Occurrence of Three Felids across a Network of Protected Areas in Thailand:
Prey, Intraguild, and Habitat Associations
Dusit Ngoprasert1, Antony J. Lynam2,11, Ronglarp Sukmasuang3, Naruemon Tantipisanuh1, Wanlop Chutipong1, Robert Steinmetz4,
Kate E. Jenks5, George A. Gale1, Lon I. Grassman Jr.6, Shumpei Kitamura7, JoGayle Howard8, Passanan Cutter9, Peter Cutter4,
Peter Leimgruber8, Nucharin Songsasen8, and David H. Reed10
1 Conservation Ecology Program, King Mongkut?s University of Technology Thonburi, Bangkhuntien, Bangkok, 10140, Thailand
2 Wildlife Conservation Society, Global Conservation Program, 2300 Southern Boulevard, Bronx, NY, 10540, U.S.A.
3 Department of Forest Biology, Faculty of Forestry, Kasetsart University, Bangkok, 10900, Thailand
4 World Wide Fund for Nature?Thailand, 2549/45 Paholyothin Road, Ladyao, Jatujak, Bangkok 10900, Thailand
5 Organismic and Evolutionary Biology, Wildlife and Fisheries Conservation, University of Massachusetts, Amherst, MA, 01003, U.S.A.
6 Feline Research Center, Caesar Kleberg Wildlife Research Institute, Texas A&M University-Kingsville, MSC 218, 700 University Blvd.,
Kingsville, TX, 78363, U.S.A.
7 The Museum of Nature and Human Activities, Yayoigaoka 6, Sanda, Hyogo, 669-1546, Japan
8 Center for Species Survival, Smithsonian Conservation Biology Institute, Front Royal, VA, 22630, U.S.A.
9 Conservation Biology Program, University of Minnesota, 187 McNeal Hall, 1985 Buford Avenue, St. Paul, MN, 55108, U.S.A.
10 Department of Biology, University of Louisville, Louisville, KY, 40292, U.S.A.
ABSTRACT
Clouded Leopard, Leopard, and Tiger are threatened felids in Southeast Asia, but little is known about the factors in?uencing their dis-
tributions. Using logistic regression, we assessed how habitat variables, prey detection patterns, and presence of intraguild predators
affect the occurrence of these felids across 13 protected areas within Thailand. Our analysis is based on data from 1108 camera-trap
locations (47,613 trap-nights). Clouded Leopard and Leopard are associated with habitat where Red Muntjac and Eurasian Wild Pig
were most likely to be present. Tiger are associated with habitat with a higher likelihood for the presence of Gaur, Eurasian Wild Pig,
and Sambar. Clouded Leopard and Tiger were both weakly associated with areas with mature evergreen forest. Besides availability of
prey, associations with potential competitors also appear to in?uence the distribution of these felids, although the strength of these
effects requires further investigation. Occurrence rates for Clouded Leopard were no different in protected areas with Leopard versus
without Leopards. Leopard had similar occurrence rates regardless of the presence of Tiger, but Leopards were less likely to be detected
at the same camera-trap points with the larger felid. Our results suggest that the two most commonly photographed prey species in the
study areas serve as key prey species, Eurasian Wild Pig for all three carnivores and Red Muntjac for Leopard and Clouded Leopard.
Abstract in Thai is available in the online version of this article.
Key words: carnivore conservation; interspeci?c interactions; Neofelis nebulosa; Panthera pardus; Panthera tigris; species distribution models.
UNDERSTANDING THE FACTORS DETERMINING THE ABUNDANCE AND
PATTERNS OF HABITAT USE by animal species of conservation
interest is integral to their management by, for example, inform-
ing habitat restoration projects (e.g., Xi et al. 2008, Trisurat et al.
2010), providing suitable areas for translocation or reintroduc-
tion (e.g., Fern?ndez et al. 2006, Klar et al. 2008, Cook et al.
2010), or examining potential connections among suitable habi-
tat patches (e.g., Muntifering et al. 2006, Trisurat et al. 2010). It
may also be possible to predict favored prey items of carnivores
by looking for associations between the presence or density of
the carnivore and the presence or density of one or more
potential prey species (Karanth et al. 2004).
It is urgent that we assess the factors impacting the distribu-
tion of top predators. Top predators have been in decline over the
last century due to a combination of habitat loss and fragmenta-
tion, loss of their prey base, and direct persecution (Karanth &
Chellam 2009, Lynam 2010, Oswell 2010). Top predators often
serve as keystone species and have been shown to be critical in
structuring ecosystems (e.g., Terborgh et al. 2001, Johnson et al.
2007, Beschta & Ripple 2009, Ritchie & Johnson 2009). In addi-
tion, top predators can serve as excellent umbrella species in con-
servation planning (Sergio et al. 2008) because of their large home
range sizes, overlap with other species of conservation concern,
and sensitivity to forest degradation (e.g., Morrison et al. 2007).
We use data collected since 1997 from 13 different protected
areas to assess potential factors affecting the distribution of the
Tiger (Panthera tigris), Leopard (Panthera pardus), and Clouded
Received 18 August 2011; revision accepted 26 February 2012.
11Corresponding author; e-mail: tlynam@wcs.org
? 2012 The Author(s) 1
Journal compilation ? 2012 by The Association for Tropical Biology and Conservation
BIOTROPICA 0(0): 1?8 2012 10.1111/j.1744-7429.2012.00878.x
Leopard (Neofelis nebulosa) across Thailand. These are the three
largest felids in Southeast Asia, weighing on average approxi-
mately 155 kg and 110 kg, 50 kg and 35 kg, and 18 kg and
13 kg for males and females of each species, respectively (Silva &
Downing 1995, Francis 2001, Grassman et al. 2005). Despite the
difference in size between the three species, we expect substantial
overlap in the range of prey sizes killed by Tiger and Leopard
and between Leopard and Clouded Leopard (e.g., Seidensticker
1976, Johnsingh 1992, Karanth & Sunquist 1995). Thus, competi-
tion among at least some pairs of the three felids is possible and
could be an important factor affecting the overall distribution or
occupancy of the three carnivores.
Our goals were to: (1) identify environmental factors associ-
ated with the current distribution of the three largest felids across
selected sites in Thailand; (2) identify the prey species with which
these felids were most closely associated; and (3) examine the
potential for interactions among the three felids.
METHODS
Camera-trap data were collected from 13 protected areas within
Thailand (Fig. S1; Tables 1 and S1). Surveys were conducted
between April 1997 and July 2010. Camera traps were set at
1108 locations across the 13 protected areas. The mean number
of trap-nights per camera was 45.6. Elevation of camera locations
ranged from 49 to 1351 m (mean = 469 m). All sites were in
protected areas, as all remaining suitable habitat for large carni-
vores in Thailand lies exclusively inside them. Hunting, a primary
threat to large carnivores and their ungulate prey, is prohibited in
all protected areas, although enforcement against poaching is
highly variable among locations (Albers & Grinspoon 1997,
Steinmetz et al. 2010a). The number of photographs for each
species and the number of camera-trap locations where the spe-
cies was detected are listed in Tables 1 and S1.
Objectives and methods varied among studies as did cam-
era spacing (Table 1). Cameras were placed in areas with a
high probability of being used by target animals, which were
typically large carnivores (e.g., trails, near water sources). We
did not measure microhabitat variables along trails near the
camera locations, although it is possible that such variables can
in?uence detection rates of predators and their prey (Harmsen
et al. 2010). Cameras were operational 24 h per d. Camera-trap
locations were not baited and trail-cameras were secured to
trees. Investigators checked their cameras every 7?14 d. We
did not include data for which functioning cameras were left
in the forest for fewer than 6 d. Greater than 98 percent of
animal photos were submitted to an external expert and com-
mittee of experienced ?eld researchers for species validation.
DATA ANALYSIS.?We used multiple logistic regression to examine
associations between various abiotic and biotic factors (the inde-
pendent variables) and detection/nondetection of the selected
species (the dependent variable). We then employed an informa-
tion-theoretic approach (AICc) to choose the model with the
highest likelihood value and to weight (through model averaging)
the relative importance of the different predictor variables (Burn-
ham & Anderson 2002). Model averaging is preferred when, as
in this case, multiple models were often reasonable ?ts to the
data. The abiotic factors included distance to forest edge (Edge),
elevation (Elev), habitat type (Habitat), distance to nearest
unsealed (open to the public) road or village (Human), slope
(Slope), average rainfall during the driest month (Rain), and dis-
tance to nearest river or stream (Water). All independent variables
were log10 transformed with the exception of Rain and Slope. All
habitat variables had pairwise correlations coef?cients <0.25 after
transformation.
Habitat was converted to a continuous variable by multiply-
ing mean tree density by mean basal area for eight habitat types
that were encountered in camera-trap study areas: grassland/
savanna, secondary forest, bamboo forest, mixed deciduous for-
est, dry dipterocarp forest, hill evergreen forest, moist evergreen
forest, and dry evergreen forest. Input values were obtained
from, and averaged across, numerous botanical studies conducted
in the same protected areas as our camera trapping (for details of
sources, see Supporting Information).
TABLE 1. Data for each of the 13 study sites including number of locations with a
trail camera (Camera Locations), total number of trap-nights for all
locations (Trap-Nights), average distance between camera locations (D), and
number of photos for Leopard (L), Clouded Leopard (CL) and Tiger
(Tiger).
Study site
Camera
locations
Trap-
nights D L CL Tiger
Bang Lang National Park
(BL)
14 463 1.2 0 4 17
Hala Bala Wildlife Sanctuary
(HB)
29 10,879 2.8 2 6 9
Huai Kha Kaeng Wildlife
Sanctuary (HKK)
173 1280 1.9 87 3 38
Kaeng Krachan National
Park (KK)
94 7699 0.6 227 0 20
Khao Ang Rue Nai Wildlife
Sanctuary (KARN)
252 5710 0.5 0 11 0
Khao Sok National Park
(KOS)
1 9 NA 0 0 0
Khao Yai National Park
(KY)
358 12,323 0.9 0 47 10
Klongsaeng Wildlife
Sanctuary (KLS)
6 113 2.1 0 0 0
Kuiburi National Park (KB) 48 2511 0.7 23 0 11
Phu Khieo Wildlife
Sanctuary (PK)
51 1212 0.6 0 6 6
Ta Phraya National Park
(TAP)
22 677 2.6 2 11 0
Thap Lan National Park
(THP)
6 187 1.8 0 0 0
ThungYai Naresuan Wildlife
Sanctuary (TY)
53 4550 0.7 1 1 19
2 Ngoprasert et al.
Abiotic factors were generated using ArcMap v.9.3 (ESRI,
Redlands, California, U.S.A.). Shape ?les of habitat type, road, vil-
lage, and stream were provided by the Thailand Department of
National Parks, Wildlife, and Plant Conservation. Elevation data
were derived from the ASTER Global Digital Elevation Model
(http://www.jspacesystems.or.jp/ersdac/GDEM/E/). Slope data
were calculated from the elevation data. Average rainfall during
the driest month was collected from rain gage stations through-
out Thailand between 1997 and 2008 and was provided by the
Thailand Meteorological Department. Rainfall for each speci?c
camera-trap location was interpolated using kriging (Kitanidis
1997).
Success of the logistic regression models was judged using
receiver operating characteristic (ROC) plots (Fielding & Bell
1997). The area under the ROC curve (AUC) is a measure of
overall ?t and varies from 0.5 (for a model that does no better
than random) to 1.0 for a perfect ?t. Although some have been
critical of using AUC when comparing the ?t of models (e.g.,
Lobo et al. 2008), AUC is thought to be especially appropriate
when comparing different models using the same data (Fielding &
Bell 1997). We do not, however, rely solely on AUC as a measure
of overall ?t. We also rely on the model averaged AICc values, the
traditional signi?cance levels from the logistic regression, and sen-
sitivity and speci?city in the ROC curve (as suggested by Lobo
et al. 2008) where appropriate.
We used logistic regression, instead of occupancy model-
ing, as our primary form of data analysis because estimates of
occupancy rates may be heavily biased by the nonrandom sam-
pling of only parts of protected areas and because of the likely
violation of closure over such a long sampling period. The
weakness of logistic regression is that absences may be due to
a lack of detection and not a true absence (MacKenzie et al.
2002). In addition to the national model using all 1108 cameras
as independent data points, a ?fty and over model was con-
structed using only the 220 camera-trap locations where the
camera spent 
 50 continuous trap-nights in the forest
(mean = 123.1). Fifty continuous trap-nights is approximately
the amount of time we calculate, from this data, to be at least
95 percent certain of capturing even the least common target
species should it occur at that location. Thus, this model
should greatly ameliorate problems associated with false
absences.
In addition to the three models described above, we built
models for the three individual protected areas with the largest
data sets. The following individual protected areas were modeled,
where N is the number of camera locations: HKK (N = 173),
KK (N = 94), and KY (N = 358). These are subsets of the
national model.
The raw proportion of camera-trap locations that detected a
species gives an index of its occurrence at a site. To quantify
associations between the three felids, we performed correlation
analysis on the occurrence of Clouded Leopards, Leopards, and
Tigers. Furthermore, to speci?cally test whether the presence of
Leopards in a protected area potentially impacted the occurrence
of Clouded Leopards, we performed a t-test comparing mean
occurrence of Clouded leopards in protected areas with and with-
out Leopards.
MODEL BUILDING PROCEDURES.?Logistic regression in combina-
tion with AICc was used to determine the relative importance of
the abiotic factors: Edge, Elev, Habitat, Human, Rain, Slope, and
Water for each of the three carnivores. After building models
using only abiotic factors, a global model was constructed using
abiotic and biotic factors. The additional biotic factors were Prey
(de?ned below) and a potential competitor. For Clouded Leop-
ards, detection or nondetection of Leopards was entered into the
model as a binary variable, whereas for Leopards, the detection
or nondetection of tigers was included in the model as a binary
variable.
Logistic regression was used to build models for potential
prey species using the seven abiotic factors. The output produced
from the logistic regression models for prey was solved to pro-
duce continuous-probability values ranging from 0.0 to 1.0. The
higher values indicate a higher probability of occurrence. The
probabilities for each prey species at each camera location were
then entered as independent variables into a logistic regression to
see which, if any, of the potential prey species were associated
with detection of the predator. The species tested for associations
were based on numerous published accounts of diet. Potential
prey for Clouded Leopard and Leopard were Hog Badger (Arct-
onyx collaris), Red Muntjac (Muntiacus muntjac), Sambar (Rusa unicol-
or), and Eurasian Wild Pig (Sus scrofa). For Tigers, Hog Badger
was replaced with Gaur (Bos gaurus). An information-theoretic
approach, corrected for small sample sizes (AICc), was then
employed to choose which species to keep in a ?nal model. To
avoid having models for prey that contained only one habitat var-
iable (and therefore completely collinear with it), we kept the prey
model with the lowest AICc value that contained at least two
habitat variables. This ?nal linear equation was speci?c to each
carnivore and treated as an explanatory variable (Prey) in further
model building.
Correlation coef?cients between Prey and the abiotic variables
were generally small (<0.40). The overall accuracy of the models was
measured using AUC and the relative importance of each factor was
determined using the information-theoretic paradigm and model
averaging (Burnham & Anderson 2002). Analyses were carried out
using JMP software v.8 (SAS Institute, Cary, North Carolina, U.S.A.).
RESULTS
The national model consistently performed better than the ?fty and
over model (Table S2). For all three species the predictor variables
chosen were nearly identical between the two models. Resampling
statistics (Table S2), which includes a measure of model uncer-
tainty, also shows the national model performing best for all three
predator species. This is consistent with Tyre et al. (2003), who
suggested that when error rates are <50 percent, greater ef?-
ciency is gained by adding more sites rather than more sampling
periods; and that the logistic regression will converge on the true
species?habitat relationships as samples size becomes very large.
Large Felid Distributions 3
All models including both abiotic and biotic factors, at the
national level and the level of individual protected areas, produced
results that were signi?cantly different from random
(AUC > 0.50; Tables 2and 3). The overall logistic regression
models were also highly signi?cant using conventional signi?cance
testing (Chi-square test, P < 0.001). The prey models were not a
direct goal of this study, but rather a means of building a better
predator model. Thus, the prey models and their details have been
placed in Table S3. The prey models ranged from poor to excel-
lent, but most were considered adequate (AUC = 0.70?0.80).
PREY ASSOCIATIONS.?The associations between felids and poten-
tial prey were highly consistent. Using prey alone produced poor
to good models, with most models being considered adequate
(AUC = 0.70?0.80). Clouded Leopards were most closely associ-
ated with places where Red Muntjac and Eurasian Wild Pig
occurred (Table 2). Leopards also were most closely associated
with places with Eurasian Wild Pig and Red Muntjac, although
there was also evidence for weaker correlations with Hog Badger
and Sambar (Table 2). Presence of tigers was associated with
Gaur, Eurasian Wild Pig, and Sambar (Table 2).
HABITAT ASSOCIATIONS.?The associations between felids and abi-
otic factors produced adequate models (AUC = 0.70?0.80), but
all improved to good (AUC = 0.80?0.90) when biotic factors
were added (Tables 3 and S4). For Clouded Leopard and Tiger,
the abiotic factors identi?ed as important were similar whether
biotic factors were included or not (Table S4). It should be
noted, however, that areas with greater than average rainfall were
positively associated with Clouded Leopard and Tiger presence in
the abiotic models only, whereas Habitat was considered moder-
ately important when biotic factors were included. Leopards
showed a preference for low elevations and areas far from roads
and villages in the abiotic only models. These models, however,
generally had poor overall ?t (Table S4).
The global model (before parameter selection) for each pred-
ator contained the seven habitat variables and Prey. The potential
competitor was left out of the national models because combina-
tions of potential competitors do not co-occur in all protected
areas, potentially biasing the analysis. The addition of the selected
habitat variables usually only marginally improved model ?t over
the models containing Prey only.
CLOUDED LEOPARD.?Clouded Leopard detection was most con-
sistently associated with higher elevations and a high likelihood of
prey presence (Table 3). Cameras that detected Clouded Leopards
were on average 144 m higher in elevation than cameras that did
not detect Clouded Leopards, although Clouded Leopards were
photographed anywhere from 90 to 1253 m in elevation.
LEOPARD.?Unlike Clouded Leopards, where the models were
somewhat idiosyncratic, the Leopard models were consistent. The
national model and the two models for individual protected areas
TABLE 2. Prey species and their relative importance values for Neofelis nebulosa,
Panthera pardus, and Panthera tigris. Model denotes whether it was the
model for all of Thailand (National) or for a particular protected area. The
importance values were determined using AICc and model averaging.
Importance values were scaled so that the most important prey species had a
value of 1.00, and all other prey species? importance values are given relative
to it. AUC represents a measure of model success relative to a random
model (AUC = 0.50).
Model Prey species included AUC
Neofelis nebulosa
National Red Muntjac (1.00), Eurasian Wild Pig (1.00) 72.8
HKK Red Muntjac (1.00), Eurasian Wild Pig (0.97) 62.3
KY Hog Badger (1.00), Red Muntjac (0.90) 70.5
Panthera pardus
National Hog Badger (1.00), Red Muntjac(0.73), Eurasian Wild Pig
(0.42)
73.2
HKK Sambar (1.00), Eurasian Wild Pig (0.95), Red Muntjac
(0.71)
60.8
KK Eurasian Wild Pig (1.00), Red Muntjac (0.56) 81.5
Panthera tigris
National Gaur (1.0), Eurasian Wild Pig (1.0), Sambar (0.36) 79.4
HKK Sambar (1.00), Gaur (0.96), Eurasian Wild Pig (0.78) 60.4
KK Eurasian Wild Pig (1.00), Sambar (0.63), Gaur (0.35) 62.6
KY Gaur (1.00), Eurasian Wild Pig (0.64), Sambar (0.28) 79.0
TABLE 3. Abiotic and biotic variables and their relative importance values (in
parentheses) for Clouded Leopard (Neofelis nebulosa), Leopard (Panthera
pardus), and Tiger (Panthera tigris). Model denotes whether it was the model
for all of Thailand (National) or for a particular protected area. The
importance values were determined using AICc and model averaging.
Importance values were scaled so that the most important variables had a
value of 1.00 and all other variables were relative to the most important.
AUC represents a measure of model accuracy relative to a random model
(AUC = 0.50). No Leopards existed historically at KY. Thus, no
potential competitor was included in that model.
Model Factors included in the model AUC
Neofelis nebulosa
National Elev (1.00), Prey (0.85), Habitat (0.61) 80.2
HKK Habitat (1.00), Prey (0.47) 75.5
KY Elev (1.00), Edge (0.77), Human (0.62), Rain
(0.45)
80.4
Panthera pardus
National Prey (1.0) 85.0
HKK Prey (1.00), Tiger (0.64) 62.1
KK Prey (1.00), Tiger (0.78) 84.3
Panthera tigris
National
model
Prey (1.00), Habitat (0.92) 79.9
HKK Water (1.00), Prey (0.59), Habitat (0.48) 63.3
KK Rain (1.00), Prey (0.64) 72.6
KY Human (1.00), Elev (0.91) 83.2
4 Ngoprasert et al.
all identi?ed presence of potential prey as the primary correlate
of Leopard detection (Table 3). Presence of tigers was negatively
correlated with Leopard presence in both individual protected
areas.
TIGER.?For tigers, Prey was the most important factor followed
closely by Habitat (Table 3). Habitat showed a similar pattern in
Tigers and Clouded Leopards; an association with mature ever-
green forests at the national level, but a correlation with grasslands
and more open areas within two individual protected areas (HKK
and KK). Protected area complexes with potential for sustaining
tiger populations in the long-term in Thailand are all >2000 km2
(Lynam 2010). In these larger reserves tigers appear to prefer core
areas where there is mature forest. In protected areas where there
is a mosaic of closed evergreen forest and semi-open deciduous
dipterocarp forest, tigers occur across this mosaic.
ASSOCIATIONS AMONG PREDATORS.?Nonparametric Spearman rank
correlations of occurrence across 10 protected areas between
Clouded Leopard and Leopard occurrence rates were 0.54
(NS), between Leopards and tigers of 0.12 (NS), and between
Clouded Leopards and tigers of 0.03 (NS). The negative corre-
lation between occurrence of Clouded Leopards and Leopards
suggests a possible negative interaction between the two species,
but the analysis is inconclusive, most likely due to small sample
size and the consequent lack of statistical power. Occurrence
rates of Clouded Leopards were not signi?cantly higher in pro-
tected areas without leopards (or at least nondetected leopards)
(mean 12.5%) than in those with leopards (9.0%) (one-tailed
t-test: tdf = 8, 0.05 = 0.42, NS).
DISCUSSION
The major ?ndings for this investigation are the following. (1)
For the three largest felids in Thailand, the presence of certain
prey species was generally the most important factor determining
presence of the predator. Clouded Leopard and Leopard presence
was associated with the presence of Red Muntjac and Eurasian
Wild Pig. For Tiger, Gaur, Eurasian Wild Pig, and Sambar
appeared to be the most important prey items. (2) Tigers do not
exclude Leopards from habitats, but Leopards do appear to avoid
tigers within their overlapping home ranges. (3) Some habitat fea-
tures were associated with high probabilities of presence for
Clouded Leopards and Tigers. Many of the habitat factors may
be related to anthropogenic disturbances and threats. The evi-
dence for this is particularly strong in Khao Yai. We elaborate on
each of these points below.
PREY.?Presence of all three felids was strongly associated with
certain prey species. Predator-prey relationships in this study are
quite robust as they are consistent among and within protected
areas, despite different models being used in each case to create
the probability map for the prey species and despite favored prey
items often having a preference for different habitat types. Care
must be taken though, because the models do not show prefer-
ence, only what potential prey species carnivores are associated
with and the strength of the association. The associations found
in the models do not replace dietary studies for speci?c locations.
Nevertheless, our results generally agree with known favored prey
determined from diet studies.
Clouded Leopard presence was consistently associated with
the presence of Red Muntjac and Eurasian Wild Pig. Little is
known of the diet of Clouded Leopards in Thailand. The only
study on the topic lists Brush-tailed Porcupine (Atherurus macrou-
rus), Indochinese Ground Squirrel (Menetes berdmorei), Red Munt-
jac, Slow Loris (Nycticebus coucang), Hog Deer (Axis porcinus), and
Sunda Pangolin (Manis javanica) as prey species (Grassman et al.
2005). Rabinowitz et al. (1987) list Sambar, Red Muntjac, Mouse
Deer (Tragulus javanicus), Bearded Pig (Sus barbatus), Common
Palm Civet (Paradoxurus hermaphroditus), various primates, and por-
cupine as prey in Borneo. Juvenile and adult Red Muntjac fall
within the size prey that would be expected to be preferred by
Clouded Leopards (i.e., approximately their own body weight),
and Choudhury (1993) witnessed Clouded Leopards stalking Eur-
asian Wild Pig. Clouded leopards are the most arboreal of the
three predators we examined, and arboreal prey, such as primates
likely constitute a greater proportion of their diet than tigers or
leopards. Our analysis only considered terrestrial prey species and
hence is less robust for clouded leopard than the other species;
this may have contributed to the prey covariate falling second to
an environmental covariate (elevation; see Table 3) in the national
model for this species.
Detection of Leopard was consistently associated with the
presence of Eurasian Wild Pig and Red Muntjac. Leopards in
both Africa and Asia exhibit considerable ?exibility in their diet,
but appear to prefer prey between 20 and 75 kg (e.g., Karanth &
Sunquist 1995, Hayward et al. 2006), with juvenile and adult Eur-
asian Wild Pig ?tting into this range. Rabinowitz (1989) suggested
that leopards ate mostly Red Muntjac in Huai Kha Kaeng
because Sambar, Gaur, and Banteng (Bos javanicus) are too large
and Eurasian Wild Pig densities too low. Grassman (1999), using
scat analysis, found Eurasian Wild Pig, Sambar, Red Muntjac,
and Hog Badger contributed the most, respectively, in biomass to
Leopard diets in Kaeng Krachan. In southern India, with habitat
similar to much of Thailand, Karanth and Sunquist (1995) found
that leopards ate Eurasian Wild Pig disproportionally little com-
pared with other prey of a similar size, whereas Wegge et al.
(2009) found that Eurasian Wild Pig was killed more often than
expected, given their relative abundance in Nepal.
Consistent with the ?ndings of studies across the region (e.g.,
Karanth et al. 2004), Tiger presence was consistently associated
with the presence of Gaur, Eurasian Wild Pig, and Sambar. Pet-
dee (2000) and Prommakul (2003) both found Sambar, Gaur,
and Banteng to be the most important prey items of Tiger in
different parts of the Western Forest Complex of Thailand
(which includes HKK and TY). We did not include Banteng in
the models because of its highly restricted distribution and occur-
rence at low densities. While Gaur and Sambar were both
strongly associated with Tiger presence, the species with the sec-
ond strongest association was Eurasian Wild Pig. Eurasian Wild
Large Felid Distributions 5
Pig does not feature prominently in tigers diets in the Western
Forest Complex (Petdee 2000, Prommakul 2003), but is known
to be important elsewhere in Thailand (Steinmetz et al. 2010a) as
well as in India, Nepal, the Russian Far East, and Sumatra
(e.g., Miquelle et al. 1996, O?Brien et al. 2003, Wegge et al. 2009).
Our results suggest that the two most commonly photo-
graphed potential prey species in the study areas, Red Muntjac
and especially Eurasian Wild Pig, serve as the major prey species
for the largest felids in Thailand. Eurasian Wild Pigs are heavily
hunted in at least some parts of the region (Duckworth et al.
1999), but they also have a high reproductive capacity and recover
quickly if protected from poaching (e.g., Steinmetz et al. 2010b).
PREDATOR ASSOCIATIONS.?Intraguild interactions play a prominent
role in shaping ecological communities and to limit population sizes
of vertebrate predators (e.g., Donadio & Buskirk 2006, Hunter &
Caro 2008, Sergio & Hiraldo 2008, Ritchie & Johnson 2009). Of
the two Clouded Leopard models at the level of individual pro-
tected areas, only Huai Kha Kaeng has Leopards. The model for
HKK failed to ?nd a relationship between detection of Leopards
and Clouded Leopards. Thus, considering this regression model
there is no evidence for Clouded Leopards avoiding Leopards.
However, considering a larger scale, we observed a negative, but
nonsigni?cant relationship between Clouded Leopard and Leopard
occurrence across protected areas. We suggest interactions between
Leopards and Clouded Leopards require further investigation.
In our study, there is no relationship between the occurrence
of Leopards and Tigers at the site level. Logistic models for two
protected areas, however, suggest that the detection of Leopards
at these sites is negatively associated with the detection of Tigers
at the scale of individual camera locations. While it has been
shown that Leopards sometimes avoid tigers (Seidensticker et al.
1990, Odden et al. 2010), this behavior appears to depend
strongly on the availability of both large and medium-sized prey,
which mitigates interference competition between Tigers and
Leopards (Karanth & Sunquist 1995). Avoidance also depends
on habitat structure, particularly the presence of trees, which
Leopards might use to escape from Tigers (Karanth & Nichols
1998). This agrees with our experience that Tigers and Leopards
sometimes co-occur within protected areas in Thailand, frequently
have overlapping home ranges, and even use the same roads and
trails (Rabinowitz 1989). For example, the highest known densi-
ties of both Leopard and Tiger within Thailand are in the same
area (Khao Nang Rum) within Huai Kha Khaeng (Simcharoen
et al. 2008). It is notable that this site also has high densities of
both medium (Red Muntjac, Eurasian Wild Pig) and large prey
(Gaur, Banteng). The models suggest ?ne scale spatial avoidance,
not displacement, of Leopards by Tigers from prime habitats.
The different outcomes for the logistic models and the
occurrence analysis have biological explanations. Considered
together, these models suggest ?ne scale spatial avoidance by
leopards toward tigers within overlapping home ranges. Leopards
are not displaced by Tigers from prime habitats, but rather avoid
them at ?ne spatial (and perhaps temporal) scales. This interpre-
tation corresponds to numerous ?eld studies that have demon-
strated that subordinate (smaller) carnivore species often respond
to potential interference competition and intraguild predation
through active, ?ne scale spatial avoidance while inhabiting home
ranges of larger species (Neale & Sacks 2001, Scognamillo et al.
2003, Steinmetz et al. 2011).
Another consideration is the effect of heavy hunting to
which most protected areas have been subjected. All the pro-
tected areas we examined were subject to hunting to varying
degrees. We could not quantify hunting pressure to model its
effect on large cat distributions. Hunting pressure, however, did
not appear to affect our results; large cats in protected areas that
were relatively heavily hunted (Kaeng Krachan) were in?uenced
by similar variables as cats in lightly hunted sites (Huai Kha Kha-
eng; Table 3). For example, leopard distributions in Huai Kha
Khaeng and Kaeng Krachan were functions of the same set of
abiotic and biotic variables (prey, and tiger distribution), and
effect sizes of these variables were also similar (Table 3). Inter-
speci?c aggression is positively correlated with food stress in ani-
mals (e.g., Polis et al. 1989, Palomares & Caro 1999). Most
populations of large cats in Thailand can be considered food-
stressed due to depletion of their prey base (Ngoprasert et al.
2007, Steinmetz et al. 2010b), and the size differential and mor-
phology of tigers and leopards lends itself potentially to high lev-
els of interference competition and interspeci?c killing (Donadio
& Buskirk 2006). That Leopards did not appear to have their
large-scale occurrence rates lowered by the presence of tigers, yet
avoided them at smaller scales, maybe because Leopards and
Tigers are forced to co-occur in the limited set of areas that sup-
port their preferred larger ungulate prey species.
ANTHROPOGENIC THREATS.?Khao Yai, which is surrounded by
human?converted habitat, is a site with particularly high levels of
disturbance (Albers & Grinspoon 1997). For Tigers, the distance to
the nearest road or village and elevation both appear in the KY
model but not the national model or any of the other models for
individual protected areas. This is noteworthy as Tigers were pres-
ent during the ?rst surveys reported in this study, but apparently
subsequently disappeared from Khao Yai (Lynam et al. 2006, Jenks
et al. 2011). For Clouded Leopard, elevation, distance to forest
edge, and distance to nearest road or village all appear in the model
for Khao Yai. Only elevation, however, appears in any other model.
Increased elevation and distance to the forest edge likely also
re?ect, at least in part, the reduced in?uence of poachers and
human encroachment rather than habitat preferences. The central
part of KY is mountainous and, due to the presence of the park
headquarters and concentration of management activities, thought
to be far less susceptible to poaching than other areas that are both
easier to traverse and more accessible from roads and villages. It
appears that the modeling process and the monitoring of the
national park via trail-cameras have accurately captured the poach-
ing pressure and ecological deterioration of large predator habitat
during the course of a decade (Lynam et al. 2006, Jenks et al. 2011).
CONSERVATION IMPLICATIONS AND CONCLUSIONS.?The current
study found that certain potential prey species were consistently
6 Ngoprasert et al.
associated with the presence of the three largest felids in Thai-
land. Red Muntjac and Eurasian Wild Pig appear to be major
prey items for Clouded Leopards and Leopards; whereas tigers
were most closely associated with Gaur and Eurasian Wild Pig.
Thus, recovery of populations or reintroduction success will likely
be heavily dependent on conserving suf?cient densities of ungu-
late prey. We also found evidence for negative associations among
the three felids. The ecological relationships among the three pre-
dators requires further study and should be carefully considered
when evaluating conservation approaches aimed primarily at
increasing densities of one of the predators (Lynam 2010, Wal-
ston et al. 2010, Wikramanayake et al. 2011). Finally, this study
demonstrates the value of collaboration among researchers in
compiling larger camera-trap data sets that ?lter out some of the
site-speci?c effects and allow the study of broader of ecological
dependence and other conservation factors.
ACKNOWLEDGMENTS
This study is dedicated to the memory of JoGayle Howard and
David Reed. We wish to thank the Clouded Leopard Project,
TRF/BIOTEC Special Program for Biodiversity Research and
Training Thailand, Kasetsart University, and the Smithsonian
Conservation Biology Institute for providing funding for the ini-
tial conference that brought all the authors together to discuss
collaboration. S. Kitamura was funded by the Mahidol University
Government Research Grant, the National Center for Genetic
Engineering and Biotechnology, the Hornbill Research Founda-
tion, and a JSPS Research Fellowship. We thank W. Duckworth
and A. Wilting for their kind assistance in identifying the carni-
vores in our data set. We thank the Wildlife Conservation Society
Thailand, World Wide Fund for Nature?Thailand, and WildAid
Thailand for allowing participants to use data collected under
their auspices. We also thank N. Bhumpakphan, R. Phoonjampa,
S. Tanasarnpaiboon. and M. Pliwsungnoen for their assistance.
SUPPORTING INFORMATION
Additional Supporting Information may be found in the online
version of this article:
TABLE S1. Number of photographs of each species and number of
camera-trap locations.
TABLE S2. Mean areas under the curve (AUC) for the receiver oper-
ating characteristic (ROC) and their 95% con?dence intervals for the three
felid species for each of the three models.
TABLE S3. The species included as prey in the predator models, the
habitat variables included in each of the prey models, and the accuracy of
the logistic regression model as measured by AUC.
TABLE S4. Abiotic variables and their relative importance values for
Clouded Leopard (Neofelis nebulosa), Leopard (Panthera pardus),
and Tiger (Panthera tigris).
FIGURE S1. Map of Thailand with each of the 13 protected
areas where camera-trapping was conducted.
Please note: Wiley-Blackwell are not responsible for the con-
tent or functionality of any supporting materials supplied by the
authors. Any queries (other than missing material) should be
directed to the corresponding author for the article.
LITERATURE CITED
ALBERS, H. J., AND E. GRINSPOON. 1997. A comparison of the enforcement of
access restrictions between Xishuangbanna Nature Reserve (China)
and KhaoYai National Park (Thailand). Environ. Conserv. 24: 351?
362.
BESCHTA, R. L., AND W. J. RIPPLE. 2009. Large predators and trophic cascades
in terrestrial ecosystems of the western United States. Biol.Conserv.
142: 2401?2414.
BURNHAM, K. P., AND D. R. ANDERSON. 2002. Model selection and multimodel
inference: A practical information-theoretic approach (2nd Edition).
Springer, Berlin.
CHOUDHURY, A. 1993. The clouded leopard in Assam. Oryx 27: 51?53.
COOK, C. N., D. G. MORGAN, AND D. J. MARSHALL. 2010. Reevaluating suitable
habitat for reintroductions: Lessons learnt from the eastern barred
bandicoot recovery program. Anim. Conserv. 13: 184?195.
DONADIO, E., AND S. W. BUSKIRK. 2006. Diet, morphology and interspeci?c
killing in carnivora. Am. Nat. 167: 524?536.
DUCKWORTH, J. W., R. E. SALTER, AND K. KHOUNBOLINE. 1999. Wildlife in Lao
PDR: 1999 Status Report. IUCN-The World Conservation Union/
Wildlife Conservation Society/Centre for Protected Areas and
Watershed Management, Vientiene.
FERN?NDEZ, N., M. DELIBES, AND F. PALOMARES. 2006. Landscape evaluation
in conservation: Molecular sampling and habitat modelling for the Ibe-
rian lynx. Ecol. Appl. 16: 1037?1049.
FIELDING, A. H., AND J. F. BELL. 1997. A review of methods for the assess-
ment of prediction errors in conservation presence/absence models.
Env. Conserv. 24: 38?49.
FRANCIS, C. M. 2001. Mammals of Thailand & South-East Asia. Asia Books
Company, Bangkok, Thailand.
GRASSMAN Jr., L. I. 1999. Ecology and behavior of the Indochinese leopard in
Kaeng Krachan National Park, Thailand. Nat. Hist. Bull. Siam Soc.
47: 77?93.
GRASSMAN Jr., L. I., M. E. TEWES, N. J. SILVY, AND K. KREETIYUTNONT. 2005.
Ecology of three sympatric felids in a mixed evergreen forest in
north-central Thailand. J. Mammal. 86: 29?38.
HARMSEN, B. J., R. J. FOSTER, S. SILVER, L. OSTRO, AND C. P. DONCASTER. 2010.
Differential use of trails by forest mammals and the implications for
camera-trap studies: A case study from Belize. Biotropica 42: 126?133.
HAYWARD, M. W., P. HENSCHEL, J. O?BRIEN, M. HOFMEYR, G. BALME, AND G.
I. H. KERLEY. 2006. Prey preferences of the leopard (Panthera pardus).
J. Zool. 270: 298?313.
HUNTER, J., AND T. CARO. 2008. Interspeci?c competition and predation in
American carnivore families. Ethol. Ecol. Evol. 20: 295?324.
JENKS, K. E., P. CHANTEAP, K. DAMRONGCHAINARONG, P. CUTTER, P. CUT-
TER, T. REDFORD, A. J. LYNAM, J.-G. HOWARD, AND P. LEIMGRUBER.
2011. Using relative abundance indices from camera-trapping to test
wildlife conservation hypotheses?an example from Khao Yai
National Park, Thailand. Trop. Conserv. Sci. 4: 113?131.
JOHNSINGH, A. J. T. 1992. Prey selection in three large sympatric carnivores in
Bandipur. Mammalia 56: 517?526.
JOHNSON, C. N., J. L. ISAAC, AND D. O. FISHER. 2007. Rarity of a top predator
triggers continent-wide collapse of mammal prey: Dingoes and marsu-
pials in Australia. Proc. R. Soc. B 274: 341?346.
KARANTH, K. U., AND R. CHELLAM. 2009. Carnivore conservation at the cross-
roads. Oryx 43: 1?2.
KARANTH, K. U., AND J. D. NICHOLS. 1998. Estimation of tiger densities in
India using photographic captures and recaptures. Ecology 79: 2852?
2862.
Large Felid Distributions 7
KARANTH, K. U., J. D. NICHOLS, N. S. KUMA, W. A. LINK, AND J. E. HINES.
2004. Tigers and their prey: Predicting carnivore densities from prey
abundance. Proc. Nat. Acad. Sci. U.S.A. 101: 4854?4858.
KARANTH, K. U., AND M. E. SUNQUIST. 1995. Prey selection by tiger, leopard,
and dole in tropical forests. J. Anim. Ecol. 64: 439?450.
KITANIDIS, P. K. 1997. Introduction to geostatistics. Cambridge University
Press, Cambridge, U.K.
KLAR, N., N. FERN?NDEZ, S. KRAMER-SCHADT, M. HERMANN, M. TRINZEN, I.
B?TTNER, AND C. NIEMITZ. 2008. Habitat selection models for Euro-
pean wildcat conservation. Biol. Conserv. 141: 308?319.
LOBO, J. M., A. JIM?NEZ-VALVERDE, AND R. REAL. 2008. AUC: A misleading
measure of the performance of predictive distribution models. Global
Ecol. Biogeogr. 17: 145?151.
LYNAM, A. J. 2010. Securing a future for wild Indochinese tigers: Trans-
forming tiger vacuums into tiger source sites. Integr. Zool. 5: 324?
334.
LYNAM, A. J., P. D. ROUND, AND W. Y. BROCKELMAN. 2006. Status of birds and
large mammals of the Dong Phayayen - Khao Yai Forest Complex.
Biodiversity Research and Training Program and Wildlife Conservation
Society, Bangkok.
MACKENZIE, D. I., J. D. NICHOLS, G. B. LACHMAN, S. DROEGE, J. A. ROYLE, AND
C. A. LANGTIMM. 2002. Estimating site occupancy rates when detection
probabilities are less than one. Ecology 83: 2248?2255.
MIQUELLE, D. G., E. N. SMIRNOV, H. G. QUIGLEY, M. G. HORNOCKER, I.G.
NIKOLAEV, AND E. N. MATYUSHIKIN. 1996. Food habits of Amur tigers
in Sikhote-AlinZapovednik and the Russian Far East: Implications for
conservation. J. Wildl. Res. 1: 138?147.
MORRISON, J. C., W. SECHREST, E. DINERSTEIN, D. S. WILCOVE, AND J. F. LAMO-
REUX. 2007. Persistence of large mammal faunas as indicators of
national human impacts. J. Mammal. 88: 1363?1380.
MUNTIFERING, J. R., A. J. DICKMAN, L. M. PERLOW, T. HRUSKA1, P. G.
RYAN, L. L. MARKER, AND R. M. JEO. 2006. Managing the matrix
for large carnivores: A novel approach and perspective from chee-
tah (Acinonyx jubatus) habitat suitability modeling. Anim. Conserv. 9:
103?112.
NEALE, J. C. C., AND B. N. SACKS. 2001. Resource utilization and interspe-
ci?c relations of sympatric bobcats and coyotes. Oikos 94: 236?
249.
NGOPRASERT, D., A. J. LYNAM, AND G. A. GALE. 2007. Human disturbance
affects habitat use and behaviour of Leopard Panthera pardus in Kaeng
Krachan National Park, Thailand. Oryx 41: 343?351.
O?BRIEN, T. G., M. F. KINNAIRD, AND H. T. WIBISONO. 2003. Crouching tigers,
hidden prey: Sumatran tiger and prey populations in a tropical forest
landscape. Anim. Conserv. 6: 131?139.
ODDEN, M., P. WEGGE, AND T. FREDRIKSEN. 2010. Do tigers displace leop-
ards? If so, why? Ecol. Res. 25: 875?881.
OSWELL, A. H. 2010. Big cat trade in Myanmar and Thailand. TRAFFIC
Southeast Asia, Petaling Jaya, Selangor, Malaysia.
PALOMARES, F., AND T. M. CARO. 1999. Interspeci?c killing among mammalian
carnivores. Am. Nat. 153: 482?508.
PETDEE, A. 2000. Feeding habits of the tiger Panthera tigris (Linnaeus) in Huai
Kha Khaeng Wildlife Sanctuary by fecal analysis. MSc dissertation,
Kasetsart University, Bangkok.
POLIS, G. A., C. A. MYERS, AND R. D. HOLT. 1989. The ecology and evolution
of intraguild predation: Potential competitors that eat each other. Ann.
Rev. Ecol. Syst. 20: 297?330.
PROMMAKUL, P. 2003. Habitat utilization and prey of the tiger Panthera tigris
(Linnaeus) in eastern ThungYai Naresuan Wildlife Sanctuary. MSc dis-
sertation (in Thai), Kasetsart University, Bangkok.
RABINOWITZ, A. 1989. The density and behavior of large cats in a dry tropical
forest mosaic in Huai Kha Khaeng Wildlife Sanctuary, Thailand. Nat.
Hist. Bull. Siam Soc. 37: 235?251.
RABINOWITZ, A., P. ANDAU, AND P. P. K. CHAI. 1987. The clouded leopard in
Malaysian Borneo. Oryx 21: 107?111.
RITCHIE, E. G., AND C. N. JOHNSON. 2009. Predator interactions, mesopredator
release and biodiversity conservation. Ecol. Lett. 12: 982?998.
SCOGNAMILLO, D., I. E. MAXIT, M. SUNQUIST, AND J. POLISAR. 2003. Coexis-
tence of jaguar (Panthera onca) and puma (Puma concolor) in a mosaic
landscape in the Venezuelan llanos. J. Zool., Lond. 259: 269?279.
SEIDENSTICKER, J. 1976. On the ecological separation between tigers and leop-
ards. Biotropica 8: 225?234.
SEIDENSTICKER, J., M. E. SUNQUIST, AND C. MCDOUGAL. 1990. Leopards living
at the edge of the Royal Chitwan National Park, Nepal. In J. C. Dan-
iel, and J. S. Serrao (Eds.). Conservation in developing countries:
Problems and prospects, pp. 415?423. Bombay Natural History Soci-
ety and Oxford University Press, Bombay.
SERGIO, F., T. CARO, D. BROWN, B. CLUCAS, H. HUNTER, J. KETCHUM, K.
MCHUGH, AND F. HIRALDO. 2008. Top predators as conservation tools:
Ecological rationale, assumptions, and ef?cacy. Annu. Rev. Ecol. Evol.
Syst. 39: 1?19.
SERGIO, F., AND F. HIRALDO. 2008. Intraguild predation in raptor assemblages:
A review. Ibis 150: 132?145.
SILVA, M., AND J. A. DOWNING. 1995. CRC handbook of mammalian body
masses. CRC Press, Boca Raton.
SIMCHAROEN, S., A. C. D. BARLOW, A. SIMCHAROEN, AND J. L. D. SMITH. 2008.
Home range size and daytime habitat selection of leopards in Huai
Kha Khaeng Wildlife Sanctuary, Thailand. Biol. Conserv. 141: 2242?
2250.
STEINMETZ, R., W. CHUTIPONG, N. SEUATURIEN, E. CHIRNGSAARD, AND M. KHA-
ENGKHETKARN. 2010a. Population recovery patterns of Southeast Asian
ungulates after poaching. Biol. Conserv. 143: 42?51.
STEINMETZ, R., D. L. GARSHELIS, W. CHUTIPONG, AND N. SEUATURIEN. 2011.
The shared preference niche of sympatric Asiatic black bears and sun
bears in a tropical forest mosaic. PLoS ONE 6(1): e14509.
DOI:10.1371/journal.pone.0014509.
STEINMETZ, R., N. SEUATURIEN, S. SRIRATTANAPORN, AND B. POONNIL. 2010b.
Experimental approaches towards tiger recovery in Kuiburi National
Park, Thailand. WWF Thailand and Department of National Parks,
Wildlife, and Plant Conservation, Bangkok, Thailand.
TERBORGH, J., L. LOPEZ, P. NU?EZ, M. RAO, G. SHAHABUDDIN, G. ORIHUELA,
M. RIVEROS, R. ASCANIO, G. H. ADLER, T. D. LAMBERT, AND L. BALBAS.
2001. Ecological meltdown in predator-free forest fragments. Science
294: 1923?1926.
TRISURAT, Y., A. PATTANAVIBOOL, G. A. GALE, AND D. H. REED. 2010. Improv-
ing the viability of large mammal populations using habitat and land-
scape models to focus conservation planning. Wild. Res. 37: 401?412.
TYRE, A. J., B. TENHUMBERG, S. A. FIELD, D. NIEJALKE, K. PARRIS, AND H. P.
POSSINGHAM. 2003. Improving precision and reducing bias in biological
surveys: Estimating false-negative error rates. Ecol. Appl. 13: 1790?1801.
WALSTON, J., J. G. ROBINSON, E. L. BENNETT, U. BREITENMOSER, AND G. A. B.
DA FONSECA. 2010. Bringing the tiger back from the brink?the six
percent solution. PLoS Biol. 8: e1000485. DOI:10.1371/journal.
pbio.1000485.
WEGGE, P., M. ODDEN, C. P. POKHAREL, AND T. STORAAS. 2009. Predator?prey
relationships and responses of ungulates and their predators to the
establishment of protected areas: A case study of tigers, leopards and
their prey in Bardia National Park, Nepal. Biol. Conserv. 142: 189?
202.
WIKRAMANAYAKE, E., E. DINERSTEIN, J. SEIDENSTICKER, S. LUMPKIN, B. PANDAV,
M. SHRESTHA, H. MISHRA, J. BALLOU, A. J. T. JOHNSINGH, I. CHESTIN, S.
SUNARTO, P. THINLEY, K. THAPA, G. JIANG, S. ELAGUPILLAY, H. KAFLEY,
N. M. B. PRADHAN, K. JIGME, S. TEAK, P. CUTTER, MD. ABDUL AZIZ,
AND T. THAN. 2011. A landscape-based conservation strategy to dou-
ble the wild tiger population. Conserv. Lett. 4: 219?227.
XI, W., R. N. COULSON, J. D. WALDRON, M. D. TCHAKERIAN, C. W. LAFON, D.
M. CAIRNS, A. G. BIRT, AND K. D. KLEPZIG. 2008. Landscape modeling
for forest restoration planning and assessment: Lessons from the
southern Appalachian Mountains. J. For. 106: 191?197.
8 Ngoprasert et al.