956; Petti-
bone 1957; Eliason 1962; Emerson 1974; Amaral 1977).

The parasitism of one polychaete by another is a poorly known phenomenon
due in part to its rare occurrence. Direct observations on such fundamental ques-
tions as to methods of reproduction, feeding, and host entry are apparently not
reported in the literature. Little can be inferred regarding these questions as most
parasitic arabellids show little or no structural modifications when compared to
free living species. A few species, however, exhibit some morphological modifi-
cations. The most aberrent parasitic arabellid, Haematocleptes terebelledis Wiren,
shows considerably morphological convergence with Veneriserva pygoclava. For
example, the mandibles and maxillary apparatus are extremely reduced, and the
reduced parapodia of Veneriserva have minute setae while those of Haemato-
cleptes lack emergent setae. Basically these species exhibit the reduction or loss
of mouthparts and locomotive structures usually associated with the adoption of
an endoparasitic mode of existence.

Sexually mature parasitic arabellids have not been recorded in the literature
(Clark 1956). This and the seeming lack of morphological modification has led
some observers to note that most endoparasitic arabellids may represent transitory
stages in the development of free living adults (Clark 1956; Emerson 1974). The
presence of gametes in V. pygoclava (eggs observed in one specimen) coupled with
obvious morphological adaptations suggests a more obligate form of parasitism.

Etymology.—The specific name (pygé, Greek = buttocks; clava, Latin = club)
refers to the shape of the pygidium.

Ecology. — The method of host penetration is unknown. V. pygoclava were found
in the ventral and lateral coelomic spaces of the hosts. No penetration of either
the gut or gut diverticulae was observed. The method of feeding is unknown.

Acknowledgments

I would like to thank Leslie Harris of SCCWRP for the loan of Aphrodita
longipalpa specimens. I would also like to thank Jerry D. Kudenov and Donald
J. Reish for their criticism of the original manuscript. I am indebted to Kristian
Fauchald for valuable comments and taxonomic discussions.

Literature Cited

Amaral, A.C. N. 1977. Um poliqueto endoparasita, Labrorostratus prolificus sp. nov. em nereideo.
Bolm Inst. oceanogr., Sao Paulo, 26:285-—292.

Clark, R. B. 1956. Capitella capitata as a commensal, with a bibliography of parasitism and com-
mensalism in the polychaetes. Ann. Mag. Nat. Hist., ser. 12, 9:433-448.

Emerson, R. R. 1974. A new species of polychaetous annelid (Arabellidae) parasitic in Diopatra
ornata (Onuphidae) from southern California. Bull. Southern California Acad. Sci., 73:1-5.

Eliason, A. 1962. Die Polychaeten der Skagerak-Expedition 1933. Zool. Bijdr, (Uppsala), 33:207-
293.

Fauchald, K. 1970. Polychaetous annelids of the families Eunicidae, Lumbrineridae, Iphitimidae,
Arabellidae, Lysaretidae and Dorvilleidae from western Mexico. Allan Hancock Monogr. Mar.
Biol., 5:1-335.

166 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

Kirkegaard, J. B. 1977. A new species of [phitime (Polychaeta: Iphitimidae) living under the tail of
Hyas (Crustacea: Decapoda) in the Oslo fjord. Pp. 199-209 in Essays on polychaetous annelids
in memory of Dr. Olga Hartman. (D. J. Reish and K. Fauchald, eds.), Allan Hancock Found.,

Univ. Southern California Press, 604 pp.

Pettibone, M. 1957. Endoparasitic polychaetous annelids of the family Arabellidae with descriptions
of new species. Biol. Bull., 113:170-187.

Pilger, J. 1971. A new species of [phitime from Cancer antennarius (Crustacea: Decapoda). Bull.
Southern California Acad. Sci., 70:84-87.

Accepted for publication 4 October 1983.

Mark M. Rossi, Dale Straughan International, 7015 Marcelle Street, Paramount,
California 90723.

INDEX TO VOLUME 83

Anglonautilus catarinae n.sp., 43

Awbrey, Frank T., Stephen Leatherwood, Edward F. Mitchell, and William Rog-
ers: Nesting Green Sea Turtles (Chelonia mydas on Isla Clari6én, Islas Re-
villagigedos, Mexico, 69

Bernard, J. Laurens, see Stephen D. Cairns

Bivens, Margaret, see Kevin J. Collins

Bradley, David W., Ross E. Landry, and Charles T. Collins: The Use of Jack-
knife Confidence Intervals with the Richards Curve for Describing Avian
Growth Patterns, 133

Cairns, Stephen D., and J. Laurens Barnard: Redescription of Janaria mirabilis,
a Calcified Hydroid from the Eastern Pacific, 1

Cameron, R. Andrew: Two Species of Lytechhinus (Toxopneustidae: Echinoidea:
Echinodermata) are Completely Cross-Fertile, 154

Collins, Charles T., see David W. Bradley

Collins, Kevin J., Scott Ralston, Traci Filak, and Margaret Bivens: The Suscep-
tibility of Oxyjulis californica to Attack by Ostracods on Three Sub-
strates, 53

Filak, Traci, see Kevin J. Collins
Fitch, John E., see David L. Stein
Fulmarus miocaenus n.sp., 84

Haaker, Peter L., David O. Parker, and Kristine C. Henderson: Observations of
Harbor Seal, Phoca vitulina Richardsi Feeding in Southern California waters,
152

Hannes, Gerald P., and'‘Susan M. Hannes: Plant Succession and Species Diversity
in the Marblehead Quarry, Ohio, 90

Hannes, Susan M., See Gerald P. Hannes

Henderson, Kristine C., see Peter L. Haaker

Hendrickx, M. E. and A. M. van der Heiden: Distributions of Seven Species of
Crustaceans Along the Pacific Coast of America, 110

Howard, Hildegarde: Additional Avian Records from the Miocene of Kern Coun-
ty, California, with the Description of a New Species of Fulmar (Aves: Pro-
cellariidae), 84

Keeley, Jon E.: Factors Affecting Germination of Chaparral Seeds, 113

Landrey, Ross E., see David W. Bradley

Leatherwood, Stephen, see Frank T. Awbrey

Lindberg, David R.: A Recent Specimen of Collisella edmitchelli from San Pedro,
California (Mollusca: Acmaeidae), 148

Miller, Walter B. and Richard L. Reeder: A New Species of Rabdotus (Gastopoda:
Pulmonata: Bulimulidae) from Arizona, 106

168 SOUTHERN CALIFORNIA ACADEMY OF SCIENCES

Paraliparis nassarum n.sp., 76
Parker, David O., see Peter L. Haaker

Rabdotus christenseni n.sp., 106

Ralston, Scott, see Kevin J. Collins

Reeder, Richard L., see Walter B. Miller

Rogers, William, see Frank T. Awbrey

Rossi, Mark M.: A New Genus and Species of Iphitimid Parasitic in an Aphroditid
(Polychaeta), with an Emendation of the Family Iphitimidae, 164

Schulman, Alisa: Humpback Whale (Megaptera novaeangliae) Sighting off Los
Angeles Harbor, Southern California, 158

Stein, David L., and John E. Fitch: Paraliparis nassarum n.sp. (Pisces, Lipari-
didae) from off Southern California with Description of Its Otoliths and
Others from North-East Pacific Liparidids, 76

Stewart, Brent S. and Pamela K. Yochem: Seasonal Abundance of Pinnipeds at
San Nicholas Island, California, 1980-1982, 121

Stewart, Joan G.: Algal Distributions and Temperature: Test of an Hypothesis
Based on Vegetative Growth Rates, 57

Sundberg, Frederick A.: Two Cretaceous Nautiloids from Baja California, Mexico,
and Southern California, 43

Thom, Ronald M.: Primary Production in Grays Harbor Estuary, Washington, 99

van der Heiden, A. M., see M. E. Hendrickx

Wicksten, Mary K.: Early Twentieth Century Records of Marine Decapod Crus-
taceans from Los Angeles and Orange Counties, California, 12

Yochem, Pamela K., see Brent S. Stewart

(a

The BULLETIN is published three times each year (April, August, and November) and includes articles in English
in any field of science with an emphasis on the southern California area. Manuscripts submitted for publication
should contain results of original research, embrace sound principles of scientific investigation, and present data in
a clear and concise manner. The current AIBS Style Manual for Biological Journals is recommended as a guide for
contributors. Consult also recent issues of the BULLETIN. Authors should strive for directness and lucidity, achieved
by use of the active voice. Special attention should be given to consistency in tense, unambiguous reference of
pronouns, and logically placed modifiers.

MANUSCRIPT PREPARATION

It is strongly recommended that, before submitting a paper, the author ask qualified persons to review it. The
author is requested to submit at /east two additional copies with the original, on 8¥2 X 11 opaque, nonerasable paper,
double spacing the entire manuscript. Do not break words at right-hand margin anywhere in the manuscript. Foot-
notes should be avoided. Manuscripts which do not conform to the style of the BULLETIN will be returned to the
author.

An abstract summarizing in concise terms the methods, findings, and implications discussed in the paper must
accompany a feature article. Abstract should not exceed 100 words.

A feature article comprises approximately five to thirty typewritten pages. Papers should usually be divided into
the following sections: abstract, introduction, methods, results, discussion and conclusions, acknowledgments, and
literature cited. Avoid using more than two levels of subheadings.

A research note is usually one to six typewritten pages and rarely utilizes subheadings. Consult a recent issue of
the BULLETIN for the format of notes. Abstracts are not used for notes.

Abbreviations: Use of abbreviations and symbols can be determined by inspection of a recent issue of the
BULLETIN. Omit periods after standard abbreviations: 1.2 mm, 2 km, 30 cm, but Figs. 1-2. Use numerals before
units of measurements: 5 ml, but nine spines (10 or numbers above, such as 13 spines). The metric system of

weights and measurements should be used wherever possible.

Taxonomic procedures: Authors are advised to adhere to the taxonomic procedures as outlined in the International
Code of Botanical Nomenclature (Lawjouw et al. 1956), the International Code of Nomenclature of Bacteria and
Viruses (Buchanan et al. 1958), and the International Code of Zoological Nomenclature (Stoll et al. 1961). Special
attention should be given to the description of new taxa, designation of holotype, etc. Reference to new taxa in
titles and abstract should be avoided.

The literature cited: Entries for books and articles should take these forms.

McWilliams, K. L. 1970. Insect mimicry. Academic Press, vii + 326 pp.

Holmes, T. Jr., and S. Speak. 1971. Reproductive biology of Myotis lucifugus. J. Mamm., 54:452-458.

Brattstrom, B. H. 1969. The Condor in California. Pp. 369-382 in Vertebrates of California. (S. E. Payne, ed.),

Univ. California Press, xii + 635 pp.

Tables and figures (line drawings, graphs, or black and white photographs) should not repeat data contained in
the text. The author must provide numbers and short legends for tables and figures and place reference to each of
them in the text. Legends should be typed on a separate sheet of paper and placed at the end of the manuscript.
Illustrations and lettering thereon should be of sufficient size and clarity to permit reduction to standard page size;
ordinarily they should be no more than twice the size of intended reduction and should not exceed 8% by 11 inches
in size. Photographs must be printed on glossy paper. Submit one photoduplicated copy of each illustration. All
illustrations accompanying Research Notes will be reduced to one column width. All half-tone illustrations will
have light screen (grey) backgrounds. Special handling such as dropout half-tones, special screens, etc., must be
requested by and will be charged to authors.

A cover illustration pertaining to an article in the issue or one of general scientific interest will be printed on the
cover of each issue. Such illustrations along with a brief caption should be sent to the Editor for review.

PROCEDURE

All manuscripts should be submitted to the Technical Editor, Jon E. Keeley, Biology Department, Occidental
College, 1600 Campus Road, Los Angeles, California 90041. Evaluation of a paper submitted to the BULLETIN
begins with a critical reading by the Editor; several referees also check the paper for scientific content, originality,
and clarity of presentation. Judgments as to the acceptability of the paper and suggestions for enhancing it are sent
to the author at which time he or she may be requested to rework portions of the paper considering these recom-
mendations. The paper then is resubmitted and may be re-evaluated before final acceptance. Authors are requested :
to submit the names, addresses and specialities of three persons who are capable of reviewing the manuscript.

Proof: The galley proof and manuscript, as well as reprint order blanks, will be sent to the author. He or she
should promptly and carefully read the proof sheets for errors and omissions in text, tables, illustrations, legends,
and bibliographical references. He or she marks corrections on the galley (copy editing and proof procedures in
Style Manual) and promptly returns both galley and manuscript to the Editor. Manuscripts and original illustrations
will not be returned unless requested at this time. All changes in galley proof attributable to the author (misspellings,
inconsistent abbreviations, deviations from style, etc.) will be charged to the author. Reprint orders are placed with
the printer, not the Editor.

CONTENTS

Factors Affecting Germination of Chaparral Seeds. By Jon E. Keeley

Seasonal Abundance of Pinnipeds at San Nicolas Island, California, 1980-
1982. By Brent S. Stewart and Pamela K. Yochem

The Use of Jackknife Confidence Intervals with the Richards Curve for:
Describing Avian Growth Patterns. By David W. Bradley, Ross E. -
Landry, and Charles T. Collins

A Recent Specimen of Collisella edmitchelli from San Pedro, California
(Mollusca: Acmaeidae). By David R. Lindberg

Research Notes

Observations of Harbor Seal, Phoca vitulina Richardsi Feeding in Southern California
Waters. By Peter L. Haaker, David O. Parker, and Kristine C. Henderson

Two species of Lytechinus (Toxopneustidae: Echinoidea: Echinodermata) are Completely Cross-
fertile. By R. Andrew Cameron

Humpback Whale (Megaptera novaeangliae) Sighting off Los Angeles Harbor, Southern Cal-
ifornia. By Alisa Schulman

A New Genus and Species of Iphitimid Parasitic in an Aphroditid (Polychaeta), with an
Emendation of the Family Iphitimidae. By Mark M. Rossi

DTG GX cat AN aes Oe Ae a Mean ns Unt ae Sees Bd Seay tare eer eee Se ee ae

COVER: Postfire skeletal remains of manzanita, Arctostaphylos glauca. Photo by Jon E. Keeley.