Andean bear density and abundance estimates — How reliable and
useful are they?
David L. Garshelis1
Minnesota Department of Natural Resources, Grand Rapids, MN 55744, USA
Abstract: Few attempts have been made to estimate numbers and densities of Andean bears
(Tremarctos ornatus). It is understandable that the many challenges involved in these efforts
have made it difficult to produce rigorous estimates. A crude estimate of ,20,000 Andean bears
was derived by extrapolating the lowest observed density of American black bears (Ursus
americanus) across the range of Andean bears. A second estimate, based on rangewide genetic
diversity, produced a wide range of values; however, the low end of the confidence interval
roughly matched the population estimate based on minimum black bear density. A mark–
recapture analysis of 3 camera-trapped bears in Bolivia also yielded a similar density (4.4–6
bears/100 km2), but overlapping home ranges of 2 radiocollared bears at that same site
suggested a higher density (.12 bears/100 km2). Neither of these estimates can be considered
reliable or representative of the wider population because of the small sample sizes. Moreover,
the effective sampling area for the camera-trapping study was uncertain. A DNA hair-trapping
mark–recapture study in Ecuador sampled a greater number of bears (n 5 25) within a larger
study area, but a male-biased sex ratio suggested that closure was violated, precluding a simple
estimate of density based on the area of the trapping grid. Also, low capture rates in what was
perceived (from incidence of bear sign) as prime bear habitat might be indicative of a sampling
bias. These issues are not simply incorporated into confidence intervals (CIs): CIs only include
uncertainty due to sampling error, not biased sampling or an ambiguous sampling area.
Whereas these (low density) estimates may provide guidance for conservation, their greatest
usefulness may be in providing directions for improvement of future studies of Andean bears, as
well as bears in Asia, which also lack rigorous population estimates.
Key words: biased estimates, camera-trapping, DNA hair-sampling, effective trapping area, genetic diversity,
mark–recapture, precision, radio-telemetry, representative sampling, Tremarctos ornatus
Ursus 22(1):47–64 (2011)
Vaughan 2009, Clark et al. 2010), to extrapolate
densities from small study sites to larger areas
(Miller et al. 1997, Boulanger et al. 2002, Robinson
et al. 2009) or to apply across large regions
(Bellemain et al. 2005, Garshelis and Noyce 2006,
Dreher et al. 2007, Kendall et al. 2009).
For a variety of reasons, these rigorous population
estimation techniques have rarely been applied to
Andean bears (Tremarctos ornatus) in South America (or, for that matter, to the bears in Asia, other
than giant pandas [Ailuropoda melanoleuca]). Logistical and financial constraints are often prohibitive,
bear densities may be too low to obtain adequate
samples, and the motivation to obtain such estimates
may be lacking because the bears are not legally
hunted. Nevertheless, there has been growing interest as well as a few recent attempts to obtain
The pursuit of estimates of animal abundance is
pervasive, partly because such estimates are useful
for management and conservation, and partly
because they satisfy a certain basic human intrigue.
In North American bear studies, population estimates are generally derived from mark–recapture
sampling approaches, either involving actual bear
captures or ‘‘recaptures’’ obtained from photographs, sightings, or DNA from hair samples. Much
effort has been and is being invested in developing
scientifically rigorous population estimation techniques, with associated confidence intervals, that can
be used to guide management of bears on small,
selected areas (Gervasi et al. 2008, Immell and
Anthony 2008, Obbard and Howe 2008, Tredick and
1
[email protected]
47
48
ANDEAN BEAR POPULATION ESTIMATES N Garshelis
abundance estimates of Andean bears using some
newly developed tools.
Here I review the current knowledge of Andean
bear abundance. Rather than simply summarizing
the few reports of population estimates, I describe
and thoroughly critique the methods and data. These
criticisms are not meant to rebuke the efforts of any
of the investigators, all of whom worked hard to
collect and analyze data, but to help evaluate the
veracity of the results and thereby aid in moving the
science of population estimation forward, especially
in the context of the Andean bear. Investigators are
often well aware of the foibles of their work, but may
be dissuaded from highlighting those in the publication process. Here, I take the role of a post-hoc
reviewer, aiming to enhance interpretations of past
works and provide guidance for future efforts.
Rangewide population estimate derived
from density of a surrogate species
While preparing a conservation action plan for
Andean bears across their range, Peyton (1999:160)
summarized their status and distribution. He indicated that a group of experts were ‘‘confident that
there are at least 18,250 wild spectacled [Andean]
bears.’’ These bears occupied at least 50 distinct
habitat fragments totaling 260,000 km2. Four of
these fragments were estimated to contain .1,000
adult bears each. The source for these estimates of
bear numbers was mentioned in an earlier report by
Peyton et al. (1998:87). The estimated total area
occupied was derived by ‘‘drawing boundaries on
topographic maps around areas where field expeditions confirmed the presence of bears.’’ These
boundaries were based on habitats where Andean
bears were known to exist. Bear abundance within
this area was then estimated by extrapolating
densities of American black bears (Ursus americanus)
to the geographic range of Andean bears.
American black bears were used as a surrogate for
Andean bears because (1) no empirical estimates of
abundance existed for Andean bears at that time, (2)
at least 24 mark–recapture-based estimates were
available for American black bears (summarized by
Garshelis 1994), and (3) American black bears and
Andean bears are primarily forest-dwelling omnivorous ursids of similar size (although we now know
that the 2 species are fairly different ecologically).
Excluding cubs of the year, American black bear
densities ranged from 7–130/100 km2, with a median
of 25 bears/100 km2. Applying this median density to
Andean bear range produced an estimate of 65,000
bears. Applying the lowest American black bear
density yielded an estimate of ,18,000 (0.07 x
260,000). Including cubs increases the estimate to
.20,000.
The extreme lowest density reported for American
black bears is from an area of marginal black bear
habitat in Alaska — one of the northernmost black
bear populations ever studied (Miller et al. 1987,
Miller 1994). This population was also hunted. Thus,
Peyton et al. (1998) were confident that the ,20,000
estimate for Andean bears, derived from this very
low black bear estimate, should be a minimal value.
Moreover, the median American black bear density
was .3x higher even though a large number of the
black bear density estimates were from hunted
populations, whereas Andean bear populations are
technically unhunted (although certainly some animals are illegally killed, especially when they
depredate livestock or crops; Jorgenson and
Sandoval-A 2005, Goldstein et al. 2006).
Rangewide population estimate derived
from genetic diversity
The degree of genetic variation in a population is a
reflection of mutation rate, genetic drift, natural
selection, immigration, and population size. A
positive relationship between genetic diversity and
population size has been corroborated with studies
from a host of species (Frankham 1996), although
the exact form of this relationship is more conceptual than empirical. In theory, however, with data on
genetic diversity and assumed values for the other
parameters, one can calculate a long-term average
effective population size (Ne). Ne is the size of an
idealized population that would produce the genetic
structure actually observed. Whereas this conceptual
definition is straightforward, the estimation of Ne
and its conversion to number of animals (N) has
been performed in a number of ways, yielding varied
results (Luikart et al. 2010). Models are also
available to detect previous population bottlenecks
from allelic frequencies. These procedures have
recently been employed, for example, to estimate
the size of ancestral brown bear (Ursus arctos)
populations (Saarma et al. 2007) and to quantify and
date the collapse of giant pandas, other carnivores,
Ursus 22(1):47–64 (2011)
ANDEAN BEAR POPULATION ESTIMATES N Garshelis
and great apes (Aspi et al. 2006, Goossens et al.
2006, Zhang et al. 2007).
Ruiz-Garcia (2003) used these genetic approaches
to produce a crude estimate of population size and
change for Andean bears. He obtained genetic
samples (hairs, bones, teeth, skin, and blood)
collected by field researchers and from zoos in the
northern part of the bear’s range: Venezuela,
Colombia, and Ecuador (although most zoo samples
were of unknown origin). He analyzed the genetic
diversity of these samples, and using populationgenetic models concluded that (1) the species was
genetically impoverished, (2) populations in Venezuela, northern Colombia, and Ecuador were genetically isolated, but (3) these conditions dated back to
ancestral Andean bears. Using estimated mutation
rates, he further calculated that these populations
diverged some 24,000 years ago, and thus the genetic
fragmentation was not a result of the European
human invasion 500 years ago.
A major finding of this study was that populations
of Andean bears did not pass through a genetic
bottleneck: Ruiz-Garcia (2003) concluded that population size has remained constant over the evolutionary history of this species. Applying a range of
mutation rates varying by more than an order of
magnitude, he calculated long-term effective population sizes. Two different mutational models produced widely varying results. For example, in one
model he obtained Ne estimates of 644–3,044 for
Colombia and 325–1,348 for Ecuador. In the second
model, his estimates were 1,888–12,875 for Colombia
and 1,429–8,017 for Ecuador.
Country-specific estimates of Ne were then expanded to rangewide estimates. Unfortunately, the
process for doing so was not explained, so cannot be
evaluated here. The full span of reported estimates
was 5,500–30,000 bears; it is unclear, though, how
the upper end of this range was calculated, given that
the maximum values for just Colombia and Ecuador
totaled nearly 21,000, while Peru and Bolivia are
reported to have the most bears (69% of area of bear
range; Peyton 1999).
To be of conservation value, the estimates of
effective population size (which is a function of
population fluctuations, sex ratio, proportion of
animals breeding, litter size, generation time, etc.)
must be converted to real population size. Many
estimates exist in the literature for the ratio of Ne to
N. Frankham (1995) compiled results from a
multitude of studies and found that estimates of Ne
Ursus 22(1):47–64 (2011)
49
tended to center around ,0.1 N; few valid (data rich)
estimates of Ne were .0.5 N. Harris and Allendorf
(1989) estimated a Ne /N ratio of 0.24–0.32 for grizzly
(brown) bears; this range encompasses the narrower
range calculated by Miller and Waits (2003) for the
Yellowstone grizzly bear population (0.27–0.32).
Applying Harris and Allendorf’s range of Ne /N
ratios to Ruiz-Garcia’s (2003) estimates of Ne would
yield total population estimates of 17,000–125,000
Andean bears. Ruiz-Garcia (2003) asserted, however, that a more credible range of Ne values was only
19,000–24,000, and also elected to use a much higher
correction factor (0.75) to convert Ne to N.
Hence, starting with samples from parts of 3
countries and initially applying 2 genetic models, a
range of plausible mutation rates, and a range of
proportions to convert Ne to N, Ruiz-Garcia
(2003:91) ultimately honed in on a value of about
25,000 Andean bears rangewide. He rejected higher
calculated values because they were inconsistent with
‘‘the IUCN census,’’ by which he meant the estimate
of Peyton (1999), derived from application of the
lowest density American black bear population to
the rangewide area of Andean bears.
It must be stressed that the estimates stemming
from genetic diversity are averaged across the span
of time that the species has existed, not a recent point
in time. Thus, if the species declined appreciably in
modern times, the long-term average would be
essentially meaningless for a current assessment.
Ruiz-Garcia (2003) concluded, though, that no
significant bottleneck occurred, so the long-term
estimate is essentially the same as the current
estimate. Given that, he argued that estimates in
the range of 100,000 bears, resulting from some
models, mutation rates, and Ne /N correction factors,
must be wrong owing to their discordance with the
‘‘census.’’ What seems clear is that Peyton’s (1999)
estimate constrained the range of results reported
from the genetic-based estimate; thus, the 2 estimates
are not truly independent. Both started with a wide
range, but were trimmed to a value near 20,000
bears.
More recent genetic studies, with larger sample
sizes from across a larger geographic area and
including more microsatellite loci (Ruiz-Garcia et
al. 2005, Viteri and Waits 2009) have found higher
genetic diversity than originally reported, which
would convert to a higher estimate of the current
bear population. Even these recent estimates of
genetic diversity may be biased low because the
50
ANDEAN BEAR POPULATION ESTIMATES N Garshelis
molecular markers employed were derived from
other bear species, which are not closely related to
Andean bears (Ruiz-Garcia et al. 2005, Viteri and
Waits 2009). Using markers that were not ascertained from the target species may inhibit the ability
to discern differences between some alleles (referred
to as ascertainment bias). Consequently, it seems
likely that the original population estimates of RuizGarcia (2003) will be substantially modified by
future derivations based on genetic diversity.
Local population estimate based on
camera captures
The Andean bear has highly individualistic color
markings, which may include variable white or buffcolored rings or partial rings around the eyes and
streaks or speckled patches on the muzzle, chin,
neck, and chest. Often these markings are bilaterally
asymmetrical. Hence, photos taken by remote
cameras of the face, neck, and chest of these bears
can be used to identify individuals. Others have
employed individual identifications based on camera-trap photos to estimate populations of various
species of striped or spotted carnivores or ungulates,
via a ‘mark’–resight approach (Karanth 1995,
Karanth and Nichols 1998, Elkan 2003, Trolle and
Kéry 2003, Silver et al. 2004, Jackson et al. 2006,
Marnewick et al. 2008, Trolle et al. 2008, Gerber et
al. 2010). The ‘marking’ phase is just an initial photo
of a naturally-marked animal; subsequent photos
represent resighting events.
This method of population estimation was exploited by Rı́os-Uzeda et al. (2007) for Andean bears
in an area straddling 2 adjacent protected areas in
northern Bolivia. They selected 17 camera-trap
locations based on presence of bear sign; habitat
that bears rarely used was discounted in the search
for potential trap sites. Chosen sites included 9 in
páramo and 8 in the forest. They set 2 cameras at
each site (to obtain opposing views of bears). They
did not use bait or scent to attract bears to the site.
Cameras were checked weekly for a period of
1 month, after which 7 photographs of 3 identifiable
bears were obtained.
Mark–recapture analysis indicated that only these
3 bears occupied this area. To convert this estimate
of abundance to density, the researchers estimated
the effective sampling area. They employed a
commonly-used technique whereby the trapping area
is increased by a buffer strip equal to K the
estimated width of an average home range (or the
radius of a circular home range) to account for the
bears’ use of areas beyond the camera sites. This
width was initially estimated based on the average
maximum distance between each individual’s photographed locations. This yielded a buffer width of
only 0.25 km around the camera-trap sites, and a
density estimate of 13.6 bears/100 km2 (95% CI 5 8–
19/100 km2). This buffer strip was obviously too
small, as a home range radius of 0.25 km would
equate to a home range area of only 0.2 km2.
A previous study in this same area by Paisley
(2001), with radiocollared bears, yielded home range
estimates of .7 km2, or an estimated radius of
1.5 km. One of these previously collared bears
appeared to have been recaptured at Rı́os-Uzeda et
al.’s (2007) camera traps. Using this home range
radius to buffer the camera-trap locations and define
the effective trapping area (Fig. 1), Rı́os-Uzeda et al.
(2007) calculated a density of only 6 bears/100km2
(3.6–8.5/100 km2). Castellanos (2004, 2011) observed
larger home ranges in Ecuador: 15 km2 for females
and 59 km2 for males. Applying Castellanos’ female
home range radius to the camera-trap area produced
a density estimate of 4.4 bears/100km2 (2.6–6.2/
100km2). Although the sex of the photographed
bears was unknown, Rı́os-Uzeda et al. (2007) elected
not to use Castellanos’ male home range estimates to
buffer the trapping area and concluded that a density
of 4.4–6 bears/100km2 was most likely.
Much recent literature on a large number of
camera-trapped species has been concerned with the
estimation of effective trapping area. One study of a
known population of leopards (Panthera pardus)
concluded that a strip width of K the mean distance
between camera captures produced a reasonably
accurate density estimate (Balme et al. 2009). Other
studies found that better results were obtained with
buffer strips equal to the average maximum distance
between captures rather than K that distance
(Parmenter et al. 2003, Sharma et al. 2010). Various
model-based approaches have also been proposed
that use the spatial distribution of all capture data to
estimate the effective area of use (Efford et al. 2004a,
2009; Gardner et al. 2009, 2010; Royle et al. 2009);
these approaches are better supported mathematically than the rather informal buffer strip calculations. Still other authors have argued that home
ranges need to be estimated independent of the
camera trap data because, especially when there are
few recaptures, the distances between them are
Ursus 22(1):47–64 (2011)
ANDEAN BEAR POPULATION ESTIMATES N Garshelis
Fig. 1. Camera-trapping sites (shown as camera
icons) used by Rı́os-Uzeda et al. (2007) to obtain a
mark–recapture estimate of the population size of
Andean bears in northern Bolivia. Three individual
bears were photographically identified. These authors used 3 varying-sized buffers, as shown
(adapted from their figure), to define the effective
trapping area and thereby estimate bear density. The
inside-most buffer (darkest shading) was based on
estimated home range radius derived from camera
captures of the same individual at different sites.
Two outer buffers were based on home ranges
estimated from 2 radiotelemetry studies (Paisley
2001; Castellanos 2004, 2011). Innermost (white)
area was assumed to be trapped, but actually shows
a sizeable void where bears would not have encountered any traps.
invariably underestimates of actual movements
(McCarthy et al. 2008). Soisalo and Cavalcanti
(2006) found that GPS-collared jaguars (Panthera
onca) moved nearly twice as far as indicated by their
recapture distances, even though camera traps were
set in places with many GPS locations. The
photographic data alone overestimated density of
jaguars by .70%. For ocelots (Leopardus pardalis),
camera-trap based buffers yielded density estimates
that were more than twice that obtained using
telemetry-based estimates of the effective area
(Dillon and Kelly 2008). Conversely, in a study of
American black bears, Tredick and Vaughan (2009)
found that the estimated density based on a
telemetry-derived buffer strip was, on one study site,
about 2x the estimate derived from spatial analysis
Ursus 22(1):47–64 (2011)
51
(Program DENSITY, Efford et al. 2004b) of the
recapture data (DNA hair samples in this case).
Tredick and Vaughan (2009) considered the latter
estimate to be more accurate in this case because the
telemetry data were too limited. Obbard et al. (2010)
concluded that maximum-likelihood-based estimators of the effective trapping area derived from
spatial distribution of recapture data (Efford et al.
2009) performed better than ad hoc boundary strips
because they include heterogeneity in detection and
also because they incorporate the variance of the
area estimate as part of the variance of the density
estimate. However, these sorts of modeling procedures require more data than are likely to be
obtained in an Andean bear camera-trapping study.
In the case of Rı́os-Uzeda et al.’s (2007) study, it is
clear that the camera recapture data alone would
have underestimated the effective trapping area. But
even the telemetry data that were used to correct the
effective area were subject to considerable error. In
Paisley’s (2001) study, the collared bears often
wandered out of range of the telemetry equipment.
Including areas where radio signals emanated but
reliable triangulated locations could not be achieved
caused the estimated home range size for 1 of 2 bears
to nearly double. Yet even this was an underestimate
because radio signals were not audible on 16–20% of
tracking attempts within the area that the bears were
known to occupy.
If, as suggested by Paisley’s (2001) locational and
presence–absence data, the real home ranges of these
bears was 20 km2 or more, the total area encompassed by Rı́os-Uzeda et al.’s (2007) camera sites was
equivalent to about 1 home range. Based on
empirical camera-trapping data with ocelots, Maffei
and Noss (2008) recommended a minimum study
area size of 4 home ranges. The smaller the study
area size relative to home range size, the less
geographic closure (Garshelis 1992), and hence the
greater the influence of the chosen buffer area on the
density estimate. In other words, the smaller the
camera-trapping area, the greater the dependence on
radiotelemetry data for estimation of home range
size and effective sampling area.
Small sampling areas are also apt to be nonrepresentative of the larger regional area by chance
alone. For example, Matthews et al. (2008) sampled
2 study sites in California with camera traps and
observed American black bear densities 7x higher on
1 site than the other. Indeed, 1 of the sites had one of
the highest densities ever reported for this species
52
ANDEAN BEAR POPULATION ESTIMATES N Garshelis
(133 bears/100 km2). The 2 areas had similar habitat
characteristics and were only 4 km apart, but during
the period of camera trapping (,6 weeks) the site
with high density had an abundance of blackberries
(Rubus discolor), which may have attracted bears. In
fact, there were only 3 bears on 1 site and 5 on the
other, so the dramatic disparity in apparent density
was really attributable to a difference of just 2 bears.
With such small numbers of animals, results are not
likely to be meaningful and may be misleading in
representing the larger surrounding area.
The number, spacing, and layout of the traps are
other important considerations. Karanth and Nichols (1998:2855) stressed that it is important ‘‘to
ensure coverage of the entire area, without leaving
holes or gaps that were sufficiently large to contain a
tiger’s movements during the sampling period.’’ In
Rı́os-Uzeda et al.’s (2007) study, a large gap
occurred in the middle of the trapping area
(Fig. 1). This area was not trapped because it was
considered marginal habitat for bears, and thus
largely unused (R. Wallace, Wildlife Conservation
Society, La Paz, Bolivia, personal communication,
2009). That seems like a logical decision in terms of
trapping efficiency, but it is difficult to reconcile,
after the fact, whether this area should have been
included in the density estimate.
Rı́os-Uzeda et al. (2007) demonstrated the potential feasibility of employing camera-based individual
identifications for population estimation of Andean
bears. But the study also highlighted shortcomings
that might be improved in future efforts of this kind,
depending on financial and logistical constraints
(e.g., larger study area with more camera traps and
no large gaps between traps). A major issue, though,
that may be difficult to control is sample size; this
depends both on bear density and their likelihood of
passing in front of a camera (which would be
enhanced with a lure, but that also influences
recapture rates and may increase heterogeneity in
probability of capture). Camera trap studies of other
species have generated population estimates from
small sample sizes (e.g., ,6 individuals; McCarthy et
al. 2008, Lynam et al. 2009, Rayan and Mohamad
2009), but in such cases individual capture heterogeneity is likely to go undetected, resulting in biased
estimates (Link 2003, 2004); moreover, the derived
CIs are likely to underestimate sampling uncertainty
because the data are too sparse to assess actual
heterogeneity. Certainly, with small numbers of
recaptures, the buffer size used to estimate density
is prone to be flawed, pointing to the need for
independent telemetry data. Notably, Rı́os-Uzeda et
al.’s (2007) estimate of $15,000 in expenses and 24
person-weeks of field effort to conduct their study
did not include the expenses and effort also needed
to obtain the telemetry data upon which they relied.
A final issue related to camera-trapping is the
ability to discriminate individuals from photographs
that are often taken at different angles and distances
from the bears and under varying ambient conditions. For Andean bears, the key is to clearly see the
markings of their face, and possibly their neck and
chest. Rı́os-Uzeda et al. (2007) used 2 cameras at
each site, positioned opposite each other, to avert the
potential problems of obtaining a photograph of
only 1 side of the face and not being able to match
that with a different photograph of the other side.
They recommended that future studies use 3 cameras
to better discriminate individuals. Rı́os-Uzeda et al.
(2007) claimed to be certain of individual identification for each of the 7 bear photographs that they
obtained. However, a problem with photographs is
that identification of individuals is subjective;
investigators may differ in their judgments about
distinguishing characteristics (Zug 2009). Had Rı́osUzeda et al. (2007) misidentified just 1 bear, their
estimates of density would have been very different.
More photos of more bears would reduce the effect
of a small number of misidentifications, but would
increase the chance of encountering many similarlooking bears; indeed, there may be a limit to the
number of individual bears that can be reliably
distinguished from photographs, especially those
taken in uncontrolled field conditions.
Local population estimate based
on radiotelemetry
Paisley (2001) radiocollared and tracked 2 bears in
essentially the same area where Rı́os-Uzeda et al.
(2007) later conducted their camera-trapping study.
Although Paisley (2001) did not attempt to estimate
density because of the small sample size, a crude
density estimate can be extracted from her data. The 2
bears were radiotracked mainly from ridgelines and
were located in the surrounding grasslands and forest,
especially along the slopes and valleys on either side of
1 main ridge in the central part of the study area. If
the bears were present in that area, their radio signals
would have been clearly heard and their locations
easily obtained by triangulation. When bears traveled
Ursus 22(1):47–64 (2011)
ANDEAN BEAR POPULATION ESTIMATES N Garshelis
beyond this area, however, they were more difficult to
locate by triangulation, so often just a single radio
bearing (or multiple non-intersecting bearings) was
obtained. Other times, when the bears traveled even
farther beyond the central study area, their radio
signals could not be heard at all.
After a full year of radiotracking, 1 bear was
located in the central study area on 64% of tracking
attempts and the other located on 70% of tracking
attempts. The portions of their home ranges within
this area highly overlapped (Fig. 2). Circumscribing
the combined area of use yields an area of ,12 km2.
Hence, during the course of a year, 1 of the bears
was present 64% of the time and the other 70% of
the time within this 12-km2 area; this yields a mean
of 0.64 + 0.70 5 1.34 bear-equivalents/12 km2, or 11
bears/100 km2. Occasionally (although rarely), some
other bears were observed in the area, so it is
probably safe to assume that 12 bears/100 km2 is a
minimum density estimate. Importantly, this represents the average density over the course of a year —
at times there might have been 3 or 4 bears present
[25–33/100km2], either by chance or due to availability of some type of food, and at other times 0 or 1
[0–8/100km2]).
Although this is not a formal estimate of density,
because it does not rigorously account for bears
other than the 2 that were collared, it certainly
represents a minimum and covers a much longer
period than the 1-month camera-trapping study in
the same area. Notably, this minimum density is 2–
3x higher than that reported in the camera-trapping
study; however, it is very close to the estimate of 13.6
bears/100 km2 that was rejected in the cameratrapping study because the buffer was deemed to be
unreasonably small (Rı́os-Uzeda et al. 2007). Ironically, the more reasonable camera-trapping buffer,
taken directly from Paisley’s (2001) telemetry data,
yielded a much lower density estimate than derived
directly from Paisley’s data.
Several possible explanations exist for the discrepancy between these 2 density estimates.
1.
2.
Density declined during the 5-year interlude
between the studies (1999–2004).
Density in this high-altitude area was lower
during the height of the dry season (Aug–
Sep), when the camera-trapping study was
conducted, than during other times of
year because the bears tended to use other
areas.
Ursus 22(1):47–64 (2011)
53
Fig. 2. Portions of home ranges of 2 Andean bears
in northern Bolivia (adapted from Paisley 2001).
Convex polygons enclose locations obtained by
triangulation over a year, and outer rectangle
(12 km2) circumscribes both polygons. The 2 bears
were located within this core area 64% and 70% of
the time that they were radiotracked, but were
outside this area the rest of the time (i.e., either their
radio signal could not be heard or the signal came
from outside this area, but location could not be
determined). These data yield a density estimate
higher than that obtained by camera trapping in the
same area 5 years later (Fig. 1).
3.
4.
5.
The camera-trapping study site included
areas, possibly of lower density, outside the
12-km2 telemetry study site.
The density estimate from the camera-trapping study was biased low due to 1 or more
of the factors described in the previous
section.
The estimates are a function of small study
areas and small sample sizes and are not
really biologically different.
Any or all of these may have contributed, but I
favor point (5) as the primary explanation.
54
ANDEAN BEAR POPULATION ESTIMATES N Garshelis
Local population estimate based on
DNA hair captures
Undoubtedly the most rigorous estimate of
abundance and density of Andean bears yet obtained
was that of Viteri (2007). Her study was a mark–
recapture design where the marks were DNA
identifications of hair samples (Woods et al. 1999,
Boulanger et al. 2004a, Lukacs and Burnham
2005a). Hair samples were collected by stringing 2
strands of barbed wire around tree trunks and
wooden posts to form a small enclosure and putting
a scent lure in the middle. Bears attracted to the lure
crawled over or under the wires, and in so doing, left
some hair with follicles (containing DNA) on the
barbs.
This study was conducted within the CayambeCoca Ecological Reserve in northern Ecuador. Hair
traps were set out in a grid pattern of 9-km2 cells
(3 km apart). This cell size was chosen based on
presumed movements of bears to afford each bear
living in the area an opportunity to encounter at
least 1 trap. A total of 103 grid cells (94 of which had
a hair trap) encompassed 927 km2, a reasonably
large study area size. Sampling was conducted over a
period of .3 months, during which time each cell
was checked and hairs collected 4 times.
Viteri (2007) obtained genotypic information for
54 samples, from which she identified 25 individual
bears. Various mark–recapture estimators (Programs CAPTURE and CAPWIRE; Otis et al.
1978, Miller et al. 2005) yielded population estimates
of 33–48 bears (95% CI for full range of estimates:
26–69). Converting these estimates to density requires population closure; accordingly, this study
was established within boundaries that were thought
to be barriers to bear movements (i.e., high ridges
and presumed marginal habitat). An analytical test
indicated that the population was closed, but this
pertains only to demographic, not geographic
closure.
A highly-skewed sex ratio in all portions of the
study area (Fig. 3) suggests that the study area was
not geographically closed: in total, 15 (65%) of the
identified bears were males and only 8 were females
(2 were unidentified as to sex). Sex ratios of captures
could be skewed by a combination of the following:
(1) an uneven sex ratio in the living population, (2)
unequal vulnerability to capture, or (3) greater
movements by the more commonly captured sex
(males in this case). This population was within a
Fig. 3. Grid cells (3 x 3 km2, n = 94) in the CayambeCoca Ecological Reserve, Ecuador, where Andean
bears were hair-trapped to produce a population
estimate (adapted from Viteri 2007). Study area cells
with no hair trap were omitted (except 1 interior cell
marked with X). Quality of habitat in each cell was
categorized as prime (class 1), moderate (class 2–3),
or poor (class 4–5), based on classes of habitat
suitability (Cuesta et al. 2003), derived by modeling
incidence of bear sign in this same area. I categorized cells as to the predominant habitat suitability
class by visually overlaying Cuesta et al.’s (2003:Fig.
3) habitat suitability map with some extrapolations to
cells outside their study area boundaries using
Viteri’s (2007) habitat map. Individual identifications
and gender of hair-trapped bears were determined
by genetic analysis (Viteri 2007). The number of
different individual bears and their sex (M = male, F =
female, U = unknown) identified in the eastern,
center, and western portions of the study site is
shown above the map. These 3 geographic areas
were delineated based on predominant habitat, for
illustrative purposes. Captures in all areas were
male-dominated, and capture rate did not correspond with habitat suitability: twice as many bears
were captured in the west, where habitat for bears
was considered mainly poor, as in the east, where
much of the habitat was considered prime, with
approximately the same trapping effort in each (n =
28 and 26 hair-traps respectively). Some bears were
detected in .1 portion of the study site, so totaling
the number captured in each geographic area (31)
exceeds the total number of different bears identified
across the whole area (25).
protected area, but even with some exploitation by
humans, the sex ratio of the living population would,
if anything, be expected to become more femalebiased because male bears are more prone to being
Ursus 22(1):47–64 (2011)
ANDEAN BEAR POPULATION ESTIMATES N Garshelis
killed when in conflict situations (Goldstein et al.
2006, Oi 2009). I suspect the capture of more males
than females partly reflected the larger home ranges
of males: the study area would be expected to
encompass a portion of more male than female home
ranges, including some males that lived mainly
outside its borders.
Skewed sex ratios in capture samples, suggesting
lack of geographic closure for wide-ranging males,
are common in studies of bears (Boulanger and
McLellan 2001, Boulanger et al. 2004a, Mowat et al.
2005). Combining the sexes can severely bias the
resulting estimates of abundance and density (Harmsen et al. 2011). One way to circumvent this is to
restrict the estimate to the sex that exhibits sufficient
closure (Harris et al. 2010).
Assuming geographic closure and simply dividing
the lowest and highest population estimates within
the 95% CI by the size of the study area yields a
range of 3–7 bears/100 km2 for Viteri’s (2007) study.
If the study area were not closed, allowing bears to
wander in and out, then the effective sampling area
(as discussed previously with regard to camera
trapping) was larger and the actual density lower
than this simple estimate. However, even without a
correction for lack of closure, these density estimates
are already quite low, and moreover, they probably
include some cubs (because 2 strands of barbed wire
were used to collect the hair samples). What is
unclear, though, is whether some bears, notably
females, might not have been attracted to the lure or
were behaviorally averse to crawling through the
barbed wire (Woods et al. 1999), making them
essentially uncatchable, at least over the short term
(M. Gibeau [Parks Canada, Lake Louise, Alberta,
Canada, personal communication, 2010] observed
this for grizzly bears filmed with remote video
cameras). If heterogeneity in capture probabilities
was the primary explanation for the skewed sex
ratio, then the population and corresponding density
estimates would have been biased low despite the
lack of geographic closure. The population estimators used in this study were designed to correct for
capture heterogeneity, but cannot account for
animals with very low capture rates (Link 2003,
2004).
A particularly enigmatic finding was that a high
proportion of the bears were captured in portions of
the study area that were previously identified as poor
Andean bear habitat, based on incidence of bear sign
(Cuesta et al. 2003), and a relatively low proportion
Ursus 22(1):47–64 (2011)
55
were captured in what had been categorized as prime
habitat (Fig. 3). Indeed, 35% of bear sign in Cuesta
et al.’s (2003) study within the same area was found
in montane cloud forest habitat (which comprised
about 5% of the study area), but only ,12% of
Viteri’s (2007) captures of identifiable bears occurred
in this habitat (my calculation); standardized for
season (May–Aug hair collection), .40% of bear
sign was found in this habitat. Conversely, the least
suitable habitat was judged to be a high-elevation
(.3,600 m) section of grassland páramo in the west
that was traversed by a road, and a low elevation
(,1,900 m) section in the east with a road and foot
path connecting 2 towns (Cuesta et al. 2003).
However, as many as 7 (28%) different bears were
captured in this anthropogenically-disturbed habitat
that was perceived, from sign, to be rarely used by
bears. This mismatch between the distribution of
sign and hair captures is difficult to reconcile, but
could suggest an under-sampling of bears in the
prime habitat (and thus underestimation of density).
Possibly the scent of the lure did not carry as far in
the forested habitat. Alternately, the detection of
sign was biased. However, sign was commonly
detected in the grassland páramo at sites distant
from roads (mean incidence 5 28% of all sign), and
scats and feeding signs in páramo tend to last
considerably longer than in cloud forest (Paisley
2001). This suggests that sign surveys would
overestimate (not underestimate) bear use of
páramo, making the apparent high rate of hair
captures in this habitat even more enigmatic. A
potential explanation is that DNA in hair samples
from cloud forest habitat degraded more quickly
than in samples from the páramo. Thus, whereas 54
samples provided identifications of 25 different
bears, another 94 samples yielded insufficient genetic
information to be included in the analysis (Viteri
2007), and these may have occurred more often in
the forested habitat.
Bias and precision of estimates
The accuracy and precision of a population
estimate are largely a function of sampling. One
can never expect to achieve a perfectly accurate
estimate, but with increased sampling, the estimated
value should approach the true value, and certainty
in the estimate should increase. Indeed, precision
typically increases (CIs become narrower) with
larger samples. However, CIs only account for
56
ANDEAN BEAR POPULATION ESTIMATES N Garshelis
variation within the sample data; they provide no
indication of bias.
Robson and Regier (1964) proposed that for
management purposes (which would include conservation monitoring), investigators should strive to
ensure that there is a 95% probability that population estimates are within 25% of true population size.
This desired precision suggests that upper and lower
limits of the 95% CI should be within 25% of the
point estimate. Assuming that the sampling is not
biased, one can calculate necessary sample sizes to
achieve this. In the simplest form of mark–recapture
population estimation (2-sample Petersen estimate),
a recapture sample that includes 16 marked animals
would tend to produce 95% confidence limits within
50% of the point estimate; a recapture sample with
,64 marked animals would be required to achieve
the desired 25% (C.J. Schwarz, 2006, An introduction to mark–recapture methods used in fisheries
management, unpublished report, Simon Fraser
University, Burnaby, British Columbia, Canada).
This sort of sample size may be impossible to attain
for Andean bears with some techniques (e.g., camera
trapping), although the same level of precision could
be achieved with smaller sample sizes within multiple
recapture sessions.
In attempting to obtain unbiased and precise
estimates, both the size and representativeness of the
sample are key. With very small samples, the
likelihood increases that the data are biased by an
odd event or unusual individual. In Paisley’s (2001)
study, for example, 1 of 2 radiocollared bears was a
conspicuously under-sized adult and the other was a
subadult, so their movements and habitat use may
not have represented the population at large (Paisley
and Garshelis 2006). Thus, although the density
estimate derived from the home range use of these
bears accurately portrayed the bear density within
the small area in which they spent most of their time,
that density may not have reflected bear density in
similar habitats across that region of Bolivia. In this
study, the bears were captured and handled, and
their weights, measurements, and body conditions
helped in assessing whether they were normal.
Camera trap photos and DNA do not enable such
an assessment.
For DNA-based studies, sex ratio and spatial
distribution of the capture samples provide the
primary indications of representativeness. Spatial
distribution of captures is expected to vary in a nonuniform environment. In fact, spatial distribution
has been used as an indicator of relative habitat
quality for other species of bears (Apps et al. 2004,
Nams et al. 2006, Proctor et al. 2010). Only if there is
a mismatch between the spatial distribution of
captures and some other independent assessment of
habitat quality (e.g., Viteri 2007 versus Cuesta et al.
2003) would there be cause for concern.
The underlying cause of a skewed sex ratio might
be revealed by comparing recapture rates of males
and females (Kendall et al. 2009). In the extreme
case, where the capture sex ratio is very lopsided
apparently due to 1 sex being under-sampled (e.g.,
females avoiding hair traps) or over-sampled (e.g.,
males routinely entering and leaving the study area),
a population estimate closer to truth might be
derived by selecting the sex with the presumably
better estimate and doubling it (Bellemain et al.
2005, Peacock et al. 2011). Whereas such an
approach may not seem analytically rigorous, it
must be stressed that analytical rigor typically
cannot overcome sampling biases, and extreme
efforts to correct such biases may be necessary.
A precise but biased estimate may be especially
misleading because the true abundance (or density)
could be well outside the CI. Sharma et al. (2010)
demonstrated this with a camera-trapping study in a
reserve with a known number of tigers (Panthera
tigris, n 5 14). By excluding subsets of the accumulated photos (from specific cameras), they simulated
the effects of varying camera trap densities on the
reliability of the population estimates. In a low
density tiger population with a low density of camera
traps, only 34% of simulated estimates contained the
true population size within their 95% CIs.
Camera trap-based population estimates are
particularly difficult to evaluate for bias because
gender information is usually lacking and, in some
cases at least, a significant number of photos may be
unidentifiable to individual because of inadequate
clarity, framing, or position of the bear (Zug 2009).
Also, bears with similar markings could be misclassified. A method to quantify potential errors in
identification and incorporate them into the CIs has
not yet been developed, but possibly could be
patterned after methods devised for genotyping
errors (Lukacs and Burnham 2005b, Petit and
Valiere 2006, Knapp et al. 2009, Wright et al.
2009). However, if these sorts of events are nonrandom (e.g., if some similar-looking individuals are
often confused), or even if they are random but
capture success is extremely low, this sort of
Ursus 22(1):47–64 (2011)
ANDEAN BEAR POPULATION ESTIMATES N Garshelis
incomplete, heterogeneous sampling can significantly bias the results (Link 2003, 2004). Even with
reasonably large capture samples, heterogeneous
sampling is common in bear studies, often resulting
in significant underestimates of population size
(Noyce et al. 2001).
To ensure reliable results from mark–recapture
models, average capture probabilities should generally be .0.2 (White et al. 1982, Boulanger et al.
2004a). For small populations (e.g., ,50 animals),
they should be higher, to account for heterogeneity,
avoid bias, and compute meaningful CIs. High trap
densities are needed to achieve such high capture
probabilities if home ranges are fairly small (Settlage
et al. 2008). Capture rates are bound to be low for
low-density Andean bear populations living in areas
where it is logistically difficult to set and check a
large number of camera traps over a wide enough
area to sample many individuals. Rı́os-Uzeda et al.
(2007) calculated a probability of capture of only
0.08 for their camera-trapping study and estimated a
population size of only 3 bears. Biologists thus face
the dilemma of whether employing an analytical
approach that is poorly equipped to handle such
small samples returns results that are reliable enough
to be of use in conserving species like the Andean
bear. If capture rates are low and sample sizes small,
investigators are unlikely to discern individual
capture heterogeneity (which is nearly ubiquitous)
and therefore would be unaware of the negative bias
in the resulting population estimate (Ebert et al.
2010, Proctor et al. 2010, Harmsen et al. 2011).
Whereas such underestimation errs on the side of
demanding more conservation attention to the
population, it also may hamper the ability to
recognize conservation efforts that have succeeded
in fostering population growth.
Reliability of Andean bear estimates
Reliability of population estimates can be gauged
from their precision, potential biases, and transparency of methods, results, and analyses (Conroy and
Carroll 2009). My review was intended to critically
evaluate these aspects of reliability for Andean bear
estimates. The issues raised here were not intended to
disparage the work of the investigators who strove,
sometimes with enormous effort, to provide information to the scientific and conservation communities that could aid in understanding and protecting
populations of these bears. Critical evaluation is a
Ursus 22(1):47–64 (2011)
57
process that helps to correct mistakes and misinterpretations and provide better direction for the
future. This process is especially important for
threatened bear populations because many of the
techniques regularly employed to track changes in
their population sizes have produced misleading or
uncertain results (Garshelis 2002). An unfortunate
irony is that the poorest data are available for the
species and populations of bears that are most
threatened.
Certainly there are ample reasons to question the
reliability of all of these estimates for Andean bears.
Peyton et al.’s (1998) estimates, based on extrapolation of American black bear densities, span a wide
range (18,000–65,000 bears) and so are much less
precise than indicated by conservative reports of just
the lower value, or a rounded-off value of ,20,000.
Some have nevertheless used the lowest density
figure to derive separate range–country estimates.
Obviously, no real assessment of reliability is
available for such estimates because they were based
on a surrogate species. Peyton et al.’s (1998) use of
the black bear density estimates filled an information
void and was arguably somewhat better than a
guess. Others have produced abundance and density
guesstimates for other species of bears based only on
occasional sightings and subjective assessments of
amount of sign that have been extrapolated over
large areas (e.g., sun bears [Helarctos malayanus],
Davies and Payne 1981; sloth bears [Melursus
ursinus], Yoganand et al. 2006).
Ruiz-Garcia’s (2003) genetic diversity-based estimates also span a wide range. However, this range
was trimmed through a process that was not easily
defensible or replicable. The apparent concordance
between the estimates of Ruiz-Garcia’s (2003) and
Peyton et al. (1998) owe in part to an effort to frame
the genetics estimates by what was considered to be
field census data. The central focus of Ruiz-Garcia’s
study was genetic heterogeneity and gene flow, but
again, in an attempt to fill a void, the methods may
have been stretched beyond their level of reliability
to generate population estimates.
Rı́os-Uzeda et al. (2007) specifically sought to
estimate density using camera-trapping, a technique
that had been successfully employed on other
species, mainly cats with distinctive pelages. Indeed,
some of the same investigators had previously used
camera traps to estimate abundance and density of
jaguars (Wallace et al. 2003) in one of the same
protected areas (Madidi National Park) where the
58
ANDEAN BEAR POPULATION ESTIMATES N Garshelis
bears were later studied. The jaguar study also
yielded a small number of captures and identifiable
individuals (but twice that of the bear study). Using
only telemetry-based buffers to estimate effective
trapping area and calculate density, Rı́os-Uzeda et
al.’s (2007) 95% CI (2.6–8.5/100 km2) was rather
imprecise; nevertheless, it suggested a low density,
equivalent to or less than the lowest reported
American black bear density, used by Peyton et al.
(1998). The issue, though, is whether the reported
95% CI actually contained the true population
density in this area, or if it was significantly biased.
This problem is not unique to this study or to bears,
but is likely pervasive for camera-trapping studies of
low-density species (Harmsen et al. 2011).
It is important to note that the reported CIs for
abundance estimates, when converted to density,
typically do not include the error associated with the
estimation of effective trapping area. So in the case
of Rı́os-Uzeda et al.’s (2007) study, the calculated
CIs assume that the effective area was a fixed, known
quantity (or, rather a choice among 3 quantities,
Fig. 1). Clearly that was not the case because the
buffer area was derived from estimates of home
range sizes, which are likely to vary substantially by
individual, sex, method of calculation, and number
of telemetry locations. Hence, the denominator in
the density calculation (area) probably has more
uncertainty than the numerator (abundance), significantly reducing the reliability of the estimate.
The density estimate based on Paisley’s (2001)
home range data was presented here with no
measure of precision. Certainly the area used by
the 2 bears (Fig. 2) was not measured without error:
potential sources of error include telemetry triangulation, sampling (even when the bears were within
this general area, they could have been in other
places when nobody was observing), and method of
depicting this portion of their home range (minimum
convex polygon). However, these sources of error are
essentially irrelevant because the area of use was
taken to be a rectangle encompassing both bears, a
very crude approximation that would tend to
underestimate density (because it includes areas
where the bears were never detected; Fig. 2). Hence
the only remaining major source of error would be
the proportion of time that the bears spent in this
area. Combining data for the 2 bears (because they
spent a similar proportion of time in the area), a
simple CI can be calculated (146 locations in 218
attempts 5 0.67, 95% CI 5 0.61–0.73, presuming the
sampling was random, which it was not, given
difficulties of sampling during the rainy season). This
suggests a range of 1.22 (2 x 0.61) to 1.46 bearequivalents in the 12-km2 area, or a density of 10.2–
12.2 bears/100 km2, which is probably far more
precise than warranted given the nature of this crude
calculation. But, especially with the addition of other
bears that were occasionally seen in this area, it
appears that this estimate is truly higher than the
camera-trapping estimate in the same area. When 2
estimates in the same area, based on different
methodologies, produce results with non-overlapping CIs, the reliability of 1 or both estimates must
be called into question.
Viteri’s (2007) estimate of Andean bear abundance
was well-designed and carefully conducted, yet the
precision of the estimate was low (within 30–50% of
point estimates) for monitoring purposes. Wide CIs,
though, are not uncommon for even the most
rigorous population estimates of bears (Miller et al.
1997, Boulanger et al. 2002). Moreover, population
trends may be gleaned from a series of estimates with
overlapping CIs (Garshelis and Hristienko 2006,
Clark et al. 2010), and estimates of population
growth (l) can even be obtained from mark–
recapture data that fail to produce a useful estimate
of abundance (e.g., Pradel model in Program
MARK; Boulanger et al. 2004b, Proctor et al. 2010).
Notably, Viteri’s (2007) estimates, converted to
densities, were equivalent to those of Rı́os-Uzeda et
al. (2007) and would have been even lower if
corrected for effective trapping area (lack of closure).
An important question, though, is whether these
results were actually biased low, given their discordance with models of habitat quality judged from
abundance of sign (Cuesta et al. 2003). Again,
conflicting results reduce confidence in reliability.
The DNA results, however, are at least highly
transparent. Geneticists routinely report error rates
associated with genotyping, and methods exist for
assessing and correcting effects of these errors on
population estimates (Lukacs and Burnham 2005b,
Petit and Valiere 2006, Knapp et al. 2009, Wright et
al. 2009). Thus, although genetic identifications
outwardly seem less apparent than photographs,
they are actually more strictly defined, due to
standardized and validated laboratory procedures
(Waits and Paetkau 2005), and are thus repeatable
and defensible. Unfortunately, individual identifications from camera trap photos are generally not
subject to the same strict standards so might be
Ursus 22(1):47–64 (2011)
ANDEAN BEAR POPULATION ESTIMATES N Garshelis
judged to be less reliable. That is not to say that less
care is taken in matching photographs. Indeed, other
studies have reported exacting protocols (Jackson et
al. 2006, Larrucea et al. 2007), and even computeraided photographic matching, which is helpful (or
even necessary) for complex pelage patterns (Kelly
2001, Speed et al. 2007, Hastings et al. 2008, Sherley
et al. 2010). In terms of Andean bears, however,
pattern matching often relies on just a few white
facial streaks, which can appear slightly different at
different angles, different lighting, or different fur
wetness, and different observers may employ different personal thresholds for distinguishing individuals. Some efforts have recently been made to better
define and standardize the process used to differentiate individuals of this species from camera-trap
photographs (Jones 2010). Improvements to this
methodology will be crucial if camera-trapping
studies are to be used to derive reliable estimates of
abundance (and survival rates), which will require
larger samples and greater certainty in recapture
events than obtained thus far.
Usefulness of population estimates
It seems almost tautological that if a population or
density estimate is reliable, it would be useful, but is
there any value in less reliable estimates? Potentially,
a collection of less rigorous estimates, all in the same
ballpark, could assist conservation assessments (e.g.,
evaluating risks and effects of poaching; SánchezMercado et al. 2008). Conversely, highly-varying
estimates might be useful for identifying relative
population strongholds and thereby defining conservation priority areas. Previously, such areas have
been identified based on incidence of sign (Peralvo et
al. 2005), but sign deposition and detection can vary
by habitat type, so difficulties may arise in comparing areas with different habitat compositions.
Depending on the particular circumstances, this sort
of bias may either be diminished or exacerbated by
attempting to progress from an index to an estimate.
For example, Rı́os-Uzeda et al. (2007) found it
difficult to obtain camera captures of bears in high
altitude grasslands because vegetation caused innumerable false camera-triggers, whereas sign was
readily observable in this habitat (Rı́os-Uzeda et al.
2005). Obviously, researchers and conservationists
must decide when the extra effort of obtaining a
population estimate is warranted, and what level of
accuracy is required.
Ursus 22(1):47–64 (2011)
59
The existing population estimates for Andean
bears, reviewed and evaluated here, are all useful in
one respect. They are a good demonstration of a
variety of approaches and highlight the kinds of
biases and other shortcomings that are likely to be
encountered in future studies. They exemplify the
desire to better understand Andean bear populations
as well as the difficulties of doing so (specifically,
issues of small sample size and sampling biases).
Others studying small bear populations have noted
the benefit of combining different sorts of sampling
— for example, setting some hair snares at feeding
sites where densities are relatively high, plucking
hairs off trees, and obtaining DNA from scats.
Doing so not only adds to the sample size, but also
helps to expose more of the population heterogeneity
(Boulanger et al. 2008, Gervasi et al. 2008, De Barba
et al. 2010, Stetz et al. 2010). If a bear eludes capture
through 1 sampling procedure, it might be captured
another way. Additionally, more reliable population
estimates may accrue from applying multiple analytical techniques, especially where sample size is
limited by small population size (Meijer et al. 2008).
Collectively, results obtained so far suggest that
Andean bears may tend to occur at lower densities
than most hunted populations of American black
bears (Garshelis 1994); in this respect, they may be
more akin to high-density interior populations of
brown bears (Schwartz et al. 2003). If this is upheld
through further work, it would indicate that these
bears have a relatively low resilience to human
exploitation.
Effective conservation strategies for Andean bears
have been developed in the absence of reliable
population estimates (Rodrı́guez et al. 2003, Castellanos et al. 2010). To that extent, it could be argued
that accurate estimates are not really necessary,
especially for a species that is not managed through
harvest. On the other hand, forecasts for the
persistence of this species within the many small,
isolated patches of habitat in which it tends to exist
depend largely on how many bears there really are
(Kattan et al. 2004).
The studies reviewed here suggest that these bears
may occur at much lower densities than once
thought, a finding, which if corroborated through
further research, has strong implications for conservation planning (e.g., assessing the value of protecting habitat fragments of a certain size or the need to
link small areas; Yerena et al. 2003, Kattan et al.
2004). However, because of the difficulty of con-
60
ANDEAN BEAR POPULATION ESTIMATES N Garshelis
ducting studies that are likely to return reliable
abundance estimates, such work should not supplant
the many other ecological, behavioral, and demographic studies that are needed to better understand
this species. Only through careful attention to the
selection of a study site, study design, representative
sampling, analytical approach, and critical scrutiny
of the results will population estimates be reasonably
reliable, but even with concern for all these details,
the many difficult circumstances associated with
studying this species will make it much harder to
obtain estimates with the same level of reliability as
achieved for North American bears. Consequently,
efforts directed toward population monitoring methods that do not require reliable population estimates
may be more useful. The same could probably be
said for most populations of bears in Asia;
accordingly, this review may be helpful to bear
biologists in that region as well.
Acknowledgments
I am especially thankful to I. Goldstein for
inviting me to present this paper at the Second
International Symposium on the Andean Bear, held
in Lima, Peru, November 2008. Symposium sponsors provided financial support. R. Wallace, M. P.
Viteri, and 3 anonymous reviewers provided helpful
comments on an earlier version of this manuscript.
Literature cited
APPS, C.D., B.N. MCLELLAN, J.G. WOODS, AND M.F.
PROCTOR. 2004. Estimating grizzly bear distribution and
abundance relative to habitat and human influence.
Journal of Wildlife Management 68:138–152.
ASPI, J., E. ROININEN, M. RUOKONEN, I. KOJOLA, AND C.
VILÀ. 2006. Genetic diversity, population structure,
effective population size and demographic history of
the Finnish wolf population. Molecular Ecology
15:1561–1576.
BALME, G.A., L.T.B. HUNTER, AND R. SLOTOW. 2009.
Evaluating methods for counting cryptic carnivores.
Journal of Wildlife Management 73:433–441.
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., AND B.N. MCLELLAN. 2001. Closure
violation in DNA-based mark–recapture estimation of
grizzly bear populations. Canadian Journal of Zoology
79:642–651.
———, G.C. WHITE, B.N. MCLELLAN, J. WOODS, M.
PROCTOR, AND S. HIMMER. 2002. A meta-analysis of
grizzly bear DNA mark–recapture projects in British
Columbia, Canada. Ursus 13:137–152.
———, S. HIMMER, AND C. SWAN. 2004b. Monitoring of
grizzly bear population trends and demography using
DNA mark–recapture methods in the Owikeno Lake
area of British Columbia. Canadian Journal of Zoology
82:1267–1277.
———, ———, J.G. WOODS, M.F. PROCTOR, AND C.
STROBECK. 2004a. Sampling design and bias in DNAbased capture–mark–recapture population and density
estimates of grizzly bears. Journal of Wildlife Management 68:457–469.
———, K.C. KENDALL, J.B. STETZ, D.A. ROON, L.P.
WAITS, AND D. PAETKAU. 2008. Multiple data sources
improve DNA-based mark–recapture population estimates of grizzly bears. Ecological Applications
18:577–589.
CASTELLANOS, A. 2004. Andean bear research in the Intag
Region, Ecuador. International Bear News 13(2):25–26.
———, J. CEVALLOS, A. LAGUNA, L. ACHIG, P. VITERI, AND
S. MOLINA. 2010. National strategy for Andean bear
conservation. Ministry of Environment of Ecuador,
Anyma Printing, Quito, Ecuador. (In Spanish.)
———. 2011. Andean bear home ranges in the Intag
Region, Ecuador. Ursus 22:65–73.
CLARK, J.D., R. EASTRIDGE, AND M.J. HOOKER. 2010.
Effects of exploitation on black bear populations at
White River National Wildlife Refuge. Journal of
Wildlife Management 74:1448–1456.
CONROY, M.J., AND J.P. CARROLL. 2009. Quantitative
conservation of vertebrates. Wiley-Blackwell, Chichester, UK.
CUESTA, F., M.F. PERALVO, AND F.T. VAN MANEN. 2003.
Andean bear habitat use in the Oyacachi River Basin,
Ecuador. Ursus 14:198–209.
DAVIES, G., AND J. PAYNE. 1981. A faunal survey of Sabah.
World Wildlife Fund Malaysia, Project 1692, Kuala
Lampur, Malaysia.
DE BARBA, M., L.P. WAITS, P. GENOVESI, E. RANDI, R.
CHIRICHELLA, AND E. CETTO. 2010. Comparing opportunistic and systematic sampling methods for noninvasive genetic monitoring of a small translocated
brown bear population. Journal of Applied Ecology
47:172–181.
DILLON, A., AND M.J. KELLY. 2008. Ocelot home range,
overlap and density: Comparing radio telemetry with
camera trapping. Journal of Zoology 275:391–398.
DREHER, B.P., S.R. WINTERSTEIN, K.M. SCRIBNER, P.M.
LUKACS, D.R. ETTER, G.J.M. ROSA, V.A. LOPEZ, S.
LIBANTS, AND K.B. FILCEK. 2007. Noninvasive estimation of black bear abundance incorporating genotyping
errors and harvested bear. Journal of Wildlife Management 71:2684–2693.
Ursus 22(1):47–64 (2011)
ANDEAN BEAR POPULATION ESTIMATES N Garshelis
EBERT, C., F. KNAUER, I. STORCH, AND U. HOHMANN. 2010.
Individual heterogeneity as a pitfall in population
estimates based on non-invasive genetic sampling: a
review and recommendations. Wildlife Biology 16:225–
240.
EFFORD, M.G. 2004a. Density estimation in live-trapping
studies. Oikos 106:598–610.
———, D.K. DAWSON, AND C.S. ROBBINS. 2004b. DENSITY: Software for analyzing capture–recapture data
from passive detector arrays. Animal Biodiversity and
Conservation 27:217–228.
———, D.L. BORCHERS, AND A.E. BYROM. 2009. Density
estimation by spatially explicit capture–recapture:
Likelihood-based methods. Pages 255–269 in D.L.
Thompson, E.G. Cooch, and M.J. Conroy, editors.
Modeling demographic processes in marked populations. Springer, New York, New York, USA.
ELKAN, P.W. 2003. Ecology and conservation of bongo
antelope (Tragelaphus eurycerus) in lowland forest,
northern Republic of Congo. Dissertation, University
of Minnesota, Minneapolis, Minnesota, USA.
FRANKHAM, R. 1995. Effective population size/adult
population size ratios in wildlife: a review. Genetical
Research 66:95–107.
———. 1996. Relationship of genetic variation to population size in wildlife. Conservation Biology 10:1500–
1508.
GARDNER, B., J.A. ROYLE, AND M.T. WEGAN. 2009.
Hierarchical models for estimating density from DNA
mark–recapture studies. Ecology 90:1106–1115.
———, ———, ———, R. RAINBOLT, AND P.D. CURTIS.
2010. Estimating black bear density using DNA data
from hair snares. Journal of Wildlife Management
74:318–325.
GARSHELIS, D.L. 1992. Mark–recapture density estimation
for animals with large home ranges. Pages 1098–1111 in
D.R. McCullough and R.H. Barrett, editors. Wildlife
2001: Populations. Elsevier Applied Science, London,
UK.
———. 1994. Density-dependent population regulation of
black bears. Pages 3–14 in M. Taylor, editor. Densitydependent population regulation in black, brown, and
polar bears. International Conference on Bear Research
and Management Monograph Series 3.
———. 2002. Misconceptions, ironies, and uncertainties
regarding trends in bear populations. Ursus 13:321–334.
———, AND H. HRISTIENKO. 2006. State and provincial
estimates of American black bear numbers versus
assessments of population trend. Ursus 17:1–7.
———, AND K.V. NOYCE. 2006. Discerning biases in a
large scale mark–recapture population estimate for
black bears. Journal of Wildlife Management
70:1634–1643.
GERBER, B., S.M. KARPANTY, C. CRAWFORD, M. KOTSCHWAR, AND J. RANDRIANANTENAINA. 2010. An assessment
Ursus 22(1):47–64 (2011)
61
of carnivore relative abundance and density in the
eastern rainforests of Madagascar using remotelytriggered camera traps. Oryx 44:219–222.
GERVASI, V., P. CIUCCI, J. BOULANGER, M. POSILLICO, C.
SULLI, S. FOCARDI, E. RANDI, AND L. BOITANI. 2008. A
preliminary estimate of the Apennine brown bear
population size based on hair-snag sampling and
multiple data source mark–recapture Huggins models.
Ursus 19:105–121.
GOLDSTEIN, I., S. PAISLEY, R. WALLACE, J.P. JORGENSON, F.
CUESTA, AND A. CASTELLANOS. 2006. Andean bear–
livestock conflicts: A review. Ursus 17:8–15.
GOOSSENS, B., L. CHIKHI, M. ANCRENAZ, I. LACKMANANCRENAZ, P. ANDAU, AND M. BRUFORD. 2006. Genetic
signature of anthropogenic population collapse in
orang-utans. PLoS Biology 4:285–291.
HARMSEN, B.J., R.J. FOSTER, AND C.P. DONCASTER. 2011.
Heterogeneous capture rates in low density populations
and consequences for capture–recapture analysis of
camera-trap data. Population Ecology 53:253–259.
HARRIS, R.B., AND F.W. ALLENDORF. 1989. Genetically
effective population size of large mammals: An assessment of estimators. Conservation Biology 3:181–191.
———, J. WINNIE, JR., S.J. AMISH, A. BEJA-PEREIRA, R.
GODINHO, V. COSTA, AND G. LUIKART. 2010. Argali
abundance in the Afghan Pamir using capture–recapture modeling from fecal DNA. Journal of Wildlife
Management 74:668–677.
HASTINGS, K.K., L.A. HIBY, AND R.J. SMALL. 2008.
Evaluation of a computer-assisted photograph-matching system to monitor naturally marked harbor seals at
Tugidak Island, Alaska. Journal of Mammalogy
89:1201–1211.
IMMELL, D., AND R.G. ANTHONY. 2008. Estimation of black
bear abundance using a discrete DNA sampling device.
Journal of Wildlife Management 72:324–330.
JACKSON, R.M., J.D. ROE, R. WANGCHUCK, AND D.O.
HUNTER. 2006. Estimating snow leopard population
abundance using photography and capture–recapture
techniques. Wildlife Society Bulletin 34:772–781.
JONES, T. 2010. Detection probability and individual
identification of the Andean bear (Tremarctos ornatus)
using camera trapping methods. Thesis, University of
Wisconsin, Madison, Wisconsin, USA.
JORGENSON, J.P., AND S. SANDOVAL-A. 2005. Andean bear
management needs and interactions with humans in
Colombia. Ursus 16:108–116.
KARANTH, K.U. 1995. Estimating tiger Panthera tigris
populations from camera-trap data using capture—
recapture models. Biological Conservation 71:333–338.
———, AND J.D. NICHOLS. 1998. Estimation of tiger
densities in India using photographic captures and
recaptures. Ecology 79:2852–2862.
KATTAN, G., O.L. HERNÁNDEZ, I. GOLDSTEIN, V. ROJAS, O.
MURILLO, C. GÓMEZ, H. RESTREPO, AND F. CUESTA.
62
ANDEAN BEAR POPULATION ESTIMATES N Garshelis
2004. Range fragmentation in the spectacled bear
Tremarctos ornatus in the northern Andes. Oryx
38:155–163.
KELLY, M. 2001. Computer-aided photographic matching
in studies using individual identification: An example
from Serengeti cheetahs. Journal of Mammalogy
82:440–449.
KENDALL, K., J. STETZ, J. BOULANGER, A.C. MAClEOD, D.
PAETKAU, AND G.C. WHITE. 2009. Demography and
genetic structure of a recovering grizzly bear population. Journal of Wildlife Management 73:3–17.
KNAPP, S.M., B.A. CRAIG, AND L.P. WAITS. 2009.
Incorporating genotyping error into non-invasive
DNA-based mark–recapture population estimates.
Journal of Wildlife Management 73:598–604.
LARRUCEA, E.S., G. SERRA, M.M. JAEGER, AND R.H.
BARRETT. 2007. Censusing bobcats using remote cameras. Western North American Naturalist 67:538–548.
LINK, W.A. 2003. Nonidentifiability of population size
from capture–recapture data with heterogeneous detection probabilities. Biometrics 59:1123–1130.
———. 2004. Individual heterogeneity and identifiability
in capture–recapture models. Animal Biodiversity and
Conservation 27:87–91.
LUIKART, G., N. RYMAN, D.A. TALLMON, M.K. SCHWARTZ,
AND F.W. ALLENDORF. 2010. Estimation of census and
effective population sizes: The increasing usefulness of
DNA-based approaches. Conservation Genetics 11:
355–373.
LUKACS, P.M., AND K.P. BURNHAM. 2005a. Review of
capture–recapture methods applicable to noninvasive genetic sampling. Molecular Ecology 14:3909–
3919.
———, AND ———. 2005b. Estimating population size
from DNA based closed capture–recapture data
incorporating genotyping error. Journal of Wildlife
Management 69:396–403.
LYNAM, A.J., A. RABINOWITZ, T. MYINT, M. MAUNG, K.T.
LATT, AND S.H.T. PO. 2009. Estimating abundance with
sparse data: Tigers in northern Myanmar. Population
Ecology 51:115–121.
MAFFEI, L., AND A.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.
MARNEWICK, K., P.J. FUNSTON, AND K.U. KARANTH. 2008.
Evaluating camera trapping as a method for estimating
cheetah abundance in ranching areas. South African
Journal of Wildlife Research 38:59–65.
MATTHEWS, S.M., R.T. GOLIGHTY, AND J.M. HIGLEY. 2008.
Mark–resight density estimation for American black
bears in Hoopa, California. Ursus 19:13–21.
MCCARTHY, K.P., T.K. 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.
MEIJER, T., K. NORÉN, P. HELLSTRÖM, L. DALÉN, AND A.
ANGERBJÖRN. 2008. Estimating population parameters
in a threatened arctic fox population using molecular
tracking and traditional field methods. Animal Conservation 11:330–338.
MILLER, C.R., AND L.P. WAITS. 2003. The history of
effective population size and genetic diversity in the
Yellowstone grizzly (Ursus arctos): Implications for
conservation. Proceedings of the National Academy of
Sciences (USA) 100:4334–4339.
———, 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.
MILLER, S.D., E.F. BECKER, AND W.B. BALLARD. 1987.
Black and brown bear density estimates using modified
capture–recapture techniques in Alaska. International
Conference on Bear Research and Management
7:23–35.
———. 1994. Black bear reproduction and cub survivorship in south-central Alaska. International Conference
on Bear Research and Management 9(1):263–273.
———, G.C. WHITE, R.A. SELLERS, H.V. REYNOLDS, J.W.
SCHOEN, K. TITUS, V.G. BARNES, JR., R.B. SMITH, R.R.
NELSON, W.B. BALLARD, AND C.C. SCHWARTZ. 1997.
Brown and black bear density estimation in Alaska
using radiotelemetry and replicated mark–resight techniques. Wildlife Monographs 133.
MOWAT, G., D.C. HEARD, D.R. SEIP, K.G. POOLE, G.
STENHOUSE, AND D.W. PAETKAU. 2005. Grizzly Ursus
arctos and black bear U. americanus densities in the
interior mountains of North America. Wildlife Biology
11:31–48.
NAMS, V.O., G. MOWAT, AND M.A. PANIAN. 2006.
Determining the spatial scale for conservation purposes
— An example with grizzly bears. Biological Conservation 128:109–119.
NOYCE, K.V., D.L. GARSHELIS, AND P.L. COY. 2001.
Differential vulnerability of black bears to trap and
camera sampling and resulting biases in mark–recapture estimates. Ursus 12:211–226.
OBBARD, M., AND E.J. HOWE. 2008. Demography of black
bears in hunted and unhunted areas of the boreal forest
of Ontario. Journal of Wildlife Management 72:869–
880.
———, ———, AND C.J. KYLE. 2010. Empirical comparison of density estimators for large carnivores. Journal
of Applied Ecology 47:76–84.
OI, T. 2009. Anthropogenic mortality of Asiatic black
bears in two populations in northern Honshu, Japan.
Ursus 20:22–29.
OTIS, D.L., K.P. BURNHAM, G.C. WHITE, AND D.R.
ANDERSON. 1978. Statistical inference from capture
data on closed animal populations. Wildlife Monographs 62.
Ursus 22(1):47–64 (2011)
ANDEAN BEAR POPULATION ESTIMATES N Garshelis
PAISLEY, S. 2001. Andean bears and people in Apolobamba, Bolivia: Culture, conflict and conservation.
Dissertation, University of Kent, Durrell Institute of
Conservation and Ecology, Canterbury, UK.
———, AND D.L. GARSHELIS. 2006. Activity patterns and
time budgets of Andean bears (Tremarctos ornatus) in
the Apolobamba Range of Bolivia. Journal of Zoology
268:25–34.
PARMENTER, R.R., T.L. YATES, D.R. ANDERSON, K.P.
BURNHAM, J.L. DUNNUM, A.B. FRANKLIN, M.T. FRIGGENS, B.C. LUBOW, M. MILLER, G.S. OLSON, C.A.
PARMENTER, J. POLLARD, E. REXTAD, T.M. SHENK, T.R.
STANLEY, AND G.C. WHITE. 2003. Small-mammal
density estimation: A field comparison of grid-based
vs. web-based density estimators. Ecological Monographs 73:1–26.
PEACOCK, E., K. TITUS, D.L. GARSHELIS, M.M. PEACOCK,
AND M. KUC. 2011. Mark–recapture using tetracycline
and genetics reveal record-high bear density. Journal of
Wildlife Management 75:in press.
PERALVO, M.F., F. CUESTA, AND F.T. VAN MANEN. 2005.
Delineating priority habitat areas for the conservation
of Andean bears in northern Ecuador. Ursus 16:
222–233.
PETIT, E., AND N. VALIERE. 2006. Estimating population
size with noninvasive capture–mark–recapture data.
Conservation Biology 20:1062–1073.
PEYTON, B., E. YERENA, D.I. RUMIZ, J. JORGENSON, AND J.
OREJUELA. 1998. Status of wild Andean bears and
policies for their management. Ursus 10:87–100.
———. 1999. Spectacled bear conservation action plan.
Pages 157–168 in C. Servheen, S. Herrero, and B.
Peyton, editors. Bears: Status survey and conservation
action plan. IUCN/SSC Bear and Polar Bear Specialist
Groups, Gland, Switzerland and Cambridge, UK.
PROCTOR, M., B. MCLELLAN, J. BOULANGER, C. APPS, G.
STENHOUSE, D. PAETKAU, AND G. MOWAT. 2010.
Ecological investigations of grizzly bears in Canada
using DNA from hair, 1995–2005: A review of methods
and progress. Ursus 21:169–188.
RAYAN, D.M., AND S.W. MOHAMAD. 2009. The importance
of selectively logged forests for tiger Panthera tigris
conservation: A population density estimate in Peninsular Malaysia. Oryx 43:48–51.
RÍOS-UZEDA, B., H. GÓMEZ, AND R.B. WALLACE. 2005.
Habitat preferences of the Andean bear (Tremarctos
ornatus) in the Bolivian Andes. Journal of Zoology
268:271–278.
———, ———, AND ———. 2007. A preliminary density
estimate for Andean bear using camera-trapping
methods. Ursus 18:124–128.
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.
Ursus 22(1):47–64 (2011)
63
ROBSON, D.S., AND H.A. REGIER. 1964. Sample size in
Petersen mark–recapture experiments. Transactions of
the American Fisheries Society 93:215–226.
RODRÍGUEZ, D., F. CUESTA, I. GOLDSTEIN, L.G. NARANJO,
O.L. HERNÁNDEZ, AND A.E. BRACHO. 2003. Ecoregional
strategy for the conservation of the spectacled bear
(Tremarctos ornatus) in the northern Andes. World
Wildlife Fund Colombia, Fundación Wii, EcoCiencia,
Wildlife Conservation Society.
ROYLE, J.A., J.D. NICHOLS, K.U. KARANTH, AND A.M.
GOPALASWAMY. 2009. A hierarchical model for estimating density in camera-trap studies. Journal of Applied
Ecology 46:118–127.
RUIZ-GARCIA, M. 2003. Molecular population genetic
analysis of the spectacled bear (Tremarctos ornatus)
in the northern Andean area. Hereditas 138:81–93.
———, P. OROZCO-tERWENGEL, A. CASTELLANOS, AND L.
ARIAS. 2005. Microsatellite analysis of the spectacled
bear (Tremarctos ornatus) across its range distribution.
Genes and Genetic Systems 80:57–69.
SAARMA, U., S.Y.W. HO, O.G. PYBUS, M. KALJUSTE, I.L.
TUMANOV, I. KOJOLA, A.A. VOROBIEV, N.I. MARKOV,
A.P. SAVELJEV, H. VALDMANN, E.A. LYAPUNOVA, A.V.
ABRAMOV, P. MANNIL, M. KORSTEN, E. VULLA, S.V.
PAZETNOV, V.S. PAZETNOV, S.V. PUTCHKOVSKIY, AND
A.M. ROKOV. 2007. Mitogenetic structure of brown
bears (Ursus arctos L.) in northeastern Europe and a
new time frame for the formation of European brown
bear lineages. Molecular Ecology 16:401–413.
SÁNCHEZ-MERCADO, A., J. FERRER-PARIS, E. YERENA, S.
GARCÍA-RANGEL, AND K.M. RODRÍGUEZ-CLARK. 2008.
Factors affecting poaching risk to vulnerable
Andean bears Tremarctos ornatus in the Cordillera de
Mérida, Venezuela: Space, parks and people. Oryx
42:437–447.
SCHWARTZ, C.C., S.D. MILLER, AND M.A. HAROLDSON.
2003. Grizzly bear Ursus arctos. Pages 556–586 in G.A.
Feldhamer, B.C. Thompson, and J.A. Chapman,
editors. Wild mammals of North America: Biology,
management, and conservation. John Hopkins University Press, Baltimore, Maryland, USA.
SETTLAGE, K.E., F.T. VAN MANEN, J.D. CLARK, AND T.L.
KING. 2008. Challenges of DNA-based mark–recapture
studies of American black bears. Journal of Wildlife
Management 72:1035–1042.
SHARMA, R.K., Y. JHALA, Q. QURESHI, J. VATTAKAVEN, R.
GOPAL, AND K. NAYAK. 2010. Evaluating capture–
recapture population and density estimation of tigers in
a population with known parameters. Animal Conservation 13:94–103.
SHERLEY, R.B., T. BURGHARDT, P.J. BARHAM, N. CAMPBELL, AND I.C. CUTHILL. 2010. Spotting the difference:
Towards fully-automated population monitoring of
African penguins Spheniscus demersus. Endangered
Species Research 11:101–111.
64
ANDEAN BEAR POPULATION ESTIMATES N Garshelis
SILVER, S.C., L.E.T. OSTRO, L.K. MARSH, L. MAFFEI, A.J.
NOSS, M.J. KELLY, R.B. WALLACE, H. GÓMEZ, AND G.
AYALA. 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.
SPEED, C.W., M.G. MEEKAN, AND C.J.A. BRADSHAW. 2007.
Spot the match — Wildlife photo-identification using
information theory. Frontiers in Zoology 4, doi:
10.1186/1742-9994-4-2.
STETZ, J.B., K.C. KENDALL, AND C. SERVHEEN. 2010.
Evaluation of bear rub surveys to monitor grizzly bear
population trends. Journal of Wildlife Management
74:860–870.
TREDICK, C.A., AND M.R. VAUGHAN. 2009. DNA-based
population demographics of black bears in coastal
North Carolina and Virginia. Journal of Wildlife
Management 73:1031–1039.
TROLLE, M., AND M. KÉRY. 2003. Estimation of ocelot
density in the Pantanal using capture–recapture analysis of camera-trapping data. Journal of Mammalogy
84:607–614.
———, A.J. NOSS, J.L.P. CORDEIRO, AND L.F.B. OLIVEIRA.
2008. Brazilian tapir density in the Pantanal: A
comparison of systematic camera-trapping and linetransect surveys. Biotropica 40:211–217.
VITERI, M.P. 2007. Conservation genetics of Andean bears
(Tremarctos ornatus) in northeastern Ecuador: Molecular tools, genetic diversity and population size. Thesis,
University of Idaho, Moscow, Idaho, USA.
———, AND L.P. WAITS. 2009. Identifying polymorphic
microsatellite loci for Andean bear research. Ursus
20:102–108.
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.
WALLACE, R.B., H. GOMEZ, G. AYALA, AND F. ESPINOZA.
2003. Camera trapping for jaguar (Panthera onca) in
the Tuichi Valley, Bolivia. Mastozoologı́a Neotropica
10:133–139.
WHITE, G.C., D.R. ANDERSON, K.P. BURNHAM, AND D.L.
OTIS. 1982. Capture–recapture and removal methods
for sampling closed populations. Report LA-8787NERP. Los Alamos National Laboratory, Los Alamos,
New Mexico, USA.
WOODS, J.G., D. PAETKAU, D. LEWIS, B.N. MCLELLAN, M.
PROCTOR, AND C. STROBECK. 1999. Genetic tagging of
free-ranging black and brown bears. Wildlife Society
Bulletin 27:616–627.
WRIGHT, J.A., R.J. BARKER, M.R. SCHOFIELD, A.C.
FRANTZ, A.E. BYROM, AND D.M. GLEESON. 2009.
Incorporating genotype uncertainty into mark–recapture-type models for estimating abundance using DNA
samples. Biometrics 65:833–840.
YERENA, E., J. PADRÓN, R. VERA, Z. MARTÍNEZ, AND D.
BIGIO. 2003. Building consensus on biological corridors
in the Venezuelan Andes. Mountain Research and
Development 23:215–218.
YOGANAND, K., C.G. RICE, A.J.T. JOHNSINGH, AND J.
SEIDENSTICKER. 2006. Is the sloth bear in India secure?
A preliminary report on distribution, threats and
conservation requirements. Journal of the Bombay
Natural History Society 103:172–181.
ZHANG, B., M. LI, Z. ZHANG, B. GOOSSENS, L. ZHU, S.
ZHANG, J. HU, M.W. BRUFORD, AND F. WEI. 2007.
Genetic viability and population history of the giant
panda, putting an end to the ‘‘evolutionary dead end’’?
Molecular Biology and Evolution 24:1801–1810.
ZUG, B. 2009. Individual identification and habitat use of
Andean bears on private lands in the Ecuadorian
Andes. Thesis, University of Wisconsin, Madison,
Wisconsin, USA.
Received: 14 November 2010
Accepted: 14 February 2011
Associate Editor: R. Harris
Ursus 22(1):47–64 (2011)