<%BANNER%>

Ecology of the Leopard (Panthera pardus) in Bori Wildlife Sanctuary and Satpura National Park, India

Permanent Link: http://ufdc.ufl.edu/UFE0019601/00001

Material Information

Title: Ecology of the Leopard (Panthera pardus) in Bori Wildlife Sanctuary and Satpura National Park, India
Physical Description: 1 online resource (135 p.)
Language: english
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: carnivore, conservation, dhole, ecology, india, leopard, panthera, pardus, tiger, wildlife
Wildlife Ecology and Conservation -- Dissertations, Academic -- UF
Genre: Wildlife Ecology and Conservation thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The ecology of the leopard (Panthera pardus) was studied from 2002 to 2006 in the Bori Wildlife Sanctuary and Satpura National Park in Madhya Pradesh, India. Density estimates of the potential prey species of leopards and its sympatric carnivores, the tiger (Panthera tigris) and the dhole (Cuon alpinus) were made using the line-transect method annually from 2002 to 2005, and for three habitat types. The results obtained by vehicle transects were compared with those of foot transects for obtaining reliable density estimates. The food habits and prey preference of leopards, tigers and dholes were quantified. Leopard density estimates for three sites in Bori Satpura and one site in Rajasthan, the Sariska Tiger Reserve, were made using camera traps and the mark-recapture method. A predictive habitat suitability map for leopards using Environmental Niche Factor Analysis (ENFA) was made at two scales and its reliability was evaluated. The environmental variables important in describing the habitat for leopards were identified and the extent and location of potential leopard habitat available for conservation action in south-central Madhya Pradesh was quantified. Chital (Axis axis) density was higher in the moist deciduous and teak dominated habitats compared to the dry deciduous habitat. Sambar (Cervus unicolor) density was higher in the teak dominated habitat. The densities of nilgai (Boselaphus tragocamelus), wild pig (Sus scrofa) and muntjac (Muntiacus muntjak) for the three habitat types were not statistically different. Annual density was lower for all prey species in 2005 as compared to 2002. Sambar was the most important prey species in the leopard?s diet. It was also the most preferred prey species by leopards, as well as by tigers and dholes. Density of leopards was estimated at 7.3, 7.5, 8.0 and 9.3 per 100 km2 for the four samples in Satpura Tiger Reserve using the half MMDM method and 4.2, 4.6, 5.3 and 6.2 per 100 km2 for the full MMDM method. The estimates for the sampled area in Sariska Tiger Reserve using the two methods were 30.9 and 20.7 per 100 km2, respectively. The results of the ENFA model showed that habitat use by leopards in Satpura was strongly associated with moist and teak forests, as well as with most prey species and was weakly negatively associated with the distance to villages. At the larger scale, in south-central Madhya Pradesh, leopard habitat was positively associated with terrain ruggedness, sambar availability and percentage of forested areas. Approximately 11500 km2 of habitat in south-central Madhya Pradesh is likely to support leopard populations. The districts with the most optimal habitat were found to be Betul, Hoshangabad and Chhindwara, which have about 2000 km2 of contiguous habitat for leopard conservation.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis: Thesis (Ph.D.)--University of Florida, 2008.
Local: Adviser: Sunquist, Melvin E.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0019601:00001

Permanent Link: http://ufdc.ufl.edu/UFE0019601/00001

Material Information

Title: Ecology of the Leopard (Panthera pardus) in Bori Wildlife Sanctuary and Satpura National Park, India
Physical Description: 1 online resource (135 p.)
Language: english
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: carnivore, conservation, dhole, ecology, india, leopard, panthera, pardus, tiger, wildlife
Wildlife Ecology and Conservation -- Dissertations, Academic -- UF
Genre: Wildlife Ecology and Conservation thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The ecology of the leopard (Panthera pardus) was studied from 2002 to 2006 in the Bori Wildlife Sanctuary and Satpura National Park in Madhya Pradesh, India. Density estimates of the potential prey species of leopards and its sympatric carnivores, the tiger (Panthera tigris) and the dhole (Cuon alpinus) were made using the line-transect method annually from 2002 to 2005, and for three habitat types. The results obtained by vehicle transects were compared with those of foot transects for obtaining reliable density estimates. The food habits and prey preference of leopards, tigers and dholes were quantified. Leopard density estimates for three sites in Bori Satpura and one site in Rajasthan, the Sariska Tiger Reserve, were made using camera traps and the mark-recapture method. A predictive habitat suitability map for leopards using Environmental Niche Factor Analysis (ENFA) was made at two scales and its reliability was evaluated. The environmental variables important in describing the habitat for leopards were identified and the extent and location of potential leopard habitat available for conservation action in south-central Madhya Pradesh was quantified. Chital (Axis axis) density was higher in the moist deciduous and teak dominated habitats compared to the dry deciduous habitat. Sambar (Cervus unicolor) density was higher in the teak dominated habitat. The densities of nilgai (Boselaphus tragocamelus), wild pig (Sus scrofa) and muntjac (Muntiacus muntjak) for the three habitat types were not statistically different. Annual density was lower for all prey species in 2005 as compared to 2002. Sambar was the most important prey species in the leopard?s diet. It was also the most preferred prey species by leopards, as well as by tigers and dholes. Density of leopards was estimated at 7.3, 7.5, 8.0 and 9.3 per 100 km2 for the four samples in Satpura Tiger Reserve using the half MMDM method and 4.2, 4.6, 5.3 and 6.2 per 100 km2 for the full MMDM method. The estimates for the sampled area in Sariska Tiger Reserve using the two methods were 30.9 and 20.7 per 100 km2, respectively. The results of the ENFA model showed that habitat use by leopards in Satpura was strongly associated with moist and teak forests, as well as with most prey species and was weakly negatively associated with the distance to villages. At the larger scale, in south-central Madhya Pradesh, leopard habitat was positively associated with terrain ruggedness, sambar availability and percentage of forested areas. Approximately 11500 km2 of habitat in south-central Madhya Pradesh is likely to support leopard populations. The districts with the most optimal habitat were found to be Betul, Hoshangabad and Chhindwara, which have about 2000 km2 of contiguous habitat for leopard conservation.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis: Thesis (Ph.D.)--University of Florida, 2008.
Local: Adviser: Sunquist, Melvin E.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0019601:00001


This item has the following downloads:


Full Text





ECOLOGY OF THE LEOPARD (Panthera pardus) IN BORI WILDLIFE SANCTUARY AND
SATPURA NATIONAL PARK, INDIA





















By

ADVAIT EDGAONKAR


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2008

































O 2008 Advait Edgaonkar






























To Aai, Baba and Vinatha.









ACKNOWLEDGMENTS

I sincerely thank my supervisor Dr. Melvin Sunquist for his guidance and patience

throughout the long period of fieldwork, for raising the funds and equipment to make the

research possible, and for going through numerous drafts of the dissertation. I would like to

acknowledge the help of my committee members Dr. Lyn Branch, Dr. Madan Oli and

Dr. Michael Binford for their help during the dissertation writing process. Dr. Kenneth Portier,

my former committee member, helped with statistics as did Dr. Mike Moulton. Delores Tilman

Caprice MacRae and Claire Williams helped with administration at the WEC, and I would like to

thank them for it.

I am grateful for the logistics support of Dave Ferguson, Fred Bagley and Mini Nagendran

at USFWS and from Beena Achankunju and Priya Ghosh at the US Embassy in New Delhi. I am

thankful to Jim Nichols and Murray Efford for valuable advice.

I would like to thank the Wildlife Institute of India, the Alumni Fellowship, the Disney

Conservation Fund and the Jennings Scholarship for funding the fieldwork in India and course

work at the University of Florida. I am grateful to Dr. Ravi Chellam who was the principal

investigator of the proj ect at WII till 2002 and to Qamar Qureshi, who took over responsibility

afterwards. I thank Dr A.J.T. Johnsingh, Dr Y.V. Jhala, Dr. K Sankar, Dr Manoj Agarwal and

Shri Vinod Thakur who helped in various ways. I would very much like to thank Ravi Kailas and

Dilip Venugopal, Deep Contractor, Bindu Raghavan, Vidya Athreya, Anirudh Belsare, Rashid

Raza, K. Ramesh, Bhaskar Acharya, Raj ah Jayapal, Abishek Harihar, Shomita Mukherj ee,

Meena Venkataraman, Priya Balasubramaniam and Gopi Sundar for help with data collection,

fruitful discussions, encouragement, support, and for their friendship. My friends Bimal Desai,

A.V.S. Prasad and Chirag Wazir contributed in numerous ways and to them I am thankful.









Nilmini Jayasena deserves special acknowledgment for her generous help during a critical phase

of the dissertation writing.

I am grateful to the Forest Department of Madhya Pradesh for their support. I would like to

thank Shri O. P. Tiwari, Shri Madhukar Chaturvedi, Shri Ramachandran, Shri S.P. Singh, Shri S.

S. Rajpoot, Shri L. K. Gupta, Shri N. D. Sharma, Shri H. P. Singh, Shri Rajeev Srivastay, Shri P.

M. Lad and Shri L.K. Chaudhary. I would especially like to thank Shri Sandeep Fellows and Shri

and Smt. Ramachandran for all their help and support in Itarsi.

I would like to thank Ramesh Yaday, Shyam Yaday, Bicchu Oju, Santosh Guttu, Gopal,

Botu and Bishram for help in the field. I would like to acknowledge my family, Vikram and

Ashok Shrotriya, and Shailaja Supekar, and Vijaya and G. Viswanathan for their support. I am

especially grateful to my parents, Jayant and Vasundhara Edgaonkar for their patience, love and

support over the years. Lastly, I thank my wife, Vinatha Viswanathan, for encouragement, ideas,

data entry, funding, companionship, help and love. This research is as much hers as it is mine.












TABLE OF CONTENTS


page

ACKNOWLEDGMENT S .............. ...............4.....


LI ST OF T ABLE S ............ ....._.. ...............9.....


LIST OF FIGURES ............_. ...._... ...............11....


AB S TRAC T ............._. .......... ..............._ 13...


CHAPTER


1 INTRODUCTION ................. ...............15.......... ......


2 DENSITY ESTIMATION OF POTENTIAL PREY SPECIES OF LARGE
CARNIVORES IN SATPURA TIGER RESERVE USING LINE-TRANSECT
SAMPLING. .............. ...............20....


Introducti on ................. ...............20.................
M ethods .............. ...............22....

Study Area ................. ........... ...............22.......
Estimation of Habitat Types ................. ...............23................
Estimation of Prey Density ................. ...............23................
Vehicle Transects .............. ...............25....
Re sults ................ ................. ...............25.......
Description of Habitat ................. ...............25................
Estimation of Density ................. ...............26................
Discussion ................. ...............27.................


3 PREY SELECTION AND THE FOOD HABITS OF TIGER, LEOPARD AND
DHOLE IN SATPURA TIGER RESERVE. .....__.....___ ..........__ ...........4


Introducti on ............. ...... ._ ...............41...
M ethod s .............. ...............42....

Study Area .............. ....___ ..... .............4
Reconstruction of Carnivore Diets .............. ...............43....

Sample Size Adequacy ................. ...............43................
Prey Biomass and Number ................. ...............44.......... .....
Estimation of Prey Selection .............. ...............44....
Dietary Overlap .............. ...............47....
R e sults................... ....... .. ..... ......... .............4
Density of Potential Prey Species............... ...............48
Sample Size Adequacy ................. ...............48................
Composition of Diet ................. ...............48................












Prey Selection ................. ...............49.................
Diet Overlap .............. ...............50....
Discussion ................. ...............50.................


4 ESTIMATION OF LEOPARD (Panthera pardus) ABUNDANCE IN INDIAN
FOREST S USING CAMERA TRAP S IN A MARK-RECAPTURE FRAMEWORK. ........64


Introducti on ................. ...............64.................
M ethods .............. ...............65....

Study Area ................. ...............65.................
Field Methods ................. ...............66.................

Analytical Methods .............. ...............67...
Estimation of population size ................. ...............67................
Estimation of leopard density............... ...............69
Re sults.................. .. ......... ...............70.......

Adequacy of Sampling .............. ...............70....
Sex Ratios ................. ...............71.................

Population Size ................. ...............71.................
Leopard Density .............. ...............71....
Discussion ................. ...............72.................
Conclusion ................ ...............75.................


5 PRESENCE-ONLY HABITAT SUITABILITY MODELS FOR LEOPARDS

(Panthera pardus) USING FIELD BASED AND REMOTELY DERIVED
VARIABLES AT TWO SPATIAL SCALES IN MADHYA PRADESH, INDIA. ........._....88


Introducti on ................. ...............88.................

Study Areas............... ...............89.
M ethods .............. ...............90....
R e sults................... ...............95.......... ......
Model Validation............... ...............9
Extensive Study Area .............. ...............95....
Satpura Tiger Reserve .............. ...............96....
Effect of Changing Resolution .............. ...............97....
Discussion ................. ...............97.................


6 CONCLU SION................ ..............11


Density of Potential Prey ................. ...............118......... .....
Preference of Prey ................. ...............118................
Density of Leopards ................. ...............119................
Habitat M odel ................. ...............119......... ......


APPENDIX A


INDICES OF UNGULATE AND CARNIVORE ABUNDANCE.............__ .........___.......121


LIST OF REFERENCES ................. ...............123................












BIOGRAPHICAL SKETCH ................. ...............135......... ......











LIST OF TABLES


Table page

2-1 Density of dominant tree species in the three habitat types along transects ................... ...30

2-2 Estimation of density parameters of potential prey by the line-transect method in the
moist deciduous habitat. ....__................. ........__. ........3

2-3 Estimation of density parameters of potential prey by the line-transect method in the
dry deciduous habitat. .............. ...............32....

2-4 Estimation of density parameters of potential prey by the line-transect method in the
teak dominated habitat. ............. ...............33.....

2-5 Estimation of overall density and its associated parameters by the line-transect
method over 4 years in the study area. .............. ...............34....._._._ ..

2-6 Density of wild ungulates at various study sites in India ................. .......__. ........._.3 5

2-7 Estimation of density parameters of potential prey by the vehicle transects assuming
Poisson variance............... ...............3

3-1 Estimation of overall density and its associated parameters by the line transect
method over 4 years in the study area ................. ...............54..............

3-2 Food habits of the leopard obtained by scat analyses ................. ......... ................55

3-3 Food habits of the tiger obtained by scat analyses ................. ...............56.............

3-4 Food habits of the dhole obtained by scat analyses. ................ ................ ......... .56

3-5 Jacobs' index values of preference for prey species in tiger diets at study sites in
India. ............. ...............57.....

3-6 Jacobs' index values of preference for prey species in leopard diets at study sites in
India. ............. ...............57.....

3-7 Jacobs' index values of preference for prey species in dhole diets at study sites in
India. ............. ...............58.....

3-8 Diet overlap between tiger, leopard and dhole using Pianka' s index. .............. ..... ........._.58

4-1 Camera-trapping effort at the study sites. .............. ...............77....

4-2 Leopard sex ratios for the different study sites. .....__.___ .... ... ._._ ... .._..._.......7

4-3 Model selection criterion and tests for Models Mo, Mh, Mb and Mt in the mark-
recapture framework and a test for population closure. .....__.___ ........___ ..............78











4-4 Population estimates for leopards at the study sites............... ...............79.

4-5 Density of leopards and estimates of sampled area using convex polygon and model
M h at the different study sites. ............. ...............79.....

4-6 Density of leopards with the associated estimated trapping area using models Mo and
M h.. . ...............80

4-7 Density estimates for leopards using different capture functions for the null models
with the MLSECR method. ..........._ ........... ............... 1....

4-8 Relative abundance index values for the 5 estimates in Satpura and Sariska Tiger
Reserves. ............. ...............82.....

5-1 Districts, sampling effort and leopard presence in the extensive study area in south-
central Madhya Pradesh ................. ...............100................

5-2 List of ecogeographical variables (EGV) with explanation and source for south-
central Madhya Pradesh. ....___................. ...............101 .....

5-3 List of ecogeographical variables (EGV) with explanation and source for the Satpura
Tiger Reserve. .............. ...............102....

5-4 Measures of evaluation for habitat models at different pixel resolutions (with cross-
validated standard deviations)............... ..............10

5-5 Correlation between ENFA factors and EGV for south-central Madhya Pradesh.. ........104

5-6 Correlation between ENFA factors and EGV for Satpura Tiger Reserve. ................... ...105

5-7 Area under various leopard-habitat categories in south-central Madhya Pradesh...........1 06

A-1 Kilometric index values of selected species using dirt trails in the monsoon from
2002 to 2005. ............. ...............121....

A-2 Encounter rates of tracks of selected carnivore species ................. ................ ...._.122










LIST OF FIGURES


Figure page

2-1 Map of the study area, showing the habitat types, transects and vehicle transects
trails ........._ ...... ...............36....

2-2 Annual densities of selected species in the study area............... ...............37..

2-2 Continued. .............. ...............38....

2-3 Comparison of density estimates between foot transects and vehicle transects in 2005
for potential prey species of large carnivores in Satpura Tiger Reserve. ..........................39

2-4 Detection function curves for vehicle and foot transects for chital, sambar, langur
and peafowl. .............. ...............40....

3-1 Map of Bori Wildlife Sanctuary and Satpura National Park, showing the location of
line transects, dirt roads and the study area. ............. ...............59.....

3-2 Relationship between sample size of scats and the percent frequency of occurrence in
tiger, leopard and dhole diet of langur, chital and sambar ................. .......................60

3-3 Relationship between the number of scats analyzed and the number of prey species
found in the diet of tiger, leopard and dhole ................. ...............61........... .

3-4 Prey taken by tiger, leopard and dhole in various body weight categories........................62

3-5 Observed and expected frequencies of prey items in scats of tiger,1eopard and dhole. ....63

4-1 Identification of leopards based on spot patterns. ........._.._.. ...._... ....._.._.......8

4-2 Rate of accumulation of new individuals in camera-trap photographs with increase in
sampling time at the four sites. ............. ...............84.....

4-3 Camera trapping in 3 sites (Churna, Kamti and Lagda) in Satpura Tiger Reserve. ..........85

4-4 Map showing camera trap locations with half MMDM and full MMDM buffers in
Sariska Tiger Reserve. ............. ...............86.....

4-5 Map showing camera trap locations with half MMDM and full MMDM buffers for
one site (Kamti)............... ...............87

5-1 Cover map of the study area in south-central Madhya Pradesh. ..........._.................107

5-2 Mosaicked landsat satellite image of the study area in south-central Madhya Pradesh..108

5-3 Landsat satellite image of the study area in Satpura Tiger Reserve. ............. ..... ........._.109












5-4 Maps of remotely derived variables for south-central Madhya Pradesh.. ................... .... 1 10

5-4 Continued ................. ...............111................


5-4 Continued ................. ...............112................


5-5 Maps of remotely derived variables for Satpura Tiger Reserve.. ................ .................1 13

5-5 Continued ................. ...............114................


5-5 Continued ................. ...............115................


5-6 Leopard habitat suitability map for south-central Madhya Pradesh. ............. .... ...........116


5-7 Leopard habitat suitability map for Satpura Tiger Reserve ................. .....................116


5-8 The predicted-to-expected frequency curves with habitat suitability values for both
the model s. ................ ............. ............ 117...









Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

ECOLOGY OF THE LEOPARD (Panthera pardus) IN BORI WILDLIFE SANCTUARY AND
SATPURA NATIONAL PARK, INDIA

By

Advait Edgaonkar
May 2008

Chair: Melvin Sunquist
Major: Wildlife Ecology and Conservation

The ecology of the leopard (Panthera pardus) was studied from 2002 to 2006 in the Bori

Wildlife Sanctuary and Satpura National Park in Madhya Pradesh, India. Density estimates of

the potential prey species of leopards and its sympatric carnivores, the tiger (Panthera tigris) and

the dhole (Cuon alpinus) were made using the line-transect method annually from 2002 to 2005,

and for three habitat types. The results obtained by vehicle transects were compared with those

of foot transects for obtaining reliable density estimates. The food habits and prey preference of

leopards, tigers and dholes were quantified. Leopard density estimates for three sites in

Bori-Satpura and one site in Raj asthan, the Sariska Tiger Reserve, were made using camera traps

and the mark-recapture method. A predictive habitat suitability map for leopards using

Environmental Niche Factor Analysis (ENFA) was made at two scales and its reliability was

evaluated. The environmental variables important in describing the habitat for leopards were

identified and the extent and location of potential leopard habitat available for conservation

action in south-central Madhya Pradesh was quantified.

Chital (Axis axis) density was higher in the moist deciduous and teak dominated habitats

compared to the dry deciduous habitat. Sambar (Cervus unicolor) density was higher in the teak

dominated habitat. The densities of nilgai (Boselaphus tragocamnelus), wild pig (Sus scrofa) and









muntj ac (M~untiacus muntjak) for the three habitat types were not statistically different. Annual

density was lower for all prey species in 2005 as compared to 2002. Sambar was the most

important prey species in the leopard's diet. It was also the most preferred prey species by

leopards, as well as by tigers and dholes. Density of leopards was estimated at 7.3, 7.5, 8.0 and

9.3 per 100 km2 for the four samples in Satpura Tiger Reserve using the half MMDM method

and 4.2, 4.6, 5.3 and 6.2 per 100 km2 for the full MMDM method. The estimates for the sampled

area in Sariska Tiger Reserve using the two methods were 30.9 and 20.7 per 100 km2,

respectively. The results of the ENFA model showed that habitat use by leopards in Satpura was

strongly associated with moist and teak forests, as well as with most prey species and was

weakly negatively associated with the distance to villages. At the larger scale, in south-central

Madhya Pradesh, leopard habitat was positively associated with terrain ruggedness, sambar

availability and percentage of forested areas. Approximately 1 1500 km2 Of habitat in south-

central Madhya Pradesh is likely to support leopard populations. The districts with the most

optimal habitat were found to be Betul, Hoshangabad and Chhindwara, which have about

2000 km2 Of contiguous habitat for leopard conservation.









CHAPTER 1
INTTRODUCTION

Humans have always been fascinated by carnivores, and our responses to them, whether

positive or negative, have been strong and emotional. Partly as a result of their food habits,

which have placed them in direct competition with us, partly because of their need for large

undisturbed areas, and for their valuable body parts, carnivores have been persecuted for many

centuries now. As a result the geographic ranges of many species have contracted, and their

populations have crashed. There is an urgent need to conserve many carnivore species, and the

first step towards this is to obtain knowledge about their basic biology: how many exist, what

they eat and where they live. There has been very little research done on most of the 37 extant

cat species of the world. This is because cats generally tend to be nocturnal, occur at low

densities, and live in remote locations. They are thus difficult and expensive to study.

The leopard has had the reputation of being one of the least studied of the large carnivores

despite being the most abundant (Hamilton 1976). The situation is hardly different even now, in

the Indian context. Most of the studies on leopards have been done in Africa (Bailey 1993;

Bertram 1982; Hamilton 1976; Jenny 1996) The sparse information on leopards in the Indian

subcontinent has mostly come from studies that focused on the tiger (Karanth & Sunquist 1995,

2000; Sunquist 1981) or the lion (Chellam 1993).

Based on estimates of density and geographic range the leopard's total effective global

population size has been estimated at greater than 50000 breeding individuals, and is listed as a

species of least concern by the IUCN red list. In India, however, it is listed in Schedule I of the

Indian Wildlife (Protection) Act, 1972, under the highest level of protection. This is because

habitat destruction, loss of wild prey, poaching for skins, bones and claws, and poisoning

carcasses of livestock killed by leopards are a significant threat to the species.









The leopard is a large sized cat, weighing on an average 58 kg for males and 37.5 kg for

females (Bailey 1993). It is the most widely distributed of the wild cats (Nowell & Jackson

1996), and is found in almost every kind of habitat, from the rainforests of the tropics to desert

and temperate regions (Kitchener 1991). It occurs from Africa through most of Asia up to the

Amur valley in Russia. The Indian subspecies, Panthera parduus fuca is found in all forested

habitats in the country, absent only in the arid deserts and above the timber line in the Himalayas

(Prater 1980). The leopard is quite adaptable with respect to habitat and food requirements,

being found in intensively cultivated and inhabited areas as well as near urban development

(Nowell & Jackson 1996). There is a wide variation in the ecology of the species across its range

and in different ecosystems.

Leopards have been found to be essentially solitary and territorial animals. They are most

likely to socialize at the carcass of large prey (Hamilton 1976). In Wilpattu, Sri Lanka, the only

social groupings seen were mother with cubs and courting pairs (Eisenberg & Lockhart 1972). In

Ruhuna National Park, also in Sri Lanka the majority of leopards observed were solitary

(Santiapillai et al. 1982). Schaller (1976) observed pairs only in three instances out of a total of

155 observations, the rest of which were of solitary leopards.

Scent marking has been conj ectured as the primary mode of communication. This

includes scraping, marking with feces and spraying of urine, which have been found in tigers to

be used most often along trails and trail intersections that serve as common boundaries between

territories (Smith et al. 1989). Communication has been speculated to serve several functions,

chief among which are to allow leopards to separate themselves in space and time, to attract the

opposite sex during courtship, and to distinguish each other by age, sex and individual status

(Bailey 1993).









Home ranges in leopards have been found to vary from being exclusive or slightly

overlapping to completely overlapping between the sexes. In Nepal, for example, the home range

of a male leopard enclosed the home ranges of several females (Seidensticker 1976), while in

Wilpattu, areas were being used exclusively by a single male and a single female (Muckenhirn &

Eisenberg 1973). Male leopards had slightly overlapping home-ranges in Thailand (Rabinowitz

1989). In Kruger National Park, South Africa, little spatial overlap between home ranges of adult

male leopards in summer has been observed and this decreased even further during the wet

season (Bailey 1993). Female home ranges also overlapped a little, while male home ranges

completely overlapped many female home ranges, as in the Nepal study (Seidensticker 1976).

Female home ranges appeared to be related to availability of prey needed to successfully raise

young ones. The juveniles share female home ranges until maturity after which they disperse and

become transient until they can find a suitable undefended portion of habitat. They can then

establish and defend a home range (Eisenberg 1986). In Asia, leopard home ranges have been

reported from Sri Lanka, Nepal and Thailand. In Wilpattu, home ranges of four leopards were

recorded as between 8 and 10.5 km2 (Muckenhirn & Eisenberg 1973), while female home ranges

between 6 and 13 km2 in Nepal (Seidensticker 1976) Thailand male leopards had ranges of 27-37

km2 and female ranges were between 11-17 km2 (Rabinowitz 1989). The low densities of

terrestrial herbivores found in rainforests may not be able to support high leopard densities.

Home range of a male leopard in wet evergreen forest in Ivory Coast was found to be 86 km2,

with partially overlapping female ranges that were up to three times smaller than the male home

range (Jenny 1996). The highest densities recorded in Kruger were 1 per 3.3 km2 where prey

biomass varied from 2932 to 6186 kg/km2 (Bailey 1993) while a crude density of 1 per 29 km2

for the entire park has been suggested (Pienaar 1969). For the Serengeti the density of leopards









was estimated at about 1 per 22 to 26.5 km2 (Schaller 1976). In Wilpattu National Park the

estimated density was 1 per 29 km2 (Muckenhirn & Eisenberg 1973). There are no published

density estimates for India.

Leopards have been shown to kill medium-sized prey, mainly impala (Aepyceros

melampus), but also take a very wide variety of small animals including hyrax, civet and

mongoose in Kruger National Park in South Africa (Bailey 1993). A wide spectrum of the

potential prey available in the Tai National Park, Ivory Coast, with about thirty species recorded

(Hoppe-Dominik 1984). In the Kalahari desert leopards have been known to take small prey like

Bat-eared foxes (Otocyon megalotis), jackals (Canzis spp), genets (Genetta spp), hares (Lepus

spp), duiker (Cephalopus spp) and porcupines (Hystrix spp) (Bothma & Le Riche 1984).

In Wilpattu leopards took chital, wild pig (Sus scrofa), sambar, langur, hare, porcupine

and domestic buffalo calves (Muckenhirn & Eisenberg 1973). In Nepal, wild pig sambar, chital,

hog deer (Axis porcinus), muntj ac and domestic cattle were part of their diet (Seidensticker et al.

1990). In the Pakistan Himalayas, leopards took mainly wild goats (Capra aegagrus) but also

preyed on livestock, hare and porcupine (Schaller 1977). In India too dietary studies have found

that leopards take a range of prey. In the Shivalik hills of Raj aji National Park analysis of scats

has shown that leopards eat chital, sambar, muntj ac, goral and livestock (Johnsingh pers comm).

In Sariska Tiger Reserve leopard scats contained rodents (Sankar & Johnsingh 2002). The

leopards on the Mundanthurai plateau have been preying mainly on sambar (Sathyakumar 1992)

while in Bandipur the leopard kills were mainly chital (Andheria et al. 2007; Johnsingh 1983). In

Gir, 40 percent of leopard scats contained chital remains while langur remains were found in

25% of the scats (Chellam 1993). Near Mumbai, leopards living near urban areas survive to a

large extent on domestic dogs and rodents (Edgaonkar & Chellam 1998). In Chapter 2, I estimate









the density of wild prey in the Satpura Tiger Reserve using the line transect method, while in

Chapter 3 I quantify the diet of leopards and its sympatric carnivore species, the dhole (Cuon

alpinus) and the tiger, and also estimate selection for the major prey species.

The dramatic reduction in tiger populations (Jhala et al. 2008) in India has also meant that

there is increasing poaching pressure on the leopard to meet the demands of the skin and bone

trade. This has been borne out by seizures of thousands of skins in recent years. In spite of this,

we do not have any reliable estimate of leopard populations in India, neither do we know

whether the population is really declining. An important first step is to estimate the densities at

which leopards are found in Indian forests. In Chapter 4, I use camera trapping and the mark-

recapture method to estimate leopard densities in three sites in the Satpura Tiger Reserve and and

one site in the Sariska Tiger Reserve.

To conserve leopards, it is necessary to first identify areas that have good leopard habitat.

In Chapter 5, I present a model that identifies leopard habitat using presence-only data for

Satpura Tiger Reserve and for the larger south- central Madhya Pradesh. The model also

identifies the habitat attributes that contribute to the likelihood of leopard presence.

The aim of this dissertation is to generate further information on the basic ecology of a

large carnivore, the leopard (Panthera pardus), which has been little studied in India. My

obj ectives are: 1) to estimate leopard prey density over four years in Satpura Tiger Reserve.

2) Quantify prey species selection by the leopard and its sympatric carnivores, dhole and

tiger. 3) Obtain density estimates for leopards using the mark-recapture framework, and 4) to

generate a leopard habitat suitability model for Satpura Tiger Reserve and for south-central

Madhya Pradesh.









CHAPTER 2
DENSITY ESTIMATION OF POTENTIAL PREY SPECIES OF LARGE CARNIVORES IN
SATPURA TIGER RESERVE USING LINE-TRANSECT SAMPLING.

Introduction

Accurate and precise estimation of animal abundance is a necessary first step to detect and

mitigate unacceptable levels of population change. Until recently there was a paucity of reliable

information on population densities of wild ungulates and other vertebrate species in India. This

paucity was attributed to funding difficulties, the politics of research and the relatively small

number of scientists engaged in long-term research (Eisenberg & Seidensticker 1976). However,

conditions in India have changed, and in the last 10-15 years the number of ungulate studies

employing rigorous methods has markedly increased and we now have population density

estimates of large herbivores from at least 10 protected areas. Density estimates of large

herbivores are available for Nagarhole (Karanth & Nichols 2000; Karanth & Sunquist 1992),

Bandipur (Johnsingh 1983; Karanth & Nichols 2000), Bhadra (Jathanna et al. 2003), Mudumalai

(Varman & Sukumar 1995), Gir (Khan et al. 1996), Melghat (Karanth & Nichols 2000), Pench

(Acharya 2008; Biswas & Sankar 2002; Karanth & Nichols 2000), Kaziranga (Karanth &

Nichols 2000), Ranthambore (Bagchi et al. 2004; Karanth & Nichols 2000) and Sariska

(Avinandan 2003; David et al. 2005).

While these efforts are laudable, India is undergoing a rapid change and by some estimates

the economy has been proj ected to grow annually at a rate of 5 percent or more for the next 30

years (Wilson & Purushothaman 2003). There is increasing pressure on natural areas and there

are reports of decline in forest quality (Lele et al. 2000). A recently released report has estimated

a population of between 1165 and 1657 tigers in India (Jhala et al. 2008), which is much lower

than estimates from just a decade earlier. A critical component for conservation of tigers and

other carnivores is the availability of wild prey (Karanth et al. 2004b). There is thus an urgent









need to establish baseline densities of prey in all protected areas in India and put in place a

monitoring program that would detect changes in their populations.

Ideally the monitoring scheme should use a method that minimizes bias and error while

maximizing precision, and has sufficient power to be able to detect population changes.

Distance-based methods have the advantage of not requiring animals to be handled, are relatively

easy to apply and give robust results if underlying assumptions are met. Violations of these

assumptions bias the resulting estimates in various ways, the details of which can be found in

Buckland et al. (2001).

A disadvantage of line transects is that a large number of observations are needed to

calculate the detection function precisely, and obtaining these is a labor intensive process.

Vehicle transects have been run along road networks as a substitute to foot transects (Ogutu et al.

2006; Ward et al. 2004). These yield a larger effort in the same time. However, the resulting

estimate may be biased as roads are not usually randomly laid with respect to the animals

(Varman & Sukumar 1995). Some species are attracted to the edge habitat created by roads while

other species may avoid the disturbance. This is likely to be area-specific depending on the

configuration of the road network and the species being monitored. It is nevertheless worth

investigating whether vehicle transects can be used to monitor populations in a given area.

The obj ectives of the present study are to 1) Estimate the density of potential prey species

of tigers, leopards and dholes using the line-transect method. 2) Evaluate changes in density over

4 years of sampling to enable monitoring of the population, and 3) Determine if vehicle transects

yield density estimates equivalent to those obtained by foot transects.









Methods


Study Area

The Satpura Tiger Reserve (22ol9' to 22 o 30' N and 77 o 56' to 78 o 20' E) is a 1428 km2

protected area located in the Hoshangabad district of Madhya Pradesh state in India. It comprises

of the Pachmarhi and Bori Wildlife Sanctuaries, and Satpura National Park. An intensive study

area of approximately 200 km2 was located in Bori Wildlife Sanctuary and Satpura National Park

(Figure 2-1). The intensive study area is a mosaic of dry and moist deciduous mixed forest. Teak

(Tectona grandis) plantations replaced mixed forests in some areas, though many of these

plantations are not pure teak, but are mixed with other species. Common tree species found there

include Palas, Butea nzonospernza; Mahua, Madhuca latifolia; Landia, Lagerstroentia parviflora;

Kari, Schleicheria oleosa; Saj, Ternzinalia arfuna and Tendu, Diospyros nzelan2oxylon .

The tiger (Panthera tigris), leopard (Panthera pardus) and the dhole (Cuon alpinus) are

carnivores of management interest in the study area. Other carnivores include jackal (Canis

aureus), striped hyena (Hyaena hyaena), sloth bear (M~ehersus ursinus), jungle cat (Felis chaus),

palm civet (Paradoxurus hernzaphroditus), small Indian civet (Viverricula indict), ruddy

mongoose (Helpestes snzithii), common mongoose (Helpestes eda~urdsi) and ratel (M~ellivora

capensis). A diverse community of ungulates and ground birds are preyed upon by the

carnivores. Potential prey for tigers, leopards and dholes include the wild pig (Sus scrofa),

chousingha (Tetracerus quadricornis), chital (Axis axis), Indian muntj ac (M~untiacus nauntiak),

sambar (Cervus unicolor), nil gai (Bosephahts tragoca~nebts) gaur (Bos guruss, the common

langur (.\Slelinimpithe us entellus), black-naped hare (Lepus nigricollis) and Indian porcupine

(Hystrix indict).









Estimation of Habitat Types

A georeferenced and orthorectified cloud-free Landsat ETM+ image for the study area was

obtained from the Global Landcover Facility (www.1andcover.org) Spectral signatures for the

classification supervision were obtained by using information from vegetation plots. A sample of

473 circular plots of 10 m radius were laid along transects and dirt trails in the study area. The

number and composition of woody tree species larger than saplings was noted inside the plot.

Five cover types were delineated. These were: moist forest, dry forest, bare ground/village, teak

dominated forest and water. Supervised classification was performed using FISHER classifier for

the study site using Idrisi Kilimanj aro (Eastman 2004). The transects were then stratified post

hoc as belonging to one of 3 different habitats: dry deciduous, moist deciduous and teak

dominated. Any transect traversing more than one habitat was allocated to the habitat type that it

most represented. A fourth habitat type, the riverine forest habitat, is found along streams. It was

considered part of the moist deciduous habitat for the purposes of stratification since it formed a

small proportion of the overall area. Density of trees along the transects was estimated from the

vegetation plots, and a Sorenson's index (Krebs 1989 ) was calculated to quantify the tree

species similarity between the three habitats.

Estimation of Prey Density

Density of the maj or prey species was estimated by the line-transect sampling method.

Twenty permanent transects were laid in the study area. The area was divided into approximately

5 km2 grids and ten grids were randomly chosen. A 2 km transect was laid in a random direction

in each grid to make a total of 10 transects. The vegetation on the transect was minimally cut so

as to allow observers to move through the forest, but not so much as to change the nature of the

habitat close to the transect. These ten transects were then supplemented in the second year by

ten more transects of 3 km length. These were laid systematically so that gaps in coverage









between the first ten transects were filled as much as possible. The location of transects is shown

in Figure 2-1. The total length of the transects was 50 km. The twenty transects were walked

repeatedly for a total effort of 1272 km. Transects were walked early in the morning and evening

in summer and winter at a speed of about 3 kmph, so that it took 40 minutes to walk a 2 km

transect and one hour for the 3 km transect. The species, group size, angle and angular distance

to the center of the group or to the individual was noted. Distance measurements were taken with

a laser rangefinder (Bushnell Yardage Pro 400) and angles were measured with a magnetic

compass. Program Distance v5 release 2 (Thomas et al. 2006) was used to estimate the density of

prey species. Two estimates of density were made: 1) A pooled density was estimated over all

habitats for each year. 2) Density was estimated pooled over all years for each habitat type. This

was calculated as a mean of the densities in each habitat type weighted by the area of each

habitat.

An exploratory analysis of the distribution of the distances was done by grouping them in

small intervals and plotting the resulting histograms as recommended by Buckland et al. (2001).

Depending on the resulting histogram, data were truncated at an appropriate distance for each

species. Evidence of heaping, spikes near the line and avoidance movements or a sharp drop-off

away from the line was investigated. The data were then grouped into appropriate intervals for

each species so that the detection function gave a good fit.

A detection-probability function was estimated from pooled data across years and habitats

for each species to maximize the number of sightings. Since all three habitats had similar tree

densities, there was no reason to believe that detections differed between habitats and years. The

data were modeled with the uniform, half normal and hazard rate models fitted with the cosine

and simple polynomial series for each species. The negative exponential model, recommended










only as a last resort, was used for the Indian peafowl since the other models did not fit well. The

model with the smallest Akaike Information Criterion (AIC) value was selected as the best-fit

model provided that the p-value for the chi-square goodness of fit for the model was greater than

0.05 (Burnham & Anderson 2003). The cluster size was calculated as a mean of observed

clusters, and variance was calculated by bootstrapping observations within transects for most

species. In the case of grey jungle fowl, muntj ac and gaur the bootstrap estimates failed to

converge. Variance was estimated empirically for these species.

Vehicle Transects

Sixty-Hyve drives were made using a 4-wheel drive vehicle on the dirt trails in the study

area, with a total effort of 388 km. Two observers were used and the vehicle was driven at

between 10 and 15 km per hour. Perpendicular distance to the sighting was estimated by

stopping in front of the animal cluster and taking a distance measurement with a laser

rangefinder to the center of the cluster. The drives overlapped with each other in spatial

coverage, and were made on one trail network. The data were analyzed as if it was from one

drive to avoid inflating degrees of freedom, and the variance was estimated using the Poisson

assumption.

Results

Description of Habitat

The teak dominated habitat had the highest stem density (798 trees/ha) and number of tree

species (42) of which 20 percent was teak (163 stems/ha). The dry deciduous habitat had 3 5 tree

species and a density of 673 stems/ha, of which the most common species was Diospyros

melan2oxylon (134 stems/ha). The moist deciduous habitat had 40 tree species (541 stems/ha) and

was dominated by bamboo (83 clumps/ha). The ten most dominant species in each habitat type

are presented in Table 2-1. The Sorensen's index of similarity between teak dominated and dry









deciduous was 0.86, between teak dominated and moist deciduous was 0.83, and between dry

deciduous and moist deciduous was 0.80.

Estimation of Density

Density estimates of potential prey species in the moist deciduous, dry deciduous and

teak dominated habitats are presented in Tables 2-2, 2-3 and 2-4. Amongst the ungulates, chital

were found in significantly higher densities in the moist deciduous and teak dominated habitat

compared to the dry deciduous habitat. Sambar was found in significantly higher densities in the

teak dominated habitat. The densities of nilgai, wild pig and muntj ac for the three habitat types

were not statistically different. The densities of black-naped hare, grey jungle fowl, red spurfowl

and Indian peafowl were also not significantly different among the three habitats. Common

langur density was highest in moist deciduous, followed by teak dominated habitat and then the

dry deciduous habitat.

When pooled over all the habitats, the number of observations of chital, sambar, nilgai,

muntj ac, wild pig and peafowl were sufficient to estimate density for each year from 2002 to

2005. Densites were lower for all species in 2005, the last year of sampling (Figure 2-2 and 2-3)

when compared to the first year, 2002. The estimates for 2002 and 2005 were statistically

different for sambar, nilgai and common langur.

Average density, weighted by area of each habitat for 11 species pooled over four years is

presented in Table 2-5. Common langur is the most abundant species. Amongst ungulates, chital

numbers were highest, followed by sambar, nilgai, wild pig, gaur and muntj ac. Ungulate

densities in Satpura Tiger Reserve are among the lower estimates when compared to other

protected areas in India (Table 2-6).

Density estimates derived from vehicle transects are given in Table 2-7. A comparison of

density estimates from vehicle transects with those from foot transects for the same year are










presented in Figure 2-3, and detection function curves for the two methods are presented in

Figure 2-4. Density estimates of chital, sambar and nilgai by foot transects were higher than

those estimated by vehicle transects, but the difference is not statistically significant. Density of

muntj ac, langur, peafowl and wild pig is greater when estimated by vehicle transects but the

difference is significant only for peafowl and langur.

Discussion

The accuracy of density estimates depends on how well the underlying assumptions are

met. The data were gathered by trained observers using a laser rangefinder and compass to

estimate the bearing and distance to the animal group. Detections near the line, as shown by the

low chi-square values for the first distance interval, were as expected for each model for all

species. There was no evidence of heaping or a sharp drop-off indicating evasive movement in

response to the observer for most species. Estimated strip width, as could be expected if distance

played a maj or role in detectability, was wider for the large sized species than for the small sized

species. A notable exception was the common langur, which was detected at larger distances.

Overall wild ungulate density in Satpura Tiger Reserve is lower than that reported for

protected areas such as Nagarhole and Bandipur in southern India and Kanha and Pench tiger

reserves in Madhya Pradesh, but is comparable to other protected areas in central India like

Tadoba, Melhghat and the Maharshtra side of Pench tiger reserve (Karanth & Nichols 2000).

In the study area, the densities of most species were lowest in the dry deciduous habitat

though some of these differences were not significant. There were fewer water sources in this

habitat and it also tended to be closer to the villages found interspersed within the study area.

These factors could be responsible for the lower densities. Nilgai, which is known to tolerate

disturbance and lack of water (Bagchi et al. 2003a), was not found in lower density in this

habitat. The teak dominant habitat had the highest density of sambar, but this area was close to










rugged terrain and also had a number of artificial waterholes. Sambar is known to prefer hilly

areas (Bhatnagar 1991) and this could have caused the higher density seen here.

Estimation of annual density shows up a pattern of density reduction for all species for the

last year. Visual inspection of the data does not show a continuous declining trend for any

species except for common langur. However, four years of data are too short a time period to

statistically estimate a trend or rate of change using methods like generalized additive modeling

(GAM) as has been done for Nagarhole (Gangadharan 2005). One study has speculated about

long-term cyclic changes associated with 3 to 10 year lagged rainfall patterns (Ogutu & Owen-

Smith 2005) in Africa, but there is little information to indicate the reason for this decline.

Possible reasons could be pressure due to over-grazing, illegal poaching or part of a natural

cyclic tendency. Personal observation did not indicate a higher degree of poaching for the last

year, nor was there a change in the number of domestic cattle over the years. It can only be

speculated that the below average rainfall in 2001 and 2002 (Mooley et al. 2007) may be

responsible for the reduction in ungulate density in 2005.

Estimates of variance of density were lowest for langur and sambar for which a large

number of observations were made, and highest for gaur, of which only 3 5 groups of which were

observed. The variance estimate is a combination of variances in the detection probability,

cluster size and encounter rate, with the encounter rate variance being the maj or component.

Encounter rate variance remained the maj or component for all the years for each species, except

for jungle fowl in 2005, where detection probability was the maj or component.

A larger effort can be achieved in a short amount of time with vehicle transects. Except for

langur and peafowl, confidence intervals for the two density estimates overlapped. The state

forest department clears viewing lanes along some dirt trails for tourism purposes. This probably









attracts peafowl, muntj ac, langur and wild pig to the cleared areas for foraging leading to an

increased estimate of density. There may some tendency for larger ungulates such as chital,

sambar and nilgai to stay away from the disturbance caused by roads, leading to reduced density

estimates, though the difference is not significant.

Even though densities of most species are moderately low in the study area, the ungulate

community is still intact and should be protected. The Bori-Satpura area is large enough to be

potentially able to support a relatively large tiger population, but to do this the ungulate prey

base will have to be enhanced. The baseline estimates generated in this study can be used to

monitor future changes in population. The last year of sampling showed lower density for all

species, and that is a matter of concern. It is therefore recommended that a monitoring program

be initiated and protection measures strengthened to arrest the putative decline in wildlife

populations in the study area. Vehicle transects can be used as the network of dirt trails is

sufficiently extensive to be able to obtain reasonably accurate results.









Table 2-1. Density of dominant tree species (individuals per hectare) in the three habitat types
along transects.
Rank Teak dominated plantation Dry deciduous teak Moist deciduous teak (N =
order (N= 73 plots) (N = 73 plots) 61 plots)
1 Tectona grandis (163) Diospyros nzelan2oxylon Bamboo species (83)
(134)
2 Diospyros nelanoxylon (85) Choloroxylon swietenia Diospyros nzelan2oxylon
(95) (79)
3 Ternzinalia alfuna (78) Ternzinalia aijuna (60) Tectona grandis (70)
4 Lagerstroentia parviflora (71) Tectona grandis (53) Ternzinalia arfjuna (32)
5 Aegle nzrmelos (41) Bamboo species (50) Saccopetahtna tonzentosunt
(29)
6 Anogeissus latifolia (40 ) Acacia catechu (42) Madhuca indica (28)
7 Zizyphus xylopara (37) Buchanania laza Lagerstroentia parviflora
(3 5) (27)
8 Saccopetahtn tonzentosunt Madhuca indica ( 29) Enablica officinalis (22)
(34 )
9 Buchanania lana (34) Lagerstroentia Choloroxylon swietenia
parviflora (22) (20)
10 Choloroxylon swietenia (3 0) Enablica officinalis (21) Zizyphus xylopara (20)










Table 2-2. Estimation of density parameters of potential prey by the line-transect method in the
moist deciduous habitat.
Species n D CV D CI D Ds Cv Ds CI Ds Model
Chital 92 8.0 14.0 6.2-10.5 2.4 12.5 2.0-3.1 Hazard
Polynomial
Sambar 68 3.8 10.6 3.0-4.4 1.8 10.1 1.4-2.0 Half-Normal
Cosine
Nilgai 26 1.4 17.3 1.1-2.1 0.7 15.2 0.5-1.0 Half-Normal
Cosine
Muntj ak 34 1.2 21.9 0.8-2.0 1.1 21.6 0.6-1.7 Half-Normal
Cosine
Wild pig 29 2.5 26.6 1.4-3.9 0.8 14.8 0.6-1.0 Half-Normal
Cosine
Black-naped hare 21 2.7 16.2 2.1-3.7 2.6 15.6 2.0-3.5 Half-Normal
Cosine
Common langur 313 39.9 10.3 34.0-51.0 9.1 9.5 8.0-11.7 Half-Normal
Cosine
Indian peafowl 29 2.0 21.1 1.3-2.9 1.3 18.7 0.8-1.7 Neg exp Cosine
Red spurfowl 20 2.7 20.4 1.7-3.6 1.5 18.8 1.0-2.0 Half-Normal
Cosine
Grey jungle fowl 34 2.9 26.9 1.5-5.6 1.6 26.2 0.8-3.0 Uniform
Polynomial
n: number of observations, D: density of individuals/km2, Ds: Density of groups/km2, CV:
coefficient of variation, CI: 95% Confidence. Sample size: 6 transects, effort: 408 km.










Table 2-3. Estimation of density parameters of potential prey by the line-transect method in the
dry deciduous habitat.
Species n D CV D CI D Ds Cv Ds CI Ds Model
Chital 27 1.9 14.5 1.4-2.4 0.6 13.4 0.5-0.4 Hazard
Polynomial
Sambar 70 3.1 10.2 2.4-3.6 1.5 9.6 1.1-1.7 Half-Normal
Cosine
Nilgai 37 1.6 17.0 1.2-2.3 0.8 14.5 0.6-1.1 Half-Normal
Cosine
Muntj ak 17 0.5 36.7 0.2-1.1 0.4 36.5 0.2-0.9 Half-Normal
Cosine
Wild pig 15 1.0 25.9 0.6-1.6 0.3 14.2 0.2-0.4 Half-Normal
Cosine
Black-naped hare 42 4.3 15.6 3.4-6.1 4.1 15.1 3.2-5.7 Half-Normal
Cosine
Common langur 155 15.7 10.4 13.4-20.2 3.6 9.6 3.1-4.6 Half-Normal
Cosine
Indian peafowl 29 1.6 21.1 1.0-2.3 1.0 18.8 0.7-1.4 Neg exp Cosine
Red spurfowl 23 2.5 20.4 1.5-3.3 1.4 18.8 0.9-1.9 Half-Normal
Cosine
Grey jungle fowl 42 2.9 18.9 1.9-4.4 1.5 17.9 1.0-2.3 Uniform
Polynomial
n: number of observations, D: density of individuals/km2, Ds: Density of groups/km2, CV:
coefficient of variation, CI: 95% Confidence. Sample size: 8 transects, effort: 513 km.










Table 2-4. Estimation of density parameters of potential prey by the line-transect method in the
teak dominated habitat.
Species n D CV D CI D Ds Cv Ds CI Ds Model
Chital 70 7.1 13.9 5.6-9.2 2.1 12.5 1.7-2.7 Hazard
Polynomial
Sambar 124 8.0 10.3 6.3-9.3 3.7 9.7 3.0-4.3 Half-Normal
Cosine
Nilgai 32 2.0 17.6 1.5-3.0 1.0 15.1 0.8-1.4 Half-Normal
Cosine
Muntj ak 12 0.5 24.6 0.3-0.9 0.4 24.3 0.2-0.8 Half-Normal
Cosine
Wild pig 19 1.9 26.2 1.1-2.9 0.6 15.6 0.4-0.8 Half-Normal
Cosine
Black-naped hare 20 3.0 15.2 2.4-4. 2.8 14.6 2.3-3.9 Half-Normal
Cosine
Common langur 169 25.1 10.4 21.4-32.2 5.7 9.6 5.0-7.4 Half-Normal
Cosine
Indian peafowl 40 3.3 20.3 2.1-4.5 2.0 18.6 1.4-2.7 Neg exp Cosine
Red spurfowl 16 2.5 20.4 1.5-3.4 1.4 18.8 0.9-1.9 Half-Normal
Cosine
Grey jungle fowl 10 1.0 36.9 0.4-2.5 0.5 36.4 0.2-1.3 Uniform
Polynomial
n: number of observations, D: density of individuals/km2, Ds: Density of groups/km2, CV:
coefficient of variation, CI: 95% Confidence. Sample size: 6 transects,effort: 351 km.










Table 2-5. Estimation of overall density and its associated parameters by the line-transect method
over 4 years in the study area.
Species n D CV D CI D Ds Cv Ds CI Ds Model
Chital 189 5.4 13.8 4.2-7.1 1.6 12.4 1.3-2.1 Hazard
Polynomial
Sambar 262 4.0 10.3 3.2-4.7 1.9 9.7 1.5-2.2 Half-Normal
Cosine
Nilgai 95 1.6 17.0 1.2-2.3 0.8 14.7 0.6-1.1 Half-Normal
Cosine
Muntj ac 63 0.8 19.0 0.6-1.2 0.7 17.3 0.5-1.1 Half-Normal
Cosine
Wild pig 63 1.8 26.2 1.1-2.9 0.6 14.5 0.4-0.7 Half-Normal
Cosine
Black-naped hare 83 3.4 15.6 2.7-4.7 3.2 15.0 2.6-4.4 Half-Normal
Cosine
Gaur 35 0.8 37.4 0.4-1.8 0.2 33.4 0.1-0.4 Half-Normal
Cosine
Common langur 637 28.3 10.3 24.1-36.3 6.4 9.5 5.7-8.3 Half-Normal
Cosine
Indian peafowl 98 2.0 20.0 1.3-2.9 1.3 17.7 0.9-1.7 Neg exp Cosine
Red spurfowl 59 2.6 20.4 1.6-3.5 1.5 18.8 1.0-1.9 Half-Normal
Cosine
Grey jungle fowl 86 2.7 17.1 1.8-3.8 1.4 16.0 1.0-2.0 Uniform
Polynomial
n: number of observations, D: density of individuals/km2, Ds: Density of groups/km2, CV:
coefficient of variation, CI: 95% Confidence. Sample size: 20 transects, effort: 1272 km.









Table 2-6. Density of wild ungulates (individuals per km2) at various study sites in India.
Place Chital Sambar Nilgai Muntj ac Wild Gaur
pig
Bandipurl 20.1 5.6 -0.7 0.7 7.0
Nagarhole2 49.1 3.4 -4.3 3.4 5.6
Pench-M.P.3 80.7 6.1 0.4 -2.6 0.3
Kanhal 49.7 1.5 -0.6 2.5
Ranthambores 31.0 17.1 11.4 -9.8
Sari ska6 27.6 8.4 5.2 -17.5
Gir7 25.2 1.8 0.4 -2.1
Bhadra4 2.3 5.8 -5.4 2.6 0.7
Melghat 2.7 -0.6 0.5 1.0
Tadobal 3.2 3.3 0.7 0.9 2.6 1.8
Pench- 5.8 5.9 0.5 -2.0 0.8
Maharshtral
Bori-Satpuras 5.4 4.0 1.6 0.8 1.8 0.8
1Karanth and Nichols (2000) 2Karanth and Sunquist (1992), 3Biswas and Sankar (2002),
4Jathanna et al. (2003), SBagchi et al. (2003) 6Avinandan, D (2003) 7Khan et al (1996) sThis
study .

Table 2-7. Estimation of density parameters of potential prey by the vehicle transects assuming
Poisson variance.
Species n D CV D CI D Ds Cv Ds CI Ds Model
Chital 51 3.7 17.0 2.6-5.2 1.1 14.6 0.8-1.4 Uniform cosine
Sambar 61 2.3 17.7 1.7-3.3 1.1 15.1 0.8-1.5 Uniform cosine
Nilgai 21 0.5 31.4 0.3-0.9 0.3 28.0 0.2-0.6 Half-Normal
Cosine
Muntj ac 35 1.3 21.9 0.8-1.9 1.0 19.6 0.7-1.5 Uniform cosine
Wild pig 18 3.3 44.9 1.4-7.9 0.8 36.8 0.4-1.7 Neg exp cosine
Common langur 148 36.0 13.7 27.5-47.0 5.1 12.0 4.1-6.5 Hazard rate
Indian peafowl 74 4.9 18.0 3.4-7.0 2.6 15.1 1.9-3.5 Half normal
n: number of observations, D: density of individuals/km2, Ds: Density of groups/km2, CV:
coefficient of variation, CI: 95% Confidence. Sample size: 1 transect, effort: 388 km.






















IIA M/ist deciduous
iik~. 1-" I Dry deciduous
Teak dominated










Figure 2-1. Map of the study area, showing the habitat types, transects and vehicle transects
trails.





8-
-; -

S -


i


2-


A.




3.5 -

3.0 -



S2.0 -

1.5 -



0.5 -


2002 2003 2004 2005
Year


2002 2003 2004 2005
Year


- -


a,1.0
-

- 0.5


,I


2002 2003 2004 2005 2002 2003 2004 2005
Year D.Year
Figure 2-2. Annual densities of selected species (individuals/km2) in the study area.
A) Chital. B) Sambar. C) Nilgai. D) Muntj ac. E) Common langur. F) Indian peafowl.
G) Wild pig. Error bars are bootstrapped 95% confidence limits.





















I


2002 2003 2004 2005
Year


60


S50


r 40

30

-g 3
a,20


10


5


S4 1


3
-



: -1

O
a 1-1


I T



2002 2003 2004 2005
Year


2002 20



2-2: Continued.


5


S4





:g2


a, 1




G.


Figure


03 2004 2005
Year






























Chital Sambar Nilgai

Species


Common langur

Species


[Z Vehicle transect
Foot transect


Muntjak Indian peafowl Wild pig
Species


Figure 2-3. Comparison of density estimates (individuals/km2) between foot transects and
vehicle transects in 2005 for potential prey species of large carnivores in Satpura
Tiger Reserve. Error bars are bootstrapped 95% confidence limits.




















CHITAL (vehcletmnsect)


CHITAL (foottfmnsect)


PERPENDICULAR DISTANCE (m)


PERPENDICULAR DISTANCE (m)


12


10-


08-













O
O








O 4

O 2








Ooo.



~02

OO


SAMBAR (foottmnsect)















)10 20 30 40 50 60 70 80 90 100
PERPENDICULAR DISTANCE (m)


S SAMBsAR(vehicle tmnsect) 1O


O oa





020






10 20 30 40 50 60 70 80 90 100 110 120
PERPENDICULAR DISTANCE (m)


LANGUR (foottmansect)


LANGUR (vehcle transect)


O 10 20 30 40 50 60 70 80 90 100 110 120
PERPENDICULAR DISTANCE (m)


20 30 40 50 60 70 80 90 100
PERPENDICULAR DISTANCE (m)


PEAFOWL (vehicle transed)


PEAFOWL (foot transect)


PERPENDICULAR DISTANCE (m) PERPENDICULAR DISTANCE (m)



Figure 2-4. Detection function curves for vehicle and foot transects for chital, sambar, langur and


peafowl .









CHAPTER 3
PREY SELECTION AND THE FOOD HABITS OF TIGER, LEOPARD AND DHOLE INT
SATPURA TIGER RESERVE.

Introduction

Information about interactions between tropical large carnivore species is scarce. Some

food habits studies have been conducted on single species (Bagchi et al. 2003b; Biswas & Sankar

2002; Edgaonkar & Chellam 2002; Reddy et al. 2004), and there is information on diet selection

and overlap between multiple species including tigers, leopards and dholes from southern India

(Johnsingh 1983; Karanth & Sunquist 1995), and on leopards and tigers (Sankar & Johnsingh

2002) and lions and leopards (Chellam 1993) in western India. Differential spatial use by tigers

and leopards was reported by one study in Nepal (Seidensticker 1976), but not in another

(Karanth & Sunquist 1995). In the neotropics, spatial avoidance of jaguars (Panthera onca) by

pumas (Puma concolor) (Scognamillo et al. 2003) was seen at fine scales, but another (Taber et

al. 1997) did not find a similar pattern. Some degree of dietary separation between pumas and

jaguars has been noted, with jaguars tending to take slightly larger prey and more peccaries

(Emmons 1987). Tigers and leopards are opportunistic stalking predators and are expected to kill

more randomly as opposed to dhole, which is a coursing predator (Schaller 1967). Dholes, or

Asiatic wild dogs, are also more diumal than tigers and leopards (Johnsingh 1983). They are

group living, coursing predators weighing about 20 kg. Anecdotal evidence exists of aggressive

interactions between the three carnivores, especially between leopards and dholes. There is need

for more data on the potential for competition and resource overlap among these maj or predators

in tropical forest assemblages over a range of resource availabilities. The present study describes

the prey taken, quantifies the dietary overlap and measures the prey selectivity of tigers, leopards

and dholes at a site where the abundance of prey is lower than other at places where food habits

of these carnivores have been studied.









Methods


Study Area

The study was conducted in the Satpura Tiger Reserve (STR). STR covers 1428 km2 in

area, and is located in the Hoshangabad district of the central Indian state of Madhya Pradesh in

India. It includes three administrative units, the Pachmarhi and Bori Wildlife Sanctuaries, and

Satpura National Park. An intensive study area of approximately 200 km2 was located in Bori

Wildlife Sanctuary and Satpura National Park (Figure 3-1).

The forest in STR (22ol9' to 22 o 30' N and 77 o 56' to 78 o 20' E) is mainly of the moist

deciduous type (Champion & Seth 1968). The intensive study area is a mosaic of dry and moist

deciduous forest dominated in many places by teak. Teak plantations replaced mixed forests in

some areas, though now even these are not pure teak, but have secondary growth of species.

Common species found there include Palas, Butea nzonospernza; Mahua, Madhuca latifolia;

Landia, Lagerstroentia parviflora; Kari, Schleicheria oleosa; Saj, Ternzinalia arfuna and Tendu,

Diospyros nzelan2oxylon .

A diverse assemblage of fauna is found including wild pig (Sus scrofa), chousingha

(Tetracerus quadricornis), chital (Axis axis), Indian muntjac (M~untiacus nauntiak), sambar

(Cervus unicolor), chinkara (Galzella gazelle), nil gai (Bosephahts tragocantehis) and gaur (Bos

guruss. The common langur (Senouspl~ithet~ us\ entellus) and rhesus macaque (Macacantulatta)

are the primates found here. Carnivores are represented by tiger (Panthera tigris), leopard

(Panthera pardus), wild dog (Cuon alpinus), jackal (Canzis aureus), striped hyena (Hyaena

hyaena), sloth bear (M~ehersus ursinus), jungle cat (Felis chaus), palm civet (Paradoxurus

hernzaphroditus), small Indian civet (Viverricula indica), ruddy mongoose (Helpestes snzithii),

common mongoose (Helpestes eda~urdsi) and ratel (M~ellivora capensis). Black-naped hare

(Lepus nigricollis), Indian porcupine (Hystrix indica), Indian giant squirrel (Ratufa indica) and










the large Indian flying squirrel (Petauristapetaurista) are some of the smaller mammals found

here.

Reconstruction of Carnivore Diets

Scats were collected opportunistically as well as systematically along animal and man-

made trails and dirt roads in the study area (Figure 3-1). Identification of tiger and leopard scats

was based on associated tracks or sign. Scat size or diameter was not used as the criterion for

discriminating between species as there is suspected to be overlap in scat size amongst the three

species and this may lead to significant misidentifieation of scats (Farrell et al. 2000). Only scats

of tigers and leopards that had associated tracks or sign near them were collected to ensure

correct identification. Scats of dholes were easy to identify as they were deposited at communal

defecation sites (Johnsingh 1983). Scats were washed and undigested remains of hair were

mounted on a slide and compared under a microscope with a reference collection of hair at the

Wildlife Institute of India following a standard protocol (Mukherj ee et al. 1994). Bird and rodent

taxa were not identified to species level. In total 193 leopard scats, 93 tiger scats and 81 dhole

scats were analyzed. The percentage frequency of occurrence of all the maj or species was

calculated along with their bootstrap confidence intervals.

Sample Size Adequacy

To check for the stability of percent frequency of occurrence in the diet, all scats for each

carnivore were randomized and the percentage frequency of occurrence of each prey item in the

diet was plotted cumulatively, at an interval of 10 scats. The number of scats required for the

frequencies to reach an asymptote was considered sufficient to quantify that prey item in the diet

reliably.










Prey Biomass and Number

The frequency of occurrence is biased towards smaller sized prey, since relatively more

scats are produced for smaller prey than larger prey. To correct for this bias, relative frequencies

of prey were converted to relative biomass consumed for tigers and leopards using an equation

estimated for cougars (Ackerman et al. 1984), and for dholes using an equation estimated for

wolves (Floyd et al. 1978). This regression equation estimates the number of field collectable

scats for a given weight of prey biomass.

These are

y = 0.38 + 0.020x (for dhole) and
y = 1.98 + 0.03 5x (for tigers and leopards)


where the independent variable x is the average weight of the prey and the dependant variable y

is the number of field collectable scats for that weight of prey. The dependant variable can then

be converted into the relative biomass of prey consumed by multiplying it by the relative

frequency of each prey species found in the scats. The relative number of each species consumed

is obtained by dividing relative biomass by the average weight of the prey species. The weight of

various prey species killed by tiger, leopard and dhole was assumed to be similar to that used in

previous research (Karanth & Sunquist 1995).

Estimation of Prey Selection

Selectivity can be defined as taking a prey at frequencies different from that expected

given its availability (Chesson 1978). If there is no selection one would expect a prey item to be

taken at relative frequencies similar to the relative frequency of its availability. Any statistically

significant deviation, whether positive or negative, would indicate, preference or avoidance of

that prey type.









Availability of a prey species is likely to be a function of its abundance, anti-predatory

behavior, habitat selection at a Eine scale and time of activity. It is assumed, as in other studies

(Bagchi et al. 2003b; Biswas & Sankar 2002; Karanth 1995), that abundance is the maj or

component of availability. Abundance was therefore estimated as the density of groups of the

maj or prey species, since the probability of encountering prey is likely to be proportional to the

density of groups, rather than of individuals (Karanth & Sunquist 1995). The prey density was

estimated by the line-transect method. Twenty permanent transects were laid in the study area.

The first ten transects were laid randomly. The area was gridded into approximately 5 km2 grids

and ten grids were randomly chosen. A 2 km transect was laid in a random direction in each grid.

The next ten transects were then laid as 3 km lines so that gaps in coverage between the first ten

transects were filled as much as possible. The location of transects is shown in Figure 3-1.

Transects were walked repeatedly for a total effort of 1272 km. The species, group size, angle

and angular distance to the individual or center of group was noted. Distance measurements were

taken with a laser rangefinder (Bushnell Yardage Pro 400) and angles were measured with a

magnetic compass. Program Distance v5 release 2 (Thomas et al. 2006) was used to estimate the

density of prey species.

Selectivity was quantified by comparing the observed frequency of each prey species in the

scats to expected frequencies (Link & Karanth 1994). Expected frequencies were derived from

the densities estimated by line transects. If a kill of species with a density di, produces hi scats,

then the proportion of scats produced when the carnivore takes prey in proportion to their density

is given by










The program SCATMAN v2.0 (Hines 2002) was used to estimate prey selection by

comparing the Hi to the the observed proportion based on random samples of predator scats. The

program uses the estimated di and hi, and the variation associated with these parameters. It

implements a parametric bootstrap designed to handle the problem of excessive Type I error

caused by comparison of estimated frequencies as opposed to exact frequencies (Link & Karanth

1994). Inputs to the program are the estimated availability and standard error of each prey

species, and the number of collectable scats that are produced by an average kill of each prey

species, along with their standard errors. High chi-square values in the output indicate that

observed frequencies are significantly different from expected, and the presence of selectivity of

prey. The contribution of each species to the total chi-square indicates whether the prey species

is taken more or less than expected.

The Jacobs' index (Jacobs 1974) has been used to estimate dietary preference in carnivores

(Hayward 2006; Hayward et al. 2006a; Hayward & Kerley 2005; Hayward et al. 2006b). It has

the advantage of being simple to compute and can be used to compare across studies easily.

Availability and utilization of prey species in other study sites in India were obtained from the

published literature. The index was computed for all the study sites using the using the formula

D=r-p
r + p 2rp

where, r is the proportion of total kills of a prey species, and p is the proportion of the total

abundance of that species. The values of the index range from +1 to -1, indicating maximum

preference and maximum avoidance respectively.

The relative number of each prey species killed was obtained by dividing the relative

biomass by the average weight of the species taken by tiger, leopard and dhole, respectively. The









mean weight of prey killed was calculated as the sum of the weight of prey species multiplied by

the proportional number taken.

Dietary Overlap

The extent of dietary overlap between all three species pairs was calculated by Pianka' s

index (Pianka 1973) The program EcoSim version 7.72 (Gotelli & Entsminger 2007) was used

on the percent frequency matrix assuming all availabilities to be equal, as well as on an electivity

matrix (Lawlor 1980), which is a matrix of frequencies of prey taken weighted by their densities.

The calculated index can take values from 0 to 1, where 1 stands for identical diets or complete

ovelap and 0 indicates completely different diets, or no overlap. The formula used for calculating

the overlap of species with species, 012 is


P27~pt,
012 = 021 = I=1




where p, is the percentage frequency of species j taken by carnivore speciesi.The index was also

calculated on the electivity matrix comprising of electivity e, where R, is the availability of prey

species j.




The program randomizes the electivity for each combination of predator and prey species to

generate a null model to compare with the observed mean index. If the mean overlap index value

is at either tail of the distribution of simulated values then it can be judged to be significantly

different than expected by chance. The density of the major prey species was derived from the

results of the line transects. Porcupine density was assumed to be similar to that reported in the

literature (Sever & Mendelssohn 1991).












Density of Potential Prey Species

The mean density of groups and individuals of the potential prey species over four years is

presented in Table 3-1. Amongst ungulates, chital were the most common, while the Indian

muntj ac had the lowest density. Ground birds likely to be found in the carnivore diet are

represented by the Indian peafowl (Pavo cristatus), grey jungle fowl (Gallus sonneratii) and red

spurfowl (Galloperdix spadicea). Their densities were similar. Overall, the common langur was

the most abundant prey species.

Sample Size Adequacy

The results of the scat analyses for various prey species in the diet of the three carnivores

show stability (Figure 3-2). For chital, sambar and langur, about 50 scats provides a stable

estimate of the percentage frequency of that prey in the diet. The sample size of scats used in the

analysis can therefore be considered adequate for quantifying the maj or species found in the diet

of these carnivores.

Composition of Diet

Leopard preyed on 10 species, the tiger took 7 species and 4 species were found in dhole

scats. It was necessary to analyze about 55 scats to detect all these species (Figure 3-3). The

percentage frequency and relative biomass of the maj or prey species in the scats of the leopard,

tiger and dhole are given in Tables 3-3, 3-4 and 3-5. It can be seen that sambar is the maj or prey

in the diet of all three predators. Chital is taken by dhole and to a lesser extent by leopard, but is

not an important component of the diet of the tiger in this study. Livestock is also not an

important constituent of the diet, especially for leopards and dholes. Rodents, birds, porcupines,

and wild pigs also do not figure in the diet of dholes. Porcupine was only taken by leopards,

while hare was taken by both leopards and dholes but not by tigers. Relative biomass and


Results









number were not estimated for the categories of bird and rodent species because of uncertainty

about their weights. However, since they are a minor component of the diet, it would have little

effect on the results.

Tigers take the highest mean weight of prey (129 kg), followed by dholes (46 kg) and

leopard (27 kg). The percentage of prey taken by tiger, leopard and dhole in various prey size

classes is presented in Figure 3-4. Leopards take prey from each size class, though they take

medium-sized prey the most. Tigers and dholes seem to specialize on large and medium-sized

prey, respectively. A small percentage of smaller sized prey is taken by both species, but dholes

do not take larger prey.

Prey Selection

Tigers (X2 = 61.5, d. f = 4, p <0.01), leopards( (2 = 52.2, d.f. = 4, p<0.01) and dholes (X2

= 54.3, d.f. = 3, p< 0.01) all exhibited overall selectivity in their diet. Figure 3-5 shows the

observed and expected frequencies of the maj or prey species in scats. Tigers significantly

preferred sambar (X2 = 60.9, p<0.01) while avoiding chital (X2= 16.1, p<0.01), langur (X2= 10.3,

p<0.01) and hare (X2= 4.1, p=0.04). Wild pig was neither preferred nor avoided (X2= 2.8, p=0. 1).

Leopards also significantly preferred sambar (X2 = 43.4, p<0.01), while avoiding langur (X2 =

20.6, p<0.01) and wild pig (X2 = 8.2, p=0.005). Chital (X2 = 0.98, p=0.4) and hare (X2 = 0.23,

p=0.64) were neither preferred nor avoided. Dholes significantly preferred both chital (X2 = 18.4,

p<0.01) and sambar (X2 = 16.7, p<0.01) avoiding hare (X2 = 7.2, p <0.01) and langur (X2 = 30.9,

p<0.01). Wild pig (X2 = 2. 1, p=0. 15) was taken in proportion to its availability.

The preference for maj or prey species by tigers, leopards and dholes in various protected

areas in India using the Jacobs' index is presented in Tables 3-5, 3-6 and 3-7. Overall, tigers

seem to take chital (mean Jacobs' index -0.05, SE 0.19, n=6 sites) and wild pig (mean Jacobs'

index 0.06, SE 0.27, n= 6 sites) approximately in proportion to their availability, though the










variance on these estimates is high. Tigers prefer sambar (mean Jacobs' index 0.38, SE 0.14, n =

6 sites) and avoid nilgai (mean Jacobs' index -0.9, SE 0.08, n = 4 sites), gaur (mean Jacobs'

index -0.45, SE 0.37, n = 4 sites) and langur (mean Jacobs' index -0.2, SE 0.16, n= 5 sites).

Leopards take chital in proportion to their availability (mean Jacobs' index 0.07, SE 0.1, n = 4

sites), prefer sambar (mean Jacobs' index 0.18, SE 0.27, n = 4 sites) and avoid gaur (mean

Jacobs' index -0.46, SE 0.37, n =3 sites), langur (mean Jacobs' index -0.21, SE 0.3, n = 3 sites)

and wild pig (mean Jacobs' index -0.12, SE 0.45, n= 4 sites). Dholes prefer chital (mean Jacobs'

index 0.20, SE 0.2, n = 4 sites) and sambar (mean Jacobs' index 0.41, SE 0.28, n = 4 sites), and

avoid gaur (mean Jacobs' index -0.80, SE 0.18, n = 4 sites), wild pig (mean Jacobs' index -0.12,

SE 0.45, n = 4 sites) and langur (mean Jacobs' index -0.80, SE 0.13, n = 3 sites).

Diet Overlap

The diet overlap (Table 3-8) exhibited a similar pattern when calculated with percent

frequency or electivity. The diets of all 3 species overlapped considerably. Tiger-leopard and

leopard-dhole diets overlapped more extensively than tiger-dhole diets, though this overlap

increased when electivity was used to calculate the index.

Discussion

Surprisingly, sambar is the preferred prey of all three species, and forms a large proportion

of the diet of the tiger in this study. Sambar has been found to be a preferred prey of tigers in

other studies also (Bagchi et al. 2003b; Biswas & Sankar 2002; Karanth & Sunquist 1995). It is a

large sized deer (about 200 kg), found in moderate densities, and is known to choose dense forest

areas (Varman & Sukumar 1995). This probably makes it more vulnerable to tiger predation,

unlike the chital. The minor role of chital in the tiger' s diet probably has to do with its habitat

selection and density. In STR, chital are found in open plain areas near villages, where there is a

lot of human disturbance. Their abundance is also not as high as that found in other national









parks in India. Their habit of congregating near human inhabitation at night has been speculated

to be the reason why they are not found in tiger diet in Bandipur (Johnsingh 1983). In Pench

National Park (Biswas & Sankar 2002) in central India and in Nepal's Royal Bardia tiger reserve

(Stoen & Wegge 1996), chital congregate in large numbers along low-lying areas. They

comprise a larger proportion of the tiger' s diet there, though they are still not highly preferred.

Wild pig are also taken less than expected, and this may be because of their low densities. In

Bardia and Nagarjunasagar (Reddy et al. 2004), wild pig were more commonly taken, and they

were found to be preferred prey of tigers in Pench (Biswas & Sankar 2002). Common langur is

also taken less than expected. In STR, langur is less important in the diet of the tiger than of the

leopard with respect to biomass and percentage frequency in scats, though relatively more langur

is taken by the tiger than by the leopard. This is because the diet of the leopard is more evenly

distributed amongst its prey species than that of the tiger. A similar pattern was seen in the

Sariska Tiger Reserve (Sankar & Johnsingh 2002), while only marginally more langur were

taken by leopards in Nagarhole (Karanth & Sunquist 1995).

Although one tiger kill of gaur was seen, gaur, nilgai and muntj ac were not found to be a

part of the diet of the tiger as measured by scat analysis. This could be because of the low density

of these species in the study area. The nilgai also prefers disturbed and open areas which are not

used by the tiger (Bagchi et al. 2003a). Livestock were also not an important component of the

diet, being found in about 5% of scats. This figure is comparable with some other studies, being

about 7% in Srisailam Tiger Reserve (Reddy et al. 2004), and 4.3% in Pench Tiger Reserve.

Leopards take chital in proportion to its availability, though it comprises about 20 % of its

biomass intake. Unlike the tiger, the leopard is also found close to human inhabitation, where

chital congregate at night. In STR, its relative lack of importance in the leopards diet may be due









to the larger mean group size of chital (6.3 per group, n=469 groups) as compared to the sambar

(2.2 per group, n=419 groups), which increases vigilance and helps avoid stalking predators. In

studies where chital is a maj or part of the diet, the chital density is quite high as compared to

sambar density (Johnsingh 1983; Karanth & Sunquist 1995). This is not the case in this study,

where densities of the two ungulates are roughly similar. Hayward et al. (2006) reviewed leopard

prey across many studies and concluded that preferred prey were likely to be in smaller groups

and in denser vegetation than avoided prey. Chital are likely to be in larger groups and in more

open vegetation than sambar, and are probably not selected because of this. In Chitwan National

Park it was observed that predation on sambar by tigers increased when chital congregated in

large herds on newly burned grasslands (Sunquist 1981). Perhaps the reason for a lack of

preference by leopards is an anti-predatory strategy of larger herd formation. Wild pig were not

an important component of the leopard' s diet in Gir National Park (Mukherj ee et al. 1994), in

Sariska National Park (Sankar & Johnsingh 2002) and in Bandipur (Johnsingh 1992) or

Nagarhole (Karanth & Sunquist 1995). In this study wild pigs were avoided, the adults are

probably dangerous prey for the leopard which likely only prey upon subadults and young. Hares

were taken in proportion to their availability by leopards and by dholes.

Along with sambar, chital is a preferred prey for the dhole. The herding behavior and

congregation by chital is not an effective strategy against a diurnal, coursing predator. Many

chases were observed, usually in the morning. The anti-predatory strategy of the chital

sometimes included running towards the village, where the dhole would not follow (Johnsingh

1983).

The diets of the three predators overlap to a great extent. The tiger diet overlaps more with

that of the leopard than the dhole because of shared inclusion of wild pig, cattle, rodents and









birds. The dhole-leopard overlap is more than the dhole-tiger overlap because the former species-

pair hunts in open areas also and both thus take a significant amount of chital, unlike the tiger.

Tigers seem to prefer large prey species that are more easily available, the mean size of

prey being 129 kg. The leopard and dhole tend to take medium sized prey. The leopard takes a

mean prey size of 27 kg, while the pack living dhole takes larger prey of 46 kg. The leopard also

takes the largest range of prey size, taking small prey like hare, birds, rodents and porcupines

that dhole did not kill in this study.










Table 3-1. Estimation of overall density and its associated parameters by the line-transect method
over 4 years in the study area.
Species n D CV D CI D Ds Cv Ds CI Ds Model
Chital 189 5.4 13.8 4.2-7.1 1.6 12.4 1.3-2.1 Hazard
Polynomial
Sambar 262 4.0 10.3 3.2-4.7 1.9 9.7 1.5-2.2 Half-Normal
Cosine
Nilgai 95 1.6 17.0 1.2-2.3 0.8 14.7 0.6-1.1 Half-Normal
Cosine
Muntj ak 63 0.8 19.0 0.6-1.2 0.7 17.3 0.5-1.1 Half-Normal
Cosine
Wild pig 63 1.8 26.2 1.1-2.9 0.6 14.5 0.4-0.7 Half-Normal
Cosine
Black-naped hare 83 3.4 15.6 2.7-4.7 3.2 15.0 2.6-4.4 Half-Normal
Cosine
Gaur 35 0.8 37.4 0.4-1.8 0.2 33.4 0.1-0.4 Half-Normal
Cosine
Common langur 637 28.3 10.3 24.1-36.3 6.4 9.5 5.7-8.3 Half-Normal
Cosine
Indian peafowl 98 2.0 20.0 1.3-2.9 1.3 17.7 0.9-1.7 Neg exp Cosine
Red spurfowl 59 2.6 20.4 1.6-3.5 1.5 18.8 1.0-1.9 Half-Normal
Cosine
Grey jungle fowl 86 2.7 17.1 1.8-3.8 1.4 16.0 1.0-2.0 Uniform
Polynomial
n: number of observations, D: density of individuals/km2, Ds: Density of groups/km2, CV:
coefficient of variation, CI: 95% Confidence. Sample size: 20 transects, effort: 1272 km.










Table 3-2. Food habits of the leopard obtained by scat analyses (N=-193 scats).
Species Weight Scats Collectable % in Bootstrapped Percent Relative
Of prey scats per kill Scat CI (95%) Biomass number
Sambar 62 102 14.9 52.8 46.1-59.6 62.2 27.1
Chital 48 39 13.1 20.2 15.0-25.9 20.7 11.6
Langur 8 21 3.5 10.9 6.7-15.5 7.0 23.7
Hare 3 11 1.4 5.7 2.6-9.3 3.2 29.2
Wild pig 37 4 11.3 2.1 0.5-4.1 1.8 1.3
Cattle 150 3 20.7 1.6 0.0-3.6 3.1 0.5
Porcupine 8 6 3.5 3.1 0.0-3.6 1.9 6.5
Rodents 0.1 6 0.05 3.1 1.0-3.6
Bird spp 5 7 2.3 3.6 1.0-6.2










Table 3-3. Food habits of the tiger obtained by scat analyses (N = 93 scats).
Species Weight Scats Collectable % in Bootstrapped Percent Relative
Of Prey scats per kill Scat CI (95%) Biomass number
Sambar 212 73 22.5 78.5 69.9-86.0 89.6 54.8
Chital 55 4 14.1 4.3 1.1-8.6 2.0 4.8
Langur 8 7 3.5 7.5 2.2-12.9 2.1 33.6
Hare 3 0 1.4 0 0
Wild pig 38 2 11.5 2.2 0.0-5.4 0.9 2.9
Cattle 180 5 21.7 5.3 1.1-10.8 5.4 3.9
Porcupine 8 0 3.5 0 0
Rodents 0.1 2 0.05 2.0 0.0-5.4
Bird spp 5 2 2.3 2.0 0.0-5.4


Table 3-4. Food habits of the dhole obtained by scat analyses (N = 81 scats).
Species Weight Scats Collectable % in Bootstrapped Percent
Of Prey scats per kill scat CI (95%) Biomass
Sambar 70 39 39.3 48.1 37.0-59.3 56.0
Chital 55 34 37.2 41.9 31.5-51.9 40.7
Langur 8 5 14.8 6.2 1.2-12.3 2.2
Hare 3 3 6.8 3.7 0.00-8.6 1.1


Relative
number
36.8
34.1
12.6
16.4
0


Wild


33.3









Table 3-5. Jacobs' index values of preference for prey species in tiger diets at study sites in
India.


Place


Chital Sambar Nilgai


Wild Gaur Langur


Bandipurl -0.30 0.07 N.P 0.77 -0.06 N.A.
Nagarhole2 -0.45 0.65 N.P. 0.68 0.26 -0.36
Pench3 0.11 0.50 -1 0.32 -1 -0.35
Ranthambore4 0.32 0.19 -0.71 -0.49 N.P -0.06
Sari ska5 0.54 0.07 -0.96 -0.71 N.P 0.30
STR6 -0.51 0.81 -1 -0.18 -1 -0.54
1Andheria et al.(2007), 2Karanth and Sunquist (1995), 3Biswas and Sankar(2002), 4Bagchi et al.
(2003b), SSankar and Johnsingh (2002), this study. NP= not present, NA not estimated.

Table 3-6. Jacobs' index values of preference for prey species in leopard diets at study sites in
India.
Place Chital Sambar Wild pig Gaur Langur
Bandipurl 0.07 -0.43 0.76 -0.42 NA
Nagarhole2 -0.02 0.62 0.27 0.05 0.10
Sari ska3 0.31 0.05 -1 NP -0.04
STR4 -0.08 0.50 -0.51 -1 -0.69
1Andheria et al.(2007), 2Karanth and Sunquist (1995), 3Sankar and Johnsingh (2002), this study.
NP= not present, NA = not estimated.









Table 3-7. Jacobs' index values of preference for prey species in dhole diets at study sites in
India.
Place Chital Sambar Wild pig Gaur Langur
Bandipurl 0.46 -0.28 0.64 -0.95 NA
Nagarhole2 -0.10 0.46 0.42 -0.94 -0.96
Pench3 -0.08 0.81 -0.56 -0.33 -0.59
STR4 0.53 0.65 -1 -1 -0.84
1Andheria et al.(2007), 2Karanth and Sunquist (1995), 3Biswas and Sankar(2002), this study.
NA = not estimated.


Table 3-8. Diet overlap between tiger, leopard and dhole using Pianka' s index.
Species Dhole Leopard
Frequency/electivity Frequency/electivity
Tiger 0.79/0.88 0.94/0.96
Dhole 0.93/0.95
























Study Area


Reservoir


River r


Figure 3-1. Map of Bori Wildlife Sanctuary and Satpura National Park, showing the location of
line transects, dirt roads and the study area.





-* Tiger
Leopard
-Y Dhole


20 40 60 80 100 120 140 160 180


20 40 60 80 100 120 140 160 180


Number of scats


Number of scats


-9 Tiger
Leopard
-7 Dhole


45-

40
20 40 60 80 100 120 140 160 180
C.
Number of scats


Figure 3-2. Relationship between sample size of scats and the percent frequency of occurrence in
tiger, leopard and dhole diet of A) Langur, B) Chital and C) Sambar.


-e Tiger
Leopard
-Y Dhole

















10-


0. hoe( 81)
a 8 Tiger (N = 93)
I I I Leopard (N = 193)

E 6-



4-








0 50 100 150 200

Num ber of scats


Figure 3-3. Relationship between the number of scats analyzed and the number of prey species
found in the diet of tiger, leopard and dhole.



































Tiger (N = 93)


100 -



80 -


Leopard (N= 193) Dhole (N= 81)

Species


Figure 3-4. Prey taken by tiger, leopard and dhole in various body weight categories.


M 10 kg
10-50 kg
S51-100 kg
>100 kg














100


-
5 0


a 40


c 0

2 20
L.


m Observed
Expected















Sambar Chital Langur Hare Wild pig

Prey Species


Sambar Chital Langur Hare Wild pig

Prey Species


S30



S20


10



0


Sambar Chital Langur Hare Wild pig

Prey Species


Figure 3-5. Observed and expected frequencies of prey items in scats of A) Tiger, B) Leopard
and C) Dhole.









CHAPTER 4
ESTIMATION OF LEOPARD (Panthera pardus) ABUNDANCE INT INDIAN
FORESTS USINTG CAMERA TRAPS INT A MARK-RECAPTURE FRAMEWORK.

Introduction

While there has been increased attention to the need for reliable estimates of

carnivore densities in India, the work has been largely restricted to tigers, Panthera tigris

(Harihar 2005; Karanth et al. 2004a; Karanth & Nichols 1998). Even basic information

on other large felids is poor, except for food habits. Leopards have been in the popular

media in India largely because of an increase in human conflicts. There is a perception

that attacks on humans have escalated in recent years, which has been attributed to

various causes, including decrease in habitat, decline in leopard prey populations,

increase in leopard densities and effects of translocations near populated areas (Athreya

et al. 2007). Unfortunately, data on leopard or prey abundances in any of the conflict

areas are lacking, and therefore the causes remain speculative.

Estimation of leopard density is, however, logistically feasible even though

leopards tend to be nocturnal, inhabit dense cover and occupy large ranges. Camera

trapping has been used in conjunction with mark-recapture techniques to estimate the

population size of species in which individuals can be uniquely identified based on the

coat patterns or other external marks. The primary method of censusing tiger, leopard and

lions (Panthera leo) by the government agency in India has been the pugmark method

(Panwar 1979). This involves taking plaster casts or paper traces of the tracks of the

targeted carnivore species in the entire survey area. The assumption is made that the

tracks of all individuals are recorded and that all individuals can be identified on the basis

of the tracings of their tracks. The method has been criticized for its subj ective nature and

the lack of incorporation of a correction for detectability (Karanth et al. 2003).









The use of statistically robust indices to monitor population trends have been

suggested, like track indices (Karanth et al. 2003), camera trapping rates (Carbone et al.

2001; Karanth & Nichols 2002) or occupancy models (MacKenzie & Nichols 2004), but

these methods do not provide an estimate of the number of individuals in the protected

area.

The mark-recapture method has long been used to estimate biological populations

(Otis et al. 1978). Recently the method has been adapted to estimate tiger populations in

India using remote camera traps. There are now estimates for tigers (Johnson et al. 2006;

Karanth & Nichols 1998; O'Brien et al. 2003), leopards (Spalton et al. 2006), jaguars

(Panthera onca) (Silver et al. 2004; Soisalo & Cavalcanti 2006) and snow leopards

(Panthera uncia) (Jackson et al. 2006) using mark recapture for other parts of the world

and it is now the accepted method. In India there are few published studies on population

estimation for carnivores other than the tiger. It is expected that more studies of leopard

abundances will soon be available for this part of the world.

Methods

Study Area

Leopard densities were estimated at three adj acent sites in Satpura Tiger Reserve

and one site in Sariska Tiger Reserve. The Satpura Tiger Reserve (22 o 19' to 22 o 30' N

and 77 o 56' to 78 o 20' E) covers 1428 km2, and is located in the Hoshangabad district of

the central Indian state of Madhya Pradesh. It consists of three administrative units, the

Pachmarhi and Bori Wildlife Sanctuaries, and Satpura National Park. The forest is mainly

the moist deciduous type (Champion & Seth 1968). The maj or ungulate fauna includes

chital (Axis axis), sambar (Cervus unicolor), Indian muntjac (M~untiacus muntjac) and

gaur (Bos guruss. The maj or carnivores are tiger, leopard, sloth bear (M~elursus ursinus)









and dhole (Cuon alpinus). The Sariska Tiger Reserve (25 o 05' to 25 o 27' N and 74ol7' to

76 o 74' E) covers 800 km2 and located in the north-western state of Raj asthan, in the

Alwar district in India. The forest is mainly the tropical dry deciduous and thorn type

(Champion & Seth 1968). The major ungulates are chital, sambar and nilgai (Bosephahts

tragocamnehis). Gaur and muntj ac do not occur there. Maj or carnivores are leopard,

striped hyena (Hyaena hyaena), jungle cat (Felis chaus) and golden j ackal (Canis

aureus). Sloth bears and dhole are absent while the tiger has recently gone locally extinct

due to illegal hunting.

Field Methods

Four sites were chosen for estimation of leopard abundances. Camera-trapping

effort at these sites ranged from 33 days (396 trap nights) to 76 days (1216 trap nights).

Three of the sites (Churna, Kamti and Lagda) were adj acent to each other in the Satpura

Tiger Reserve in central India (Figure 4-3) while the fourth was in Sariska Tiger Reserve

(Figure 4-4). Camera trap locations were chosen after reconnaissance to maximize the

probability of getting photos of leopards. Locations close to villages or on routes where

there was a great deal of human movement were excluded to minimize the possibility of

theft. Trailmaster 1550 (Goodson Associates, Lenexa, Kansas) camera traps with

Olympus and Canon autofocus cameras were deployed at all sites. At two sites (Churna

and Sariska), a one-camera setup was used at most stations, and a two-camera setup was

used at a few stations. These two- camera setup locations were changed when both flanks

of individuals in that area were obtained. At the other two sites each camera location had

a two-camera setup to photograph both flanks at the same time. Camera traps were

activated at dusk and deactivated at dawn. The minimum interval between two photos

was 6 seconds. Camera sensors were placed at a height that allowed photographs of









smaller species like black-naped hare (Lepus nigricollis) and grey jungle fowl (Gallus

sonneratii) .

Analytical Methods

All photos were scanned, printed (Fig. 2-1) and the flank of each leopard

photographed was compared to every other leopard photo. Printouts of the photos were

scrutinized under a magnifying glass to identify patterns of similar looking spots. Photos

that were underexposed due to the leopard being farther away from the camera, or where

the coat patterns were distorted because the individual was not approximately parallel to

the camera, were difficult to identify. Difficult photos were enlarged and matched on the

computer after some image processing to enhance contrast and brightness. If a pattern

was detected then a separate area of the flank was checked to confirm the identity of the

leopard. Leopards whose identities could not be confirmed were discarded and not used

in the analysis. Sometimes photos of both flanks were available, usually in cases where

two cameras were used. In one case a clear photograph of the face was available to link

the two flanks. In these cases the identity of the leopard was unambiguous, and the

leopard was included in analyses of both flanks. The number of individuals obtained from

the right flank and the left flank were compared and the dataset with the greater number

of individuals was used for the analysis.

Estimation of population size

For all sites, capture histories were developed using each day as the sampling

occasion. The capture history for each individual leopard consisted of a row vector of t

entries where t is the number of trapping occasions for each site. Each entry takes a value

of either 1 or 0 depending on whether the individual leopard was photographed on that

particular occasion or not. The entire matrix of observations for all the leopards, called










the X matrix (Otis et al. 1978) was used to estimate the population, N, and its standard

error. Program CAPTURE2 (Hines 1994) was used for the estimation. CAPTURE2

estimates the population parameters under various assumptions of the sources of

variability in capture probabilities. These are: none (Mo), individual heterogeneity (Mh),

behavioral heterogeneity (Mb) and time (Mt). The null model, Mo, corresponds to the case

which assumes that the capture probability across all individuals is the same. Model Mh

assumes that each individual has its own capture probability, and this differs from that of

all other individuals. Model Mb aSSumes that the capture probability varies after the

individual is caught for the first time, and becomes either trap shy or trap happy. Model

Mt refers to change in capture probability from one occasion to another. Models Mbh, Mth,

Mtb and Mtbh, aSSume that variation in capture probability is explained by a combination

of these sources of variation. Goodness-of-fit tests and tests of models Mo vs Mh, Mo vs,

Mb, Mo vs Mt was calculated using program CAPTURE2 where enough data was

available. A model selection procedure which scores the models according to

appropriateness using a discriminant function criterion was used (Otis et al. 1978;

Rexstad & Burnham 1991). Model Mo, the simplest model, is sensitive to violations of

the assumption of similar individual capture probabilities, so when this model was

selected, the parameters computed using the next best model have also been presented.

The test for population closure computed by program CAPTURE2 was used to detect

violation of this assumption. Also, in Churna where trapping was conducted for 150 days,

two estimates were obtained for 75 days each to enable the closure assumption to be

maintained within these two shorter sessions.









Estimation of leopard density


The Effective Trapping Area (ETA) method: Density, D, is defined as N/A,


where N iis the estimated number of leopards and A is the estimated area in which the

sampling was conducted. This area is typically the area encompassed by the trapping

grid, plus a strip of buffer around it (Dice 193 8), to obtain an ETA (Figure 4-5). The

buffering was done using both concave and convex polygons. Boundary width was

calculated using the mean maximum distance moved (full-MMDM), and half-MMDM

(Parmenter et al. 2003), to get a total of four ETAs (concave-MMDM, convex-MMDM,

concave-half MMDM, convex-half MMDM). MMDM and its standard error were

approximated by the mean of the maximum distance between two photos of each

individual leopard for all leopards photographed at more than one camera trap location.

Any portion of the ETA that lay outside the boundaries of the Tiger Reserve was

subtracted using a GIS package. A relatively small area (26 km2) was sampled in Sariska,

and so the data from this site were not used in the MMDM estimation.


The Spatially Explicit Maximum Likelihood method: Efford (2004) estimated Dj

directly from trapping data by a simulation of the trapping process. This removes the

need for a buffer width around the trapping area. The process uses the location of each

trap and includes a sub-model for the distribution of individuals and another sub-model

for the capture process. The distribution of individuals is modeled by a homogeneous

Poisson process. The capture process models the probability of capturing an individual in

a particular trap given the location of its unknown home-range center. The capture

probability is modeled using the spatial analog of the detection function (Buckland 2001).

The half-normal, hazard rate and negative exponential detection functions can be used.









These functions use the independent parameters g(0) for overall efficiency of detection

and a for spatial scale. Incorporation of sources of heterogeneity (individual-based, time-

based and behavior-based) is possible in these parameters, as in conventional capture-

recapture. However these increase the number of parameters that need to be estimated.

Because only a few animals were detected, only the null models for both parameters were

used, denoted as g(0)[.] o[.], the dot denoting lack of heterogeneity. The method assumes

that 1) Trap placement is random with respect to location of home ranges, and home

ranges are randomly oriented. 2) Home ranges do not change for the duration of the

trapping and the population is demographically closed. 3) Home-range centers have a

Poisson distribution, and 4) Individuals are independently detected. (Efford et al. in

review) provides details of the method. The software Density 4. 1 (Efford 2007) was used

to calculate the densities and associated variances using all three detection functions. The

Akaike Information Criterion (AIC) was used to select between the models, the model

with the lowest AIC being selected.

Results

A total of 288 leopard photos were obtained, twenty were unidentifiable and were

removed from the analysis. Of the identifiable photos 141 were of the left flank and 127

were of the right flank. Sampling intensity varied between sites, being lowest in Sariska,

and highest in Churna (Table 4-1).

Adequacy of Sampling

A measure of the adequacy of sampling is if new individuals are no longer

photographed with additional sampling. Figure 4-1 shows the addition of new leopards

for the 4 sites using the left flank. The shape of the curves and the number of individuals

identified were similar for the right and the left flanks. An asymptote was reached for the









sites in Satpura Tiger Reserve by 6 weeks. No asymptote was reached in Sariska

suggesting that further sampling would have yielded photographs of additional new

individuals.

Sex Ratios

The sex ratios are female biased in all areas except Kamti. The average ratio is 1.7

(SE 0.38) females per male (Table 4-2).

Population Size

The model selection criterion chose Mo for 2 sites and Mh for 3 sites. When Mo was

chosen the Mh model selection value was not much lower, though the difference was

significant or marginally non-significant (Table 4-3). Mo is not recommended because it

is sensitive to departures from the assumption of no individual-based heterogeneity

(Karanth & Nichols 1998), though both models have been presented. All Mh WeTO

estimated with the jackknife estimator, which is robust and has performed well in

simulation studies (Burnham & Overton 1979). Test for population closure was not

significant for all the sites, indicating that the assumption of demographic closure was not

violated. A high proportion of the estimated population was photographed, ranging from

69 to 89 percent for the Mh model and 69 to 100 percent for the Mo model. Population

sizes, capture probability and estimated proportion photographed for both estimators are

given in Table 4-4.

Leopard Density

Table 4-5 gives the estimation of density of leopards per 100 km2 at the different

sites using the convex polygon to calculate the Effective Trapping Area method with full-

MMDM and half-MMDM. Estimated density is dependant on the method used to

calculate the strip width and the polygon. The densities calculated using all combinations









of concave and convex polygon with half MMDM and full MMDM are presented in

Table 4-6. Concave polygon with half MMDM gave the smallest effective trapping area

and consequently the highest density, while the convex polygon with MMDM gave the

largest effective trapping area and therefore the lowest density at each site.

Using the maximum-likelihood-spatially-explicit-captr-eatr method, the

lowest AIC values for Churna and Kamti were obtained by the four parameter hazard rate

model, while for Lagda and Sariska the three parameter half normal model was selected.

Densities obtained by this method are given in Table 4-7.

The relative abundance index (Table 4-8) was also highest for Sariska followed

by the second session at Churna.

Discussion

Ideally, it is desirable to obtain photos of both flanks of the body so that

identification of individuals is unambiguous. When camera numbers are limited, it seems

possible to obtain unambiguous photographs of both flanks for a large proportion of the

population using two cameras at a few locations, while using one camera at the remaining

locations, provided the trapping goes on for a long period. This would maximize

coverage of the area with the available number of cameras. The individual identification

of leopards from photographs was found to be quite easy except when the animals walked

farther away from the cameras, resulting in underexposed photos. This was likely to

happen when the distance between the two sensors was more than 10 meters.

Tigers were sometimes observed to avoid camera traps, leaving the trail just before

the camera location and getting back on the trail afterwards. Other studies have also

observed this behaviour (Wegge et al. 2004). On the evidence of tracks, leopards were

never observed to avoid cameras traps, and showed no response to the flash. Leopards of









both sexes were photographed while standing or sitting in front of the camera, and did not

rapidly move away. Sometimes more than one photograph was taken at the same time,

indicating that the leopard stayed in that position for at least 6 seconds after the flash of

the first photograph. However, rates of photo-captures for males seemed to be

consistently higher than for females. The existence of heterogeneity in capture

probabilities with respect to gender is possible. In Kruger National Park for instance, it

was easier to capture males as opposed to females in box traps (Bailey 1993).

The calculation of effective area of sampling is a noteworthy issue in the estimation

of density using camera traps. There is generally no measurement of the home range of

the sample of individuals used in the estimation. It has been recommended that half the

mean maximum distance moved (MMDM) be used as the buffer for estimation of

densities (Wilson & Anderson 1985). A recent study on jaguars comparing MMDM

obtained by telemetry to half MMDM and full MMDM found that the full MMDM

results were much closer to densities based on actual movement rates, and that half

MMDM seemed to overestimate densities (Soisalo & Cavalcanti 2006). There is still not

enough data available for movement in leopards to advocate a shift to full MMDM. Also,

as the area of the buffer increases, it is more likely to include habitat that is unsuitable for

the species and unrepresentative of the probability of capture at the camera trap location.

In this study densities obtained using the convex polygon-full MMDM gave results

that were similar to the MLSECR method at most sites, while densities calculated using

half MMDM were much larger (Table 4-8). The density of leopards was highest in

Sariska Tiger Reserve, where tigers have been extirpated recently (Sankar et al. 2005),

while it was lowest at Lagda, which had the highest activity of tigers amongst all sites









(pers obs) though it is not a high density tiger area. Variation in density of carnivores is

associated with density of prey as shown for tigers (Karanth et al. 2004b), and other

carnivores (Carbone et al. 1999). However, other factors, like human disturbance

(Woodroffe 2000) and tiger presence (Seidensticker 1976; Sunquist 1981) may also play

a role, although there is some evidence that leopard densities may not be unduly

depressed by presence of other large carnivores (Marker & Dickman 2005).

Leopard density estimates are available for various parts of the world, but from

different methods. For the Serengeti it was about 3.8-4.5 per 100 km2 (Schaller 1976), for

Kruger it was about 3.4 per 100 km2 (Pienaar 1969). It was estimated at about 7. 1 in the

rain forest of the Ivory Coast (Jenny 1996), and in Wilpattu National Park in Sri Lanka it

was estimated as about 3.4 (Eisenberg & Lockhart 1972). In Namibia, a mean of 10.5 (SE

4.0) inside protected areas (n =6), and 2.1 (SE 1.6) outside protected areas has been

reported (Marker & Dickman 2005).

Photocapture rates calculated per 100 trap nights for 4 sites in India ranged from a

low of 0. 18 for Kaziranga National Park, a medium 2.3 for Pench National Park to a high

of 5.44 in Nagarhole National Park (Karanth & Nichols 1998). Estimates of Relative

Abundance Index (RAI) for the present study, ranging from 2.2 to 6.8 (Table 4-7) seem

to be within the range found in other areas in India.

The second session in Churna, conducted in spring-summer, had higher capture

probability than the first session, conducted in winter-spring. Camera traps were mostly

placed along topographic contours, where leopard signs were high, and water tended to

be found. It is possible that movement of leopards around such places increased in









summer when water sources in the hills dried up, leading to the higher capture

probability.

RAI has been recommended for tigers when there is not enough data for mark

recapture sampling (Carbone et al. 2001). In low leopard density areas it takes a long

time to get a sufficient number of captures to use in the mark-recapture framework. If the

assumption of population closure is severely violated, then the RAI may be used as a

substitute for density estimation. An index can also be used on species that do not have

individually identifiable markings. However, the difference between the RAI estimates

and density estimates is noteworthy. The second session at Churna and the session at

Sariska have similar RAI values, but the density at Sariska is much higher. Similarly, the

RAI of the first session at Churna is almost 3 times lower than the RAI of the second

session, but density estimates are not significantly different. This indicates that RAI does

not seem to index density in a reliable way, as noted elsewhere also (Jennelle et al. 2002;

Maffei et al. 2004).

Conclusion

Sariska Tiger Reserve has the highest densities despite having a history of human

disturbance and poaching. This may be related to the recent removal of the tiger from

Sariska and the occupation of prime habitats by the leopard. Another reason could be the

smaller spatial extent of the effective trapping area in Sariska. It may be that the

distribution of individuals in Sariska is more patchy and that the lower density areas were

not surveyed. In such a scenario, comparing Sariska to another study site will not be

useful and the parameter values should be limited to monitoring the same site over time.

Leopard RAI values in Bori-Satpura are comparable to other study sites in India. The









leopard RAI in Satpura, where tiger density is relatively low, seem to be higher than at

Kanha and Pench, which have higher tiger densities.

The estimates provided by the mark-recapture framework give us a relatively

robust measure of population size, but the estimation of density is still problematic given

the uncertainty involved with estimation of the effective trapping area. The spatially

explicit maximum likelihood method offers a solution to that problem, but modeling

heterogeneity is more complicated with low population sizes, since the number of

parameters to be estimated is high. The precision derived in the present study makes it

difficult to detect changes in population density. It is logistically difficult to both sample

at an intensity that obtains high precision and at a large enough spatial scale for a species

of this size. Given these limitations, serious investigation should be made into the use of

indices to monitor population changes with greater precision, though RAI does not seem

to be the appropriate index in these study sites.











Table 4-1. Camera-trapping effort (in trap nights) at the study sites.
Site Number of camera Number of Effort
trapping stations Nights ( trap nights)
Churna (session 1) 16 76 1216
Churna (session 2) 16 75 1200
Kamti 20 52 1040
Lagda 20 33 660
Sari ska 12 33 396



Table 4-2. Leopard sex ratios for the different study sites.
Site Males Females Sex ratio
(no of females
per male)
Churna sessionn) 4 6 1.5
Churna sessionn) 3 8 2.7
Kamti 7 4 0.6
Lagda 3 5 1.7
Sari ska 3 6 2.0












Table 4-3. Model selection criterion and tests for Models Mo, Mh, Mb and Mt in the mark-recapture framework and a test for
population closure for the different study sites.
Site Model selection Mo vs Mh Mo vs Mt Mo vs Mb Mh Goodness of Closure test
criterion fit
Mo Mh Mb Mt X2 df p X2 df p X2 df p X2 df p z p
Churna 1.0 0.94 0.51 0.0 Not done 2.5 76 1.00 1.2 1 0.26 85.4 76 0.22 -0.6 0.28
session

Churna 0.93 1.00 0.46 0.0 5.8 2 0.05 14.5 72 1.0 0.01 1 0.91 106.5 72 0.00 -0.13 0.45
session

Kamti 0.96 1.00 0.44 0.0 3.6 1 0.06 3.9 82 1.0 0.03 1 0.86 97.4 82 0.11 -1.33 0.09
Lagda 0.93 1.00 0.38 0.0 Not done 8.0 32 0.99 0.00 1 0.97 59.03 32 0.00 -1.36 0.09
Sari ska 1.00 0.91 0.42 0.0 Not done 2.2 28 1.0 Test failed 32.91 28 0.24 1.24 0.11










Table 4-4. Population estimates for leopards at the study sites.


Site


Estimate (Mo)


Estimate (Mh)


0.92


0.79


N ASE

1413.6


p

0.03

0.08

0.03
0.08
0.04


N ASE p

1211.5 0.02


Churna
session
Churna
session
Kamti
Lagda
Sari ska


1.0 1110.14 0.07


0.79 1412.6


1.0
1.0
0.69


1010.81
810.76
1313.8


0.03
0.07
0.04


0.83
0.89
0.69


1213.7
916.9
1314.4


Table 4-5. Density of leopards and estimates
Mh at the different study sites.
Sites Churna


sessionn)
152.2


of sampled area using convex polygon and model


Churna
sessionn)


Kamti Lagda Sariska


Estimates
Effective area (half
MMDM) m2.
Density (per 100 km2)

Effective area (full
MMDM) m2.
Density (per 100 km2)


149.2 119.3 122.7


44.4


8.012.5

230.8

5.314.7


9.3+2.0 7.512.8 7.315.1 30.9112.1


223.6 195.0 210.9


66.2


6.211.6 4.6+2.0 4.213.1 20.7110.0










Table 4-6. Density of leopards with the associated estimated trapping area using models Mo and


Mh
Polygon
method


DensityiSE (per 100 km2)


Site


Strip
method


ETA
(km2)


185.6
77.9
230.8
152.2
176.8
73.0
223.6
149.2
179.5
92.3
195.0
119.3
194.8
97.7
210.9
122.7
54.6
21.1
66.2
44.4


Churna Concave
session Concave
Convex
Convex
Churna Concave
session Concave
Convex
Convex
Kamti Concave
Concave
Convex
Convex
Lagda Concave
Concave
Convex
Convex
Sariska Concave
Concave
Convex
Convex


MMDM
MMDM/2
MMDM
MMDM/2
MMDM
MMDM/2
MMDM
MMDM/2
MMDM
MMDM/2
MMDM
MMDM/2
MMDM
MMDM/2
MMDM
MMDM/2
MMDM
MMDM/2
MMDM
MMDM/2


6.4+2.0
15.413.3
5.211.6
7.811.7
6.211.1
15.011.6
4.910.8
7.310.8
4.511.3
8.71.8
4.111.2
6.71.4
4.113.4
8.211.8
3.811.7
6.511.4
23.8~111
61.6125
19.619.1
29.2110.8


6.612.5
15.714.8
5.314.7
8.012.5
7.912.0
19.114.1
6.211.6
9.3+2.0
5.012.1
9.713.6
4.6+2.0
7.512.8
4.613.4
9.116.4
4.213.1
7.315.1
25.2112.1
65.1125.5
20.7110.0
30.9112.1










Table 4-7. Density estimates for leopards (number/100 km2) USing different capture functions for
the null models with the MLSECR method.


Site

Churna
session
Churna
session
Churna
session
Churna
session


Capture
function
Hazard

Negative
exponential
Half
normal
Hazard


Model No of
param
g0[.]o[.] 4

g0[.]o [.] 3

g0[.]o [.] 3

g0[.]o [.] 4

g0[.]o [.] 3

g0[.]o [.] 3


Log
likelihood
-172.29


AIC


AAIC Density


SE

3.21


352.59


0 7.21


-173.52 353.04

-175.03 356.07

-364.57 737.15


0.45

3.48


6.62 2.73


5.92


2.3

1.37

1.25

1.14

0.05
1.61

1.58

1.41

1.23

0.19
7.0

0.4
6.57


0 4.04


Churna Half
session normal
Churna Negative
session exponential
Kamti Hazard
Kamti Half
normal
Kamti Negative
exponential
Lagda Half
normal
Lagda Negative
exponential
Lagda Hazard
Sariska Half


-369.75 745.49 8.34

-372.9 751.8 14.65


3.83

3.51

4.67
4.15


g0[.]o [.]
g0[.]o [.]

g0[.]o [.]

g0[.]o [.]

g0[.]o [.]

g0[.]o [.]
g0[.]o [.]

g0[.]o [.]
g0[.]o [.]


-179.45
-183.64


366.9
373.28


0
6.38


3 -183.81 373.62 6.72 4.08


3 -149.7 305.4

3 -149.92 305.83


0 3.27


0.43

0.78
0

0.81
0.85


3.11

3.44
14.58

20.08
12.65


-149.09 306.18
-73.75 153.49


normal
Hazard
Negative
exponential


Sari ska
Sari ska


-73.15
-74.17


154.3
154.34










Table 4-8. Relative abundance index values for the 5 estimates in Satpura and Sariska Tiger
Reserves.


Site


No of camera trap
locations


No of independent RAI (per 100 SE of RAI
captures trap nights)
27 2.2 0.65
80 6.7 1.85
39 3.9 0.71
24 3.8 0.89
27 6.8 2.21


Churna (session 1)
Churna (session 2)
Kamti
Lagda
Sari ska
















































Figure 4-1. Identification of leopards based on spot patterns. The first two photos are of the same
leopard, the third photo is of a different leopard.














83















er 12








- O






.0 4 -1 +~rI II-C Kamti
I I --O Lagda
-9- Churna session
2 ~-A-~- Churna sessions
'si --M- Sariska



0 2 4 6 8 10 12 14

Weeks

Figure 4-2. Rate of accumulation of new individuals in camera-trap photographs with increase in
sampling time at the four sites.




































Kamti
10~~ ~ *ioetr C hurna
0 10Kiloeter / Map of Satpura Tiger
NReserve


Figure 4-3. Camera trapping in 3 sites (Churna, Kamti and Lagda) in Satpura Tiger Reserve.
















































Carnera trap locations
Shalf-MMDM buffer
M full-MMDM b~uffer
0 30 KilDm etf5.
I Map of Sarlska Tiger
Reserve

Figure 4-4. Map showing camera trap locations with half MMDM and full MMDM buffers in
Sariska Tiger Reserve.


Sariska Tiger
Reserve


rm r rm rim ~alrr





























Bonl Wildlife
Sanctuary









0 10 Kilometers


p,,,,,,r~l, ,,,, -rl


Kamti camera trap
locations
haF-MMDM buffer
full-MMDM buffer
Satpura Tiger
Reserve


N


Figure 4-5. Map showing camera trap locations with half MMDM and full MMDM buffers for
one site (Kamti).









CHAPTER 5
PRESENCE-ONLY HABITAT SUITABILITY MODELS FOR LEOPARDS (Panthera pardus)
USINTG FIELD BASED AND REMOTELY DERIVED VARIABLES AT TWO SPATIAL
SCALES INT MADHYA PRADESH, INDIA.

Introduction

Knowledge of the distribution and habitat requirements of a species are essential to

formulate conservation strategies. While some species are considered habitat generalists, they are

still vulnerable to habitat loss and fragmentation. These factors along with prey depletion and

poaching are responsible for the decline of the tiger (Panthera tigris) across its geographic

distribution (Sunquist et al. 1999). It has been estimated that the tiger exists in only 7 percent of

its historical range (Dinerstein et al. 2007). The leopard is a wide-ranging large carnivore that is

less susceptible to disturbance, is a generalist with respect to habitat requirements, and can

survive on a wide range of prey species (Sunquist & Sunquist 2002). Unlike the tiger, which

needs a high biomass of large-sized prey (Karanth & Sunquist 1995), the leopard has been

known to survive on domestic dogs and rodents in the absence of wild prey populations

(Edgaonkar & Chellam 2002). As tiger populations in India have declined, leopard populations

have also come under increased poaching pressure. Conserving leopards in this environment will

require a quantification of habitat requirements and identification of potential habitat availability

in India. Good habitats for leopards can then be given conservation priority in protection and

management strategies.

Categorizing suitable leopard habitat requires information at multiple scales. First-order

selection (Johnson 1980) refers to the distribution of a species with respect to geographical

space. Large-scale species distribution models can be used to guide conservation strategies

(Guisan et al. 2006; Hirzel et al. 2004; Mladenoff & Sickley 1998; Seoane et al. 2006).

Techniques like logistic regression (Karlsson et al. 2007; Woolf et al. 2002) and generalized









linear models (GLM) (Austin 2007; Bustamante & Seoane 2004) use the information from

multivariate measurements of habitat variables at locations with species presence and at locations

where the species is absent (Guisan & Zimmermann 2000; Meynard & Quinn 2007). This

information is used to derive a probability of species presence at each location. Though these

methods are preferred when absence data are reliable (Brotons et al. 2004), logistic regression

models are known to be sensitive to even low levels of non-detections (Gu & Swihart 2004).

Leopards are not only rare and secretive, they are also crepuscular (Sunquist & Sunquist 2002)

and without intensive effort there is a high likelihood of non-detections in areas where leopards

are present, contaminating the absence data. Presence-only models are a way of dealing with this

problem. This paper uses environmental niche factor analysis (ENFA) (Hirzel et al. 2001), a

presence-only environmental habitat-envelope based method to create habitat suitability maps for

the leopard in south-central India. ENFA has been used successfully to model the distributions

and habitat suitability of a variety of taxa: dung beetles (Chefaoui et al. 2005), corals (Bryan &

Metaxas 2007), reptiles (Santos et al. 2006), birds (Braunisch & Suchant 2007; Brotons et al.

2004; Olivier & Wotherspoon 2006; Reutter et al. 2003; Titeux et al. 2007), ungulates (Dettki et

al. 2003; Traill & Bigalke 2007) and carnivores (Mestre et al. 2007).

The objectives of this paper are: 1) To develop predictive habitat suitability maps for

leopards at two scales and evaluate their reliability; 2) To identify the environmental variables

important in describing the habitat for this species, and 3) To quantify the extent and location of

potential leopard habitat available for conservation action in south-central Madhya Pradesh.

Study Areas

The extensive study area covers 52971 km2 (Table 5-1) and includes thirteen districts in

south-central Madhya Pradesh; it comprises about 18 percent of the state of Madhya Pradesh.

Altitudes in the study area range from 215 to 1312 m. Annual rainfall for the state averages 1143









mm, with rainfall decreasing from the eastern part of the state to the west. The landscape is a

mosaic of forests, agriculture, villages and small and large towns (Figure 5-1). The main crops

are wheat, soybean, sorghum, sugarcane and pulses. The forests are mainly teak dominated, as

well as dry and moist deciduous forests. The climate is cool in winter and very hot in summer,

with temperatures ranging from 2-450 C. The largest river in region is the Narmada. The two

main protected areas within the landscape are the Satpura and the Pench Tiger Reserves (Figure

5-2).

The intensive study site (Figure 5-3) consists of a 433 km2 area of moist and dry deciduous

forests along with some teak plantations located inside the Satpura Tiger Reserve (STR). It is

located in the center of the extensive study site. Topography ranges from relatively flat to very

steep slopes and cliffs; altitudes range from 300 to 1315 m. There are 7 small forest villages

within its boundaries. Details of the intensive study are given in previous chapters.

Methods

In the STR study site, visual sightings of prey species were obtained from walking 20

straight-line transects (630 km) through the forest, and from driving-transects using a 4-wheel

drive vehicle at 10-15 kmph along a network of dirt trails (369.5 km). An encounter rate was

calculated as the number of sightings per kilometer using all sightings of potential prey species.

Potential prey species include chital (Axis axis), sambar (Cervus unicolor), langur

(.\'llininsithe us\ entellus), wild pig (Sus scrofa), and a small-prey category comprising hare

(Lepus nigricollis), peafowl (Pavo cristatus), red spurfowl (Galloperdix spadicea) and grey

jungle fowl (Gallus sonneratii). This encounter rate was then divided into 5 categories. The first

category was 0 encounter rate, and the other 4 were based on equal quantiles (25th, 50th, 75th and

100th). Photos of prey from the camera-trap stations were converted into a rate per trap night, and









also similarly divided into 5 increasing categories based on equal quantiles. The two data sources

(encounter rates and photo trap rates) were then assumed to be equivalent indices of prey

abundance and were subsequently merged. Distance-weighted interpolation of the categorical

index was then done using the INTERPOL module of Idrisi Kilimanj aro v14.02 (Eastman 2004)

to obtain a prey map.

Sampling for evidence of leopard presence was done using kills, tracks, scrapes and

camera-trap photos. All quantitative measures were degraded into presence-absence measures to

reduce biases introduced by different sampling efforts. Multiple instances of presence within a

one-hectare plot were combined to reduce spatial autocorrelation, which can lead to bias in

precision estimates for habitat models (Diniz et al. 2003).

Secondary data were obtained as part of a j oint Wildlife Institute of India- Proj ect Tiger

initiative to monitor tiger populations in India in 2006. A total of 2582 beats were sampled. A 3

to 4-km-long transect was located in each beat, and each beat was a walked a total of three times

by the local forest guard in charge of the beat. The average size of each beat was 20 km2. Data on

presence of livestock signs, encounter rate of prey and sign of leopards were collected. Digitized

beat maps were obtained and the centroid of each beat was used to approximate the location of

leopard presence in the beat.

Ecogeographical variable (EGV) maps of prey encounter rates for sambar, nilgai

(Bosephahts tragocamnehis) and wild pig using the INTERPOL module of Idrisi were created.

The encounter rate of leopard sign was converted to a binary variable of presence and

pseudoabsences. A buffer width of 3000 m was applied to create a 9 km2 patch of leopard-

presence pixels around each point. Female leopard home ranges are known to vary from 6 to 30

km2 in Africa (Bailey 1993) and average 17 km2 in Nepal (Odden & Wegge 2005), so 9 km2 was









considered a conservative estimate of the area in which presence could be assumed in the forest

beat. Only beats where some evidence of leopards was detected were retained for the analysis.

All beats where evidence of leopards was not detected were discarded from the dataset.

The elevation layer was obtained from the 90-m resolution DEMs created from the SRTM

mission data by the CGIAR-CSI (http://srtm.csi .cgiar. org ). Using Idrisi, a slope map and a

ruggedness map (using the standard deviation of mean elevation in a 3 x 3 moving window) was

created from the DEM. A moving window of 3x3 has been used to create a similar index of

terrain ruggedness to model mountain lion habitat in Montana (Riley & Malecki 2001)

The extensive study area encompassed parts of 6 Landsat ETM+ images. Georeferenced

and orthorectified cloud-free images dating between 2002 and 2004 were obtained from the

Global Landcover Facility (http://glcf.umiacs.umd. edu ). Those parts of the extensive study area

found in each of these images were classified into four cover types: agriculture, bare

ground/urban, forest and water. These were then mosaicked together to obtain the cover map.

For the STR area 5 cover types were delineated. These were: moist forest, dry forest, bare

ground/village, teak dominated forest and water. Spectral signatures for the classification

supervision were obtained by using information from 473 vegetation plots in the Satpura Tiger

Reserve, and with visual inspection of satellite imagery using Google Earth

(http//:www. earth.google.com) for the extensive study area. Supervised classification was

performed using FISHER classifier for both the study sites using the Idrisi GIS package. Using

the CircAnn module of Biomapper, the Boolean maps of each land cover type was converted to

percent frequency in a 20 km2 circular moving window for the extensive study area (Figure 5-4)

and 1 km2 for the STR site (Figure 5-5). The models were made at two pixel resolutions: 1000 m









for the extensive study area and 100 m for the STR study site. A list of all the EGVs for both

study areas is given in Table 5-2 and 5-3.

To observe the effect of pixel resolution on the accuracy of the models, the ENFA analysis

was repeated at the 200-m, 300-m and 500-m resolution for the STR area, and at 1000-m without

the buffer, at 2000-m, 3000-m and 5000-m for the extensive study area.

The relationship between the distribution of leopard presence patches and a set of mapped

ecogeographical variables was analyzed using ENFA. The program Biomapper v3.2 (Hirzel et al.

2006a) was used. Biomapper needs two types of data to calculate habitat suitability. The first is a

map of locations where the species has been detected, and the next are a set of quantitative raster

maps describing the environment as used by the species under investigation. This presence-only

modeling technique describes the ecological niche of a species by computing uncorrelated

factors from a comparison of values of ecogeographical variables in the entire study area and

their values at the site where the species is known to be present. The first ENFA factor

maximizes the absolute value of the marginality, defined as the standardized difference between

the species mean and the global mean of each of the EGVs. The first factor explains how the

species niche differs most from the available conditions. The first factor also explains all the

marginality and some of the specialization. Specialization is defined as the ratio of the overall

variance to the species variance for all the EGVs, and describes how restricted is the usage of the

species of that variable compared to its availability. Details on the calculation of marginality and

specialization are given in Hirzel et al. (2002). The subsequent factors maximize the

specialization. A high absolute value of the correlation of the variable with the specialization

factor indicates that the species niche breadth is narrow with respect to that variable. There are as

many factors as there are variables, but they successively explain a decreasing amount of the









specialization. The number of factors used to calculate the habitat suitability was decided using

MacArthur' s broken-stick criterion (Hirzel et al. 2002) .

The habitat-suitability map was evaluated for its predictive accuracy by internal area

adjusted frequency cross-validation (Fielding & Bell 1997). Leopard presence were

geographically stratified and randomly partitioned into 10 sets. Nine partitions were used to

compute a habitat suitability model and the left-out partition was used to validate it on

independent data. This process was repeated 10 times, each time by leaving out a different

partition. This process resulted in ten different habitat-suitability maps. Each map was

reclassified into 4 bins, where each bin covered some proportion of the total study area (Ai) and

contained some proportion of the left-out validation points (Ni).The area-adjusted frequency for

each bin was computed as Fi = Ni /Ai. The expected Fi was 1 for all bins if the model was

completely random. If the model is good, low values of habitat suitability should have a low F

(below 1) and high values a high F (above 1) with a monotonic increase in between. The

monotonicity of the curve was measured with a Spearman rank correlation on the Fi in a moving

window, termed as the continuous Boyce Index (Boyce et al. 2002; Hirzel et al. 2006b).

Validation of the models was also done using the Absolute Validation Index (AVI) and the

Contrast Validation Index (CVI). AVI is the proportion of validation points that have a habitat

selection of >=50. Possible values the index can take range from 0 to 1. The higher the value the

more accurate is the model. CVI is calculated at AVI minus the AVIchance, which is the AVI

one would expect from chance alone, and is a measure of departure from randomness of the

model. Possible value the index can take range from 0 to AVI. One criticism of the presence

models is that they yield too optimistic results (Zaniewski et al. 2002). This problem was

mitigated by using breaks in the predicted-to-expected ratio frequency curves to define 4 habitat









classes (Hirzel et al. 2006). The map was then reclassified using the new bins into unsuitable,

marginal, suitable and optimal habitat.

Results

Model Validation

Overall the habitat suitability models for both STR and the extensive study area were

equally accurate. They both showed similar values of AVI, indicating that the proportional

accuracy in classifying presence points in the evaluation partition was similar for south-central

Madhya Pradesh and for Satpura Tiger Reserve. CVI values showed that the model had some

difficulty in discriminating between the suitability map and a purely random model. This is

consistent with the generalist nature of the species. Both the continuous Boyce Index values were

high, indicating good predictive power for both the models, but the extensive study area model

had better predictive power (Table 5-4). The predicted-to-expected frequency curves showed

higher variance for good habitat than for bad habitat with both models (Figure 5-8), the

inflections in the curves were used to guide the selection of bins to reclassify the habitat

suitability maps for the two areas (Figure 5-6 and 5-7).

Extensive Study Area

The marginality value was 1.25 and the tolerance value was 0.92, indicating that leopards

were using conditions that were different from the mean environmental values, and that the

leopard was more of a generalist in using a wide range from the EGVs. Seven factors were

retained. The first factor accounted for 100 % of the marginality, while all 7 factors accounted

for 100 % of the marginality and 80% of the specialization. The marginality coefficients show

that leopard habitat was more positively correlated with sambar distribution, terrain ruggedness

and percentage of forests. It was less strongly correlated with altitude, slope, NDVI and nilgai

and wild pig encounter rates. Leopard distribution was negatively correlated with presence of










agriculture and urban-bare ground land cover types. Livestock presence, an indicator of human

disturbance, was a weak negative correlate. The specialization factor indicated that the leopard

used a restricted niche with respect to the availability of percentage frequency of urban-bare

ground and agriculture, but not when compared with the availability of elevation, ruggedness and

slope measured at the 1 km scale across the big study area (Table 5-5). The amount of suitable,

marginal, unsuitable and optimal habitat in each district is given Table 5-7. Maps of the EGVs

are shown in Figure 5-4.

Satpura Tiger Reserve

The marginality value was 0.67, indicating that leopards were using conditions not too

different from the mean environmental values. Tolerance was also relatively high (0.56),

indicating that the leopard was found in areas that had a wide range of values of the EGVs. Four

factors were retained, the first factor accounting for 100 % of the marginality. The four factors

explained 79 % of the specialization. The marginality factor was strongly positively loaded with

the coefficient for tassled-cap 'greenness', an index of above ground biomass (Crist & Kauth

1986) and percentage frequency of moist forests and teak dominated forest. It was also positively

correlated with distance to water and the encounter rate of cervids sambarr and chital), wild pig

and small-sized prey. The positive correlation with langur encounter rate was weak. Leopard

presence was negatively correlated with elevation, slope and frequency of bare ground pixels.

The negative loading with respect to distance from village was weak. The specialization factor

indicated that elevation was used in a more restricted way than was available in the study area, as

was the frequency of the moist, dry and teak forests (Table 5-6). Maps of the EGVs are shown in

Figure 5-5.









Effect of Changing Resolution

For the extensive study area the best model was at the 1000-m scale with buffer. It gave the

highest continuous Boyce Index value. The coarsest resolution model was the most inaccurate.

At a resolution of 5 km and a moving window size of 225 km2, the continuous Boyce Index

reduced to 0.55. Changing the resolution did not change the AVI, CVI and the Boyce Index

much at all the other resolutions. For the STR area, the best model was the 100-m resolution with

a moving window of 1-km2, followed by the 200-m model. The effect of increasing the moving

window scale degraded accuracy slightly. The 300-m and 500-m resolution models had lower

Boyce Index values (Table 5-4). The habitat suitability maps for the two areas were made from

the best models.

Discussion

The leopard is an adaptable species, being able to live in a wide variety of environmental

conditions. This is reflected in the marginality and tolerance values for the model of the STR

area, where almost all the area is potential leopard habitat. Habitat use by leopards in Satpura

was strongly associated with moist and teak forests, as well as with most prey species, except the

langur, with which it was only weakly associated. This is because more langur are seen in open

areas, closer to villages, and along roads, rather than in denser forest areas (pers obs), perhaps as

an anti-predatory strategy, and they do not comprise a large proportion of the leopard' s diet in

this area (Chapter 3). Leopard presence had a weak negative association with the distance to

villages. That means it was found closer to villages than average, though this tendency was

weak. Unlike tigers, which are shy and prone to move away from disturbance, leopards are

known to be bold and not uncommonly found in proximity to human habitats, where they prey

upon livestock (Odden & Wegge 2005). Though they are tolerant of human presence, they are

not unaffected by disturbance, as the extensive study area model showed, with leopard habitat









being negatively associated with bare ground/urban land use frequency. In Thailand leopard

activity has been shown to be negatively correlated with distance from villages (Ngoprasert et al.

2007). Leopard habitat was negatively correlated with urban-bare ground and agriculture land

cover types as also with livestock presence. At the large scale, good leopard habitat was seen to

be more associated with terrain ruggedness, sambar availability and percentage of forested areas,

and less associated with nilgai and wild pig prey availability. Both the latter species are known to

be crop pests and able to live close to human inhabitation (Sekhar 1998), and this probably

contributes to the observed pattern. Cougar (Puma concolor) abundance has also been shown to

be affected by prey availability, terrain ruggedness and forest cover at the landscape scale (Riley

& Malecki 2001).

The larger spatial area model had a higher predictive accuracy than the smaller scale as

quantified by the higher continuous Boyce Index (Table 5-4). This is possibly because the

Satpura Tiger Reserve has relatively little disturbance and is a less heterogeneous area given its

smaller size. Given the high density of leopards in the area (Chapter 4) and that they require

relatively large tracts of contiguous habitat (Marker & Dickman 2005), they probably move

through and spend time in habitats that are not highly preferred, but are still inhabitable.

Consequently, very few areas in the Reserve are likely to be completely unsuitable for leopards.

The change in resolution seemed to have a similar impact on models of both study areas.

Coarse pixel resolutions, at 300 m and 500 m for STR and 5 km for the extensive study area,

degraded the accuracy of the models. The scale of the circular moving window for frequency of

land use cover did not change accuracy appreciably, except at the very largest spatial resolution

(225 km2)









The habitat model was used to estimate the area occupied by various habitat categories in

the 13 districts in south-central Madhya Pradesh (Table 5-7). 'Optimal' habitat was 5.2% of the

study area, ranging from 0.5 to 8 percent of each district. As an absolute measure it can be said

that approximately 1 1500 km2 Of habitat is likely to support leopard populations. The districts

with the most optimal habitat are Betul, Hoshangabad and Chhindwara. These districts are

geographically adj acent to each other and constitute a compact block of about 2000 km2 Of

optimal habitat. The Satpura Tiger Reserve lies in Hoshangabad district and is already protected,

but Betul and Chhindwara districts can be prioritized when allocating resources for leopard

conservation efforts in Madhya Pradesh. In conclusion the ENFA model seems to work better at

larger spatial areas for a generalist species like the leopard. It is a useful tool to explore the

characteristics of the leopard' s niche as well as to produce habitat suitability maps that can aid in

conservation management.










Table 5-1. Districts, sampling effort and leopard presence in the extensive study area in south-
central Madhya Pradesh.
District Sampled Area (km2) Number of transects Transects with Leopard presence
Balaghat 419.9 50 0
Betul 10041.5 622 23
Bhopal 57.1 1 0
Chhindwara 11815.8 553 57
Dewas 1296.7 30 0
East Nimar 2104.4 100 8
Harda 3329.0 163 2
Hoshangabad 6734.1 351 111
Jabalpur 258.5 27 2
Narmsimhapur 4420.3 107 17
Raisen 3293.9 92 16
Sehore 3266.8 112 23
Seoni 5891.4 361 41

































Same as above
Same as above
Calculated.




Same as above

Same as above


Same as above


Calculated.


None
None
Square root



Box-Cox

Box-Cox


Box-Cox


None


Table 5-2. List of ecogeographical variables (EGV)
central Madhya Pradesh.
Ecogeographical Explanation
Variables


Elevation



Ruggedness
(Elevation standard
deviation)
Slope
NDVI


Forest





Urban/bareground
Agriculture
Livestock




Nilgai encounter
rate (ER)
Sambar ER



Wild pig ER


Distance to Water


with explanation and source for south-


Transformation


Source


DEM in meters at 100 m
spatial resolution, averaged
to 1 km spatial resolution.
Calculated with a moving
window of 3x3 cells from
DEM.
Calculated from DEM
Calculated from bands 3 and
4 of Landsat ETM +
imagery .
Percentage frequency of
cells with forests,
urban/bareground and
agriculture in a circular
window area 25 km2
Same as above
Same as above
Di stance-weighted
interpolation of encounter
rate (number seen/km) from
line transects.
Same as above

Interpolated encounter rate
(number seen/km) from line
transects.
Interpolated encounter rate
(number seen/km) from line
transects.
Distance to the nearest
water source in meters.


None


None


None
None


None


SRTM data


Calculated


Calculated
Calculated


Supervised classification
of Landsat ETM+ imagery
to obtain landcover;
Frequency calculated.










Table 5-3. List of ecogeographical variables (EGV) with explanation and source for the Satpura
Tiger Reserve.


Ecogeographical
Variables
Cervid Encounter
Rate (ER)



Langur ER
Wild pig ER
Small Prey ER


Bare ground




Dry forest
Teak dominated
forest
Moist forest
Tassled-cap
'greenness'

Elevation

Slope
Distance from
village
Distance to water


Explanation


Transformation Source


Interpolated encounter
rate (number seen/km)
from line transects and
vehicle transects
Same as above
Same as above
Interpolated ER of jungle
fowl, spur fowl, peafowl
and black-naped hare
Percentage frequency of
cells with in a circular
window of area 1 km2

Same as above
Same as above

Same as above
The first band of tassled-
cap transform using
Landsat ETM + imagery.
DEM in meters at 100m
resolution
Calculated from DEM
Distance to the nearest
village
Distance from the nearest
water source mn meters


B ox-Cox




B ox-Cox
Box-Cox
None


B ox-Cox




Box-Cox
None

B ox-Cox
None


B ox-Cox

Box-Cox
B ox-Cox

B ox-Cox


This study




This study
This study
This study


Supervised classification of
Landsat ETM+ imagery to
Obtain landcover; frequency
calculated.
Same as above
Same as above

Same as above
Calculated.


SRTM data

Calculated.
Calculated

Calculated.









Table 5-4. Measures of evaluation for habitat models at different pixel resolutions (with cross-
validated standard deviations).


Study
Site


Model
Resolution


Circular
moving
window
Size


AVI CVI Continuous
Boyce
Index


STR

STR

STR

STR

STR

SC
Madhya
Pradesh
SC
Madhya
Pradesh
SC
Madhya
Pradesh
SC
Madhya
Pradesh
SC
Madhya
Pradesh


100 m 1 km2 0.51
(0.11)
100 m 54 km2 0.48
(0.14)
200 m 56 km2 0.50
(0.19)
300 m 52 km2 0.49
(0.22)
500 m 56 km2 0.49
(0.17)
1000 m, 20 km2 0.48
with buffer (0.12)


0.30
(0.11)
0.33
(0.13)
0.40
(0.19)
0.34
(0.21)
0.34
(0.16)
0.33
(0.11)

0.33
(0.14)

0.35
(0.18)

0.29
(0.10)

0.20
(0.10)


0.75
(0.18)
0.69
(0.35)
0.74
(0.25)
0.36
(0.39)
0.63
(0.26)
0.91
(0.13)

0.72
(0.32)

0.78
(0.24)

0.77
(0.17)

0.55
(0.25)


1000 m 21 km2


0.48
(0.15)


2000 m 84 km2 0.50
(0.19)

3000 m 81 km2 0.52
(0.11)


5000 m 225 km2


0.48
(0.11)


Note: AVI measures proportional accuracy in classifying habitat and ranges from 0 to 1. Higher
values of AVI denote a more accurate model. CVI measures the difference between the model
and a random model, with values ranging from 0 to AVI. High values of CVI indicate a model
that is very different from random. The Boyce index measures the correlation between habitat
suitability values and the area adjusted frequency of presence points in the habitat map.










Table 5-5. Correlation between ENFA factors and EGV for south-central Madhya Pradesh. The
percentages quantify the amount of specialization attributed to the factor.


Factor2
(22%)


Factor
(12%)


Factor4
(10%)
****


FactorS'
(9%) >
****


Factor
(8%)


Factor*
(7%)


EGV


Factor 1


++++


Elevation
Elevation
standard
deviation
Slope
NDVI
Forest
Bare
gsround/urban
Agriculture
Livestock ER
Nilgai ER
Sambar ER
Wild pig ER
Distance to
water


0 0
* *****
* ****


0 0 0 0
*** *** **


*
-++ 0 0
-++ n


***


0 ****
0 0
0 *
* **


** *****


** ***


Note: 'For the marginality factor, the + symbol indicates that leopards presence was associated
with values higher than average, and vice versa for -. The number of signs indicates the strength
of the relationship. For the specialization factor, indicates that leopards were found in
narrower range of values than available. The number of indicates the narrowness of the range.
A 0 indicates low specialization. Factor 1 accounts for all the marginality.










Table 5-6. Correlation between ENFA factors and EGV for Satpura Tiger Reserve. The
percentages quantify the amount of specialization attributed to the factor.
EGV Factor1 (22%) Factor2 (42%) Factor3 (9%/) Factor4 (6%)
Elevation --- ****** 0 0
Slope -- 0 0 *
Langur ER + 0 0
Cervid ER ++ 0 0 0
Pig ER ++ 0 0 0
Small-prey ER +++ 0 0 *
Tassled cap 'greenness' ++++ ** ****** ********
Teak Forest +++ **** **** ***
Moist Forest +++ ***** ***** ****
Dry Forest --- **** **
Bare ground ---- *** ***
Distance to water +++ 0 ** *
Distance to village ** **
Note: +For the marginality factor, the + symbol indicates that leopards presence was associated
with values higher than average, and vice versa for -. The number of signs indicates the strength
of the relationship. + For the specialization factor, indicates that leopards were found in
narrower range of values than available. The number of indicates the narrowness of the range.
A 0 indicates low specialization.Factor 1 accounts for all the marginality.










Table 5-7. Area under various leopard-habitat categories in south-central Madhya Pradesh.


Di stri ct


Unsuitable
(km2)
118.3
5031.5
50.1
5621.2
776.2
1141.3
2187.3
3325.6
38.1
3265.4
2464.1
1590.6
2951.5


Marginal
(km2)
210.6
2387.9
7.0
3528.2
288.8
453.3
490.4
1577.5
172.5
638.8
535.5
882.5
1745.0


Suitable
(km2)
86.2
1172.1
0
1931.6
166.5
343.0
415.2
1371.0
28.1
303.9
282.8
583.7
892.6


Optimal
(km2)
5.0
857.5
0
744.1
66.2
168.5
238.7
465.3
20.0
215.6
14.0
212.6
306.9


Balaghat
Betul
Bhopal
Chhindwara
Dewas
East Nimar
Harda
Hoshangabad
Jabalpur
Narmsimhapur
Raisen
Sehore
Seoni



















1,
i.---

1;.5

-- r

-

-J .


'L
'7-


I~-I---






'L-
I'


I_
...
Mddnya ~~rlesh

'i I



"L, t ...

r i ?;
e. ----
-
~I .
-~- I


IM. 3j-!round /



M Agn.ulture
I Wa-.~





100
km *'n


r
i ".L.-.7
II-
I


F i '
-'




ri


~I

$~- '


.


A

i-
s.


:-..~-.


I


Figure 5-1. Cover map of the study area in south-central Madhya Pradesh.



























































107





~I~ ~~~



-:;


~1~1~3-


100 km


Figure 5-2. Mosaicked landsat satellite image of the study area in south-central Madhya Pradesh.































Figure 5-3. Landsat satellite image of the study area in Satpura Tiger Reserve.
















15
13
19
I'u'~-C ~-~--~r~g~ its
Is
33 'L
39
''
56
~Y~ ~1~Y~-~ Tu~d~~.( h~r~3a,


Isa
rJ IP(
A C Ilw


~j~ 100 km
100krn



,,
93
Is


,, s
~! ~i~?
:N 61
r se
n itjil
63
st
ss
76 (D6
63 1 111
as CD- D lip


i~i- 100 km
100km

Figure 5-4. Maps of remotely derived variables for south-central Madhya Pradesh. A) Frequency

of forests. B) Frequency of urban/bareground. C) Frequency of agriculture. D)

Distance to water sources. E: Elevation. F: Slope G: Ruggedness (Std deviation of

elevation). H) NDVI. I) Nilgai abundance index J) Sambar abundance index. K)

Wild pig abundance index. L) Livestock abundance index.











0D.M
1191
2382



19054



'om-o
9 09-10
1 IM-12
1 > 12


G
100 km


100km


Figure 5-4. Continued.


1

~1 P
.~b: ~u-oo~





























m 1-san K -2


100 km 0a ll-n41 5 4379 100km 0 l(-2434528 4379
4a ue- ea2( g 4988 _,
m eaz2 im' 6922- tonal


~~J~-.0,.1~i i 9 .. Mco-o23
1*. 0 47 079



II 7145-24379L
100 km 45431P-6L
6 922- TWooD 100 km


Figure 5-4.Continued.


















Fiue -. as freoel ervd aialsfo atua ierRsev. )Frqenyo





Figu~bregr5.Msound BFrequency ofre moriast foresat.ur C)e Freequen of dyfores.DFrquency o
of teak forest. E) Cervid abundance index F) Wild pig abundance index. G) Langur
abundance index. H) Small prey abundance index. I) Elevation. J) Slope. K) Tassel
cap 'greenness'. L) Distance to village. M) Distance to water.















E
II
II


C`


II
01


Figure 5-5. Continued.


CrYP)
5




























~, ...
81.- I:
I*1-ap
~an.~













b



6~:L
I..~....
I ~..


iI)- -


Y I+-en
1~1
~ : I r~

I Iin

r













11~4
I lo~.~a
I )l(lN Ullr
Yt='~~
Ilsnm~-
111BI~ 1*111


I ~~a

Oraiaman
o ~31~-Y196P
I wl~rp-llmlr



"


Figure 5-5. Continued.





* Sampled beats
Habitat categories
SUnsuitable
I Marginal
SSuitable
I Optimal
I Water


Figure 5-6. Leopard habitat suitability map for south-central Madhya Pradesh.


1
;Be~8


I


.. H biat categories

I Marginal
SSuitable
I Optimal
I Water
*Sampled
in a n ICatiDRS


Figure 5-7. Leopard habitat suitability map for Satpura Tiger Reserve.















south central
M,~~adhya
Pradesh







labitats utabiry

4 _~.Satpura Tiger
SReserve







0 510 15 2 2 3 3 4 4 5 060m 707580 5 0 9
Hblantt u tabilty



Figure 5-8. The predicted-to-expected frequency curves with habitat suitability values for both
the models.










































117









CHAPTER 6
CONCLUSION

Density of Potential Prey

Rigorous estimates of ungulate density were not available until now for the Satpura Tiger

Reserve. The results of this study indicate that wildlife populations are lower than those in

relatively well protected parks in India such as Nagarhole and Bandipur in southern India and

Kanha and Pench tiger reserves in Madhya Pradesh. They are however in line with estimates

from other central Indian protected areas like Tadoba, Melghat and Pench (Maharashtra).

Though densities of most prey species are low in the study area, the ungulate community is still

intact. Since the area is relatively large, a higher degree of protection and habitat management

should increase prey and be able to support a relatively large population of carnivores. The

coefficients of variation around the density estimates of most species generated by this study are

low enough to be useful to monitor population changes. A note of caution should be sounded

since the last year of sampling showed lower densities for all species. A monitoring program

needs to be instituted to make sure that this population is not continuously declining. This

monitoring program should involve distance sampling and can use vehicles to take advantage of

the extensive trail network.

Preference of Prey

Leopards, dholes and tigers strongly prefer sambar in this study area. The density of

sambar can be enhanced further using suitable habitat management techniques, but the main

requirement is likely to be effective protection from poaching. Because the food habits of

leopards, dholes and tigers overlap to a great extent, any increase in density of medium and large

biomass will benefit all the predators.









Density of Leopards

There is no other published information on density of leopards in India, so comparison of

these results with other areas is difficult. Indices computed for other parks are in the same range

as ones calculated for this study. Index based approaches such as the number of photographs per

night (RAI) have been recommended to be used for tigers when there are not enough data for

mark recapture sampling (Carbone et al. 2001). In this study, the RAI did not index leopard

density well and it cannot be recommended. Therefore the results of indices available for other

study areas should be interpreted with caution. At the present time there appears to be no

substitute to the mark-recapture framework used in this study. It is recommended that leopard

densities be estimated across all protected areas and a reliable index be developed so that large

areas in the country can be monitored for population changes. There is urgent need for more

research to accurately estimate the effective trapping area. The spatially explicit maximum

likelihood method offers hope for the future, but with the low population sizes it is difficult to

obtain high precision, making it difficult to detect changes in population density. The logistics

involved in sampling at a high enough intensity to get good precision and at a large enough

spatial scale for a species of this size are very difficult. Given these limitations, there is urgent

need for a calibrated index to be developed to monitor population changes of leopards.

Habitat Model

The leopard is an adaptable species, being able to live in a wide variety of environmental

conditions. The habitat model showed that moist deciduous and teak dominant forests had a

higher association with leopards. Prey densities, especially those of sambar, were also higher in

these habitat types. The larger spatial scale model showed that leopards were negatively

associated with land cover associated with human use. Though it has a reputation for tolerating

human presence, leopard densities are negatively affected by disturbance and presence of









agriculture. Given that it is a large-sized carnivore species that requires relatively large tracts of

contiguous habitat, the model predicted a compact block of about 2000 km2 Of optimal habitat in

the districts of Betul, Hoshangabad and Chhindwara. In addition, approximately 1 1500 km2 Of

habitat is likely to support leopard populations in south-central Madhya Pradesh. It is

recommended that protection for this habitat be adequately strengthened and resources

prioritized accordingly when managing leopard conservation efforts in Madhya Pradesh.











APPENDIX A
INDICES OF UNGULATE AND CARNIVORE ABUNDANCE


Table A-1. Kilometric index values (number of individuals per km.) of selected species using dirt
trails in the monsoon from 2002 to 2005 with bootstrapped 95% C.I.


Species

Chital

Sambar

Indian muntj ac

Nilgai

Wild pig

Common
langur
Indian peafowl


2002
N=9696.2 km
1.26
(1.01-1.54)
0.43
(0.36-0.52)
0.08
(0.06-0.10)
0.13
(0.09-0.18)
0.43
(0.30-0.58)
1.28
(1.05-1.54)
0.30
(0.24-0.35)


2003
N=9910.3 km
0.77
(0.64-0.92)
0.31
(0.25-0.37)
0.04
(0.03-0.06)
0.08
(0.05-0.10)
0.36
(0.27-0.46)
1.14
(0.98-1.3)
0.34
(0.29-0.38)


2004
N=91188.1 km
1.39
(1.12-1.71)
0.19
(0.15-0.24)
0.04
(0.03-0.05)
0.06
(0.04-0.08)
0.24
(0.15-0.37)
2.51
(2.22-2.79)
0.26
(0.22-0.31)


2005
N=9811.8 km
1.48
(1.16-1.87)
0.14
(0.10-0.19)
0.04
(0.03-0.06)
0.06
(0.03-0.09)
0.17
(0.07-0.24)
2.39
(0.21-0.28)
0.24
(0.19-0.29)











Table A-2. Encounter rates of tracks of selected carnivore species (frequency per 500 m section),


with bootstrapped 95 %
Time period Leopard Dhole
Nov'O3- 0.14 (
Feb'O4 (0.08- (0
0.19) 0
Mar'O4- 0.13 (
Jun'O4 (0.10- (0
0.16) 0


C.I. (November 2003 to June 2006).


Tiger
0.06
(0.03-

0.03
(0.02-.05)


Sloth bear Jungle cat
0.22 (0.17- 0.10
0.28) (0.06-
0.14)
0.43 (0.39- 0.09
0.47) (0.07-
0.11)


Palm civet
0.32 (0.26-
0.39)

0.35 (0.31-
0.39)



0.22 (0.20-
0.26)


0.27 (0.25-
0.30)


0.45 (0.42-
0.47)


0.33 (0.30-
0.35)


Nov' O4-
Feb'05


Mar'05-
Jun'05


Nov'05-
Feb'O6


Mar'O6-
Jun'O6


0.11
(0.09-
0.13)

0.03
(0.02-
0.04)

0.11
(0.10-
0.13)

0.12
(0.10-
0.14)


0.07
(0.05-
0.09)

0.03
(0.02-
0.04)

0.14
(0.12-
0.15)

0.08
(0.06-
0.09)


0.08
(0.06-
0.10)

0.02
(0.01-
0.02)

0.02
(0.02-
0.03)

0.02
(0.01-
0.03)


0.10 (0.08-
0.13)


0.36 (0.33-
0.39)


0.11 (0.01-
0.13)


0.26 (0.23-
0.28)


0.02
(0.01-
0.03)

0.06
(0.05-
0.09)

0.12
(0.10-
0.14)

0.09
(0.08-
0.11)









LIST OF REFERENCES


Acharya, B. B. 2008. The Ecology of the Dhole or Asiatic Wild Dog (Cuon alpinus) In Pench
Tiger Reserve, Madhya Pradesh. Page 115. Saurashtra University Rajkot.

Ackerman, B. B., F. G. Lindzey, and T. P. Hemker. 1984. Cougar Food-Habits in Southern Utah.
Journal of Wildlife Management 48: 147-15 5.

Andheria, A. P., K. U. Karanth, and N. S. Kumar. 2007. Diet and prey profiles of three sympatric
large carnivores in Bandipur Tiger Reserve, India. Journal of Zoology 273:169-175.

Athreya, V. R., S. S. Thakur, S. Chaudhari, and A. V. Belsare. 2007. Leopards in human-
dominated areas: a spillover from sustained translocations into nearby forests. Journal of
the Bombay Natural History Society 104 13-18.

Austin, M. 2007. Species distribution models and ecological theory: A critical assessment and
some possible new approaches. Ecological Modelling 200:1-19.

Avinandan D. 2003. Food habits of the tiger (Pathera tigris tigris) in Sariska Tiger Reserve,
Rajasthan. Page 63. Saurashtra University, Dehradun.

Bagchi, S., S. P. Goyal, and K. Sankar. 2003a. Niche relationships of an ungulate assemblage in
a dry tropical forest. Journal of Mammalogy 84:981-988.

Bagchi, S., S. P. Goyal, and K. Sankar. 2003b. Prey abundance and prey selection by tigers
(Panthera tigris) in a semi-arid, dry deciduous forest in western India. Journal of
Zoology 260:285-290.

Bagchi, S., S. P. Goyal, and K. Sankar. 2004. Herbivore density and biomass in a semi-arid
tropical dry deciduous forest of western India. Journal of Tropical Ecology 20:475-478.

Bailey, T. N. 1993. The African Leopard: Ecology and Behavior of a Solitary Felid. Columbia
University Press.

Bertram, B. C. R. 1982. Leopard ecology as studied by radio tracking. Symp. zool. Soc.
Lond:341-352.

Bhatnagar, Y. V. 1991. Habitat preference of sambar (Cervus unicolor) in Rajaji National Park.
Page 51. Saurashtra University, Rajkot.

Biswas, S., and K. Sankar. 2002. Prey abundance and food habit of tigers (Panthera tigris tigris)
in Pench National Park, Madhya Pradesh, India. Journal of Zoology 256:41 1-420.

Bothma, J. D. P., and E. A. N. Le Riche. 1984. Aspects of the ecology and the behaviour of the
leopard Panthera ~~P~~Pp ardusPPP~~PP in the Kalahari Desert. Koedoe 27:259-279.










Boyce, M. S., P. R. Vernier, S. E. Nielsen, and F. K. A. Schmiegelow. 2002. Evaluating resource
selection functions. Ecological Modelling 157:281-300.

Braunisch, V., and R. Suchant. 2007. A model for evaluating the 'habitat potential' of a landscape
for capercaillie Tetrao urogallus: a tool for conservation planning. Wildlife Biology
13:21-33.

Brotons, L., W. Thuiller, M. B. Araujo, and A. H. Hirzel. 2004. Presence-absence versus
presence-only modelling methods for predicting bird habitat suitability. Ecography
27:437-448.

Bryan, T. L., and A. Metaxas. 2007. Predicting suitable habitat for deep-water gorgonian corals
on the Atlantic and Pacific Continental Margins of North America. Marine Ecology-
Progress Series 330:113-126.

Buckland, S. T., D. R. Anderson, K. P. Burnham, J. L. Laake, and D. L. Borchers 2001.
Introduction to distance sampling: Estimating abundance of biological populations.
Oxford University Press.

Burnham, K. P., and D. Anderson 2003. Model Selection and Multi-Model Inference. Springer.

Burnham, K. P., and W. S. Overton. 1979. Robust estimation of population-size when capture
probabilities vary among animals. Ecology 60:927-936.

Bustamante, J., and J. Seoane. 2004. Predicting the distribution of four species of raptors (Aves :
Accipitridae) in southern Spain: statistical models work better than existing maps. Journal
of Biogeography 31:295-306.

Carbone, C., S. Christie, K. Conforti, T. Coulson, N. Franklin, J. R. Ginsberg, M. Griffiths, J.
Holden, K. Kawanishi, M. Kinnaird, R. Laidlaw, A. Lynam, D. W. Macdonald, D.
Martyr, C. McDougal, L. Nath, T. O'Brien, J. Seidensticker, D. J. L. Smith, M. Sunquist,
R. Tilson, and W. N. W. Shahruddin. 2001. The use of photographic rates to estimate
densities of tigers and other cryptic mammals. Animal Conservation 4:75-79.

Carbone, C., G. M. Mace, S. C. Roberts, and D. W. Macdonald. 1999. Energetic constraints on
the diet of terrestrial carnivores. Nature 402:286-288.

Champion, H. G., and S. K. Seth 1968. A revised survey of the forest types of India. Manager of
publications, Government of India, Delhi.

Chefaoui, R. M., J. Hortal, and J. M. Lobo. 2005. Potential distribution modelling, niche
characterization and conservation status assessment using GIS tools: a case study of
Iberian Copris species. Biological Conservation 122:327-338.

Chellam, R. 1993. The ecology of the asiatic lion (Panthera leo persica). Saurashtra University,
Raj kot.










Chesson, J. 1978. Measuring Preference in Selective Predation. Ecology 59:211-215.

Crist, E. P., and R. J. Kauth. 1986. The tasseled cap de-mystified. Photogrammetric Engineering
and Remote Sensing 52:81-86.

David, A. C., Q. Qureshi, S. P. Goyal, K. Sankar, V. B. Mathur, Pannalal, A. Verma, and D.
Patial. 2005. Estimating Herbivore Abundance using Line Transect Method in Sariska
Tiger Reserve. Page 10. Wildlife Institute of India, Dehradun, India.

Dettki, H., R. Lofstrand, and L. Edenius. 2003. Modeling habitat suitability for moose in coastal
northern Sweden: Empirical vs process-oriented approaches. Ambio 32:549-556.

Dice, L. R. 1938. Some Census Methods for Mammals. The Journal of Wildlife Management
2:119-130.

Dinerstein, E., C. Loucks, E. Wikramanayake, J. Ginsberg, E. Sanderson, J. Seidensticker, J.
Forrest, G. Bryja, A. Heydlauff, S. Klenzendorf, P. Leimgruber, J. Mills, T. G. O'Brien,
M. Shrestha, R. Simons, and M. Songer. 2007. The fate of wild tigers. Bioscience
57:508-514.

Diniz, J. A. F., L. M. Bini, and B. A. Hawkins. 2003. Spatial autocorrelation and red herrings in
geographical ecology. Global Ecology and Biogeography 12:53-64.

Eastman, R. 2004. Idrisi Kilimanjaro v14.02. Clark University.

Edgaonkar, A., and R. Chellam. 2002. Food habit of the leopard, Panthera pardus, in the Sanj ay
Gandhi National Park, Maharashtra, India. Mammalia 66:353-360.

Edgaonkar, A. J., and R. Chellam. 1998. A preliminary study on the ecology of the leopard,
Panthera parduus fsc in the Sanjay Gandhi National Park, Maharashtra Wildlife
Institute of India, Dehradun, India:39.

Efford, M. 2004. Density estimation in live-trapping studies. Oikos 106:598-610.

Efford, M. 2007. Density: spatially explicit capture recapture, Otago.

Efford, M., D. Borchers, and A. Byrom. in review. Density estimation by spatially explicit
capture-recapture: likelihood-based methods. Environmental and Ecological Statistics.

Eisenberg, J. F. 1986. Life history strategies of the felidae: Variations on a common theme.
Pages 293-303 in S. D. Miller, and D. D. Everett, editors. Cats of the World of the World:
Biology, Conservation, and Management. National Wildlife Foundation, Washington,
D.C.

Eisenberg, J. F., and M. Lockhart. 1972. An ecological reconnaissance of Wilpattu National
Park. Smithsonian Contributions to Knowledge, (Zoology) 101:1-118.










Eisenberg, J. F., and J. Seidensticker. 1976. Ungulates in southern Asia: A consideration of
biomass estimates for selected habitats. Biological Conservation 10:293-308.

Emmons, L. H. 1987. Comparative feeding ecology of felids in a neotropical rainforest.
Behavioral Ecology and Sociobiology 20:271-283.

Farrell, L. E., J. Roman, and M. E. Sunquist. 2000. Dietary separation of sympatric carnivores
identified by molecular analysis of scats. Molecular Ecology 9: 1583-1590.

Fielding, A. H., and J. F. Bell. 1997. A review of methods for the assessment of prediction errors
in conservation presence/absence models. Environmental Conservation 24:38-49.

Floyd, T. J., L. D. Mech, and P. A. Jordan. 1978. Relating Wolf Scat Content to Prey Consumed.
Journal of Wildlife Management 42:528-532.

Gangadharan, A. 2005. Density Estimation and Time Trend Analysis of Large Herbivores in
Nagarhole, India. Page 82. University of St. Andrews.

Gotelli, N. J., and G. L. Entsminger. 2007. EcoSim: Null models software for ecology. Acquired
Intelligence Inc. & Kesey-Bear, Jericho, VT 05465.

Gu, W. D., and R. K. Swihart. 2004. Absent or undetected? Effects of non-detection of species
occurrence on wildlife-habitat models. Biological Conservation 116:195-203.

Guisan, A., A. Lehmann, S. Ferrier, M. Austin, J. M. C. Overton, R. Aspinall, and T. Hastie.
2006. Making better biogeographical predictions of species' distributions. Journal of
Applied Ecology 43:386-392.

Guisan, A., and N. E. Zimmermann. 2000. Predictive habitat distribution models in ecology.
Ecological Modelling 135:147-186.

Hamilton, P. H. 1976. The movements of leopards in Tsavo National Park, Kenya, as determined
by radio-tracking. University of Nairobi.

Harihar, A. 2005. Population, Food Habits and Prey Densities of Tiger in Chilla Range, Rajaji
National Park, Uttaranchal, India. Saurashtra University, Rajkot.

Hayward, M. W. 2006. Prey preferences of the spotted hyaena (Crocuta crocuta) and degree of
dietary overlap with the lion (Panthera leo). Journal of Zoology 270:606-614.

Hayward, M. W., P. Henschel, J. O'Brien, M. Hofmeyr, G. Balme, and G. I. H. Kerley. 2006a.
Prey preferences of the leopard (Panthera pardus). Journal of Zoology 270:298-3 13.

Hayward, M. W., and G. I. H. Kerley. 2005. Prey preferences of the lion (Panthera leo). Journal
of Zoology 267:309-322.










Hayward, M. W., J. O'Brien, M. Hofmeyr, and G. I. H. Kerley. 2006b. Prey preferences of the
African wild dog Lycaon pictus (Canidae : Carnivora): Ecological requirements for
conservation. Journal of Mammalogy 87: 1122-1 13 1.

Hines, J. 1994. CAPTURE2 version 070503. 1333 Patuxent Wildlife Research Center,USGS.
http://www. mbr-pwrc. usgs.gov. Accessed on 1 September 2007.

Hines, J. 2002. SCATMAN version 2. Patuxent Wildlife Research Center, USGS.
http://www. mbr-pwrc. usgs.gov. Accessed on 1 September 2007.

Hirzel, A. H., J. Hausser, D. Chessel, and N. Perrin. 2002. Ecological-niche factor analysis: How
to compute habitat-suitability maps without absence data? Ecology 83:2027-2036.

Hirzel, A. H., J. Hausser, and N. Perrin. 2006a. Biomapper version 3.2. Laboratory for
Conservation Biology University of Lausanne, Switzerland.

Hirzel, A. H., V. Helfer, and F. Metral. 2001. Assessing habitat-suitability models with a virtual
species. Ecological Modelling 145:111-121.

Hirzel, A. H., G. Le Lay, V. Helfer, C. Randin, and A. Guisan. 2006b. Evaluating the ability of
habitat suitability models to predict species presence. Ecological Modelling 199:142-
152.

Hirzel, A. H., B. Posse, P. A. Oggier, Y. Crettenand, C. Glenz, and R. Arlettaz. 2004. Ecological
requirements of reintroduced species and the implications for release policy: the case of
the bearded vulture. Journal of Applied Ecology 41:1 103-1 116.

Hoppe-Dominik, B. 1984. Prey Frequency of the Leopard, Panthera ~~P~~Pp ardus,~~PP~~ in the Tai-National-
Park of the Ivory-Coast. Mammalia 48:477-487.

Jackson, R. M., J. D. Roe, R. Wangchuk, and D. O. Hunter. 2006. Estimating snow leopard
population abundance using photography and capture-recapture techniques. Wildlife
Society Bulletin 34:772-781.

Jacobs, J. 1974. Quantitative Measurement of Food Selection Modification of Forage Ratio and
Ivlevs Electivity Index. Oecologia 14:413-417.

Jathanna, D., K. U. Karanth, and A. J. T. Johnsingh. 2003. Estimation of large herbivore
densities in the tropical forests of southern India using distance sampling. Journal of
Zoology 261:285-290.

Jennelle, C. S., M. C. Runge, and D. I. MacKenzie. 2002. The use of photographic rates to
estimate densities of tigers and other cryptic mammals: a comment on misleading
conclusions. Animal Conservation 5:119-120.










Jenny, D. 1996. Spatial organization of leopards Panthera pardus in Tai National Park, Ivory
Coast: Is rain forest habitat a 'tropical haven'? Journal of Zoology 240:427-440.

Jhala, Y. V., R. Gopal, and Q. Qureshi. 2008. Status of the Tigers, Co-predators, and Prey in
India. Page 151. National Tiger Conservation Authority, Govt. of India, New Delhi, and
Wildlife Institute of India, Dehradun.

Johnsingh, A. J. T. 1983. Large mammalian prey--predators in Bandipur. Journal of the Bombay
Natural History Society 80:1-57.

Johnsingh, A. J. T. 1992. Prey Selection in 3 Large Sympatric Carnivores in Bandipur.
Mammalia 56:517-526.

Johnson, A., C. Vongkhamheng, M. Hedemark, and T. Saithongdam. 2006. Effects of human-
carnivore conflict on tiger (Panthera tigris) and prey populations in Lao PDR. Animal
Conservation 9:421-430.

Johnson, D. H. 1980. The comparison of usage and availability measurements for evaluating
resource preference. Ecology 61:65-71.

Karanth, K. U. 1995. Estimating Tiger Panthera tigris populations from camera-trap data using
capture-recapture models. Biological Conservation 71:333-338.

Karanth, K. U., R. S. Chundawat, J. D. Nichol, and N. S. Kumar. 2004a. Estimation of tiger
densities in the tropical dry forests of Panna, Central India, using photographic capture-
recapture sampling. Animal Conservation 7:285-290.

Karanth, K. U., and J. D. Nichols. 1998. Estimation of tiger densities in India using photographic
captures and recaptures. Ecology 79:2852-2862.

Karanth, K. U., and J. D. Nichols. 2000. Ecological Status and Conservation of Tigers in India:
Final Technical Report submitted to US Fish and Wildlife Service, Washington DC, and
Wildlife Conservation Society, New York. Centre for Wildlife Studies, Bangalore,
Bangalore.

Karanth, K. U., J. D. Nichols, N. S. Kumar, W. A. Link, and J. E. Hines. 2004b. Tigers and their
prey: Predicting carnivore densities from prey abundance. Proceedings of the National
Academy of Sciences of the United States of America 101:4854-4858.

Karanth, K. U., J. D. Nichols, J. Seidensticker, E. Dinerstein, J. L. D. Smith, C. McDougal, A. J.
T. Johnsingh, R. S. Chundawat, and V. Thapar. 2003. Science deficiency in conservation
practice: the monitoring of tiger populations in India. Animal Conservation 6: 141-146.

Karanth, K. U., and M. E. Sunquist. 1992. Population-Structure, Density and Biomass of Large
Herbivores in the Tropical Forests of Nagarahole, India. Journal of Tropical Ecology
8:21-35.











Karanth, K. U., and M. E. Sunquist. 1995. Prey Selection by Tiger, Leopard and Dhole in
Tropical Forests. Journal of Animal Ecology 64:439-450.

Karanth, K. U., and M. E. Sunquist. 2000. Behavioural correlates of predation by tiger (Panthera
tigris), leopard (Panthera pardus) and dhole (Cuon alpinus) in Nagarahole, India. Journal
of Zoology 250:255-265.

Karanth, U., and J. Nichols 2002. Monitoring Tigers and Their Prey. Center for Wildlife Studies.

Karlsson, J., H. Broseth, H. Sand, and H. Andren. 2007. Predicting occurrence of wolf territories
in Scandinavia. Journal of Zoology 272:276-283.

Khan, J. A., R. Chellam, W. A. Rodgers, and A. J. T. Johnsingh. 1996. Ungulate densities and
biomass in the tropical dry deciduous forests of Gir, Guj arat, India. Journal of Tropical
Ecology 12:149-162.

Kitchener, A. C. 1991. The natural history of the wild cats. Christopher Helm, A and C Black,
London.

Krebs, C. J. 1989 Ecological methodology. Harper and Row, New York.

Lawlor, L. R. 1980. Structure and Stability in Natural and Randomly Constructed Competitive
Communities. American Naturalist 116:394-408.

Lele, U., N.Tuinar, S. A. Hussain, A. Zazneta, and L. Kelly. 2000. The World Bank forest
strategy:striking the right balance.

Link, W. A., and K. U. Karanth. 1994. Correcting for Overdispersion in Tests of Prey
Selectivity. Ecology 75:2456-2459.

MacKenzie, D. I., and J. D. Nichols. 2004. Occupancy as a surrogate for abundance estimation.
Animal Biodiversity and Conservation 27:461-467.

Maffei, L., E. Cuellar, and A. Noss. 2004. One thousand jaguars (Panthera onca) in Bolivia's
Chaco? Camera trapping in the Kaa-lya National Park. Journal of Zoology 262:295-304.

Marker, L. L., and A. J. Dickman. 2005. Factors affecting leopard (Panthera pardus) spatial
ecology, with particular reference to Namibian farmlands. South African Journal of
Wildlife Research 35: 105-115.

Mestre, F. M., J. P. Ferreira, and A. Mira. 2007. Modelling the distribution of the European
Polecat M~ustelaputorius in a Mediterranean agricultural landscape. Revue D Ecologie-la
Terre et la Vie 62:35-47.










Meynard, C. N., and J. F. Quinn. 2007. Predicting species distributions: a critical comparison of
the most common statistical models using artificial species. Journal ofBiogeography
34:1455-1469.

Mladenoff, D. J., and T. A. Sickley. 1998. Assessing potential gray wolf restoration in the
northeastern United States: A spatial prediction of favorable habitat and potential
population levels. Journal of Wildlife Management 62:1-10.

Mooley, D. A., B. Parthasarathy, K. Rupa Kumar, N. A. Sontakke, A. A. Munot, and D. R.
Kothawale. 2007. IITM Indian regional/subdivi sional Monthly Rainfall data set (IITM-
IMR).

Muckenhirn, N., and J. F. Eisenberg. 1973. Ranges and predation in the Ceylon leopard. Pages
142-175 in R. L. Eaton, editor. The Worlds Cats. Vol 1: Ecology and Conservation.
World Wildlife Safari, winstone, Oregon.

Mukherjee, S., S. P. Goyal, and R. Chellam. 1994. Standardization of Scat Analysis Techniques
for Leopard (Panthera pardus) in Gir-National-Park, Western India. Mammalia 58: 139-
143.

Ngoprasert, D., A. J. Lynam, and G. A. Gale. 2007. Human disturbance affects habitat use and
behaviour of Asiatic leopard Panthera pardus in Kaeng Krachan National Park,
Thailand. Oryx 41:343-351.

Nowell, K., and P. Jackson. 1996. Wild Cats: status survey and Conservation Plan, Gland,
Switzerland.

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. Animal Conservation
6:131-139.

Odden, M., and P. Wegge. 2005. Spacing and activity patterns of leopards Panthera pardus in
the Royal Bardia National Park, Nepal. Wildlife Biology 11:145-152.

Ogutu, J. O., N. Bhola, H. P. Piepho, and R. Reid. 2006. Efficiency of strip- and line-transect
surveys of African savanna mammals. Journal of Zoology 269: 149-160.

Ogutu, J. O., and N. Owen-Smith. 2005. Oscillations in large mammal populations: are they
related to predation or rainfall? African Journal of Ecology 43:332-339.

Olivier, F., and S. J. Wotherspoon. 2006. Modelling habitat selection using presence-only data:
Case study of a colonial hollow nesting bird, the snow petrel. Ecological Modelling
195:187-204.

Otis, D. L., K. P. Burnham, G. C. White, and D. R. Anderson. 1978. Statistical inference from
capture data on closed populations. Wildlife Monographs 62:1-135.











Panwar, H. S. 1979. A note on tiger census techniques based on pug mark tracings. Indian
Forester: 18-3 6.

Parmenter, R. R., T. L. Yates, D. R. Anderson, K. P. Burnham, J. L. Dunnum, A. B. Franklin, M.
T. Friggens, B. C. Lubow, M. Miller, G. S. Olson, C. A. Parmenter, J. Pollard, E.
Rexstad, T. M. Shenk, T. R. Stanley, and G. C. White. 2003. Small-mammal density
estimation: A field comparison of grid-based vs. web-based density estimators.
Ecological Monographs 73:1-26.

Pianka, E. R. 1973. The Structure of Lizard Communities. Annual Review of Ecology and
Systematics 4:53-74.

Pienaar, U. V. 1969. Predator-prey relationships amongst the larger mammals of the Kruger
National Park. Koedoe 12:108-176.

Prater, S. H. 1980. The book of Indian animals. Bombay Natural History Society, Bombay.

Rabinowitz, A. R. 1989. The density and behaviour of large cats n a dry tropical forest in Huai
Kha Khaeng Wildlife Sanctuary, Thailand. Natural History Society Bulletin Siam Society
37:235-251.

Reddy, H. S., C. Srinivasulu, and K. T. Rao. 2004. Prey selection by the Indian tiger (Panthers
tigris tigris) in Nagarjunasagar Srisailam Tiger Reserve, India. Mammalian Biology
69:384-391.

Reutter, B. A., V. Helfer, A. H. Hirzel, and P. Vogel. 2003. Modelling habitat-suitability using
museum collections: an example with three sympatric Apodemus species from the Alps.
Journal of Biogeography 30:581-590.

Rexstad, E., and K. P. Burnham. 1991. User' s guide for interactive program CAPTURE.
Abundance estimation of closed animal populations. Colorado State University, Fort
Collins, Colorado, USA 29.

Riley, S. J., and R. A. Malecki. 2001. A landscape analysis of cougar distribution and abundance
in Montana, USA. Environmental Management 28:317-323.

Sankar, K., S. P. Goyal, and Q. Qureshi. 2005. Assessment of status of tiger (Panthera tigris) in
Sariska Tiger Reserve, Raj asthan: A Report submitted to the Proj ect Tiger, Ministry of
Environment & Forests, Govt. of India, New Delhi. Wildlife Institute of India, Dehra
Dun

Sankar, K., and A. J. T. Johnsingh. 2002. Food habits of Tiger (Panthera tigris) and Leopard
(Panthera pardus) in Sariska Tiger Reserve, Raj asthan, India, as shown by scat analysis.
Mammalia 66:285-289.










Santiapillai, C., M. R. Chambers, and N. Ishwaran. 1982. The Leopard Panthera P~Pp ardus-fuscaP~~P~~P
(Meyer 1794) in the Ruhuna National-Park, Sri Lanka, and Observations Relevant to Its
Conservation. Biological Conservation 23:5-14.

Santos, X., J. C. Brito, N. Sillero, J. M. Pleguezuelos, G. A. Llorente, S. Fahd, and X. Parellada.
2006. Inferring habitat-suitability areas with ecological modelling techniques and GIS: A
contribution to assess the conservation status of Vipera lata~stei. Biological Conservation
130:416-425.

Sathyakumar, S. 1992. Food habits of leopard (Panthera pardus) on Mundanthurai plateau,
Tamil Nadu, India. Tiger Paper XX:8-9.

Schaller, G. B. 1967. The Deer and the Tiger. University of Chicago Press.

Schaller, G. B. 1976. The Serengeti Lion: A Study of Predator-Prey Relations. University Of
Chicago Press.

Schaller, G. B. 1977. The mountain monarchs: wild sheep and goats of the Himalaya. University
of Chicago Press, Chicago.

Scognamillo, D., I. E. Maxit, M. Sunquist, and J. Polisar. 2003. Coexistence of jaguar (Panthera
onca) and puma (Puma concolor) in a mosaic landscape in the Venezuelan Ilanos.
Journal of Zoology 259:269-279.

Seidensticker, J. 1976. On the Ecological Separation between Tigers and Leopards. Biotropica
8:225-234.

Seidensticker, J., M. Sunquist, and C. McDougal. 1990. Leopards living at the edge of Royal
Chitwan National park. Pages 415-423 in J. C. Daniel, and J. S. Serrao, editors.
Conservation in Developing Countries: Problems and Prospects Bombay Natural History
Society, Bombay.

Sekhar, N. U. 1998. Crop and livestock depredation caused by wild animals in protected areas:
the case of Sariska Tiger Reserve, Rajasthan, India. Environmental Conservation 25:160-
171.

Seoane, J., J. H. Justribo, F. Garcia, J. Retamar, C. Rabadan, and J. C. Atienza. 2006. Habitat-
suitability modelling to assess the effects of land-use changes on Dupont's lark
Chersophilus duponti: A case study in the Layna Important Bird Area. Biological
Conservation 128:241-252.

Sever, Z., and H. Mendelssohn. 1991. Spatial movement patterns of porcupines (Hystrix indict).
Mammalia 55:187-205.










Silver, S. C., L. E. T. Ostro, L. K. Marsh, L. Maffei, A. J. Noss, M. J. Kelly, R. B. Wallace, H.
Gomez, and G. Ayala. 2004. The use of camera traps for estimating jaguar Panthera onca
abundance and density using capture/recapture analysis. Oryx 38:148-154.

Smith, J. L. D., C. McDougal, and D. Miquelle. 1989. Scent marking in free-ranging tigers,
Panthera tigris. Animal Behaviour 37:1-10.

Soisalo, M. K., and S. M. C. Cavalcanti. 2006. Estimating the density of a jaguar population in
the Brazilian Pantanal using camera-traps and capture-recapture sampling in combination
with GPS radio-telemetry. Biological Conservation 129:487-496.

Spalton, J. A., H. M. al Hikmani, D. Willis, and A. S. B. Said. 2006. Critically Endangered
Arabian leopards Panthera pardus nimr persist in the Jabal Samhan Nature Reserve,
Oman. Oryx 40:287-294.

Stoen, O. G., and P. Wegge. 1996. Prey selection and prey removal by tiger (Panthera tigris)
during the dry season in lowland Nepal. Mammalia 60:363-373.

Sunquist, M., K. U. Karanth, and F. Sunquist. 1999. Ecology, behaviour and resilience of the
tiger and its conservation needs. Pages 5-18 in J. Seidensticker, editor. Riding the tiger:
tiger conservation in human-dominated landscapes. Cambridge University Press.

Sunquist, M. E. 1981. The Social Organization of Tigers (Panthera tigris) in Royal Chitawan
National Park, Nepal. Smithsonian Institution Press.

Sunquist, M. E., and F. Sunquist, editors. 2002. Wild cats of the world. University of Chicago
Press.

Taber, A. B., A. J. Novaro, N. Neris, and F. H. Colman. 1997. The food habits of sympatric
jaguar and puma in the Paraguayan Chaco. Biotropica 29:204-213.

Thomas, L., J. L. Laake, S. Strindberg, F. F. C. Marques, S. T. Buckland, D. L. Borchers, D. R.
Anderson, K. P. Burnham, S. L. Hedley, J. H. Pollard, J. R. B. Bishop, and T. A.
Marques. 2006. Distance 5.0 Release 2. Research Unit for Wildlife Population
Assessment, University of St. Andrews, UK.

Titeux, N., M. Dufrene, J. Radoux, A. H. Hirzel, and P. Defourny. 2007. Fitness-related
parameters improve presence-only distribution modelling for conservation practice: The
case of the red-backed shrike. Biological Conservation 138:207-223.

Traill, L. W., and R. C. Bigalke. 2007. A presence-only habitat suitability model for large
grazing African ungulates and its utility for wildlife management. African Journal of
Ecology 45:347-354.










Varman, K. S., and R. Sukumar. 1995. The Line Transect Method for Estimating Densities of
Large Mammals in A Tropical Deciduous Forest An Evaluation of Models and Field
Experiments. Journal of Biosciences 20:273-287.

Ward, A. I., P. C. L. White, and C. H. Critchley. 2004. Roe deer Capreohes capreohes behaviour
affects density estimates from distance sampling surveys. Mammal Review 34:315-319.

Wegge, P., C. P. Pokheral, and S. R. Jnawali. 2004. Effects of trapping effort and trap shyness on
estimates of tiger abundance from camera trap studies. Animal Conservation 7:251-256.

Wilson, D., and R. Purushothaman. 2003. Dreaming With BRICs: The Path to 2050. Global
Economics Paper No: 99. Goldman Sachs.

Wilson, K. R., and D. R. Anderson. 1985. Evaluation of Two Density Estimators of Small
Mammal Population Size. Journal of Mammalogy 66:13-21.

Woodroffe, R. 2000. Predators and people: using human densities to interpret declines of large
carnivores. Animal Conservation 3:165-173.

Woolf, A., C. K. Nielsen, T. Weber, and T. J. Gibbs-Kieninger. 2002. Statewide modeling of
bobcat, Lynx rufus, habitat in Illinois, USA. Biological Conservation 104:191-198.

Zaniewski, A. E., A. Lehmann, and J. M. C. Overton. 2002. Predicting species spatial
distributions using presence-only data: a case study of native New Zealand ferns.
Ecological Modelling 157:261-280.









BIOGRAPHICAL SKETCH

Advait Edgaonkar was born in 1970 in Pune, India. He got his B.Sc. degree in zoology and

biochemistry from St. Xavier' s College, Mumbai. He then obtained a post graduate diploma in

forestry management from the Indian Institute of Forest Management, Bhopal and an M. Sc. in

wildlife science from the Wildlife Institute of India, Dehradun. He is married to Vinatha

Viswanathan. Advait hopes to spend the rest of his life doing wildlife research in India.





PAGE 1

1 ECOLOGY OF THE LEOPARD ( Panthera pardus ) IN BORI WILDLIFE SANCTUARY AND SATPURA NATIONAL PARK, INDIA By ADVAIT EDGAONKAR A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2008

PAGE 2

2 2008 Advait Edgaonkar

PAGE 3

3 To Aai, Baba and Vinatha.

PAGE 4

4 ACKNOWLEDGMENTS I sincerely thank m y supervisor Dr. Melv in Sunquist for his guidance and patience throughout the long period of fiel dwork, for raising the funds and equipment to make the research possible, and for going through numerous drafts of the dissertation. I would like to acknowledge the help of my committee memb ers Dr. Lyn Branch, Dr. Madan Oli and Dr. Michael Binford for their help during the dissertation writing process. Dr. Kenneth Portier, my former committee member, helped with statis tics as did Dr. Mike M oulton. Delores Tilman Caprice MacRae and Claire Williams helped with administration at the WEC, and I would like to thank them for it. I am grateful for the logistics support of Da ve Ferguson, Fred Bagley and Mini Nagendran at USFWS and from Beena Achankunju and Priya Ghosh at the US Embassy in New Delhi. I am thankful to Jim Nichols and Murray Efford for valuable advice. I would like to thank the Wild life Institute of India, the Alumni Fellowship, the Disney Conservation Fund and the Jennings Scholarship for funding the fi eldwork in India and course work at the University of Florida. I am grat eful to Dr. Ravi Chellam who was the principal investigator of the project at WII till 2002 and to Qamar Qureshi, who took over responsibility afterwards. I thank Dr A.J.T. Johnsingh, Dr Y.V. Jhala, Dr. K Sankar, Dr Manoj Agarwal and Shri Vinod Thakur who helped in various ways. I would very much like to thank Ravi Kailas and Dilip Venugopal, Deep Contractor, Bindu Raghava n, Vidya Athreya, Anirudh Belsare, Rashid Raza, K. Ramesh, Bhaskar Acharya, Rajah Ja yapal, Abishek Harihar, Shomita Mukherjee, Meena Venkataraman, Priya Balasubramaniam and Gopi Sundar for help with data collection, fruitful discussions, encouragement, support, and for their friendship. My friends Bimal Desai, A.V.S. Prasad and Chirag Wazir contributed in numerous ways and to them I am thankful.

PAGE 5

5 Nilmini Jayasena deserves speci al acknowledgment for her generous help during a critical phase of the dissertation writing. I am grateful to the Forest Department of Ma dhya Pradesh for their support. I would like to thank Shri O. P. Tiwari, Shri Madhukar Chaturve di, Shri Ramachandran, Shri S.P. Singh, Shri S. S. Rajpoot, Shri L. K. Gupta, Shri N. D. Sharma Shri H. P. Singh, Shri Rajeev Srivastav, Shri P. M. Lad and Shri L.K. Chaudhary. I would especial ly like to thank Shri Sandeep Fellows and Shri and Smt. Ramachandran for all thei r help and support in Itarsi. I would like to thank Ramesh Yadav, Shya m Yadav, Bicchu Oju, Santosh Guttu, Gopal, Botu and Bishram for help in the field. I woul d like to acknowledge my family, Vikram and Ashok Shrotriya, and Shailaja Supekar, and Vijaya and G. Viswanathan for their support. I am especially grateful to my parents, Jayant a nd Vasundhara Edgaonkar for their patience, love and support over the years. Lastly, I thank my wife, Vinatha Viswanathan, for encouragement, ideas, data entry, funding, companionship, help and love. Th is research is as much hers as it is mine.

PAGE 6

6 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........9 LIST OF FIGURES.......................................................................................................................11 ABSTRACT...................................................................................................................................13 CHAP TER 1 INTRODUCTION..................................................................................................................15 2 DENSITY ESTIMATION OF POTENT IAL PREY SPECIES OF LARGE CAR NIVORES IN SATPURA TIGER RESERVE USING LINE-TRANSECT SAMPLING............................................................................................................................20 Introduction................................................................................................................... ..........20 Methods..................................................................................................................................22 Study Area.......................................................................................................................22 Estimation of Habitat Types............................................................................................ 23 Estimation of Prey Density.............................................................................................. 23 Vehicle Transects............................................................................................................25 Results.....................................................................................................................................25 Description of Habitat..................................................................................................... 25 Estimation of Density...................................................................................................... 26 Discussion...............................................................................................................................27 3 PREY SELECTION AND THE FOOD HA BITS OF TIGER, LEOPARD AND DHOL E IN SATPURA TIGER RESERVE........................................................................... 41 Introduction................................................................................................................... ..........41 Methods..................................................................................................................................42 Study Area.......................................................................................................................42 Reconstruction of Carnivore Diets.................................................................................. 43 Sample Size Adequacy....................................................................................................43 Prey Biomass and Number..............................................................................................44 Estimation of Prey Selection........................................................................................... 44 Dietary Overlap...............................................................................................................47 Results.....................................................................................................................................48 Density of Potential Prey Species.................................................................................... 48 Sample Size Adequacy....................................................................................................48 Composition of Diet........................................................................................................ 48

PAGE 7

7 Prey Selection..................................................................................................................49 Diet Overlap....................................................................................................................50 Discussion...............................................................................................................................50 4 ESTIMATION OF LEOPARD ( Panthera pardus ) ABUNDANCE IN I NDIAN FORESTS USING CAMERA TRAPS IN A MARK-RECAPTURE FRAMEWORK......... 64 Introduction................................................................................................................... ..........64 Methods..................................................................................................................................65 Study Area.......................................................................................................................65 Field Methods..................................................................................................................66 Analytical Methods......................................................................................................... 67 Estimation of population size................................................................................... 67 Estimation of leopard density................................................................................... 69 Results.....................................................................................................................................70 Adequacy of Sampling....................................................................................................70 Sex Ratios........................................................................................................................71 Population Size................................................................................................................ 71 Leopard Density..............................................................................................................71 Discussion...............................................................................................................................72 Conclusion..............................................................................................................................75 5 PRESENCE-ONLY HABITAT SUITABILITY MODELS FOR LEOPARDS ( Panthera pardus ) USING FIELD BASE D AND REMOTELY DERIVED VARIABLES AT TWO SPATIAL SCALES IN MADHYA PRADESH, INDIA............... 88 Introduction................................................................................................................... ..........88 Study Areas.............................................................................................................................89 Methods..................................................................................................................................90 Results.....................................................................................................................................95 Model Validation.............................................................................................................95 Extensive Study Area...................................................................................................... 95 Satpura Tiger Reserve..................................................................................................... 96 Effect of Changing Resolution........................................................................................97 Discussion...............................................................................................................................97 6 CONCLUSION..................................................................................................................... 118 Density of Potential Prey...................................................................................................... 118 Preference of Prey.................................................................................................................118 Density of Leopards..............................................................................................................119 Habitat Model.................................................................................................................. .....119 APPENDIX A INDICES OF UNGULATE AND CARNIVORE ABUNDANCE ............................................. 121 LIST OF REFERENCES.............................................................................................................123

PAGE 8

8 BIOGRAPHICAL SKETCH.......................................................................................................135

PAGE 9

9 LIST OF TABLES Table page 2-1 Density of dominant tree species in the three habitat type s along transects. ..................... 30 2-2 Estimation of density parameters of poten tial prey by the line-transect m ethod in the moist deciduous habitat......................................................................................................31 2-3 Estimation of density parameters of poten tial prey by the line-transect m ethod in the dry deciduous habitat......................................................................................................... 32 2-4 Estimation of density parameters of poten tial prey by the line-transect m ethod in the teak dominated habitat.......................................................................................................33 2-5 Estimation of overall density and its a ssociated param eters by the line-transect method over 4 years in the study area................................................................................ 34 2-6 Density of wild ungulates at various study sites in India................................................... 35 2-7 Estimation of density parameters of poten tial p rey by the vehicle transects assuming Poisson variance.................................................................................................................35 3-1 Estimation of overall density and its asso ciated param eters by the line transect method over 4 years in the study area................................................................................ 54 3-2 Food habits of the leopard obtained by scat analyses. ....................................................... 55 3-3 Food habits of the tiger obtained by scat analyses. ............................................................ 56 3-4 Food habits of the dhole obtained by scat analyses. .......................................................... 56 3-5 Jacobs index values of preference for prey species in tiger di ets at study sites in India. ......................................................................................................................... .........57 3-6 Jacobs index values of preference for prey species in leopard di ets at stu dy sites in India.......................................................................................................................... .........57 3-7 Jacobs index values of preference for prey species in dhole diet s at study sites in India. ......................................................................................................................... .........58 3-8 Diet overlap between tiger, leop ard and dhole using Piankas index. ............................... 58 4-1 Camera-trapping effort at the study sites...........................................................................77 4-2 Leopard sex ratios for the different study sites.................................................................. 77 4-3 Model selection criterio n and tests for Models Mo, Mh, Mb and Mt in the markrecapture framework and a test for population closure...................................................... 78

PAGE 10

10 4-4 Population estimates for leopards at the study sites........................................................... 79 4-5 Density of leopards and estimates of sam pled area using convex polygon and model Mh at the different study sites............................................................................................ 79 4-6 Density of leopards with the associated estimated trapping area using models Mo and Mh.......................................................................................................................................80 4-7 Density estimates for leopards using diffe rent capture functions for the null m odels with the MLSECR method.................................................................................................81 4-8 Relative abundance index va lues f or the 5 estimates in Satpura and Sariska Tiger Reserves.............................................................................................................................82 5-1 Districts, sampling effort and leopard pr esence in the extensive study area in southcentral Madhya Pradesh. .................................................................................................. 100 5-2 List of ecogeographical variables (EGV) with explan ation and source for southcentral Madhya Pradesh. .................................................................................................. 101 5-3 List of ecogeographical va riab les (EGV) with explanation and source for the Satpura Tiger Reserve...................................................................................................................102 5-4 Measures of evaluation for habitat models at different pixel reso lutions (with crossvalidated standard deviations)..........................................................................................103 5-5 Correlation between ENFA factors and EGV for south-central Madhya Pradesh.. ........104 5-6 Correlation between ENFA factors a nd EGV for Satpura Tiger Reserve.. ..................... 105 5-7 Area under various leopard-habitat cate gories in south-central Madhya Pradesh. .......... 106 A-1 Kilometric index values of selected speci es us ing dirt trails in the monsoon from 2002 to 2005....................................................................................................................121 A-2 Encounter rates of tracks of selected carnivore species. .................................................. 122

PAGE 11

11 LIST OF FIGURES Figure page 2-1 Map of the study area, showing the habita t types, transects an d vehicle transects trails....................................................................................................................................36 2-2 Annual densities of selected species in the study area.. ..................................................... 37 2-2 Continued.................................................................................................................. .........38 2-3 Comparison of density estimates between foot transects and vehicle transects in 2005 for potential prey species of large ca rnivores in S atpura Tiger Reserve........................... 39 2-4 Detection function curves for vehicle a nd foot transects for chital, sam bar, langur and peafowl........................................................................................................................40 3-1 Map of Bori Wildlife Sanc tuary and Satpura National Pa rk, showing the location of line transects, dirt ro ads and the study area. ...................................................................... 59 3-2 Relationship between sample size of scats and the percent freque ncy of occurrence in tiger, leopard and dhole diet of langur, chital and sam bar................................................. 60 3-3 Relationship between the number of scats an alyzed and the number of prey species found in the diet of tiger, leopard and dhole...................................................................... 61 3-4 Prey taken by tiger, leopard and dhole in various body weight categories........................ 62 3-5 Observed and expected frequencies of prey item s in scats of tiger,leopard and dhole..... 63 4-1 Identification of leopards based on spot patterns...............................................................83 4-2 Rate of accumulation of new individuals in cam era-trap photographs with increase in sampling time at the four sites........................................................................................... 84 4-3 Camera trapping in 3 sites (Churna, Ka m ti and Lagda) in Satpura Tiger Reserve........... 85 4-4 Map showing camera trap locations w ith half MMDM and full MMDM buffers in Sariska T iger Reserve........................................................................................................ 86 4-5 Map showing camera trap locations w ith half MMDM and full MMDM buffers for one site (Kam ti)..................................................................................................................87 5-1 Cover map of the study area in south-central Madhya Pradesh. ......................................107 5-2 Mosaicked landsat satellite im age of the study area in south-central Madhya Pradesh.. 108 5-3 Landsat satellite image of the study area in Satpura Tiger Reserve. ............................... 109

PAGE 12

12 5-4 Maps of remotely derived variab les f or south-central Madhya Pradesh......................... 110 5-4 Continued.................................................................................................................. .......111 5-4 Continued.................................................................................................................. .......112 5-5 Maps of remotely derived vari ables f or Satpura Tiger Reserve...................................... 113 5-5 Continued.................................................................................................................. .......114 5-5 Continued.................................................................................................................. .......115 5-6 Leopard habitat suitability map for south-central Madhya Pradesh. ............................... 116 5-7 Leopard habitat suitability m ap for Satpura Tiger Reserve............................................. 116 5-8 The predicted-to-expected frequency curves with habita t suita bility values for both the models..................................................................................................................... ...117

PAGE 13

13 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy ECOLOGY OF THE LEOPARD ( Panthera pardus ) IN BORI WILDLIFE SANCTUARY AND SATPURA NATIONAL PARK, INDIA By Advait Edgaonkar May 2008 Chair: Melvin Sunquist Major: Wildlife Ecology and Conservation The ecology of the leopard ( Panthera pardus ) was studied from 2002 to 2006 in the Bori Wildlife Sanctuary and Satpura Na tional Park in Madhya Pradesh, India. Density estimates of the potential prey species of leopards a nd its sympatric carnivores, the tiger ( Panthera tigris) and the dhole (Cuon alpinus ) were made using the line-transe ct method annually from 2002 to 2005, and for three habitat types. The results obtained by vehicle transects were compared with those of foot transects for obtaining reliable density es timates. The food habits and prey preference of leopards, tigers and dholes were quantified. Leopard density estimates for three sites in Bori-Satpura and one site in Rajasthan, the Sari ska Tiger Reserve, were made using camera traps and the mark-recapture method. A predictive ha bitat suitability map for leopards using Environmental Niche Factor Analysis (ENFA) was made at two scales and its reliability was evaluated. The environmental variables important in describing the habitat for leopards were identified and the extent and location of poten tial leopard habitat avai lable for conservation action in south-central Ma dhya Pradesh was quantified. Chital ( Axis axis ) density was higher in the moist de ciduous and teak dominated habitats compared to the dry deciduous habitat. Sambar ( Cervus unicolor ) density was higher in the teak dominated habitat. The densities of nilgai (Boselaphus tragocamelus), wild pig ( Sus scrofa ) and

PAGE 14

14 muntjac ( Muntiacus muntjak ) for the three habitat types were not statistically different. Annual density was lower for all prey species in 2005 as compared to 2002. Sambar was the most important prey species in the le opards diet. It was also the most preferred prey species by leopards, as well as by tigers and dholes. Dens ity of leopards was estimated at 7.3, 7.5, 8.0 and 9.3 per 100 km2 for the four samples in Satpura Ti ger Reserve using the half MMDM method and 4.2, 4.6, 5.3 and 6.2 per 100 km2 for the full MMDM method. The estimates for the sampled area in Sariska Tiger Reserve using th e two methods were 30.9 and 20.7 per 100 km2, respectively. The results of the ENFA model showed that habita t use by leopards in Satpura was strongly associated with moist a nd teak forests, as well as with most prey species and was weakly negatively associated with the distance to villages. At the larger scale, in south-central Madhya Pradesh, leopard habitat was positively associated with terrain ruggedness, sambar availability and percentage of fo rested areas. Approximately 11500 km2 of habitat in southcentral Madhya Pradesh is likely to support leopard populations. The districts with the most optimal habitat were found to be Betul, Hoshangabad and Chhindwara, which have about 2000 km2 of contiguous habitat for leopard conservation.

PAGE 15

15 CHAPTER 1 INTRODUCTION Hum ans have always been fa scinated by carnivores, and our responses to them, whether positive or negative, have been strong and emoti onal. Partly as a result of their food habits, which have placed them in direct competition with us, partly because of their need for large undisturbed areas, and for their valuable body parts, carnivores have been persecuted for many centuries now. As a result the geographic ranges of many specie s have contracted, and their populations have crashed. There is an urgent need to conserve many carnivore species, and the first step towards this is to obtain knowledge about their basi c biology: how many exist, what they eat and where they live. There has been ve ry little research done on most of the 37 extant cat species of the world. This is because cats generally tend to be noc turnal, occur at low densities, and live in remote locations. Th ey are thus difficult and expensive to study. The leopard has had the reputation of being one of the least studied of the large carnivores despite being the most abundant (Hamilton 1976). The situation is hardly different even now, in the Indian context. Most of the studies on leopards have been done in Africa (Bailey 1993; Bertram 1982; Hamilton 1976; Jenny 1996) The sparse information on leopards in the Indian subcontinent has mostly come from studies that focussed on the tiger (Karanth & Sunquist 1995, 2000; Sunquist 1981) or the lion (Chellam 1993). Based on estimates of density and geographic range the leopards total effective global population size has been estimated at greater than 50000 breeding individuals, and is listed as a species of least concern by the IUCN red list. In India, however, it is listed in Schedule I of the Indian Wildlife (Protection) Act, 1972, under th e highest level of prot ection. This is because habitat destruction, loss of wild prey, poachi ng for skins, bones and claws, and poisoning carcasses of livestock killed by leopards ar e a significant threat to the species.

PAGE 16

16 The leopard is a large sized cat, weighing on an average 58 kg for males and 37.5 kg for females (Bailey 1993). It is the most widely di stributed of the wild cats (Nowell & Jackson 1996), and is found in almost every kind of habitat, from the rainforests of the tropics to desert and temperate regions (Kitchener 1991). It occurs from Africa through most of Asia up to the Amur valley in Russia. The Indian subspecies, Panthera pardus fusca is found in all forested habitats in the country, absent on ly in the arid deserts and above the timber line in the Himalayas (Prater 1980). The leopard is quite adaptable with respect to habitat and food requirements, being found in intensively cultivated and inhabited areas as well as near urban development (Nowell & Jackson 1996). Th ere is a wide variation in the eco logy of the species across its range and in different ecosystems. Leopards have been found to be essentially sol itary and territorial animals. They are most likely to socialize at the carca ss of large prey (Hamilton 1976). In Wilpattu, Sri Lanka, the only social groupings seen were mother with cubs and courting pairs (Eisenberg & Lockhart 1972). In Ruhuna National Park, also in Sri Lanka the ma jority of leopards observed were solitary (Santiapillai et al. 1982). Schaller (1976) observed pairs only in three inst ances out of a total of 155 observations, the rest of whic h were of solitary leopards. Scent marking has been conjectured as the primary mode of communication. This includes scraping, marking with feces and spraying of urine, which have been found in tigers to be used most often along trails and trail intersecti ons that serve as common boundaries between territories (Smith et al. 1989) Communication has been speculate d to serve several functions, chief among which are to allow leopards to separate themselves in space and time, to attract the opposite sex during courtship, and to distinguish each other by age, sex and individual status (Bailey 1993).

PAGE 17

17 Home ranges in leopards have been found to vary from being exclusive or slightly overlapping to completely overlapping between the sexes. In Nepal, for example, the home range of a male leopard enclosed th e home ranges of several females (Seidensticker 1976), while in Wilpattu, areas were being used exclusively by a single male and a single female (Muckenhirn & Eisenberg 1973). Male leopards had slightly ove rlapping home-ranges in Thailand (Rabinowitz 1989). In Kruger National Park, South Africa, little spatial overlap betwee n home ranges of adult male leopards in summer has been observed and this decreased even further during the wet season (Bailey 1993). Female home ranges also overlapped a little, while male home ranges completely overlapped many female home ranges, as in the Nepal study (Seidensticker 1976). Female home ranges appeared to be related to ava ilability of prey needed to successfully raise young ones. The juveniles share female home ranges unt il maturity after which they disperse and become transient until they can find a suitable undefended portion of habitat. They can then establish and defend a home range (Eisenberg 1 986). In Asia, leopard home ranges have been reported from Sri Lanka, Nepal a nd Thailand. In Wilpattu, home ranges of four leopards were recorded as between 8 and 10.5 km2 (Muckenhirn & Eisenberg 1973), while female home ranges between 6 and 13 km2 in Nepal (Seidensticker 1976) Thai land male leopards had ranges of 27-37 km2 and female ranges were between 11-17 km2 (Rabinowitz 1989). The low densities of terrestrial herbivores found in rainforests may not be able to support high leopard densities. Home range of a male leopard in wet evergreen forest in Ivory Coast was found to be 86 km2, with partially overlapping female ranges that we re up to three times smaller than the male home range (Jenny 1996). The highest densities recorded in Kruger were 1 per 3.3 km2 where prey biomass varied from 2932 to 6186 kg/km2 (Bailey 1993) while a crude density of 1 per 29 km2 for the entire park has been suggested (Pienaar 1969). For the Serengeti the density of leopards

PAGE 18

18 was estimated at about 1 per 22 to 26.5 km2 (Schaller 1976). In Wilpattu National Park the estimated density was 1 per 29 km2 (Muckenhirn & Eisenberg 1973). There are no published density estimates for India. Leopards have been shown to kill medium-sized prey, mainly impala ( Aepyceros melampus ), but also take a very wide variety of small animals including hyrax, civet and mongoose in Kruger National Park in South Af rica (Bailey 1993). A wide spectrum of the potential prey available in the Ta i National Park, Ivory Co ast, with about thir ty species recorded (Hoppe-Dominik 1984). In the Kalahari desert leopards have been known to take small prey like Bat-eared foxes ( Otocyon megalotis ), jackals ( Canis spp ), genets ( Genetta spp ), hares ( Lepus spp), duiker (Cephalopus spp ) and porcupines ( Hystrix spp ) (Bothma & Le Riche 1984). In Wilpattu leopards took chital, wild pig ( Sus scrofa ), sambar, langur, hare, porcupine and domestic buffalo calves (Muckenhirn & Eisenberg 1973). In Nepal, wild pig sambar, chital, hog deer ( Axis porcinus), muntjac and domestic cattle were pa rt of their diet (Seidensticker et al. 1990). In the Pakistan Himalayas, leopards took mainly wild goats ( Capra aegagrus ) but also preyed on livestock, hare and porcupine (Schalle r 1977). In India too diet ary studies have found that leopards take a range of prey. In the Shivalik hills of Rajaji National Park analysis of scats has shown that leopards eat chital, sambar, muntjac, goral and livestock (Johnsingh pers comm ). In Sariska Tiger Reserve leopard scats contained rodents (Sankar & Johnsingh 2002). The leopards on the Mundanthurai plateau have been preying mainly on sambar (Sathyakumar 1992) while in Bandipur the leopard kills were mainly chital (Andheria et al. 2007; Johnsingh 1983). In Gir, 40 percent of leopard scats contained chit al remains while langur remains were found in 25% of the scats (Chellam 1993). Near Mumbai, leopards living n ear urban areas survive to a large extent on domestic dogs and rodents (Edgaonkar & Chellam 1998). In Chapter 2, I estimate

PAGE 19

19 the density of wild prey in the Satpura Tiger Reserve using the line tr ansect method, while in Chapter 3 I quantify the diet of leopards and its sympatric carnivore species, the dhole ( Cuon alpinus ) and the tiger, and also estimate se lection for the major prey species. The dramatic reduction in tiger populations (Jhala et al. 2008) in India has also meant that there is increasing poaching pressure on the leop ard to meet the demands of the skin and bone trade. This has been borne out by se izures of thousands of skins in recent years. In spite of this, we do not have any reliable estimate of leopard populations in India, neither do we know whether the population is really declining. An important first step is to estimate the densities at which leopards are found in Indi an forests. In Chapter 4, I us e camera trapping and the markrecapture method to estimate leopard densities in three sites in the Satpura Tiger Reserve and and one site in the Sariska Tiger Reserve. To conserve leopards, it is neces sary to first identify areas that have good leopard habitat. In Chapter 5, I present a model that identifie s leopard habitat using presence-only data for Satpura Tiger Reserve and for the larger sout hcentral Madhya Pradesh. The model also identifies the habitat attribut es that contribute to the li kelihood of leopard presence. The aim of this dissertation is to generate further information on the basic ecology of a large carnivore, the leopard ( Panthera pardus ), which has been little studied in India. My objectives are: 1) to estimate leopard prey density over four years in Satpura Tiger Reserve. 2) Quantify prey species selection by the le opard and its sympatri c carnivores, dhole and tiger. 3) Obtain density estimates for leopards using the mark-recapture framework, and 4) to generate a leopard habitat suit ability model for Satpura Tiger Reserve and for south-central Madhya Pradesh.

PAGE 20

20 CHAPTER 2 DENSITY ESTIMATION OF POTENTIAL PREY SPECIES OF LARGE CARNIVORES IN SATPURA TIGER RE SERVE USI NG LINE-TRANSECT SAMPLING. Introduction Accurate and precise estim ation of animal abunda nce is a necessary first step to detect and mitigate unacceptable levels of population change. Until recently there was a paucity of reliable information on population densities of wild ungulates and other vertebrate species in India. This paucity was attributed to funding difficulties, the politics of re search and the relatively small number of scientists engaged in long-term rese arch (Eisenberg & Seidensticker 1976). However, conditions in India have changed, and in the last 10-15 years the number of ungulate studies employing rigorous methods has markedly incr eased and we now have population density estimates of large herbivores from at least 10 protected areas. Density estimates of large herbivores are available for Nagarhole (Kar anth & Nichols 2000; Karanth & Sunquist 1992), Bandipur (Johnsingh 1983; Karanth & Nichols 2000), Bhadra (Jatha nna et al. 2003), Mudumalai (Varman & Sukumar 1995), Gir (Khan et al. 1996) Melghat (Karanth & Nichols 2000), Pench (Acharya 2008; Biswas & Sankar 2002; Karanth & Nichols 2000), K aziranga (Karanth & Nichols 2000), Ranthambore (Bagchi et al. 2004; Karanth & Nichols 2000) and Sariska (Avinandan 2003; David et al. 2005). While these efforts are laudable, India is unde rgoing a rapid change and by some estimates the economy has been projected to grow annually at a rate of 5 percent or more for the next 30 years (Wilson & Purushothaman 2003). There is increasing pressure on na tural areas and there are reports of decline in forest quality (Lele et al. 2000). A recently rele ased report has estimated a population of between 1165 and 1657 tigers in India (Jhala et al. 2008), which is much lower than estimates from just a decade earlier. A critical component for conservation of tigers and other carnivores is the availability of wild prey (Karanth et al. 2004b). Th ere is thus an urgent

PAGE 21

21 need to establish baseline densities of prey in all protected areas in India and put in place a monitoring program that would dete ct changes in their populations. Ideally the monitoring scheme should use a me thod that minimizes bias and error while maximizing precision, and has sufficient power to be able to detect population changes. Distance-based methods have the advantage of not requiring animals to be handled, are relatively easy to apply and give robust results if underlying assumptions are met. Violations of these assumptions bias the resulting estimates in vari ous ways, the details of which can be found in Buckland et al. (2001). A disadvantage of line transects is that a la rge number of observations are needed to calculate the detection function precisely, and obtaining these is a labor intensive process. Vehicle transects have been run along road networks as a substitute to foot transects (Ogutu et al. 2006; Ward et al. 2004). These yield a larger e ffort in the same time. However, the resulting estimate may be biased as roads are not usuall y randomly laid with respect to the animals (Varman & Sukumar 1995). Some species are attracted to the edge habitat created by roads while other species may avoid the disturbance. This is likely to be areaspecific depending on the configuration of the road netw ork and the species being monitored. It is nevertheless worth investigating whether vehicle tr ansects can be used to monitor populations in a given area. The objectives of the present st udy are to 1) Estimate the dens ity of potential prey species of tigers, leopards and dholes using the line-transect method. 2) Evaluate changes in density over 4 years of sampling to enable monitoring of the population, and 3) Determin e if vehicle transects yield density estimates equivalent to those obtained by foot transects.

PAGE 22

22 Methods Study Area The Satpura Tiger Reserve (22o19' to 22 o 30' N and 77 o 56' to 78 o 20' E) is a 1428 km2 protected area located in the Hoshangabad district of Madhya Prades h state in India. It comprises of the Pachmarhi and Bori Wildlife Sanctuaries, and Satpura National Park. An intensive study area of approximately 200 km2 was located in Bori Wildlife Sa nctuary and Satpura National Park (Figure 2-1). The intensive study area is a mosaic of dry and moist deciduous mixed forest. Teak ( Tectona grandis) plantations replaced mixed forests in some areas, though many of these plantations are not pure teak, but are mixed with other species. Common tree species found there include Palas, Butea monosperma; Mahua, Madhuca latifolia ; Landia, Lagerstroemia parviflora; Kari, Schleicheria oleosa ; Saj, Terminalia arjuna and Tendu, Diospyros melanoxylon The tiger (Panthera tigris ), leopard ( Panthera pardus ) and the dhole ( Cuon alpinus) are carnivores of management interest in the study area. Other carnivor es include jackal ( Canis aureus ), striped hyena ( Hyaena hyaena), sloth bear ( Melursus ursinus ), jungle cat ( Felis chaus ), palm civet ( Paradoxurus hermaphroditus ), small Indian civet ( Viverricula indica ), ruddy mongoose ( Herpestes smithii ), common mongoose ( Herpestes edwardsi ) and ratel ( Mellivora capensis ). A diverse community of ungulates and ground birds are preyed upon by the carnivores. Potential prey for tigers, le opards and dholes include the wild pig ( Sus scrofa ), chousingha ( Tetracerus quadricornis ), chital ( Axis axis ), Indian muntjac ( Muntiacus muntjak ), sambar ( Cervus unicolor ), nilgai ( Bosephalus tragocamelus ) gaur ( Bos gaurus), the common langur ( Semnopithecus entellus ), black-naped hare ( Lepus nigricollis ) and Indian porcupine ( Hystrix indica).

PAGE 23

23 Estimation of Habitat Types A georeferenced and orthorectified cloud-free Landsat ETM+ im age for the study area was obtained from the Global Landcover Facility (www .landcover.org). Spectral signatures for the classification supervision were obtained by using information from vegetation plots. A sample of 473 circular plots of 10 m radius were laid along transe cts and dirt trails in the study area. The number and composition of woody tree species larger than saplings was noted inside the plot. Five cover types were delineated. These were: mois t forest, dry forest, bare ground/village, teak dominated forest and water. Supervised classifi cation was performed using FISHER classifier for the study site using Idrisi Kilimanjaro (Eastman 2004). The transects were then stratified post hoc as belonging to one of 3 different habi tats: dry deciduous, moist deciduous and teak dominated. Any transect traversing more than one habitat was allocat ed to the habitat type that it most represented. A fourth habitat type, the riverine forest habitat, is found along streams. It was considered part of the moist deciduous habitat fo r the purposes of stratifi cation since it formed a small proportion of the overall ar ea. Density of trees along the tr ansects was estimated from the vegetation plots, and a Sorensons index (Krebs 1989 ) was calculated to quantify the tree species similarity between the three habitats. Estimation of Prey Density Density of the m ajor prey species was es timated by the line-transect sampling method. Twenty permanent transects were laid in the st udy area. The area was divided into approximately 5 km2 grids and ten grids were randomly chosen. A 2 km transect was laid in a random direction in each grid to make a total of 10 transects. The vegetation on the transect was minimally cut so as to allow observers to move through the forest, but not so much as to change the nature of the habitat close to the transect. Th ese ten transects were then suppl emented in the second year by ten more transects of 3 km length. These were la id systematically so that gaps in coverage

PAGE 24

24 between the first ten transects were filled as much as possible. The location of transects is shown in Figure 2-1. The total length of the transects was 50 km. The tw enty transects were walked repeatedly for a total effort of 1272 km. Transects were walked early in the morning and evening in summer and winter at a speed of about 3 km ph, so that it took 40 mi nutes to walk a 2 km transect and one hour for the 3 km transect. The species, group size, angle and angular distance to the center of the group or to the individual was noted. Distance measurements were taken with a laser rangefinder (Bushnell Yardage Pro 400) and angles were measured with a magnetic compass. Program Distance v5 release 2 (Thomas et al. 2006) was used to estimate the density of prey species. Two estimates of density were made: 1) A pooled density was estimated over all habitats for each year. 2) Density was estimated pooled over all years for each habitat type. This was calculated as a mean of the densities in eac h habitat type weighted by the area of each habitat. An exploratory analysis of the distribution of the distances was done by grouping them in small intervals and plotting the resulting hist ograms as recommended by Buckland et al. (2001). Depending on the resulting histogram, data were truncated at an appropriate distance for each species. Evidence of heaping, spikes near the lin e and avoidance movements or a sharp drop-off away from the line was investigated. The data we re then grouped into appropriate intervals for each species so that the dete ction function gave a good fit. A detection-probability function was estimated from pooled data across years and habitats for each species to maximize the number of sighti ngs. Since all three hab itats had similar tree densities, there was no reason to believe that detections diffe red between habitats and years. The data were modeled with the uniform, half normal and hazard rate models fitted with the cosine and simple polynomial series for each species. The negative exponentia l model, recommended

PAGE 25

25 only as a last resort, was used fo r the Indian peafowl si nce the other models did not fit well. The model with the smallest Akaike Information Crit erion (AIC) value was sele cted as the best-fit model provided that the p-value for the chi-square goodness of fit for the model was greater than 0.05 (Burnham & Anderson 2003). The cluster size was calculated as a mean of observed clusters, and variance was calcu lated by bootstrapping observations within transects for most species. In the case of grey jungle fowl, mun tjac and gaur the bootstrap estimates failed to converge. Variance was estimated em pirically for these species. Vehicle Transects Sixty-five drives were m ade using a 4-wheel drive vehicle on the dirt trails in the study area, with a total effort of 388 km. Two observers were used and the vehicle was driven at between 10 and 15 km per hour. Perpendicular distance to the sighting was estimated by stopping in front of the animal cluster and taking a distance measurement with a laser rangefinder to the center of the cluster. The drives overlapped with each other in spatial coverage, and were made on one trail network. Th e data were analyzed as if it was from one drive to avoid inflating degrees of freedom, a nd the variance was estimated using the Poisson assumption. Results Description of Habitat The teak dom inated habitat had the highest stem density (798 trees/ha) and number of tree species (42) of which 20 percent was teak (163 stems/ha). The dr y deciduous habitat had 35 tree species and a density of 673 stems/ha, of which the most common species was Diospyros melanoxylon (134 stems/ha). The moist deciduous habita t had 40 tree species (541 stems/ha) and was dominated by bamboo (83 clumps/ha). The ten mo st dominant species in each habitat type are presented in Table 2-1. The Sorensens index of similarity between teak dominated and dry

PAGE 26

26 deciduous was 0.86, between teak dominated an d moist deciduous was 0.83, and between dry deciduous and moist deciduous was 0.80. Estimation of Density Density estim ates of potential prey species in the moist deciduous, dry deciduous and teak dominated habitats are presented in Tabl es 2-2, 2-3 and 2-4. Among st the ungulates, chital were found in significantly higher densities in the moist deciduous and teak dominated habitat compared to the dry deciduous habitat. Sambar was found in significantly higher densities in the teak dominated habitat. The dens ities of nilgai, wild pig and muntjac for the three habitat types were not statistically different. The densities of black-naped hare, grey jungle fowl, red spurfowl and Indian peafowl were also not significan tly different among the three habitats. Common langur density was highest in moist deciduous, fo llowed by teak dominated habitat and then the dry deciduous habitat. When pooled over all the habitats, the number of observations of chital, sambar, nilgai, muntjac, wild pig and peafowl were sufficient to estimate density for each year from 2002 to 2005. Densites were lower for all species in 2005, the last year of sampling (Figure 2-2 and 2-3) when compared to the first year, 2002. The estimates for 2002 and 2005 were statistically different for sambar, nilgai and common langur. Average density, weighted by area of each habita t for 11 species pooled over four years is presented in Table 2-5. Common langur is the most abundant species. Amongst ungulates, chital numbers were highest, followed by sambar, nilg ai, wild pig, gaur and muntjac. Ungulate densities in Satpura Tiger Reserve are among th e lower estimates when compared to other protected areas in India (Table 2-6). Density estimates derived from vehicle transe cts are given in Tabl e 2-7. A comparison of density estimates from vehicle transects with th ose from foot transects for the same year are

PAGE 27

27 presented in Figure 2-3, and dete ction function curves for the two methods are presented in Figure 2-4. Density estimates of chital, sambar and nilgai by foot tran sects were higher than those estimated by vehicle transect s, but the difference is not stat istically significant. Density of muntjac, langur, peafowl and w ild pig is greater when estimat ed by vehicle transects but the difference is significant onl y for peafowl and langur. Discussion The accuracy of density estim ates depends on how well the underlying assumptions are met. The data were gathered by trained observers using a laser rangefinder and compass to estimate the bearing and distance to the animal group. Detections near the line, as shown by the low chi-square values for the first distance inte rval, were as expected for each model for all species. There was no evidence of heaping or a sharp drop-off indicating evasive movement in response to the observer for most species. Estimated strip width, as could be expected if distance played a major role in detectabil ity, was wider for the large sized species than for the small sized species. A notable exception was the common langur, which was dete cted at larger distances. Overall wild ungulate density in Satpura Tige r Reserve is lower than that reported for protected areas such as Nagar hole and Bandipur in southern I ndia and Kanha and Pench tiger reserves in Madhya Pradesh, but is comparable to other protected areas in central India like Tadoba, Melhghat and the Maharshtra side of Pe nch tiger reserve (Karanth & Nichols 2000). In the study area, the densities of most spec ies were lowest in th e dry deciduous habitat though some of these differences were not signifi cant. There were fewer water sources in this habitat and it also tended to be closer to the villages found inte rspersed within the study area. These factors could be responsible for the lower densities. Nilg ai, which is known to tolerate disturbance and lack of water (Bagchi et al. 2003a), was not found in lower density in this habitat. The teak dominant habita t had the highest density of sambar, but this area was close to

PAGE 28

28 rugged terrain and also had a number of artificial waterholes. Sambar is known to prefer hilly areas (Bhatnagar 1991) and this could ha ve caused the higher density seen here. Estimation of annual density shows up a pattern of density reduction for all species for the last year. Visual inspection of the data does not show a continuous declining trend for any species except for common langur. However, four years of data are too short a time period to statistically estimate a trend or rate of change using methods like generalized additive modeling (GAM) as has been done for Nagarhole (Ganga dharan 2005). One study has speculated about long-term cyclic changes associated with 3 to 10 year lagged rainfall patterns (Ogutu & OwenSmith 2005) in Africa, but there is little inform ation to indicate the reason for this decline. Possible reasons could be pressu re due to over-grazing, illegal po aching or part of a natural cyclic tendency. Personal observation did not indicate a higher degree of poaching for the last year, nor was there a change in the number of domestic cattle over the years. It can only be speculated that the below average rainfall in 2001 and 2002 (Mooley et al. 2007) may be responsible for the reduction in ungulate density in 2005. Estimates of variance of density were lowe st for langur and sambar for which a large number of observations were ma de, and highest for gaur, of wh ich only 35 groups of which were observed. The variance estimate is a combination of variances in the detection probability, cluster size and encounter rate, with the encounter rate variance bei ng the major component. Encounter rate variance remained the major component for all the years for each species, except for jungle fowl in 2005, where detection probability was the major component. A larger effort can be achieved in a short amount of time with vehicle transects. Except for langur and peafowl, confidence intervals for the two density estimates overlapped. The state forest department clears viewing lanes along some dirt trails fo r tourism purposes. This probably

PAGE 29

29 attracts peafowl, muntjac, langur and wild pig to the cleared ar eas for foraging leading to an increased estimate of density. There may some tendency for larger ungulates such as chital, sambar and nilgai to stay away from the distur bance caused by roads, leading to reduced density estimates, though the differen ce is not significant. Even though densities of most species are moderately low in th e study area, the ungulate community is still intact and should be protected. The Bori-Sat pura area is large enough to be potentially able to support a relatively large tiger population, but to do this the ungulate prey base will have to be enhanced. The baseline es timates generated in this study can be used to monitor future changes in population. The last year of sampling showed lower density for all species, and that is a matter of concern. It is therefore recommended that a monitoring program be initiated and protection meas ures strengthened to arrest the putative decline in wildlife populations in the study area. Vehicl e transects can be used as the network of dirt trails is sufficiently extensive to be able to obtain reasonably accurate results.

PAGE 30

30 Table 2-1. Density of dominant tr ee species (individuals per hectar e) in the three habitat types along transects. Rank order Teak dominated plantation (N= 73 plots) Dry deciduous teak (N = 73 plots) Moist deciduous teak (N = 61 plots) 1 Tectona grandis (163) Diospyros melanoxylon (134) Bamboo species (83) 2 Diospyros melanoxylon (85) Choloroxylon swietania (95) Diospyros melanoxylon (79) 3 Terminalia arjuna (78) Terminalia arjuna (60) Tectona grandis (70) 4 Lagerstroemia parviflora (71) Tectona grandis (53) Terminalia arjuna (32) 5 Aegle marmelos ( 41) Bamboo species (50) Saccopetalum tomentosum (29) 6 Anogeissus latifolia (40 ) Acacia catechu (42) Madhuca indica (28) 7 Zizyphus xylopara (37) Buchanania lanzan (35) Lagerstroemia parviflora (27) 8 Saccopetalum tomentosum (34 ) Madhuca indica ( 29) Emblica officinalis (22) 9 Buchanania lanzan (34) Lagerstroemia parviflora (22) Choloroxylon swietania (20) 10 Choloroxylon swietania (30) Emblica officinalis (21) Zizyphus xylopara (20)

PAGE 31

31 Table 2-2. Estimation of density parameters of po tential prey by the line-transect method in the moist deciduous habitat. Species n D CV DCI D Ds Cv DsCI Ds Model Chital 92 8.0 14.06.2-10.52.412.52.0-3.1 Hazard Polynomial Sambar 68 3.8 10.63.0-4.41.810.11.4-2.0 Half-Normal Cosine Nilgai 26 1.4 17.31.1-2.10.715.20.5-1.0 Half-Normal Cosine Muntjak 34 1.2 21.90.8-2.01.121.60.6-1.7 Half-Normal Cosine Wild pig 29 2.5 26.61.4-3.90.814.80.6-1.0 Half-Normal Cosine Black-naped hare 21 2.7 16.22.1-3.72.615.62.0-3.5 Half-Normal Cosine Common langur 313 39.9 10.334.0-51.09.19.58.0-11.7 Half-Normal Cosine Indian peafowl 29 2.0 21.11.3-2.91.318.70.8-1.7 Neg exp Cosine Red spurfowl 20 2.7 20.41.7-3.61.518.81.0-2.0 Half-Normal Cosine Grey jungle fowl 34 2.9 26.91.5-5.61.626.20.8-3.0 Uniform Polynomial n: number of observations, D: density of individuals/km2, Ds: Density of groups/km2, CV: coefficient of variation, CI: 95% Confidence. Sample size: 6 tran sects, effort: 408 km.

PAGE 32

32 Table 2-3. Estimation of density parameters of po tential prey by the line-transect method in the dry deciduous habitat. Species n D CV DCI D Ds Cv DsCI Ds Model Chital 27 1.9 14.51.4-2.40.613.40.5-0.4Hazard Polynomial Sambar 70 3.1 10.22.4-3.6 1.59.61.1-1.7Half-Normal Cosine Nilgai 37 1.6 17.01.2-2.30.814.50.6-1.1Half-Normal Cosine Muntjak 17 0.5 36.70.2-1.10.436.50.2-0.9Half-Normal Cosine Wild pig 15 1.0 25.90.6-1.60.314.20.2-0.4Half-Normal Cosine Black-naped hare 42 4.3 15.63.4-6.14.115.13.2-5.7Half-Normal Cosine Common langur 155 15.7 10.413.4-20.23.69.63.1-4.6Half-Normal Cosine Indian peafowl 29 1.6 21.11.0-2.31.018.80.7-1.4Neg exp Cosine Red spurfowl 23 2.5 20.41.5-3.31.418.80.9-1.9Half-Normal Cosine Grey jungle fowl 42 2.9 18.91.9-4.41.517.91.0-2.3Uniform Polynomial n: number of observations, D: density of individuals/km2, Ds: Density of groups/km2, CV: coefficient of variation, CI: 95% Confidence. Sample size: 8 tran sects, effort: 513 km.

PAGE 33

33 Table 2-4. Estimation of density parameters of po tential prey by the line-transect method in the teak dominated habitat. Species n D CV DCI D Ds Cv DsCI Ds Model Chital 70 7.1 13.95.6-9.22.112.51.7-2.7Hazard Polynomial Sambar 124 8.0 10.36.3-9.33.79.73.0-4.3Half-Normal Cosine Nilgai 32 2.0 17.61.5-3.01.015.10.8-1.4Half-Normal Cosine Muntjak 12 0.5 24.60.3-0.90.424.30.2-0.8Half-Normal Cosine Wild pig 19 1.9 26.21.1-2.90.615.60.4-0.8Half-Normal Cosine Black-naped hare 20 3.0 15.22.4-4.2.814.62.3-3.9Half-Normal Cosine Common langur 169 25.1 10.421.4-32.25.79.65.0-7.4Half-Normal Cosine Indian peafowl 40 3.3 20.32.1-4.52.018.61.4-2.7Neg exp Cosine Red spurfowl 16 2.5 20.41.5-3.41.418.80.9-1.9Half-Normal Cosine Grey jungle fowl 10 1.0 36.90.4-2.50.536.40.2-1.3Uniform Polynomial n: number of observations, D: density of individuals/km2, Ds: Density of groups/km2, CV: coefficient of variation, CI: 95% Confidence. Sample size: 6 tran sects,effort: 351 km.

PAGE 34

34 Table 2-5. Estimation of overall density and its as sociated parameters by the line-transect method over 4 years in the study area. Species n D CV DCI D Ds Cv DsCI Ds Model Chital 189 5.4 13.84.2-7.11.612.41.3-2.1Hazard Polynomial Sambar 262 4.0 10.33.2-4.71.99.71.5-2.2Half-Normal Cosine Nilgai 95 1.6 17.01.2-2.30.814.70.6-1.1Half-Normal Cosine Muntjac 63 0.8 19.00.6-1.20.717.30.5-1.1Half-Normal Cosine Wild pig 63 1.8 26.21.1-2.90.614.50.4-0.7Half-Normal Cosine Black-naped hare 83 3.4 15.62.7-4.73.215.02.6-4.4Half-Normal Cosine Gaur 35 0.8 37.40.4-1.80.233.40.1-0.4Half-Normal Cosine Common langur 637 28.3 10.324.1-36.36.49.55.7-8.3Half-Normal Cosine Indian peafowl 98 2.0 20.01.3-2.91.317.70.9-1.7Neg exp Cosine Red spurfowl 59 2.6 20.41.6-3.51.518.81.0-1.9Half-Normal Cosine Grey jungle fowl 86 2.7 17.11.8-3.81.416.01.0-2.0Uniform Polynomial n: number of observations, D: density of individuals/km2, Ds: Density of groups/km2, CV: coefficient of variation, CI: 95% Confidence. Sample size: 20 transects, effort: 1272 km.

PAGE 35

35 Table 2-6. Density of wild ungulates (individuals per km2) at various study sites in India. Place Chital Sambar Nilgai Muntjac Wild pig Gaur Bandipur1 20.1 5.6-0.70.7 7.0 Nagarhole2 49.1 3.4-4.33.4 5.6 Pench-M.P.3 80.7 6.10.4 -2.6 0.3 Kanha1 49.7 1.5-0.62.5 Ranthambore5 31.0 17.111.4 -9.8 Sariska6 27.6 8.45.2 -17.5 Gir7 25.2 1.80.4 -2.1 Bhadra4 2.3 5.8-5.42.6 0.7 Melghat1 2.7-0.60.5 1.0 Tadoba1 3.2 3.30.70.92.6 1.8 PenchMaharshtra1 5.8 5.90.5 -2.0 0.8 Bori-Satpura8 5.4 4.01.60.81.8 0.8 1Karanth and Nichols (2000) 2Karanth and Sunquist (1992), 3Biswas and Sankar (2002), 4Jathanna et al. (2003), 5Bagchi et al. (2003) 6Avinandan, D (2003) 7Khan et al (1996) 8This study. Table 2-7. Estimation of density parameters of potential prey by the ve hicle transects assuming Poisson variance. Species n D CV DCI D Ds Cv DsCI Ds Model Chital 51 3.7 17.02.6-5.21.114.60.8-1.4Uniform cosine Sambar 61 2.3 17.71.7-3.31.115.10.8-1.5Uniform cosine Nilgai 21 0.5 31.40.3-0.90.328.00.2-0.6Half-Normal Cosine Muntjac 35 1.3 21.90.8-1.91.019.60.7-1.5Uniform cosine Wild pig 18 3.3 44.91.4-7.90.836.80.4-1.7Neg exp cosine Common langur 148 36.0 13.727.5-47.05.112.04.1-6.5Hazard rate Indian peafowl 74 4.9 18.03.4-7.02.615.11.9-3.5Half normal n: number of observations, D: density of individuals/km2, Ds: Density of groups/km2, CV: coefficient of variation, CI: 95% Confidence. Sample size: 1 transect, effort: 388 km.

PAGE 36

36 Figure 2-1. Map of the study area, showing the habitat types, transects and vehicle transects trails.

PAGE 37

37 A. Year 2002200320042005 Density (individuals per km2 ) 2 4 6 8 10 B. Year 2002200320042005 Density (individuals per km2 ) 2 3 4 5 6 7 8 9 D. Year 2002200320042005 Density (individuals per km2 ) 0.0 0.5 1.0 1.5 2.0 2.5 C. Year 2002200320042005 Density (individuals per km2 ) 0.5 1.0 1.5 2.0 2.5 3.0 3.5 Figure 2-2. Annual densities of se lected species (individuals/km2) in the study area. A) Chital. B) Sambar. C) Nilgai. D) Mun tjac. E) Common langur. F) Indian peafowl. G) Wild pig. Error bars are bootstrapped 95% confidence limits.

PAGE 38

38 E. Year 2002200320042005 Density (individuals per km2 ) 10 20 30 40 50 60 F. Year 2002200320042005 Density (individuals per km2 ) 0 1 2 3 4 5 G. Year 2002200320042005 Density (individuals per km2 ) 0 1 2 3 4 5 Figure 2-2: Continued.

PAGE 39

39 Species ChitalSambarNilgai Density (individuals per km2) 0 1 2 3 4 5 6 Vehicle transect Foot transect Species Common langur Density (individuals per km2) 0 10 20 30 40 50 Species MuntjakIndian peafowlWild pig Density (individuals per km2) 0 2 4 6 8 10 Figure 2-3. Comparison of density estimates (individuals/km2) between foot transects and vehicle transects in 2005 for pot ential prey species of la rge carnivores in Satpura Tiger Reserve. Error bars are boot strapped 95% confidence limits.

PAGE 40

40 LANGUR (vehicle transect)0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 0102030405060708090100PERPENDICULAR DISTANCE (m)DETECTION PROBA B PEAFOWL (vehicle transect)0.0 0.2 0.4 0.6 0.8 1.0 1.2 0102030405060708090100PERPENDICULAR DISTANCE (m)DETECTION PROB A SAMBAR (vehicle transect)0.0 0.2 0.4 0.6 0.8 1.0 1.2 0102030405060708090100110120PERPENDICULAR DISTANCE (m)DETECTION PROB CHITAL (vehicle transect)0.0 0.2 0.4 0.6 0.8 1.0 1.2 0102030405060708090100110120PERPENDICULAR DISTANCE (m)DETECTION PROB CHITAL (foot transect)0.0 0.2 0.4 0.6 0.8 1.0 1.2 0102030405060708090100PERPENDICULAR DISTANCE (m)DETECTION PRO B SAMBAR (foot transect)0.0 0.2 0.4 0.6 0.8 1.0 1.2 0102030405060708090100PERPENDICULAR DISTANCE (m)DETECTION PROBA B LANGUR (foot transect)0.0 0.2 0.4 0.6 0.8 1.0 1.2 0102030405060708090100110120PERPENDICULAR DISTANCE (m)DETECTION PROBA B PEAFOWL (foot transect)0.0 0.2 0.4 0.6 0.8 1.0 1.2 01 02 03 04 05 06 0PERPENDICULAR DISTANCE (m)DETECTION PROB A Figure 2-4. Detection function curves for vehicle and foot transects for chital, sambar, langur and peafowl.

PAGE 41

41 CHAPTER 3 PREY SELECTION AND THE F OOD HABITS OF TIGER, LEOPARD AND DHOLE IN SATPURA TIGER RESERVE. Introduction Inform ation about interactions between tropi cal large carnivore species is scarce. Some food habits studies have been conducted on single species (Bagchi et al. 2003b; Biswas & Sankar 2002; Edgaonkar & Chellam 2002; Reddy et al. 2004), and there is information on diet selection and overlap between multiple species including tigers, leopards and dholes from southern India (Johnsingh 1983; Karanth & Sunquist 1995), and on leopards and tigers (Sankar & Johnsingh 2002) and lions and leopards (Chellam 1993) in we stern India. Differential spatial use by tigers and leopards was reported by one study in Ne pal (Seidensticker 1976), but not in another (Karanth & Sunquist 1995). In the neotropics, spatial avoidance of jaguars ( Panthera onca) by pumas ( Puma concolor ) (Scognamillo et al. 2003) was seen at fine scales, but another (Taber et al. 1997) did not find a similar pattern. Some de gree of dietary separa tion between pumas and jaguars has been noted, with jagua rs tending to take slightly la rger prey and more peccaries (Emmons 1987). Tigers and leopards are opportunistic stalking predat ors and are expected to kill more randomly as opposed to dhole, which is a coursing predator (Sch aller 1967). Dholes, or Asiatic wild dogs, are also more diurnal than tigers and leopards (Johnsingh 1983). They are group living, coursing predators weighing about 20 kg. Anecdotal evidence exists of aggressive interactions between the three car nivores, especially between le opards and dholes. There is need for more data on the potential for competition and resource overlap among these major predators in tropical forest assemblages over a range of resource availabilities. The present study describes the prey taken, quantifies the diet ary overlap and measures the prey selectivity of tigers, leopards and dholes at a site where the abundance of prey is lower than other at places where food habits of these carnivores have been studied.

PAGE 42

42 Methods Study Area The study was conducted in the Satpura Ti ger Reserve (STR). STR covers 1428 km2 in area, and is located in the Hosha ngabad district of the central I ndian state of Madhya Pradesh in India. It includes three administrative units, the Pachmarhi and Bori Wildlife Sanctuaries, and Satpura National Park. An intensive study area of approximately 200 km2 was located in Bori Wildlife Sanctuary and Satpura National Park (Figure 3-1). The forest in STR (22o19' to 22 o 30' N and 77 o 56' to 78 o 20' E) is mainly of the moist deciduous type (Champion & Seth 1968). The intensive st udy area is a mosaic of dry and moist deciduous forest dominated in many places by tea k. Teak plantations replaced mixed forests in some areas, though now even these are not pure t eak, but have secondary growth of species. Common species found there include Palas, Butea monosperma; Mahua, Madhuca latifolia ; Landia, Lagerstroemia parviflora; Kari, Schleicheria oleosa ; Saj, Terminalia arjuna and Tendu, Diospyros melanoxylon A diverse assemblage of fauna is found including wild pig ( Sus scrofa ), chousingha ( Tetracerus quadricornis ), chital ( Axis axis ), Indian muntjac ( Muntiacus muntjak ), sambar ( Cervus unicolor), chinkara ( Gazella gazella ), nilgai ( Bosephalus tragocamelus) and gaur ( Bos gaurus ). The common langur ( Semnopithecus entellus ) and rhesus macaque (Macaca mulatta ) are the primates found here. Carniv ores are represented by tiger (Panthera tigris), leopard ( Panthera pardus ), wild dog ( Cuon alpinus ), jackal ( Canis aureus), striped hyena ( Hyaena hyaena), sloth bear ( Melursus ursinus), jungle cat ( Felis chaus ), palm civet ( Paradoxurus hermaphroditus), small Indian civet ( Viverricula indica ), ruddy mongoose ( Herpestes smithii ), common mongoose (Herpestes edwardsi ) and ratel ( Mellivora capensis ). Black-naped hare ( Lepus nigricollis ), Indian porcupine ( Hystrix indica ), Indian giant squirrel ( Ratufa indica ) and

PAGE 43

43 the large Indian flying squirrel (Petaurista petaurista ) are some of the smaller mammals found here. Reconstruction of Carnivore Diets Scats were collected opportuni stically as well as system a tically along animal and manmade trails and dirt roads in the study area (Fi gure 3-1). Identification of tiger and leopard scats was based on associated tracks or sign. Scat size or diameter was not used as the criterion for discriminating between species as there is suspec ted to be overlap in scat size amongst the three species and this may lead to significant misidentification of scats (Farrell et al. 2000). Only scats of tigers and leopards that had associated tracks or sign near them were collected to ensure correct identification. Scats of dhol es were easy to identify as th ey were deposited at communal defecation sites (Johnsingh 1983). Scats were washed and undigested remains of hair were mounted on a slide and compared under a microscope with a reference colle ction of hair at the Wildlife Institute of India following a standard protocol (Mukherjee et al. 1994). Bird and rodent taxa were not identified to speci es level. In total 193 leopard scats, 93 tiger scats and 81 dhole scats were analyzed. The percentage frequenc y of occurrence of all the major species was calculated along with their boot strap confidence intervals. Sample Size Adequacy To check for the stability of percent frequency of occurrence in the diet, all scats for each carnivo re were randomized and the percentage fre quency of occurrence of each prey item in the diet was plotted cumulatively, at an interval of 10 scats. The number of scats required for the frequencies to reach an asymptote was considered sufficient to quantify that prey item in the diet reliably.

PAGE 44

44 Prey Biomass and Number The frequency of occurrence is b iased toward s smaller sized prey, since relatively more scats are produced for smaller prey than larger pr ey. To correct for this bias, relative frequencies of prey were converted to relative biomass consumed for tigers and leopards using an equation estimated for cougars (Ackerman et al. 1984), an d for dholes using an equation estimated for wolves (Floyd et al. 1978). This regression equa tion estimates the number of field collectable scats for a given weight of prey biomass. These are y = 0.38 + 0.020 x (for dhole) and y = 1.98 + 0.035 x (for tigers and leopards) where the independent variable x is the average weight of the prey and the dependant variable y is the number of field collectable scats for that weight of prey. The depe ndant variable can then be converted into the relative biomass of prey consumed by multiplying it by the relative frequency of each prey species found in the scat s. The relative number of each species consumed is obtained by dividing relative biomass by the average weight of the prey species. The weight of various prey species killed by tiger leopard and dhole was assumed to be similar to that used in previous research (K aranth & Sunquist 1995). Estimation of Prey Selection Selectivity can be defined as taking a prey at frequencies different from that expected given its availability (Chesson 1978). If there is no selection one would expect a prey item to be taken at relative frequencies similar to the relative frequency of its availa bility. Any statistically significant deviation, whether posit ive or negative, would indicate, preference or avoidance of that prey type.

PAGE 45

45 Availability of a prey species is likely to be a function of its abundance, anti-predatory behavior, habitat selection at a fi ne scale and time of activity. It is assumed, as in other studies (Bagchi et al. 2003b; Biswas & Sankar 2002; Karanth 1995), that abundance is the major component of availability. Abunda nce was therefore estimated as the density of groups of the major prey species, since the probability of encoun tering prey is likely to be proportional to the density of groups, rather than of individuals (Karanth & Sunquist 1995). The prey density was estimated by the line-transect method. Twenty perm anent transects were laid in the study area. The first ten transects were laid randomly. The area was gridded into approximately 5 km2 grids and ten grids were randomly chosen. A 2 km transect was laid in a random direction in each grid. The next ten transects were then laid as 3 km lines so that gaps in coverage between the first ten transects were filled as much as possible. The location of tr ansects is shown in Figure 3-1. Transects were walked repeatedly for a total effort of 1272 km. The species, group size, angle and angular distance to the indi vidual or center of group was noted. Distance measurements were taken with a laser rangefinder (Bushnell Yardag e Pro 400) and angles were measured with a magnetic compass. Program Distance v5 release 2 (Thomas et al. 2006) was used to estimate the density of prey species. Selectivity was quantified by comparing the obser ved frequency of each prey species in the scats to expected frequencies (Link & Karanth 1994). Expected frequencies were derived from the densities estimated by line transects. If a kill of speciesi with a density di produces i scats, then the proportion of scats produc ed when the carnivore takes prey in proportion to their density is given by i ii ii id d

PAGE 46

46 The program SCATMAN v2.0 (Hines 2002) was used to estimate prey selection by comparing the i to the the observed proportion based on ra ndom samples of predator scats. The program uses the estimated di and i, and the variation associated with these parameters. It implements a parametric bootstrap designed to handle the problem of excessive Type I error caused by comparison of estimated frequencies as opposed to exact frequencies (Link & Karanth 1994). Inputs to the program are the estimated av ailability and standard error of each prey species, and the number of collectable scats that are produced by an aver age kill of each prey species, along with their standard errors. High chi-square values in the output indicate that observed frequencies are significantl y different from expected, and the presence of selectivity of prey. The contribution of each species to the tota l chi-square indicates wh ether the prey species is taken more or less than expected. The Jacobs index (Jacobs 1974) has been used to estimate dietary pref erence in carnivores (Hayward 2006; Hayward et al. 2006a; Hayward & Kerley 2005; Hayward et al. 2006b). It has the advantage of being simple to compute and can be used to compare across studies easily. Availability and utilization of prey species in other study sites in India were obtained from the published literature. The index wa s computed for all the study sites using the using the formula rppr pr D 2 where, r is the proportion of total kills of a prey species, and p is the proportion of the total abundance of that species. The values of the index range from +1 to -1, indicating maximum preference and maximum a voidance respectively. The relative number of each prey species killed was obtained by dividing the relative biomass by the average weight of the species taken by tiger, leopard and dhole, respectively. The

PAGE 47

47 mean weight of prey killed was calculated as th e sum of the weight of prey species multiplied by the proportional number taken. Dietary Overlap The extent of dietary overlap between all three s pecies pa irs was calculated by Piankas index (Pianka 1973) The program EcoSim vers ion 7.72 (Gotelli & Entsminger 2007) was used on the percent frequency matrix assuming all availab ilities to be equal, as well as on an electivity matrix (Lawlor 1980), which is a matrix of frequenc ies of prey taken weighted by their densities. The calculated index can take valu es from 0 to 1, where 1 stands for identical diets or complete ovelap and 0 indicates completely different diets, or no overlap. The formula used for calculating the overlap of species1 with species2, O12 is n i n ii i iipp pp OO1 22 1 21 12))((1 2 12 where pij is the percentage frequency of species j taken by carnivore speciesi.The index was also calculated on the electivity matrix comprising of electivity eij where Rj is the availability of prey species j. j ij ijR p e The program randomizes the electivity for each co mbination of predator and prey species to generate a null model to compare with the observe d mean index. If the mean overlap index value is at either tail of the distribu tion of simulated values then it ca n be judged to be significantly different than expected by chance. The density of the major prey species was derived from the results of the line transects. Porc upine density was assumed to be similar to that reported in the literature (Sever & Mendelssohn 1991).

PAGE 48

48 Results Density of Potential Prey Species The mean density of groups and individuals of the potential prey spec ies over four years is presented in Table 3-1. Amongst ungulates, chital were the most common, while the Indian muntjac had the lowest density. Ground birds likely to be found in the carnivore diet are represented by the Indian peafowl ( Pavo cristatus ), grey jungle fowl ( Gallus sonneratii ) and red spurfowl ( Galloperdix spadicea ). Their densities were similar. Overall, the common langur was the most abundant prey species. Sample Size Adequacy The results of the scat analyses for various prey species in the diet of the three carnivores show stability (Figure 3-2). For chital, samb ar and langur, about 50 scats provides a stable estimate of the percentage frequency of that prey in the diet. The sample size of scats used in the analysis can therefore be considered adequate for quantifying the major species found in the diet of these carnivores. Composition of Diet Leopard preyed on 10 species, the tiger took 7 species and 4 species were found in dhole scats. It was necessary to analyze about 55 scat s to detect all these species (Figure 3-3). The percentage frequency and relative biomass of the major prey species in the scats of the leopard, tiger and dhole are given in Tables 3-3, 3-4 and 3-5. It can be seen that sambar is the major prey in the diet of all three predators. Chital is ta ken by dhole and to a lesser extent by leopard, but is not an important component of the diet of the tiger in this study. Livestock is also not an important constituent of the diet, especially for leopards and dholes. Rodents, birds, porcupines, and wild pigs also do not figure in the diet of dholes. Porcupine was only taken by leopards, while hare was taken by both leopards and dhol es but not by tigers. Relative biomass and

PAGE 49

49 number were not estimated for the categories of bird and rodent species because of uncertainty about their weights. However, since they are a minor component of the diet, it would have little effect on the results. Tigers take the highest mean weight of prey (129 kg), followed by dholes (46 kg) and leopard (27 kg). The percentage of prey taken by tiger, leopard and dhole in various prey size classes is presented in Figure 3-4. Leopards take prey from each size class, though they take medium-sized prey the most. Tigers and dholes seem to specialize on large and medium-sized prey, respectively. A small percentage of smaller sized prey is taken by both species, but dholes do not take larger prey. Prey Selection Tigers ( 2 = 61.5, d. f = 4, p <0.01), leopards ( 2 = 52.2, d.f. = 4, p<0.01) and dholes ( 2 = 54.3, d.f. = 3, p< 0.01) all exhibited overall sel ectivity in their diet. Figure 3-5 shows the observed and expected frequencies of the major prey species in scats. Tigers significantly preferred sambar ( 2 = 60.9, p<0.01) while avoiding chital ( 2= 16.1, p<0.01), langur ( 2= 10.3, p<0.01) and hare ( 2= 4.1, p=0.04). Wild pig was ne ither preferred nor avoided ( 2= 2.8, p=0.1). Leopards also significantl y preferred sambar ( 2 = 43.4, p<0.01), while avoiding langur ( 2 = 20.6, p<0.01) and wild pig ( 2 = 8.2, p=0.005). Chital ( 2 = 0.98, p=0.4) and hare ( 2 = 0.23, p=0.64) were neither preferred nor avoided. D holes significantly pr eferred both chital ( 2 = 18.4, p<0.01) and sambar ( 2 = 16.7, p<0.01) avoiding hare ( 2 = 7.2, p <0.01) and langur ( 2 = 30.9, p<0.01). Wild pig ( 2 = 2.1, p=0.15) was taken in propor tion to its availability. The preference for major prey species by tiger s, leopards and dholes in various protected areas in India using the Jacobs index is presen ted in Tables 3-5, 3-6 and 3-7. Overall, tigers seem to take chital (mean Jacobs index -0.05, SE 0.19, n=6 sites) and wild pig (mean Jacobs index 0.06, SE 0.27, n= 6 sites) approximately in proportion to their availability, though the

PAGE 50

50 variance on these estimates is high. Tigers prefer sambar (m ean Jacobs index 0.38, SE 0.14, n = 6 sites) and avoid nilgai (mean Jacobs index -0.9, SE 0.08, n = 4 sites), gaur (mean Jacobs index -0.45, SE 0.37, n = 4 sites) and langur (m ean Jacobs index -0.2, SE 0.16, n= 5 sites). Leopards take chital in proportion to their av ailability (mean Jacobs index 0.07, SE 0.1, n = 4 sites), prefer sambar (mean Jacobs index 0.18, SE 0.27, n = 4 sites) and avoid gaur (mean Jacobs index -0.46, SE 0.37, n =3 sites), langur (mean Jacobs i ndex -0.21, SE 0.3, n = 3 sites) and wild pig (mean Jacobs index -0.12, SE 0.45, n= 4 sites). Dholes prefer chital (mean Jacobs index 0.20, SE 0.2, n = 4 sites) and sambar (mean Jacobs index 0.41, SE 0.28, n = 4 sites), and avoid gaur (mean Jacobs index -0.80, SE 0.18, n = 4 sites), wild pig (mean Jacobs index -0.12, SE 0.45, n = 4 sites) and langur (mean J acobs index -0.80, SE 0.13, n = 3 sites). Diet Overlap The diet overlap (Table 3-8) exhibited a si milar pattern when calculated with percent frequency or electivity. The diets of all 3 species overlapped considerab ly. Tiger-leopard and leopard-dhole diets overlapped more extensively than tiger-dhole diets, though this overlap increased when electivity was used to calculate the index. Discussion Surprisingly, sambar is the preferred prey of all three species, and forms a large proportion of the diet of the tiger in this study. Sambar has been found to be a preferred prey of tigers in other studies also (Bagchi et al. 2003b; Biswas & Sankar 2002; Karanth & Sunquist 1995). It is a large sized deer (about 200 kg), found in moderate densities, and is known to choose dense forest areas (Varman & Sukumar 1995). Th is probably makes it more vulnerable to tiger predation, unlike the chital. The minor role of chital in the tigers diet probably has to do with its habitat selection and density. In STR, ch ital are found in open plain areas ne ar villages, where there is a lot of human disturbance. Thei r abundance is also not as high as that found in other national

PAGE 51

51 parks in India. Their habit of congregating near human inhabitati on at night has been speculated to be the reason why they are not found in tiger diet in Bandipur (Johnsingh 1983). In Pench National Park (Biswas & Sankar 2002) in central India and in Nepals Royal Bardia tiger reserve (Stoen & Wegge 1996), chital congregate in large numbers along low-lying areas. They comprise a larger proportion of the tigers diet there, though they are st ill not highly preferred. Wild pig are also taken less than expected, and this may be because of their low densities. In Bardia and Nagarjunasagar (Reddy et al. 2004), wild pig were mo re commonly taken, and they were found to be preferred prey of tigers in Pench (Biswas & Sankar 2002). Common langur is also taken less than expected. In STR, langur is less important in the diet of the tiger than of the leopard with respect to biomass and percentage frequency in scats, thoug h relatively more langur is taken by the tiger than by the leopard. This is because the diet of the leopard is more evenly distributed amongst its prey species than that of the tiger. A similar pattern was seen in the Sariska Tiger Reserve (Sankar & Johnsingh 2002), while only marginal ly more langur were taken by leopards in Nagarhole (Karanth & Sunquist 1995). Although one tiger kill of gaur was seen, gaur, nilgai and muntjac were not found to be a part of the diet of the tiger as measured by scat analysis. This c ould be because of the low density of these species in the study area. The nilgai also prefers distur bed and open areas which are not used by the tiger (Bagchi et al. 2003a). Livestock were also not an important component of the diet, being found in about 5% of scats. This figu re is comparable with some other studies, being about 7% in Srisailam Tiger Reserve (Reddy et al. 2004), and 4.3% in Pench Tiger Reserve. Leopards take chital in proportion to its av ailability, though it comprises about 20 % of its biomass intake. Unlike the tiger, the leopard is also found close to human inhabitation, where chital congregate at night. In STR, its relative lack of importance in the leopards diet may be due

PAGE 52

52 to the larger mean group size of chital (6.3 pe r group, n=469 groups) as co mpared to the sambar (2.2 per group, n=419 groups), which increases vigi lance and helps avoid stalking predators. In studies where chital is a major part of the diet, the chital density is quite high as compared to sambar density (Johnsingh 1983; Karanth & Sunqui st 1995). This is not the case in this study, where densities of the two ungulates are roughly similar. Hayward et al. (2006) reviewed leopard prey across many studies and concluded that preferred prey were likely to be in smaller groups and in denser vegetation than avoided prey. Chital are likely to be in la rger groups and in more open vegetation than sambar, and are probably not selected becaus e of this. In Chitwan National Park it was observed that predation on sambar by tigers increased when chital congregated in large herds on newly burned grasslands (Sunqui st 1981). Perhaps the reason for a lack of preference by leopards is an anti-p redatory strategy of larger he rd formation. Wild pig were not an important component of the leopards diet in Gir National Park (Mukherjee et al. 1994), in Sariska National Park (Sankar & Johnsingh 2002) and in Bandipur (Johnsingh 1992) or Nagarhole (Karanth & Sunquist 1995). In this study wild pigs were avoided, the adults are probably dangerous prey for the leopard which likely only prey upon subadults and young. Hares were taken in proportion to their availability by leopards and by dholes. Along with sambar, chital is a preferred pr ey for the dhole. The herding behavior and congregation by chital is not an effective strategy against a di urnal, coursing predator. Many chases were observed, usually in the morni ng. The anti-predatory strategy of the chital sometimes included running towards the village, where the dhole would not follow (Johnsingh 1983). The diets of the three predators overlap to a great extent. The tig er diet overlaps more with that of the leopard than the dhole because of sh ared inclusion of wild pig, cattle, rodents and

PAGE 53

53 birds. The dhole-leopard overlap is more than the dhole-tiger overlap because the former speciespair hunts in open areas also a nd both thus take a significant am ount of chital, unlike the tiger. Tigers seem to prefer large prey species that are more easily available, the mean size of prey being 129 kg. The leopard and dhole tend to take medium sized prey. The leopard takes a mean prey size of 27 kg, while the pack living dhol e takes larger prey of 46 kg. The leopard also takes the largest range of prey size, taking small prey like hare birds, rodents and porcupines that dhole did not kill in this study.

PAGE 54

54 Table 3-1. Estimation of overall density and its as sociated parameters by the line-transect method over 4 years in the study area. Species n D CV DCI D Ds Cv DsCI Ds Model Chital 189 5.4 13.84.2-7.11.612.41.3-2.1Hazard Polynomial Sambar 262 4.0 10.33.2-4.71.99.71.5-2.2Half-Normal Cosine Nilgai 95 1.6 17.01.2-2.30.814.70.6-1.1Half-Normal Cosine Muntjak 63 0.8 19.00.6-1.20.717.30.5-1.1Half-Normal Cosine Wild pig 63 1.8 26.21.1-2.90.614.50.4-0.7Half-Normal Cosine Black-naped hare 83 3.4 15.62.7-4.73.215.02.6-4.4Half-Normal Cosine Gaur 35 0.8 37.40.4-1.80.233.40.1-0.4Half-Normal Cosine Common langur 637 28.3 10.324.1-36.36.49.55.7-8.3Half-Normal Cosine Indian peafowl 98 2.0 20.01.3-2.91.317.70.9-1.7Neg exp Cosine Red spurfowl 59 2.6 20.41.6-3.51.518.81.0-1.9Half-Normal Cosine Grey jungle fowl 86 2.7 17.11.8-3.81.416.01.0-2.0Uniform Polynomial n: number of observations, D: density of individuals/km2, Ds: Density of groups/km2, CV: coefficient of variation, CI: 95% Confidence. Sample size: 20 transects, effort: 1272 km.

PAGE 55

55 Table 3-2. Food habits of the leopard obtai ned by scat analyses (N =193 scats). Species Weight Of prey Scats Collectable scats per kill % in Scat Bootstrapped CI (95%) Percent Biomass Relative number Sambar 62 102 14.952.846.1-59.662.2 27.1 Chital 48 39 13.120.215.0-25.920.7 11.6 Langur 8 21 3.510.96.7-15.57.0 23.7 Hare 3 11 1.45.72.6-9.33.2 29.2 Wild pig 37 4 11.32.10.5-4.11.8 1.3 Cattle 150 3 20.71.60.0-3.63.1 0.5 Porcupine 8 6 3.53.10.0-3.61.9 6.5 Rodents 0.1 6 0.053.11.0-3.6Bird spp 5 7 2.33.61.0-6.2-

PAGE 56

56 Table 3-3. Food habits of the tiger obta ined by scat analyses (N = 93 scats). Species Weight Of Prey Scats Collectable scats per kill % in Scat Bootstrapped CI (95%) Percent Biomass Relative number Sambar 212 73 22.578.569.9-86.089.6 54.8 Chital 55 4 14.14.31.1-8.62.0 4.8 Langur 8 7 3.57.52.2-12.92.1 33.6 Hare 3 0 1.40 -0 0 Wild pig 38 2 11.52.20.0-5.40.9 2.9 Cattle 180 5 21.75.31.1-10.85.4 3.9 Porcupine 8 0 3.50-0 0 Rodents 0.1 2 0.052.00.0-5.4Bird spp 5 2 2.32.00.0-5.4Table 3-4. Food habits of the dhole obtain ed by scat analyses (N = 81 scats). Species Weight Of Prey Scats Collectable scats per kill % in scat Bootstrapped CI (95%) Percent Biomass Relative number Sambar 70 39 39.348.137.0-59.356.0 36.8 Chital 55 34 37.241.931.5-51.940.7 34.1 Langur 8 5 14.86.21.2-12.32.2 12.6 Hare 3 3 6.83.70.00-8.61.1 16.4 Wild pig 38 3 33.30-0 0

PAGE 57

57 Table 3-5. Jacobs index values of preference for prey species in tiger diets at study sites in India. Place Chital Sambar Nilgai Wild pig Gaur Langur Bandipur1 -0.30 0.07N.P0.77-0.06N.A. Nagarhole2 -0.45 0.65N.P.0.680.26-0.36 Pench3 0.11 0.50-10.32-1-0.35 Ranthambore4 0.32 0.19-0.71-0.49N.P-0.06 Sariska5 0.54 0.07-0.96-0.71N.P0.30 STR6 -0.51 0.81-1-0.18-1-0.54 1Andheria et al.(2007), 2Karanth and Sunquist (1995), 3Biswas and Sankar(2002), 4Bagchi et al. (2003b), 5Sankar and Johnsingh (2002), 6this study. NP= not present, NA = not estimated. Table 3-6. Jacobs index values of preference for prey species in leopard diets at study sites in India. Place Chital Sambar Wild pig Gaur Langur Bandipur1 0.07 -0.430.76-0.42NA Nagarhole2 -0.02 0.620.270.050.10 Sariska3 0.31 0.05-1NP-0.04 STR4 -0.08 0.50-0.51-1-0.69 1Andheria et al.(2007), 2Karanth and Sunquist (1995), 3Sankar and Johnsingh (2002), 4this study. NP= not present, NA = not estimated.

PAGE 58

58 Table 3-7. Jacobs index values of preference fo r prey species in dhole diets at study sites in India. Place Chital Sambar Wild pig Gaur Langur Bandipur1 0.46 -0.280.64-0.95NA Nagarhole2 -0.10 0.460.42-0.94-0.96 Pench3 -0.08 0.81-0.56-0.33-0.59 STR4 0.53 0.65-1-1-0.84 1Andheria et al.(2007), 2Karanth and Sunquist (1995), 3Biswas and Sankar(2002), 4this study. NA = not estimated. Table 3-8. Diet overlap between tiger, leopard and dhole using Piankas index. Species Dhole Frequency/electivity Leopard Frequency/electivity Tiger 0.79/0.88 0.94/0.96 Dhole 0.93/0.95

PAGE 59

59 Figure 3-1. Map of Bori Wildlife Sanctuary and Satpura National Park, showing the location of line transects, dirt roads and the study area.

PAGE 60

60 A. Number of scats 20406080100120140160180 Estimated Langur (%) in diet 0 2 4 6 8 10 Tiger Leopard Dhole B. Number of scats 20406080100120140160180 Estimated Chital (%) in diet 0 10 20 30 40 50 Tiger Leopard Dhole C. Number of scats 20406080100120140160180 Estimated Sambar (%) in diet 40 45 50 55 60 65 70 75 80 Tiger Leopard Dhole Figure 3-2. Relationship between sample size of s cats and the percent frequency of occurrence in tiger, leopard and dhole diet of A) Langur, B) Chital and C) Sambar.

PAGE 61

61 Number of scats 0 50 100 150 200 Cumulative number of species 0 2 4 6 8 10 12 Dhole (N = 81) Tiger (N = 93) Leopard (N = 193) Figure 3-3. Relationship between the number of scats analyzed and the number of prey species found in the diet of tiger, leopard and dhole.

PAGE 62

62 Species Tiger (N = 93)Leopard (N= 193)Dhole (N= 81) Prey (%) in scats 0 20 40 60 80 100 <10 kg 10-50 kg 51-100 kg >100 kg Figure 3-4. Prey taken by tiger, leopard and dhole in various body weight categories.

PAGE 63

63 C. Prey Species SambarChitalLangurHareWild pig Frequency in dhole scats 0 10 20 30 40 50 Observed Expected B. Prey Species SambarChitalLangurHareWild pig Frequency in leopard scats 0 20 40 60 80 100 Observed Expected A. Prey Species SambarChitalLangurHareWild pig Frequency in tiger scats 0 10 20 30 40 50 60 70 Observed Expected Figure 3-5. Observed and expected frequencies of prey items in scats of A) Tiger, B) Leopard and C) Dhole.

PAGE 64

64 CHAPTER 4 ESTIMATION OF LEOPARD ( Panthera pardus) ABUNDANCE IN I NDIAN FORESTS USING CAMERA TRAPS IN A MARK-RECAPTURE FRAMEWORK. Introduction While there has been increased attention to the need for reliable estimates of carnivore densities in India, the work has been la rgely restricted to tigers, Panthera tigris (Harihar 2005; Karanth et al. 2004a; Karanth & Nichols 199 8). Even basic information on other large felids is poor, except for food habits. Leopards have been in the popular media in India largely because of an increas e in human conflicts. There is a perception that attacks on humans have es calated in recent years, which has been attributed to various causes, including decrease in hab itat, decline in leopard prey populations, increase in leopard densities and effects of translocations near populated areas (Athreya et al. 2007). Unfortunately, data on leopard or prey abundances in any of the conflict areas are lacking, and therefore th e causes remain speculative. Estimation of leopard density is, however logistically feasible even though leopards tend to be nocturnal, inhabit dens e cover and occupy large ranges. Camera trapping has been used in conjunction with mark-recapture techniques to estimate the population size of species in which individua ls can be uniquely identified based on the coat patterns or other external marks. The primary method of censusing tiger, leopard and lions ( Panthera leo ) by the government agency in Indi a has been the pugmark method (Panwar 1979). This involves taking plaster ca sts or paper traces of the tracks of the targeted carnivore species in the entire survey area. The assumption is made that the tracks of all individuals are reco rded and that all individuals can be identified on the basis of the tracings of thei r tracks. The method has been critic ized for its subjective nature and the lack of incorporation of a correction for detectability (Karanth et al. 2003).

PAGE 65

65 The use of statistically robust indices to monitor populaton trends have been suggested, like track indices (Karanth et al. 2003 ), camera trapping rates (Carbone et al. 2001; Karanth & Nichols 2002) or occupancy models (MacKenzie & Nichols 2004), but these methods do not provide an estimate of the number of individuals in the protected area. The mark-recapture method has long been used to estimate biological populations (Otis et al. 1978). Recently the method has been adapted to estimate tiger populations in India using remote camera traps. There are now estimates for tigers (Johnson et al. 2006; Karanth & Nichols 1998; O'Brien et al. 2003 ), leopards (Spalton et al. 2006), jaguars ( Panthera onca ) (Silver et al. 2004; Soisalo & Cavalcanti 2006) and snow leopards ( Panthera uncia ) (Jackson et al. 2006) using mark recapture for other parts of the world and it is now the accepted method. In India there are few published studies on population estimation for carnivores other than the tiger. It is expected that more studies of leopard abundances will soon be available for this part of the world. Methods Study Area Leopard densities were estimated at thr ee adjacent sites in Satpura Tiger Reserve and one site in Sariska Tiger Rese rve. The Satpura Tiger Reserve (22 o 19' to 22 o 30' N and 77 o 56' to 78 o 20' E) covers 1428 km2, and is located in the Hoshangabad district of the central Indian state of Madhya Pradesh. It consists of three administrative units, the Pachmarhi and Bori Wildlife Sanctuaries, and Sa tpura National Park. The forest is mainly the moist deciduous type (Champion & Seth 1968). The major ungulate fauna includes chital ( Axis axis), sambar ( Cervus unicolor ), Indian muntjac ( Muntiacus muntjac ) and gaur ( Bos gaurus). The major carnivores are ti ger, leopard, sloth bear ( Melursus ursinus)

PAGE 66

66 and dhole ( Cuon alpinus). The Sariska Tiger Reserve (25 o 05' to 25 o 27' N and 74o17' to 76 o 74' E) covers 800 km2 and located in the north-western state of Rajasthan, in the Alwar district in India. The forest is mainly the tropica l dry deciduous and thorn type (Champion & Seth 1968). The major ungulat es are chital, sambar and nilgai ( Bosephalus tragocamelus ). Gaur and muntjac do not occur th ere. Major carnivores are leopard, striped hyena ( Hyaena hyaena ), jungle cat ( Felis chaus) and golden jackal ( Canis aureus). Sloth bears and dhole are absent while the tiger has recently gone locally extinct due to illegal hunting. Field Methods Four sites were chosen for estimati on of leopard abundances. Camera-trapping effort at these sites ranged from 33 days ( 396 trap nights) to 76 da ys (1216 trap nights). Three of the sites (Churna, Kamti and Lagda) were adjacent to each other in the Satpura Tiger Reserve in central India (Figure 4-3) wh ile the fourth was in Sariska Tiger Reserve (Figure 4-4). Camera trap lo cations were chosen after re connaissance to maximize the probability of getting photos of leopards. Locati ons close to villages or on routes where there was a great deal of hu man movement were excluded to minimize the possibility of theft. Trailmaster 1550 (Goodson Associates Lenexa, Kansas) camera traps with Olympus and Canon autofocus cameras were depl oyed at all sites. At two sites (Churna and Sariska), a one-camera setup was used at most stations, and a two-camera setup was used at a few stations. These twocamera setup locations were changed when both flanks of individuals in that area were obtained. At the other two sites each camera location had a two-camera setup to photograph both flanks at the same time. Camera traps were activated at dusk and deactiv ated at dawn. The minimum interval between two photos was 6 seconds. Camera sensors were placed at a height that a llowed photographs of

PAGE 67

67 smaller species like black-naped hare ( Lepus nigricollis ) and grey jungle fowl ( Gallus sonneratii ). Analytical Methods All photos were scanned, printed (Fig. 2-1) and the flank of each leopard photographed was compared to every other le opard photo. Printouts of the photos were scrutinized under a magnifying glass to identif y patterns of similar looking spots. Photos that were underexposed due to the leopard be ing farther away from the camera, or where the coat patterns were distorted because the individual was not approx imately parallel to the camera, were difficult to identify. Diffi cult photos were enlarged and matched on the computer after some image processing to e nhance contrast and bri ghtness. If a pattern was detected then a separate area of the fla nk was checked to confir m the identity of the leopard. Leopards whose identities could not be confirmed were di scarded and not used in the analysis. Sometimes photos of both fla nks were available, usually in cases where two cameras were used. In one case a clear photograph of the face was available to link the two flanks. In these cases the identity of the leopard was unambiguous, and the leopard was included in analys es of both flanks. The number of individuals obtained from the right flank and the left flank were comp ared and the dataset w ith the greater number of individuals was used for the analysis. Estimation of population size For all sites, capture histories were developed using each day as the sampling occasion. The capture history for each individu al leopard consisted of a row vector of t entries where t is the number of trapping occasions for each site. Each entry takes a value of either 1 or 0 depending on whether the individual leopard wa s photographed on that particular occasion or not. The entire matrix of observations for all the leopards, called

PAGE 68

68 the X matrix (Otis et al. 1978) wa s used to estimate the population, N, and its standard error. Program CAPTURE2 (Hines 1994) was used for the estimation. CAPTURE2 estimates the population parameters under va rious assumptions of the sources of variability in capture probabilities. These are: none (Mo), individual heterogeneity (Mh), behavioral heterogeneity (Mb) and time (Mt). The null model, Mo, corresponds to the case which assumes that the capture probability ac ross all individuals is the same. Model Mh assumes that each individual has its own capture probability, and this differs from that of all other individuals. Model Mb assumes that the capture probability varies after the individual is caught for the first time, and becomes either trap shy or trap happy. Model Mt refers to change in capture probability from one occasion to another. Models Mbh, Mth, Mtb and Mtbh, assume that variation in capture pr obability is explained by a combination of these sources of variation. Goodnessof-fit tests and te sts of models Mo vs Mh, Mo vs, Mb, Mo vs Mt was calculated using program CA PTURE2 where enough data was available. A model selection procedure which scores the models according to appropriateness using a discriminant functi on criterion was used (Otis et al. 1978; Rexstad & Burnham 1991). Model Mo, the simplest model, is se nsitive to violations of the assumption of similar individual captur e probabilities, so when this model was selected, the parameters computed using the next best model have also been presented. The test for population closure computed by program CAPTURE2 was used to detect violation of this assumption. Also, in Chur na where trapping was conducted for 150 days, two estimates were obtained for 75 days each to enable the closure assumption to be maintained within these two shorter sessions.

PAGE 69

69Estimation of leopard density The Effective Trapping Area (ETA) method: Density,D, is defined as N/A, where Nis the estimated number of leopards and A is the estimated area in which the sampling was conducted. This area is typi cally the area encompassed by the trapping grid, plus a strip of buffe r around it (Dice 1938), to obtai n an ETA (Figure 4-5). The buffering was done using both concave a nd convex polygons. Boundary width was calculated using the mean maximum dist ance moved (full-MMDM), and half-MMDM (Parmenter et al. 2003), to get a total of four ETAs (concave-MMDM, convex-MMDM, concave-half MMDM, convex-half MMDM). MMDM and its standard error were approximated by the mean of the maximum distance between two photos of each individual leopard for all leopards photographe d at more than one camera trap location. Any portion of the ETA that lay outside the boundaries of the Tiger Reserve was subtracted using a GIS package. A relatively small area (26 km2) was sampled in Sariska, and so the data from this site we re not used in the MMDM estimation. The Spatially Explicit Maximum Like lihood method: Efford (2004) estimated D directly from trapping data by a simulation of the trapping process. This removes the need for a buffer width around the trapping ar ea. The process uses the location of each trap and includes a sub-model for the distri bution of individuals and another sub-model for the capture process. The distribution of individuals is modeled by a homogeneous Poisson process. The capture process models the probability of capturing an individual in a particular trap given th e location of its unknown home-range center. The capture probability is modeled using the spatial analog of the detection function (Buckland 2001). The half-normal, hazard rate and negative e xponential detection func tions can be used.

PAGE 70

70 These functions use the independent parameters g(0) for overall effi ciency of detection and for spatial scale. Incorporation of sour ces of heterogeneity (individual-based, timebased and behavior-based) is possible in th ese parameters, as in conventional capturerecapture. However these increase the number of parameters that need to be estimated. Because only a few animals were detected, only the null models for both parameters were used, denoted as g(0)[.] [.], the dot denoting lack of heterogeneity. The method assumes that 1) Trap placement is random with respect to location of home ranges, and home ranges are randomly oriented. 2) Home range s do not change for the duration of the trapping and the population is demographically closed. 3) Home-range centers have a Poisson distribution, and 4) I ndividuals are independently de tected. (Efford et al. in review) provides details of the method. The software Density 4.1 (Efford 2007) was used to calculate the densities and associated variances using all three detection functions. The Akaike Information Criterion (AIC) was used to select between the models, the model with the lowest AIC being selected. Results A total of 288 leopard photos were obtain ed, twenty were unidentifiable and were removed from the analysis. Of the identifiable photos 141 were of the left flank and 127 were of the right flank. Sampling intensity vari ed between sites, bei ng lowest in Sariska, and highest in Churna (Table 4-1). Adequacy of Sampling A measure of the adequacy of sampling is if new individuals are no longer photographed with additional sampling. Figure 4-1 shows the addition of new leopards for the 4 sites using the left flank. The shape of the curves and the number of individuals identified were similar for the right and the left flanks. An asymptote was reached for the

PAGE 71

71 sites in Satpura Tiger Reserve by 6 weeks. No asymptote was reached in Sariska suggesting that further sampling would have yielded photographs of additional new individuals. Sex Ratios The sex ratios are female biased in all ar eas except Kamti. Th e average ratio is 1.7 (SE 0.38) females per male (Table 4-2). Population Size The model selection criterion chose Mo for 2 sites and Mh for 3 sites. When Mo was chosen the Mh model selection value was not much lower, though the difference was significant or marginally nonsignificant (Table 4-3). Mo is not recommended because it is sensitive to departures from the assumption of no individual-based heterogeneity (Karanth & Nichols 1998), though both models have been presented. All Mh were estimated with the jackknife estimator, wh ich is robust and has performed well in simulation studies (Burnham & Overton 1979). Test for population closure was not significant for all the sites, indicating that the assumption of demogr aphic closure was not violated. A high proportion of the estimat ed population was photographed, ranging from 69 to 89 percent for the Mh model and 69 to 100 percent for the Mo model. Population sizes, capture probability and estimated pr oportion photographed for both estimators are given in Table 4-4. Leopard Density Table 4-5 gives the estimation of density of leopards per 100 km2 at the different sites using the convex polygon to calculate the Effective Trapping Area method with fullMMDM and half-MMDM. Estimated density is dependant on the method used to calculate the strip width and the polygon. The densities calcul ated using all combinations

PAGE 72

72 of concave and convex polygon with half MMDM and full MMDM are presented in Table 4-6. Concave polygon with half MMDM ga ve the smallest effective trapping area and consequently the highest density, while the convex polygon w ith MMDM gave the largest effective trapping area and theref ore the lowest density at each site. Using the maximum-likelihood-spatially-e xplicit-capture-recapture method, the lowest AIC values for Churna and Kamti were obtained by the four parameter hazard rate model, while for Lagda and Sariska the thr ee parameter half normal model was selected. Densities obtained by this method are given in Table 4-7. The relative abundance index (Table 4-8) was also highest for Sariska followed by the second session at Churna. Discussion Ideally, it is desirable to obtain photos of both fl anks of the body so that identification of individuals is unambiguous When camera numbers are limited, it seems possible to obtain unambiguous photographs of both flanks for a large proportion of the population using two cameras at a few locations while using one camera at the remaining locations, provided the trapping goes on for a long period. This would maximize coverage of the area with the available num ber of cameras. The individual identification of leopards from photographs was found to be qu ite easy except when the animals walked farther away from the cameras, resulting in underexposed photos. This was likely to happen when the distance between the two sensors was more than 10 meters. Tigers were sometimes observed to avoid cam era traps, leaving the trail just before the camera location and getting back on the trail afterwards. Other studies have also observed this behaviour (Wegge et al. 2004). On the evidence of tracks, leopards were never observed to avoid cameras traps, and showed no response to the flash. Leopards of

PAGE 73

73 both sexes were photographed while standing or si tting in front of the camera, and did not rapidly move away. Sometimes more than one photograph was taken at the same time, indicating that the leopard stayed in that position for at least 6 seconds after the flash of the first photograph. However, rates of photo-captures for males seemed to be consistently higher than for females. Th e existence of heterogeneity in capture probabilities with respect to gender is possibl e. In Kruger National Pa rk for instance, it was easier to capture males as opposed to females in box traps (Bailey 1993). The calculation of effective area of sampli ng is a noteworthy issue in the estimation of density using camera traps. There is gene rally no measurement of the home range of the sample of individuals used in the estim ation. It has been recommended that half the mean maximum distance moved (MMDM) be used as the buffer for estimation of densities (Wilson & Anderson 1985). A recent study on jaguars comparing MMDM obtained by telemetry to half MMDM and full MMDM found that the full MMDM results were much closer to densities base d on actual movement rates, and that half MMDM seemed to overestimate densities (Soisalo & Cavalcanti 2006). There is still not enough data available for movement in leopard s to advocate a shift to full MMDM. Also, as the area of the buffer increases, it is more likely to include habitat that is unsuitable for the species and unrepresentative of the probabil ity of capture at the camera trap location. In this study densities obtained usi ng the convex polygon-full MMDM gave results that were similar to the MLSECR method at mo st sites, while densities calculated using half MMDM were much larger (Table 4-8) The density of leopards was highest in Sariska Tiger Reserve, where tigers have been extirpated recently (Sankar et al. 2005), while it was lowest at Lagda, which had th e highest activity of tigers amongst all sites

PAGE 74

74 (pers obs) though it is not a high density tiger area. Variation in density of carnivores is associated with density of prey as show n for tigers (Karanth et al. 2004b), and other carnivores (Carbone et al. 1999). However, other factors, like human disturbance (Woodroffe 2000) and tiger presence (Seide nsticker 1976; Sunquist 1981) may also play a role, although there is some evidence that leopard densities may not be unduly depressed by presence of other larg e carnivores (Marker & Dickman 2005). Leopard density estimates are available fo r various parts of the world, but from different methods. For the Serenge ti it was about 3.8-4.5 per 100 km2 (Schaller 1976), for Kruger it was about 3.4 per 100 km2 (Pienaar 1969). It was estimated at about 7.1 in the rain forest of the Ivory Coast (Jenny 1996), and in Wilpattu National Park in Sri Lanka it was estimated as about 3.4 (Eisenberg & Loc khart 1972). In Namibia, a mean of 10.5 (SE 4.0) inside protected areas (n =6), and 2.1 (SE 1.6) outside protected areas has been reported (Marker & Dickman 2005). Photocapture rates calculated per 100 trap nights for 4 sites in India ranged from a low of 0.18 for Kaziranga National Park, a me dium 2.3 for Pench National Park to a high of 5.44 in Nagarhole National Park (Karanth & Nichols 1998). Estimates of Relative Abundance Index (RAI) for the present study, ranging from 2.2 to 6.8 (Table 4-7) seem to be within the range found in other areas in India. The second session in Churna, conducted in spring-summer, had higher capture probability than the first session, conducted in winter-spring. Camera traps were mostly placed along topographic contours, where leopa rd signs were high, and water tended to be found. It is possible that movement of leopards around such places increased in

PAGE 75

75 summer when water sources in the hills dried up, leading to the higher capture probability. RAI has been recommended for tigers when there is not enough data for mark recapture sampling (Carbone et al. 2001). In low leopard density areas it takes a long time to get a sufficient number of captures to use in the mark-recapture framework. If the assumption of population closure is severely violated, then the RAI may be used as a substitute for density estimation. An index can also be used on species that do not have individually identifiable markings. Howeve r, the difference between the RAI estimates and density estimates is noteworthy. The s econd session at Churna and the session at Sariska have similar RAI values, but the densit y at Sariska is much higher. Similarly, the RAI of the first session at C hurna is almost 3 times lower than the RAI of the second session, but density estimates ar e not significantly different. Th is indicates that RAI does not seem to index density in a reliable way, as noted elsewhere also (Jennelle et al. 2002; Maffei et al. 2004). Conclusion Sariska Tiger Reserve has the highest dens ities despite having a history of human disturbance and poaching. This may be relate d to the recent removal of the tiger from Sariska and the occupation of prime habitats by the leopard. Another reason could be the smaller spatial extent of the effective trapping area in Sariska. It may be that the distribution of individuals in Sariska is more patchy and that the lower density areas were not surveyed. In such a scenario, compari ng Sariska to another study site will not be useful and the parameter values should be limited to monitoring the same site over time. Leopard RAI values in Bori-Satpura are comp arable to other study sites in India. The

PAGE 76

76 leopard RAI in Satpura, where tiger density is relatively low, seem to be higher than at Kanha and Pench, which have higher tiger densities. The estimates provided by the mark-recapture framework give us a relatively robust measure of population size, but the estima tion of density is still problematic given the uncertainty involved with estimation of the effective trapping area. The spatially explicit maximum likelihood method offers a solution to that problem, but modeling heterogeneity is more complicated with low population sizes, since the number of parameters to be estimated is high. The prec ision derived in the present study makes it difficult to detect changes in population density. It is logistically difficult to both sample at an intensity that obtains high precision a nd at a large enough spat ial scale for a species of this size. Given these limitations, serious i nvestigation should be made into the use of indices to monitor population changes with grea ter precision, though RAI does not seem to be the appropriate in dex in these study sites.

PAGE 77

77 Table 4-1. Camera-trapping effort (in trap nights) at the study sites. Site Number of camera trapping stations Number of Nights Effort ( trap nights) Churna (session 1) 16761216 Churna (session 2) 16751200 Kamti 20521040 Lagda 2033660 Sariska 1233396 Table 4-2. Leopard sex ratios fo r the different study sites. Site Males Females Sex ratio (no of females per male) Churna (session1) 4 6 1.5 Churna (session2) 3 8 2.7 Kamti 7 4 0.6 Lagda 3 5 1.7 Sariska 3 6 2.0

PAGE 78

78 Table 4-3. Model selection crit erion and tests for Models Mo, Mh, Mb and Mt in the mark-recapture framework and a test for population closure for the different study sites. Model selection criterion Mo vs Mh Mo vs Mt Mo vs Mb Mh Goodness of fit Closure test Site Mo Mh Mb Mt 2 dfp 2 df p 2 dfp 2 df p z p Churna session 1 1.0 0.94 0.51 0.0 Not done2.5761.001.210.26 85.4760.22-0.60.28 Churna session 2 0.93 1.00 0.46 0.0 5.820.0514.5721.00.0110.91 106.5720.00-0.130.45 Kamti 0.96 1.00 0.44 0.0 3.610.063.9821.00.0310.86 97.4820.11-1.330.09 Lagda 0.93 1.00 0.38 0.0 Not done8.0320.990.0010.97 59.03320.00-1.360.09 Sariska 1.00 0.91 0.42 0.0 Not done2.2281.0Test failed 32.91280.24-1.240.11

PAGE 79

79 Table 4-4. Population estimates for leopards at the study sites. Estimate (Mo) Estimate (Mh) Site p N Mt 1 N SE p N Mt 1N SE Churna session1 0.03 0.92 12.50.020.7914.6 Churna session2 0.08 1.0 11.140.070.7914.6 Kamti 0.03 1.0 10.810.030.8312.7 Lagda 0.08 1.0 8.760.070.899.9 Sariska 0.04 0.69 13.80.040.6913.4 Table 4-5. Density of leopards and estimates of sampled area using convex polygon and model Mh at the different study sites. Sites Estimates Churna (session1) Churna (session2) Kamti Lagda Sariska Effective area (half MMDM) km2. Density (per 100 km2) 152.2 8.0.5 149.2 9.3.0 119.3 7.5.8 122.7 7.3.1 44.4 30.9.1 Effective area (full MMDM) km2. Density (per 100 km2) 230.8 5.3.7 223.6 6.2.6 195.0 4.6.0 210.9 4.2.1 66.2 20.7.0

PAGE 80

80 Table 4-6. Density of leopards with the asso ciated estimated trapping area using models Mo and Mh. DensitySE (per 100 km2 ) Site Polygon method Strip method ETA (km2) Mo Mh Concave MMDM 185.66.4.06.6.5 Concave MMDM/2 77.915.4.315.7.8 Convex MMDM 230.85.2.65.3.7 Churna session1 Convex MMDM/2 152.27.8.78.0.5 Concave MMDM 176.86.2.17.9.0 Concave MMDM/2 73.015.0.619.1.1 Convex MMDM 223.64.9.86.2.6 Churna session2 Convex MMDM/2 149.27.3.89.3.0 Concave MMDM 179.54.5.35.0.1 Concave MMDM/2 92.38.7.89.7.6 Convex MMDM 195.04.1.24.6.0 Kamti Convex MMDM/2 119.36.7.47.5.8 Concave MMDM 194.84.1.44.6.4 Concave MMDM/2 97.78.2.89.1.4 Convex MMDM 210.93.8.74.2.1 Lagda Convex MMDM/2 122.76.5.47.3.1 Concave MMDM 54.623.825.2.1 Concave MMDM/2 21.161.665.1.5 Convex MMDM 66.219.6.120.7.0 Sariska Convex MMDM/2 44.429.2.830.9.1

PAGE 81

81 Table 4-7. Density estimate s for leopards (number/100 km2) using different capture functions for the null models with the MLSECR method. Site Capture function Model No of param Log likelihood AIC AIC Density SE Churna session1 Hazard g0[.] [.] 4-172.29352.590 7.21 3.21 Churna session1 Negative exponential g0[.] [.] 3-173.52353.040.45 6.62 2.73 Churna session1 Half normal g0[.] [.] 3-175.03356.073.48 5.92 2.3 Churna session2 Hazard g0[.] [.] 4-364.57737.150 4.04 1.37 Churna session2 Half normal g0[.] [.] 3-369.75745.498.34 3.83 1.25 Churna session2 Negative exponential g0[.] [.] 3-372.9751.814.65 3.51 1.14 Kamti Hazard g0[.] [.] 4-179.45366.90 4.67 0.05 Kamti Half normal g0[.] [.] 3-183.64373.286.38 4.15 1.61 Kamti Negative exponential g0[.] [.] 3-183.81373.626.72 4.08 1.58 Lagda Half normal g0[.] [.] 3-149.7305.40 3.27 1.41 Lagda Negative exponential g0[.] [.] 3-149.92305.830.43 3.11 1.23 Lagda Hazard g0[.] [.] 4-149.09306.180.78 3.44 0.19 Sariska Half normal g0[.] [.] 3-73.75153.490 14.58 7.0 Sariska Hazard g0[.] [.] 4-73.15154.30.81 20.08 0.4 Sariska Negative exponential g0[.] [.] 3-74.17154.340.85 12.65 6.57

PAGE 82

82 Table 4-8. Relative abundance index values fo r the 5 estimates in Satpura and Sariska Tiger Reserves. Site No of camera trap locations No of independent captures RAI (per 100 trap nights) SE of RAI Churna (session 1) 16 272.2 0.65 Churna (session 2) 16 806.7 1.85 Kamti 20 393.9 0.71 Lagda 20 243.8 0.89 Sariska 12 276.8 2.21

PAGE 83

83 Figure 4-1. Identification of leopa rds based on spot patterns. The fi rst two photos are of the same leopard, the third photo is of a different leopard.

PAGE 84

84 Figure 4-2. Rate of accumulation of new individuals in camera-trap photographs with increase in sampling time at the four sites.

PAGE 85

85 Figure 4-3. Camera trapping in 3 sites (Churna, Kamti and Lagda) in Satpura Tiger Reserve.

PAGE 86

86 Figure 4-4. Map showing camera trap locations with half MMDM and full MMDM buffers in Sariska Tiger Reserve.

PAGE 87

87 Figure 4-5. Map showing camera trap locations with half MMDM and full MMDM buffers for one site (Kamti).

PAGE 88

88 CHAPTER 5 PRESENCE-ONLY HABITAT SUITABILITY MODELS FOR LEOPARDS ( Panthera pardus ) USING FIE LD BASED AND REMOTELY DERI VED VARIABLES AT TWO SPATIAL SCALES IN MADHYA PRADESH, INDIA. Introduction Knowledge of the distribution and habitat requirements of a species are essential to formulate conservation strategies. While some species are considered habitat generalists, they are still vulnerable to habitat loss and fragmentation. These factors along with prey depletion and poaching are responsible for the decline of the tiger ( Panthera tigris ) across its geographic distribution (Sunquist et al 1999). It has been estimated that th e tiger exists in only 7 percent of its historical range (Dinerstein et al. 2007). The leopard is a wide -ranging large carnivore that is less susceptible to disturbance, is a generalist with respect to habitat requirements, and can survive on a wide range of prey species (Sunqui st & Sunquist 2002). Unlike the tiger, which needs a high biomass of large-sized prey (K aranth & Sunquist 1995), the leopard has been known to survive on domestic dogs and rodents in the absence of w ild prey populations (Edgaonkar & Chellam 2002). As tiger populations in India have declined, leopard populations have also come under increased poaching pressure Conserving leopards in this environment will require a quantification of habitat requirements and identificati on of potential habitat availability in India. Good habitats for leopa rds can then be given conserva tion priority in protection and management strategies. Categorizing suitable leopard habitat requires information at multiple scales. First-order selection (Johnson 1980) refers to the distribution of a species with respect to geographical space. Large-scale species distri bution models can be used to guide conservation strategies (Guisan et al. 2006; Hirzel et al. 2004; Mladenoff & Sickley 1998; Seoane et al. 2006). Techniques like logistic regressi on (Karlsson et al. 2007; Woolf et al. 2002) and generalized

PAGE 89

89 linear models (GLM) (Austin 2007 ; Bustamante & Seoane 2004) use the information from multivariate measurements of habitat variables at locations with species presence and at locations where the species is absent (Guisan & Zimmermann 2000; Meynard & Quinn 2007). This information is used to derive a probability of species presence at each location. Though these methods are preferred when absence data are reli able (Brotons et al. 2004), logistic regression models are known to be sensitive to even lo w levels of non-detections (Gu & Swihart 2004). Leopards are not only rare and secretive, they are also crepuscular (S unquist & Sunquist 2002) and without intensive effort th ere is a high likelihood of non-dete ctions in areas where leopards are present, contaminating the absence data. Presen ce-only models are a way of dealing with this problem. This paper uses environmental niche fa ctor analysis (ENFA) (Hirzel et al. 2001), a presence-only environmental habita t-envelope based method to create habitat suitability maps for the leopard in south-central India. ENFA has been used successfully to model the distributions and habitat suitability of a variety of taxa: dung beetles (Chefaoui et al. 2005), corals (Bryan & Metaxas 2007), reptiles (Santos et al. 2006), bi rds (Braunisch & Suchant 2007; Brotons et al. 2004; Olivier & Wotherspoon 2006; Reutter et al. 2003; Titeux et al. 2007), ungulates (Dettki et al. 2003; Traill & Bigalke 2007) an d carnivores (Mestre et al. 2007). The objectives of this paper are: 1) To deve lop predictive habitat suitability maps for leopards at two scales and evaluate their reliab ility; 2) To identify the environmental variables important in describing the habitat for this specie s, and 3) To quantify th e extent and location of potential leopard habitat availa ble for conservation action in south-central Madhya Pradesh. Study Areas The extensive study area covers 52971 km2 (Table 5-1) and include s thirteen districts in south-central Madhya Pradesh; it comprises about 18 percent of the state of Madhya Pradesh. Altitudes in the study area range from 215 to 1312 m. Annual rainfall for the state averages 1143

PAGE 90

90 mm, with rainfall decreasing from the eastern part of the state to the west. The landscape is a mosaic of forests, agriculture, villages and sm all and large towns (Figure 5-1). The main crops are wheat, soybean, sorghum, sugarcane and pulses. The forests are mainly teak dominated, as well as dry and moist deciduous forests. The climat e is cool in winter a nd very hot in summer, with temperatures ranging from 2-45 C. The largest river in re gion is the Narmada. The two main protected areas within the landscape are th e Satpura and the Pench Tiger Reserves (Figure 5-2). The intensive study site (Figur e 5-3) consists of a 433 km2 area of moist and dry deciduous forests along with some teak plantations located inside the Satpura Tiger Reserve (STR). It is located in the center of the ex tensive study site. Topography ranges from relatively flat to very steep slopes and cliffs; altitudes range from 300 to 1315 m. There are 7 small forest villages within its boundaries. Details of the intens ive study are given in previous chapters. Methods In the STR study site, visual sightings of pr ey species were obtained from walking 20 straight-line transects (630 km) through the fore st, and from driving-transects using a 4-wheel drive vehicle at 10-15 kmph along a network of dirt trails (369.5 km). An encounter rate was calculated as the number of sightings per kilomete r using all sightings of potential prey species. Potential prey species include chital ( Axis axis ), sambar ( Cervus unicolor ), langur ( Semnopithecus entellus ), wild pig ( Sus scrofa ), and a small-prey category comprising hare ( Lepus nigricollis ), peafowl ( Pavo cristatus ), red spurfowl ( Galloperdix spadicea ) and grey jungle fowl ( Gallus sonneratii ). This encounter rate was then di vided into 5 categories. The first category was 0 encounter rate, and the other 4 were based on equal quantiles (25th, 50th, 75th and 100th). Photos of prey from the camera-trap stations were converted into a rate per trap night, and

PAGE 91

91 also similarly divided into 5 in creasing categories based on equal quantiles. The two data sources (encounter rates and photo trap rates) were then assumed to be equiva lent indices of prey abundance and were subsequently merged. Distan ce-weighted interpolation of the categorical index was then done using the INTERPOL module of Idrisi Kilimanjaro v14.02 (Eastman 2004) to obtain a prey map. Sampling for evidence of leopard presence wa s done using kills, tracks, scrapes and camera-trap photos. All quantitative measures were degraded into presence-absence measures to reduce biases introduced by different sampling effo rts. Multiple instances of presence within a one-hectare plot were combined to reduce spatial autocorrelation, which can lead to bias in precision estimates for habitat models (Diniz et al. 2003). Secondary data were obtained as part of a jo int Wildlife Institute of IndiaProject Tiger initiative to monitor tiger popula tions in India in 2006. A total of 2582 beats were sampled. A 3 to 4-km-long transect was located in each beat, a nd each beat was a walked a total of three times by the local forest guard in charge of the beat. The average size of each beat was 20 km2. Data on presence of livestock signs, encounter rate of prey and sign of leopards were collected. Digitized beat maps were obtained and the centroid of each beat was used to approximate the location of leopard presences in the beat. Ecogeographical variable (EGV) maps of pr ey encounter rates for sambar, nilgai ( Bosephalus tragocamelus) and wild pig using th e INTERPOL module of Idrisi were created. The encounter rate of leopard sign was convert ed to a binary variable of presences and pseudoabsences. A buffer width of 3000 m was applied to create a 9 km2 patch of leopardpresence pixels around each point. Female leopard home ranges are known to vary from 6 to 30 km2 in Africa (Bailey 1993) and average 17 km2 in Nepal (Odden & Wegge 2005), so 9 km2 was

PAGE 92

92 considered a conservative estimate of the area in which presence could be assumed in the forest beat. Only beats where some evidence of leopard s was detected were retained for the analysis. All beats where evidence of le opards was not detected were discarded from the dataset. The elevation layer was obtained from the 90m resolution DEMs created from the SRTM mission data by the CGIAR-CSI ( http://srtm.csi.cgiar.org ). Using Idrisi, a slope m ap and a ruggedness map (using the standard deviation of mean elevation in a 3 x 3 moving window) was created from the DEM. A moving window of 3x3 has been used to create a similar index of terrain ruggedness to model mountain lion ha bitat in Montana (Rile y & Malecki 2001) The extensive study area encompassed parts of 6 Landsat ETM+ images. Georeferenced and orthorectified cloud-free images dating between 2002 and 2004 were obtained from the Global Landcover Facility ( http://glcf.umiacs.umd.edu ). Those parts of the extensive study area found in each of these im ages were classified into four cover types: agriculture, bare ground/urban, forest and water. These were then mosaicked together to obtain the cover map. For the STR area 5 cover types were delineated. These were: moist forest, dry forest, bare ground/village, teak dominated forest and water. Spectral signatures for the classification supervision were obtained by using information from 473 vegetati on plots in the Satpura Tiger Reserve, and with visual inspection of satellite imagery using Google Earth (http//: www.earth.google.com ) for the extensive study area. Supervised classification was perform ed using FISHER classifier for both the study sites using the Idrisi GIS package. Using the CircAnn module of Biomapper, the Boolean ma ps of each land cover type was converted to percent frequency in a 20 km2 circular moving window for the extensive study area (Figure 5-4) and 1 km2 for the STR site (Figure 5-5). The models were made at two pixel resolutions: 1000 m

PAGE 93

93 for the extensive study area and 100 m for the ST R study site. A list of all the EGVs for both study areas is given in Table 5-2 and 5-3. To observe the effect of pixel resolution on th e accuracy of the models, the ENFA analysis was repeated at the 200-m, 300-m and 500-m resolu tion for the STR area, and at 1000-m without the buffer, at 2000-m, 3000-m and 5000-m for the extensive study area. The relationship between the distribution of le opard presence patches and a set of mapped ecogeographical variables was analyzed using EN FA. The program Biomapper v3.2 (Hirzel et al. 2006a) was used. Biomapper needs two types of data to calculate habitat suita bility. The first is a map of locations where the species has been detect ed, and the next are a se t of quantitative raster maps describing the environment as used by the sp ecies under investigation. This presence-only modeling technique describes the ecological ni che of a species by computing uncorrelated factors from a comparison of values of ecogeogr aphical variables in th e entire study area and their values at the site wher e the species is known to be present. The first ENFA factor maximizes the absolute value of the marginality, defined as the standardized difference between the species mean and the global mean of each of the EGVs. The first factor explains how the species niche differs most from the available co nditions. The first factor also explains all the marginality and some of the specialization. Special ization is defined as the ratio of the overall variance to the species variance for all the EGVs, a nd describes how restricted is the usage of the species of that variable compared to its availab ility. Details on the calcul ation of marginality and specialization are given in Hi rzel et al. (2002). The subs equent factors maximize the specialization. A high absolute value of the corre lation of the variable with the specialization factor indicates that the species niche breadth is narrow with respect to that variable. There are as many factors as there are variables, but they successively explain a decreasing amount of the

PAGE 94

94 specialization. The number of fact ors used to calculate the habita t suitability was decided using MacArthurs broken-stick crit erion (Hirzel et al. 2002) The habitat-suitability map was evaluated fo r its predictive accur acy by internal area adjusted frequency cross-validation (Fie lding & Bell 1997). Leopard presences were geographically stratified and randomly partitioned into 10 sets. Nine partitions were used to compute a habitat suitability model and the left -out partition was used to validate it on independent data. This process was repeated 10 times, each time by leaving out a different partition. This process resulted in ten different habitat-suitability maps. Each map was reclassified into 4 bins, where each bin covered some proportion of the total study area (Ai) and contained some proportion of the left-out validat ion points (Ni).The areaadjusted frequency for each bin was computed as Fi = Ni /Ai. The expected Fi was 1 for all bins if the model was completely random. If the model is good, low values of habitat suitabi lity should have a low F (below 1) and high values a high F (above 1) with a monotonic incr ease in between. The monotonicity of the curve was measured with a Spearman rank correlation on the Fi in a moving window, termed as the continuous Boyce Inde x (Boyce et al. 2002; Hi rzel et al. 2006b). Validation of the models was also done using the Absolute Validation Index (AVI) and the Contrast Validation Index (CVI). AVI is the prop ortion of validation points that have a habitat selection of >=50. Possible values the index can take range from 0 to 1. The higher the value the more accurate is the model. CVI is calculated at AVI minus the AVIchance, which is the AVI one would expect from chance alone, and is a measure of departure from randomness of the model. Possible value the index can take range from 0 to AVI. One criticism of the presence models is that they yield too optimistic results (Zaniewski et al. 2002). This problem was mitigated by using breaks in the predicted-to-expected ratio frequency curves to define 4 habitat

PAGE 95

95 classes (Hirzel et al. 2006). The map was then r eclassified using the new bins into unsuitable, marginal, suitable and optimal habitat. Results Model Validation Overall the habitat suitability models fo r both STR and the extensive study area were equally accurate. They both showed similar values of AVI, indicating that the proportional accuracy in classifying presence points in the ev aluation partition was similar for south-central Madhya Pradesh and for Satpura Tiger Reserve. CVI values showed that the model had some difficulty in discriminating between the suitab ility map and a purely random model. This is consistent with the generalist na ture of the species. Both the c ontinuous Boyce Index values were high, indicating good predictive power for both th e models, but the extensive study area model had better predictive power (Table 5-4). The predicted-to-expected frequency curves showed higher variance for good habitat than for bad ha bitat with both models (Figure 5-8), the inflections in the curves were used to guide th e selection of bins to reclassify the habitat suitability maps for the two areas (Figure 5-6 and 5-7). Extensive Study Area The marginality value was 1.25 and the tolerance value was 0.92, indicating that leopards were using conditions that were different from the mean environmental values, and that the leopard was more of a generalist in using a wide range from the EGVs. Seven factors were retained. The first factor accounted for 100 % of the marginality, while all 7 factors accounted for 100 % of the marginality and 80% of the speci alization. The marginality coefficients show that leopard habitat was more pos itively correlated with sambar distribution, terrain ruggedness and percentage of forests. It was less strongly correlated with altitude, slope, NDVI and nilgai and wild pig encounter rates. Leopard distributio n was negatively correlated with presence of

PAGE 96

96 agriculture and urban-bare ground land cover types. Livestock pres ence, an indicator of human disturbance, was a weak negative correlate. The specialization factor indicated that the leopard used a restricted niche with respect to the ava ilability of percentage frequency of urban-bare ground and agriculture, but not when compared with the av ailability of elevation, ruggedness and slope measured at the 1 km scale across the big study area (Table 5-5). The amount of suitable, marginal, unsuitable and optimal habitat in each di strict is given Table 5-7. Maps of the EGVs are shown in Figure 5-4. Satpura Tiger Reserve The marginality value was 0.67, indicating that leopards were using conditions not too different from the mean environmental values Tolerance was also relatively high (0.56), indicating that the leopard was found in areas that had a wide range of values of the EGVs. Four factors were retained, the firs t factor accounting for 100 % of th e marginality. The four factors explained 79 % of the specialization. The margin ality factor was strongly positively loaded with the coefficient for tassled-cap greenness, an index of above ground biomass (Crist & Kauth 1986) and percentage frequency of moist forests and teak dominated forest. It was also positively correlated with distance to water and the encounte r rate of cervids (sambar and chital), wild pig and small-sized prey. The positive correlation wi th langur encounter rate was weak. Leopard presence was negatively correlated with elevat ion, slope and frequenc y of bare ground pixels. The negative loading with respect to distance fr om village was weak. The specialization factor indicated that elevation was used in a more restricted way than was available in the study area, as was the frequency of the moist, dry and teak fore sts (Table 5-6). Maps of the EGVs are shown in Figure 5-5.

PAGE 97

97 Effect of Changing Resolution For the extensive study area the be st model was at the 1000-m scal e with buffer. It gave the highest continuous Boyce Index value. The coar sest resolution model was the most inaccurate. At a resolution of 5 km and a moving window size of 225 km2, the continuous Boyce Index reduced to 0.55. Changing the resolution did not change the AVI, CVI and the Boyce Index much at all the other resolutions. For the STR ar ea, the best model was the 100-m resolution with a moving window of 1-km2, followed by the 200-m model. The effect of increasing the moving window scale degraded accuracy slightly. The 300-m and 500-m resolution models had lower Boyce Index values (Table 5-4). The habitat suitability maps for the two areas were made from the best models. Discussion The leopard is an adaptable species, being able to live in a wide variety of environmental conditions. This is reflected in the marginality and tolerance values for the model of the STR area, where almost all the area is potential leopard habitat. Habitat use by leopards in Satpura was strongly associated with moist and teak forests, as well as with most prey species, except the langur, with which it was only weakly associated. This is because more langur are seen in open areas, closer to villages, and along roads, rather than in denser forest areas ( pers obs ), perhaps as an anti-predatory strategy, and they do not comp rise a large proportion of the leopards diet in this area (Chapter 3). Leopard presence had a w eak negative associati on with the distance to villages. That means it was found closer to villages than average, though this tendency was weak. Unlike tigers, which are shy and prone to move away from disturbance, leopards are known to be bold and not uncommonly found in proxi mity to human habitats, where they prey upon livestock (Odden & Wegge 2005). Though they ar e tolerant of human presence, they are not unaffected by disturbance, as the extensive study area model showed, with leopard habitat

PAGE 98

98 being negatively associated with bare ground/ urban land use frequency. In Thailand leopard activity has been shown to be negatively correlat ed with distance from villages (Ngoprasert et al. 2007). Leopard habitat was negatively correlated with urban-bare ground and agriculture land cover types as also with livestock presence. At the large scale, good leopar d habitat was seen to be more associated with terrain ruggedness, sambar availability and percentage of forested areas, and less associated with nilgai and wild pig prey availability. Both the latter species are known to be crop pests and able to live close to human inhabitation (S ekhar 1998), and this probably contributes to the obser ved pattern. Cougar ( Puma concolor ) abundance has also been shown to be affected by prey availability, terrain ruggedness and forest cover at the landscape scale (Riley & Malecki 2001). The larger spatial area model had a higher pr edictive accuracy than the smaller scale as quantified by the higher continuous Boyce Index (Table 5-4). This is possibly because the Satpura Tiger Reserve has relatively little distur bance and is a less heterogeneous area given its smaller size. Given the high density of leopards in the area (Chapter 4) and that they require relatively large tracts of contiguous habita t (Marker & Dickman 2005) they probably move through and spend time in habitats that are not highly preferred, but are still inhabitable. Consequently, very few areas in the Reserve are lik ely to be completely unsuitable for leopards. The change in resolution seemed to have a similar impact on models of both study areas. Coarse pixel resolutions, at 300 m and 500 m for STR and 5 km for the extensive study area, degraded the accuracy of the models. The scale of the circular moving window for frequency of land use cover did not change accuracy appreciably, except at the very la rgest spatial resolution (225 km2).

PAGE 99

99 The habitat model was used to estimate the area occupied by various ha bitat categories in the 13 districts in south-central Madhya Pradesh (Table 5-7). Optimal ha bitat was 5.2% of the study area, ranging from 0.5 to 8 percent of each di strict. As an absolute measure it can be said that approximately 11500 km2 of habitat is likely to support leopard populations. The districts with the most optimal habitat are Betul, Ho shangabad and Chhindwara. These districts are geographically adjacent to each other and c onstitute a compact block of about 2000 km2 of optimal habitat. The Satpura Tiger Reserve lies in Hoshangabad district an d is already protected, but Betul and Chhindwara district s can be prioritized when al locating resources for leopard conservation efforts in Madhya Pradesh. In conclu sion the ENFA model seems to work better at larger spatial areas for a generalist species like th e leopard. It is a useful tool to explore the characteristics of the leopards niche as well as to produce habitat suitability maps that can aid in conservation management.

PAGE 100

100 Table 5-1. Districts, sampling effort and leopard presence in th e extensive study area in southcentral Madhya Pradesh. District Sampled Area (km2)Number of transectsTransects with Leopard presence Balaghat 419.9 50 0 Betul 10041.5 622 23 Bhopal 57.1 1 0 Chhindwara 11815.8 553 57 Dewas 1296.7 30 0 East Nimar 2104.4 100 8 Harda 3329.0 163 2 Hoshangabad 6734.1 351 111 Jabalpur 258.5 27 2 Narmsimhapur 4420.3 107 17 Raisen 3293.9 92 16 Sehore 3266.8 112 23 Seoni 5891.4 361 41

PAGE 101

101 Table 5-2. List of ecogeographical variables (EGV) with explanati on and source for southcentral Madhya Pradesh. Ecogeographical Variables Explanation TransformationSource Elevation DEM in meters at 100 m spatial resolution, averaged to 1 km spatial resolution. None SRTM data Ruggedness (Elevation standard deviation) Calculated with a moving window of 3x3 cells from DEM. None Calculated Slope Calculated from DEM None Calculated NDVI Calculated from bands 3 and 4 of Landsat ETM + imagery. None Calculated Forest Percentage frequency of cells with forests, urban/bareground and agriculture in a circular window area 25 km2. None Supervised classification of Landsat ETM+ imagery to obtain landcover; Frequency calculated. Urban/bareground Same as above None Same as above Agriculture Same as above None Same as above Livestock Distance-weighted interpolation of encounter rate (number seen/km) from line transects. Square root Calculated. Nilgai encounter rate (ER) Same as above Box-Cox Same as above Sambar ER Interpolated encounter rate (number seen/km) from line transects. Box-Cox Same as above Wild pig ER Interpolated encounter rate (number seen/km) from line transects. Box-Cox Same as above Distance to Water Distance to the nearest water source in meters. None Calculated.

PAGE 102

102 Table 5-3. List of ecogeographical variables (E GV) with explanation a nd source for the Satpura Tiger Reserve. Ecogeographical Variables Explanation TransformationSource Cervid Encounter Rate (ER) Interpolated encounter rate (number seen/km) from line transects and vehicle transects Box-Cox This study Langur ER Same as above Box-Cox This study Wild pig ER Same as above Box-Cox This study Small Prey ER Interpolated ER of jungle fowl, spur fowl, peafowl and black-naped hare None This study Bare ground Percentage frequency of cells with in a circular window of area 1 km2 Box-Cox Supervised classification of Landsat ETM+ imagery to obtain landcover; frequency calculated. Dry forest Same as above Box-Cox Same as above Teak dominated forest Same as above None Same as above Moist forest Same as above Box-Cox Same as above Tassled-cap greenness The first band of tassledcap transform using Landsat ETM + imagery. None Calculated. Elevation DEM in meters at 100m resolution Box-Cox SRTM data Slope Calculated from DEM Box-Cox Calculated. Distance from village Distance to the nearest village Box-Cox Calculated Distance to water Distance from the nearest water source in meters Box-Cox Calculated.

PAGE 103

103 Table 5-4. Measures of evaluation for habitat models at different pixel resolutions (with crossvalidated standard deviations). Study Site Model Resolution Circular moving window size AVI CVI Continuous Boyce Index STR 100 m1 km2 0.51 (0.11) 0.30 (0.11) 0.75 (0.18) STR 100 m54 km2 0.48 (0.14) 0.33 (0.13) 0.69 (0.35) STR 200 m56 km2 0.50 (0.19) 0.40 (0.19) 0.74 (0.25) STR 300 m52 km2 0.49 (0.22) 0.34 (0.21) 0.36 (0.39) STR 500 m56 km2 0.49 (0.17) 0.34 (0.16) 0.63 (0.26) SC Madhya Pradesh 1000 m, with buffer 20 km2 0.48 (0.12) 0.33 (0.11) 0.91 (0.13) SC Madhya Pradesh 1000 m21 km2 0.48 (0.15) 0.33 (0.14) 0.72 (0.32) SC Madhya Pradesh 2000 m84 km2 0.50 (0.19) 0.35 (0.18) 0.78 (0.24) SC Madhya Pradesh 3000 m81 km2 0.52 (0.11) 0.29 (0.10) 0.77 (0.17) SC Madhya Pradesh 5000 m225 km2 0.48 (0.11) 0.20 (0.10) 0.55 (0.25) Note: AVI measures proportional accuracy in classifying habitat and ranges from 0 to 1. Higher values of AVI denote a more accurate model. CVI measures the difference between the model and a random model, with values ranging from 0 to AVI. High values of CVI indicate a model that is very different from random. The Boyce index measures the correlation be tween habitat suitability values and the area adjusted freque ncy of presence points in the habitat map.

PAGE 104

104 Table 5-5. Correlation between ENFA factors an d EGV for south-central Madhya Pradesh. The percentages quantify the amount of speci alization attributed to the factor. EGV Factor1 (12%)+ Factor2 (22%) Factor3 (12%) Factor4 (10%) Factor5 (9%) Factor6 (8%) Factor7 (7%) Elevation + 00*********** ** Elevation standard deviation ++++ 00*0* 0 Slope ++ 00000 0 NDVI ++ ************* ** Forest +++++ ****************** ******* Bare ground/urban -******************** *** Agriculture ---************************** ***** Livestock ER *****0**** ** Nilgai ER +++ 00*00 0 Sambar ER ++++ *0*0* Wild pig ER ++ *********** ** Distance to water 0 *0****** ** Note: +For the marginality factor, the + symbol indi cates that leopards presence was associated with values higher than average, and vice versa for -. The number of signs indicates the strength of the relationship. For the specialization factor, indi cates that leopards were found in narrower range of values than av ailable. The number of indicates the narrowness of the range. A 0 indicates low specialization. Factor 1 accounts for all the marginality.

PAGE 105

105 Table 5-6. Correlation between ENFA factor s and EGV for Satpura Tiger Reserve. The percentages quantify the amount of speci alization attributed to the factor. EGV Factor1 + (22%)Factor2 (42%)Factor3 (9%) Factor4 (6%) Elevation ---****** 0 0 Slope -0 0 Langur ER + 0 0 Cervid ER ++ 0 0 0 Pig ER ++ 0 0 0 Small-prey ER +++ 0 0 Tassled cap greenness ++++ ******** ******** Teak Forest +++ ******** *** Moist Forest +++********** **** Dry Forest --**** ** Bare ground ---* *** *** Distance to water +++ 0 ** Distance to village 0 ** ** Note: +For the marginality factor, the + symbol indi cates that leopards presence was associated with values higher than average, and vice versa for -. The number of signs indicates the strength of the relationship. For the specialization factor, indi cates that leopards were found in narrower range of values than av ailable. The number of indicates the narrowness of the range. A 0 indicates low specialization.Factor 1 accounts for all the marginality.

PAGE 106

106 Table 5-7. Area under various leopard-habitat ca tegories in south-centr al Madhya Pradesh. District Unsuitable (km2) Marginal (km2) Suitable (km2) Optimal (km2) Balaghat 118.3 210.686.25.0 Betul 5031.5 2387.91172.1857.5 Bhopal 50.1 7.000 Chhindwara 5621.2 3528.21931.6744.1 Dewas 776.2 288.8166.566.2 East Nimar 1141.3 453.3343.0168.5 Harda 2187.3 490.4415.2238.7 Hoshangabad 3325.6 1577.51371.0465.3 Jabalpur 38.1 172.528.120.0 Narmsimhapur 3265.4 638.8303.9215.6 Raisen 2464.1 535.5282.814.0 Sehore 1590.6 882.5583.7212.6 Seoni 2951.5 1745.0892.6306.9

PAGE 107

107 Figure 5-1. Cover map of the study ar ea in south-central Madhya Pradesh.

PAGE 108

108 Figure 5-2. Mosaicked landsat sate llite image of the study area in south-central Madhya Pradesh.

PAGE 109

109 Figure 5-3. Landsat satellite image of the study area in Satpura Tiger Reserve.

PAGE 110

110 Figure 5-4. Maps of remotely derived variables for south-central Madhya Pradesh. A) Frequency of forests. B) Frequency of urban/bareground. C) Fre quency of agriculture. D) Distance to water sources. E: Elevation. F: Slope G: Ruggedness (Std deviation of elevation). H) NDVI. I) Nilgai abundance index J) Sambar abundance index. K) Wild pig abundance index. L) Livestock abundance index.

PAGE 111

111 Figure 5-4. Continued.

PAGE 112

112 Figure 5-4.Continued.

PAGE 113

113 Figure 5-5. Maps of remotely derived variables for Satpura Ti ger Reserve. A) Frequency of bareground. B) Frequency of moist forest. C) Frequency of dry forest. D) Frequency of teak forest. E) Cervid abundance index F) Wild pig abundance index. G) Langur abundance index. H) Small prey abundance index. I) Elevation. J) Slope. K) Tassel cap greenness. L) Distance to village. M) Distance to water.

PAGE 114

114 Figure 5-5. Continued.

PAGE 115

115 Figure 5-5. Continued.

PAGE 116

116 Figure 5-6. Leopard habitat suitability map for south-central Madhya Pradesh. Figure 5-7. Leopard habitat suitabil ity map for Satpura Tiger Reserve.

PAGE 117

117 Figure 5-8. The predicted-to-expect ed frequency curves with hab itat suitability values for both the models.

PAGE 118

118 CHAPTER 6 CONCLUSION Density of Potential Prey Rigorous estimates of ungulate density were not available until now for the Satpura Tiger Reserve. The results of this study indicate that wildlife populations are lower than those in relatively well protected parks in India such as Nagarhole and Bandipur in southern India and Kanha and Pench tiger reserves in Madhya Prades h. They are however in line with estimates from other central Indian protected areas li ke Tadoba, Melghat and Pench (Maharashtra). Though densities of most prey species are low in the study area, the ungulate community is still intact. Since the area is relatively large, a hi gher degree of protection and habitat management should increase prey and be able to support a relatively la rge population of carnivores. The coefficients of variation around th e density estimates of most sp ecies generated by this study are low enough to be useful to monitor population changes. A note of caution should be sounded since the last year of sampling showed lower densities for all species. A monitoring program needs to be instituted to make sure that this population is not continuously declining. This monitoring program should involve distance sampling and can use vehi cles to take advantage of the extensive trail network. Preference of Prey Leopards, dholes and tigers st rongly prefer sambar in this study area. The density of sambar can be enhanced further using suitable habitat management techniques, but the main requirement is likely to be effective protecti on from poaching. Because the food habits of leopards, dholes and tigers overlap to a great extent, any increase in density of medium and large biomass will benefit all the predators.

PAGE 119

119 Density of Leopards There is no other published info rmation on density of leopards in India, so comparison of these results with other areas is difficult. Indices computed for other parks are in the same range as ones calculated for this study. Index based approaches such as the number of photographs per night (RAI) have been recommended to be used for tigers when there are not enough data for mark recapture sampling (Carbone et al. 2001). In this study, the RAI did not index leopard density well and it cannot be reco mmended. Therefore the results of indices available for other study areas should be interpreted with caution. At the present time there appears to be no substitute to the mark-recapture framework used in this study. It is recommended that leopard densities be estimated across all protected areas and a reliable i ndex be developed so that large areas in the country can be monitored for population changes. There is urgent need for more research to accurately estimate the effectiv e trapping area. The spatially explicit maximum likelihood method offers hope for the future, but w ith the low population sizes it is difficult to obtain high precision, maki ng it difficult to detect changes in population density. The logistics involved in sampling at a high enough intens ity to get good precision and at a large enough spatial scale for a species of this size are very difficult. Given these limitations, there is urgent need for a calibrated index to be develope d to monitor population changes of leopards. Habitat Model The leopard is an adaptable species, being able to live in a wide variety of environmental conditions. The habitat model s howed that moist deciduous and teak dominant forests had a higher association with leopards. Prey densities, especially those of sambar, were also higher in these habitat types. The larger spatial scal e model showed that leopards were negatively associated with land cover associated with hum an use. Though it has a reputation for tolerating human presence, leopard densities are negative ly affected by disturba nce and presence of

PAGE 120

120 agriculture. Given that it is a large-sized carnivore species that requ ires relatively large tracts of contiguous habitat, the model predic ted a compact block of about 2000 km2 of optimal habitat in the districts of Betul, Hoshangabad and C hhindwara. In addition, approximately 11500 km2 of habitat is likely to support le opard populations in south-cen tral Madhya Pradesh. It is recommended that protection for this habita t be adequately strengthened and resources prioritized accordingly when managing leopard conservation efforts in Madhya Pradesh.

PAGE 121

121 APPENDIX A INDICES OF UNGULATE AN D C ARNIVORE ABUNDANCE Table A-1. Kilometric index values (number of indi viduals per km.) of selected species using dirt trails in the monsoon from 2002 to 2005 with bootstrapped 95% C.I. Species 2002 N=9696.2 km 2003 N=9910.3 km 2004 N=91188.1 km 2005 N=9811.8 km Chital 1.26 (1.01-1.54) 0.77 (0.64-0.92) 1.39 (1.12-1.71) 1.48 (1.16-1.87) Sambar 0.43 (0.36-0.52) 0.31 (0.25-0.37) 0.19 (0.15-0.24) 0.14 (0.10-0.19) Indian muntjac 0.08 (0.06-0.10) 0.04 (0.03-0.06) 0.04 (0.03-0.05) 0.04 (0.03-0.06) Nilgai 0.13 (0.09-0.18) 0.08 (0.05-0.10) 0.06 (0.04-0.08) 0.06 (0.03-0.09) Wild pig 0.43 (0.30-0.58) 0.36 (0.27-0.46) 0.24 (0.15-0.37) 0.17 (0.07-0.24) Common langur 1.28 (1.05-1.54) 1.14 (0.98-1.3) 2.51 (2.22-2.79) 2.39 (0.21-0.28) Indian peafowl 0.30 (0.24-0.35) 0.34 (0.29-0.38) 0.26 (0.22-0.31) 0.24 (0.19-0.29)

PAGE 122

122 Table A-2. Encounter rates of tr acks of selected carnivo re species (frequency per 500 m section), with bootstrapped 95 % C.I. (November 2003 to June 2006). Time period Leopard Dhole Tiger Sloth bear Jungle cat Palm civet NovFeb 0.14 (0.080.19) 0.24 (0.180.31) 0.06 (0.030.11) 0.22 (0.170.28) 0.10 (0.060.14) 0.32 (0.260.39) MarJun 0.13 (0.100.16) 0.15 (0.120.18) 0.03 (0.02-.05) 0.43 (0.390.47) 0.09 (0.070.11) 0.35 (0.310.39) NovFeb 0.11 (0.090.13) 0.07 (0.050.09) 0.08 (0.060.10) 0.10 (0.080.13) 0.02 (0.010.03) 0.22 (0.200.26) MarJun 0.03 (0.020.04) 0.03 (0.020.04) 0.02 (0.010.02) 0.36 (0.330.39) 0.06 (0.050.09) 0.27 (0.250.30) NovFeb 0.11 (0.100.13) 0.14 (0.120.15) 0.02 (0.020.03) 0.11 (0.010.13) 0.12 (0.100.14) 0.45 (0.420.47) MarJun 0.12 (0.100.14) 0.08 (0.060.09) 0.02 (0.010.03) 0.26 (0.230.28) 0.09 (0.080.11) 0.33 (0.300.35)

PAGE 123

123 LIST OF REFERENCES Acharya, B. B. 2008. The Ecology of the Dhole or Asiatic W ild Dog ( Cuon alpinus ) In Pench Tiger Reserve, Madhya Pradesh. Page 115. Saurashtra University Rajkot. Ackerman, B. B., F. G. Lindzey, and T. P. He mker. 1984. Cougar Food-Habits in Southern Utah. Journal of Wildlife Management 48:147-155. Andheria, A. P., K. U. Karanth, and N. S. Kumar. 2007. Diet and prey profiles of three sympatric large carnivores in Bandipur Tiger Reserve, India. Journal of Zoology 273:169-175. Athreya, V. R., S. S. Thakur, S. Chaudhari, and A. V. Belsare. 2007. Leopards in humandominated areas: a spillover from sustained tran slocations into nearby forests. Journal of the Bombay Natural History Society 104 13-18. Austin, M. 2007. Species distri bution models and ecological theo ry: A critical assessment and some possible new approaches. Ecological Modelling 200:1-19. Avinandan D. 2003. Food habits of the tiger ( Pathera tigris tigris ) in Sariska Tiger Reserve, Rajasthan. Page 63. Saurashtra University, Dehradun. Bagchi, S., S. P. Goyal, and K. Sankar. 2003a. Ni che relationships of an ungulate assemblage in a dry tropical forest. Journal of Mammalogy 84:981-988. Bagchi, S., S. P. Goyal, and K. Sankar. 2003b. Prey abundance and prey selection by tigers ( Panthera tigris ) in a semi-arid, dry deciduous fore st in western India. Journal of Zoology 260:285-290. Bagchi, S., S. P. Goyal, and K. Sankar. 2004. Herbivore density and biomass in a semi-arid tropical dry deciduous forest of western India. Journal of Tropical Ecology 20:475-478. Bailey, T. N. 1993. The African Leopard: Ecology and Behavior of a Solitary Felid. Columbia University Press. Bertram, B. C. R. 1982. Leopard ecology as studied by radio tracking. Symp. zool. Soc. Lond:341-352. Bhatnagar, Y. V. 1991. Habitat preference of sambar (Cervus unicolor ) in Rajaji National Park. Page 51. Saurashtra University, Rajkot. Biswas, S., and K. Sankar. 2002. Prey abundance and food habit of tigers ( Panthera tigris tigris ) in Pench National Park, Madhya Pradesh, India. Journal of Zoology 256:411-420. Bothma, J. D. P., and E. A. N. Le Riche. 1984. Aspects of the ecology and the behaviour of the leopard Panthera pardus in the Kalahari Desert. Koedoe 27:259-279.

PAGE 124

124 Boyce, M. S., P. R. Vernier, S. E. Nielsen, a nd F. K. A. Schmiegelow. 2002. Evaluating resource selection functions. Ecological Modelling 157:281-300. Braunisch, V., and R. Suchant. 2007. A model fo r evaluating the 'habitat potential' of a landscape for capercaillie Tetrao urogallus : a tool for conservation planning. Wildlife Biology 13:21-33. Brotons, L., W. Thuiller, M. B. Araujo, and A. H. Hirzel. 2004. Presence-absence versus presence-only modelling methods for predic ting bird habitat su itability. Ecography 27:437-448. Bryan, T. L., and A. Metaxas. 2007. Predicting su itable habitat for deep-water gorgonian corals on the Atlantic and Pacific Continental Margins of North America. Marine EcologyProgress Series 330:113-126. Buckland, S. T., D. R. Anderson, K. P. Burnham, J. L. Laake, and D. L. Borchers 2001. Introduction to distance sampling: Estima ting abundance of biological populations. Oxford University Press. Burnham, K. P., and D. Anderson 2003. Model Se lection and Multi-Model Inference. Springer. Burnham, K. P., and W. S. Overton. 1979. R obust estimation of population-size when capture probabilities vary among animals. Ecology 60:927-936. Bustamante, J., and J. Seoane. 2004. Predicting the distribution of four speci es of raptors (Aves : Accipitridae) in southern Spain: statistical models work better than existing maps. Journal of Biogeography 31:295-306. Carbone, C., S. Christie, K. Conforti, T. Coulson, N. Franklin, J. R. Ginsberg, M. Griffiths, J. Holden, K. Kawanishi, M. Kinnaird, R. La idlaw, A. Lynam, D. W. Macdonald, D. Martyr, C. McDougal, L. Nath, T. O'Brien, J. Seidensticker, D. J. L. Smith, M. Sunquist, R. Tilson, and W. N. W. Shahruddin. 2001. The use of photographic rates to estimate densities of tigers and other cr yptic mammals. Animal Conservation 4:75-79. Carbone, C., G. M. Mace, S. C. Roberts, and D. W. Macdonald. 1999. Energetic constraints on the diet of terrestrial carnivores. Nature 402:286-288. Champion, H. G., and S. K. Seth 1968. A revised su rvey of the forest types of India. Manager of publications, Government of India, Delhi. Chefaoui, R. M., J. Hortal, and J. M. Lobo. 2005. Potential distribution modelling, niche characterization and conservation status a ssessment using GIS tools: a case study of Iberian Copris species. Biological Conservation 122:327-338. Chellam, R. 1993. The ecology of the asiatic lion ( Panthera leo persica ). Saurashtra University, Rajkot.

PAGE 125

125 Chesson, J. 1978. Measuring Preference in Selective Predation. Ecology 59:211-215. Crist, E. P., and R. J. Kauth. 1986. The tassele d cap de-mystified. Photogrammetric Engineering and Remote Sensing 52:81-86. David, A. C., Q. Qureshi, S. P. Goyal, K. Sa nkar, V. B. Mathur, Pannalal, A. Verma, and D. Patial. 2005. Estimating Herbivore Abundance us ing Line Transect Method in Sariska Tiger Reserve. Page 10. Wildlife Institute of India, Dehradun, India. Dettki, H., R. Lofstrand, and L. Edenius. 2003. Modeling habitat suitability for moose in coastal northern Sweden: Empirical vs proc ess-oriented approaches. Ambio 32:549-556. Dice, L. R. 1938. Some Census Methods for Mammals. The Journal of Wildlife Management 2:119-130. Dinerstein, E., C. Loucks, E. Wikramanayake, J. Ginsberg, E. Sanderson, J. Seidensticker, J. Forrest, G. Bryja, A. Heydlauff, S. Klenzendorf, P. Leimgruber, J. Mills, T. G. O'Brien, M. Shrestha, R. Simons, and M. Songer. 2007. The fate of wild tigers. Bioscience 57:508-514. Diniz, J. A. F., L. M. Bini, and B. A. Hawkin s. 2003. Spatial autocorrel ation and red herrings in geographical ecology. Global Ecology and Biogeography 12:53-64. Eastman, R. 2004. Idrisi Kiliman jaro v14.02. Clark University. Edgaonkar, A., and R. Chellam. 2002. Food habit of the leopard, Panthera pardus, in the Sanjay Gandhi National Park, Maharashtra, India. Mammalia 66:353-360. Edgaonkar, A. J., and R. Chellam. 1998. A preliminary study on the ecology of the leopard, Panthera pardus fusca in the Sanjay Gandhi Nationa l Park, Mahara shtra Wildlife Institute of India, Dehradun, India:39. Efford, M. 2004. Density estimati on in live-trapping studies. Oikos 106:598-610. Efford, M. 2007. Density: spatially explicit capture recapture, Otago. Efford, M., D. Borchers, and A. Byrom. in review. Density estimation by spatially explicit capturerecapture: likelihood-based methods. Environmental and Ecological Statistics. Eisenberg, J. F. 1986. Life history strategies of the felidae: Variations on a common theme. Pages 293-303 in S. D. Miller, and D. D. Evere tt, editors. Cats of the World of the World: Biology, Conservation, and Management. National Wildlife Foundation, Washington, D.C. Eisenberg, J. F., and M. Lockhart. 1972. An ecological reconnaissance of Wilpattu National Park. Smithsonian Contributions to Knowledge, (Zoology) 101:1-118.

PAGE 126

126 Eisenberg, J. F., and J. Seidensticker. 1976. Ungulates in southern Asia: A consideration of biomass estimates for selected ha bitats. Biological Conservation 10:293-308. Emmons, L. H. 1987. Comparative feeding ecolo gy of felids in a neotropical rainforest. Behavioral Ecology and Sociobiology 20:271-283. Farrell, L. E., J. Roman, and M. E. Sunquist. 200 0. Dietary separation of sympatric carnivores identified by molecular analys is of scats. Molecular Ecology 9:1583-1590. Fielding, A. H., and J. F. Bell. 1997. A review of methods for th e assessment of pr ediction errors in conservation presence/absence models. Environmental Conservation 24:38-49. Floyd, T. J., L. D. Mech, and P. A. Jordan. 1978. Relating Wolf Scat Cont ent to Prey Consumed. Journal of Wildlife Management 42:528-532. Gangadharan, A. 2005. Density Estimation and Time Trend Analysis of Large Herbivores in Nagarhole, India. Page 82. University of St. Andrews. Gotelli, N. J., and G. L. Entsminger. 2007. Ec oSim: Null models software for ecology. Acquired Intelligence Inc. & Kesey-B ear, Jericho, VT 05465. Gu, W. D., and R. K. Swihart. 2004. Absent or undetected? Effect s of non-detection of species occurrence on wildlife-habitat mo dels. Biological Conservation 116:195-203. Guisan, A., A. Lehmann, S. Ferrier, M. Austin, J. M. C. Overton, R. Aspinall, and T. Hastie. 2006. Making better biogeogra phical predictions of species distributions. Journal of Applied Ecology 43:386-392. Guisan, A., and N. E. Zimmermann. 2000. Predic tive habitat distributi on models in ecology. Ecological Modelling 135:147-186. Hamilton, P. H. 1976. The movements of leopards in Tsavo National Park, Kenya, as determined by radio-tracking. University of Nairobi. Harihar, A. 2005. Population, Food Habits and Prey Densities of Tiger in Chilla Range, Rajaji National Park, Uttaranchal, India. Saurashtra University, Rajkot. Hayward, M. W. 2006. Prey pref erences of the spotted hyaena ( Crocuta crocuta ) and degree of dietary overlap with the lion ( Panthera leo ). Journal of Zoology 270:606-614. Hayward, M. W., P. Henschel, J. O'Brien, M. Hofmeyr, G. Balme, and G. I. H. Kerley. 2006a. Prey preferences of the leopard (Panthera pardus ). Journal of Zoology 270:298-313. Hayward, M. W., and G. I. H. Kerley 2005. Prey preferences of the lion ( Panthera leo ). Journal of Zoology 267:309-322.

PAGE 127

127 Hayward, M. W., J. O'Brien, M. Hofmeyr, and G. I. H. Kerley. 2006b. Prey preferences of the African wild dog Lycaon pictus (Canidae : Carnivora): Ec ological requirements for conservation. Journal of Mammalogy 87:1122-1131. Hines, J. 1994. CAPTURE2 version 070503.1333 Patuxent Wildlife Research Center,USGS. http://www.mbr-pwrc.usgs.gov. Accessed on 1 September 2007. Hines, J. 2002. SCATMAN version 2. Pa tuxent Wildlife Research Center, USGS. http://www.mbr-pwrc.usgs.gov. Accessed on 1 September 2007. Hirzel, A. H., J. Hausser, D. Chessel, and N. Perrin. 2002. Ecological-niche factor analysis: How to compute habitat-suitability ma ps without absence data? Ecology 83:2027-2036. Hirzel, A. H., J. Hausser, and N. Perrin. 2006a. Biomapper version 3.2. Laboratory for Conservation Biology Universi ty of Lausanne, Switzerland. Hirzel, A. H., V. Helfer, and F. Metral. 2001. Assessing habitat-suitability models with a virtual species. Ecological Modelling 145:111-121. Hirzel, A. H., G. Le Lay, V. Helfer, C. Randi n, and A. Guisan. 2006b. Evaluating the ability of habitat suitability models to predict species presences. Ecological Modelling 199:142152. Hirzel, A. H., B. Posse, P. A. Oggier, Y. Crettenand, C. Glenz, and R. Arlettaz. 2004. Ecological requirements of reintroduced species and the im plications for release policy: the case of the bearded vulture. Journal of Applied Ecology 41:1103-1116. Hoppe-Dominik, B. 1984. Prey Frequency of the Leopard, Panthera pardus in the Tai-NationalPark of the Ivory-Coast. Mammalia 48:477-487. Jackson, R. M., J. D. Roe, R. Wangchuk, and D. O. Hunter. 2006. Estimating snow leopard population abundance using photography and cap ture-recapture techniques. Wildlife Society Bulletin 34:772-781. Jacobs, J. 1974. Quantitative Measurement of Food Selection Modification of Forage Ratio and Ivlevs Electivity Index. Oecologia 14:413-417. Jathanna, D., K. U. Karanth, and A. J. T. Johnsingh. 2003. Estimation of large herbivore densities in the tropical forests of southern India using distance sampling. Journal of Zoology 261:285-290. Jennelle, C. S., M. C. Runge, and D. I. MacKenzie. 2002. The use of photographic rates to estimate densities of tigers and other cr yptic mammals: a comment on misleading conclusions. Animal Conservation 5:119-120.

PAGE 128

128 Jenny, D. 1996. Spatial organization of leopards Panthera pardus in Tai National Park, Ivory Coast: Is rain forest habitat a 't ropical haven'? Journal of Zoology 240:427-440. Jhala, Y. V., R. Gopal, and Q. Qureshi. 2008. St atus of the Tigers, Co-predators, and Prey in India. Page 151. National Tige r Conservation Authority, Govt. of India, New Delhi, and Wildlife Institute of India, Dehradun. Johnsingh, A. J. T. 1983. Large mammalian prey--predators in Bandipur. Journal of the Bombay Natural History Society 80:1-57. Johnsingh, A. J. T. 1992. Prey Selection in 3 Large Sympatric Carnivores in Bandipur. Mammalia 56:517-526. Johnson, A., C. Vongkhamheng, M. Hedemark, a nd T. Saithongdam. 2006. Effects of humancarnivore conflict on tiger ( Panthera tigris ) and prey populations in Lao PDR. Animal Conservation 9:421-430. Johnson, D. H. 1980. The comparison of usage and availability measurements for evaluating resource preference. Ecology 61:65-71. Karanth, K. U. 1995. Estimating Tiger Panthera tigris populations from camera-trap data using capture-recapture models. Biological Conservation 71:333-338. Karanth, K. U., R. S. Chundawat, J. D. Nic hol, and N. S. Kumar. 2004a. Estimation of tiger densities in the tropical dr y forests of Panna, Central In dia, using photographic capturerecapture sampling. Animal Conservation 7:285-290. Karanth, K. U., and J. D. Nichols. 1998. Estima tion of tiger densities in India using photographic captures and recaptures. Ecology 79:2852-2862. Karanth, K. U., and J. D. Nichols. 2000. Ecologica l Status and Conservation of Tigers in India: Final Technical Report submitted to US Fish and Wildlife Service, Washington DC, and Wildlife Conservation Society, New York. Centre for Wildlife Studies, Bangalore, Bangalore. Karanth, K. U., J. D. Nichols, N. S. Kumar, W. A. Link, and J. E. Hines. 2004b. Tigers and their prey: Predicting carnivore densities from prey abundance. Proceedings of the National Academy of Sciences of the United States of America 101:4854-4858. Karanth, K. U., J. D. Nichols, J. Seidensticker, E. Dinerstein, J. L. D. Smith, C. McDougal, A. J. T. Johnsingh, R. S. Chundawat, and V. Thap ar. 2003. Science deficiency in conservation practice: the monitoring of tiger populat ions in India. Animal Conservation 6:141-146. Karanth, K. U., and M. E. Sunquist. 1992. Population-Structure, Density and Biomass of Large Herbivores in the Tropical Forests of Naga rahole, India. Journal of Tropical Ecology 8:21-35.

PAGE 129

129 Karanth, K. U., and M. E. Sunquist. 1995. Prey Selection by Tiger, Leopard and Dhole in Tropical Forests. Journal of Animal Ecology 64:439-450. Karanth, K. U., and M. E. Sunquist. 2000. Beha vioural correlates of predation by tiger ( Panthera tigris), leopard ( Panthera pardus) and dhole ( Cuon alpinus) in Nagarahole, India. Journal of Zoology 250:255-265. Karanth, U., and J. Nichols 2002. Monitoring Tigers and Their Prey. Center for Wildlife Studies. Karlsson, J., H. Broseth, H. Sand, and H. Andr en. 2007. Predicting occurren ce of wolf territories in Scandinavia. Journal of Zoology 272:276-283. Khan, J. A., R. Chellam, W. A. Rodgers, and A. J. T. Johnsingh. 1996. Ungulate densities and biomass in the tropical dry deciduous forests of Gir, Gujarat, India. Journal of Tropical Ecology 12:149-162. Kitchener, A. C. 1991. The natu ral history of the w ild cats. Christopher Helm, A and C Black, London. Krebs, C. J. 1989 Ecological met hodology. Harper and Row, New York. Lawlor, L. R. 1980. Structure and Stability in Natural and Randomly Constructed Competitive Communities. American Naturalist 116:394-408. Lele, U., N.Tuinar, S. A. Hussain, A. Zazneta, and L. Kelly. 2000. The World Bank forest strategy:striking th e right balance. Link, W. A., and K. U. Karanth. 1994. Corr ecting for Overdispersi on in Tests of Prey Selectivity. Ecology 75:2456-2459. MacKenzie, D. I., and J. D. Nichols. 2004. O ccupancy as a surrogate for abundance estimation. Animal Biodiversity and Conservation 27:461-467. Maffei, L., E. Cuellar, and A. Noss. 2004. One thousand jaguars (Panthera onca ) in Bolivia's Chaco? Camera trapping in the Kaa-Iya National Park. Journal of Zoology 262:295-304. Marker, L. L., and A. J. Dickman. 2005. Factors affecting leopard ( Panthera pardus ) spatial ecology, with particular refe rence to Namibian farmlands South African Journal of Wildlife Research 35:105-115. Mestre, F. M., J. P. Ferreira, and A. Mira. 2007. Modelling the distri bution of the European Polecat Mustela putorius in a Mediterranean agricultural landscape. Revue D Ecologie-la Terre et la Vie 62:35-47.

PAGE 130

130 Meynard, C. N., and J. F. Quinn. 2007. Predicting species distributions: a critical comparison of the most common statistical models using artificial species. J ournal of Biogeography 34:1455-1469. Mladenoff, D. J., and T. A. Sickley. 1998. A ssessing potential gray wo lf restoration in the northeastern United States: A spatial prediction of favorable habitat and potential population levels. Journal of Wildlife Management 62:1-10. Mooley, D. A., B. Parthasarathy, K. Rupa Kuma r, N. A. Sontakke, A. A. Munot, and D. R. Kothawale. 2007. IITM Indian regional/subdiv isional Monthly Rainfall data set (IITMIMR). Muckenhirn, N., and J. F. Eisenberg. 1973. Ranges and predation in the Ceylon leopard. Pages 142-175 in R. L. Eaton, editor. The Worlds Cats. Vol 1: Ecology and Conservation. World Wildlife Safari, Winstone, Oregon. Mukherjee, S., S. P. Goyal, and R. Chellam. 1994. Standardizati on of Scat Analysis Techniques for Leopard (Panthera pardus) in Gir-National-Park, Western India. Mammalia 58:139143. Ngoprasert, D., A. J. Lynam, and G. A. Gale 2007. Human disturbance affects habitat use and behaviour of Asiatic leopard Panthera pardus in Kaeng Krachan National Park, Thailand. Oryx 41:343-351. Nowell, K., and P. Jackson. 1996. Wild Cats: status survey and Conservation Plan, Gland, Switzerland. O'Brien, T. G., M. F. Kinnaird, and H. T. Wibisono. 2003. Crouching tigers, hidden prey: Sumatran tiger and prey populations in a tr opical forest landscape. Animal Conservation 6:131-139. Odden, M., and P. Wegge. 2005. Spacing and activity patterns of leopards Panthera pardus in the Royal Bardia National Park, Nepal. Wildlife Biology 11:145-152. Ogutu, J. O., N. Bhola, H. P. Piepho, and R. Reid. 2006. Efficiency of stripand line-transect surveys of African savanna mammals. Journal of Zoology 269:149-160. Ogutu, J. O., and N. Owen-Smith. 2005. Oscillat ions in large mammal populations: are they related to predation or rainfall? African Journal of Ecology 43:332-339. Olivier, F., and S. J. Wotherspoon. 2006. Modelling habitat sele ction using pres ence-only data: Case study of a colonial hollow nesting bi rd, the snow petrel. Ecological Modelling 195:187-204. Otis, D. L., K. P. Burnham, G. C. White, and D. R. Anderson. 1978. Statistical inference from capture data on closed populat ions. Wildlife Monographs 62:1-135.

PAGE 131

131 Panwar, H. S. 1979. A note on tiger census te chniques based on pug mark tracings. Indian Forester:18-36. Parmenter, R. R., T. L. Yates, D. R. Anderson, K. P. Burnham, J. L. Dunnu m, A. B. Franklin, M. T. Friggens, B. C. Lubow, M. Miller, G. S. Olson, C. A. Parmenter, J. Pollard, E. Rexstad, T. M. Shenk, T. R. Stanley, and G. C. White. 2003. Small-mammal density estimation: A field comparison of grid-b ased vs. web-based density estimators. Ecological Monographs 73:1-26. Pianka, E. R. 1973. The Structure of Lizard Communities. Annual Review of Ecology and Systematics 4:53-74. Pienaar, U. V. 1969. Predator-prey relationshi ps amongst the larger mammals of the Kruger National Park. Koedoe 12:108-176. Prater, S. H. 1980. The book of Indian animals. Bombay Natural History Society, Bombay. Rabinowitz, A. R. 1989. The density and behaviour of large cats n a dry tropical forest in Huai Kha Khaeng Wildlife Sanctuary, Thailand. Natu ral History Society Bulletin Siam Society 37:235-251. Reddy, H. S., C. Srinivasulu, and K. T. Ra o. 2004. Prey selection by the Indian tiger ( Panthers tigris tigris ) in Nagarjunasagar Srisailam Tiger Reserve, India. Mammalian Biology 69:384-391. Reutter, B. A., V. Helfer, A. H. Hirzel, and P. Vogel. 2003. Modelling ha bitat-suitability using museum collections: an exam ple with three sympatric Apodemus species from the Alps. Journal of Biogeography 30:581-590. Rexstad, E., and K. P. Burnham. 1991. Use rs guide for interac tive program CAPTURE. Abundance estimation of closed animal populations. Colorado State University, Fort Collins, Colorado, USA 29. Riley, S. J., and R. A. Malecki. 2001. A lands cape analysis of cougar distribution and abundance in Montana, USA. Environmental Management 28:317-323. Sankar, K., S. P. Goyal, and Q. Qureshi. 2005. Assessment of status of tiger ( Panthera tigris) in Sariska Tiger Reserve, Rajasthan: A Report su bmitted to the Project Tiger, Ministry of Environment & Forests, Govt. of India, New Delhi. Wildlife Institute of India, Dehra Dun Sankar, K., and A. J. T. Johnsingh. 2002. Food habits of Tiger ( Panthera tigris ) and Leopard ( Panthera pardus) in Sariska Tiger Reserve, Rajasthan, India, as shown by scat analysis. Mammalia 66:285-289.

PAGE 132

132 Santiapillai, C., M. R. Chambers, and N. Ishwaran. 1982. The Leopard Panthera pardus-fusca (Meyer 1794) in the Ruhuna National-Park, Sri Lanka, and Observations Relevant to Its Conservation. Biological Conservation 23:5-14. Santos, X., J. C. Brito, N. Sillero, J. M. Pleguezu elos, G. A. Llorente, S. Fahd, and X. Parellada. 2006. Inferring habitat-suitability areas with ecological modelling techniques and GIS: A contribution to assess the conservation status of Vipera latastei Biological Conservation 130:416-425. Sathyakumar, S. 1992. Food habits of leopard ( Panthera pardus ) on Mundanthurai plateau, Tamil Nadu, India. Tiger Paper XX:8-9. Schaller, G. B. 1967. The Deer and the Tiger. University of Chicago Press. Schaller, G. B. 1976. The Serengeti Lion: A Stu dy of Predator-Prey Relations. University Of Chicago Press. Schaller, G. B. 1977. The mountain monarchs: wild sheep and goats of the Himalaya. University of Chicago Press, Chicago. Scognamillo, D., I. E. Maxit, M. Sunquist, a nd J. Polisar. 2003. Coexistence of jaguar (Panthera onca ) and puma (Puma concolor ) in a mosaic landscape in the Venezuelan llanos. Journal of Zoology 259:269-279. Seidensticker, J. 1976. On the Ecological Sepa ration between Tigers and Leopards. Biotropica 8:225-234. Seidensticker, J., M. Sunquist, and C. McDouga l. 1990. Leopards living at the edge of Royal Chitwan National park. Pages 415-423 in J. C. Daniel, and J. S. Serrao, editors. Conservation in Developing Countries: Probl ems and Prospects Bombay Natural History Society, Bombay. Sekhar, N. U. 1998. Crop and livestock depredation caused by wild animal s in protected areas: the case of Sariska Tiger Reserve, Rajast han, India. Environmental Conservation 25:160171. Seoane, J., J. H. Justribo, F. Garcia, J. Retamar, C. Rabadan, and J. C. Atienza. 2006. Habitatsuitability modelling to a ssess the effects of land-use changes on Dupont's lark Chersophilus duponti : A case study in the Layna Impor tant Bird Area. Biological Conservation 128:241-252. Sever, Z., and H. Mendelssohn. 1991. Spatial movement patterns of porcupines (Hystrix indica ). Mammalia 55:187-205.

PAGE 133

133 Silver, S. C., L. E. T. Ostro, L. K. Marsh, L. Maffei, A. J. Noss, M. J. Kelly, R. B. Wallace, H. Gomez, and G. Ayala. 2004. The use of camera traps for estimating jaguar Panthera onca abundance and density using capture/recapture analysis. Oryx 38:148-154. Smith, J. L. D., C. McDougal, and D. Miquell e. 1989. Scent marking in free-ranging tigers, Panthera tigris. Animal Behaviour 37:1-10. Soisalo, M. K., and S. M. C. Cavalcanti. 2006. Estimating the density of a jaguar population in the Brazilian Pantanal using camera-traps a nd capture-recapture sampling in combination with GPS radio-telemetry. Biological Conservation 129:487-496. Spalton, J. A., H. M. al Hikmani, D. Willis, and A. S. B. Said. 2006. Critically Endangered Arabian leopards Panthera pardus nimr persist in the Jabal Samhan Nature Reserve, Oman. Oryx 40:287-294. Stoen, O. G., and P. Wegge. 1996. Prey selection and prey removal by tiger (Panthera tigris) during the dry season in lo wland Nepal. Mammalia 60:363-373. Sunquist, M., K. U. Karanth, and F. Sunquist. 1999. Ecology, behaviour and resilience of the tiger and its conservation needs. Pages 5-18 in J. Seidensticker, editor. Riding the tiger: tiger conservation in human-dominated la ndscapes. Cambridge University Press. Sunquist, M. E. 1981. The Social Organization of Tigers ( Panthera tigris ) in Royal Chitawan National Park, Nepal. Smithsonian Institution Press. Sunquist, M. E., and F. Sunquist, editors. 2002. Wild cats of the world. University of Chicago Press. Taber, A. B., A. J. Novaro, N. Neris, and F. H. Colman. 1997. The food habits of sympatric jaguar and puma in the Paraguayan Chaco. Biotropica 29:204-213. Thomas, L., J. L. Laake, S. Strindberg, F. F. C. Marques, S. T. Buckland, D. L. Borchers, D. R. Anderson, K. P. Burnham, S. L. Hedley, J. H. Pollard, J. R. B. Bishop, and T. A. Marques. 2006. Distance 5.0 Release 2. Research Unit for Wildlife Population Assessment, University of St. Andrews, UK. Titeux, N., M. Dufrene, J. Radoux, A. H. Hi rzel, and P. Defourny. 2007. Fitness-related parameters improve presence-on ly distribution modelling for conservation practice: The case of the red-backed shrike. Biological Conservation 138:207-223. Traill, L. W., and R. C. Bigalke. 2007. A pr esence-only habitat suit ability model for large grazing African ungulates and its utility for wildlife management. African Journal of Ecology 45:347-354.

PAGE 134

134 Varman, K. S., and R. Sukumar. 1995. The Line Transect Method for Estimating Densities of Large Mammals in A Tropical Deciduous Forest An Evaluation of Models and Field Experiments. Journal of Biosciences 20:273-287. Ward, A. I., P. C. L. White, and C. H. Critchley. 2004. Roe deer Capreolus capreolus behaviour affects density estimates from distance sampling surveys. Mammal Review 34:315-319. Wegge, P., C. P. Pokheral, and S. R. Jnawali. 20 04. Effects of trapping effort and trap shyness on estimates of tiger abundance from came ra trap studies. Animal Conservation 7:251-256. Wilson, D., and R. Purushothaman. 2003. Dr eaming With BRICs: The Path to 2050. Global Economics Paper No: 99. Goldman Sachs. Wilson, K. R., and D. R. Anderson. 1985. Ev aluation of Two Density Estimators of Small Mammal Population Size. Journal of Mammalogy 66:13-21. Woodroffe, R. 2000. Predators and people: using human densities to interpret dec lines of large carnivores. Animal Conservation 3:165-173. Woolf, A., C. K. Nielsen, T. Weber, and T. J. Gibbs-Kieninger. 2002. Statewide modeling of bobcat, Lynx rufus habitat in Illinois, USA. Biological Conservation 104:191-198. Zaniewski, A. E., A. Lehmann, and J. M. C. Overton. 2002. Predicting species spatial distributions using presence-only data: a ca se study of native New Zealand ferns. Ecological Modelling 157:261-280.

PAGE 135

BIOGRAPHICAL SKETCH Advait Edgaonkar was born in 1970 in Pune, Indi a. He got his B.Sc. degree in zoology and biochem istry from St. Xaviers College, Mumbai. He then obtained a post graduate diploma in forestry management from the Indian Institute of Forest Management, Bhopal and an M.Sc. in wildlife science from the Wildlife Institute of India, Dehradun. He is married to Vinatha Viswanathan. Advait hopes to spend the rest of his life doing wildlife research in India.