Transcript
Page 1: ESTIMATION OF OCELOT DENSITY IN THE … · Key words: camera-trapping, capture–recapture density estimates, individual recognition,Leopar- ... combination of distinguishing characters

607

Journal of Mammalogy, 84(2):607–614, 2003

ESTIMATION OF OCELOT DENSITY IN THE PANTANAL USINGCAPTURE–RECAPTURE ANALYSIS OF CAMERA-TRAPPING DATA

MOGENS TROLLE* AND MARC KERY

Mammal Department, Zoological Museum, University of Copenhagen, Universitetsparken 15, 2100,Copenhagen, Denmark (MT)

Patuxent Wildlife Research Center, U.S. Geological Survey, 11510 American Holly Drive,Laurel, MD 20708, USA (MK)

Neotropical felids such as the ocelot (Leopardus pardalis) are secretive, and it is difficultto estimate their populations using conventional methods such as radiotelemetry or signsurveys. We show that recognition of individual ocelots from camera-trapping photographsis possible, and we use camera-trapping results combined with closed population capture–recapture models to estimate density of ocelots in the Brazilian Pantanal. We estimated thearea from which animals were camera trapped at 17.71 km2. A model with constant captureprobability yielded an estimate of 10 independent ocelots in our study area, which translatesto a density of 2.82 independent individuals for every 5 km2 (SE 1.00).

Key words: camera-trapping, capture–recapture density estimates, individual recognition, Leopar-dus pardalis, ocelot, the Pantanal

Neotropical felids, such as the ocelot(Leopardus pardalis), are secretive and dif-ficult to study in the field. Estimates of pop-ulation size are especially challenging. Es-timates based on track observations are fail-ure prone and unreliable (Karanth 1995,1999). Radiotelemetry is constrained by thesmall number of animals that can be taggedsimultaneously, the uncertainty about howmany individuals are not tagged, and thehigh costs and efforts involved (Karanth1995, 1999).

Recently, automatic camera trapping incombination with capture–recapture statis-tical modeling has been used to estimatepopulation sizes of wild carnivores. Tiger(Panthera tigris) populations have beenstudied successfully using the natural vari-ation in fur patterns between individual ti-gers (Franklin et al. 1999; Karanth 1995;Karanth and Nichols 1998). Martorello etal. (2001) captured and marked black bears(Ursus americanus) and used subsequent

* Correspondent: [email protected]

phototrapping data to estimate populationsize.

Karanth (1995) suggested that using nat-ural variation in fur markings also had po-tential applicability for other secretivemammals with distinctive markings. In thisstudy, we use this technique to estimatedensity of ocelots in the Brazilian Pantanal.A few studies have estimated density ofocelots based on radiotracking data (Em-mons 1988; Ludlow and Sunquist 1987).Our study is the first published attempt touse camera-trapping data and capture–re-capture methodology to estimate populationsize of felids in South America.

Except for the jaguar (Panthera onca),the ocelot is the largest spotted cat of SouthAmerica. Ocelots are generally solitary,nocturnal, and crepuscular with some diur-nal activity, feed on a variety of vertebrates,mainly smaller mammals, and inhabit arange of habitat types (Emmons and Feer1997; Murray and Gardner 1997). In thePantanal study area, camera trapping re-

Page 2: ESTIMATION OF OCELOT DENSITY IN THE … · Key words: camera-trapping, capture–recapture density estimates, individual recognition,Leopar- ... combination of distinguishing characters

608 Vol. 84, No. 2JOURNAL OF MAMMALOGY

vealed that ocelots were common, and theywere recorded in all nonflooded terrestrialhabitats. Three females radiotracked in thesouthern Pantanal relatively close to ourstudy area had a minimum home range of0.8–1.6 km2 (Crawshaw and Quigley 1989).Emmons (1988) found that in PeruvianAmazon rainforest, radiotracked ocelotsmost often chose different pathways on se-quential nights and that they visited the en-tire home range boundaries every 2–4 days.

MATERIALS AND METHODS

The study was conducted in the upper Rio Ne-gro basin of the southeastern part of the Pantanalfloodplain, Mato Grosso do Sul, SW Brazil. Thestudy site is a research and conservation reserveof Universidade para o Desenvolvimento do Es-tado e da Regiao do Pantanal (UNIDERP) andadjacent areas (headquarters at GPS position19830.4039S; 55836.7919W). The area consistedof a mosaic of open and closed, and wet and dryhabitat types including gallery forest, semideci-duous forest islands with the understory domi-nated by the palm Schealea phalerata, drywoodland and savanna, seasonally floodedgrassland, and marshes. The habitats of the Pan-tanal are described in more detail in Prance andSchaller (1982) and Ratter et al. (1988).

Ocelot camera-trapping data were obtainedduring a general 3-month mammal survey,May–July 2001. We used 6 infrared trail moni-tors (Trailmasters: 5 passive monitors, modelTM550, and 1 active monitor, model TM1550—Goodson and Associates, Inc., Lenexa, Kansas)with camera kits (Trailmaster model TM35-1)using adapted, automatic, weatherproof, 35-mmYashica cameras with automatic flash. The ac-tive system uses an infrared beam between atransmitter and a receiver and is triggered whenthe beam is broken. The passive system consistsonly of a transmitter that monitors a wedge-shaped infrared field and is triggered by warm-blooded animals moving through the wedge. Onthe passive TM550 monitors, we covered the in-frared sensor with tape, leaving only a 1-cm ver-tical gap in the center (assuring that animalswould be in the center of the photograph whentriggering the trap). We used the following sen-sitivity settings: P 5 2 and Pt 5 2. (The infraredwedge is divided into a number of ‘‘windows.’’P is the number of windows that must be broken

by a warm-blooded animal for the trap to betriggered. Pt is the number of seconds allowedfor the animal to break these windows.) The ad-vantages of the passive system are that it ischeaper than the active system and is easier toset up. The disadvantages are that in an open,tropical area like the Pantanal, the trap may betriggered by shadows of branches moving in thewind in front of the sensor and that it is difficultfor the infrared sensor to ‘‘see’’ warm-bloodedanimals when environmental temperature ishigh. To avoid these problems, we programmedunits to work only from 1700–0700 h, which weassumed to be the primary period of activity ofocelots. The advantages of the active system arethat even in an open area like the Pantanal, itcan work both day and night and one can easilychoose the minimum size of animals one wantsto monitor by setting the height above theground of the infrared beam. Trailmonitors wereset at a height of approximately 20 cm so thatocelots passing the trap stations would alwaysbe recorded. Camera delay (the minimum timebetween 2 photographs) was set at a short period(0.5 min) which, combined with bait, gave a bet-ter chance of getting photographs of both flanksof individual ocelots during one capture. Weplaced stations in all nonflooded habitats at sitessuch as trails and corridors that appeared to benatural travel routes for ocelots. Sardines in oilwere used as bait.

To identify individual ocelots from the cam-era-trapping photographs obtained, we used acombination of distinguishing characters includ-ing the patterns of rosettes, spots and stripes onflanks, sex, length and banding pattern of tails,‘‘hanging’’ bellies of lactating females, slimbodies of young individuals, and notched ears(Figs. 1 and 2). Males were identified by thepresence of testes.

Statistical methods.—We divided our camera-trapping data into twelve 1-week periods, eachconstituting a ‘‘trapping occasion.’’ This gave acapture history for each ocelot consisting of astring of ones and zeroes indicating whether theindividual was camera trapped (1) or not (0) dur-ing each 1-week period (see Table 1). To esti-mate abundance, we used the program CAP-TURE (obtainable at www.cnr.colostate.edu/;gwhite/software.html as of 30 April 2002) toimplement capture–recapture models for closedpopulations (Otis et al. 1978). Closed-populationmodels assume that a population remains un-

Page 3: ESTIMATION OF OCELOT DENSITY IN THE … · Key words: camera-trapping, capture–recapture density estimates, individual recognition,Leopar- ... combination of distinguishing characters

May 2003 TROLLE AND KERY—OCELOT DENSITY IN BRAZIL 609

FIG. 1.—Features allowing recognition of individual ocelots, as shown by 2 camera-trapping pho-tographs of the same male at different sites and in different postures, with enlargements of detailsfrom flank, tail, and inner thigh. a) Lateral view. b–d) Details from photograph (a). e) Lateroposteriorview of same male. f–h) Details from photograph (e). Arrows indicate specific distinguishing furmarkings (see text).

changed during the study period, i.e., that thereare no gains or losses of individual ocelots (theclosure assumption).

We checked the closure assumption (the as-sumption that our ocelot population did notchange significantly during the study period) in2 ways. First, we applied the closure test imple-mented in CAPTURE. Because this test isknown to have low power and to be sensitive totrap response (G. C. White et al., in litt.), wealso applied an alternative procedure. We usedprogram MARK (White and Burnham 1999) totest the open-population Cormack–Jolly–Sebermodel against a constrained model in which ap-parent survival was fixed at 1. The latter modelagain represents a closed-population model, soa comparison between these 2 models tests

whether the closed-population assumption isreasonable for this data set (Stanley and Burn-ham 1999).

Program CAPTURE estimates abundance un-der 7 models that differ in their assumptionsabout capture probability. The simplest model,M0, assumes a constant capture probabilityacross all occasions and animals. Model Mt

(where t is time) assumes that capture probabil-ity varies between occasions, e.g., due to chang-ing weather conditions. Model Mb (behavior) al-lows trap response, i.e., capture probability maybe different between the 1st capture and all sub-sequent recaptures of an animal. For example,baiting traps might have caused ocelots to returnmore quickly to a trap site than they would oth-erwise have done. Model Mh (heterogeneity) as-

Page 4: ESTIMATION OF OCELOT DENSITY IN THE … · Key words: camera-trapping, capture–recapture density estimates, individual recognition,Leopar- ... combination of distinguishing characters

610 Vol. 84, No. 2JOURNAL OF MAMMALOGY

FIG. 2.—Camera-trapping photographs of 6 ocelots, indicating characters used in distinguishingindividuals. a) Female (note lack of testes), very short tail, notches in both ears, long series of rosettesmelted together; b) female, slender body, medium long tail, isolated medium-sized rosettes; c) lac-tating female (note hanging belly), short tail, large, round rosettes on upper flank; d) female, longtail, weakly defined rosettes and many small spots and lines; e) male (more strongly built thanfemales), long tail, long rosette on shoulder, additional rosettes on flank circular; f) male (note testes),very long tail, long narrow rosettes.

sumes that each animal had its own probabilityof being captured, e.g., that individuals withlarger home ranges are exposed to more traps.In addition, CAPTURE allows estimation under3 models that are pairwise combinations of thesesources of variation in capture probability (mod-els Mth, Mbh, and Mtb).

To identify an adequate model for estimation,we used goodness-of-fit tests, between modeltests, and the model selection algorithm provid-ed in program CAPTURE. We report estimatesfrom program CAPTURE of capture probability,population size, and the standard error of pop-ulation size based on the most adequate model.

To convert the estimate of population size intoan estimate of density, we followed the proce-dure adopted by Karanth and Nichols (1998).We first calculated a core area as the minimumconvex polygon defined by all trapping stations.This core area was unlikely to contain the entirehome range of all trapped ocelots. Instead, it islikely that some ocelots had home ranges thatextended beyond the core area. To account forthat, we added a boundary strip to obtain thetotal area from which our animals were taken.Strip width was given by half the mean maxi-mum distance moved by ocelots caught on morethan 1 trap. This ad hoc approach has little the-

Page 5: ESTIMATION OF OCELOT DENSITY IN THE … · Key words: camera-trapping, capture–recapture density estimates, individual recognition,Leopar- ... combination of distinguishing characters

May 2003 TROLLE AND KERY—OCELOT DENSITY IN BRAZIL 611

TABLE 1.—Summary of camera-trapping re-sults. The capture history consists of 12 captureoccasions (i.e., twelve 1-week periods). A ‘1’indicates that an individual was camera-trappedduring a particular 1-week period and ‘0’ that itwas not. Maximum distance moved is given forindividuals camera-trapped at more than 1 site.

IndividualCapturehistory

Maximumdistance

moved (km)

Adult females

f1f2f3f4

000100001000000111000010000000011000000010000110

2.41.30.31.0

Adult males

m1m2m3m4

000110000100010000000000000000010000000000000001

2.1—0.3—

Subadults

s1 101000000000 1.1

oretical justification, but it appeared to workwell in simulation studies of Wilson and Ander-son (1985), and it was the only available meansof estimating boundary strip width in our situ-ation. Density was then obtained by dividing thepopulation size estimate by the estimated totalarea. Formulas for these estimators and theirvariances can be found in Karanth and Nichols(1998).

RESULTS

During the study period, 30 trapping sta-tions were camera trapped successfully,covering a minimal convex polygon area of9.26 km2 fairly regularly. The total camera-trapping effort was about 450 camera-trap-ping nights. Both the active and the passiveTrailmaster systems worked well for oce-lots. Sardines in oil proved to be quite at-tractive to ocelots, as evidenced by manyphotographs of animals sniffing or eatingthe bait. Fifty-five camera-trapping photo-graphs of ocelots were obtained on 29 cap-tures and from 13 sites. (We defined a cap-ture as a record of 1 individual. Because ofthe bait, the ocelots sometimes remained in

front of a camera trap long enough for morethan 1 photograph to be taken during thesame capture. A trapping occasion wascomposed of a 1-week period. During some1-week periods, more than 1 capture wasobtained of an individual.)

Recognition of individual ocelots.—Many distinguishing characters can befound in the markings of ocelots. Fig. 1shows an example of 16 characters allow-ing positive identification of a male ocelotphotographed at 2 different sites and in twodifferent body postures. The series of ro-settes on the middle and upper flanks andupper shoulder (often in combination withshort lines or dots, or both) are generallythe best features allowing certain identifi-cation of individuals. The pattern seen onthe lower part of the shoulder varies greatlyaccording to the position of the front leg.Frequently, long rosettes followed by a spe-cific pattern of smaller rosettes and dots arecharacteristic (Fig. 1). Tail length, numberof bands, and banding pattern (e.g., the spe-cific combination of thin, thick, and brokenbands) are often helpful characters. The pat-tern of large stripes and smaller dots on theinner thigh may also be helpful features.Fig. 2 shows examples of distinguishingcharacters of 6 additional individuals.

All but 3 of the 55 photographs wereidentified. Nine individuals could be distin-guished: 4 adult males, 4 adult females, and1 subadult. The capture history for each in-dividual is given as part of Table 1.

Density estimate.—Both closure testswere consistent with the assumption thatthe ocelot population was closed for the du-ration of the study (test in CAPTURE: z 520.963, P 5 0.17; test using MARK: x2 51.384, d.f. 5 1, P 5 0.23). The model se-lection algorithm in CAPTURE selectedmodel M0, with constant capture probabili-ty, as most appropriate. Its selection crite-rion was 1.0 compared with 0.81 for thenext best model, Mh. Direct hypothesis testsbetween model M0 and the competing mod-els Mt, Mb, and Mh gave no reason to reject

Page 6: ESTIMATION OF OCELOT DENSITY IN THE … · Key words: camera-trapping, capture–recapture density estimates, individual recognition,Leopar- ... combination of distinguishing characters

612 Vol. 84, No. 2JOURNAL OF MAMMALOGY

the assumption of a constant capture prob-ability (P . 0.83).

The estimated capture probability per oc-casion and individual was 0.16. The result-ing population size estimate was 10 ocelots(SE 5 1.36) with a 95% confidence interval(CI) of 9 to 14. The estimated probabilityof catching an ocelot at least once duringthe entire study period is given by 1 2 (12 0.1641)12 5 0.88 or alternatively by theratio of total number of animals caught toestimated population size, 9/10 5 0.90 (dif-ference due to rounding).

The minimal convex polygon core areaof our camera trapping was 9.26 km2. Thetotal area including the boundary strip of0.6 km measured 17.71 km2. The estimateddensity of ocelots was 2.82 independent in-dividuals per 5 km2 (SE 5 1.00; 95% CI 50.86 to 4.79). The wide CI reflects both alow capture probability (per trapping occa-sion) and a relatively small trapping areaand population size.

DISCUSSION

Biological considerations.—We found adensity of ocelots of 2.82/5 km2 in our Pan-tanal study area. Only a few attempts havepreviously been made to estimate ocelotdensity. Based on radiotelemetry data(home range configurations) and availablehabitats, Ludlow and Sunquist (1987) esti-mated density at 1.9 resident ocelots per 5km2 in the Venezuelan Llanos. Emmons(1988) estimated density at 4 resident oce-lots per 5 km2 in the Peruvian Amazon.

Methodological considerations.—In thedesign of a capture–recapture study to es-timate density using closed-populationmodels, it is desirable to achieve constantand high capture probabilities and shorttrapping sessions (of about 1.5–2 months).The former yields unbiased and preciseabundance estimates. The latter makes itmore probable that the closure assumptionis met, which is difficult to test (Kendall1999). Factors that could affect ocelot cap-ture probability are trap density, durationand definition of the trapping occasion, spa-

tial arrangement of the traps, and trappingdetails, such as whether baits are used.

Our study had a trap density per trappingoccasion of 6/16.3 km2, corresponding to 1trap for every 2.7 km2. We achieved a cap-ture probability per occasion of 0.16. Thiswas probably adequate because a captureprobability of 0.1 is usually cited as thelower limit for which meaningful estimatesof population size can be obtained (Otis etal. 1978).

In a capture–recapture study, it is essen-tial to appropriately define the trapping oc-casion. We chose 1-week intervals as ourtemporal units (occasions). It is importantthat for each trapping occasion the wholestudy area is covered well with traps andthat there are no ‘‘gaps’’ where individualsare not exposed to any traps at all. Karanthand Nichols (1998) had an alternative studydesign, which may be more appropriatethan ours. They had 4–6 traplines that ranthrough an entire study area, each traplinewith 12–15 camera-trapping points. Theirtrapping occasion was defined as the 4–6consecutive nights it took to trap all trap-lines for 1 night each. An entire trappingsession lasted for 9–16 occasions.

In conclusion, we recommend the fol-lowing. Trapping sessions should be keptrelatively short for the closed population as-sumption to be met. Each trapping occasionmay include camera trapping in several(e.g., up to 5) subareas, which allows alarger area to be studied and ensures thatthe study area is covered equally by cameratraps at each capture occasion. The studyarea should be covered with a trapping-sta-tion density that yields a capture probabilityof at least 0.1. The home range of everytarget individual within the study areashould contain at least a few traps. If eco-nomically feasible, 2 cameras should beused on each trapping site so that bothflanks of animals are photographed on eachcapture.

If the same study area is revisited after alonger period (e.g., months or a year), Pol-lock’s robust capture–recapture design can

Page 7: ESTIMATION OF OCELOT DENSITY IN THE … · Key words: camera-trapping, capture–recapture density estimates, individual recognition,Leopar- ... combination of distinguishing characters

May 2003 TROLLE AND KERY—OCELOT DENSITY IN BRAZIL 613

be applied, which permits estimation of sur-vival and migration rates in addition toabundance. For a discussion of this usefuldesign, see Kendall et al. (1997) and Wil-liams et al. (2002).

In conclusion, our study successfully ap-plies the use of automatic camera trappingand capture–recapture modeling of the pop-ulation size, initiated in tiger studies, to an-other elusive carnivore. We note that, es-pecially at low densities, a host of charac-ters may serve as individual markings, suchas sex, length of whiskers, scars, and mu-tilations (for example, see Fig. 2a). Thus,the use of the techniques used in this studyis probably not limited to animals with suchstrong fur markings as the ocelot or the ti-ger. Study design, however, deserves suffi-cient consideration before initiating anystudy so that as many of the model as-sumptions as possible are met and that ef-ficient use can be made of the data to modelpopulation size and other demographictraits.

RESUMO

A jaguatirica (Leopardus pardalis), bemcomo os outros felinos Neotropicais, saoanimais discretos e arredios, o que dificultaestimar suas populacoes utilizando metodosconvencionais tais como a radio-telemetriaou a identificacao atraves de sinais diretosou indiretos (fezes, pegadas). Demonstra-mos que e possıvel o reconhecimento de in-divıduos de jaguatirica atraves do uso defotografias registradas pelo uso de ‘came-ras-trap’. Utilizamos os resultados obtidoscom o uso das ‘cameras-trap’ combinadoscom modelos de captura e recapturas parapopulacoes fechadas para estimar a densi-dade populacional das jaguatiricas no Pan-tanal do Brasil. A area estimada dos ani-mais registrados foi de 17.71 km2. Um mo-delo de probabilidade de captura constanteproduziu uma estimativa de 10 jaguatiricasem nossa area de estudo, o que se traduzpara uma densidade de 2.82 indivıduos in-dependentes por 5 km2 (SE 1.00).

ACKNOWLEDGMENTS

We appreciate helpful discussions with J.Nichols. The fieldwork was a collaboration be-tween the Zoological Museum, University ofCopenhagen, Universidade para o Desenvolvi-mento do Estado e da Regiao do Pantanal (UN-IDERP), and Museu Nacional, UniversidadeFederal do Rio de Janeiro. The project was madeeconomically possible by the support of, amongothers, WWF-Denmark/Novo Nordisk, Zoolog-ical Museum of Copenhagen, University of Co-penhagen, and Copenhagen Zoo. We are verythankful to UNIDERP for allowing the study totake place at Fazenda Sta Emılia and for impor-tant logistical support. For essential help in pre-paring the project and logistical support in Brazilwe are grateful to J. A. de Oliveira and N. M.R. Guedes. We also thank E. Guimaraes, H. J.Baagøe, J. Fjeldsa, and M. Andersen. Finally,we thank the librarian at Patuxent, W. Manning,for her excellent services, A. Bezerra for trans-lating the summary into Portuguese, and 2 anon-ymous reviewers.

LITERATURE CITED

CRAWSHAW, P. G., JR., AND H. B. QUIGLEY. 1989. Noteson ocelot movement and activity in the Pantanal re-gion, Brazil. Biotropica 21:377–379.

EMMONS, L. H. 1988. A field study of ocelots (Felispardalis) in Peru. Revue d’Ecologie la Terre et laVie 43:133–157.

EMMONS, L. H., AND F. FEER. 1997. Neotropical rain-forest mammals: a field guide. 2nd ed. University ofChicago Press, Chicago, Illinois.

FRANKLIN, N., BASTONI, SRIYANTO, D. SISWOMARTONO,J. MANANSANG, AND R. TILSON. 1999. Last of theIndonesian tigers: a cause for optimism. Pp. 130–147 in Riding the tiger: tiger conservation in human-dominated landscapes (J. Seidensticker, S. Christie,and P. Jackson, eds.). Cambridge University Press,Cambridge, United Kingdom.

KARANTH, K. U. 1995. Estimating tiger Panthera tigrispopulations from camera trap data using capture-re-capture models. Biological Conservation 71:333–338.

KARANTH, K. U. 1999. Counting tigers, with confi-dence. Pp. 350–353 in Riding the tiger: tiger con-servation in human-dominated landscapes (J. Sei-densticker, S. Christie, and P. Jackson, eds.). Cam-bridge University Press, Cambridge, United King-dom.

KARANTH, K. U., AND J. D. NICHOLS. 1998. Estimationof tiger densities in India using photographic cap-tures and recaptures. Ecology 79:2852–2862.

KENDALL, W. L. 1999. Robustness of closed capture-recapture methods to violations of the closure as-sumption. Ecology 80:2517–2525.

KENDALL, W. L., J. D. NICHOLS, AND J. E. HINES. 1997.Estimating temporary emigration using capture-re-

Page 8: ESTIMATION OF OCELOT DENSITY IN THE … · Key words: camera-trapping, capture–recapture density estimates, individual recognition,Leopar- ... combination of distinguishing characters

614 Vol. 84, No. 2JOURNAL OF MAMMALOGY

capture data with Pollock’s robust design. Ecology78:563–578.

LUDLOW, M. E., AND M. E. SUNQUIST. 1987. Ecologyand behavior of ocelots in Venezuela. National Geo-graphic Research 3:447–461.

MARTORELLO, D. A., T. H. EASON, AND M. R. PELTON.2001. A sighting technique using cameras to esti-mate population size of black bears. Wildlife SocietyBulletin 29:560–567.

MURRAY, R. L., AND G. L. GARDNER. 1997. Leoparduspardalis. Mammalian Species 548:1–10.

OTIS, D. L., K. P. BURNHAM, G. C. WHITE, AND D. R.ANDERSON. 1978. Statistical inference from capturedata on closed animal populations. Wildlife Mono-graph 62:1–135.

PRANCE, G. T., AND G. B. SCHALLER. 1982. Preliminarystudy of some vegetation types in the Pantanal, MatoGrosso, Brazil. Brittonia 34:228–251.

RATTER, J. A., A. POTT, V. J. POTT, C. N. DA CUNHA,AND M. HARIDASAN. 1988. Observations on woody

vegetation types in the Pantanal and Corumba, Bra-zil. Notes from the Royal Botanic Garden of Edin-burgh 45:503–525.

STANLEY, T. R., AND K. P. BURNHAM. 1999. A closuretest for time-specific capture-recapture data. Envi-ronmental and Ecological Statistics 6:197–209.

WHITE, G. C., AND K. P. BURNHAM. 1999. ProgramMARK: survival estimation from populations ofmarked animals. Bird Study 46, supplement:120–138.

WILLIAMS, B. K., J. D. NICHOLS, AND M. J. CONROY.2002. Analysis and management of animal popula-tions. Academic Press, San Diego, California.

WILSON, K. R., AND D. R. ANDERSON. 1985. Evaluationof two density estimators of small mammal popu-lation size. Journal of Mammalogy 66:13–21.

Submitted 17 December 2001. Accepted 11 June 2002.

Associate Editor was William L. Gannon.


Recommended