Content uploaded by Krishnendu Mondal
Author content
All content in this area was uploaded by Krishnendu Mondal on Aug 16, 2022
Content may be subject to copyright.
ORIGINAL PAPER
Factors influencing the distribution of leopard in a semiarid
landscape of Western India
Krishnendu Mondal &K. Sankar &Qamar Qureshi
Received: 8 May 2012 /Accepted: 15 October 2012 / Published online: 2 November 2012
#Mammal Research Institute, Polish Academy of Sciences, Białowieża, Poland 2012
Abstract Previously, factors governing distribution of leop-
ard (Panthera pardus fusca) in forest habitats of the Indian
subcontinent were unknown. The present study assessed the
influence of different ecogeographic variables determining
the distribution of leopards in and around Sariska Tiger
Reserve through MaxEnt habitat suitability model based
on camera trapping method. Camera trapping was used to
collect presence/absence information in the study area from
December 2008 to June 2010. Information of 11 macro-
habitat characteristics and variables (habitat types, prey
species, Normalized Difference Vegetation Index (NDVI),
elevation, livestock, village, water source, etc.) were col-
lected along with leopard presence data. The probability of
presence of leopards increased with decreasing distance to
water and increasing encounter rate of peafowl (Pavo cris-
tatus), chital (Axis axis), sambar (Rusa unicolor), and wild
pig (Sus scrofa). It was found that the probability of pres-
ence of leopards increased with increasing area of Zizyphus
mixed forest patches and NDVI. Results of this study
showed that the probability of presence of leopards was
higher in habitat types with intermediate cover, high wild
prey base, and water sources. They also indicated that leop-
ards are not always ‘generalists’showing some degree of
specialization, at least in their choice of habitat, and this
information is useful for conserving leopard in human-
dominated landscapes.
Keywords Panthera pardus fusca .Distribution modeling .
MaxEnt .Habitat variables .Camera trapping .Dry
deciduous forest
Introduction
The leopard Panthera pardus, due to the generalist nature of
its ecology has abilities to adapt to different habitats and
prey and to alter their behavior in close proximity to humans
(Myers 1986; Hamilton 1976). Despite being one of the
most abundant felids, it is considered as one of the least
studied of the large carnivores (Hamilton 1976). Most of the
studies on leopards have been conducted in Africa (Schaller
1972; Hamilton 1976; Bertram 1982; Bailey 1993; Jenny
1996; Ray and Sunquist 2001; Henschel et al. 2005). In
India, the leopards belonging to subspecies P. pardus fusca,
have been studied addressing conflicts with humans
(Edgaonkar and Chellam 1998; Athreya et al. 2007; Goyal
and Chauhan 2006), population estimation (Mondal 2006,
2011; Sankar et al. 2009; Edgaonkar 2008; Harihar et al.
2009; Kalle et al. 2011), and food habits (Sathyakumar
1992; Karanth and Sunquist 1995; Edgaonkar and Chellam
1998; Sankar and Johnsingh 2002; Ramesh et al. 2009;
Mondal et al. 2011). No other study is actually available
on the factors governing the distribution of leopards in the
Indian subcontinent. The present study assessed the influ-
ence of different ecogeographic variables determining leop-
ards' distribution in and around Sariska Tiger Reserve
through a habitat suitability model.
Habitat suitability maps are basically computed by fitting
a relevant statistical or numerical model to environmental
data and species distribution data. Classical methods (e.g.,
logistic regression, discriminant analysis, Generalized
Linear Model, etc.) need both species presence and absence
data. Presences attest a good habitat and absences attest a
Communicated by: Krzysztof Schmidt
Electronic supplementary material The online version of this article
(doi:10.1007/s13364-012-0109-6) contains supplementary material,
which is available to authorized users.
K. Mondal (*):K. Sankar :Q. Qureshi
Wildlife Institute of India,
Dehradun, India
e-mail: krishtigris@yahoo.co.in
K. Sankar
e-mail: sankark@wii.gov.in
Acta Theriol (2013) 58:179–187
DOI 10.1007/s13364-012-0109-6
bad habitat. An “absence”(0lack of observation) may have
three causes: (1) the species is present but was not detected
(false absence), (2) the habitat is suitable, but the species is
not yet/no more present (false absence), and (3) the habitat is
actually not suitable (true absence) (Hirzel et al. 2002).
Categorizing suitable habitat for leopard requires informa-
tion at multiple scales. First-order selection (Johnson 1980)
refers to the distribution of a species with respect to geo-
graphical space. This information is used to derive a prob-
ability of species presence at each location. Leopards are not
only rare and secretive, they are also crepuscular (Sunquist
and Sunquist 2002) and without intensive effort there is a
high likelihood of nondetections in areas where leopards are
actually present, contaminating the absence data. Presence-
only models are a way of dealing with this problem. The
present study used the maximum entropy model (Phillips et
al. 2004), a presence-only environmental habitat-based
method, to create habitat suitability maps for leopards in
Sariska Tiger Reserve and a surrounding 5 km buffer area.
Study area
The study was carried out in Sariska Tiger Reserve (Sariska
TR), Rajasthan (74°17′Eto76°34′E and 25°5 to 27° 33 N)
from December 2008 to June 2010. The total area of the Tiger
Reserve is 881 km
2
, of which 274 km
2
is notified National
Park. There are 31 villages within the Tiger Reserve boundary
and ten out of them are in the National Park. The human
population is over 1,700 in the villages of the National Park
along with a population of 10,000 livestock including buffalo,
cow, goat, and sheep. In the entire Sariska TR, the human
population is around 6,000 and the livestock population is
more than 20,000 (Sankar et al. 2009). The vegetation of this
region is tropical dry deciduous forest and tropical thorn forest
(Champion and Seth 1968). The climate is subtropical, char-
acterized by a distinct winter (October–February), summer
(March–June), and monsoon (July–September). The average
annual rainfall is 700 mm, occurring mostly during July–
September. The wild ungulates found in Sariska are chital
(Axis axis), sambar (Rusa unicolor), nilgai (Boselaphus trag-
ocamelus), and wild pig (Sus scrofa). Apart from leopard,
other carnivores present are tiger (Panthera tigris), striped
hyena (Hyaena hyaena), golden jackal (Canis aureus), jungle
cat (Felis chaus), desert cat (Felis silvestris), common mon-
goose (Herpestes edwardsi), small Indian mongoose (H.
auropunctatus), ruddy mongoose (H. smithi), palm civet
(Paradoxurus hermaphroditus), small Indian civet
(Viverricula indica), and ratel (Mellivora camensis).
Rhesus monkey (Macaca mulatta) and common langur
(Semnopithecus entellus) are the two primates found.
Porcupine (Hystrix indica), rufous-tailed hare (Lepus
nigricollis ruficaudatus), and peafowl (Pavo cristatus)
also occur in Sariska (Sankar et al. 2009).
Material and methods
The camera trapping method was adopted to collect pres-
ence/absence information along with direct sightings data of
leopards in the intensive study area from December 2008 to
June 2010. The entire Sariska TR and adjoining 5-km areas
were divided into 2× 2-km
2
grid cells (n0574). A total of
165 locations were selected for the placement of camera
traps in the intensive study area (Sariska National Park)
and adjoining areas inside Sariska TR. All the 574 grid cells
could not be sampled with camera traps because of exten-
sive human disturbance and inaccessibility and out of 574
grids, only 165 grids were found suitable for camera trap-
ping. One pair of camera was deployed in one 2 × 2-km
2
grid
cell (n0165). Each camera trap location was considered as
the representative of leopard presence/absence data in that
grid cell. The rest of the grid cells were sampled through
trail survey (total effort 0206 km) and the Global
Positioning System (GPS) locations of indirect leopard ev-
idence (pugmark, scat, kill, etc.) were recorded whenever
encountered. Camera trapping was done covering ~600 km
2
area and care had been taken to sample all the habitat types
through camera traps. Camera traps were operated 25–54
nights in each grid cell. Records on presence /absence of
leopards were converted into digital data in Geographic
Information System using program ArcGIS (Environmental
Systems Research Institute, Redlands, CA, USA) and were
stored on the base map prepared for the Sariska TR and
adjoining 5-km areas. If leopards were detected at least once
at a site over the entire sampling duration, then the species was
recorded as present ‘1’or ‘0’otherwise. Individual leopards
were also identified based on their rosette pattern (Mondal et
al. 2012;Sankaretal.2009) to estimate the number of indi-
viduals operating in the study area in the study period. Spatial
data which was generated for Sariska TR (Sankar et al. 2009)
was used for preparing spatial layers on habitat features rele-
vant for each species. A total of 11 macrohabitat character-
istics and variables were considered for the analysis (Table 1).
Encounter rates of five prey species (chital, sambar, wild pig,
common langur, and peafowl) in the study area were included
as variables. The other variables include Digital Elevation
Model (DEM), anthropogenic variables (i.e., Euclidean dis-
tance from villages and roads), a habitat variable (i.e., vege-
tation types), and Normalized Differential Vegetation Index
(NDVI) and a hydrological variable (i.e. Euclidean distance to
water). NDVI basically indicates the chlorophyll content
[NDVI 0(NIR −VIS) / (NIR + VIS); VIS and NIR stand
for the spectral reflectance measurements acquired in the
visible (red) and near-infrared regions] of an area, which is
180 Acta Theriol (2013) 58:179–187
an important limiting factor for the species distribution in
tropical dry deciduous forest in summer season. These varia-
bles were chosen on the basis of field knowledge and infor-
mation on leopard biology (Prater 1980;Schaller1967). The
distribution of livestock and nilgai were correlated (Pearson's
correlation00.673, p00.001, N0574); hence, the distribution
of nilgai was removed from the habitat suitability analysis.
The distance to the nearest village was also highly correlated
with the distribution of livestock (Pearson's correlation0
0.991, p00.000, N0574), and hence, it was removed from
the analysis.
Mapping of vegetation types was done earlier in Sariska
based on remotely sensed data of Landsat-7 ETM + imagery
for the month of September 2007. Geocoded False Color
Composite (FCC) on a 1:50,000 scale for the entire Sariska
TR and the adjoining 5-km area (Sankar et al. 2009). Nine
vegetation and land cover classes were delineated and
mapped with 80 % accuracy (Sankar et al. 2009). Area
occupied by each vegetation type was extracted grid cell-
wise from a vegetation map. Separate layers were prepared
for each vegetation type, thereby computing the area for
each habitat variable. Digital data on contour and drainage
were used to create a DEM on the basis of interpolation in
the program ArcGIS 9.2 (Environmental Systems Research
Institute, Redlands, CA, USA). All village locations and
water points were recorded using GPS during the field work.
The Euclidean distance was calculated for each grid center
from the nearest water sources and villages. The encounter
rates of different wild prey species and livestock, obtained
from line transects, were then extracted in each grid cell. A
number of 72 transects were laid covering the entire study
area and the researchers walked three to five times in early
morning to record the encounter rates of wild and domestic
prey species. All the habitat variable characteristics were
then converted into ASCII layers for the entire grid. The
data were analyzed using program MaxEnt (Phillips et al.
2006). “MaxEnt is a program for modeling species distribu-
tions from presence-only species records, which minimizes
the relative entropy between two probability densities (one
estimated from the presence data and one from the land-
scape) defined in covariate space”(Elith et al. 2011). While
the MaxEnt model runs, it keeps track of which environ-
mental variables are contributing to fitting the model. Each
step of the MaxEnt algorithm increases the gain of the
model by modifying the coefficient for a single feature;
the program assigns the increase in the gain to the environ-
mental variable(s) that the feature depends on. “The pro-
gram MaxEnt's predictive performance is consistently
competitive with the highest performing methods”(Elith et
al. 2011) and has been utilized extensively for modeling
species distributions. Published literatures cover “diverse
aims (finding correlates of species occurrences, mapping
Table 1 List of variables used to predict habitat suitability model for leopard in Sariska Tiger Reserve, Western India through MaxEnt analysis
Variables Variable type Parameter Source
1. Habitat Anogeissus-dominated
forest
Percentage of area in each grid cell (m
2
) Land use and land cover map from Landsat-7
ETM + data (source: Sankar et. al. 2009)
Boswellia-dominated forest Percentage of area in each grid cell (m
2
)
Zizyphus mixed forest Percentage of area in each grid cell (m
2
)
Butea-dominated forest Percentage of area in each grid cell (m
2
)
Acacia-dominated forest Percentage of area in each grid cell (m
2
)
Scrubland Percentage of area in each grid cell (m
2
)
Barren land Percentage of area in each grid cell (m
2
)
Agricultural land Percentage of area in each grid cell (m
2
)
Water body Percentage of area in each grid cell (m
2
)
NDVI (summer) Mean value in each grid cell Wildlife Institute GIS cell
2. Prey
information
Chital Encounter rate in each grid cell (number/km) Line transect data (2007)
Sambar Encounter rate in each grid cell (number/km)
Wild pig Encounter rate in each grid cell (number/km)
Nilgai Encounter rate in each grid cell (number/km)
Livestock Encounter rate in each grid cell (number/km)
Common langur Encounter rate in each grid cell (number/km)
Peafowl Encounter rate in each grid cell (number/km)
3. Anthropogenic Distance from village Distance (m) to nearest village from the center of
each grid cell (m)
Village and road map, WII
4. Topographical DEM Mean value in each grid cell Wildlife Institute GIS cell
5. Hydrological Distance from water bodies Distance (m) to nearest water point from the center
of each grid cell (m)
Field data, WII
Acta Theriol (2013) 58:179–187 181
current distributions, and predicting to new times and pla-
ces) across many ecological, evolutionary, conservation and
biosecurity applications”(Tinoco et al. 2009; Young et al.
2009; Monterroso et al. 2009; Elith et al. 2011).
The hypothesis in the present study was ‘leopards are
present when the area is suitable.’The hypothesis was
validated with the area under the curve (AUC), which is
the probabilistic ratio between sensitivity and specificity of
the hypothesis. In the present study, the receiver operating
characteristic (ROC) (Fig. 1) curve predicted the AUC with
the present data and random prediction. The ROC curve is a
statistical accuracy measure of a model given by the type I
error (probability of rejecting the hypothesis when it is
actually true, α) vs. 1−type II error (probability of rejecting
the hypothesis, when it is actually false, 1 −β) or power of
the analysis. The main advantage of ROC analysis is
thattheareaundertheROCcurve(AUC)providesa
single measure of model performance and is indepen-
dent of any particular choice of threshold. ROC analysis
has recently been applied to a variety of classification
problems in the evaluation of models of species distri-
butions (Elith et al. 2002; Fielding and Bell 1997). To
understand the importance of each ecogeographic vari-
able for predicting the distribution of leopard in Sariska
TR and adjoining areas, a jackknife test was performed
in the MaxEnt program. In the jackknife test, each
variable was used alone (in isolation) to test its predic-
tive power and the full model was also used excluding
each variable, in turn, to calculate the loss of informa-
tion from each variable.
Results
In total, 158 photographic captures were obtained from 36
individual leopards. Detection probability of leopard in the
entire sampled area was estimated to be 0.22 (program
PRESENCE 4.0; Hines 2006). In the present study, the
ROC curve showed that the MaxEnt model could predict a
site as unsuitable when it is actually unsuitable with a power
(1−β) of 0.9 (with 90 % accuracy), while the probability of
committing an error of not predicting it as unsuitable was
0.2 (Fielding and Bell 1997).
The MaxEnt-predicted model gave the percent contribu-
tion and permutation importance of each variable, which
matter most for the leopard distribution in Sariska TR and
adjoining 5-km areas (Table 2). The effects of different
variables determining leopards' presence, which contributed
most (>5 %) for predicting habitat suitability model for
leopards, were given in Appendix Figs. 1to 8. Each
Appendix figure (Appendix Figs. 1to 8) represents a differ-
ent MaxEnt model created using only the corresponding
variable. Each Appendix figure shows the probability of
occurrence of leopards modeled with a particular ecogeo-
graphic variable. These plots reflect the dependence of
predicted suitability for leopard on the selected variable in
the Sariska TR and adjoining areas. The distances to water
from leopard presence locations contributed maximally
(19.8 %) to the habitat suitability model for leopards.
From the response curve of the leopard (Appendix Fig. 1),
it was observed that the probability of leopard presence was
negatively correlated with the increasing distance to water
Fig. 1 The prediction of area
under receiver operating
characteristic (ROC) curve
(AUC) for determination habi-
tat suitability model for leopard
in Sariska Tiger Reserve, West-
ern India
182 Acta Theriol (2013) 58:179–187
sources. Among the prey species, distribution of peafowl
(11.9 %) and chital (11.5 %) contributed most to the habitat
suitability model for leopards followed by livestock (7 %),
sambar(5.9%),wildpig(5.2%),andcommonlangur
(0.8 %). The probability of presence of leopards increased
with increasing encounter rate of peafowl, chital, sambar,
and wild pig (Appendix Figs. 2,3,7and 8). Initially, the
probability of presence of leopards decreased with increas-
ing encounter rate of domestic livestock but further in-
creased with increase of encounter rate of domestic
livestock (Appendix Fig. 6), which may be attributed to
the presence of ten villages in the National Park area.
Among the vegetation types, Zizyphus mixed forest con-
tributed most (8.4 %) to predict leopards' distribution fol-
lowed by Boswellia-dominated forest (4.7 %), agricultural
land (4.4 %), Anogeissus-dominated forest (2.8 %), scrub-
land (1.5 %), Acacia-dominated forest (1.5 %), barren land
(1.3 %), and Butea-dominated forest (0.4 %) (Table 2). It
was found that the probability of presence of leopards in-
creased with increasing area of Zizyphus mixed forest
patches (Appendix Fig. 4). NDVI, in summer, directly con-
tributed to leopard distribution. It was observed that the
probability of presence of leopards increased with increas-
ing NDVI, i.e., increasing “greenness,”in summer
(Appendix Fig. 5).
The role of each variable (with or without) was examined
in the jackknife test against the MaxEnt prediction gain
(Fig. 2). It was observed that the ecogeographic variable
with the highest predictive power, when used in isolation
was the sambar, which, therefore, appeared to have the most
useful information by itself for the prediction of the distri-
bution of leopards. The ecogeographic variable that de-
creased the probability of prediction of the distribution of
leopards most, when it was omitted was agricultural land,
which, therefore, appeared to have useful information that
was not present in other variables. Agricultural land was
found to be negatively correlated with the probability of
presence of leopard.
Considering the contribution of each ecogeographic var-
iable for the prediction of the distribution of leopards, a
composite habitat suitabality value was obtained for each
grid cell by the program MaxEnt, which was then repre-
sented as MaxEnt habitat suitability model for leopards in
Sariska TR and adjoining 5-km areas (Fig. 3).
Discussion
In the present study, the habitat suitability model for leop-
ards in the Sariska TR and surrounding area was predicted
using presence only data. Systematic biological survey data
tend to be sparse and/or limited in coverage. Species records
are available, albeit in the form of presence-only records in
literature records and herbarium and museum databases.
Many of these databases represent well over a century of
public and private investment in biological science and are
an important source of species occurrence data. The desire
to maximize the utility of such resources has spawned an
array of Species Distribution Model (SDM) methods for
modeling presence-only data. MaxEnt (Phillips et al. 2006;
Phillips and Dudık2008) is one such method followed in
the present study. Expanding the use of presence-only data
for modeling species distributions has prompted wide dis-
cussion about the sorts of distributions (e.g., potential vs.
realized) that can be modeled with presence-only data in
contrast to presence–absence data (Hirzel and Le Lay 2008;
Soberon and Nakamura 2009; Lobo et al. 2010). As men-
tioned in several of these articles, the subject is complex
because of the interplay of data quality (amount and accu-
racy of species data; ecological relevance of predictor vari-
ables; and availability of information on disturbances,
dispersal limitations, and biotic interactions), modeling
method, and scale of analysis (Elith et al. 2011).
Habitat models of ‘generalist’big cats demonstrate asso-
ciations with certain elevations, aspects, ruggedness and
vegetation types, and negative correlations with proximity
to roads and human density (Ortega-Huerta and Medley
1999; Hatten et al. 2005; Carroll and Miquelle 2006;
Linkie et al. 2006). In comparison with these models, the
present study showed similar findings even in a smaller
landscape. The variable ‘water’was found to be the most
Table 2 Percent contribution of each biogeographical variable predict-
ing leopard distribution model in Sariska Tiger Reserve, Western India
Variable Percent
contribution
Permutation
importance
Dist to water 19.8 13.6
Peafowl 11.9 18.9
Chital 11.5 0.7
Zizyphus mixed forest 8.4 3.8
NDVI (summer) 8.1 6.4
Livestock 7 4.2
Sambar 5.9 12.3
Wild pig 5.2 0
Elevation 4.8 6
Boswellia-dominated
forest
4.7 5.7
Agricultural land 4.4 8.9
Anogeissus-dominated
forest
2.8 10.3
Scrubland 1.5 2.1
Acacia-dominated forest 1.5 3.2
Barren land 1.3 0.9
Common langur 0.8 0.4
Butea-dominated forest 0.4 2.5
Acta Theriol (2013) 58:179–187 183
important factor contributing (19.8 %) to the leopard distri-
bution in Sariska TR and adjoining areas. Sources of water
are known as the most prominent limiting factor and play an
important role in dry deciduous habitat for species distribu-
tion (Schaller 1972; Kruuk 1972). In Sariska TR, the central
part of the National Park area proved to be the most suitable
habitat for leopards because it held more numbers of water
bodies (artificial and natural), leading to the highest pres-
ence of leopards in that area.
Carnivore distribution and densities are clearly linked to
prey distribution and abundance (Carbone and Gittleman
2002). Peafowl was the most abundant prey species in the
entire Sariska TR. The density of peafowl in the entire
Sariska TR was 60.2/km
2
and in the National Park area it
was 100.7 to 121.4/km
2
, which was predicted as the most
suitable habitat for leopards. The densities of chital and
sambar in the National Park area were comparatively higher
than the peripheral area and the MaxEnt predicted leopards'
distribution was positively correlated with both chital and
sambar. Mondal et al. (2011) found that chital and sambar
contributed most in leopards' diet and were the most pre-
ferred prey species in the study area. Evidently, the distri-
bution of chital and sambar was found to be the key factors
for the prediction of suitable habitat for leopards in Sariska
TR. Although wild pig was consumed by leopards less than
its availability (Mondal et al. 2011), the habitat suitability
model for leopard showed positive correlation with the
distribution of wild pig because its distribution was limited
Fig. 2 Jackknife test of
ecogeographic variables to test
their predictive power
explaining leopards'
distribution in Sariska Tiger
Reserve, Western India.
Environmental variables:
acadom Acacia-dominated
forest, agrilnd agricultural land,
anodom Anogeissus-dominated
forest, barlnd barren land,
bosdom Boswellia-dominated
forest, butdom Butea-
dominated forest, demmn mean
of Digital Elevation Model,
distwtr distance to nearest water
source, livstck livestock,
scrblnd scrubland, sndvimn
mean of Normalized Difference
Vegetation Index in summer,
zizmix Zizyphus mixed forest
Fig. 3 Habitat suitability map for leopards in Sariska Tiger Reserve
and surrounding 5 km areas in Western India
184 Acta Theriol (2013) 58:179–187
to the National Park area. Although the distribution of
livestock contributed 7 % for the prediction of leopard's
suitability model, it showed no significant trend (neither
positive nor negative). As there are 31 villages located
inside the Sariska TR, the distribution of leopards was
influenced by neither the presence of village nor the distri-
bution of livestock.
In the present study, the NDVI in summer contributed
8.1 % for the prediction of suitability model of leopard in
Sariska TR and revealed a positive correlation with leopard
presence. “Various studies have shown that the NDVI inte-
grates the influence of climatic variables (e.g. rainfall and
evapotranspiration) and other environmental factors (Cihlar
et al. 1991) and is related to the distribution of both plant
and animal species diversity”(Walker et al. 1992;
Gavashelishvili and Lukarevskiy 2008). “The NDVI corre-
lates directly with photosynthetically active biomass or veg-
etation productivity (Tucker and Sellers 1986; Reed et al.
1994), hence it accounts for biomass of wild ungulates and
other herbivores in forested areas”(Loe et al. 2005;
Pettorelli et al. 2005; Gavashelishvili and Lukarevskiy
2008). The NDVI in different seasons were correlated;
hence, only the NDVI in summer was taken into analysis.
In a dry deciduous area, like Sariska TR, the NDVI in
summer is more important than in other seasons because
the available vegetation biomass and productivity become
critical in peak summer and drought period. Thus, the NDVI
values indicated the presence of food (i.e., herbivores) and
water for leopards, as well as cover (shrubs and trees) as
important for thermal protection, reproduction, escape, and
stalking prey.
Zizyphus mixed forest with combination of several edible
grass species, allows assemblage of ungulates throughout
the year, which, in turn, helps leopards to stalk and kill. The
availability of Zizyphus mixed forest was very limited to the
valley areas of Sariska National Park (only 7 % of the total
habitat of Sariska TR), but it was used more than its avail-
ability (p<0.05) by leopards (Mondal 2011). The availabil-
ity of Boswellia-dominated forest and Anogeissus-
dominated forest was found throughout the Sariska TR but
considering the other ecogeographical factors (prey base and
NDVI), the most suitable habitat of leopard was predicted in
the Sariska National Park area. Agricultural land and barren
land showed negative correlation with leopards' presence.
Though ‘scrubland’contributed very little (1.5 %) to deter-
mine suitable habitat for leopards, it showed slight positive
correlation with the presence of leopard, as scrubland pro-
vides day time refuge and cover for leopard in dry decidu-
ous thorn forest (Kruuk 1972). Acacia-dominated forest and
Butea-dominated forest contributed less (1.5 % and 0.4 %,
respectively) for predicting the suitable habitat of leopards,
and hence showed no detectable correlation with the pres-
ence of leopard.
Considering the contributions of prey species and avail-
ability of different habitat types for the prediction of the
distribution of leopard, it was found that leopards at Sariska
TR selected habitats where it was easier to catch prey rather
than in areas where prey was more common. In Sariska TR,
the habitat types with Zizyphus woodland and savanna
grassland hold more prey density than any other habitat
types. Leopards are widely perceived to favor the densest
habitats available for hunting (Hes 1991; Sunquist and
Sunquist 2002;Bailey1993). For example, of 50 kills
recorded in the Kruger National Park (Bailey 1993), 46 %
were found in dense riparian vegetation, 44 % in medium to
dense thornbush thickets, and 10 % in open habitats.
Hayward et al. (2006) suggested a similar trend, with leop-
ards preferring to hunt in dense environments. However, in
the present study, out of 82 kills recorded, 60 were in
Zizyphus mixed forest, Butea mixed forest and scrubland.
However, results from this study showed that the probability
of presence of leopards was higher in habitat types with
intermediate cover levels, even though areas with dense
vegetation may have higher abundance of available prey.
Leopards are visual hunters, relying heavily on sight and to
a lesser extent on hearing to detect prey (Sunquist and
Sunquist 2002;Bailey1993). Thicker habitat types on
Sariska TR (riparian forest, Anogeissus forest with Grewia
understory or Acacia forest with Adathoda understory) may
have reduced the chance of an encounter with prey suffi-
ciently to negate any benefits accrued from increased cover
for stalking. Additionally, even if prey is successfully
detected, increased vegetation density may not always ben-
efit a stalking predator. Very thick cover can either impede
the progress of a stalk by obstructing a clear view of the
target or increase the chance of detection by prey (because
of a noisier approach by the predator) and may hamper the
final chase even if a predator gets within charging distance
(Leyhausen 1979).
Although the leopard has the widest habitat tolerance of
any large felid and is more resilient than others, this cat
species faces various anthropogenic threats (Sunquist and
Sunquist 2002). The widely held perception that leopards
are ‘super-generalists’with little need for dedicated conser-
vation action is increasingly criticized (Spong et al. 2000;
Balme and Hunter 2004; Ray et al. 2005). The results in the
present study indicated that leopards showed some degree of
specialization, at least in their choice of habitat.
Acknowledgments We thank Rajasthan Forest Department for
granting permission to work in Sariska, as part of ‘Ecology of Leopard’
project conducted by Wildlife Institute of India (WII). We thank the
director and dean, WII for encouragement and support they provided
for the study. We thank Sutirtha Dutta and Swati for helping in data
analysis. We thank our field assistants, Jairam, Omi, and Ramesh, for
their assistance in the field.
Acta Theriol (2013) 58:179–187 185
References
Athreya VR, Thakur SS, Chaudhari S, Belsare AV (2007) Leopards in
human dominated areas: a spillover from sustained translocations
into nearby forests. J Bombay Nat Hist Soc 104:13–18
Bailey TN (1993) The African leopard: ecology and behavior of a
solitary felid. Columbia University Press, New York
Balme GA, Hunter LTB (2004) Mortality in a protected leopard pop-
ulation, Phinda Private Game Reserve, South Africa: a population
in decline? CC Africa Ecol J 6:5e11
Bertram BCR (1982) Leopard ecology as studied by radio tracking.
Symp Zool Soc Lond 49:341–352
Carbone C, Gittleman JL (2002) A common rule for the scaling of
carnivore density. Science 295:2273–2276. doi:10.1126/
science.1067994
Carroll C, Miquelle DG (2006) Spatial viability analysis of Amur tiger
Panthera tigris altaica in the Russian Far East: the role of pro-
tected areas and landscape mat
Champion HG, Seth SK (1968) A revised survey of forest types of
India. Manager of Publications, Government of India, New Delhi,
p 404
Cihlar J, St-Laurent L, Dyer JA (1991) The relation between normalized
differences vegetation index and ecological variables. Remote Sens
Environ 35:279–298. doi:10.1016/0034-4257(91)90018-2
Edgaonkar A (2008) Ecology of the leopard (Panthera pardus fusca)
in Satpura National Park and Bori Wildlife Sanctuary. Ph.D.
Thesis, Saurashtra University, Rajikot
Edgaonkar A, Chellam R (1998) A preliminary study on the ecology of
the leopard (Pathera pardus fusca) in the Sanjay Gandhi National
Park, Maharashtra. RR-98/002. Wildlife Institute of India, Dehra
Dun, p 33
Elith J, Burgman MA, Regan HM (2002) Mapping epistemic uncer-
tainties and vague concepts in predictions of species distribution.
Ecol Model 157:313–329. doi:10.1016/S0304-3800(02)00202-8
Elith J, Phillips SJ, Hastie T, Dudik M, Chee YE, Yates CJ (2011) A
statistical explanation of MaxEnt for ecologists. Divers Distrib
17:43–57. doi:10.1111/j.1472-4642.2010.00725.x
Fielding AH, Bell JF (1997) A review of methods for the assessment of
prediction errors in conservation presence/absence models. Envi-
ron Conserv 24:38–49
Gavashelishvili A, Lukarevskiy V (2008) Modelling the habitat require-
ments of leopard Panthera pardus in west and central Asia. J Appl
Ecol 45:579–588. doi:10.1111/j.1365-2664.2007.01432.x
Goyal SP, Chauhan D (2006) Status and ecology of leopard in Pauri
Garhwal: ranging patterns and reproductive biology of leopard
(Panthera pardus) in Pauri Garhwal. Comprehensive Report 2000
to 2006. Wildlife Institute of India, Dehra Dun
Hamilton PH (1976) The movements of leopards in Tsavo National
Park, Kenya, as determined by radio-tracking. M.Sc. Thesis,
University of Nairobi, Kenya
Harihar A, Pandav B, Goyal SP (2009) Density of leopard (Panthera
pardus) in Chilla Range of Rajaji National Park, Uttarakhand,
India. Mammalia 73:68.71. doi:10.1515/MAMM.2009.007
Hatten JR, Averill-Murray A, Van Pelt WE (2005) A spatial model of
potential jaguar habitat in Arizona. Journal of Wildlife Manage-
ment, 69, 1024–1033. doi: http://dx.doi.org/10.2193/0022-541X
(2005)069[1024:ASMOPJ]2.0.CO;2 DOI:10.2193/0022-541X
(2005)069%5b1024:ASMOPJ%5d2.0.CO;2
Hayward MW, Henschel P, O'Brien J, Hofmeyr M, Balme G, Kerley
GIH (2006) Prey preferences of the leopard (Panthera pardus). J
Zool 270:298–313. doi:10.1111/j.1469-7998.2006.00139.x
Henschel P, Abernathy KA, White LJT (2005) Leopard food habits in
the Lopé National Park, Gabon, Central Africa. Afr J Ecol 43:21–
28. doi:10.1111/j.1365-2028.2004.00518.x
Hes L (1991) The leopards of Londolozi. Struik Winchester, Cape Town
Hirzel AH, Le Lay G (2008) Habitat suitability modeling and niche
theory. J Appl Ecol 45:1372–1381. doi:10.1111/j.1365-
2664.2008.01524.x
Hirzel AH, Hausser J, Chessel D, Perrin N (2002) Ecological niche
factor analysis: how to compute habitat-suitability maps without
absence data? Ecology 87:2027–2036. http://dx.doi.org/10.1890/
0012-9658(2002)083[2027:ENFAHT]2.0.CO;2
Jenny D (1996) Spatial organization of leopards Panthera pardus in Tai
National Park, Ivory Coast: is rainforest habitat a ‘tropical haven’?J
Zool 240:427–440. doi:10.1111/j.1469-7998.1996.tb05296.x
Johnson DH (1980) The comparison of usage and availability measure-
ments for evaluating resource preference. Ecology 61:65–71
Kalle R, Ramesh T, Qureshi Q, Sankar K (2011) Density of tiger and
leopard in a tropical deciduous forest of Mudumalai Tiger Reserve,
southern India as estimated using photographic capture–recapture
sampling. Acta Theriol. doi:10.1007/s13364-011-0038-9
Karanth KU, Sunquist ME (1995) Prey selection by tiger, leopard and
dhole in tropical forests. J Anim Ecol 64:439–450
Kruuk H (1972) The spotted hyena. Chicago. The University of Chi-
cago Press, Chicago
Leyhausen P (1979) Cat behavior: the predatory and social behaviour
of domestic and wild cats. Garland STPM, New York
Linkie M, Chapron G, Martyr DJ, Holden J, Leader-Williams N (2006)
Assessing the viability of tiger subpopulations in a fragmented
landscape. J Appl Ecol 43:576–586. doi:10.1111/j.1365-
2664.2006.01153.x
Lobo JM, Jimenez-Valverde A, Hortal J (2010) The uncertain nature of
absences and their importance in species distribution modelling.
Ecography 33:103–114. doi:10.1111/j.1600-0587.2009.06039.x
Loe L, Bonenfant C, Mysterud A (2005) Climate predictability and
breeding phenology in red deer: timing and synchrony of rutting
and calving in Norway and France. J Anim Ecol 74:579–588.
doi:10.1111/j.1365-2656.2005.00987.x
Mondal K (2006) Leopard and ungulate abundance estimation in Rajaji
National Park, Uttranchal. Masters Thesis. Forest Research Insti-
tute (Deemed University), Dehradun, p 95
Mondal K (2011) Ecology of leopard (Panthera pardus fusca)in
Sariska Tiger Reserve, Rajasthan, India. Ph.D. Thesis submitted
to Saurashtra University, Rajkot, India, p 235
Mondal K, Gupta S, Qureshi Q, Sankar K (2011) Prey selection and
food habits of leopard (Panthera pardus fusca) in Sariska Tiger
Reserve, Rajasthan, India. Mammalia 75:201–205. doi:10.1515/
MAMM.2011.011
Mondal K, Gupta S, Bhattacharjee S, Qureshi Q, Sankar K (2012)
Response of leopards to re-introduced tigers in Sariska Tiger
Reserve, Western India. Int J Biodivers Conserv 4(5):228–236.
doi:10.5897/IJBC12.014
Monterroso P, Brito JC, Ferreras P, Alves PC (2009) Spatial ecology of
the European wildcat in a Mediterranean ecosystem: dealing with
small radio-tracking datasets in species conservation. J Zool
279:27–35
Myers N (1986) Conservation of Africa's cats: problems and opportu-
nities. In: Miller SD, Everett DD (eds) Cats of the world. National
Wildlife Federation, Washington, DC, pp 437–457
Ortega-Huerta MA, Medley KE (1999) Landscape analysis of jaguar
(Panthera onca) habitat using sighting records in the Sierra De
Tamaulipas, Mexico. Environ Conserv 26:257–269
Pettorelli N, Vik JO, Mysterud A (2005) Using the satellite-derived
normalized difference vegetation index (NDVI) to assess ecolog-
ical effects of environmental change. Trends Ecol Evol 20:503–
510
Phillips SJ, Dudık M (2008) Modeling of species distributions with
MaxEnt: new extensions and a comprehensive evaluation. Ecog-
raphy 31:161–175. doi:10.1111/j.0906-7590.2008.5203.x
Phillips SJ, Dudık M, Schapire RE (2004) A maximum entropy ap-
proach to species distribution modeling. In: Proceedings of the
186 Acta Theriol (2013) 58:179–187
21st International Conference on Machine Learning. ACMPress,
New York, pp 655–662
Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy
modeling of species geographic distributions. Ecol Model
190:231–259. doi:10.1016/j.ecolmodel.2005.03.026
Prater SH (1980) The book of Indian animals. Bombay Natural History
Society, Bombay
Ramesh T, Snehalatha V, Sankar K, Qureshi Q (2009) Food habits and
prey selection of tiger and leopard in Mudumalai Tiger Reserve,
Tamil Nadu, India. J Sci Trans Environ Technovation 2:170–181
Ray JC, Sunquist ME (2001) Trophic relations in a community of
African rainforest carnivores. Oecologia 127:395–408.
doi:10.1007/s004420000604
Ray JC, Hunter LTB, Zigouris J (2005) Setting conservation and
research priorities for larger African carnivores. WCS Working
Paper 24. Wildlife Conservation Society, New York
Reed BC, Brown JF, VanderZee D, Loveland TR, Merchant JW, Ohlen
DO (1994) Measuring phenological variability from satellite im-
agery. J Veg Sci 5:703–714. doi:10.2307/3235884
Sankar K, Johnsingh AJT (2002) Food habits of tiger (Panthera tigris)
and leopard (P. pardus) in Sariska Tiger Reserve, Rajasthan,
India, as shown by scat analysis. Mammalia 66(2):285–289
Sankar K, Qureshi Q, Mondal K, Worah D, Srivastava T, Gupta S, Basu S
(2009) Ecological studies in Sariska Tiger Reserve, Report submit-
ted to National Tiger Conservation Authority. Govt. of India, New
Delhi and Wildlife Institute of India, Dehra Dun, p 145
Sathyakumar S (1992) Food habits of leopard (Panthera pardus)on
Mundanthurai plateau, Tamil Nadu, India. Tiger Paper 20:8–9
Schaller GB (1967) The deer and the tiger. University of Chicago Press
Schaller GB (1972) The Serengeti lion. University Chicago Press,
Chicago
Soberon J, Nakamura M (2009) Niches and distributional areas: con-
cepts, methods, and assumptions. Proc Natl Acad Sci U S A
106:19644–19650
Spong G, Hellborg L, Creel S (2000) Sex ratio of leopards taken in
trophy hunting: genetic data from Tanzania. Conserv Genet
1:169e171. doi:10.1023/A:1026543308136
Sunquist ME, Sunquist F (eds) (2002) Wild cats of the world. Univer-
sity of Chicago Press, Chicago
Tinoco BA, Astudillo PX, Latta SC, Graham CH (2009) Distribution,
ecology and conservation of an endangered Andean humming-
bird: the violet-throated metaltail (Metallura baroni). Bird Con-
serv Int 19:63–76. doi:10.1017/S0959270908007703
Tucker CJ, Sellers PJ (1986) Satellite remote sensing of primary
production. Int J Remote Sens 7:1133–1135. doi:10.1080/
01431168608948944
Walker RE, Stoms DM, Estes JE, Cayocca KD (1992) Relationships
between biological diversity and multi-temporal vegetation index
data in California. ASPRS ACSM, Albuquerque
Young BF, Franke I, Hernandez PA, Herzog SK, Paniagua L, Tovar C,
Valqui T (2009) Using spatial models to predict areas of endemism
and gaps in the protection of Andean slope birds. Auk 126:554–565
Acta Theriol (2013) 58:179–187 187