Management and Conservation Article

Argali Abundance in the Afghan Pamir
Using Capture–Recapture Modeling
From Fecal DNA

RICHARD B. HARRIS,1 Department of Ecosystem and Conservation Science, University of Montana, Missoula, MT 59812, USA

JOHN WINNIE, JR.,2 Department of Ecosystem and Conservation Science, University of Montana, Missoula, MT 59812, USA

STEPHEN J. AMISH, Department of Biological Sciences, University of Montana, Missoula, MT 59812, USA

ALBANO BEJA-PEREIRA, Centro de Investigação em Biodiversidade e Recursos Genéticos, Universidade do Porto, Campus Agrário de Vairão, 4485-661
Vairão, Portugal

RAQUEL GODINHO, Centro de Investigação em Biodiversidade e Recursos Genéticos, Universidade do Porto, Campus Agrário de Vairão,
4485-661 Vairão, Portugal

VÂNIA COSTA, Centro de Investigação em Biodiversidade e Recursos Genéticos, Universidade do Porto, Campus Agrário de Vairão,
4485-661 Vairão, Portugal

GORDON LUIKART,3 Department of Biological Sciences, University of Montana, Missoula, MT 59812, USA, and Centro de Investigação em
Biodiversidade e Recursos Genéticos, Universidade do Porto, Campus Agrário de Vairão, 4485-661 Vairão, Portugal

ABSTRACT Estimating population size in a mark–recapture framework using DNA obtained from remotely collected genetic samples
(e.g., feces) has become common in recent years but rarely has been used for ungulates. Using DNA extracted from fecal pellets, we estimated

the size of an argali (Ovis ammon) population that was believed to be isolated from others within the Big Pamir Mountains, Afghanistan, an

area where access was difficult and expensive. We used closed-capture models to estimate abundance, and Pradel models to examine closure

assumptions, both as implemented in Program MARK. We also made visual counts of argali in the Big Pamirs, allowing comparison of count

indices of abundance with modeled estimates. Our model-averaged estimate for female argali in the Big Pamir was 172 (95% CI 5 117–232),

which was about 23% higher than our best assessment using uncorrected visual counts. However, mark–recapture models suggested that males

were not a closed population; thus, we were unable to provide a meaningful estimate of overall population size. Males either suffered much

higher mortality than females during the sampling period, or, more likely, males moved in and out of the Big Pamir area. Although information

from DNA did not provide a clear overall population estimate, it suggested that the Big Pamir was not isolated from other argali populations,

which could not have been confirmed with visual observations alone. Estimating argali population size using mark–recapture models and fecal

DNA is feasible but may be too expensive for frequent monitoring of large and remote populations. Our study demonstrates the importance of

sex identification and separate abundance estimation for each sex, especially if movement ecology differs by sex.

KEY WORDS abundance estimate, Afghanistan, argali, fecal samples, mark–recapture, noninvasive sampling, Ovis ammon.

Estimating population size using well-established statistical
techniques that deal with imperfect detectability is generally
considered preferable to uncorrected counts (Anderson
2001, Williams et al. 2002, White 2005). Yet for many
wide-ranging ungulate species, particularly those inhabiting
remote, mountainous habitats, uncorrected index counts
remain the staple (e.g., Magomedov et al. 2003, Harris and
Loggers 2004, Schaller and Kang 2008). The 2 most
common classes of models to deal with imperfect detect-
ability are distance sampling, and physically marking
animals and later recapturing or resighting them (capture–
mark–recapture [CMR]). For mountain ungulates, both
distance sampling and CMR are logistically difficult to
apply without gross violations of crucial assumptions, very
small samples sizes, or both. Although having an estimate of
abundance is not necessarily as important as having a
measure of population trend or an understanding of the
important covariates of population vigor, there are circum-
stances in which simply having a population estimate is

important (sensu Caughley 1977). Conservation and
management options for small or isolated populations often
depend on knowing the approximate population size,
particularly when these populations are of interest to legal
or illegal hunters.

The argali (Ovis ammon) is the epitome of a species for
which obtaining a population estimate, as differentiated from
an abundance index, remains a largely unresolved challenge.
Counts of individual argali are more easily obtained than for
many other species, so visual counts have usually been the
basis of abundance assessments. However, argali are also
capable of long-distance movements and usually move away
from observers at distances far in excess of those that allow
individual recognition. Argali are group-living ungulates, but
group size and composition are usually fluid, with individuals
often leaving or joining groups daily (Schaller 1998,
Fedosenko and Blank 2005, Harris 2007). With the
exception of a few distinctive older rams or occasional
animals with deformities or distinctive markings, argali
within broad age and sex classes look alike. This, together
with the difficulty of approaching argali, makes identification
of individuals by direct observation subject to great
uncertainty. Thus, although argali are easy to count,
interpreting these counts as population sizes is fraught with
error: some animals go undetected in any survey (thus biasing

1
E-mail: rharris@montana.com

2
Present address: Ecology Department, Montana State University,

Bozeman, MT 59715, USA
3

Present address: Flathead Lake Biological Station, University of
Montana, Polson, MT 59860, USA

Journal of Wildlife Management 74(4):668–677; 2010; DOI: 10.2193/2009-292

668 The Journal of Wildlife Management N 74(4)



counts low), whereas argali roaming behavior may cause
duplicate counts of individuals (thus biasing counts high).

Estimating abundance with CMR models using DNA
microsatellites for individual identification has attracted
interest in recent years (Lukacs and Burnham 2005a). Using
DNA is attractive because it can often be obtained without
physically handling individual animals (Taberlet et al. 1999).
Advances in field study design (Boulanger et al. 2004, 2008)
and the ability of models to account for both field and
laboratory limitations (Lukacs and Burnham 2005b, Petit
and Valiere 2006, Lukacs et al. 2007, Knapp et al. 2009)
continue to be made. The use of CMR with remotely
collected samples (sensu Garshelis 2006) to estimate
abundance has become common among some taxa but
remains uncommon for others. Bears (Ursus spp.) pose
particular problems to investigators interested in abundance,
but bears can easily be induced to provide hair samples, so
DNA-based CMR has become a standard part of bear
biologists’ tool kit (e.g., Bellemain et al. 2005, Solberg et al.
2006, Kendall et al. 2008). Most other uses of DNA-based
CMR have been for other carnivores, including canids
(Creel et al. 2003, Prugh et al. 2005), felids (Perez et al.
2006, Ruell et al. 2009), and mustelids (Wilson et al. 2003,
Mulders et al. 2007, Williams et al. 2009). In theory,
remotely based CMR estimation could be used for any
species (e.g., raptors, primates; Rudnick et al. 2008,
Guschanski et al. 2009), but its use for free-ranging
artiodactyls has thus far been uncommon (but see Fickel
and Hohmann 2006, Valiere et al. 2006).

As part of a broader study of the conservation status of
argali in the Islamic Republic of Afghanistan, our
objectives were to estimate abundance of argali (O. ammon
polii, ‘‘Marco Polo sheep’’) in a portion of the Wakhan
District within Badakhshan Province locally termed Big
Pamir. The Big Pamir was historically reserved as a
hunting area for King Zahir of Afghanistan and during
the late 1960s and 1970s was named the Pamir-i-Buzurg
(Big Pamir) Wildlife Reserve and operated by the Afghan
Tourist Organization (Petocz 1973, 1978). Based on
previous surveys and anecdotal information, we suspected
that argali in the Big Pamir range had declined
considerably from their abundance during the 1970s and
that argali were largely isolated (Petocz 1973, Petocz et al.
1978; G. B. Schaller, Wildlife Conservation Society, unpub-
lished report). There were no recent reports of argali
inhabiting areas to the south of the Big Pamir (either within
Afghanistan’s Wakhan District or in the Hindu Kush range
forming the border with Pakistan) or elsewhere to the west
within Badakhshan Province (Habibi 1997, Fitzherbert and
Mishra 2003). Similarly, reports suggested only occasional
individual sightings of argali to the east of the Big Pamir to
approximately 37u209N, 74u209E, where a larger, seemingly
more robust population was known to inhabit the Little Pamir
Mountains some 120 km away (G. B. Schaller, unpublished
report; B. Habib, Wildlife Conservation Society, unpublished
report). It was unknown whether argali in the Big Pamir were
demographically or genetically linked with animals to the
north of the Panj (Amu Darya) River in Tajikistan.

STUDY AREA

The Wakhan Corridor was a 15–68-km-wide section of
Afghanistan extending southwest to northeast in the far
northeast of the country (Fig. 1). Bordered on the north by
Tajikistan, China on the east, and Pakistan to the south, the
region geographically termed the Pamir Knot, formed by the
confluence of the Pamir, Hindu Kush, and Karakoram
mountain ranges, was characterized by high, broad valleys
(pamirs) bordered by steep, rugged mountains.

The approximately 1,600-km2 Big Pamir study area
(Fig. 1) was located roughly at the midpoint of the Wakhan
Corridor and bordered Tajikistan to the north. Valleys
bordered by high ridges ascend from the Panj River (which
downstream is called the Amu Darya) in the northwest
(approx. 3,000 m) to the crest of the Big Pamir range that
rises to .6,400 m to the southeast. At their lowest, near the
Panj River, the mountains and ridges are generally rounded,
punctuated by heavily eroded, steep gullies. To the south-
east, at higher elevations, the terrain is rugged and steep,
with occasional small lakes and ponds in the upper valleys
and numerous glaciers and permanent snowfields on higher
mountain slopes and peaks.

South- and southwest-facing slopes were dominated by
sparse grass (primarily Agrostis, Poa, Festuca, and Agropyron
spp.), and sage (Artemisia spp.) communities, with sedge
(Carex and Kobresia spp.) meadows in wetter sites (D.
Bedunah, University of Montana, unpublished data).
North- and northeast-facing slopes were similar but tended
to be wetter and contained more extensive areas of sedge
meadows. Except for occasional, small sedge meadows at
wet sites, there was little vegetation above 4,900 m, and we
did not see argali at elevations .5,000 m.

Weather data for the study area are scarce, but the Library
of Congress Country Studies reported that the Wakhan
Corridor received ,100 mm of rainfall annually and
classified the area as arid to semi-arid. This assessment
accords with our experiences in the field: rain was rare, brief,
and light, from spring through fall, and snows were light,
rarely accumulating to depths .15 cm in the winter.

The Big Pamir was used extensively by Wakhi livestock
herders, primarily from late spring to early fall, although one
herding operation remained year-round, high in the Aba
Khan valley near the northeast boundary of the study area.
Domestic sheep and goats were typically grazed in the mid-
to lower elevations (3,000 m to 4,000 m), whereas cattle and
yaks were grazed near the heads of the valleys, often to
elevations in excess of 4,850 m.

METHODS

Field Procedures
We conducted field work (visual counts and fecal sampling)
in 4 discrete sessions: 24 June–13 July 2007 (summer 2007),
13 November–8 December 2007 (autumn 2007), 15
January–8 February 2008 (winter 2008), and 17 May–10
July 2008 (summer 2008). Long intervals between sampling
sessions were not ideal for CMR estimation; however, they
were necessitated by logistical constraints and other research

Harris et al. N Argali Abundance in the Afghan Pamir 669



objectives. We conducted field work on foot, horseback, or
yak-back. Because argali moved frequently through difficult
terrain and our own movements were circumscribed by the
valley systems separated by steep ridges, we made no attempt
to impose a standardized geographic sampling regime.
Instead, we attempted to survey for argali within all major
drainages of the Big Pamir during each field session,
generally walking to high vantage points to search for
animals during the early morning and late afternoon.

We adopted a noninvasive approach to CMR estimation,
using DNA extracted from fecal samples as individual
marks. We collected 3 fecal pellets from each pellet group
when we encountered fecal pellets we were reasonably
certain were freshly deposited by argali. We only collected
pellets adjacent to each other within the group, reducing to
inconsequential the probability of a sample containing DNA
from .1 argali. We avoided collecting from pellet groups
scattered over more than approximately 0.1 m

2
or that

appeared to have been deposited while the animal was
moving. Low quality samples with malformed or broken
pellets were not extracted. We did not sample fecal pellets

that appeared to have been produced by lambs of the year;
thus, our abundance estimates from DNA-based CMR
modeling apply to animals

L

1 year of age. We took Global
Positioning System (GPS) locations for each sample, unless
samples were within approximately 3 m of an existing GPS
fix, in which case we recorded the same location, and noted
the date, time, and name of the collector. We stored fecal
pellets in sterile 30-cm centrifuge tubes with securely fitting
screw-tops to which sporks (plastic, ovoid-shaped protuber-
ances with fork-like tines attached to the inside of the cap)
were attached, allowing individual handling of each sample
without risk of contamination (Evergreen Scientific, Los
Angeles, CA). We added approximately 4 parts 95% ethanol
for each part fecal material.

We counted observed argali and attempted to assess
whether animals were unique from others previously
observed during that session. When possible, we classified
animals as adult females (age

L

2 yr), lambs (either sex),
yearlings (either sex), and adult males (age

L

2 yr). We
made these counts while collecting fecal samples as well as
while making observations related to assessing argali

Figure 1. Afghanistan, showing the location of the Wakhan District in northeast part of the country, separating Tajikistan (to the north) from Pakistan (to
the south). Inset: Wakhan District, showing relative locations of the Big Pamir and Little Pamir ranges and Waghjir Valley. Arrow identifies the Big Pamir
study area, where we studied argali, summer 2007 to summer 2008, roughly corresponding with the boundary (light outline) of the former Pamir-i-Buzurg
Wildlife Reserve.

670 The Journal of Wildlife Management N 74(4)



productivity, habitat use, and responses to disturbance.
Because we had no way to quantify the uncertainty
surrounding duplicate counts arising from movements and
the formation of new and dissolution of existing groups, we
developed upper and lower bounds to the number of argali
we observed in each session. For the upper bound, we
deleted only observations of animals that we observed
temporally and geographically so closely to other observa-
tions of the same sex and age class that we felt certain they
represented identical individuals; we counted all other
observations. For the lower bound, we deleted observations
that could plausibly have resulted from movements of
previously recorded animals. Neither index accounted for
animals we did not visually observe during the session.

Genotyping
Genetic work was conducted in 2 laboratories. We
conducted initial work at Centro de Testagem Molecular
(CTM/CIBIO), Portugal, where we extracted DNA from
fecal samples using the DNeasyTM Blood Kit (Qiagen,
Valencia, CA) modified to include an initial wash of one
fecal pellet for 15 minutes in 350 mL of lysis buffer (0.1 M
Tris-HCl, 0.1 M EDTA, 0.01 M NaCl, 1% N-lauroyl
sarcosine, pH 7.5). We used approximately 200 mL of the
lysis buffer directly in the extraction protocol as if the
sample were blood (as in Maudet et al. 2004).

At CTM, we co-amplified 8 microsatellite DNA markers
in 3 multiplex polymerase chain reaction (PCR) amplifica-
tions (MP1: MAF33, ADC; MP2: MAF36, FCB266, and
FCB304; MP3: GLYCAM, KRT2, and LIF) using the
QIA multiplex mix (Qiagen). We performed amplifications
on a MJR DYAD PTC220 DNA Thermal Cycler following
QIA mix protocol for 40 cycles at 3 annealing temperatures
(54u C, 62u C, and 57u C for MP1, MP2, and MP3,
respectively). Total reaction volume was 10 mL, including
5 mL of the Qiagen PCR Master Mix, 1 mL of primer mix,
and 2 mL of DNA. We accomplished reactions with 3
primers for each locus, following the M13-tailed primer
method (Oetting et al. 1995). Fluorescently labeled DNA
fragments were visualized on an ABI3130xl DNA analyzer
(Applied Biosystems, Foster City, CA) and chromatograms
were analyzed using GeneMapper 4.0 software (Applied
Biosystems) by 2 independent experts at scoring micro-
satellite profiles. We initially screened all samples twice with
MP1 to quantify the quality of nuclear DNA. We selected
samples that showed reliable amplifications and matching
genotypes to continue the genotyping process. We inde-
pendently regenotyped selected samples 3 to 4 times.

We conducted the remaining lab work and genetic analysis
at the University of Montana Conservation Genetics
Laboratory, Missoula, Montana, USA. We repeated 3
markers initially run in Portugal as a data quality check,
and we screened 7 additional microsatellite loci, plus the
amelogenin sex identification locus (Pidancier et al. 2006).
We optimized a multiplex and one PCR and performed 10-
mL reactions on MJR PTC200 thermocyclers using touch-
down profiles. Each reaction contained 2.5 mL of template
DNA, 4.5 mL of QIA multiplex mix (Qiagen), and either

1 mL of 103 primer mix, or 1 mL of 2 pM forward and
reverse primers. We used 2 touch-down profiles with 35
cycles, one with an initial annealing temperature of 63u C
stepping down to 58u C and another starting at 58u C and
stepping down to 53u C. We visualized fluorescently labeled
DNA fragments on an ABI 3130xl automated capillary
sequencer (Applied Biosystems) in the Murdock DNA
Sequencing Facility at the University of Montana. We
determined allele sizes using the ABI GS600LIZ ladder
(Applied Biosystems). Chromatograms were viewed and
analyzed using GeneMapper software v3.7 (Applied Bio-
systems) by 2 independent researchers.

We determined sex by PCR amplification of the
amelogenin gene as in Pidancier et al. (2006). We obtained
2 PCR products (approx. 315 base pairs [bp] and approx.
359 bp) for males but only the longer product for females.
We based consensus genotypes on multiple sample runs and
the following rules: 1) for a sample to be heterozygous at a
locus we had to observe both alleles twice and 2) for a
sample to be homozygous, we had to observe the allele 3
times. We randomly chose 10% of samples for re-extraction
and repeat genotyping to monitor for errors; we detected no
genotype differences or errors. Due to the large size of
fragments at the amelogenin locus, we determined con-
sensus genotypes as above with the following changes: we
provisionally accepted heterozygotes (M) if we observed a
male band only once; we classified homozygotes where we
observed only the female band ,3 times (e.g., of 3
independent PCRs) as of unknown sex, and we accepted
as males genotypes where we observed only the male band

L

3 times.
We ran principal correspondent analysis (PCoA) and

multilocus genotype matching in GENALEX (Peakall and
Smouse 2006) to identify outliers due to potential tube
mishandling (in the lab or field), genotyping errors, or non-
argali samples and to identify recaptures. We computed
amplification success rate, false allele rate, and allelic drop-
out rate as in Luikart et al. (2008). We found loci
contributing significantly more unique individuals than
expected with DROPOUT (McKelvey and Schwartz
2005). We estimated expected heterozygosity, tested for
gametic (linkage) disequilibrium, and assessed departures
from Hardy–Weinberg proportions (using exact tests and a
Markov chain) as implemented in GENEPOP 3.4 (Ray-
mond and Rousset 1995). We computed the probability that
randomly drawn, unrelated individuals would have identical
genotypes (probability of identity [PID]) with Api-Calc
(Ayres and Overall 2004).

Estimation of Abundance
We used closed capture modeling in Program MARK to
estimate population size, selecting among plausible models
of capture variation. Except for one reference model, we
assumed throughout that recapture probability would not
differ from probability of initial capture (i.e., c 5 p) because
we could envision no situation in which our sampling
activity would affect probability of subsequent capture (i.e.,
we included no models that allowed for a behavioral

Harris et al. N Argali Abundance in the Afghan Pamir 671



response to initial capture). We examined models in which
capture probability varied by sex and by session. We also
examined a reduced model in which we classified sessions
into low and high capture effort (sessions 1 and 3 were low
effort and sessions 2 and 4 were high effort according to
total no. of samples collected in each session, see below).

Because we occasionally obtained recaptures within our
defined capture sessions, we also estimated abundance in a
session-free context using Program CAPWIRE (Miller et
al. 2005). Program CAPWIRE uses a continuous sampling
approach and thus potentially can make use of recaptures
occurring within a given capture session.

To examine the assumptions of geographic and demo-
graphic population closure (which, under our sampling
design, were confounded and could not be distinguished),
we examined the fit of a series of Pradel models (Pradel
1996) to the same data. In addition to unconstrained Pradel
models (for open populations), we systematically con-
strained each model by fixing the apparent survival term
(Q) to one (thus closing the population to subtractions) and
the recruitment term ( f ) to zero (thus closing the
population to additions). Our interest was not in either of
these demographic terms (or in the derived parameter, the
finite per capita rate of increase, l), but rather in the relative
fit of models with differing closure assumptions to our data.
We used both Akaike’s Information Criterion corrected for
small sample size (AICc) and likelihood ratio tests to
examine the weight of evidence for closure (Cooch and
White 2006).

RESULTS

Visual counts of argali varied, depending on the amount of
time we spent in the field, habitats used by the animals,
weather conditions, and our level of uncertainty in judging
observations to be duplicates of animals previously tallied.
Our counts of yearlings were not always reliable, depending
on the distance between the animals and observer. Counts of
adult females (Table 1) varied from as few as 40 in summer
2007 (with all possible duplicates removed) to as many as
179 in summer 2008 (allowing for some possibility of
duplication). Counts of adult males varied from as few as 21
(in autumn 2007) to as many as 82 (in summer 2008). We
typically were unable to classify 15–20% of animals we

observed (Table 1). We knew of no way to objectively
extract one best estimate from these data. However,
assuming rough closure among females, taking the lower
bounds of the autumn 2007 and summer 2008 counts as
only slightly conservative, and adding half the number of
yearlings (assuming a 50:50 sex ratio) from these time
periods, our direct observations suggested about 140 females
aged

L

1 year present in the Big Pamir during 2007–2008.

Fecal Sampling, DNA Extraction, and Genotyping
We collected 392 fecal samples (61 in summer 2007, 134 in
autumn 2007, 62 in winter 2008, and 135 in summer 2008)
we judged to be freshly deposited by argali, of which 249
samples yielded consensus genotypes at

L

6 of 12 loci.
Amplification success was relatively low (85%), but false
allele (1%) and allelic drop-out (5%) rates were also
relatively low (Luikart et al. 2008).

To improve precision, we dropped 2 problematic loci
(ADC, LIF) and required remaining samples to share

L

6 of
10 loci in common. We dropped LIF because .20% of
consensus genotypes were missing data at this locus. We
removed ADC from the analysis because DROPOUT
found it identified significantly (P , 0.05) more unique
individuals than expected. These measures to improve
precision reduced our genetic sample size to 232 consensus
genotypes.

Genotyping success was higher in winter than other
seasons (although handling and storage differences may also
have influenced genotyping success). We identified 32
multilocus genotypes from 61 samples analyzed (52%), 53
genotypes from 134 samples (43%), 45 genotypes from 62
samples (73%), and 33 genotypes from 68 samples (49%) in
summer 2007, autumn 2007, winter 2008, and summer
2008, respectively.

We identified 147 unique genotypes from these 232
consensus genotypes. We removed 4 outlier individuals
identified by PCoA because they were closely related to
samples collected in Tajikistan from urials (Ovis orientalis).
Of the remaining 143 genotypes, we could not reliably
classify 12 to sex based on the amelogenin sex identification
locus. For 7 of these 12, however, our field knowledge of the
fecal pellets in question was sufficient to assign a sex (i.e.,
only one sex known to be in the vicinity of the specific area
within a few days of the collection), leaving 5 for which we

Table 1. Number of argali, classified by sex and age, that we visually observed during each session, Big Pamir Mountain Range, Afghanistan, summer 2007
through summer 2008. Upper row in each case represents a conservative estimate of the number of unique animals seen, removing groups from the cumulative
total when duplication was a possibility, and may be underestimates; lower row in each case represents the maximum number of unique animals we observed,
removing only certain duplicate observations, and thus are most likely overestimates. Counts during winter 2008 were conducted by a field assistant who did
not provide sufficient details to derive lower and upper bounds and thus are not reported here.

Sampling period Type of estimate

Argali classification

Ad F Yearlings Lambs Ad M Unclassified Total

Summer 2007 Lower bound 40 31 4 34 9 118
Upper bound 69 54 8 67 12 210

Autumn 2007 Lower bound 110 28 5 21 15 179
Upper bound 113 31 5 21 15 185

Summer 2008 Lower bound 106 45 11 73 20 255
Upper bound 179 77 24 82 27 389

672 The Journal of Wildlife Management N 74(4)



could not determine sex. Our raw data for all subsequent
CMR analysis (and min. population sizes) thus consisted of
138 uniquely identified argali (82 F, 56 M). Of the 138 sex-
identified consensus genotypes, none were mismatched by
one allele, one pair was mismatched by one allele at 2 loci,
and one pair was mismatched by one allele at 4 loci.

Heterozygosity was high (mean expected heterozygosity 5
0.74; range 5 0.52–0.85 among loci), as was the number of
alleles per locus (x̄ 5 9.5, range 5 6–16). Probability that 2
randomly selected, unrelated individuals would have iden-
tical genotypes (PID) was low (4.5 3 10211 for 10 loci, 1.1
3 1026 for 6 loci), as was probability of randomly selected
siblings having identical genotypes (1.1 3 1024, 4.4 3 1023

for 10 and 6 loci, respectively). Marker FCB226 deviated
from Hardy–Weinberg proportions (FIS 5 0.388, P ,
0.001), probably due to a null allele. One pair of loci
revealed gametic disequilibrium (MAF48 and FCB304, P ,
0.01).

Abundance Estimation
We tallied 163 captures (including recaptures but excluding
intrasession recaptures, see below) of the 138 individually
identified argali, which formed the basis of closed-capture
estimation. Recaptures among the 4 sessions were uncom-
mon. We captured 63 of the 82 identified females only once,
we recaptured 18 once (i.e., captured them twice), and
recaptured one twice. Of the 56 identified males, we
recaptured only 3 animals after initial identification (we

recaptured 1 animal once and 2 animals twice, Table 2). In
addition, we tallied 30 intrasession recaptures that did not
contribute to estimates using Program MARK: 1 in summer
2007, 10 in autumn 2007, 8 in winter 2008, and 11 in
summer 2008.

The top ranked closed-capture model of population
abundance allowed capture probability to vary by both sex
and session (Table 3). The second ranked model was
similar, with session categorized by effort. Other models
enjoyed less support. Model averaging produced a point
estimate of 172.4 female argali, with a 95% confidence
interval of 112.81–231.95 (CI coverage/estimate 5 0.69).
Approximately 8% of total variance was attributed to model
uncertainty. The same procedures produced a model-
averaged point estimate of 248.6 male argali (95% CI 5
49.5–447.8). We suspected this estimate for males was
positively biased because 1) we never visually observed more
males than females in the Big Pamir study area during any
sampling session, 2) our molecular analyses identified more
females than males, and 3) although as yet unstudied for
argali, males of similar species generally have lower survival
than females, suggesting that females should outnumber
males, at least among adults (Toı̈go and Gaillard 2003).
Thus, we conducted post hoc Pradel modeling to investigate
closure.

Pradel modeling with various assumptions regarding
closure did not result in an unambiguous signal but generally
supported the hypothesis that females could be acceptably
modeled in a closed-population framework but males could
not. In most comparisons, models with Q, f, or both
constrained for males (i.e., no additions or subtractions for
M) ranked lower than corresponding unconstrained models
(Table 4; with models ranked by deviance rather than AICc,
because constrained models will tend to have lower AICc
due to having fewer parameters, independent of model fit).
Likelihood ratio tests comparing constrained with uncon-
strained models for males were not significant, but these
tests are known to lack power, particularly when the number
of sessions is low (e.g., only 4 in our case) and capture
heterogeneity is present. Model comparisons provided little
evidence to reject the hypothesis that females were a closed
population during the time period, although the weakness of
our tests must be kept in mind.

Table 2. Captures and recaptures of identified argali during the 4 sessions,
Big Pamir, Afghanistan, 2007–2008 (excluding recaptures within sessions).
Sessions are 1: Summer 2007, 2: Autumn 2007, 3: Winter 2008, 4: Summer
2008. The last column also includes one male captured in sessions 1, 2, and
3; and one male and one female captured during sessions 1, 2, and 4.

Session Sex
Initial

capture

Recaptured
in session

Total captured
during session2 3 4

1 F 9 5 5 1 20
M 9 2 1 0 12

2 F 22 0 3 3 29
M 22 0 0 0 24

3 F 26 0 0 2 36
M 7 0 0 0 9

4 F 10 0 0 0 17
M 15 0 0 0 16

Table 3. Top ranked candidate models of probability of capture (p) of female argali in the Big Pamir Mountains, Afghanistan, 2007–2008, based on
Akaike’s Information Criterion (AICc), model weights (wi), number of parameters (K ), deviance, and their point estimates of the number of females and
standard errors. Estimates of male abundance are not included in this table.

Modela DAICc wi K Deviance Point estimate F SE

p 5 c + (g + s) 0.00 0.776 10 37.85 170.17 28.87
p 5 c + (g + effort) 3.12 0.163 6 49.22 174.24 29.92
p 5 c + (g) 6.58 0.029 4 56.76 174.74 30.05
p 5 c + (s) 7.17 0.022 6 53.28 213.35 36.79
p 5 c + (.) 9.41 0.007 3 61.62 216.46 37.51
p 5 c (effort) 10.88 0.003 4 61.06 216.25 37.46
Weighted average 172.38
95% CI 112.81–231.95

a
Abbreviations: c 5 probability of recapture, g 5 sex, s 5 session (capture occasion), effort 5 occasions grouped by similar numbers of fecal pellets

collected, (.) 5 constant.

Harris et al. N Argali Abundance in the Afghan Pamir 673



In CAPWIRE, equal catchability models were always
more preferred (using likelihood ratio tests) over the 2
innate rates (i.e., mixture) models. Including all recaptures
(regardless of session or distance from other captures, n 5
117 captures) resulted in an estimate of 153 female argali
(95% CI 5 120–202; CI coverage/estimate 5 0.54), lower
than the model averaged estimate using closed captures in
Program MARK, but with a narrower confidence interval.
However, we noted that the spatial distribution of within-
session recaptures differed significantly (Kolmogorov–Smir-
nov 2-sample test of equality, P , 0.001) from that of
between-session recaptures (Fig. 2). In fact, 19 of 30
within-session recaptures occurred when sampling from
groups of argali that had been bedded at one site (and thus
assigned an identical location). We considered that
recaptures obtained under such circumstances violated the
independence assumption underlying all continuous time
models that we know of, and we thus rejected this estimate
as having false precision. Removing within-session recap-
tures that had identical locations (i.e., were not indepen-
dent) and rerunning Program CAPWIRE resulted in an
estimate of 208 female argali (95% CI 5 148–299; CI
coverage/estimate 5 0.73), higher than suggested by our
closed-capture models. Because our exploration of closure
using Pradel models suggested that males were not a closed
population, we made no attempt to estimate male
abundance using CAPWIRE.

DISCUSSION

Although argali are not difficult to count through direct
ground-based observations, interpreting surveys based on
count tallies is fraught with difficulty. Beyond the well-
explored issue that not all animals will be detected, argali
move more quickly than ground-based observers, who must
therefore deal with the possibility of tallying the same
individuals multiple times. Capture–mark–recapture estima-
tion, using DNA obtained from fecal samples, offers a
potential solution to these problems. Our closed-capture
abundance point estimate for females (N̂ 5 172, 95% CI 5

113–232) was about 23% higher than our best interpolation
of our series of visual counts and considerably higher than
most counts obtained during individual field sessions. More
importantly, our estimate was based on models we under-
stood and assumptions we could articulate.

Equally importantly, our exploration of closure assump-
tions provided evidence that male abundance varied among
time periods. Our closed capture models thus overestimated
the number of males within the sample area, but we lacked
information with which to identify a superpopulation to
which the estimate might apply (Kendall 1999). We
hypothesize that most variation in male abundance was
due to movements in and out of the Big Pamir area rather
than mortality. Undetected poaching and natural mortality
no doubt occurred during the study, and one would expect
mortality rates of males to exceed those of females. Our
data, however, gave us no indication that mortality was
substantially greater among males than females. We
observed directly (Table 1) and captured (Table 4) more
males in summer 2008 than in summer 2007, contrary to
what we would expect if males were declining throughout
the time period. Shortly after our summer 2007 session (and
prior to our autumn 2007 session in which we tallied few
M), B. Habib and Z. Moheb of our survey team observed
106–191 male argali (depending on assumptions about
duplicate counts) but no females, during a week in the
Waghjir Valley, about 130 km southeast of the Big Pamir.
We can reasonably assume that argali males, unencumbered

Table 4. Support for Pradel open population models of argali population
abundance in the Big Pamir Mountains, Afghanistan, 2007–2008, with
selected parameters constrained to produce closed or partially closed
models. Fully closed models have apparent survival (Q) constrained to 1.0
(no subtractions) and recruitment ( f ) constrained to 0.0 (no additions).
Models are shown in pairs, in ascending order of deviance. Shown are
Akaike’s Information Criterion (AICc), number of parameters (K ),
deviance, and results of likelihood ratio tests, x

2
, df (difference between

the pair of models), and the probability of obtaining a x
2

value this large or
larger if there is no difference in model fit.

Model AICc K Deviance x
2 df P

M Q, open, rest
closed 546.56 5 38.334 1.835 1 0.1756

Both sexes closed 546.27 4 40.169
M open, F closed 548.26 6 37.871 2.298 2 0.3170
Both sexes closed 546.27 4 40.169
Both sexes open 552.56 8 37.790 2.379 4 0.6663
Both sexes closed 546.27 4 40.169
F open, M closed 550.47 6 40.087 0.082 2 0.9599
Both sexes closed 546.27 4 40.169

Figure 2. Distances (binned to nearest kilometer) between remote
recaptures of the same individual argali, Big Pamir study area, Wakhan
District, Afghanistan, summer 2007 through summer 2008 (F: shaded bars;
M: open bars). (A) Recaptures within a given sampling session. (B)
Recaptures among sampling sessions.

674 The Journal of Wildlife Management N 74(4)



by lambs and interested in maximizing mating opportunities
during rut, travel greater distances than do females, as males
do in most polygynous artiodactyls. We cannot identify
sources or destinations of itinerant males (nor exact time
periods of movement) but find the overall data persuasive
that males moved in and out of the Big Pamir freely. Thus,
even had we a field method that overcame the issues of
imperfect detectability and possible duplicate counting, an
estimate of male abundance based solely on direct observa-
tions could easily have misled us into interpreting it as
meaningful for the Big Pamir range in isolation.

Our study design was compromised by logistical con-
straints (it being expensive and time-consuming to mount
expeditions to the area), along with competing research
objectives (e.g., argali habitat use during specific seasons).
Ideally in closed-capture CMR estimation, sampling
sessions would be shorter, separated by briefer intervening
time intervals, and more numerous than we report here. Had
we been able to conduct multiple capture sessions within a
shorter time span during which males remained on the study
area, we might have avoided the problem of population
closure and thus succeeded in generating population
estimates for both sexes. For males, such an estimate would
have been meaningful only for that limited time duration,
however. Because we lacked information on when males
moved into and out of the area, we had no way to optimize
the timing of our CMR work, even had logistics not been
limiting. The robust design (Pollock 1982, Kendall and
Pollock 1992), employing secondary sampling sessions
nested within primary sessions, would likely have solved
these problems and yielded better information, but we
lacked the resources to intensify our sampling sufficiently to
implement it.

Remotely obtained genetic data do not necessarily con-
form to traditional designs of mark–recapture studies, in
which a capture may occur, at most, once during each
session. Continuous-session models that make use of within
session recaptures such as CAPWIRE offer the potential for
increased precision (Miller et al. 2005, Lukacs et al. 2007,
Puechmaille and Petit 2007, Robinson et al. 2009). In our
case, however, we found that most (63%) within-session
recaptures actually contained no new information. Because
they originated from fecal samples deposited in the same
location (and at a similar time) as other samples from that
individual, they reflected our inability to distinguish .1
defecation of an individual from those of multiple
individuals, rather than anything about the abundance of
animals. Had we ignored this spatio-temporal independence
problem, our results would have contained falsely high
precision. After removing within-session recaptures that
were not independent, Program CAPWIRE produced an
estimate with slightly less precision than model averaged
closed captures in MARK. We urge investigators to
consider whether samples collected closely in time and (or)
space are truly biologically independent. If not, the ability of
traditional closed-capture models (e.g., in MARK) to
consider heterogeneity in capture probability explicitly
remains an advantage.

Although we were disappointed that we did not develop
an abundance estimate for both sexes in the Big Pamir
range, we were heartened to find evidence that our initial
assumption of population isolation was evidently wrong.
Further genetic information from this population (G.
Luikart, University of Montana, unpublished data) suggests
considerable gene flow with argali in other areas, notwith-
standing that these are apparently separated from the Big
Pamir by argali-free regions.

Most analyses of mark–recapture data consider males and
females separately because they are likely to differ in capture
probability. However, genetic data, particularly from low-
quality samples such as feces, sometimes allow discrimination
among more individuals (based on microsatellites) than it can
reliably identify to sex. There may thus be a temptation to
increase overall sample size by ignoring sex, considering males
and females together. Our study reminds investigators that in
addition to differing in capture probability (and thus adding
undesirable variation to N̂), sexes may differ in their
conformance with closure assumptions. Abundance assess-
ments of species that are sexually dimorphic behaviorally
should consider them separately.

These insights from molecular data came at some cost.
Because other aspects of our argali study (e.g., G. Luikart,
unpublished data; J. Winnie, Jr., University of Montana,
unpublished data) required considerable field work and
employed many of the same personnel, estimating the added
cost of our DNA-based CMR estimate over unadjusted
visual counts is not straightforward. Including the costs of
supplies, shipping, and personnel, a reasonable estimate is
approximately US$150/sample. A stand-alone survey de-
signed to obtain similar data would have to budget for field
work and personnel over and above this. Obtaining better
precision than we did would require larger samples, more
sampling sessions, or both. However, given the remote nature
of the area and the added expenses associated with handling
animals, the use of traditional mark–recapture or telemetry
studies to answer the questions of population size or sex-
biased movement patterns would be even more expensive.

Much of the laboratory costs involved development and
optimization of genotyping assays (in each of 2 independent
laboratories). Future analyses would probably cost less (e.g.,
US$100/sample). Costs will likely continue to decline with
technological advancements (Beja-Pereira et al. 2009).
Further, future argali analyses might be conducted with
less replication if higher quality samples could be collected,
for example by sampling only in the winter or when fecal
pellets are better formed and contain better-quality DNA
(Maudet et al. 2004, Luikart et al. 2008; for review see Beja-
Pereira et al. 2009). Although collecting high quality fecal
samples for genetic analysis in remote areas can be a
challenge, it is also easy to integrate into field surveys
without many of the additional costs associated with
traditional mark–recapture or telemetry studies.

MANAGEMENT IMPLICATIONS

Both CMR estimates and our visual counts confirmed that
the number of argali occupying the Big Pamir range is small

Harris et al. N Argali Abundance in the Afghan Pamir 675



and that this portion of argali range is appropriately viewed
as a serious conservation concern. Further monitoring and
careful management is needed. Although Petocz (1973)
considered the Big Pamir to have contained .172 adult
females in the early 1970s, differences in estimation
methods lead us to caution against interpreting our
estimate as necessarily indicating a dramatic decline since
then. Capture–mark–recapture modeling, as well as asso-
ciated genetic information (G. Luikart, unpublished data)
suggested that argali inhabiting the Big Pamir are not
highly inbred or genetically isolated. With such low
numbers, genetic (and possibly occasional demographic)
linkage with other populations, including those in Tajiki-
stan or China, has probably functioned to help maintain
this population.

Remote (noninvasive) estimation of abundance using CMR
models and genetic data from fecal samples is increasingly
feasible and is often the only reliable option for wide-ranging
ungulates living in difficult habitats such as argali. Such
estimates overcome many of the problems inherent in
uncorrected visual counts and in traditional CMR ap-
proaches. However, although fecal pellets are easy to collect,
rigorous laboratory analyses are costly (especially when initial
optimization is required), and statistical analyses must be
considered carefully. We suspect our approach will be useful
primarily for populations of particular concern, which
typically will be small and well defined.

ACKNOWLEDGMENTS

Our study was part of the Afghanistan Biodiversity
Conservation Program of the Wildlife Conservation Society
(WCS), supported by the United States Agency for
International Development (USAID). G. Luikart was
partially supported by the Portuguese American Foundation
for Development and Centro de Investigação em Biodiver-
sidade e Recursos Genéticos-Universidade do Porto. R.
Godinho and A. Beja-Pereira were supported by post-
doctoral grants (SFRH/BPD/36021/2007 and SFRH/
BPD/38096/2007, respectively) from the Portuguese
Science Foundation. For field assistance we thank B. Habib,
Z. Moheb, Sabir, and A. Khairzad. We also thank S.
Ostrowski, D. Bedunah, S. Nikzad, Z. Ejlasi, I. Farahmand,
Q. Sahar, K. Sediqi, K. Sidiqi, G. Sediq, A. Ahamad, R.
King, and L. Yook for their able assistance. We thank A.
Dehgan, P. Zahler, P. Smallwood, P. Bowles, and A. Simms
for advice and administrative assistance, and G. White and
P. Lukacs for analytical help.

LITERATURE CITED

Anderson, D. R. 2001. The need to get the basics right in wildlife field
studies. Wildlife Society Bulletin 29:1294–1297.

Ayres, K. L, and A. D. J. Overall. 2004. API-CALC 1.0: a computer
program for calculating the average probability of identity allowing for
substructure, inbreeding and the presence of close relatives. Molecular
Ecology Notes 4:315–318.

Beja-Pereira, A., R. Oliveira, P. C. Alves, M. K. Schwartz, and G. Luikart.
2009. Advancing ecological understandings through technological
transformations in noninvasive genetics. Invited review. Molecular
Ecology Resources 9:1279–1301.

Bellemain, E., J. E. Swenson, D. Tallmon, S. Brunberg, and P. Taberlet.
2005. Estimating population size of elusive animals with DNA from
hunter-collected feces: four methods for brown bears. Conservation
Biology 19:150–161.

Boulanger, J. B., B. N. McLellan, J. G. Woods, M. F. Proctor, and C.
Strobeck. 2004. Sampling design and bias in DNA-based capture–mark–
recapture population and density estimates of grizzly bears. Journal of
Wildlife Management 68:457–469.

Boulanger, J. B., G. C. White, M. Proctor, G. Stenhouse, G. MacHutchon,
and S. Himmer. 2008. Use of occupancy models to estimate the influence
of previous live captures on DNA-based detection probabilities of grizzly
bears. Journal of Wildlife Management 72:589–595.

Caughley, G. 1977. Analysis of Vertebrate Populations. Wiley and Sons,
Chichester, United Kingdom.

Cooch, E., and G. White. 2006. Program MARK: a gentle introduction.
Fifth edition. Colorado State University, Fort Collins, USA. ,http://
www.phidot.org/software/mark/docs/book/.. Accessed 1 Mar 2009.

Creel, S., G. Spong, J. L. Sands, J. Rotella, J. Zeigle, L. Joe, K. M. Murphy,
and D. Smith. 2003. Population size estimation in Yellowstone wolves
with error-prone noninvasive microsatellite genotypes. Molecular Ecol-
ogy 12:2003–2009.

Fedosenko, A. K., and D. A. Blank. 2005. Ovis ammon. Mammalian
Species 773:1–15.

Fickel, J., and U. Hohmann. 2006. A methodological approach for non-
invasive sampling for population size estimates in wild boars (Sus scrofa).
European Journal of Wildlife Research 52:28–33.

Fitzherbert, A., and C. Mishra. 2003. Afghanistan Wakhan Mission
technical report. United Nations Environment Program, Nairobi, Kenya.

Garshelis, D. 2006. On the allure of noninvasive genetic sampling: putting
a face to the name. Ursus 17:109–123.

Guschanski, K., L. Vigilant, A. McNeilage, M. Gray, E. Kagoda, and M.
Robbins. 2009. Counting elusive animals: comparing field and genetic
census of the entire mountain gorilla population of Bwindi Impenetrable
National Park, Uganda. Biological Conservation 142:290–300.

Habibi, K. 1997. Afghanistan. Pages 204–211 in D. M. Shackleton and the
IUCN/SSC Caprinae Specialist Group, editors. Wild sheep and goats
and their relatives. Status survey and conservation action plan for
Caprinae. International Union for Conservation of Nature, Gland,
Switzerland.

Harris, R. B. 2007. Wildlife conservation in China: preserving the habitat
of China’s Wild West. M. E. Sharpe, Armonk, New York, USA.

Harris, R. B., and C. O. Loggers. 2004. Status of Tibetan plateau mammals
in Yeniugou, China. Wildlife Biology 10:121–129.

Kendall, K. C., J. B. Stetz, D. A. Roon, L. P. Waits, J. B. Boulanger, and
D. Paetkau. 2008. Grizzly bear density in Glacier National Park,
Montana. Journal of Wildlife Management 72:1693–1705.

Kendall, W. L. 1999. Robustness of closed capture–recapture methods to
violations of the closure assumption. Ecology 80:2517–2525.

Kendall, W. L., and K. H. Pollock. 1992. The Robust Design in capture-
recapture studies: a review and evaluation by Monte Carlo simulation.
Pages 31–43 in D. R. McCullough and R. H. Barrett, editors. Wildlife
2001: populations. Elsevier, London, United Kingdom.

Knapp, S. M., B. A. Craig, and L. P. Waits. 2009. Incorporating
genotyping error into non-invasive DNA based mark–recapture popula-
tion estimates. Journal of Wildlife Management 73:598–604.

Luikart, G., S. Zundel, D. Rioux, C. Miquel, K. A. Keating, J. T. Hogg, B.
Steele, K. Foresman, and P. Taberlet. 2008. Low genotyping error rates
for microsatellite multiplexes and noninvasive fecal DNA samples from
bighorn sheep. Journal of Wildlife Management 72:299–304.

Lukacs, P. M., and K. P. Burnham. 2005a. Review of capture recapture
methods applicable to noninvasive sampling. Molecular Ecology
14:3909–3919.

Lukacs, P. M., and K. P. Burnham. 2005b. Estimating population size from
DNA-based closed capture–recapture data incorporating genotyping
error. Journal of Wildlife Management 69:396–403.

Lukacs, P. M., L. S. Eggert, and K. P. Burnham. 2007. Estimating
population size from multiple detections with noninvasive genetic data.
Wildlife Biology in Practice 3:83–92.

Magomedov, M.-R., E. G. Akhmedov, W. A. Wall, and A. E. Subbotin.
2003. Current status and population structure of argalis (Ovis ammon
L., 1758) in Central Asia. Beitrage zur Jagd und Wildforschung 28:
151–163.

676 The Journal of Wildlife Management N 74(4)



Maudet, C., G. Luikart, D. Dubray, A. von Hardenberg, and P. Taberlet.
2004. Low genotyping error rates in wild ungulate faeces sampled in
winter. Molecular Ecology Notes 4:772–775.

McKelvey, K. S., and M. K. Schwartz. 2005. DROPOUT: a program to
identify problem loci and samples for noninvasive genetic samples in a
capture–mark–recapture framework. Molecular Ecology Notes 5:716–718.

Miller, C. R., P. Joyce, and L. Waits. 2005. A new method for estimating
the size of small populations from genetic mark–recapture data.
Molecular Ecology 14:1991–2005.

Mulders, R., J. Boulanger, and D. Paetkau. 2007. Estimation of population
size for wolverines Gulo gulo at Daring Lake, Northwest Territories, using
DNA based mark–recapture methods. Wildlife Biology 13(Suppl. 2):
38–51.

Oetting, W. S., H. K. Lee, D. J. Flanders, G. L. Wiesner, T. A. Sellers, and
R. A. King. 1995. Linkage analysis with multiplexed short tandem repeat
polymorphisms using infrared fluorescence and M13 tailed primers.
Genomics 30:450–458.

Peakall, R., and P. E. Smouse. 2006. GENALEX 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.

Petit, E., and N. Valiere. 2006. Estimating population size with
noninvasive capture–mark–recapture data. Conservation Biology
20:1062–1073.

Petocz, R. G. 1973. Marco Polo sheep (Ovis ammon poli) of the Afghan
Pamir: a report of biological investigations in 1972–1973. Food and
Agriculture Organization, Rome, Italy.

Petocz, R. G. 1978. Report on the Afghan Pamir part 3: management plan
for the Big Pamir Wildlife Reserve. United Nations Development
Program/Food and Agriculture Organization/Government of Afghani-
stan, Kabul, Afghanistan.

Petocz, R. G., K. Habibi, A. Jamil, and A. Wassey. 1978. Report on the
Afghan Pamir Part 2: biology of Marco Polo Sheep (Ovis ammon poli).
United Nations Development Program/Food and Agriculture Organiza-
tion/Department of Forests and Range, Ministry of Agriculture, Kabul,
Afghanistan.

Pidancier, N., S. Jordan, G. Luikart, and P. Taberlet. 2006. Evolutionary
history of the genus Capra (Mammalia, Artiodactyla): discordance
between mitochondrial DNA and Y-chromosome phylogenies. Molec-
ular Phylogenetics and Evolution 40:739–749.

Pollock, K. H. 1982. A capture–recapture design robust to unequal
probability of capture. Journal of Wildlife Management 46:757–760.

Pradel, R. 1996. Utilization of capture-mark–recapture for the study of
recruitment and population growth rate. Biometrics 52:703–709.

Prugh, L. R., C. E. Ritland, S. M. Arthur, and C. J. Krebs. 2005.
Monitoring coyote population dynamics by genotyping faeces. Molecular
Ecology 14:1585–1596.

Puechmaille, S. J., and E. J. Petit. 2007. Empirical evaluation of non-
invasive capture–mark–recapture estimation of population size based on a
single sampling session. Journal of Applied Ecology 44:843–852.

Raymond, M., and F. Rousset. 1995. An exact test for population
differentiation. Evolution 49:1280–1283.

Robinson, S. J., L. P. Waits, and I. D. Martin. 2009. Estimating abundance
of American black bears using DNA-based capture–mark–recapture
models. Ursus 20:1–11.

Rudnick, J. A., T. E. Katzner, E. A. Bragin, and J. A. DeWoody. 2008. A
non-invasive genetic evaluation of population size, natal philopatry, and
roosting behavior of non-breeding eastern imperial eagles (Aquila heliaca)
in central Asia. Conservation Genetics 9:667–676.

Ruell E. W., S. P. D. Riley, M. R. Douglas, J. P. Pollinger, and K. R.
Crooksa. 2009. Estimating bobcat population sizes and densities in a
fragmented urban landscape using noninvasive capture–recapture sam-
pling. Journal of Mammalogy 90:129–135.

Schaller, G. B. 1998. Wildlife of the Tibetan Steppe. University of Chicago
Press, Chicago, Illinois, USA.

Schaller, G. B., and A. L. Kang. 2008. Status of Marco Polo sheep Ovis
ammon polii in China and adjacent countries: conservation of a vulnerable
subspecies. Oryx 42:100–106.

Solberg, K. H., E. Bellemain, O.-M. Drageset, P. Taberlet, and J. E.
Swenson. 2006. An evaluation of field and non-invasive genetic methods
to estimate brown bear (Ursus arctos) population size. Biological
Conservation 128:158–168.

Taberlet, P., L. P. Waits, and G. Luikart. 1999. Noninvasive genetic
sampling: look before you leap. Trends in Ecology and Evolution 14:323–
327.

Toı̈go, C., and J.-M. Gaillard. 2003. Causes of sex-biased adult survival in
ungulates: sexual size dimorphism, mating tactic or environment
harshness? Oikos 101:376–384.

Valiere, N., C. Bonenfantk, C. Toıgo, G. Luikart, J.-M. Gaillard, and F.
Klein. 2007. Importance of a pilot study for non-invasive genetic
sampling: genotyping errors and population size estimation in red deer.
Conservation Genetics 8:69–78.

White, G. C. 2005. Correcting wildlife counts using detection probabilities.
Wildlife Research 32:211–216.

Williams, B. K., J. D. Nichols, and M. J. Conroy. 2002. Analysis and
management of animal populations. Academic Press, San Diego,
California, USA.

Williams B. W., D. R. Etter, D. W. Linden, K. F. Millenbah, S. R.
Winterstein, and K. T. Scribner. 2009. Noninvasive hair sampling and
genetic tagging of co-distributed fishers and American martens. Journal
of Wildlife Management 73:26–34.

Wilson, G. J., A. C. Frantz, L. C. Pope, T. J Roper, T. A. Burke, C. L.
Cheesman, and R. J. Delahay. 2003. Estimation of badger abundance
using faecal DNA typing. Journal of Applied Ecology 40:658–666.

Associate Editor: Latch.

Harris et al. N Argali Abundance in the Afghan Pamir 677