CARNIVORES RESEARCH IN 2007-2020 WESTERN FOREST COMPLEX
ค ำน ำ
การอนุรักษ์และจัดการสัตว์ป่าและถิ่นอาศัยนั้น จ าเป็นต้องด าเนินการภายใต้ข้อมูลพื้นฐาน ทางวิชาการ เพื่อให้การวางแผนการอนุรักษ์และการจัดการสัตว์ป่ามีแนวทางการด าเนินงานที่มีประสิทธิภาพ เหมาะสมกับสัตว์ป่าแต่ละชนิดและในแต่ละพื้นที่ การศึกษาวิจัยด้านสัตว์ป่าเพื่อให้ทราบข้อมูลทางด้านประชากร ชีววิทยา นิเวศวิทยา ลักษณะถิ่นอาศัยของสัตว์ป่า จึงเป็นข้อมูลที่ส าคัญส าหรับก าหนดและวางแผนการอนุรักษ์ สัตว์ป่า เพื่อให้การบริหารจัดการทรัพยากรธรรมชาติของประเทศมีประสิทธิภาพยิ่งขึ้น กลุ่มงานวิจัยสัตว์ป่า ภายใต้การด าเนินงานของสถานีวิจัยสัตว์ป่าเขานางร า ซึ่งเป็นหน่วยงานใน ส านักอนุรักษ์สัตว์ป่า กรมอุทยานแห่งชาติ สัตว์ป่า และพันธุ์พืช ได้รับมอบหมายภารกิจส าคัญในการสนับสนุน ข้อมูลวิชาการเพื่อช่วยงานจัดการด้านสัตว์ป่าและถิ่นอาศัยในพื้นที่ป่าอนุรักษ์ โดยสถานีวิจัยสัตว์ป่าเขานางร า ได้ด าเนินโครงการศึกษาวิจัยต่าง ๆ เพื่อศึกษาความหลากหลายทางชีวภาพ ชีววิทยาและนิเวศวิทยาของสัตว์ป่า ในพื้นที่ป่าอนุรักษ์ รวมถึงการส ารวจและติดตามสถานภาพสัตว์ป่าที่ส าคัญในพื้นที่มรดกโลกทางธรรมชาติ เขตรักษาพันธุ์สัตว์ป่าทุ่งใหญ่ - ห้วยขาแข้ง ซึ่งเป็นส่วนหนึ่งของผืนป่าตะวันตก เป็นพื้นที่ที่มีความหลากหลาย ทางชีวภาพและความชุกชุมของสัตว์ป่าสูง ทั้งกลุ่มสัตว์เลี้ยงลูกด้วยนม นก สัตว์เลื้อยคลาน และสัตว์สะเทินน้ า สะเทินบก อีกทั้งยังเป็นแหล่งอาศัยที่ส าคัญของสัตว์ป่าที่ใกล้สูญพันธุ์ และมีความส าคัญต่อระบบนิเวศในกลุ่ม ของสัตว์ผู้ล่าขนาดใหญ่ ทั้งเสือโคร่ง เสือดาว รวมถึงสัตว์กีบขนาดใหญ่ที่เป็นเหยื่อของสัตว์ผู้ล่าเหล่านั้น ผลงานวิจัยที่ได้ด าเนินการมาอย่างต่อเนื่องเป็นระยะเวลามากกว่า 10 ปี เป็นที่ยอมรับทั้งในประเทศและระดับ นานาชาติ และมีส่วนส าคัญในการสนับสนุนการบริหารจัดการพื้นที่มรดกโลกทางธรรมชาติเขตรักษาพันธุ์สัตว์ป่า ทุ่งใหญ่ - ห้วยขาแข้ง กรมอุทยานแห่งชาติ สัตว์ป่า และพันธุ์พืช หวังเป็นอย่างยิ่งว่า หนังสือ “Carnivore Research in Western Forest Complex 2007 - 2020” ฉบับนี้ ซึ่งเป็นการรวบรวมผลงานวิจัยด้านสัตว์ป่าในพื้นที่เขตรักษาพันธุ์สัตว์ป่า ห้วยขาแข้ง - ทุ่งใหญ่นเรศวร ที่ได้รับการตีพิมพ์และเผยแพร่เพื่อเป็นข้อมูลในการอนุรักษ์และการจัดการสัตว์ป่า ทั้งในระดับประเทศและระดับนานาชาติ จะเป็นแนวทางการด าเนินงานวิจัยด้านสัตว์ป่าให้กับพื้นที่ป่าอนุรักษ์อื่น ๆ รวมถึงเป็นการรวบรวมข้อมูลให้สะดวกต่อการค้นหาและน ามาใช้ เพื่อน าไปสู่การวางแผนการจัดการเพื่ออนุรักษ์ และคุ้มครองสัตว์ป่าและถิ่นอาศัยในพื้นที่มรดกโลกทางธรรมชาติเขตรักษาพันธุ์สัตว์ป่าทุ่งใหญ่ - ห้วยขาแข้ง อย่างมีประสิทธิภาพต่อไป
นายธัญญา เนติธรรมกุล อธิบดีกรมอุทยานแห่งชาติ สัตว์ป่า และพันธุ์พืช
INTRODUCTION
Conservation and management of wildlife and their habitats require basic academic research on planning and guidelines for each species and area. Wildlife research on population, biology, ecology and habitat are important to wildlife conservation and natural resource management planning. Wildlife Research Division which administrate Khao Nang Rum Wildlife Research Station, under Department of National Parks, Wildlife and Plant Conservation, is established to support academic research for management of wildlife and their habitats in protected areas. Khao Nang Rum Wildlife Research Station focuses research on biodiversity, biology and ecology of wildlife, including key wildlife species surveying and monitoring in Thung Yai- Huai Kha Khaeng Wildlife Sanctuaries, one of the World Heritage sites. It is an area with high biodiversity and abundance of wildlife including mammals, birds, amphibians and reptiles, also the endangered species habitats. There are mainly large predators in the ecosystem such as tigers and leopards along with their large preys. The contribution of Khao Nang Rum Wildlife Research Station, which has continually been done for more than 10 years and accepted by international research community, also supports the management of Thung Yai-Huai Kha Khaeng Wildlife Sanctuary World Heritage. Department of National Parks, Wildlife and Plant Conservation hopefully that the publish of “Carnivores Research in Western Forest Complex 2007 – 2020 ”, a collection of wildlife research in Thung Yai-Huai Kha Khaeng Wildlife Sanctuaries, will be distribute and becomes a guideline research in other protected areas. Additionally, this easy-to-access collection of research can be used in effective management planning for wildlife conservation and protection in Thung Yai-Huai Kha Khaeng Wildlife Sanctuary World Heritage.
Mr. Thanya Netithammakun Director General Department of National Parks, Wildlife and Plant Conservation
CONTENTSCONTENTS
Page
Monitoring of the Leopard Population at Khao Nang Rum In Huai Kha Khaeng 1 Wildlife Sanctuary Saksit Simcharoen and Somphot Duangchantrasiri
How many tigers Panthera tigris are there in Huai Kha Khaeng Wildlife Sanctuary, 14 Thailand? An estimate using photographic capture-recapture sampling Saksit Simcharoen, Anak Pattanavibool, K. Ullas Karanth, James D. Nichols and N. Samba Kumar
Home range size and daytime habitat selection of leopards in Huai Kha Khaeng 21 Wildlife Sanctuary, Thailand Saksit Simcharoen, Adam C.D. Barlow, Achara Simcharoen, and James L.D. Smith
Female tiger Panthera tigris home range size and prey abundance: important 30 metrics for management Achara Simcharoen, Tommaso Savini, George A. Gale, Saksit Simcharoen, Somphot Duangchantrasiri, Somporn Parkpien and James L.D. Smith
Ecological factors that influence sambar (Rusa unicolor) distribution and abundance 38 in western Thailand: implications for tiger conservation Achara Simcharoen, Tommaso Savini, George A. Gale, Erin Roche, Vijak Chimchome and James L. D. Smith
Non-Panthera cat records from big cat monitoring in Huai Kha Khaeng Wildlife 45 anctuary Saksit Simcharoen, Mayuree Umponjan, Somphot Duangchantrasiri and Anak Pattanavibool
Dynamics of a low-density tiger population in Southeast Asia in the context 50 of improved law enforcement Somphot Duangchantrasiri, Mayuree Umponjan, Saksit Simcharoen, Anak Pattanaviboo, Soontorn Chaiwattana, Sompoch Maneerat, N. Samba Kumar, Devcharan Jathanna, Arjun Srivathsa, and K. Ullas Karanth
A context-sensitive correlated random walk: A new simulation model for movement 60 Sean C. Ahearn, Somayeh Dodge, Achara Simcharoen, Glenn Xavier And James L.D. Smith
Ecological Covariates at Kill Sites Influence Tiger (Panthera tigris) Hunting Success 82 in Huai Kha Khaeng Wildlife Sanctuary, Thailand Somporn Pakpien, Achara Simcharoen, Somphot Duangchantrasiri, Vijak Chimchome, Nantachai Pongpattannurak and James L. D. Smith
CONTENTSCONTENT S
Tiger and leopard diets in western Thailand: Evidence for overlap and potential 89 consequences Achara Simcharoen, Saksit Simcharoen, Somphot Duangchantrasiri, Joseph Bump and James L.D. Smith
Impact of prey occupancy and other ecological and anthropogenic factors on 95 tiger distribution in Thailand's western forest complex Somphot Duangchatrasiri, Pornkamol Jornburom, Sitthichai Jinamoy, Anak Pattanvibool, James E. Hines, Todd W. Arnold John Fieberg and James L. D. Smith
Weights of gaur (Bos gaurus) and banteng (Bos javanicus) killedby tigers in Thailand 105 Supawat Khaewphakdee, Achara Simcharoen, Somphot Duangchantrasiri Vijak Chimchome, Saksit Simcharoen and James L. D. Smith
Spatial and temporal analysis of leopards (Panthera pardus), their prey and tigers 113 (Panthera tigris) in Huai Kha Khaeng Wildlife Sanctuary, Thailand Apinya Saisamorn, Prateep Duengkae, Anak Pattanavibool, Somphot Duangchantrasiri, Achara Simcharoen and James L.D. Smith
Diet composition of the golden jackal (Canis aureus L.) during the dry season 123 in west-central Thailand Saksit Simcharoen, Achara Simcharoen, Sasitorn Hasin, Francesca Cuthbert and J. L. David Smith
Diet of the Large Indian Civet (Viverra zibetha L., 1758) in west-central Thailand 135 Saksit Simcharoen, Achara Simcharoen, Sasitorn Hasin, Francesca Cuthbert and J. L. David Smith
1
Thai J. For. 27 : 68-80 (2008) ÇÒÃÊÒÃǹÈÒʵà 27 : 68-80 (2551)
Original article
Monitoring of the Leopard Population at Khao Nang Rum In Huai Kha Khaeng Wildlife Sanctuary
Saksit Simcharoen1 Somphot Duangchantrasiri1
1 Research Division, Department of National Park, Wildlife and Plant Conservation, Pahonyotin Road, Chatuchak, Bangkok 10900 Received: November 20, 2007 Accepted: June 18, 2008
ABSTRACT
Leopard population was monitered using capture-recapture technique. Camera traps were set up for 3 years (1996-1999) over a 115.88 km2 area (A) around Khao Nang Rum Wildlife Research Station, Huai Kha Khaeng Wildlife Sanctuary. Eighteen leopards were photographed including 4 adult females, 3 sub-adult females, 3 adult males, 3 sub-adult males and 5 males that could not be identified by age class. Four black leopards were individually identified: one leopard was collared and 3 leopards were identified by their size, sex, photograph time and locations. Closure test indicated that the leopard population was closed (p>0.05). The estimation of population size of leopard using model Mh are 10, 10 and 11 leopards for three sessions respectively. Log- normal-based 95% confidence interval ranged from 10 to 29 leopards for the first session, from 9 to 17 leopards for the second session and from 11 to 26 leopards for the third session. Estimated leopard densities were decreased from 7.88 ± 5.82 to 5.21 ± 3.12 and 4.86 ± 2.29 leopards/ 100 km2 respectively. The average leopard density in this study was 5.98 leopards/ 100 km2. This highly leopard density indicated that the Western Forest Complex is an important area for leopard conser- vation. Moreover, data on the abundance and density of leopard will help us to understand their present status in the study area and evaluate habitat quality and success of the management.
Keywords: Leopard, Panthera pardus, Population, Huai Kha Khaeng Wildlife Sanctuary
INTRODUCTION
The leopard is an important predator loss, depletion of prey and direct hunting. It has in the tropical rainforest ecosystem. It plays an been listed on Appendix I of the Convention of important role in predator-prey interactions. It International Trade of Endangered Species of regulates prey populations and maintains Wild Fauna and Flora (CITES) since 1983. natural selection by influencing prey behavior Monitoring of leopard population is important and directly affects the vigor of prey populations to the planning and management to maintain and indirectly, the health of the whole ecosystem. them. However, leopard is an endangered large felid Leopard is secretive and usually noc- which shows a declining trend due to habitat turnal animals. Thus, methods for estimating 2
ÇÒÃÊÒÃǹÈÒʵà 2 : 68-80 (2551) 69 abundance of wild cats such as tiger and manage and conserve the leopard in the wild. leopard in the past used counting tracks of Data on the abundance and density of leopard individual animal identified by measurement of will help us to understand their present status their tracks (Riordan, 1998) and track shape in the study area and evaluate habitat quality (Panwar, 1979). However, this method has and success of the management. These factors serious problems associated with field surveys are the principal objectives of this study. and some assumptions are erroneous (Karanth 1987, 1988, 1995). Capture-recapture is a common approach MATERIALS AND METHODS to investigating the abundance of marked animals; capture histories of animals can be Study Area used for determining the number of undetected The study was carried out in the forests animals present. There are statistical models to around Khao Nang Rum Wildlife Research be employed for estimating the both of open Station (N. Lat. 15º25´-15º31´ and E. Long. and closed populations. In a closed population, 99º15´-99º20´) within Huai Kha Khaeng Wildlife it is assumed that there are no births/deaths/ Sanctuary (Figure 1), which forms part of an 2 immigration/emigration during the time of the 18,727 km protected area network known as survey. the Western Forest Complex. The area has rugged, Camera trapping is probably the best hilly topography, mean annual temperature in possible available technique to build up capture 20 years from 1986 to 2005 was 24.64ºc and histories that can be analyzed in readily available annual rainfall of 1447 mm. The area supports capture-recapture software such as CAPTURE four vegetation type dry evergreen forest, hill (Otis et al., 1978; White et al., 1982) and evergreen forest, mixed deciduous forest and MARK (White and Burnham, 1999). This approach dry dipterocarp forest depending on rainfall already has been used for leopards in many patterns and edaphic factors (Srikosamatara, area (Jenny, 1996; Stander et al., 1997; 1993; Tunhikorn et al., 2004). Khorozyan, 2003; Kostyria et al., 2003; Spalton Camera Trapping et al., 2006) and also has been applied to other This study was begun to document cryptic felids such as tiger (Panthera tigris) the presence of leopards and other mammals. (Karanth, 1995; Karanth and Nichols, 1998; Consequently camera trapping was done on Kawanishi and Sunquist, 2004), jaguar an ad hoc basis, without strictly following optimal (Panthera onca) (Silveira et al., 2003; Wallace survey protocols (Nichols and Karanth, 2002). et al., 2003; Maffei et al., 2004; Silver, 2004), However, The following field protocols were snow leopard (Uncia uncia) (Spearing, 2002) helpful for analyzing the data within a formal and ocelot (Trolle and Kerv, 2003). Moreover, capture-recapture framework. they are linked with movement, distribution, Camera trapping was carried out for activity patterns, habitat use and reproductive 3 years (1996-1999) over a 115.88 km2 area information that are important for wildlife (A) around Khao Nang Rum Wildlife Research conservation (Spalton et al., 2006; Silveira et Station (Figure 2), Huai Kha Khaeng Wildlife al., 2003). Sanctuary. The area (A) was determined by a The majority of past research has taken minimum convex polygon around all camera place in sub-Saharan Africa, India, Sri Lanka locations. Camera traps were deployed for 3-4 and Nepal. In Thailand, there has been little months a year. research concerning the abundance, behavior Trailmaster® (Goodson Associates, and some aspects of leopard ecology. However, Inc., Kansas, USA) camera traps were set on the detailed knowledge of leopard abundance the trails and roads where leopard tracks or has not been fully studied and is necessary to other secondary signs were frequently found. 3
70 Thai J. For. 27 : 68-80 (2008)
Figure 1. Map of Huai Kha Khaeng Wildlife Sanctuary and the study area. 4
ÇÒÃÊÒÃǹÈÒʵà 2 : 68-80 (2551) 71 Camera trap sites, camera trap areas, and buffer areas. buffer areas, and trap camera sites, trap Camera Figure 2. 5
72 Thai J. For. 27 : 68-80 (2008)
Camera traps were mounted on trees 3-4 m size using the computer program CAPTURE from the path, with the infrared beam set 45 (Otis et al., 1978; White et al., 1982; Rexstad cm above the ground (Karanth, 1995; Karanth and Burnham, 1991). The program estimates and Nichols, 1998; Kostyria, 2003). Camera abundance of closed population under seven traps were set out in pairs to capture the opposing, models (Mh, Mb, Mt, Mbh, Mth, Mtb, Mtbh that asymmetrical spot patterns of the passing vary by h-heterogeneity, b-behavior, t-time) leopards (Karanth, 1995). The distance between using M0, in which probabilities are constant is unit pairs was 1.0-3.0 km, with GPS locations null the hypothesis (White et al., 1982). In taken for all sites. Cameras were checked CAPTURE, the closure test for the number of every 2-3 days to change batteries and film individual is constant during the overall study and to document the presence of animal tracks. period and is computed for the subset of the Leopard Identification data defined by the capture frequencies. However, Leopards are patterned animals that the validity test can not be devised because of we can individually identify from their natural problems in behavioral responses and time markings. Rosettes and spot patterns of individual trends (White et al., 1982). However, we assumed leopard are asymmetrical on the two flanks. Both that the population closed because the short sides of each animal had to be photographed period. The study would end before significant simultaneously. The photographs are needed immigration could occur. In this study, we used for clear identification. However the obtained, model Mh in which capture histories vary by photographs viewed only on one side, can be individual heterogeneity because the estimator ˆ linked to the left or right profile from camera for model Mh (the jackknife, Nh ) is the most ˆ ˆˆˆ trapping and radio collared leopard to identify robust of the five estimators ( Nbhbt,,, NNN0 ) individual leopards. Unclear animal photographs (White et al., 1982). In addition, this model was on one side were not used for analysis. Black widely used for estimate abundance for large leopards could not be identified clearly by their cats (Karanth and Nichols, 1998; Kostyria et spot pattern. However, they were identified by al., 2003) that are territorial animals in which their size, sex and photograph locations which home range size and trapping depend on social relate to photograph time. position and spatial location of the animal Animal photographs were organized (Karanth, 1995). separately for each individual leopard and the Estimating Density date, time, photograph location, age class and Animal density can be estimated using sex were recorded. Leopards were given all abundance estimates from CAPTURE. Density ID from their collar number or new ID was is defined as: created for non-collared animal. Estimating Abundance Dˆˆ= N/() AWˆ (1) The capture history was used to describe capture frequencies. Data were recorded in Dˆ is the estimated animal density, Nˆ an X matrix consisting of i animals in rows and is the estimated abundance and AWˆ () is the t trapping occasions in columns, assuming a estimated effective area in which photographed value of either "0" if the animal was not photo- animals live; W is the boundary strip of width graphed or a "1" if it was photographed. We that is added around the perimeter of the area used five trapping occasions for all sites with in which the camera traps were set. each occasion containing 3 day. The capture Variance in animal density is calculated histories of individual leopards were used in from variance in estimated abundance and the framework of capture-recapture theory to effective area: estimate capture probabilities and population 6
ÇÒÃÊÒÃǹÈÒʵà 27 : 68-80 (2551) 73
⎡⎤115.88 km2. The research activition for each varˆ AWˆ 2 ⎢⎥()() varˆ ()N var(ˆ DDˆˆ ) =+ session were carried out for 3-4 months during ⎢⎥⎡⎤2 ˆ 2 (2) the dry season. A total of 2,094 trap nights ⎢⎥AWˆ N ⎣⎦⎣⎦() included 650 trap nights between December Variance in estimated abundance was 1996 and March 1997, 620 trap nights between computed by program CAPTURE. The effective December 1997 and February 1998 and 824 area includes the area where camera trap sites trap nights between December 98 and March 99. This traps sampled each location, on an were plotted and connected on the edge to form average for about 15 days. Leopards were the perimeter and the buffer area. The buffer detected 106 times during 2,094 trap-nights; 43, width (Wˆ ) was calculated using the mean ˆ 39 and 24 times for the three sessions respectively maximum distance moved by leopards ( d ) (Table 1). between camera trap sites divided by two Abundance (Karanth and Nichols, 1998). Eighteen leopards were captured including 4 adult females, 3 sub-adult females, ˆ ˆ (3) W = d / 2 3 adult males, 3 sub-adult males and 5 males Variance in buffer width was calculated that could not be identified by age class. From using variance in mean maximum distance a total of 27 black leopard photographs, 15 moved by leopards. Let d denote the maximum photographs were separated for 4 individual i leopards; one leopard had a collar and 3 leopards distance moved between camera trap sites for were identified by their size, sex, photographed animal i and m denote the number of animals time and locations. L055 and L195 were that cameras trapped at least twice: captured every session. Five leopards were ˆ varˆˆ()Wdˆ =× 0.25 var (4) captured for two sessions and eleven leopards ( ) were capture during only one session. 2 m ˆ Capture frequency between sexes = dd11− ∑ i ) was difference: males were photographed 1.81 varˆ dˆ = ( (5) ( ) mm()−1 times more than females. The mean capture rate of males was 0.77 detections/ 100 trap Variance in estimated effective area was nights for adults and 0.18 detections/100 trap calculated following Karanth and Nichols (1998): nights for sub-adults. The mean female capture 2 rate was 0.43 and 0.10 detections/100 trap- ˆ 2 ˆˆ varˆˆ()AW()=+ 4π ( c W ) var () W (6) nights for adults and sub-adults respectively. The average number of photograph locations This estimation approximates each of the for individual leopards were 0.56, 0.28, 0.13 and sampled areas as a circle in which c is a constant 2 0.10 locations/100 trap nights for adult males, ˆ ˆ that is calculated from AW()=+π c W adult females, sub-adult males and sub-adult ˆ () ; AW() is the estimated effective area that females respectively. Both mean capture rate includes the camera trap area and the buffer and number of photograph locations ranked from area and Wˆ is the buffer width. most to least value as determined by number of detections and photographed locations per 100 RESULTS AND DISCUSSION trap-nights are: adult male, adult female, sub-adult male and sub-adult female. The highest capture Camera Trap Effort rate male was L055 which was photographed Camera traps were set up around 1.88 detections/100 trap-nights; L390, the Khao Nang Ram WRS ranging from 39 to 56 highest capture rate female, was photographed locations per session and covered an area of 0.61 detections/100 trap-nights. 7
74 Thai J. For. 27 : 68-80 (2008) Table 1. Summary statistics for camera trap data on leopards in HKK Table 8
ÇÒÃÊÒÃǹÈÒʵà 2 : 68-80 (2551) 75
Closure tests indicated that leopard area and the boundary strip were 126.93, population was closed (Table 2). The estima- 191.76 and 226.44 km2 (Figure 2). In these areas, tion of population size of leopard using model the estimated population sizes were 10, 11 and 11
Mh are 10, 10 and 11 leopards for the three leopards respectively. We used these estimates sessions respectively. Log-normal-based 95% in conjunction with Equation 1 to estimate leopard confidence interval ranged from 10 to 29 leop- density. We found that the estimated leopard ards for the first session, from 9 to 17 leopards densities decreased from 7.88 ± 5.82 to for the second session and from 11 to 26 leop- 5.21 ± 3.12 and 4.86 ± 2.29 leopards/ 100 km2 ards for the third session. Estimated capture respectively (Table 4). The average leopard probability over all sampling occasions varied density in this study was 5.98 leopards/ 100 km2. between 0.80 and 0.91. Estimated values of Discussions average capture probability were varied ranging Results of the study indicated that, 126 from 0.33 to 0.44 (Table 2). photographs of leopards during 2,094 trap-nights Density (6.02 photographs/100 trap-nights) which was For the three sessions, the camera trap higher than the ones which were obtained from areas were 42.05, 57.86 and 96.93 km2 respec- the study on leopard in Kaeng Khachan National tively. The number of animals which the camera Park (KKNP), Thailand (3.25 photographs/100 trapped at least twice was six for all three trap-nights, Ngoprasert, 2004) and in Southwest sessions. Average maximum distances moved Primorski Krai (4.69 photographs/100 trap- by photographed leopards ranged from 3.19 to nights, Kostyria et al., 2003). In addition, the 4.65 km. The estimated boundary strip widths estimated capture probabilities of this study were 1.59(SD=0.30), 2.33(SD=0.40) and (0.44, 0.42 and 0.33) are higher than the results 1.81(SD=0.15) km respectively (Table 3). The were obtained from the study in KKNP (0.27, estimated effective area from the first to the using model Mh) and Southwest Primorski Krai third sessions, which included the camera trap (0.20, using model Mh).
Table 2. Estimated abundance and capture probabilities of leopards in HKK under model Mh of program CAPTURE
The average maximum distances low temperature and snow cover in winter, between capture sites, was varied from 3.19 which are extremal conditions for leopard to 4.65 km, and with the significance level which (Kostyria et al., 2003). These conditions was lower than the one obtained form the South- probably impacted the distances which were west Primorski Krai (9.70 km). Southwest higher than the ones obtained from the other Primorski Krai is in the temperate zone with study areas including KKNP and this study. 9
76 Thai J. For. 27 : 68-80 (2008)
Table 3. Calculated effective area using half the mean maximum distance moved by leopards caught on more than one occasion
Table 4. Estimated leopard density and detectable change using program CAPTURE and calculated effective area A(W)
Moreover, the obtained male leopard Estimated density using a framework photographs was 1.81 times which was more of capture-recapture averaged 5.98 leopards/ than female leopard photographs. In the same 100 km2. The densities were 7.88 ± 5.82 to way, Ngoprasert (2004) reported that capture 5.21 ± 3.12 and 4.86 ± 2.29 leopards/ 100 km2, frequency between sexes was differenced. which were close to densities using radio track- Male leopards had recapture more than female ing in the same period (6.3 leopards/100 km2). leopards (T-test, p=0.01). Santiapillai et al. The average density using radio tracking method (1982) reported that male leopards were visually were 10.1, 6.9, 3.7 and 4.5 leopard/100 km2 observed more frequently (72%) than female from 1996 to 1999 respectively. In the same leopards in Ruguna National Park parallels the times and area, both methods gave nearly observations of Muckenhirn and Eisenberg densities and with a decreasing trend. Further- (1973) in Wilpattu National Park and Bailey more the average leopard densities were greater (1993) in Kruger National Park. Bailey (1993) than in KKNP (4.78 leopards/100 km2) reported that males were captured with less (Ngoprasert, 2004) and much greater than in trapping effort than females and were observed Southwest Primorski Krai (1.2 ± 0.2 leopards/ more often (63%) than females along tourist 100 km2) (Kostryria et al., 2003). In Huai Kha roads although, females were observed more Khaeng Wildlife Sanctuary, Rabinowitz (1989) often (68%) and more than males along fire- estimated leopard density was lower (4 leopards/ break roads. He suggested that the females 100 km2) than the obtainul result from this study. avoided contact with human more than the Based on the comparison with the another males. The data support the contention that the method, the average leopard density in this study females were shyer than males (Santiapillai et was lower than in Tsavo National Park, Kenya al., 1982). (Hamilton, 1976); Cape Province, South Africa 10
ÇÒÃÊÒÃǹÈÒʵà 2 : 68-80 (2551) 77
(Norton and Henley; 1987) and Tai National covered an area of 115.88 km2 around Khao Park, Ivory Coast (Jenny, 1996) but higher than Nang Rum Research Station, Huai Kha in Sri Lanka (Clark, 1901); Wilpattu National Khaeng Wildlife Sactuary. Eighteen leopards Park, Sri Lanka (Eisenberg and Lockhart, were photographed included 4 adult females, 3 1972); Serengeti National Park, Tanzania sub-adult females, 3 adult males, 3 sub-adult (Schaller, 1972; Cavallo, 1993); Kalahari males and 5 males that could not be identified Desert, Southern Africa (Bothma and Le Riche, by age class. From a total of 27 black leopard 1984); Stellenbosch, Cape Province (Norton and photographs, 15 photographs were separated Lawson, 1985); Kruger National Park, South for 4 individual leopards; one leopard had a Africa (Bailey, 1993); and North-eastern Namibia collar and 3 leopards were identified by their (Stander et al., 1997). size, sex, photographed time and locations. The trend of leopard density decreased Closure tests indicated that leopard throughout the study period which during the population was closed. The estimates of same period the tiger density increased from population size of leopard using model M are 1.71 0.99 to 2.27 1.12 and 2.94 2.26 tigers/ h ± ± ± 10, 10 and 11 leopards for the three sessions 100 km2 respectively. Seidensticker (1976) recorded respectively. Log-normal-based 95% confidence that a tiger may have appropriated a kill from interval ranged from 10 to 29 leopards for the a leopard, and he believed that social dominance first session, from 9 to 17 leopards for the second is a major factor in tiger-leopard interaction. He found that leopard used areas not which were session and from 11 to 26 leopards for the third frequented by tiger in order to minimize their chance session. Estimated leopard densities decreased of encounter. Moreover, we found a female from 7.88 ± 5.82 to 5.21 ± 3.12 and 4.86 ± 2.29 2 black leopard was killed in the study area. The leopards/ 100 km respectively. The average 2 animal died as a result of one powerful bite to the leopard density was 6.0 leopards/ 100 km . chest over the heart which is only one clue and we found female tiger tracks around it. We believed that this leopard was killed by a tiger. ACKNOWLEDGEMENTS The average estimated density in this study can roughly estimate the leopard population We would like to express our great size in Huai Kha Khaeng Wildlife Sanctuary, appreciation to the National Park, Wildlife and which covered an area of 2780.14 km2, as 166 Plant Conservation Department for giving me leopards. If the habitat quality in the Western the opportunity to work and study at Khao Nang Forest Complex, which covered an area of Rum Wildlife Research Station. We gratefully 18,727 km2, were the as same as in this study acknowledge the late Mr. Mark Graham and area, estimated leopard population size was WWF (Thailand) for supporting the camera 1,120 leopards. The data indicated that The trap. We are particularly appreciated Prof. Western Forest Complex was an important Dr.Ullas Karanth, Director Wildlife Conser- area for leopard conservation. Moreover, data vation Society (India), for explanation and on the abundance and density of leopard will suggestion about the estimation of leopard helped us to understand their present status in population. We especially thank Mr. Onsa the study area and evaluate habitat quality and Norrasan, Mr. Precha Prommakun and Mr. success of the management. Pakawat Ponak for their help in the field study. We would like to give special thank to Ms. Ardith Eudey who helped editing our English CONCLUSION AND manuscript. Our appreciation also is due to Ms RECOMMENDATION Namkhang Saelee for completing our manu- script. We would like to thank everyone who Monitoring of leopard population was helped and supported this study. 11
78 Thai J. For. 27 : 68-80 (2008)
REFERENCES
Bailey, T. N. 1993. The African Leopard Karanth, K. U. 1987. Tiger in India: a critical Ecology and Behavior of a Solitary review of field censuses, pp. 118-131. Felid. Columbia University Press, In R. L. Tilson and U. S. Seal, eds. New York. Tigers of the World: The Biology, Bothma, J. D. P. and E. A. N. Le Riche. 1984. Biopolitics, Management and Aspects of the ecology and the Conservation of an endangered behavior of the leopard Panthera species. Noyes Publications, Park pardus in the Kalahari Desert. Ridge, NJ, USA. Supplement to Koedoe. :259-279. ______1988. Analysis of predator-prey Carbone, C., S. Christie, K. Conforti, T. balance in Bandipur tiger reserve with Coulson, N. Franklin, J. R. Ginsberg, reference to census reports. Journal M. Griffiths, J. Holden, K. Kawanishi, of the Bombay Natural History M. Kinnaird, R. Laidlaw, A. Lynam, Society 85: 1-8. D. W. MacDonald, D. Martyr, C. ______1995. Estimating tiger (Panthera McDougal, L. Nath, T. O. O’Brian, J. tigris) populations from camera-trap Seidensticker, J. L. D. Smith, M. data using capture-recapture models. Sunquist, R. Tilson, and W.N.Wan Biological Conservation 71: 333- Shahruddin. 2001. The use of 338. photographic rates to estimate ______and J. D. Nichols. 1998. Estimation densities of tigers and other cryptic of tiger densities in India using mammals. Animal Conservation 4: photographic captures and recaptures. 75-80. Ecology 79: 2852-2862. Cavallo, J. 1993. A study of leopard behaviour Kawanishi, K. and M. Sunquist. 2004. and ecology in the Seronera Valley, Conservation status of tigers in a Serengeti National park. Serengeti Wildlife Research Centre Scientific primary rainforest of peninsular Report 1990-1992. Malaysia. Biological Conservation Clark, A. 1901. Sport in the Low-country in 120: 333-348. Ceylon. Tisara Prakasakyo, Dehiwela, Khorozyan, I. 2003. Camera Photo-trapping Sri Lanka. of the Endangered Leopards Eisenberg, J.F. and M. Lockhart. 1972. An (Panthera pardus) in Armenia: a ecological reconnaissance of Wipattu Key Element of Species Status National Park, Ceylon. Smithsonian Assessment. Contributions to Zoology 101: 1- Kostyria, A. V., A. S. Skorodelov, D. G. 118. Miquelle, V.V. Aramilev, D. Hamilton, P. H. 1976. The movements of McCullough. 2003. Report on a Leopards in Tsavo National Park, census of Far-Eastern Leopards Kenya as Determined by Radio Using Camera Traps in South-west Tracking. M. Sc. Thesis, University Primorskii Province, Winter 2002- of Nairobi. 2003. Wildlife Conservation Society. Jenny, D. 1996. Spatial organization of leopards Maffei, L., E. Cuellar and A. Noss. 2004. One (Panthera pardus) in Tai National thousand jaguars (Panthera onca) in Park, Ivory Coast: is rain forest habitat Bolivia’s Chaco? camera trapping in a tropical haven? Journal of Zoology Kaa-Iya National Park. Journal of 240: 427-440. Zoology 262: 295-304. 12
ÇÒÃÊÒÃǹÈÒʵà 2 : 68-80 (2551) 79
Muckenhirn, N. A. and J. F. Eisenberg. 1973. Rexstad, E. and K. P. Burnham. 1991. User’ Home ranges and predation of Ceylon s Guide for Interactive Program leopard, pp. 142-175. In R.L. Eaton, CAPTURE: AbundanceEstimation ed. The World’s cats 1(1). World of Closed Animal Populations. Wildlife Safari, Winston, Oregon. Colorado State University, Corolado, Ngoprasert, D. 2004. Effects of Roads, USA. Selected Environmental Variables Riordan, P. 1998. Unsupervised recognition of and Human Disturbance on Asiatic individual tigers and snow leopards Leopard (Panthera pardus) in from their footprints. Animal Conservation 12 : 252-262. Kaeng Krachan National Park. M. Santiapillai, C., M. R. Chambers and N. Sc. Thesis, King Mongkut’s University Ishwaran. 1982. The leopard, of Technology Thonburi. Panthera pardus fusca (Meyer, Nichols, J. D. and K. U. Karanth, 2002. 1794), in the Ruhuna National Park, Statistical concepts: Estimating Sri Lanka, and observations relevant absolute densities of Tigers using to its conservation. Biological Capture-recapture sampling, pp. 121- Conservation 23: 5-14. 137. In K. U. Karanth and J. D. Schaller, G. B. 1972. The Serengeti Lion. Nichols, eds. Monitoring Tigers and University of Chicago Press, London. their Prey: A manual for Researchers, 480 pp. Managers and Conservationists in Seidensticker, J. 1976. On the ecological Tropical Asia. Bangalore: Centre for separation between tigers and Wildlife Studies. leopards. Biotropica 8 (4): 225-234. Norton, P. M. and A. B. Lawson. 1985. Radio Silveira, L., A. T. A. Jacomo, and J. A. F. Diniz- tracking of leopards and caracals in Filho. 2003. Camera trap, line transect the Stellenbosh area, Cape Province. census and track surveys: a South African Journal of Wildlife comparative evaluation. Biological Research 15: 17-24. Conservation 114: 351-355. ______and S. R. Henley. 1987. Home range Silver, S. C. 2004. The use of camera traps for and movements of male leopard in the estimating Jaguar (Panthera onca) Cerdarberg wilderness area, Cape abundance and density using capture- recapture analysis 38: 148-154. Province. South African Journal of . Oryx Spalton, J. A., H. M. Al Hikmani, D. Willis and Wildlife Research 17 (2): 41-48. A. S. B. Said. 2006. Critically Otis, D. L., K. P. Burnham, G. C. White and endangered Arabian leopards D. R. Anderson. 1978. Statistical Panthera pardus nimr persist in the inference from capture data on Jabal Samhan Nature Reserve, Oman. closed animal populations. Wildlife Oryx 40 (3): 287-294. Monographs 62: 1-35. Spearing, A. 2002. A note on the prospects Panwar, H. S. 1979. A note on tiger census for snow leopard census using technique based on pugmark tracings. photographic capture. Proceeding Tigerpaper 6: 16-18. of the snow leopard survival summit, Rabinowitz, A. 1989. The density and behavior Snow leopard survival summit, ISLT. of large cats in a dry tropical forest Srikosamatara, S. 1993. Density and biomass mosaic in Huai Kha Khaeng Wildlife of large herbivores and other mammals Sanctuary, Thailand. Natural History in a dry tropical forest, western Bulletin of the Siam Society 37 (2): Thaiand. Journal of Tropical 235-251. Ecology 9: 33-43. 13
80 Thai J. For. 27 : 68-80 (2008)
Stander, P. E., P. J. Haden, Kaqece and Ghau. Wallace, R. B., H. Gomez, G. Ayala and F. 1997. The ecology of asociality in Espinoza. 2003. Camera trapping for Nabibian leopards. Journal of jaguar (Panthera onca) in the Tuichi Zoology, London 242: 343-364. valley, Bolivia. Journal of Neotropical Trolle, M., and M. Kerv. 2003. Estimation of Mammals 10: 133-139. ocelot density in the pantanal using White, G. C., D. R. Anderson, K. P. Burnham capture-recapture analysis of camera and D. L. Otis. 1982. Capture- trap data. Journal of Mammalogy 84: Recapture and Removal Methods 607-614. for Sampling Closed Populations. Tunhikorn, S., J.L.D. Smith, T. Prayurasiddhi, Los Alamos National Laboratory M. Graham, P. Jackson and P. Cutter Publication LA-8787-NERP, NM, (Eds.). 2004. Saving Thailand’s USA. tigers: An action plan. Bangkok: ______and K. P. Burnham. 1999. Program Ministry of Natural Resources and MARK: Survival rate estimation from Environment. Department of National both live and dead encounters. Bird Park, Wildlife, and Plant Conservation. Study 46: 120-139. 14
Oryx Vol 41 No 4 October 2007
How many tigers Panthera tigris are there in Huai Kha Khaeng Wildlife Sanctuary, Thailand? An estimate using photographic capture-recapture sampling
Saksit Simcharoen, Anak Pattanavibool, K. Ullas Karanth, James D. Nichols and N. Samba Kumar
Abstract We used capture-recapture analyses to esti- Western Forest Complex c. 720 tigers. Although based mate the density of a tiger Panthera tigris population in on field protocols that constrained us to use sub-optimal the tropical forests of Huai Kha Khaeng Wildlife Sanctuary, analyses, this estimated tiger density is comparable to Thailand, from photographic capture histories of 15 distinct tiger densities in Indian reserves that support moderate individuals. The closure test results (z 5 0.39, P 5 0.65) prey abundances. However, tiger densities in well- provided some evidence in support of the demographic protected Indian reserves with high prey abundances closure assumption. Fit of eight plausible closed models to are three times higher. If given adequate protection we the data indicated more support for model Mh,which believe that the Western Forest Complex of Thailand incorporates individual heterogeneity in capture probabil- could potentially harbour .2,000 wild tigers, highlight- ities. This model generated an average capture probability ing its importance for global tiger conservation. The p^ 5 0.42 and an abundance estimate of NbðSEb½Nb Þ 5 19 monitoring approaches we recommend here would be (9.65) tigers. The sampled area of AbðWÞðSEb½AbðWÞ Þ 5 useful for managing this tiger population. 477.2 (58.24) km2 yielded a density estimate of DbðSEb½Db Þ 5 3.98 (0.51) tigers per 100 km2. Huai Kha Khaeng Wildlife Keywords Camera traps, capture-recapture models, Sanctuary could therefore hold 113 tigers and the entire Panthera tigris, population estimation, Thailand, tiger.
Introduction to obtain reliable estimates of tiger densities at a large number of sites across the 1.2 million km2 geographic The tiger Panthera tigris is categorized as Endangered on range of the species (Seidensticker et al., 1999). the IUCN Red List (IUCN, 2006) and during the past Thailand is a key tiger range state, with 25% of its land 3 decades substantial efforts have been invested in tiger area under forest cover, 16% of it being managed un- conservation by governments and non-governmental der wildlife and national park protection legislation agencies. However, these efforts are constrained by (Pattanavibool & Dearden, 2002). In addition, increasing a lack of reliable data on the distribution as well as societal wealth and an official commitment to science- densities of wild tiger populations. Furthermore, dis- based tiger conservation (Tunhikorn et al., 2004) make semination of putative ‘tiger numbers’ (Jackson, 1993), Thailand a critical region for tiger conservation. Conse- often based on guesswork or demonstrably faulty meth- quently, attempts have been made to map accurately the ods (Karanth, 1987, 1988; Karanth et al.,2003)masksareal distribution of tiger populations in Thailand from field scarcity of reliable data. Therefore, there is an urgent need surveys (Rabinowitz, 1993, 1999; Smith et al., 1999; Tunhikorn et al., 2004; WEFCOM, 2004). However, to
Saksit Simcharoen Wildlife Research Division, Department of National Park, use such maps for managing wild tiger populations there Plant, and Wildlife Conservation, Paholyotin Road, Chatuchak, Bangkok is an additional need to estimate densities and sizes of 10900, Thailand. individual tiger populations at specific sites. This critical
Anak Pattanavibool Wildlife Conservation Society -Thailand Program, P.O.Box need has been enunciated in Thailand’s national action 170, Laksi, Bangkok 10210, Thailand. plan for tigers (Tunhikorn et al., 2004). The national plan 2 K. Ullas Karanth (Corresponding author) and N. Samba Kumar Wildlife also identifies the 18,000 km Western Forest Complex, Conservation Society-India Program, Centre for Wildlife Studies, 26-2, Aga which contains 17 protected areas, including Huai Kha Abbas Ali Road (Apt: 430), Bangalore, Karnataka-560 042, India. E-mail Khaeng Wildlife Sanctuary, as the most important tiger [email protected] conservation area in the country. James D. Nichols US Geological Survey, Patuxent Wildlife Research Center, Although reliable estimation of tiger abundance is Laurel, Maryland 20708, USA. difficult because of their elusive behaviour and naturally Received 1 February 2006. Revision requested 25 September 2006. low densities, recent development of automated camera Accepted 20 December 2006. traps and their application within a formal framework
447 ª 2007 FFI, Oryx, 41(4), 447–453 doi:10.1017/S0030605307414107 Printed in the United Kingdom 15
448 S. Simcharoen et al.
of capture-recapture population sampling (see Karanth Kha Khaeng Wildlife Sanctuary (Fig. 1). The area is et al., 2004b, for a review) have enabled investigators to rugged and hilly over altitudes of 200-1,600 m, has an obtain rigorous density estimates in India (Karanth & annual temperature range of 10-35°C and annual pre- Nichols, 1998; Karanth et al., 2004a,c), Nepal (Wegge cipitation of c. 1,500 mm. It supports four vegetation et al., 2004), Malaysia (Kawanishi & Sunquist, 2004) and types: dry deciduous dipterocarp forests, mixed decidu- Indonesia (O’Brien et al., 2003). Unlike previous tiger ous forest, dry evergreen forest, and hill evergreen forest, monitoring approaches based on footprint total counts depending on rainfall patterns and edaphic factors (Panwar, 1980), radio-telemetry (Sunquist, 1981; Smith, (Srikosamatara, 1993; Tunhikorn et al., 2004; WEFCOM, 1993) or raw photographic trapping rates (Carbone 2004). From earlier food habit studies in the area (Petdee, et al., 2001), capture-recapture methods can effectively 2000), principal prey species of tigers are wild pig Sus deal with the typical inability of surveys to detect all scrofa,sambarCervus unicolor,commonmuntjacMuntiacus individual tigers present in an area (i.e. detection prob- muntjac,bantengBos javanicus and gaur Bos frontalis.Other ability P ,1; Williams et al., 2002). Photographic capture- potential tiger prey include wild buffalo Bubalus bubalis recapture sample surveys of tigers conducted in habitats and Malayan tapir Tapirus indicus. ranging from evergreen, semi-deciduous and deciduous forests to alluvial grasslands (O’Brien et al., 2003; Karanth et al., 2004a; Kawanishi & Sunquist, 2004; Wegge et al., Methods 2004) show that reliable estimates can be generated at Field methods relatively low densities of 2-3 tigers per 100 km2, although their variances tend to be large because of the The original goal was to document the presence of tigers small number of traps typically deployed in such studies. and other mammals in the area using camera-trap A recent study (Karanth et al., 2006) that integrated techniques. Therefore, trapping was done on an ad hoc photo-capture data across space and time employing basis without employing recommended survey proto- the Robust Design (Pollock et al., 1990; Lebreton et al., cols (Karanth et al., 2002; Nichols & Karanth, 2002). 1992; Kendall et al., 1997; Williams et al., 2002) demon- Twelve Trailmaster (Goodson & Associates, Lenexa, strated the power of capture-recapture analyses to detect USA) and 10 CamTrakker (CamTrakker, Georgia, USA) changes in the temporal dynamics of a tiger population. units were deployed to cover a 211 km2 area using 103 However, prior to this study, there has not been an trap locations (Fig. 1). estimate of tiger abundance in Thailand based on Trapping was carried out from 9 February 2004 to capture-recapture analyses. Here we present the results 1 February 2005 using 14 clusters of trap locations. These of a post hoc capture-recapture analysis of camera trap clusters are analogous to trapping blocks (Nichols & survey data collected in Huai Kha Khaeng Wildlife Karanth, 2002), with each block consisting of c. seven Sanctuary during 2004-2005. The objectives of our anal- trap locations. The sampling effort varied among blocks: ysis were to: (1) Assess the potential for employing eight locations were trapped for .20 days, 49 locations camera trap surveys in the semi-deciduous forests that for 16-19 days, 12 locations for c. 15 days and the form a large proportion of tiger habitat in Thailand remaining 34 locations were trapped for ,15 days. On (Tunhikorn et al., 2004). (2) Analyse the tiger photo- average there were c. 15 trap-days at each location, and capture data in a formal capture-recapture sampling this trapping effort was uniform across the study area. framework (Otis et al., 1978; White et al., 1982; Williams The moving of traps among blocks did not follow a strict et al., 2002) to generate estimates of capture probability, pre-designed sequence and was driven by logistics as population size, effectively sampled area and tiger well as opportunities for setting traps at tiger kill sites. density based on survey protocols developed in India However, in combination, data from all these blocks (Karanth et al., 2002; Nichols & Karanth, 2002). (3) covered the area evenly (Fig. 1). Assess whether tiger densities in Huai Kha Khaeng are Of particular concern for the analysis was the long comparable to densities recorded in ecologically similar survey duration of 12 months, resulting in the possibility semi-deciduous forest sites in India (Karanth et al., of the sampled tiger population being demographically 2004c). (4) Examine the general implications of our open (Otis et al., 1978). Given the high turnover of results for understanding tiger ecology and monitoring individuals in tiger populations (Karanth et al., 2006), wild tiger populations in Thailand. such lack of closure could bias estimates of population size. However, the following aspects of the survey encouraged us to attempt a post-hoc statistical analysis Study area of these data under a formal capture-recapture sampling This study was carried out in the forests around Khao framework: (1) There were two opposing cameras at Nang Rum research station within the 2,780 km2 Huai each trap location, at a distance of c. 3-5 m from the
ª 2007 FFI, Oryx, 41(4), 447–453 16
Tiger density in tropical forests of Thailand 449
Fig. 1 The Khao Nang Rum camera-trap survey area in Huai Kha Khaeng Wildlife Sanctuary. Inset shows the Sanctuary’s location (HKK) in Thailand. anticipated path of moving tigers at c. 45 cm height, that capture data from well spaced locations were which obtained good photographs of both flanks, en- included in every sampling occasion. We constructed abling unambiguous identification of individuals. (2) The five sampling occasions based on the calendar dates on camera trap locations were selected based on signs of which each location was trapped (Otis et al., 1978; Karanth past tiger activity to maximize capture probabilities, & Nichols, 1998). Because of low capture rates, tiger resulting in a relatively large number (n 5 17) of in- photo-capture data from three successive calendar dates dividual tigers being photo-captured. (3) The maximum at each trapping location were combined before being spacing between any two trap locations was ,2.3 km, assigned to a specific sampling occasion. We thus ensured thus ensuring that there were no holes in the sampled that equal trapping effort was expended and the entire area and that every tiger in the sampled population had area was sampled during each of the sampling occasions. a non-zero probability of being photo-captured during The individual tiger capture histories in the standard each sampling occasion. X-matrix format (Otis et al., 1978; White et al., 1982) were analysed using models developed for closed popula- tions (Otis et al., 1978; White et al., 1982) implemented in Analysis the software CAPTURE (Rexstad & Burnham, 1991). We Given the potential for lack of demographic closure tested the population closure assumption against our (Karanth et al., 2002; Williams et al., 2002) in the data we data. The closure test (Otis et al., 1978; White et al., 1982) would have preferred to use open model analyses implemented in CAPTURE is based on the number of (Karanth et al., 2006). However, the lack of simultaneous sample periods separating the times of first and last natural temporal coverage of the entire survey area capture for each animal caught at least twice. If animals (because we had to construct our sampling occasions are entering and/or leaving the sampled population as described above) precluded this option. Therefore, we during the survey period, the time between first and last constructed closed model capture histories following captures should be shorter on average than if all animals survey design 4 of Nichols & Karanth (2002), ensuring were present during the entire survey period.
ª 2007 FFI, Oryx, 41(4), 447–453 17
450 S. Simcharoen et al.
The capture-recapture models implemented in CAP- Table 1 Capture histories of tigers photo-trapped in Huai Kha TURE consider potential effects of behavioural response Khaeng Wildlife Sanctuary, Thailand, during 2004-2005.
of tigers to camera trapping (e.g. trap-avoidance: model Sampling occasion Mb), time-specific variation (e.g. weather changes over Identification no. 12345Age/sex* the 3-day sampling occasions: model Mt), and heteroge- neity among individual animals (e.g. caused by factors HKT-101 11111F such as territorial status or trap access: model Mh), as well HKT-102 10111F
as more complex models such as Mbh,Mth,Mtb and Mtbh HKT-103 10101F that incorporate occurrence of the effects of heterogeneity, HKT-104 11111F HKT-105 11111F trap response and time in different combinations. HKT-106 11111M We fitted the null model M0 and each of the above seven HKT-107 00011M models to our data using CAPTURE (Rexstad & Burnham, HKT-108 01011F 1991) and examined results of goodness-of-fit and between- HKT-109 10000F model tests, and the overall discriminant function, to HKT-11000100M HKT-11100001F guide the selection of an appropriate model for the data. HKT-11201000U The selected model was then used for estimating cap- HKT-11300100F b ture probabilities bp and abundance N. We estimated the HKT-11511000C effectively sampled area using an approach evaluated by HKT-11601000C Wilson & Anderson (1985), and computed tiger densities HKT-11400001M HKT-11710100F by dividing the population size by the sampled area. This computational approach is fully described elsewhere *F, female .12 months; M, male .12 months; U, unknown sex .12 (Karanth & Nichols, 1998; Nichols & Karanth, 2002). months; C, cubs ,12 months (not included in the capture-recapture analysis).
Results between photo-captures was 0.90-16.05 km, with a mean Photographic captures of tigers value of 7.11 km. Using the approach described more In a total sampling effort of 1,509 trap-days we obtained fully elsewhere (Karanth & Nichols, 1998; Nichols & 124 tiger photographs (59 right flanks, 57 left flanks, four Karanth, 2002), we estimated the effectively sampled area b b b 2 frontal, four rear) of 15 individual tigers judged to as AðWÞðSE½AðWÞ Þ 5 477.2 (58.24) km . be .12 months of age (10 females, four males, one of un- known sex). Individual tigers could be clearly identified Tests for population closure and model selection from stripe patterns (Karanth et al., 2002) and were given unique identification numbers (HKT-101-HKT-117). The statistical test for population closure implemented We obtained both left and right profile photos for 12 in CAPTURE (Rexstad & Burnham, 1991) was consistent individuals, and three more animals were identified with the assumption that our tiger population was from only left profiles. Capture data for two cubs were closed during the survey period (z 5 0.39, P 5 0.65). excluded from the analysis. Because of the long (12 months) survey period, we The capture histories generated from the field survey would have liked to consider open models as well but (Table 1) show that the number of individuals caught the ad hoc field sampling design precluded this possi-
was small (Mt+1 515), as expected in a low to medium bility. We assumed that our data supported the closure density tiger population (Karanth et al., 2004c). Four assumption, albeit not strongly. animals were caught in all five sampling occasions, one The test for presence of individual heterogeneity in
was caught in four occasions, two animals were caught capture probabilities showed that the null model M0 was thrice, two others twice and six individual tigers were rejected in favour of the model incorporating heterogeneity 2 caught only once. We expected this low recapture rate Mh (v 5 10.07, df 5 1, P ,0.002). The goodness-of-fit
for several individuals to induce substantial uncertainty test results for models Mh and Mb (incorporating trap- in our estimates. response behaviour) provided no evidence of lack of fit (v2 5 3.85, df 5 4, P 5 0.43 and v2 5 2.57, df 5 4, P 5 0.64, respectively). The tests also did not reject the Estimates of effectively sampled area 2 null model M0 in favour of alternative models Mb (v 5
The polygon formed by the outer-most camera traps 0.77, df 5 1, P 5 0.38) or Mt (time-specific variation in (Fig. 1) was 211 km2. For the 10 individual tigers that capture probabilities; v2 5 2.86, df 5 4, P 5 0.58).
were caught more than once, the maximum distance Model Mbh, which accommodates heterogeneity as well as
ª 2007 FFI, Oryx, 41(4), 447–453 18
Tiger density in tropical forests of Thailand 451 trap response was not favoured over the more par- Based on comparisons of this ad hoc study with 2 simonious Mh model (v 5 0.67, df 5 2, P 5 0.72). earlier surveys in India that employed more rigorous The overall discriminant function model selection field protocols (Karanth et al., 2002; Nichols & Karanth, algorithm in CAPTURE (Rexstad & Burnham, 1991) 2002), we recommend the following modifications to scored the competing models as: M0 5 0.88, Mh 5 1.00, future camera trap studies of tigers in the area: (1) The Mb 5 0.38, Mbh 5 0.57, Mt 5 0.0, Mth 5 0.41, Mtb 5 number of camera traps employed in this study was
0.37, Mtbh 5 0.64. The higher scoring model Mh is more small (10-15). To improve robustness of the statistical likely to have generated the observed capture history inferences of tiger abundance we recommend deploy- data in comparison to lower scoring models. This choice ment of at least 40-50 traps, so that the sampled area, the of model Mh in statistical tests reported above is consis- potential number of tiger-exposed traps, and recapture tent with past results (Karanth et al., 2004c) as well as rates can all be increased. (2) The camera trap survey aspects of tiger biology. Resident breeding tigers maintain duration should be shorter, preferably ,6 weeks, to home ranges that overlap between the sexes. Addi- avoid potential violation of population closure assump- tionally, some individuals in the population are non- tions. Furthermore, a pre-designed field survey protocol breeding ’floaters’, which may not have stable home (Nichols & Karanth, 2002), which can generate data ranges (Sunquist, 1981; Smith, 1993; Karanth & Sunquist, amenable to straightforward construction of capture 2000). These space use patterns, as well as location of our histories, should be employed. A larger number of traps camera traps in relation to home ranges of individuals, would make it easier to implement such a survey design. were likely to induce differences in capture probabilities (3) It would be useful to sample this tiger population among individual tigers. photographically on an annual basis to estimate its size and density, as well as other parameters such as longer Estimates of capture probability, tiger population term rates of survival, recruitment, and permanent and size and density temporary emigration. Robust Design and other recent refinements in capture-recapture analyses (Pollock et al., The tiger capture histories (Table 1) were used to 1990; Lebreton et al., 1992; Kendall et al., 1997; Williams generate parameter estimates under model M using h et al., 2002) facilitate such analyses (Karanth et al., 2006). the jackknife estimator (Burnham & Overton, 1978; Otis Reliable monitoring of the responses of tiger population et al., 1978) implemented in CAPTURE, which per- dynamics to threats and conservation interventions formed well in earlier photographic capture studies of can be an effective component of long-term adaptive tigers (Karanth & Nichols, 1998; Karanth et al., 2004c). management. The estimated average capture probability per sampling The observed mean density of c. 4 tigers per 100 km2 occasion (bp) was 0.42. The total population size estimate in this study was comparable to the density of 3.3-7.3 (Nb) was 19 tigers with a standard error ðSEb½Nb Þ of 3.9 tigers per 100 km2 measured in ecologically similar tigers. Based on the sampled area AbðWÞðSEb½AbðWÞ Þ 5 disturbed semi-deciduous forests such as Tadoba, 477.2 (58.24) km2, the estimated density of tigers in the Bhadra, Melghat, Pench and Panna reserves in India area was DbðSEb½Db Þ 5 3.98 (0.51) tigers per 100 km2. (Karanth et al., 2004c). However, better protected Indian These estimates exclude cubs ,12 months in age, which reserves that are ecologically comparable to Huai Kha generally comprise 20-25% of wild tiger populations Khaeng, such as Kanha, Bandipur and Nagarahole, (Smith, 1993; Kenny et al., 1995). support tiger densities that are thrice as high (c.12 tigers per 100 km2). The Huai Kha Khaeng landscape Discussion lacks an abundant, social cervid such as the chital Axis We have demonstrated in this study that non-invasive axis that accounts for 13-95% of prey numbers recorded photographic sampling is a potentially useful method in Indian reserves. However, Eld’s deer Cervus eldi, for estimating densities of tigers in the tropical forests of which was extirpated from Huai Kha Khaeng in histor- the Western Forest Complex in Thailand and therefore ical times, is such a species. probably for other similar areas in South-east Asia. Our study area of 477 km2 around Khao Nang Rum Ecological factors, such as climate, topography and research station forms 17% of the area of Huai Kha present tiger density levels permit the application of Khaeng Wildlife Sanctuary and 2.7% of the Western this method. The overall probability of capturing a tiger Forest Complex. Prima facie, this area appears to sup- present in the sampled area during the entire survey port low densities of ungulate prey (Srikosamatara, b (Mt+1/N 5 0.79) was ,1. Therefore, it is critical to 1993), and consequently a relatively low density of c.4 use the capture-recapture sampling-based approach tigers per 100 km2. If the entire landscape surrounding (Williams et al., 2002) to deal with the fact that not all Khao Nang Rum research station supports comparable tigers present in the study area are likely to be detected. tiger densities, Huai Kha Khaeng Sanctuary could hold
ª 2007 FFI, Oryx, 41(4), 447–453 19
452 S. Simcharoen et al.
113 tigers, and the entire Western Forest Complex c. 720 W.N. (2001) The use of photographic rates to estimate tigers. densities of tigers and other cryptic mammals. Animal Conservation 4 Given the similarity of vegetation types, climate and , , 75–79. IUCN (2006) 2006 IUCN Red List of Threatened Species. IUCN, prey composition between semi-deciduous forests of Gland, Switzerland [http://www.redlist.org, accessed 21 India and Thailand, their ecological productivities August 2007]. should be comparable. Furthermore, given the similarity Jackson, P. (1993) The status of the tiger in 1993. Cat News, 19, in composition of their ungulate prey assemblages, 5–11. potential maximum prey densities and hence tiger Karanth, K.U. (1987) Tigers in India: a critical review of field censuses. In Tigers of the World: The Biology, Biopolitics, densities should also be similar. Based on tiger density Management and Conservation of an Endangered Species (eds R.L. data from well protected Indian reserves (Karanth et al., Tilson & U.S. Seal), pp. 118–132. Noyes Publications, Park 2004c), we speculate that Huai Kha Khaeng Sanctuary Ridge, USA. could potentially hold 338 tigers, and the entire Western Karanth, K.U. (1988) Analysis of predator-prey balance in Forest Complex .2,000 tigers, highlighting the impor- Bandipur Tiger Reserve with reference to census reports. Journal of the Bombay Natural History Society, 85, 1–8. tance of this area for global tiger conservation. Major Karanth, K.U., Chundawat, R.S., Nichols, J.D. & Kumar, new conservation initiatives followed on from this N.S. (2004a) Estimation of tiger densities in the tropical study, in particular improved law enforcement under dry forests of Panna, Central India, using photographic the joint initiatives of the Thailand government and the capture-recapture sampling. Animal Conservation, 7 Wildlife Conservation Society, and we have also imple- , 285–290. Karanth, K.U., Kumar, N.S. & Nichols, J.D. (2002) Field surveys: mented an improved camera-trap monitoring system estimating absolute densities of tigers using capture- that employs standard closed model photographic recapture sampling. In Monitoring Tigers and their Prey: A capture-recapture sampling of ,60 days duration Manual for Researchers, Managers and Conservationists in (Karanth et al., 2002) using 136 trap sites to sample Tropical Asia (eds K.U. Karanth & J.D. Nichols), pp. 139–152. effectively an area of 1,260 km2. Centre for Wildlife Studies, Bangalore, India. Karanth, K.U. & Nichols, J.D. (1998) Estimating tiger densities in India from camera trap data using photographic captures and recaptures. Ecology, 79, 2852–2862. Acknowledgements Karanth, K.U., Nichols, J.D. & Kumar, N.S. (2004b) We are grateful to the Department of National Park, Photographic sampling of elusive mammals in tropical forests. In Sampling Rare or Elusive Species (ed. W.L. Wildlife, and Plant Conservation, Government of Thompson), pp. 229–247. Island Press, Washington, DC, USA. Thailand, for supporting this work. We gratefully ac- Karanth, K.U., Nichols, J.D., Kumar, N.S., Link, W.A. & Hines, knowledge encouragement received from the following J.E. (2004c) Tigers and their prey: predicting carnivore officials of the department: Chatchawan Pisdamkhom, densities from prey abundance. Proceedings of National Soontoon Chaiwattana, Kalyanee Boonkerd, Boosabong Academy of Sciences (USA), 101, 4854–4858. Karanth, K.U., Nichols, J.D., Seidensticker, J., Dinerstein, E., Kanchanasakha, and Suchitra Changtragoon. We thank Smith, J.L.D., McDougal, C., Johnsingh, A.J.T., Chundawat, the Wildlife Conservation Society, New York, for sup- R.S. & Thapar, V. (2003) Science deficiency in conservation porting the involvement of AP, KUK and NSK in this practice: the monitoring of tiger populations in India. Animal study and for partial funding, and the US Geological Conservation, 6, 141–146. Survey’s Patuxent Wildlife Research Center for support- Karanth, K.U., Nichols, J.D., Kumar, N.S. & Hines, J.E. (2006) Assessing tiger population dynamics using photographic ing the involvement of JDN. We acknowledge WWF– capture-recapture sampling. Ecology, 87, 2925–2937. Thailand for providing 10 camera traps to start the Karanth, K.U. & Sunquist, M.E. (2000) Behavioural correlates of study. We are grateful to our enthusiastic team of predation by tiger (Panthera tigris), leopard (Panthera pardus) research assistants, Boonyang Srichan, Sompoad Daung- and dhole (Cuon alpinus) in Nagarahole, India. Journal of sirichantra and Somporn Pakpein for their dedicated Zoology, London, 250, 255–265. Kawanishi, K. & Sunquist, M.E. (2004) Conservation status of fieldwork. We are particularly grateful to James E. Hines tigers in a primary rainforest of Peninsular Malaysia. for assistance with data analysis. Biological Conservation, 120, 329–344. Kendall, W.L., Nichols, J.D. & Hines, J.E. (1997) Estimating References temporary emigration and breeding proportions from capture-recapture data with Pollock’s Robust Design. Ecology, Burnham, K.P. & Overton, W.S. (1978) Estimation of the size 78, 563–578. of a closed population when capture probabilities vary Kenny, J.S., Smith, J.L.D., Starfield, A.M. & McDougal, C.W. among animals. Biometrika, 65, 625–633. (1995) The long-term effects of tiger poaching on population Carbone, C., Christie, S., Conforti, K., Coulson, T., Franklin, N., viability. Conservation Biology, 9, 1127–1133. Ginsberg, J.R., Griffiths, M., Holden, J., Kawanishi, K., Lebreton, J.D., Burnham, K.P., Clobert, J. & Anderson, D.R. Kinnaird, M., Laidlaw, R., Lynam, A., MacDonald, D.W., (1992) Modelling survival and testing biological hypotheses Martyr,D.,McDougal,C.,Nath,L.,O’Brien,T.,Seidensticker,J., using marked animals: a unified approach with case studies. Smith, D., Sunquist, M., Tilson, R. & Wan Shaharuddin, Ecological Monographs, 62, 67–118.
ª 2007 FFI, Oryx, 41(4), 447–453 20
Tiger density in tropical forests of Thailand 453
Nichols, J.D. & Karanth, K.U. (2002) Statistical concepts: Srikosamatara, S. (1993) Density and biomass of large estimating absolute densities of tigers using capture- herbivores and other mammals in a dry tropical forest, recapture sampling. In Monitoring Tigers and their Prey: western Thailand. Journal of Tropical Ecology, 9, 33–43. A Manual for Researchers, Managers and Conservationists Sunquist, M.E. (1981) Social organisation of tigers Panthera tigris in Tropical Asia (eds K.U. Karanth & J.D. Nichols), pp. in Royal Chitwan National Park, Nepal. Smithsonian 121–137. Centre for Wildlife Studies, Bangalore, India. Contributions to Zoology, 336, 1–98. O’Brien, T.G., Kinnaird, M.F. & Wibisono, H.T. (2003) Crouching Tunhikorn, S., Smith, J.L.D., Prayurasiddhi, T., Graham, M., tigers, hidden prey: Sumatran tigers and prey populations Jackson, P. & Cutter, P. (eds) (2004) Saving Thailand’s in a tropical forest landscape. Animal Conservation, 6, Tigers: An Action Plan. Ministry of Natural Resources and 131–139. Environment, Department of National Parks, Wildlife and Otis, D.L., Burnham, K.P., White, G.C. & Anderson, D.R. Plant Conservation, Bangkok, Thailand. (1978) Statistical inference from capture data on closed WEFCOM (2004) GIS Database and its Applications for animal populations. Wildlife Monographs, 62, 1–135. Ecosystem Management. The Western Forest Complex Panwar, H.S. (1980) A note on tiger census technique based Ecosystem Management Project, Department of National on pugmark tracings. Cheetal, 22, 40–46. Park, Wildlife, and Plant Conservation, Bangkok, Pattanavibool, A. & Dearden, A. (2002) Fragmentation and Thailand. wildlife in montane evergreen forest, northern Thailand. Wegge, P., Pokheral, C.P. & Jnawali, S.R. (2004) Effects of Biological Conservation, 107, 155–164. trapping effort and trap shyness on estimates of tiger Petdee, A. (2000) Feeding habits of the tiger (Panthera tigris) abundance from camera trap studies. Animal Conservation, in Huai Kha Khaeng Wildlife Sanctuary by fecal analysis.MSc 7, 251–256. thesis, Faculty of Forestry, Kasetsart University, Bangkok, White, G.C., Anderson, D.R., Burnham, K.P. & Otis, D.L. (1982) Thailand. Capture-Recapture Removal Methods for Sampling Closed Pollock, K.H., Nichols, J.D., Brownie, C. & Hines, J.E. (1990) Populations. Los Alamos National Laboratory publication no. Statistical inference for capture-recapture experiments. LA-8787-NERP. Los Alamos, USA. Wildlife Monographs, 107, 1–97. Williams, B.K., Nichols, J.D. & Conroy, M.J. (2002) Analysis and Rabinowitz, A. (1993) Estimating the Indochinese tiger Panthera Management of Animal Populations. Academic Press, San tigris corbetti population in Thailand. Biological Conservation, Diego, USA. 65, 213–217. Wilson, K.R. & Anderson, D.R. (1985) Evaluation of two density Rabinowitz, A. (1999) The status of the Indochinese tiger: estimators of small mammal population size. Journal of separating fact from fiction. In Riding the Tiger: Tiger Mammalogy, 66, 13–21. Conservation in Human Dominated Landscapes (eds J. Seidensticker, S. Christie & P. Jackson), pp. 148–165. Cambridge University Press, Cambridge, UK. Rexstad, E. & Burnham, K.P. (1991) User’s Guide for Biographical sketches Interactive Program CAPTURE. Colorado Cooperative Wildlife Research Unit, Colorado State University, Fort Saksit Simchareon’s area of interest is the study of tigers and Collins, USA. leopards in the dry forests of Thailand using radio telemetry Seidensticker, J., Christie, S. & Jackson, P. (1999) Preface. and camera trapping. He has carried out intensive ecolog- In Riding the Tiger: Tiger Conservation in Human Dominated ical studies on these species over the past 3 years. Anak Landscapes (eds J. Seidensticker, S. Christie & P. Jackson), pp. Pattanavibool is interested in examining large mammal xv–xix. Cambridge University Press, Cambridge, UK. ecology and conservation issues in Thailand within a land- Smith, J.L.D. (1993) The role of dispersal in structuring the scape ecological framework. K. Ullas Karanth developed Chitwan tiger population. Behaviour, 124, 165–195. camera trap surveys in India with a focus on integrating Smith, J.L.D., Tunhikorn, S., Tanhan, S., Simcharoen, S. & them with modern animal sampling methods. James D. Kanchanasaka, B. (1999) Metapopulation structure of tigers in Nichols works on development of rigorous sampling and Thailand. In Riding the Tiger: Tiger Conservation in Human analytical methodologies for assessing wildlife populations. Dominated Landscapes (eds J. Seidensticker, S. Christie & N. Samba Kumar specializes in developing field protocols P. Jackson), pp. 166–175. Cambridge University Press, for surveying large mammals in Asian forests. Cambridge, UK.
ª 2007 FFI, Oryx, 41(4), 447–453 21
Author's personal copy
BIOLOGICAL CONSERVATION 141 (2008) 2242– 2250
available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/biocon
Home range size and daytime habitat selection of leopards in Huai Kha Khaeng Wildlife Sanctuary, Thailand
Saksit Simcharoena, Adam C.D. Barlowb,*, Achara Simcharoenc, James L.D. Smithb aResearch Division, Department of National Park, Wildlife and Plant Conservation, Pahonyotin Road, Chatuchak, Bangkok 10900, Thailand bFisheries, Wildlife and Conservation Biology Department, University of Minnesota, 1980 Folwell Avenue, St. Paul, MN 55108, United States cConservation Areas Regional Office 12, Kositai Road, Muang, Nakhornsawan 60000, Thailand
ARTICLE INFO ABSTRACT
Article history: Leopards (Panthera pardus) are endangered in South East Asia yet little is known about Received 23 November 2007 which resources need to be secured for their long-term conservation or what numbers of Received in revised form this species this region can support. This study uses radio telemetry to investigate seasonal 14 June 2008 variation in habitat selection and home range size of Leopards in Huai Kha Khaeng Wildlife Accepted 22 June 2008 Sanctuary, Thailand. Over a five year period, 3690 locations were recorded from nine indi- Available online 10 August 2008 viduals. The mean ± standard error of fixed kernel home range size for six adult females was 26 ± 8.2 km2, for two adult males was 45.7 ± 14.8 and for two sub-adult females was Keywords: 29 km2 ± 5.5. Adult female wet and dry season home range sizes did not differ significantly. Compositional analysis One adult male showed an increase in home range size from dry to wet seasons. Estimated Fixed kernel density was 7 adult females/100 km2, which suggests 195 adult female leopards living in Leopard Huai Kha Khaeng alone, thus highlighting the larger Western Forest Complex’s potential Panthera pardus contribution to leopard conservation. Compositional analysis of second and third order Telemetry habitat selection suggested mixed deciduous and dry evergreen forest types, flat slope and areas close to stream channels are important landscape features for leopards. These results can help formulate a much needed conservation strategy for leopards in the region. Ó 2008 Elsevier Ltd. All rights reserved.
1. Introduction such as the leopard have a chance of long-term survival (Trisurat, 2006). Without research-based conservation plan- The leopard (Panthera pardus) is a widespread but endangered ning in the region, remaining leopard populations face an large felid. Habitat destruction, poaching and prey depletion uncertain future (Rabinowitz, 1989; Weber and Rabinowitz, have created a discontinuous patchwork of leopard popula- 1996). tions throughout Asia, Africa, the Middle East and south east- An essential component of conservation planning is iden- ern Europe (Bailey, 1993; Uphyrkina et al., 2001). Persecution tifying important resources that relate to population persis- of leopards has led to listing of the species on Appendix 1 of tence (Alldredge and Ratti, 1992; Marker and Dickman, the Convention of International Trade of Endangered Species 2005). The ‘‘importance’’ of a resource can be measured by of Wild Fauna and Flora. its perceived contribution to an animal’s survival or reproduc- In Thailand, leopards are particularly threatened with hab- tion (Garshelis, 2000). However, for short-term conservation itat loss and prey depletion (Rabinowitz, 1989; Grassman, planning it is necessary to identify key resources during the 1999), and there are few contiguous areas left where large cats considerable time needed to investigate habitat-specific
* Corresponding author: Tel.: +1 612 626 1213. E-mail addresses: [email protected] (S. Simcharoen), [email protected] (A.C.D. Barlow), [email protected] (A. Simcharoen). 0006-3207/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.biocon.2008.06.015 22
Author's personal copy
BIOLOGICAL CONSERVATION 141 (2008) 2242– 2250 2243
demographic parameters for long-lived animals with low lined in this paper. Our investigation builds substantially on reproduction rates such as the leopard (Nielson et al., 2006). this earlier work by taking advantage of the largest sample Importance can potentially be inferred by ‘habitat selection’’ size of its kind in Asia, a longer study time and the develop- which refers to a behavioral response resulting in dispropor- ment of more advanced analysis techniques. This study does tionate use of habitat types that may affect the animal’s fit- not identify causal factors that drive spatial dynamics or den- ness (Block and Brennan, 1999). Habitat selection may vary sity. Instead it focuses on providing baseline data on popula- due to seasonal variation in resource availability, so studies tion status and identifies important landscape features for should take this into account to identify resources important leopards that can be used for conservation planning on a na- at only certain times of year (Schooley, 1994). tional or regional scale. Although habitat selection for leopards is covered by some studies (Bailey, 1993; Marker and Dickman, 2005), information 2. Materials and methods is lacking for Asia. A camera trap study in Kaeng Krachen National Park in Thailand found that leopard habitat use in- 2.1. Study area creased with distance from human habitation (Ngoprasert et al., 2007). However, no study in Thailand has conclusively The study site consisted of a valley and surrounding hills in identified, either seasonally or otherwise, leopard selection Huai Kha Khaeng Wildlife Sanctuary around Khao Nang of other geographical features likely to influence their distri- Rum Wildlife Research Station and Sub Pha Pa Guard Post bution, such as forest type, topography or water courses. (Fig. 1). Average annual rainfall for the area, as recorded at In addition, management will benefit from improved Khao Nang Rum, was 1447 mm. The rainfall pattern delin- understanding of leopard home range, which can be consid- eates wet (May–October) and dry (November–April) seasons ered a more-or-less restricted area where an animal moves (Walker and Rabinowitz, 1992). Elevation varies from 121 m during its normal activities (Harris et al., 1990). Information at the valley bottom up to the highest surrounding hill at on leopard home range size will allow managers to assess 1350 m. HKK is a well protected sanctuary with human activ- population status and model responses to different threat ities largely limited to research, tourism and an annual mush- scenarios. Information regarding leopards’ spatial require- room harvest. HKK also experiences annual fires originating ments and habitat needs can then be used with existing satel- in adjacent agricultural land. These fires are generally of lite coverages to guide management of the species on a low intensity and affect a small proportion of the sanctuary. landscape-level, as has been done for tigers (Panthera tigris) Forest types include mixed deciduous, dry deciduous dip- (Sanderson et al., 2006). terocarp, dry evergreen and hill evergreen (Rabinowitz, 1989). Considering the leopard’s elusive nature and naturally low In addition to leopard, HKK’s large mammal assemblage in- densities, radio telemetry is probably the best available cludes tiger (Panthera tigris), Asiatic black bear (Ursus thibet- means to investigate both habitat selection and home range anus), Malayan sun bear (Helarctos malayanus), banteng (Bos size (Bailey, 1993). This technique has been used extensively javanicus), gaur (Bos gaurus), sambar (Cervus unicolor), Malaysian in the leopard’s African range (Norton and Lawson, 1985; tapir (Tapirus indicus) and elephant (Elephas maximas)(Rabino- Bailey, 1993; Jenny, 1996; Mizutani and Jewell, 1998; Marker witz and Walker, 1991; Srikosamatara, 1993). The 2780 km2 and Dickman, 2005). In Asia it has been used in Nepal HKK Wildlife Sanctuary is part of the larger 18,727 km2 Wes- (Seidensticker, 1976; Sunquist, 1983; Odden and Wegge, tern Forest Complex (WEFCOM), and is made up of 17 protected 2005), India (Karanth and Sunquist, 1995, 2000) and Thailand areas. This whole area is one of the largest intact contiguous (Rabinowitz, 1989; Grassman, 1999). Rabinowitz (1989) col- forests left in South East Asia (Trisurat, 2006). lected telemetry data on two leopards in Huai Kha Khaeng Wildlife Sanctuary, and Grassman (1999) collected data on 2.2. Leopard capture three leopards in Kaeng Krachen National Park. Both studies suggested there may be slight increases in male leopard home Wooden box traps (2 m long, 0.8 m wide, 1 m high) were bai- range size during the wet season, possibly in response to ted with live or dead chickens and placed along jeep tracks wider dispersion of prey (Rabinowitz, 1989; Grassman, 1999). and animal paths. Placement was based on presence of sec- We conducted a radio telemetry study in Huai Kha Khaeng ondary sign such as tracks and scrapes, which indicated reg- Wildlife Sanctuary (HKK), Thailand to investigate seasonal ular leopard travel routes (Rabinowitz, 1989). All traps were variation in (1) home range size and (2) selection of vegeta- checked on a daily basis. Captured leopards were immobilized tion, slope and stream habitat types. We chose to focus on using Ketamine (5 mg/kg) and Xylazine (2 mg/kg), delivered second and third order habitat selection, which reflects an by a jab stick. The Xylazine was later antagonized with animal’s choice of a home range within the landscape and Yohimbine (0.05 mg/kg). Leopards were fitted with a VHF radio habitat patches within a home range, respectively (Johnson, collar (Telonic Inc., Mesa, Arizona and Advanced Telemetry 1980). We also included analysis of combined season habitat Systems, Minnesota, USA) and released at the site of capture. selection to determine if this pooled data source would have Leopards were observed until they regained full conscious- failed to identify temporally important resources, as has been ness and walked away. noted for other studies (Schooley, 1994). To give an indication of population status for the area, we estimated density of 2.3. Home range estimation adult female leopards. Rabinowitz’s (1989) previous study of leopards in HKK Each day, an attempt was made to locate all radio-collared established the trapping and monitoring methodology out- leopards using standard telemetry procedures (Mech, 1983). 23 Author's personal copy
2244 BIOLOGICAL CONSERVATION 141 (2008) 2242– 2250
Fig. 1 – Map of Western Forest Complex and Huai Kha Khaeng Wildlife Sanctuary, Thailand.
To minimize possible error resulting from a moving animal, tive purposes with other studies the minimum convex three teams, positioned at different locations, took bearings polygons (MCP) were also calculated. of the same leopard simultaneously. Initial bearings were The computer program BIOTAS v1.03 (Ecological Software generally taken from ridge tops or other vantage points before Solutions, Florida) was also used to determine 95% MCP and descending to the valley floor to acquire additional bearings if 95% FK home ranges (Worton, 1989), with the FK smoothing necessary. All points were recorded using a geographical posi- factor determined by least squares cross validation (LSCV). tioning system. Seaman and Powell (1996) found that FK estimators gave little To reduce the likelihood of autocorrelation, which is the bias when smoothed by LSCV, and made good distribution lack of independence between pairs of observations at given estimates compared to the adaptive kernel approach. Leopard distances in time or space (Legendre, 1993), only one radio home ranges were determined for each study year (combined location per day was used for each leopard. If more than seasons) and for wet and dry seasons within each year. one location was taken for an individual leopard in a day then Kernohan et al. (2001) noted that kernel estimators performed the location taken closest to 12 p.m. was used. well, even with as little as 50 data points, so this was used as To assess home range size we used the fixed kernel (FK) the minimum threshold for calculating a home range within a method (Worton, 1989), which is not effected by grid size or time period. A mean home range size was estimated for each placement (Silverman, 1986), and can estimate density isop- leopard, and a grand mean for all leopards in the same demo- leths of any shape (Seaman and Powell, 1996). For compara- graphic group. A Wilcoxon signed rank t-test (a = 0.05), was 24
Author's personal copy
BIOLOGICAL CONSERVATION 141 (2008) 2242– 2250 2245
used to test for differences between wet and dry home range Leban, 1999). Compositional analysis overcomes the problems sizes. that can result from (1) assessing habitat selection without The density of adult females was calculated by construct- taking into account non-independence of proportions (also ing a 100% MCP around all adult female locations for the year known as the unit sum constraint), and (2) inappropriate level with the most collared animals of that demographic group. of sampling and sample size (Aebischer et al., 1993). Telemetry error was estimated by the distance between The minimum number of individuals needed for composi- estimated (determined by radio telemetry and program tional analysis is six (Aebischer et al., 1993), thus for this LOASTM 4.0 b, Ecological Software Solutions) and actual (re- study all leopards with adequate number of locations (n =9) corded using a GPS) locations of a spare radio collar that were pooled together, irrespective of demographic group. was moved around the forest. Mean telemetry error, calcu- If selection was significantly non random (p < 0.05), the lated by distance between estimated and actual collar posi- habitat types were ranked from most to least selected, using tions (n = 28) was 150 ± 76 m (range = 300–30 m). This level a matrix of mean and standard deviation of log ratio differ- of error was acceptable considering that it was lower than ences for all habitat types. that used by Dickson and Beier (2002), whose approach we Classification and analysis of GIS coverages was done followed. using GIS software ArcView 3.3 (ESRI, Redlands, California). The number of unsuccessful location attempts was not re- The study area was defined by a minimum convex polygon corded, so missing data were calculated as the total number around the combined locations of all leopards, which totaled of days when a collar was active but not located. 172.3 km2. Habitat classes analyzed were vegetation, slope, and 2.4. Habitat selection water courses (Fig. 2). Vegetation class included three broad types; dry dipterocarp, dry evergreen and mixed deciduous In this study ‘‘habitat type’’ (e.g. dipterocarp forest) represents forest. There was also a small amount of hill evergreen, a category of variation in characteristics within a ‘‘habitat but since this comprised <1% of the study area it was class’’, which signifies a physical, biological or geographical merged with dry evergreen; the habitat type with most sim- coverage (e.g. vegetation). For statistical analysis of habitat ilar characteristics. selection using location data, we used the compositional The slope class was created using a 20 m elevation con- analysis approach (Aitchison, 1986; Aebischer et al., 1993) tours to define slope angle for each 30 m grid square. Slope using Program Resource Selection for Windows (RSW 1.00Ó class was divided into three types measured in degrees from
Fig. 2 – Habitat class, types and composition in the Huai Kha Khaeng study area. 25
Author's personal copy 2246 BIOLOGICAL CONSERVATION 141 (2008) 2242– 2250 level; flat slope type was 0–12 degrees, moderate was 13–24 3. Results degrees, and steep was 25–49 degrees. The water course class was constructed using a 150 m buf- 3.1. Animal capture and telemetry effort fer around all streams and rivers in the study area. Area with- in this buffer was termed stream type, whilst outside was Ten leopards were captured (seven females and three males). termed dry type. Although somewhat arbitrary, we consid- All animals recovered from immobilization and were released ered the 150 m buffer width to take into account resources with a VHF radio collar. The radio-collared leopards were fol- (such as prey and travel routes) that may be distributed close lowed over five years (1994–1999) with each animal tracked for to the water channels (Grassman, 1999). a mean of 25 ± 16 months. One male leopard had too few loca- To circumvent the problems associated with using tions (12) to be included in further analysis. In total, 3690 loca- point data for habitat selection analysis (Rettie and tions were recorded from the remaining nine leopards. The McLoughlin, 1999), a 150 m buffer (the mean estimated mean number of daily locations for each of these leopards telemetry error) was created around each leopard location. was 337 ± 234 (Table 1). This was done to take into account both telemetry error The mean percentage of available tracking days when and the potential importance of habitat close to the ani- locations were acquired (not including M855 whose home mal’s position (Rettie and McLoughlin, 1999; Dickson and range was not estimated) was 44.5% and ranged from 31 to Beier, 2002). 73.7% (Table 1). This included days when no attempt was Following Dickson and Beier (2002), we looked at second made to acquire a location and days when, despite efforts, order selection using compositional analysis to compare pro- no location was obtained. portions of available habitat types in the study area to se- lected habitat within all area encompassed by buffered 3.2. Home range size leopard locations for combined, wet (May–October) and dry (November–April) season locations. Adult female leopards (n = 6) had a mean annual 95% FK For third order selection we compared the proportion of home range of 26 ± 8.2 km2, a wet season home range (n = 5) available habitat in multiyear 95% fixed kernel home ranges of 25.8 ± 7.8 km2 and a dry season home range (n = 6) of (for combined, wet and dry locations) to the selected habitat 29.2 km2 (±12.5). Mean 95% FK combined seasons’ home range encompassed by the buffered locations within the home size for the two sub-adult females still living in their natal range (Dickson and Beier, 2002). areas was 29 ± 5.5 km2. A Wilcoxon paired rank t-test For both orders of selection, results can be unsound if (a = 0.05) revealed no significant difference (p = 0.06) between one or more animals are not recorded in one of the habitat wet and dry season adult female 95% FK home ranges. Of note types within a habitat class, in which case the zero value was an old adult female F025 who expanded her home range can be replaced by a number less than 0.1 times the next from 30.9 km2 in year one to 76.7 km2 in year two followed by lowest observed value (Aebischer et al.,1993). However, in a reduction to 14.4 km2 in year three, moving between back this study all leopards were recorded in all classes of all hab- and forth between the adjacent territories of F046 and F195. itat types. Both sub-adult females resided within their mother’s
Table 1 – Tracking effort and Fixed Kernal home range sizes of study leopards LPD Age Mo. Loc. Aq. (%) Combined seasons Wet season Dry season n 95% FK (km2) Range (km2) n 95% FK (km2) Range (km2) n 95% FK (km2) Range (km2)
F025 A 31 412 50.8 3 40.7 (±32.3) 14.4–76.7 2 38.8 (±34.5) 14.4–63.2 2 46.4 (±19.5) 32.6–60.2 F035 A 23 275 41.7 3 20.7 (±4.9) 15.5–21.5 2 18 (±5.9) 13.8–22.1 1 22.7 22.7–22.7 F046 A 55 848 50.8 5 28.9 (±13.3) 8.7–45.49 4 24.7 (±8) 17.5–36 4 34.9 (±7.6) 28.4–45.3 F195 A 40.5 603 40.5 5 26 (±9.1) 18.1–38.3 2 23 (±4.4) 19.9–26.3 3 38.2 (±20.9) 20.4–61.2 F345 A 30 302 31 3 17.8 (±7) 11.9–25.5 1 24.5 – 3 16.7 (±5.4) 11.9–22.6 F390 A 13 160 41.8 1 22 – 0 – – 1 15.9 –
Mean 32 433 43 6 26.0 (±8.2) – 5 25.8 (±7.8) – 6 29.2 (±12.5) – F035 SA 4.5 154 73.7 1 25 – – – 0 - – F467 SA 28.5 353 40.53 3 32.9 (±2.6) 30.2–35.5 2 30.6 (±2.5) 28.8–32.4 2 27.4 (±4.1) 24.4–30.3
Mean 16.5 254 57 2 29 (±5.5) – 1.0 30.6 – 1 27.4 1 M055 A 35 422 31.2 3 56.1 (±8.5) 46.5–62.6 2 72.3 (±4.16) 69.4–75.3 2 55.2 (±21) 40.4–70 M550 A 13 168 43.2 1 35.2 – 0 – – 1 41.3 – M855 A 2 12 20 0 – – 0 – – 0 – –
Mean 16.5 201 32 2 45.7 (±14.8) – 1 72.3 – 2 48.3 (±9.8) –
Note: LPD is leopard identity, F = female, M = male, A = adult, SA = sub-adult, Mo. = number of months tracked, Aq. % = the percentage of available tracking days when locations were aquired, n is the number of periods with >50 recorded locations, and for the mean is the number of leopards. Leopard F035 has been listed twice; once as a sub-adult and once as an adult after she had moved out of her natal area and established another home range elsewhere. 26 Author's personal copy
BIOLOGICAL CONSERVATION 141 (2008) 2242– 2250 2247
territories and had home ranges within the general size range of the paired t-test results testing selection between habitats of adult females. Sub-adult F035 then dispersed and estab- were significant (Table 3). lished a home range to the North East of her natal area, and was classed as an adult for the later period. The two adult 3.3.2. Slope males had annual home ranges of 56.1 ± 8.5 km2 and At the second order level slope types were selected signi- 35.2 km2. One adult male mean wet season home range ficantly differently from random for combined (k = 0.14, 72.3 ± 4.16 km2 was larger than its mean dry season home p < 0.001), wet (k = 0.04, p < 0.001) and dry (k = 0.08, p < 0.0001) range of 55.2 ± 21 km2 (Table 1). periods. In all cases, low slope was ranked first followed by The mean 95% MCP for adult females for combined, wet medium and steep. For each period the difference between and dry seasons were 22.8 ± 8.6 km2, 21.4 ± 12.1 km2 and rankings of each slope type was significant (p < 0.05) (Table 3). 20.2 ± 7.9 km2, respectively. For the same time periods, sub- Third order selection analysis inferred significant differ- adult females had 95% MCPs of 24.2 ± 6.9 km2, 17.4 km2, and ences for combined (k = 0.32, p < 0.05), and dry (k = 0.14, 15.3 km2. The two males had combined, wet and dry 95% p < 0.001) seasons. As slope increased selection decreased, MCPs of 36.7 ± 10.3 km2, 47.5 km2 and 32.2 ± 7.1 km2 (Table 2). with flat slope always being selected significantly more Estimated density from year three, when five adult fe- (p < 0.05) than moderate and steep slope (Table 3). males were recorded, was 7 adult female leopards/100 km2. 3.3.3. Water courses 3.3. Habitat selection At the second order level, leopards selected significantly more stream type (the area encompassed by the 150 m buffer sur- 3.3.1. Vegetation rounding all streams and rivers) for combined (k = 0.76, Evaluating second order selection (study area vs. 95% FK), p < 0.0001), wet (k = 0.21, p < 0.001) and dry (k = 0.07, leopards showed significant habitat selection for combined p < 0.0001) seasons. At the third order level there was no sig- (k = 0.4, p < 0.05), and wet seasons (k = 0.29, p < 0.05) . In both nificant difference between proportions of selected and avail- cases dry dipterocarp was ranked last and selected signifi- able stream habitat types (Table 3). cantly less (p < 0.05) than mixed deciduous and dry evergreen. The combined season result ranked dry evergreen before 4. Discussion mixed deciduous, whereas for the wet season mixed decidu- ous was ranked first. However, in both cases, the difference 4.1. Home range size between mixed deciduous and dry evergreen was not signifi- cant, i.e. the ranking of these habitat types can be considered We found no discernable difference in adult female 95% FK interchangeable (Aebischer et al., 1993)(Table 3). home range sizes for wet and dry seasons, despite differing For third order (95% FK vs. buffered locations within 95% environmental conditions which potentially affect resource FK), only wet season locations produced a significant overall distribution and availability. Similar findings were noted for result (k = 0.04, p < 0.05); dry dipterocarp was selected the leopards in Namibia (Marker and Dickman, 2005). Two previ- most, followed by dry evergreen and mixed deciduous. None ous telemetry studies in Thailand suggested male leopards
Table 2 – Minimum convex polygon home range sizes of study leopards LPD Age Combined seasons Wet season Dry season n 95% MCP (km2) Range (km2) n 95% MCP (km2) Range (km2) n 95% MCP (km2) Range (km2)
F025 A 3 42.5 (±28.9) 22–63 2 42.5 (±28.9) 22–63 2 31.9 (±20.0) 17.7–46 F035 A 3 14.0 (±4.5) 10.8–17.2 2 14.0 (±4.5) 10.8–17.2 1 10.6 – F046 A 5 14.0 (±9.7) 15.4–21.5 4 14.0 (±9.7) 15.4–21.5 4 19.5 (±4.7) 13.3–24.5 F195 A 5 16.1 (±0.6) 15.7–16.6 2 16.1 (±0.6) 15.7–16.6 3 22.3 (±10.1) 13.4–33.3 F345 A 3 20.3 – 1 20.3 – 3 12.4 (±8.8) 6.7–22.47 F390 A 1 – – 0– – 1 24.4 –
Mean A 6 21.4 (±12.1) – 5 21.4 (±12.1) – 6 20.2 (±7.9) –
F035 SA 1 – – – – 0– – F467 SA 3 17.4 (±2.7) 15.5–19.3 2 17.4 (±2.7) 15.5–19.3 2 15.3 (±2.9) 13.2–17.3 Mean SA 2 17.4 – 1.0 17.4 – 1 15.3 –
M055 A 3 47.5 (±15.0) 36.8–58 2 47.5 (±15.0) 36.8–58 2 37.2 (±2.0) 35.8–38.6 M550 A 1 – – 0– – 1 27.2 – M855 A 0 – – 0– – 0– –
Mean A 2 47.5 – 1 47.5 – 2 32.2 (±7.1) –
Note: LPD is leopard identity, F = female, M = male, A = adult, SA = sub-adult, Leopard F035 has been listed twice; once as a sub-adult and once as an adult after she had moved out of her natal area and established another home range elsewhere. 27 Author's personal copy
2248 BIOLOGICAL CONSERVATION 141 (2008) 2242– 2250
Table 3 – Compositional analysis results of leopard habitat selection Habitat class Selection order Season k p Habitat type ranking
Vegetation 2nd Combined 0.4 <0.05 Dry EG > mix. decid dry dipt 2nd Wet 0.29 <0.05 Mix. decid > dry EG dry dipt 2nd Dry 0.53 0.06 – 3rd Combined 0.8 0.30 3rd Wet 0.04 <0.05 Dry dipt > dry EG > mix. decid 3rd Dry 0.74 0.69 –
Slope 2nd Combined 0.14 <0.001 Flat medium steep 2nd Wet 0.17 <0.001 Flat medium steep 2nd Dry 0.08 <0.0001 Flat medium steep 3rd Combined 0.32 <0.05 Flat moderate steep 3rd Wet 0.56 0.07 3rd Dry 0.14 <0.001 Flat moderate > steep
Stream 2nd Combined 0.76 <0.0001 Stream dry 2nd Wet 0.21 <0.001 Stream dry 2nd Dry 0.07 <0.0001 Stream dry 3rd Combined 0.69 0.07 – 3rd Wet 0.94 0.49 – 3rd Dry 0.82 0.12 –
k is the Wilk’s lamda statistic computed from the matrix of log ratio differences and p = probability of their being no overall habitat selection. Habitats are listed in order of rank, from most to least selected, with > > indicating a significant difference in selection between two habitat types. For vegetation class dry eve. = dry evergreen, mix. dec. = mixed deciduous and dry dip. = dry dipterocarp. For slope class flat = 0–12 degrees, moderate = 13–24 degree and steep = 25–49 degrees. For stream class stream = all area within 150 m from water courses and dry = all area not within 150 m from water courses.
may increase home range sizes in the wet season 4.2. Habitat selection (Rabinowitz, 1989; Grassman, 1999). This is supported by data in this study from one male leopard which had larger home Discerning leopard resource requirements by identifying ranges recorded in the wet season (72.3 ± 4.16 km2) compared what habitat types individuals are selecting, presumes that to the dry season (55.2 ± 21 km2). Grassman (1999) suggested animals choose those which will increase survival and repro- that this effect may be in response to relatively diffuse distri- duction potentials. However, when attempting to characterize bution of prey in the wet season, but in HKK adult female home selection there may be problems associated with two main ranges recorded in our study, which should also reflect prey underlying assumptions: that recorded observations can be density (Bailey, 1993), were not affected in the same way. used to infer habitat selection, and that evidence of selection The mean MCP home range for adult females reported in is related to fitness and population growth (Porter and this study (22.8 km2 ± 8.6) is the largest so far recorded in Asia. Church, 1987; Alldredge and Ratti, 1992; Garshelis, 2000). Nev- Compared to the other studies in Thailand, Rabinowitz (1989) ertheless, investigating how a species selects different fea- recorded one MCP home range of 11.4 km2 (42 locations) in tures of its environment is an essential step towards HKK, and Grassman (2000) one of 8.8 km2 (92 locations) in assigning importance to those features, with the acknowl- Kaeng Krachan National Park. Rabinowitz’s (1989) female edged caveat that subsequent measurements of survival leopard home range size was within the range recorded in and reproduction parameters will be needed (Garshelis, 2000). our study, but substantially below the mean. Likewise we re- This study focused on daytime habitat selection and leop- corded larger adult male MCP home ranges (44 km2 ± 14.6 and ards may select habitat differently at the third order level dur- 29.4 km2) compared to 27 km2 (Rabinowitz, 1989) and 17.3– ing the night; an area for further research. A larger sample 18 km2 (Grassman, 1999). However, the difference between size would be required to determine variation in selection studies was probably also influenced by study time and sam- for different demographic groups; in this study male, female, ple size. adult and sub-adult leopards were pooled together. The individual home ranges we recorded showed high var- Also, the inferences of the selection results would have iability and are probably underestimates, considering that on been strengthened by understanding prey distribution with average we did not acquire locations on 55.5% of available respect to habitat types, but no comparable prey abundance tracking days for each leopard. The use of satellite collars in data set was collected during the study period. future studies should make comparisons between data sets Acknowledging the caveats, the results of this study sug- easier, cut necessary study time and avoid some of the prob- gest that mixed deciduous and dry evergreen are potentially lems associated with missing data and telemetry error. important habitats for leopards, which selected home ranges The density estimate of 7 adult female leopards/100 km2 within the landscape with proportionally more of those vege- recorded is still probably a conservative estimate of carrying tation types than available for the combined and wet season capacity, considering there may have been undetected indi- periods. However, the third order selection was more ambig- viduals residing in the same area. uous; within wet season home ranges (the only significant 28 Author's personal copy
BIOLOGICAL CONSERVATION 141 (2008) 2242– 2250 2249
result), dry dipterocarp was ranked fist, followed by dry ever- Plant Conservation Department. Additional funding for the green and mixed deciduous, but the difference between each project came through generous donations from World Wild- rankings was not significant. life Fund (Thailand). We are very grateful to the hard work The results for the slope and stream habitat classes were of many people that assisted with data collection in the more conclusive; low gradient and stream areas seem to be field. These included assistant researchers Prathom important landscape features for leopards when choosing a Boontawee, Preecha Prommakun, and Somporn Pakpien home range. This result is consistent with the findings of and Khao Nang Rum staff Onsa Norrasan, Saiphech Ngoprasert et al. (2007) who found that leopard use of habitat Toomnoi and Thaworn Thadwijit. Dr. Utis Kutintara was greater closer to streams. Within home ranges leopards (Kasetsart University), Chatchawan Pisdamkham (former also selected low gradient areas but did not show any prefer- chief of HKK) and Dr. Alan Rabinowitz, (Wildlife Conserva- ence to stream habitat. However, stream and slope class are tion Society) all provided valuable guidance to this study. probably correlated to some degree so these results should This preparation of this paper was aided by Namkhang not be considered as totally independent. Saelee and improved by comments from Dave Garshelis, In general, analysis of the pooled data correctly identified Peter Cutter, Francie Cuthbert and Christina Greenwood. selected habitat types identified by the seasonal data. However, for second order wet season selection, mixed decid- uous and dry evergreen were correctly identified as important but were ranked differently. For third order selection of vege- REFERENCES tation, the pooled data failed to identify the wet season selec- tion but, as noted before, this result was not conclusive. Aebischer, N.J., Robertson, P.A., Kenward, R.E., 1993. 4.3. Conservation implications Compositional analysis of habitat use from animal radio- tracking data. Ecology 74, 1313–1325. Aitchison, J., 1986. The statistical analysis of compositional data. Thailand’s protected area system mostly comprises small, Chapman and Hall, London, England. fragmented, rugged, higher altitude forest areas. With the Alldredge, J.R., Ratti, J.T., 1992. Further comparison of some exception of parts of the Western Forest Complex, including statistical techniques for analysis of resource selection. the study area, large, extensive, contiguous lowland forest Journal of Wildlife Management 56, 1–9. areas are now mostly gone due to conversion into agricultural Bailey, T.N., 1993. The African Leopard. Ecology and behavior of a land (Trisurat, 2006). The continued fragmentation of remain- solitary field. Colombia University Press, New York. Block, W.M., Brennan, L.A., 1999. The habitat concept in ing forests is a serious threat to Thailand’s large mammal fauna ornithology: theory and applications. Current Ornithology 11, (Pattanavibool and Dearden, 2002). Other leopard-range coun- 35–91. tries in Southeast Asia that still have large areas of lowland for- Dickson, B.G., Beier, L.R., 2002. Home-range and habitat selection est should consider initiating conservation strategies that by adult cougars in Southern California. Journal of Wildlife ensure good representation of this rich and diverse habitat. Management 66, 1235–1245. Even though much of the potential leopard habitat in Thai- Garshelis, D.L., 2000. Delusions in habitat evaluation: measuring use, selection, and importance. In: Boitani, L., Fuller, T.K. land is legally protected, wildlife managers must still decide (Eds.), Research Techniques in Animal Ecology; Controversies how to prioritize use of limited resources. Focusing conserva- and Consequences. Columbia University Press, New York, pp. tion efforts on umbrella species such as tiger, leopard and ele- 111–164. phant may be the most effective and immediately available Grassman, L.I., 1999. Ecology and behavior of the Indochinese strategy to improve future prospects for much of the remain- leopard in Kaeng Krachan National Park, Thailand. Natural ing terrestrial ecosystems in the country. Based on the rough History Bulletin of the Siam Society 47, 77–93. density estimates in this study, the 2780 km2 of HKK alone Harris, S.W., Cresswell, J., Forde, P.G., Trewhella, W.J., Woollard, T., Wray, S., 1990. Home range analysis using could hold at least 195 adult female leopards, and the 2 radio-tracking data; a review of problems and techniques 18,727 km of WEFCOM many times that number. This high- particularly as applied to mammals. Mammal Review 20, lights the area’s importance as a priority conservation area 97–123. for leopards as well as for tigers (Simcharoen et al., 2007). Jenny, D., 1996. Spatial organization of leopards (Panthera pardus) The leopard home range and habitat selection data pre- in Tai National Park, Ivory Coast: is rain forest habitat a sented in this study can be used to identify other areas within tropical haven? Journal of Zoology 240, 427–440. the existing protected areas network important for leopards. Johnson, D.H., 1980. The comparison of usage and availability measurements for evaluating resource preference. Ecology 61, This information, in combination with previously accumu- 65–71. lated knowledge on leopards from the region (Rabinowitz, Karanth, K.U., Sunquist, M.E., 1995. Prey selection by tiger, leopard 1989; Grassman, 1999; Ngoprasert et al., 2007), should contrib- and dhole in tropical forests. Journal of Animal Ecology 64, ute to the formulation of a national landscape-level conserva- 439–450. tion plan for the species. Karanth, K.U., Sunquist, M.E., 2000. Behavioral correlates of predation by tiger (Panthera tigris), leopard (Panthera pardus) and dhole (Cuon alpinus) in Nagarahole, India. Journal of Acknowledgments Zoology 250, 255–265. Kernohan, B.J., Gitzen, R.A., Millspaugh, J.J., 2001. Analysis of animal space use and movements. In: Marzluff, J.M. (Ed.), This research was only possible with the kind permission Radio Tracking and Animal Populations. Academic Press, San and logistical support of the National Parks, Wildlife and Diego. 29
Author's personal copy
2250 BIOLOGICAL CONSERVATION 141 (2008) 2242– 2250
Leban, F., 1999. Resource Selection for Windows 1.00. Sanderson, E., Forrest, J., Loucks, C., Ginsberg, J., Dinerstein, E., Legendre, P., 1993. Spatial autocorrelation: trouble or new Seidensticker, J., Leimgruber, P., Songer, M., Heydlauff, A., paradigm? Ecology 74, 1659–1673. O’Brien, T., Bryja, G., Klenzendorf, S., Wikramanayake, E., 2006. Marker, L.L., Dickman, A.J., 2005. Factors affecting leopard Setting Priorities for the Conservation and Recovery of Wild (Panthera pardus) spatial ecology, with particular reference to Tigers: 2005–2015. The Technical Assessment. WCS, WWF, Namibian farmlands. South African Journal of Wildlife Smithsonian, NFWF–STF, Washington DC. Research 35, 105–115. Schooley, R.L., 1994. Annual Variation in Habitat Selection: Mech, L.D., 1983. Handbook of animal radio-tracking. University Patterns Concealed by Pooled Data. Journal of Wildlife of Minnesota Press, Minneapolis. Management 58, 367–374. Mizutani, F., Jewell, P.A., 1998. Home-range and movements of Seaman, D.E., Powell, R.F., 1996. An evaluation of the accuracy of leopards on a livestock ranch in Kenya. Journal of Zoology 244, kernel estimators for home range analysis. Ecology 77, 2075– 269–286. 2085. Ngoprasert, D., Lynam, A.J., Gale, G.A., 2007. Human disturbance Seidensticker, J., 1976. On the ecological separation between affects habitat use and behaviour of Asiatic leopard Panthera tigers and leopards. Biotropica 8, 225–234. pardus in Kaeng Krachan National Park Thailand. Oryx 41, 343– Silverman, B.W., 1986. Density estimation for statistics and data 351. analysis. Chapman and Hall, London, United Kingdom. Nielson, S.E., Stenhouse, G.B., Boyce, M.S., 2006. A habitat-based Simcharoen, S., Pattanavibool, A., Karanth, K.U., Nichols, J.D., framework for grizzly bear conservation in Alberta. Biological Kumar, N.S., 2007. How many tigers Panthera tigris are there in Conservation 130, 217–229. Huai Kha Khaeng Wildlife Sanctuary, Thailand? An estimate Norton, P.M., Lawson, A.B., 1985. Radio tracking of leopards and using photographic capture-recapture sampling. Oryx 41, 447– caracals in the Stellenbosch area, Cape Province. South 453. African Journal of Wildlife Research 15, 17–24. Srikosamatara, S., 1993. Density and biomass of large herbivores Odden, M., Wegge, P., 2005. Spacing and activity patterns of and other mammals in a dry tropical forest, western Thailand. leopards Panthera pardus in the Royal Bardia National Park, Journal of Tropical Ecology 9, 33–43. Nepal. Wildlife Biology 11, 145–152. Sunquist, M., 1983. Dispersal of three radio tagged leopards. Pattanavibool, A., Dearden, P., 2002. Fragmentation and wildlife in Journal of Mammalogy 64, 337–341. montane evergreen forests, northern Thailand. Biological Trisurat, Y., 2006. Applying gap analysis and a comparison index Conservation 107, 155–164. to evaluate protected areas in Thailand. Environmental Porter, W.F., Church, K.E., 1987. Effects of environmental pattern Management 39, 235–245. on habitat preference analysis. Journal of Wildlife Uphyrkina, O., Johnson, W.E., Quigley, H., Miquelle, D., Marker, L., Management 51, 681–685. Bush, M., O’Brian, S.J., 2001. Phylogenetics, genome diversity Rabinowitz, A., 1989. The density and behavior of large cats in a and origin of modern leopard, Panthera pardus. Molecular dry tropical forest mosaic in Huai Kha Khaeng Wildlife Ecology 10, 2617–2633. Sanctuary, Thailand. Natural History Bulletin of the Siam Walker, S., Rabinowitz, A., 1992. The small-mammal community Society 37 (2), 235–251. of a dry-tropical forest in central Thailand. Journal of Tropical Rabinowitz, A., Walker, S.R., 1991. The carnivore community in a Ecology 8, 57–71. dry tropical forest mosaic in Huai Kha Khaeng Wildlife Weber, W., Rabinowitz, A., 1996. A global perspective on large Sanctuary, Thailand. Journal of Tropical Ecology 7, 37–47. carnivore conservation. Conservation Biology 10, Rettie, J.W., McLoughlin, P.D., 1999. Overcoming radio telemetry 1046–1054. bias in habitat-selection studies. Canadian Journal of Zoology Worton, B.J., 1989. Kernel methods for estimating the utilization 77, 1175–1184. distribution in home range studies. Ecology 70, 164–168. 30
Female tiger Panthera tigris home range size and prey abundance: important metrics for management
A CHARA S IMCHAROEN,TOMMASO S AVINI,GEORGE A. GALE,SAKSIT S IMCHAROEN S OMPHOT D UANGCHANTRASIRI,SOMPORN P AKPIEN and J AMES L.D. SMITH
Abstract Tigers Panthera tigris are highly threatened South Asia to the temperate forests of Far East Russia and continue to decline across their entire range. Actions (Sunquist et al., 1999). However, the species is highly to restore and conserve populations need to be based on threatened across its entire range by poaching, habitat loss science but, in South-east Asia, information on ecology and prey depletion (Kenny et al., 1995; Wikramanayake and behaviour of tigers is lacking. This study reports the et al., 1998; Karanth et al., 2004). Its distribution is greatly relationship between the home range size of female tigers restricted; populations that are viable, at least in the short and prey abundance, using data from radio-collared tigers in term, occur only in Bangladesh, Bhutan, India, Indonesia, Huai Kha Khaeng Wildlife Sanctuary, Thailand, and Malaysia, Nepal, Russia, and Thailand (Seidensticker, 2010; published data from other studies. A total of 11 tigers, four Walston et al., 2010). The species is extinct in Bali, Java, males and seven females, were fitted with global position- southern China and central Asia, and populations in other ing system collars, to estimate home ranges using 95 and range countries are now severely reduced and at imminent 100% minimum convex polygons (MCP). Prey abundance risk of extinction because of their small size (Carroll & was estimated by faecal accumulation rates. The mean home Miquelle, 2006). range size of male tigers was 267 and 294 km2 based on 95 and In response to these significant declines wildlife 100% MCPs, respectively; the mean female home range size managers are attempting a number of strategies, alone or was 70 and 84 km2, respectively. Territories of male and in combination, to increase the number of tigers. Efforts female tigers had little overlap, which indicated both sexes include restoring tiger habitat, re-establishing habitat were territorial. Mean densities of the prey species sambar connectivity, restoring prey populations, reducing poaching Rusa unicolor, barking deer Muntiacus muntjac and large of tigers and their prey, and reintroducing animals to the bovids were 7.5, 3.5 and 3.0 km−2, respectively. When female wild (Seidensticker, 2010). All of these management options home range size and prey abundance were compared at six require an understanding of the relationship of tigers to locations in Thailand, and at other sites in India, Nepal, their prey, and of the influence of ecological variables on Bangladesh and Russia, a significant negative correlation both tigers and their prey. Although ecological variables can was found between prey abundance and home range size. affect tigers directly, many are likely to have a pronounced Monitoring this relationship can provide managers with influence on the density of the tiger’s prey. Current models metrics for setting conservation goals. suggest that, in the absence of heavy poaching, the density of tigers is primarily determined by prey abundance Keywords Huai Kha Khaeng Wildlife Sanctuary, Panthera (Karanth et al., 2004; Wegge et al., 2009; Harihar et al., tigris, protected area management, satellite telemetry 2009). Documentation of a positive relationship between tiger populations and their prey does not, however, explain the behavioural mechanisms underlying population growth Introduction or resilience. As reported for many solitary felids, breeding female tigers in Nepal and Russia occupy defended home he tiger Panthera tigris occurs in a wide range ranges (Seidensticker, 1976; Sunquist, 1981; Smith et al., 1987; of climates and habitats, from the tropical forests of T Goodrich et al., 2010) and individual females have the sole responsibility for raising their young, using prey they acquire within their home range. Although prey abundance ACHARA SIMCHAROEN* (Corresponding author), TOMMASO SAVINI and GEORGE A. GALE School of Bioresources and Technology, King varies widely between these two countries, females in both Mongkut’s University of Technology, Thonburi, Thailand. E-mail simtom@ localities are usually able to establish breeding territories windowslive.com large enough to raise young. Because previous studies have SAKSIT SIMCHAROEN,SOMPHOT DUANGCHANTRASIRI and SOMPORN PAKPIEN demonstrated that female home range size is a function Department of National Parks, Wildlife and Plant Conservation, Thailand of prey density (Smith et al., 1987; Miquelle et al., 2010) we JAMES L.D. SMITH Department of Fisheries, Wildlife and Conservation Biology, ’ University of Minnesota, USA hypothesize that this relationship holds across the tiger s range and that mean female home range size may be used *Also at: Department of National Parks, Wildlife and Plant Conservation, Thailand to calculate local carrying capacity for tigers. Received 31 May 2012. Revision requested 13 August 2012. The number of female territories in a population Accepted 10 October 2012. determines that population’s recruitment potential
© 2014 Fauna & Flora International, Oryx, Page 1 of 8 doi:10.1017/S0030605312001408 http://journals.cambridge.org Downloaded: 25 Mar 2014 IP address: 115.67.101.28 31
2 A. Simcharoen et al.
and ultimately its viability and resilience. Thus, a measure- ment of the number of breeding female tigers that can exist in a given area (i.e. carrying capacity) is a useful metric for managers. This measurement would be particularly useful if tiger carrying capacity could be approximated by estimating the abundance of prey in a unit of habitat. With this information, managers would be able to predict the response of breeding female home range sizes to changes in prey abundance, as well as to identify if tiger poaching, prey poaching or habitat quality were limiting tiger population size. We used data from six resident breeding female tigers fitted with global positioning system (GPS) satellite collars in Huai Kha Khaeng Wildlife Sanctuary in western Thailand during 2005–2011 to explore the relationship between female home range size and prey density. We compared our results to those from studies in four other protected areas in which tiger home range size and prey abundance have been previously estimated. The objectives of this study are to (1) estimate home range size and home range overlap of tigers in western Thailand, (2) correlate female tiger home range size and prey density across the tiger’s range, and (3) demonstrate the significance of female home range size and prey abundance as two important metrics for reserve managers.
Study area
FIG. 1 Huai Kha Khaeng Wildlife Sanctuary, with the prey The study was conducted in the Huai Kha Khaeng Wildlife survey transects in six female tiger home ranges. The Sanctuary (Fig. 1), which, in conjunction with the wildlife dark-shaded area on the inset indicates the location of the sanctuaries Thung Yai East and West, is designated as a main map in Thailand, on the border with Myanmar. UNESCO World Heritage Site. This protected area complex supports the largest tiger population in Thailand (DNP, tiger prey species occur in high numbers in the area: banteng 2010). In Huai Kha Khaeng Wildlife Sanctuary the estimated Bos javanicus, gaur Bos gaurus, sambar Rusa unicolor and tiger density is 3.98 per 100 km2 (Simcharoen et al., 2007). wild boar Sus scrofa (Srikosamatara, 1993). The sanctuary is 2,780 km2, altitudes are 200–1,600 m, annual temperature range is 8–38° C and mean annual Methods rainfall is 1,375 mm (Khao Nang Rum Wildlife Research Station, unpubl. data). Normally the lowest temperatures Capture and radio collaring occur in January and the highest in April. The dry season (November–April) has a mean rainfall of 298 mm and the During June 2005–August 2011 we captured tigers in parts wet season (May–October) a mean of 1,088 mm. The area of Huai Kha Khaeng Wildlife Sanctuary that represent has four main vegetation types, the occurrence of which the range of habitat types and topographic features in depends on rainfall patterns and edaphic conditions: the Sanctuary. Healthy live domestic cows were placed mixed deciduous forest (48%), dry evergreen forest (25%), along trails where repeated recent tiger scent marks hill evergreen forest (13%) and dry deciduous dipterocarp occurred. When a cow was killed three or four leg-hold forest (7%; WEFCOM, 2004). snares were set near the cow to capture the tiger when it In addition to the tiger, other carnivores in the sanctuary returned to feed the next night. The snares were designed include the leopard Panthera pardus, clouded leopard by P. Pratumratanatarn, an experienced technician with Neofelis nebulosa, marbled cat Pardofelis marmorata, the Department of National Parks, Wildlife and Plant leopard cat Prionailurus bengalensis, Asiatic jackal Canis Conservation. Captured tigers were anaesthetized with a aureus, wild dog Cuon alpinus, Asiatic black bear Ursus mixture of tiletamine HCL and zolazepam HCL (Zoletil; thibetanus and sun bear Helarctos malayanus. Four major Virbac Laboratories, Carros, France) at a dose of 4 mg kg−1
© 2014 Fauna & Flora International, Oryx, 1–8
http://journals.cambridge.org Downloaded: 25 Mar 2014 IP address: 115.67.101.28 32
Tiger home range and prey abundance 3
(Kreeger & Arnemo, 2007). At the time of capture we prey density is an important factor determining the size recorded sex, age and reproductive status. The age of each of tiger home ranges. However, we include 100% MCP tiger was estimated from tooth eruption, tooth wear and estimates here also to facilitate comparison with other staining. Reproductive status of females was classified as studies. nulliparous if nipples were pink; dark nipples indicated To determine if a female home range was used a female had produced young. Animal handling and exclusively by a single female to obtain food for herself immobilization were undertaken in accordance with the and her young, we defined neighbouring females as those University of Minnesota IACUC protocol 0906A67489. that had adjacent (with insufficient space or evidence for Breeding tigers were fitted with one of three radio collar there to be another female living in between) and potentially models: Advanced Telemetry Systems (ATS) GPS collar overlapping home ranges during the same time period. model G2000 (Isante, Minnesota, USA), Telonics Argos Overlaps were based on the frequency a female used the Terrestrial Transmitter (Mesa, Arizona, USA), or Vectronic same geographical locations as her female neighbours Aerospace GmbH GPS Plus (Berlin, Germany). The ATS (at least once per month). As we did not have data for all collars were programmed to acquire locations every 2 hours. female home ranges surrounding our study females we Telonics and Vectronic collars were programmed to could only estimate the minimum percentage overlap (100% fi acquire locations every hour. All radio collars were MCPs), de ned here as HRij 5 Aij/Ai, where HRij is the released with drop-off mechanisms. ATS and Telonics proportion of individual i’s home range that is overlapped models had to be recovered to download data but the by individual j’s home range, Ai is the area of individual Vectronic model transmitted data via Iridium satellites and i’s home range and Aij is the area of overlap between the the data were available from the Vectronic website. Output two individuals’ home ranges (Kernohan et al., 2001). The of all collars included date, time, latitude, longitude, altitude home range sizes and degree of overlap of radio collared and fix status. We used location data from both two- and tigers were calculated using the ArcView v. 3.3 (ESRI, three-dimensional fixes. Redlands, USA) extension package Home Range (Rodgers & Carr, 1998).
Home range size and overlap Prey abundance
We used 95 and 100% minimum convex polygons (MCP) We estimated prey abundance during the dry season to estimate home range size of tigers (Jennrich & Turner, (December–April) for 2009–2011. In addition, prey abun- 1969). Because the 100% MCP can be influenced by outliers, dance in the entire Sanctuary has been monitored since the 95% MCP is often preferred, to avoid inflation of 2006 as part of sanctuary management; these data sug- the estimate (Harris et al., 1990; White & Garrott, 1990). gested that prey density was approximately constant during The MCP is also the most widely used home range estimator 2006–2011 (Wildlife Conservation Society, unpubl. data). (Harris et al., 1990), which facilitates comparisons with We used faecal accumulation rate techniques (Bailey & other studies. To determine the number of locations needed Putman, 1981) to estimate prey abundance in the Sanctuary. to obtain stable home range areas we conducted a bootstrap Four c. 100 km2 areas that encompassed the six female tiger simulation in which home range size was estimated from home ranges were selected to represent the range of habitat randomly selected locations. We began with 50 locations types and topographic features in the Sanctuary. Two sites and increased the sample size by 50 until all radio locations were in the central valley of the Sanctuary along the lower were included in the calculations. We considered the and upper portions of the Huai Kha Khaeng River. The home range established when the number of days or the lower site was characterized by mixed deciduous forest. number of fixes reached an asymptote (Harris et al., 1990). The upper site was a combination of mixed deciduous The kernel method, in contrast to the MCP, is best as an and dry evergreen forest and the habitat was more rugged estimator of the probability of use within a home range than at the lower site. A third site, near the Khao Nang Rum and thus can address questions related to third-order Wildlife Research Station, was a drier area away from the selection (sensu Johnson, 1980). But the kernel method can Huai Kha Khaeng River. It was composed of a combination be problematic for estimating home range size, especially of dry dipterocarp and mixed deciduous habitat. The fourth as the number of locations increases with the use of GPS site was in the northern part of the Sanctuary, and also collars (Downs & Horner, 2008). Kernel estimates are away from the Huai Kha Khaeng River. It was characterized particularly inflated when there is a large number of by mixed deciduous and dry evergreen habitat in locations along a ‘hard’ boundary, which is often the case rugged terrain. for animals, such as tigers, that defend home ranges (Downs A total of 90 square line transects, 200 × 200 m, were & Horner, 2008). Therefore, we believe that the 95% MCP randomly located in each of the four sites (Fig. 1) to survey is a suitable estimator for testing our hypothesis that gaur and banteng, as dung of these species is relatively
© 2014 Fauna & Flora International, Oryx, 1–8 http://journals.cambridge.org Downloaded: 25 Mar 2014 IP address: 115.67.101.28 33
4 A. Simcharoen et al.
rare. In addition, 40 circular plots of 20 m2 were placed 20 m between female home range size and prey abundance at the apart on each of these transects, to sample the more local level. Across the tiger’s range we reviewed information common ungulates (sambar, barking deer Muntiacus on tiger home range size and prey abundance from five muntjac and wild boar). All ungulate pellet groups were studies in Nepal, India, Bangladesh and Russia that had counted and removed from all transects and circular plots. used radio telemetry to estimate home range size. In all Following removal, new faecal accumulation was measured studies home range size was estimated by either the 95 or for a period of 30 days (over this time period no entire 100% MCP. Using data from earlier studies we estimated pellet groups were lost to decomposition or other natural prey biomass for each site as the product of prey density processes during the dry season). A pellet group consisted and mean body weight (Dhungel & O’Gara, 1991; Karanth of .15 pellets. Pellet groups that overlapped were dis- & Sunquist, 1992). We then used simple linear regression to tinguished by size and colour, which became lighter as predict home range size in relation to prey abundance at the pellets aged. local level and Spearman’s rank correlation to determine if We searched for gaur and banteng dung piles up to 2 m the same relationship holds at a range-wide level (our results either side of the transects. Observed piles were counted from Huai Kha Khaeng Wildlife Sanctuary combined with and marked, and the perpendicular distance from the published data). Probabilities , 0.05 were considered transect was measured. We could not distinguish gaur and significant. banteng dung and therefore combined the data, as ‘large bovid’. To determine absolute abundance of prey species we used published defecation rates of large bovids (9.5 dung Results piles per day) and barking deer (7.5 pellet groups per day; Srikosamatara, 1993; Sukmasuang, 2001). For sambar we Home range size and overlap recorded the defecation rate of a group of six animals of A total of 11 tigers, seven resident breeding females and four mixed sex and age (one male, two females and three young) resident breeding males, were fitted with GPS collars. One at Khao Pratapchang Wildlife Breeding Centre. To control female and one male were collared twice, at separate times, for possible seasonal differences in defecation rate, observa- and home range estimates were made for each period for tions at the Centre were made in the same season as the field each individual as size and shape of home ranges changed study. During a 5-day period deer grazed undisturbed; in a over time. The mean number of days used for home range second 5-day period deer were stimulated so that they were analysis was 152 ± SD 72 days (range 77–358 days). The mean more active than normal. Deer were fed a diet of natural time to reach an asymptote in home range size was browse that was cut near the breeding centre, to provide 77 ± SD 35 days and the mean number of fixes to reach an a diet similar in species and cellulose composition to their asymptote was 418 ± SD 160 locations. The mean number of natural diet. The mean number of pellet groups per locations used to estimate home range was 1,502 ± SD 1,215. individual per day was 9.5 ± SD 1.41 (n 5 6). We calculated Mean home range (95% MCP) of the resident females deer and barking deer density as D 5 (P/A)/(Df Ds), where (n 5 8) was 70.2 ± SD 33.2 km2 and that of the resident P 5 the number of pellet groups found on the second visit, males was 267.6 ± SD 92.4 km2 (n 5 5; Table 1). Mean home A 5 the total sample area (km2), Df 5 the defecation rate range estimated with the 100% MCP was 84.2 ± SD 40.8 km2 and Ds 5 the number of days between the two visits. (n 5 8) and 294.1 ± SD 100.3 km2 (n 5 5) for the resident Distance v. 5.0 (Thomas et al., 2006) was used to estimate the males and females, respectively (Table 1). The ratio of male density of dung, which was then used to estimate large bovid to female tiger home range size was 3.8 : 1. abundance. Data were right truncated at 1.2 m and grouped Home range overlap was examined from three pairs of into 0.2 m intervals to improve model fit. Uniform and adjacent females and one pair of adjacent males. Mean half-normal models for the detection function were fitted overlap was 4.5 ± SD 3% for the females and 18% for the against the data, using cosine adjustments (Buckland et al., males. We used five female home ranges within three male 2001). The best model selection and adjustment terms were home ranges to examine overlap between male home ranges based on Akaike’s Information Criterion (AIC; Akaike, and sympatric female home ranges. Male home ranges 1974). We calculated large bovid density BD 5 Dd/(DfDs), overlapped nearly the entire sympatric home range of the where Dd 5 bovid dung density, Df 5 defecation rate of females (90 ± SD 15%, n 5 5). bovids, and Ds 5 number of days between the two visits.
Prey abundance Home range size and prey abundance Mean densities of sambar and barking deer were We estimated prey biomass in six resident female tiger 7.5 ± SD 6 and 3.5 ± SD 2.6 km−2 over the four sites. Our home ranges in the Sanctuary to determine the correlation sample of wild boar dung (n 5 77) was insufficient for
© 2014 Fauna & Flora International, Oryx, 1–8
http://journals.cambridge.org Downloaded: 25 Mar 2014 IP address: 115.67.101.28 34
Tiger home range and prey abundance 5
TABLE 1 Estimates, using 95 and 100% minimum convex polygons (MCP), of the home range size of 11 individual collared tigers (one male and one female were collared twice) in Huai Kha Khaeng Wildlife Sanctuary, Thailand (Fig. 1).
Home range size (km2) Tiger Data collection period No. of fixes 95% MCP 100% MCP Male 1 May–Aug. 2005 542 246.1 291.2 Male 2 Feb.–Oct. 2008 333 174.8 197.0 Male 3 Apr.–June 2009 1,235 281.2 289.3 Male 4 Apr.–July 2009 1,899 218.5 234.0 Male 5 June–Nov. 2011 2,659 417.5 459.0 Female 1 Feb.–July 2005 470 57.0 69.84 Female 2 Jan.–Dec. 2005 529 52.9 78.2 Female 3 Feb. 2007–Feb. 2008 738 122.3 155.9 Female 4 Feb.–June 2010 1,537 117.1 133.5 Female 5 Mar.–July 2010 2,126 61.6 75.3 Female 6 Dec. 2009–July 2010 4,585 31.0 36.6 Female 7 Aug.–Nov. 2010 608 75.6 78.3 Female 8 May.–Nov. 2011 2,274 44.1 46.0 precise density estimation and therefore we used the density of wild boar in the Sanctuary (2.4 ± SD 0.05 km−2) estimated by Sukmasuang (2001). To estimate large bovid density by distance sampling we first tested models of dung observa- bility based on distance from the transects. During the surveys 105 bovid dung piles were detected. The best model for observability based on the lowest AIC score was a uniform cosine distribution. The bovid dung density was estimated to be 860 ± SD 126 km−2 and, from this, bovid density was estimated to be 3.0 ± SD 0.4 km−2.
Home range size and prey abundance FIG. 2 Correlation between the home range size of six female The simple linear regression indicated there was a sig- tigers and prey biomass in Huai Kha Khaeng Wildlife Sanctuary fi (Fig. 1). The linear regression indicates that the home range ni cant negative correlation between size of female fi 2 size of female tigers is signi cantly correlated (P , 0.05) with home range and prey biomass (r 5 0.70,P, 0.05,n5 6). prey density. This relationship appeared to be log-linear as a log transformation of the home range size gave a better (Litvaitis et al., 1986) and Canadian lynx Lynx canadensis fi 2 t (r 5 0.85,P, 0.05,n5 6; Fig. 2). At a regional scale (Ward & Krebs, 1985). Schaller (1972), Orsdol et al. (1985) (Table 2) we found the same relationship (Spearman and Hayward et al. (2007) also demonstrated this relation- rs 5 0.88,P, 0.005,n5 6). ship for large African felids. To test our hypothesis that a female’s home range Discussion should be large enough to support the prey requirements for a female and her cubs from birth until they disperse at Our results support the hypothesis that variation in the c. 18 months (Smith et al., 1987), we needed a home range home range size of female tigers is partially explained by estimator that defined the area in which a female acquires variation in prey density. The relationship appears to be food. We also needed to test the degree to which females log-linear, probably reflecting the biological upper and have exclusive home ranges. Choice of home range lower limits of home range sizes within the study area. This estimators depends on the question being addressed relationship indicates that home range size of female tigers is (Kie et al., 2010) and therefore we calculated home range relatively constant when prey biomass is . 5,000 kg km−2, as both the 95 and 100% MCPs (Smith et al., 1987; indicating minimum home ranges of 10–20 km2 (Fig. 2). Chundawat et al., 1999; Karanth & Sunquist, 2000; In carnivores a direct relationship between higher prey Goodrich et al., 2010; Barlow et al., 2011). For female tigers abundance and smaller home range sizes has been home ranges are not only related to prey abundance reported in many species, including bobcat Felis rufus but are exclusively used by an individual female.
© 2014 Fauna & Flora International, Oryx, 1–8 http://journals.cambridge.org Downloaded: 25 Mar 2014 IP address: 115.67.101.28 35
6 A. Simcharoen et al.
Tigers in Russia and Nepal exhibit territorial behaviour as a strategy to ensure access to adequate prey to raise their young to dispersal age (Smith et al., 1987, Goodrich et al., 2010). This behaviour is also common in other solitary female carnivores (Sandell, 1989). We did not document aggressive behaviour among females as an indicator of territorial behaviour but we did find that territorial overlap among females is small. For example, three pairs of adjacent females had an overlap of only 4%, which is similar Reza et al. (2002) Prey study Bhattarai & Kindlman (2012) This study to estimates from Nepal, where the overlap of females was 3.5–7% (Smith et al., 1987), and Russia, where female tigers had a mean overlap of 9% (Goodrich et al., 2010). Factors other than prey, such as poaching of tigers, may reduce competition among females and result in home ranges larger than needed to supply food resources. These conditions may confound the negative relationship between female tiger home range size and prey abundance. When density of tigers was low in Nepal during periods Barlow et al. (2011) Reference Karanth & Sunquist (2000)Smith et al. (1987) Karanth & Sunquist (1995) Goodrich et al. (2010)This study Stephens et al. (2006) Chundawat et al. (1999) Chundawat & Sharma (2008) when habitat was recovering from human disturbance, tiger home ranges were larger than needed and this allowed females to make room for their dispersing daughters (Smith et al., 1987). ) Home range study 2 − In contrast it is widely reported that home range size for male solitary carnivores is not influenced by food require- s range, including this study. 436 Biomass density (kg km 2,511 ments, as males attempt to hold territories that encompass as many females as possible (Sandell, 1989). The home range size, and thus the number of females that a male may mate with, is constrained by the energy and time requirements expended visiting females to track their reproductive status, and patrolling boundaries to discourage intruding males that may kill cubs (Sandell, 1989). Mean male home range size in Huai Kha Khaeng Wildlife Sanctuary was 3.8 times larger than the mean female home range. This ratio is ) Prey abundance
2 similar to the ratio of home range size between male and
12 Distance sampling 4,297 female tigers in Russia and Nepal (Smith et al., 1987; Goodrich et al., 2010). Although our sample size for the correlation between home range and prey is small, it is currently the best MCP home range size (km information available. Tracking the relationship between female home range size and prey abundance is a useful tool for managers because it provides them with a metric for 2 7 54 21 Distance sampling 4,257 setting management goals. Hayward et al. (2007) suggested that the next step in refining the relationship between female home range size and prey abundance is to analyse the Male Female Male Female Method No. of animals 1 1 243 27 Distance sampling 4,057 1 1 32 17 Distance sampling 5,482 6 21 1,160 394ecological Track count factors that predict prey abundance. Detailed information on female home range size, prey abundance and the ecological factors that determine prey abundance will allow managers to develop appropriate management actions. For example, when female home range sizes are larger than needed to supply adequate prey for breeding females to raise their young, managers should investigate the possibility that tigers are being poached. Alternatively, 2 Summary data for tiger home range size and prey abundance at six study sites across the tiger ’ when prey abundance is below the potential carrying ABLE Panna, India T Study site Nagarahole, India Chitwan, Nepal Sundarbans, Bangladesh Sikhote-Alin, Russia Huai Kha Khaeng, Thailand 4 7capacity 267 predicted 70by Pellet count environmental variables, managers
© 2014 Fauna & Flora International, Oryx, 1–8
http://journals.cambridge.org Downloaded: 25 Mar 2014 IP address: 115.67.101.28 36
Tiger home range and prey abundance 7
should examine whether prey are being poached. In this mangrove ecosystem for the species’ conservation. Oryx, contrast, if the mean size of home ranges is at the minimum 45, 125–128. size given the prey density, then the only management BHATTARAI, B.P. & KINDLMANN,P.(2012) Habitat heterogeneity as the key determinant of the abundance and habitat preference option to increase the viability of the tiger population is to of prey species of tiger in the Chitwan National Park, Nepal. enlarge the area available. If in the next few years female Acta Theriologica, 57, 89–97. home range size increases in Huai Kha Khaeng Wildlife BUCKLAND, S.T., ANDERSON, D.R., BURNHAM, K.P., LAAKE, J.L., Sanctuary, managers should investigate if it is because of BORCHERS, D.L. & THOMAS,L.(2001) Introduction to Distance declining prey abundance or an increase in tiger poaching. Sampling: Estimating Abundance of Biological Populations. Oxford University Press, Oxford, UK. Faced with continued threats to tigers, conservationists CARROLL,C.&MIQUELLE, D.G. (2006) Spatial viability analysis of have focused on strategies that emphasize conserving source Amur tiger Panthera tigris altaica in the Russian Far East: the role of populations (Walston et al., 2010) or entire landscapes protected areas and landscape matrix in population persistence. (Sanderson, 2010). These efforts provide important guide- Journal of Applied Ecology, 43, 1056–1068. lines, and strategic conservation planning is essential. CHAPRON, G., MIQUELLE, D.G., LAMBERT, A., GOODRICH, J.M., LEGENDRE, S. & CLOBERT,J.(2008) The impact on tigers of Poaching, reduced carrying capacity and the loss of the poaching versus prey depletion. Journal of Applied Ecology, fi land area available to tigers have been identi ed as the key 45, 1667–1674. threats to the species (Kenny et al., 1995; Chapron et al., CHUNDAWAT, R.S., GOGATE, N. & JOHNSINGH, A.J.T. (1999) 2008; Seidensticker, 2010). Managers can use the relation- Tigers in Panna: preliminary results from an Indian tropical ship between female home range size and prey abundance, dry forest. In Riding the Tiger: Tiger Conservation in Human-Dominated Landscapes (eds J. Seidensticker, S. Christie along with long-term camera-trap monitoring of tiger – fi & P. Jackson), pp. 123 129. Cambridge University Press, populations, to examine the threats speci c to tigers in each Cambridge, UK. protected area. CHUNDAWAT, R.S. & SHARMA,K.(2008) Tiger prey in a tropical dry forest: an assessment of abundance and of biomass estimation derived from distance sampling. Journal of the Bombay Natural Acknowledgements History Society, 105, 64–72. DHUNGEL, S.K. & O’GARA, B.W. (1991) Ecology of the hog deer in We acknowledge the Department of National Parks, Royal Chitwan National Park, Nepal. Wildlife Monographs, 119, 3–40. Wildlife and Plant Conservation, and especially DNP (2010) Thailand Tiger Action Plan 2010–2022. Department of Chatchawan Pitdumkam and Theerapat Prayurasiddhi, for National Parks, Wildlife and Plant Conservation, Bangkok, supporting this work. We also appreciate the help provided Thailand. by the former chiefs of Huai Kha Khaeng Wildlife DOWNS, J.A. & HORNER, M.W. (2008)Effects of point pattern Sanctuary, Sunthorn Chaiwatana and Apicha Yusombune. shape on home-range estimates. Journal of Wildlife Management, – Funding was provided by PTTEP (Thailand), the U.S. Fish 72, 1813 1818. GOODRICH, J.M., MIQUELLE, D.G., SMIRNOV, E.N., KERLEY, L.L., and Wildlife Service, The Royal Golden Jubilee PhD QUIGLEY, H.B. & HORNOCKER, M.G. (2010) Spatial structure Programme (Thailand) and the Biodiversity Research and of Amur (Siberian) tigers (Panthera tigris altaica) on Sikhote-alin Training Programme (Thailand). We are also grateful to Biosphere Zapovednik, Russia. Journal of Mammalogy, 91, 737–748. research assistants Onsa Norasarn, Thavorn Thadvichit, HARIHAR, A., PANDAV,B.&GOPAL, S.P. (2009) Responses of tiger Kerkpol Wongchu, Boonyang Srichan, Yingbune (Panthera tigris) and their prey to removal of anthropogenic influences in Rajaji National Park, India. European Journal Chongsomchai, Kerati Phetong, Nattakarn Pengmark, of Wildlife Research, 55, 97–105. Namkang Saelee and Parinyakorn Worawan for their help HARRIS, S., CRESSWELL, W.J., FORDE, P.G., TREWHELLA, W.J., with the capture of tigers and field data collection. We are WOODLARD, T. & W RAY,S.(1990) Home-range analysis using particularly grateful to Francesca J. Cuthbert, Erin Roche, radio-tracking data—a review of problems and techniques Anak Pattanavibool, Vijak Chimchom, John Fieberg particularly as applied to the study of mammals. Mammal Review, 20, 97–123. and Adam Barlow for their comments the article and for HAYWARD, M.W., O’BRIEN,J.&KERLEY, G.I.H. (2007) Carrying suggestions regarding data analysis. capacity of large African predators: predictions and tests. Biological Conservation, 139, 219–229. JENNRICH, R.I. & TURNER, F.B. (1969) Measurement of non-circular References home range. Journal of Theoretical Biology, 22, 227–237. JOHNSON, D.H. 1980. The comparison of usage and availability AKAIKE, H. (1974) A new look at the statistical model identification. measurements for evaluating resource preference. Ecology, 61, 65–71. IEEE Transactions on Automatic Control, 19, 716–723. KARANTH, K.U., NICHOLS, J.D., KUMAR, N.S., LINK, W.A. & HINES, J. BAILEY, R.E. & PUTMAN, R.J. (1981) Estimation of fallow deer E. (2004) Tigers and their prey: predicting carnivore densities from (Dama dama) populations from faecal accumulation. Journal prey abundance. Proceedings of the National Academy of Sciences of Applied Ecology, 18, 697–702. of the United States of America, 101, 4854–4858. BARLOW, A.C.D., SMITH, J.L.D., AHMAD, I.U., HOSSAIN, A.N.M., KARANTH, K.U. & SUNQUIST, M.E. (1992) Population structure, RAHMAN,M.&HOWLADER, A. (2011) Female tiger Panthera tigris density and biomass of large herbivores in the tropical forests home range size in the Bangladesh Sundarbans: the value of of Nagarahole, India. Journal of Tropical Ecology, 8, 21–35.
© 2014 Fauna & Flora International, Oryx, 1–8 http://journals.cambridge.org Downloaded: 25 Mar 2014 IP address: 115.67.101.28 37
8 A. Simcharoen et al.
KARANTH, K.U. & SUNQUIST, M.E. (1995) Prey selection by tiger, SMITH, J.L.D., MCDOUGAL, C.W. & SUNQUIST, M.E. (1987) Female leopard and dhole in tropical forests. Journal of Animal Ecology, land tenure system in tigers. In Tigers of the World (eds R.L. Tilson 64, 439–450. & U.S. Seal), pp. 97–109. Noyes Publications, Saddle River, USA. KARANTH, K.U. & SUNQUIST, M.E. (2000) Behavioural correlates SRIKOSAMATARA, S. (1993) Density and biomass of large herbivores of predation by tiger (Panthera tigris), leopard (Panthera pardus) and other mammals in a dry tropical forest, western Thailand. and dhole (Cuon alpinus) in Nagarahole, India. Journal of Zoology, Journal of Tropical Ecology, 9, 33–43. 250, 255–265. STEPHENS, P.A., ZAUMYSLOVA, O.YU., MIQUELLE., D.G., KENNY, J.S., SMITH, J.L.D., STARFIELD, A.M. & MCDOUGAL, C. MYSLENKOV, A.I. & HAYWARD, G.D. (2006) Estimating population (1995) The long-term effects of tiger poaching on population density from indirect sign: track counts and the Formozov– viability. Conservation Biology, 9, 1127–1133. Malyshev–Pereleshin formula. Animal Conservation, 9, 339–348. KERNOHAN, B.J., GITZEN, R.A. & MILLSPAUGH, J.J. (2001) Analysis SUKMASUANG,R.(2001) Ecology of barking deer (Muntiacus spp.) of animal space use and movements. In Radio Tracking and Animal in Huai Kha Khaeng Wildlife Sanctuary. PhD thesis. Kasetsart Populations (eds J.J. Millspaugh & J.M. Marzluff), pp 125–166. University, Bangkok, Thailand. Academic Press, San Diego, USA. SUNQUIST, M.E. (1981) The social organization of tigers (Panthera KIE, J.G., MATTHIOPOULOS, J., FIEBERG, J., POWELL, R.A., tigris) in Royal Chitwan National Park, Nepal. Smithsonian CAGNACCI, F., M ITCHELL, M.S. et al. (2010) The home-range Contributions to Zoology, 336, 1–98. concept: are traditional estimators still relevant with modern SUNQUIST, M.E., KARANTH, K.U. & SUNQUIST,F.(1999) Ecology, telemetry technology? Philosophical Transactions of the Royal behaviour and resilience of the tiger and its conservation needs. Society B, 365, 2221–2231. In Riding the Tiger (eds J.C.S. Seidensticker & P. Jackson), pp. 5–18. KREEGER, T.J. & ARNEMO, J.M. (2007) Handbook of Wildlife Chemical Cambridge University Press, Cambridge, UK. Immobilization. Sunquest Print, China. THOMAS, L., LAAKE, J.L., STRINDBERG, S., MARQUES, F.F.C., LITVAITIS, J.A., SHERBURNE, J.A. & BISSONETTE, J.A. (1986) BUCKLAND, S.T., BORCHERS, D.L. & MARQUES, T.A. (2006) Bobcat habitat use and home range size in relation to prey density. Distance 5.0. Release 2. Research Unit for Wildlife Population The Journal of Wildlife Management, 50, 110–117. Assessment, University of St. Andrews, UK. MIQUELLE, D.G., GOODRICH, J.M., SMIRNOV, E.N., STEPHENS, P.A., WALSTON, J., ROBINSON, J.G., BENNETT, E.L., BREITENMOSER, U., ZAUMYSLOVA, O.Y., CHAPRON, G. et al. (2010) The Amur Tiger: FONSECA, G.A.B., GOODRICH, J. et al. (2010) Bringing the tiger A Case Study of Living on the Edge. Oxford University Press, back from the brink—the six percent solution. PLoS Biology, Oxford, UK. 8,e1000485. ORSDOL, K.G., HANBY, J.P. & BYGOTT, J.D. (1985) Ecological correlates WARD, R.M.P. & KREBS, C.J. (1985) Behavioural response of lynx to of lion social organization (Panthera leo). Journal of Zoology, 206, declining snowshoe hare abundance. Canadian Journal of Zoology, 97–112. 63, 2817–2824. REZA, A.H.M.A., FEEROZ, M.M. & ISLAM, M.A. (2002) Prey species WEFCOM (WESTERN FOREST COMPLEX)(2004) GIS Database and density of Bengal tiger in the Sundarbans. Journal of the Asiatic its Applications for Ecosystem Management. The Western Forest Society of Bangladesh, Science, 28, 35–42. Complex Ecosystem Management Project, Department of National RODGERS, A.R. & CARR, A.P. (1998) HRE: the Home Range Park, Wildlife, and Plant Conservation, Bangkok, Thailand. Extension for Arc View. User’s Manual. Ontario Ministry of Natural WEGGE, P., O DDEN, M., POKHAREL, Pd.C. & STORAAS, T. (2009) Resources, Peterborough, Canada. Predator–prey relationships and responses of ungulates and their SANDELL,M.(1989) The mating tactics and spacing behaviour of predators to the establishment of protected areas: a case study solitary carnivores. In Carnivore Behaviour, Ecology and Evolution of tigers, leopards and their prey in Bardia National Park, Nepal. (ed. J.L. Gittleman), pp. 164–182. Cornell University Press, Biological Conservation, 142, 189–202. New York, USA. WHITE, G.C. & GARROTT, R.A. (1990). Analysis of Wildlife Radio SANDERSON, E.W. (2010) Setting priorities for tiger conservation: Tracking Data. Academic Press, San Diego, USA. 2005–2015. In Tigers of the World: The Science, Politics, and WIKRAMANAYAKE, E.D., DINERSTEIN, E., ROBINSON, J.G., Conservation of Panthera tigris (eds R. Tilson & P.J. Nyhus), KARANTH, U., RABINOWITZ, A., OLSON, D. et al. (1998) An pp. 143–161. Academic Press, San Diego, USA. ecology-based method for defining priorities for large mammal SCHALLER, G.B. (1972) The Serengeti Lion: A Study of Predator–Prey conservation: the tiger as case study. Conservation Biology, Relations. University of Chicago Press, Chicago, USA. 12, 865–878. SEAMAN, D.E. & POWELL, R.A. (1996) An evaluation of the accuracy of kernel density estimators for home range analysis. Ecology, 77, 2075–2085. Biographical sketches SEIDENSTICKER,J.(1976) On the ecological separation between tigers and leopards. Biotropica, 8, 225–234. A CHARA S IMCHAROEN specializes in tiger ecology and tiger SEIDENSTICKER,J.(2010) Saving wild tigers: a case study in biodiversity movement behaviour. T OMMASO S AVINI’ S research includes the loss and challenges to be met for recovery beyond 2010. Integrative behavioural ecology of primates and pheasants. G EORGE A. GALE Zoology, 5, 285–299. conducts research on the ecology and population dynamics of birds SIMCHAROEN, S., PATTANAVIBOOL, A., KARANTH, K.U., and mammals. S AKSIT S IMCHAROEN studies the ecology and habitat NICHOLS, J.D. & KUMAR, N.S. (2007) How many tigers Panthera use of tigers and leopards. S OMPHOT D UANGCHANTRASIRI is in tigris are there in Huai Kha Khaeng Wildlife Sanctuary, Thailand? charge of tiger monitoring in the Western Forest Complex of Thailand. An estimate using photographic capture–recapture sampling. Oryx, S OMPORN P AKPIEN undertakes research on tigers and their prey. 41, 447–453. J AMES L. D. SMITH investigates landscape-scale tiger ecology.
© 2014 Fauna & Flora International, Oryx, 1–8
http://journals.cambridge.org Downloaded: 25 Mar 2014 IP address: 115.67.101.28 38
Conservation & Ecology RAFFLES BULLETIN OF ZOOLOGY 62: 100–106 Date of publication: 11 March 2014 http://zoobank.org/urn:lsid:zoobank.org:pub:3528409A-62CE-4F2D-8713-06D47FF98585
Ecological factors that infl uence sambar (Rusa unicolor) distribution and abundance in western Thailand: implications for tiger conservation
Achara Simcharoen1, Tommaso Savini1, George A. Gale1, Erin Roche2, Vijak Chimchome3 & James L. D. Smith4
Abstract. Prey density is declining throughout the tiger’s (Panthera tigris) range and knowledge of the ecological factors that affect prey distribution and abundance remains surprisingly limited for this globally endangered species. In this study, we examined the ecological variables infl uencing the abundance of sambar (Rusa unicolor), the dominant prey species for the tiger across its global southern range. We also identifi ed the scale at which these variables impact sambar distribution in Huai Kha Khaeng Wildlife Sanctuary, a high tiger density site in Southeast Asia. The fecal pellet group accumulation method was used to estimate an index of sambar abundance. Pellet groups were counted along 360 line transects randomly placed among four approximately 100 km2 sites that encompassed six female tiger home ranges. The relationship between sambar pellet-group counts and 10 environmental variables was investigated using generalised linear mixed models. The sambar abundance index was negatively associated with distance to the largest river in the study area, elevation, and the amount of dry deciduous dipterocarp forest cover. Distribution and abundance of sambar were positively associated with relatively fl at areas of river valleys, presumably due to the quality of vegetation available for foraging and greater visibility for detecting predators compared to other portions of the study area. This study is the fi rst to identify the importance of wide alluvial valleys to tiger prey and suggests this habitat is critical for securing one of the largest tiger source populations in Southeast Asia.
Key words. Panthera tigris corbetti, tiger prey, habitat selection, Huai Kha Khaeng Wildlife Sanctuary
INTRODUCTION 1995; Hayward et al., 2012), studies (e.g., Ngampongsai, 1987; Padmalai et al., 2003; Matsubayashi et al., 2007 Approximately 60% of the published studies on tiger Bhattarai & Kindlmann, 2012) have rarely quantifi ed the (Panthera tigris) diet report that the sambar (Rusa unicolor) habitat preferences of the tiger’s primary prey, the sambar. is the dominant prey species in terms of biomass in South Understanding the habitat requirements of this species is and Southeast Asia (e.g., Seidensticker & McDougal, clearly needed to predict the distribution and carrying capacity 1993; Karanth & Sunquist, 1995; Biswas & Sankar, 2002). of tigers, and where appropriate, to manage the habitat to Additionally, Hayward et al. (2012) reported that sambar improve conditions for preferred prey. The sambar is also an are one of the two most preferred prey species throughout important prey species because its widespread distribution the entire range of tigers, and hypothesized that sambar in Asia largely overlaps that of the tiger in South Asia, importance in the tiger diet is a consequence of the nearly South China and Southeast Asia (Corbett & Hill, 1992). 1:1 predator to prey weight ratio between these two species, Across its range, the sambar is highly adaptive; it occurs a common relationship reported for large cats. Ackerman in a wide diversity of habitats, ranging from ocean shores (1986) provided an ecological rationale for this ratio in the to subalpine regions, and consumes a varied diet including mountain lion (Puma concolor) whereby a female needs coarse grasses, woody browse, broad-leaved foliage, fruit, and to kill prey her size, or larger, to meet her own energetic partially submerged water plants (Geist, 1998). In Thailand, requirements as well as those of her maturing offspring sambar is the largest of the cervid species and historically which can weigh as much or more than she does prior to it had the widest distribution in the region (Lekagul & their independence. Despite previous research on tiger diet McNeely, 1977; Francis, 2008). Given the signifi cance of and the proposed optimum prey size (Karanth & Sunquist, sambar in the diet of tigers and widespread documentation of prey depletion across the tiger’s range, including Thailand (Ramakrishnan et al., 1999; Johnson et al., 2006; Datta et
1Conservation Ecology Program, King Mongkut’s University of Technology, Thonburi, al., 2008), a better understanding of the ecological factors Thailand; Email: [email protected] (*corresponding author) that infl uence sambar abundance and spatial distribution is 2Department of Biological Sciences, University of Tulsa, USA needed in critical tiger habitats. The objectives of this study 3Department of Forest Biology, Kasetsart University, Thailand were to: 1) assess an index of abundance and distribution of 4Department of Fisheries, Wildlife and Conservation Biology, University of Minnesota, sambar in the Huai Kha Khaeng Wildlife Sanctuary (HKK) USA a key site for tigers in western Thailand; and 2) identify © National University of Singapore ecological factors associated with sambar distribution. This ISSN 2345-7600 (electronic) | ISSN 0217-2445 (print) information will provide conservation managers with a set
100 39
RAFFLES BULLETIN OF ZOOLOGY 2014 of metrics to select and evaluate management actions for within four sites where home ranges of tiger have been sambar and tigers. studied (Fig. 1). This deer survey was primarily designed to investigate food availability in different tiger home ranges MATERIAL AND METHODS to represent the array of ecological diversity within the sanctuary. These sites cover about 14% of the total sanctuary Study area. This study was conducted in HKK, which area and represent the three major forest types present (we is located in western Thailand (15°00'–15°50'N, 99°00'– excluded hill evergreen forest which covers approximately 99°19'E) and represents the core of a forest complex 13% of the sanctuary). Although not optimal, we are confi dent consisting of 17 contiguous protected areas (collectively that data obtained in this survey can be cautiously used to known as the Western Forest Complex or WEFCOM). This explain the sambar distribution in the entire sanctuary. Each region supports the largest tiger population in Thailand (DNP, of the four sites is approximately 100 km2 encompassing six 2010). The sanctuary is 2,780 km2 and elevation varies from female tiger home ranges. These sites were also chosen to 200–1,600 m. It has a tropical monsoonal climate with an represent the range of ecological conditions in HKK that annual temperature ranging from 8°C in January to 38°C might refl ect differences in both deer density and tiger home in April. Total rainfall averages 1,386 mm (2000–2011), range size. Two sites (KBD and KYD) were in the central but most (78%) occurs in the wet season (May–October). valley of the sanctuary along the lower and upper portions From late November until April, dry season forest fi res are of the Huai Kha Khaeng River. The other sites (KNR and common. This seasonal variation in temperature and rainfall KPP) were drier areas away from Huai Kha Khaeng River. results in a general dry deciduous forest mosaic. Depending For this project, deer sampling did not include the full on rainfall patterns, edaphic factors, and fi re frequency, range of the ecosystems found in the WEFCOM, but was four primary vegetation types occur: mixed deciduous limited to a subsample due to the intensive sampling needed forest (48%); dry evergreen forest (25%); hill evergreen to estimate sambar numbers and the limited budget and forest (13%); and dry deciduous dipterocarp forest (7%) personnel. Specifi cally, we did not sample the roughly 22% (WEFCOM, 2004). A central feature of HKK is the 100 of HKK above 900 m because sambar are rarely observed km long Huai Kha Khaeng River, which drains the central in this steep terrain (Trisurat et al., 2010). At each site we valley from north to south (Fig. 1). Many temporary and two smaller permanent streams originate in the rugged hills and empty into the main river. The lower part of this central valley is wide and less steep and is characterised by richer alluvial soils than the upper part of the valley. The HKK tiger population is currently estimated to be between 59–77 breeding individuals (DNP, 2010). Mean female home range size is 70 km2 and mean male home range size is 267 km2 (Simcharoen et al., 2014). The fi ve major prey species of the tiger in HKK are banteng (Bos javanicus), sambar, gaur (B. gaurus), wild boar (Sus scrofa), and red muntjac (Muntiacus muntjak), but banteng, sambar, and gaur account for 87% of the biomass consumed by tigers (Petdee, 2000). In large parts of WEFCOM, however, sambar and other large prey species of tigers are absent or in decline (Steinmetz et al., 2010).
Sambar distribution and relative abundance. An index of sambar abundance was estimated by using the fecal pellet group accumulation method. This technique has been widely applied in ungulate habitat studies to obtain both absolute estimates as well as indices of deer abundance (Rogers et al., 1958; Mitchell et al., 1977; Bailey & Putman, 1981). The method is typically applied in habitats in which animals are diffi cult to count directly or where the assumptions of distance sampling are likely violated (e.g., in dense habitat animals secretively move from sight before being observed; Smart et al., 2004; Wegge & Storaas, 2009). Because > 70% of the habitat in HKK (e.g., dry evergreen forest, mixed deciduous forest) has dense ground cover, we assumed that sambar would be diffi cult to count directly. Field work was conducted during the dry season (November–April) between 2009 and 2011.
Fig.1 Location of Huai Khaeng Wildlife Sanctuary in western Within the Wildlife Sanctuary we focused our investigation Thailand and the 360 sampling units contained within the four on deer distribution and relative abundance in areas located areas represent six female tiger home ranges.
101 40
Simcharoen et al.: Ecological factors for sambar prey of tiger randomly located 90 transects each forming a 200 × 200 m variables were then added to the model. We also investigated square, totalling 800 m in diameter, for a total of 360 square the interaction between the distance from Huai Kha Khaeng transects across the four sites. In each square transect, we River and the elevation as a variable because our previous place 40 circular plots, each measuring 20 m2, approximately experience suggested that both tigers and sambar avoided 20 m apart (for a total of 3,600 plots per site and 14,400 higher elevations near the main river. We accepted the model plots in total). We cleared plots of all pellet groups and then with the lowest AIC value as the best representation of the returned to the plots after 30 days to estimate the rate of relationship between pellet group counts and ecological pellet group accumulation. Pellet groups were counted and factors (Burnham & Anderson, 2010). The analysis was distinguished from each other following Simcharoen et al. performed using the MASS package (Venables & Ripley, (2014). We considered the total number of pellet groups 2002) in program R (R Core Team, 2012). divided by the total number of transects as an index of sambar abundance and number of sambar pellet groups within each RESULTS transect as an index of habitat use. Sambar distribution and index of abundance. We searched Ecological variables. Based on previous work by Trisurat et sambar pellet groups on a total of 360 square transects in al. (2010) and our knowledge of the species, we identifi ed four sites representing six female tiger home ranges that 10 ecological variables which we hypothesized to potentially ranged from 200–900 msl in altitude and included three main infl uence sambar distribution and abundance in HKK. Nine habitat types recorded in HKK: 75.3% MD, 8.9% DD, and were derived using a GIS database prepared by the Western 15.7% DE. Habitat proportions for each site were 72% MD, Forest Complex Ecosystem Management Project, Department 5% DD, and 23% DE for KPP; 73% MD, 9% DD, and 18% of National Parks, Wildlife and Plant Conservation DE for KYD; 70% MD, 21% DD, and 9% DE for KNR; (WEFCOM, 2004). Variables included terrain ruggedness, and 88% MD, 2% DD, and 10% DE for KBD. The median slope, elevation (m), distance to all streams (m), distance to slope for the square transects ranged from 0–23%. Sambar Huai Kha Khaeng River (m), distance to a salt lick (m), and pellet groups were found in 63% of all square transects. In average low slope patch size (areas with slopes shallower total, we recorded 1,041 pellets groups (102 in KPP, 444 in than 10%). Because Rotenberry et al. (2006) emphasised KYD, 26 in KNR, and 469 in KBD). Per habitat type we the need to consider multiple scales in habitat use studies recorded 48 pellet groups in DD forest, 246 in DE forest, we measured slope and the average low slope patch size at and 747 in MD forest. The mean number (±SD) of pellet two scales (150 and 600 m radius from the center of each groups recorded in all square transects was 2.9±4.6. The transect). These scales are our best estimate of the range of mean number of pellets in a pellet-group was 32±14. The areas in which ecological variables would attract a sambar relative abundance of sambar in HKK was 3,615±3,180 pellet to the vicinity of our transects. Vegetation type was recorded groups km-2. A Mann-Whitney U test was used to examine during our fi eld surveys and classifi ed for each of the 1,440 the effect of distance to the main river on sambar pellet circular plots. For each circular plot we recorded vegetation abundance. There was a signifi cant difference in the median type as either mixed deciduous forest (MD), dry deciduous number of sambar pellet groups in the square transects of dipterocarp forest (DD), dry evergreen forest (DE) or bamboo KBD and KYD, sites located close to the Huai Kha Khaeng forest (BB). The habitat type for each square transect was River (Fig. 1; median=3 pellet groups square transect-1; assigned as the proportion of the habitat types recorded in abundance=6,340±245 pellets group km-2), and the median its 40 circular plots. Bamboo was later lumped with mixed in the other two sites KNR and KPP, located further from deciduous forests both in our square transects and in the the Huai Kha Khaeng River (Fig. 1), ( median=0 pellets overall forest type availability map in the WEFCOM GIS. group square transect-1; abundance=889±746 pellets group Terrain ruggedness was measured as a vector ruggedness km-2); (Mann-Whitney U test, Z=−9.31, p=0.05, n=180). measure (VRM) using an ArcView script. Ruggedness values range from 0 (no terrain variation) to 1 (complete terrain Ecological variables. Results from the correlation tests variation). The value of natural terrain ranges from 0–0.4 suggested that sambar pellet abundance had a strong negative (Sappington et al., 2007). We defi ned “slope” as the median association with the distance to the Huai Kha Khaeng slope and the “average low slope patch size” as the total low River (r=−0.34, p<0.0001). Sambar pellets were negatively slope area within a 150 m and 600 m radius. associated with the occurrence of dry deciduous dipterocarp forest (r=−0.16, p=0.001) and positively associated with Data analysis. We investigated the relationship between the average low slope patch size within a 600 m radius sambar pellet-group counts (response variable) and the above (r=0.16, p=0.001) and 150 m radius of square transect centers 10 environmental variables using generalised linear mixed (r=0.14, p=0.004). Terrain ruggedness was discarded from models with negative binomial error terms. We did not the variables tested because it was highly correlated with standardise these variables when performing the regression slope (r=0.76) and slope was more correlated with pellet analyses. To reduce possible effects of multicollinearity, we group abundance. The remaining nine ecological factors discarded one independent variable of each tested pair if the were deemed to be uniquely associated with pellet group between variable association had an r-value>0.5 (Torres et al., counts (Table 1). Following model selection, only three of the 2011).The choice of which correlated variable to remove was ecological variables (dist. hkk, DD, and elev) were included in based on the relative strength of its Pearson correlation with the best-fi t model (Table 2). Our top model suggests that the the frequency of pellet counts. The remaining independent interaction between distance to the main river and elevation
102 41
RAFFLES BULLETIN OF ZOOLOGY 2014
Table 1. Variables used to identify factors potentially affecting sambar (Rusa unicolor) distribution and their levels of signifi cance (Pearson correlations).
Variable p-value r-value Distance to main stream (dist.hkk) <0.001 −0.340 Average low slope patch size of 600 radius (patch.600) 0.001 0.166 Habitat type : dry dipterocarp (DD) 0.001 −0.162 Average low slope patch size of150 radius (patch.150) 0.004 0.140 Elevation (elev). 0.080 −0.092 Habitat type : Mixed deciduous (MD) 0.100 0.086 Distance to salt lick (salt lick) 0.500 −0.035 Slope 600 radius (slp.600) 0.729 −0.018 Slope 150 radius (slp.150) 0.733 −0.017 Habitat type : Dry evergreen (DE) 0.770 0.015 Distance to any stream (dist.st) 0.995 0.0003
Table 2. Candidate models of sambar (Rusa unicolor) occurrence in Huai Kha Khaeng Wildlife Sanctuary, Thailand. Here we report
Akaike’s Information Criterion values (AIC), the difference in AIC rank relative to the top model (∆AIC), the relative model weights (wi), the number of parameters in the model (k), and the model deviance (Dev.). Full variable names and abbreviations are listed in Table 1.
Candidate Model AIC ∆AIC wi k Dev. Dist.hkk*elev+DD 1446.2 0 0.56 6 359.2 Dist.hkk+DD 1447.9 1.7 0.24 4 360.1 Dist.hkk+DD+patch.600 1448.7 2.5 0.16 5 360.4 Dist.hkk+patch.600 1452.5 6.3 0.02 4 361.3
Table 3. Regression coeffi cient estimates and standard errors (SE) for in this habitat type however, we do not have evidence to the top-supported models for sambar abundance (Table 2, model 1). support this interpretation.
Variable Estimate SE Sambar abundance. Sambar is an important tiger prey intercept 2.417e+00 3.307e-01 species and is preferred in the diet of tigers throughout dist.hkk −5.860e-04 1.296e-04 Southeast and South Asia (Hayward et al., 2012). Historically, elev −1.714e-03 6.793e-04 sambar occurred throughout the tiger’s range, with the − DD 2.445e-02 1.040e-02 exception of northern China and the Far East of Russia dist.hkk*elev 6.222e-07 2.838e-07 where it is replaced by a close relative, the red deer, Cervus elaphus, (Miquelle et al., 2010, Hayward et al., 2012). affect sambar abundance (regression parameter estimate for However, sambar is one of several large mammal species the interaction of dist.hkk*elev.=6.22×10-7, SE=2.84×10-7, that has recently experienced major declines primarily due p<0.01). Sambar abundance was greatest at low elevations to habitat degradation and poaching (Wikramanayake et al., near the Huai Kha Khaeng River (Fig. 2; Table 3). Sambar 1998, Linkie et al., 2003, O’Brien et al., 2003) and is now abundance was negatively associated with increases in DD consider globally threatened (Timmins et al., 2008). Pellet habitat (regression parameter estimate for DD=−2.44 x10-2, groups were used as an index of sambar abundance. Although SE=2.84×10-2, p<0.01) it is important not to assume that indices are automatically linked to the actual abundance of the species, we felt that DISCUSSION fecal accumulation provides a reasonable estimate of relative abundance because it appeared that decomposition rates were This study examined sambar distribution and index of similar across habitat types. abundance at landscape and micro scales in relation to habitat variables and it focused on habitats typical of the Ecological variables. We examined nine ecological variables core area of the tiger population in Thailand’s Western hypothesized to predict sambar pellet abundance. Pellet group Forest Complex. Our results indicated that the distribution abundance appeared to be correlated with distance to Huai and index of abundance of sambar in HKK was related to Kha Khaeng River. The river valley is topographically fl atter distance from the main river (Huai Kha Khaeng River) and than the rest of the sanctuary and it has a relatively high elevation; abundance was greater in areas closest to the main percentage of grass species (Kruuk et al., 1994). Shrestha river at lower elevations where the predominant habitat (2004) reported that ungulates in lowland Nepal prefer type is mixed deciduous forest. In addition, dry deciduous similar low-lying areas, particularly fl ood plains with grass dipterocarp was negatively correlated with the index of and riverine forests. Trisurat et al. (2010) noted that sambar abundance. This could be related to a lower defecation rate avoid steep terrain and prefer open habitat; Bagchi et al.
103 42
Simcharoen et al.: Ecological factors for sambar prey of tiger
(2003) also suggested that sambar preferred grassland and water itself that makes the main river attractive to sambar; dense shrubs closer to water in India. Additionally, McKay instead, a complex of ecological parameters derived from & Eisenberg (1974) noted that sambar are sedentary and do other geologic and geographic features of the valley likely not shift their ranges seasonally. Our study aimed to model produce a desirable mixed deciduous forest habitat. Gallery sambar distribution in relation to ecological variables and to forests, the typical habitat along the Huai Kha Khaeng river, identify the importance of wide alluvial valleys to tiger prey. described as both seasonally fl ooded areas and non-fl ooded areas of mixed deciduous forest (Chimchome et al., 1998), Distance to the main river had the highest correlation with might have a higher food availability due to the constant pellet group abundance, but it is important to note that, in water supply (Budke et al., 2008) compared to the same contrast, there was no correlation of deer pellet abundance habitat type found away from this permanent water source. to distance of smaller permanent streams. It is clearly not Moreover, this habitat had denser understudy vegetation
Fig. 2 The distribution of elevations at which transects were placed (a). The distribution of distances from the Huai Kha Khaeng River (HKK) at which transects were placed (b). The combinations of elevation and distance to HKK River at which pellet groups were found (each dot represents a single transect) (c). The number of pellet groups found in relation to distance from the HKK River at three elevations (where dots represents transects) (d). Dots represent transects and lines represent the predicted number of pellet groups using our top-supported model (Table 2, model 1) solved at mean covariate values and one of three elevations to illustrate the interaction of elevation with distance to the HKK River.
104 43
RAFFLES BULLETIN OF ZOOLOGY 2014 which might offer better concealment from predators. On a ACKNOWLEDGEMENTS geological time scale, gradual erosion has likely deposited richer soils near the main river resulting in the growth of We thank the U.S. Fish and Wildlife Service (Rhinoceros more nutritional ungulate forage in this zone. Similarly, higher and Tiger Conservation Fund), PTT Exploration and soil moisture is more likely at lower elevations near the main Production Public Company Limited, Royal Golden Jubilee river. These two factors are represented in the interaction Grant, and Biodiversity Research Training Thailand, Grant between the main river and elevation. Soil quality and number BRT-T-254010 for funding this research. We also moisture might also be expected to affect food availability thank the Department of National Parks, Wildlife and Plant following Shrestha’s (2004) observation of sambar preference Conservation for study permission. Todd Arnold provided for the fl ood plains of Nepal and similarly in India sambar’s statistical advice and Mayuree Umpormjan provided GIS preference for lower elevations (Bagchi et al., 2003). data. Saksit Simcharoen, Sompot Duangchantrasiri, Somporn Pakpien, and Peter Cutter helped develop the prey survey Although overall presence of water was not highly correlated design in HKK. Kirati Phetthong, Thani Daoruang, Dolroman with sambar numbers, lack of water combined with poorer Chatson, and KNR staff organised the prey fi eld surveys. soils found in dry deciduous dipterocarp forest may account Finally, we thank Anak Pattanavibool and Francie Cuthbert for the lower density of sambar in this vegetation type. The for comments on earlier drafts of this manuscript. negative response to dry deciduous dipterocarp may refl ect a seasonal shift of sambar away from this habitat. For LITERATURE CITED example, Srikosamatara (1993) reported that sambar density in HKK in the dry season was lower than the wet season. Ackerman BB, Lindzey FG & Hemker TP (1986) Predictive However, even in the wet season ground cover is limited energetics model for cougars. In: Miller SD & Everett D (eds.) and grass is much sparser in dry deciduous dipterocarp forest Cats of the World: Biology, Conservation, and Management. National Wildlife Federation, Washington DC. Pp. 333–352. (Smitinand, 1977). The lower relative abundance of sambar Bagchi S, Goyal SP & Sankar K (2003) Prey abundance and in dry deciduous dipterocarp forest versus mixed deciduous prey selection by tigers (Panthera tigris) in a semi-arid, dry forest is further indicated by a preliminary analysis of long- deciduous forest in western India. Journal of Zoology, London, term camera trapping data where detection in DD was two 260: 285–290. times lower than in MD (S. Duangchantrasiri, unpublished Bailey RE & Putman RJ (1981) Estimation of fallow deer (Dama data). Finally, a study in Nagarahole India demonstrated that dama) populations from faecal accumulation. Journal of Applied the density of sambar in mixed deciduous forest habitat was Ecology, 18: 697–702. Bhattarai BP & Kindlmann P (2012) Habitat heterogeneity as the seven times that of dry deciduous dipterocarp forest (Karanth key determinant of the abundance and habitat preference of & Sunquist, 1992). Therefore, a number of factors appear prey species of tiger in the Chitwan National Park, Nepal. to make this forest type less attractive to sambar than the Acta Theriol, 57: 89–97. alluvial valleys. Biswas S & Sankar K (2002) Prey abundance and food habit of tigers (Panthera tigris tigris) in Pench National Park, Madhya Conservation implications. Our results have subtle but Pradesh, India. Journal of Zoology, London, 256: 411–420. important implications for sambar and tiger conservation. Budke JC, Jarenkow JA & Oliveira-Filho AT (2008) Tree community features of two stands of riverine forest under different fl ooding With widespread prey depletion occurring globally in regimes in Southern Brazil. Flora, 203: 162–174. response to habitat degradation and poaching (Sanderson et Burnham KP & Anderson DR (2010). Model Selection and Multi- al., 2002; Linkie et al., 2003; O’Brien et al., 2003), increased model Inference: A Practical Information-Theoretic Approach. patrolling and other forms of management (e.g., restoration Springer-Verlag, New York. 488 pp. of degraded land, management on trans-boundary lines Corbett GB & Hill JE (1992) The Mammals of the Indomalayan between tiger ranges) to reduce human impacts are needed. Region. A Systematic Review. Oxford University Press, USA. Given fi nancial constraints, it is important to target sambar 488 pp. Chimchome V, Vidhidharm A, Simchareon S, Bumrungsri S & management efforts in areas where there is likely to be a Poonswad P (1998) Comparative study of the breeding biology reasonable return for conservation efforts. For example, and ecology of two endangered hornbill species in Huai Kha investment to restore sambar numbers in rugged terrain in Khaeng Wildlife Sanctuary,Thailand. In: Poonswad P (ed) Kuiburi National Park, Thailand, met with little success The Asian Hornbills: Ecology and Conservation. Biodiversity (Steinmetz et al., 2009). Our study suggests that even with Research and Training Program, National Center for Genetic signifi cant effort, such rugged habitat is not likely to support a Engineering and Biotechnology. Pp. 111–136. dense population of sambar. We recommend that prime areas Datta A, Anand MO & Naniwadekar R (2008) Empty forests: large carnivore and prey abundance in Namdapha National Park, to target should be the wide interior valleys of the WEFCOM, north-east India. Biological Conservation, 141: 1429–1435. and elsewhere in Asia where the habitat is more favorable. DNP (2010) Thailand Tiger Action Plan 2010–2022. Department In WEFCOM, some of these valleys, though they occur in of National Parks, Wildlife and Plant Conservation, Bangkok, wildlife sanctuaries, are currently used by small villages Thailand. 55 pp. that inhabited the area before it was designated as a wildlife Francis CM (2008). A Field Guide to the Mammals of Thailand sanctuary. Our results suggest that these valleys should be a and South-East Asia. New Holland Publishers, UK. 392 pp. high priority for management because the habitat is suitable Geist V (1998) The Deer of the World, Their Evolution, Behavior, and Ecology. Stackpole Books, Mechanicsburg, MA. 432 pp. and lie within the area of the largest source population of Hayward MW, Jędrzejewski W & Jêdrzejewska B (2012) Prey tigers in Southeast Asia (Walston et al., 2010). preferences of the tiger Panthera tigris. Journal of Zoology, 296: 221–231.
105 44
Simcharoen et al.: Ecological factors for sambar prey of tiger
Johnson A, Vongkhamheng C, Hedemark M & Saithongdam T Sappington JM, Longshore KM & Thompson DB (2007) (2006) Effects of human-carnivore confl ict on tiger (Panthera Quantifying landscape ruggedness for animal habitat analysis: tigris) and prey populations in Lao PDR. Animal Conservation, a case study using bighorn sheep in the Mojave Desert. Journal 9: 421–430. of Wildlife Management, 71: 1419–1426. Karanth KU & ME Sunquist (1992) Population structure, density Seidensticker J & McDougal C (1993) Tiger predatory behavior, and biomass of large herbivores in the tropical forests of ecology and conservation. Zoological Society of London Nagarahole, India. Journal of Tropical Ecology, 8: 21–35. Symposium, 65: 105–125. Karanth KU & ME Sunquist (1995) Prey selection by tiger, leopard Shrestha MK (2004) Relative Ungulate Abundance in a Fragmented and dhole in tropical forests. Journal of Animal Ecology, 64: Landscape: Implications for Tiger Conservation. Unpublished 439–450. PhD Thesis, University of Minnesota, USA. 99 pp. Kruuk H, Kanchanasaka B, O’Sullivan S & Wanghongsa S (1994) Simcharoen A, Savini T, Gale GA, Simcharoen S, Duangchantrasiri Niche separation in three sympatric otters Lutra perspicillata, S, Pakpien S & Smith JLD (2014) Female tiger home range L. lutra and Aonyx cinerea in Huai Kha Khaeng, Thailand. size and prey abundance: important management metrics. Biological Conservation, 69: 115–120. Oryx, in press. Lekagul B & McNeely JA (1977) Mammals of Thailand. Kulusapa, Smart JCR, Ward AI & White PCL (2004) Monitoring woodland Bangkok. 760 pp. deer populations in the UK: an imprecise science. Mammal Linkie M, Martyr DJ, Holden J, Yanuar A, Hartana AT, Sugardjito J Review, 34: 99–114. & Leader-Williams N (2003) Habitat destruction and poaching Smitinand T (1977). Vegetation and ground cover of Thailand. threaten the Sumatran tiger in Kerinci Seblat National Park, Department of Forest Biology, Kasetsart University, Technical Sumatra. Oryx, 37: 41–48. Paper, 1: 160–171. Matsubayashi H, Lagan P, Abd. Sukor JR & Kitayama K (2007) Srikosamatara S (1993) Density and biomass of large herbivores Seasonal and daily use of natural licks by sambar deer (Cervus and other mammals in a dry tropical forest Western, Thailand. unicolor) in a Bornean tropical rain forest. Tropics, 17: 81–85. Journal of Tropical Ecology, 9: 33–43. McKay GM & Eisenberg JF (1974) Movement patterns and habitat Steinmetz R, Seuaturien N, Chutipong W, Chamnankit C & Poonil utilization of ungulates in Ceylon. In: Geist V & Walther F (eds.) B (2009) The Ecology and Conservation of Tigers and their The Behavior of Ungulates and its Relation to Management. Prey in Kuiburi National Park, Thailand. WWF Thailand and IUCN Publication, Morges, Switzerland. Pp.708–721. Department of National Parks, Wildlife and Plant Conservation, Miquelle DG, Goodrich JM, Smirnov EN, Stephens PA, Zaumyslova Bangkok, Thailand. 58 pp. OY, Chapron G, Kerley L, Murzin AA, Hornocker MG & Steinmetz R, Chutipong W, Seuaturien N, Chirngsarrd E & Quiley HB (2010) The Amur tiger: a case study of living Khaengkhetkarn M (2010) Population recovery patterns on the edge. In: Loveridge AJ & Macdonald WD (eds.) The of Southeast Asian ungulates after poaching. Biological Biology and Conservation of Wild Felid. Oxford University, Conservation, 143: 42–51. Oxford, UK. Pp. 325–340. Timmins RJ, Steinmetz R, Sagar Baral H, Samba Kumar N, Mitchell B, Rowe JJ, Ratcliffe PR & Hinge M (1977) The Ecology Duckworth JW, Anwarul Islam Md, Giman B, Hedges S, Lynam of Red Deer: a Research Review Relevant to Their Management AJ, Fellowes J, Chan BPL & Evans T (2008) Rusa unicolor. in Scotland. Institute of Terrestrial Ecology, Cambridge, UK. IUCN Red List of Threatened Species. Version 2013.1. www. 74 pp. iucnredlist.org (Accessed 06 September 2013) Ngampongsai C (1987) Habitat use by the sambar (Cervus unicolor) Torres RT, Santos J, Linnell JDC, Virgós E & Fonseca C (2011) in Thailand: a case study for Khao-Yai National Park. In: Factors affecting roe deer occurrence in a Mediterranean Wemme CM (ed.) Proceedings of the Biology and Management landscape, Northeastern Portugal. Mammalian Biology, 76: of Cervidae, Smithsonian Institution Press, USA. Pp. 289–298. 491–497. O’Brien TG, Kinnaird MF & Wibisono HT (2003) Crouching Trisurat Y, Pattanavibool A, Gale GA & Reed DH (2010) Improving tigers, hidden prey: Sumatran tiger and prey populations in a the viability of large-mammal populations by using habitat and tropical forest landscape. Animal Conservation, 6: 131–139. landscape models to focus conservation planning. Wildlife Padmalal UKGK, Takatsuki S & Jayasekara P (2003) Food habits Research, 37: 401–412. of sambar Cervus unicolor at the Horton Plains National Park, Venables WN & Ripley BD (2002) Modern Applied Statistics with Sri Lanka. Ecological Research, 18: 775–782. S. Fourth Edition. Springer, New York, 495pp. Petdee A (2000) Feeding Habits of the Tiger Panthera tigris Walston J, Robinson JG, Bennett EL, Breitnmoser U, Fonseca (Linnaeus) in Huai Kha Khaeng Wildlife Sanctuary by Fecal GAB, Goodrich J, Gumal M, Hunter L, Johnson A, Karanth Analysis. Unpublished MSc Thesis, Kasetsart University, KU, Williams NL, MacKinnon K, Miquelle D, Pattanavibool Bangkok, 92 pp. [In Thai] A, Poole C, Rabinowitz A, Smith JLD, Stokes EJ, Stuart SN, R Core Team (2012) R: A Language and Environment for Statistical Vongkhamheng C & Wibisono H (2010) Bringing the tiger back Computing. R Foundation for Statistical Computing, Vienna, from the brink- the six percent solution. PLoS Biology, 8: 1–4. Austria. http://www.R-project.org (Accessed 06 September WEFCOM (2004) GIS Database and its Applications for Ecosystem 2013). Management. The Western Forest Complex Ecosystem Ramakrishnan U, Coss RG & Pelkey NW (1999) Tiger decline Management Project, Department of National Park, Wildlife, caused by the reduction of large ungulate prey: evidence and Plant Conservation, Bangkok, Thailand. 228 pp. from a study of leopard diets in southern India. Biological Wegge P & Storaas T (2009) Sampling tiger ungulate prey by Conservation, 89: 113–120. the distance method: lessons learned in Bardia National Park, Rogers G, Julander O & Robinette WL (1958) Pellet-group counts Nepal. Animal Conservation, 12: 78–84. for deer census and range-use index. Journal of Wildlife Wikramanayake ED, Dinerstein E, Robinson JG, Karanth U, Management, 22: 193–199. Rabinowitz A, Olson D, Mathew T, Hedao P, Conner M, Rotenberry JT, Preston KL & Knick ST (2006) GIS-based niche Hemley G & Bolze D (1998) An ecology-based method for modeling for mapping species habitat. Ecology, 87: 1458–1464. defi ning priorities for large mammal conservation: the tiger as Sanderson EW, Jaiteh M, Levy MC, Redford KH, Wannebo AV case study. Conservation Biology, 12: 865–878. &Woolmer G (2002) The human footprint and the last of the wild. Bioscience, 52: 891–904.
106 45 original contribution
SAKSIT SIMCHAROEN1, MAYUREE UMPONJAN2*, SOMPHOT DUANGCHANTRASIRI1 AND law, Thailand’s Office of Natural Resources ANAK PATTANAVIBOOL2 and Environmental Policy and Planning has reported, based on expert opinions, the sta- Non-Panthera cat records from tus of threatened species in Thailand and list- ed jungle cat and flat-headed cat as ‘critically big cat monitoring in Huai Kha endangered’, marbled cat as ‘endangered’, and clouded leopard, fishing cat and Asiatic Khaeng Wildlife Sanctuary golden cat as ‘vulnerable’ species; leopard cat is the only species considered nationally A camera-trapping deployment for tiger Panthera tigris monitoring in Huai Kha Khaeng of least concern (Nabhitabhata & Chan-ard Wildlife Sanctuary HKK, in the Western Forest Complex WEFCOM of Thailand, was 2005). carried out intensively between 2005 and 2009. The deployment’s annual setup in- Non-Panthera cats in the wild in Thailand cluded an average of 162 camera-trap locations with more than 2,000 trap-nights and have received less attention than the two covered almost 1,000 km2. Many other wildlife species were photographed including large cats, tiger and leopard. Leopard cat was small and medium (non-Panthera) cats. This analysis explores the potential use of studied in HKK in the late 1980s (Rabinowitz the system to monitor cat species other than tiger and leopard Panthera pardus. In 1990). From the late 1990s to mid 2000s came five years, leopard and tiger, major targets of the deployment, were camera-trapped a string of publications: leopard cat in Kaeng in 653 and 483 notionally independent events respectively. Among non-Panthera cats, Krachan National Park, Southern Thailand leopard cat Prionailurus bengalensis was the most common, with 155 events. Inde- (Grassman 1998), clouded leopard in Khao Yai pendent events of three other non-Panthera cats were rare: ten of Asiatic golden cat National Park, Northeastern Thailand (Austin Catopuma temminckii, six of mainland clouded leopard Neofelis nebulosa, and only & Tewes 1999), and leopard cat and marbled two of marbled cat Pardofelis marmorata. Leopard cat in HKK used mixed deciduous cat in Phu Khieo Wildlife Sanctuary, Northern forest heavily and showed an obvious crepuscular and nocturnal activity pattern. central Thailand (Grassman & Tewes 2000, The camera-trapping deployment for tigers in HKK could be used to monitor leopard 2002, Grassman et al. 2005). Since 2005, re- cats, but different deployment designs would be necessary for other non-Panthera sources and man power have been heavily in- cats at this site. vested in conservation of Panthera species es- pecially tiger (Simcharoen et al. 2007, Lynam South-east Asia is home to nine small and cats are under-represented in field studies 2010, Stokes 2010), in Thailand’s Western 31 medium cat species (i.e. excluding genus Pan- (Grassman et al. 2005). Four of the seven spe- Forest Complex WEFCOM. thera). Of these, seven occur in Thailand (all cies are categorised as globally threatened WEFCOM is categorised as a Tiger Conserva- those of mainland Southeast Asia): jungle cat by The IUCN Red List of Threatened Species. tion Landscape Class I (one that has habitat to Felis chaus, leopard cat, fishing cat Prionai- In Thai law, marbled cat is listed as ‘endan- support at least 100 tigers, evidence of breed- lurus viverrinus, flat-headed cat P. planiceps, gered’ and the rest as ‘protected’ under the ing, minimal-moderate levels of threat, and Asiatic golden cat, marbled cat and clouded Wildlife Preservation and Protection Act B. E. conservation measures in place), and Global leopard (Wilson & Mittermeier 2009). In Thai- 2535 (A. D. 1992) (Wildlife Conservation Divi- priority (highest probability of persistence of land as in much of the world, non-Panthera sion 1992, Boonboothara 1996). Besides the tiger populations over the long term; Diner- stein et al. 2006). Within WEFCOM, HKK is a core area where tiger and leopard ecology has been thoroughly studied, and populations estimated (Simcharoen et al. 2007, 2008). Camera-trapping started in a systematic man- ner in 2005, following the setup described in Karanth & Nichols (2002). Although designed for tigers, the deployment also photographed non-Panthera cats and many other species. This study uses by-catch from the long-term camera-trapping deployment in HKK to (1) examine the records of non-Panthera cats, and present what can be learned about status and natural history, and (2) discuss whether the programme generates sufficient non-Panthera cat records to allow these species’ conservation status to be moni- tored using such deployments.
Study Area Fig. 1. Location of Huai Kha Khaeng Wildlife Sanctuary, the major habitat types, and the Huai Kha Khaeng Wildlife Sanctuary (15°00’- locations of camera-traps. 15°50’ N/99°00’- 99°19’ E) is one of the best-
Non-Panthera cats in South-east Asia 46 Simcharoen et al.
2005 2006 2007
2008 2009
32
Fig. 2. Locations of camera-traps in Huai Kha Khaeng Wildlife Sanctuary showing where Fig. 3. Camera-trap points where Asiatic leopard cat was detected (red dots) and not detected (black dots) each year during 2005– golden cat, clouded leopard and marbled 2009. The background shows forest types. cat were detected in HKK WS 2005-2009.
known protected areas in Thailand (WEFCOM mountain range plays an important role in more rain in the west and less in the east, 2004; Fig. 1). It covers 2,780 km2 and is part blocking the southwest monsoon flowing in a variation causing significant differences in of a much bigger (18,000 km2) protected area from Myanmar. The southern part of HKK is vegetation type. network called the Western Forest Complex generally lower with many small hills of 700- HKK consists of mixed deciduous forest over WEFCOM. HKK was declared a wildlife sanc- 800 m high (Forest Research Centre 1997). almost half of the sanctuary. The other fo- tuary in 1972. Currently there are 19 ranger The climate is a mix of tropical and sub-trop- rest types include dry evergreen (25%), hill stations, located mostly along the eastern ical, has three seasons: the hot dry season evergreen (14%), dry dipterocarp (7%) and boundary, to protect HKK from poaching and of March-April with average temperature of bamboo forest (4%) (WEFCOM 2004). The land encroachment (WEFCOM 2004). 24°-38° C, the rainy season of May-October open dominant forest types of mixed decidu- HKK is part of the Dawna Range, north of the with 23°-34° C, and the cool dry season of ous and dry dipterocarp occur at elevations Tenasserim Range, separating northwestern December-February with 18°-21°C (Forest of 450-900 m. The forest is sometimes mixed Thailand from Myanmar. HKK topography is Research Centre 1997). The average annual with bamboos (major bamboo species: Bam- more mountainous to the north and west of rainfall is about 1,500 mm with the minimum busa arundincea, B. burmanica, Dendrocala- the area, with ridges exceeding 1,000 m. This in January and maximum in October. There is mus strictus, Gigantochloa albociliata). The
CATnews Special Issue 8 Spring 2014 47 non-Panthera cats in Huai Kha Khaeng Wildlife Sanctuary dominant tree species in the crown layer in- Table 1. The number of notionally independent events for cat species during camera- clude Afzelia xylocarpa, Tetrameles nudifora, trapping in Huai Kha Khaeng Wildlife Sanctuary during 2005-2009. When both cameras Lagerstroemia tomentosa, L. duperreanna, in a pair photographed an animal, this is recorded as only one record. Shorea obtusa, S. siamensis, Dipterocarpus Number of notionally independent events obtusifolis and D. tuberculatus (Forest Re- Species 2005 2006 2007 2008 2009 Total search Centre 1997). Tiger 107 68 91 111 106 483 Methods Leopard 133 138 139 115 128 653 For this study, data from camera-trapping Clouded leopard 22 1016 collected between 2005 and 2009 were ana- Asiatic golden cat 03 12410 lysed. The deployments occurred mainly in Leopard cat 9 24 56 12 54 155 the two open dominant forest types, given Marbled cat 10 0102 that the main target species was the tiger. Total camera-trap-nights 2,241 2,020 2,467 2,804 2,731 Tigers prefer open forests where, with their grass base, large ungulates such as gaur staff, with support from two wildlife biolo- of small and medium cat (Table 1). No do- Bos gaurus and banteng B. javanicus mostly gists, with more than five years of experience mestic cats Felis catus were captured during reside (Prayurasiddhi 1997). The camera- of camera trapping, in case of doubts. All these surveys. Tables 1 and 2 also contain trapping areas covered about 1,000 km2 of photographs of cats were scanned, put into results for tiger and leopard, for comparison this near-optimal tiger habitat. Almost 80% a database and identification of all photo- with the smaller species; detailed analysis of of camera-trap locations were in mixed de- graphs listed as non-Panthera were assessed Panthera data will be published elsewhere. ciduous and dry dipterocarp forests, 17% in independently by J. W. Duckworth. Records degraded evergreen, and the rest in other were calculated in terms of: 1) number of in- Number of notionally independent events vegetation types. dependent events, and 2) number of camera- Each year the camera-traps were deployed Several camera-trap models, including trap stations detecting the species. To assess for more than 2,000 trap-nights with a total CamTrakker, Bushnell and Scoutguard, were conservation status, the photographs at one of 12,263 trap-nights over the five years. set up following the standard method used camera-trap station are not independent if The numbers of independent events for non- for monitoring tigers, detailed in Karanth & they show the same animal. This problem is Panthera cats are much lower than Panthera Nichols (2002). Camera-traps were located reduced by presenting the number of camera- cats (Table 1). Leopard cat was the most fre- 33 mainly along forest roads and animal trails, trap stations recording the species, although quently detected small cat. Clouded leopard and at salt licks. At each location camera- even this will not exclude non-independent and golden cat events ranged from very few traps were set in a pair, each unit 3-5 m from records if multiple camera-trap stations are to none per year; marbled cat was detected the path and about 45 cm above ground. No within a typical individual’s home range. No- only twice (Supporting Online Material SOM bait was used. Camera-traps were set to tionally independent events are defined as Table T1). function throughout the 24-hour cycle. one or more photographs of one or more ani- The spacing between camera-trap locations mals of the same species at a given camera- Number of camera-trap stations detecting was about 3-4 km, based on female tiger trap location, separated by no more than 30 the species home-range (Karanth & Nichols 2002). With minutes. Between 150 and 190 camera-trap stations about 180 camera-trap locations each year, Camera-trap locations were overlaid with a were set each year, covering almost 1,000 trapping was divided into eight blocks of habitat map interpreted from LANDSAT 5 TM km2. Leopards and tigers were the most 20-25 trapping locations. The camera-traps 2002 (WEFCOM 2004) to determine the veg- widely detected cat species (Table 2). Among were left in each block for 15-20 days before etation cover at each location. non-Panthera cats, leopard cat had the wid- being relocated to another block. Two blocks est detection, but even so each year less than were sampled simultaneously. Trapping nor- Results one-sixth of camera-trap stations detected mally started in January and finished by mid Tiger-focussed camera-trapping in HKK be- leopard cats. The other three cats were found May. For an optimal setting of cameras, lo- tween 2005 and 2009 captured four species at very few stations. cations within a block were moved slightly between years. Thus, spacing between ca- Table 2. The number of camera-trap stations recording each species in Huai Kha mera-trap locations used in different years Khaeng Wildlife Sanctuary during 2005-2009. was frequently well below 3-4 km, and the Number of stations where the species were recorded total number of camera-trap locations at Species 2005 2006 2007 2008 2009 Total which some species were found over the five years exceeded the 180 total camera-trap lo- Tiger 58 46 52 67 64 287 cations per year. Leopard 77 61 76 61 63 338 The total of camera-trap-nights is the sum Clouded leopard 121015 of the number of nights each pair of cam- Asiatic golden cat 0312410 eras was open functioning at all camera- Leopard cat 9 14 29 7 33 92 trap locations. Species identification from Marbled cat 100102 photographs was carried out by the project Total camera-trap locations 155 136 156 180 186
Non-Panthera cats in South-east Asia 48 Simcharoen et al.
basis for the 1989 report warrants a review. Jungle cat apparently occurs predominantly in deciduous forest in South-east Asia ( Duck- worth et al. 2005), so parts of HKK might be expected to support it. However, no records were obtained from this intensive camera- trapping survey, mostly in deciduous forest, despite reasonable trapping rates described in other studies (e.g. Gray et al. 2014), sug- gesting that jungle cat is rare or even absent from HKK. The other small cat of Thailand, the flat-headed cat, does not occur this far north (Wilting et al. 2010).
Fig. 4. Clouded leopard on 3 June 2006, 23:49 h. Habitat: Mixed deciduous forest. Small cat community In HKK, leopard cat is common but golden Leopard cat habitat use and activity pattern Discussion cat, clouded leopard and marbled cat were Leopard cat was the only small cat with suf- Leopard cat all recorded only rarely. Focused camera- ficient camera-trap records (92 locations in five Leopard cat is the only small cat species so far trapping in HKK’s evergreen forests might years) for an analysis of habitat use (SOM T2). studied intensively in multiple parts of Thai- find these three species more often, but they Caution is required in interpretation because land (Rabinowitz 1990, Grassman et al. 2005). are evidently rare in HKK’s deciduous forest. patterns may be biased by the selection of Similarly, it is the only species with enough Observations in other areas suggest that camera-trap locations, and refer only to the camera-trap detections in HKK for a confident leopard cat population increases when larger late dry season. Almost 70% of camera-trap discussion of abundance and habitat use at predators, such as golden cat and clouded locations with leopard cat detection were in the site, albeit only for the late dry season. leopard, are eliminated (Wilson & Mittermei- mixed deciduous forest (SOM T2; Fig. 2), while It was photographed in many habitat-types, er 2009). Release of leopard cat population the other two open canopy forest types, dry coinciding with its generally wide habitat use with reduction of interspecific competition 34 dipterocarp (10%) and degraded dry evergreen (Wilson & Mittermeier 2009). In HKK the high from golden cat and marbled cat is plausible, forests (15%) were used to a lesser extent. encounter rates in mixed deciduous forest because the three species presumably share Leopard cat was also the only small cat spe- may simply reflect disproportionate survey similar small prey such as rodents, reptiles, cies with enough data to allow for the analysis effort. However, the low encounter rate in birds, amphibians and insects. However, it is of activity patterns. At least in the late dry sea- dry dipterocarp forest relative to survey ef- less likely for clouded leopard, which preys son, it is nocturnal, with the main activity start- fort corroborates earlier findings in HKK that on larger animals such as porcupines (Hystri- ing after 18:00 h and peaking during 19:00 h it uses mixed deciduous and dry evergreen cidae), pigs Sus spp., young sambar Rusa uni- - 22:00 h and fluctuating from 22:00 h to 06:00 forests more than dry dipterocarp forest with color, muntjacs Muntiacus spp., chevrotains h. It is almost inactive by day (SOM Figure F1). its lower dry-season grass base, and thus Tragulus spp. and palm civets (Paradoxurinae) lower cover and prey (Rabinowitz 1990). Wet- (Wilson & Mittermeier 2009). In this study, Morphology of Asiatic golden cat season surveys, when dry dipterocarp forest clouded leopard seems to use evergreen for- Of the 10 records of golden cat, seven were has rich understorey growth, might reveal a est more frequently than leopard cat, which is of golden animals and three of grey ones. very different habitat use. found more in deciduous forest. These results found leopard cat to be cre- puscular and nocturnal, with very few pho- Conclusions and management implica- tographs by day. Radio-collared leopard cats tions in Phu Khieo Wildlife Sanctuary, northeast- Intensive camera-trap deployment for tigers ern Thailand, in more evergreen habitats, in Huai Kha Kheang Wildlife Sanctuary from showed somewhat more daytime activity, 2005 to 2009 captured six cat species: tiger, while still being mainly crepuscular and noc- leopard, clouded leopard, golden cat, marbled turnal (Grassman et al. 2005). cat and leopard cat. Tiger and leopard were recorded often. Of the non-Panthera cats, Other non-Panthera cats leopard cat was found commonly whereas Fishing cat was reported in the Master Plan golden cat, clouded leopard and marbled cat of HKK in 1989 (Thailand Faculty of Forestry were rarely found. 1989). It was not detected in the 2005-2009 Thus, camera-trapping for tigers in Huai Kha camera-trap deployment, which covered Khaeng Wildlife Sanctuary provides useful large areas including near streams, and data to study abundance patterns, activity seems very unlikely to occur there presently. rhythms, and habitat use of leopard cat, but Fig. 5. Leopard cat on 23 April 2006, Because individuals of this species are often data are too sparse for a similar analysis of 14:59 h. Habitat: Mixed deciduous forest. misidentified, (Duckworth et al. 2009), the clouded leopard, golden cat and marbled cat.
CATnews Special Issue 8 Spring 2014 49 non-Panthera cats in Huai Kha Khaeng Wildlife Sanctuary
Moreover, annual numbers of leopard cat independent events fluctuated considerably during this five-year study, making it diffi- cult to use this method to assess population trends during short periods of time. To monitor clouded leopard, golden cat and marbled cat, other camera-trapping study de- signs would need to be experimented with, such as placing more camera-trap stations in evergreen forests, or around fruiting trees with high rodent concentration.
Acknowledgements Fig. 6. Asiatic golden cat on 16 March 2006 08:04 h. Habitat: Hill evergreen forest. Funding for this study was from US Fish and Wild- life Service: Rhino and Tiger Conservation Fund, galensis) in a subtropical evergreen forest Simcharoen S., Pattanavibool A., Karanth K. U., Panthera, and Liz Claiborne and Art Ortenberg in southern Thailand. Zoological Society La Nichols J. D. & Kumar N. S. 2007. How many Foundation. We are grateful to the Thai govern- Torbiera, Scientific Report 4, 9-21. tigers Panthera tigris are there in Huai Kha ment’s Department of National Parks, Wildlife and Grassman, L. I. & Tewes M. E. 2000. Marbled cat Khaeng Wildlife Sanctuary, Thailand? An es- Plant Conservation and the Wildlife Conservation pair in northeastern Thailand. Cat News 33, 24. timate using photographic capture-recapture Society (WCS) for the logistics and technical sup- Grassman, L. I., & Tewes M. E. 2002. Marbled cat sampling. Oryx 41, 447-453. port. Special thanks go to the staff of Kao Nang in northeastern Thailand. Cat News 36, 19-20. Stokes E. J. 2010. Improving effectiveness of pro- Ram Wildlife Research Station, of WCS Thailand Grassman Jr L. I., Tewes M. E., Silvy N. J. & Kreeti- tection efforts in tiger source sites: developing and of WCS India. yutanont K. 2005. Spatial organization and diet a framework for law enforcement monitoring of leopard cat (Prionailurus bengalensis) in using MIST. Integrative Zoology 5, 363-377. References northern-central Thailand. Journal Zoological Thailand Faculty of Forestry 1989. The master plan Austin S. C. & Tewes M. E. 1999. Ecology of the Society of London, 266, 45-54. for Huai Kha Khaeng Wildlife Sanctuary. Fac- clouded leopard in Khao Yai National Park, Gray T. N. E., Phan C., Pin C. & Prum S. 2014. The ulty of Forestry, Kasetsart University, Bangkok. 35 Thailand. Cat News 31, 17-18. status of jungle cat and sympatric small cats in WEFCOM 2004. GIS Database and its applications Boonboothara K. 1996. Wildlife Preservation and Cambodia’s Eastern Plains. Cat News Special for ecosystem management. The Western For- Protection Act B.E.2535. The Information Of- Issue 8, 31-35. est Complex Ecosystem Management Project fice, the Royal Forest Department, Bangkok. IUCN. 2011. The IUCN Red List of Threatened Spe- (WEFCOM), Department of National Parks, Dinerstein E., Loucks C., Heydlauff A., Wikra- cies. Version 2011.2 http://www.iucnredlist.org. Wildlife and Plant Conservation, Bangkok. manayake E., Bryja G., Forrest J., Ginsberg Karanth K. U. & Nichols J. D. 2002. Monitoring ti- Wildlife Conservation Division 1992. The Wildlife J., Klenzendorf S., Leimgruber P., O’Brien T., gers and their prey. Centre of Wildlife Studies, Preservation and Protection Act 1992. The Roy- Sanderson E., Seidensticker J. & Songer M. Bangalore, India. al Forest Department, Bangkok. In Thai. 2006. Setting priorities for the conservation Lynam A. J. 2010. Securing a future for wild In- Wilson D. E. & Mittermeier R. A. (Eds). 2009. and recovery of wild tigers: 2005-2015. A dochinese tigers: transforming tiger vacuums Handbook of the mammals of the world. Vol. 1. User’s Guide. WWF, WCS, Smithsonian, and into tiger source sites. Integrative Zoology 5, Carnivores. Lynx Ediciones, Barcelona. NFWF-STF, Washington, D.C. and New York. 324-334. Wilting A., Cord A., Hearn A. J., Mohamed A., Duckworth J. W., Poole C. M., Tizard R. J., Walston Nabhitabhata J. & Chan-ard, T. 2005. Thailand Red Traeholdt C., Cheyne S. M., Sunarto S., Jayasi- J. L. & Timmins R. J. 2005. The Jungle Cat Felis Data: Mammals, Reptiles, and Amphibians. lan M.-A., Ross J., Shapiro A. C., Sebastian A., chaus in Indochina: a threatened population of Office of Natural Resources and Environmental Dech S., Breitenmoser C., Sanderson J., Duck- a widespread and adaptable species. Biodiver- Policy and Planning, Bangkok. worth J. W. & Hofer H. 2010. Modelling the sity and Conservation 14, 1263-1280. Prayurasiddhi T. 1997. The ecological separation species distribution of flat-headed cats (Prion- Duckworth J. W., Shepherd C. R., Semaidi G., of gaur (Bos gaurus) and banteng (Bos javani- ailurus planiceps), an Endangered South-East Schauenberg P., Sanderson J., Roberton S. I., cus) in Huai Kha Khaeng Wildlife Sanctuary, Asian small felid. PLoS One 5, 1-18. O’Brien T. G., Maddox T., Linkie M., Holden J., Thailand. PhD thesis, University of Minnesota, & Brickle N. W. 2009. Does the fishing cat in- U.S.A. Supporting Online Material SOM Tables T1, T2 and habit Sumatra? Cat News 51, 4-9. Rabinowitz A. R. 1990. Notes on the behavior and Figure F1 are available at www.catsg.org/catnews Forest Research Centre. 1997. Application of re- movements of leopard cats Felis bengalensis mote sensing and GIS for monitoring forest in a dry tropical forest mosaic in Thailand. Bio- 1 Department of National Parks, Wildlife and land use change in Huai Kha Khaeng Wildlife tropica 22, 397-403. Plant Conservation, 61 Paholyotin Road, Cha- Sanctuary. Final Report to the Royal Forest Simcharoen S., Barlow A., Simcharoen A. & Smith tuchak, Bangkok, 10900, Thailand Department by Faculty of Forestry, Kasetsart J. D. 2008. Home range size and daytime habi- 2 Wildlife Conservation Society Thailand Pro- University, Bangkok. tat selection of leopards in Huai Kha Khaeng gram, 55/295 Muangthong Thani, Project 5, Grassman L. I. 1998. Movements and prey se- Wildlife Sanctuary, Thailand. Biological Con- Chaengwattana Road, Pakkred, Nonthaburi, lection of the leopard cat (Prionailurus ben- servation 141, 2242-2250. 11120, Thailand *
Non-Panthera cats in South-east Asia 50 Contributed Paper Dynamics of a low-density tiger population in Southeast Asia in the context of improved law enforcement
Somphot Duangchantrasiri,∗ Mayuree Umponjan,† Saksit Simcharoen,∗ Anak Pattanavibool,†‡ Soontorn Chaiwattana,∗ Sompoch Maneerat,∗ N. Samba Kumar,§∗∗ Devcharan Jathanna,§ Arjun Srivathsa,§ ¶ and K. Ullas Karanth§†† ∗Department of National Parks, Wildlife and Plant Conservation, Paholyotin Road, Chatuchak, Bangkok 10110, Thailand †Wildlife Conservation Society, Thailand Program, 55/295 Muangthong Thani 5, Chaengwattana Road, Pakkred, Nonthaburi 10210, Thailand ‡Faculty of Forestry, Department of Conservation, Kasetsart University, Bangkok 10900, Thailand §Centre for Wildlife Studies, 1669, 31st Cross, 16th Main, Banashankari 2nd Stage, Bengaluru 560 070, India ∗∗Wildlife Conservation Society, India Program, 1669, 31st Cross, 16th Main, Banashankari 2nd Stage, Bengaluru 560 070, India ††Wildlife Conservation Society, Global Conservation Program, 2300 Southern Boulevard, Bronx, NY 10460, U.S.A.
Abstract: Recovering small populations of threatened species is an important global conservation strategy. Monitoring the anticipated recovery, however, often relies on uncertain abundance indices rather than on rigorous demographic estimates. To counter the severe threat from poaching of wild tigers (Panthera tigris), the Government of Thailand established an intensive patrolling system in 2005 to protect and recover its largest source population in Huai Kha Khaeng Wildlife Sanctuary. Concurrently, we assessed the dynamics of this tiger population over the next 8 years with rigorous photographic capture-recapture methods. From 2006 to 2012, we sampled across 624–1026 km2 with 137–200 camera traps. Cameras deployed for 21,359 trap days yielded photographic records of 90 distinct individuals. We used closed model Bayesian spatial capture-recapture methods to estimate tiger abundances annually. Abundance estimates were integrated with likelihood-based open model analyses to estimate rates of annual and overall rates of survival, recruitment, and changes in abundance. Estimates of demographic parameters fluctuated widely: annual density ranged from 1.25 to 2.01 tigers/100 km2, abundance from 35 to 58 tigers, survival from 79.6% to 95.5%, and annual recruitment from 0 to 25 tigers. The number of distinct individuals photographed demonstrates the value of photographic capture–recapture methods for assessments of population dynamics in rare and elusive species that are identifiable from natural markings. Possibly because of poaching pressure, overall tiger densities at Huai Kha Khaeng were 82–90% lower than in ecologically comparable sites in India. However, intensified patrolling after 2006 appeared to reduce poaching and was correlated with marginal improvement in tiger survival and recruitment. Our results suggest that population recovery of low-density tiger populations may be slower than anticipated by current global strategies aimed at doubling the number of wild tigers in a decade.
Keywords: abundance estimation, camera traps, carnivores, overhunting, patrolling, population dynamics, spatial capture-recapture models
La Din´amica de una Poblacion´ de Tigres con Baja Densidad en el Sureste de Asia dentro del Contexto de una Aplicacion´ Mejorada de la Ley Resumen: Recuperar las poblaciones pequenas˜ de las especies amenazadas es una importante estrategia global de conservacion.´ Sin embargo, monitorear la recuperacion´ esperada generalmente depende de ´ındices
¶Address correspondence to A. Srivathsa, email [email protected] Paper submitted May 4, 2015; revised manuscript accepted October 10, 2015. 1 Conservation Biology, Volume 00, No. 0, 1–10 C 2015 Society for Conservation Biology DOI: 10.1111/cobi.12655 51 2 Tiger Population Dynamics in Thailand inciertos de abundancia en lugar de estimados demograficos´ rigurosos. Para contrarrestar la gran amenaza causada por la cacer´ıa furtiva de tigres (Panthera tigris), el Gobierno de Tailandia establecio´ un sistema intensivo de patrullaje en 2005 para proteger y recuperar la poblacion´ fuente mas´ grande en el Santuario Huai Kha Khaeng. Simultaneamente,´ evaluamos las dinamicas´ de esta poblacion´ de tigres durante los siguientes ocho anos˜ con rigurosos m´etodos fotograficos´ de captura-recaptura. De 2006 a 2012 muestreamos a lo largo de 624–1026 km2 con 137–200 trampas camara.´ Las camaras´ desplegadas durante 21,359 d´ıas de trampa produjeron registros fotograficos´ de 90 individuos distinguibles. Usamos m´etodos espaciales de captura- recaptura y modelo bayesiano cerrado para estimar anualmente la abundancia de los tigres. Los estimados de abundancia estuvieron integrados por analisis´ de modelo abierto basados en la probabilidad para estimar la tasa anual y las tasas generales de supervivencia, reclutamiento y cambios en la abundancia. Los estimados de los parametros´ demograficos´ fluctuaron ampliamente: la densidad anual vario´ entre 1.25 y 2.01 tigres/ 100 km2, la abundancia entre 35 a 58 tigres, la supervivencia entre 79–6 y 95.5% y el reclutamiento anual de 0 a 25 tigres. El numero´ de individuos distinguibles que fue fotografiado demuestra el valor de los m´etodos de captura-recaptura para la evaluacion´ de las dinamicas´ poblacionales de especies raras y elusivas que son identificables gracias a marcas naturales. Posiblemente por causa de la presion´ ejercida por la caza furtiva, la densidad general de los tigres en Huai Kha Khaeng fue 82–90% mas´ baja que en sitios ecologicamente´ comparables de India. Sin embargo, el patrullaje intensivo despu´es de 2006 parecio´ reducir la caza furtiva y estuvo correlacionado con el mejoramiento marginal de la supervivencia y reclutamiento de los tigres. Nuestros resultados sugieren que la recuperacion´ de las poblaciones de tigres con baja densidad puede ser mas´ lenta de lo esperado por las estrategias globales actuales enfocadas en la duplicacion´ del numero´ de tigres en una d´ecada.
Palabras Clave: carn´ıvoros, din´amicas poblacionales, estimacion´ de la abundancia, exceso de caza, modelos espaciales de captura-recaptura, patrullaje, trampas c´amara
Introduction 3000–4000 individuals in <200 years. Direct killing for commerce or because of human–carnivore conflict, Protected areas have facilitated the persistence and recov- overhunting of prey, and habitat loss have driven this ery of many large mammal species (Margules & Pressey decline (Walston et al. 2010; Wikramanayake et al. 2000; Brooks et al. 2009). In the context of escalating 2011). Currently, about 70% of surviving wild tigers are threats (Schipper et al. 2008) and consequent population concentrated within <7% of remaining habitat. These are declines (Ceballos & Ehrlich 2002), models of “source source sites that support breeding populations, without population recovery” have gained support (Walston et al. which tigers cannot survive across wider landscapes 2010). This approach relies on the measurement of pop- (Walston et al. 2010). Although Southeast Asia contains ulation recoveries of target species as its central metric of approximately 50% of remaining tiger habitat, no study effectiveness (Walpole et al. 2001; Williams et al. 2002; of population dynamics exists for any tiger population in Soule et al. 2005). However, rigorous empirical studies this region. Responding to threats, after 2006, the gov- of ongoing species recoveries under this model (Walsh ernment of Thailand adopted a tiger conservation model & White 1999) are scarce. based on source population recovery. In collaboration Large carnivores have long been flagships of species with the Wildlife Conservation Society (WCS), it estab- recovery efforts, often serving as umbrellas for overall lished systematic, intensified foot patrols in Huai Kha biodiversity (Caro & O’Doherty 1999). Given their nat- Khaeng (HKK) Wildlife Sanctuary to reduce hunting pres- urally low population densities, wide-ranging behaviors, sures on tigers and prey species (WCS-Thailand 2007). and elusiveness, rigorously monitoring their populations Simultaneously, a scheme for rigorous tiger population is often difficult (Palomares et al. 2010). In addition to monitoring developed in India (Karanth & Nichols 2002; abundance, estimation of demographic vital rates (sur- Karanth et al. 2006, 2011a) was initiated. These matched vival, recruitment) that drive carnivore population re- conservation interventions offer a rare opportunity to ex- sponses is also critical (Hebblewhite et al. 2003; Lambert amine whether a correlative relationship exists between et al. 2006; Balme et al. 2009). Because of methodological increased law enforcement efforts and tiger abundance. challenges, many recovery programs rely on surrogate in- Camera-trap surveys and capture-recapture (CR) meth- dices of carnivore abundance, threat levels, and efficacy ods have expanded the scope of noninvasive study of of conservation efforts instead of using rigorous demo- carnivores with identifiable natural markings (Williams graphic assessments. et al. 2002; Karanth et al. 2006). Because of high rates The tiger (Panthera tigris) represents a good case of mortality, reproduction, and turnover (Karanth et al. study of endangerment of a flagship species. Its global 2006; Goodrich et al. 2008), tiger populations need to be range has contracted by approximately 93%, and assessed annually. Such annual surveys, combined with its global population has declined to approximately open-model analyses across years, can yield estimates of
Conservation Biology Volume 00, No. 0, 2015 52 Duangchantrasiri et al. 3 density and abundance and vital rates that drive changes in these factors (Karanth et al. 2006; van de Kerk et al. 2013). We applied methodological advances in spatial capture–recapture (SCR) (Borchers & Efford 2008; Royle et al. 2009) in the first-ever study of tiger population dynamics in Southeast Asia. In our study of tigers in HKK Wildlife Sanctuary, Thailand from 2005 to 2012, we sought to estimate tiger photo-capture probabilities, tiger density and abundance, and rates of population change from 2005 to 2012; estimate survival and recruitment rates, which drive population dynamics; examine spa- tial and temporal patterns in tiger-population parameters; and explore the possible effect of the recovery strategy employed in HKK on these population parameters. We hypothesized that illegal hunting of tigers and prey species depressed tiger densities in HKK prior to 2006 (Karanth et al. 2004). We anticipated that tiger numbers in HKK would improve in response to increased protec- tion efforts and sought to measure the rate and magnitude of this response.
Methods
Study Area HKK (2780 km2) is embedded in the 18,000 km2 West- ern Forest Complex (WEFCOM) in Thailand (15°00’N– 15°50’N and 99°00’E–99°19’E) (Fig. 1). The area sup- ports dry-deciduous Dipterocarp forest, mixed decidu- ous forest, dry-evergreen forest, hill evergreen forest, and bamboo-mixed secondary forest, depending on local rain- fall, soil, and past human disturbances. To counter threats from illegal hunting and logging, HKK was established as a wildlife sanctuary in 1972, and several villages were Figure 1. Location of Huai Kha Khaeng (HKK) resettled from within HKK from 1972 to 1991. However, Wildlife Sanctuary in the Western Forest Complex there are about 30 villages along HKK’s eastern bound- (WEFCOM), Thailand. ary that are a significant source of illegal hunting (WCS- Thailand 2007; Department of National Parks 2010). The sanctuary supports one of the largest tiger popula- ing was likely the cause for depressed densities, suggest- tions in Southeast Asia (Simcharoen et al. 2007). It is also ing that increased law enforcement efforts were required. an important conservation area for several threatened We conducted our study after Simcharoen et al.’s (2007). ungulates that are tiger prey, including gaur (Bos gau- rus), banteng (Bos javanicus), and Eld’s deer (Cervus Field Survey eldii) (Bhumpakphan 1997; Prayurasiddhi 1997). Other principal prey species are wild pig (Sus scrofa), sam- Simcharoen et al. (2007) surveyed 477 km2 in HKK from bar (Rusa unicolor), red muntjac (Muntiacus muntjak), 2004 to 2005. In 2005, we surveyed an adjacent area. and, potentially, wild buffalo (Bubalus bubalis) and These 2 surveys did not follow CR sampling protocols Malayan tapir (Tapirus indicus) (Petdee 2000). Livestock because survey durations far exceeded the recommended are limited to the edges outside HKK (Jotikapukkana et al. 45–60 d, violating the population closure assumption 2010). (Karanth & Nichols 2002). Therefore, we included 24 Simcharoen et al. (2007) estimated the potential carry- tigers detected just before our survey only as a starting ing capacity of tigers at approximately 700 in HKK and data point of our systematic surveys, which spanned the approximately 2000 in the WEFCOM landscape, which subsequent 7 years from 2006 to 2012. highlight the region’s global importance for tiger conser- We positioned paired camera traps along anticipated vation. Their results, however, raised concerns that tiger tiger travel routes to simultaneously photograph both densities were low at <4 tigers/100 km2 and that poach- flanks. Trap locations were selected to maximize photo
Conservation Biology Volume 00, No. 0, 2015 53 4 Tiger Population Dynamics in Thailand
Table 1. Details of camera-trap sampling from a survey of wild tigers in Huai Kha Khaeng Wildlife Sanctuary (HKK), Thailand, from 2005 to 2012. No. of days Sampling No. of trap Trap-array Effort No. of tigers Cumulative no. of 2 Year of sampling midpoint t (years) locations area (km ) (trap days) detected (Mt+1) tigers detected 2005 481 6 Oct 2004 1.41 155 524 910 24 24 2006 185 4 Mar 2006 1.04 137 624 2156 27 36 2007 135 20 Mar 2007 1.01 156 991 2597 26 41 2008 145 23 Mar 2008 0.92 180 982 2999 34 50 2009 149 20 Feb 2009 1.08 194 1094 2919 33 61 2010 143 22 Mar 2010 0.97 181 934 2935 29 70 2011 128 13 Mar 2011 1.02 183 905 2974 32 83 2012 144 19 Mar 2012 – 200 1026 3869 35 90
captures of tigers and were spaced at around 1.5– SCR analyses, we used daily sampling occasions; a trap 3 km apart (Karanth & Nichols 2002, 2010). The number deployment matrix indicated locations that were active of trap locations varied by year (Table 1). Because of (or not) on specific days. For the CAPTURE analyses, human-power and equipment limitations, we used stan- sampling occasions were constructed by combining data dard block trapping (Karanth & Nichols 2002). We used across blocks (i.e., captures from the first day from all shorter secondary sampling periods, nested within the blocks were assigned to the first sampling occasion). 8 annual primary periods (Karanth et al. 2011a, 2011b). Tiger abundance (Nˆ ) (population size) for each year However, logistical constraints of surveying the rugged, was obtained by multiplying Dˆ by the study area size roadless terrain resulted in our secondary samples ex- (2780 km2). We used the cell-specific (pixel-level) tiger ceeding the 45–60 d recommended duration (Table 1) densities obtained annually from SCR analyses (Royle (Karanth & Nichols 2002). et al. 2009) to examine variations in tiger density within the HKK study area. The rate of population change be- tween successive years was estimated from the ratios Analyses Nˆ t+1/Nˆ t for the corresponding interval t (Karanth et al. We used the pattern-matching software EXTRACTCOM- 2006; Karanth et al. 2011b). We also performed conven- PARE (Hiby et al. 2009) to compare tiger photographs and tional closed-model CR analyses implemented in program subsequently verified the matched individuals visually. CAPTURE to generate a second set of density and abun- These unique tiger identities generated standard individ- dance estimates. ual capture-history matrices with daily samples (Karanth For estimating recapture and survival probabilities et al. 2011a). SCR analyses require additional information for tigers between successive annual surveys, we used on locations of traps and captures. As per standard tiger- Cormack–Jolly–Seber (CJS) models (Cormack 1964; Jolly survey protocols (Karanth et al. 2004, 2006), because 1965; Seber 1965). The apparent survival rate (φ) ac- of low capture rates for cubs, all demographic data and counts for losses from both mortality and permanent analyses pertained to tigers >1 year of age. emigration (Williams et al. 2002). Recapture probability Unlike conventional CR methods that rely on an ad hoc (p) here is the probability that an individual is captured in buffer area around the trap array to deal with geographic a primary sampling occasion, following its initial capture. closure, SCR models directly integrate animal movement The 4 biologically plausible scenarios for the tiger popu- data into modeling of the detection process. They also lation, implemented in program MARK (Cooch & White better model individual heterogeneity in capture proba- 2009), included models in which survival and recapture bility arising from the spatial configuration of tiger home probabilities varied over time [φ(t),p(t)]; survival and re- ranges in relation to the trap array (Royle et al. 2009). capture probabilities were constant over time [φ(.),p(.)]; We relied on a Bayesian MCMC-based SCR approach only recapture probability varied over time [φ(.),p(t)], (Royle et al. 2013) implemented in program SPACECAP and only survival varied over time [φ(t),p(.)]. We assessed (Gopalaswamy et al. 2012) to estimate tiger densities (Dˆ ). the fit of these 4 CR models to our capture-history data We set the state–space S at 4600 km2 based on a home- with Akaike’s information criterion corrected for small range diameter 3 times larger than observed in field stud- sample sizes (AICc ) (Burnham & Anderson 2002). ies (Royle et al. 2013). We used 52,000 MCMC iterations We first converted the annual survival rate estimates φˆ with an initial burn-in of 2000. The augmentation value ( t) derived from the open model analysis to estimates (maximum number of animals that could exist within of survival over the interval between successive primary φˆ t the state space) was set at 150, approximately 5 times sampling occasions ( t ). We then used annual estimates the number of tigers we captured each year. We used the of tiger abundance (Nˆ ) and survival to compute annual Geweke (1992) statistic reported by program SPACECAP recruitment arising from in situ reproduction and immi- to check for convergence of the MCMC chains. For the gration. Recruitment (Bˆ ) was computed for each interval
Conservation Biology Volume 00, No. 0, 2015 54 Duangchantrasiri et al. 5 from 2006 to 2012 as the difference between estimated directly as higher tiger survival rates. The increase in abundance Nˆ t+1 in year t + 1 and the expected number of survival rate could manifest indirectly, such that in the φˆ t survivors (i.e., Nt t ) from the previous year t (Pollock longer term tiger survival rate would improve as a result et al. 1990). of increased prey densities. Similarly, we expected that Considered together, our above estimates of density, the effect of increased patrolling on recruitment (entry of abundance, population change, survival, and recruitment new individuals >1 year old into the population, either in provided us with direct, comprehensive measures of the situ or through immigration) could be direct (increased dynamics of the wild tiger population in HKK at annual cub survival because of decreased hunting of adults) or intervals and averaged across the entire study period. indirect with a time lag (due to increased cub survival and immigration resulting from higher prey densities). We could not explore potential time lags in the response Assessment of Protection Measures of tiger survival or recruitment in relation to patrol efforts The management interventions concurrent with our because patrol data were available only for 2005 onwards. camera-trapping studies were driven by administrative We tried to test plausible relationships between tiger and social contingencies. Therefore, our examination of survival and recruitment rates, and between annual pa- the relationship between monitoring and management trol effort and rates of detection of poaching incidents interventions is purely observational, typical of conser- (as a measure of patrolling efficacy) using the temporal vation science. Although it was not possible to deter- symmetry modeling approach (Pradel 1996). We con- mine causal relationships due to the lack of a suitable sidered population models in which survival probability experimental design (replication, random assignments (φˆ), recapture probability (ˆp), and per capita recruitment of treatments, and controls), we nonetheless looked for rate ˆf either varied over time or were constant (but not correlative support for the hypothesis that increased law the model in which all 3 varied over time, due to con- enforcement efforts results in increased tiger abundance. founding of parameters). We also modeled survival and Based on theoretical models of tiger population viabil- recruitment as functions of total annual patrol effort and ity (Karanth & Stith 1999) and field studies (Karanth et al. detection rates of poaching incidents. 2004), we hypothesized that poaching of ungulates for local trade and of tigers for global commerce are drivers Results of tiger decline in HKK. The increased patrols in HKK Estimates of Tiger Capture Probabilities, from 2006 to 2012 aimed to reduce poaching, which in Densities, and Abundance turn would stabilize the tiger population and eventually lead to the recovery to its potential carrying capacity. For each year (excluding 2005), depending on logisti- Anecdotal information from antipoaching efforts and cal feasibility, we sampled approximately the same area law enforcement experience in HKK indicated that small (Fig. 2; Table 1). Our surveys yielded photo captures of groups of hunters temporarily camped in the forest, typ- 90 individual tigers (Table 1) (33 males, 50 females, 7 ically operating on foot. They poison carcasses of wild tigers of unknown sex). prey to lure and subsequently kill tigers (Department The movement parameter in the SCR model showed of National Parks 2013). After 2006, the authorities ex- tiger movement parameter ranged from 3.424 to panded the patrol range, intensified patrol networks, and 6.249 km and the expected basal encounter rate of an collected data on spatial and temporal deployment of pa- individual in a sampling occasion ranged from 0.050 to trols. Information on law enforcement, including data on 0.016 (Table 2). We were unable to perform separate cases of poaching of prey animals, was gathered and main- analyses by sex because of sample size constraints. Tiger tained using the MIST patrolling system software (Stokes densities Dˆ ranged from 1.27 (SD 0.19) to 2.09 (SD 0.39) 2010). The annual number of patrol days increased from tigers/100 km2. Population size Nˆ generally ranged from 1031 to 3316 and distances patrolled increased from 41 (SE 1.3) to 58 (SE 2.1) animals, except in 2011 when 5,979 km to 12,907 km (Fig. 3). there was a steep decline to 35 (SE 1.0) animals (Table The patrol teams recorded direct observations of 2). Estimated spatial densities of tigers in HKK from 2005 poaching and evidence of past occurrences, such as pres- to 2012 are presented in Fig. 2. The year-to-year change ence of snares or traps, camps, and animal carcasses. ratio for successive years t and t + 1 ranged from 0.70 to Poachers were arrested either during patrols or later 1.58. The geometric mean of the finite rate of population based on prior intelligence, following which the wildlife change over the entire study period was 0.99 (SE 0.23), managers prosecuted them. Thus, substantial effort was indicating a virtually stable tiger population. invested to deter, prevent, and catch poachers. We speculated that increases in patrol effort and in Survival Rates and Recruitment encounter rates of poaching signs of tigers and their prey would reflect intensified patrolling in the corresponding The best-fit CJS model (Supporting Information) held year and that the effects of these patrols would manifest capture probability constant across years, but allowed
Conservation Biology Volume 00, No. 0, 2015 55 6 Tiger Population Dynamics in Thailand
Figure 2. Estimated spatial pattern of tiger densities per 0.336 km2 pixel (generated with program SPACECAP) and camera-trap locations in Huai Kha Khaeng Wildlife Sanctuary from 2005 to 2012.
Table 2. Annual estimates of tiger densities (Dˆ ), abundance (Nˆ ), and model parametersa with posterior standard deviations (SD)b based on photographic capture data from Huai Kha Khaeng Wildlife Sanctuary (HKK), Thailand, 2005–2012. Year σˆ (SD) λˆ0 (SD) ψˆ (SD) Dˆ (SD) Dˆ 95% CI Nˆ (SE) 2005 4.080 (0.38) 0.050 (0.007) 0.483 (0.10) 1.82 (0.35) 1.19–2.53 50.60 (1.85) 2006 3.424 (0.37) 0.020 (0.004) 0.545 (0.11) 2.09 (0.39) 1.33–2.81 58.10 (2.06) 2007 4.512 (0.49) 0.013 (0.003) 0.385 (0.08) 1.46 (0.25) 0.97–1.95 40.59 (1.32) 2008 5.288 (0.40) 0.016 (0.002) 0.386 (0.06) 1.54 (0.21) 1.12–1.95 42.81 (1.11) 2009 4.275 (0.36) 0.016 (0.003) 0.427 (0.07) 1.69 (0.24) 1.23–2.16 46.98 (1.27) 2010 3.635 (0.29) 0.026 (0.004) 0.414 (0.07) 1.60 (0.26) 1.14–2.08 44.48 (1.37) 2011 6.249 (0.57) 0.012 (0.002) 0.323 (0.06) 1.27 (0.19) 0.88–1.60 35.31 (1.00) 2012 3.495 (0.21) 0.023 (0.003) 0.392 (0.06) 2.01 (0.26) 1.51–2.53 55.88 (1.37) a Parameters: movement, σˆ; basal encounter rate, λˆ0; proportion of MCMC-data augmented individuals that are real, (ψˆ ). bValues derived from spatial capture–recapture analyses in program SPACECAP. See “Methods” section for descriptions of spatial capture– recapture model parameters and details on abundance estimation. survival rates to vary [φ(t),p(.)]. Overall capture proba- Protection Measures bility, annual survival and interval survival rates for tigers Official reports showed that patrol teams detected 88 car- were estimated under this model (Table 3). Annual sur- casses of prey species. Three incidents of tiger poaching vival rates ranged from 79.6% to 95.5% (Table 3), re- were detected and successfully investigated, although it is sulting in an overall survival rate of 82% (derived from likely that not all such cases were detected. Overall, inten- the constant survival model [φ(.),p(.)]) (Supporting Infor- sification of foot patrols appeared to reduce incidences mation). Annual recruitment of new individuals into the of poaching (Fig. 3). The negative correspondence be- population (0–24 tigers) varied, but it increased gradually tween increased law enforcement and extent of poaching during the study, except for a drop during the 2010–2011 was supported by different types of evidence, including interval (Table 3).
Conservation Biology Volume 00, No. 0, 2015 56 Duangchantrasiri et al. 7
Table 3. Estimates of tiger abundance (N ), recapture probability ( pˆ ), annual survival (φˆ ), interval survival (φˆ t), and recruitment (Bˆ ) from open-model Cormack–Jolly–Seber analysesa based on photographic capture data from for Huai Kha Khaeng Wildlife Sanctuary (HKK), Thailand, from 2005 to 2012.b
Year Nˆ (SE) ˆp (SE) φˆ (SE) φˆ t (SE) Bˆ (SE) 2005 50.60 (1.85) – 0.80 (0.08) 0.73 (0.08) – 2006 58.10 (2.06) 0.87 (0.03) 0.80 (0.08) 0.79 (0.08) 0 (5.03) 2007 40.59 (1.32) 0.87 (0.03) 0.96 (0.06) 0.95 (0.06) 5.17 (2.86) 2008 42.81 (1.11) 0.87 (0.03) 0.87 (0.05) 0.88 (0.05) 9.32 (2.67) 2009 46.98 (1.27) 0.87 (0.03) 0.59 (0.08) 0.57 (0.08) 16.81 (4.22) 2010 44.48 (1.37) 0.87 (0.03) 0.86 (0.05) 0.86 (0.05) 0 (2.70) 2011 35.31 (1.00) 0.87 (0.03) 0.90 (0.09) 0.90 (0.09) 24.05 (3.40) 2012 55.88 (1.37) 0.87 (0.03) – – – aAnalyses performed using the selected model [φ(t), p(.)] in program MARK. bSee “Methods” section for details on how abundance, interval survival, and recruitment were computed.
Figure 3. Foot-patrol efforts and encounter rates of poaching incidents from 2006 to 2012 in Huai Kha Khaeng Wildlife Sanctuary. number of poaching incidents (Pearson’s correlation r = patrolling efficacy) positively affected recruitment, the −0.39; P < 0.01) and detection of old poaching camps effect was very weak, as indicated by the small estimated (r = –0.55; P < 0.01), traps and snares (r =−0.87; P < β-coefficient (βˆ (SE) = 0.05 (0.02)) and the fact that the 0.01), and poached animal carcasses (r = –0.35; P < 0.2). 95% CIs straddled zero (−0.007 to 0.108). Models with Increased patrolling was positively correlated to numbers survival rate as a function of total annual patrol effort or of arrested poachers (r = 0.29; P > 0.2), although the encounter rate of poaching signs received little support. relationship was not significant. Because no clear increases in population density were Discussion evident from 2005 to 2012, we could not directly ex- plore effects of increased patrolling on tiger abundance Utility of Noninvasive Capture–Recapture Sampling or density. The temporal symmetry analysis indicated Our results showed that the number of individual tigers that a single model did not receive clear support from captured M + (and camera-trapping rates) did not < t 1 the data because 5 models had AICC 2 (Supporting vary monotonically with tiger densities or abundance Information). Among these, the models with the low- (Table 1). We were also able to map spatial density vari- est AICc value included survival probability as a func- ations within the HKK study area for each year (Fig. 2). tion of time and per capita recruitment as a function of Because a fully developed SCR modeling approach for poaching sign encounter rate (Supporting Information). demographically open populations does not currently Although poaching sign encounter rates (a surrogate for exist, we integrated our closed-model SCR estimates of
Conservation Biology Volume 00, No. 0, 2015 57 8 Tiger Population Dynamics in Thailand abundance with likelihood-based, nonspatial, open- number of breeding females, in adult or cub survival, model CR analyses. However, open-model extensions and in subadult dispersal (Sunquist 1981; Smith 1993; of the SCR currently under development may permit Goodrich et al. 2008). addressing questions of demography more thoroughly With over 50 tigers, HKK probably supports the largest in terms of age, sex-specific social organization, and land source population of wild tigers outside the Indian sub- tenure. continent, highlighting its strategic importance for range- wide tiger recovery. Tiger survival rates were generally Study Limitations stable and reasonably high across all annual intervals in our study, except 2009–2010 when survival dipped to The potential violation of demographic closure assump- 59%. The overall survival rate of 82% at HKK compares fa- tion in our secondary samples (Table 1) may have resulted vorably with the rate of 77% observed in the high-density in some overestimation of tiger abundance due to under- protected population in Nagarahole (Karanth et al. 2006). estimation of basal encounter rate and potential turnover Annual recruitment rates of tigers (from in situ repro- of individuals. Given that the sampling durations were duction and immigration) showed a gradual increase and similar over time (excluding 2005), we expect compa- attained relatively higher levels toward the end of our rable magnitudes of bias in basal encounter rate and in study period. We speculate that the dip in survival in estimates of abundance across all years. Therefore, esti- 2009–2010 likely caused the drop in recruitment in sub- mated rates of recruitment and population change may sequent intervals, depressing the population size as ob- be relatively unbiased. Our tiger density estimates were served in 2011. Subsequent rebound of survival rates in lower than those reported by Simcharoen et al. (2007), 2010–2011 and 2011–2012 resulted in a large recruitment likely because the 477 km2 area they sampled supports of 24 tigers into the population in 2011–2012, raising the the highest tiger density within HKK due to its history of population size to 56 tigers by 2012. protection and their reliance on conventional CR analysis Although per capita recruitment into the sampled pop- with the half MMDM buffer probably resulted in density ulation appeared to respond favorably to increased detec- estimates higher than SCR analyses (Sharma et al. 2010; tion of poaching signs, the influence was not significant. Karanth et al. 2011a). We contend that tiger densities Responses in survival rates of adult tigers to increased pa- in HKK were not substantially higher in the area we trolling may still occur indirectly through increased prey sampled before our study. Variation in the number of densities, which is likely to manifest after a time lag of trap locations over time was fully accounted for in the several years. Although recruitment into the population SCR analyses, which included the spatial and temporal of tigers >1 year old may increase slightly through greater deployment of camera traps. The CJS analysis was robust cub survival, we speculate that far greater increases in to increase in sampled area after 2005 because capture survival are likely to occur after a time lag of several histories were conditioned on first captures and there years (through cub survival and immigration) in response was no strong reduction in the trap array area over time. to increase and stabilization of prey densities following Estimation of recruitment and rate of population change increased patrol effort. Overall, our results do not pro- were both sensitive to increases in the sampled area; vide unambiguous evidence of an increase in the tiger therefore, we excluded data from the first year of sam- population size in HKK between 2006 and 2012. pling (2005) when estimating these quantities. Finally, our indices of poaching intensity were subject to the same problems that weaken all indices of abundance. Fu- Conservation and Management Implications ture monitoring of poaching would benefit from survey Excessive hunting is a recognized driver of species de- protocols that enable these poaching intensity indices to clines globally. Establishment of effective protected areas be corrected for detectability. has been pivotal to reversing species declines and pre- venting extinctions of many taxa (Hoffmann et al. 2010). Tiger Population Dynamics in Huai Kha Khaeng In this context, the endangerment of the iconic tiger Tiger densities in HKK currently appear to be well below has received significant conservation focus for nearly potential levels (Supporting Information), similar to den- 5 decades. Walston et al. (2010) argue that preventive sities in Indian reserves such as Pench (Madhya Pradesh), law enforcement focused on tiger population sources Tadoba, and Bhadra (3.3–4.9 tigers/100 km2 from 2000 to (typically located in protected areas) is the most effec- 2004), which had relatively short histories of protection tive species recovery strategy, whereas Wikramanayake (Karanth et al. 2004). Tiger densities showed a substan- et al. (2011) advocate efforts across wider landscapes. Re- tial fluctuation of 62% (1.3–2.1 tigers/100 km2) over the gardless of these nuances, rigorous assessment and long- 7 years of our study (Table 2). The fluctuation across 9 term monitoring of tiger population response is critical years was even higher at 152% (8.6–21.7 tigers/100 km2) as demonstrated by our results. in Nagarahole, India (Karanth et al. 2011b), despite much Even expensive and intensive conservation interven- higher densities. Such fluctuations are inherent to tiger tions are rarely accompanied by such assessments (Walsh populations and arise from demographic stochasticity in & White 1999; Sutherland et al. 2004; Balme et al. 2009).
Conservation Biology Volume 00, No. 0, 2015 58 Duangchantrasiri et al. 9
Historical evidence from tiger reserves in India and Nepal of capture probabilities, abundance, trap array, buffer suggests that 10–15 years of intensive protection of width, sampled area, and tiger densities derived source sites is required before prey populations attain from closed-model conventional capture–recapture anal- ecologically optimal densities. Thereafter, range sizes of yses (Appendix S2), model-selection statistics from breeding female tigers shrink in response (Sunquist 1981; Pradel’s (1996) temporal symmetry analyses in HKK Smith 1993; Karanth et al. 1999), leading to denser pack- (Appendix S3), and spatial intensity of patrols in HKK ing of breeders and increased tiger abundance. from 2006 to 2012 (Appendix S4) are available as part of We found that antipoaching efforts in HKK intensified the on-line article. The authors are solely responsible for substantially after 2006. These efforts were correlated the content and functionality of these materials. Queries with declines in evidence of poaching and possibly led (other than absence of the material) should be directed to improvements in tiger survival and recruitment rates, to the corresponding author. although we were not able to demonstrate a clear re- lationship. A major criminal gang engaged in tiger and Literature Cited prey poaching was apprehended in 2011 (Department of Balme G, Slotow R, Hunter L. 2009. Impact of conservation interven- National Parks 2013). In one incident, camera-trap survey tions on the dynamics and persistence of a persecuted leopard (Pan- teams detected poaching of 3 tigers. Vigorous follow up thera pardus) population. Biological Conservation 142:2681–2690. by managers led to apprehension of the poachers. It is Bhumpakphan N. 1997. Ecological characteristics and habitat utilization plausible that the drop in tiger survival during 2008– of gaur (Bos gaurus H. Smith, 1827) in different climatic sites. Thesis, Kasetsart University, Bangkok, Thailand. 2009 was a consequence of increased levels of organized Borchers DL, Efford M. 2008. Spatially explicit maximum likelihood poaching. Increases in tiger densities were not evident methods for capture–recapture studies. Biometrics 64:377–385. after 7 years, possibly due to these continued poaching Brooks TM, Wright SJ, Sheil D. 2009. Evaluating the success of conser- pressures or because of limitations in our methods. vation actions in safeguarding tropical forest biodiversity. Conserva- Targeted management and increased protection have tion Biology 23:1448–1457. Burnham KP, Anderson DR. 2002. Model selection and multimodel enabled rapid recovery of populations in some felid inference: a practical information-theoretic approach. Springer- species after negative impacts were controlled (Lindzey Verlag, New York. et al. 1992; Balme et al. 2009). In contrast, our results indi- Caro TM, O’Doherty G. 1999. On the use of surrogate species in con- cate that population recoveries of tigers in the face of pre- servation biology. Conservation Biology 13:805–814. vailing levels of poaching pressures in Southeast Asia are Ceballos G, Ehrlich PR. 2002. Mammal population losses and the ex- tinction crisis. Science 296:904–907. likely to be much slower and uncertain, as seen in HKK. Cooch EG, White GC. 2009. Using MARK- a gentle introduction. Avail- Therefore, the goal of doubling wild tiger populations in able from http://www.phidot.org/software/ (accessed April 2013). 10 years set by the Global Tiger Initiative (2013) appears Cormack RM. 1964. Estimates of survival from the sighting of marked to be unsupported on the basis of current evidence. animals. Biometrika 51:429–438. Department of National Parks. 2010. Wildlife conservation in Thailand. Department of National Parks Report, Wildlife and Plant Conserva- Acknowledgments tion, Bangkok. Department of National Parks. 2013. Wildlife conservation in Thailand. We thank the Wildlife Conservation Society, New York, Department of National Parks Report, Wildlife and Plant Conserva- for supporting our research. We are grateful to the tion, Bangkok. following donors for funding support: The U.S. Fish Geweke J. 1992. Evaluating the accuracy of sampling-based approaches to the calculation of posterior moments. Pages 169–193 in Bernardo and Wildlife Service, U.S. Department of State, Liz Clai- JM, editor. Bayesian statistics 4. Oxford University Press, Oxford, borne and Art Ortenberg Foundation, Panthera, Save UK. the Tiger Fund, Disney Worldwide Conservation Fund, Global Tiger Initiative. 2013. Global Tiger Recovery Program Implemen- National Geographic Society, and Petroleum Authority tation Plan 2013–14. The World Bank, Washington, DC. of Thailand: Production & Exploration (PTT-EP). We Goodrich J, Kerley L, Smirnov E, Miquelle D, McDonald L, Quigley H, Hornocker M, McDonald T. 2008. Survival rates and causes of also acknowledge support from the Department of Na- mortality of Amur tigers on and near the Sikhote-Alin Biosphere tional Parks, Wildlife and Plant Conservation (Thailand), Zapovednik. Journal of Zoology 276:323–329. The Royal Thai Police and Kasetsart University-Thailand. Gopalaswamy AM, Royle JA, Hines JE, Singh P, Jathanna D, Kumar We are grateful to J.D. Nichols, J.A. Royle, and A.M. NS, Karanth KU. 2012. Program SPACECAP: software for estimating Gopalaswamy for useful discussions on methods and to animal density using spatially explicit capture–recapture models. Methods in Ecology and Evolution 3:1067–1072. V.R. Goswami for analytical assistance. Hebblewhite M, Percy M, Serrouya R. 2003. Black bear (Ursus ameri- canus) survival and demography in the Bow Valley of Banff National Park, Alberta. Biological Conservation 112:415–425. Supporting Information Hiby L, Lovell P, Patil N, Kumar NS, Gopalaswamy AM, Karanth KU. 2009. A tiger cannot change its stripes: using a three-dimensional model to match images of living tigers and tiger skins. Biology Letters Model-selection statistics for the open-model Cormack– 5:383–386. Jolly–Seber analyses used to estimate survival and de- Hoffmann M, et al. 2010. The impact of conservation on the status of tection probability (Appendix S1), annual estimates the world’s vertebrates. Science 330:1503–1509.
Conservation Biology Volume 00, No. 0, 2015 59 10 Tiger Population Dynamics in Thailand
Jolly GM. 1965. Explicit estimates from capture-recapture data with Prayurasiddhi T. 1997. The ecological separation of gaur (Bos gaurus) both death and immigration-stochastic model. Biometrika 52:225– and banteng (Bos javanicus) in Huai Kha Khaeng Wildlife Sanc- 247. tuary, Thailand. Wildlife Conservation, University of Minnesota, St. Jotikapukkana S, Berg A,˚ Pattanavibool A. 2010. Wildlife and human Paul, Minnesota. use of buffer-zone areas in a wildlife sanctuary. Wildlife Research Royle JA, Chandler RB, Sollmann R, Gardner B. 2013. Spatial capture- 37:466–474. recapture. Academic Press, New York. Karanth KU, Nichols JD. 2002. Monitoring tigers and their prey: a Royle JA, Karanth KU, Gopalaswamy AM, Kumar NS. 2009. Bayesian manual for wildlife researchers, managers and conservationists in inference in camera trapping studies for a class of spatial capture- tropical Asia. Centre for Wildlife Studies, Bangalore. recapture models. Ecology 90:3233–3244. Karanth KU, Nichols JD. 2010. Non-invasive survey methods for assess- Schipper J, Chanson JS, Chiozza F, Cox NA, Hoffmann M, Katariya ing tiger populations. Pages 241–261 in Tilson RL, Nyhus PJ, editors. V, Lamoreux J, Rodrigues AS, Stuart SN, Temple HJ. 2008. The Tigers of the world: science, politics and conservation of Panthera status of the world’s land and marine mammals: diversity, threat, tigris. Elsevier, New York. and knowledge. Science 322:225–230. Karanth KU, Nichols JD, Kumar NS. 2011a. Estimating tiger abundance Seber GAF. 1965. A note on multiple-recapture census. Biometrika from camera trap data: field surveys and analytical issues. Pages 97– 52:249–259. 117 in O’Connell AF, Nichols JD, Karanth KU, editors. Camera traps Sharma RK, Jhala Y, Qureshi Q, Vattakaven J, Gopal R, Nayak K. 2010. in animal ecology: methods and analyses. Springer, Tokyo. Evaluating capture-recapture population and density estimation of Karanth KU, Nichols JD, Kumar NS, Jathanna D. 2011b. Estimation of tigers in a population with known parameters. Animal Conservation demographic parameters in a tiger population from long-term cam- 13:94–103. era trap data. Pages 145–161 in O’Connell AF, Nichols JD, Karanth Simcharoen S, Pattanavibool A, Karanth KU, Nichols JD, Kumar NS. KU, editors. Camera traps in animal ecology: methods and analyses. 2007. How many tigers Panthera tigris are there in Huai Kha Springer, Tokyo. Khaeng Wildlife Sanctuary, Thailand? An estimate using photo- Karanth KU, Nichols JD, Kumar NS, Hines JE. 2006. Assessing tiger pop- graphic capture-recapture sampling. Oryx 41:447–453. ulation dynamics using photographic capture-recapture sampling. Smith JLD. 1993. The role of dispersal in structuring the Chitwan tiger Ecology 87:2925–2937. population. Behaviour 124:165–195. Karanth KU, Nichols JD, Kumar NS, Link WA, Hines JE. 2004. Tigers Soule ME, Estes JA, Miller B, Honnold DL. 2005. Strongly interacting and their prey: predicting carnivore densities from prey abundance. species: conservation policy, management, and ethics. BioScience Proceedings of the National Academy of Sciences of the United 55:168–176. States of America 101:4854–4858. Stokes EJ. 2010. Improving effectiveness of protection efforts in tiger Karanth KU, Stith BM. 1999. Prey depletion as a critical determi- source sites: developing a framework for law enforcement monitor- nant of tiger population viability. Pages 100–113 in Seidensticker ing using MIST. Integrative Zoology 5:363–377. J, Christie S, Jackson P, editors. Riding the tiger: tiger conserva- Sunquist ME. 1981. The social organization of tigers (Panthera tigris) tion in human dominated landscapes. Cambridge University Press, in Royal Chitwan National Park, Nepal. Smithsonian Contributions Cambridge, United Kingdom. to Zoology 336:1–98. Karanth KU, Sunquist ME, Chinnappa KM. 1999. Long term monitoring Sutherland WJ, Pullin AS, Dolman PM, Knight TM. 2004. The need of tigers: lessons from Nagarahole. Pages 114–122 in Seidensticker for evidence-based conservation. Trends in Ecology & Evolution J, Christie S, Jackson P, editors. Riding the tiger: tiger conserva- 19:305–308. tion in human dominated landscapes. Cambridge University Press, van de Kerk M, de Kroon H, Conde DA, Jongejans E. 2013. Carnivora Cambridge, United Kingdom. population dynamics are as slow and as fast as those of other mam- Lambert CMS, Wielgus RB, Robinson HS, Katnik DD, Cruickshank HS, mals: implications for their conservation. PLOS ONE 8: (e70354) Clarke R, Almack J. 2006. Cougar population dynamics and viability DOI: 10.1371/journal.pone.0070354. in the Pacific Northwest. Journal of Wildlife Management 70:246– Walpole MJ, Morgan-Davies M, Milledge S, Bett P, Leader-Williams N. 254. 2001. Population dynamics and future conservation of a free-ranging Lindzey FG, Van Sickle WD, Laing SP, Mecham CS. 1992. Cougar popu- black rhinoceros (Diceros bicornis) population in Kenya. Biological lation response to manipulation in southern Utah. Wildlife Society Conservation 99:237–243. Bulletin 20(2):224–227. Walsh PD, White LJ. 1999. What it will take to monitor forest elephant Margules CR, Pressey RL. 2000. Systematic conservation planning. Na- populations. Conservation Biology 13:1194–1202. ture 405:243–253. Walston J, Robinson JG, Bennett EL, Breitenmoser U, da Fonseca GA, Palomares F, Rodrigues A, Revilla E, Lopez-Bao J, Calzada J. 2010. As- Goodrich J, Gumal M, Hunter L, Johnson A, Karanth KU. 2010. sessment of the conservation efforts to prevent extinction of the Bringing the tiger back from the brink—the six percent solution. Iberian lynx. Conservation Biology 25:4–8. PLOS Biology 8: (e1000485) DOI: 10.1371/journal.pbio.1000485. Petdee A. 2000. Feeding habits of the tiger (Panthera tigris) in Huai WCS-Thailand. 2007. Building a monitoring system for tiger conserva- Kha Khaeng Wildlife Sanctuary by fecal analysis. Faculty of Forestry, tion in the Western Forest Complex, Thailand. A final report to US Kasetsart University, Bangkok. Fish and Wildlife Service. WCS Thailand, Bangkok. Pollock KH, Nichols JD, Brownie C, Hines JE. 1990. Statistical infer- Wikramanayake E, Dinerstein E, Seidensticker J, Lumpkin S, Pandav ence for capture-recapture experiments. Wildlife Monographs 107: B, Shrestha M, Mishra H, Ballou J, Johnsingh A, Chestin I. 2011. 3–97. A landscape-based conservation strategy to double the wild tiger Pradel R. 1996. Utilization of capture-mark-recapture for the study population. Conservation Letters 4:219–227. of recruitment and population growth rate. Biometrics 52:703– Williams BK, Nichols JD, Conroy MJ. 2002. Analysis and management 709. of animal populations. Academic Press, California, USA.
Conservation Biology Volume 00, No. 0, 2015
View publication stats 60
A context-sensitive correlated random walk: A new simulation model for movement
Sean C. Ahearna,∗ Somayeh Dodgeb,d∗, Achara Simcharoenc, Glenn Xavierd, James L.D. Smithb
International Journal of Geographical Information Science (in press, August 2016)
aCity University of New York – Hunter College bUniversity of Minnesota, Twin Cities cDepartment of National Parks, Plant Conservation, Thailand dUniversity of Colorado, Colorado Springs
Abstract Computational Movement Analysis focuses on the characterization of the trajectory of individuals across space and time. Various analytic techniques, including but not limited to random walks, brownian motion models, and step selection functions have been used for modeling movement. These fall under the rubric of signal models which are divided into deterministic and stochastic models. The difficulty of applying these models to the movement of dynamic objects (e.g. animals, humans, vehicles) is that the spatiotemporal signal pro- duced by their trajectories a complex composite that is influenced by the ge- ography through which they move (i.e. the network or the physiography of the terrain), their behavioral state (i.e. hungry, going to work, shopping, tourism, etc.), and their interactions with other individuals. This signal reflects multi- ple scales of behavior from the local choices to the global objectives that drive movement.preprint In this research we propose a stochastic simulation model that in- corporates contextual factors (i.e. environmental conditions) that affect local choices along its movement trajectory. We show how actual GPS observations can be used to parameterize movement and validate movement models, and argue that incorporating context is essential in modeling movement.
keywords: Movement model; stochastic models, agent-based simulation; en- vironmental context; behavior; movement pattern, scale.
∗Corresponding authors. Email: [email protected], [email protected]
1 61
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016
1 Introduction
Movement is essential to almost all organisms. The advent of inexpensive and ubiqui- tous positioning technologies such as Global Positioning Systems (GPS) has resulted in unprecedented datasets that can be used for quantitative assessment of the move- ment of individuals (i.e. animals, humans) and their collective dynamics (Galton, 2005). As a consequence, the study of movement has gained significant momentum in Geographic Information Science (GIScience) and its applications in movement ecol- ogy (Demˇsaret al., 2015; Holyoak et al., 2008), mobility and transportation (Tribby et al., 2016; Song et al., 2015), behavioral studies (Gonz´alezet al., 2008; Sang et al., 2011; Torrens et al., 2012), and public health (Glasgow et al., 2014; Lu and Fang, 2014), to name but a few. Numerous deterministic and stochastic methods have been developed to model movement, generate synthetic trajectories, and analyze patterns of movement (Schick et al., 2008; Laube, 2014; Dodge et al., 2016). These methods include various exten- sions of time geography approaches (Winter and Yin, 2010; Hornsby and Egenhofer, 2002; Miller, 1991), L´evyflights (Jiang et al., 2009; Rhee et al., 2011), random walks (Ahearn et al., 2001; Batty et al., 2003; Gautestad and Mysterud, 2005; Codling et al., 2008; Technitis et al., 2014), brownian bridges (Horne et al., 2007; Kranstauber et al., 2012; Buchin et al., 2012), Step Selection Functions (Squires et al., 2013; Thurfjell et al., 2014), and hidden Markov models (Franke et al., 2004). These models mainly employ information about movement parameters (e.g. speed, distance, turn angle) and time to simulate possible trajectories in space and time (i.e random walks and L´evyflights), as well as to quantify a probable space accessible or used by dynamic objects (i.e. time geography and brownian bridge models). Most studies concentrate on the pattern of space use and visit probability (Downs and Horner, 2009; Kie et al., 2010; Benhamou and Riotte-Lambert, 2012; Song and Miller, 2014) while others have examined the complex interaction between movement behavior and variation in envi- ronmental conditions (Ovaskainen, 2004; Morales et al., 2005; Moorcroft et al., 2006; Forester et al., 2007; Ovaskainen et al., 2008; Ahearn et al., 2010). While significant progress has been made, a key question with respect to existing models is how well do they capture: the local choices that agents make as they move through their environment, their interaction with other individuals (Yuan and Nara, 2015), and their own geographic strategies for resource usage (Ahearn and Smith, 2005; Dodge etpreprint al., 2014; Dodge, 2016). The spatiotemporal signal produced by movement trajectories is in fact a complex composite that reflects multiple scales of behavior from the local choices to the global objectives that drive movement. Move- ment encapsulates several components: the internal states and behavioral states of the moving entity, and the space and context (i.e. environmental factors) through which it moves in time (Dodge, 2016). Existing approaches often confound movement patterns associated with external factors with those patterns associated with global scale behaviors. Accurate simulation models for movement should incorporate and deconstruct the environmental and behavioral drivers of movement for more reliable
2 62
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016
and realistic representations. Simulation of movement is essential for studying and predicting patterns and behavioral responses of moving agents to varying conditions and their interactions with other individuals. This study introduces a context-sensitive simulation model for movement based on the concept of the correlated random walk (Codling et al., 2008). The proposed model is different from existing approaches because it integrates environmental context into modeling movement as an integral part of each local choice along the movement tra- jectory. The strength of our approach is that the model is parameterized and validated using real movement data. The parametrization of the model is achieved through the development of probability distributions that relate actual movement to its context. We assess our proposed model on movement of endangered tigers (Panthera tigris) in West-central Thailand. The case study investigates the influence of geography and physiography (i.e. the shape of home range and landscape characteristics) on tiger’s movement characteristics.
2 Related Work on Modeling Movement
Random walks have broadly been used to model change or movement at local and global scales (David and Perry, 2013). A two dimensional or spatial random walk describes the probability of moving in a direction from the current position in space. Codling et al. (2008) provide a comprehensive review of different random walk models applied in biology and ecology applications. The simplest version of random walk is called the uncorrelated random walk. In this model, the next direction of movement is independent of the direction of the previous movement step. In contrast, the correlated random walk (CRW) involves a persistence in the direction of successive movement steps by introducing a correlation to the preceding directions (Turchin, 1998). Previous studies suggest that the movement of organisms is less random and CRWs are better suited to represent movement (David and Perry, 2013). In addition to the local persistence in direction, random walks can involve an external bias to maintain a global direction of movement from an origin towards a destination (Ahearn et al., 2001; Technitis et al., 2014). Random walks form the building block of many agent- based simulation models for movement (Ahearn et al., 2001; Batty et al., 2003; Tang and Bennett,preprint 2010; Torrens et al., 2012; Technitis et al., 2014). Ahearn et al. (2001) built an agent-based model to simulate the collective dynamics of male and female tigers (with or without their cubs), and their interactions with prey. Their simulation is based on a correlated random walk by introducing external biases (i.e. location of other tigers and location of prey) that are dynamic and a function of the animal’s state. Gautestad and Mysterud (2005) proposed a multi-scale random walk (MSR) which does not make the assumption of a low-order Markovian process (i.e. the current state is a function of the immediate previous state) that many models do. Their work uses the MSR to model the multiple spatiotemporal scales over which animals operate (Gautestad and Mysterud, 2010). In another study, Technitis et al.
3 63
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016
(2014) introduced a point to point correlated random walk model (i.e. for given origin and destination locations) based on a set of assumptions on speed capacities of moving individuals and a time budget, akin to time geography approaches. Random walk models are often used to generate a trajectory (i.e. sequence of loca- tions over time). In contrast, time geography and brownian bridge movement models mainly quantify a continuos space accessible by a moving individual. As such these models can represent space utilization distribution or visitation probability in space and/or in time. Time geography models are based on the concept of H¨agerstrand’s space-time prism (H¨agerstrand,1970). In time geography, accessibility is defined as a space that a moving entity can possibly reach given a time budget and a maximum speed (Miller, 1991). A variety of time geography approaches have been proposed to model movement in a planar space (Winter and Yin, 2010; Song and Miller, 2014), or in a network space (Song et al., 2015). The brownian bridge model quantifies the probability of being in a location between a start (A) and an endpoint (B) taking into account the distance between the points and the overall time that it takes to traverse from A to B (Horne et al., 2007; Kranstauber et al., 2012). Most existing simulation models are parametrized using a set of assumptions and rules based on the dynamic capacities of the moving agents (e.g. speed capacities, time budget, turn angle patterns). It is now possible to calibrate models based on the dis- tribution functions and correlations of movement parameters (e.g. speed, turn angle) computed from real observations, thus obviating the need for such assumptions. An- other limitation of existing movement models is that many of them ignore the internal states, behaviors, and context (i.e. external factors) that results in a set of movement patterns (Dodge, 2016; Dodge et al., 2016). They mainly consider dynamic capacities of moving individuals and external biases such as destination or interactions between moving individuals (Miller, 2015). A number of researchers have begun to incorpo- rate context into their understanding and modeling of movement (Ovaskainen, 2004; Morales et al., 2005; Moorcroft et al., 2006; Forester et al., 2007; Ovaskainen et al., 2008). Ovaskainen (2004) and Ovaskainen et al. (2008) used a segmented landscape with different diffusion coefficients for each landscape type with separate specification for the boundary condition. Their research provides a strong case for understanding movement in heterogeneous landscapes. Moorcroft et al. (2006) proposed a mecha- nistic home range model for carnivores that incorporates conspecific avoidance, rough terrain avoidance and habitat selection to derive home range patterns. The models resulting probabilitypreprint density functions were tested against the home range derived from observations of coyote GPS locations and demonstrated a strong “goodness of fit”. Schick et al. (2008) provide an excellent summary of these models and other existing approaches for modeling animal movement. The models they reviewed focus on three areas: simulating realistic movement, understanding organism-environment interaction and its effect on movement, and inferential models for predicting move- ment where data may be incomplete. In conclusion they find that while models have attempted to understand organism-environment interaction, “none of these models has an ability to test for how landscape features actually influence the movement
4 64
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016 process” (Schick et al., 2008). They propose a conceptual model that explicitly in- corporates state, location, and habitat suitability as factors governing the choice of an individual as it moves from place to place. A more recent development in understanding the relationship between environ- ment and movement is the Step Selection Function (SSF) (Thurfjell et al., 2014). The mechanism for this model is the Resource Selection Function (RSF) which usually uses a Logistic Regression to relate the probability of selecting the next resource unit to a set of environmental correlates defined by their frequency of use and availability (Thurfjell et al., 2014). While this is a powerful new method to understand these relationships, it has been used in a limited way to simulate movement. For instance, it has been applied to create a probability surface of use for each resource unit to calculate a least-cost path to generate point to point movement (Squires et al., 2013). This in essence is a deterministic model of movement in contrast to the stochastic model we propose in this paper.
3 Methodology
This section presents a new context-sensitive correlated random walk (CsCRW) method and compares it with a standard correlated random walk (CRW) to better understand the contribution of geography and contextual factors to an individual’s trajectory. The models are implemented in a Monte Carlo agent-based simulation environment in which trajectories are simulated within an area of interest (e.g. defined by the shape of a home range (Powell and Mitchell, 2012)). The simulated trajectories are used to generate a visitation probability surface for evaluation and comparison of the models, as described below.
3.1 Drivers of Movement Movement occurs in response to the internal state of a moving agent (e.g. the physi- ological state of being hungry) which results in a change in its behavioral states (e.g. hunting, patrolling) (Ahearn and Smith, 2005). The behavioral states of a moving agent are realized as patterns of movement that occur at different spatial and tempo- ral scales (Dodge, 2016). Environmental factors (e.g. slope, vegetation density, prey density and the existencepreprint of trails) often affect local scale choices in movement (e.g. direction and rate of movement). These choices are made based on local conditions under the broader constraints of more global scale behaviors driven by internal states of the moving agent (Figure 1). In this paper, context refers to environmental fac- tors that may influence the movement of an agent at local scales. Understanding the contribution of individual’s behavior at different scales to the observable patterns of movement is critical for their proper interpretation. In this paper we focus on the local scale choices affected by environmental context; how to calibrate them and how to validate their importance.
5 65
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016
internal state behavioral state (e.g. biological state) (e.g. movement goal)
environmental correlates (e.g. slope, vegetation)
local choices
patterns of movement
Figure 1: Drivers of movement
3.2 Correlated random walk (CRW) The algorithm implemented here (Figure 2) is a modified version of the standard CRW (Turchin, 1998) to simulate a trajectory T (equation 1). The algorithm first selects a random starting point (Step 1) and picks a persistence value for direction (p) from a random uniform distribution (Step 2). If it is less than the input value of persistence in direction, as calculated from the actual GPS observations, then the agent persist in the previous direction with a small probability of deviation (e.g. standard deviation 10 degrees, Step 3.a). If it is more, then the agent can move with a much larger turn angle (e.g. standard deviation 45 degrees, Step 3.b) to a new direction (α, equation 2), to reduce back tracking. Next, a distance d is selected from the probability density function derived from GPS observations (Step 4, equation 3). The agent is then moved for distance d in direction α forming a vector (Step 5, equation 4). This process is repeated to generate a trajectory in the area of interest using the same numberpreprint of points as the actual GPS trajectory (Figure 2). The formal representation of the process is as follows:
6 66
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016
T = {(x0, y0, t0), (x1, y1, t1), .., (xi−1, yi−1, ti−1), (xi, yi, ti), ..., (xn, yn, tn)} (1) ( ◦ 2 10 if p ≥ persistence αi = αi−1 + N(0, σ ), where σ = (2) 45◦ otherwise 2 2 di = t ∗ χ (µ), where χ (µ) is the distribution of speed obtained from GPS(3)
xi,t = d ∗ cos αi + xi−1,t−1 (4)
yi,t = d ∗ sin αi + yi−1,t−1
1. pick a random start 2. pick a direction location in persistence probability the area of value (p) start interest
x0,y0 yes p ≥ persistence
no
3.a. pick a random direction (a) from a normal distribution
Highest direction probability is in the direction of movement
probability Lowest direction probability 0 10-10 next turn angle primary movement direction
3.b. pick a random direction (a)
probability in each direction probability 0 45-45 primary movement direction next turn angle
4. pick a distance (d) from 5. move d meters in
a c c 2 distribution the selected xj,yj direction (a) a d xi,yi µ
preprintyes no end no_pts > MAX
Figure 2: Correlated Random Walk (CRW)
3.3 Context-sensitive correlated random walk (CsCRW) The CsCRW is a modification of the CRW algorithm to account for contextual fac- tors (i.e. slope) in generating a context-sensitive trajectory (T c, equation 5). The
7 67
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016
movement of the agent across space is simulated on the per-pixel basis rather than by a vector as in the CRW model, to incorporate local choices made by the agent based on context as it moves through space. Contextual factors are modeled by using a probability density function P (c) to determine the next probable move (Step 5.1, Figure 3, equation 6). This function is approximated by a chi-square distribution χ2(µ), derived from the relationship between the agent’s locations (i.e. GPS observa- tions) and the environmental context values (i.e slope use). The CsCRW algorithm follows the same process described for the CRW until Step 4 in Figure 2. Instead of moving along a vector in direction α for distance d, the CsCRW moves one pixel at a time based on context (Step 5.1, Figure 3). At each step, the pixel which is in di- rection α and its two adjacent neighbors become possible choices for the agent’s next move (Step 5.2). Among the three choices, the pixel whose context value is closest to the random context variable (c) selected from P (c) is chosen (Step 5.3, equation 7), formally:
c T = {(x0, y0, c0, t0), .., (xi−1, yi−1, ci−1, ti−1), (xi, yi, ci, ti), ..., (xn, yn, cntn)} (5) P (c) = χ2(µ), where χ2(µ) is distribution of context values used by an agent(6)
ci = Cj| min(|Cj − c|), where j ∈ J = {1, 2, 3} (7)
(xi,t, yi,t) = (X(ci), Y (ci)), where X, Y are the coordinates of choice ci (8)
The agent is moved to the selected pixel (Step 6, equation 8 ) and the process continues until the selected distance d is reached. When that happens it returns to Step 2 of CRW (Figure 2). This process is repeated to generate a trajectory in the area of interest using the same number of points as the actual GPS trajectory. This approach could in fact include any number of contextual factors (m) that can be combined through a joint probability function (P(c) in equation 10), formally:
ci = {c1, c2, ..., cm}, where m is the number of contextual factors (9) m Y P (c) = P (ck) (10) k=1 3.4 Assessmentpreprint of models To assess the CRW and CsCRW models a surface for probability of visitation is created for each model in the area of interest (e.g. home range). Each surface is created using a Monte Carlo simulation of 1,000 trajectories and intersecting the trajectories with the raster of the area of interest and accumulating the number of visits for each cell. To evaluate how well the surface for probability of visitation captures the actual visitation probability of the moving entity, its GPS trajectory is compared with a random trajectory of equivalent size. Our premise for evaluation is that the best model will result in higher cell counts along the GPS trajectory, than for the random trajectory.
8 68
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016
5.1. pick a c c 2 random context variable (c) step 4., CRW
µ
5.3. select a pixel with 5.2. take three pixels according to the a context value selected direction from step 3. closest to c (from 5.1.)
6. move to the selected pixel
(xj,yj)
(xi,yi)
is distance d no reached? current pixel (xi,yi) yes
step 2., CRW
Figure 3: Context-sensitive Correlated Random Walk (CsCRW)
4 Case Study and Results
4.1 Dataset To evaluate our proposed methods we use movement data for two female tigers tracked in Huai Kha Kaeng Wildlife Sanctuary in Thailand. Both tigers were fitted with a Vectronic Aerospace GmbH GPS Plus collar. Tiger T 7203 was tracked between December 2009 and July 2010 with a sampling rate of one hour, and tiger T 10727 was tracked between December 2012 and January 2015 with a sampling rate of four hours. Following removal of outliers and erroneous GPS observations, T 7203 consisted a total of 4874 points and T 10727 consisted of 3946 GPS points. Figure 4 shows trajectories and computed homepreprint ranges of the two tigers, and the elevation maps of the areas. The home range of the tigers are calculated using both a 100% Minimum Convex Polygon (MCP) (Mohr, 1947) (shown in magenta in Figure 4) and a Characteristic Hull Polygon (CHP) (Duckham et al., 2008) (shown in dark blue). The characteristic hull polygon is generated from the Delaunay triangulation (Okabe et al., 1992) of the set of GPS observations. The CHP is extracted from the Delaunay triangulation through the removal of triangles determined by the chi-shape algorithm, resulting in a bounding hull with non-convex edges (Duckham et al., 2008). This method was selected for the analysis because it is one of the most parsimonious methods for determining a home range, as can be seen in the comparison with the commonly used
9 69
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016
MCP approach (Downs and Horner, 2009) (Figure 4). The generated home ranges using the Delaunay triangulation method resulted in a 34.26 km2 boundary for tiger T 7203 (93.7% of MCP), and a 99.92 km2 boundary for T 10727 (89.1% of MCP). As seen in Figure 4, the home ranges of the tigers show a diversity of terrain. Both tiger tracks are annotated with elevation and slope information from the digital elevation model (DEM) of the study areas obtained from ASTER GDEM dataset1 (30 meter resolution) using a bilinear interpolation (Dodge et al., 2013).
Vietnam ! Myanmar ! !! !! ! ! ! ! ! !! ! ! !! !!!!! ! ! ! !!!!!!!! !!!!!!! !! !!! ! !! ! ! ! !! ! ! ! !! !!!!! ! !! !! ! ! !! ! ! ! ! ! !! ! ! !! ! ! ! !! !!! !!!!!!!!!!!!!!!!! ! ! !! !!!!!!!!!! ! ! !! ! ! ! ! ! ! Thailand ! ! ! !! ! ! !!! !! ! ! ! ! ! !! ! ! ! ! !!! ! ! ! ! ! ! ! !!!!! !!!! !!! ! ! !!!! !!!!!!!!!!!! ! ! ! !! ! ! ! !!! !! !! !!!!!!!!!!! ! !!!!!! ! !!!!!! ! !! ! !!!!!!!!!!!!!! ! !! !!!!! !!!!!!!!!!!! ! !!! ! ! ! ! !!!!!!!!! ! Cambodia ! !! ! ! !!!!!!!!!!! !!!! ! !!!!!!!!!!!!! !! !! ! ! ! ! !! ! ! ! ! ! ! ! ! ! !!!!!!!! ! ! !! !!!! ! ! ! !!!!!!! ! !!!! ! ! !! !!!!!!!!!!!!!!! ! !!!!!! ! !! !! ! !! ! ! ! ! ! !!!!!!! ! ! ! ! ! !!!!! ! ! ! ! ! ! !! !! ! ! !! !!!!!!!!!!! ! !!! ! ! !! ! ! !! !!!! !! !!!! !! !! ! ! !!! ! !!! !!!!!!!! ! ! ! !!!!! ! ! ! ! !! !!!! ! ! !!! ! !! ! ! !! ! !!!!!!!!!! ! ! ! !!!!!!!!!!!! ! ! !!! ! ! ! ! !!! !! ! !! ! ! ! ! !!!!!!!!!!!!!!!! ! ! ! ! ! !!!!!!! ! !!!! ! !!!!!!!! ! !!! !!!!!!!!!! !! !!!! !!!!! ! ! !!!!! !!!! !!! !!! ! ! !!!!!!!! ! ! !!! ! ! ! !!!!! !!!! ! ! !!!!!!!!!! !! ! !! ! !!!! !! ! !!! ! !!!!!!! !! ! !!! !!! ! !! ! !!!!!!!!! !!!!!!! ! !! !!!! ! ! !!! !!! ! !!!!!! ! ! ! ! !! !!! ! ! !!! ! ! !!!!!!!! ! ! !! !! ! ! ! ! ! !! !! ! !!!! !! ! ! ! !!! ! ! !! !!!!!!!!!! ! !! ! ! ! ! !!!!!!!!!! ! !!!!!!!!!!!! !! ! ! ! ! !! !!! ! !!!! !!!!!!! ! ! ! !! !!!!! ! !!! ! ! ! !! ! ! !!!!!!!!!!!!!! ! !! ! ! ! !!! ! ! !!! !!!!!! !! !!!! ! ! !!!! ! !!! !! !!!! ! ! ! ! !! ! ! ! !!!! ! !! !!! ! ! ! !! !!!!!!!!!!! ! !!!!!!!! ! ! ! !! !!! ! ! ! ! ! !! ! ! !!!!! ! ! ! ! ! ! ! ! ! ! !! !! ! ! ! ! !! ! !!!! !!! ! !! !! ! !! ! !!! ! !! !!! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !!! ! ! ! ! ! !!! !! ! !!!! ! ! !! ! ! ! ! ! ! !!!!!! ! ! ! !!!! !!!!!!!! ! ! !!!!!! ! ! ! ! ! ! ! !!!!!!!!!!!!!!!!!!! !!! !!!!!!!! ! !!!!! !! ! !!!!!!!!!!!!!!!!!!!!!!!! !! !!!! !! ! !!!!! !!!!!!!!! !! !!!!!!!!!!!!!!! !! !!!!!! !! !!!! ! ! !!!!!!!!!!!!!! !! !!!!!!!!!! ! ! !! !! ! !!!!!!!!!! !! ! ! !!!!!!!!! !!! ! !! !! ! ! !!!!! !! ! !!!! ! ! ! ! ! ! ! ! ! !! ! ! ! ! !! ! ! ! !!! ! ! ! !!!!!!!!! !! ! ! !! ! !! !!! ! !!!!!!!! !! ! ! !! !!!!!! ! ! ! !!!! ! ! !! !!!! ! ! !!!!!!!!!!!! ! ! !!!!!!!!!! ! ! !!!! ! !!!!! ! !! ! ! !!!!!!!!!!!!! ! !!! ! !!! ! !! !!! !!!!!! ! !!!! !!! !! !!! !! ! ! ! !!!! ! !! ! ! ! !! ! ! ! !!!!!!!!!! ! ! !!! ! !! ! ! !! !! ! ! ! ! ! ! ! ! ! !! !! ! !!! !! !! ! ! ! ! ! ! ! !!!!!!! ! !!!!! ! !!!! ! ! ! ! !! ! ! ! ! !! !! !! !! !!! !!!!!!!!!!!! ! ! !! ! !!!!!!!! ! ! ! ! ! !! ! ! !! !!! !!!!! !!! ! !! !!! ! ! ! ! ! !!!!! !!!! !! !!!! ! !! !! ! ! ! ! ! ! !!! ! ! ! !! ! ! !!!! !!!! ! ! !!! ! !! ! ! ! ! ! ! ! ! ! ! ! ! !!!!!! ! ! ! ! ! ! ! ! !!! ! ! ! !!! ! !!!! !! ! ! ! !!!! ! ! ! !!!!!!!!!!!! ! !!! ! ! !! ! ! ! ! ! ! !! ! !! ! ! ! ! ! ! ! ! !!!! ! ! ! ! ! ! ! !!!! ! !! !!!!! ! ! ! !! ! !!! ! ! !! !! !! !!!!! !! ! ! ! !!!!!!!!!!!!!!!!! !!!!!! ! !! !! ! !! !!! ! !! ! ! ! ! ! !!!! !! !! !!! ! ! !! ! !!!! ! ! ! ! ! !! ! ! !!!! ! !!!!!!!!!! ! ! ! ! ! ! !!! !!!!!!!!!!! !!!! ! ! ! ! !!!!!!!!!!!!! ! ! ! !! ! !!!!!!!!! !!! !!!!!!!!!!!! ! ! ! ! !!! ! ! !! !!!!!!!!! !! ! ! ! !!!! ! ! !! !!!! ! ! !!!!!!! !! !! !!!!!!!!!!!!!!!!! ! !! !! ! ! ! !!!!! !! ! !!!! !! ! ! !!!!! ! ! ! !!!! !!! ! ! ! ! ! ! ! !! ! !!!!! ! ! ! ! ! !!!!!!!! ! ! ! !! !!!!!!!! ! ! Observed GPS Point !!! !!!!!!!!!! ! ! !!! ! ! ! ! ! ! ! !!!! !!! ! ! ! !! !!!!! ! !!! ! !!! ! !!! ! ! ! Observed Trajectory !!! !!!! ! !! !!!! ! !!! ! ! !!!! !!!!!!!! ! ! !!!!! !! !!! !!!!!!!!!!! ! ! !! !!! ! ! ! !!!!! !! !! !!! ! !!!!!!!!! ! ! ! ! ! !! ! !!!! !! ! ! !!!!!!!!!!!!!!!!!!!!!!!! ! ! ! !!!! ! ! Characteristic Hull Polygon ! !!!!!!!!!!!!!! ! !!!! ! ! !! ! !!!!! !!!! ! ! !! !!!!!!!!!!!!!! !!!!!!!!! ! !!!! ! ! ! ! !!! !!!!!!! ! !! ! ! ! ! !!!! ! !!!!!!!!!!!!!!!!!!! ! !!! ! ! ! ! !!!!!! !! ! MCP 100% !! ! !!!! !!!! !! ! ! ! ! ! ! ! ! ! !! ! ! ! ! ! !!! !! ! ! ! !! ! Elevation (m) ! !!! !!!! ! !!! !!!!!!!!! ! ! !!! ! !!!!!!! !!! ! !!!!!! High : 1946 ! ! !!!!!!!!! ! ! ! ! !!! !!!!!! ! !!! ! Low : 93 ! ! ! !! 0 0.5 1 2 Km ! ! !!
(a) T7203 trajectory and home range
!! ! !! !! ! ! Vietnam ! Myanmar !! ! ! ! ! ! ! !
! !!! ! ! ! ! Thailand ! ! ! Cambodia ! ! !! ! ! ! ! ! ! ! !!!!! ! ! !!!! ! !!!! ! ! ! ! ! ! ! !!!!!!!!! ! ! ! ! ! ! ! ! !! ! ! ! ! ! !!!!!! ! ! ! ! ! ! ! ! ! ! ! ! !! ! ! ! ! !!! ! !! ! ! ! ! ! ! ! !! !! ! ! ! ! ! ! ! ! !! ! !!!!!!!!! ! ! ! ! !!! ! !!!!!!!! ! ! !!!! !! ! !!! ! ! ! !!!!!! !!! !! !!!!! !! ! ! ! !! ! ! !!!!! !!!! ! ! ! ! !!!!!! ! !!!!!!!! ! !!!!!! !! !! ! ! !!! !! ! !!!!!!!!! !!!!! !!!!!! !!!! ! !! ! ! ! ! !!!! !! !! !! ! !! ! ! !!!!!!!!!!!!!! !!!!! ! !! ! !!!!!! ! !! ! !! !!! !!!!!!!! !!! !!!!! ! !!! !!! !! ! !! ! !!! !!! !!!!! ! !!!! !! ! ! ! ! !!!! ! ! !!! !!! ! ! !!!!!! !!!! ! !! !! ! !!! ! !!!!!!!!!!!! ! !!!!!!!!!!!!!!!!! !! ! !!! ! ! !! ! !! ! !!!! !!!!! ! ! ! ! ! ! ! ! ! ! !! !! !! !!!!! !! ! ! ! !!!!!!! ! !!!! ! ! ! ! ! ! !! ! ! ! ! ! !!! !!!! ! ! ! ! ! !!!!!!! ! ! ! ! !! ! ! ! ! ! ! !!! !!! !! ! ! !!!! !! !! ! ! ! ! !!!!! ! ! ! ! ! !!!!!!! !!!!!!! ! ! ! ! !! ! ! !!!! !!!!!!!!! !!!! ! !!!!!!! ! ! ! !! !! ! ! !!!! ! ! !!!!!! ! !!!!!! ! ! ! !! ! ! !!!!!! ! ! ! ! !! ! ! ! !!!! ! !!!! !! !! ! !!!!! !! ! ! ! !!!! !! !! !!!! ! ! ! !! ! !!! ! !!!!!!!! !!!!!!! ! ! ! ! ! !!!!! ! !!!!!!!!! !!! ! ! !!!!!!!!!!!!! ! !!!!!!! ! !!! !! !!!!!! !!!!! !! ! !!!!!!! ! ! !!! !!! ! !! !!!!! !!! ! !!!!!!! !! ! ! !!! ! ! !!!! ! !! ! !!! ! !! !! !!! ! ! ! ! !!!! ! !!!!!!!! !!!! ! ! ! !! !! !!!! ! !! ! ! !!! !!!! ! ! !!!!! ! !!! !!!!!!!! !! !!! !!! ! !!!!!! !!!!!!!!!!!!!! ! !!!! !! ! ! ! ! ! ! ! !!!!!!!!!! !! ! ! !!! ! ! ! !!! ! ! ! ! !! !!!!!!!! !!! ! ! ! ! !! ! ! ! ! ! ! !!!! ! ! !! ! ! ! !!! ! ! ! ! !! !!!! !! !!!!! !!! ! !! !!! ! !!! ! ! ! !!!!! !! ! !! ! ! ! !!! ! !! ! !! ! ! !! ! !!! !!! ! ! ! !!!! ! !! ! ! ! ! ! !! !!! !!!! ! !!!! ! ! !! !!!!!!!! ! ! !!!! !!!!! !! ! ! ! ! ! ! ! ! ! ! ! !!!!!!!!! ! !!!!!!!!!!!!!! !! ! ! ! ! ! ! !! ! !! ! !!!! ! ! ! ! ! ! ! !! ! !!!!! !! ! ! ! ! !!!!!! ! ! ! ! ! ! !! ! ! ! !! ! ! ! ! !!!!!!!!!!!!!! !!!! ! ! !!! !! ! ! ! !!!! ! ! !!! ! ! ! ! ! !!!!!! !!!!! !!!!!!!!! ! ! ! ! ! !!! !! !! ! !! !!!!! ! !!! !!!!!! !! ! ! ! ! ! ! !!! ! !!! !!!!!!!!! !!!! !! !! !!! ! ! ! ! !!!!!!!!!!!! ! !! ! !!!! ! ! ! !! ! ! !!! ! ! ! ! !! ! !! ! !!! ! !! ! ! ! ! ! ! !! ! ! !!!!!!! !!!!!!!! ! ! ! !! ! ! ! !! ! !! !!!!! !!!! ! ! ! !! ! ! !! ! !! ! ! ! ! !!!!!!!!!!! ! ! ! ! !!! ! !!! ! ! !! ! ! !! !!!!! ! !! ! ! !! !!!!!!!!! !! ! !!! ! !!!!!!!!!! ! ! ! ! !!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!! ! !! ! ! ! ! !! ! ! ! ! ! !! !! ! ! ! !!!!!! ! !! !!! ! ! ! !!! ! !!! !!!! ! ! ! ! !! ! ! !!! !!!! ! ! !! ! ! !!!!!!!!!!!!!!!!! !!! !!! ! !!! ! ! !! !!!!!!!! !! ! !!!!! ! ! !!!!!! !! ! !!! !!!! !! !!!!! !! ! ! ! !!!!!! ! ! ! !!!!!!!!!! ! ! !!!!!!!!!!!!!!!! !!!!!!!! !!!!!!!!!! !! ! ! !! ! ! !!!!!! ! !!!! !!!!!!!!!!!!! ! ! ! ! ! !!!! !! !! ! !!! ! !!!!!!!!!!!!!!! !!! !! ! ! ! ! !! ! !!! ! ! ! !!!! ! ! ! ! ! ! ! ! ! ! ! !!!!! !!!! ! ! !! ! ! ! ! ! ! ! !!!!!!!!!!!!!! ! ! ! !! ! !! ! ! !! ! ! !!!!!! !! !! !!! ! !!!! !! ! ! !! ! ! !!!!!!!!!! ! ! !! !!!! !!! ! ! !!! ! ! ! ! ! !! ! ! !! ! !! ! !!! !!! ! ! ! ! ! !!! ! !! ! ! ! !! ! ! ! ! ! ! !! !! ! !!! ! ! ! ! !! ! !!!! ! !! !!!! ! ! ! ! !! !!!!!!!!! ! ! ! ! !!!! !!!! !! ! ! ! !!!!! ! ! ! !! ! ! !! Observed GPS Point ! !! !!!!! ! !!! ! ! ! !!! !!!! ! !!! ! ! ! ! ! ! ! !! !! ! !! ! !! ! !!! ! ! ! !!!! ! ! !! ! ! ! ! Observed Trajectory !! ! !! ! ! ! !!!! ! !!! ! ! ! ! !! ! ! !!!!!!!! !!! !! ! ! ! !!!! ! ! ! ! !! !!!!! ! !!!!!!!! ! !!! ! ! !!! ! ! ! ! !! !! ! ! ! ! Characteristic Hull Polygon !! !!!!!!!! !!!!! ! ! ! ! ! !!! !! ! !!! ! ! ! ! ! !!!! !!! !! ! ! ! ! !!!! ! ! ! !!!!!!!! !! ! ! ! ! !!!! ! ! preprint! ! !! !!! ! !! ! ! !!!! !!! MCP 100% ! ! !!!!!!!! ! ! ! ! !!! ! ! !! ! !! ! ! ! ! ! ! ! ! ! ! Elevation (m) ! ! ! ! !! ! ! ! ! ! !!!! !! ! High : 1946 ! ! ! !!! !! ! ! ! !!! ! !!!! !!! Low : 93 0 1 2 4 Km !!
(b) T10727 trajectory and home range
Figure 4: The trajectories and home ranges of two female tigers, and the elevation maps of the study areas.
1Data source: http://asterweb.jpl.nasa.gov/gdem.asp
10 70
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016
4.2 Model Parameterization Three model parameters are calibrated from analysis of the tigers GPS data. These parameters are: directional persistence, speed of movement, and slope preference. Directional persistence and speed of movement are used to parametrize both the CRW and CsCRW models, while the slope persistence is used for the CsCRW model, as an example of a context variable. Directional persistence is calculated as the ratio of the total number of turn angles between −20◦ and +20◦ to the total number of GPS vectors and resulted in a value of 0.3. To parametrize speed of movement, a probability density function is created using 1-hourly GPS observations and estimated by a χ2 distribution with a mean of 0.24 km/hr (Figure 5). To parameterize slope preference probability density functions are created from the slope annotated GPS observations of the two tigers, and were estimated with χ2(µ = 5.8◦) for T 7203 and χ2(µ = 6.1◦) for T 10727, as shown in Figure 6. It is important to remark that the data of the two tigers resulted in very similar mean estimates for the χ2 distributions of their slope use. It is necessary to note that since speed and directional persistence are spatial- temporal measures, they are computed only using the 1-hour data set for parame- terizing both models. However, since slope preference is a spatial phenomenon, the temporal resolution of the data has no effect on estimation of slope use. Therefore the differing temporal resolutions of the two tigers do not impact the model parame- terization or the results of the validations.
T7203 T10727 0.15 0.15 mean = 0.24 km/h chi−square chi−square mu = 5.8 deg mu = 6.1 deg 0.10 0.10 density density density 0.05 0.05 0 1 2 3 4 5 6
0.0 0.5 1.0 1.5 2.0
speed values (km/h) preprint0.00 0.00 0 5 10 15 20 25 30 0 10 20 30 40 Figure 5: Kernel density plot (a) slope values (degree) (b) slope values (degree) of speed values obtained from GPS observations, estimated Figure 6: Kernel density plots of slope values obtained with a χ2 distribution of µ = from GPS observations of T 7203 and T 10727, fitted with 0.24 km/hr χ2 distributions of µ = 5.8◦ and µ = 6.1◦, respectively.
11 71
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016
4.3 Results 4.3.1 Correlated random walk (CRW) Speed distribution and directional persistence obtained from model parameterization are used as input to calculate the distance and direction for each successive move in the correlated random walk model (as described in Section 3.2). Figures 7(a) and 8(a) show probability of visitation surfaces, generated through a Monte Carlo simulation using a CRW model (as described in Section 3.4), for tigers T 7203 and T 10727. The resolution of the generated probability of visitation surfaces is the same as the DEM (30 meters). The highest visitation counts (shown in red) are away from the borders of the home ranges and the areas near the borders show lower visitation counts in blue. As seen in the Figures 7(a) and 8(a), this model captures the impact of geography (i.e. location within the home range) on visitation probability.
4.3.2 Context-sensitive correlated random walk (CsCRW) The sensitivity of tiger movement to slope (i.e. slope preference) for the CsCRW model is parametrized for the two tigers using χ2 distributions of µ = 5.8◦ and µ = 6.1◦ as shown in Figure 6. These functions are used to incorporate slope selection for each successive move in the CsCRW model (as described in Section 3.3). Figures 7(b) and 8(b) show probability of visitation surfaces, generated through a Monte Carlo simulation using a CsCRW model (as described in Section 3.4), for tigers T 7203 and T 10727. The resolution of the generated probability of visitation surfaces is the same as the DEM (30 meters). As seen in the figures, this model captures the impact of terrain on the visitation probability of the agent. The areas in blue (i.e. low visitation) correspond to the high slope areas or hilltops, which are difficult to reach for tigers. These rougher terrains, as highlighted on slope maps on Figures 7(c) and 8(c), result in the creation of corridors of high visitation for the CsCRW model (i.e. higher visitation counts shown in red) along river bottoms, valleys, and ridge lines.
4.4 Assessment Two tiger trajectories from GPS observations are used for validation. For each tiger, a random trajectory of the same number of points as the tiger trajectory is generated (i.e. 4874 pointspreprint for T 7203, and 3946 points for T 10727). The visitation count for each pixel along the tiger trajectory and the random trajectory is extracted from the probability of visitation surfaces of both models (i.e. CRW and CsCRW) as described in Section 3.4. The results are presented as boxplots in Figure 9 and summarized using means, medians and their differences for the tiger and random trajectories (Table 1). Both CRW and CsCRW models result in significant higher means and medians for both tiger GPS trajectories as compared to the random trajectories, using a T- squared test (p < .0001). The CsCRW model shows much higher mean and median differences for visitation counts (i.e. diff in Table 1) than the CRW model for both tigers. For T 7203, the CsCRW mean difference is 245.4 (visits per cell), while the
12 72
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016
Figure 7: Visitation counts for the home range of tiger T 7203 obtained from (a) CRW and (b) CsCRW models; and (c) the slope map for tiger T 7203. The area of the home range is 34.26 km2.
Table 1: Comparative assessment of CRW and CsCRW models for tigers T 7203 and T 10727. CRW CsCRW tiger random diff tiger random diff mean 1002.0 928.5 73.5 1144.0 898.6 245.4 T7203 median 1056.0 1028.0 28.0 908.0 684.5 223.5 mean 385.7 349.9 35.8 367.7 307.9 59.8 T10727 median 396.0 378.0 18.0 355.0 307.0 48.0
CRW mean difference is 73.5. For T 10727, the CsCRW mean difference is 59.8, while the CRW mean difference is 35.8. The CRW model shows a low variance (Figure 9) because it captures the relative uniformity in space use by the tiger towards the middle of its home range. More visits occur toward the center of the home range and away from the boundary due to the need to cross the middle of the home range when moving from one part of the home range to another. However, as seen on Figures 7(a) and 8(a) the modelpreprint does not reflect the lower space use and movement corridors used by the tigers in the rougher terrain of the home ranges (illustrated in Figure 4). In contrast, the CsCRW model has a much higher variance with a pronounced skew above the median (Figure 9). This can be attributed to the fact that the CsCRW model incorporates context (i.e. slope) into the choice of movement and better captures the corridors which tigers use more frequently. For instance, high visitation corridors, shown by the CsCRW model, in the eastern slopes of T 7203 (Figures 7) and in the northwestern part of the home range of T 10727 (Figure 8), coincide with the actual tiger movement track as seen in Figure 4. The more distinct corridors of modeled surfaces can also be observed in the more rugged portions of both home ranges. This
13 73
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016
Figure 8: Visitation counts for the home range of tiger T 10727 obtained from (a) CRW and (b) CsCRW models; and (c) the slope map for tiger T 10727. The area of the home range is 99.92 km2. is not observed in the CRW model for either tiger’s home range. It is important to note that due to the difference in the size of the two home ranges, and therefore the difference in the scales of two figures, the influence of micro-topography is seen more pronouncedly in Figure 7 than Figure 8.
5 Discussion
The underlying principle of this paper is that modeling trajectories using geometric metrics (e.g. path sinuosity, step length, turn angle, etc.) ignores the fact that a trajectory is a complex and composite signal of various behaviors that occur across multiple spatial and temporal scales. The goal of this paper is to demonstrate that local choices (i.e. decisions and behavior at local scales) play a major role in where a moving agent (e.g. tiger) is likely to go. Two stochastic models are developed and examined, which estimate the space use and movement by an individual. These models simulate agents to move through space using two algorithms, a correlated random walk (CRW) and a context-sensitive correlated random walk (CsCRW). The CRW results in apreprint space use that is more pronounced away from the boundary (in this case the tiger’s home range). It captures the global patterns of space use for an area by an individual. In contrast, the CsCRW model incorporates the contextual factors (in this case slope) which influence the strategies for local movements by an individual. The premise of our analysis is that local scale effects must be understood and modeled before regional and global scale behaviors can be deduced from a trajectory. For instance, a tiger that is patrolling its home range boundary (i.e. global scale behavior), still makes choices at local scales (i.e. following contour lines or moving along natural trails) to traverse its home range. There may be external biases driving
14 74
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016
CRW t7203 CsCRW t7203 visitation counts visitation counts 0 1000 2000 3000 0 1000 2000 3000
tiger random tiger random
CRW t10727 CsCRW t10727 visitation counts visitation counts 0 200 400 600 800 0 200 400 600 800
tiger random tiger random
Figure 9: Boxplots of visitation counts of CRW and CsCRW models for random trajectories and tigers T 7203 and T 10727 tiger’s global trajectory, but how the tiger reaches its goal-oriented destination is influenced by local choices. Therefore we make a distinction between local choice models (i.e. CsCRW) and biased CRW. The external biases constrain the possible paths to the goal but the local choices are largely independent of theses biases. Other approaches such as Step Selection Function (SSF) can also be used to understand the relationshipspreprint between local scale environmental correlates and movement choice based on available resources, however, their use has been confined to deterministic models for movement (Squires et al., 2013) in contrast to the stochastic approach we have taken in this research. A number of contextual factors play a role in local movement patterns and space use of an individual. The strength of the CsCRW model is that it can incorporate multiple contextual factors. In the case of the tiger this could include slope, vege- tation density, prey density or proximity to the boundary. Parametrization of these factors can be made using actual field observations and GPS locations. In fact, the strength of our approach is that the models are calibrated and validated through
15 75
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016 actual observations. In the case of the CsCRW model implemented in this research the parametrization is done using the tiger’s occurrence on different slope classes. A similar process could be used to parametrize other contextual information that con- tributes to tiger movement choices. In this study, we validated our model using two different trajectories of two tigers. These trajectories were different in terms of their geography (i.e. geometry of their home ranges), physiography (i.e. terrain type and topographic features), as well as the temporal resolution of their GPS observations. The comparative analysis of the two home ranges suggests that the proposed simula- tion model successfully incorporates the influence of geography and topography (i.e. context) on space use and local choices made by the tigers. Movement is driven by an individual’s state and the associated behaviors that occur at different spatial and temporal scales (Ahearn et al., 2001; Gautestad and Mysterud, 2005). The resulting trajectory of movement is therefore a complex com- posite (i.e. signal) that is influenced by geography through which individuals move (i.e. the network or the physiography of the terrain), their behavioral state (i.e. hungry, going to work, shopping, tourism, etc.) and their interaction with other individuals. The study of movement aims at understanding these components and the scales over which they occur (Dodge, 2016; Dodge et al., 2016). In this research we use a spatiotemporal stochastic simulation procedure that incorporates space and environmental context as two important components of the signal associated with a trajectory. We see context as a driver for the local choices that individuals may make, that are nested within more global scale objectives driven by the animals behavioral state. For instance a tiger patrols its home range at certain intervals, however the path it takes may be influenced by the characteristics of the terrain and vegetation. Thus the pattern of movement is a reflection of both of these factors at two very dif- ferent scales. In essence by modeling these contextual parameters we are separating them out from the patterns of movement which are attributed to the agents behaviors state.
6 Conclusion and Future Work
The main contribution of this study is to integrate environmental context in a spa- tiotemporal simulation model for movement, and to use real tracking observations to parameterize andpreprint validate the simulation. Three parameters (i.e. directional per- sistence, slope use, speed) were calibrated in this research from GPS tiger tracking data of two tigers. Two different models (CRW and CsCRW) were developed and evaluated for estimating probability of visitation (or space use) by a tiger within its home range. The CsCRW model shows the most promise for its ability to incorporate multiple contextual variables that influence movement choices at local scales in space and time. The spatiotemporal simulation proposed in this research incorporates the three important components of movement: space, time, and context as in the framework suggested by Dodge (2016). This research did not model the state and objectives of
16 76
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016 individuals and the resulting higher level behaviors (i.e. patrolling, hunting, tourism, shopping) which drive large scale movement patterns. Future extension of this method will incorporate these larger scale behaviors.
Acknowledgments
S. Dodge and G. Xavier work was supported under the 2015 UCCS Committee on Research and Creative Works (CRCW) award and the UCCS College of Letters, Arts, and Sciences Student-Faculty Research Awards. The authors wish to thank the reviewers for their insightful and constructive comments.
References
Ahearn, S. C., David, J. L., Joshi, A. R. and Ding, J. (2001), ‘TIGMOD: an individual-based spatially explicit model for simulating tiger/human interaction in multiple use forests’, Ecological Modelling 140, 81–97.
Ahearn, S. C. and Smith, J. (2005), Modeling the interaction between humans and animals in multiple use forests: a case study of Panthera tigris, in D. Maguire, M. Goodchild and M. Batty, eds, ‘GIS, Spatial Analysis, Modeling’, ESRI, Red- lands, California, chapter 18, pp. 387–403.
Ahearn, S. C., Smith, J., Simchareon, A., Simchareon, S. and Garcia, J. (2010), Modeling the relationship between patterns of movement of Panthera tigris and its behavioral states, in B. Gottfried, P. Laube, A. Klippel, N. Van de Weghe and R. Billen, eds, ‘the 1st Workshop on Movement Pattern Analysis, MPA’10’, Zurich Switzerland, pp. 143–146.
Batty, M., DeSyllas, J. and Huxbury, E. (2003), ‘The discrete dynamics of small-scale spatial events: agent-based models of mobility in carnivals and street parades’, International Journal of Geographical Information Science 17(7), 673 – 697. URL: http://discovery.ucl.ac.uk/242/
Benhamou, S. and Riotte-Lambert, L. (2012), ‘Beyond the utilization distribution: Identifying homepreprint range areas that are intensively exploited or repeatedly visited’, Ecological Modelling 227, 112–116. URL: http://dx.doi.org/10.1016/j.ecolmodel.2011.12.015
Buchin, K., Sijben, S., Arseneau, T. and Willems, E. P. (2012), ‘Detecting move- ment patterns using Brownian bridges’, Proceedings of the 20th International Conference on Advances in Geographic Information Systems pp. 119–128.
Codling, E. A., Plank, M. J. and Benhamou, S. (2008), ‘Random walk models in biology’, Journal of The Royal Society Interface 5(25), 813–834.
17 77
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016
David, O. and Perry, G. L. W. (2013), Spatial simulation: exploring pattern and process, Wiley-Blackwell. URL: http://onlinelibrary.wiley.com/book/10.1002/9781118527085
Demˇsar,U., Buchin, K., Cagnacci, F., Safi, K., Speckmann, B., Van de Weghe, N., Weiskopf, D. and Weibel, R. (2015), ‘Analysis and visualisation of movement: an interdisciplinary review’, Movement Ecology 3(1).
Dodge, S. (2016), From observation to prediction: The trajectory of movement re- search in GIScience, in H. Onsrud and W. Kuhn, eds, ‘Advancing Geographic Information Science: The Past and Next Twenty Years’, GSDI Association Press, chapter 9, pp. 123–136.
Dodge, S., Bohrer, G., Bildstein, K., Davidson, S. C., Weinzierl, R., Bechard, M. J., Barber, D., Kays, R., Brandes, D., Han, J. et al. (2014), ‘Environmental drivers of variability in the movement ecology of turkey vultures (cathartes aura) in north and south america’, Philosophical Transactions of the Royal Society of London B: Biological Sciences 369(1643), 20130195.
Dodge, S., Bohrer, G., Weinzierl, R., Davidson, S. C., Kays, R., Douglas, D., Cruz, S., Han, J., Brandes, D. and Wikelski, M. (2013), ‘The environmental-data au- tomated track annotation (Env-DATA) system: linking animal tracks with envi- ronmental data’, Movement Ecology 1(1), 3.
Dodge, S., Weibel, R., Ahearn, S. C., Buchin, M. and Miller, J. A. (2016), ‘Analysis of movement data’, International Journal of Geographical Information Science 30(5), 1–10. URL: http://dx.doi.org/10.1080/13658816.2015.1132424
Downs, J. A. and Horner, M. W. (2009), ‘A characteristic-hull based method for home range estimation’, Transactions in GIS 13(5-6), 527–537.
Duckham, M., Kulik, L., Worboys, M. and Galton, A. (2008), ‘Efficient generation of simple polygons for characterizing the shape of a set of points in the plane’, Pattern Recognition 41(10), 3224–3236. Forester, J. D., Ives,preprint A. R., Turner, M. G., Anderson, D. P., Fortin, D., Beyer, H. L., Smith, D. W. and Boyce, M. S. (2007), ‘State-space models link elk movement patterns to landscape characteristics in Yellowstone National Park’, Ecological Monographs 77(2), 285–299.
Franke, A., Caelli, T. and Hudson, R. J. (2004), ‘Analysis of movements and be- havior of caribou (Rangifer tarandus) using hidden Markov models’, Ecological Modelling 173(2-3), 259–270. URL: http://www.sciencedirect.com/science/article/pii/S0304380003003983
18 78
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016
Galton, A. (2005), Dynamic collectives and their collective dynamics, in ‘Spatial Information Theory, Lecture Notes in Computer Science’, Vol. 3693, Springer, pp. 300–315. URL: http://www.springerlink.com/index/514wxvd6fe70y9w4.pdf Gautestad, A. O. and Mysterud, I. (2005), ‘Intrinsic scaling complexity in animal dispersion and abundance’, The American naturalist 165(1), 44–55. Gautestad, A. O. and Mysterud, I. (2010), ‘The home range fractal: from random walk to memory-dependent space use’, Ecological Complexity 7(4), 458–470. Glasgow, M. L., Rudra, C. B., Yoo, E.-H., Demirbas, M., Merriman, J., Nayak, P., Crabtree-Ide, C., Szpiro, A. A., Rudra, A., Wactawski-Wende, J. and Others (2014), ‘Using smartphones to collect time–activity data for long-term personal- level air pollution exposure assessment’, Journal of Exposure Science and Envi- ronmental Epidemiology . Gonz´alez,M. C., Hidalgo, C. A. and Barab´asi,A.-L. (2008), ‘Understanding individ- ual human mobility patterns’, Nature 453(7196), 779–82. URL: http://www.ncbi.nlm.nih.gov/pubmed/18528393 H¨agerstrand,T. (1970), ‘What about people in Regional Science?’, Papers of the Regional Science Association 24(1), 6–21. Holyoak, M., Casagrandi, R., Nathan, R., Revilla, E. and Spiegel, O. (2008), ‘Trends and missing parts in the study of movement ecology’, Proceedings of the National Academy of Sciences of the United States of America 105(49), 19060–5. Horne, J. S., Garton, E. O., Krone, S. M. and Lewis, J. S. (2007), ‘Analyzing animal movements using Brownian bridges’, Ecology 88(9), 2354–2363. Hornsby, K. and Egenhofer, M. (2002), ‘Modeling moving objects over multiple gran- ularities’, Annals of Mathematics and Artificial Intelligence 36(1), 177–194. Jiang, B., Yin, J. and Zhao, S. (2009), ‘Characterizing the human mobility pattern in a large street network’, Physical Review E(80), 021136. Kie, J. G., Matthiopoulos,preprint J., Fieberg, J., Powell, R. A., Cagnacci, F., Mitchell, M. S., Gaillard, J.-M. and Moorcroft, P. R. (2010), ‘The home-range concept: are traditional estimators still relevant with modern telemetry technology?’, Philosophical Transactions of the Royal Society of London B: Biological Sciences 365(1550), 2221–2231. Kranstauber, B., Kays, R., Lapoint, S. D., Wikelski, M. and Safi, K. (2012), ‘A dynamic Brownian bridge movement model to estimate utilization distributions for heterogeneous animal movement’, Journal of Animal Ecology 81(4), 738–746. Laube, P. (2014), Computational movement analysis, Springer.
19 79
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016
Lu, Y. and Fang, T. (2014), ‘Examining personal air pollution exposure, intake, and health danger zone using time geography and 3D geovisualization’, ISPRS International Journal of Geo-Information 4(1), 32–46. URL: http://www.mdpi.com/2220-9964/4/1/32/
Miller, H. J. (1991), ‘Modeling accessibility using space-time prism concepts within geographical information systems’, International journal of geographical infor- mation systems 5(3), 287–301. URL: http://www.tandfonline.com/doi/abs/10.1080/02693799108927856
Miller, J. A. (2015), ‘Towards a better understanding of dynamic interaction metrics for wildlife: a null model approach’, Transactions in GIS 19(3), 342–361.
Mohr, C. O. (1947), ‘Table of equivalent populations of north american small mam- mals’, The American Midland Naturalist 37(1), 223–249.
Moorcroft, P. R., Lewis, M. A. and Crabtree, R. L. (2006), ‘Mechanistic home range models capture spatial patterns and dynamics of coyote territories in Yellow- stone’, Proceedings of the Royal Society – Biological Sciences 273(1594), 1651– 1659.
Morales, J. M., Fortin, D., Frair, J. L. and Merrill, E. H. (2005), ‘Adaptive models for large herbivore movements in heterogeneous landscapes’, Landscape Ecology 20(3), 301–316. URL: http://link.springer.com/10.1007/s10980-005-0061-9
Okabe, A., Boots, B. and Sugihara, K. (1992), Spatial Tessellations: Concepts and Applications of Voronoi Diagrams, John Wiley & Sons, Inc., New York, NY, USA.
Ovaskainen, O. (2004), ‘Habitat-specific movement parameters estimated using mark- recapture data and a diffusion model’, Ecology 85(1), 242–257.
Ovaskainen, O., Rekola, H., Meyke, E. and Arjas, E. (2008), ‘Bayesian methods for analyzing movements in heterogeneous landscapes from mark–recapture data’, Ecology 89(2),preprint 542–554. Powell, R. A. and Mitchell, M. S. (2012), ‘What is a home range?’, Journal of Mam- malogy 93(4), 948–958.
Rhee, I., Shin, M., Hong, S. and Lee, K. (2011), ‘On the levy-walk nature of human mobility’, IEEE/ACM Transactions on Networking (TON) 19(3), 630–643. URL: http://dl.acm.org/citation.cfm?id=2042974
Sang, S., O’Kelly, M. and Kwan, M.-P. (2011), ‘Examining commuting patterns: results from a journey-to-work model disaggregated by gender and occupation’, Urban Studies 48(5), 891–909.
20 80
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016
Schick, R. S., Loarie, S. R., Colchero, F., Best, B. D., Boustany, A., Conde, D. A., Halpin, P. N., Joppa, L. N., McClellan, C. M. and Clark, J. S. (2008), ‘Under- standing movement data and movement processes: current and emerging direc- tions.’, Ecology letters 11(12), 1338–50. URL: http://www.ncbi.nlm.nih.gov/pubmed/19046362 Song, Y. and Miller, H. J. (2014), ‘Simulating visit probability distributions within planar space-time prisms’, International Journal of Geographical Information Science 28(1), 104–125. URL: http://www.tandfonline.com/doi/abs/10.1080/13658816.2013.830308 Song, Y., Miller, H. J., Zhou, X. and Proffitt, D. (2015), ‘Modeling visit probabili- ties within network-time prisms using Markov techniques’, Geographical Analysis 48(1), 18?42. URL: http://doi.wiley.com/10.1111/gean.12076 Squires, J. R., DeCesare, N. J., Olson, L. E., Kolbe, J. A., Hebblewhite, M. and Parks, S. A. (2013), ‘Combining resource selection and movement behavior to predict corridors for Canada lynx at their southern range periphery’, Biological Conservation 157, 187–195. URL: http://dx.doi.org/10.1016/j.biocon.2012.07.018 Tang, W. and Bennett, D. (2010), ‘Agent-based modeling of animal movement: A review’, Geography Compass 4(7), 682–700. Technitis, G., Safi, K. and Weibel, R. (2014), ‘From A to B, randomly: An endpoint- to-endpoint random trajectory generator for animal movement’, International Journal Geographical Information Science 29(6), 912–934. Thurfjell, H., Ciuti, S. and Boyce, M. S. (2014), ‘Applications of step-selection func- tions in ecology and conservation’, Movement Ecology 2(1), 4. URL: http://www.movementecologyjournal.com/content/2/1/4 Torrens, P. M., Nara, A., Li, X., Zhu, H., Griffin, W. a. and Brown, S. B. (2012), ‘An extensible simulation environment and movement metrics for testing walking behavior in agent-based models’, Computers, Environment and Urban Systems 36(1), 1–17.preprint URL: http://dx.doi.org/10.1016/j.compenvurbsys.2011.07.005 Tribby, C. P., Miller, H. J., Brown, B. B., Werner, C. M. and Smith, K. R. (2016), ‘As- sessing built environment walkability using activity-space summary measures’, Journal of transport and land use 9(1), 187–207. Turchin, P. (1998), Quantitative analysis of movement: measuring and modeling popu- lation redistribution in animals and plants, Sinauer Associates, Sunderland, MA, USA. URL: http://www.getcited.org/pub/100319196
21 81
A context-sensitive correlated random walk Ahearn et al., IJGIS 2016
Winter, S. and Yin, Z.-C. (2010), ‘Directed movements in probabilistic time geog- raphy’, International Journal of Geographical Information Science 24(9), 1349– 1365. URL: http://www.tandfonline.com/doi/abs/10.1080/13658811003619150
Yuan, M. and Nara, A. (2015), Space-Time Integration in Geography and GIScience: Research Frontiers in the US and China, Springer Netherlands, Dordrecht, chap- ter Space-Time Analytics of Tracks for the Understanding of Patterns of Life, pp. 373–398.
preprint
22 82
Research Article
Tropical Conservation Science Volume 10: 1–7 Ecological Covariates at Kill Sites ! The Author(s) 2017 Reprints and permissions: Influence Tiger (Panthera tigris) sagepub.com/journalsPermissions.nav DOI: 10.1177/1940082917719000 Hunting Success in Huai Kha Khaeng journals.sagepub.com/home/trc Wildlife Sanctuary, Thailand
Somporn Pakpien1,2, Achara Simcharoen1, Somphot Duangchantrasiri1, Vijak Chimchome2, Nantachai Pongpattannurak2, and James L. D. Smith3
Abstract Despite significant knowledge of tiger ecology, information on hunting behavior is limited because tigers hunt in habitats where they are difficult to observe. From May 2013 to June 2015, we visited kill sites of eight female radio-collared tigers (Panthera tigris) to identify prey species of this species in Huai Kha Khaeng Wildlife Sanctuary, Thailand. At 150 kill sites, 11 mammalian species were identified from skeletal remains or hair samples. Sambar (Rusa unicolor), banteng (Bos javanicus), and gaur (Bos gaurus) composed 95.1% of tiger prey biomass. A subset of 87 kill sites was paired with 87 randomly selected sites within the home ranges of five of the eight radio-collared tigers to determine the influence of prey abundance and other ecological variables on hunting success. At each site, geomorphic and ecological covariates were sampled in 900 m2 square plots. A generalized linear model was used to investigate differences between kill sites and random sites. Mean relative prey abundance at kill sites was significantly lower than relative prey abundance at random sites (77.8 and 139.3 tracks/ha, respectively) indicating tigers did not kill in areas of higher relative prey abundance. Model selection was used to examine 12 landscape features that potentially influence kill site location. In the best model, low shrub cover and high crown cover were highly significant; tree density was included in this model but was not significant. This is the first study to demonstrate that kill location requires a combination of landscape features to first detect and then successfully stalk prey.
Keywords cluster locations, Huai Kha Khaeng Wildlife Sanctuary, hunting success, tiger kill site characteristics, tiger prey
Animals should seek habitat with adequate food, cover, second order habitat selection which focuses on where nest/den sites, or other resources critical for survival female tigers settle and the relationship of prey abun- (Manly, McDonald, Thomas, McDonald, & Erickson, dance to their territory size. These authors found an 2002). For female felids, sufficient food to raise young inverse relationship between female territory size and is often their primary resource need, and natural selection is expected to drive foraging decisions to optimize food 1 intake and minimize energy expenditure (Krebs & Department of National Parks, Wildlife and Plant Conservation, Thailand 2Faculty of Forestry, Kasetsart University, Thailand Davies, 1993). Food demands of female tigers increase 3Department of Fisheries, Wildlife, and Conservation Biology, University of rapidly as cubs mature and mothers continue to be the Minnesota, USA primary provider until their young are approximately Received 30 April 2017; Revised 8 June 2017; Accepted 9 June 2017 1.5 years old; at this time, male offspring are often larger than their mothers (Smith, McDougal, & Corresponding Author: Achara Simcharoen, Department of National Parks, Wildlife and Plant Miquelle, 1989). Johnson (1980) proposed a hierarchical Conservation, Protected Area Regional Office 12, 19/47 Kositai Road, model as a framework by which animals efficiently meet Nakhonsawan, Thailand. their resource needs. Simcharoen et al. (2014) studied Email: [email protected]
Creative Commons Non Commercial CC BY-NC: This article is distributed under the terms of the Creative Commons Attribution-NonCommercial 4.0 License (http://www.creativecommons.org/licenses/by-nc/4.0/) which permits non-commercial use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/open-access-at-sage). 83
2 Tropical Conservation Science the abundance of large prey. Within a female’s territory, (Figure 1) which is characterized by mixed deciduous, dry third and fourth order resource selection includes selec- dipterocarp, and dry evergreen forest. The average tion of denning, resting, and hunting sites. Here, we focus annual rainfall (1375 mm) is divided into a wet season on kill site selection. (May–October), with a mean of 1088 mm of rain and a Many studies have shown that high prey abundance dry season (November to April) with a mean of 298 mm. is the primary factor that predicts hunting success The temperature reaches 40 C in April at the end of the (Litvaitis, Sherburne, & Bissonette, 1986; Murray, dry season. The tiger is the largest carnivore in this eco- Boutin, O’Donoghue, & Nams, 1995; Palomares, system and its density ranges from 1.25 to 2.01 tigers/km2 Delibes, Revilla, Calzada, & Fedriani, 2001; Spong, (Duangchantrasiri et al., 2016). Phetdee (2000) identified 2002). Alternatively, Hebblewhite, Merrill, and 16 prey species consumed by tigers but the primary prey McDonald (2005) propose a landscape hypothesis that were animals 100 kg, represented by large ungulates predators prefer habitat where it is easier to kill their pri- including sambar (Rusa unicolor), banteng (Bos javani- mary prey. Following Hollings (1959), Hebblewhite et al. cus), gaur (B. gaurus), and water buffalo (Bubalus buba- simplify predation into the instantaneous probability of lis), which characterize 89.8% of tiger diet in this region. encounter followed by the conditional probability of suc- cessfully killing prey. This landscape hypothesis suggests Data Collection and Analysis that landscape features such as slope, ruggedness, and various aspects of horizontal cover first favor prey detec- Kill site data. Potential kill sites were identified using cluster tion and once prey is detected, catchability will be favored analysis of hourly location data from eight female satellite (Hebblewhite et al., 2005). Our study examines resource radio-collared tigers (Vectronic Aerospace GmbH, selection by hunting female tigers to evaluate the import- Germany; radio collaring was in accordance with the ance of prey abundance and landscape attributes that University of Minnesota IACUC protocol 0906A67489). If affect hunting success (Stephens & Krebs, 1986). we obtained >2 locations with consecutive movement dis- Once a prey animal is detected, felids typically tances < 100 m within 48 h, sites were identified as locations approach it using available vegetative cover (Elliott, where a tiger had potentially killed a prey animal (Figure 2; Cowan, & Hollings, 1977). Concealment allows them to Miller et al., 2013). We investigated these sites on foot and if hunt by ambushing prey by stalking and then chasing a kill was located, species, sex, and age class of prey were prey for a short distance (Kruuk, 1986; Caro & identified from skeletal material, hair, and hoofs (Lekagul & Fitzgibbon, 1992; Sunquist & Sunquist, 1989). Tigers McNeely, 1977; Phetdee, 2000). We used Phetdee’s (2000) (Panthera tigris) employ this strategy, stalking or pelage and skeletal size criteria to classify juvenile versus ambushing prey from cover (Schaller, 1967; Sunquist & adult. Gaur, banteng, and sambar were classified as adult Sunquist, 1989). Cover, however, conceals predators so when > 9 months old, and wild boar were classified as that hunting success is improved and also reduces prey adult when 6 months old. Kills were found at 150 sites but detection by predators (Balme, Hunter, & Slotow, 2007). only 87 of these, used by 5 female satellite radio-collared Thus, the landscape hypothesis must balance first detec- tigers, were investigated to study kill site characteristics. At tion of prey and then stalking success. Both of these fac- kill sites, we identified the actual kill site, which could be tors are components of Holling’s (1959) theoretical identified from the drag marks or the presence of the rumen, framework; predators should select habitats to maximize which is usually removed before the animal is dragged. Kill both aspects of hunting success (Hayward & Kerley, sites were compared with 87 randomly sampled sites 2005). The objectives of our study were as follows: (a) (excluding kill sites) from the home ranges of the five identify the main prey species consumed by tigers and female collared tigers. (b) determine relative importance of prey abundance and the ecological variables that influence hunting suc- Prey abundance data. To test the hypothesis that tigers kill cess of tigers in Huai Kha Khaeng Wildlife Sanctuary, prey in areas of high prey abundance, we assessed relative Thailand. prey abundance at both kill and random sites. At each site, we searched for tracks and dung of sambar, banteng, gaur, water buffalo, wild boar (Sus scrofa), and muntjac Method (Muntiacus muntjac) within four 10 -m radius subplots Study Area which were oriented in cardinal directions 30 m from the site center. An independent sample t-test was used to com- The study was conducted between May 2013 to June 2015 pare the relative prey abundance at kill and random sites. in Huai Kha Khaeng Wildlife Sanctuary, Thailand ( 15 310 N, 99 160 E) which is located in the eastern portion Ecological covariate data. To examine factors that influence of the Western Forest Complex (WEFCOM). The study kill site location, we chose 12 ecological covariates that was concentrated in the northern part of the Sanctuary we hypothesized might influence tiger hunting success. 84
Pakpien et al. 3
Figure 1. Map of Huai Kha Khaeng Wildlife Sanctuary showing the locations of 150 sites where tiger-killed prey were found and identified and 87 kill and 87 random sites where ecological correlate data were measured to investigate tiger kill site characteristics.
A30 30 m plot was placed at each kill and random site to (R Core Team, 2015). The importance of ecological vari- quantify eight of these ecological features. Vegetation struc- ables in the top performing models was assessed based on ture at these plots was characterized as number of shrubs, their respective z-values, the associated probability of each percent of crown cover, number of barrier features (e.g., variable’s beta coefficient, and the 94.5% confidence inter- fallen logs, lianas), bamboo clump density, basal area of val of the beta estimates. trees, tree density, visibility, and slope (Table 1). Visibility at both kill and random sites were measured as percentage visible of a 50 100-cm board placed at the center of each Results plot (Nudds, 1977). These variables were combined with Prey Species and Prey Abundance four geographic variables generated using ArcMap 9.3 (Esri, Redlands, CA) that were also hypothesized to influ- Eleven prey species were identified at 150 kill sites ence catchability; these were distances to permanent stream, (Table 2). Sambar, banteng, and gaur composed 95.1% distance to seasonal stream, distance to salt lick, and eleva- of biomass of these kills. Analysis of hair texture, teeth, tion. Collinearity between these 12 variables was assessed and skeletal remains indicated that tigers killed adult prey prior to analysis using Spearman’s rank correlation test. 67.6% of the time or three times more frequently than A set of generalized linear models with binomial error dis- juvenile prey (22.7%); we could not classify age of 9.7% tributions and a logit link function was used to evaluate the of kills. Mean relative abundance of prey sign at 87 kill ecological variables that best differentiate kills sites from sites (77.9 sign/ha, n ¼ 77, SD ¼ 72.3) was significantly random sites. The most likely model was selected using lower than at 87 randomly selected nonkill sites (139.2 Akaike’s information criteria (Burnham & Anderson, sign/ha, n ¼ 77, SD ¼ 112.3 sign; t ¼ 4.04, df ¼ 129.79, 2002). Statistical analyses were performed in R software p ¼ .00009). 85
4 Tropical Conservation Science
Ecological Covariates To identify the ecological variables that best predict kill site characteristics, we first examined colinearity and eliminated distance to salt licks, which was highly corre- lated with elevation (r ¼ .73) as well as distance to per- manent streams (r ¼ .87). The model with lowest Akaike information criteria included three variables: low shrub density, high crown cover, and low tree density (Table 3). The deviation of z-values from zero and their associated p values indicated the strength of each variable in the top two models. In the top model, the strongest deviation of z from zero was shrub cover (z ¼ 2.813, p > jzj¼.0049).
Table 2. The Number of Tiger’s Kills and Biomass of Species in 150 Kills in Huai Kha Khaeng Wildlife Sanctuary, Thailand.
Species No. kills Weight Biomass (%)
Sambar 76 212 55.52 Banteng 27 287 26.70 Gaur 13 287 12.86 Wild boar 18 37 2.29 Elephant 3 200 2.07 Porcupine 5 8 0.14 Muntjac 3 20 0.21 Hog badger 2 10 0.07 Serow 1 30 0.10 Figure 2. (a) We found this sambar kill by visiting sites where a Pangolin 1 3 0.01 tiger was located three or more times in consecutive 1 hr GPS fixes. (b) We occasionally placed camera traps at kill sites to obtain Langur 1 9 0.03 additional information on a tiger’s condition and, for females, their Note. Mean weight of animals killed by tiger are from Karanth and Sunquist reproductive status (photo: Thailand Tiger Project). (1995).
Table 1. Description of the 12 Ecological Covariates That Were Used for Comparison of Kill and Random Sites.
Potential impact on Name of variable Unit Description kill site selection
Tree density N/ha Total number of trees diameter > 4.5 cm in 30 30 plot Detection of prey Basal area of trees BA/ha Total basal area of trees diameter > 4.5 cm in 30 30 plot Detection of prey Shrub cover N/ha Total number of shrubs height 30–100 cm in 30 30 plot Detection of prey Bamboo N/ha Total number of bamboo clumps in 30 30 plot Impacts prey escape Barrier feature N/ha Total of fallen trees and climbers in 30 30 plot Impacts prey escape Crown cover % Mean percentage of crown cover in 30 30 plot Shade impacts prey vigilance using convex spherical densitometer Visibility % Mean percentage of visibility in each cardinal direction (30 m) Detection of prey Slope % Mean of slope measured by clinometer in cardinal direction Impacts hunting success Elevation m. Generated using ArcMap 9.3 Impacts hunting success Distance to permanent m Generated using ArcMap 9.3 Impacts prey abundance stream Distance to seasonal stream m Generated using ArcMap 9.3 Impacts prey abundance Distance to salt lick m Generated using ArcMap 9.3 Impacts prey abundance
Note. We identify potential impact of each variable on kill site selection. 86
Pakpien et al. 5
Table 3. Summary Statistics for the Top Four Models of Kill Site We initially thought that high prey abundance would Characteristics With an Accumulative Weight of 0.90 for Tigers in be a good predictor of kill sites because several studies on Huai Kha Khaeng Wildlife Sanctuary, Thailand. kill site habitat selection by large carnivores support the Independent Delta_ AICc Cum. hypothesis that kill site location within an animal’s home variables K AICc AICc Wt Wt LL range is largely influenced by prey abundance (Davidson et al., 2012). Furthermore, a strategy to hunt in areas of cvþtrþsh 4 224.64 0.00 0.52 0.52 108.20 high prey abundance, especially by adult females that cvþtrþshþbr 5 226.21 1.57 0.24 0.76 107.92 need to meet the energetic demands of feeding their off- cvþsh 3 228.04 3.41 0.09 0.85 110.95 spring, should optimize energy gained at the lowest risk cvþshþbr 4 229.21 4.58 0.05 0.90 110.49 cost (Heurich et al., 2016). Our study, however, did not support the prey abundance-hypothesis that killing suc- Note. AIC: Akaike information criteria; cv: Crown cover; tr: Tree density; sh: Shrub cover; br: Barrier feature. Resource use was estimated from tiger kill cess, and thus energy, are maximized by hunting in areas sites and resource availability was estimated from random locations that of high prey abundance. On the contrary, we found that were chosen from tiger locations within their home range. kill sites had a significantly lower prey abundance than random sites located along a tiger’s route of travel. We do not know the extent to which prey may have avoided kill The next most important variable was crown cover sites, but found no literature indicating prey shift their (z ¼ 2.721, p > jzj¼.0065). There was only weak support range. for the third variable, tree density (Table 3). The second Thus, our findings led us to evaluate an alternative set best model included a fourth ecological correlate, bar- of landscape hypotheses that certain habitat attributes rier cover, but its z-value was not significant (>0.05). are more important to killing success than prey abun- The third- and fourth-ranked models were subsets dance. Several previous studies also support landscape of Models 1 and 2 and garnered weights of 9% and hypotheses that carnivores select habitats where prey 5%. These top four models had an accumulative weight are more susceptible to predation (Balme et al., 2007; of 90%). Belotti et al., 2013; Davidson et al., 2012; Hopcraft, Sinclair, & Packer, 2005). Of the 12 ecological correlates we examined to explain Discussion tiger hunting location, low cover was the most important Identification of Prey at Kill Sites variable in all of the top-ranked models. This was surpris- ing because tigers, lions (Panthera leo), and other felids We identified only 11 mammalian prey species at kill sites favor stalking to within a short distance followed by a as compared with 16 species reported in an analysis of relatively short chase. Thus, adequate cover is essential. scats (Phetdee, 2000) from the same area. Larger prey in Lions and tigers accelerate faster than many of their prey, our study (sambar, gaur, and banteng) composed 95.1% but their top speed peaks much below that of their prey of the biomass of kills we identified, which was higher so they must initiate an attack at a close distance (Elliott than the 88% biomass of these prey reported by Phetdee et al., 1977). Furthermore, their large, muscular body (2000). It is not surprising that prey identified from makes it energetically costly to maintain high speed kill sites are biased toward larger prey species because over a long chase. we identified kill sites by noting a sequence of clumped However, large cats must first detect their prey, thus 1-hr interval locations. Small kills could be processed they need an optimal combination of habitat structure to before we noticed a clump. Also we visited kill sites first locate, and then successfully stalk and ambush their a mean of 8 days after a kill and the scattering of prey (Lamprecht, 1978; Murphy & Ruth, 2010). To small kills made them more difficult to find. However, understand the role of different ecological covariates in identifying smaller prey is less important to under- large carnivore hunting success, Hebblewhite et al. (2005) standing kill site selection because, as shown by scat suggest using Holling’s (1959) theoretical framework to surveys, these animals compose 8.5% of the tiger’s diet decompose hunting success into two components: first, Phetdee (2000). the instantaneous probability of encounter followed by Three of our 87 kills were elephants (Elephas max- the conditional probability of a successful stalk leading imus) < 1 year in age, which is an observation not to a kill. In this context, less cover at a site with generally reported in the previous scat study (Phetdee, 2000). We high cover would be favored to increase the initial prob- speculate that young elephants were not found in the past ability of detection. The median cover at kill sites was because, prior to the mid-1990s, elephant poaching was 10,410 shrubs/ha. Given that kills are made at sites with widespread. With improved management beginning in considerably less cover than random sites (med- the 1990s, elephant numbers and recruitment have ian ¼ 14,190 shrubs/ha), areas with lower cover have ade- increased (Kanchanasaka, 2010). quate cover for tigers to successfully hunt. 87
6 Tropical Conservation Science
Table 4. The Beta Coefficients for the Best Model Which Included Low Shrub Cover, High Crown Cover and Low Tree Density.
Estimate coefficient Std. error z value Pr(>jzj) 95% CI