GIANT PANDA BEHAVIOR ACROSS a COUPLED HUMAN and NATURAL SYSTEM by Vanessa Hull a DISSERTATION Submitted to Michigan State Univer

GIANT PANDA BEHAVIOR ACROSS A COUPLED HUMAN AND NATURAL SYSTEM

Vanessa Hull

A DISSERTATION

Submitted to Michigan State University in partial fulfillment of the requirements for the degree of

Fisheries and Wildlife – Doctor of Philosophy

2014

ABSTRACT

GIANT PANDA BEHAVIOR ACROSS A COUPLED HUMAN AND NATURAL SYSTEM

Vanessa Hull

Animals interact with their environments in complex ways across space and time. These interactions are in turn influenced by human activities that take place across dynamic coupled human and natural systems. Understanding such phenomena is important for the management of endangered species, which face increasing threats from human influences on fragile landscapes.

This dissertation presents investigations of the behavior and ecology of the endangered giant panda (Ailuropoda melanoleuca), an elusive species whose behavior is not fully understood. The objective of this work is to better understand giant panda space use and habitat selection across a coupled human and natural system and in turn provide recommendations for management of pandas and their habitat for the future.

My research team used global positioning system (GPS) collars to track individual giant pandas in Wolong Nature Reserve, Sichuan, China. We integrate these data with other diverse sources including field surveys, geographic information systems (GIS), and remotely sensed imagery to address a number of key questions relevant to giant panda ecology. These include (a) identifying key patterns and complexities in existing giant panda habitat selection studies

(Chapter 2); (b) exploring giant panda space use using novel model-based approaches (Chapter

3); and (c) investigating habitat use and selection patterns by individual pandas across continuous space (Chapter 4). We then build on this knowledge to explore emerging issues in management of giant pandas and their habitat, including (a) the efficacy of zoning designations for spatially segregating giant panda conservation and human development (Chapter 5) and (b)

the impact of livestock grazing, an increasing but understudied threat to giant pandas and their habitat (Chapter 6).

By synthesizing previous literature, we found evidence of interactive effects of different habitat characteristics on panda habitat selection (e.g. slope and forest disturbance), variation in selection across different selection levels (e.g. geographic range vs. home range), and differences in habitat use based on habitat availability (e.g. declining use with increasing availability of secondary forests, Chapter 2). We also found that pandas occupied small home ranges (2.8 – 6 km2) made up of several small core areas and displayed significant dynamic spatio-temporal interactions with neighboring individuals (Chapter 3). Pandas used a broader range of habitat characteristics than previously understood, such as steep slopes and non-forest areas, while solar radiation was a significant predictor of both habitat use and selection (Chapter 4). Zoning designations designed to spatially segregate pandas and humans had mixed effects, being successful at containing infrastructural development but not all human activities, while also leaving around 50% of suitable panda habitat in vulnerable zones (Chapter 5). Domestic, free- ranging horses had a measurable negative impact on giant pandas and their habitat by overlapping spatially with suitable giant panda habitat, displaying similar habitat selection patterns as pandas, and consuming large amounts of bamboo (Chapter 6). Our results provide a novel perspective on giant panda ecology and conservation at the individual animal level, information which can inform efforts to manage the increasingly degraded habitat to promote long-term sustainability of the species. The integrative model-based approaches used throughout the dissertation also provide a potential framework for studies on other species around the globe facing similar human pressures across coupled human and natural systems.

This dissertation is dedicated to all of my friends in Wolong For graciously inviting me into their homes and their lives

ACKNOWLEDGMENTS

This dissertation would not have been possible without the assistance of many wonderful people. All of the work presented is a result of a collaborative team effort involving a number of dedicated individuals, principally involving a partnership between the Center for Systems

Integration and Sustainability (CSIS) at Michigan State University, the Research Center for Eco-

Environmental Sciences in the Chinese Academy of Sciences, and the China Conservation and

Research Center for the Giant Panda (CCRCGP) in Wolong Nature Reserve, China.

I would first like to thank my major advisor, Dr. Jianguo Liu, for being an incredible mentor and for giving me the opportunity to fulfill my childhood dream of working with giant pandas. He took a leap of faith in sending a 22-year-old American woman he had only briefly met over to the remote Wolong Nature Reserve in China and in doing so changed the trajectory of my life. From my first hike up the mountain with him in Wolong 10 years ago to the final preparations of this document and everything in between, I have learned so much and will carry each lesson forward in my career and life. I have especially appreciated his unwavering confidence in me, his attention to detail, his broad vision, and his commitment to excellence.

I would also like to thank my other committee members- Dr. Gary Roloff, Dr. Ashton

Shortridge, and Dr. Andrés Viña for their guidance, support, patience, and thoughtful insights at each step of this research. They each helped to create an open and warm atmosphere in which to engage in discussion and debate throughout our many formal and informal meetings, providing me with the tools I needed to complete this work. I also wish to acknowledge Dr. William Taylor for being a generous mentor, for providing unique opportunities for me to grow, and for inviting me on many fishing trips. I thank Dr. Scott Bearer for thoughtfully paving the way for me to

vi conduct ecological field work in Wolong and providing so much guidance on both research techniques and life in the Reserve. I thank Dr. Anita Morzillo for her friendship and for serving as an invaluable mentor for me as a woman and as a scientist. I also thank the rest of the members of the Center for Systems Integration and Sustainability (CSIS) for creating a warm and intellectually stimulating working environment, especially Neil Carter, Xiaodong Chen,

Guangming He, Shuxin Li, Yu Li, Junyan Luo, Wei Liu, Bill McConnell, Sue Nichols, Nils

Peterson, Nicholas Reo, Yin-Phan Tsang, Mao-Ning Tuanmu, and Wu Yang.

I owe an enormous debt of gratitude to so many friends, colleagues, and mentors over in

China. I would like to especially thank Dr. Zhiyun Ouyang, deputy director of the Research

Center for Eco-Environmental Science (RCEES), Chinese Academy of Sciences, for providing invaluable administrative, logistical, and scientific support throughout all of my time in China. I admire him for being among the most influential figures in conservation in China, a challenge he handles with confident grace. I also thank Dr. Weihua Xu from the RCEES for his expertise and guidance over the years, for his friendship and thoughtful words of encouragement, and for always welcoming me in Beijing.

I also am incredibly thankful and lucky to have been paired with Dr. Jindong Zhang from the RCEES as a partner with whom to conduct our respective dissertations. Jindong and I have shared so much over the years, including cold days by the fire at Wuyipeng, tandem rides on a motorcycle on the streets of Wolong with our GPS tracking equipment in tow, and long hikes in the remote woods in the rain, snow, and mud. We would never have captured our study animals nor gotten this research off the ground if it were not for the diligence of Jindong. He also provided a special talent for negotiating complex local relationships, a drive to seek out new data, and confidence and persistence to analyze and write-up our findings.

vii

Wolong Nature Reserve became a home away from home for me over the course of my numerous months-long trips to the field and that is because of the welcoming spirit and kindness of people living and working there. I thank reserve director Dr. Hemin Zhang for his guidance and support of my work and career over the years, his congeniality and openness via invitations for dinners and meetings, and his strong leadership in taking on the task of leading a globally important nature reserve through challenging times. I also thank other leaders from the China

Research and Conservation Center for the Giant Panda (CCRCGP), including Desheng Li,

Xiaoping Zhou, Rongping Wei, Chunxiang Tang, Yan Huang, and Guiquan Zhang for their administrative and logistical support and for being so welcoming towards me over the years.

Special thanks to Chunxiang Tang for assistance with veterinary supplies and personnel and Yan

Huang for providing generous field support and for his commitment to exchanging scientific ideas.

I would like to especially thank the members of the research lab at CCRCGP for their commitment to hard work, for their guidance, and for their friendship. I am grateful for my mentors Jinyan Huang and Shiqiang Zhou who both played integral roles as partners in this research. I wish to acknowledge Jinyan Huang for his strong leadership, his extensive knowledge on plant ecology, and his patience. I also admire Shiqiang Zhou for his limitless energy, his passion for field work, and his strong work ethic. I am indebted to Rengui Li and Dian Liu for their eagerness to participate in our research and their commitment to being team players. I thank

CCRCGP veterinarians Yanxi Cheng, Linhua Deng, and Caiwu Li for their extensive help with providing veterinary care. I also thank the entire team of CCRCGP staff and researchers involved in the panda reintroduction program who generously shared their time, resources, and friendship

viii while in the field. I also thank other Reserve administrators for their ongoing support and guidance, including Jian Yang.

This study would not have been possible without the help of numerous local people who diligently assisted with field work. I sincerely thank the honorary postdocs, Shumin Fan and

Wenbin Yang, for putting in years of dedicated and arduous field work. Shumin Fan developed a habit of grabbing me at just the right moment to prevent me from falling down the mountainside and Wenbin Yang's jovial spirit always kept things light and fun in even the most challenging conditions. I would like to extend a special thank you to Youfu Wang for always taking care of me- for cooking hundreds of meals for me, for taking care of the field station under difficult conditions, for lots of talks and laughter, and for teaching me how to play mahjong (and occasionally letting me win). I also thank Shumin Fan, Wenbin Yang, and Youfu Wang and their spouses and extended families for graciously inviting me into their homes and allowing me to become a part of their families year after year, especially on the Chinese New Year holiday. I have them to thank for a lifetime of memories of Tibetan dancing by the fire, decadent meals eaten in warm farmhouses, and firecrackers ignited on the rooftops. I also thank Qunyin Wang for her leadership in orchestrating the transport of countless pounds of equipment up to our field station, for her role as temporary caretaker at the field station, and for her daily demonstration of quiet strength that I came to admire. I also thank the many others who participated in the demanding tasks of transporting field equipment, building animal traps, and capturing our study animals.

I am also indebted to a number of generous funding sources that made this work possible.

These include- the National Science Foundation (NSF) including the Graduate Research

Fellowship program (GRFP), the National Aeronautics and Space Administration (NASA) Earth

ix and Space Science Fellowship (NESSF) program, the Michigan State University Distinguished

Fellowship Program, the William W. and Evelyn M. Taylor International Engagement Program, the Rocky Mountain Goat Fellowship, the International Association for Bear Research and

Management Research and Conservation Grant, the National Natural Science Foundation of

China (40901289), the State Key Laboratory of Urban and Regional Ecology, the Research

Center for Eco-Environmental Sciences, Chinese Academy of Sciences (SKLURE2008-1), and the Giant Panda International Collaboration Fund (Grant SD0681).

Furthermore, I thank my friends for their ongoing support and guidance. I extend a special thank you to my best friend Michele Zager for the long talks, her non-judgmental ear, her exquisite artwork, and her unwavering commitment to our lifelong sisterhood. I am also grateful for the presence of the late Wilhelmina Yonkman in my life, who taught me so much about friendship, generosity, and family, and for her son and daughter-in-law Leonard and Alice

Yonkman for their prayers, friendship, and ongoing support. Thanks also goes to my friend Jiang

Ming for cheering me up on rough days in Wolong and to my Chinese family in Beijing for providing me with a warm respite from the field and for teaching me about Chinese culture. I acknowledge my Kenya forever friends Jenny Chipault and Maluszka Slabinski-Schmidt for sharing their thoughts, laughs, and guidance over the years and Marisa Rinkus and Amy Damrow for their camaraderie and for sharing their academic ideas, advice, and aspirations with me. I also sincerely thank Saori Shiokawa and Amanda Gutowski for providing loving and attentive childcare for my son so that I could complete this dissertation.

I am also immensely grateful for the support of my family. I thank my parents-in-law

Yuzo and Ayako Fujimoto for their kindness and for welcoming me into their family, and my brother-in-law and sister-in-law Seiji and Yuko Fujimoto for their support and kinship. I thank

x my three brothers Cameron, Justin, and Patrick Hull for not giving me a hard time as the sole girl in the family and for humoring me with endless panda gifts at Christmas. I sincerely thank my parents Diane and Edward Hull, who planted the seeds for this research long ago by giving me my first giant panda stuffed animal at age 6 and for supporting my young passion for the animal year after year until I had more pandas in my bedroom in high school than probably exist in the wild. They have always encouraged me and believed in me and never worried too much about my many trips to remote areas of China, knowing that this would make me happy and fulfilled.

I also thank my husband Masanori Fujimoto for giving me a ride in the cold on a snowy day in Michigan many years ago, for lending me his lucky fishing lure to help me catch the pandas, and for his endless patience during my many long trips away from him. His big heart and big laugh have kept me going through all of the ups and downs of this project. I thank my young son Haruyoshi for bringing a new and warm season of spring into my life and for his sweet disposition and patience while I finished this work. I hope to one day take him to Wolong so that he can appreciate the rich experience that I have had thanks to the contributions from so many.

TABLE OF CONTENTS

LIST OF TABLES ...... xv

LIST OF FIGURES ...... xvi

CHAPTER 1 INTRODUCTION ...... 1 Background ...... 2 The Big Picture ...... 2 The Giant Panda- a Global Icon for Conservation ...... 3 Threat of Humans ...... 5 Current Research Gaps ...... 8 Dissertation Overview ...... 9 A New Opportunity ...... 9 Study Area ...... 10 Data Collection ...... 13 Dissertation Outline ...... 14 APPENDIX ...... 17

CHAPTER 2 A SYNTHESIS OF GIANT PANDA HABITAT SELECTION ...... 22 Abstract ...... 23 Introduction ...... 24 Methods...... 27 Results ...... 29 Scope of Habitat Selection Studies on the Giant Panda ...... 29 Single Geophysical Factors ...... 30 Single Vegetation Factors ...... 31 Single Disturbance Factors ...... 32 Multivariate Habitat Selection Factors ...... 34 Interactive Habitat Selection Factors ...... 35 Habitat Selection across Selection Levels ...... 36 Discussion ...... 37 Implications for Giant Panda Ecology and Management ...... 37 Future Directions ...... 42 APPENDIX ...... 44

CHAPTER 3 SPACE USE BY ENDANGERED GIANT PANDAS ...... 54 Abstract ...... 55 Introduction ...... 56 Materials and Methods ...... 58

xii

Study Area and Study Animals ...... 58 Giant Panda Home Range Estimation ...... 59 Core Area Estimation ...... 60 Spatial Interactions among Pandas ...... 61 Results ...... 61 Giant Panda Home Range Estimation ...... 61 Core Area Estimation ...... 62 Spatial Interactions among Pandas ...... 62 Discussion ...... 63 APPENDIX ...... 66

CHAPTER 4 AN INDIVIDUAL ANIMAL PERSPECTIVE ON GIANT PANDA HABITAT USE AND SELECTION ...... 74 Abstract ...... 75 Introduction ...... 76 Methods...... 78 Study Area and Panda Subjects ...... 78 Estimating Panda Use ...... 80 Habitat Characteristics ...... 81 Habitat Use Modeling ...... 82 Extensions to Habitat Suitability and Habitat Selection Analyses ...... 84 Results ...... 85 Discussion ...... 86 APPENDIX ...... 91

CHAPER 5 EVALUATING THE EFFICACY OF ZONING DESIGNATIONS FOR PROTECTED AREA MANAGEMENT ...... 98 Abstract ...... 99 Introduction ...... 100 Materials and Methods ...... 104 Study Area ...... 104 Zoning and Pandas ...... 107 Zoning and Human Impacts ...... 110 Proposed Zoning Revisions for Panda Conservation...... 112 Results ...... 113 Zoning and Pandas ...... 113 Zoning and Human Impacts ...... 115 Proposed Zoning Revisions for Panda Conservation...... 117 Discussion ...... 118 Efficacy of Zoning in Wolong ...... 118 Zoning as a Conservation Tool ...... 122 APPENDIX ...... 125

xiii

CHAPTER 6 IMPACT OF LIVESTOCK ON GIANT PANDAS AND THEIR HABITAT ...... 132 Abstract ...... 133 Introduction ...... 134 Methods...... 136 Study Area ...... 136 Study Subjects ...... 137 Data Collection and Analysis...... 139 Results ...... 144 Spatial Distribution of Horses ...... 144 Comparison of Space Use and Habitat Selection of Horses and Pandas ...... 145 Impact of Horses on Bamboo and Pandas ...... 146 Discussion ...... 147 APPENDIX ...... 152

CHAPTER 7 CONCLUSIONS ...... 162

REFERENCES ...... 168

xiv

LIST OF TABLES

Table 2.1. Summary of studies on giant panda habitat selection reviewed from the English and Chinese language literature...... 49

Table 2.2. Summary of giant panda habitat selection with respect to various geophysical, vegetation, and disturbance factors from 23 studies reviewed in the English and Chinese language literature (listed by mountain range). An “x” indicates the factor was a significant correlate of habitat selection; an “o” indicates the factor was not significant, and a blank space means that the factor was not considered in the study...... 51

Table 2.3. Habitat characteristics deemed important to giant panda selection by habitat selection level (sensu Johnson (1980))...... 53

Table 3.1. Summary of 5 giant pandas tracked with GPS collars in Wolong Nature Reserve, China...... 71

Table 3.2. Home ranges and core areas for 5 giant pandas monitored using GPS collars in Wolong Nature Reserve, China. Home ranges are based on 95% isopleths. Pan Pan and Long Long were monitored for 7 and 6 months, respectively, and the remaining pandas were monitored for 1 year. Core areas were estimated using the core area estimation method outlined in Vander Wal and Rodgers (2012). Surface area correction was performed based on overlay with a digital elevation model...... 72

Table 3.3. Minta’s (1992) test for dynamic space use interactions among the 5 giant pandas monitored. LAA and LBB represent the relationship between observed and expected use of the area shared by the 2 animals (area of overlap of their respective home ranges) by each individual in the pair, where A is the first animal and B is the second animal listed. Lixn is a measure of simultaneous use of the shared area by the animals (ratio of simultaneous use and avoidance to solitary use and avoidance). For all coefficients, values <0 represent avoidance and values >0 represent attraction (in bold)...... 73

Table 4.1. Summary of study pandas and GPS collar performance over the one year period included in this study...... 96

Table 4.2. Contribution of habitat variables to predicting giant panda habitat use across their utilization distributions (n=5)...... 97

Table 6.1. Summary of four domestic, free-ranging horse herds monitored in giant panda habitat in Wolong Nature Reserve, China...... 160

Table 6.2. Results of the randomization test for the k-select analysis on habitat selection by pandas and horse herds. Tests were based on 10,000 randomization steps, at which the first eigenvalues of observed data were compared to those from randomized datasets. Marginality vectors for each variable represent differences between mean used and mean available habitats...... 161

LIST OF FIGURES

Figure 1.1. A giant panda. Photo credit: Bin Liu, 2005...... 18

Figure 1.2. Distribution of giant pandas over time. Pre-historic estimates are derived from a map in Loucks et al. (2001) but adjusted for additional fossil evidence detailed in Jin et al. (2007). Estimates for 1980 and 1990 are derived from a map in Reid and Gong (1999) that summarizes data in Zhu and Long (1983). Present estimates are derived from a supervised habitat classification by Viña et al. (2010)...... 19

Figure 1.3. The Hetaoping study area for wild giant panda GPS collar research. Elevation is derived from a digital elevation model acquired by the National Aeronautics and Space Administration’s (NASA) Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER)...... 20

Figure 1.4. Broad overview of this dissertation...... 21

Figure 2.1. Giant panda selection for slope by slope class. Values shown are Vanderploeg and Scavia relative electivity indices (Vanderploeg and Scavia 1979) divided into studies showing (a) quadratic and (b) linear and nonlinear decreasing trends. Slopes were estimated in mid-sized field plots (10 x 4 m2, 20 x 2 m2, 20 x 20 m2, or 30 x 30 m2)...... 45

Figure 2.2. Giant panda habitat use in relationship to habitat availability for (a) topographic slope, (b) bamboo cover, and (c) secondary forest. Asterisks in (c) represent significant differences at the P = 0.05 level (determined via χ2 goodness of fit tests on the distribution of used versus available habitats (Neu et al. 1974))...... 46

Figure 2.3. Giant panda selection for bamboo cover across 6 studies. Bamboo cover was measured using visual estimation in fixed area 20 x 20 m or 30 x 30 m plots...... 48

Figure 3.1. Study area for GPS collar tracking of giant pandas in Wolong Nature Reserve, China. We derived elevations from a digital elevation model obtained from The Advanced Spaceborne Thermal Emission and Reflection Radiometer...... 67

Figure 3.2. Giant panda 95% home ranges estimated using the biased random bridge model. ... 68

Figure 3.3. Space use within the home ranges for 5 giant pandas estimated using the biased random bridge model...... 69

Figure 3.4. Space use interactions among 3 giant pandas. (a) The HR index represents the percent of the first animal’s utilization distribution overlapping with the second animal’s utilization distribution at a given isopleth. (b) All instances of close distances between individuals (<200 m) recorded at simultaneous time points throughout the time period of the study...... 70

xvi

Figure 4.1. Hetaoping study area for giant panda GPS collar research in Wolong Nature Reserve, China. Forest layer is derived from supervised classification of Landsat TM imagery (Viña et al. 2011)...... 92

Figure 4.2. Fit of simultaneous autoregressive (SAR) models depicted by actual versus predicted log-transformed response variables (utilization distributions) for each panda...... 93

Figure 4.3. Proportion of GPS collared panda's utilization distributions in different habitat suitability classes for (a) elevation, (b) forest, and (c) slope. Classes were derived from Liu et al. (1999)...... 94

Figure 4.4. Habitat selection by GPS-collared giant pandas for various biogeophysical characteristics at two selection levels- within the home range level (top row) and at the home range level (bottom row). Habitat use of each class was calculated as the proportion of the utilization distribution. Habitat availability was calculated as the proportion of habitat available within individual home ranges (top row) and within the entire study area (bottom row). Lines below each plot represent significant differences in selection across levels determined via randomization tests...... 95

Figure 5.1. Zoning designations in Wolong Nature Reserve, Sichuan, China in 1998 (original) and 2009 (most recent). The core zone is designated as an area where the main priority is biodiversity conservation. No human activities are permitted in the core zone and limited human activity is allowed in the buffer zone, while human activities (including infrastructure development) are permitted in the experimental zone...... 126

Figure 5.2. Distribution of giant panda habitat suitability classes across core, buffer, and experimental zones in Wolong Nature Reserve in 1997 (year before zoning designation). Habitat suitability was derived from the criteria established in Liu et al. (1999) and reported in Liu et al. (2001), which considers panda habitat as a combination of suitable slopes, elevations (both derived from a DEM) and forest cover (derived from Landsat imagery). Distribution is shown for (a) the entire reserve and (b) only the portions of the reserve within the giant panda’s elevational range (2,000-3,300 m)...... 127

Figure 5.3. Distribution of wild giant panda signs obtained in Wolong Nature Reserve (n= 487 signs) as part of the 2000-2004 National Giant Panda Census in relationship to management zones (core, buffer, and experimental). The panda distribution area was estimated using a 95% kernel (h= 1000). Also shown is the distribution of 3 horse herds monitored. Inset map shows a GPS collar study on 2 wild giant panda females (Mei Mei and Pan Pan) and one of the horse herds...... 128

Figure 5.4. Percentage of houses, roads, tourism facilities, and livestock in each management zone (core, buffer, and experimental) in Wolong Nature Reserve. House locations (n= 1060) were measured with GPS units in 2002, roads were obtained from government documents, tourism facilities (n=19) were recorded with GPS units in 2006, and livestock (three herds of horses only) were monitored using GPS collars and field sampling. . 129

Figure 5.5. Forest cover across zoning designations in Wolong Nature Reserve in 1974 (year before reserve establishment), 1997 (year before zoning designations), and 2007. Forest

xvii cover was derived from Landsat TM imagery and analyzed with respect to areal coverage in each zone. Error bars on the forest classification of the 1974 and 1997 images represent the area taken up by “unclassified” areas (areas with excessive clouds) that could have either been forest or non-forest...... 130

Figure 5.6. Proposed focal areas recommended to be considered for zoning revisions in Wolong Nature Reserve for improved giant panda conservation. Zoning designations (core, buffer, and experimental) from the most recent version (2009) are presented along with focal experimental zones (areas of experimental zone that should be considered for conversion to buffer and/or core zone) and focal buffer zones (areas of buffer zone that should be considered for partial or full conversion to core zone). Both focal zones represent areas that support giant pandas and are also outside of existing human establishments. Letters represent focal experimental (A, B, C) and buffer (D, E, F) zones of particular importance that are recommended for revision to better protect the panda population...... 131

Figure 6.1. Distribution of four domestic, free-ranging horses and pandas studied in Wolong Nature Reserve, China. The polygons represent the home ranges derived using the biased random bridge model (Benhamou 2011) with the exception of the Fangzipeng herd in which the MCP method was used on transect data due to lack of available GPS collar data. The probability distribution of pandas across the Reserve was obtained by conducting a bivariate normal probability density estimation on giant panda signs obtained from the most recent census on giant pandas conducted in 2001 (State Forestry Administration 2006)...... 153

Figure 6.2. Proportion of horse home ranges across giant panda habitat suitability classes. Data pertain to home ranges of four domestic, free-ranging horse herds inhabiting forested areas in Wolong Nature Reserve, China. Percent suitable habitat is derived from a suitability index (Liu et al. (1999), which is a composite of forest cover, elevation and slope (each shown individually on the right). Forest was derived from supervised classification of Landsat TM (2007) imagery and elevation and slope from a Digital Elevation Model (DEM) acquired by the National Aeronautics and Space Administration’s (NASA) Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER)...... 154

Figure 6.3. Proportion of horse home ranges in each percent volume contour of the predicted probability distribution of giant pandas in Wolong Nature Reserve. The probability distribution was obtained by conducting a bivariate normal probability density estimation on giant panda signs obtained from the most recent census on giant pandas conducted across their entire range in 2001 (State Forestry Administration 2006). Percent volume contours ranged from 5 to 100% (with 5% representing the top 5% probability of occurrence), although no portion of any horse home range fell within the 5-70% range...... 155

Figure 6.4. Home range area obtained through home range isopleths for three domestic free- ranging horses and three wild pandas monitored using GPS collars in Wolong Nature Reserve. Home ranges were calculated using the biased random bridge approach (Benhamou 2011). Values depicted are the means and standard deviations...... 156

Figure 6.5. K-select analysis on marginality of habitat selection vectors among 3 wild giant pandas (Mei Mei, Zhong Zhong, and Chuan Chuan) and 3 horse herds roaming in giant

xviii panda habitat (Yusidong, Papagou, and Qicenglou). Plots include (a) bar chart of eigenanalysis showing proportion of marginality explained by each vector, (b) direction of the first two factorial axes, (c) variable loadings on the first factorial plane, and (d) marginality vectors for all 6 subjects on the first factorial plane. The notation “d” in (b), (c), and (d) indicates the distance of each grid cell...... 157

Figure 6.6. Estimated percent of bamboo eaten by domestic, free-ranging horses in field plots (30 m x 30 m) distributed in two areas [Fangzipeng (n= 49) and Yusidong (n= 57)] of giant panda habitat in Wolong Nature Reserve, China...... 158

Figure 6.7. Number of panda signs observed during repeated sampling in transects at Fangzipeng during three periods prior to horse occupancy and four periods after horse occupancy. Sampling periods denoted with a star (*) represent twice as many signs as were actually found during the survey (i.e., estimates were doubled to account for a ½ search effort performed during these periods)...... 159

xix

CHAPTER 1

INTRODUCTION

1 Background

The Big Picture

Humans have altered the earth in numerous profound ways over the last century, to the extent that this period of the earth's history has come to be known as the anthropocene (Crutzen

2006). As the human population has increased exponentially over the last hundred years, so have a multitude of threats to the natural world such as pollution, desertification, deforestation, and collapse of ecosystems (Liu and Raven 2010; Pimm and Raven 2000). Because humans and nature are inextricably linked in coupled human and natural systems throughout the globe (Liu et al. 2007a), such natural threats in turn directly affect human populations by contributing to social unrest, war, disease, and poverty (Liu et al. 2007b). As a result, ensuring long-term sustainability of the increasingly degraded earth is one of the ultimate challenges of our time (Clark and

Dickson 2003).

Investigations into how animal species around the world are faring under such circumstances can provide a unique window into understanding global sustainability challenges.

The situation for many animals is dire, as reflected in the current extinction rates that are several thousand times higher than background levels (Ceballos et al. 2010). Anthropogenic threats responsible for such declines are numerous, including overharvesting, invasive species introductions, and global climate change (Pimm and Raven 2000). However, among the most widespread and severe threats involve human pressure on animal habitats– the very lifelines that support animal existence (Pimm and Raven 2000). Such threats are manifested as habitat loss, habitat degradation, and habitat fragmentation and affect nearly every corner of the globe.

Investigating the complex relationships between animals and their habitats is one of the cornerstones of ecology. Examples of key lines of inquiry include characterization of an animal's

2 cognitive perception of heterogeneous space (Menzel 1991), an individual's ability to adapt to changing resource limitations (Arthur et al. 1996), a species' evolutionary strategy (Southwood

1977), a habitat's capacity to support a given number of individuals (MacLeod et al. 1996), and a habitat's suitability for a given species' persistence (Hirzel et al. 2002). This information is of urgent need for endangered species whose dwindling habitats face mounting pressures from humans and whose survival depends on active management and often rapid decision-making by managers.

Studies that track individual animals across time and space can provide a comprehensive picture of how animals relate to their habitats across their entire home ranges, in addition to an understanding of how human activities have directly impacted the lives of individual animals

(Kernohan et al. 2001; Marzluff et al. 2001). Recent advances in Global Positioning System

(GPS) animal collars have resulted in increased spatial and temporal resolution of such data, allowing researchers to ask complex questions related to the impact of spatial and temporal variability on animal behavior (Cagnacci et al. 2010). Such advances make it possible to understand complex coupled human and natural systems in a novel way from the perspective of an animal in real time as it negotiates contested space (Martin et al. 2010).

The Giant Panda- a Global Icon for Conservation

The giant panda (Ailuropoda melanoleuca) is a globally important animal species that is in need of such pioneering efforts. This large mammal is perhaps most well-known for having a unique black-and-white coloration that gives it recognizable and expressive facial features

(Figure 1.1). The giant panda is a "living fossil" that has existed since the time of the Fourth Ice

Age (Peng et al. 2007). The panda belongs to the Ailuropoda genus which first appeared in the

3 fossil record in the late Pliocene era in caves of southern China (~2.4 million years ago, Jin et al.

2007). The Ailuropoda evolved from the ursid Ailurarctos lufengensis, a bear first appearing in the late Miocene age (~7-8 million years ago) in China's Yunnan province (Hunt 2004). Having survived the large-scale die-off of mammals in the early Pleistocene era, the Ailuropoda genus later flourished and by the late Pleistocene era (15 thousand years ago) was found in the fossil record across China, Myanmar, Laos, Vietnam, and Thailand (Hunt 2004; Jin et al. 2007).

A topic of particular scientific interest in this species is its unique adaptations to bamboo, a plant that makes up over 99% of its diet (Schaller et al. 1985). Panda’s carnivore-like short digestive tract and lack of rumen mean that the panda can only digest less than 20% of its food

(Schaller et al. 1985). However, pandas have adapted over evolutionary time with the development of an enlarged wrist bone to allow for manipulating bamboo stems, a large skull and jawbone for mastication of rough plant material (Schaller et al. 1985), and a series of enzymes in the gut to break down cellulose (Zhu et al. 2011). Pandas have also adapted behavioral strategies to maximize energy intake by selecting for different bamboo species and plant parts at different times of year based on bamboo nutritional characteristics (Hu and Wei

2004; Schaller et al. 1985).

Giant pandas also have several other notable characteristics. They are believed to be solitary, living in separate but overlapping home ranges roughly 4-10 km2 in size (Pan et al. 2001;

Schaller et al. 1985). The male home range is slightly larger and encompasses the ranges of several females (Pan et al. 2001; Schaller et al. 1985). Pandas gather in groups during the mating season every spring, when multiple males aggressively compete for rights to mate with estrous females (Pan et al. 2001). Females raise one young at a time (even after sometimes birthing two cubs) and invest 1.5-2 years in care of altricial young (Schaller et al. 1985). Pandas monitor the

4 activities of conspecifics through a complex scent marking system (Schaller et al. 1985).

Individuals advertize their presence by depositing anogenital gland secretions on tree trunks throughout the habitat (Schaller et al. 1985; Swaisgood et al. 1999). Other pandas interpret such signs, gathering information such as deposition time and cues to the identity and estrus state of the individual (Schaller et al. 1985; Swaisgood et al. 1999).

Threat of Humans

Despite having flourished during much of the Pleistocene era, pandas have recently experienced significant declines. The modern day distribution of giant pandas declined to roughly 262,000 km2 by the year 1800, distributed across just 5 provinces in southwestern China

(Reid and Gong 1999; Zhu and Long 1983, Figure 1.2). The species' range fell to 124,000 km2 in the year 1900, and is limited to 21,300 km2 today (Reid and Gong 1999; Viña et al. 2010; Zhu and Long 1983). The panda's range is also currently fragmented into over 30 isolated patches across 6 mountain ranges in China's Gansu, Sichuan, and Shaanxi provinces (Hu 2001). This range decline is also reflected in population estimates. The giant panda population size prior to the 20th century is unknown, but was estimated to be around 3,000 individuals in 1950 (Hu 2001).

Range-wide surveys indicate declines to 2,400 in the 1970’s, 1,100 in the 1980’s, and 1,600 individuals in the 2000’s (Hu 2001).

Declines in panda habitat and population are mainly the result of growth of the human population and associated human impacts (Liu et al. 2001; Reid and Gong 1999). As the world's most populous nation, China has experienced a population growth from roughly 580 million in the 1950's (Banister 1987) to over 1.3 billion people today (20% of the world's total, Liu and

Diamond 2005). Along with the population increase, there have also been marked shifts in

5 human behavior since the onset of the industrialized era, with an exponential increase in gross domestic product (GDP, World Bank 2007) and a rapid increase in per-individual resource use with declining household size (Liu et al. 2003). Threats to giant pandas are varied and can be tied to both challenges in meeting basic livelihood needs of the economically poor rural people living within giant panda habitat and outside pressures from development companies and external markets (Liu et al. 1999).

Timber harvesting was the most significant threat to the giant panda during the 1960’s to early 1990’s. Timber harvesting was banned across all natural forests in China in 1998, but the legacy effect is still present across giant panda habitat (State Forestry Administration 2006) and small-scale fuelwood collection still occurs even inside nature reserves (Liu et al. 1999). Timber harvesting was the most frequently encountered disturbance in the most recent giant panda census, encountered in 28% of 34,187 plots across the giant panda range (more than double any other disturbance type, 41% of all disturbances, State Forestry Administration 2006). Panda use is affected because bamboo may not grow in such areas or may grow in overly dense and non- nutritious patches (Bearer et al. 2008; Schaller et al. 1985). However, pandas have been shown to make use of forest stands that have not been cut for at least 37 years and are recovering (Bearer et al. 2008).

Land cultivation has also taken place in panda habitat for several hundred years (Pan et al.

2001). This activity most severely affects low elevation regions of panda habitat, but in the last several decades has extended up to the higher elevations (Schaller et al. 1985). In addition, road construction is a major source of fragmentation. For instance, the once connected Minshan mountain range, which now supports roughly 44% of the panda population, is divided into three separate patches due to the creation of major highways and associated human activities

6 surrounding them (State Forestry Administration 2006). Livestock grazing is another prominent threat, as it was the second most frequently observed in the most recent giant panda census (11% of 34,187 plots, 17% of all disturbances) and also the most current, with 93% of occurrences ongoing (State Forestry Administration 2006). Livestock can threaten pandas by consuming bamboo and trampling vegetation (Ran et al. 2003). Tourism also threatens giant pandas, a threat that has increased in recent years. As of 2000, around 80% of China’s nature reserves had developed tourism programs and almost 16% attracted more than 100,000 tourists per year (Li and Han 2001). Tourism can impact giant pandas and their habitat via facility construction, in addition to noise and garbage (Liu 2012). Other threats including medicinal herb collection, bamboo harvesting, and mining have been less studied but are frequently observed across giant panda habitat (State Forestry Administration 2006).

Considering these various and ongoing threats to the giant panda and its habitat, sound management is required in order to ensure the species long-term survival. A number of management initiatives have been put in place to protect the species, including the establishment of over 70 nature reserves covering nearly 50% of the giant panda's current range (State Forestry

Administration 2006; Viña et al. 2010). In addition, several policies have been put in place to protect pandas both inside and outside of nature reserves, including strict punishments for poaching (Reid and Gong 1999) and a timber harvesting ban in all natural forests (Zhang et al.

2000). Payments for ecosystem services (PES) programs have also been established to reward local people for participating in conservation initiatives, such as by monitoring forest parcels against illegal timber harvesting (Natural Forest Conservation Program) or planting trees in once-cultivated areas (Grain-to-Green Program, Liu et al. 2008). These initiatives have made progress in protecting the giant panda's habitat, but more could be done to improve their efficacy,

7 in part by developing a better understanding of wild giant panda behavior, which in many ways lags behind policymaking.

Current Research Gaps

What is known about the behavior of giant pandas comes from studies conducted on the species in the 1980’s, when much of the panda ecology was explored for the first time (Hu et al.

1985; Pan et al. 2001; Schaller et al. 1985). These ambitious studies remain as the cornerstones of panda research to this day. However, such studies were also constrained by the fact that giant pandas are elusive and spend the majority of their time in dense bamboo thickets that provide low visibility for human observers (Schaller et al. 1985). Thus, most of the knowledge on panda behavior has been gained from analysis of wildlife signs (e.g. feces, eaten bamboo, footprints).

But it is difficult to tie such signs to individual pandas and gain an appreciation for spatial extent and timeline of habitat use by individual pandas. A limited number of radio collars were deployed on giant pandas in the 1980’s that helped to fill this knowledge gap (Pan et al. 2001;

Schaller et al. 1985), but the data had poor location accuracy due to challenges in triangulating in difficult terrain. In addition, the technology available at the time did not allow for these data to be spatially analyzed or linked to information on habitat characteristics (such as those derived from remotely sensed imagery). Since all collaring of giant pandas was banned by the Chinese government from 1995-2006, no telemetry studies have been implemented on giant pandas until recently.

As a result, current management plans and policies for this species rely on a largely simplified view of panda behavior. Models often use broad habitat suitability classes to predict panda presence or absence (Liu et al. 2004). Other models that attempt to convert

8 presence/absence to a representation of individual pandas use a “cookie-cutter” value to represent panda home range (such as 4-6 km2), or a singular value that is reproduced in an identical fashion across all panda habitat (Linderman et al. 2004; Loucks et al. 2003), which adds little context to understanding the intensity of use across that space and the relative suitability or importance of different areas within an animal’s range. Perhaps the greatest danger with the simplifications in these models is that they may lead to erroneous assumptions about the impacts of humans on wild pandas, a topic that has been extensively explored on an abstract level and on panda habitats, but never empirically tied to individual wild pandas.

Dissertation Overview

A New Opportunity

We obtained special permission from the Chinese government to deploy GPS collars on 5 wild giant pandas in Wolong Nature Reserve, Sichuan province, China. This unique opportunity allowed us to adopt novel analytical techniques to understand giant panda space use and habitat selection at a fine temporal and spatial scale. Although the sample size is necessarily small for this endangered and government-protected species, several other studies on rare species have been restricted to similar sample sizes in the past and have still been able to contribute new and meaningful findings about understudied and rare animals (Gill et al. 2008; Kramer-Schadt et al.

2004; Miller et al. 2010). Nonetheless, our results should not be extrapolated to the entire giant panda population and should be interpreted with appropriate caution. That said, this is among the first telemetry studies conducted on giant pandas since 1995 and among the first to use GPS collars on the species (see also Zhang et al. (2014) for another recent effort in Foping Nature

Reserve).

9 The objective of this dissertation is to examine behavior of wild giant pandas across a coupled human and natural system. Our focus is on interpreting patterns of space use and habitat selection revealed from our novel GPS collar dataset and in turn integrating these data with a diverse collection of other data sources to inform the design of conservation and management strategies for this endangered species.

Study Area

The study area is located in Wolong Nature Reserve (102°52’ – 103°24’E, 30°45’ –

31°25’N), Sichuan, China. This 2,000 km2 protected area is in the southeastern portion of the

Qionglai mountain range, located in the center of the province. Wolong is one of the first nature reserves established in China in 1963 after the Third National People’s Congress issued a directive in 1956 to set aside areas for conservation (Harkness 1998). The main objective of

Wolong laid out in the first management plan was to provide a refuge for ‘the protection of [the] giant panda, other valuable rare animals and plants, and the typical natural ecosystem’ (Ministry of Forestry 1998: X).

Located in a transition zone between the Sichuan basin and the Tibetan highlands, the reserve is known for having a large elevational range spanning 1,200 to 6,250 m (Schaller et al.

1985). Steep cliffs and narrow valleys are part of the high topographic relief in this area that have arisen as a result of frequent tectonic activity taking place along the Longmen mountain fault line since the third-century glacial events (Wolong Administration Bureau 2004). The local climate is typical of the Qinghai-Tibetan Highland belt, with a temperature averaging at around

9˚C throughout the year (range of -12˚C to 30˚C) and a high humidity, averaging 80% (Wolong

Administration Bureau 2004).

10 Wolong is within one of the global biodiversity hotspots (Liu et al. 2003; Myers et al.

2000), a product of the wide elevation range, humid climate, complex topography, and extensive watersheds in the Qionglai mountains (Mackinnon 2008). Perhaps the most well known species present is the giant panda. Wolong is the third-largest reserve for giant pandas based on area and contains approximately 10% (154) of the total giant panda population (State Forestry

Administration 2006). Aside from giant pandas, there are thousands of other plant and animal species (Li et al. 1992). In fact, Wolong alone ‘contain(s) more plants, butterflies, amphibians, and birds than in most European countries’ (Mackinnon 2008: 96).

There is a distinct vertical zonation pattern of habitat along the elevational gradient, as habitat types gradually change from evergreen broadleaf forest, evergreen and deciduous broadleaf forest, mixed coniferous and deciduous broadleaf forest, subalpine coniferous forest, alpine meadow and thicket, and finally rock as elevation increases (Schaller et al. 1985). Pandas are mainly found at the mid-elevations of 2000 to 3300 m (Schaller et al. 1985). The habitat in this zone includes evergreen and deciduous broadleaf, coniferous and deciduous broadleaf, and subalpine coniferous forests, each having abundant bamboo in the understory (Schaller et al.

1985).

Wolong was inhabited by people centuries before it became a nature reserve, among the earliest inhabitants believed to be Tibetan people who arrived during the late seventeenth century

(Ghimire 1997). As such, the original management plan of Wolong contained several sections related to poverty alleviation in the local rural communities, making human livelihoods a central underpinning of reserve focus and activities (Ministry of Forestry 1998). Today there are nearly

5,000 local residents living within the reserve, many of whom are ethnic minorities (e.g. Tibetan,

Qiang) adopting farming lifestyles. Much like the rest of China, the human population in Wolong

11 has increased significantly in recent years from 1582 people in 1949 (Li et al. 1992) to roughly

4,900 people in 2012. The residents interact with the natural environment in ways that impact giant panda conservation, mainly via land cultivation, animal husbandry, timber harvesting, fuelwood collection, and medicinal herb collection (Liu et al. 1999; State Forestry

Administration 2006). There is also a provincial road running through the reserve (303), which supports various forms of transportation of goods and tourists, fueling the local economy. Local people also participate in payments for ecosystem services (PES) programs, namely the Natural

Forest Conservation Program (NFCP), the Grain-to-Green Program (GTGP), and Grain-to-

Bamboo Program, all of which provide monetary rewards or subsidies for participating in conservation endeavors (Liu et al. 2008; Yang et al. 2013).

The specific area where the study took place is in the northeastern portion of Wolong and is called Hetaoping (Figure 1.3). It is roughly 40 km2 in size extending from the captive giant panda breeding center next to the main road (1,800 m) up along a steep mountainous incline to

3,100 m. Hetaoping is also located between two villages to the south (Sancun) and northeast

(Laoyashan) and a provincial highway (303) to the west. The rugged terrain prevents frequent human visitation to the home ranges occupied by the pandas, but camera trapping revealed the occasional presence of people in the forest.

Hetaoping supports a large number of giant pandas, as camera trapping and DNA extracted from field-collected feces suggest a local population of 16-25 individuals (unpublished data). Such surveys also indicated that the area was inhabited by other large mammals, including the tufted deer (Elaphodus cephalophus), serow (Capricornis milneedwardsii), and sambar (Rusa unicolor), although none compete directly with pandas for space and food (Schaller et al. 1985).

The study area includes mixed deciduous and coniferous forest and subalpine coniferous forest.

12 Common tree species include Chinese walnut (Juglans cathayensis), mono maple (Acer mono), hemlock (Tsuga longibracteata) and spruce (Picea asperata). Bamboo is also prevalent throughout the understory. The main bamboo species are arrow (Sinarundinaria fangiana), umbrella (Fargesia robusta) and Yushan (Yushania bravipaniculata) bamboo.

Data Collection

Five wild giant pandas were captured in 2010 and 2011 at Hetaoping and fitted with GPS collars before being released. Pandas may be referred to by researcher given names throughout the dissertation- Pan Pan, Mei Mei, Zhong Zhong, Long Long, and Chuan Chuan. Study pandas included 4 females and 1 male (Chuan Chuan) and all adults except for 1 sub-adult female (Long

Long). Staff members at the China Conservation and Research Center for the Giant Panda

(CCRCGP) were responsible for animal safety. We obtained an exemption from the Institutional

Animal Care and Use Committee (IACUC) from Michigan State University (MSU) to conduct this dissertation because no individuals affiliated with MSU had any direct contact with the animals being cared for by the CCRCGP. Details about collar specifications and data collection procedures are discussed in forthcoming chapters.

In addition to the GPS collar data, several other data sources are used throughout the dissertation to characterize giant panda behavior across a coupled human and natural system.

These include published statistics on habitat selection by pandas, digitized maps from Wolong's management plans, remotely sensed imagery, animal and habitat data obtained from field surveys, and GPS collar data from livestock in the reserve. These sources of data will be discussed in detail in the relevant forthcoming chapters.

13 Dissertation Outline

This dissertation consists of seven main chapters (Figure 1.4). Following this current

Chapter 1 which serves as the background and dissertation overview, Chapter 2 presents a literature review and quantitative synthesis of published studies on giant panda habitat selection.

This chapter is important because there is a lack of consensus in the literature on fundamental components of giant panda habitat selection. This chapter also explores for the first time several complexities in giant panda habitat selection such as multivariate effects, interactions among habitat factors, and selection across levels. My coauthors and I also put forth recommendations for improvement of habitat selection studies on giant pandas in the future, guidance which also informed the development of the remainder of this dissertation. This chapter has been accepted for publication in the journal Ursus and is largely written in this journal's style.

Chapters 3 and 4 explore ecological patterns revealed by studies on the GPS-collared giant pandas. Chapter 3 is an investigation of space use. In this chapter, we employ model-based approaches to characterize home range, core areas, and space use interactions among giant pandas. This chapter is important because it sets the stage for understanding the extent and structure of giant panda space use prior to delving into panda-habitat relationships in more detail in forthcoming chapters. This chapter is in review with the Journal of Mammalogy and is largely written in this journal's style. Chapter 4 builds on Chapter 3 by analyzing variation in habitat use and selection by the collared pandas across their home ranges. Further, one of the main outputs from Chapter 3 (the biased random bridge utilization distribution) is used as a response variable for spatial autoregressive resource utilization functions (RUF) presented in Chapter 4 with several biogeophysical variables included as model predictors. We also relate panda habitat use

14 to habitat suitability models and habitat selection patterns, the latter based on availability of habitats across both the home range and entire study area.

Chapters 5 and 6 both utilize foundational principles established in the previous three chapters to explore special topics relevant to informing management decisions. These chapters integrate the GPS collar data with other diverse data sources to answer key management-driven questions. Chapter 5 tackles the issue of zoning in protected areas. We examine the efficacy of the zoning scheme put in place in Wolong to spatially segregate human development and biodiversity conservation. We conduct spatial overlays of the zoning scheme with a number of different types of human and natural data collected in the reserve, including data from GPS- collared pandas and livestock. We also pinpoint specific areas where the zoning scheme could be revised to better protect the endangered giant panda in the future. This chapter was published in

Biological Conservation (Hull et al. 2011b) and largely conforms to this journal's style.

Chapter 6 builds on Chapter 5 by further exploring an emerging but severely understudied human impact on giant pandas- livestock grazing. We look specifically at a new livestock threat in Wolong, that of domestic, free-ranging horses grazing in forests. We integrate data from the GPS-collared pandas with other field and GIS data to examine a number of different aspects of the issue, including the distribution of horses with respect to giant panda habitat, differences between habitat selection patterns of horses and pandas, bamboo consumption by horses, and panda use of horse-affected areas. We synthesize these findings to provide an integrated on-the-ground understanding of this emerging management issue. This chapter was published in the Journal for Nature Conservation (Hull et al. 2014) and largely conforms to this journal's style. In Chapter 7, I synthesize information from all previous chapters

15 and reflect on possible future directions for research on giant panda behavior in coupled human and natural systems in the future.

APPENDIX

17 Figure 1.1. A giant panda. Photo credit: Bin Liu, 2005.

18 Figure 1.2. Distribution of giant pandas over time. Pre-historic estimates are derived from a map in Loucks et al. (2001) but adjusted for additional fossil evidence detailed in Jin et al. (2007). Estimates for 1980 and 1990 are derived from a map in Reid and Gong (1999) that summarizes data in Zhu and Long (1983). Present estimates are derived from a supervised habitat classification by Viña et al. (2010).

19 Figure 1.3. The Hetaoping study area for wild giant panda GPS collar research. Elevation is derived from a digital elevation model acquired by the National Aeronautics and Space Administration’s (NASA) Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER).

20 Figure 1.4. Broad overview of this dissertation.

CHAPTER 2

A SYNTHESIS OF GIANT PANDA HABITAT SELECTION

In collaboration with

Gary Roloff, Jindong Zhang, Wei Liu, Shiqiang Zhou, Jinyan Huang, Weihua Xu, Zhiyun

Ouyang, Hemin Zhang, and Jianguo Liu

Abstract

The giant panda (Ailuropoda melanoleuca) is a global conservation icon, but its habitat selection patterns are poorly understood. We synthesized previous studies on giant panda habitat selection.

We confirmed that pandas generally selected forests with moderate to high bamboo densities, mid-elevations, both primary and secondary forests, and areas more distant from human activities. Pandas did not select steep slopes. We also highlight the interactive effects among different habitat components, such as weaker selection for gentle slope and large patch size in disturbed secondary forests compared to primary forests. Pandas selected for land cover and disturbance at the level of the geographic range and selected for variables such as slope and bamboo density at the level of the home range. Furthermore, selection for higher bamboo cover did not change with bamboo availability, but selection against secondary forest declined as availability of this forest type increased. Our results have implications for the conservation of pandas, particularly the need for inclusion of areas previously seen as less suitable (e.g. moderate slopes and secondary forest) in protected area and habitat restoration planning.

Introduction

The study of habitat has important implications for understanding resource needs and ongoing threats to the most endangered ursid in the world, the giant panda (Ailuropoda melanoleuca). Various habitat factors have been explored, including characteristics of the geophysical environment (e.g., elevation, topographic slope and aspect; Liu et al. 1999; Schaller et al. 1985), vegetation structure (e.g., bamboo and tree cover; Reid and Hu 1991; Tuanmu et al.

2011; Viña et al. 2008), natural disturbances (Linderman et al. 2006; Viña et al. 2011), and human impacts (e.g., timber harvesting, livestock grazing; Pan et al. 2001; Hull et al. 2011b; Liu et al. 2001). Generally, studies to date have focused on habitat use (i.e., panda occupancy of a given area with certain environmental conditions) and few have explicitly addressed habitat selection (i.e. use as a function of availability). Habitat selection studies are needed to better understand how pandas prioritize where to spend time, given limited available options.

Giant pandas are currently found in the forests of southwestern China. Human impacts have relegated the remaining 1,600 wild giant pandas to small and fragmented areas totaling roughly 21,300 km2 (State Forestry Administration 2006; Viña et al. 2010). Pandas were once distributed throughout the lowlands of western China, but today are limited to a fraction of their historical range in six fragmented, mountainous regions (Wei et al. 2012). Panda habitat is currently being managed to help sustain the population, via creation of nature reserves and implementation of payments for ecosystem services (PES) programs that recruit locals for forest monitoring and replantation (Viña et al. 2010; Liu et al. 2008).

Giant pandas are a specialist species with bamboo comprising about 99% of their diet

(Schaller et al. 1985). Pandas consume >60 bamboo species across their range (Hu and Wei,

2004) and forage for up to 14 hours per day (Schaller et al. 1985). Despite possessing specialized

enzymes for digesting cellulose in the gut (Zhu et al. 2011), pandas have a short, carnivorous digestive tract that lacks compartments for rumination so passage rates are high (Schaller et al.

1985). Nutrient uptake is also low due to the low nutrient quality of bamboo (Schaller et al.

1985). This limitation means that pandas need to be highly selective when choosing habitats to fulfill their foraging needs.

Panda habitat is mixed coniferous-deciduous forest with bamboo dominating the understory (up to 90% of the understory cover) with a varying mid- and overstory tree species mix and structure (Bearer et al. 2008). These assemblages provide opportunities for pandas to use diverse bamboo species, ages, and plant parts as environmental conditions change (Reid et al.

1989). The high availability of bamboo throughout the mountain ranges in southwestern China allows pandas to occur at higher densities than many other bear species throughout the world

(Garshelis 2004).

Whereas giant panda reliance on bamboo is well documented, other elements of panda habitat selection are not well understood (Liu et al. 2005; Hull et al. 2011a). Pandas occupy remote, inaccessible areas with thick vegetation, making research logistically difficult.

Additionally, the Chinese government imposed an 11-year ban (1995 – 2006) on telemetry of giant pandas for animal safety reasons (Durnin et al. 2004). Yang et al. (2006) has summarized panda habitat selection patterns from published studies, but the authors did not present a quantitative analysis and also did not differentiate habitat selection and use, thus confounding the two concepts and obscuring the effect of habitat availability. In addition, the paper was published before many recent contributions to the habitat selection literature (i.e., before 10 studies included in the current review). Habitat selection is important to because it represents an

expression of animal behavior (i.e., a choice) that is presumably linked to how animals respond to different habitat availability (Manly et al. 2002; Nielsen et al. 2010).

Habitat availability characterized by variables such as slope, aspect, elevation, and forest disturbance varies considerably across the geographic range of pandas (Zhang and Hu 2000; Hu and Wei 2004; Yang et al. 2006). Panda habitats tend to occur as multiple insular areas associated with different mountain ranges. For instance, panda habitat in the Qinling mountain range tends to occur at lower elevations and is flatter when compared to the rest of panda range due to both natural and human factors (Hu and Wei 2004; Tuanmu et al. 2012). Hu and Wei

(2004) and Wang et al. (2010) also noted that the available habitat in the Xiangling mountain ranges was more fragmented than habitats in the rest of panda range. Although several authors have identified differences in habitat availability throughout the range of pandas, quantitative studies on selection of those habitats are lacking (but see Zhang and Hu 2000 for a comparison between Dafengding and Yele Reserves).

We sought to better understand panda-habitat relationships by synthesizing existing studies, emphasizing habitat selection. To our knowledge, this is the first effort at quantitatively analyzing findings across the available panda habitat selection studies. In doing so, we characterized selection of giant pandas with respect to available geophysical, vegetation, and disturbance conditions. We also synthesized the complexities of the habitat selection process for pandas, including: (1) multivariate effects, (2) interactions among different habitat factors, and (3) variation in selection across habitat selection levels (from geographic range to within-home range; Johnson 1980). Exploration of habitat selection complexities is needed because most previous panda research focused on single variable relationships at one selection level, thus potentially oversimplifying the habitat selection process for this species. We conclude with a

discussion of current weaknesses and future directions for research on giant panda habitat selection that could facilitate conservation planning for this species.

Methods

We performed literature searches in English (ISI Web of Science and Google Scholar) and Chinese (Wangfang Data and the Chinese National Science Digital Library) to find publications that described giant panda habitat. We sought references in refereed journals, university theses and dissertations, books, government documents, and edited book chapters. We used the key words “giant panda” and “habitat selection” and “giant panda” and “habitat”. We filtered individual references to determine relevancy to our study. The inclusion criterion was that the study provided data or results on habitat selection, i.e., compared the environmental conditions of areas used by pandas to conditions available (or not used) in the landscape. Many references (n > 40) solely analyzed habitat use, some of which incorrectly (n = 10) adopted the term habitat selection in the title or abstract. Three references satisfied the habitat selection criterion but were excluded from our synthesis because of low sample sizes (i.e., < 50 used or available data points), which were further reduced when the authors of those studies divided the data into ≥ 4 habitat classes (e.g., types of forest, magnitudes of slope). We summarized studies with respect to design and implementation and identified three habitat factors that affect pandas: geophysical, vegetation, and disturbances (Table 2.1). We chose these categories based on previous panda research (Liu et al. 1999). We noted which measured habitat components were found to significantly relate to panda habitat selection for each study (Table 2.2). We also explored interactive effects among different habitat components.

In some instances, we used additional analyses to compare habitat selection across studies. One analysis that was used in many studies was the Vanderploeg and Scavia relativized electivity index (VS relativized index; Vanderploeg and Scavia 1979). This index is robust and stable across varying magnitudes of resource availabilities (Lechowicz 1982; Manly et al. 2002).

The index ranges from -1 (strong selection against) to 1 (strong selection for), with values around

0 indicating that habitats are being used in proportion to their availability (i.e., no selection). We calculated this index on topographic slope, bamboo cover, and forest age for studies that provided adequate data (Ran et al. 2003; Kang et al. 2011b). We also constructed plots that portrayed habitat use versus availability for these same three variables. These habitat characteristics were chosen due to a combination of available data and known importance to panda habitat selection. We divided slope into 5 discrete categories (<5°, 6–20°, 21–30°, 31–40°, and >40°) and bamboo cover into 4 discrete categories (0–25%, 25–50%, 50–75%, and 75–100% cover) based on the most common delineations used in the studies. We also performed a χ2 goodness-of-fit test (Neu et al. 1974) to evaluate panda selection of primary versus secondary forest. We also classified habitat selection according to level. Selection levels range from first order (i.e., factors influencing occurrence across the species' range), second order (i.e., factors influencing selection of the home range), third order (i.e. factors influencing selection of habitats within the home range), and fourth order (i.e., micro-site features such as food items or shelters selected within a home range; sensu Johnson 1980).

Results

Scope of Habitat Selection Studies on the Giant Panda

We located 23 studies (Table 2.1) and deemed those conducted within the same reserve as separate observations if the researchers, time frames, or spatial extents differed. In nine instances, we found 2 or 3 references that used the same dataset and we counted these as a single study (Table 2.1). Most studies (52%) were conducted in areas of 20–500 km2. The remaining studies with the largest areas included the panda range-wide dataset (Ye 2008; Wang et al. 2010), two province-wide datasets (Gansu Provincial Forestry Department of Wildlife Conservation

2004; Zhang et al. 2011), and two mountain range-wide datasets (Qi et al. 2009, 2011, 2012;

Wang 2008; Wang et al. 2009; Table 2.1). All studies except one were conducted at the population level (i.e., individual pandas were not identified). Most studies (74%) documented panda occurrence via indirect evidence (e.g., feces, tree marks, partially-eaten food) and paired these used locations with available locations sampled nearby. Of these, most (76%) located sample plots along transects but few (35%) provided detailed information on how availability plots were located (e.g., minimum distance to used plots, choice of location), which hindered our ability to draw inference. Survey transects and corresponding plots tended to be opportunistically located along established human and animal travel routes, potentially introducing bias (see Discussion section for further discussion on methodological issues). For those studies not relying on field plots to characterize availability (26%), some selected random locations across the study area (8%) or the entire study area (17%) using a geographic information system (GIS).

Panda habitat selection studies have analyzed 46 habitat factors, including 8 geophysical,

30 vegetation, and 8 disturbance factors (Table 2.2). Around 52% of the studies included

geophysical, vegetation, and disturbance factors simultaneously. The most frequently studied factor was vegetation (91% of studies), followed by geophysical (87%) and disturbance (61%).

The most common methods used for data analysis were the VS relativized index (35% of studies) and the χ2 test (35%). Other types of analyses included Mahalanobis distance, regression, discriminate function analysis, and ecological niche factor analysis (ENFA).

Single Geophysical Factors

Pandas selected for gentle and moderate slopes with abundant solar radiation and mid- elevations. Of the 18 studies that considered slope, 78% found significant effects of slope on giant panda habitat selection (Table 2.2), with 56% documenting a negative relationship. We summarized the VS relativized index from the slope data from 10 studies with available raw data and found that 70% showed positive selection for gentle to moderate slopes with 30% showing a quadratic relationship (i.e., peak in selection at moderate slopes; Figure 2.1). Steep slopes (>40°) represented <20% of available panda habitat and pandas selected <6% of these steeper areas

(Figure 2.2a). We found the greatest variation in available habitat for moderate slopes (5°–20° and 21-30°), which made up between 19–41% and 20–54% of the landscape, respectively.

Availability of these slope classes did not appear to consistently influence selection.

Of the 16 studies that analyzed panda habitat selection with respect to aspect, 69% found a significant result, but selected aspects varied by study area (Table 2.2). Liu et al. (2011) found that solar radiation was a positive predictor of panda habitat selection and suggested that solar radiation may be a more direct indicator of selection than aspect alone. Elevation was significant in 9 of 11 studies, as pandas selected mid-elevations where palatable bamboo was typically found (2,000 – 3,000 m), a pattern that likely reflects both biogeophysical limitations that

prevent bamboo growth at extreme high and low elevations and human impacts at low elevations.

In one study (Qi et al. 2011) elevation (and slope) was more important to habitat selection for females than males. Hillside position was analyzed in 8 studies and pandas generally selected ridges (63% of the studies), upper slopes (50%) and mid-slopes (50%) over valley basins (Table

2.2).

Single Vegetation Factors

Pandas selected forest areas with larger patch sizes, but individual structural components of those forests were not consistently important (Table 2.2). Selection for specific forest types varied across 13 studies, with some (46%) showing selection for both coniferous and mixed coniferous-deciduous forest and others for one of those two types (15% each). Landscape metrics were evaluated in 2 studies which showed that pandas tended to select larger, more connected patches (Bearer et al. 2008; Wang et al. 2010).

The structural attributes of forests were quantified using various tree, shrub, and bamboo based vegetation metrics (Table 2.2). Selection of tree and shrub structural attributes was not consistent among studies. For example, only 29% of the studies (n = 14) that evaluated canopy cover found a significant effect on panda habitat selection, 63% (of 8 studies) found that tree diameter was important, and 67% (of 6 studies) found that tree height was important (Table 2.2).

Similarly mixed results were found for shrub-related metrics (Table 2.2). For example, 67% (of 6 studies) found that shrub cover was significant, and 50% (of 4 studies) found that shrub height was significant (Table 2.2). Collectively, our results indicate that tree and shrub attributes are inconsistent determinants of panda habitat selection.

Pandas consistently selected areas with higher bamboo cover. Of the 9 studies that examined this variable, all demonstrated significant positive effects (Table 2.2). We summarized the VS relativized index for 6 studies and found that pandas did not select areas with < 25% bamboo cover but selected for areas with 50–100% cover (Figure 2.3). A non-linear relationship was identified, with selection reaching an asymptote at 75% for most studies with available data

(Figure 2.3). This finding was supported by studies that found selection for the most dense patches (Wei et al. 1999; Bearer 2005) and others which revealed no selection for the most dense patches in favor of moderate densities (Wei et al. 1996; Tang and Hu 1998). Selection for bamboo density varied by season (Reid and Hu 1991) and bamboo species (Liu et al. 2005). The structural attributes of bamboo, i.e., height and diameter, were also significant. In most cases, taller and thicker bamboos were positively related to panda selection. Pandas consistently selected areas with higher bamboo cover (i.e., >50% cover) regardless of availability across panda range (Figure 2.2b) indicating that bamboo cover is a useful range-wide determinant of panda habitat selection.

Single Disturbance Factors

Pandas generally avoided areas of persistent human activity (Table 2.2). All studies (n =

3) that evaluated the effects of farmland on panda habitat selection found a negative relationship.

In 6 studies that considered distance to human activity (active roads or village) pandas selected areas farther from such locations (Bearer et al. 2008; Wang 2008; Wang et al. 2008; Feng et al.

2009; Qi et al. 2012). Qi et al. (2011) found that abandoned logging roads were positively related to pandas' (particularly females') habitat selection, suggesting that human activity is the primary deterrent to road use. Two studies considered livestock grazing and reported that pandas did not select areas used by livestock (Ran et al. 2004a; Zeng et al. 2002). Poaching (mainly of

ungulates) or herb collection did not appear to affect panda habitat selection (Ran et al. 2004a;

Zeng et al. 2002).

Pandas selected both primary and secondary forests, given that these forests offered a suitable bamboo resource. Most studies that investigated forest disturbance (70%) considered it as a binary variable: primary forest (no timber harvest) or secondary forest (a forest that has re- grown after timber harvest). Forest structure observed across these two forest types varied considerably across giant panda range, with some secondary forests supporting bamboo communities similar to what is typically found in primary forests under the appropriate conditions (i.e., a moderate amount of overstory canopy closure (35–70%; Bearer 2005)).

Pandas selected primary forests over secondary forests in 6 of 10 studies (60%) we evaluated (Bearer et al. 2008; Ran et al. 2003; Gansu Provincial Forestry Department of Wildlife

Conservation 2004; Ran et al. 2004a; Wang et al. 2006; Zhang et al. 2011). Bearer et al. (2008) suggested that pandas exhibit a non-linear selection for forest age, with selection of primary and secondary (31–100 years post-harvest) forests over forests cut within the last 30 years. Forests cut within 30 years made up 55% of all available plots, but only 16% of used plots (Bearer et al.

2008). Pan et al. (2001) compared primary and secondary forests that were cut within 6 years where 35–70% crown closure was retained and found no difference in panda habitat selection.

The authors also suggested (but did not quantify) that pandas responded negatively to more intensive forest harvesting in the form of clearcutting. Qi et al. (2012) examined selective logging versus clearcuts and reported that pandas were located closer to selectively logged forests (especially for females) and in areas with lower frequency of clearcuts. Studies (n = 2) that included intensively managed plantations as another forest type found that pandas did not

select plantations (Bearer et al. 2008; Gansu Provincial Forestry Department of Wildlife

Conservation 2004).

Availability of secondary forests ranged from 26–76% (Figure 2.2c). Pandas used secondary forests significantly less than their availability when these forests made up <60% of available habitat, but used them in proportion to their availability in two of three studies that had >60% secondary forests available.

Multivariate Habitat Selection Factors

Nine studies provided insights into the integrated importance of different variables in habitat selection of giant pandas using multivariate analyses. Bamboo cover and occurrence was consistently important over most other variables analyzed (Bearer et al. 2008; Kang et al. 2013;

Wang et al. 2008; Wang 2008; Zhang et al. 2011). Forest attributes (e.g., % canopy cover, canopy height) were important in some studies (Bearer et al. 2008; Qi et al. 2009; Wang et al.

2008; Wang 2008; Zhang et al. 2011), but not others (e.g., Kang et al. 2013; Zhang et al. 2009).

Multivariate studies also identified distance to human disturbances including active roads

(Bearer et al. 2008), villages (Feng et al. 2009; Wang et al. 2008), and crops (Wang 2008) as important positive contributors to panda habitat selection.

Unlike the consistent multivariate selection patterns for bamboo, forest availability, and human disturbance, multivariate selection based on elevation, slope, and aspect varied widely across studies. Slope is often cited as an important predictor of panda habitat selection, however it was among the most important variables in only two of the multivariate studies (Feng et al.

2009, Zhang et al. 2009). Studies that conducted both univariate and multivariate analyses (n = 3) found that some significant single variables were no longer significant when analyzed with other

variables (Bearer et al. 2008; Kang et al. 2013; Zhang et al. 2009). These single variables included tree and shrub size and bamboo density and height (after controlling for slope and proportion of old shoots, Zhang et al. 2009), and several tree structural attributes after controlling for bamboo cover, basal area, and overstory height (Bearer et al. 2008). Our findings draw attention to the potential dangers in giant panda habitat selection studies of analyzing habitat relationships without regard for multiple variables and how they potentially interact.

Interactive Habitat Selection Factors

As suggested by results from multivariate studies, understanding giant panda habitat selection is complicated by interactions among different habitat characteristics. Such interactive effects were only analyzed in 6 studies included in our review. In one study, areas where human impacts were less prominent, pandas selected lower elevations (Feng et al. 2009).

Bamboo cover and bamboo species composition also interact to influence panda habitat selection. Liu et al. (2005) found that bamboo cover significantly affected habitat selection for

Fargesia qinlingensis and not Bashania fargesii. Similarly, Bearer (2005) found that bamboo cover significantly affected habitat selection for Fargesia robusta, but not Bashania fabri. Bearer

(2005) also found that selection of slope varied among areas with different bamboo species (at different elevations). Pandas selected for the lowest slopes when foraging on B. fabri, but did not select the lowest slopes when foraging on F. robusta, potentially reflecting topographic differences in the sites that support each species (i.e., F. robusta occurred on slightly steeper slopes (mean ± SE, 24.5 ± 0.9°) than B. fabri (19.6 ± 1.2°)).

Slope also interacted with forest age. Bearer (2005) reported that slope was significant for predicting panda habitat selection in primary and recently cut forests [< 10 years] but not in moderate-aged secondary forests. Mean slope of used plots was significantly higher in the former

age classes, but did not differ significantly across ages in the available plots. Similarly, Ran et al.

(2004b) noted that pandas selected moderate slopes and avoided the steepest slopes in primary forests, but showed no selection for slope when in secondary forests. Variation existed in slope selection among different seasons in a single study (Reid and Hu 1991). Slope was not the only variable that interacted with forest age, as Bearer et al. (2008) found that distance to road was only a positive correlate of giant panda habitat selection in younger forests (<30 years) and not older forests.

Of the two studies that analyzed landscape metrics (e.g., edge density, patch size; Bearer et al. 2008; Wang et al. 2010), both found that patch size interacted with forest characteristics.

When pandas selected dense forests they chose larger patches that were closer together, and more contiguous than unselected habitat (Wang et al. 2010). Similarly, when selecting primary forests, pandas chose larger patches than unselected habitat (Bearer et al. 2008). In contrast, patch size was not important for selection of sparse (forests subjected to intense logging, Wang et al. 2010) or secondary forests (Bearer et al. 2008).

Habitat Selection across Selection Levels

Nearly half (48%) of the reviewed studies were conducted at a first-order selection level.

In first-order selection researchers compared use and availability across large spatial extents (e.g., whole reserves), often with coarse measures of habitat availability (e.g., forest and non-forest land cover types). The variables significant in predicting first order giant panda habitat selection included distances to human disturbances (e.g., villages, active roads), land cover type (e.g., forest or non-forest), elevation, and the presence of bamboo (Table 2.3).

To date no studies have differentiated among second and third order habitat selection, hence we combined these two levels for our synthesis (52% of studies). Factors that consistently predicted habitat selection at the second and third levels included slope, position on hillside (e.g. ridge vs. valley), bamboo cover, bamboo density, and distance to human disturbance (Table 2.3).

Fourth order, or selection of specific resources within a home range, was largely beyond the scope of our paper, but factors looked at have included bamboo and den trees (Table 2.3, see also

Hu and Wei, 2004 and Zhang et al. 2007).

No studies effectively compared habitat selection across multiple orders of selection within a single study. Although Qi et al. (2012) looked at multiple selection orders, the analytical approach they used was sensitive to different spatial extents thereby confounding inference on the behavioral processes of the animals (Hirzel et al. 2002). Kang et al. (2013) investigated selection at two spatial scales (both within the third selection order) and found that predictors of habitat selection at the feeding site scale (1 m2, bamboo density and diameter) differed from predictors at the larger habitat scale (9000 m2, proportion of young bamboo and bamboo cover,

Kang et al. 2013).

Discussion

Implications for Giant Panda Ecology and Management

We synthesized information about habitat selection of the giant panda across a large number of studies throughout panda range. By isolating habitat selection studies from the greater number of studies that described panda habitat use, we characterized choices that pandas make when multiple habitats are available. Our synthesis indicated that giant pandas are more flexible in their habitat selection choices than previously thought, with this flexibility likely

related to the availability of preferred habitat components (see Garshelis 2000 for a discussion on habitat selection and preference). First order habitat selection (i.e., the geographic range; Johnson

1980) by giant pandas provides a perspective on the habitats that are available in their human- influenced landscapes. Habitat variables consistently selected at the geographic range included bamboo presence, forest cover, and areas not in close proximity to human communities. These variables are likely inter-related, with increased human activity corresponding to less bamboo and forest cover. Elevation was another variable that helped define panda habitat selection across their geographic range, but the range of selected elevations was variable and depended on location. In general, pandas have been relegated to steeper mid-elevations in many areas throughout their range because humans occupy the lowlands and in many cases have converted the habitats to development and agriculture. High availability of bamboo throughout the mid- elevations appears to support pandas at high densities in a variety of different habitat conditions.

Characteristics of topographic slope, bamboo, and human disturbance influenced habitat selection by pandas at mid-levels (i.e., home range and within home ranges; Johnson 1980).

Moderate and steep slopes have frequently been proposed as limiting panda habitat selection because it is energetically costly to traverse steeper mountainsides (Schaller et al. 1985; Liu et al.

1999). However, we found that pandas selected for a broader range of slopes at mid-levels than previously documented. The importance of slope was reduced in some multivariate models, likely because slope was correlated to other more important variables such as forest type or bamboo cover. Additionally, although the avoidance of steep slopes by pandas is consistent across studies, we also found consistent selection for gentle and moderate slopes that contrasts with how slope is typically represented (i.e., as monotonic and linear) in current habitat suitability models (e.g., Liu et al. 1999). Some studies in our review found stronger selection for

moderate slopes over gentle slopes but we caution that this finding potentially relates to interacting factors such as higher human disturbance on gentle slopes. We also found that selection for slope varied by bamboo species, season, forest type and study area. In the future, we recommend that habitat suitability models recognize that selected slopes include a wider range than previously modeled (0–30°), while areas with even steeper slopes might be useable according to conditional criteria (depending on other factors present).

Our results also point to nuances in panda habitat selection for other geophysical variables aside from slope. Slope and aspect relate to the amount of solar radiation striking a surface, a variable that was a positive predictor of panda selection in Liu et al. (2011). The authors hypothesized that low amounts of solar radiation could be a limiting factor for plant

(specifically bamboo) growth. Selection for ridges, upper and mid slopes likely reflects the combined effects of human activity in valley bottoms and preferred scent marking locations for pandas along easily traversable ridgelines (Schaller et al. 1985).

Our results on selection for bamboo across all levels of habitat selection confirm its importance for pandas. Our novel contribution to this previously well-documented relationship is that pandas selected for higher bamboo cover regardless of availability across the landscape.

Other habitat factors we investigated may only be important with respect to how they correlate with conditions suitable for bamboo occurrence, growth, and diversity. Wang et al. (2010) found that bamboo cover varied significantly with elevation and aspect but not slope, canopy cover, or position on the mountainside. Bearer at al. (2008) also found that bamboo cover was related to other habitat characteristics (e.g., elevation, overstory cover, slope) but the effects varied by bamboo species. Bamboo density is another related but significant variable, but pandas selected moderate densities in some areas and high densities in other areas. The former finding may

reflect the fact that extremely high density bamboo patches can be more difficult to traverse and also may contain less palatable food for pandas (Schaller et al. 1985). We caution that bamboo density and cover are likely confounded by bamboo age and diameter and suggest further exploration of these relationships.

Beyond consistent selection for bamboo, our findings indicate that broad generalizations on panda habitat selection are likely inappropriate and that researchers and managers should cautiously transfer findings from one study area to another. We found that pandas do not consistently select for specific tree structures at mid-levels, likely because bamboo occurs in forests with many different configurations of middle and overstory vegetation structure. In particular, canopy cover was a poor determinant of panda habitat selection according to our review, suggesting that suitable forest attributes cannot be detected using simple percentage cover measurements. Indeed, our results suggest that panda habitat selection for tree-related characteristics is likely context-dependent and related to complex interactions between tree- and bamboo-related variables (Taylor et al. 2004).

Our findings suggest that habitat selection results should be cautiously applied to management of giant pandas and their habitat. Current management can potentially benefit from improved prioritization of important habitat areas. For example, the creation of nature reserves, delineation of zoning schemes within nature reserves, population monitoring, and planning for potential future habitat restoration can all be improved through better understanding panda habitat selection. However, habitat selection results alone should not cause managers to discount habitats that are being used less than their availability (Garshelis 2000). For example, our review suggests that areas with steep slopes (>40°) and low bamboo cover (<50%) should be ranked

lower but not excluded from panda conservation activities, that is, these areas can be used by pandas and thus may contribute towards conservation.

Similarly, while some have recently advocated for an increased management focus on primary forests (e.g., Zhang et al. 2011), our results suggest that secondary forests in some locations play an important habitat role, particularly in landscapes where secondary forests are a dominant forest type. Although more study is needed to link the population demographics of pandas to primary and secondary forests, habitat restoration to connect fragmented patches of habitat using secondary forests may be worthwhile. In addition to our evidence that pandas select secondary forests in some study areas, other studies conducted on habitat use alone also demonstrate that pandas use secondary forests after sufficient time has passed (Pan et al. 2001;

Yang et al. 2006). The value of secondary forests for pandas likely depends on a complex combination of bamboo occurrence and growth, historical and current land use practices, and proximity to chronic human disturbances. Our finding that pandas did not select plantation forests is not surprising, considering that many replanted forests in the panda range are plantations of dense, exotic monocultures that do not support bamboo growth (Bearer et al. 2005;

Lu et al. 2007). Plantations designed to have lower tree densities may allow for bamboo growth and support pandas, but this requires further study.

Undoubtedly, the studies we reviewed have added considerably to the understanding of panda ecology. We note that our findings are influenced by differences and biases in sampling designs across studies and hence, generalizations on panda habitat selection should be cautiously used. The biggest shortcoming of studies surveyed in our review was that survey plots were often opportunistically located along pre-selected transects and that these transects served as established human and animal travel routes for monitoring. Such routes do not represent

randomly available habitats, potentially biasing data collection to areas that are accessible and easy to traverse. Additionally, few studies spatially or temporally replicated sampling for pandas and hence detection probability could not be estimated (McDonald 2004). These issues can be better addressed in the future by using rigorous design-based approaches that explicitly produce data that can be used to estimate detection probability. An effective design-based sampling strategy might involve systematically sampling among different strata of high and low panda density (as in Qi et al. (2009)). Probability of detection can be estimated a variety of ways including multiple survey methods on the same area (such as pairing telemetry with field surveys), mark-recapture methods, spatially replicated surveys in the same sampling area, or by double-sampling the same area using the same method but with different observers (McDonald

2004). We caution that areas of panda range that are difficult to access are under-sampled in the current literature because most studies are based on researchers encountering sign. Hence the scope of inference for most studies does not apply to a random sample of panda habitats.

Future Directions

Future work should more closely examine different types of human impacts on panda habitat selection, as only about 60% of the studies we reviewed looked at these effects. Many of these studies used distance-based GIS measures rather than more detailed field observations and hence the magnitude of human activity was potentially lost. Livestock grazing and tourism are two examples of emerging threats to the giant panda that specifically warrant future research. We also recommend that future work should better understand how habitat selection varies across mountain ranges; a topic better explored using a single, consistent dataset with a standard way of

defining habitat availability (e.g. used plots paired with an equal number of available plots located a set distance away).

We recommend that future panda habitat work use multivariate approaches that examine interactions among variables. One area of inquiry that warrants exploration is how giant panda habitat selection is affected by the interaction between distance to human communities and forest disturbances, as most forest disturbances occur closer to roads and villages. Also, the effect of spatial structure of habitat components on panda selection continues to be understudied. For example, the differences we noted between selection for cover in Fargesia and Bashania bamboo could be related to Fargesia growing in clumps that exhibit wide variation in cover across space, while Bashania is more uniformly distributed (making bamboo cover in an individual plot less important). Spatial structure of habitat components also differentiates habitat selection between the second- and third-order levels, which we aggregated in our review due to the limited amount of telemetry data that were available.

Lastly, we recommend that future panda habitat selection research incorporate density dependent effects. It is unknown how density affects giant panda habitat selection across space.

Density dependent effects are important for determining the extent to which pandas adapt to fewer available resources in competitive environments by altering their selection patterns.

Ultimately, the research community should work toward formulating an understanding of how habitat characteristics relate to panda fitness (Garshelis 2000). A better understanding of panda habitat selection processes is crucial for maintaining provision of panda habitats in the future.

APPENDIX

Vanderploeg and Scavia index Scavia and Vanderploeg -0.7 -0.8 0 5 10 15 20 25 30 35 40 45 >45 Slope (°)

0.4

0.3 0.2 0.1 Bearer et al. 2008 0 Kang et al. 2013 -0.1 Forestry-Gansu 2004 -0.2 -0.3 Reid and Hu 1991 -0.4 Tang and Hu 1998 -0.5 -0.6 Wei et al. 1996

Vanderploeg and Scavia index Scavia and Vanderploeg -0.7 Wei et al. 1999 -0.8 0 5 10 15 20 25 30 35 40 45 >45 Slope (°)

a. 1

0.8

0.6 slope (°)

0.4 0–5 / 0–10

5–20 Proporion used Proporion / 10–20

0.2 21–30 31–40 > 40 0 0 0.5 1 Proportion available

b. 1

0.8

0.6

bamboo 0.4 cover (%)

Proporion used Proporion 0–25 0.2 25–50 50–75 75–100 0 0 0.5 1

Proportion available

Figure 2.2 cont'd

Pan et al. 0.8 2001

Zeng et al. Kang et al. 2002 0.6 2011b * Bearer et al. 2008 0.4 * Wang et al. 2006

Proporion used Proporion * Ran et al. 2004b 0.2 * Ran et al. 2003 * Forestry- 0 Gansu 2004 0 0.5 1 Proportion available

Figure 2.3. Giant panda selection for bamboo cover across 6 studies. Bamboo cover was measured using visual estimation in fixed area 20 x 20 m or 30 x 30 m plots. 0.6

Bearer 2005 0.4 Forestry-Gansu 2004 0.2 Kang et al. 2011b 0 Kang et al. 2013 -0.2 Ran et al. 2003 -0.4 Wang et al. 2006

-0.6 Vanderploeg and Scavia index Scavia and Vanderploeg -0.8

-1 25 50 75 100 Percentage bamboo cover (%)

Table 2.1. Summary of studies on giant panda habitat selection reviewed from the English and Chinese language literature.

Study Locationb Area (km2)c Sampling date Data typed Use / availabilitye Plot size (m2) 1. Bearer 2005; Wolong NR 2000 2001 – 2003 F, G feces / no feces 30 x 30 Bearer et al. 2008 2. Feng et al. 2009 part of Qinling mntns. 5700 2006 – 2007 G sign / GIS 85 x 85

3. Forestry-Gansu 2004a Gansu Province 4000 1999 – 2001 F sign /no sign 20 x 20

4. Kang et al. 2011a; Wanglang NR 460 1997 – 2009 F sign / no sign 20 x 20 Kang et al. 2011b 5. Kang et al. 2013 Wanglang NR 320 2012 F feces / no feces 20 x 20

6. Liu 2001; Liu et al. Foping NR 290 1991 – 1995 F, G collar / GIS 10 x 10, 30 x 30 2005; Liu et al. 2011 7. Pan et al. 2001 part of Qinling mntns. 15 spring 1987 F sign / no sign Variablef

8. Qi et al. 2009; Qi et al. Liangshan mntns. 10,067 2005 – 2007 G sign / GIS 30 x 30 2011; Qi et al. 2012 9. Ran et al. 2003 Yele NR 200 June 2001 F sign / no sign 20 x 20

10. Ran et al. 2004a Xiaoxiangling mntns. 400 2001 F sign / no sign 20 x 20

11. Ran et al. 2004b Baoxing County 1700 2001 F sign / no sign 20 x 20

12. Reid and Hu 1991 Wolong NR 25 1986 – 1987 F feces / no feces 20 x 2

13. Tang and Hu 1998 Yele NR 24 April 1994 F sign / no sign 10 x 4

14. Wang et al. 2006 Baishuijiang NR 77 June 2005 F sign / no sign 20 x 20

15. Wang 2003 Liaoxiancheng NR 126 fall 2002 F sign / no sign 20 x 20

Table 2.1. cont'd

Study Locationb Area (km2)c Sampling date Data typed Use / availabilitye Plot size (m2) 16. Wang et al. 2010; entire panda range 160,000 2000 – 2001 G sign / GIS 250 x 250 Ye 2008

17. Wang et al. 2008 Pingwu county 5,959 1998 F, G sign / GIS 20 x 20, 30 x 30

18. Wang 2008; Minshan mntns 9,569 1999 – 2007 F, G sign / GIS 20 x 20, 30 x 30 Wang et al. 2009 19. Wei et al. 1996 Mabian Dafengding NR 25 1991 – 1992 F feces / no feces 20 x 2

20. Wei et al. 1999; Yele NR 25 1994 – 1996 F feces / no feces 20 x 2 Wei et al. 2000 21. Zeng et al. 2002; Wanglang NR 300 April 1998 F sign / no sign unkn. Guo 2003

22. Zhang et al. 2006; Fengtongzhai NR 20 2002 – 2003 F feces / no feces 20 x 2 Zhang et al. 2009

23. Zhang et al. 2011 Sichuan Province ~110,000 1999 – 2003 F sign / no sign 20 x 20 a abbreviated here and in other tables and figures; full author name is "Gansu Provincial Forestry Department of Wildlife Conservation" blocations of the study are expressed in terms of the most specific place identifier and include names of nature reserves (NR), counties, mountain ranges, or provinces. csize of study area corresponds to only the area sampled. dhabitat data were obtained from field survey (F) or a GIS (G). esign means all animal sign including feces, foraging site, footprint, hair, animal sighting; feces means feces only. flarge quadrats that were searched for signs; quadrats ranged from 5 – 29 ha

not considered in the study.

tion

Study DBH rub

Geophysical elevation slope Aspect shape slope position hillside water to distance index moisture index radiation solar Vegetation type forest metrics landscape area basal cover canopy height tree density tree dispersion tree DBH tree density stump tree density log fallen dispersion log fallen height shrub cover shrub density shrub dispersion shrub sh height understory presence bamboo species bamboo cover bamboo height bamboo density bamboo diameter bamboo shoot old proportion stem young proportion stem old proportion broken stems proportion bamboo dead proportion type growth bamboo condi bamboo Disturbance disturbance forest farmland grazing livestock collection herb poaching disturbance of severity road to distance village to distance Liangshan Qi et al. 2009; Qi et al. 2011; Qi et al. x x x x x x x

20121 Wei et al. 1996 x x x x x x Minshan Kang et al. 2011a; x x x x x x x o x o o x x x x x x x Kang et al. 2011b 1 o x2 o o x2 o x2 x x2 x o x2 x x x x x Kang et al. 2013 Zeng et al. 2002 x o x x x x x o o x Guo 2003 Wang et al. 2006 x x x x o x o x x x x x x

Wang et al. 20081 x o o x x x x x x

Wang 2008; Wang x o o o x x x x x x et al. 20091 Qionglai Bearer 2005; Bearer et al. o x x x o x x o x x x x x x x

20081 Ran et al. 2004b x o x x o x x x x x x x

Table 2.2 cont'd

to village to

oportion young stem young oportion

elevation slope Aspect shape slope position hillside water to distance index moisture index radiation solar Vegetation type forest metrics landscape area basal cover canopy height tree density tree dispersion tree DBH tree density stump tree density log fallen dispersion log fallen height shrub cover shrub density shrub dispersion shrub DBH shrub height understory presence bamboo species bamboo cover bamboo height bamboo density bamboo diameter bamboo shoot old proportion pr stem old proportion broken stems proportion bamboo dead proportion type growth bamboo condition bamboo Disturbance disturbance forest farmland grazing livestock collection herb poaching disturbance of severity road to distance distance Study Geophysical Reid and Hu 1991 x x x x x x Zhang et al. 2006; Zhang et al. x o o o o x2 o o o o x2 x2 x2 x

20091 Qinling Feng et al. 20091 x x o o x o x x o x x

Liu 2001; Liu et al. x x o o o o x x x x o x x x x 2005; Liu et al. 2011 Pan et al. 2001 o Wang 2003 x x x x x o x Lesser Xiangling Ran et al. 2003 o x x x x o o o x x x Ran et al. 2004a x x o o o Tang and Hu 1998 x x x x Wei et al. 1999; x x x x x x o Wei et al. 2000 Multiple mountains Forestry-Gansu x x x x x x o x x x x x x o x x 2004 Zhang et al. 20111 o x o x o x x x

Wang et al. 2010; x x Ye 20081

1 multivariate study 2 significant in univariate analysis only

Table 2.3. Habitat characteristics deemed important to giant panda selection by habitat selection level (sensu Johnson (1980)).

Level of Selection Habitat Characteristics Selection in multiple studies Selection in some but not all studies First-order forest land cover type gentle slope geographic range bamboo presence elevation far from human disturbance village/town/road/cropland

Second-order gentle/moderate slope aspect (orientation depends on study home range mid-slope/upper mountain area) Third-order high bamboo cover mixed forest within home range moderate/high bamboo density coniferous forest far from human disturbance old-growth forest village/town/road/cropland canopy cover tree height tree DBH high proportion old shoots

Fourth-order bamboo shoots taller bamboo resources younger bamboo culms thicker bamboo bamboo leaves larger diameter den trees bamboo species with highest nutrients

CHAPTER 3

SPACE USE BY ENDANGERED GIANT PANDAS

In collaboration with

Jindong Zhang, Shiqiang Zhou, Jinyan Huang, Rengui Li, Dian Liu, Weihua Xu, Yan Huang,

Zhiyun Ouyang, Hemin Zhang, and Jianguo Liu

Abstract

Studies on animal space use can reveal insights into how animals interact with one another and their environment. Research on the space use patterns of the endangered giant panda (Ailuropoda melanoleuca) in China has nevertheless lagged behind that of many other species, as a government moratorium prevented telemetry data collection on pandas from 1995 to 2006. We studied 5 giant pandas using GPS telemetry and estimated home ranges, core areas, and space use using model–based approaches. Panda home range was 6 km2 for the male and spanned 2.8 to 4.7 km2 for the 4 females. Pandas occupied several small core areas that they revisited after time lags of up to several months. Pandas also displayed significant space use interactions, especially among a male and 2 different females across several weeks during a fall season, a time of year not previously thought to involve extensive inter–panda interaction.

Introduction

Research on how animals distribute across space provides insights into resource allocation and intrapopulation competition for resources (Kernohan et al. 2001; Powell 2012).

Studies on animal space use can also inform conservation of endangered animals by revealing the spatial requirements for individuals, number of animals that a given area can support, and sensitivity of individuals to disturbances (Hull et al. 2011b; Macdonald and Rushton 2003; Viña et al. 2010).

The study of space use is important for the giant panda (Ailuropoda melanoleuca), endemic to China and the world’s most endangered ursid. Owing to an increase in human population size and associated human impacts (Chen et al. 2010; Liu and Raven 2010; Liu et al.

2013), the giant panda population is now limited to 1,600 individuals inhabiting a mere 21,300 km2 in over 20 fragmented, mountainous forests in southwestern China (Loucks et al. 2001; State

Forestry Administration 2006; Viña et al. 2008). As obligate bamboo foragers, giant pandas forage in mixed deciduous and coniferous forests with bamboo prevalent in the understory

(Schaller et al. 1985; Tuanmu et al. 2010; Viña et al. 2010). Pandas have no known major predators aside from humans and their daily activities are mainly structured around eating bamboo (for up to 14 hours per day) and sleeping for much of the remaining time (Schaller et al.

1985). Unlike many other ursids, pandas do not hibernate in winter. Pandas are largely solitary and are believed to interact with their neighbors outside of the yearly mating season mainly through scent communication (Schaller et al. 1985). Pandas deposit and interpret scent marks from their anogenital glands on trees dispersed throughout the habitat which convey information on their size, sex, estrus state, rank, and identity (Swaisgood et al. 2004).

Studying giant pandas is difficult because they are elusive, avoid humans and live in rugged, forested landscapes with dense vegetation and poor visibility (Hull et al. 2011a;

Linderman et al., 2006). Three telemetry studies from the 1980s and early 1990s with sample sizes of 5 (Pan et al. 2001; Schaller et al. 1985; Yong et al. 2004) and 22 pandas (Pan et al. 2001) documented small home ranges for mammals of comparable size (4–29 km2, minimum convex polygons or ellipses). Pandas did not defend spatial areas (as territories) and did not patrol the peripheries of their home ranges (Pan et al. 2001; Schaller et al. 1985). Home ranges were slightly larger in males than females, home ranges overlapped extensively, and pandas displayed a distinct seasonal pattern of a summer range and winter range separated by a few hundred meters in elevation. Home ranges expanded as pandas matured to adulthood, home ranges of females shrunk during years when they reproduced, male cubs inherited parts of their mothers’ home range as adults, males had a dispersed space use pattern for securing mates, and females had a concentrated pattern for rearing young (Pan et al. 2001).

This early research nonetheless suffered from low spatial accuracy and data constrained to daytime and good weather conditions. Then from 1995 until 2006, the Chinese government banned all telemetry of giant pandas due to concerns over animal safety (Durnin et al. 2004).

Using global positioning system (GPS) telemetry on pandas, Zhang et al. (2014) estimated

Brownian bridge home ranges of 8–11 km2 (n=4). Despite considerable spatial overlap among panda home ranges, individuals did not exhibit significant dynamic (spatio–temporal) interactions (Zhang et al. 2014). Zhang et al. (2014) did not investigate core areas (areas of concentrated use within the home range), a topic also understudied in previous literature.

We were afforded a rare opportunity to conduct a GPS collar study on giant panda space use, allowing us to explore several understudied aspects of giant panda space use. These included

estimation of home range, core areas, and space use interactions among pandas using novel data with a high spatial and temporal resolution. This study informs giant panda biology and conservation by revealing new insights into how these elusive animals occupy space.

Materials and Methods

Study Area and Study Animals

We studied pandas in Wolong Nature Reserve (102°52’ – 103°24’E, 30°45’ – 31°25’N),

Sichuan, China. Wolong is within a global biodiversity hotspot (Liu et al. 2003). The reserve takes up a 2,000 km2 area containing approximately 10% of the total giant panda population (Liu et al. 2001). Panda habitat consists of mixed coniferous and deciduous broad–leaved forests and subalpine coniferous forests (Schaller et al. 1985). The area within Wolong where the pandas were collared was called Hetaoping and located in the northeastern section of the reserve (Figure

3.1). Camera trapping and genetic testing of DNA extracted from feces collected throughout the study area suggest a total of 16–25 pandas present (in litt.).

We captured five free–ranging giant pandas, outfitted them with GPS collars, and released them at their capture sites in 2010 and 2011 (Table 3.1). We anesthetized pandas for short periods using weight–dependent doses of ketamine deposited via compressed air guns.

Staff members at the China Conservation and Research Center for the Giant Panda (CCRCGP) were responsible for the pandas’ safety. We used 12–channel Lotek GPS_4400 M GPS Collars

(Lotek Engineering Inc., Newmarket, Ont., Canada) to monitor the pandas. Collars weighed approximately 1.2 kg and recorded longitude, latitude, and elevation every 4 hours. Collars indexed activity (movement of pandas' heads along the X and Y axes every 5 minutes). Study animals included three adult females, one sub–adult (1.5–5 year old) female, and one adult male.

Research followed ASM guidelines (Sikes 2011) and was approved by an institutional animal use and care committee.

We excluded data collected within 1 week after capture to minimize effects of capture on behavior. Collars falling off pandas and collar damage limited data to 184–700 days/panda

(Table 3.1). Static testing on collars prior to deployment in a variety of habitat conditions revealed a fix acquisition rate of over 90% for each collar (n=30 habitat locations). Failure to record fixes could not be correlated to measured habitat characteristics (e.g. slope, forest cover).

Actual fix acquisition rate of collars while on the pandas was much lower (31–54%, Table 3.1).

Missed fixes were sporadic and intermixed with successful fixes at least once every 10 days and normally at least every 3 days for all pandas except the male, whose collar malfunctioned and did not record data for 2 lengthy periods (14 November 2011–25 December 2011 and 27 March

2012–6 May 2012). Position error of collars relative to a differentially corrected GPS unit averaged 16 to 23 m (n = 30 locations per collar).

Giant Panda Home Range Estimation

We estimated home range using a biased random bridge model (Benhamou 2011). We chose this model because it incorporates animal movement into the estimation and also includes an advection component for preferred locations and directions of movement (Benhamou 2011).

We set the parameters Tmax (maximum step duration), hmin (location uncertainty parameter), and Lmin (minimum distance between successive locations) as 36 hours, 10 m, and 20 m, respectively. The diffusion parameter (D) was chosen using the plug–in method (Benhamou

2011) and was taken as the average across all animals to allow for better inter–individual comparison (D = 0.85 m2/s). We also used the collars’ activity data to adjust the time from total

time elapsed to “active” time since the previous location (i.e. proportion of time with nonzero activity as measured by the collar activity sensor).

For this portion of the analysis and for the core area analysis, we limited data input to a 1 year period for those individuals with more than 1 year of data available (two adult females and the male, 11 April 2011–11 April 2012) to improve the comparability of models across individuals, but we also provide the full home ranges in the text for the reader's reference. We estimated home ranges on a grid size of 30 m (to roughly correspond to the available digital elevation model) and defined as that area encompassing the smallest 95% of the predicted utilization distribution (Laver and Kelly 2008). We also performed a surface area correction by overlaying home ranges (and core areas) on a digital elevation model (DEM) obtained from The

Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER, 29 m x 29 m resolution) and subsequently calculating the surface area using a triangulation method outlined by Jenness (2004). We compared our results to those obtained using the 95% minimum convex polygon (MCP) method for comparison to early panda studies.

Core Area Estimation

We defined core areas as areas within home ranges falling below the threshold where percent of home range area increases faster than the predicted probability of use (Seaman and

Powell 1990; Vander Wal and Rodgers 2012). We also calculated the proportion of the home range found in the core area, the number of distinct core areas, the number of revisits to each core area, and the time between revisits. We considered a revisit to have occurred if a panda left a core area and was gone for at least 7 days before returning (a period of time greater than all

gaps in fix acquisition except for one 10 day period for one female (Zhong) and one 42 day period for the male).

Spatial Interactions among Pandas

We calculated the proportion of animal i’s home range within animal j’s (Fieberg and

Kochanny 2005) and tested for spatial and temporal interaction (Minta 1992). We calculated the coefficients LAA and LBB, which compare the actual use by each animal (A or B) of their shared space to a predicted use of the shared space under a random use condition (no attraction or avoidance), given the proportion of the area of the home range occupied by the shared space. We also calculated the coefficient Lixn, which compares simultaneous presence/absence in the shared space versus solitary presence/absence. A χ2 test was used to make statistical inference about whether the frequency of the two pandas' use of the shared space differed from random. We used only data pertaining to time periods in which data were available on both individuals of a pair.

We also isolated instances when pandas were less than 200 meters from one another to pinpoint the timing of instances of potential direct social interaction. All data analyses were conducted in the R statistical computing software (R Development Core Team 2005), mainly using the

"adehabitatHR" package (Calenge 2011b). Minta's test was conducted in R using DITools (Long

2012).

Results

Giant Panda Home Range Estimation

The giant pandas’ biased random bridge home ranges spanned 2.75 to 6.04 km2 and averaged 4.38 ± 1.24 km2 ( ± SD) after performing surface area correction (Figure 3.2, Table

3.2). The male had the largest surface–area corrected home range size at 6.04 km2. For the 3 pandas who had more than one year of available data, home ranges calculated on the full datasets spanned 4.59 to 7.17 km2 and averaged 5.76 + 1.31 km2 after surface area correction. MCP home ranges were larger than surface area-corrected, biased random bridge home ranges (7.08 + 4.35 km2, Table 3.2).

Core Area Estimation

Pandas showed a slow increase in home range size with home range isopleth level

(Figure 3.3). The pandas extensively used relatively small areas and used large portions of their home ranges infrequently. Core areas encompassed the 66–69% predicted isopleth across pandas, but this amounted to only 21–34% of the total home range area (Table 3.2). With surface area correction, total core areas were 0.77 to 1.53 km2 divided among 16–39 separate cores (Table

3.2). Pandas revisited 1 to 10 of their cores for a total of 1 to 25 return visits (Table 3.2). The male revisited more core areas and had a much higher revisit frequency than the females (10 of

16 cores revisited, 25 return visits). Pandas revisited a core area after a time frame of 8–191 days

(average of 55.15 ± 58.58 days).

Spatial Interactions among Pandas

We found 17–35% total home range overlap among one adult female, the sub–adult female, and the adult male (Figure 3.3a). The greatest overlap was between the latter two pandas, as 48% of the sub–adult's 40% home range isopleth (area containing the top 40% of the cumulative probability distribution) was within the male's home range and 28% of the male's 40% isopleth was within the sub–adult female's home range (Figure 3.3a). In addition, the adult

female and sub–adult female overlapped with one another, as 28% of the sub–adult's 30% home range isopleth was within the adult female's home range (Figure 3.4a).

The male–female dyads formed by these same three individuals exhibited significant simultaneous attraction to shared space (Lixn in Table 3.3). Pandas were not attracted to the space they shared if the other individual in the pair was not also present (LAA and LBB were not positive and significant) and the male significantly avoided the space he shared with the adult female unless she was also present (LBB was negative and significant but Lixn was positive and significant). Significant simultaneous avoidance also occurred among two adult females. The majority of the 52 recorded inter–animal distances that were less than 200 m for the duration of the study occurred from late July to early October (Figure 3.4b). An adult female and sub–adult female were in close proximity to one another during a 2 week period in late August and the sub–adult female and adult male were in close proximity over the course of a 2 week period in late July and throughout the month of September.

Discussion

Panda home ranges are small relative to other terrestrial mammal species of comparable body size, including up to several hundred kilometers smaller than other bears (Garshelis 2004).

This pattern likely occurs because the panda’s low energy, yet abundantly available bamboo food source makes it advantageous to limit cost of travel, while maximizing intake at a given location

(Schaller et al. 1985). Nonetheless, our estimates of pandas' home range sizes continued to increase through 1 year of telemetry and did not reach an asymptote after 16 months for 2 of 3 pandas, suggesting that our study (and all previous studies that did not calculate beyond yearly home ranges) may not approximate the full home ranges.

This study is the first to use model–based approaches to define core areas for pandas. The core areas we described differ markedly from core areas identified in other species concentrating in a limited number of centralized or high quality locations (Powell 2000). The pattern of multiple core areas used by each panda is unique and reflects pandas’ unique foraging patterns, as they move from one patch of bamboo to the next, concentrating time in areas of adequate new bamboo shoots (Schaller et al. 1985). Our results suggest that core areas are twice the size predicted by Schaller et al. (1985) using a grid–based approach. Schaller et al. (1985) and Pan et al. (2001) both reported only female pandas using core areas, but Yong et al. (2004) reported core areas in both sexes. Our pandas revisited core areas, often after long absences (e.g. 6 months), suggesting that pandas have strong spatial memories (Tarou et al. 2004). Our male panda's more frequent revisits than females may be a mating strategy to monitor multiple females throughout the year (Schaller et al. 1985, Pan et al. 2001).

This study is the first to report significant dynamic space use interactions (those including both spatial and temporal components) among pandas. Such interactions are common in animals living in social groups (e.g. white–tailed deer (Bertrand et al. 1996), brown hyena (Miller 2012)), but not at the close distances we observed in species believed to be largely solitary (e.g. jaguar

(de Azevedo and Murray 2007), lynx (Poole 1995)). The pandas spent several weeks in close proximity to one another in the fall, an unexpected finding because the mating season is between

March and May. Anecdotal accounts have suggested a “pseudo” mating season in September and

October (Lan et al. 2003). The pandas may also belong to a family group, particularly the adult female and sub–adult female that were often observed together. Pan et al. (2001) noted unusually high spatial overlap between family group members even into adulthood, an otherwise

understudied phenomenon in this solitary species. We hope further GPS collar research on pandas affords larger sample sizes to further explore these findings.

APPENDIX

Figure 3.2. Giant panda 95% home ranges estimated using the biased random bridge model.

Figure 3.3. Space use within the home ranges for 5 giant pandas estimated using the biased random bridge model.

1 0.9 0.8 Animal Pairs 0.7 Zhong:Long 0.6

Long:Zhong 0.5 Zhong:Chuan HR 0.4 Chuan:Zhong 0.3 Long:Chuan Chuan:Long 0.2 0.1 0 20 40 60 80 Home range isopleth

b. 250

200

Animal Pairs 150 Zhong:Long Zhong:Chuan 100 Mei:Zhong

Long:Chuan Distance (m) Distance 50

0 1-Apr-11 1-Jul-11 1-Oct-11 1-Jan-12 Time period

Table 3.1. Summary of 5 giant pandas tracked with GPS collars in Wolong Nature Reserve, China. Pan Pan Long Long Mei Mei Zhong Zhong Chuan Chuan Sex female female female female male Age adult sub–adult adult adult adult Start date 18 April 2010 11 April 2011 18 April 2010 11 April 2011 11 April 2011 Days monitored 219 184 700 485 487 Total fixes recorded 507 458 1285 699 1588 Fix acquisition rate* 0.39 0.41 0.31 0.24 0.54 *Rates were higher for the 1 year period used for the first analysis in Table 3.2 (0.47 for Mei Mei, 0.30 for Zhong Zhong, and 0.87 for Chuan Chuan.

Pan Pan Long Long Mei Mei Zhong Zhong Chuan Chuan Home range (km2) Minimum convex polygon (MCP) 2.81 5.03 4.02 11.76 11.79 Biased random bridge movement model 2.29 4.02 3.23 4.05 5.08 Surface area corrected 2.75 4.73 3.68 4.71 6.04 Core area Size (km2) 0.67 0.86 0.96 1.33 1.30 Surface area corrected 0.77 1.23 1.10 1.53 1.5 Proportion core (%) 29.2 21.5 34.0 32.8 25.6 Isopleth volume of core (%) 66.5 69.3 66.0 68.7 67.3 Number of separate cores 16 34 31 39 16

Number of cores revisited 2 1 3 6 10 Total number of return visits 2 1 4 7 25 Days elapsed between revisits ( ± SD) 44.75 ± 32.41 80 68.5 ± 62.78 127.67 ± 84.56 32.54 ± 31.29 Area per separate core (km2, ± SD) 0.07 ± 0.09 0.03 ± 0.06 0.04 ± 0.09 0.03 ± 0.08 0.08 ± 0.17

Animal pair LAA LBB Lixn Mei–Pan –0.09 0.22 0.11 Mei–Zhong –1.05** 0.41** –0.76* Mei–Long 0.57 –0.53 –0.87 Mei–Chuan –– –– –– Zhong–Long 0.56** –1.32** –0.04 Zhong–Chuan 0.21 –0.9** 0.83** Long–Chuan –0.04 –0.02 0.74** *p<0.05, **p<0.01

CHAPTER 4

AN INDIVIDUAL ANIMAL PERSPECTIVE ON GIANT PANDA HABITAT USE AND

SELECTION

In collaboration with

Jindong Zhang, Andrés Viña, Jinyan Huang, Shiqiang Zhou, Hemin Zhang, Zhiyun Ouyang, and

Jianguo Liu

Abstract

Animals make choices about where to spend their time in complex and dynamic landscapes, choices that reveal information about their biology that in turn guides their conservation. We conducted a novel individual-based analysis of habitat use and selection by elusive and endangered giant pandas (Ailuropoda melanoleuca) monitored using GPS collars.

We constructed spatial autoregressive resource utilization functions (RUF) to model the relationship between the pandas' utilization distributions and various habitat characteristics across a continuous space. We subsequently extended our analysis to (a) habitat suitability models and (b) habitat selection analyses. The latter examined use with respect to availability of habitat types both within the home range and across the entire study area. Results reveal several new insights, including use of a broader range of habitat characteristics than previously understood for the species, particularly steep slopes and non-forest areas. Solar radiation (+) and elevation (-) were significant predictors of panda habitat use. Both solar radiation and slope were significant predictors of panda selection, as pandas selected against steeper slopes and lower solar radiation in proportion to their availability in both the home range and entire study area.

Our results have implications for modeling and managing the habitat of this understudied endangered species by revealing how individual pandas relate to habitat and make choices that differ from assumptions made in broad scale models. Our approach also informs studies of habitat use on other species by highlighting the value of a spatial autoregressive RUF approach in species for which a complete picture of individual-level habitat use and selection across space is otherwise lacking.

Introduction

The relationship between animals and their habitats is a central component of wildlife ecology (Morrison et al. 2006). One important area of research involves understanding the behavior of individual animals as they use habitats distributed across heterogeneous space

(Marzluff et al. 2004). Such studies often reveal a wealth of information that cannot be obtained by population-level surveys, including information on fine-scale variation over space and time

(Marzluff et al. 2004). Research has also been extended to the study of habitat selection, or the choice of habitats relative to their availability on the landscape (Manly et al. 2002). Such studies can inform conservation efforts of endangered species by revealing the full range of resources used by animals that may be missed by population surveys alone, while also pinpointing which types of habitats are selected in higher proportion to their availability, thus potentially warranting increased conservation focus (Hebblewhite and Haydon 2010; Schofield et al. 2007).

This is an important research topic for the endangered giant panda (Ailuropoda melanoleuca). Endemic to mountainous forests found in southwestern China, giant pandas are the most endangered ursid on earth. Pandas are a largely solitary species which have no known predators aside from humans (Schaller et al. 1985). Currently limited to a mere 21,300 km2 of estimated suitable habitat (Viña et al. 2008), the 1,600 remaining giant pandas have faced human threats including road construction, timber harvesting, tourism, and livestock grazing (State

Forestry Administration 2006). The remaining panda habitat is defined by the existence of bamboo, their main food source, which makes up over 99% of their diet (Schaller et al. 1985).

Bamboo occurs in mixed deciduous and coniferous forests in areas that are often rugged, with steep mountainsides and rapidly changing elevation.

Aside from bamboo, panda habitat suitability is most commonly defined by three variables- forest, slope, and elevation (Liu et al. 1999). Pandas use forests located at the mid- elevations that provide suitable conditions for bamboo growth (Schaller et al. 1985). Slope is also one of the most important habitat characteristics for pandas, as pandas use areas of low or moderate slope for energetically efficient traveling (Reid and Hu 1991; Schaller et al. 1985).

Pandas also use areas with high solar radiation, choosing warmer topographic aspects (Liu et al.

2011) in addition to areas farther from focal areas of human activity such as roads (Bearer et al.

2008) and villages (Wang et al. 2008) and in areas not recently subjected to timber harvesting

(Bearer et al. 2008). Pandas also select many of these same characteristics at a higher proportion than what is available (e.g. higher bamboo cover and higher solar radiation) or at a lower proportion than what is available (e.g. steep slopes, Chapter 3).

Despite this information, many gaps remain in our understanding of this elusive species.

Previous research was mainly conducted at the population level, with little known about habitat use and selection of individual giant pandas. This is in large part due to a government moratorium on all giant panda telemetry from 1995-2006 (Durnin et al. 2004), which limited the information available on behavior of individual pandas that otherwise avoid humans and hide in dense vegetation. As a result, information on habitat use and selection is derived from wildlife sign detected along transects sampled throughout the habitat. Such an approach, while valuable, leaves little appreciation for variation in intensity of habitat use and selection because it is derived as a binary (presence/non-presence) variable and cannot be designated to individual animals. This limitation was overcome in a recent study by Zhang et al. (2012) involving habitat use of GPS-collared giant pandas which found significant effects of several geophysical factors

(including elevation and slope). But several questions remain to be answered, including the

relationship between habitat use and both existing habitat suitability models and habitat selection.

In addition, there has only been one individual-level assessment of habitat selection in the species (Liu et al. 2011; Liu et al. 2005). This was a valuable work, but was derived from pairing radio telemetry locations (with limited spatial accuracy) to other locations in a reserve, leaving remaining questions about variation in habitat use across the home range and definition of available habitats.

To begin filling these knowledge gaps, this chapter reports the results of a novel GPS collar study examining habitat use and selection by individual giant pandas. We adopted a multivariate spatial autoregressive modeling approach to investigate the biogeophysical factors related to continuous predictions of habitat use by individual giant pandas. We also sought to (a) relate habitat use to existing habitat suitability models for giant pandas and (b) investigate habitat selection (use/availability) by individual giant pandas at within-home range and at-home range selection levels for the first time. This study generates new information on the biology of this endangered species, specifically by providing necessary individual context for understanding how pandas relate to their complex environments, in turn informing conservation by helping to prioritize management of the remaining limited panda habitat.

Methods

Study Area and Panda Subjects

The study area is located in Wolong Nature Reserve (102°52’ – 103°24’E, 30°45’ –

31°25’N), Sichuan, China. Home to 10% of the total wild giant panda population (Liu et al.

2001), the reserve contains ample forest stretching across mountains with steep slopes (beyond

50°, Schaller et al. 1985). The study was conducted in the northeastern portion of the reserve in

an area known as Hetaoping (Figure 4.1). Roughly 40 km2 in size and spanning an elevational range of 1,800 to 3,100 m, the study area includes mixed deciduous and coniferous forests and subalpine coniferous forests with bamboo dominating their understories. Common tree species include Chinese walnut (Juglans cathayensis), mono maple (Acer mono), hemlock (Tsuga longibracteata) and spruce (Picea asperata), while the main bamboo species are arrow

(Sinarundinaria fangiana), umbrella (Fargesia robusta) and Yushan (Yushania bravipaniculata) bamboo.

Camera trapping and genetic testing of DNA extracted from field-collected feces suggest that Hetaoping supports a local population of 16-25 pandas (Zhang et al. unpublished). Camera trapping also revealed the presence of other large mammals, including the tufted deer (Elaphodus cephalophus), serow (Capricornis milneedwardsii), and sambar (Rusa unicolor), although none are believed to be direct competitors with pandas for space and food (Schaller et al. 1985).

Hetaoping is also an area located between two villages to the southwest and northeast and a provincial highway to the west (Figure 4.1). The steep slopes prevent frequent visits from humans, but camera trapping revealed their occasional presence in the forest.

Five giant panda individuals were captured in 2010 and 2011 at Hetaoping, fitted with

GPS collars, and released (Table 4.1). Capturing was accomplished using anesthetization dart guns loaded with weight-dependent doses of ketamine. Animals were handled for short (~30 minute) periods. Staff members at the China Conservation and Research Center for the Giant

Panda (CCRCGP) were responsible for animal safety. The study pandas included 4 females and

1 male, all adults except for a sub-adult female.

Pandas were fitted with Lotek GPS_4400 M collars (Lotek Engineering Inc., Newmarket,

Ont., Canada). The collars weighed about 1.2 kg and recorded longitude, latitude, and elevation

once every four hours. Collars also measured activity (movement of pandas' heads along the X and Y axes every 5 minutes). Number of days of monitoring varied across individuals (Table 4.1) due to either the collar falling off of the animal or collar damage. Data collected within one week after an individual's release were excluded to minimize bias introduced by the capture/release event. Static testing on collars ahead of deployment in various habitat conditions revealed a fix acquisition rate of over 90% for each collar (n=30 habitat locations). Fix acquisition was not correlated to habitat characteristics (e.g. slope, forest cover). Fix acquisition rate of collars while worn by pandas was lower (30-70%, Table 4.1). This is likely due to animal behavior (e.g. antenna obstruction while panda was sleeping or eating). A successful fix occurred at least once every 10 days (and usually at least once every 3 days) for all pandas except the male, whose collar malfunctioned and did not record data for 2 longer periods (11/14/2011-12/25/2011 and

3/27/2012-4/11/2012). Positional error of collars compared to a differentially corrected GPS unit averaged 16 to 23 m across individuals (n= 30 locations per collar). Since pre-deployment testing showed no significant difference in location error of 2-D (3 satellites) versus 3-D (4 or more satellites) locations, we included both in the analysis. However, we excluded data for which the elevation estimate was inaccurate (measuring below 1,000 m, n= 11% of all observations), as these fixes appeared to also have inaccurate longitude and latitude measurements.

Estimating Panda Use

A utilization distribution (UD) is a probability density function representing a continuous prediction of an animal's frequency of use across space (Van Winkle 1975). We estimated the

UD of each individual by applying the biased random bridge (BRB) movement model to the

GPS locations (Benhamou 2011). This is a stochastic model composed of a biased random walk

in which the probability of being found at a particular location is dependent on the starting and ending locations and the time elapsed between them. An advection-diffusion component accounts for animals having a higher probability of drifting toward certain directions, given prior behavior. We estimated the probability density function using a circular bivariate normal distribution. The diagonal of the variance-covariance matrix of this model was the diffusion coefficient (D) (Benhamou 2011). Parameters Tmax (maximum step duration), hmin (location uncertainty parameter), and Lmin (minimum distance between successive locations) were set as

36 hours, 10 m, and 20 m, respectively. We used a standard diffusion parameter (D) chosen using the plug–in method by taking the average across all animals (D = 0.85 m2/s, Benhamou

2011). We also used the collars’ activity data to correct for “active” time since the previous location (i.e. only the proportion of time with nonzero activity measured using the collars' activity sensors, Benhamou 2011). We defined the extent of space use by each animal at the 95%

UD boundary (commonly defined as the animal's home range (Laver and Kelly 2008)). Home ranges of the individuals studied ranged from 2.8 to 6 km2 during the time period analyzed

(Chapter 2).

Habitat Characteristics

We examined 5 habitat characteristics relevant for habitat use of giant pandas. These included slope, elevation, forest cover, terrain position, and solar radiation. Slope, elevation, and forest cover are commonly used in giant panda habitat suitability mapping for the species.

Pandas are believed to use gentle slopes due to ease of travel and mid-elevation forested areas due to suitability for bamboo growing conditions (Liu et al. 1999). Topographic position has been hypothesized to be an important predictor of panda use in the past, with pandas using ridges

intensively for travel and scent communication (Ran 2003; Wang et al. 2006). Solar radiation has been hypothesized to be an important predictor of panda use, with pandas using warmer areas more intensively than cooler areas (Liu et al. 2011).

Slope, elevation, topographic position, and solar radiation were derived from a Digital

Elevation Model (DEM) acquired by the National Aeronautics and Space Administration’s

(NASA) Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER, 29 m resolution). Topographic position was calculated using the topographic position index (TPI), a measure of the difference between the elevation in a pixel and the average elevation in the surrounding pixels (we chose a 9-pixel neighborhood area) calculated using the Land Facet

Corridor Designer in ArcGIS (Jenness et al. 2012). Higher values represent mountain ridges and lower values represent valleys. Solar radiation was estimated using the Area Solar Radiation tool in ArcGIS (with a 200 m sky size and a year-long estimation using monthly intervals). The forest/non-forest layer was derived from a supervised classification (with an 82.6% accuracy) of

Landsat TM imagery (30 x 30 m resolution) acquired in 2007 (Viña et al. 2011). We did not include distance to human disturbance (road, household) as a predictor because it was highly correlated with elevation for some pandas and not meaningful for fine-scale analyses since human establishments are spatially segregated from pandas on the outskirts of the study area.

Habitat Use Modeling

We designed resource utilization functions (RUF) to examine predictors of habitat use within the home ranges of the panda subjects. The RUF approach involves characterizing the relationship between the utilization distribution (UD) of an animal and a set of spatially-explicit habitat characteristics in a regression model (Marzluff et al. 2004). A separate RUF model was

built for each panda individual but model results were later combined for a population-level analysis.

The type of RUF model we used was a simultaneous autoregressive model (SAR). We chose this model because our data were spatially autocorrelated, violating the assumption of independent observations that is required in an ordinary least squares (OLS) model. The SAR model accounted for spatial autocorrelation in the data by including a non-zero covariance structure to produce more accurate coefficient estimates (Lichstein et al. 2002). The form of the model was as follows:

y = A + ρW(Y-A) + ε

The response variable y was the predicted probability of use, W was a spatial neighbor matrix, A was the vector of independent variables related to previously mentioned habitat characteristics, and ρ was an interaction parameter indicating the amount of autocorrelation between neighboring points (Bailey and Gatrell 1995). ρ was defined by the inter-point distance over which neighborhood values were spatially autocorrelated (Lichstein et al. 2002). This distance was determined by visual interpretation of semivariograms which showed autocorrelation up to 400 m for all individuals except the male (900 m). The response variables were log10- transformed to meet assumptions. We tested multiple interaction terms among habitat characteristics but did not include them in the final model because they were not consistently significant across pandas. We tested for multicollinearity using the variance inflation factor (VIF) and found no significant multicollinearity (all VIF <3). We assessed model fit of the best model by plotting the actual UD to that predicted in the autoregressive model (since there is no R2 value in autoregressive models).

We combined model results for a population-level assessment of resource use using a method outlined in Marzluff et al. (2004). Partial regression coefficients were standardized to account for differences in measurement scale across individuals. Mean standardized regression coefficients were calculated as:

where is the partial regression coefficient, sxj is the standard deviation of the variable measured and sRUF is the standard deviation of the utilization distribution. We then tested the null

hypothesis that each differed from 0 using a t test.

Extensions to Habitat Suitability and Habitat Selection Analyses

We examined the relationship between the UD and three habitat characteristics commonly used in habitat suitability models for this species (slope, elevation, and forest cover).

We performed a spatial overlay to determine the proportion of each animal's UD in different habitat suitability classes according to a four-class habitat suitability scheme outlined in Liu et al.

(1999), i.e., unsuitable, moderately suitable, suitable and highly suitable.

We assessed the pandas' habitat selection according to a method proposed in Millspaugh et al. (2006) as an extension of compositional analysis (Aebischer et al. 1993). In this method, the habitat use of discrete habitat types is compared to the proportion of the habitat types that are available for the animal to choose from. We created 2-6 classes for each habitat variable by dividing the full range of the variable into equal intervals (for all except forest/non-forest, for which there were only two classes). We summed the UD values (up to the 95% UD) by habitat type to represent proportional use of each type by the animal. We calculated availability as the proportion of area taken up by each habitat type. We included two measures of availability- one

calculated over each animal's home range and one including the entire study area (defined by drawing a minimum convex polygon around all panda home ranges and buffering this region by

500 m). These corresponded to within-home range and at-home range selection (second and third order selection in Johnson (1980)).

We calculated and plotted the difference between the log-transformed used and available habitats across classes. We used the Wilks' lambda statistic to test for significant selection of each habitat characteristic and subsequently performed a ranking analysis on each habitat class.

We ran both analyses using randomization tests (n=500 runs). We performed an eigenanalysis on selection ratios to test for the assumption of similar selection patterns across individuals (as required for running the previous compositional analysis). All analyses were performed using the

R statistical computing software (R Development Core Team 2005) mostly using the

"adehabitatHS" package (Calenge 2011a).

Results

Elevation and solar radiation were significant predictors of habitat use among individuals

(Table 4.2). All pandas used low elevations more intensively than high elevations, and most pandas (except one adult female) used areas receiving higher solar radiation more intensively than low solar radiation. Slope, terrain position, and forest were not consistently significant predictors of habitat use among individuals. Across pandas, all models had a significant spatial autocorrelation component (all α significant at p<0.01) and were a significantly better fit from an ordinary least squares multiple regression model (∆AIC > 2). Plots of actual versus predicted use revealed that the fit of the models was generally good, but varied across pandas (Figure 4.2).

The majority of the giant pandas' utilization distributions were found within habitats predicted to be suitable habitat using current habitat suitability models for the species (Figure

4.3). Pandas used elevations completely within highly suitable and suitable elevation ranges

(1500-3250 m). However, 18-42% of their utilization distributions were in areas classified as non-forest, a habitat type typically deemed unsuitable for their inhabitance. In addition, 14-26% of their utilization distributions were found in areas previously deemed too steep to be suitable panda habitat (above 30°).

With regard to habitat selection, there was no statistically significant pattern of overall selection of any of the habitat variables, but several significant effects were seen at the individual class-level in the habitat ranking analysis (Figure 4.4). Non-forest was significantly preferred over forest at the within-home range selection level but not at the at-home range level. At both selection levels, areas of steep slopes and low solar radiation were not selected relative to others.

For elevation and terrain position index, significant differences were seen only among the smallest classes and others (and the largest classes and others for elevation only) but only at the at-home range level. Pandas selected against valleys and low elevation areas. The eigenanalysis on selection ratios revealed differences in habitat selection patterns across individuals, suggesting that the significance tests should be interpreted with caution.

Discussion

This study makes several new contributions to understanding habitat use and selection of the endangered giant panda. Although the sample size is small and results should be cautiously interpreted, this is a common challenge for rare and government-protected species facing restrictions on telemetry permits (Kramer-Schadt et al. 2004; Miller et al. 2010). Because pandas

are an elusive species whose habitat use has mainly been studied at the population level in the past, several new insights can be drawn from this study.

One of the main contributions of the study is the characterization of spatial variation in habitat use by individual giant pandas. Our approach allowed us to go beyond the typical binary response variable of used vs. non-used or used vs. available. By modeling panda habitat use across the entire home range, we found that pandas used a wider range of resources than previously appreciated or detected in transect surveys. For instance, moderate to severely steep slopes (over 30°) made up from 14 to 26% of the pandas' utilization distributions. In the past, these sloped areas have been labeled as "marginally suitable" to "unsuitable". This delineation has profound implications for modeling and management of panda habitat over large scales in that it may potentially result in an underestimation of available suitable habitat, while also leaving some steeper habitat unprotected if it is discounted by managers (as also discussed in

Chapter 2).

Our study also demonstrated the important distinction between habitat use and habitat selection with respect to topographic slope. Pandas did not use gentle slopes more than steep slopes. Slope was not significantly related to habitat use even in a model by itself with other variables removed. We did not find any consistent significant correlations or interactions between slope and other variables we investigated. It is therefore likely that pandas used some areas of moderate and steep slopes intensively because they were otherwise valuable to them, potentially due to aspects of the bamboo stands we did not measure. Nonetheless, pandas did select against steep slopes with respect to their availability in the home range, and to a greater extent, their availability across the whole study area. This avoidance of the steepest slopes likely relates to navigation challenges and potentially unavailability of bamboo in these extreme areas.

Results showing pandas used higher solar radiation and selected against lower solar radiation is in keeping with previous literature (Liu et al. 2011). Pandas likely prefer warmer areas because they allow for retention of body heat and thus maximization of thermal energy in the cool mountains or because warmer areas may support more bamboo growth. Our study adds a new angle to this previously documented phenomenon by demonstrating the importance of solar radiation in a multivariate model framework for the first time, showing its importance over several other key habitat variables.

The fact that pandas used low elevations with a greater intensity than high elevations likely reflects differences in panda foraging patterns on the low elevation umbrella bamboo compared to high elevation arrow bamboo, a phenomenon that has not been demonstrated using telemetry data before. Pandas gain a greater energetic return from foraging on the carbohydrate- rich umbrella bamboo (Schaller et al. 1985) and would therefore be predicted to forage longer in one location on this diet according to optimal foraging theory. In addition, umbrella bamboo has a more spatially clumped (patchy) distribution that promotes more intense localized foraging compared to the ubiquitous arrow bamboo found at the higher elevation. This spatial pattern may also in part explain the fact that the relationship between elevation and panda habitat use was opposite in sign from the relationship between elevation and habitat selection. In other words, pandas used lower elevation areas more intensively but used them in lower proportion relative to their overall availability in the study area. Pandas also spent more time at the higher elevation

(two thirds of the year), only coming down to the lower elevation for the umbrella bamboo shooting season.

Another important observation from this study was that the panda utilization distributions included a measurable amount of areas classified as non-forest, which goes against the prevailing

understanding of forest being a requirement for panda inhabitance (Liu et al. 1999). This finding is likely related to pandas making use of shrublands that still contain bamboo for consumption.

In fact, forest gaps can promote bamboo growth in otherwise good biophysical conditions due to removal of competition for resources from the overstory (Taylor et al. 2004). This finding also relates to the issue of spatial scale of measurement, as many of the non-forest areas found within the panda home ranges were small (e.g. one or a few 30 x 30 m2 pixels) and surrounded by forest.

The fact that non-forest was not selected at the home range selection level (only the within home range level) likely reflects the fact that there were unsuitable (and larger) non-forest patches present at the fringes of the study area associated with human communities. Use and selection of forest gaps and shrublands probably depends on a complex array of factors such as forest disturbance history and spatial pattern of habitat, complexities which require further study.

Further research should also be done on the ground to attempt to detect and quantify clandestine human activities such as hunting or herb collection that may occur throughout the habitat and affect panda habitat use and selection. Livestock grazing is another human impact that may warrant further study. Livestock grazing occurred in one portion of our study area but only overlapped minimally with the study pandas, thus making it difficult to draw robust conclusions about the effect of livestock on pandas using the modeling framework in this study.

However, the effect of livestock grazing across panda habitat in the reserve is further explored using a mixed methods approach in Chapter 6.

In addition to contributing to giant panda ecology and conservation, our study also informs research on habitat use and selection across other species. Our work demonstrates the utility of the RUF approach for species in which a clear understanding of the entire picture of habitat use and selection across the home ranges of individuals is lacking. The significance of

spatial autocorrelation in our models was also revealing. This component is often overlooked in

RUF models despite the fact that neighboring points are not independent. We recommend an autoregressive approach for other studies using this framework for more accurate model specification. Our study also demonstrated that establishing a nuanced understanding of habitat use and selection by individual animals across space can be valuable for clarifying models and assumptions made about animal-habitat relationships at broader scales.

APPENDIX

Figure 4.1. Hetaoping study area for giant panda GPS collar research in Wolong Nature Reserve, China. Forest layer is derived from supervised classification of Landsat TM imagery (Viña et al. 2011).

Figure 4.2. Fit of simultaneous autoregressive (SAR) models depicted by actual versus predicted log-transformed response variables (utilization distributions) for each panda.

Table 4.1. Summary of study pandas and GPS collar performance over the one year period included in this study. Pan Long Mei Zhong Chuan Sex female female female female male Age adult sub-adult adult adult adult Start date 4/18/2010 4/11/2011 4/18/2010 4/11/2011 4/11/2011 Days monitored 219 184 365 365 351 Total fixes recorded 507 458 961 458 1473 Fix acquisition rate 0.39 0.41 0.47 0.30 0.70

Table 4.2. Contribution of habitat variables to predicting giant panda habitat use across their utilization distributions (n=5). factor standardized 95% CI P ( = 0) # pandas + - elevation -0.17 -0.23 -0.11 0.001 5 slope 0.03 -0.00 0.05 0.07 3 TPI 0.11 -0.03 0.24 0.10 4 1 solar 0.08 0.01 0.16 0.04 4 forest -0.02 -0.05 0.00 0.07 3

CHAPER 5

EVALUATING THE EFFICACY OF ZONING DESIGNATIONS FOR PROTECTED

AREA MANAGEMENT

In collaboration with

Weihua Xu, Wei Liu, Shiqiang Zhou, Andrés Viña, Jindong Zhang, Mao-Ning Tuanmu, Jinyan

Huang, Marc Linderman, Xiaodong Chen, Yan Huang, Zhiyun Ouyang, Hemin Zhang, and

Jianguo Liu

Abstract

Protected areas worldwide are facing increasing pressures to co-manage human development and biodiversity conservation. One strategy for managing multiple uses within and around protected areas is zoning, an approach in which spatial boundaries are drawn to distinguish areas with varying degrees of allowable human impacts. However, zoning designations are rarely evaluated for their efficacy using empirical data related to both human and biodiversity characteristics. To evaluate the effectiveness of zoning designations, we developed an integrated approach. The approach was calibrated empirically using data from

Wolong Nature Reserve, a flagship protected area for the conservation of endangered giant pandas in China. We analyzed the spatial distribution of pandas, as well as human impacts

(roads, houses, tourism infrastructure, livestock, and forest cover change) with respect to zoning designations in Wolong. Results show that the design of the zoning scheme could be improved to account for pandas and their habitat, considering the amount of suitable habitat outside of the core zone (area designated for biodiversity conservation). Zoning was largely successful in containing houses and roads to their designated experimental zone, but was less effective in containing livestock and was susceptible to boundary adjustments to allow for tourism development. We identified focus areas for potential zoning revision that could better protect the panda population without significantly compromising existing human settlements. Our findings highlight the need for evaluating the efficacy of zoning in other protected areas facing similar challenges with balancing human needs and conservation goals, not only in China but also around the world.

Introduction

Human activities have caused massive losses in Earth’s biodiversity over the last few centuries in what has been termed “the sixth extinction” (Leakey and Lewin 1996). The current rate of extinctions attributed to humans is markedly above background rates observed in the fossil record (Barnosky et al. 2011). The underlying causes of this phenomenon lie in the cascading effects of anthropogenic activities such as habitat destruction, overharvesting, invasive species, and greenhouse gas emissions (Diamond 2005; Pimm et al. 1995). Today, few ecosystems are untouched by humans, such that they can be conceptualized as coupled human and natural systems (CHANS) in which the human and natural components are intricately linked

(Liu et al. 2007b). Appreciating the interactions, feedbacks, heterogeneity, thresholds, and surprises that arise in CHANS helps to better understand, model, and derive management recommendations for complex systems across the globe (Liu et al. 2007a).

Alarming trends in ecosystem degradation have inspired multi-faceted conservation initiatives over the last few decades, one of the most salient being the establishment of protected areas, or set aside areas for biodiversity conservation where human activities are limited or controlled (DeFries et al. 2007). There are now over 100,000 protected areas across the world, covering nearly 13% of the global land area (Jenkins and Joppa 2009; WDPA 2009). Success of protected areas in achieving conservation goals has been mixed, as not all have been able to function effectively in an increasingly human dominated world (Andam et al. 2008; Babcock et al. 2010; Liu et al. 2001; Wittemyer et al. 2008).

Over the last two decades, there has been a noticeable shift in the design and conceptualization of protected areas across the world (Naughton-Treves et al. 2005). Once seen as areas of pristine wildlife habitat where no human impacts should be allowed, protected areas

100

are now increasingly designated as areas where multiple-use activities occur which include, but are not limited to, biodiversity conservation (Geneletti and van Duren 2008; Gonzales et al. 2003;

Naughton-Treves et al. 2005). This shift occurred as a result of both the recognition of the reality of increasing human presence in once untouched ecosystems and the realization of the ethical dilemma involved in removing basic resource access rights to rural poor communities living in biodiversity hotspots (Naughton-Treves et al. 2005).

Because the Earth has become occupied by increasingly industrialized societies, there is a lack of space available worldwide to accommodate all goals and there is a need to more strategically designate the spatial extents of competing activities. The increasing prevalence of multiple-use protected areas across the globe has prompted managers to initiate careful design strategies founded upon zoning schemes that designate specific areas along a gradient from fully off-limits to humans to fully-available for multiple human activities (Geneletti and van Duren

2008). Zoning is increasingly being used in design of one of the most common types of protected areas- nature reserves. Zones may be set up as part of larger-scale biodiversity planning over areas that encompass a number of once singularly-managed small nature reserves, with areas bordering reserves designated as buffers between reserves and neighboring human development zones (Eigenbrod et al. 2009). It has been argued that in an increasingly human- dominated world, zoning designations serve an important purpose in mitigating conflicts between competing uses for limited resources by establishing guidelines for multiple use of shared space (Hjortso et al. 2006; Sabatini et al. 2007).

Zoning has featured prominently throughout the scientific literature on protection of marine reserves for designation of specific areas for varying levels of fishing or recreation in diverse ocean systems across the world (Agardy 2010). Surprisingly, discussions on zoning are

101

less prominent in comparable literature on terrestrial reserves (Geneletti and van Duren 2008).

Across both terrestrial and marine systems, a variety of tools have been developed to design zoning schemes, including Marxan-based decision support tools (Watts et al. 2009), simulated annealing (Sabatini et al. 2007), and spatially-explicit ecosystem-based tools (Salomon et al.

2002). However, there are much fewer examples of empirical studies investigating the efficacy of existing zoning schemes, including recent studies on the marine ecosystem in the Great

Barrier Reef Marine Park (Kenchington and Day 2011) and large mammals in a Central African protected area system (Remis and Kpanou 2011).

The main criticism of zoning designations is that there is no clear mechanism to operationalize them on the ground, thus they often become “paper maps” that exist in management plans but have no meaning with respect to realized activities (Sabatini et al. 2007).

This can occur especially in developing nations, where protected areas lack funding and personnel to design and enforce guidelines about proper activities for each designated zoning area (Sabatini et al. 2007).

Challenges in designing and enforcing zoning schemes are particularly significant in

China, the world’s fastest growing economy during the past three decades that also has seen explosive growth in the number of nature reserves (Liu and Raven 2010; State Forestry

Administration 2006). Nature reserves in China are set up such that many encompass areas already inhabited by rural human communities (Jim and Xu 2002). The presence of humans in and around the protected areas has often threatened their effectiveness, in some cases causing ecosystem decline despite the protected status (Harris 2008; Liu et al. 2001). Faced with the challenge of balancing development and conservation needs, the government mandated that all nature reserves in China be divided into three zones: core, buffer, and experimental (The State

102

Council 1994). This scheme appears to be largely inspired by the same configuration initiated by

UNESCO and the IUCN for biosphere reserves, areas of regional-based management of resources that bisect human and natural areas and are designed to promote both biodiversity conservation and sustainable development (Batisse 1997; McNeely 1994). The core zone is designed to protect natural ecosystems and the experimental zone is set up to allow for human development, with the buffer zone positioned in-between the two in order to soften the impacts of humans on natural ecosystems (McNeely 1994; Yu and Jiang 2003).

In practice, however, many reserves in China have not followed this mandate, considering that some reserves lack buffer zones, while others place buffer zones in locations that are not in-between the core and experimental zones (thus defeating their purpose, Liu and Li

2008). Even some reserves that do follow the specified 3-zone framework do not design the zones according to the stated conservation goals, because the designations are based solely on proximity to human settlements, as opposed to wildlife habitat quality or suitability (Jim and Xu

2004; Liu and Li 2008). In addition, the mandate may actually induce further environmental degradation, as some reserves with no human populations within their borders have designated an experimental zone for future development (Liu and Li 2008). Spatial context of zoning that extends beyond individual protected areas is also not adequately considered, as the core zone of one reserve may not be contiguous with core zones of neighboring reserves (Xiao et al. 2011).

Furthermore, there has not been a concerted attempt to evaluate the efficacy of existing zoning designations in China’s reserves for meeting their intended goals of balancing human activities with biodiversity conservation. Here we attempt to fill this gap using the world- renowned Wolong Nature Reserve, as a case study. Wolong is an ideal case study for examining this issue because it is a flagship nature reserve that other protected areas across China look to as

103

an example when shaping future policy (Liu et al. 2001). We specifically focused on the effectiveness of the zoning scheme for this reserve with respect to conserving the endangered giant panda (Ailuropoda melanoleuca, TSN: 621845), an animal revered as a national treasure for which more than 60 nature reserves have been established across its geographic range (Viña et al. 2010). The panda is ideal for such an analysis because it is a flagship species that garners significant conservation attention (Liu et al. 2001) and may also be considered an umbrella species, since its habitat encompasses areas with among the greatest biodiversity per unit area in the world (Mackinnon 2008). We analyzed the spatial distribution of both pandas and human impacts across the different zoning designations in the reserve. We then proposed adjustments of the zoning boundaries that would better meet giant panda conservation needs without significantly compromising existing human settlements. We discuss the role of zoning in the greater toolbox of conservation approaches and explore the implications of the findings for protected areas across the world that face similar and growing challenges of balancing human needs and biodiversity conservation.

Materials and Methods

Study Area

The study area is Wolong Nature Reserve (102°52’ – 103°24’E, 30°45’ – 31°25’N,

Figure 5.1), Sichuan, China. Established in 1975, the reserve consists of a 2,000 km2 area that supports approximately 10% of the total giant panda population (Zhang et al. 1997). Aside from giant pandas, there are over 2,200 animal species and around 4,000 plant species that are found within the reserve (Tan et al. 1995), a high level of biodiversity that is related to the large elevation range spanning 1,200 to 6,250 m (Schaller et al. 1985). The topography in the reserve

104

is characterized by steep cliffs alternating with narrow valleys, with slopes exceeding 50 degrees in some places (Schaller et al. 1985). At the elevation ranges where giant pandas are most commonly found (2,000- 3,300 m), the habitat consists of mixed coniferous and deciduous broad-leaved forest and subalpine coniferous forest (Schaller et al. 1985).

There are also nearly 5,000 local residents living within the reserve. The residents are mostly ethnic minorities of predominantly Tibetan (in addition to Qiang) descent who partake in farming-based lifestyles (Ghimire 1997). The residents interact with the natural environment in ways that impact the giant pandas and their habitat, mainly via land cultivation, animal husbandry, timber harvesting, fuelwood collection, and medicinal herb collection (Liu et al. 1999;

State Forestry Administration 2006). There is also a provincial-level road running through the reserve (303) which supports various forms of transportation of goods and people, and in so doing fuels the local economy.

The reserve is managed under the Wolong Administration Bureau, which is presided over by both the State Forestry Administration and the Sichuan provincial-level government. The earliest zoning scheme for Wolong was conceptualized in the late 1970s, as reserve officials differentiated between two distinct regulatory zones- one for human development and one for protection of nature (Wolong Nature Reserve 2005). The regulation called for the complete relocation of one of the villages in an attempt to contain human development, a part of the plan which never materialized (Ghimire 1997). A more integrated version of the zoning scheme was later formalized as part of the “Wolong Nature Reserve Master Plan” released in 1998. The zoning scheme is designed to help achieve the objectives of the Reserve, which were first identified as focused “mainly [on] the protection of [the] giant panda, other valuable rare animals and plants, and the typical natural ecosystem there” (Ministry of Forestry 1998: IX) in addition

105

to “meet[ing] the country’s modernization construction and sustainable development strategies”

(Ministry of Forestry 1998: X). The zoning scheme includes the core, buffer, and experimental zones according to the national guidelines for reserve planning. No human activity is permitted in the core zone and limited (but not clearly defined) human activity is allowed in the buffer zone, while human development is allowed in the experimental zone. While the specific methodology behind zoning boundary designation is not explained in the plan, the factors that went into zoning considerations included tourism, agriculture, distance to roads, elevation, wildlife, vegetation, scientific research activities and "specific regulatory rules" (Ministry of Forestry

1998). Parameters were assigned to 2 x 2 km cells drawn across the reserve with respect to each factor and were then combined in a clustering algorithm to generate the final zoning designations

(Ministry of Forestry 1998, Figure 5.1).

The zoning designation was largely based upon proximity to the provincial road that cuts through the reserve. The provincial road outlines areas where human development has already taken place and will be permitted to continue (mainly areas closest to the road). In the original zoning designation, the majority of the reserve consisted of core zone (1,416 km2, 70% of total), followed by buffer (434 km2, 21% of total), and experimental (183 km2, 9% of total, Figure 5.1).

The buffer zone located in between the core and experimental zones was notably narrow in some places, with 23% of the distances in between the core and experimental zones being less than

500 m.

There have been minor adjustments to the zoning boundaries in Wolong over the years, but the overall shape and distribution remains the same (Figure 5.1). A revised zoning scheme was drawn up after the May 12, 2008 Wenchuan Earthquake in order to account for post- earthquake reconstruction, with all reconstruction designated to be contained in the experimental

106

zone. The revisions resulted in little change from the original, with only a 0.4% decrease in buffer zone and 0.4% and 0.2% increases in experimental and core zones, respectively. These changes amounted to slight revisions in zoning designations at three locations throughout the reserve (Figure 5.1). One change involved the conversion of a long valley from experimental to core zone after tourism development was deemed infeasible (middle circle), while two other changes involved a slight extension of buffer and experimental zones to allow for planned future tourism development (right and left-most circles). Unless otherwise stated, we used the original

1998 version for analysis in the study, as it is the one that has been operational for the majority of the time period evaluated.

Zoning and Pandas

The distribution of pandas with respect to the zoning scheme in Wolong was evaluated using three complementary approaches: (1) assessment of panda habitat suitability across zones,

(2) spatial overlay of panda census data (derived from fecal counts obtained during transect surveys) and zones, and (3) summary of behavior of two individual GPS-collared giant pandas with respect to zones. The three different approaches were used in order to strengthen the assessment of the relationship between zoning designations and pandas, considering that each approach has different strengths and limitations. The first approach (habitat suitability index) is informative in that it provides a broad-scale assessment of potential habitat. The second approach (panda census) is a more direct measure of panda habitat use than the first and comprises the most comprehensive dataset in existence on panda distribution across the reserve.

The third approach (GPS collar study) captures the behavior of individual pandas in a

107

temporally-explicit way that has rarely been demonstrated before. Together, the three approaches provide a comprehensive picture of panda distribution with respect to the zones.

For the first approach, we used a giant panda habitat suitability index established by Liu et al. (1999) and reported in Liu et al. (2001). This index was a multiplicative function of three main environmental variables (elevation, slope, and forest) that contribute to habitat suitability for pandas. Specifically, pandas prefer gentle slopes, moderate elevation ranges and forested areas (Liu et al. 2001). Elevation and slope were derived from a digital elevation model (DEM) of Wolong (90 x 90 m resolution), while forest was derived from land cover classifications of

Landsat Thematic Mapper (TM) images (30 x 30 m resolution) acquired in 1997 (pre-zoning,

Liu et al. 2001; Viña et al. 2007). We summarized habitat suitability index by zone by determining the percentage of pixels in each habitat suitability class in each zone.

For the second approach, we used data from the 3rd National Giant Panda Census, the most recent comprehensive evaluation of the distribution of giant pandas across their geographic range, which was conducted from 2000-2004 (State Forestry Administration 2006). We used all observed locations of panda signs from the survey that took place in Wolong (in May and June of 2001), which represent observations of panda presence (but are not tied to individual pandas, n= 487). This is the only comprehensive dataset for panda distribution available across the entire

Reserve and it is currently the standard used for panda management assessment and decision making by central, provincial and local governments. This dataset is likely biased to more accessible areas that are easier to search. In other words, there was probably a higher proportion of the experimental and buffer areas searched for panda signs as opposed to the more remote core areas. To roughly define the known distribution pattern of giant pandas in the reserve, we generated a bivariate normal kernel density contour (Bailey and Gatrell 1995, bandwidth h=1000

108

m). The kernel allowed us to delineate a polygon with a boundary encompassing 95% of the panda signs. We chose the 95% contour because it would provide a conservative estimate of habitat use while capturing the majority of the distribution area. A spatial overlay was conducted to determine the density of panda signs found within the portion of the kernel in each of the zoning designations (number of panda signs divided by the kernel area).

For the third approach, we used data acquired by global positioning system (GPS) collars

(12-channel GPS_4400 M, Lotek Engineering Inc., Newmarket, Ont., Canada) placed on two wild, female pandas (Mei Mei and Pan Pan) in Wolong. Although this is a small sample size, this is one of the first GPS collar studies and one of the first panda tracking studies done on the species since radio tracking studies were conducted in the 1980s and early 1990s. We obtained special permission to collar only a small number of pandas, marking the end of a 15 year-long government-enforced ban on all telemetry of giant pandas. The collars were programmed to collect GPS fixes every four hours during the period of April 18, 2010 to April 12, 2011 (Mei

Mei) and April 18, 2010- November 25, 2010 (Pan Pan, had a shortened time frame since collar fell off). We used data acquired at least one week after the pandas were collared, in order to reduce bias from the possible effect of the collaring event on panda behavior. Fix acquisition rates of the collars were 44% and 40%, for Mei Mei and Pan Pan, respectively. Considering the short dispersal distances for this species (normally less than 500 m straight distance a day), we do not believe that loss of fixes created a bias with respect to spatial distribution across zones.

Field testing against a differentially-corrected GPS unit revealed that the locations recorded by the GPS collars were 95% accurate within a distance of 60 m. The locations from the collars were overlayed with zoning designations and summarized with respect to presence inside, as well as distance from the nearest zone. The area where the pandas were found was not

109

influenced by the zoning adjustments that occurred in the Reserve and was also not significantly influenced by the earthquake that occurred in the area in 2008.

Zoning and Human Impacts

Human impacts were analyzed with respect to the zoning designations in Wolong by characterizing (a) human activities and (b) forest cover dynamics across space. We adopted a spatial overlay framework for the majority of the analysis because we felt this approach would best integrate the available human and panda-related data with the spatial configuration of the zones. This analysis provided a means to evaluate the suitability of the design of the zoning scheme (spatial component), but also in some instances provided a means to evaluate the effectiveness of zones once put in place (temporal component). Human activities included roads, houses, tourism facilities, and livestock. Forest cover dynamics were analyzed with respect to forest cover change over time (which could reflect a combination of timber harvesting, fuelwood collection, forest monitoring, reforestation, and afforestation).

With regard to human activities, any paved surface accessible to four-wheeled vehicles was considered a road. Roads were all established prior to the zoning designation and no new roads have been built since then. Houses were drawn from the 2001 Wolong household survey

(to be generally consistent with 1998 zoning designations) and house locations were recorded using a GPS unit in 2002. We also analyzed data on house locations from 2006 as a second post- zoning time point to determine whether houses were effectively contained within the experimental zone over time. Tourism facilities included any infrastructure built for tourism activities (which were all built after the zoning designation in 1998). We obtained georeferenced locations of each type and calculated the percentage of roads, houses, and tourism facilities

110

located in each zone. We also calculated the distance of each house, tourism facility, and 100 m stretch of road to the nearest core zone using the Proximity Tool in ArcGIS 9.2 (ESRI 2006).

Livestock was assessed post-zoning only using three domestic horse herds owned by local residents and maintained in three separate areas in the Reserve. While this constitutes a small snapshot of the effects of livestock across the reserve, we illustrate it here as a case study to inspire further inquiry. The herd monitored most intensively (Herd 1, n=22) was monitored from July 2010 to April 2011 using a GPS collar fitted on a member of the herd. A second herd

(Herd 2, n= 15) was monitored in the same manner for a shorter period from June 2011 to July

2011. The collars were identical to those used on the pandas (see section 2.2 for collar description). The third herd (Herd 3, n= 16) was monitored by fied surveys only, through the establishment of 5 transects running through a roughly 1 km2 area where we observed them to roam over the previous one year period. We recorded the presence or absence of any horse sign

(feces or eaten bamboo) in 30 x 30 m plots every 100 meters along these transects. We then summarized percentage of GPS points (Herds 1 and 2) or field plots with horse presence (Herd 3) across zones.

With respect to forest cover change, Landsat TM images of the reserve in 1974, 1997, and 2007 were classified into forest and non-forest covers using supervised and unsupervised classification approaches (for details see: Linderman et al. 2005; Liu et al. 2001; Viña et al.

2007). The year 1974 was the year prior to the Reserve establishment and the year 1997 was the year prior to the zoning designation. We included the 1974 time point in order to provide historical context for forest cover dynamics in this reserve and across the as-yet undesignated zones. Previous studies have shown marked declines in forest cover in Wolong from 1974 to

1997 and then some degree of recovery after 2001 (Liu et al. 2001; Viña et al. 2007; Viña et al.

111

2011). Recent recovery is associated with the implementation of two conservation/restoration programs: the Natural Forest Conservation Program (NFCP) and the Grain-to-Green Program

(GTGP, Viña et al. 2011). However, these changes have not been analyzed with respect to differences across zones. We analyzed forest change across zones by performing a spatial overlay of zoning designations and forest cover.

Proposed Zoning Revisions for Panda Conservation

We set out to identify areas in the current zoning scheme that could be redrawn for the specific purpose of improving the conservation of the endangered giant panda. We used the 95% kernel of panda signs obtained from the latest panda census (generated in part 2.2) to represent the main area occupied by pandas in the reserve (panda presence layer). We chose to use the kernel for the panda presence layer as opposed to the actual census points because we believe that the buffered region around known panda locations helped to account for potential movement of the animals over space. We then generated a corresponding spatial extent for human establishments in the Reserve (human presence layer). To create this human presence layer, we combined spatial locations of roads, tourism facilities, and houses with buffers of different distances around them. We created a 200 m buffer around all roads, which matched the width of the experimental zone along the road in the existing zoning designation. This distance seemed reasonable because in the case where roadsides were not lined by existing human establishments, the steepness of mountainsides on either side of the road made human presence drop off over short distances. We chose a larger buffer of 500 m around tourism facilities and households to account for activities such as farming and infrastructure development that could comprise larger areas.

112

We then overlayed the panda presence layer, the human presence layer, and the zoning boundaries layer (in this one instance we used the most recent 2009 version). The main purpose of proposing a potential revision was to identify specific areas of experimental and buffer zone that were contained within the panda presence layer but currently outside of the human presence layer. In other words, we sought to identify areas that could be better protected by the zoning designation (i.e. converted from experimental to buffer or buffer to core) for panda conservation without significantly compromising existing human settlements.

Results

Zoning and Pandas

The zoning scheme in Wolong was not distributed in a way that maximized protection of the endangered giant panda in the core zone. During 1997 (the year prior to zone designation), around 54% of highly suitable habitat laid outside the core zone (40% in the buffer and 14% in the experimental zones, Figure 5.2a). A similar distribution across zones was found for suitable panda habitat (47%, 41%, and 12% in core, buffer, and experimental zones, respectively). The core zone contained a high percentage of unsuitable habitat (comprising 78% of all unsuitable habitat). The reason for this distribution is that the core zone included a large area of high elevation, i.e., non-forested areas above the tree line that are not considered panda habitat. If we isolate just the elevational range at which pandas are primarily found (2,000-3,300 m, Schaller et al. 1985), we find that the remaining unsuitable habitat is more evenly distributed across zones

(Figure 5.2b). However, the overall pattern of distribution of habitat suitability classes remains otherwise the same, with a considerable amount of suitable habitat existing outside of the core zone.

113

The data on panda occurrence supports the general pattern of pandas being not limited to just the core zone. The panda presence data obtained from the 3rd National Giant Panda Census

(State Forestry Administration 2006) showed that 58% of panda signs were found in the core, 38% in the buffer, and 5% in the experimental zone (Figure 5.3). Considering the kernel representing estimated panda distribution area in Wolong, there was nearly equal density of panda signs found in the core and buffer zones (1.03 and 1.01 signs per km2), with about half the density in the experimental zone (0.46 signs per km2).

With respect to the GPS-collared pandas, both pandas were also not limited to the core zone and in fact spent most of the time in the buffer zone (Figure 5.3, inset map). Mei Mei had

58% of fixes in the buffer zone, 32% in the core zone, and 1% in the experimental zone. There was not a particular time of year that she preferred the buffer over the core zone, as she used both intermittently. The boundary between the core and buffer zone happened to run down the center of her roughly 3 km2 range for the time period in question (the boundary lay along a valley bottom and Mei Mei split her time along mountainsides on both sides of the valley). Mei Mei’s use of the experimental zone occurred because her range was buttressed up against a livestock grazing area (which formed the outer boundary of the experimental zone). Pan Pan spent no time in the core zone during the course of the study, while 99% of her fixes were located in the buffer zone and 1% in the experimental zone. Pan Pan’s use of the experimental zone occurred when she was distributed at a lower elevation, a mere one hundred meters of map distance to the main road during a time in which the low elevation umbrella bamboo (Fargesia robusta) shoots were emerging. In both pandas the 1% use of experimental zone occurred very close to zone boundaries and could be interpreted as resulting from errors in the GPS collar or the zoning map.

Nonetheless, the close proximity to experimental zone is of importance. In fact, Mei Mei and Pan

114

Pan were located on average 912 ± 470 m and 940 ± 531 m (mean ± SD) from the nearest experimental zone, respectively.

Zoning and Human Impacts

Zoning had mixed results with respect to containing human activities in the experimental zone (Figure 5.4). Houses were completely contained within the experimental zone (during both

2002 and 2006). Roads and tourism facilities were mostly contained within the experimental zone (87% and 79%, respectively). Tourism facilities were distributed closest to the core zone

(63% within 1 km), followed by roads (53% within 3 km) and houses (68% within 4 km, for map see Figure 5.1).

Of the 4 tourism facilities located outside the experimental zone, two of the three sites in the buffer zone included a long-term scientific monitoring station that is occasionally inhabited by controlled (and minimal) numbers of birders and a recently constructed panda observation station for tourists that is no longer used as a result of the May 12, 2008 Wenchuan earthquake.

The third site in the buffer zone was a major tourism attraction currently in construction, which required a revision in zoning, such that the 2009 zoning scheme now designates the area as experimental zone. The one site located in the core zone was a scenic destination along the small portion of the main road which falls inside the core zone, although in a high elevation, non- forested area that does not constitute giant panda habitat. We believe this portion of road was misclassified due to the use of an inaccurate provincial road layer during the zoning designation.

In contrast to the houses, roads, and tourism facilities, livestock were not well contained in the experimental zone (Figure 5.4). Herd 1 (the herd monitored over the longest time period) spent most of its time in the buffer zone (52%) followed by the core (33%) and experimental

115

(15%) zones. There was not a particular time of year that the herd was found in one zone or another. Instead, the herd was located over a roughly 2-3 km2 region where the buffer zone was particularly narrow (100 m wide at the narrowest point) and thus the herd would migrate freely back and forth through core, buffer, and experimental zones all within a span of 500 meters and often spend parts of a single day in more than one zone (Figure 5.3, inset map). Herd 2 was distributed in a roughly 0.17 km2 area between experimental (70%) and buffer (30%) zones during the short time (one month) it was monitored. All horse signs spanning the distribution of

Herd 3 were distributed solely in the buffer zone (100%) across a 0.8 km2 area roughly 1.5 km away from both the nearest experimental and core zones.

Forest cover change also varied across zones (Figure 5.5). The forest cover loss that occurred from 1974 to 1997 (both periods prior to zoning designation) was highest in the area that would later be designated the experimental zone, followed by the buffer and core zones. At the time of designation, the core zone inherently had the lowest forest cover (30%) compared to buffer (62%) and experimental (44%) due to the fact that this contained large areas at high elevations and above the tree line. After zoning designation, both the buffer and core zones experienced forest recovery such that the resulting percent forest cover exceeded the estimated percentage once existing in 1974 by around 4%. On the other hand, the forest recovery in the experimental zone during the 1997-2007 period, while nearly equivalent in overall magnitude to the other two zones (~12%), did not have as measurable of an impact when considering its potential for supporting forest, since the forest cover in this zone remained 8% lower than it once was in 1974.

116

Proposed Zoning Revisions for Panda Conservation

We identified approximately 37 km2 of current experimental zone in the reserve that is outside of existing human establishment but also inside of areas identified as having panda presence (focal experimental zone, Figure 5.6). These areas are identified here as deserving consideration for future adjustment to become either buffer or core zone (or a combination of the two). Three of these areas deserve mention (A, B, and C in Figure 5.6). Areas A and B are both valleys that have apparently been left open as experimental areas for potential tourism development in the future. Area C is the contentious area also discussed as part of our GPS collar component of this paper. This is the area of the reserve where there is among the narrowest width of buffer zone (100 m) and one in which a narrow strip of core zone extends out between surrounding human establishments. Area C also partly overlaps with an existing grazing area, such that there should be further discussion to determine the exact location where the boundary should be drawn.

We also identified approximately 178 km2 of current buffer zone in the reserve that is outside of existing human establishment but also inside of areas identified as having panda presence (focal buffer zone, Figure 5.6). Certainly, some of this area should remain as buffer zone, especially when adjacent to an experimental zone. However, of note in Figure 5.6 is the large width of buffer zone in some places, extending up to 4 km away from the nearest experimental zone. We identify two areas in Figure 5.6 (D and E) where consideration should be made for extending the core zone to account for panda presence. In addition, considering the strong presence of pandas in area F (an area which is currently entirely buffer zone), we propose that discussion should be initiated to create a new region of core zone here.

117

Discussion

Efficacy of Zoning in Wolong

This study showed that the design of zoning designations put in place in Wolong Nature

Reserve for multiple-use management could be improved with respect to the goal of protecting the endangered giant panda population. All three approaches we used to analyze the distribution of giant pandas and their habitat with respect to the zones suggested that the buffer zone (and some areas of the experimental zone) serves an important role in supporting the giant panda population. However, the regulations on the types and degrees of human activities that are allowed in the buffer zones are not clearly defined in the management and policy arenas at the national level in China (Liu and Li 2008). Thus, there is a degree of vulnerability of the panda population when it is not limited to the fully protected core zone.

It is important to underline that giant pandas are just one species of the thousands present in this reserve. Although the giant panda is often a considerable focus for management and policy making, further research is required to investigate the efficacy of zoning for other plant and animal species, most of which have insufficient data available to draw conclusions. It is worth noting, however, that rare and endangered plant and animal species inhabiting the high- elevation (above tree-line) areas in the reserve appear to be well contained in the core zone of this zoning scheme, including the snow leopard (Uncia uncial), blue sheep (Pseudois nayaur), and red poppywort (Meconopsis punicea) (Schaller et al. 1985; State Forestry Administration

1999; Wolong Nature Reserve Management 1987). On the other hand, other rare and endangered species that have a higher degree of overlap with the giant panda’s forest habitat may be facing similar risks with respect to zoning, including the Asiatic black bear (Urus thibetanus),

118

red panda (Ailurus fulgens), dove tree (Davidia involucrata) and the Katsura tree

(Cercidiphyllum japonicum) (Schaller et al. 1985; Wolong Nature Reserve Management 1987).

Another question that is important to address concerns the issue of efficacy of zoning designation with respect to enforcing restrictions on human activity across zones. Perhaps the most positive outcome of the zoning scheme from the perspective of biodiversity conservation is that no new houses or roads were built outside of the experimental zone after the zoning designation was put in place. One could argue that it would not be feasible or practical to construct roads and houses in the high elevation parts of the core zone, regardless of whether a zoning designation prohibited such construction. However, several areas of buffer zone and some parts of the core zone (e.g. areas surrounding C and D in Figure 5.6) are at low elevations and in close proximity to existing human establishments, meaning that further human development would be conceivable in these areas if it were not prohibited by the zoning scheme.

It is also promising that the majority of tourism infrastructure was contained within the experimental zone. However, the efficacy of the zoning designation was put in question with the extension of the experimental zone to allow for a new major tourism attraction in one area. This revision was balanced out by returning one undeveloped area to core zone in another part of the reserve, but it is debatable as to whether those areas are of the same value for panda conservation.

This is an issue that is by no means limited to Wolong, as several other nature reserves in China have put forth requests for zoning revisions to allow for future development (Hubei

Environmental Protection Agency 2008). Such a practice reflects an insufficiency of the policy governing the rules for zoning designation and readjustment at the national level, which could stand to threaten the efficacy of this tool for conservation.

119

Another important finding of this study was that zoning was largely ineffective in regulating livestock grazing. While the livestock issue warrants further study at a larger scale, the three herds we monitored in this study, and particularly Herd 1 which roamed freely across all three zones, provided revealing information. Livestock have significant impacts on panda habitat through their destruction of bamboo, the panda’s main food source (Ran 2003) and have been identified as one of the most significant threats to panda habitat across their entire range in the latest National Giant Panda Census (State Forestry Administration 2006). However, there is currently no policy in place that specifically tackles this issue. Livestock are difficult to regulate on the ground because they are less visible than a tourism facility or a house. But on the other hand, livestock are domesticated and can be effectively managed using clear policy frameworks

(unlike a wild pest or invasive species). Our study indicates that zoning is apparently ineffective at containing livestock and thus other conservation measures, such as conservation incentive programs (similar to the existing Natural Forest Conservation Program, NFCP (Liu et al. 2008)), should be considered when addressing this issue. Strategies that impose steep penalties such as imprisonment have been successful at regulating activities such as poaching of giant pandas (Lü and Kemf 2001), but may not be realistic or ethical when projected onto livestock grazing.

It is also important to note that the core and buffer zones did appear to experience more relative improvement in forest cover than the experimental zone, given historical levels of forest cover estimated in 1974. However, such improvements cannot be fully attributed to the zoning scheme and may instead be more closely linked to other conservation policies such as the NFCP

(Viña et al. 2011). This conservation payment program was implemented around the same time as the zoning scheme and may have more directly impacted forest cover change by altering the behavior of individuals inhabiting local households as they responded to changes in ecosystem

120

services (e.g. fuelwood availability) and land use dynamics (e.g. cropland and transportation changes, Chen et al. 2009; Viña et al. 2011). Further studies should strive to study the complex interactions between these conservation policies and that of zoning.

To better understand the inherent complexities in such a system, it is instructive to consider it as a Coupled Human and Natural System (CHANS, Liu et al. 2007a; Liu et al. 2007b).

Zoning informs this paradigm by highlighting the extent to which regulatory designations are difficult to design on the ground when the human and natural components of a system interact and cannot be completely separated into distinct, exclusive zones. The ideal core zone would be designed just for biodiversity conservation alone and the ideal experimental zone for human activities alone, with the buffer zone representing an area where there would be some overlap between the two systems. However, in reality, we found that large parts of the core zone were uninhabitable by both humans and pandas because they were located above the tree line and lacking in key resources that both depend upon. At the same time, parts of the experimental zone were ideal for both humans and pandas and were places where both interacted across shared space (such as with livestock and pandas both sharing the same 3-5 km2 area or with pandas inhabiting areas close to tourism facilities, roads, or houses).

One could argue that the revised zoning designations we propose in this study have little value when considering the formidable challenges with enforcement of human activities across zones. While we recognize and discuss such challenges in the Methods section, we do not believe that these challenges should warrant discounting the value of zoning schemes. Instead, we believe that the successes we documented here with the exclusion of development in the core zone mean that efforts should be made toward improving the design. We hope that the areas of buffer and experimental zone identified in this study to be considered for revision can help aid in

121

pinpointing locations for future panda conservation focus. This is particularly evident when considering area F in Figure 5.6. The zoning in this area was recently amended to allow for the development of a tourism facility. This amendment comprises a large tract of buffer zone that appears to support a large number of pandas (according to the most recent panda census). It is located in close proximity to a proposed linkage area identified by Xu et al. (2006) to be ideal for establishing a corridor to help decrease fragmentation between Wolong and neighboring panda habitat outside the reserve. Such corridors are important, considering that population viability analyses on the species suggest that dispersal among fragmented sub-populations is crucial for long-term survival (Zhou and Pan 1997; Zhu et al. 2010). That being said, we do not suggest that our identified focus areas for zoning revision comprise an ideal design. In fact, further groundwork should be done to obtain a more detailed picture of the costs and benefits to both humans and pandas for zoning adjustment at each specific site.

Zoning as a Conservation Tool

Considering these findings, it is important to take a step back to ask the question of the role of zoning in protected areas in China (and across the world) while recognizing both its strengths and limitations. The challenges in zoning designations are unlikely unique to Wolong.

In fact, Wolong is regarded by many to be a flagship nature reserve and one that has had measurable success in conservation and management (State Forestry Administration 2006), while the challenges with zoning appear to be more severe in many other nature reserves in

China (Liu and Li 2008).

The strengths of zoning lie in its ability to shape development of tourism and transportation infrastructure at the hands of development companies. The challenge here is to

122

ensure that biodiversity is adequately accounted for in the original design of zoning boundaries when it has the tendency to be ignored (Liu and Li 2008). It is also important to ensure that adjustments of boundaries are performed with a set of well-defined rules and regulations that consider both conservation and development needs, as opposed to simply re-drawing the lines when development is desired. In Wolong, for example, despite revisions performed to account for the construction of tourism facilities, there have been no apparent attempts to revisit the zoning scheme in a corresponding way with respect to biodiversity protection.

The biggest limitation of zoning schemes is that they are inherently difficult to enforce on the ground when it comes to individual animal and human behaviors because it may be difficult to draw “lines in the sand” where one zone begins and the another ends. While buffer zones can help in this regard, by serving as “fuzzy” boundaries, their effectiveness is limited when there are no physical boundaries separating zones. Animals and plants certainly do not observe the designations, yet creating man-made boundaries (e.g., fences) is usually not practical. Humans may also not be aware of the designations, considering that some of our social surveys in

Wolong with local residents revealed that many of them were unaware of other recent government policies (He et al. 2009). This observation reiterates the central message of the

CHANS framework, which is that humans and natural systems are inherently coupled, such that simply drawing boundaries on a map to attempt to extricate them is difficult.

Ethical issues also come into play when considering the prospect of completely barring local residents (who are often economically poor) from using vital natural resources in their neighborhoods (Melick et al. 2007). Therefore, conservation payment programs that are currently in place in nature reserves of China such as the NFCP and the GTGP, which provide monetary subsidies to residents for their participation in conservation programs may be more

123

suitable for regulating these types of human activities (Liu et al. 2008). Thus, it is important for reserve managers to further strengthen these types of programs when the zoning scheme itself may not be adequately addressing all biodiversity goals.

In conclusion, zoning is one of many tools that are not a ‘cure all’ for conservation problems. No one policy can address all complexities of conservation challenges in a human- dominated world; rather a portfolio of different policies is needed. In our study area, zoning has proven effective as a tool in itself by preventing haphazard human development throughout an area of high biodiversity and one that supports a conservation icon, the giant panda. However, it is also important to recognize where zoning falls short as a method for achieving conservation goals (in our case with managing livestock and when readjustment was not clearly regulated), such that other methods may be needed to fill in the gaps.

In order for zoning to be effective, it must be implemented in a transparent way and in a way that allows for regulated and sound adjustments to be made in response to changing conditions in today’s human-dominated systems (Geneletti and van Duren 2008; Villa et al.

2002). It is especially important to adopt an adaptive approach when considering the implications of climate change, a phenomenon that can cause species’ ranges to shift outside of their inscribed management zones and thus require revisions to management plans (Hannah et al.

2007; Murphy et al. 2010). The complexity of interacting human and natural components in our system underscores the importance of such procedures for effective zoning with respect to multiple-use systems. Considering these challenges, we advocate for a CHANS approach to investigating the efficacy of zoning designations throughout nature reserves worldwide in order to better understand both human and natural factors that govern the success of this measure within the context of broader conservation initiatives.

124

APPENDIX

125

126

100% 90%

80% 70% 60% 50% core 40% buffer 30%

Percent of total area totalof Percent experimental 20% 10% 0% highly suitable moderately unsuitable suitable suitable Habitat suitability b.

100% 90%

80% 70% 60% 50% core 40% buffer 30%

Percent of total area totalof Percent experimental 20% 10% 0% highly suitable moderately unsuitable suitable suitable Habitat suitability

127

128

100%

90% 80% 70% 60% 50% 40% core 30% buffer

20% experimental Percentage in each in zone Percentage 10% 0% Houses Roads Tourism Livestock facilities (three herds) Human activity

129

Figure 5.5. Forest cover across zoning designations in Wolong Nature Reserve in 1974 (year before reserve establishment), 1997 (year before zoning designations), and 2007. Forest cover was derived from Landsat TM imagery and analyzed with respect to areal coverage in each zone. Error bars on the forest classification of the 1974 and 1997 images represent the area taken up by “unclassified” areas (areas with excessive clouds) that could have either been forest or non-forest.

40 1974 30 1997 2007

20 Percent of total area (%) total of area Percent 10

0 experimental buffer core Zone

130

131

CHAPTER 6

IMPACT OF LIVESTOCK ON GIANT PANDAS AND THEIR HABITAT

In collaboration with

Jindong Zhang, Shiqiang Zhou, Jinyan Huang, Andrés Viña, Wei Liu, Mao-Ning Tuanmu,

Rengui Li, Dian Liu, Weihua Xu, Yan Huang, Zhiyun Ouyang, Hemin Zhang, and Jianguo Liu

132

Abstract

Livestock production is one of the greatest threats to biodiversity worldwide. However, impacts of livestock on endangered species have been understudied, particularly across the livestock-wildlife interface in forested protected areas. We investigated the impact of an emerging livestock sector in China’s renowned Wolong Nature Reserve for giant pandas. We integrated empirical data from field surveys, remotely sensed imagery, and GPS collar tracking to analyze (1) the spatial distribution of horses in giant panda habitat, (2) space use and habitat selection patterns of horses and pandas, and (3) the impact of horses on pandas and bamboo

(panda’s main food source). We discovered that the horse distribution overlapped with suitable giant panda habitat. Horses had smaller home ranges than pandas but both species showed similarities in habitat selection. Horses consumed considerable amounts of bamboo, and may have resulted in a decline in panda habitat use. Our study highlights the need to formulate policies to address this emerging threat to the endangered giant panda. It also has implications for understanding livestock impacts in other protected areas across the globe.

133

Introduction

One of the most significant drivers of global land use change is livestock production, a fast growing economic sector currently affecting 23 of the 35 global biodiversity hotspots

(Rindfuss et al. 2008; Steinfeld et al. 2006). Livestock can have a profound impact on biodiversity by promoting habitat loss and degradation, global climate change, pollution, spread of invasive species, and disease transmission (Steinfeld et al. 2006). Livestock also may directly compete with wildlife species for limited food and space, in turn threatening their survival

(Madhusudan 2004; Mishra et al. 2004). Vulnerable to such competition are herbivorous species

(particularly threatened/endangered species) that share similar dietary restrictions and foraging strategies as livestock (Beck and Peek 2005; Namgail et al. 2007; Young et al. 2005).

One endangered animal species that may be affected by livestock production is the giant panda, a large herbivorous mammal and international symbol for biodiversity conservation (Hull et al. 2011b). The ca. 1,600 remaining wild pandas are native to the mixed deciduous and coniferous forests in southwestern China (State Forestry Administration 2006), where they inhabit isolated mountain ranges fragmented by the activities of a growing human population, including farming, road construction and timber harvesting (Chen et al. 2010). Over 60 nature reserves have been established to protect giant pandas (Viña et al. 2010), but reserves may not always provide a sufficient regulatory framework to prevent further panda habitat degradation

(Liu et al. 2001).

In the most comprehensive survey of giant pandas and their habitat to date, the 3rd

National Giant Panda Survey, surveyors spanned the entire geographic range of the species to document evidence of both giant panda habitat use and human disturbance (State Forestry

Administration 2006). In this survey, out of ten different types of human disturbance identified,

134

livestock grazing was the second most commonly encountered type (11% of 34,187 plots, 17% of all disturbances), behind only timber harvesting (28% of the plots, 41% of all disturbances).

However, timber harvesting was determined to be a legacy effect (i.e., not occurring at the time of survey) in more than 90% of the observed cases due to a successful national timber harvesting ban (State Forestry Administration 2006). On the other hand, 93% of livestock grazing incidences were deemed to be ongoing at the time of the survey. Livestock grazing was also the most prevalent disturbance in one recent study spanning the entire Minshan mountain range

(livestock disturbance found in 19% of over 1,600 sample plots (Wang 2008)). However, as far as we know, there is little monitoring and management of livestock production in the panda’s geographic range, even inside nature reserves.

Despite the recorded prevalence of livestock across giant panda habitat, research on the nature of the impacts on pandas and their habitat is limited to a small number of case studies in the Chinese literature (Kang et al. 2011c; Ran 2003, Ran et al. 2003, and Ran et al. 2004). These studies reiterated the findings of the Third Giant Panda Survey about the prevalence of livestock disturbance in panda habitat, in addition to showing that there is some overlap in the habitat selection of pandas and livestock. However, many questions remain regarding the space use and habitat selection of individual livestock animals, spatial distribution of livestock impacts on panda habitat and the nature of the impacts. Of particular concern is whether livestock could threaten the sustainability of the giant panda’s main food, understory bamboo, a food source that is not believed to be threatened by any other animal competitor (Schaller et al. 1985).

We set out to fill these information gaps in Wolong Nature Reserve, a flagship reserve for giant panda research and a driver of policy making for the conservation of this endangered species (Tuanmu et al. 2010; Viña et al. 2008). We analyzed data obtained from forest surveys,

135

remote sensing and Global Positioning System (GPS) collar telemetry to investigate the effects of an emerging livestock sector in this Reserve – the rearing of domestic horses. Our primary interest was in determining whether forest encroachment by horses could threaten the giant panda by occupying suitable habitat and consuming bamboo. Our objectives were to: (1) assess horse distribution with respect to panda habitat suitability and panda distribution, (2) compare space use and habitat selection patterns of horses and wild pandas, and (3) analyze the impact of the horse herds on bamboo biomass and on panda habitat use.

Methods

Study Area

Wolong Nature Reserve is located in Wenchuan County, Sichuan province, China

(102°52' to 103°24'E, 30°45' to 31°25'N, Schaller et al. 1985, Figure 6.1). It is one of the largest reserves for the conservation of giant pandas (2,000 km2) and harbors 10% of the total wild giant panda population (Liu et al. 2001; State Forestry Administration 2006). There are also over

10,000 plant and animal species found in the Reserve (Tan et al. 1995) owing to its wide elevational range (1,200 to 6,250 m, Schaller et al. 1985). The giant pandas mainly inhabit mixed deciduous broadleaved coniferous forests at intermediate elevations of 2,250 to 2,750 where they forage on bamboo, which can cover up to 95% of the forest understory area (Schaller et al. 1985). Giant pandas are solitary mammals and are obligate bamboo foragers, with bamboo making up over 99% of their diet throughout all seasons of the year (Schaller et al. 1985).

The Reserve is also home to nearly 5,000 human residents who are mainly farmers. With respect to livestock, residents raise cattle, pigs, goats, and yaks for meat (Ghimire 1997). Yaks make up the largest group of livestock, with over 3,000 animals in the Reserve, followed by goats and

136

cows (~1,500 animals each, Wolong Nature Reserve 2008). Historically, horse rearing was rare in the Reserve (less than 25 horses occurred in the entire Reserve as recently as 1998). Although horses now make up the smallest proportion of all livestock animals at just under 350 heads (in approximately 20-30 herds), the number of horses has increased tenfold from 1996 to 2008

(Wolong Nature Reserve 1996, 2008). The growth in this sector can be attributed to telecoupling processes (Liu et al. 2013) such as selling horses to regions far away from Wolong and to strengthening agricultural business exchanges between Wolong residents and those in Xiaojin township (located outside of giant panda habitat and adjacent to the Reserve on its western side), where horse rearing is prominent.

Horses are supposed to be contained year round in existing grazing areas. However, in recent years, some horse herds have been excluded from grazing areas because they over- consumed grasses. As a result, horse herders have sent their horses to nearby forests (and panda habitat) to graze separately from the cattle. In these forests we have observed horses to forage largely on bamboo, since it is the most available plant matter present in the understory. To our knowledge, this practice has occurred with at least four horse herds in Wolong. Horse herders only visit their herds approximately once per month and do not spatially contain their activities.

Study Subjects

We monitored four focal herds of horses (hereafter Yusidong, Qicenglou, Papagou and

Fangzipeng, after the name of the local regions where they graze) inhabiting giant panda habitat in Wolong. These herds were not chosen as representative of all livestock production systems occurring across the entire reserve but for analyzing the emerging trend of forest encroachment by horses (i.e., the herds chosen inhabit forests in the Reserve). All herds have only recently

137

been introduced to their respective forest areas (Table 6.1). The Yusidong, Qicenglou, and

Papagou herds were monitored by placing a GPS collar on one member of each herd. Of those, the Yusidong herd is the largest (23 horses). The Fangzipeng herd was monitored using field surveys only (since additional GPS collars were not available).

We also monitored three wild giant pandas (two adult females— Mei Mei and Zhong

Zhong and one adult male— Chuan Chuan) using the same type of GPS collars. These pandas occupied an area in close proximity to the Yusidong herd (Figure 6.1). While these pandas constitute a small sample, like many other endangered species (Gill et al. 2008; Miller et al.

2010), it is not feasible for the government to give permits to study many individuals using GPS collars. Nevertheless, this is the first spatially explicit account of sympatric panda and horse behavior using high accuracy GPS collar telemetry. The use of wild pandas for this project was approved by the State Forestry Administration of China. The China Conservation and Research

Center for the Giant Panda (CCRCGP) was responsible for all animal care procedures. Efforts were made to limit disturbance to animals to short periods required for initial anesthetization and collar deployment. To keep the time period constant across individuals, we restricted the data used in this analysis to a one-year period (between 6/15/2011 and 6/15/2012 for all pandas and horses except Mei Mei, a female panda monitored from 6/15/2010 to 6/15/2011).

All GPS collars were 12-channel GPS 4400 M models (Lotek Engineering Inc.,

Newmarket, Ont., Canada) and were scheduled to record fixes every 4 hours. Fix acquisition rates of the collars were 84, 99, and 90% for the Yusidong, Qicenglou, and Papagou horse herds, respectively and 51%, 16%, and 54%, for Mei Mei, Zhong Zhong, and Chuan Chuan, respectively. The lower fix acquisition rate for pandas is largely due to their behavior (e.g., time spent sleeping or feeding while inclined on the back may block satellite reception). Before

138

deployment, we performed field tests of the collars’ fix acquisition rates (n= 30 plots), and all exhibited rates above 90%, with no detectable effects of landscape characteristics (e.g., topographic position, canopy cover). Tests of locational accuracy against a differentially- corrected GPS unit showed that locations obtained by the GPS collars were 95% accurate within

60 m.

Data Collection and Analysis

(1) Spatial Distribution of Horses

We assessed giant panda habitat suitability in areas occupied by horses. Habitat suitability (divided into four suitability classes) was obtained by a multiplicative function of elevation, forest cover, and slope (Liu et al. 1999). Pandas require forest, are limited to certain mid-range elevations (around 2,000-3,300 m in Wolong) due to bamboo growing patterns, and prefer gentle slopes (<45˚) for ease of travel (Schaller et al. 1985). The forest cover layer was derived from a supervised classification of Landsat TM imagery (30 x 30 m resolution) from

2007 (Viña et al. 2011) into forest versus non-forest (with an 82.6% accuracy). The slope and elevation layers were obtained from a Digital Elevation Model (DEM) acquired by the National

Aeronautics and Space Administration’s (NASA) Advanced Spaceborne Thermal Emission and

Reflection Radiometer (ASTER, 29 m resolution). We conducted spatial overlays to summarize the proportion of each horse home range within each panda habitat suitability class.

We assessed the spatial relationship between areas occupied by horses and the predicted probability distribution of pandas across the Reserve. Panda distribution prior to horse occupancy was estimated by conducting a bivariate normal kernel density estimation (Bailey and

Gatrell 1995, bandwidth h=1000 m) on all panda signs surveyed during the most recent nation-

139

wide giant panda census in 2001 (conducted prior to horse occupancy, State Forestry

Administration 2006). In this survey, surveyors searched for panda signs along transects established along elevational gradients within 2-6 km2 habitat blocks distributed across all panda habitat in the Reserve. We summarized the proportion of each horse home range within each percent volume contour of the predicted panda distribution (5-100% contours, with 5% representing the top 5% probability of occurrence).

Horse home ranges were estimated by constructing 95% probability distributions around the locations obtained from each of the GPS collars in each herd using the biased random bridge approach (Benhamou 2011). This approach uses a probability density function to predict an animal’s probability of use of an area based on the angle of movement from one time point to the next using a biased random walk (Benhamou 2011). A biased random walk refers to the condition in which the distribution of angles is not uniform due to a preference on the part of the animal. In the case of the Fangzipeng herd (which was not monitored with GPS collars), we established a set of 5 transects (0.5 to 1 km each) running through an area of ca. 1 km2 where we observed the herd to roam, and recorded the presence or absence of horse signs (e.g., feces or eaten bamboo) in 30 x 30 m plots (n=49) every 100 meters along these transects (for sampling details see (3) below). We then created a minimum convex polygon (MCP) that covered 95% of the plots with horse signs to represent a rough estimate of their home range. Despite the lower accuracy of this method, we included the Fangzipeng herd due to the historical significance of its location as the core study region of decades-long giant panda research (Schaller et al. 1985). All home range analyses were conducted using the R software (R Development Core Team 2005) and the “adehabitatHR” package for R (Calenge 2011b).

140

(2) Comparison of Space Use and Habitat Selection of Horses and Pandas

We compared the home ranges estimated for horses with those estimated for the three wild GPS-collared giant pandas in order to contrast how these two species use space. For the pandas, we used the same home range estimation method used in the first three horse herds

(biased random bridge approach). We excluded the fourth herd (Fangzipeng) from this comparison due to unavailability of GPS collar data. We compared horses’ and pandas’ home range sizes using isopleths. We also compared the core area sizes, which were calculated using the core area estimation method outlined in Vander Wal and Rodgers (2012).

We then compared habitat selection by horses and pandas, using the k-select analysis method (Calenge et al. 2005). This multivariate method is an eigenanalysis of the marginality in the data. Marginality refers to the “squared Euclidean distance between the average habitat conditions used by an organism and the average habitat conditions available to it” (Calenge et al.

2005: 145). The eigenanalysis reveals the “linear combination of habitat variables for which the average marginality is the greatest” (Calenge et al. 2005: 143). The approach is particularly useful for identifying differences in habitat selection among individuals of a group, rather than averaging the variation across individuals (Calenge et al. 2005). We also performed randomization tests (with n= 10,000 randomization steps) to determine the significance of the results. At each step, the k-select analysis was repeated and the first eigenvalue of the observed data was compared to the randomized data set. This method tests whether the marginality vector for each animal is significantly different from what would be expected under a random use and also determines the effect of each variable on the overall marginality. The k-select analysis was conducted using the “adehabitatHS” package for R (Calenge 2011a).

141

We defined “available habitat” in our analysis as a single large area that encompassed the home ranges of all study animals. This area was delineated by the smallest polygon that enclosed locations of all study animals, expanded by a 500-meter buffer area to allow the inclusion of border effects. We defined “used habitat” as a random selection of one fix per day from all fixes obtained for each animal. This random selection was used to account for temporal autocorrelation and offset differences in GPS fix acquisition rates among individuals.

Variables included in the habitat selection analysis were slope, elevation, topographic position, solar radiation, forest cover and distance to nearest household. These variables were chosen due to their importance for panda habitat use as demonstrated in previous studies (Bearer et al. 2008; Hu 2001; Liu et al. 2011). Slope, elevation and forest cover were obtained as previously described. Topographic position was measured using the topographic position index

(TPI), which is a measure of the difference between the elevation in a pixel and the average elevation in the surrounding pixels. The TPI was calculated on a 9-pixel neighborhood area

(chosen to represent the smallest window around a given pixel) using the Land Facet Corridor

Designer in ArcGIS (Jenness et al. 2012) based on the same elevation layer described previously

(i.e., DEM derived from ASTER data). Solar radiation was calculated using the Area Solar

Radiation tool in ArcGIS (assuming a 200 m sky size and a year-long calculation using monthly intervals). Distance to nearest household was calculated using ArcGIS while using the DEM to calculate the surface distance between each pixel. Household locations were obtained by our research group in 2003 using GPS receivers (Chen et al. 2009; Linderman et al. 2005).

142

(3) Impact of Horses on Bamboo and Pandas

To assess the impact of horses on bamboo, we established plots (30 x 30 m) every 100 meters along transects throughout the home ranges of the horse herds in two of the sites studied

(Fangzipeng n = 49 and Yusidong n = 57). Transects were selected according to a spatial orientation which would allow an even coverage of the affected area and also capture variations in elevation and vegetation cover. We visually estimated the percent cover of arrow bamboo

(Bashania fangiana) (the only bamboo species present in these areas), eaten in each plot. We also counted the number of bamboo culms foraged in 1 x 1 m subplots located at the center of all

Fangzipeng plots (n = 49) and at the center of a random sample of the plots at Yusidong (n = 10).

If a subplot had no bamboo, we left the entry blank (yielding a total of n = 38 subplots with bamboo). Consumption of bamboo by horses was visually conspicuous and readily distinguishable from those of other animals. For instance, horses foraged along the tops of bamboo culms and ate leaves growing from the top few nodes of the plant. Contrary to native ungulates and giant pandas, horses did not selectively choose individual bamboo culms but instead foraged on the majority of the culms located in a given area, thus giving all culms a stunted appearance.

To assess the impact of horses on panda habitat use, we conducted a temporal analysis of frequency of panda signs before and after horse occupancy. This analysis was only conducted for Fangzipeng, since this was the only area for which we had a repeated sampling dataset. We summarized the number of panda signs found in our repeated sampling transects in three visits prior to horse occupancy (November 2006, February 2007, and April 2007) and four visits after horse occupancy (January 2008, October 2008, June 2009, and October 2009) and performed a two sample t-test to determine whether there was a significant difference in the number of feces

143

found in the two periods. Although the exact months in which sampling was conducted varied in the pre- and post-horse occupancy sampling periods, differences of only one month (e.g. October vs. November) are not believed to create a bias, since panda signs remain visible for several months. For all periods, we searched exhaustively for panda signs (feces or eaten bamboo) within 100 m on either side of each transect line. If multiple signs were observed in the same 30 x 30 m plot, they were counted as a single observation. There was a lower search effort in the sampling periods in June 2008 and October 2009 (one day of sampling instead of two) due to a recent earthquake, whereby the same amount of area was surveyed but with less time spent in the search. Therefore, for these two sampling periods, we multiplied the number of panda signs by two prior to the analysis to account for the 50% lower search effort. The earthquake caused no damage to the Fangzipeng habitat area [i.e., 0% habitat loss; Ouyang et al. (2008)] and did not result in discernible avoidance by pandas. During our long-term sampling of this study area, we did not observe or detect any other disturbance aside from horses that could conceivably alter panda habitat use.

Results

Spatial Distribution of Horses

Approximately 50% of three of the horse herd home ranges were located in highly suitable or suitable giant panda habitat (Figure 6.2). All herds were found at suitable elevations for panda inhabitance. With the exception of the Qicenglou herd, the majority of the horses’ home ranges were located in forests (58-88%) with suitable slopes for the pandas (63-86%).

Although the Qicenglou herd was distributed in comparably less suitable panda habitat than the other herds (mainly due to the existence of non-forest), the impact of this herd was the most

144

widespread, as this herd was moved by the herders to four distinct locations over a three year period (see Figure 6.1). All herds, except Qicenglou, were distributed completely within the predicted probability distribution of giant pandas. However, the herds never distributed within the central part of the panda distributions, as they were mainly distributed within the 75-85% percent volume contours (Figure 6.3, for spatial distribution see Figure 6.1).

Comparison of Space Use and Habitat Selection of Horses and Pandas

The utilization distributions of horses were smaller in area than those of pandas across all home range isopleths above 60 (Figure 6.4). Panda home range sizes (the 95 level isopleths) were 2.2, 3.4 and 5.6 km2 for Mei Mei, Zhong Zhong, and Chuan Chuan (the male), respectively, while horse home range sizes were 1.1, 1.0, and 1.6 km2 for Yusidong, Papagou, and Qicenglou herds, respectively. Core areas were also generally smaller for horses (Y= 0.3 km2, P= 0.3 km2, and Q= 0.6 km2) than pandas (M= 0.6 km2, Z= 0.9 km2, and C= 1.4 km2).

Randomization tests revealed that the first eigenvalue obtained in the k-select analysis was significantly larger than expected under random habitat use (λ1 = 1.39, p<0.0001), making further analysis of habitat use patterns informative. Habitat use was significantly non-random for all horses and pandas (Table 6.2). All pandas and horses showed selection for gentle slopes and areas with high solar radiation (Table 6.2). All animals showed selection for higher elevations except the Qicenglou herd, which exhibited an opposite trend (i.e., selection for lower elevations). All pandas showed selection for areas of higher topographic position (e.g. mountain ridges), but only one of the three horse herds showed the same pattern. Only three subjects showed significant habitat selection with respect to distance to the nearest household, which included the male panda Chuan Chuan (positive) and two of the three horse herds (both negative).

145

Results with respect to selection for forest cover were also mixed, as one panda selected positively for forest cover (Zhong Zhong), one horse herd and one panda (Chuan Chuan) selected against forest cover, and the rest displayed no selection.

With regard to the eigenanalysis, the first two eigenvalues captured the majority of the marginality in the data (Figure 6.5a). The first axis represented mainly elevation, while the second axis captured mainly forest cover (Figures 6.5b and 6.5c). As is evident from the configuration of the arrows representing each animal on the projection of the marginality vectors on the first factorial plane (Figure 6.5d), the pandas had similar habitat selection patterns with one another as they selected for high elevation areas and high topographic position. One horse herd (Yusidong) had similar patterns as the pandas, and an almost identical niche as the panda

Mei Mei. The Papagou herd diverged slightly from the pandas by selecting more strongly for high solar radiation and low slope and less strongly for high elevation. The Qicenglou herd diverged strongly from all other animals by selecting against forested areas and for areas closer to the nearest household.

Impact of Horses on Bamboo and Pandas

Horses had an impact on the arrow bamboo in both areas studied. At Fangzipeng, we estimated that the horses foraged more than 20% of the bamboo in 18 of the 49 plots, with 5 of those plots experiencing more than 75% bamboo foraged by horses (Figure 6.6). At Yusidong, we estimated that horses foraged more than 20% of bamboo in 28 of the 57 plots and over 75% of the bamboo in 2 plots (Figure 6.6). In the 1 x 1 m subplots, horses either foraged over 90% of the available bamboo culms (52% of subplots) or less than 1% of the available bamboo culms

(34% of subplots), with little moderate foraging in between these two extremes (i.e., remaining

146

12% of subplots). Horses also appear to have had a significant impact on panda habitat use, considering that the number of plots containing panda signs in our repeated sampling prior to the introduction of the horses (2006 and 2007) at Fangzipeng was 5 to 10 times higher than the post- horse occupancy surveys in 2008 and 2009 (Figure 6.7).

Discussion

This study documented an emerging threat to the endangered giant panda and its habitat: livestock grazing. Logistical constraints prevented us from following a greater number of pandas and horse herds; thus our inference space is relatively narrow and results should be interpreted with caution. However, our findings shed light on the potential consequences of this emerging conservation issue. Horses in particular may be poised to be incompatible with giant panda conservation goals due to their large food consumption rates and their ability to live for long periods of time un-monitored by humans in forested areas. Our results provide the first evidence to support this hypothesis by demonstrating horse use of suitable giant panda habitat, potential of some overlap with the niche of the giant panda, high bamboo consumption rates, and negative effect on panda occupancy.

One of our main findings was that horse herds engaged in forest encroachment are distributed in suitable panda habitat. The Fangzipeng area in particular is historically important for scientific research on the giant panda and has been portrayed as consistently harboring several pandas (Schaller et al. 1985). Fangzipeng supports at least two den trees that have been used by giant panda mothers to rear their young, not only in the past (Schaller et al. 1985) but also in recent years (according to our field observations) prior to horse occupancy. However, when we visited the den trees after horse occupancy, we saw only horse droppings where panda

147

droppings were once prevalent. Although we lacked a control site with which to compare the degree of decline in panda signs after horse occupancy, our results at Fangzipeng are important and useful to provide guidance for further research on the mechanisms (e.g., direct avoidance, competition for bamboo) behind the effects of horse occupancy on giant panda occupancy.

Our habitat suitability analysis also showed that the current methods used to classify and predict suitable panda habitat may be insufficient. Giant panda habitat suitability models often rely on binary classifications of forest versus non-forest as a primary measure to delineate areas suitable for panda inhabitance (Liu et al. 1999; Liu et al. 2004; Wang et al. 2010). Our findings suggest that it is also important to include other human disturbances in addition to timber harvesting (An et al. 2006; He et al. 2009; Linderman et al. 2006; Tuanmu et al. 2011), as current forested areas in otherwise suitable giant panda habitat may be subjected to threats such as livestock grazing that may only be detected using field surveys.

Our habitat selection analysis showed both similarities and differences in habitat selection by pandas and horses. Individual variation in selection patterns was also notable within both the pandas and horses. Areas most likely to be contested due to selection by both horses and pandas appear to be those with low slopes and high solar radiation. The Yusidong herd best demonstrated the potential for niche overlap between horses and pandas, while both seemed to group together in the k-select analysis due to their similar selection with respect to elevation, slope, and solar radiation. It is not surprising that this herd demonstrated the greatest similarity with the monitored pandas, as it is located in closest proximity to the monitored pandas compared to the other herds. On the other hand, the divergent pattern of the Qicenglou herd relative to the monitored pandas and its selection against forested areas suggests that there is

148

potential to avoid niche overlap between horses and pandas if herders appropriately choose non- forested areas to allow their horses to roam and subsequently monitor them effectively.

Our field observations show that bamboo foraging by horses could potentially threaten the availability of food for the giant pandas, if horse populations are unchecked. This is a significant finding because in the past, it has generally been accepted that giant pandas have no strong competitors for their bamboo food source (Schaller et al. 1985; Taylor and Qin 1987).

Other wild animal species that consume bamboo across the giant panda geographic range (e.g. tufted deer, sambar) do so in small quantities and as part of more varied diets (Schaller et al.

1985). Other species relying on bamboo such as the red panda select for different types of micro-habitats compared to giant pandas, displaying a degree of niche diversification (Zhang et al. 2006).

As non-ruminants, horses need to consume up to 10 kg of plant biomass a day, which is

60% larger than ruminants such as cattle (Menard et al. 2002)). In our study areas, horses ate arrow bamboo, which provides around 0.4 to 0.9 kg of plant biomass per m2 (Taylor and Qin

1987). However, horses only grazed the tops of bamboo stems and leaves, accounting for roughly 10% of each plant. At this rate, a herd of 20 horses could potentially graze on up to 20% of bamboo culms in a 1 km2 area in a year. Although the daily consumption rate of horses is about the same as giant pandas, the level of intensity of foraging per unit of area by horses is significantly larger due to their higher densities, in addition to their smaller home ranges as compared to the pandas. Although exhibiting some overlap in their home ranges, pandas are solitary and are believed to be distributed at densities of about one individual per 2 km2 in this reserve (or 19 pandas in a 35 km2 area in one study (Schaller et al. 1985)). Panda foraging itself does not significantly threaten bamboo populations (Pan et al. 2001), perhaps in part due to the

149

adequate spacing of pandas, in addition to the fact that pandas alternate their foraging behavior to select for different bamboo plant parts and species throughout the year (Pan et al. 2001; Wei et al.

2011). However, the nearly 50-fold higher density of horses introduced by humans presents a markedly higher amount of foraging pressure on localized bamboo patches. Further studies are needed in order to determine potential long-term effects of horse grazing on bamboo and ascertain whether horse foraging may promote or discourage new bamboo growth.

In addition to direct impacts from foraging, horse herding may also have other impacts on pandas that need to be further studied, including avoidance due to sound and odor disturbances.

Other potential impacts to also explore in the future include disease transmission across the livestock-wildlife interface, a topic that has been understudied in pandas despite early warnings of potential risks (Hu 1981). Since we observed horse carcasses in close proximity to panda signs and horse feces in the pandas’ water sources, this issue should be seriously considered in future research.

While a directive is in place requiring livestock to be kept within designated grazing areas inside of giant panda nature reserves, our observations [and those of others across the entire giant panda geographic range (State Forestry Administration 2006)] suggest that this rule is not always operational on the ground. Indeed, in other areas of the giant panda geographic range exhibiting less management than Wolong and supporting higher numbers of livestock, the situation may be more serious (State Forestry Administration 2006). In order to better manage livestock management issues inside reserves, we propose that they be treated as part of a

Coupled Human and Natural System (CHANS) so that the complex interactions between people and nature are fully appreciated (Liu et al. 2007a). For instance, the financial needs of the local people should be taken into account, especially considering that animal husbandry serves as an

150

important mechanism for combating poverty across western China and may be needed to fill an income gap created after timber harvesting was banned in the late 1990s (Melick et al. 2007). It may be possible to link livestock management efforts with other successful conservation incentive programs implemented across China (see Liu et al. 2008) by providing monetary incentives to local farmers for participating in conservation.

Our study is unique in that it captures the nature of a human disturbance towards the start of its emergence rather than after it has become a problem entrenched in a system. The value of such a study is that it can drive future management measures. For instance, after our concerns about the impact of horses were brought to the managers of Wolong Nature Reserve, the administration called for removal of horses from the reserve. This study thus demonstrates the importance of on the ground assessments of livestock impact on forests and direct ties to endangered species relying on such forests for their survival.

151

APPENDIX

152

153

154

155

156

Figure 6.5. K-select analysis on marginality of habitat selection vectors among 3 wild giant pandas (Mei Mei, Zhong Zhong, and Chuan Chuan) and 3 horse herds roaming in giant panda habitat (Yusidong, Papagou, and Qicenglou). Plots include (a) bar chart of eigenanalysis showing proportion of marginality explained by each vector, (b) direction of the first two factorial axes, (c) variable loadings on the first factorial plane, and (d) marginality vectors for all 6 subjects on the first factorial plane. The notation “d” in (b), (c), and (d) indicates the distance of each grid cell.

157

158

159

Table 6.1. Summary of four domestic, free-ranging horse herds monitored in giant panda habitat in Wolong Nature Reserve, China.

Horse herd n Year introduced Type of analysis Fangzipeng 16 2007 Field survey Papagou 12 2004 GPS collar Qicenglou 5 2011* GPS collar Yusidong 23 2004 GPS collar, field survey *but previously held as part of a larger herd (n=20) at Huangcaoping and Laowashan; was moved back to Huangcaoping on 2/7/2012 (see Figure 6.1)

160

Pandas Horses Mei Zhong Chuan Papagou Qicenglou Yusidong Tests of marginality Marginality 1.97* 0.83* 3.46* 1.59* 3.62* 2.01* Selection of habitat variables (marginality vectors) elevation 0.82* 0.53* 0.83* 0.46* -0.88* 0.85* slope -0.78* -0.28* -0.60* -0.60* -0.40* -0.46* solar radiation 0.81* 0.51* 1.13* 0.81* 0.58* 1.03* terrain position 0.19* 0.38* 0.66* 0.35* 0.18 0.01 forest cover -0.14 0.22* -0.26* -0.15 -1.22* -0.12 distance to house 0.05 0.09 0.79* -0.46* -0.91* 0.00

161

CHAPTER 7

CONCLUSIONS

162

Animal species around the world face mounting threats from humans in today's globalized and industrialized era (Pimm and Raven 2000). At the core of this global challenge is the alarming rate of degradation and fragmentation of habitats, the fragile lifelines that sustain animal populations (Hoekstra et al. 2005; Pimm and Raven 2000). Research that contextualizes animal habitat within coupled human and natural systems highlights the connections between humans and the individual animals navigating contested landscapes, thus informing conservation efforts geared toward achieving sustainability (Liu et al. 2007a). Such research is important for the endangered giant panda, a species recognized as a global conservation icon facing growing threats from human development in its remaining fragmented habitat in southwestern China (Liu et al. 2001).

In this dissertation, I along with my colleagues sought to integrate behavioral data on individual giant pandas tracked using global positioning system (GPS) collars with other diverse sources of information describing how pandas relate to habitat. We in turn sought to integrate novel information about the ecology of pandas to on-the-ground management of the coupled human and natural system in Wolong Nature Reserve. The results of our analyses challenge previous assumptions made about how pandas relate to habitat, highlight new aspects of how pandas behave and respond to humans, and bring to light new questions that warrant future research. Our findings also inform investigations of animal behavior in other coupled human and natural systems by highlighting the utility of behavioral analyses of individual animals for testing long-held assumptions about animal-habitat relationships and for assessing the efficacy of management tools.

In Chapter 2, our synthesis of published studies on giant panda habitat selection identified trends, complexities, and gaps in existing knowledge that informed the development of the

163

remainder of the dissertation. We found that panda habitat selection was a complex process that involved interactions among different habitat characteristics and varied with habitat availability and selection level. Pandas generally selected forested areas found at moderate elevations that supported ample bamboo. Selection for topographic slope was nonlinear and not as significant as many other variables in multivariate analyses. Selection for primary (unharvested) versus secondary (harvested) forest varied across studies and selection against secondary forest decreased as availability of this forest type increased. Our findings collectively demonstrated the plasticity of panda habitat selection and the potential need to broaden criteria for suitable habitat for the species for protected area planning and habitat restoration. Our work also identified gaps in existing literature that informed subsequent chapters in the dissertation, including the need for a multivariate and multi-selection level approach to analyzing panda-habitat relationships

(explored in Chapter 4) and the need to more fully investigate the effect of human impacts on giant panda behavior (investigated in Chapters 5 and 6).

Our analysis of space use by giant pandas in Chapter 3 was useful by itself and also served as a necessary stepping stone for subsequent integration with habitat characteristics in

Chapter 4 and management data in Chapters 5 and 6. Tracking individual pandas with GPS collars provided high location accuracy, an advantage over much previous work done on the species. Use of biased random bridge movement models allowed us to delineate the utilization distributions of pandas across 2.8-6 km2 home ranges. We also documented for the first time the existence of several (16-39) small core areas distributed across the home range that pandas returned to after delays of weeks to several months, reflective of the panda's bamboo foraging strategy. We documented significant dynamic spatio-temporal interactions among neighboring pandas, a first for the species and a finding that gives credence to previously understudied

164

hypotheses regarding close ties among panda groups, despite being otherwise solitary.

Significant male-female interactions occurred during a surprising time of year (fall), suggesting a possible psudo-mating season deserving of further study.

In Chapter 4 we integrated the panda utilization distributions obtained in Chapter 3 with data on various habitat characteristics to understand habitat use and selection across the home range. Our spatial autoregressive resource utilization functions (RUF) revealed that panda habitat use was significantly negatively correlated to elevation and significantly positively correlated to solar radiation. One new finding in this chapter was that a large portion of the pandas' home ranges were in steep slopes (14-26%) and non-forest areas (18-42%) deemed marginally suitable or unsuitable in standard suitability models for the species. When examining habitat use with respect to habitat availability (i.e. habitat selection), we found that solar radiation was a positive predictor and slope was a negative predictor at both the within-home range and at home range selection level. Our results provided a rare individual-animal perspective on giant panda habitat use and selection, which supported some findings discussed in Chapter 2 at the population level, but also contextualized and challenged assumptions commonly made in broad scale models.

In Chapter 5, we built on the space use and habitat selection analyses from the previous chapters by adopting an integrated approach to analyzing the efficacy of a zoning scheme put in place in Wolong Nature Reserve to co-manage conservation and human development. We found that the zoning scheme was not optimally distributed to protect giant panda habitat across the reserve, while the individual giant pandas monitored using GPS collars moved freely throughout zones. We also found that while the zoning scheme succeeded at containing development in the form of house and road construction, it fell short in regulating individual human activities (e.g. livestock grazing) that are more difficult to monitor. We also identified several specific areas for

165

zoning revision (including the GPS collar study area) in order to better protect the panda population and also highlighted the need to integrate zoning with other management tools such as Payments for Ecosystem Services (PES) programs for better protection of panda habitat.

In Chapter 6, we expanded upon the issue of livestock grazing that was briefly addressed in Chapter 5 by conducting an in-depth assessment of the impact of an emerging livestock sector in Wolong— the rearing of domestic, free-ranging horses. We documented the encroachment of horses into forests that contained suitable habitat for pandas (including in the GPS collar study area) and detailed the impact of horse foraging on bamboo. We also illustrated the similar habitat selection patterns among horses and pandas and a decline in panda use of one highly suitable habitat after horse occupancy. We discussed the incompatibility of horse rearing with panda conservation and the need to seek other, creative ways to provide opportunities for rural residents of panda habitat to earn income without compromising panda habitat.

Collectively, this dissertation revealed new insights about giant panda behavior that can help guide conservation. However, the research also brought to light several remaining unknowns and areas needing future work. A priority should be placed on validating our findings using a larger sample size of giant panda individuals potentially monitored over a longer period of time. Such an exercise would not only strengthen the power of statistical tests but also potentially incorporate a greater range of possible behavioral patterns and habitat preferences. In addition, it would be worthwhile to link space use patterns seen in Chapter 3 to genetic data to determine kin relations among neighboring pandas, integrate bamboo data into habitat use and selection models like those presented in Chapter 4, investigate a greater diversity of human activities across zoning areas in analyses building off of Chapter 5, and explore a holistic picture of the distribution of numerous different livestock rearing practices expanding on Chapter 6.

166

In addition, future work should seek to more comprehensively integrate human behaviors into panda habitat studies to better understand how the movements and actions of individual people affect animals across the landscape. This could be accomplished using creative approaches such as participatory mapping to pinpoint areas of high human use not otherwise detected using satellite imagery or land use maps (see Rambaldi et al. (2006)) or by modeling movement patterns of human actors using handheld GPS technology (see Ashbrook and Starner

(2003)) to get a clearer picture of how humans move in relationship to pandas across space.

These data could then be linked to other information about humans and their communities obtained via social surveys (e.g. demographics, economics, values, social norms). These types of approaches would be novel not only for panda research, but for animal ecology as a whole, a field that would be enriched by drawing upon a coupled human and natural systems approach to better understand behaviors of co-occurring humans and animals.

167

REFERENCES

168

REFERENCES

Aebischer, N.J., Robertson, P.A., Kenward, R.E., 1993. Compositional analysis of habitat use from animal radio-tracking data. Ecology 74, 1313-1325.

Agardy, T., 2010. Ocean zoning- Making marine management more effective. Earthscan, London, UK.

An, L., He, G., Liang, Z., Liu, J., 2006. Impacts of demographic and socioeconomic factors on spatio-temporal dynamics of panda habitat. Biological Conservation 15, 2343-2363.

Andam, K.S., Ferraro, P.J., Pfaff, A., Sanchez-Azofeifa, G.A., Robalino, J.A., 2008. Measuring the effectiveness of protected area networks in reducing deforestation. Proceedings of the National Academy of Sciences 105, 16089-16094.

Arthur, S.M., Manly, B.F.J., McDonald, L.L., Garner, G.W., 1996. Assessing habitat selection when availability changes. Ecology 77, 215-227.

Ashbrook, D., Starner, T., 2003. Using GPS to learn significant locations and predict movement across multiple users. Personal and Ubiquitous Computing 7, 275-286.

Babcock, R.C., Shears, N.T., Alcala, A.C., Barrett, N.S., Edgar, G.J., Lafferty, K.D., McClanahan, T.R., Russ, G.R., 2010. Decadal trends in marine reserves reveal differential rates of change in direct and indirect effects. Proceedings of the National Academy of Sciences 107, 18256-18261.

Bailey, T.C., Gatrell, A.C., 1995. Interactive spatial data analysis. Prentice Hall, Harlow, England.

Banister, J., 1987. China's changing population. Stanford University Press. Redwood City, CA, USA.

Barnosky, A.D., Matzke, N., Tomiya, S., Wogan, G.O.U., Swartz, B., Quental, T.B., Marshall, C., McGuire, J.L., Lindsey, E.L., Maguire, K.C., Mersey, B., Ferrer, E.A., 2011. Has the Earth's sixth mass extinction already arrived? Nature 471, 51-57.

Batisse, M., 1997. Biosphere reserves: A challenge for biodiversity conservation and regional development. Environment: Science and Policy for Sustainable Development 39, 6-33.

Bearer, S., 2005. The effects of forest harvesting on giant panda habitat use in Wolong Nature Reserve, China. PhD Dissertation, Michigan State University, East Lansing, Michigan, USA.

Bearer, S., Linderman, M., Huang, J., An, L., He, G., Liu, J. 2008. Effects of fuelwood collection and timber harvesting on giant panda habitat use. Biological Conservation 141(2), 385- 393.

169

Beck, J.L., Peek, J.M., 2005. Diet composition, forage selection, and potential for forage competition among elk, deer, and livestock on aspen-sagebrush summer range. Rangeland Ecology and Management 58(2), 135-147.

Benhamou, S., 2011. Dynamic approach to space and habitat use based on biased random bridges. PLoS ONE 6, e14592.

Bertrand, M.R., Denicola, A.J., Beissinger, S.R., Swihart, R.K., 1996. Effects of parturition on home ranges and social affiliations of female white-tailed deer. Journal of Wildlife Management 60, 899-909.

Cagnacci, F., Boitani, L., Powell, R.A., Boyce, M.S., 2010. Animal ecology meets GPS-based radiotelemetry: A perfect storm of opportunities and challenges. Philosophical Transactions of the Royal Society B: Biological Sciences 365, 2157-2162.

Calenge, C., 2011a. Analysis of habitat selection by animals. adehabitatHS Package for R, Version 0.3.

Calenge, C., 2011b. Home range estimation. adehabitatHR Package for R, Version 0.3.

Calenge, C., Dufour, A.B., Maillard, D., 2005. K-select analysis: A new method to analyse habitat selection in radio-tracking studies. Ecological Modelling 186(2), 143-153.

Ceballos, G., Garcia, A., Ehrlich, P.R., 2010. The sixth extinction crisis: Loss of animal populations and species. Journal of Cosmology 8, 1821-1831.

Chen, X., Lupi, F., He, G., Ouyang, Z., Liu, J., 2009. Factors affecting land reconversion plans following a payment for ecosystem service program. Biological Conservation 142, 1740- 1747.

Chen, X., Lupi, F., Viña, A., He, G., Liu, J., 2010. Using cost-effective targeting to enhance the efficiency of conservation investments in payments for ecosystem services. Conservation Biology 24, 1469-1478.

Clark, W.C., Dickson, N.M., 2003. Sustainability science: The emerging research program. Proceedings of the National Academy of Sciences 100, 8059-8061.

Crutzen, P.J., 2006. The “anthropocene”. Pages 13-18 in Ehlers, E., Krafft, T., editors. Earth system science in the Anthropocene- Emerging issues and problems. Springer-Verlag, Berlin, Germany.

De Azevedo, F.C.C., Murray, D.L., 2007. Spatial organization and food habits of jaguars (Panthera onca) in a floodplain forest. Biological Conservation 137, 391-402.

DeFries, R., Hansen, A., Turner, B.L., Reid, R., Liu, J.G., 2007. Land use change around protected areas: Management to balance human needs and ecological function. Ecological Applications 17, 1031-1038.

170

Diamond, J., 2005. Collapse: How societies choose to fail or succeed. Penguin Books, New York, NY, USA.

Durnin, M.E., Swaisgood, R.R., Czekala, N., Zhang, H., 2004. Effects of radiocollars on giant panda stress-related behavior and hormones. Journal of Wildlife Management 68, 987- 992.

Eigenbrod, F., Anderson, B.J., Armsworth, P.R., Heinemeyer, A., Jackson, S.F., Parnell, M., Thomas, C.D., Gaston, K.J., 2009. Ecosystem service benefits of contrasting conservation strategies in a human-dominated region. Proceedings of the Royal Society of London, Series B: Biological Sciences 276, 2903-2911.

ESRI, 2006. ArcGIS Desktop. Version 9.2. ESRI Inc., Redlands, California, USA.

Feng, T., van Manen, F.T., Zhao, N., Li, M., Wei, F., 2009. Habitat assessment for giant pandas in the Qinling Mountain Region of China. Journal of Wildlife Management 73, 852-858.

Fieberg, J., Kochanny C.O., 2005. Quantifying home-range overlap: The importance of the utilization distribution. Journal of Wildlife Management 69, 1346-1359.

Gansu Provincial Forestry Department of Wildlife Conservation, 2004. Report of the third giant panda survey in Gansu Province, Lanzhou, Gansu, China. (In Chinese).

Garshelis, D.L., 2000. Delusions in habitat evaluation: Measuring use, selection, and importance. Pages 111-163 in Boitani, L., Fuller, T.K., editors. Research techniques in animal ecology- Controversies and consequences. Columbia University Press, New York, New York, USA.

Garshelis, D. L., 2004. Variation in ursid life histories: Is there an outlier? Pages 53-73 in Lindburg, D., Baragona, K., editors. Giant pandas: Biology and conservation. University of California Press, Berkeley, California, USA.

Geneletti, D., van Duren, I., 2008. Protected area zoning for conservation and use: A combination of spatial multicriteria and multiobjective evaluation. Landscape and Urban Planning 85, 97-110.

Ghimire, K.B., 1997. Conservation and social development: An assessment of Wolong and other panda reserves in China. Pages 187-213 in Ghimire, K.B., Pimbert, M.P., editors. Social change and conservation: Environmental politics and impacts of national parks and protected areas. Earthscan Publications, London, UK.

Gill, R., Saucedo Galvez, C., Aldridge, D., Morgan, G., 2008. Ranging behaviour of huemul in relation to habitat and landscape. Journal of Zoology 274, 254-260.

Gonzales, E.K., Arcese, P., Schulz, R., Bunnell, F.L., 2003. Strategic reserve design in the central coast of British Columbia: Integrating ecological and industrial goals. Canadian Journal of Forest Research-Revue Canadienne De Recherche Forestiere 33, 2129-2140.

171

Guo, H., 2003. Impacts of human disturbance on the giant panda (Ailuropoda melanoleuca) and its habitats at the Wanglang Nature Reserve. Masters Thesis, Sichuan University, Chengdu, China. (In Chinese).

Hannah, L., Midgley, G., Andelman, S., Araujo, M., Hughes, G., Martinez-Meyer, E., Pearson, R., Williams, P., 2007. Protected area needs in a changing climate. Frontiers in Ecology and the Environment 5, 131-138.

Harkness, J., 1998. Recent trends in forestry and conservation of biodiversity in China. The China Quarterly 156, 911-934.

Harris, R.B., 2008. Wildlife conservation in China- Preserving the habitat of China's wild west. M.E. Sharpe, Inc., Armonk, New York, USA.

He, G., Chen, X., Bearer, S., Colunga, M., Mertig, A., An, L., Zhou, S., Linderman, M., Ouyang, Z., Gage, S., Li, S., Liu, J., 2009. Spatial and temporal patterns of fuelwood collection in Wolong Nature Reserve: Implications for panda conservation. Landscape and Urban Planning 92, 1-9.

Hebblewhite, M., Haydon, D.T., 2010. Distinguishing technology from biology: A critical review of the use of GPS telemetry data in ecology. Philosophical Transactions of the Royal Society B-Biological Sciences 365, 2303-2312.

Hirzel, A., Hausser, J., Chessel, D., Perrin, N., 2002. Ecological-niche factor analysis: How to compute habitat-suitability maps without absence data? Ecology 83, 2027-2036.

Hjortso, C.N., Straede, S., Helles, F., 2006. Applying multi-criteria decision-making to protected areas and buffer zone management: A case study in the Royal Chitwan National Park, Nepal. Journal of Forest Economics 12, 91-108.

Hoekstra, J.M., Boucher, T.M., Ricketts, T.H., Roberts, C., 2005. Confronting a biome crisis: Global disparities of habitat loss and protection. Ecology Letters 8, 23-29.

Hu, J., 1981. Research on the biology of giant panda, snub-nosed monkey and takin in Wolong Nature Reserve. Sichuan People's Publishing House, Chengdu, China. (In Chinese).

Hu, J., 2001. Research on the giant panda. Shanghai Science and Education Publishing House, Shanghai, China. (In Chinese).

Hu, J., Schaller, G.B., Pan, W., Zhu, J., 1985. Wolong's giant panda. Sichuan Publishing House of Science and Technology, Sichuan, China. (In Chinese).

Hu, J., F. Wei., 2004. Comparative ecology of giant pandas in the five mountain ranges of their distribution in China. Pages 137-148 in Lindburg, D., Baragona, K., editors. Giant pandas: Biology and conservation. University of California Press, Berkeley, California, USA.

172

Hubei Environmental Protection Agency, 2008. Activities of seven environmental protection departments for stregthening protected area management. .

Hull, V., Shortridge, A., Liu, B., Bearer, S., Zhou, X., Huang, J., Zhou, S., Zhang, H., Ouyang, Z., Liu, J., 2011a. The impact of giant panda foraging on bamboo dynamics in an isolated environment. Plant Ecology 212, 43-54.

Hull, V., Xu, W., Liu, W., Zhou, S., Viña, A., Zhang, J., Tuanmu, M.-N., Huang, J., Linderman, M., Chen, X., Huang, Y., Ouyang, Z., Zhang, H., Liu, J., 2011b. Evaluating the efficacy of zoning designations for protected area management. Biological Conservation 144, 3028-3037.

Hull, V., Zhang, J., Zhou, S., Huang, J., Viña, A., Liu, W., Tuanmu, M.-N., Li, R., Liu, D., Xu, W., 2014. Impact of livestock on giant pandas and their habitat. Journal for Nature Conservation 22(3), 256-264.

Hunt, R.M., 2004. A paleontologist's perspective on the origin and relationships of the giant panda. Pages 45-52 in Lindburg, D., Baragona, K., editors. Giant pandas- Biology and conservation. University of California Press, Berkeley, CA, USA.

Jenkins, C.N., Joppa, L., 2009. Expansion of the global terrestrial protected area system. Biological Conservation 142, 2166-2174.

Jenness, J.S., 2004. Calculating landscape surface area from digital elevation models. Wildlife Society Bulletin 32, 829-839.

Jenness, J., Brost, B, Beier, P., 2011. Land Facet Corridor Designer: Extension for ArcGIS. Jenness Enterprises. Available at: .

Jim, C.Y., Xu, S.S.W., 2002. Stifled stakeholders and subdued participation: Interpreting local responses toward Shimentai Nature Reserve in South China. Environmental Management 30, 327-341.

Jim, C.Y., Xu, S.S.W., 2004. Recent protected-area designation in China: An evaluation of administrative and statutory procedures. The Geographical Journal 170, 39-50.

Jin, C., Ciochon, R.L., Dong, W., Hunt, R.M., Liu, J., Jaeger, M., Zhu, Q., 2007. The first skull of the earliest giant panda. Proceedings of the National Academy of Sciences 104, 10932- 10937.

Johnson, D.H., 1980. The comparison of usage and availability measurements for evaluating resource preference. Ecology 61, 65-71.

Kang, D., Kang, W., Tan, L., Li, J., 2011a. The habitat selection of giant panda in Wanglang Nature Reserve, Sichuan Province, China. Acta Ecologica Sinica 31, 401-409. (In Chinese).

173

Kang, D., Zhao, Z., Guo, W., Tan, L., Kang, W., Li, J., 2011b. Habitat selection attributes of giant panda. Chinese Journal of Applied Ecology 22, 519-525. (In Chinese).

Kang, D., Zhao, L., Song, G., 2011c. Competition relationship between giant panda and livestock in Wanglang National Nature Reserve, Sichuan. Journal of Northeast Forestry University 39(7), 74-76. (In Chinese).

Kang, D., Yang, H.W., Li, J.Q., Chen, Y.P., Zhao, L.J., 2013. Habitat use by giant pandas Ailuropoda melanoleuca in the Wanglang Nature Reserve, Sichuan, China. Zoological Studies 52, 1-6. (In Chinese).

Kenchington, R., Day, J., 2011. Zoning, a fundamental cornerstone of effective Marine Spatial Planning: Lessons learnt from the Great Barrier Reef, Australia. Journal of Coastal Conservation 15, 271-278.

Kernohan, B.J., Gitzen, R.A., Millspaugh, J.J., 2001. Analysis of animal space use and movements. Pages 126-166 in Millspaugh, J.J., Marzluff, J.M., editors. Radio tracking and animal populations. Academic Press, San Deigo, California, USA.

Kramer-Schadt, S., Revilla, E., Wiegand, T., Breitenmoser, U., 2004. Fragmented landscapes, road mortality and patch connectivity: Modelling influences on the dispersal of Eurasian lynx. Journal of Applied Ecology 41, 711-723.

Lan, J., Lu, W., Zhang, Z., Li, G., Wang C., Yu, J., Wang, J., Huang, X., 2003. Fall estrus and mating in a giant panda. Sichuan Journal of Zoology 22, 9-11. (In Chinese).

Laver, P.N., Kelly, M.J., 2008. A critical review of home range studies. Journal of Wildlife Management 72, 290-298.

Leakey, R.E., Lewin, R., 1996. The sixth extinction: Patterns of life and the future of humankind. Anchor Books, New York, USA.

Lechowicz, M.J., 1982. The sampling characteristics of electivity indices. Oecologia 52, 22-30.

Li, C., Zhou, S., Xiao, D., Chen, Z., Tian, Z., 1992. A general description of Wolong Nature Reserve. Pages 313-325 in Wolong Nature Reserve, Sichuan Normal College, editors. The animal and plant resources and protection of Wolong Nature Reserve. Sichuan Publishing House of Science and Technology, Chengdu, Sichuan, China.

Li, W., Han, N., 2001. Ecotourism management in China's nature reserves. AMBIO: A Journal of the Human Environment 30, 62-63.

Lichstein, J.W., Simons, T.R., Shriner, S.A., Franzreb, K.E., 2002. Spatial autocorrelation and autoregressive models in ecology. Ecological Monographs 72, 445-463.

Linderman, M., Bearer, S., An, L., Tan, Y., Ouyang, Z., Liu, J., 2004. The effects of understory bamboo on broad-scale estimates of giant panda habitat. Biological Conservation 121, 383-390.

174

Linderman, M.A., An, L., Bearer, S., He, G., Ouyang, Z., Liu, J., 2005. Modeling the spatio- temporal dynamics and interactions of households, landscapes, and giant panda habitat. Ecological Modelling 183, 47-65.

Linderman, M., Bearer, S., An, L., He, G., Ouyang, Z., Liu, J., 2006. Interactive effects of natural and human disturbances on vegetation dynamics across landscapes. Ecological Applications 16, 452-463.

Liu, J., Ouyang, Z., Taylor, W.W., Groop, R., Tan, K., Zhang, H., 1999. A framework for evaluating the effects of human factors on wildlife habitat: The case of giant pandas. Conservation Biology 13, 1360-1370.

Liu, J., Linderman, M., Ouyang, Z., An, L., Yang, J., Zhang, H., 2001. Ecological degradation in protected areas: The case of Wolong Nature Reserve for giant pandas. Science 292, 98- 101.

Liu, J., Daily, G.C., Ehrlich, P.R., Luck, G.W., 2003. Effects of household dynamics on resource consumption and biodiversity. Nature 421, 530-533.

Liu, J., Diamond, J., 2005. China's environment in a globalizing world. Nature 435, 1179-1186.

Liu, J., Dietz, T., Carpenter, S.R., Alberti, M., Folke, C., Moran, E., Pell, A.N., Deadman, P., Kratz, T., Lubchenco, J., Ostrom, E., Ouyang, Z., Provencher, W., Redman, C.L., Schneider, S.H., Taylor, W.W., 2007a. Complexity of coupled human and natural systems. Science 317, 1513-1516.

Liu, J., Dietz, T., Carpenter, S.R., Folke, C., Alberti, M., Redman, C.L., Schneider, S.H., Ostrom, E., Pell, A.N., Lubchenco, J., Taylor, W.W., Ouyang, Z., Deadman, P., Kratz, T., Provencher, W., 2007b. Coupled human and natural systems. AMBIO: A Journal of the Human Environment 36, 639-649.

Liu, J., Li, S., Ouyang, Z., Tam, C., Chen, X., 2008. Ecological and socioeconomic effects of China's policies for ecosystem services. Proceedings of the National Academy of Sciences 105, 9477-9482.

Liu, J., Raven, P., 2010. China's environmental challenges and implications for the world. Critical Reviews in Environmental Science and Technology 40, 823-851.

Liu, J., Hull, V., Batistella, M., DeFries, R., Dietz, T., Fu, F., Hertel, T.W., Izaurralde, R.C., Lambin, E.F., Li, S., Martinelli, L.A., McConnell, W.J., Moran, E.F., Naylor, R., Ouyang, Z., Polenske, K.R., Reenberg, A., de Miranda Rocha, G., Simmons, C.S., Verberg, P.H., Vitousek, P.M., Zhang, F., Zhu, C., 2013. Framing sustainability in a telecoupled world. Ecology and Society 18(2), 26. Liu, W., 2012. Patterns and impacts of tourism development in a coupled human and natural system. PhD Dissertation, Department of Fisheries and Wildlife. Michigan State University, East Lansing, MI, USA.

175

Liu, X., 2001. Mapping and modelling the habitat of giant pandas in Foping Nature Reserve, China. PhD Dissertation, Division of Agriculture, Conservation and Environment (ACE), Wageningen University, Enschede, The Netherlands.

Liu, X., Bronsveld, M.C., Skidmore, A.K., Wang, T., Dang, G., Yong, Y., 2004. Mapping habitat suitability for giant pandas in Foping Nature Reserve, China. Pages 176-186 in Lindburg, D., Baragona, K., editors. Giant pandas- Biology and conservation. University of California Press, Berkeley, CA, USA.

Liu, X., Toxopeus, A.G., Skidmore, A.K., Shao, X., Dang, G., Wang, T., Prins, H.H.T., 2005. Giant panda habitat selection in Foping Nature Reserve, China. Journal of Wildlife Management 69, 1623-1632.

Liu, X., Li, J., 2008. Scientific solutions for the functional zoning of nature reserves in China. Ecological Modelling 215, 237-246.

Liu, X., Cheng, X., Skidmore, A.K., 2011. Potential solar radiation pattern in relation to the monthly distribution of giant pandas in Foping Nature Reserve, China. Ecological Modelling 222, 645-652.

Long, J. A., 2012. A suite of R tools for exploring dynamic interaction patterns in wildlife telemetry data. Version 0.1.

Loucks, C.J., Lü, Z., Dinerstein, E., Wang, D., Olson, D.M., Zhu, C., Wang, D., 2001. Giant pandas in a changing landscape. Science 294, 1465.

Loucks, C.J., Lü, Z., Dinerstein, E., Wang, D., Fu, D., Wang, H., 2003. The giant pandas of the Qinling Mountains, China: A case study in designing conservation landscapes for elevational migrants. Conservation Biology 17, 558-565.

Lu, X., Jiang, Z., Li, C., 2007. Comparative habitat use by giant panda, Ailuropoda melanoleuca in selectively logged forests and timber plantations. Folia Zoologica 56, 137-143.

Lü, Z., Kemf, E., 2001. Giant pandas in the wild: A WWF species status report. WWF International, Gland, Switzerland.

Macdonald, D.W., Rushton, S., 2003. Modelling space use and dispersal of mammals in real landscapes: A tool for conservation. Journal of Biogeography 30, 607-620.

Mackinnon, J., 2008. Species richness and adaptive capacity in animal communities: Lessons from China. Integrative Zoology 3, 95-100.

MacLeod, S.B., Kerley, G.I.H., Gaylard, A., 1996. Habitat and diet of bushbuck Tragelaphus scriptus in the Woody Cape Nature Reserve: Observations from faecal analysis. South African Journal of Wildlife Research 26, 19-25.

Madhusudan, M.D., 2004. Recovery of wild large herbivores following livestock decline in a tropical Indian wildlife reserve. Journal of Applied Ecology 41(5), 858-869.

176

Manly, B.F.J., McDonald, L.L., Thomas, D.L., McDonald, T.L., Erickson, W.P., 2002. Resource selection by animals—Statistical design and analysis for field studies. Second edition. Kluwer Academic Publishers, Dordrecht, The Netherlands.

Martin, J., Basille, M., Van Moorter, B., Kindberg, J., Allaine, D., Swenson, J.E., 2010. Coping with human disturbance: Spatial and temporal tactics of the brown bear (Ursus arctos). Canadian Journal of Zoology 88, 875-883.

Marzluff, J.M., Knick, S.T., Millspaugh, J.J., 2001. High-tech behavioral ecology: Modeling the distribution of animal activities to better understand wildlife space use and resource selection. Pages 310-326 in Millspaugh, J.J., Marzluff, J.M., editors. Radio tracking and animal populations. Academic Press, San Diego, CA, USA.

Marzluff, J.M., Millspaugh, J.J., Hurvitz, P., Handcock, M.S., 2004. Relating resources to a probabilistic measure of space use: Forest fragments and steller's jays. Ecology 85, 1411- 1427.

McDonald, L.L., 2004. Sampling rare populations. Pages 11-42 in Thompson, W.L., editor. Sampling rare or elusive species. Island Press, Washington, DC, USA.

McNeely, J.A., 1994. Protected areas for the 21st century: Working to provide benefits to society. Biodiversity and Conservation 3, 390-405.

Melick, D., Yang, X., Xu, J., 2007. Seeing the wood for the trees: How conservation policies can place greater pressure on village forests in southwest China. Biodiversity and Conservation 16, 1959-1971.

Menard, C., Duncan, P., Fleurance, G., Georges, J.-Y., Lila, M., 2002. Comparative foraging and nutrition of horses and cattle in European wetlands. Journal of Applied Ecology 39(1), 120-133.

Menzel, C.R., 1991. Cognitive aspects of foraging in Japanese monkeys. Animal Behaviour 41, 397-402.

Miller, C.S., Hebblewhite, M., Goodrich, J.M., Miquelle, D.G., 2010. Review of research methodologies for tigers: Telemetry. Integrative Zoology 5, 378-389.

Miller, J.A., 2012. Using spatially explicit simulated data to analyze animal interactions: A case study with brown hyenas in northern Botswana. Transactions in GIS 16, 271-291.

Millspaugh, J.J., Nielson, R.M., McDonald, L., Marzluff, J.M., Gitzen, R.A., Rittenhouse, C.D., Hubbard, M.W., Sheriff, S.L., 2006. Analysis of resource selection using utilization distributions. Journal of Wildlife Management 70, 384-395.

Ministry of Forestry, 1998. Wolong Nature Reserve Master Plan, Beijing, China. (In Chinese).

Minta, S.C., 1992. Tests of temporal and spatial interaction among animals. Ecological Applications 2, 178-188.

177

Mishra, C., Van Wieren, S.E., Ketner, P., Heitkonig, I.M.A., Prins, H.H.T., 2004. Competition between domestic livestock and wild bharal Pseudois nayaur in the Indian Trans- Himalaya. Journal of Applied Ecology 41(2), 344-354.

Morrison, M.L., Marcot, B., Mannan, W., 2006. Wildlife-habitat relationships: Concepts and applications. Island Press, Washington, DC, USA.

Murphy, K., Huettmann, F., Fresco, N., Morton, J., 2010. Connecting Alaska landscapes into the future. U.S. Fish and Wildlife Service, and the University of Alaska. .

Myers, N., Mittermeier, R.A., Mittermeier, C.G., Da Fonseca, G.A., Kent, J., 2000. Biodiversity hotspots for conservation priorities. Nature 403, 853-858.

Namgail, T., Fox, J.L., Bhatnagar, Y.V., 2007. Habitat shift and time budget of the Tibetan argali: The influence of livestock grazing. Ecological Research 22(1), 25-31.

Naughton-Treves, L., Holland, M.B., Brandon, K., 2005. The role of protected areas in conserving biodiversity and sustaining local livelihoods. Annual Review of Environmental Resources 30, 219-252.

Neu, C.W., Byers, C.R., Peek, J.M., 1974. A technique for analysis of utilization availability data. Journal of Wildlife Management 38, 541-545.

Nielsen, S.E., McDermid, G., Stenhouse, G.B., Boyce, M.S., 2010. Dynamic wildlife habitat models: Seasonal foods and mortality risk predict occupancy-abundance and habitat selection in grizzly bears. Biological Conservation 143, 1623-1634.

Ouyang, Z., Xu, W., Wang, X., Wang, W., Dong, R., Zheng, H., Li, D., Li, Z., Zhang, H., Zhuang C., 2008. Impact assessment of Wenchuan earthquake on ecosystems. Acta Ecologica Sinica 28(12), 5801-5809. (In Chinese).

Pan, W., Lü, Z., Zhu, X., Wang, D., Wang, H., Long, Y., Fu, D., Zhou, X., 2001. A chance for lasting survival. Beijing University Press, Beijing, China. (In Chinese).

Peng, R., Zeng, B., Meng, X., Yue, B., Zhang, Z., Zou, F., 2007. The complete mitochondrial genome and phylogenetic analysis of the giant panda (Ailuropoda melanoleuca). Gene 397, 76-83.

Pimm, S.L., Russell, G.J., Gittleman, J.L., Brooks, T.M., 1995. The future of biodiversity. Science 269, 347-350.

Pimm, S.L., Raven, P., 2000. Biodiversity: Extinction by numbers. Nature 403, 843-845.

Poole, K.G., 1995. Spatial organization of a lynx population. Canadian Journal of Zoology 73, 632-641.

178

Powell, R.A., 2000. Animal home ranges and territories and home range estimators. Pages 65- 110 in Boitani, L., Fuller, T.K., editors. Research techniques in animal ecology: Contraversies and consequences. Columbia University Press, New York, NY, USA.

Powell, R.A., 2012. Movements, home ranges, activity, and dispersal. Pages 188-217 in Boitani, L., Powell, R.A., editors. Carnivore ecology and conservation: A handbook of techniques. Oxford University Press, London, UK.

Qi, D., Hu, Y., Gu, X., Li, M., Wei, F., 2009. Ecological niche modeling of the sympatric giant and red pandas on a mountain-range scale. Biodiversity and Conservation 18, 2127-2141.

Qi, D., Zhang, S., Zhang, Z., Hu, Y., Yang, X., Wang, H., Wei, F., 2011. Different habitat preferences of male and female giant pandas. Journal of Zoology 285, 205-214.

Qi, D., Zhang, S., Zhang, Z., Hu, Y., Yang, X., Wang, H., Wei, F., 2012. Measures of giant panda habitat selection across multiple spatial scales for species conservation. Journal of Wildlife Management 76, 1092-1100.

R Development Core Team, 2005. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.

Rambaldi, G., Kyem, P.A.K., McCall, M., Weiner, D., 2006. Participatory spatial information management and communication in developing countries. The Electronic Journal of Information Systems in Developing Countries 25, 1-9.

Ran, J., 2003. Habitat selection by giant pandas and grazing livestock in the Xiaoxiangling Mountains of Sichuan Province. Journal of Sun Yatsen University Social Science Edition 23, 2253-2259. (In Chinese).

Ran, J., Liu, S., Wang, H., Sun, Z., Zeng, Z., Liu, S., 2003. Effect of grazing on giant pandas' habitat in Yele Nature Reserve. Acta Theriologica Sinica 23, 288-294 (In Chinese).

Ran, J., Liu, S., Wang, H., Zeng, Z., Sun, Z., Liu, S., 2004a. A survey of disturbance of giant panda habitat in the Xiaoxiangling Mountains of Sichuan Province. Acta Theriologica Sinica 24, 277-281. (In Chinese).

Ran, J., Zeng, Z., Wang, H., Liu, S., Wang, H., Liu, S., 2004b. A comparative study on habitat preference of giant pandas in primary and secondary forests. Journal of Beijing Forestry University 26, 8-14. (In Chinese).

Reid, D.G., Hu, J., Sai, D., Wei, W., Huang, Y., 1989. Giant panda Ailuropoda melanoleuca behavior and carrying capacity following a bamboo die-off. Biological Conservation 49, 85-104.

Reid, D.G., Hu, J., 1991. Giant panda selection between Bashania fangiana bamboo habitats in Wolong Reserve, Sichuan, China. Journal of Applied Ecology 28, 228-243.

179

Reid, D.G., Gong, J., 1999. Giant panda conservation action plan. Pages 241-254 in Servheen, C., Herrero, S., Peyton, B., editors. Bears: Status survey and conservation action plan. IUCN/SSC Bear and Polar Bear Specialist Groups, Gland, Switzerland.

Remis, M.J., Kpanou, J.B., 2011. Primate and ungulate abundance in response to multi-use zoning and human extractive activities in a Central African Reserve. African Journal of Ecology 49, 70-80.

Rindfuss, R.R., Entwisle, B., Walsh, S.J., An, L., Badenoch, N., Brown, D.G., Deadman, P., Evans, T.P., Fox, J., Geoghegan, J., Gutmann, M., Kelly, M., Linderman, M., Liu, J., Malanson, G.P., Mena, C.F., Messina, J.P., Moran, E.F., Parker, D.C., Parton, W., Prasartkul, P., Robinson, D.T., Sawangdee, Y., Vanwey, L.K., Verburg, P.H., 2008. Land use change: Complexity and comparisons. Journal of Land Use Science 3(1), 1-10.

Sabatini, M.d.C., Verdiell, A., Rodriguez Iglesias, R.M., Vidal, M., 2007. A quantitative method for zoning of protected areas and its spatial ecological implications. Journal of Environmental Management 83, 198-206.

Salomon, A.K., Waller, N.P., McIlhagga, C., Yung, R.L., Walters, C., 2002. Modeling the trophic effects of marine protected area zoning policies: A case study. Aquatic Ecology 36, 85-95.

Schaller, G.B., Hu, J., Pan, W., Zhu, J., 1985. The giant pandas of Wolong. University of Chicago Press, Chicago, Illinois, USA.

Schofield, G., Bishop, C.M., MacLean, G., Brown, P., Baker, M., Katselidis, K.A., Dimopoulos, P., Pantis, J.D., Hays, G.C., 2007. Novel GPS tracking of sea turtles as a tool for conservation management. Journal of Experimental Marine Biology and Ecology 347, 58-68.

Seaman, D.E., Powell, R.A., 1990. Identifying patterns and intensity of home range use. International Conference on Bear Research and Management 8, 243-249.

Sikes, R.S., Gannon, W.L., The Animal Care and Use Committee of the American Society of Mammalogists., 2011. Guidelines of the American Society of Mammalogists for the use of wild mammals in research. Journal of Mammalogy 92, 235-253.

Southwood, T., 1977. Habitat, the templet for ecological strategies? The Journal of Animal Ecology 46, 337-365.

State Forestry Administration, 1999. National focal protected wild plant list of China, Beijing, China. (In Chinese).

State Forestry Administration, 2006. The 3rd national survey report on giant panda in China. Science Publisher, Beijing, China. (In Chinese).

180

Steinfeld, H., Gerber, P., Wassenaar, T., Castel, V., Rosales, M., de Haan, C., 2006. Lifestock's long shadow- Environmental issues and options. Food and Agriculture Organization (FAO), Rome, Italy.

Swaisgood, R.R., Lindburg, D., White, A.M., Zhang, H., Zhou, X., 2004. Chemical communication in giant pandas: Experimentation and application. Pages 106-120 in Lindburg, D., Baragona, K., editors. Giant pandas: Biology and conservation. University of California Press, Berkeley, California, USA.

Swaisgood, R.R., Lindburg, D.G., Zhou, X.P., 1999. Giant pandas discriminate individual differences in conspecific scent. Animal Behaviour 57, 1045-1053.

Tan, Y., Ouyang, Z., Zhang, H., 1995. Spatial characteristics of biodiversity in Wolong Nature Reserve. China's Biosphere Reserve 3, 19-24. (In Chinese).

Tang, P., Hu, J., 1998. Habitat selection of giant pandas in Yele Natural Reserve. Pages 33-36 in Hu, J., Wu, Y., editors. The resources and protection of vertebrates. Sichuan Scientific and Technological Press, Chengdu, Sichuan, China. (In Chinese).

Tarou, L.R., Snyder, R.J., Maple, T.L., 2004. Spatial memory in the giant panda. Pages 101-105 in Lindburg, D., Baragona, K., editors. Giant pandas: Biology and conservation. University of California Press, Berkeley, CA, USA.

Taylor, A.H., Qin, Z., 1987. Culm dynamics and dry matter production of bamboos in the Wolong and Tangjiahe giant panda reserves, Sichuan, China. Journal of Applied Ecology 24, 419-433.

Taylor, A.H., Huang, J., Zhou, S., 2004. Canopy tree development and undergrowth bamboo dynamics in old-growth Abies-Betula forests in southwestern China: A 12-year study. Forest Ecology and Management 200, 347-360.

The State Council, 1994. Bylaws of nature reserves in the People's Republic of China, Beijing, China. (In Chinese).

Tuanmu, M.-N., Viña, A., Bearer, S., Xu, W., Ouyang, Z., Zhang H., Liu, J., 2010. Mapping understory vegetation using phenological characteristics derived from remotely sensed data. Remote Sensing of Environment 114, 1833-1844.

Tuanmu, M.-N., Viña, A., Roloff, G., Liu, W., Ouyang, Z., Zhang, H., Liu, J., 2011. Temporal transferability of wildlife habitat models: Implications for habitat monitoring. Journal of Biogeography 38, 1510-1523.

Tuanmu, M.-N., Viña, A., Winkler, J.A., Li, Y., Xu, W., Ouyang, Z., Liu, J., 2012. Climate change impacts on understory bamboo species and giant pandas in China's Qinling Mountains. Nature Climate Change 3, 249-253.

Van Winkle, W., 1975. Comparison of several probabilistic home-range models. Journal of Wildlife Management 39, 118-123.

181

Vander Wal, E., Rodgers, A.R., 2012. An individual-based quantitative approach for delineating core areas of animal space use. Ecological Modelling 224, 48-53.

Vanderploeg, H., Scavia, D., 1979. Two electivity indices for feeding with special reference to zooplankton grazing. Journal of Fisheries Research Board of Canada 36, 362-365.

Villa, F., Tunesi, L., Agardy, T., 2002. Zoning marine protected areas through spatial multiple- criteria analysis: The case of the Asinara Island National Marine Reserve of Italy. Conservation Biology 16, 515-526.

Viña, A., Bearer, S., Chen, X., Hu, G., Linderman, M., An, L., Zhang, H., Ouyang, Z., Liu, J., 2007. Temporal changes in giant panda habitat connectivity across boundaries of Wolong Nature Reserve, China. Ecological Applications 17, 1019-1030.

Viña, A., Bearer, S., Zhang, H., Ouyang, Z., Liu, J., 2008. Evaluating MODIS data for mapping wildlife habitat distribution. Remote Sensing of Environment 112, 2160-2169.

Viña, A., Tuanmu, M.-N., Xu, W., Li, Y., Ouyang, Z., DeFries, R., Liu, J., 2010. Range-wide analysis of wildlife habitat: Implications for conservation. Biological Conservation 143, 1960-1969.

Viña, A., Chen, X., McConnell, W., Liu, W., Xu, W., Ouyang, Z., Zhang, H., Liu, J., 2011. Effects of natural disasters on conservation policies: The case of the 2008 Wenchuan Earthquake, China. Ambio 40, 274-284.

Wang, J., Huang, H., Teng, J., Yang, W., Ma, T., 2006. Habitat selection by giant pandas in west of Baishuijiang Nature Reserve. Sichuan Journal of Zoology 25, 771-775. (In Chinese).

Wang, T., 2003. Habitat analysis for giant panda in Laoxiancheng Nature Reserve in the Qinling Mountains China. Master's Thesis, Geo-information Science and Earth Observation, International Institute for Geo-information Science and Earth Observation, Enschede, The Netherlands.

Wang, T., Ye, X., Skidmore, A.K., Toxopeus, A.G., 2010. Characterizing the spatial distribution of giant pandas (Ailuropoda melanoleuca) in fragmented forest landscapes. Journal of Biogeography 37, 865-878.

Wang, X., 2008. The impact of human disturbance on giant panda habitat in the Minshan mountains. PhD Dissertation, State Key Laboratory of Urban and Regional Ecology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing, China. (In Chinese).

Wang, X., Xu, W., Ouyang, Z., Liu, J., Xiao, Y., Chen, Y., Zhao, L., Huang, J., 2008. Application of ecological-niche factor analysis in habitat assessment of giant pandas. Acta Ecologica Sinica 28, 821-828. (In Chinese).

182

Wang, X., Xu, W., Ouyang, Z., 2009. Integrating population size analysis into habitat suitability assessment: Implications for giant panda conservation in the Minshan Mountains, China. Ecological Research 24, 1101-1109.

Watts, M.E., Ball, I.R., Stewart, R.S., Klein, C.J., Wilson, K., Steinback, C., Lourival, R., Kircher, L., Possingham, H.P., 2009. Marxan with Zones: Software for optimal conservation based land- and sea-use zoning. Environmental Modelling and Software 24, 1513-1521.

WDPA, 2009. World Database on Protected Areas (WDPA) Annual Release 2009, February 2009. .

Wei, F., Zhou, A., Hu, J., Wang, W., Yang. G., 1996. Habitat selection by giant pandas in Mabian Dafengding Reserve. Acta Theriologica Sinica 16, 241-245. (In Chinese).

Wei, F., Feng, Z., Wang, Z., 1999. Habitat selection by giant pandas and red pandas in Xiangling Mountains. Acta Zoologica Sinica 45, 57-63. (In Chinese).

Wei, F., Feng, Z., Wang, Z., Hu, J., 2000. Habitat use and separation between the giant panda and the red panda. Journal of Mammalogy 81, 448-455.

Wei, F., Zhang, Z., Hu, J., 2011. Research advances and perspectives on the ecology of wild giant pandas. Acta Theriologica Sinica 31(4), 412-421 (In Chinese).

Wei, F., Hu, Y., Zhu, L., Bruford, M.W., Zhan, X., Zhang, L., 2012. Black and white and read all over: The past, present and future of giant panda genetics. Molecular Ecology 21, 5660- 5674.

Wittemyer, G., Elsen, P., Bean, W.T., Burton, A.C.O., Brashares, J.S., 2008. Accelerated human population growth at protected area edges. Science 321, 123-126.

Wolong Administration Bureau, 2004. The chronicle of Wolong Nature Reserve. Sichuan Science and Technology Press, Chengdu, China.

Wolong Nature Reserve, 1996. Wolong Nature Reserve agricultural survey. Wolong Administration Bureau, Sichuan, China. (In Chinese).

Wolong Nature Reserve, 2005. History of the development of Wolong Nature Reserve. Sichuan Science Publisher, Chengdu, China. (In Chinese).

Wolong Nature Reserve, 2008. Wolong Nature Reserve annual agricultural report- 2008, Wolong Administration Bureau, Sichuan, China. (In Chinese).

Wolong Nature Reserve Management, 1987. Wolong vegetation and resource plants. Sichuan Publishing House of Science and Technology, Sichuan, China. (In Chinese).

World Bank, 2007. World development indicators. World Bank, Washington, DC, USA.

183

Xiao, J., Xu, W.H., Kang, D.W., Li, J.Q., 2011. Nature reserve group planning for conservation of giant pandas in North Minshan, China. Journal for Nature Conservation 19, 209-214.

Xu, W., Ouyang, Z., Viña, A., Zheng, H., Liu, J., Xiao, Y., 2006. Designing a conservation plan for protecting the habitat for giant pandas in the Qionglai mountain range, China. Diversity and Distributions 12, 610-619.

Yang, C.-h., Zhang, H.-m., Zhou, X.-p., Wang, P.-y., Wang, X.-m., 2006. Review of habitat selection in the giant panda (Ailuropoda melanoleuca). Acta Ecologica Sinica 26, 3442- 3453. (In Chinese).

Yang, W., Liu, W., Viña, A., Luo, J., He, G., Ouyang, Z., Zhang, H., Liu, J., 2013. Performance and prospects of payments for ecosystem services programs: Evidence from China. Journal of Environmental Management 127, 86-95.

Ye, X., 2008. Characterizing the spatial distribution of giant pandas in China using MODIS data and landscape metrics. Master's Thesis, International Institute for Geo-Information Science and Earth Observation (ITC), Enschede, The Netherlands.

Yong, Y., Liu, X., Wang, T., Skidmore, A.K., Prins, H.H.T., 2004. Giant panda migration and habitat utilization in Foping Nature Reserve, China. Pages 159-169 in Lindburg, D., Baragona, K., editors. Giant pandas: Biology and conservation. University of California Press, Berkeley, California, USA.

Young, T.P., Palmer, T.M., Gadd, M.E., 2005. Competition and compensation among cattle, zebras, and elephants in a semi-arid savanna in Laikipia, Kenya. Biological Conservation 122(2), 351-359.

Yu, G.-Z., Jiang, Z.-G., 2003. Buffer zones: Pattern, function and design principles. Biodiversity Science 11, 256-261. (In Chinese).

Zeng, Z., Yue, B., Ran, J., Liu, S., Chen, Y., Jiang, S., 2002. Panda's exploitation of habitats at the Wanglang Nature Reserve. Journal of Sichuan University (Natural Science Edition) 39, 1140-1144. (In Chinese).

Zhang, H., Li, D., Wei, R., Tang, C., Tu, J., 1997. Advances in conservation and studies on reproductivity of giant pandas in Wolong. Sichuan Journal of Zoology 16, 31-33. (In Chinese).

Zhang, P., Shao, G., Zhao, G., Master, D.C.l., Parker, G.R., Dunning, J.B., Li, Q., 2000. China's forest policy for the 21st century. Science 288, 2135-2136.

Zhang, Z., J. Hu., 2000. A study on the giant panda's habitat selection. Journal of Sichuan Teachers College Natural Science Edition 21, 18-21. (In Chinese).

Zhang, Z., Wei, F., Li, M., Hu, J., 2006. Winter microhabitat separation between giant and red pandas in Bashania faberi bamboo forest in Fengtongzhai Nature Reserve. Journal of Wildlife Management 70, 231-235.

184

Zhang, Z., Swaisgood, R.R., Wu, H., Li, M., Yong, Y., Hu, J., Wei, F., 2007. Factors predicting den use by maternal giant pandas. Journal of Wildlife Management 71, 2694-2698.

Zhang, Z., Zhan, X., Yan, L., Li, M., Hu, J., Wei, F., 2009. What determines selection and abandonment of a foraging patch by wild giant pandas (Ailuropoda melanoleuca) in winter. Environmental Science Pollution Research 16, 79-84.

Zhang, Z., Swaisgood, R.R., Zhang, S., Nordstrom, L.A., Wang, H., Gu, X., Hu, J., Wei, F., 2011. Old-growth forest is what giant pandas really need. Biology Letters 7, 403-406.

Zhang, Z., Sheppard, J.K., Swaisgood, R.R., Wang, G., Nie, Y., Wei, W., Zhao, N., Wei, F., 2014. Ecological scale and seasonal heterogeneity in the spatial behaviors of giant pandas. Integrative Zoology 9, 46-60.

Zhou, Z., Pan, W., 1997. Analysis of the viability of a giant panda population. Journal of Applied Ecology 34, 363-374.

Zhu, J., Long, Z., 1983. The vicissitudes of the giant panda. Acta Zoologica Sinica 29, 93-104. (In Chinese).

Zhu, L., Wu, Q., Dai, J., Zhang, S., Wei, F., 2011. Evidence of cellulose metabolism by the giant panda gut microbiome. Proceedings of the National Academy of Sciences 108, 17714- 17719.

Zhu, L., Zhan, X., Meng, T., Zhang, S., Wei, F., 2010. Landscape features influence gene flow as measured by cost-distance and genetic analyses: A case study for giant pandas in the Daxiangling and Xiaoxiangling Mountains. BMC Genetics 11, 1-11.

185