USE OF PHOTOGRAPHIC IDENTIFICATION IN CAPTURE-RECAPTURE STUDIES OF MEDITERRANEAN MONK SEALS

MARINE MAMMAL SCIENCE, 16(4):767-793 (October 2000) 0 2000 by the Society for Marine Mammalogy

USE OF PHOTOGRAPHIC IDENTIFICATION IN

MEDITERRANEAN MONK SEALS CAPTURE-RECAPTURE STUDIES OF

JAUME FORCADA~ ALEX AGUILAR

Department of Animal Biology, Faculty of Biology,

University of Barcelona, Diagonal 645, E-08028 Barcelona, Spain

E-mail: [email protected]

ABSTRACT

The use of photo-identification and its reliability in capture-recapture studies of Mediterranean monk seals were assessed using slides collected in the colony at Cap Blanc, western Sahara, from 1993 to 1996. Five tests indicated that researchers involved in photo-identification were proficient in matching slides of identified seals, consistent in classifying the side of the seal shown in slides and in assigning the morphological stage of the seal, and that changes of markings over a period of three years were insufficient to affect matching success. The certainty of identifying a seal was not dependent on the number of slides used but on distinctiveness of the markings and the quality of the slides taken. Capture-recapture abundance estimates were biased upwards when including poor quality slides. The exclusive use of excellent- and good- quality slides provided the best estimates. The proportion of distinctive seals varied between morphological stages and was significantly lower in juveniles. When including the identification histories of juveniles, the heterogeneity of capture probabilities was higher. Therefore, abundance estimates were less biased when all juveniles were considered as non-distinctive seals. Reliable abundance estimates required a balance between duration of capture occasions and time interval between these.

Key words: abundance estimate, Cap Blanc, capture-recapture, field techniques, Mediterranean monk seals, Monacbns monacbus, photo-identification, western Sahara.

The Mediterranean monk seal (Monachus monachtls) is a phocid species with subtropical distribution. The present range extends from the Black Sea, in the east, to the Atlantic coast of Northwest Africa. This range encompasses the

Current address: National Marine Fisheries Service, Southwest Fisheries Science Center, P. 0. Box 271, La Jolla, California 92038, U.S.A.

767

768 MARINE MAMMAL SCIENCE. VOL. 16. NO. 4. 2000

historical distribution areas of the species (Kenyon 198l), although numbers have been greatly reduced. The world population is estimated to be 415-615 animals, and a number of local populations appear to be at risk of extinction in the next 60 yr (Brasseur et al. 1997). Therefore, there is a need to implement monitoring programs within the framework of conservation management strategies. The assessment of the population status is an essential part of these programs and precise and accurate estimates of abundance are critical for the development of these strategies.

Photo-identification techniques aimed at obtaining individual marking data of marine mammals have been developed (e.g., Wiirsig and Wiirsig 1977, Katona and Krauss 1979, Hiby and Lovell 1990, Beck and Reid 1995) and successfully applied to a number of species and populations (e.g., Bigg 1982, Clapham and Mayo 1990, Jones 1990, Yochem et al. 1990). These techniques avoid handling individuals, and thus they are convenient when animals are difficult to capture or when non-invasive techniques are required. They are particularly useful when working on endangered species and can be used to produce capture-recapture data to estimate population parameters. IWC (1990) provides a comprehensive review of the applications of photo-identification to population, behavioral, and biological studies in marine mammals.

The Mediterranean monk seal has short fur which allows ready identification of a large subset of the population through their natural markings (Marches- saux 1989). These marks, scars, and pelage patterns unique to individuals, are apparent because the fur is short. Photo-identification has been used to identify individuals in different regions of the species range: cg., the Sporadas Islands and Zakynthos, Greece (Hiby and Jeffery 1987), the Cap Blanc peninsula, western Sahara (Marchessaux 1989, Cantos et al. 1998), and the Madeira Is- lands (Beudels 1991). However, not any of these studies have carried out proper assessment of the reliability of the technique.

In some cases capture-recapture data derived from photo-identification may be biased and may not be easily applied unless carefully designed and validated (e.g., Hammond 1990, Agler 1992). The most critical assumptions underlying capture-recapture studies are that the involved population is often geograph- ically closed, sometimes even demographically closed, that all individuals have equal capture probabilities, that all observed marks are reported, and that marks are permanent over a given study period (IWC 1990). These assumptions may not be met, for example, when marks change over time (Carlson e t a/. 1990), the ability of observers to compare sets of photographs varies, or the sampling coverage of the species’ range is insufficient (Agler 1992). Pre- vious studies indicate that the reliability of photo-identification methods de- pends on the species involved, the sampling strategy, and the kind of natural marks used to identify individuals (Arnbom 1987, Hiby and Lovell 1990).

This paper addresses the reliability of using photo-identification data obtained from 1993 to 1996 in the Cap Blanc peninsula (western Sahara). The efficiency of the photographers and researchers who compared and cataloged identification slides, the proportion of identifiable individuals, and the representativeness of photo-identification surveys were all evaluated. The influence

FORCADA AND AGUILAR: MONK SEALS 769

of photographic quality and seal distinctiveness on capture-recapture abundance estimates was also evaluated using a representative subset of slides. We addressed photographic quality and distinctiveness in terms of obtaining ho- mogeneity of capture probabilities. We associated different distinctiveness to the fact that some individuals were hard to identify because they had few natural marks. Thus, distinctiveness was a quality of the seal. Photographic quality referred to the quality of photo-identification slides.

We provide abundance estimates for the colony at Cap Blanc in 1995, as an example of the use of our methods. However, this is not an objective of this article, since the population analyses are presented elsewhere (Forcada et al. 1999).

METHODS

Photographic Collection and Cataloging

The photo-identification material was collected from the colony at the Cap Blanc Peninsula, located in the southern fringe of the western Sahara (20'40'- 21"20'N, 17"00'-17°20'W) (Fig. 1) between March 1993 and May 1996. During the study period the population was composed of at least 228 individuals (those identified photographically). In population counts and identification fieldwork, seals were assigned to three different morphological stages: young seals, medium and large gray seals, and adult black males (GonzPlez et 61. 1997).

Photo-identification surveys were conducted each year during spring and summer. Annual sampling periods selected for the capture-recapture analysis to estimate abundance comprised between 16 and 34 effective field days: 16 d in May-June 1993, 30 d in March-May 1994, 32 d in April-June 1995, and 34 d in April-July 1996. In each survey, seals were approached from the cliff tops where the main haul-out areas were located. Most of these areas are narrow beaches inside caves, which have an open entrance to the sea. Color slides (ISO@ 100 and 200) of individual monk seals were taken with 35-mm cameras using telephoto lenses.

In each survey photographers attempted to obtain complete sequences of photographs from all views of each swimming seal, targeting identifying features such as scars, wounds, and coloration patterns. At the laboratory, five different views of each seal, coded A, B, C, D, and X (Fig. 2) were finally selected as the most useful to characterize and thus identify individuals. How- ever, the right side of the head, view B, was the most frequently photographed and a high proportion of good or excellent quality slides were obtained from this view. This view provided enough identifying features for unequivocal recognition of seals. Thus, it was treated as the principal identification view. Each slide was coded according to its photographic quality (Q) and distinctiveness (D) (Fig. 3). The quality of slides was evaluated as excellent (l), good (2), or poor (3) according to the focus, glare, angle and distance from the seal, and the proportion of the individual photographed, as proposed by Hammond

770 MARINE MAMMAL SCIENCE. VOL. 16. NO. 4, 2000

17' 00'

Tarf el Cuerguerat

[ Zone 4

Breeding caves #1 and #3

CG 5 B

P 3

:a 2

Cap Blanc

Figure 1. Map of Cap Blanc Peninsula, western Sahara, showing caves, beaches, and hauling-out areas used by monk seals.

(1986). Distinctiveness, a quality of the seal, was reflected by the degree of visibility of permanent marks and was evaluated according to the ease of individual identification. Seal identification was then graded as unquestionable (l), certain (2), or uncertain (3). A 1 was given to seals with big and numerous marks, or with a single or few very characteristic marks; a 2 was given to seals

FORCADA AND AGUILAR: MONK SEALS 77 1

A

D

Figure 2. fication slides.

Principal views used in identification of individuals from photo-identi-

with numerous distinctive marks; a 3 was given to seals with only a few small marks. The percentage of good and excellent slides for each view was between 70 and 82. The percentage of slides coded with unquestionable and certain distinctiveness was between 64 and 80. This percentage represented 228 identified seals.

Well-focused and clear slides of seals from each survey were selected to create a catalog of type specimens. After completing slide examination from each new survey, previously unidentified individuals were assigned a new code and the catalog was updated with the most recent and best slides of all the individuals. An individual was assumed identifiable as soon as it acquired enough distinctive marks to become a code 2 of distinctiveness, and its history in the catalog started from that point. If it was unidentifiable in a previous survey (e.g., distinctiveness 31, it was not considered captured (or marked and released) for the first time in the survey. To minimize missing identifications, matches were confirmed by two or more experienced researchers and non- matched slides were viewed against the catalog at least three times. The proportion of seals in the catalog not properly identified, and included as different individuals in subsequent catalog updates (missing matches) was analyzed by annual reviews, which were conducted by three researchers. Based on these reviews, we calculated the percentage of duplicate entries of individuals ( i e . , proportion of individuals entered as different animals), with respect to the total number of seals identified in a year. In this way we were able to address the significance of a potential bias in abundance estimates.

Reliability test-This test was designed to evaluate the usefulness of photo- identification to identify the same individual using natural marks, and to identify potential sources of error in the ability to recapture a photo-identified

772 MARINE MAMMAL SCIENCE, VOL. 16, NO. 4, 2000

Figwe 3 . Photographs showing seals of different distinctiveness (D1 : unquestionable; D2: certain; D3: uncertain), and slides of different photographic qualities (41: excellent; 4 2 : good; 43: poor).

seal. We selected 14 seals tagged with plastic tags (for details on tagging see Gazo et al. 2000), and created two samples of photo-identification slides. In the first sample (I), there was a slide of each of the 14 tagged seals when they were first captured, and 14 slides of other different seals. In the second sample (11), there were 42 slides of the 14 tagged seals, corresponding to all the photographic recaptures from 1993 to 1996. Because each of these slides was associated with a seal also recaptured by reading a tag, recaptures were known to correspond to captured seals. In sample I, seals were captured when they


were youngsters (less than 1 yr old) or juveniles (from 1 to 1.5 yr old); i,e., each seal appeared as when captured for the first time. In sample 11, seals were recaptured at various stages: youngsters, juveniles, or medium-sized seals (small adult seals), and seals recaptured from one to six times, over different periods of time, and different morphological stages. Thus, potential covariates affecting photo-identification were morphological stage, number of recaptures, proportion of times recaptured (i.e., ratio of times recaptured to number of total possible recaptures in the sample), number of morphological stage changes since first captured (e.g., a seal captured as a youngster and recaptured as a medium-sized seal went through two changes), and interval of time in months between capture and recapture.

A researcher usually involved in monk seal photo-identification was given sample 11, where every slide was unassociated with the other slides in the sample. Thus, he could not know that some slides corresponded to the same seal. The researcher then compared each slide in sample I1 to sample I, to look for matches. The researcher did not receive any information on the seals represented by the slides. We calculated the proportion of correct matches, and used generalized additive models of the binomial family (Hastie and Tibshirani 1990) in a logistic regression analysis. The predictors were the five covariates listed above, treated as factors, except number of recaptures, proportion of times recaptured, and interval of time, that were treated as linear or smooth terms, using cubic smoothing splines. The response variable was a correct match (1) or an incorrect match (0). We used a stepwise selection method, based on the analysis of deviance and the AIC (Chambers and Hastie 1992) in order to select the best model. The degrees of freedom in the smooth terms were selected by cross validation. We used the best model to assess the effect of covariates relevant to the photo-identification process.

Test of the researchers’ ability t o identify seals-This test had two objectives: (1) to analyze the ability of researchers to match and catalog slides, and (2) to assess the effectiveness of different views to characterize individuals. For the test, we created an experimental catalog (catalog A) and an experimental sample of slides (sample A). Catalog A contained 20 individuals, each represented by five different high-quality slides corresponding to views A, B, C, D, and X. Sample A was composed of 126 slides from 24 different seals, and 15 of these seals were also included in catalog A, while nine were not. This sample was organized into seven different groups of six individuals, and each individual within a group was represented by one to five slides of the same view. Because there were only five different views, two of the groups contained six individuals represented by a combination of the five views selected at random.

Three researchers usually involved in monk seal photo-identification examined sample A and were asked to compare it with the catalog A. When producing a match they were required to assign the appropriate identification code. When an animal was not found in the catalog, they assigned it a new code. In this way they followed the same routine of identifying and cataloging seals as in the laboratory.

Researchers also assigned a morphological stage to the seals shown in the


slides and identified the view (from A to X) shown. To do this, catalog A contained the same proportion of slides of the three main morphological stages as the main catalog: juveniles, gray seals (medium and large sized gray seals) and black males. These stages follow the classification by Gonzalez et al. (1997).

Scores of correct matches of each of the seven groups from the subsample were compared and ranked for each researcher to determine the views in which identification was more reliable and frequent. We used the kappa statistic (Agresti 1990) to measure the agreement between the scores obtained by different observers, and also between each observer and the original scoring (from the catalog). In the kappa statistic, the difference between the observed proportion of cases in which the observers agree and the proportion expected by chance is divided by the maximum difference possible between the observed and expected proportions, given the marginal totals. A value of 1 indicates perfect agreement. A value of 0 indicates that agreement is no better than chance (Agresti 1990). The agreement between the classification values by morphological stages and sides given by researchers during the test, and the actual values given for the real catalog was also determined by the kappa statistic.

Minimum number of slides t o ensure identijcation-Some species of marine mammals (e.g., bowhead whales) require a minimum number of photographs of each individual to be sure that an identifiable individual has not been overlooked (Rugh et al. 1992). A test was designed to ascertain whether this minimum number of slides was also required for Mediterranean monk seals.

We selected a group of 50 seals with many photo-identification slides of view B. We selected five slides from each of these individuals to create an experimental catalog (catalog A2). We also created an experimental “field sample” (sample A2) with the following composition: 10 individuals with 5 slides, 10 with 4 , 10 with 3, 10 with 2 and 10 with 1. All the slides comprising sample A2 were randomly mixed and one of the most experienced researchers examined and compared them with catalog A2. We then calculated the proportion of correct matches obtained for each of the groups with an equal number of slides. These proportions were plotted +gainst the number of identification slides per individual to examine differences among groups. We tested if the probability of matching correctly all the seals was equal for each of the five groups. We used a Pearson’s chi-square statistic to perform a test of proportions, under the null hypothesis that the probability of matching correctly all the seals independent of the number of slides was equal for all groups.

Permanence of Marks

Mediterranean monk seals have variable amounts and types of natural marks depending on the age and sex of individuals. This variability, and the change of marks over time, are likely to make poorly marked seals difficult to reidentify. This may lead to the overestimation of population size in capture-recapture experiments (Hammond 1986). A test was carried out by three researchers


to assess the effect of the change of marks on matching success. Seventeen individually identified seals seen in three consecutive years after first capture (from 1993 to 1996) were selected. Two slides of each seal, one from 1993 and the other from 1996, were selected and were mixed together. Researchers were asked to again match the individuals, and we estimated their proportions of correct matches.

Assessment of Photographic Quality

Photographic quality and its ejfict on capture-recapture abundance estimates- Juveniles were less likely to be identified because their marks were subtle and only became apparent with good quality pictures. We evaluated the proportion of slides with poor quality pictures in each morphological group to identify groups poorly represented in the photo-identification sampling. We selected a sample (sample Bl) comprising 200 slides of view B, 115 of which were from medium- to large-sized gray seals (called gray seals hereafter), 35 from juveniles, and 50 from adult black males (black males hereafter). These proportions of morphological stages matched those of the general photo-identification catalog. Proportions were estimated as the ratio of quality three slides to the total number of slides in each group.

The use of poor quality slides is likely to decrease the ability to match identified seals. This effect, which is comparable to misreading a tag on a marked animal, biases capture-recapture abundance estimates to some extent (Hammond 1986). A simple population analysis was conducted to investigate whether such an effect occurred in the monk seal data sets. In this analysis, capture-recapture abundance estimates were produced from data from all morphological stages, selected from surveys conducted in 1995 (Table l) , and distributed in three different sets. Each data set comprised identifications from slides of varying stages of photographic qualities (Q): Ql, and Ql-3, from all morphologic stages. Abundance was estimated independently for each data set using Chapman’s modified Lincoln-Petersen index (Chapman 195 1)

(MI + N n , + 1) (m, + 1)

where MI is the number of marked seals in the population just before the ith sample is taken, n, is the size of the ith sample and m, is the number of marked individuals in a,. The average abundance estimate (Chapman 1952) was obtained from the equation

- 1 N, =

where s was 3, the number of surveys conducted in 1995. We selected Chap- man’s index to simplify this analysis. We decided to use more robust estimators to produce more reliable abundance estimates (see below). Variance was estimated as in Seber (1982).

Tabl

e 1.

M

odel

sel

ectio

n pr

oces

s of

log

istic

regr

essi

on a

naly

sis

with

dat

a fr

om r

elia

bilit

y te

st. I

nitia

l mod

el was a

mod

el o

f ide

ntif

icat

ion

as re

spon

se, a

nd n

umbe

r of r

ecap

ture

s, p

ropo

rtio

n of

tim

es re

capt

ured

, mor

phol

ogic

al st

age,

dur

atio

n of

inte

rval

bet

wee

n ca

ptur

e and

reca

ptur

e,

and

chan

ge in

mor

phol

ogic

al g

roup

as

linea

r pre

dict

ors.

Tab

le s

how

s su

mm

ary

of r

esul

ts o

f an

alys

is o

f de

vian

ce, w

here

ter

ms

are

chan

ged

or

elim

inat

ed in

ste

pwis

e se

lect

ion

proc

ess.

Fina

l te

rm i

s fa

ctor

or

cova

riate

as

reta

ined

by

final

mod

el, e

ither

lin

ear o

r no

n-lin

ear t

erm

. Sim

ple

nam

es o

f co

varia

tes

mod

eled

as

linea

r ter

ms,

and

s(c

ovar

iate

) ind

icat

ed s

moo

th te

rm.

Res

idua

l R

esid

ual

Initi

al t

erm

Fi

nal t

erm

df

D

evia

nce

df

devi

ance

A

IC

Full

linea

r mod

el

34

27.8

4 41

.8

Inte

rval

s(

1nte

rval

) -2

.9

-6.3

28

.4

8.79

34

.0

Mor

phol

ogic

al st

age

1.92

2.

31

30.3

1 11

.09

32.5

C

hang

e of

sta

ge

1.03

0.

67

31.3

5 11

.76

31.1

Prop

ortio

n s(

Prop

ortio

n)

-2.7

-1

2.75

31

.27

15.0

8 34

.5


The best estimate, considered to be that of greatest accuracy and precision, was selected by the minimum residual sum of squares (RSS), a procedure recommended by Burnham et al. (1995) to assess the best fit of models to the data. We used the statistic

where fi is the estimated abundance and N, is the true population size estimated from the data set of excellent photographic quality (el), and s is the number of surveys.

Distinctiveness of individuals and its $fict on capture-recapture abundznce estimates-Some studies have shown that the ability to identify an individual with low distinctiveness decreases with decreasing photographic quality. Con- versely, highly distinctive individuals can be recognized with poor quality pictures (Arnbom 1987, Friday 1997). This leads to the potential mismatching of individuals. In order to address this question we determined the correlation between distinctiveness codes and photographic quality obtained from the slides in sample B1, using the Pearson’s product moment correlation test.

Recapture rates of individuals from sample B1 were estimated and compared between morphological stages. These rates were obtained from the identification (capture) histories of seals. They were calculated as the number of times a seal had been resighted divided by the number of surveys in which it could potentially be resighted. To avoid confounding with emigration, recapture rates were estimated from five capture occasions with a duration of eight days each, conducted in 1994. During this period, the individuals involved in the experiment did not change between morphological stages. Recapture rates were also regressed against predicted distinctiveness for each morphological stage. In this analysis, a significant trend was interpreted as an association between distinctiveness and the ability to resight a given seal.

Proportion of distinctive seals in the population-We estimated the mean proportion of distinctive seals from individuals with good or excellent quality slides. A seal was considered to be distinctive when it had visible and durable marks. The proportion of distinctive seals in the ith survey (d,) was estimated as the number of individuals with view B graded with certainty 1 or 2, divided by all the individuals captured (every captured seal was represented by a single slide). We carried out a series of consecutive surveys, each considered as an independent sample, and calculated from them a mean for the proportion of distinctive seals in the population (a. The variance of d was estimated as

and was considered to be unbiased, assuming that d, was independent between consecutive surveys. This parameter was further used to scale up capture-


recapture abundance estimates correcting for the proportion of non-distinctive seals. A simple corrected estimator of abundance was derived as

where fid is the estimated abundance of the distinctive population. The variance of the corrected abundance was estimated using the Delta method (e.g., Seber 1982) as

h

Ability of obseweys to grade photographic quality and seal distinctiveness-An experiment was conducted to test the consistency of researchers in grading a sample of slides for quality (Q) and distinctiveness (D). Researchers were asked to grade 126 slides of seals (sample A) which had already been classified before their inclusion in the catalog. The agreement between the grading of observers and those of the catalog was tested with the kappa statistic. We also estimated the correlation between the quality and distinctiveness codes given by the different observers.

Assessment of Sampling Representativeness and EfFciency

An insufficient sampling effort may cause some photo-identification surveys to have data sets which are poorly representative of the sampled population. If good-quality slides are not obtained, all the marked seals present may not be reported. This can cause a positive bias in abundance estimates (Otis et al. 1978). On the other hand, low sample sizes lead to imprecise abundance estimates. Therefore, we investigated the quality of our data in terms of pro- viding capture histories in multisample capture-recapture experiments.

We estimated the increase in average capture probabilities for an increasing number of consecutive sampling days. Capture probabilities were expressed as the proportion of identified seals subsequently recaptured (Pollock et al. 1990) from data collected in 1995 at Cap Blanc. Similarly, we estimated the opti- mum number of days between different capture occasions that allowed some turnover of seals at the haul-outs. Capture probabilities were estimated as above, as the proportion of identified seals that were recaptured on a certain day of an interval. But, in this case, capture probabilities were estimated at intervals of 1-15 d between the capture and recapture. In both estimates, the mean probability was obtained from 15 replicated estimates of capture probabilities. In each replicate, capture probabilities at each interval were averaged over the 15 replicates.

We further combined capture periods and intervals between periods of op- timum durations, as decided by the above methods, to produce abundance estimates. The estimates were produced with Chao’s Mrh estimator (Chao et al. 1992), which is robust to heterogeneity and time variation in capture proba-

FORCADA A N D AGUILAR: MONK SEALS 779

bilities. We selected the maximum number of capture occasions according to the maximum capture probabilities and minimum optimal interval duration between these.

RESULTS

Photographic CoZZection and Cataloging

At the end of the study period the catalog contained 228 seals, 30 of which corresponded to juveniles, 163 to gray seals, and 35 to black males. Duplicated entries (false negative matching errors) detected in annual reviews accounted for 4.3% of the total number of seals in the catalog. The minimum interval between the entry dates of duplicate slides was one year.

Reliability test-The results indicated that photo-identification was valid in capture-recapture experiments, especially for juveniles and older stages. The effect of time between recaptures in matching success was significant only in the youngest seals, and seals more easily identified were those recaptured more frequently. In total, 30 slides from the 42 (71%) of sample I1 were matched correctly with slides from sample I. All seals in sample I were correctly matched at least once with recapture slides in sample 11. Of the slides incor- rectly coded, all were youngsters or juveniles of less than two years of age. This is consistent with fewer natural marks in these stages. The logistic regression analysis indicated that in the best model, in terms of AIC, the response was a function of the interval of time between capture and recapture as a smooth term, the number of recaptures as a linear term, and the proportion of times recaptured as a smooth term (Table 1). The same model including morphological stage as a linear term had a similar AIC value, indicating that this effect was also significant. That is, younger seals were less often recaptured. Interval and morphological stage were not accepted as an interaction term by the modeling process. However, the captures and recaptures of older seals had longer intervals in between, and in all cases individuals were correctly identified. The proportion of times recaptured was higher for individuals with correct matches which, in general, were older seals.

Test for the researchers’ ability to identify seals-All three researchers made 100% correct matches and identified 100% of the individuals which were new to the catalog. The time employed to complete the test was 160 and 175 min for the two most experienced researchers and five hours for the less experienced. We did not observe differences in scores of correct matches for the different groups of individuals, thus we conclude that each of the five sides may be used as a main identification view.

Kappa values, expressing the agreement between the actual views and morphological stages shown in the slides and those that were assigned by the researchers during the test (Table 2 ) , were similar for the three researchers, indicating substantial or almost perfect agreement in views and fair agreement in morphological stages, according to the classification of the kappa statistic values of Landis and Koch (1977). This suggests that experience did not have


Table 2. Agreement between actual seal view and morphological stages as classified in photo-identification catalog, and classification by each observer during test of researchers’ ability to classify slides. Kappa statistic values used as measure of agreement.

Morphological View of Researchers stages the seal

A (experienced) K = 0.35 K = 0.82 B (experienced) K = 0.69 K = 0.76 C (less experience) K = 0.45 K = 0.78

a significant effect on the identification of views or morphological stage. The low agreement found in morphological stage classification was mainly due to the difficulties of telling medium-sized seals from large gray seals from the slides, and not being able to appreciate the real size of the seal as in the field.

Minimum number of slides to ensure identifiation-The researcher produced 100% correct matches for the groups with a variable number of identification slides of each individual. These results suggest no differences in the probability of identifying an individual depending on the number of photo-identification slides available. Consequently, provided that slides are of good or excellent quality, photo-identification of a given seal should be possible with only a single slide of view B.

Permanence of Marks

The three researchers who undertook the test correctly grouped all 17 pairs of slides corresponding to individuals taken in different years. This indicates that the effect of changes in the marks of the seals over a period of three years was not enough to preclude recognition by experienced researchers. Therefore, the probability of detecting a previously captured seal in the photo-identification catalog is considered to be high.

Assessment of Photographic Quality

Photographic quality and its &$ct on capture-recapture abundance estimates- Table 3 shows the estimated proportion of poor-quality slides by morpholog-

Table 3. Proportion of slides with poor photographic quality by morphological group, from sample of slides representative of catalog of photo-identified seals. Pro- portions were significantly different ( P < 0.0001) according to Pearson’s x2 statistic to test differences in proportions.

Morphological stage Total Qs Q,ITotal

Juveniles 35 28 0.086 (0.047) Grey seals 115 115 0.200 (0.037) Black males 50 24 0.520 (0.071) All stages 200 167 0.260 (0.0312)


Table 4. Capture-recapture abundance estimates and residual sum of squares (RSS) when slides of different photographic quality discarded. Photographic quality codes: 1 = excellent, 2 = good, 3 = poor.

Photographic Sample Abundance Residual sum quality size (% CV) of squares 1 111 221 (16.20) - 1-2 185 249 (8.00) 0.008 1-3 210 266 (6.62) 0.020

ical stage derived from sample B1. Proportions were significantly different ( P < 0.0001) according to a Pearson’s x2 statistic to test differences in proportions. The proportion for black males was significantly higher than for the other stages, and that of juveniles significantly lower. Consequently, we accept that black males were poorly identified in the slides. A greater effort (in days) should be made in the field to better account for this stage. That is, increasing the number of days in the field would allow for more opportunities to obtain high-quality slides.

Capture-recapture abundance estimates decreased when lower-quality slides were omitted (Table 4), while the residual sum of squares (RSS) of the fitted models decreased as slides of poor quality were discarded. Moreover, the RSS of the estimate obtained from slides of good and excellent quality was very low compared with the full data set, which included poor-quality slides. Estimated abundance was 6% higher when using all the slides. This implies that maintaining a large sample size, while it increases precision, introduces a positive bias in the estimate. Therefore, the estimate obtained when using only good- and excellent-quality slides provided the best balance for the trade- off between bias and loss of precision. These results agree with those of a similar approach implemented by Friday (1997) in a humpback whale photo- identification study, where omitting poor-quality slides also increase accuracy in population estimates.

Distinctiveness of individuals and its gfict on capture-recapture abundance estimates-The codes of distinctiveness in sample B1 (Table 5) were independent of photographic quality, according to a Pearson’s product moment correlation

Table 5 . Observed and expected (in parentheses) number of slides of different photographic quality and distinctiveness in sample of 200 slides in experimental catalog (catalog B l), representative of Mediterranean monk seal photo-identification catalog.

Photographic quality 1 2 3 All

Distinctiveness 1 18 (16.9) 33 (32.1) 18 (20) 69 2 18 (17.6) 33 (33.5) 21 (20) 72

59 3 13 (14.5) 27 (27.5) 19 (17.1) All 49 93 58 200

782 MARINE MAMMAL SCIENCE, VOL. 16, NO. 4 , 2000

Table 6. Analysis of proportion of identifiable individuals from representative sample of photo-identified seals at colony of Cap Blanc, obtained in 1995. Standard errors given in parentheses. Values with asterisk significantly different with P < 0.05.

~~~~~ ~ ~~

Proportion of Morphological distinctive

stage seals (d) Juveniles 0.466 (0.066)* Gray seals 0.866 (0.089) Black males 0.895 (0.049) All stages 0.781 (0.061)

(P = 0.219). The test was one-tail (a positive Correlation), because there was no reason to think that seals having worst-quality slides would be classified as less distinctive than the rest.

Proportion of distinctive seals in the population-According to the estimated proportions of distinctive seals by stages, juveniles were significantly (P < 0.05) less distinctive than other seals (Table 6). This indicates that their identification is less likely to be reliable.

The estimated abundance of Mediterranean monk seals for 1995, using good- and excellent-quality slides (Table 4), was corrected for poorly marked seals. The selected correction factor accounted for non-distinctive seals from all morphological stages (Table 6) and was 0.781 (SE = 0.061). With this factor, abundance was estimated as 319 individuals (CV = 0.112). Log-normal 95% confidence intervals were estimated as 265 and 409 for the low and high limits respectively.

From the five surveys conducted in 1994, mean recapture rates were estimated as 0.36 (SE = 0.036) for juveniles, 0.55 (SE = 0.021) for gray seals, and 0.67 (SE = 0.036) for black males. These rates were significantly different across morphological stages (ANOVA, F = 19.45, P C 0.001).

No significant correlation between recapture rates and distinctiveness was found in either gray seals or black males, as shown by the results from the regression analysis. In juveniles, contrary to expectation, there was a nearly significant increase in recapture with lower distinctiveness (r = 0.327, P = 0.058). It should be noted, however, that the small sample size available for this stage (n = 35) precludes confirming whether this effect is real or just a consequence of the cataloging process. That is, since juveniles were less distinctive in general, only those more distinctive were well represented in the catalog, and this in sample B1.

Because juveniles recapture rates were very low, a new capture-recapture abundance estimate for 1995 was produced excluding them from the analysis. This estimate, based upon 165 identified individuals, was 219 seals (CV = 0.083). As in the previous estimate, it was corrected for the proportion of poorly marked individuals. However, on this occasion all captured juveniles were included as non-distinctive seals. The new correction factor was estimated

FORCADA AND AGUILAR MONK SEALS 783

as 0.698 (SE = 0.053) and the abundance estimate was 313 individuals (CV = 0.113). Log-normal 95% confidence limits were 258 and 399.

Ability of observers t o grade photographic quality and seal distinctiveness-When compared to the catalog values, the scores of photographic quality assigned by researchers during the grading test showed fair agreement for the two most experienced researchers, and poor agreement (less than expected by chance) for the less experienced researcher (Table 7). The agreement in coding photographic quality between the two experienced researchers was moderate, but the agreement between these and the less experienced researcher was slight and fair, respectively.

The agreement between the distinctiveness codes of the catalog and those assigned by the researchers followed a similar pattern-the two most experienced researchers obtained fair agreement, while the less experienced researcher obtained only slight agreement (Table 7). The agreement between experienced researchers was moderate, and between these and the less experienced researcher was fair. The correlation between quality and distinctiveness codes assigned during the test was high for all the researchers, indicating that distinctiveness was likely to be influenced by the poor quality of slides.

Assessment of Sampling Representativeness and EfJiciency

The analysis of recapture probabilities showed a sharp increase with increasing number of days per capture occasion (Fig. 4, 5A). This result was similar for juveniles, gray seals, and black males although, on average, recapture was lower for juveniles. For periods longer than six days the values stabilized for all stages, except juveniles. Recapture probabilities were high (>0.3) for gray seals and black males in periods equal to or above four days.

Mean capture probabilities stabilized at values which were not significantly lower than 0.05 by the end of the first seven days of survey (Fig. 5B). This indicated that seals were mixed at random at the haul-outs, and the probability of resighting seals which had already been identified was not higher than that of those which had not been seen before. Consequently, the choice of an interval length of 5-8 d between capture occasions allowed some turnover of seals at the haul-out sites, increasing the number of different seals captured.

We used the 1995 sampling period to examine the effect of variation in the duration of capture occasions, and the interval between these, on identification histories for capture-recapture abundance estimates. Abundance estimates were similar for 6, 10, and 14 capture occasions (Table 8). The mean capture probabilities increased with an increasing number of occasions, but differences were less apparent between the results of using 10 and 14 occasions. With 14 occasions, there was more heterogeneity in capture probabilities. In addition, with increasing number of capture occasions, the interval length between these was reduced, allowing for less turnover of seals. This was identified as the main source of heterogeneity. By chance, the total number of captured seals (n = 154) did not change when organizing the capture occasions. This made the precision in the three abundance estimates very similar.

Table

7.

Agr

eem

ent b

etw

een

actu

al p

hoto

grap

hic

qual

ity a

nd d

istin

ctiv

enes

s cod

es o

f slid

es g

iven

whe

n cl

assif

ied

in M

edite

rran

ean

mon

k se

al c

atal

og (

MSC

), an

d th

ose

assi

gned

dur

ing

test

of

abili

ty o

f re

sear

cher

s to

gra

de s

lides

, and

agr

eem

ent b

etw

een

rese

arch

ers

duri

ng t

est.

Kap

pa s

tatis

tics,

and

cor

rela

tion

coef

ficie

nts

of q

ualit

y (Q

) and

dis

tinct

iven

ess (

D) d

epic

ted

for e

ach

rese

arch

er in

volv

ed in

tes

t.

Cor

rela

tion

Qua

lity

Dis

tinct

iven

ess

coef

ficie

nt

Res

earc

her

A

B

C

MSC

A

B

C

M

SC

of Q

and

D

0.47

(P

< 0

.001

) 0.

40 (

P <

0.0

01)

0.16

(P

< 0

.033

)

A (

expe

rienc

ed)

-

0.41

0.

12

0.41

-

0.56

0.

28

0.40

B

(ex

perie

nced

) -

0.24

0.

42

-

0.40

0.

40

C (

less

exp

erie

nced

) -

0.16

-

0.30

FORCADA AND AGUILAR MONK SEALS 785

0.7

0.6

0.5

9 0.4

.m 0 %

g 0.3

2 0.2

k

2 0.1

1 c1

0

0

0 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 1 4 1 5

Duration of capture occasion (days)

l----all- gray seals-. - - - black males- - - - - -,juveniles 1 Figure 4. Increase of recapture probabilities with increasing number of days per

capture occasion. Mean capture probabilities estimated as proportion of identified seals recaptured at end of capture occasion. Mean probability at cumulative intervals of 1- 15 d obtained from 15 replicated sets of capture histories. Error bars correspond to 95% confidence intervals.

A (1'11"1".1

0.6

.g 0.5 3

0.4

e 0.3

0.2

I

L I

8 2 0.1

0.01, , , , , , , , , , , , , , ,I 1 4 7 10 13 16

Time in days

B

0.4 '

0.3

0.2

0.1

0.0 1, , , , , , , , , , , , , , ,

1 4 7 10 13 16 Time in days

Figwe 5 . Comparison of mean recapture probabilities for variable number of days per capture occasion (A), pooling across morphological stages, and mean capture probabilities estimated as proportion of identified seals recaptured on last day of a certain interval (B). Mean probability at cumulative intervals of 1-15 d obtained from 15 replicated sets of capture histories. Error bars correspond to 95% confidence limits.


Table 8. Capture-recapture abundance estimates of Mediterranean monk seal at Cap Blanc, using variable number of capture occasions of different duration in days, and intervals between occasions of different duration. Data corresponds to period April- June 1995, where population assumed to be closed. Total number of marked animals was 154, and estimates corrected by proportion of unmarked animals.

Interval (days) Number of Days per between Mean

capture capture capture capture Abundance occasions occasions occasions probability escimate

6 5 7 0.14 309 (%CV = 12.5; 95%CI: 250-405)

10 3 3-4 0.09 311 (%CV = 12.2; 95%CI: 252-404)

14 2 3 0.06 313 (%CV = 12.2; 95%CI: 254-407)

DISCUSSION

Photo-identification proved to be a reliable tool when applied to Mediter- ranean monk seals. The tests applied to the field data indicated that most of the population can be unequivocally identified over time periods of at least a few years and that methods used to examine, code, and catalog slides are generally satisfactory. However, the study revealed the existence of certain limitations in particular aspects of the process. These need to be addressed if reliable population parameters, particularly using capture-recapture methods, are to be determined.

Photographic Collectzon and Cataloging

The errors reported in the analyses of photo-identification data are missed matches and falsely identified individuals (IWC 1990). Only an effect of missed matches was observed, and further addressed during the routine catalog reviews. When researchers failed to reidentify an already cataloged seal, they assigned a new code to the new slide, thus creating a duplicate entry. As duplicates positively bias capture-recapture abundance estimates, their rate of occurrence needs to be estimated prior to any analysis. A method to assess such a rate is to estimate the probability of missing a match using a reliability or matching test (Carlson et af . 1990). Annual reviews of the catalog, looking for duplicates allowed us to estimate this rate at a 0.043 per year. Most duplicates involved juveniles. Since identification histories of juveniles could be excluded from the capture-recapture analysis, this source of bias could be avoided.

The estimated annual percentage of duplicates is unlikely to create a significant bias in abundance estimates if the sampling period considered is short- er than one year (e.g., when demographic closure can be assumed). For longer periods, the bias could be of the same magnitude as the percentage of dupli-


cates. This indicates that any long-term population analysis would require comprehensive screening to detect duplicates. Alternatively, the ratio of duplicates to the total number of seals in a sample can be used to correct capture- recapture abundance estimates.

Matching errors can be produced by a change of natural marks over time or the use of poor quality identification slides in which marks are not clear (IWC 1990, Dufault and Whitehead 1995). In our study the misclassification of the view of a seal in a slide was identified as an additional factor. This is likely to make recognition of marks more difficult and the seal (represented by a slide) may be discarded when a particular view is favored for identification. Nevertheless, the almost perfect agreement in the tests to assign views suggest that only a few errors are to be expected during the cataloging process. Thus, the misclassification of views is unlikely to have a significant effect on estimates of population parameters.

Morphological stages assigned to slides during the tests were quite consistent with classifications made in the field. However, confounding between medium-sized seals and large grey seals was common in the test, and it can be minimized only by classification of seals in the field. Moreover, because distinctiveness and recapture rates of the various morphological stages varies, such consistency is critical for the estimation of population parameters based on stage-structured populations.

Pemanence o f Marks Our reliability test showed that the effect of time on the high number of

correct matches was low and was most important in the identification of very young animals, for which natural marks are scarce. Low levels of success in the matching test would demonstrate that changes in marks over time are influential. In our three-year study the test produced 100% correct matches, indicating that short-term photo-identification studies are feasible for the species. The match of a seal identified in 1997 with pictures taken in 1984 by Marchessaux indicates that at least some marks are permanent enough to sup- port long-term studies. However, in 1997 this individual carried many additional marks which were not present in 1984. Owing to the continuous addition of new marks and the potential disappearance of older ones, a continuous monitoring of individuals and an update of the catalog with recent slides is necessary when attempting long-term studies.

Further research is needed to assess the rate at which marks change. In individuals with long-standing, clearly identifiable features, the rate of change of secondary marks may be investigated in the same way as a tag-loss experiment (IWC 1990, Dufault and Whitehead 1995). However, definitive and accurate conclusions about mark change rates will be obtained only when a large sample of identified seals is studied over a long time period.

Assessment o f Photographic Quality Photographic quality atid its ejjjct on capture-recapture abundznce estimutes-

Researchers were not highly consistent in assigning codes of photographic


quality, as shown by the results of the grading test. However, the agreement with the cataloged codes was acceptable. A method based on measures of photographic quality can help avoid a more subjective classification of slides. Otherwise, capture-recapture abundance estimates may be biased, as shown by the results of the analysis. In other words, the probability of missed matches will increase if poor-quality slides are coded as good quality ones, and then population estimates would be positively biased. Conversely, a highly conser- vative coding would lead to smaller sample sizes and less precise estimates.

As noted by Burnham et al. (199S), the biggest difficulty in the analysis of the residual sums of squares is identifying a true population size. In our case, the true population size was taken from the set of excellent-quality slides. This procedure has the risk of confounding the bias introduced by seal misidenti- fication with that introduced by the misclassification of slides. Nevertheless, as the sample of slides selected for the analysis was coded according to mea- sured parameters of photographic quality, the potential for bias in coding slides is assumed to be consistent among the different quality levels. Therefore, the RSS procedure, which in our case implies the removal of poor-quality slides, is considered to provide the best trade-off between bias and precision. To further assess and minimize the problem, the grading of slides with quality measures should ideally be carried out by more than one researcher to allow validation; e.g., by conducting a test to assess the agreement between researchers (Friday 1997).

It could also be argued that if the best data set is the one composed exclu- sively of excellent-quality slides, other slides of less quality should not be used. However, the proportion of excellent quality slides was only 25% on average of all the slides obtained in the field. Therefore, abundance estimates would be compromised by the lack of precision, and would be unrepresentative because the total number of captured individuals would be low. Abundance estimators cannot produce unbiased estimates with samples with poor coverage (Pollock et al. 1990), and a population assessment can thus be compromised.

The estimated proportion of poor-quality slides was low for juveniles and grey seals. This result was expected given their behavior. Individuals of these stages remained in waters close to haul-outs and their interaction with other seals was limited. Therefore, they were easily accessible to photographic sampling.

In contrast, black males were less easily photographed. Most of them haul out inside caves but interact with other males when in water. In these inter- actions, black males chase other approaching males until they swim away, making photographic capture of excluded individuals more difficult. This is further complicated by their spatial and temporal fidelity to specific territories (Marchessaux 1989). Taking this into account, the identification of a representative sample of black males would require comprehensive surveying of their distribution range at Cap Blanc. As part of this range includes areas of difficult access (Gonziilez et al. 1997), photo-identification surveys were lengthened to allow turnover of black males in the caves. A consequence of this was an increase in capture probabilities, and also in abundance, as seen


in the last row of Table 8, since black males were appropriately represented in the sample of captured seals.

Distinctiveness of individuals and its effect on capture-recapture abunlnce estimates-The estimated distinctiveness codes indicated that the proportion of unidentified seals was moderately low for all morphological stages with the exception of juveniles. Up to 36% of the latter stage could not be identified in the surveys, whereas only 18% of grey seals and 10% of black males remained unidentified. Natural marks are acquired with age as a result of in- teractions with other individuals and the environment. For this reason, marks in juveniles are non-existent or far more subtle than in adults in which, similarly to the Hawaiian congeneric species, fighting and interaction resulting in scars is more frequent (Hiruki et a/. 1993).

Because they are less well marked, juveniles had lower recapture rates compared to grey seals and black males. Therefore, when combined with those of the other morphological stages in a capture-recapture experiment, the identification histories of juveniles would introduce heterogeneity. However, if the histories of poorly marked seals are removed from the experiment, the resulting abundance estimate would not fully account for this stage and would therefore be negatively biased.

A reliable population estimate would require reaching a compromise between the bias caused by this heterogeneity and that resulting from the removal of poorly marked juveniles. A possible solution is to include all juveniles in the calculation of the correction factor which accounts for non-distinctive seals in the capture-recapture models. In doing this, we also estimated the corrected abundance without the identification histories of juveniles, but this estimate was not significantly different from the estimate including the juvenile histories. The apparent reason is that identification histories of juveniles represented only 11% of the total histories, therefore the bias introduced was small (approximately 1.8%). In this situation, the power of the statistical tests used to compare the estimates was too low to detect the differences produced by the bias and the effect of the heterogeneity is difficult to appraise. Con- sequently, omitting the identification histories in the estimation of abundance of well-marked seals as a precautionary measure is likely to provide less biased abundance estimates of Mediterranean monk seals.

Assessment of Sampling Coverage and Efiiency

When the number of captured seals in a survey is high, the precision of capture-recapture abundance estimates is also high. In addition, estimators for closed populations tend to be less biased. In our case, the sample size used in the estimates of 1995 was high, compared to the total population size. The proportion of captured seals was also high (154 from 228 marked over four years of study), and thus, we assume that our abundance figures are in agreement with the real size of the colony. However, it is worth noting that an increasing number of capture occasions introduced more heterogeneity in capture probabilities as a result of the decrease in the interval length between


capture occasions. Therefore, robust estimators were preferred to assess abundance, and the best balance between duration of capture occasions and interval length has to be accounted for when considering the performance of different abundance estimators.

Even if differences in capture probabilities can be accommodated in some models for closed populations (Otis et al. 1978), time specificity and heterogeneity should be reduced as much as possible. In this respect, the present analysis suggests that a strategy to follow in future surveys on the species would be to expand sampling periods in order to increase capture probabilities. Moreover, the number of capture occasions between surveys within each sampling period should be enlarged to increase sample size of seals from stages with higher proportion of poor quality slides, and thus, lower recapture probabilities.

Capture-recapture studies on the Hawaiian monk seal have shown better results (2. e., more precise and less-biased population parameter estimates) using plastic tags to mark animals (Gilmartin et al. 1993, Craig and Ragen 1999). However, two circumstances made our study difficult to compare to those on the Hawaiian monk seal. Firstly, the hauling-out behavior, distribution, and easy access to reading the tags on the seals in Hawaiian islands gave capture probabilities above 0.85, even for young animals, and close to 1 in many occasions (Craig and Ragen 1999). This situation of almost maximum sampling coverage is unusual and not comparable to that of the Mediterranean monk seal in all its range. Secondly, the Hawaiian monk seal studies combined the use of plastic tags and identification through natural marks in the same data sets (Craig and Ragen 1999), a situation which leads to some degree of heterogeneity, but which is rather unimportant, given the high recapture probabilities observed. In this regard, our results are not strictly comparable again. However, and given the difficulty of access to the Mediterranean monk seals, we believe that the use of photo-identification can be a very valid sampling, non-disruptive technique. If the sampling design is adequate in terms of coverage and representativeness, abundance estimates can be precise and unbiased (Forcada 2000). Because of the lack of efficient methods of surveying populations of this species, as well as their extremely low densities, the current world estimates are no more than best educated guesses, based on minimum counts of individuals. Thus, our estimated numbers are difficult to compare with the total numbers given by Brasseur et al. (1997). Nevertheless, the techniques we describe could be used in many populations of this species that are still accessible, such as that of Desertas Islands, in Madeira, or in selected areas of Greece and Turkey.

ACKNOWLEDGMENTS

Thanks are due to the colleagues of the University of Barcelona who were involved in the photo-identification field and laboratory work: E. Badosa, G. Cantos, E. Grau, and R. Samaranch. They also participated in the reliability tests and contributed sug- gestions to improve the analysis. P. S. Hammond (Sea Mammal Research Unit, St.

FORCADA AND AGUILAR: MONK SEALS 79 1

Andrews, UK), J. Derry (University of Edinburgh, UK) and L. M. GonzLlez (General Directorate for Nature Conservation, Ministry of the Environment, Spain), and two anonymous referees reviewed draft versions of the manuscript and contributed valuable comments. The teams from Isifer and the University of Las Palmas were of great assistance in the fieldwork. The study was made possible thanks ro financing from European Commission LIFE Projects B4-3200/94/741 and B4-3200/96/5 10.

LITERATURE CITED

AGLER, B, A. 1992. Testing the reliability of photographic identification of individual fin whales (Balaenoptera phyralm). Report of the International Whaling Commis- sion 42:731-737.

AGRESTI, A. 1990. Categorical data analysis. John Wiley & Sons, Inc., New York, NY. ARNBOM, T. 1987. Individual identification of sperm whales. Report of the Interna-

tional Whaling Commission 37:20 1-204. BECK, C. A. AND P. REID. 1995. An automated photo-identification catalog for studies of

the life history of the Florida manatee. Pages 120-134 in T. J. OShea, B. B. Ack- erman and H. F. Percival, eds. Population biology of the Florida manatee. National Biological Service Information and Technology Report 1, Washington, DC.

BEUDELS, R. C. 1991. The monk seal register: Its structure and organisation. Seminar on Conservation of the Mediterranean monk seal. Pages 61-64. Technical and scientific aspects. Strasbourg, 19 June 1991 <TPVS91,25>. Secretariat Memo- randum prepared by the Directorate of Environment and Local Authorities. Coun- cil of Europe.

BIGG, M. A. 1982. An assessment of killer whale (Orcinw o w ) stocks off Vancouver Island, British Columbia. Report of the International Whaling Commission 32: 655-666.

BRASSEUR, S. M. J. M., P. H. J. REIJNDERS AND G. VERRIOPOULOS. 1997. Mediterranean monk seal. Pages 12-26 in P. H. J. Reijnders, G. Verriopoulos and S. M. J. M. Brasseur, eds. Status of pinnipeds relevant to the European Union. IBN Scientific Contributions 8, Wageningen, The Netherlands.

BURNHAM, K. P., G. C. WHITE AND D. R. ANDERSON. 1995. Model selection strategy in the analysis of capture-recapture data. Biometrics 5 1 :888-898.

CANTOS, G., R. SAMARANCH, E. BADOSA, E. GRAU, J. FORCADA AND A. AGUILAR. 1998. Photo-identification catalogue of Mediterranean monk seals (Monachus monachw) from the western Sahara. Pages 300-301 in P. G. H. Evans, E. C. M. Parsons and S. L. Clark, eds. European Research on Cetaceans 11. Proceedings of the 1 1 th Conference of the European Cetacean Society, Stralsund, Germany, 10-1 2 March 1997.

CARLSON, C. A., C. A. MAYO AND H. WHITEHEAD. 1990. Changes in the ventral fluke pattern of the humpback whale (Meguptera nwaeangliae), and its effect on matching; evaluation of its significance to photo-identification research. Report of the International Whaling Commission (Special Issue 12):105-112.

CHAMBERS, J. M., AND T. J. HASTIE. 1992. Statistical models in S. Wardsworth and Brooks/ Cole Advanced Books and Software, California, USA.

CHAO, A. S.-M. LEE AND S.-L. JENG. 1992. Estimating population size for capture- recapture data when capture probabilities vary by time and individual animal. Biometrics 48:201-216.

CHAPMAN, D. G. 1951. Some properties of the hypergeometric distribution with applications to zoological censuses. University of California Publications on Statistics

CHAPMAN, D. G. 1952. Inverse multiple and sequential sample censuses. Biometrics

CLAPHAM, P. J., AND C. A. MAYO. 1990. Reproduciton of humpback whales (Megaptera

1:131-160.

81286-306.

7 92 MARINE MAMMAL SCIENCE, VOL. 16, NO. 4, 2000

nwaeungliae) observed in the Gulf of Maine. Report of the International Whaling Commission (Special Issue 12):171-176.

CRAIG, M. P., AND T. J. RAGEN. 1999. Body site, survival, and decline of juvenile Hawaiian monk seals, Monachus schauinslandi. Marine Mammal Science 15:786- 809.

DUFAULT, S., AND H. WHITEHEAD. 1995. An assessment of changes with time in the marking patterns used for photo-identification of individual sperm whales, Physeter macrocepbalus. Marine Mammal Science 11:335-343.

FORCADA, J., P. S. HAMMOND AND A. AGUILAR. 1999. Status of the Mediterranean monk seal in the western Sahara and the implications of a mass mortality event. Marine Ecology Progress Series 188:249-261.

FORCADA, J. 2000. Can population surveys show if the Mediterranean monk seal colony at Cap Blanc is declining in abundance? Journal of Applied Ecology 37:171-181.

FRIDAY, N. 1997. Evaluating photographic capture-recapture estimates of abundance of North Atlantic humpback whales. Dissertation, University of Rhode Island, USA.

GAZO, M., F. APARICIO, M. A. CEDENILLA, J. E LAYNA AND L. M. GONZALEZ. 2000. Pup survival in the Mediterranean monk seal (Monachus monachus) colony of the Cab0 Blanco Peninsula (Western Sahara-Mauritania). Marine Mammal Science 16: 158- 168.

GILMARTIN, W. G., T. C. JOHANOS AND L. L. EBERHARDT. 1993. Survival rates for the Hawaiian monk seal (Monacbw scbauinslandi). Marine Mammal Science 9:407- 420.

GONZALEZ, L. M., A. AGUILAR, L. F. L~PEZ-JURADO AND E. GRAU. 1997. Status and distribution of the Mediterranean monk seal (Monachus monachus) on the “Cabo Blanco” Peninsula (western Sahara-Mauritania) in 1993-1994. Biological Conser- vation 80:22 5-23 3.

HAMMOND, P. S. 1986. Estimating the size of naturally marked whale populations using capture-recapture techniques. Reports of the International Whaling Commission (Special Issue 8):253-282.

HAMMOND, P. S. 1990. Heterogeneity in the Gulf of Maine? Estimating humpback whale population size when capture probabilities are not equal. Report of the International Whaling Commission (Special Issue 12): 135-1 39.

HASTIE, T. J., AND R. TIBSHIRANI. 1990. Generalized additive models. Chapman and Hall, London.

HIBY, A. R., AND J. S. JEFFERY. 1987. Census techniques for small populations, with special reference to the Mediterranean monk seal. Zoological Symposium 58: 193- 210.

HIBY, A. R., AND P. LOVELL. 1990. Computer aided matching of natural marks: A prototype system for gray seals. Report of the International Whaling Commission (Special Issue 12):57-62.

HIRUKI, L. M., 1. STIRLING, W. G. GILMARTIN, T. C. JOHANOS AND B. L. BECKER. 1993. Significance of wounding to female reproductive success in Hawaiian monk seals (Monachus scbauinslandi) at Laysan Island. Canadian Journal of Zoology 7 1 :469- 474.

IWC. 1990. Report of the workshop on individual recognition and the estimation of cetacean population parameters. Report of the International Whaling Commission (Special Issue 12):3-17.

JONES, M. L. 1990 The reproductive cycle in gray whales based on photographic re- sightings of females in the breeding grounds from 1977-82. Report of the In- ternational Whaling Commission (Special Issue 12):177-182.

KATONA, S. K., AND S. KRAUSS. 1979. Photographic identification of individual humpback whales (Megaptera nwaeanglzae): Evaluation and analysis of the technique. US Department of Commerce, NTIS Publication, Springfield, VA.

KENYON, K. W. 1981. Monk seals Monachus Fleming, 1822. Handbook of marine


mammals. Volume 2. Seals in S. H. Ridgway and R. J. Harrison, eds. Academic Press, London.

LANDIS, J. R., AND G. G. KOCH. 1977. The measurement of observer agreement for categorical data. Biometrics 33:159-174.

MARCHESSAUX, D. 1989. Recherche sur la Biologie, 1’Ecologie et les statut de Phoque moine, Monacbus monarhus. GIS Posidonie Publications, ISBN No2-905 540-1 3- 3, Marseille, France.

OTIS, D. L., K. P. BURNHAM, G. C. WHITE AND D. R. ANDERSON. 1978. Statistical inference from capture data on closed animal populations. Wildlife Monographs 62:l-135.

POLLOCK, K. H., J. D. NICHOLS, C. BROWNIE AND J. E. HINES. 1990. Statistical inference for capture-recapture experiments. Wildlife Monographs 107: 1-97.

RUGH, D. J., H. W. BRAHAM AND G. W. MILLER. 1992. Methods for photographic identification of bowhead whales, BaLaena mysticetus. Canadian Journal of Zoology 70617-624.

SEBER, G. A. F. 1982. Estimation of animal abundance and related parameters. Second edition. Griffin and Co. Ltd., London.

WURSIG, B., AND M. WURSIG. 1977. The photographic determination of group size, composition and stability of coastal porpoises (Tursiops trmratus). Science 198:

YOCHEM, P. K., B. S. STEWART, M. MINA, A. ZORIN, V. SADOVOV AND A. YABLOKOV. 1990. Non-metrical analyses of pelage patterns in demographic studies of harbor seals. Report of the International Whaling Commission (Special Issue 12):87-90.

Received: 28 October 1999 Accepted: 2 May 2000

7 5 5-7 56.

USE OF PHOTOGRAPHIC IDENTIFICATION IN CAPTURE-RECAPTURE STUDIES OF MEDITERRANEAN MONK SEALS

Documents