View
212
Download
0
Category
Preview:
Citation preview
BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit publishers, academic institutions, researchlibraries, and research funders in the common goal of maximizing access to critical research.
Comparison of noninvasive genetic and camera-trapping techniques for surveyingsnow leopardsAuthor(s): Jan E. Janečka, Bariushaa Munkhtsog, Rodney M. Jackson, Galsandorj Naranbaatar, David P.Mallon, and William J. MurphySource: Journal of Mammalogy, 92(4):771-783. 2011.Published By: American Society of MammalogistsDOI: http://dx.doi.org/10.1644/10-MAMM-A-036.1URL: http://www.bioone.org/doi/full/10.1644/10-MAMM-A-036.1
BioOne (www.bioone.org) is a nonprofit, online aggregation of core research in the biological, ecological, andenvironmental sciences. BioOne provides a sustainable online platform for over 170 journals and books publishedby nonprofit societies, associations, museums, institutions, and presses.
Your use of this PDF, the BioOne Web site, and all posted and associated content indicates your acceptance ofBioOne’s Terms of Use, available at www.bioone.org/page/terms_of_use.
Usage of BioOne content is strictly limited to personal, educational, and non-commercial use. Commercial inquiriesor rights and permissions requests should be directed to the individual publisher as copyright holder.
Comparison of noninvasive genetic and camera-trapping techniquesfor surveying snow leopards
JAN E. JANECKA,* BARIUSHAA MUNKHTSOG, RODNEY M. JACKSON, GALSANDORJ NARANBAATAR, DAVID P. MALLON, AND
WILLIAM J. MURPHY
Department of Veterinary Integrative Biosciences, College of Veterinary Medicine and Biomedical Sciences, Texas
A&M University, College Station, TX 77843-4458, USA (JEJ, WJM)
Institute of Biology, Mongolian Academy of Sciences and Irbis Mongolian Centre, Jukov Avenue 77, Ulaanbaatar 51,
Mongolia (BM, GN)
Snow Leopard Conservancy, 18030 Comstock Avenue, Sonoma, CA 95476, USA (RMJ)
School of Biology, Chemistry and Health Science, Manchester Metropolitan University, Manchester M1 5GD, United
Kingdom (DPM)
* Correspondent: jjanecka@cvm.tamu.edu
The endangered snow leopard (Panthera uncia) is widely but sparsely distributed throughout the mountainous
regions of central Asia. Detailed information on the status and abundance of the snow leopard is limited because
of the logistical challenges faced when working in the rugged terrain it occupies, along with its secretive nature.
Camera-trapping and noninvasive genetic techniques have been used successfully to survey this felid. We
compared noninvasive genetic and camera-trapping snow leopard surveys in the Gobi Desert of Mongolia. We
collected 180 putative snow leopard scats from 3 sites during an 8-day period along 37.74 km of transects. We
then conducted a 65-day photographic survey at 1 of these sites, approximately 2 months after scat collection. In
the site where both techniques were used noninvasive genetics detected 5 individuals in only 2 days of
fieldwork compared to 7 individuals observed in the 65-day camera-trapping session. Estimates of population
size from noninvasive genetics ranged between 16 and 19 snow leopards in the 314.3-km2 area surveyed,
yielding densities of 4.9–5.9 individuals/100 km2. In comparison, the population estimate from the 65-day
photographic survey was 4 individuals (adults only) within the 264-km2 area, for a density estimate of 1.5 snow
leopards/100 km2. Higher density estimates from the noninvasive genetic survey were due partly to an inability
to determine age and exclude subadults, reduced spatial distribution of sampling points as a consequence of
collecting scats along linear transects, and deposition of scats by multiple snow leopards on common sites.
Resulting differences could inflate abundance estimated from noninvasive genetic surveys and prevent direct
comparison of densities derived from the 2 approaches unless appropriate adjustments are made to the study
design.
Key words: abundance, Gobi Desert, microsatellites, Mongolia, monitoring, Panthera uncia, population survey, scat
E 2011 American Society of Mammalogists
DOI: 10.1644/10-MAMM-A-036.1
The endangered snow leopard (Panthera uncia) occupies
particularly inaccessible mountainous habitat in central Asia,
covering an area of more than 1.2 million km2 across 12
countries (Nowell and Jackson 1996). Snow leopards are elusive
and sparsely distributed, making them extremely difficult to
observe and study in the wild (Nowell and Jackson 1996). As a
result, reliable information is lacking on the number of snow
leopards remaining, locations of peripheral and core popula-
tions, and areas where they are in decline. However, such
knowledge is critical for the conservation of this flagship
species. Considerable effort is underway to determine the status
of the species across its extensive and highly fragmented range
(Ale et al. 2007; Jackson et al. 2009; Janecka et al. 2008; Lovari
et al. 2009; McCarthy et al. 2008; Xu et al. 2008). Survey
methods used must generate reliable quantitative data and be
time efficient, easily repeatable, and affordable.
To date, researchers have relied heavily on field sign to
detect and monitor snow leopards (Ale et al. 2007; Fox et al.
w w w . m a m m a l o g y . o r g
Journal of Mammalogy, 92(4):771–783, 2011
771
1991; Hussain 2003; Jackson and Hunter 1996; Mallon 1991;
McCarthy and Munkhtsog 1997; Schaller et al. 1987, 1988;
Xu et al. 2008). Spoors, scrapes, scats, and scent-sprayed rocks
constitute evidence used for presence–absence surveys and
abundance indexes (Jackson and Hunter 1996). The high
marking frequency of snow leopards, conspicuousness of the
deposited sign, and use of predictable marking sites located
along ridgelines, saddles, and cliff bases facilitate this
approach (Ahlborn and Jackson 1988). However, many
environmental and biological factors (e.g., time of year,
structure of substrate, available scraping sites, breeding status
of females, age structure, social status, and home-range
overlap—Ahlborn and Jackson 1988; Jackson and Hunter
1996) confound the relationship between snow leopard density
and distribution and frequency of sign, making comparisons
between sites difficult. In addition, sign surveys do not
account for differences in detection probabilities or distinguish
between individual snow leopards, making it largely impos-
sible to derive quantitative population estimates with this
approach.
Photographic and molecular methods incorporating nonin-
vasive sampling offer more-robust techniques for monitoring
carnivore populations (Karanth and Nichols 1998; Kohn et al.
1999). Camera trapping has been used extensively for
estimating abundance and population size of many felids,
including ocelots (Leopardus pardalis—Dillon and Kelly
2008), jaguars (Panthera onca—Soisalo and Cavalcanti
2006), leopards (Panthera pardus—Balme et al. 2009), and
tigers (Panthera tigris—Karanth and Nichols 1998). In
northern India, Jackson et al. (2006) demonstrated the
effectiveness of this approach on snow leopards and concluded
that it was a useful tool for estimating population size where
density exceeds 2 or 3 individuals/100 km2. However, camera
trapping requires substantial investment in field time and
equipment, along with careful attention to the distribution and
placement of cameras. Low captures and detection probabil-
ities as a result of insufficient sampling lead to wide
confidence intervals and poor precision of capture–mark–
recapture estimates of population size (Thompson 2004).
Ruggedness of snow leopard habitat and generally poor access
within study sites increases the amount of time and effort
required to set and maintain camera stations. This limits the
size of areas that can be surveyed within the 45- to 55-day
period recommended for maintaining population closure
(Jackson et al. 2006, 2009).
Noninvasive genetic techniques also enable estimation of
population size and density and are becoming more common
for wildlife monitoring (Schwartz et al. 2007; Waits and
Paetkau 2005). This approach has been used to estimate
population size in diverse species, including tigers (Mondol et
al. 2009), bobcats (Lynx rufus—Ruell et al. 2009), coyotes
(Canis latrans—Kohn et al. 1999), and elephants (Loxodonta
cyclotis—Eggert et al. 2003). The simplest technique is to
sample scats deposited within a study site and attribute the scat
to specific individuals by genotyping variable microsatellite
loci. Two approaches are used for deriving population
estimates from noninvasive genetic data collected in this
manner. The 1st uses multiple sampling occasions and a
capture–mark–recapture model similar to camera-trapping
surveys. The 2nd approach derives population estimates from
the number of times different individuals are observed among
samples collected during a single period. In a tiger study in
Bandipur National Park (India) capture–mark–recapture
estimates of population size from noninvasive genetic
techniques were similar to those from camera trapping
(Mondol et al. 2009).
Efficacy of detecting multiple snow leopards via genetic
analysis of scats recently has been illustrated across diverse
geographic areas (Janecka et al. 2008; Lovari et al. 2009;
McCarthy et al. 2008). McCarthy et al. (2008) found a
correlation between the frequency of snow leopard sign (total
sign per kilometer) and number of individuals genetically
detected via analysis of scats. However, neither estimator was
correlated with camera-trapping results from 2 of 3 study sites
(McCarthy et al. 2008). In addition, McCarthy et al. (2008) did
not derive population estimates from genetic analysis of scats
and so did not directly compare abundance from noninvasive
genetic and camera-trapping techniques.
Both noninvasive genetic and camera techniques are
promising for generating quantitative information on distribu-
tion and abundance of snow leopards. Cost-effective and
reliable methods for estimating snow leopard abundance to
better target conservation initiatives are urgently needed
(Jackson and Fox 1997; McCarthy and Chapron 2003). The
objective of our study was to compare the 2 most common
noninvasive approaches for deriving density estimates in terms
of effort, cost, and ability to detect and recapture individual
snow leopards in the Gobi Desert of Mongolia. Comparison
of results obtained by these approaches, along with the
advantages and disadvantages of each, are important for
making decisions on the investment of resources for
monitoring snow leopards. A priori knowledge of the actual
abundance of snow leopards in the area was not available;
therefore, we made comparisons using information on snow
leopard biology and ecology and previous estimates of
population density.
MATERIALS AND METHODS
Study area.—The distribution of snow leopards in Mongolia
extends into the Gobi Desert (Bannikov 1954). In this region
numerous relatively isolated, low-elevation (1,500–2,500 m)
mountains run west to east, south of the larger Gobi Altai
range. We surveyed Tost Uul (i.e., Tost Mountains;
43u10.89N, 101u34.89E) and Noyon Uul (43u7.89N,
102u0.009E) located approximately 250 km west of Dalan-
zadgad (Fig. 1). These rugged massifs rise out of the desert
and are composed of exposed rock and steep slopes broken by
seasonal drainages. In Dalanzadgad the mean annual precip-
itation for the period between 1961 and 2004 was 1.31 cm
with most falling in July and August (Cheng et al. 2011;
Sternberg et al., in press). Temperatures were typically lowest
772 JOURNAL OF MAMMALOGY Vol. 92, No. 4
in January (monthly mean 214.5uC, 1961–2004) and highest
in July (monthly mean 21.6uC, 1961–2004—Cheng et al.
2011).
The vegetation is sparse on mountain massifs, with feather
grass (Stipa gobica and S. glareosa), low shrubs (Caragana
spp. and Artemisia frigida), and herbaceous plants (Ajania
spp. and Scorzonera capito) most common. In valleys and
gullies the bush Amygdalus mongolica is often dominant.
Snow leopards prey primarily on ibex (Capra sibirica) and
argali (Ovis ammon) in the Gobi Desert, although the argali is
now rare in many areas. Small mammals also are utilized by
snow leopards and include the Tolai hare (Lepus tolai) and
Pallas’s pika (Ochotona pallasi). Sympatric carnivores that
potentially compete for resources include the gray wolf (Canis
lupus), red fox (Vulpes vulpes), and Eurasian lynx (Lynx
lynx—Bannikov 1954; Bold and Dorjzunduy 1976). Pastoral-
ists primarily graze goats and sheep; they use desert slopes
during summer but live in winter–spring camps closer to the
mountains from October through May. Both the local people
and wildlife are dependent on a small number of wells and
springs because of limited available water.
Scat collection.—Snow leopards were not handled directly
during this project. Three sites were surveyed by collecting
scats that later were analyzed genetically. The 1st site (Noyon)
was in Noyon Uul and the other 2 sites (Tost A and Tost B) in
Tost Uul, ,125 km to the west of Noyon (Figs. 1 and 2).
Knowledgeable local residents guided field teams to areas
actively used by snow leopards. We focused on sites with
known or suspected snow leopard activity to maximize
probability of detection because sampling random locations
would provide low detection rates insufficient for population
size estimates (Otis et al. 1978; Pollock et al. 1990).
Two teams concurrently searched for snow leopard sign
(scrapes, tracks, and scent marks) along different transects
traversing ridgelines, saddles, and rocky outcrops. When sign
was observed, each team followed wildlife trails along well-
defined topographic features and collected snow leopard scats
that were present. Species identification of scats in the field
was based on diameter, segmentation, odor, and proximity to
other snow leopard sign (Jackson and Hunter 1996). Only
scats believed to be of snow leopard origin were collected. A
small sample of each scat (,1 cm2) was stored in 15-ml
centrifuge tubes with silica desiccant. Total numbers of scats,
scrapes, and tracks observed were noted. Length of each
transect was dependent on terrain, amount of scat, and
available time. Transects were continued until approximately
18 samples were collected or 5 km of terrain were covered.
Limits were imposed to avoid uneven representation among
transects and to ensure sufficient time for teams to return to
base camp.
Species, sex, and individual identification of scats.—DNA
was extracted from scats using the Qiagen Stool DNA
extraction kit (Qiagen, Valencia, California) following manu-
facturer recommendations. Species-specific polymerase chain
reaction (PCR) or restriction enzyme assays were not available
for all sympatric carnivores present in our study sites. We
therefore identified species that deposited scats by PCR
amplifying and sequencing a 148-base pair (bp) segment of
the mitochondrial cytochrome-b gene using previously de-
scribed methods (Farrell et al. 2000; Janecka et al. 2008).
Sequences observed (GenBank accessions HQ897983–
HQ897986) were aligned with reference taxa, and a neighbor-
joining phylogeny was reconstructed in PAUP* 4.0b10
(Janecka et al. 2008; Swofford 2003). A scat was identified
when its haplotype was in a monophyletic clade with a
reference species supported by 100% bootstrap values and
exhibited ,3% sequence divergence (Janecka et al. 2008).
Genetically identified snow leopard scats were genotyped at
7 microsatellite loci (PUN082, PUN100, PUN124, PUN132,
PUN225, PUN229, and PUN327—Janecka et al. 2008). Loci
FIG. 1.—Map showing scat transects and camera-trap locations
sampled in the Gobi Desert of Mongolia. Scat transects were buffered
with one-half mean maximum distance moved (MMDM) so that they
appear more noticeable.
FIG. 2.—Map showing the distribution of scat transects and
camera-trap locations within the Tost A study site. Buffers used to
estimate effective study areas for the 2 respective surveys also
are shown.
August 2011 JANECKA ET AL.—SURVEYING SNOW LEOPARD POPULATIONS 773
were originally characterized in the domestic cat by Menotti-
Raymond et al. (1999) and assayed in snow leopards by Waits
et al. (2007), but the primers used were redesigned from snow
leopard flanking sequences and placed closer to microsatellite
repeats to minimize genotyping errors as a result of PCR
failure, allele dropout, and false alleles (Janecka et al. 2008).
Previous analysis of error rates suggested that 3 independent
PCR replicates were sufficient to minimize genotyping errors
to an acceptable level (Janecka et al. 2008). All 7 loci were
variable in the snow leopard population studied.
Microsatellites were PCR amplified using the methods of
Janecka et al. (2008) and genotyped on an ABI 3730 (Applied
Biosystems, Inc., Foster City, California) in the Molecular
Cytogenetics and Genomics Laboratory (Texas A&M Univer-
sity). Samples were handled in a dedicated pre-PCR area, and
barrier-tips were used to minimize contamination. All PCRs
and genotyping were done in triplicate and analyzed with
extraction blanks. Samples that did not yield �2 consistent
replicates at each locus were discarded from analyses. Alleles
had to be observed �2 times in the 3 PCR replicates to be
included in the consensus genotype. Only samples with
consensus genotypes generated for all 7 loci were included
in the analysis.
Sex identification was performed with felid-specific
primers (Murphy et al. 1999) that amplified a 200-bp intronic
segment of the AMELY gene on the Y chromosome (Janecka
et al. 2008). Amplifications were performed in triplicate
along with 1 male positive control, 1 female positive control,
and 1 negative control. PCR products were run on a 1.5%
agarose gel, stained with ethidium bromide, and visualized
under ultraviolet light. A sample was identified as male if at
least 2 of 3 PCRs gave strong amplification of the AMELY
marker. Samples that yielded PCR product in only 1 PCR
were not assigned to either sex. If no amplification of the
AMELY marker occurred in all 3 replicates, the individuals
were considered females. However, to avoid misidentifying
male samples as female due to poor DNA quality or quantity,
only scats that were genotyped successfully with the
complete 7-microsatellite panel were identified positively as
female.
Analysis of microsatellite data.—Level of PCR failure and
allele dropout were monitored closely, and scat samples with
missing consensus genotypes at �1 loci were excluded to
avoid genotyping errors. We thus excluded 21 snow leopard
scat samples. We estimated PCR failure and allele dropout
rates for 39 scats included in population analyses. We also
estimated the quality index for each locus and sample, and
global quality index as recommended by Miquel et al. (2006).
Quality index can range from 0 to 1; the higher the index,
the lower the error rate. We discarded all samples that
had a quality index , 0.80. Allele number (AN), observed
heterozygosity (HO), expected heterozygosity (HE), and
Hardy–Weinberg equilibrium were calculated in GENALEX
6.0.9 (Peakall and Smouse 2006). The sequential Bonferroni
technique was used to correct for multiple comparisons in
Hardy–Weinberg equilibrium tests (Rice 1989). Probability of
identity (Paetkau and Strobeck 1994) was estimated in
GENALEX for unrelated individuals (PID-unr) and siblings
(PID-sibs—Waits et al. 2001).
Population size estimates from scat survey.—Numbers of
snow leopards detected were compared across transects.
Capture–mark–recapture population estimates were inappro-
priate for our sample set because scats were collected in only 1
survey session. Estimating population size by pooling
noninvasive samples into 1 temporal period previously has
provided robust estimates across diverse taxa and study sites
(Eggert et al. 2003; Kohn et al. 1999; Miller et al. 2005; Ruell
et al. 2009); we therefore used this approach. However, one of
the major weaknesses of pooling scat samples is that the
temporal period remains undefined because the length of time
scats persist in the field is uncertain. In the Gobi the time
period represented can be 2–6 months, based on previous field
experience.
The 1st method used was the maximum-likelihood
estimator of CAPWIRE, which calculates the likelihood of
data fitting hypothesized population sizes (Miller et al. 2005).
CAPWIRE takes into account capture heterogeneity by
assigning individuals into 1 of 2 classes (type A and type B)
and estimating relative capture probabilities. The likelihood
ratio test was used to determine if our data fit the equal capture
probability model or the 2 innate rates model with a rejection
value of 0.05 (based on 1,000 bootstrap replicates). Confi-
dence intervals (CIs) around the population size were
estimated from 1,000 bootstrap replicates.
The 2nd method involved the use of rarefaction estimators
that derive an asymptote (the estimate of population size) from
the curve of the cumulative number of individuals detected
versus number of samples genotyped. These approaches do
not incorporate capture heterogeneity. Three proposed formu-
las, originally described by Kohn et al. (Kohn—1999), Eggert
et al. (Eggert—2003), and D. Chessel (Chessel—Valiere
2002), can be used for calculating the rarefaction curve; all
were implemented in GIMLET 1.3.1 (Valiere 2002). GIMLET
was used to generate 1,000 pseudoreplicates from random
reiterations of sampling order. Population estimates were
derived in R version 1.7.1 (Ihaka and Gentleman 1996) from
the asymptote of the rarefaction equations fitting the data from
the pseudoreplicates generated by GIMLET.
To define area surveyed we buffered transects with mean
maximum distance moved (MMDM) and one-half MMDM
between locations where the same individual was genetically
‘‘recaptured’’ (Karanth and Nichols 2002; Wilson and
Anderson 1985). Polygons were generated from buffers and
areas estimated in ArcGIS 9.3 (ESRI, Inc., Redlands,
California). Density was derived from population estimates
divided by area of buffered transects.
Camera-trapping methods.—Two months after scat collec-
tion a camera-trapping survey described by Jackson et al.
(2009) was conducted in the Tost A site (Figs. 1 and 2). The
approach was based on methods refined during a 2-year
camera-trap study undertaken in the Indian Trans-Himalaya
(Jackson et al. 2006). We deployed 18 camera stations, each
774 JOURNAL OF MAMMALOGY Vol. 92, No. 4
2–4 km apart (Fig. 2). Two film (analog) cameras triggered by
a TrailMaster 1550 active-infrared sensor (20-s delay, 24-h
continuous operation; Goodson & Associates, Inc., Lenexa,
Kansas) were concealed within natural rock cairns at each
station. Camera traps were positioned along snow leopard
travel routes (e.g., ridgelines, valley edges, and passes); no
baits or lures were used as attractants. We did not relocate
traps during the survey because of limited access and to
maximize the probability of observing all individuals present
within the surveyed area. Camera traps were checked every 2–
10 days to ensure that they were functional and to replace film.
Results of camera trapping are decribed in Jackson et al.
(2009).
Population size estimates from camera-trapping survey.—
We distinguished individual snow leopards from pelage
patterns using the criteria of Jackson et al. (2006). Program
CAPTURE (Karanth and Nichols 1998; White et al. 1982) was
used to estimate population size with the capture–mark–
recapture approach under 1-day, 3-day, 5-day, and 7-day
sampling occasions; details and results of the capture–mark–
recapture analysis are provided in Jackson et al. (2009). Cubs
were excluded from the analysis. Population closure was
tested with the Stanley and Burnham (1999) method. To
define effective area surveyed and estimate snow leopard
densities, polygons were generated in ArcGIS 9.3 by buffering
individual camera traps and the minimum convex polygon of
camera traps using one-half MMDM and MMDM distances
between successive captures of individuals (Karanth and
Nichols 2002).
Cost comparison of surveys.—Effort put forth during
surveys was estimated by multiplying number of trained
personnel and days required to complete work (i.e., person-
days). We calculated total costs of surveys by summing
expenditures for domestic airfares (Ulaanbaatar to Dalanzad-
gad), ground transportation to study sites, food and accom-
modation for team members, expendable field and laboratory
supplies (e.g., film, batteries, and scat collection kits), and
salaries of local staff. Expenditures in the laboratory were
estimated for DNA extraction and species identification
(extraction kits, PCR, and sequencing), individual identifica-
tion (PCR and microsatellite genotyping), and sex identifica-
tion (PCR and agarose electrophoresis). Remote-sensing
cameras, laboratory equipment, indirect costs, and salaries of
staff from the United States were excluded.
RESULTS
Noninvasive genetic survey.—Thirteen transects were sur-
veyed from 8 to 16 March 2007 (Fig. 1). Of the 180 scats
collected we obtained genetic species identification for 81%
(Table 1). Number of scrapes per kilometer ranged from 3.46 in
Noyon to 10.56 in Tost B (Table 2). Tracks were observed in 2
of 3 sites, with 1.01 tracks/km in Noyon and 0.30 tracks/km in
Tost A. The majority of scats collected were red fox (n 5 83),
and the snow leopard was the next most common species
detected (n 5 60). Three wolf or dog scats were observed.
We genotyped the verified snow leopard scats with 7
microsatellites. We used rigorous criteria for allele sizing and
discarded all ambiguous samples (21 of 60) from the analysis.
For the 39 scats used for population analysis our mean PCR
failure rate was 2.44% (range, 0.85–3.42%), and allele
dropout was 0.83% (range, 0.00–1.79%; Appendix I). Mean
quality index for loci was 0.971 (range, 0.957–0.991) and for
scats also 0.971 (range, 0.857–1.00; Appendix I). Global
quality index was 0.966. Our high quality values were a direct
result of excluding all questionable genotypes from the
analysis. Complete genotypes were obtained for 39 samples
(65%), among which a total of 15 unique genotypes were
observed (Table 1).
All 7 loci were polymorphic, with AN ranging from 2 to 5
(X 5 3.6) and HE from 0.18 to 0.70 (X 5 0.51; Appendix II).
No additional snow leopard samples were available from the
study population that would permit an independent estimate of
allelic frequencies. We therefore estimated allelic frequencies
and genetic parameters from unique genotypes observed
during our survey. All loci were in Hardy–Weinberg
TABLE 1.—Species and individual identity (ID) of putative snow leopard scats collected along transects located in Noyon Uul and Tost Uul of
the Gobi Desert. Species ID refers to the numbers of scats genetically attributed to the listed species. n 5 number of scats, X 5 mean number of
scats per transect, NA 5 not applicable.
Scat
Noyon (4 transects) Tost A (4 transects) Tost B (3 transects) Total (11 transects)
n (%) X n X n (%) X n (%) X
All scats 39 (NA) 9.8 84 (NA) 21.0 57 (NA) 19.0 180 (NA) 16.4
Species ID 32 (82) 8.0 70 (83) 17.5 44 (77) 11.0 146 (81) 13.3
Snow leopard 13 (41) 3.3 33 (47) 8.3 14 (32) 3.5 60 (41) 5.5
Red fox 19 (59) 4.8 37 (53) 9.3 27 (61) 6.8 83 (57) 7.5
Wolf 0 (0) 0 0 (0) 0.0 3 (7) 0.8 3 (2) 0.3
Snow leopard scat
Individual ID 10 (77) 2.5 19 (58) 4.8 10 (71) 2.5 39 (65) 3.6
Individuals 6 (NA) 2.0 5 (NA) 3.0 4 (NA) 2.3 15 (NA) 2.5
Males 3 (NA) 1.0 3 (NA) 2.0 3 (NA) 1.3 9 (NA) 1.5
Females 3 (NA) 1.0 2 (NA) 1.0 1 (NA) 1.0 6 (NA) 1.0
August 2011 JANECKA ET AL.—SURVEYING SNOW LEOPARD POPULATIONS 775
equilibrium. The PID-unr was 0.00016 and PID-sib was 0.01801
with all 7 loci. In conjunction with the Y-linked marker we
identified 9 male and 6 female snow leopards.
Population size estimates.—Fifteen snow leopards were
genetically detected; numbers in each of the 3 sites varied
from 4 (3 males and 1 female) in Tost B to 6 (3 males and 3
females) in Noyon (Table 1). Mean number of captures was
2.7 per snow leopard. Six snow leopards (40%) were observed
only once, and the other 10 a mean of 3.5 times. The most
common male was observed 8 times and the most common
female 5 times. Within the camera-trapped site (Tost A), scat
sampling detected a total of 5 individuals, with 3 males and 1
female observed 3 or more times and 1 female observed only
once. The male that was identified in 8 scats was detected at
this site.
We combined data across all 3 sites to estimate population
size within the entire area sampled. The CAPWIRE program
selected the 2 innate rates model model based on the
difference between the likelihood scores (P 5 0.046).
Population size derived from the maximum-likelihood method
was 19 (95% CI 5 11–26). Mean number of observations per
individual was 2.7, and relative capture probability of type A
versus type B individuals was 4.15. The rarefaction population
size estimates were 16 (95% CI 5 12–19 [Chessel]), 17 (95%
CI 5 12–31 [Eggert]), and 24 (95% CI 5 15–95 [Kohn]).
Previous studies have shown that the Kohn method overes-
timates sizes of small populations (Eggert et al. 2003; Frantz
and Roper 2006; Valiere 2002). In our data set it also produced
unrealistically wide confidence intervals; thus, we excluded it
from subsequent analyses.
We derived density estimates from the CAPWIRE and
Chessel results, which provided an upper and lower range.
Scat transects buffered using one-half MMDM (2.84 km)
yielded survey areas of 155.5 km2 (Noyon), 108.0 km2 (Tost
A), and 59.8 km2 (Tost B), for a total of 323.3 km2. Buffers
provided density estimates of 4.9 (95% CI 5 3.6–5.8) and 5.9
(95% CI 5 3.4–8.0) snow leopards/100 km2. When the 5.69-
km MMDM buffer was used, the total area was 859.2 km2,
yielding densities of 1.8 (95% CI 5 1.3–2.2) and 2.2 (95% CI
5 1.3–3.0) snow leopards/100 km2.
We examined possible effects of violations of population
closure resulting from scats deposited well before our survey
(.2 months prior) by reestimating population size with the
CAPWIRE and Chessel methods after excluding scats
classified as old. During collection, scats were divided into
3 subjective categories: fresh—surface was completely intact
and glossy; recent—majority of surface was intact, very dry,
and slightly cracked; and old—surface was cracked and
visibly deteriorated. Although it was not possible to gauge the
absolute age of scats, this exercise provided insight into the
effects temporal distribution of scat deposition had on
population estimators. We removed 10 scats that were old,
based on our classification scheme. This removal reduced the
number of individuals detected to 13; however, population
estimates were similar to those derived when all scats were
included (CAPWIRE: n 5 19, 95% CI 5 13–29; Chessel: n 5
14, 95% CI 5 10–18).
Camera-trapping effort and capture success.—A contiguous
block of habitat in Tost A (Figs. 1 and 2) was camera trapped
over 65 consecutive days using 18 camera stations (12 May–16
July 2007; 1,114 total trap nights). It took 7 days to deploy all
camera stations. The survey period began after all cameras were
activated. Snow leopards were observed during 47 capture
events (120 total photographs), leading to a capture rate of
10.77 snow leopard photos/100 trap nights (Jackson et al. 2009).
In 13 of 47 capture events a specific individual could not be
identified. Seven snow leopards were detected, including 1
female (SL-1) with 3 cubs (,12 months of age), 2 adult males
(SL-2 and SL-4), and 1 adult (SL-3) of undetermined sex. Three
adults (SL-1, SL-2, and SL-4) and 3 cubs were detected within
the first 4 trap nights. It took an additional 18 nights to detect the
remaining adult snow leopard male, SL-3. The individual
capture rate was 0.63 individuals (including cubs)/100 trap
nights (Jackson et al. 2009).
Capture–mark–recapture history and population esti-
mates.—We captured SL-1 seven times, SL-2 sixteen times,
TABLE 2.—Indexes of snow leopard abundance and effort expended for the 3 sites surveyed using noninvasive genetic techniques in the Gobi
Desert. Sign data were not recorded for 2 of the 5 transects sampled in Noyon. ID 5 identity, NA 5 not applicable.
Abundance index Noyon N (SE) Tost A N (SE) Tost B N (SE) X
Scats (visual ID)/kma 5.13 (2.49) 7.85 (1.00) 7.34 (0.97) 6.77
Scats (genetic ID)/km 1.69 (0.52) 2.74 (0.75) 1.63 (0.69) 2.02
Scrapes/km 3.46 (0.57) 10.45 (3.47) 10.56 (3.03) 8.16
Tracks/km 1.01 (0.32) 0.30 (0.30) 0 0.44
Total signb/km 6.17 (1.15) 13.48 (3.56) 12.20 (3.66) 10.62
Detected individualsc/km 1.08 (0.37) 0.97 (0.21) 0.87 (0.12) 0.97
Detected individuals/100 km2 d 3.86 (NA) 4.63 (NA) 6.68 (NA) 5.06
Effort
Total transect length (km) 8.01 13.39 6.01 9.14
Person-days 4 4 3 3.67
Transect km/100 km2 d 1.00 1.24 1.47 1.24
a Length of transect.b Sum of scrapes, tracks, and snow leopard scats that were genetically verified.c Snow leopards detected by genotyping scat.d Area estimated by buffering transects with one-half mean maximum distance moved (MMDM).
776 JOURNAL OF MAMMALOGY Vol. 92, No. 4
SL-3 twice, and SL-4 four times. Capture rates declined after
40 days of trapping, particularly for SL-1 and her cubs. Cubs
were not detected after the initial 40 days. Fifty-two percent of
captures were made at 4 trap sites that were located on
actively used travel routes with high abundance of snow
leopard sign. Four camera stations (22%) did not detect any
snow leopards (Jackson et al. 2009).
Data for the capture–mark–recapture analysis are provided
in Jackson et al. (2009) and Appendixes III and IV. Although
CAPTURE indicated a closed population over the entire 65-
day period, the Stanley–Burnham test (Stanley and Burnham
1999) supported closure only for ,57 days. Relative fit tests
could not reject the null Mo model (constant capture
probabilities) over the Mt (variation by time) and Mb
(behavioral response) models. Expected values were too small
to test the Mh (heterogeneity) model and Mh versus Mbh
(behavioral response and heterogeneity).
The simplest model, Mo, was most appropriate for our data
based on these tests. The model selection criteria derived by
CAPTURE also indicated that model Mo provided the best fit
(Appendix III). The Mh model, in which capture probabilities
are heterogeneous as a result of factors including sex, age,
activity patterns, and social status, was 2nd (model selection
criteria 5 0.84–0.95; Appendix III). However, small sample
size does not provide sufficient power to discriminate between
null and more complex models; therefore, we estimated
abundance using the 2 best-fitting models. Population size
estimates were 4 (95% CI 5 4–4) adult snow leopards with
both Mo and Mh (Appendix IV). The narrow confidence
interval indicated that we detected all snow leopards present
during the survey. Capture probabilities were relatively high,
ranging from 0.118 for 1-day trapping occasions to 0.563 for
7-day occasions (Appendix IV), further supporting this
conclusion.
The one-half MMDM was 3.38 km, yielding an effective
area of 264 km2 for individually buffered camera stations and
294 km2 when the minimum convex polygon of stations was
buffered. This resulted in densities of 1.5 (95% CI 5 1.5–1.5)
and 1.4 (95% 5 1.4–1.4) snow leopards/100 km2, respectively
(Table 3). Using the MMDM (6.75 km) as a buffer distance
reduced densities to 0.8 (95% CI 5 0.8–0.8) and 0.7 (95% CI
5 0.7–0. 7) snow leopards/100 km2, respectively.
Comparison of noninvasive genetic and camera-trapping
surveys.—Within Tost A, scats were collected on only 2 field
days compared to a total of 65 days of sampling for camera
trapping (Table 4). Total effort was 149 person-days for
camera trapping of 1 site versus 59 person-days for the genetic
survey of 3 sites (Noyon, Tost A, and Tost B). Genetic
analysis costs at Texas A&M University were $10/sample for
extraction and sequencing (n 5 180 scats), $30/sample for
individual identification (60 snow leopard scats), and $3/
sample for sex identification (60 snow leopard scats).
Excluding equipment purchases, air travel, and United States
salaries, our total survey expenditures were $5,230 for the
genetic survey and $10,800 for camera trapping.
DISCUSSION
Noninvasive genetic survey.—Genetic analysis revealed that
the majority of scats collected were deposited by red fox; only
41% were attributed to snow leopards. This high misidenti-
fication rate was observed despite only sampling scats that
were judged to be from our target species based on their
appearance and associated sign. Red fox scats were found
frequently (76.3% of the total) on active snow leopard scrape
sites. Misidentification of carnivore scat has been reported in
previous field studies incorporating genetics (Davison et al.
2002; Farrell et al. 2000; Perez et al. 2006). The error rate for
snow leopards was 35% in Ladakh (Janecka et al. 2008), 51%
in Mongolia (Janecka et al. 2008), and 41% in China and
TABLE 3.—Camera-trapping density estimates of snow leopards in Tost Uul derived with 2 different buffering methods. Data from Jackson et
al. (2009). Density estimates are individuals per 100 km2. NA 5 not applicable.
Buffer Buffer width (km)
MCP buffereda Individual stations buffered
Area (km2) Density 6 95% CI Area (km2) Density 6 95% CI
One-half MMDMb 3.38 294.4 1.4 6 0 264.0 1.4 6 0
MMDM 6.77 552.5 0.7 6 0 536.9 0.8 6 0
Mean home-range radiusc 7.23 593.6 0.7 6 0 NA NA
a MCP 5 minimum convex polygon.b MMDM 5 mean maximum distance moved by snow leopards between photographic captures.c Radius of average snow leopard home range in Mongolia (from McCarthy et al. 2008).
TABLE 4.—Cost comparison for noninvasive genetic and camera-
trapping surveys of snow leopards in the Gobi Desert of Mongolia.
Camera trapping was conducted in only 1 site (Tost A), whereas the
noninvasive genetic survey included 3 sites (Tost A, Tost B,
and Noyon).
Cost Noninvasive genetic survey Camera-trapping survey
Financial ($US)a 5,230b 10,800
Effort (person-days) 59 150
Fieldc 20 133
Data analysis 39 18
No. team members 3 2 or 3
a Excludes international air travel, equipment costs (i.e., cameras, sensors, thermal
cyclers, etc.), and salaries of United States staff.b DNA analysis costs: $10/sample for species identification, $3/sample for sex
identitfication, and $30/sample for individual identification.c Person-days for camera set up, take down, and monitoring throughout the 65-day
survey; genetic survey includes time required for sampling transects and traveling
between collection sites.
August 2011 JANECKA ET AL.—SURVEYING SNOW LEOPARD POPULATIONS 777
Kyrgyzstan (McCarthy et al. 2008), suggesting that this is a
prevalent problem with carnivore fecal surveys.
The proportion of misidentified scats varied among
transects, even within the same site (e.g., 38–89% for Tost
B). Confirmed snow leopard scats per kilometer were 2- to 3-
fold lower than those based on visual identification. Because
snow leopard scat is misidentified frequently in the field, and
the proportion might not be consistent across transects, its use
in surveys should be discounted unless corroborated by
genetic analysis. Trained dogs could be incorporated into
snow leopard surveys to improve species-specific scat
detection (Mackay et al. 2008).
Lowest numbers of scrapes were detected in Noyon. In this
area scrapes per kilometer (3.46 scrapes/km) were similar to
those previously observed in Ladakh, India (up to 2.6 scrapes/
km—Fox et al. 1991) and Nepal (2.46 scrapes/km, range 0.28–
4.96 scrapes/km—Ale et al. 2007). Other studies reported only
total sign per kilometer after incorporating visually identified
scats, making it difficult to make additional comparisons
(Hussain 2003; Lovari et al. 2009; McCarthy et al. 2008).
Scrape frequency was 3-fold higher in the Tost sites than in
Noyon, even though numbers of detected individuals per
kilometer were similar (1.1 versus 0.9 individuals/km, respec-
tively). Jackson et al. (2006) noted that sign indexes likely
underestimate abundance based on their camera-trapping data.
Marking behavior in snow leopards is behaviorally, seasonally,
and site-dependent (Ahlborn and Jackson 1988). Such factors
likely influence scraping frequency, limiting the utility of this
type of sign for indexes of abundance, a conclusion consistent
with the discrepancies we observed.
The genetic survey was effective for documenting presence
of snow leopards. This species was detected on 11 of 13
transects sampled across 3 different sites in 8 days of
fieldwork. We obtained enough individual recaptures to make
preliminary estimates of population size. The maximum-
likelihood estimator of CAPWIRE can in some cases
outperform rarefaction methods because it incorporates
heterogeneity in capture probabilities (Miller et al. 2005).
This is largely dependent on how individual capture
probabilities fit the 2-category model, compared to the equal
probability assumed in rarefaction approaches.
We used both models (rarefaction and 2-category) because
we did not have a priori knowledge of capture probabilities.
Our estimate derived using the heterogeneity model was not
significantly different from the rarefaction results. The long
persistence of scat in dry, cold environments makes it difficult
to determine whether pooling samples into 1 temporal period
violated population closure, potentially inflating size esti-
mates. However, excluding samples defined as old did not
change our results substantially.
Camera-trapping survey.—The camera-trapping rate in Tost
Uul was relatively high (10.77 snow leopard photos/100 nights
compared to 0.09, 0.93, and 2.37 snow leopard photos/100
nights in the Tien Shan Mountains of Kyrgyzstan and China—
McCarthy et al. 2008), and 3.5 snow leopard photos/100
nights in the western Sayan Mountains of Russia (Subbotin
and Istomov 2009). Photographic captures were similar to
those observed over 2 successive years in Ladakh, India (9.41
and 15.11 photos/100 nights—Jackson et al. 2006). Capture
probabilities in our survey ranged from 0.12 for 1-day
sampling occasions to 0.56 for 7-day occasions (see Appendix
IV); in contrast, tiger capture probabilities ranged from 0.04
(3-day occasions) to 0.26 (4- to 6-day occasions) for several
surveys (Karanth et al. 2004; Karanth and Nichols 1998). The
mean interval between snow leopard captures was 9.3 nights,
lower than in the Ladakh study (14.5 nights) and substantially
less than the mean for tiger captures (99.4 nights, 19 studies),
but consistent with intervals for areas where tiger densities
approach 10 individuals/100 km2 (Carbone et al. 2001).
Repeated use of well-defined travel routes by snow leopards in
the Gobi appears to enable high capture rates when
appropriate camera-trapping sites are selected.
Unfortunately, high capture rates and detection probabilities
do not necessarily lead to more robust population estimates. It
also is necessary to obtain sufficient samples over an area
large enough to encompass a significant portion of the
population (Otis et al. 1978; Thompson 2004). This is
especially problematic when surveying rare and difficult to
detect carnivores. The small area we were able to survey
within the 65-day period exemplifies the difficulty of meeting
this goal. The narrow confidence intervals and selection of the
simplest capture model suggest that although we may have
detected nearly all individuals within the specific study site
because of the uniform trap layout and high capture rates, our
sample was likely insufficient for estimating population size
with the capture–mark–recapture model (Otis et al. 1978;
White et al. 1982). In addition, the relevance of our estimate
for overall abundance in the region is difficult to assess,
because the area covered was limited.
White et al. (1982) recommended using closed population
models to estimate population size with more than 20
individuals. Even a cursory review of the literature indicates
that camera-trapping surveys of felids rarely attain this
threshold. There is no snow leopard survey that we are aware
in which .10 individuals were observed during an indepen-
dent session (Jackson et al. 2006; McCarthy et al. 2008;
Subbotin and Istomov 2009). Few surveys of other cat species
have had larger sample sizes. For example, only 3 of 15
camera-trapping surveys of tigers had .20 photo-captured
individuals (Harihar et al. 2009). We found only 1 survey of
jaguars that detected .11 individuals (Soisalo and Cavalcanti
2006). Ocelots numbered ,10 in 3 published camera-trap
studies (Haines et al. 2006; Maffei and Noss 2008; Trolle and
Kery 2003). This is a direct result of the limited resources
researchers have when conducting camera-trapping surveys of
species with inherently low densities.
Comparison of noninvasive genetic and camera-trapping
methods.—Each approach provided information on snow
leopard presence–absence, the minimum number of individ-
uals detected, and estimates of abundance. At 1 of the sites
(Tost A) we conducted both surveys for direct comparison of
results. During the noninvasive genetic survey of this site we
778 JOURNAL OF MAMMALOGY Vol. 92, No. 4
collected 84 scats on 4 transects over the course of only 2 days.
Genetic identification confirmed that 33 originated from snow
leopards. This species was detected on all 4 transects, with a
total of 3 males and 2 females being observed. In Tost A we
found many active scrapes. Two months later the camera-
trapping survey identified 7 individuals (including 3 cubs);
however, it took 11 field days (7 days to deploy cameras and 4
trap nights) to detect 6 and 25 field days to observe all 7.
Recaptures were common with both methods, enabling
population estimates. Population size derived from camera
trapping (4 adults, with additional 3 cubs) was similar to the
number of individuals (5) genetically observed among the
scats collected at this site.
Several important factors must be considered when
comparing camera-trapping and genetic surveys. First, sub-
adults typically are excluded from capture–mark–recapture
estimates. In contrast, age of individuals cannot be inferred
from scat; thus, the data set might include subadults. Camera
trapping also provides date and time of each capture allowing
a definition of sampling periods, but unless the same transects
are sampled repeatedly, no definite temporal bounds exist for
scat surveys. If scats persist in the field for .2 months,
violations to population closure could exist. However, the
number of individuals genetically detected in Tost A was
below that observed by camera trapping, suggesting that
migrants and transients might not significantly affect the
genetic survey. Scat sampling can fail to detect individuals
that have lower marking activity, as was the case with
subadult wolves (Marucco et al. 2009).
Density.—The geographic extent of our study population was
uncertain. This is a problem faced in many wildlife surveys
(Wilson and Anderson 1985), because most cannot cover the
entire geographic area encompassed by a population, often for
logistical reasons alone. Researchers, therefore, typically use
density to compare abundance and population trends, which
requires a careful definition of effective area surveyed (Wilson
and Anderson 1985). One of the most difficult problems to
overcome when monitoring populations is defining this
parameter critical for assessing density and study effort.
A common approach is to buffer the locations sampled
(camera stations or scat transects) with a distance represen-
tative of movement patterns in the area (Wilson and Anderson
1985). Subsequent densities therefore are influenced strongly
by the distribution of the sampled points, the concordance
between the buffered area and home ranges, and the metric
used as the buffer (Balme et al. 2009; Maffei and Noss 2008).
Much debate surrounds how this metric, a surrogate for
movement distances, should be derived (Balme et al. 2009;
Dillon and Kelly 2008; Soisalo and Cavalcanti 2006; Wilson
and Anderson 1985). Previously, studies have shown that
using derivations of the mean of the maximum distance
between successive captures yields reasonable densities
(Jackson et al. 2006; Karanth and Nichols 1998; Silver et al.
2004; Wilson and Anderson 1985).
Some studies also have used mean home-range diameter or
summary statistics of directly observed movements when
telemetry data were available (Balme et al. 2009; Dillon and
Kelly 2008; Soisalo and Cavalcanti 2006). In most snow
leopard field sites telemetry is logistically not feasible
(Jackson and Ahlborn 1989; McCarthy et al. 2005). Applying
telemetry data to different sites or periods could lead to
densities biased by previously observed movement patterns.
We thus chose to use one-half MMDM and MMDM for
estimating our effective areas and densities. In a study of leopards
one-half MMDM provided meaningful density estimates from
camera-trapping data (Balme et al. 2009). However, concurrent
global positioning system telemetry and camera-trapping jaguar
studies found that one-half MMDM underestimated distances
moved, significantly inflating density (Soisalo and Cavalcanti
2006). In another study MMDM buffers resulted in ocelot
densities consistent with radiotracking data (Dillon and Kelly
2008). These discrepancies likely are caused by site-specific
differences in habitat distribution, individual movement, and
distribution of sampling locations.
Our estimates varied from 1.5 individuals/100 km2 from
camera trapping to 5.9 individuals/100 km2 from the
noninvasive genetic survey (using one-half MMDM). Bold
and Dorjzunduy (1976) estimated 4.4 individuals/100 km2 in
Tost Uul, although they did not explain clearly how this
estimate was derived. Snow leopard densities have ranged
from a low of 0.2 individuals/100 km2 in SaryChat
Zapovednik, Kyrgyzstan (McCarthy et al. 2008) to 4.5
individuals/100 km2 (2002) and 8.5 individuals/100 km2
(2003) in Hemis National Park, India (Jackson et al. 2006).
Jackson et al. (2006) noted that narrow spacing of camera
stations in 2003 likely led to an overestimation of density.
Methods used to define effective areas have been inconsistent
among snow leopard surveys. The few studies reporting
density have used either one-half MMDM (Jackson et al.
2009), mean home-range diameter inferred from ungulate
density (McCarthy et al. 2008), or an ad hoc estimation of
study area size (Lovari et al. 2009). This lack of consistency
prevents direct comparisons. A method of defining study areas
should be used consistently across snow leopard surveys
conducted by different research groups.
Despite comparable one-half MMDM distances from scat
and photo captures (2.84 km versus 3.38 km, respectively), the
effective areas covered by the 2 surveys contributed to
substantial differences in density. Densities from the genetic
survey were nearly 4-fold greater than from camera trapping
(,5.4/100 km2 versus 1.4/100 km2). This can be attributed
largely to 2 main factors. First, cubs were excluded in
population estimates from camera trapping; in contrast, all
detected individuals were used in the genetic survey. Second,
the broader spacing of camera stations compared to scats
collected along transects resulted in a greater effective area.
This is clearly seen in Tost A, where the camera survey
covered a 2.5-fold larger area (Fig. 2). In addition, multiple
snow leopards frequently mark and deposit scats in common
sites such as prominent outcrops and saddles (Ahlborn and
Jackson 1988). When transects are selected preferentially to
cover these sites, a large portion of the population can be
August 2011 JANECKA ET AL.—SURVEYING SNOW LEOPARD POPULATIONS 779
detected within a relatively small area. Finally, the persistence
of scat in cold, dry environments makes it difficult to define
the length of the survey when collected scats are pooled into 1
sampling period, although excluding scats judged to be old did
not significantly change our estimates. The combination of
these factors, particularly the reduced size of the area covered
in each site (Fig. 2), seems to inflate density derived from scat
transects. To minimize such bias in noninvasive genetic
surveys scat transects need to be separated spatially and
oriented to maximize the area surveyed. It can be advanta-
geous to sample a greater number of shorter transects (Ahlborn
and Jackson 1988). These modifications should be combined
with a stratified sampling design that ensures a more uniform
distribution of sampled points than in our study.
Because of the uncertainties involved in deriving density
estimates, Jackson et al. (2009) recommended that surveys
report basic capture indexes (e.g., minimum number of
individuals detected) and trapping effort (e.g., photographs
per 100 trap nights). Similarly, some standardized metrics also
should be used in genetic surveys. We propose reporting
number of genetically verified snow leopard scats per
kilometer and individual snow leopards per kilometer, along
with derived densities. Information on the spatial distribution
of transects and effort per site defined as transect kilometer per
square kilometer also is necessary for meaningful compari-
sons. This would facilitate comparisons between surveys
undertaken in different areas and habitats without relying on
possibly questionable density estimates.
Logistical advantages and disadvantages of methods.—To
attain the target of 20 individuals proposed by White et al.
(1982) a snow leopard survey must cover a large contiguous
area. In regions where densities are ,3 snow leopards/100 km2
field teams would need to survey approximately 700 km2 while
avoiding home-range–size gaps. Given poor access, rough
terrain, and often high altitude of snow leopard habitat, this
presents a logistically daunting task. Assuming 1.5 camera
stations/25 km2, availability of 20 stations (each with 2
cameras), and the ability to set 3 stations daily, it would take
approximately 7 days to set 20 stations covering only 350 km2.
After six 3-day sampling occasions (areas with low capture
probability would likely require more) it would take another
10 days to move the stations and survey an additional 350 km2
over another 18-day period. Seven more days would be required
to take down the stations. A field team would have difficulty
covering ,700 km2 within 60 days. In contrast, a noninvasive
genetic survey could be conducted in a shorter time frame.
Assuming 2 transects within each 25-km2 block, 28 blocks
sampled at a rate of 1 block per day, and 4 additional days for
travel between more distant sites, a genetic survey could cover
the same ,700-km2 area within approximately 32 days.
Noninvasive genetic surveys do not require specialized or
expensive field equipment. The ability to complete transects
rapidly, and the relatively high detection rates, enable more
sites and larger areas to be surveyed. Further, scat surveys
yield DNA samples of populations, thereby greatly expanding
the scope of information that can be generated. The cost also is
substantially lower, in our case ,50% that of camera trapping
even though we sampled 2 additional sites. However, genetic
surveys involve sophisticated laboratory facilities, trained
personnel, and additional time and financial cost to complete
analysis after the field session. Some countries within the
distribution of the snow leopard currently lack the necessary
laboratories for genetic analysis and have restrictions on the
export of fecal material.
Conservation implications.—Quantitative status and distri-
bution surveys of snow leopards are notably difficult compared
to those of other rare carnivores, such as tigers, leopards, or
jaguars that tend to have greater road access in their study areas.
Camera trapping has offered a major advance for research and
conservation of snow leopards. However, this approach is
hampered by the rugged terrain, poor access, field logistics, and
subsequent small sample sizes. Our results suggest that
population estimates in larger areas are more attainable with
noninvasive genetic surveys because of the ability to detect
multiple individuals with less field effort.
Examination of genetic data also can offer invaluable
insight into relatedness, dispersal, population structure, and
landscape connectivity (Schwartz et al. 2007), which is critical
for conservation planning. The information camera trapping
can provide on these topics is limited. Samples collected
during wide-ranging, noninvasive genetic surveys can be used
to address these population-level questions. This approach will
be useful for identifying core snow leopard populations and
landscape-level mapping of corridors that likely play a key
role in the long-term persistence of snow leopards, particularly
in disjunct mountain ranges.
More research is needed to establish optimal strategies for
scat sampling, given different snow leopard densities, habitat
factors, and terrain conditions. Behavior and social structure
influence marking and scat deposition and need to be
understood to model capture probabilites (Marucco et al.
2009). Concurrent genetic surveys and telemetry studies could
explore various sampling strategies under specific environ-
mental and biological conditions (Balme et al. 2009).
However, because of financial and logistical constraints, such
studies will be limited to only a few sites. For results to be
widely applicable research among groups should be coordi-
nated so that the selected sites are representative of snow
leopard populations across most of the range.
Our results suggest that the collection and genetic analysis
of scats offers a more efficient means than camera trapping for
detecting snow leopards in terms of field effort, total effort,
and financial expenditure. The main factor is the reduction in
time required to sample populations. Time spent in the field by
researchers in many areas of central Asia occupied by snow
leopards is severely constrained. Sample size and number of
sampled sites principally limit population surveys. The ability
to collect a greater number of samples over a larger area can
yield better population estimates. Noninvasive genetics,
therefore, has a greater potential for monitoring snow leopards
on a larger scale than camera trapping. The main challenges
that need to be addressed are factors that influence scat
780 JOURNAL OF MAMMALOGY Vol. 92, No. 4
deposition, violations to population closure, absence of age
information, optimal sampling schemes, and insufficient
capacity for genetic analysis in many countries within the
distribution of the snow leopard.
ACKNOWLEDGMENTS
We appreciate funding generously provided by the Snow Leopard
Conservancy, The National Geographic Society, Leonard X. Bosack
Foundation, Bette M. Kruger Charitable Foundation, Wildlife
Conservation Network, California Institute of Environmental Studies,
Ken and Gabrielle Adelman, Texas A&M University, and an
anonymous individual. We thank the Institute of Biology, Mongolian
Academy of Sciences, and Irbis Mongolia Center for providing permits
and in-country staff. We are grateful for assistance provided in the field
by N. Mijiddorj, Dorji, and Shaarav, and aid from J. Robinson, A.
Royle, J. Hines (Patuxent Wildlife Research Center, United States
Geological Survey), K. Nowell (Cat Action Treasury), and E. Williams
and J. Jencek (San Francisco Zoo). We thank K. Logan and 1
anonymous reviewer for helpful suggestions for our manuscript. San
Francisco Zoo generously provided samples for positive controls. Scat
samples were exported from Mongolia under Mongolia State
Specialized Supervision Inspectorate Agency Permit 271470.
LITERATURE CITED
AHLBORN, G., AND R. JACKSON. 1988. Marking in free-ranging snow
leopards in west Nepal: a preliminary assessment. Pp. 25–49 in
Fifth International Snow Leopard Symposium (H. Freeman, ed.).
International Snow Leopard Trust and Wildlife Institute of India,
Dehradun, India.
ALE, S. B., P. YONZON, AND K. THAPA. 2007. Recovery of snow leopard
Uncia uncia in Sagarmantha (Mount Everest). Oryx 41:89–92.
BALME, G. A., L. T. B. HUNTER, AND R. SLOTOW. 2009. Evaluating
methods for counting cryptic carnivores. Journal of Wildlife
Management 73:433–441.
BANNIKOV, A. G. 1954. Mlekopitayushchie Mongol’skoi Marodnoi
Respubliki (Mammals of the Mongolian People’s Republic).
Academy of Sciences, Moscow, Russia.
BOLD A., AND S. DORJZUNDUY. 1976. Report on snow leopards in the
southern spurs of the Gobi Altai. Proceedings of the Institute of
General and Experimental Biology—Ulaanbaatar 11:27–43.
CARBONE, C., ET AL. 2001. The use of photographic rates to estimate
densities of tiger and other cryptic mammals. Animal Conservation
4:75–79.
CHENG, Y., M. TSUBO, T. Y. ITO, E. NISHIHARA, AND M. SHINODA. 2011.
Impact of rainfall variability and grazing pressure on plant diversity
in Mongolian grasslands. Journal of Arid Environments 75:471–476.
DAVISON, A., J. D. S. BIRKS, R. C. BROOKES, T. C. BRAITHEWAITE, AND J.
E. MESSENGER. 2002. On the origin of faeces: morphological versus
molecular methods for surveying rare carnivores from their scats.
Journal of Zoology (London) 257:141–143.
DILLON, A., AND M. J. KELLY. 2008. Ocelot home range, overlap and
density: comparing radio telemetry with camera-trapping. Journal
of Zoology (London) 275:391–398.
EGGERT, L., J. EGGERT, AND D. WOODRUFF. 2003. Estimating
population sizes for elusive animals: the forest elephants of
Kakum National Park, Ghana. Molecular Ecology 12:1389–1402.
FARRELL, L. E., J. ROMAN, AND M. E. SUNQUIST. 2000. Dietary
separation of sympatric carnivores identified by molecular analysis
of scats. Molecular Ecology 9:1583–1590.
FOX, J. L., S. P. SINHA, R. S. CHUNDAWAT, AND P. K. DAS. 1991. Status
of the snow leopard Panthera uncia in northwest India. Biological
Conservation 55:283–298.
FRANTZ, A. C., AND T. J. ROPER. 2006. Simulations to assess the
performance of different rarefaction methods in estimating
population size using small datasets. Conservation Genetics
7:315–318.
HAINES, A. M., J. E. JANECKA, M. E. TEWES, L. I. GRASSMAN, AND P.
MORTON. 2006. The importance of private lands for ocelot
Leopardus pardalis conservation in the United States. Oryx 40:1–5.
HARIHAR, A., B. PANDAV, AND S. K. GYAL. 2009. Subsampling
photographic capture–recapture data of tigers (Panthera tigris) to
minimize closure violation and improve estimate precision: a case
study. Population Biology 51:471–479.
HUSSAIN, S. 2003. The status of snow leopard in Pakistan and its
conflict with local farmers. Oryx 37:26–33.
IHAKA, R., AND R. GENTLEMAN. 1996. R: a language for data analysis
and graphics. Journal of Computational and Graphical Statistics
5:299–314.
JACKSON, R., AND G. AHLBORN. 1989. Snow leopards (Panthera uncia)
in Nepal: home range and movements. National Geographic
Research 5:161–175.
JACKSON, R., AND J. L. FOX. 1997. Snow leopard conservation:
accomplishments and research priorities. Pp. 128–145 in Proceed-
ings of the 8th International Snow Leopard Symposium (R.
Jackson and A. Ahmad, eds.). International Snow Leopard Trust
and WWF–Pakistan, Seattle, Washington.
JACKSON, R., AND D. O. HUNTER. 1996. Snow leopard survey and
conservation handbook. International Snow Leopard Trust and
United States Geological Survey, Biological Resources Division,
Fort Collins, Colorado.
JACKSON, R. M., B. MUNKHTSOG, D. P. MALLON, G. NARANBAATAR, AND
K. GERELMAA. 2009. Camera-trapping snow leopards in the Tost
Uul region of Mongolia. Cat News 51:18–21.
JACKSON, R., J. D. ROE, R. WANGCHUK, AND D. O. HUNTER. 2006.
Estimating snow leopard population abundance using photography and
capture–recapture techniques. Wildlife Society Bulletin 34:772–781.
JANECKA, J. E., ET AL. 2008. Population monitoring of snow leopards
using noninvasive collection of scat samples: a pilot study. Animal
Conservation 11:401–411.
KARANTH, K. U., R. S. CHUNDAWAT, J. D. NICHOLS, AND N. S. KUMAR.
2004. Estimation of tiger densities in the tropical dry forests of
Panna, Central India, using photographic capture–recapture
sampling. Animal Conservation 7:285–290.
KARANTH, K. U., AND J. D. NICHOLS. 1998. Estimation of tiger
densities in India using photographic captures and recaptures.
Ecology 79:2852–2862.
KARANTH, K. U., AND J. D. NICHOLS. 2002. Monitoring tigers and their
prey: a manual for researchers, managers and conservationists in
tropical Asia. Centre for Wildlife Studies, Bangalore, India.
KOHN, M., E. YORK, D. KAMRADT, G. HAUGHT, R. SAUVAJOT, AND R.
WAYNE. 1999. Estimating population size by genotyping faeces.
Proceedings of the Royal Society of London, B. Biological
Sciences 266:657–663.
LOVARI, S., ETAL. 2009. Restoring a keystone predator may endanger a prey
species in a human-altered ecosystem: the return of the snow leopard to
Sagarmatha National Park. Animal Conservation 12:559–570.
MACKAY, P., D. A. SMITH, R. LONG, AND M. PARKER. 2008. Scat
detection dogs. Pp. 183–222 in Noninvasive survey methods for
carnivores (R. Long, P. Mackay, J. Ray, and W. Zielinski, eds.).
Island Press, Washington, D.C.
August 2011 JANECKA ET AL.—SURVEYING SNOW LEOPARD POPULATIONS 781
MAFFEI, L., AND N. J. NOSS. 2008. How small is too small? Camera
trap survey areas and density estimates for ocelots in the Bolivian
Chaco. Biotropica 40:71–75.
MALLON, D. P. 1991. Status and conservation of large mammals in
Ladakh. Biological Conservation 56:110–119.
MARUCCO, F., D. H. PLETSCHER, L. BOITANI, M. K. SCHWARTZ, K. L.
PILGRIM, AND J. D. LEBRETON. 2009. Wolf survival and population
trend using non-invasive capture–recapture techniques in the
Western Alps. Journal of Applied Ecology 46:1003–1010.
MCCARTHY, K., T. FULLER, M. MING, T. M. MCCARTHY, L. WAITS, AND
K. JUMABAEV. 2008. Assessing estimators of snow leopard
abundance. Journal of Wildlife Management 72:1826–1833.
MCCARTHY, T. M., AND G. CHAPRON (EDS.). 2003. Snow leopard
survival strategy. International Snow Leopard Trust and the Snow
Leopard Network, Seattle, Washington.
MCCARTHY, T. M., T. K. FULLER, AND B. MUNKHTSOG. 2005.
Movements and activities of snow leopards in southwestern
Mongolia. Biological Conservation 124:527–537.
MCCARTHY, T. M., AND B. MUNKHTSOG. 1997. Preliminary assessment
of snow leopard sign surveys in Mongolia. Pp. 57–65 in Eighth
International Snow Leopard Symposium (R. Jackson and A.
Ahmad, eds.). International Snow Leopard Trust and WWF–
Pakistan, Seattle, Washington.
MENOTTI-RAYMOND, M., ET AL. 1999. A genetic linkage map of
microsatellites in the domestic cat (Felis catus). Genomics 57:9–23.
MILLER, C. R., P. JOYCE, AND L. P. WAITS. 2005. A new method for
estimating the size of small populations from genetic mark–
recapture data. Molecular Ecology 14:1991–2005.
MIQUEL, C., E. BELLEMAIN, C. POILLOT, J. BESSIERE, A. DURAND, AND P.
TABERLET. 2006. Quality indexes to asses the reliability of
genotypes in studies using noninvasive sampling and multiple-
tube approach. Molecular Ecology Notes 6:985–988.
MONDOL, S., K. U. KARANTH, N. B. KUMAR, A. M. GOPALASWAMY, A.
ANDHERIA, AND U. RAMAKRISHNA. 2009. Evaluation of non-invasive
genetic sampling methods for estimating tiger population size.
Biological Conservation 142:2350–2360.
MURPHY, W. J., S. SUN, Z. Q. CHEN, J. PECON-SLATTERY, AND S. J.
O’BRIEN. 1999. Extensive conservation of sex chromosome
organization between cat and human revealed by parallel radiation
hybrid mapping. Genome Research 9:1223–1230.
NOWELL, K., AND P. JACKSON. 1996. Wild cats: status survey and
action plan. IUCN—World Conservation Union, Gland, Switzer-
land.
OTIS, D. L., K. P. BURNHAM, G. C. WHITE, AND D. R. ANDERSON. 1978.
Statistical inference from capture data on closed animal popula-
tions. Wildlife Monographs 62:1–135.
PAETKAU, D., AND C. STROBECK. 1994. Microsatellite analysis of
genetic-variation in black bear populations. Molecular Ecology
3:489–495.
PEAKALL, R., AND P. E. SMOUSE. 2006. GENEALEX 6: genetic
analysis in Excel. Population genetic software for teaching and
research. Molecular Ecology Notes 6:288–295.
PEREZ, I., E. GEFFEN, AND O. MOKADY. 2006. Critically endangered
Arabian leopards Panthera pardus nimr in Israel: estimating
population parameters using molecular scatology. Oryx 40:295–301.
POLLOCK, K. H., J. D. NICHOLS, C. BROWNIE, AND J. E. HINES. 1990.
Statistical inference for capture–recapture experiments. Wildlife
Monographs 107:1–97.
RICE, W. R. 1989. Analyzing tables of statistical tests. Evolution
43:223–225.
RUELL, E. W., S. P. RILEY, M. R. DOUGLAS, J. P. POLLINGER, AND K. R.
CROOKS. 2009. Estimating bobcat population sizes and densities in
a fragmented urban landscape using noninvasive capture–recapture
sampling. Journal of Mammalogy 90:129–135.
SCHALLER, G. B., H. LI, T. REN, AND M. QIU. 1987. The snow leopard
in Xinjiang, China. Oryx 22:197–204.
SCHALLER, G. B., J. R. REN, AND M. J. QIU. 1988. Status of the snow
leopard Panthera uncia in Qinghai and Gansu provinces, China.
Biological Conservation 45:179–194.
SCHWARTZ, M. K., G. LUIKART, AND R. S. WAPLES. 2007. Genetic
monitoring as a promising tool for conservation and management.
Trends in Ecology & Evolution 22:25–33.
SILVER, S. C., ET AL. 2004. The use of camera traps for estimating
jaguar (Panthera onca) abundance and density using capture/
recapture analysis. Oryx 38:148–154.
SOISALO, M. K., AND S. M. C. CAVALCANTI. 2006. Estimating the
density of a jaguar population in the Brazilian Pantanal using
camera-traps and capture–recapture sampling in combination with
GPS radio-telemetry. Biological Conservation 129:487–496.
STANLEY, T. R., AND K. P. BURNHAM. 1999. A closure test for time-
specific capture–recapture data. Environmental and Ecological
Statistics 6:197–209.
STERNBERG, T., D. THOMAS, AND N. MIDDLETON. In press. Drought
dynamics on the Mongolian steppe, 1970–2006. International
Journal of Climatology.
SUBBOTIN, A. E., AND S. V. ISTOMOV. 2009. The population and status
of snow leopards Uncia uncia (Felidae, Carnivora) in the western
Sayan mountain range. General Biology 6:846–849.
SWOFFORD, D. L. 2003. PAUP*: phylogenetic analysis using
parsimony (*and other methods), version 4b 10. Sinauer Associ-
ates, Inc., Publishers, Sunderland, Massachusetts.
THOMPSON, W. L. (ED.). 2004. Sampling rare or elusive species:
concepts, designs and techniques for estimating population
parameters. Island Press, Washington, D.C.
TROLLE, M., AND M. KERY. 2003. Estimation of ocelot density in the
Pantanal using capture–recapture analysis of camera-trapping data.
Journal of Mammalogy 84:607–614.
VALIERE, N. 2002. GIMLET: a computer program for analysing genetic
individual identification data. Molecular Ecology Notes 2:377–379.
WAITS, L. P., V. A. BUCKLEY-BEASON, W. E. JOHNSON, D. ONORATO,
AND T. MCCARTHY. 2007. A select panel of polymorphic
microsatellite loci for individual identification of snow leopards
(Panthera uncia). Molecular Ecology Notes 7:311–314.
WAITS, L. P., G. LUIKART, AND P. TABERLET. 2001. Estimating the
probability of identity among genotypes in natural populations:
cautions and guidelines. Molecular Ecology 10:249–256.
WAITS, L. P., AND D. PAETKAU. 2005. Noninvasive genetic sampling
tools for wildlife biologists: a review of applications and
recommendations for accurate data collection. Journal of Wildlife
Management 69:1419–1433.
WHITE, G. C., D. R. ANDERSON, K. P. BURNHAM, AND D. L. OTIS. 1982.
Capture–recapture and removal methods for sampling closed popula-
tions. Los Alamos National Laboratory, Los Alamos, New Mexico.
WILSON, K. R., AND D. R. ANDERSON. 1985. Evaluation of two density
estimators of small population size. Journal of Mammalogy 66:13–21.
XU, A., ET AL. 2008. Status and conservation of snow leopard Panthera
uncia in theGouliRegion,KunlunMountains,China.Oryx42:460–463.
Submitted 2 February 2010. Accepted 21 January 2011.
Associate Editor was Jonathan A. Jenks.
782 JOURNAL OF MAMMALOGY Vol. 92, No. 4
APPENDIX I
Microsatellite genotyping error rates observed among 39 snow
leopard scats individually identified in our study and used to estimate
population size. The quality index was estimated using the methods
of Miquel et al. (2006). PCR 5 polymerase chain reaction.
Locus PCR failure (%) Allele dropout (%) Quality index
PUN082 3.42 1.11 0.957
PUN100 0.85 0.00 0.991
PUN124 3.42 0.96 0.957
PUN132 1.71 0.00 0.983
PUN225 2.56 1.79 0.966
PUN229 2.56 1.10 0.966
PUN327 2.56 0.00 0.974
X 2.44 0.83 0.971
APPENDIX II
Microsatellite diversity observed among snow leopards sampled
with scats in Noyon Uul and Tost Uul of the Gobi Desert. AN 5 total
number of alleles, AE 5 effective number of alleles, HO 5 observed
heterozygosity, HE 5 expected heterozygosity, HWE (P) 5 Hardy–
Weinberg equilibrium P-value, PID-unr 5 probability 2 unrelated
individuals will have an identical genotype.
Locus AN AE HO HE HWE (P) PID-unr
PUN082 2 2.0 0.67 0.50 0.19 0.37
PUN100 5 2.8 0.60 0.65 0.56 0.17
PUN124 4 3.4 0.93 0.70 0.44 0.15
PUN132 3 1.2 0.20 0.18 0.98 0.67
PUN225 3 2.1 0.53 0.53 0.83 0.33
PUN229 5 2.9 0.60 0.66 0.43 0.18
PUN327 3 1.6 0.47 0.37 0.71 0.44
X 3.6 2.3 0.57 0.51 0.59 0.33
APPENDIX III
Selection criterion values estimated by program CAPTURE for
different capture probability models. Mo 5 constant capture
probabilities; Mh 5 heterogeneity; Mb 5 behavioral response; Mbh
5 behavioral response and heterogeneity; Mt 5 variation by time;
Mth 5 variation by time and heterogeneity; Mtb 5 variation by time
and behavioral response; Mtbh 5 variation by time, behavioral
response, and heterogeneity. Data from Jackson et al. (2009).
Sampling
Occasion
Capture model
Mo Mh Mb Mbh Mt Mth Mtb Mtbh
1-day 1.00 0.95 0.37 0.60 0.00 0.31 0.44 0.64
3-day 1.00 0.95 0.41 0.65 0.00 0.34 0.46 0.69
5-day 1.00 0.84 0.26 0.58 0.00 0.42 0.31 0.65
7-day 1.00 0.80 0.23 0.55 0.00 0.40 0.26 0.63
APPENDIX IV
Capture probability and population size estimates of snow leopards camera trapped in Tost Uul using the 2 best-fitting capture models. Mo 5
constant capture probabilities, Mh 5 heterogeneity. Data from Jackson et al. (2009).
Sampling occasion
Mo Mh
Capture probability
Estimated abundance
6 SE (95% CI) Capture probability
Estimated abundance
6 SE (95% CI)
1-day 0.118 4 6 0.063 (4–4) 0.118 4 6 0.375 (4–4)
3-day 0.303 4 6 0.066 (4–4) 0.303 4 6 0.266 (4–4)
5-day 0.432 4 6 0.090 (4–4) 0.432 4 6 0.215 (4–4)
7-day 0.563 4 6 0.080 (4–4) 0.563 4 6 0.165 (4–4)
August 2011 JANECKA ET AL.—SURVEYING SNOW LEOPARD POPULATIONS 783
Recommended