Acta Theriologica

, Volume 56, Issue 4, pp 335–342

Density of tiger and leopard in a tropical deciduous forest of Mudumalai Tiger Reserve, southern India, as estimated using photographic capture–recapture sampling


    • Wildlife Institute of India
  • Tharmalingam Ramesh
    • Wildlife Institute of India
  • Qamar Qureshi
    • Wildlife Institute of India
  • Kalyanasundaram Sankar
    • Wildlife Institute of India
Original Paper

DOI: 10.1007/s13364-011-0038-9

Cite this article as:
Kalle, R., Ramesh, T., Qureshi, Q. et al. Acta Theriol (2011) 56: 335. doi:10.1007/s13364-011-0038-9


Density of tiger Panthera tigris and leopard Panthera pardus was estimated using photographic capture–recapture sampling in a tropical deciduous forest of Mudumalai Tiger Reserve, southern India, from November 2008 to February 2009. A total of 2,000 camera trap nights for 100 days yielded 19 tigers and 29 leopards within an intensive sampling area of 107 km2. Population size of tiger from closed population estimator model Mb Zippin was 19 tigers (SE = ±0.9) and for leopards Mh Jackknife estimated 53 (SE = ±11) individuals. Spatially explicit maximum likelihood and Bayesian model estimates were 8.31 (SE = ±2.73) and 8.9 (SE = ±2.56) per 100 km2 for tigers and 13.17 (SE = ±3.15) and 13.01 (SE = ±2.31) per 100 km2 for leopards, respectively. Tiger density for MMDM models ranged from 6.07 (SE = ±1.74) to 9.72 (SE = ±2.94) per 100 km2 and leopard density ranged from 13.41 (SE = ±2.67) to 28.91 (SE = ±7.22) per 100 km2. Spatially explicit models were more appropriate as they handle information at capture locations in a more specific manner than some generalizations assumed in the classical approach. Results revealed high density of tiger and leopard in Mudumalai which is unusual for other high density tiger areas. The tiger population in Mudumalai is a part of the largest population at present in India and a source for the surrounding Reserved Forest.


Large felidsCamera trapsSpatially explicit capture–recapture modelsMudumalai Tiger Reserve


Large carnivore population estimates have always been influential in the allocation of resources for conservation efforts and for evaluating the success of conservation programs (Nowell and Jackson 1996; Karanth and Nichols 2000). Often, abundance has been a state variable of interest to wildlife managers assessing the impacts of management actions and it is commonly the central variable in long-term monitoring programs. Reliable estimates of status and population trends are critical for the conservation of large terrestrial carnivores as they play an important role in providing benchmark data for future management decisions. Over the years, camera trapping in a capture–recapture framework has proven to be a credible and practical technique for quantitative estimation of carnivore populations in varied habitats. In India, this method has also been used for estimating tiger densities in a number of protected areas (Sharma et al. 2010; Jhala et al. 2008; Karanth 1995; Karanth and Nichols 1998, 2000, 2002; Karanth et al. 2004, 2006). The use of camera traps to estimate tiger density was first proposed by Karanth (1995) and received further development by Karanth and Nichols (1998, 2000, 2002). Since then, it has been widely used on several individually identifiable carnivore species (O'Brien et al. 2003; Trolle and Kéry 2003; Karanth et al. 2004, 2006; Silver et al. 2004; Wegge et al. 2004; Soisalo and Cavalcanti 2006). Leopard densities achieved from photographic capture–recapture sampling method have been reported in only few parts of the country (Harihar et al. 2009; Edgaonkar et al. 2007; Sankar et al. 2008; Mondal 2006; Chauhan et al. 2005a).

Mudumalai Tiger Reserve (TR) is a typical tropical forest that was established to conserve an integral assemblage of large mammals with diverse habitats. Mudumalai TR, which is a part of the Western Ghats Complex, has large and contiguous forest tracts with the adjoining forest areas thereby forming a crucial linkage for genetic exchange between the Western Ghat animal populations and supporting one of the largest breeding populations (source populations) of tiger and leopard in India (Jhala et al. 2008). This forest possesses a high prey density (Ramesh et al. 2009) thereby having the potential to sustain high density source populations of tiger and leopard provided they are devoid of anthropogenic disturbance. Food habits of tiger and leopard in Mudumalai revealed their coexistence by feeding on a broad choice of prey species (Ramesh et al. 2009).

Our objective in this study was to determine the abundance of tiger and leopard in Mudumalai using camera traps and apply our findings as a basis for a long-term population monitoring program.

Study area

The study was carried out from November 2008 to February 2009 in a newly created Mudumalai Tiger Reserve (321 km2), which lies on the northeastern and northwestern slopes of the Nilgiris descending to the Mysore plateau and placed at the tri-junction of Kerala, Karnataka, and Tamil Nadu, India (Fig. 1). It is located between 11°32′ and 11°43′ N latitudes and 76°22′ and 76°45′ E longitudes (Suresh et al. 1996). The intensive study area (107 km2) comprises tropical moist and tropical dry deciduous forest forming the central region of the core zone, i.e, the notified national park. The intensive study area was chosen as it has the potential to hold high large felid densities due to the high prey biomass available (8,365 kg/km2, Ramesh et al. 2009). The terrain of the reserve is gently undulating with elevation ranging from 485 to 1,266 m (Varman and Sukumar 1995). The climate is moderate with a decreasing rainfall gradient from the west and south to the east and north (Venkataraman et al. 2005). It experiences cold weather during the month of December till the beginning of January and hot weather in March and April. According to Champion and Seth (1968), the vegetation is classified into southern tropical dry thorn forest, southern tropical dry deciduous forest, southern tropical moist deciduous forest, southern tropical semi-evergreen forest, moist bamboo brakes, and riparian fringe forest. Tiger Panthera tigris, leopard Panthera pardus, and Asiatic wild dog or dhole Cuon alpinus are sympatric large carnivores in the study area. The potential prey species of the tiger and leopard in Mudumalai TR are chital or spotted deer Axis axis, sambar Rusa unicolor, muntjac or barking deer Muntiacus muntjak, wild pig Sus scrofa, Asian chevrotain or mouse deer Moschiola meminna, and gaur Bos gaurus, among the ungulates. Asian elephant Elephas maximus, and sloth bear Melursus ursinus are well distributed throughout the park. Black-naped hare Lepus nigricollis, bonnet macaque Macaca radiata, common langur Semnopithecus entellus, Indian porcupine Hystrix indica, Malabar giant squirrel Ratufa indica, and peafowl Pavo cristatus are other prey species. Domestic livestock (cattle, buffalo, and goat) occur in village areas present inside the reserve.
Fig. 1

Locations of 20 camera trap sites for estimation of tiger and leopard densities in Mudumalai Tiger Reserve, southern India

Material and methods

Capture–recapture field sampling

Field work began with a preliminary reconnaissance to identify locations with indirect signs (pugmarks, tracks, scats, scrapes, rake marks, scent deposits, kills) indicative of frequent tiger and leopard movement and consequently identify appropriate camera trap sites as prescribed by Karanth and Nichols (2002) and Jhala et al. (2008). Potential Global Positioning System points of camera trap stations were then mapped using ArcView GIS 3.2 (ESRI, Redlands, CA, USA) and further plotted on the survey of India (R.F. 1:50,000) topographic map. We identified 20 camera trap sites and systematically placed cameras with a mean inter-camera trap distance of 2.38 km in order to cover the intensive study area without leaving any large gaps in the trap array which may result in missing any tiger or leopard. The size of the minimum convex polygon (MCP) formed by joining the peripheral camera locations was 107 km2 (Fig. 1). At each site, a pair of passive infrared analog camera traps DeerCam® DC300 (DeerCam, Park Falls, WI, USA) was placed opposite each other between 5 and 8 m from the center of the trail so as to photograph both flanks of tiger and leopard. The camera delay was set at a minimum of 15 s. Cameras were mounted in wooden boxes at a height of 35 cm, above the ground, perpendicular to the expected direction of animal movement. Cameras were deployed along forest roads, trails, stream beds, and near water holes. These cameras have heat and motion sensors to detect animal presence. Due to independent sensors in each camera, on some occasions we failed to get pictures of both flanks largely due to the delay in response of the sensor, which resulted in single flank pictures. Due to the difference in right and left flank pictures, we used the flank which yielded maximum unique individuals for abundance estimation. No bait or lure was used at any location to attract tigers or leopards. Trapping was done for 100 days from November 2008 to February 2009 amounting to a total of 2,000 trap nights (20 locations × 100 days). One trap night was a 24-h period during which a camera was functional. Owing to the good network of roads, all trapping sites were checked on a daily basis. Each camera unit was given a unique identification number (e.g., CT1A, CT1B,…, CT20A, CT20B) and each film roll was marked (e.g., CT 1A/Roll1, CT 1B/Roll1,…, CT20A/Roll1, CT20B/Roll1). Each camera unit recorded the date and time of the photograph.

Capture–recapture data analyses

Every tiger and leopard photographed was given a unique identification number after examining their unique stripe and rosette patterns on the flanks, limbs, and forequarters (Schaller 1967; McDougal 1977; Karanth 1995). Individual capture histories of tiger and leopard were developed in a standard “X-matrix format” (Otis et al. 1978; Nichols 1992) keeping 100 sampling occasions for both species. One critical assumption for the closed population estimate is that the population should be demographically and geographically closed (Otis et al. 1978; Rexstad and Burnham 1991). Therefore, we tested for population closure using program CloseTest 3 (Stanley and Burnham 1999). Capture data of tiger and leopard in “X-matrix format” (Otis et al. 1978; Nichols 1992) was analyzed using models developed for closed populations. In program CAPTURE (Rexstad and Burnham 1991), analysis was done for the right flank of tiger and left flank of leopard which yielded maximum number of unique individuals (Mt + 1). Due to different behavioral response to photo traps by tigers and leopards (Karanth and Nichols 1998; Wegge et al. 2004), we used the Mb Zippin estimator (Otis et al. 1978; Karanth 1995; Karanth and Nichols 1998) to estimate tiger population (N) while Mh Jackknife method (Otis et al. 1978) for assessment of leopard population. The appropriate model was selected based on discriminant function score (Rexstad and Burnham 1991). The density was estimated using likelihood-based spatially explicit capture–recapture (SECR) model in program DENSITY 4.4 (Efford 2009) and Bayesian-based SECR model using R-WinBugs code (Royle et al. 2009) and SPACECAP 1 (Singh et al. 2010). The likelihood-based SECR model has two components: state mode which describes the distribution of animal's home range centers in the landscape and observation model which describes capture probability as a function of the distance from the home range center to the trap (Pledger and Efford 1998; Efford 2009; Efford et al. 2004). In this analysis, Poisson distribution was assumed, buffer of 6 km was used, and proximity trap option was chosen which allowed for multiple captures on the same occasion. Half normal function was fitted to the distance between the home range center and trap. Program SPACECAP 1 was used to achieve spatial Bayesian estimate (Royle et al. 2009) with Bernoulli distribution and trap response absent for leopard and present for tiger as behavioral response was observed for tiger. Spatial Bayesian method incorporates data augmentation to deal with unobserved capture histories where data was augmented with more than double the estimated population and uniform priors were used (Royle and Dorazio 2008). We generated systematic home range centers in an area contained within 6 km buffer (larger than mean maximum distance moved, MMDM; Karanth and Nichols 1998) around camera traps. The large buffer around the sampled area was used to ensure inclusion of all individual home ranges within a reach of cameras (Royle and Dorazio 2008). In SECR models buffer width of 3–6 km varied till estimates stabilized, buffer and augmentation values were changed till these two parameters did not seem to cause changes in density estimates. The effective trapping area for traditional models was calculated by full MMDM and half mean maximum distance moved (MMDM/2) as described by Karanth and Nichols (1998, 2002).


Capture success

The total sampling effort of 2,000 trap nights for 100 days yielded 69 tiger photographs (44 right flanked and 25 left flanked) and 119 leopard photographs (56 right flanked and 63 left flanked). Nineteen individual tigers were identified from right flank photos and 16 individual tigers from left flank photos. Twenty-three individual leopards were identified from right flank photos and 29 individual leopards from left flank photos (Table 1).
Table 1

Summaries of spatially explicit capture–recapture model parameters for estimating tiger and leopard densities with camera trapping in Mudumalai Tiger Reserve, southern India



Best model

D ± SEa

Sigma ± SEb

g0/Lambda ± SEc

Psi ± SEd

N(X) ± SEe


Max likelihood

Half normal

8.31 ± 2.73

2.72 ± 0.52

0.005 ± 0.003


Mt + 1e = 19


Half normal

8.90 ± 2.56

0.6564 ± 0.259

0.005 ± 0.002

0.503 ± 0.149

65 ± 19


Max likelihood

Half normal

13.17 ± 3.15

2.19 ± 0.298

0.008 ± 0.002


Mt + 1e = 29


Half normal

13.01 ± 2.31

0.4114 ± 0.109

0.008 ± 0.002

0.739 ± 0.141

95 ± 17

aDensity of individuals/100 km2

bSpatial scale parameter

cDetection probability (frequentist) and lambda expected encounter frequency (Bayesian)—at trap location considered as home range center

dData augmentation parameter

ePopulation size of individuals having their activity centers within the effective trapping area

Statistical tests for population closure in program CloseTest 3 (Stanley and Burnham 1999) supported the population closure assumption both for tiger (χ2 = 29.99, P = 1.0) and leopard (χ2 = 43.03, P = 1.0).

Estimates of population sizes and densities

The capture probability (p-hat) of tiger was 0.037 for right flank data; population size (N) based on model Mb Zippin for right flank was 19 (SE = ±0.9). Leopard capture probability was 0.012 and population size based on model Mh Jackknife for left flank was 53.3 (SE = ±11, Table. 2). The average p-hat was low in comparison to other areas (Table 3). Spatially explicit maximum likelihood and Bayesian model density estimates were 8.31 (SE = ±2.73) and 8.9 (SE = ±2.56)/100 km2 for tigers and 13.17 (SE = ±3.15) and 13.01(SE = ±2.31)/100 km2 for leopards, respectively (Table 1). Tiger density for MMDM models ranged from 6.07 (SE = ±1.74) to 9.72 (SE = ±2.94)/100 km2 and leopard density ranged from 13.41 (SE = ±2.67) to 28.91 (SE = ±7.22)/100 km2, respectively (Table 2).
Table 2

Tiger and leopard abundance estimates and statistical parameters from conventional MMDM models in Mudumalai Tiger Reserve, southern India


Best model



N ± SEb (95% CI)

Mrd ± SEc


A(W)e (km2)

D ± SEf


Mb Zipping



19.0 ± 0.9 (19–24)

1.81 ± 0.43

3.86 ± 0.78


6.07 ± 1.74

Mt + 1 = 19





9.72 ± 2.94


Mh Jackknifei



53.3 ± 11.0 (39–85)

1.98 ± 0.43

3.44 ± 1.03


13.41 ± 2.67

Mt + 1 = 29





28.91 ± 7.22

aCapture probability

bPopulation size

cMean recapture distance (in kilometers)

dMean maximum distance moved (in kilometers)

eEffective trapping area (in square kilometers)

fDensity of individuals/100 km2

gTrap response model

hHalf mean maximum distance moved (in kilometers)

iIndividual heterogeneity model

Table 3

Trap nights (C), effective trapping area [A(W) km2], number of unique individuals (Mt + 1), capture probability (p-hat) population size with associated standard error (N ± SE), and density with associated standard error (D ± SE) of tiger and leopard from camera trap studies across India and parts of Southeast Asia

Study site


A(W) km2

Mt + 1


N ± SE

D ± SE



Mudumalai (PS)a





19 ± 0.9

9.72 ± 2.94


Pakke (Chauhan et al. 2006)





4 ± 2.56

1.15 ± 0.8


Chilla (Harihar 2005)





4 ± 0.08

3.01 ± 0.71







12 ± 1.97

3.27 ± 0.59







9 ± 1.93

3.42 ± 0.84







6 ± 1.41

4.94 ± 1.37


Ranthambore (Chauhan et al. 2005b)





21 ± 6.1

5.83 ± 2.01







24 ± 6.09

6.67 ± 1.85







29 ± 9.65

6.94 ± 3.23







20 ± 4.41

7.29 ± 2.54







28 ± 7.29

11.46 ± 4.20







29 ± 3.77

11.92 ± 1.71







34 ± 9.9

11.97 ± 3.71


Kanha (Sharma et al. 2010)





12 ± 0.59

12.53 ± 2.09







28 ± 4.51

16.76 ± 2.96


Corbett (Contractor 2007)




108 ± 4.56

19.20 ± 1.6


Mudumalai (PS)





53.3 ± 11

28.91 ± 7.22


Sariska (Sankar et al. 2008)





14.0 ± 0.2

7 ± 0.2


Satpura (Edgaonkar et al. 2007)





9 ± 2.6–14 ± 6.9

7 ± 2.1–10 ± 5.1


Chilla (Harihar et al. 2009)





13 ± 6.02

14.99 ± 6.9


Southeast Asia

Lao PDR (Johnson et al. 2006)





2 ± 0.66–4 ± 1.54



Bhutan (Wang and Macdonald 2009)





8 ± 2.12

0.52 ± 0.05


Malaysia (Kawanishi and Sunquist 2004)





5 ± 1.92–7 ± 2.44

1.1 ± 0.52–1.98 ± 0.77


Sumatra (O'Brien et al. 2003)





13 ± 3.66



Myanmar (Lynam et al. 2009)





2 ± 0.05–3 ± 1.54



Bhutan (Wang and Macdonald 2009)





16 ± 2.91

1.04 ± 0.01


Thailand (Simcharoen and Duangchantrasiri 2008)





10 ± 1.76–11 ± 2.48

4.86 ± 2.29–7.88 ± 5.82


T tiger, L leopard

aPresent study

bKaranth et al. 2004

In spatially explicit models, we observed a significant difference in density estimate with different buffer width. In Bayesian method augmentation size also affected density estimate. The population augmentation in Bayesian method was done till the psi value and its upper 97.5th quantile was away from the value of 1. We found that population augmentation by twice the estimated population is a good starting point to evaluate the condition of psi. We analyzed data for SECR models with buffer increment of 1 km and found that buffer width equivalent to MMDM is the best guess to start with as density estimates start stabilizing from that buffer size. Tiger and leopard densities stabilized at 6 km buffer width. In Bayesian method, tiger density with 4 km buffer was 15.2/100 km2 and at 6 km buffer was 8.91/100 km2, i.e, 59% difference and leopard density had shown 63% difference from 4 km (20.66/100 km2) to 6 km (13.01/100 km2) buffer. In the case of likelihood approach, this difference was 5% for tiger and 4% for leopard at similar buffer widths. Bayesian analysis had a longer computational time (14–28 h on SPACECAP 1or R-WinBugs code) than the likelihood approach (20–50 min in DENSITY 4.4 and R-SECR) on the computer having Intel Core 2 processor with 4 GB of RAM.


We have implemented likelihood-based (Efford et al. 2004) and Bayesian (Royle et al. 2009) SECR models along with conventional MMDM methods in this paper. The conventional MMDM methods were used here for comparison with earlier studies. MMDM models were found to be inconsistent in estimating density in comparison to the spatial models. The SECR methods handle the capture information by modeling the capture distance and are therefore better in estimating effective area under which population size and density are estimated than classical methods where defining the effective area is difficult and density estimates are not reliable. Although SECR models provide more reliable results than conventional MMDM methods, researchers should be aware that the Bayesian methods are extremely sensitive to the buffer width and augmentation size. Likelihood estimates are comparatively less sensitive to buffer width as the former approach is based directly on parameters estimating density unlike Bayesian, which derives density estimates from the likely population capturable in the given buffer width.

For simpler density analysis, the likelihood approach seems to be appropriate as it is faster and both spatial methods have not shown any significant difference in terms of density estimation. We feel that the Bayesian approach might have an advantage in studies which will model encounter history data conditioned on the covariates in space or time.

Considering the high prey biomass (8,365 kg/km2) available in Mudumalai TR (Ramesh et al. 2009), estimates of tiger and leopard density arrived from spatial estimators appear to be ecologically realistic (Carbone and Gittleman 2002; Karanth et al. 2004). The density estimates for tiger and leopard are comparable with other areas in the country (Table 3) and are indicative of habitat productivity and importance of this protected area for conservation of both predators. One of the drawbacks of this study was the smaller trapping area though the effective sample area is the median value of studies carried out in India (Table 3). It will be appropriate to have a larger trapping area (Karanth and Nichols 1998; Jhala et al. 2008) like in recent studies where effective trapping area was >300 km2 for example in Melghat, Corbett, Panna, Tadoba, and Bandipur (Table 3). Since previous studies reported densities from MMDM/2 method, we compared these values for tiger and leopard from this study to those reported in other parts of the species distribution range (Table 3). Tiger density in various parts of the country ranged from 2.0 to 20 individuals per 100 km2 (Karanth et al. 2004; Jhala et al. 2008; Harihar 2005); therefore, the density observed in Mudumalai TR was relatively high (Table 3). Similarly, leopard density determined in our study was also high in comparison to other reported estimates in the country (Sankar et al. 2008; Edgaonkar et al. 2007; Harihar et al. 2009; Table 3). Leopard densities reported from other parts of the world (e.g., 3.8–4.5/100 km2 in Serengeti, Schaller 1972; 3.4 in Kruger, Pienaar 1969; and 3.4 in Wilpattu National Park, Sri Lanka, Eisenberg and Lockhart 1972) did not use formal population sampling approaches thereby lacking comparable estimates for making statistical inferences (Karanth and Nichols 1998). Leopard density is fairly high in spite of good tiger density and this trend (Qureshi, unpublished data) seems to be due to high prey biomass of cervids (Ramesh et al. 2009) and heterogeneous landscape, which provide opportunity for coexistence of these two large predators at higher densities.

Mudumalai TR is a part of Bandipur–Nagarahole–Mudumalai–Wynad landscape unit with an estimated tiger population of 267 (207–327) individuals occupying 9,087 km2 (Jhala et al. 2008) and harbors probably the largest continuous tiger population in the world. The state of Tamil Nadu also has a large leopard population distributed within 14,484 km2 (Jhala et al. 2008). Mudumalai TR being the most important conservation unit linking western and eastern parts of Western Ghats requires habitat protection along with regular monitoring of large carnivores and their prey population with the incorporation of robust scientific methods. This study has provided important baseline information for long-term monitoring of tiger and leopard in Mudumalai TR which is one of the few areas where both predators occur in good densities. Future studies should aim at understanding the factors governing high density of both felids.


We thank the Tamil Nadu Forest Department for granting permission to conduct this study in Mudumalai Tiger Reserve. This research was undertaken as part of “Sympatric carnivore studies,” funded by Wildlife Institute of India. We sincerely thank the director and dean, Wildlife Institute of India, for their support and encouragement in carrying out this study. We are grateful to our field assistants Madan, James, and Ketan for all their hard work and whole-hearted cooperation in the field. We thank the two anonymous reviewers and editors for their critical review and suggestions which significantly improved this manuscript.

Copyright information

© Mammal Research Institute, Polish Academy of Sciences, Białowieża, Poland 2011