Integration of Remote Sensing, Modeling, and Field Approaches for Rangeland Management and Endangered Species Conservation in Central Asia

SUNY College of Environmental Science and Forestry Digital Commons @ ESF

Dissertations and Theses

Spring 5-17-2018

Integration of remote sensing, modeling, and field approaches for rangeland management and endangered species conservation in Central Asia

Mikhail Paltsyn [email protected]

Follow this and additional works at: https://digitalcommons.esf.edu/etds

Recommended Citation Paltsyn, Mikhail, "Integration of remote sensing, modeling, and field approaches for rangeland management and endangered species conservation in Central Asia" (2018). Dissertations and Theses. 36. https://digitalcommons.esf.edu/etds/36

This Open Access Dissertation is brought to you for free and open access by Digital Commons @ ESF. It has been accepted for inclusion in Dissertations and Theses by an authorized administrator of Digital Commons @ ESF. For more information, please contact [email protected], [email protected]. INTEGRATION OF REMOTE SENSING, MODELING, AND FIELD APPROACHES FOR

RANGELAND MANAGEMENT AND ENDANGERED SPECIES CONSERVATION

IN CENTRAL ASIA

Mikhail Yu. Paltsyn

A thesis submitted in partial fulfillment of the requirements for the Doctor of Philosophy Degree State University of New York College of Environmental Science and Forestry Syracuse, New York May 2018

Department of Environmental and Forest Biology

Approved by: James P. Gibbs, Major Professor Kelley J. Donaghy, Chair, Examining Committee Neil H. Ringler, Department Chair S. Scott Shannon, Dean, The Graduate School

ACKNOWLEDGMENTS

This thesis was completed in collaboration with many wonderful people I am very grateful to. First of all, I am very grateful to James P. Gibbs, my Major Professor, for his overall support of my work, wise guidance, and kind assistance on all stages of the thesis development. I would like to thank Giorgos Mountrakis for his wonderful lectures on remote sensing and spatial analysis, bright ideas, advice on methodology, and kind review of my thesis chapters. I very thankful to Liza Iegorova, my friend, faithful assistant and co-author of my publications, for her great help and hard work on all stages of modeling and analysis of my data.

I am extremely grateful to members of my Steering Committee: Jacqueline Frair, for her inspiring course on the Applied Wildlife Science and fair comments on the thesis chapters; Paul Hirsch, for his kind guidance on conservation ethics, community-based conservation, tradeoffs between conservation and development, and his wonderful course on the climate change issues; Rodney Jackson, for his inspiration to devote 15 years of my life to conservation of snow leopards; and Kelley Donaghy for her kind chairing of my Examining Committee.

I am very thankful to my field collaborators and friends who spent countless days, shared unforgettable camp fires with me in the mountains of Altai-Sayan, and provided invaluable help in collecting data for this thesis: Sergey Spitsyn, Alexander Kuksin, Stepan Denisov, Aduchy Beletov, Sandrash Samoilov, Atay Ayatkhan, Munkhtogtogh Ochirjav, Onon Yondon, Andrey Chelyshev, Leonid Baylagasov, Yury Robertus, Khairat, Bagdat, Jukhan, Totembek, and Hamish Gibbs.

I am grateful to my parents, Yury and Elena Paltsyn, for their overall support, love and understanding. Finally, I am very thankful to my wife Svetlana Kozlova, and my sons Egor, Kirill, and Nikolay for their love, care, and inspiration to complete this thesis.

This work was funded by the United State Agency for International Development (USAID) Climate Change Resilient Development Program (grant # CCRDCS0007), National Aeronautics and Space Administration (NASA) Land Cover Land Use Change Program (grant # NNX15AD42G), and World Wide Fund for Nature (grants # WWF566, WWF706, and WWF837).

TABLE OF CONTENTS LIST OF TABLES ...... iv LIST OF FIGURES ...... v ABSTRACT ...... vii INTRODUCTION ...... 1 CHAPTER 1: ESTIMATION AND PREDICTION OF GRASSLAND COVER IN WESTERN MONGOLIA USING MODIS-DERIVED VEGETATION INDICES ...... 6 CHAPTER 2: INTEGRATING TRADITIONAL ECOLOGICAL KNOWLEDGE AND REMOTE SENSING FOR MONITORING RANGELAND DYNAMICS IN THE ALTAI MOUNTAIN REGION ...... 21 CHAPTER 3: FORECASTING OF SNOW LEOPARD (PANTHERA UNCIA) AND ALTAI ARGALI (OVIS AMMON AMMON) HABITAT DYNAMICS IN THE ALTAI-SAYAN ECOREGION UNDER A CHANGING CLIMATE ...... 39 CHAPTER 4: TIGER RE-ESTABLISHMENT POTENTIAL TO FORMER CASPIAN TIGER (PANTHERA TIGRIS VIRGATA) RANGE IN CENTRAL ASIA ...... 62 CHAPTER 5: SNOW LEOPARD CONSERVATION IN RUSSIA: LESSONS LEARNED OVER 15 YEARS ...... 83 CONCLUSIONS...... 101 APPENDICES ...... 104 APPENDIX 1. SUPPLEMENTARY MATERIALS FOR CHAPTER 3 ...... 104 APPENDIX 2: REPORT MODELING OF TIGER AND ITS PREY POPULATIONS IN THE BALKHASH LAKE REGION AS THE BASIS FOR ADAPTIVE MANAGEMENT OF THE TIGER REINTRODUCTION PROGRAM IN KAZAKHSTAN ...... 117 VITA ...... 163

iii

LIST OF TABLES

Chapter 1 Table 1. Pixel reliability for 92, 1x1 km sampled blocks for MODIS Terra and Aqua datasets for May, June, and July 2013 in the Sailugem area of western Mongolia…………………………….11 Table 2. Linear (ax+b) and second-order polynomial regressions (ax2+bx+c) describing the relationship between July 2013 Aqua- and Terra-derived MODIS vegetation indices and percentage of vegetation cover on 1x1 km blocks with QA=0 sampled in Sailugem Range in western Mongolia, July-August 2013…………………………………………………………….12 Table 3. Coefficient of determinations (R²) for second-order polynomial regressions describing the relationship between May, June and July 2013 Aqua- and Terra-derived MODIS vegetation indices and percentage of vegetation cover on 1x1 km blocks with QA=0 (see Table 1) sampled in Sailugem Range, western Mongolia, July-August 2013………………………………………….13

Chapter 2 Table 1. Parameters and coefficients of determinations (R²) from multiple linear regression relating herder-derived forage value to % vegetation cover and vegetation height on 1x1 km blocks (n=49) sampled in Sailugem Range in western Mongolia, July-August 2013……………………………29 Table 2. Parameters of linear regressions describing relationship between herder perception of summer pasture quality, herder agreement (Gini-Simpson Index), NDVI anomalies in June-August in the 5 km buffer zones of herder camps in Mongolia, Russia, and China portions of the Altai Mountains over the past decade…………………………………………………………………..30 Table 3. Summary of rangeland conditions across the Altai Mountain study area including mean NDVI, NDVI CV, percentage of herders that could report on pasture conditions back to 2006, and significance of NDVI influence on herder perception of pasture quality and level of agreement in Mongolia, Russia, China, and Kazakhstan over previous decade………………………………...34

Chapter 3 Table 1. Topographic and climate variables used for distribution projections for snow leopard, Altai argali, forest and desert/semidesert cover in the Altai-Sayan Ecoregion under future climate scenarios………………………………………………………………………………………….43 Chapter 4 Table 1. Estimated maximum population densities of wild boar, Bukhara deer and roe deer in the Central Asia biomes……………………………………………………………………………...65 Table 2. Estimated values of N0 and r = for wild boar, Bukhara deer, and roe deer in the Ili- Balkhash region…………………………………………………………………………………..66 Table 3. Estimated year of Caspian tiger extinction in Central Asian and adjacent countries……68

LIST OF FIGURES

Chapter 1 Figure 1. Location of 92, 1 km2 sampling blocks measured in July-August 2013 to contrast MODIS-derived vegetation indices and percentage of vegetation ground cover in grasslands and semi-deserts of Sailugem Range, western Mongolia………………………………………………9 Figure 2. Transects used for sampling vegetation cover on each of 92 1x1km blocks aligned with the MODIS grid in the Sailugem area of western Mongolia, July-August 2013…………………10 Figure 3. First and second order polynomial regressions describing the relationship between July 2013 Aqua- and Terra-derived MODIS vegetation indices and percentage of vegetation cover at sampling sites with QA=0 in western Mongolia………………………………………………….11 Figure 4. Maps of vegetation cover for Sailugem Range, western Mongolia, in July 2013, extrapolated from 2nd order polynomial regressions presented in Table 2………………………12

Chapter 2 Figure 1. Study area in the Altai Mountains including Ulaankhus and Nogoonnuur somons of Bayan-Olgii Aimag, Mongolia; the Kosh-Agach district of Altai Republic, Russia; the Katon- Karagay district of Eastern Kazakhstan Region, Republic of Kazakhstan; and Altai prefecture of Xinjiang Autonomous Region, China……………………………………………………………23 Figure 2. Ecological contrasts among study regions in the Altai Mountains in terms of mean NDVI and its temporal coefficient of variance (CV, June-August 2006-2015) for grasslands around summer camps (5 km buffer) of herders interviewed about summer pasture conditions in Mongolia, Russia, China, and Kazakhstan……………………………………………………….24 Figure 3. Sampling design used for estimating of vegetation cover and herder-derived forage value on each of 92, 1x1km blocks aligned with MODIS grid in the Sailugem area of western Mongolia portion of the Altai Mountains in July-August 2013……………………………………………...25 Figure 4. Relationships among herders’ estimation of forage value, Terra-derived MODIS NDVI, % vegetation cover, and vegetation height at 49 sampling sites in Sailugem Range of the western Mongolia portion of the Altai Mountains during July-August 2013……………………………...28 Figure 5. Temporal patterns of herder perceptions of grassland quality and satellite-derived measures………………………………………………………………………………………….31

Chapter 3 Figure 1. Predicted snow leopard and Altai argali habitat, forest cover and desert cover models for current conditions and RCP 2.6 and 8.5 scenarios for CCSM4 GCM in the Altai-Sayan Ecoregion………………………………………………………………………………………...46 Figure 2. Projected changes of area covered by forest in current habitat, area of habitat, number and average size of habitat patches for snow leopard and Altai argali in the Altai-Sayan Ecoregion in 2060-2080 predicted by CCSM4, HadGEM2-ES, and MIROC-ESM GCMs under RCP 2.6-8.5 scenarios………………………………………………………………………………………….48 Figure 3. Projected changes in snow leopard and Altai habitat in the Altai-Sayan Ecoregion in 2060-2080 predicted by CCSM4 GCM under RCP 2.6 and 8.5 scenarios…….………………….49

Chapter 4 Figure 1. Documented sightings of Caspian tigers in 10-20th centuries and reconstructed 19-20th century range based on review of historic accounts……………………………………………....68 Figure 2. Present distribution of tugay and reed ecosystems in Central Asia…………………...70 Figure 3. Projected dynamics of tugay woodland and reed thicket ecosystem in the Ili-Balkhash region under three management scenarios………………………………………………………..72 Figure 4. Dynamics of tiger habitat area and prey carrying capacity predicted in the Ili-Balkhash region over the nearest 50 years under three management scenarios……………………………73 Figure 5. Projections for prey populations for wild boar, Bukhara deer, and roe deer and carrying capacity for tigers in three TMUs of Ili-Balkhash region under 3 management scenarios………..74 Figure 6. Population projections for tigers at different scenarios and schemes of introduction in 3 TMUs in the Ili-Balkhash region

Chapter 5 Figure 1. Snow leopard habitat and known populations in Russia……………………………….84 Figure 2. Number of livestock killed by snow leopards in western Tyva, Russia, in 2000-2013 based on reports by local herders…………………………………………………………………88 Figure 3. Snow leopard female with a wire snare around its neck in Sayano-Shushensky Nature Reserve…………………………………………………………………………………………...92

ABSTRACT

M.Yu. Paltsyn. Integration of remote sensing, modeling, and field approaches for rangeland management and endangered species conservation in Central Asia. 170 pages, 10 tables, 21 figures, 2 appendices, 2018. Scientific Style Guide used.

Integration of robust scientific approaches and on-the-ground conservation practice to “bridge the gap” between biologists and field managers is a perennial challenge in biodiversity conservation. In this thesis I present five, related case studies of integrating key scientific approaches (remote sensing techniques, habitat modeling and suitability analysis, and population modeling) with field practices to facilitate sustainable and locally accepted rangeland management, support conservation of snow leopard and Altai argali, and suggest options for tiger restoration in Central Asia. My synthesis of these case studies reveals that to advance regional long-term conservation initiatives, conservation science has to address relevant conservation problem directly, suggest solutions and recommendations that can be implemented by conservation managers given their capacity levels, fit into local knowledge systems as they pertain to the ecosystems under consideration, and focus on sharing lessons learned across projects.

Key Words: Altai argali, adaptive management, climate change, conservation, Caspian tiger, grasslands, lessons learning, Maxent, population modeling, snow leopard, traditional ecological knowledge, vegetation index.

M.Yu. Paltsyn Candidate for the degree of Doctor of Philosophy, May 2018 Major Professor: James P. Gibbs, Ph.D. Department of Environmental and Forest Biology State University of New York College of Environmental Science and Forestry Syracuse, New York

vii

INTRODUCTION

Conservation Biology is often called a crisis discipline that has as its main goal development of practical solutions for long-term conservation of the Earth’s species and ecosystems (Soulé 2007) in the face of key threats for biodiversity such as overexploitation, habitat destruction, invasive species, and climate change (Caughley 1994). To identify these practical solutions conservation biology needs a complex multi-disciplinary approach that includes not only biology at its core but also the intersection of many other fields of knowledge including geography, natural resource management, sociology, economics, politics, and business (Groom, Meffe and Carroll 2005; Soulé 2007; Van Dyke 2008). As with many other crisis disciplines, the conservation biology needs to suggest practical solutions under conditions of uncertainty and often without a sufficient knowledge base available (Soulé and Wilcox 1980). To effectively decrease the uncertainty and risk of failure, conservation should be based on the best available science and the best practical experience, a synergy often called an adaptive management, that is, integration of strategic planning, management, and monitoring to test hypotheses (assumptions) on the management results in order to learn from conservation practice and adapt conservation approaches to make them more effective (Ludwig et al. 1993; Parma et al. 1998; Salafsky et al. 2002; Gunderson 2008; Allen et al. 2011). Unfortunately, this integration of sound science and with conservation practice is rare and has even been called a “great divide” or “research-implementation gap” between academic conservation biology and field conservation management (Redford & Taber 2000; Pullin et al. 2003; Sutherland et al. 2004; Sunderland et al. 2009; Shackleton et al. 2009).

Integration of robust scientific approaches and on-the-ground conservation practice to “bridge the gap” (Sunderland et al. 2009) is the overall topic of my thesis that focuses on the Central Asia, mainly Altai-Sayan Ecoregion part, where I have spent more than 20 years of my professional life working as a field biologist, ranger, and manager of conservation programs. Having observed multiple cases of “conservation projects” that included only research without direct input to real conservation as well as many cases of “action first, science later” (Parma et al. 1998) experiences with conservation initiatives in Central Asia and other parts of the world have made the adaptive management concept – integration of science and field conservation – the cornerstone of my professional development as a conservation practitioner.

The geographic area of my main conservation experience – the Central Asia – is a huge region that encompasses Kazakhstan, Kyrgyzstan, Tajikistan, Uzbekistan, Turkmenistan, Mongolia, southern part of Russia (South Siberia) and northwestern part of China (Xinjiang) (Bedunah et al. 2004). The area is very diverse in terms of geography, climate, and biodiversity: it has an elevation range between 154 m below sea level (Turpan Basin) and almost 8,000 m above sea level (highest peaks of Pamir); annual precipitation between 25 mm in the deserts and more than 1,000 mm at mountain tops (Bedunah et al. 2004). The region has 112 types of ecosystems - unique tugai forests in the river valleys, world renowned wetlands, deserts, vast steppes, conifer forests, alpine meadows and glaciers (WWF 2006). The region is home for 7,000 species of vascular plants, 900 species of vertebrate animals and more than 20,000 species of invertebrate animals (WWF 2006). Central Asia is also a homeland for such charismatic megafauna as snow leopard (Panthera uncia), common leopard (Panthera pardus), argali (Ovis ammon), Saiga (Saiga tatarica), and many others (WWF 2006). The legendary Caspian tiger (Panthera tigris virgata) once populated the largest

1 geographical range of any tiger subspecies mainly in Central Asia and disappeared only 50-60 years ago (Geptner and Sludski 1972). Wildlife must persist in Central Asia where about 75% of the entire region is occupied by rangelands that provide pastures for livestock as well as food, homes, and water for humans, and that play a great ecological, economical, and cultural role in the life of millions nomadic and semi-nomadic people of the vast region (De Haan et al., 1997; Bedunah et al. 2004).

The Altai-Sayan Ecoregion - the key focus of this thesis - is one of the most precious and diverse parts of Central Asia located at the borders of Russia, Kazakhstan, Mongolia, and China. The Altai-Sayan is one of the 200 priority areas delineated by World Wide Fund for Nature (WWF) for conservation of the Earth’s biodiversity (Olson & Dinerstein, 2002). The area has rich species diversity encompassing about 10,000 of species of plants, animals and fungi, including 3,500 of vascular plant and 680 vertebrate animal species (Kupriyanov et al. 2003; UNDP/GEF 2007). Global importance of the Altai-Sayan Ecoregion was underlined by two nominations of the UNESCO World Nature Heritage sites – Golden Mountains of Altai and Uvs Nuur Basin (UNESCO World Heritage Convention, 2013). Conservation of this area offers an extraordinary opportunity to preserve and manage sustainably over 1,000,000 km² of practically pristine transboundary landscape in the very center of Eurasia (WWF 2012). Biodiversity in the Altai- Sayan Ecoregion is threatened by various anthropogenic pressures, which are likely to increase during the next decades, including hydropower generation, mining, linear infrastructure development, overgrazing, wildlife poaching, and logging (WWF 2012). The combined effects of these threats may intensify due to climate change (Batima 2006; Kokorin et al. 2011).

The ecoregion and the Central Asia as whole have been in the focus of different international conservation programs and projects since 1998-1999 (UNDP/GEF 2005, 2006; WWF 2012; Snow Leopard Working Secretariat 2013; UNEP/CMS 2014) but very few of them have practiced adaptive management approach and integrated robust conservation science into the field practice, and vice versa. For example, despite heavy dependence of local herder communities on rangelands for their livelihoods most of the region lacks any coherent policy and monitoring scheme to help balance rangeland use by pastoralists and their livestock with conservation of the significant biodiversity values of the region (Maroney 2005; Bailagasov 2011; Addison et al., 2012; Naidansuren and Bayasgalan, 2012; WWF 2012). In contrast, rangeland monitoring schemes that could inform adaptive management of these grassland-dominated ecosystems, no matter how scientifically rigorous when disconnected from traditional ecological knowledge often have low perceived relevance by local communities and hence limited acceptance for use in locally based decision-making (Fernandez-Gimenez 2000; Selemani et al. 2012).

Rapid climate change is also one of the major threats to wildlife in the region (Batima 2006; Kokorin et al. 2011). Despite the high value of many Altai-Sayan wildlife species including snow leopards and argali as flagship and umbrella species for national and international conservation projects in this and other regions of Central Asia no conservation strategies exist for addressing the needs of these species to adapt to impending climate change. Currently existing conservation strategies for snow leopard and argali only consider climate change as one of the many threats for these species and declare importance of addressing climate change but do not consider what may happen with the species under various climate change scenarios and do not suggest climate-smart conservation measure to address the issue (Paltsyn et al., 2011; Paltsyn et al., 2012; WWF 2012;

CMS 2013; Snow Leopard Working Secretariat 2013). Predicting umbrella species response to different climate change scenarios could provide conservation managers and planners with much more reliable insight on possible changes in the species habitats, covered large portions of landscape (Gillson et al. 2013). More specifically, such predictive models could provide a sound basis for adaptive management approach in conservation as a tool to decrease uncertainty associated with the impact of climate change (IPCC 2007, 2014).

Overhunting and habitat destruction are other major threats to wildlife in the region. The Caspian tiger extinction is a good example of their combined impact. Despite the grim history of extinction of the tiger, two factors have combined last two decades to raise the possibility of tiger restoration to Central Asia. First, the breakup of the Soviet Union and introduction of market economies in newly established states lead to the recovery of tiger habitats in some areas as many state- sponsored agricultural programs along the rivers have been abandoned as non-profitable (WWF 2014). Second, tiger phylogenetics (Driscoll et. al. 2009) recently revealed that Caspian tigers were extremely close relatives to the still extant Amur tiger (P. t. altaica) which might therefore serve as suitable “analog” form for restoration of tigers to Central Asia. Modest recovery of Caspian tiger habitats along with new perspectives on tiger phylogenetics triggered wider discussions about possible tiger introduction to Central Asia (Driscoll et al. 2011, 2012; WWF 2014) and resulted in development of the WWF Tiger Reintroduction Program in Central Asia (WWF 2014). However, the program could not start until recently because it lacked sound scientific foundation based on habitat suitability analysis and population modeling at different management scenarios – key elements for the program’s adaptive management.

Integrating experiences and insights among regional wildlife conservation projects focused on mitigating multiple threats to wildlife in the region is an extremely valuable tool for advancing endangered species conservation both regionally and range-wide yet, paradoxically, it is widely recognized that conservation practitioners generally do not share their experiences in published form (Redford & Taber, 2000; Pullin et al. 2003; Sutherland et al. 2004; Sunderland et al., 2009). This results in a huge volume of collective conservation experience being lost for future projects. Moreover, unexpected outcomes and project failures regularly occur in endangered species conservation projects given the complex socio-economic and uncertainty issues associated with the species’ conservation. Yet these are very rarely reported despite their high practical value for others working in similar situations (Sunderland et al. 2009). This is the case for Altai-Sayan Ecoregion, as well, that has decades of conservation experience yet never has it been analyzed and shared in terms of highlighting factors in implementation associated with of failure and success in the species conservation programs. This is particularly true for the snow leopard, one of the most charismatic yet least known species in the region.

The overall hypothesis of my thesis is that science has to be fully integrated in regional long-term conservation initiatives to inform the project design, implementation and learning. More specifically, the science has (1) to address relevant conservation problem directly, (2) suggest solutions and recommendations that can be really implemented by conservation managers and (3) fit into local knowledge systems as they pertain to the ecosystems under consideration. To examine my overall hypothesis and in attempt to cover the “research-implementation gaps” identified above I developed five chapters for this thesis that incorporate different aspects of adaptive management approach as applied to wildlife and rangeland conservation, using the species and habitats of the

Altai-Sayan as my focus. Chapters 1 and 2 deal with scientific approaches for baseline analysis and monitoring in sustainable and locally accepted rangeland management, Chapter 3 and 4 have key focus on application of biodiversity science for strategic planning of endangered species conservation and restoration programs, and Chapter 5 considers learning and sharing of lessons from species conservation programs. All the chapters are briefly introduced below.

Chapter 1 “Estimation and Prediction of Grassland Cover in Western Mongolia using MODIS- derived Vegetation Indices” explores applicability of the MODIS Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) for rangeland management in the western Mongolia. In this study we contrasted performance of NDVI and EVI metrics obtained from Aqua and Terra, the two satellite platforms carrying the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor, to evaluate whether satellite datasets can be useful for rangeland managers in western Mongolia to monitor and predict summer pasture conditions at the regional level using indicators acceptable for local herders. This chapter was published as Paltsyn, M.Yu., J.P. Gibbs, L.V. Iegorova, G. Mountrakis. 2017. Estimation and Prediction of Grassland Cover in Western Mongolia Using MODIS-Derived Vegetation Indices. Rangeland Ecology & Management 70(6): 723-729 https://www.sciencedirect.com/science/article/pii/S1550742417300453 .

Chapter 2 “Integrating Traditional Ecological Knowledge and Remote Sensing for Monitoring Rangeland Dynamics in the Altai Mountain Region” analyzes applicability of herders’ traditional ecological knowledge (TEK) for monitoring of rangeland dynamics in the Altai Mountains along with remote sensing techniques. In this chapter I explore whether herder-derived estimates of pasture forage value at the peak of growing season corroborate satellite-derived vegetation indices in the Altai Mountain region thereby enabling re-expression of satellite data in a form more compatible with herder perceptions. This chapter has been submitted for publication in Environmental Management as Paltsyn, M.Yu., Gibbs, J.P., Mountrakis, G. Integrating Traditional Ecological Knowledge and Remote Sensing for Monitoring Rangeland Dynamics in the Altai Mountain Region.

Chapter 3 “Forecasting of snow leopard (Panthera uncia) and Altai argali (Ovis ammon ammon) future habitat dynamic in the Altai-Sayan Ecoregion under a changing climate” explores snow leopard and Altai argali habitat projections in 2060-2080 under four climate change scenarios to identify climate-smart measures for conservation of both species in the region. The chapter is prepared as a manuscript Paltsyn, M.Yu., Iegorova, L.V., and J.P. Gibbs. Forecasting of snow leopard (Panthera uncia) and Altai argali (Ovis ammon ammon) habitat dynamics in the Altai- Sayan Ecoregion under a changing climate for publication in the Biological Conservation and integration in the WWF conservation strategy in the Altai-Sayan Ecoregion.

Chapter 4 “Tiger Re-establishment Potential to Former Caspian Tiger (Panthera tigris virgata) Range in Central Asia” is the case study of endangered species restoration that examines the biological feasibility of re-establishment of tigers in Central Asia in the context of increasing water consumption and security concerns of local communities in the Ili-Balkhash region, Kazakhstan. The chapter was published as Chestin, I.E., Paltsyn, M.Yu., Pereladova, O.B., Iegorova, L.V., Gibbs, J.P. 2017. Tiger re-establishment potential to former Caspian tiger (Panthera tigris

4 virgata) range in Central Asia. Biological Conservation. Vol. 205. Pp. 42-51 https://www.sciencedirect.com/science/article/pii/S0006320716308151.

Chapter 5 “Snow Leopard Conservation in Russia: Lessons Learned Over 15 Years” summarizes key lessons learned from snow leopard conservation projects in Altai-Sayan Ecoregion as well as factors of the project success and failure in implementation of anti-poaching and snow leopard – herder conflict mitigation initiatives. The chapter was published as Paltsyn M., Poyarkov A., Spitsyn S., Kuksin A., Istomov S., Gibbs J.P., Jackson R. M., Castner J.L., Kozlova S., Malykh S., Korablev M., Zvychaynaya E., Rozhnov V. 2016. Chapter 40. Northern Range: Russia. In: McCarthy T. and Mallon D., Editors. “Snow Leopards. Biodiversity of the World: Conservation from Genes to Landscapes”. Elsevier. Pp. 501-511. https://www.elsevier.com/books/snow- leopards/nyhus/978-0-12-802213-9 .

CHAPTER 1: ESTIMATION AND PREDICTION OF GRASSLAND COVER IN WESTERN MONGOLIA USING MODIS-DERIVED VEGETATION INDICES

Abstract: Spectral indices derived from satellite observations, such as the Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI), are widely used for grassland monitoring and management around the globe. In this study we contrasted performance of NDVI and EVI metrics obtained from Aqua and Terra, the two satellite platforms carrying the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor, for estimating grassland cover measured at ground level on 92, 1km x 1km blocks distributed from semi-desert to high montane grasslands in the Sailugem Range of western Mongolia where overgrazing and overstocking of domestic livestock is a concern for pastureland management. We also explored utility of late spring (May) vegetation indices for forecasting vegetation cover at the peak of growing season (July). Vegetation indices developed using MODIS 1 km monthly data (MOD13A3 and MYD13A3) were strongly related to on-the-ground field estimates of the percentage of vegetation cover in July (74-85% variation explained), with second-order polynomial regressions demonstrating better fit to the data than first-order regressions, Aqua vegetation indices (VIs) explaining slightly more variance than Terra’s VIs, and NDVI performing comparably to EVI for both Aqua and Terra. Both Aqua and Terra vegetation indices for May were highly predictive of July vegetation cover (R²=0.80-0.84). We conclude that monthly MODIS NDVI and EVI datasets can be useful for rangeland managers in western Mongolia to monitor and predict summer pasture conditions at the regional level where science-based guidance on grazing policy and practices is much needed.

Introduction

Grasslands occupy over one quarter of the Earth’s terrestrial surface, excluding Antarctica and Greenland (Olson et al., 1983; Groombridge, 1992; White et al., 2000). Grasslands also support much biological diversity, including the greatest plant species richness at smaller spatial grains (<50 m²) of any ecosystem (Wilson et al., 2012) and the highest concentrations of herbivorous animals (White et al., 2000). In addition to acting as sink or source for atmospheric CO₂ (Frank, 2002) thereby playing an important role in regulation of regional and global climate (Yatagai and Yasunari, 1995), grasslands are an integral component for sustaining water quality, soil conservation, agriculture and recreation (National Research Council, 1994; White et al., 2000; Marsett et al., 2006). Last, > 20 million households depend on grasslands as pastures for livestock (De Haan et al., 1997). In aggregate, the tremendous ecological, economic, social and cultural values of grasslands make them a critical global resource.

Management of grassland resources at local, regional and global levels requires timely monitoring of indicators reliably reflecting vegetation cover, productivity, biomass, and plant species composition (White et al., 2000). Grassland monitoring and management is challenging because grasslands often occupy vast and remote areas such that field surveys of grassland conditions are expensive. Moreover, grasslands demonstrate high variability in response to climate variables and anthropogenic impact thereby requiring frequent sampling (Reeves et al., 2001; Tueller, 2001; Marsett et al., 2006).

Since the 1970s satellite-borne remote sensing has been used for monitoring and research of grassland and rangeland ecosystems (Booth and Tueller, 2003; Liu et al., 2005). Satellite-derived Vegetation Indices (VIs), representing arithmetic combination of two or more bands reflecting spectral characteristics of vegetation, have been widely applied for phenological monitoring, vegetation classification, and derivation of structural vegetation parameters, such as plant greenness, vigour, productivity, biomass and leaf area index (Huete et al., 2002). Normalized

Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) are the most often used global-based VIs for monitoring of terrestrial vegetation (Huete et al., 2002) and have been applied in grasslands and rangelands research and monitoring around the world, e.g., in Mongolia and Japan (Purevdorj et al., 1998), China (Akiyama and Kawamura, 2007), USA, Africa and Mongolia (Boone et al., 2007), Italy (Colombo et al., 2011), and Brazil (Ferreira et al., 2004).

NDVI and EVI have different limitations for monitoring vegetation cover. EVI is a modified expression of the NDVI metric with improved sensitivity to high biomass regions and vegetation monitoring ability due to reduction of atmospheric influence (Huete and Justice, 1999) yet EVI is more sensitive to topography-induced uncertainty (Matsushita et al., 2007). For this reason, NDVI offers optimal performance in mountain areas (Matsushita et al., 2007), but is subject to uncertainty due to different atmospheric and canopy background conditions (Liu and Huete, 1995).

The Moderate Resolution Imaging Spectroradiometer (MODIS) sensor is one of the most widely used sources for NDVI and EVI data by researchers and environmental managers for grassland assessment and monitoring (Huete et al., 2002; Li et al., 2010; Cui et al., 2012). NDVI and EVI exhibit variable performance for different grassland areas and regions (Huete et al., 2002; Ferreira et al., 2004; Kawamura et al., 2005; Guo et al., 2007; Li et al., 2010; Luo et al., 2010; Zhao and Ma, 2011; Cui et al., 2012; Nakano et al., 2013) warranting further exploration of the relative performance of these indices for different applications.

MODIS sensors are installed on two satellite platforms – Terra and Aqua – that complement each other by both collecting NDVI and EVI daily data (Wang et al., 2007). Studies have revealed comparable performance of Terra and Aqua vegetation indexes (Yang et al., 2006; Wang et al., 2007), however, others have reported inconsistencies in the satellites’ indexes values (Yang et al., 2006; Wang et al., 2012; Djavidnia et al. 2010). An understanding of the relationships between VI values derived from Terra MODIS and Aqua MODIS is critical for refining their use for environmental monitoring yet still poorly developed (Yang et al., 2006; Wang et al., 2007; Colombo et al., 2011).

The issue of sound remote sensing monitoring of grasslands is particularly germane in Mongolia where rangelands make up roughly three-fourths of the country’s land area, providing pasture for some 56 million head of livestock, and supporting livelihoods for some 26% of the country’s inhabitants while generating about 13.5% of Mongolia’s GDP (Erdenesan, 2016). However, an estimated 75% of Mongolia’s pastureland is overgrazed and overstocked with domestic livestock such that significant degeneration caused both by anthropogenic and climate factors is a major concern for decision-makers (Stumpp et al., 2005; Naidansuren and Bayasgalan, 2012; Hilker, et al. 2014). Our objectives were to (1) compare performance of Aqua and Terra satellites and (2) NDVI versus EVI to explain variation in percentage of vegetation cover in the western Mongolia grasslands and (3) to explore the utility of late spring VI values for predicting percentage of vegetation cover at the peak of growing season (July) as a proxy of pasture quality and forage availability. Our work combined NDVI and EVI derived from Terra MODIS and Aqua MODIS 1 km resolution monthly data tiles (products MOD13A3 and MYD13A3) with our own on-the- ground estimation of vegetation cover at 92 sites sampled throughout the Sailugem Range in western Mongolia. Our work tests the hypotheses that Aqua and Terra sensors as well as NDVI and EVI perform equally well in the conditions of short-grass steppe of western Mongolia, and

7 that early season NDVI and EVI values are positively correlated with mid-season grassland vegetation cover within a given year.

Materials and Methods

Study Area Our geographical focus was the Sailugem Range of western Mongolia, a high montane grassland region traditionally used for intensive livestock grazing by Kazakh people while having high value for conservation of wildlife including endangered Altai argali (Ovis ammon ammon) and Siberian ibex (Capra sibirica) (Maroney, 2005; Paltsyn et al., 2011). The region as well as the entire country still lacks policy and adequate management practices as well as rangeland monitoring system to advance sustainable range use that balances conservation values and pastoralist livelihoods (Addison et al., 2012; Naidansuren and Bayasgalan, 2012). Total number of livestock in the area increased by 47% and reached historical maxima in the last 25 years: from 503,486 in 1990 to 742,084 in 2014 due mainly to 350% increase of number of goats, which increased from 100,082 in 1990 to 355,793 in 2015 (Department of Statistics under the Government of Bayan-Olgii Aimag, 1990-2014). The increase in livestock populations in Sailugem Range, as well as entire Mongolia, has been caused by two factors: (1) collapse of centralized economy followed by mass unemployment after the breakdown of the Soviet Union (Reading et al., 2006) and (2) high global demand for cashmere produced by goats (Naidansuren and Bayasgalan, 2012). Increasing goat herds cause severe degradation of pastures due to goats’ destructive grazing habits leaving a shortage of forage for other livestock and wild ungulate species (Naidansuren and Bayasgalan, 2012). Livestock grazing in Sailugem Range even occurs in the Silkhemiin Nuruu National Park, established for protection of highly endangered Altai argali and snow leopard protection (Paltsyn et al., 2011). Local authorities have rights to regulate livestock number and distribution, however, they have very low capacity and resources for adequate rangeland management (Fernandez-Gimenez et al., 2008) and try to avoid unpopular livestock regulation decisions. All sampled sites occurred in the southern part of Sailugem Range, Bayan-Olgii Aimag, Mongolia, along the border with the Russian Federation (49º11' N - 49º46' N, 88º44' E - 90º26' E) (Fig. 1). The area is characterized by undulating mountains and hills, wide intermountain depressions and high plateaus within elevation range between 1500 and 3400 m above sea level (average elevation – 2350 m). Climate is strongly continental with short growing season (April-August), severe winters and ca. 250-400 mm of annual precipitation with average annual temperature -6º to -7º C (Hilbig, 1995; Modina, 1997). The landscape is dominated by alpine short-steppe grasslands (dominated by Cobresia), with higher altitudes (2900-3400 m above sea level) covered by barren rock and moss, lichen, and low shrubs and intermountain depressions (1500-1800 m) by dry steppe (dry wormwood - grass steppes with Caragana and bunchgrass Achnatherum), semi-deserts (dominated by Stipa and Caragana) and, in some areas, by wetlands.

Figure 1. Location of 92, 1 km2 sampling blocks measured in July-August 2013 to contrast MODIS-derived vegetation indices and percentage of vegetation ground cover in grasslands and semi-deserts of Sailugem Range, western Mongolia.

Analytical Approach To obtain remotely sensed estimates of conditions of ground vegetation, we accessed Aqua and Terra MODIS Vegetation Indices Monthly L3 Global 1km SIN Grid V006 datasets (MYD13A3 and MOD13A3, tile h23v04) for May, June, and July 2013 downloaded from the NASA’s EOSDIS Reverb web-site (http://reverb.echo.nasa.gov/reverb/). For field sampling the MODIS tiles were re-projected to WGS 1984 UTM Zone 45 projection and clipped to our study area. Longitude and latitude of the center of each 1x1 km pixel in the study area were calculated in ArcGIS 10.2.2 (ESRI, 2014). To obtain co-located and temporally coincident estimates of ground conditions where for MODIS-derived vegetation indices, between 18 July -3 August 2013 we sampled the percentage of vegetation cover at 92, 1km x 1km blocks (Fig. 1), each of which represented a single pixel of MODIS 1 km resolution Monthly Vegetation Index tile. All sampled blocks were located on relatively flat terrain 1495-2860 m above sea level that we purposely selected to represent the full range of grassland cover present in the region from montane grassland tundra to dry grassland and semi-desert mediated by the logistical constraints of traveling through this remote region. To select the maximal range of grasslands cover for sampling we used re-projected MODIS EVI and NDVI images and topographic maps of the study area.

To estimate percentage of vegetation cover in each sampled block we first navigated to the block center and then walked an ever-increasing and spiral-shaped, 4800-m-long transect guided by a hand-held GPS unit accurate to ca. 10 m (Fig. 2). We stopped every 25 paces along the transect to make point measurements of vegetation cover (green vegetation versus bare ground or dry

9 vegetation litter at the sampling point touched by a pin placed at the leading edge of the surveyor’s foot). Point estimates accumulated along a given transect were used to calculate proportion of points that were vegetated among all points sampled within a given sampling block (average n = 215 points per sampled block, range = 167-255 depending on stride length of surveyor).

Figure 2. Transects used for sampling vegetation cover on each of 92 1x1km blocks aligned with the MODIS grid in the Sailugem area of western Mongolia, July-August 2013. Black frame represents a 1x1 km sampling block, red lines depict segments of a 4800-m-long sampling transect, and arrows the direction of movement during sampling.

Estimates of the percentage of vegetation cover for selected blocks were associated with respective NDVI and EVI values of appropriate pixels for May, June, and July 2013 Terra MODIS (MOD13A3) and Aqua MODIS (MYD13A3) 1 km resolution Monthly Vegetation Index tiles. First- and second-order polynomial least squares regressions relating Terra and Aqua NDVI and EVI datasets to on-the-ground vegetation cover were fit and model performance contrasted with model coefficient of determination (R² and R²adj), root mean square error (RMSE), and Akaike’s Information Criterion (AICc). Best regressions were used to extrapolate maps of vegetation cover in our study area using ArcGIS 10.2.2 (ESRI, 2014).

Results The number of reliable pixels among the 92 1x1 km blocks sampled on-the-ground varied among the MODIS datasets for May, June, and July 2013 (Table 1) ranged from highest for Terra in June (n=82) to lowest for Terra in July (n=51). Only pixels of highest reliability (MODIS QA = 0, or “Good data”) were used to build regression models (NASA LP DAAC, 2016).

Table 1. Pixel reliability (NASA LP DAAC, 2016) for 92, 1x1 km sampled blocks for MODIS Terra and Aqua datasets for May, June, and July 2013 in the Sailugem area of western Mongolia. Pixel reliability Aqua Terra (QA) May June July May June July 0 - Good data 69 52 59 70 82 51

1 - Marginal data 20 40 33 22 10 41

3 – Cloudy 3 0 0 0 0 0

Vegetation cover estimated in the field varied from 4% to 100% (average 57%) on the 92 sampled blocks. The ground-level vegetation cover in July was strongly and positively correlated with Terra and Aqua MODIS EVI and NDVI sampled in the same month (> 71% of variance explained, Table 2). Top-ranked models were second-order polynomial regressions, which demonstrated better fit (ΔAICc > 2.0 units) to the data than did first order regressions in all cases. Second-order polynomial regressions for the Aqua VI explained ~8% more variance in vegetation cover and had smaller RMSE than Terra for both NDVI and EVI in July. However, these same regressions with NDVI and EVI had <1% difference in amount of variation in vegetation cover explained for both Aqua and Terra (Table 2, Fig. 3).

Figure 3. First (in green) and second order polynomial (in blue) regressions describing the relationship between July 2013 Aqua- and Terra-derived MODIS vegetation indices and percentage of vegetation cover at sampling sites with QA=0 (n=59 for Aqua and n=51 for Terra) in western Mongolia.

Table 2. Linear (ax+b) and second-order polynomial regressions (ax2+bx+c) describing the relationship between July 2013 Aqua- and Terra-derived MODIS vegetation indices and percentage of vegetation cover on 1x1 km blocks with QA=0 (n=59 for Aqua and n=51 for Terra) sampled in Sailugem Range in western Mongolia, July-August 2013.

Satellite Vegetation Order a b c R²adj RMSE AICc ΔAICc Index 1st 135.3 12.4 NA 0.795 11.6 461.2 15.0 NDVI 2nd -293.9 327.1 -10.3 0.844 10.1 446.2 0 Aqua 1st 193.8 13.9 NA 0.803 11.4 459.0 10.3 EVI 2nd -554.1 430.2 -3.3 0.837 10.3 448.7 0 1st 130.7 18.3 NA 0.711 12.8 409.3 7.9 NDVI 2nd -258.1 295.9 -1.7 0.758 11.6 401.4 0 Terra 1st 188.3 18.3 NA 0.739 12.2 404.1 2.1 EVI 2nd -367.6 343.9 6.3 0.755 11.7 402.0 0 * - p < 0.0001 for all regressions.

Maps of the vegetation cover calculated using 2nd order polynomial regressions on the base of Aqua and Terra NDVI and EVI reflects similar rangeland conditions (Fig. 4).

Figure 4. Maps of vegetation cover for Sailugem Range, western Mongolia, in July 2013, extrapolated from 2nd order polynomial regressions presented in Table 2.

In terms of predicting future ground vegetation conditions, strong positive correlations were evident between ground estimates in July 2013 and satellite NDVI and EVI monthly data for May and June 2013 (Table 3) with > 79% of variance explained for May-estimated VIs and >78% for

June-estimated VIs. For May and June 2013 Terra performed slightly better than Aqua for both NDVI and EVI (1-3% more variance explained). EVI performed slightly better than NDVI for May and June for both Aqua and Terra (< 2% more variance explained). Notably R² values for Terra in May were higher than for July 2013, the month when on-the-ground conditions were sampled (Table 3).

Table 3. Coefficient of determinations (R²) for second-order polynomial regressions describing the relationship between May, June and July 2013 Aqua- and Terra-derived MODIS vegetation indices and percentage of vegetation cover on 1x1 km blocks with QA=0 (see Table 1) sampled in Sailugem Range, western Mongolia, July-August 2013. ퟐ ퟐ ퟐ Satellite Vegetation 푹푴풂풚 푹푱풖풏풆 푹푱풖풍풚 Index NDVI 0.798 0.783 0.850 Aqua EVI 0.813 0.803 0.843 NDVI 0.826 0.807 0.768 Terra EVI 0.840 0.810 0.765 p < 0.0001 for all regressions

Discussion Our analysis demonstrated that 1 km resolution monthly Aqua and Terra MODIS NDVI and EVI for July 2013 reliably indexed the degree of vegetation cover in the grasslands of western Mongolia for the same month. Clearly second-order polynomial regressions describe relationships between VIs and percentage of vegetation cover better than first-order models; however, second-order polynomial regressions should be used with caution for areas with dense vegetation cover (percentage of vegetation cover higher than 80%), especially for NDVI, because at such thresholds the predicted relationships seemed to saturate (Fig. 3). Moreover, saturation effects are clearly evident in the maps of model-extrapolated vegetation cover (Fig. 4): maximal percentage of vegetation cover does not exceed 80-87% even for areas with high and dense vegetation, e.g. wetlands in the north-east corner of our study area. Furthermore, areas with very low vegetation cover (<5%), such as deserts and rocks at high elevation, can generate negative projections when the 2nd polynomial order models are applied. Thus, these statistical models are most appropriate for typical short-grass Mongolian mountain and dry steppes with vegetation cover 10-80%. Our results are consistent with the results of Purevdorj et al. (1998) who reported for other Mongolian grasslands that second-order polynomial regression outperformed first-order models in describing the relationship between percentage of vegetation cover and NDVI derived from simulated Advanced Very High Resolution Radiometer (AVHRR) data. However, Li et al. (2010) reported that first-order models performed better than second-order in reflecting relationships between percentage of vegetation cover and VIs (NDVI and EVI) derived from MODIS VI 16 day 250 m resolution datasets in the grasslands of Northern Hebei Province in China. Similarly, Cui et al. (2012) demonstrated that a first-order model described variance of percentage of vegetation cover based on the EVI derived from MODIS Terra/Aqua Surface Reflectance Daily L2G Global 500m and 1 km SIN Grid datasets better than a power regression. Despite these contrasting study outcomes, these and our study indicate that monthly MODIS-derived VIs provide useful inference about actual ground conditions in the region studied.

Aqua MODIS EVI and NDVI for July outperformed Terra’s VIs to describe vegetation cover in the same month, however, this difference was not statistically significant. Similarly, a study by

Yang et al. (2006) reported comparable performance of Terra and Aqua 8-day reflectance and Leaf Area Index (LAI) products over large areas but revealed noticeable differences at smaller scales (a few kilometres) due possibly to atmospheric effects (Yang et al. 2006). Wang et al. (2007) found that NDVI values derived from Terra MODIS and Aqua MODIS 250 m resolution 16-day datasets covered north-western part of China are similar for different vegetation types. Wang et al. (2012) found higher uncertainty in estimation of vegetation using Terra MODIS NDVI time series for forest and tundra ecosystems in North America and explained this fact by Terra MODIS calibration degradation. Djavidnia et al. (2010) also reported inconsistent estimations of ocean chlorophyll-a concentrations from Terra and Aqua MODIS sensors.

In our study the slight difference between performance of Aqua and Terra may be explained by several factors. First, Aqua and Terra monthly 1 km resolution data are calculated using MODIS 16-day 1 km resolution datasets that overlap the month (Didan and Huete, 2006). Thus, for Aqua MODIS monthly July 2013 datasets three 16-day tiles were used that cover the period from June 18 to August 4. For producing of July 2013 Terra MODIS dataset the 16-day tiles covered period of June 26 – August 12. Different atmospheric conditions in these days could affect the monthly datasets produced from Aqua and Terra data. Second, Aqua and Terra satellites have 3 hour difference in equatorial crossing time and reflect different atmospheric conditions over the same area that can affect all levels of the datasets, including monthly data (NASA, 2016). Third, Aqua has an ascending orbit whereas Terra has a descending orbit (NASA, 2016) such that MODIS data are resampled differently using Aqua versus Terra datasets for the same area. For these reasons Aqua and Terra VI monthly data yield slightly different indices of ground conditions depending on season and their performance may therefore vary for any particular application.

We did not detect any meaningful biological difference in the performance of July NDVI and July EVI both for Aqua and Terra in reflecting ground-estimated vegetation cover in our study area. Multiple publications demonstrate that MODIS NDVI and EVI exhibit variable performance for different regions and grassland areas. Huete et al. (2002) reported that MODIS-derived NDVI generated a higher range of values over semiarid sites but a lower range over the humid forested sites in South-East USA and Brazil. However, both NDVI and EVI had a similar range in values for the intermediate mesic grasslands (Huete et al., 2002). A related study in Northern Hebei Province, China, reported that MODIS-NDVI was more often correlated with on-the-ground measurements of vegetation cover in grassland, shrub and forest areas, than MODIS-EVI (Li et al., 2010). Classification of grasslands of Inner Mongolia (China) based on MODIS-EVI data provided better results than classification based on MODIS-NDVI dataset (Zhao and Ma, 2011).

Our studies contrasted with the wider literature suggest the importance of exploring performance of MODIS NDVI and EVI and assess suitability for a particular area. Tan et al. (2008) showed that MODIS NDVI is more sensitive to small variations in vegetation and performs better in sparse vegetation areas, than MODIS EVI over the North America continent. Classification of general vegetation domains (herbaceous, woody, and forested) in Brazilian savannah using MODIS NDVI and EVI datasets showed better performance of NDVI (75% of data correctly classified) over EVI (71% of data correctly classified), although EVI better performed for separation of grassland and shrub biomes (Ferreira et al., 2004). MODIS-EVI demonstrated strongest correlation with field samples of dry biomass and vegetation cover among other 17 VIs (including NDVI) in grasslands of the Qinghai–Tibet Plateau, China (Cui et al., 2012). A study of different vegetation types in

Northwest China demonstrated similar correlations of MODIS NDVI and EVI values with data collected with spectrometer in the field for arid and semi-arid areas, while in relatively high grass meadows NDVI saturated and EVI performed better (Guo et al., 2007). Our conclusions about ecologically insignificant differences in performance of NDVI versus EVI for reflecting percentage of vegetation cover in mountain grasslands of western Mongolia are consistent with results of some studies in similar environmental conditions that report similarity of the indexes (e.g., Huete et al. 2002 and Guo et al. 2007).

High correlations between percentage of vegetation cover in July 2013 and MODIS VIs in May and June 2013 imply that on-the ground conditions of grasslands in Sailugem range in July probably can be predicted from VIs measured 1-2 months earlier. July to early August is the peak of short growing season in western Mongolia when NDVI and EVI have maximal values reflecting highest vegetation growth, percentage of vegetation cover and biomass (Chu and Guo, 2012; Cui et al., 2012; Nakano et al., 2013). Our study revealed that MODIS NDVI and EVI datasets for May and June could be reliably used to make projections of maximal annual vegetation cover on the mountain pastures, at least during 2013 when our fieldwork was conducted. May-based projections of the vegetation cover and biomass might be of higher importance because they allow predictions of summer pasture quality before Mongolian herders with their livestock depart for grazing areas (herders arrive at summer pastures in the Sailugem region at the middle of June and stay there until the middle of August). Forecasting rangeland conditions reliably ahead at least one month can help avoid livestock concentration in areas with unfavourable pasture conditions.

Implications We conclude that monthly Aqua and Terra NDVI and EVI datasets that are readily available and that do not require complex processing can be useful for rangeland managers and decision makers of western Mongolia as a tool for monitoring and predicting of summer pasture conditions (on the base of percentage of vegetation cover) at the regional level. Given our findings preference in this case probably should be given to Aqua VIs (that performed slightly better than Terra VIs to reflect vegetation cover in our study area in July) and 2nd order polynomial models. Although vegetation cover is only one of many parameters that can be monitored and used for grassland management in western Mongolia, vegetation cover is the most frequently used indicator in remote sensing to measure grassland production and level of pasture degradation (Liu et al., 2004; Li et al., 2010; Cui et al., 2012) because it strongly reflects the ecological value of grasslands, especially in the highland and arid landscapes (Guo et al., 2006; Cui et al., 2012; Yan and Lu, 2015). Also, as was demonstrated by the research of Fernandez-Gimenez and Allen-Diaz (1999) vegetation cover was the only parameter that demonstrated the most reliable response to grazing and precipitation across mountain steppe, steppe, and desert-steppe in Central Mongolia. At the same time vegetation cover has strong positive correlation with the above ground biomass in rangelands (R² = 0.60-0.96) (Li et al., 2006; Al-Bakri and Abu-Zanat, 2007; Eckert and Engesser, 2013; Yan and Lu, 2015). Furthermore, as was reported by Fernandez-Gimenez (1997 and 2000), Mongolian herders generally rely on vegetation cover in their traditional assessment of pasture quality; therefore, this indicator can be better understood and accepted by local communities than other grassland parameters. Our studies suggest that decision makers in western Mongolia responsible for environmental monitoring and grassland management could benefit from incorporating simple vegetation cover analysis based on MODIS VIs in their decision-making process that will be consistent with traditional herders’ knowledge and practices. More specifically, our preliminary

15 findings indicate MODIS VI’s can enable broad-scale and rapid monitoring of summer pasture conditions at the regional level and also potentially be converted to practical and understandable advice for herders on best locations of summer camps with sufficient advance notice (at least 1 month) to influence herder decision-making each year. Currently community-based pasture management and community ownership of large sets of seasonal pastures (khoshuun) as well as strengthening of community-based rangeland management organizations with allocation of full management rights to them may be a feasible option to sustainable pasture management in Mongolia where private property on pastureland is unconstitutional (Fernandez-Gimenez et al., 2008; Naidansuren and Bayasgalan, 2012; Bruegger et al., 2014). In this case a simple rangeland monitoring tool with ability to monitor and predict conditions of summer pastures using traditional herder indicators (e.g., vegetation cover) could be quite valuable for decision making by rangeland management organizations on pasture use. Further research would be useful to understand the stability of the patterns we observed in 2013 over different years and MODIS VI data of different resolution (250 and 500 m) and especially to corroborate our conclusion that NDVI and EVI estimates for May and June can predict percentage of vegetation cover in the peak of growing season in other years and in other regions. Together such inquiry would support more fully participatory decision-making for sustainable management of grazing areas in western Mongolia where the ramifications for biodiversity conservation and sustainability of herder livelihoods are so significant.

References Addison, J., Friedel, M., Brown, C., Davies, J., Waldron, S., 2012. A critical review of degradation assumptions applied to Mongolia’s Gobi Desert. Rangeland Journal. 34, 125–137. Akiyama, T., Kawamura, K., 2007. Grassland degradation in China: Methods of monitoring, management and restoration. Grassland Science. 53, 1-17. doi:10.1111/j.1744- 697X.2007.00073.x. Al-Bakri, J.T., Abu-Zanat, M.M., 2007. Correlating Vegetation Cover and Biomass of a Managed Range Reserve with NDVI of SPOT-5 HRV. Jordan Journal of Agricultural Science. 3, No.1. Boone, R.B., Lackett, J.M., Galvin, K.A., Ojima, D.S., Tucker, C.J., 2007. Links and broken chains: evidence of human-caused changes in land cover in remotely sensed images. Environmental Science and Policy. 10,135-149. Booth, D.T., Tueller, P.T., 2003. Rangeland Monitoring Using Remote Sensing. Arid Land Research and Management. 17, 455–467. Bruegger, R.A., Jigjsuren, O., Fernandez-Gimenez, M.E., 2014. Herder Observations of Rangeland Change in Mongolia: Indicators, Causes, and Application to Community-Based Management. Rangeland Ecology and Management. 67,119–131.

Chu, T., Guo, X., 2012. Characterizing Vegetation Response to Climatic Variations in Hovsgol, Mongolia Using Remotely Sensed Time Series Data. Earth Science Research. 1, 279-290. Colombo, R., Busetto, L., Fava, F., Di Mauro, B., Migliavacca, M., Cremonese, E., Galvagno, M., Rossini, M., Meroni, M., Cogliati, S., Panigada, C., Siniscalco, C., Morra di Cella, U., 2011. Phenological monitoring of grassland and larch in the Alps from Terra and Aqua MODIS images. Italian Journal of Remote Sensing. 43, 83 - 96. Cui, X., Guo, Z.G., Liang, T.G., Shen, Y.Y., Liu, X.Y., Liu, Y., 2012. Classification management for grassland using MODIS data: a case study in the Gannan region, China. International Journal of Remote Sensing. 33, 3156–3175. De Haan, C., Steinfeld, H., Blackburn, H., 1997. Livestock and the Environment: Finding a Balance. Brussels: Commission on of the European Communities, Food and Agricultural Organization of the United Nations, and the World Bank. Didan, K., Huete, A., 2006. MODIS Vegetation Index Product Series Collection 5 Change Summary. TBRS Lab, The University of Arizona. http://landweb.nascom.nasa.gov/QA_WWW/forPage/MOD13_VI_C5_Changes_Docume nt_06_28_06.pdf. Djavidnia, S., Melin, F., Hoepffner, N., 2010. Comparison of global ocean colour data records. Ocean Science. 6, 61-76. Eckert, S., Engesser, M., 2013. Assessing vegetation cover and biomass in restored erosion areas in Iceland using SPOT satellite data. Applied Geography. 40, 179-190 Erdenesan, E., 2016. Livestock Statitics in Mongolia. FAO Asia and Pacific Commission on Agricultural StatisticsTwenty-Sixth Session, 15-19 February 2016. Thimphu, Bhutan. http://www.fao.org/fileadmin/templates/ess/documents/apcas26/presentations/APCAS- 16-6.3.5_-_Mongolia_-_Livestock_Statistics_in_Mongolia.pdf. Fernandez-Gimenez, M., 2000. The role of Mongolian nomadic pastoralists’ ecological knowledge in rangeland management. Journal of Ecological Applications. 10, 1318-1326. doi:10.1890/1051-0761(2000)010[1318:TROMNP]2.0.CO;2. Fernandez-Gimenez, M.E., 1997. Landscapes, livestock, and livelihoods: social, ecological, and land-use change among the nomadic pastoralists of Mongolia. PhD Dissertation, University of California, Berkeley, CA. Fernandez-Gimenez, M., Allen-Diaz, B., 1999. Testing a non-equilibrium model of rangeland vegetation dynamics in Mongolia. Journal of Applied Ecology. 36, 871-885 Fernandez-Gimenez, M. E., Kamimura, A., Batbuyan, B., 2008. Implementing Mongolia’s land law: progress and issues. In: Final report to the Central Asian Legal Exchange. Nagoya, Japan: Central Asian Legal Exchange, Nagoya University. 45 p. Ferreira, L.G., Yoshioka, H., Huete, A., Sano, E.E., 2004. Optical characterization of the Brazilian Savanna physiognomies for improved land cover monitoring of the cerrado biome: preliminary assessments from an airborne campaign over an LBA core site. Journal of Arid Environments. 56, 425–447. Frank, A., 2002. Carbon dioxide ﬂuxes over a grazed prairie and seeded pasture in the Northern Great Plains. Environmental Pollution. 116, 397–403. Groombridge, B., 1992. Global Biodiversity: Status of the Earth's Living Resources. London: Chapman and Hall. Guo, Z.G., Liang, T.G., Liu, X.Y., Niu, F.J., 2006. A new approach to grassland management for the arid Aletai region in Northern China. The Rangeland Journal. 28, 97–104.

Guo, N., Lanzhou, C.M.A., Wang, X., Cai, D., Yang, J., 2007. Comparison and evaluation between MODIS vegetation indices in Northwest China. Geoscience and Remote Sensing Symposium, 2007. IGARSS 2007. IEEE International. Barcelona: IEEE. 3366 - 3369. Hilbig, W., 1995. The Vegetation of Mongolia. Amsterdam, Netherlands: SPB Academic Publishing. Hilker, T., Natsagdorj, E., Waring, R.H., Lyapustin, A., Wang, Y., 2014. Satellite observed widespread decline in Mongolian grasslands largely due to overgrazing. Global Change Biology. 20, 418–428. doi:10.1111/gcb.12365. Huete, A., Didan, K., Miura, T., Rodriguez, E.P., Gao, X., Ferreira, L.G., 2002. Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sensing of Environment. 83, 195-213. Huete, A.R., Justice, C., 1999. MODIS vegetation index (MOD13) algorithm theoretical basis document. Version 3. https://modis.gsfc.nasa.gov/data/atbd/atbd_mod13.pdf Kawamura, K., Akiyama, T., Yokota, H., Tsutsumi, M., Yasuda, T., Watanabe, O., Wang, S., 2005. Comparing MODIS vegetation indices with AVHRR NDVI for monitoring the forage quantity and quality in Inner Mongolia grassland, China. Grassland Science. 51, 33– 40. Li, X.R., Jia, X.H., Dong, G.R., 2006. Influence of desertification on vegetation pattern variations in the cold semi-arid grasslands of Qinghai-Tibet Plateau, North-west China. Journal of Arid Environments. 64, 505-522 Li, Z., Li, X., Wei, D., Xu, X., Wang, H., 2010. An assessment of correlation on MODIS-NDVI and EVI with natural vegetation coverage in Northern Hebei Province, China. Procedia Environmental Sciences. 2, 964–969. Liu, H.Q., Huete, A.R., 1995. A feedback based modification of the NDV I to minimize canopy background and atmospheric noise. IEEE Transactions on Geoscience and Remote Sensing. 33, 457-465. Liu, Y.S., Hu, Y.C., Peng, L.Y., 2005. Accurate Quantification of Grassland Cover Density in an Alpine Meadow Soil Based on Remote Sensing and GPS. Pedosphere. 15 (6), 778-783. Luo, L., Wang, Z., Ren, C., Song, K., Li, X., 2010. Models for estimation of grassland production and spatial inversion based on MODIS data in Songnen Plain. Transactions of the Chinese Society of Agricultural Engineering. 26, 182-187. Maroney, R.L., 2005. Conservation of argali Ovis ammon in western Mongolia and the Altai- Sayan. Biological Conservation. 121, 231-241. Marsett, R.C., Qi, J., Heilman, P., Biedenbender, S.H., Watson, M.C., Amer, S., Weltz, M., Goodrich, D., Marsett, R., 2006. Remote Sensing for Grassland Management in the Arid Southwest. Rangeland Ecological Management. 59, 530–540. Matsushita, B., Yang, W., Chen, J., Onda, Y., Qiu, G., 2007. Sensitivity of the Enhanced Vegetation Index (EVI) and Normalized Difference Vegetation Index (NDVI) to Topographic Effects: A Case Study in High-Density Cypress Forest. Sensors. 7, 2636- 2651. Modina, T.D. 1997 . Climate of Altai Republic. Novosibirsk: NPU. Naidansuren, E., Bayasgalan, O., 2012. An Economic Analysis of the Environmental Impacts of Livestock Grazing in Mongolia. Research Report, Singapore: Economy and Environment Program for Southeast Asia.

Nakano, T., Bavuudorj, G., Urianhai, N.G., Shinoda, M., 2013. Monitoring aboveground biomass in semiarid grasslands using MODIS images. Journal of Agricultural Meteorology. 69 (1), 33-39. NASA, 2016. NASA Distributed Active Archive Center (DAAC) at NSDIC. MODIS Data. Accessed July 5, 2016. https://nsidc.org/data/modis/terra_aqua_differences. National Research Council, 1994. Rangeland health: new methods to classify, inventory, and monitor rangelands. Washington, DC: National Academy Press. Olson, J.S., Watts, J.A., Allison, L. J., 1983. Carbon in Live Vegetation of Major World Ecosystems. Report ORNL-5862, Tennessee: Oak Ridge National Laboratory. Paltsyn, M.Yu., Spitsyn, S.V., Kuksin, A.N., Lkhagvasuren, B., Onon, Yo., Munkhtogtogh, O., 2011. Conservation of Altai argali in the transboundary area of Russia and Mongolia. Krasnoyarsk: WWF-Russia and UNDP/GEF Project “Biodiversity Conservation in the Russian portion of Altai-Sayan Ecoregion". [in Russian] Purevdorj, TS., Tateishi, R., Ishiyama, T., Honda, Y., 1998. Relationships between percent vegetation cover and vegetation indices. International Journal of Remote Sensing. 19(18), 3519-3535. doi:10.1080/014311698213795. Reeves, M.C., Winslow, J.C., Running, S.W., 2001. Mapping weekly rangeland vegetation productivity using MODIS algorithms. Journal of Range Management. 54, A90–A105. Stumpp, M., Wesche, K., Retzer, V., Miehe, G., 2005. Impact of grazing livestock and distance from water source on soil fertility in southern Mongolia. Mountain Research and Development. 25, 244–251. Tan, B., Morisette, J.T., Wolfe, R.E., Feng, G., Ederer, G.A., Nightingale, J., Pedelty, J.A., 2008. Vegetation Phenology Metrics Derived from Temporally Smoothed and Gap-Filled MODIS Data. Geoscience and Remote Sensing Symposium, 2008. IGARSS 2008. IEEE International. III, 593-596. Tueller, P.T., 2001. Remote sensing of range production and utilization. Journal of Range Management. 54, A77–A89. Wang, D., Morton, D., Masek, J., Wu, A., Nagol, J., Xiong, X., Levy, R., Vermote, E., Wolfe, R., 2012. Impact of sensor degradation on the MODIS NDVI time series. Remote Sensing of Environment. 119, 55-61. Wang, J., Guo, N., Wang, X., Yang, J., 2007. Comparisons of normalized difference vegetation index from MODIS Terra and Aqua data in northwestern China. IEEE. 3390-3393. White, R.P., Murray, S., Rohweder, M., 2000. Pilot Analysis of Global Ecosystems: Grassland Ecosystems. Technical Report, Washington DC: World Resources Institute. Wilson, J.B., Peet, R.K., Dengler, J., Partel, and M., 2012. Plant species richness: the world records. Journal of Vegetation Science. 23, 796–802. doi:10.1111/j.1654- 1103.2012.01400.x. Yan, Y., Lu, X., 2015. Is grazing exclusion effective in restoring vegetation in degraded alpine grasslands in Tibet, China? PeerJ 3. doi: 10.7717/peerj.1020 Yang, W., Shabanov, N.V., Huang, D., Wang, W., Dickinson, R.E., Nemani, R.R., Knyazikhin, Y., Myneni, R.B., 2006. Analysis of leaf area index products from combination of MODIS Terra and Aqua data. Remote Sensing of Environment. 104, 297–312. Yatagai, A., Yasunari, T., 1995. Interannual variations of summer precipitation in the arid/semi- arid regions in China and Mongolia: their regionality and relation to the asian summer monsoon. Journal of the Meteorological Society of Japan. 73, 909–923.

Zhao, B., Ma, L., 2011. Multi-source data complex classification of grassland: case study of Inner Mongolia grassland. In: Proceedings of SPIE - The International Society for Optical Engineering 8006:91. November 2011. DOI: 10.1117/12.902734

CHAPTER 2: INTEGRATING TRADITIONAL ECOLOGICAL KNOWLEDGE AND REMOTE SENSING FOR MONITORING RANGELAND DYNAMICS IN THE ALTAI MOUNTAIN REGION

Abstract: Integrating traditional ecological knowledge (TEK) with remote sensing capabilities to monitor rangeland dynamics could lead to more acceptable, efficient and beneficial rangeland management schemes for all stakeholders of grazing systems. We contrasted pastoralists’ perception of summer pasture quality in the Altai Mountains of Central Asia with Normalized Difference Vegetation Index (NDVI) metrics obtained from Terra Moderate Resolution Imaging Spectroradiometer (MODIS) satellite sensor. Herders’ estimates of forage value at 49, 1x1 km grassland blocks sampled in July 2013 were positively and strongly (> 63% of variance explained) related to NDVI (MODIS 1 km monthly data, MOD13A3) as well as % vegetation cover (>62% explained) and to a lower degree to vegetation height (>28% explained). NDVI temporal anomaly explained > 11% of variance in estimates of pasture quality recalled among 187 herders of pasture quality over the period 2006-2016; substantial variation in strength of recall occurred among countries (Russia, Mongolia, China) based in part of herder dependence on herding as a livelihood and occurrence of extreme climate events. We conclude that herder-derived estimates of pasture forage value at the peak of growing season generally corroborate satellite-derived vegetation indices in the Altai Mountain region thereby enabling re-expression of satellite data in a form more compatible with herder perceptions.

Introduction Grasslands occupy >25% of the Earth’s surface (Olson, Watts and Allison 1983; White, Murray and Rohweder 2000) and support >20 million households as pastures for their livestock (De Haan, Steinfeld, and Blackburn 1997). Large-scale monitoring of grassland ecosystems has been enabled by satellite-borne remote sensing since the 1970s (Booth and Tueller 2003; Liu, Hu and Peng 2005) via vegetation indices derived from grassland vegetation’s spectral responses. Vegetation indexes have provided a basis for phenological monitoring, vegetation classification, and derivation of structural vegetation parameters, such as plant greenness, biomass and leaf area index (Huete et al. 2002). A complementary but vastly underutilized source of information for grassland assessment and monitoring are pastoralists themselves. Pastoralists’ continuous interaction with the environment and their keen observation skills yield a practical understanding of grassland ecology and dynamics often referred as traditional ecological knowledge (TEK) (Spooner 1973; Niamir-Fuller 1995; Fernandez-Gimenez 2000). Pastoralists’ TEK includes an ability to efficiently characterize pasture quality, livestock carrying capacity, optimal timing for grazing, palatability of different plant species for livestock, protective features of landscape, seasonal forage distribution and availability, and other dimensions of grasslands and their use (Fernandez-Gimenez 2000; Oba and Kotile 2001; Ghorbani et al. 2013; Egeru et al. 2015), which directly influences how they use the resource (Homewood and Rodgers 1989; Mills, et al. 2002). Pastoralists’ TEK is still poorly documented, however, which makes its integration into grassland research, management, and policy development problematic (Fernandez-Gimenez 2000; Oba 2012) or simply neglected (Brown 1971; Roba and Gufu 2009). Most importantly, rangeland monitoring schemes, no matter how scientifically rigorous, when disconnected from TEK often have low perceived relevance by local communities and hence limited acceptance for use in locally based decision-making (Fernandez- Gimenez 2000; Selemani et al. 2012). Integrating TEK and remote sensing for monitoring rangeland dynamics could make monitoring and management schemes more acceptable and beneficial for all stakeholders of grazing systems

21 including herder communities, regional decision makers and ecologists (Robbins 2003). Over the last two decades some efforts have been made to integrate remote sensing and TEK for survey, monitoring and management of different ecosystems (Hellier, Newton and Gaona 1999; Robbins 2003; Huntington et al. 2004; Naidoo and Hill 2006; Lauera and Aswanib 2008; Maynard et al. 2008; Polfus, Heinemeyer, and Hebblewhite 2014), but few studies were devoted to contrasting of herder knowledge and satellite-derived data in grassland research and monitoring (Klein et al. 2014; Egeru et al. 2015; Fernandez-Gimenez et al. 2015). Clearly the mechanics for integrating TEK and remote sensing for grassland monitoring and management require further evaluation as practical cases of integration remain rare (Sulieman and Ahmed 2013; Egeru et al. 2015; Fernandez-Gimenez et al. 2015). Our study was focused on summer pastures in the Altai Mountains (parts of Mongolia, Russia, China, and Kazakhstan) over 10 years (2006-2016) where herder communities are heavily dependent on rangelands for their livelihoods. Most of the region lacks any coherent policy or monitoring scheme to help balance rangeland use by pastoralists and their livestock with conservation of the significant biodiversity values of the region (Maroney 2005; Bailagasov 2011; Addison et al., 2012; Naidansuren and Bayasgalan, 2012; WWF 2012). To address this research gap, our study’s objective was to elucidate the association between pastoralists’ TEK and satellite- derived MODIS Normalized Difference Vegetation Index (NDVI) of grassland condition. Our purpose was to explore the degree of corroboration provided by the (1) spatial relationship between satellite-based assessment of the grassland quality and on-the-ground evaluation by local herders employing traditional measures of forage value for livestock, and (2) temporal relationship between NDVI and herder assessments in the context of herders’ ability to reconstruct historical rangeland conditions. We tested two hypotheses critical to the concept that satellite- and herder- derived indices of rangeland quality are compatible: (a) herders’ estimates of grassland forage value for livestock positively correlate with satellite-derived indices across a strong spatial gradient of grassland productivity; and (b) herder’s perception of summer pasture quality for a particular year positively correlates temporally with satellite-derived indices for the same pastures over the time frame of a decade. Correspondence of herder perceptions and satellite-derived indices in a predictable manner would create opportunity for re-expressing satellite-derived indices in terms relevant to herders and potentially enable efficient propagation of TEK-adjusted satellite pasture monitoring and management over large landscapes.

Materials and Methods Study Area Our study area in the Altai Mountains (47º35' - 50º13' N, 85º32' - 90º43' E) incorporated spatial adjacent segments of Mongolia (Ulaankhus and Nogoonnuur somons (districts) of Bayan-Olgii Aimag), Russia (Kosh-Agach district of Altai Republic), China (Altai Prefecture of Xinjiang Autonomous Region), and the Republic of Kazakhstan (Katon-Karagay district of Eastern Kazakhstan Region) (Fig. 1). This is high mountain grassland region in the very centre of the Eurasian continent traditionally used by Kazakh and Altaian people for livestock grazing. The Altai is also recognized as a UNESCO World Heritage Site (UNESCO 1992-2016) for its landscape diversity and high value for conservation of endangered species, including snow leopard (Panthera uncia) and Altai argali (Ovis ammon ammon) (Maroney 2005; Paltsyn et al. 2012).

The area is characterized by mountains, hills and intermountain depressions (elevation range: 800 - 4,506 m). In the Mongolian and Russian parts of our study area the landscape is dominated by mountain grasslands, whereas in the Kazakhstan and China sectors coniferous forests interspersed with grasslands are more prominent. The region has a strong continental climate with a short growing season (April-August), severe winters, wide annual precipitation range (from 121 mm in Bayan-Olgii, Mongolia, to 441 mm in Katon-Karagay, Kazakhstan), and average annual temperature from -4.0º (Kosh-Agach, Russia) to 2.8ºC (Altay, China) (unpublished data from Bayan-Olgii, Kosh-Agach, Katon-Karagay, and Altay meteorological stations for 2005-2014). Mean grassland productivity on summer pastures, as captured by MODIS NDVI values, varies inversely with its temporal variance such that the regions with the highest mean NDVI have the lowest annual variation (e.g., Kazakhstan segment) and vice versa (e.g., Mongolia segment, Fig. 2).

Figure 1. Study area in the Altai Mountains including Ulaankhus and Nogoonnuur somons of Bayan-Olgii Aimag, Mongolia; the Kosh-Agach district of Altai Republic, Russia; the Katon- Karagay district of Eastern Kazakhstan Region, Republic of Kazakhstan; and Altai prefecture of Xinjiang Autonomous Region, China. Point locations are given of 1 km2 sampling blocks assessed in July-August 2013 (gray blocks) and also for summer camps (black dots) where herders were interviewed in 2015-2016 about summer pasture quality in 2006-2016.

0.12

Mongolia

2015 -

Russia

0.07

August August 2006 -

Kazakhstan

China NDVI CV June 0.02 0.3 0.4 0.5 0.6 0.7 0.8 Mean NDVI June-August 2006-2015

Figure 2. Ecological contrasts among study regions in the Altai Mountains in terms of mean NDVI and its temporal coefficient of variance (CV, June-August 2006-2015) for grasslands around summer camps (5 km buffer) of herders interviewed about summer pasture conditions in Mongolia, Russia, China, and Kazakhstan.

Analytical approach Contrasting herder-derived estimates of grassland forage value and Terra MODIS NDVI data We obtained co-located and temporally coincident estimates of rangeland condition for both MODIS-derived Normalized Difference Vegetation Index (NDVI) and herder-derived forage values at 92, 1x1 km blocks sampled in 2013 (Fig. 1). All sampled blocks were located between 1495 to 2860 m above sea level that we purposely selected to represent the full range of grassland cover present in the region from mountain grassy tundra to dry grasslands and semi-deserts. NDVI is a frequently used index for grassland monitoring (Huete et al., 2002) and to obtain remotely- sensed estimates of conditions of ground vegetation, we accessed Terra MODIS NDVI Monthly L3 Global 1km SIN Grid V005 dataset (MOD13A3, tile h23v04) for July 2013 downloaded from the NASA’s EOSDIS Reverb web-site (http://reverb.echo.nasa.gov/reverb/). For field sampling the MODIS tile was re-projected to WGS 1984 UTM Zone 45 projection and clipped to our study area. Coordinates of centers of every 1x1 km pixel in the study area were calculated in ArcGIS 10.2.2 (ESRI 2014).

Between July 18 - August 3, 2013 we sampled herder-derived forage value as well as percentage of vegetation cover, vegetation height, and percentage of grass and forbs at the same 92 sites where MODIS NDVI estimates were gathered (Fig. 1). To estimate the vegetation parameters in each sampled block a rangeland ecologist walked simultaneously with a local herder in an ever- increasing, spiral-shaped, 4800-m-long transect guided by a hand-held GPS unit achieving approximately equal sampling coverage of each block via the transect design (Fig. 3). A total of 4

24 herder-ecologist pairs completed the survey work. The ecologist-herder pair stopped every 25 paces along the transect to enable the herder to make a point estimate of vegetation forage value (0 - low, 1 – medium, and 2- high). At the same point the ecologist made a point measurement of vegetation cover (vegetation versus bare ground), vegetation height (none: 0 cm, low: 0.1-5.0 cm, medium: 5.1-15.0 cm, and high: > 15.0 cm) and vegetation composition (none, grass or forb) at the sampling point touched by a pin placed at the leading edge of the surveyor’s foot. The point estimates accumulated along a given transect in each block were used to calculate (a) herder- derived forage value for a given sampling block as weighted average of points in three categories; (b) proportion of points that were vegetated among all sampled points within a given sampling block; (c) average vegetation height as a weighted average of points in four height categories; and (d) proportion of points that were grass or forb among all sampled points within given sampling block (average n = 215 points per sampled block, range = 167-255 depending on stride length of surveyors).

Figure 3. Sampling design used for estimating of vegetation cover and herder-derived forage value on each of 92, 1x1km blocks aligned with MODIS grid in the Sailugem area of western Mongolia portion of the Altai Mountains in July-August 2013. Black frame represents a 1x1 km sampling block; red lines depict segments of a 4800-m-long sampling transect and arrows the direction of movement during sampling.

Linear regressions were used to contrast the relationship (based on the coefficient of determination, R²adj) of herder-derived vegetation forage estimates to other grassland parameters measured by the ecologist and to the spatially co-located and temporally co-incident Terra NDVI values. Only pixels of highest reliability (QA = 0 “Good data”) (NASA LP DAAC, 2016) within the MODIS NDVI dataset were used for the regressions (n=49 from sampled 92 blocks). We used the same regression approach to contrast herders’ estimate of forage value with percentage of vegetation cover, vegetation height and percentage of grass and forb in the vegetation cover on the same 49 sampling blocks to understand how these vegetation parameters influence herder’s estimates on the grassland forage value.

Evaluating correspondence between herders’ recollection of summer pasture quality and satellite- derived indices over the previous decade

We compiled time series of herder recollections of rangeland quality during July-August 2015 and August 2016 through interviews with 187 local herders on summer pastures throughout the Altai Mountain region: in Ulaankhus and Nogoonnuur somons of Bayan-Olgii Aimag, Mongolia (n=53); Kosh-Agach district of Altai Republic, Russia (n=51); Katon-Karagay district of Eastern Kazakhstan region, Kazakhstan (n=34); and Altai prefecture of Xingjian, China (n=49) (Fig. 1). Herders were asked to recall an estimate of the quality their summer pasture for each year in sequence between 2006-2015 (2006-2016 in China) using the following scale: “bad”, “average”, and “good”. Location of the herders’ summer camps were recorded with a hand-held GPS unit (Fig.1) enabling us to link the historical record of herder recollection of rangeland condition with the satellite-derived estimates for the same summer pastureland. We extracted mean annual NDVI values for June-August (the period each year when herders locate to summer pastures) for 2006-2015 for areas < 5 km from locations of each summer camp of interviewed herders using inverse distance weighted average:

푛 1 ∑푖=1 ⁄ 2푁퐷푉퐼푖 푑푖 푁퐷푉퐼 퐼퐷푊 푚푒푎푛 = 푛 1 , ∑푖=1 ⁄ 2 푑푖 where 푑푖 is a distance (meters) from pixel i center to the interviewed herder’s camp. We extracted from herder interviews: 1) percentage of herders who were able to recall pasture conditions for each particular year in 2006-2015 (2006-2016 in China) and 2) average estimate from among those herders reporting of summer pasture quality for each year calculated as percentage of mean perception over the same 10 years (data not available for Kazakhstan, see Results), coded as bad = 0, average = 1, and good = 2. The Gini-Simpson index (Jost 2006) was used as a measure of level of herder agreement about estimates of the pasture quality in any given year. Simple linear and multiple regressions were used to relate herders’ perception and agreement on pasture quality with NDVI anomaly for each year calculated as percentage of the mean NDVI for a site over the same 10 years (2006-2015, 2006-2016 in China) to which herders’ recollections pertained.

Results Contrasting herder-derived estimates of grassland forage value and Terra MODIS NDVI data Vegetation parameters across the 49, 1x1km grassland blocks with highest reliability (QA = 0 “Good data”) varied as follows: % vegetation cover – from 9% to 100% (mean = 60%), average vegetation height – from 0 to 10 cm (mean = 4 cm), % grass – from 22% to 90% (mean = 55%), and percentage of forb – from 11% to 93% (mean = 44%). Average herder-derived forage value varied from 0.28 to 1.91 (mean = 1.11 with theoretical minimum = 0 and maximum = 2). Herders’ estimates of forage value in July 2013 were strongly and positively correlated with % vegetation cover (>62% of variance explained, Figure 3 b) and vegetation height (>28% of variance

26 explained, Figure 3 c) measured on those same sampling blocks. No relationship (P > 0.05) was observed between forage value and % grass or % forb in vegetation cover. Multiple linear regression with standardized variables indicated that percentage of vegetation cover exerted > x2 higher influence on herder perception of pasture forage value than did vegetation height (Table 1). In terms of relationships between ground-level assessments and satellite-derived estimates, herders’ estimates of forage value in July 2013 were strongly and positively correlated with Terra MODIS NDVI sampled in the same month (> 63% of variance explained, Figure 4 a). Both % vegetation cover and vegetation height correlated positively with NDVI but with different degrees of variance explained (69% and 11%, respectively, Fig. 4 d and e).

2 2

1.5 1.5

1 1

derived value forage derived value forage 0.5 - - 0.5

R² = 0.6431 R² = 0.6223 Herder Herder 0 0 0 0.2 0.4 0.6 0 20 40 60 80 100 NDVI % of vegetation cover

p<0.0001 p<0.0001 a b 2 100

80 1.5 60 1 40

R² = 0.2888 derived value forage

- 0.5 R² = 0.6967

% of vegetation vegetation of % cover 20

Herder 0 0 5 10 0 0.2 0.4 0.6 NDVI Vegetation height, cm p<0.0001 p<0.0001 c d 10

Vegetation height, Vegetation height, cm R² = 0.1177 0 0 0.2 0.4 0.6 NDVI p<0.05 e Figure 4. Relationships among herders’ estimation of forage value, Terra-derived MODIS NDVI, % vegetation cover, and vegetation height at 49 sampling sites in Sailugem Range of the western Mongolia portion of the Altai Mountains during July-August 2013.

Table 1. Parameters and coefficients of determinations (R²) from multiple linear regression relating herder-derived forage value to % vegetation cover and vegetation height on 1x1 km blocks (n=49) sampled in Sailugem Range in western Mongolia, July-August 2013. Variables standardized to permit comparison of relative contributions of each to variation in herder-derived forage value.

Parameter Coefficient Std. Err. t P Intercept 0.000 0.081 0.00 1.000 % of Veg. Cover 0.684 0.088 7.775 <0.0001 Vegetation Height 0.285 0.088 3.245 0.002 Model F = 51.84; P <0.0001; R²adj = 0.679

Evaluating herders’ recollection of summer pasture quality over 2006-2016 Over the decade-long assessment period (2006-2015 [2006-2016 for China]), 98% of interviewed herders in Mongolia and China, 86% in Russia, and only 6% in Kazakhstan could report about the pasture conditions back to 2006 (lack of herder recollection in Kazakhstan precluded further analysis of temporal patterns of herder assessments). Mean herder perception of pasture quality in a particular year (percentage of mean perception of pasture quality for 10 years (2006-2015 [2006- 2016 for China])) varied from 33.6% (2010) to 157.8% (2013) in Mongolia (CV = 0.408); from 84.8% (2015) to 104.5% (2012) in Russia (CV = 0.056); and from 39.3% (2015) to 132.1% (2006) in China (CV = 0.286). NDVI anomaly varied over the same period from 85.0% (2008) to 125.1% (2013) in Mongolia (CV = 0.112), from 85.4% (2008) to 115.7% (2013) in Russia (CV = 0.082), and from 91.6% (2008) to 105.4 (2016) in China (CV = 0.034) (Fig. 5, a-c). Herder perceptions and satellite indices of annual variation in grassland productivity were similar insofar as NDVI anomaly explained > 11% of variance in average herder score of pasture quality (P < 0.05, Fig. 5 d, Table 2, A). When data were subset separately for Mongolia, Russia, and China, NDVI anomaly explained >30% of the variance in Mongolia where this regression was nearly significant (P = 0.056) (Table 2, B), only ~4% of variance in the herder perception in Russia (P = 0.280) (Table 2, C), and ~7% of variance in China (P = 0.552) (Table 2, D). Similarly, NDVI anomaly combined with years lapsed from 2016 explained >32% of variation in herder agreement on pasture conditions (as measured by Gini-Simpson Index, Table 2, E; Fig. 5 e) indicating that there was more agreement among herders with more years lapsed. When data were subset separately for Mongolia, Russia, and China years lapsed remained a significant predictor of herder agreement on pasture quality for both Mongolia and Russia whereas NDVI anomaly remained so for Mongolia only (Table 2, F and G). In case of China both predictors – NDVI anomaly and years lapsed – were not significant (Table 2, H). Increased herders’ agreement on the pasture quality (reduced Gini- Simpson Index) with years lapsed back from 2015 for both Russia and Mongolia was associated with herders tending to rank pasture quality to an average level of as “medium” with increasingly years lapsed (Fig. 4, e). Notably, variability of Gini-Simpson Index was higher in Mongolia (CV = 0.210) than in Russia (CV = 0.187) and China (CV=0.139) (Table 2, F; Fig. 5, d).

Table 2. Parameters of linear regressions describing relationship between herder perception of summer pasture quality, herder agreement (Gini-Simpson Index), NDVI anomalies in June-August in the 5 km buffer zones of herder camps in Mongolia, Russia, and China portions of the Altai Mountains over the past decade. Parameter Coefficient SE t P Herder perception of pasture quality versus NDVI anomaly A. All countries: Intercept -35.776 61.271 -0.584 0.564 NDVI anomaly 1.358 0.611 2.223 0.034 B. Mongolia: Intercept -125.688 101.608 -1.237 0.251 NDVI anomaly 2.257 1.011 2.234 0.056 C. Russia: Intercept 74.016 22.507 3.289 0.011 NDVI anomaly 0.260 0.224 1.158 0.280 D. China: Intercept 270.853 276.617 0.979 0.353 NDVI anomaly -1.709 2.765 -0.618 0.552 Herder agreement versus NDVI anomaly and years lapsed E. All countries: Intercept 0.525 0.116 4.522 0.0001 NDVI anomaly -0.002 0.001 -1.555 0.131 Years lapsed -0.012 0.003 -4.034 0.0004 F. Mongolia: Intercept 0.799 0.103 7.736 0.0001 NDVI anomaly -1.341 0.299 -4.487 0.003 Years lapsed -0.021 0.003 -5.954 0.0006

G. Russia: Intercept 0.334 0.115 2.911 0.023 NDVI anomaly 0.001 0.001 0.446 0.669 Years lapsed -0.017 0.003 -5.686 0.001 H. China: Intercept 0.326 0.531 0.615 0.556 NDVI anomaly -4.514E-05 0.005 -0.009 0.993 Years lapsed -0.003 0.005 -0.495 0.634 A: F = 4.94; p=0.034; R²adj = 0.116 B: F = 4.99; p=0.056; R²adj = 0.307 C: F = 1.34; p=0.280; R²adj = 0.037 D: F = 0.38; p=0.552; R²adj = 0.066 E: F = 8.14; p =0.002; R²adj = 0.322 F: F = 18.97; p =0.002; R²adj = 0.800 G: F = 18.30; p =0.002; R²adj = 0.794 H: F = 0.16; p =0.850; R²adj = 0.200

150 150 150 130 130 130 110 110 110 90 Anomaly (%) Anomaly 90

90 (%) Anomaly Anomaly (%) Anomaly 70 70 70 50 50 50 30 30 30 2005 2010 2015 2005 2010 2015 2005 2010 2015 Years Years Years a. Mongolia (gray - herders’ b. Russia (gray - herders’ c. China (gray - herders’ perception, perception, black – NDVI anomaly) perception, black – NDVI anomaly) black – NDVI anomaly) 180 0.500 140 0.400

100 2016 mean) 2016

- 0.300

60 Index Simpson

- 0.200 Gini from 2006from R² = 0.1455 R² = 0.313

20 0.100 Annual herder perception (% (% perception Annual herder 80 90 100 110 120 130 0 2 4 6 8 10 Annual NDVI (% from 2006-2016 mean) Years lapsed d e 50 0.450 140

0.400 40 120 0.350 30

Simpson Index 0.300 - 100

20 % anomaly, NDVI

Gini 0.250 # of herdersreporting of #

"medium" qualitypasture"medium" 10 0.200 80 0 5 10 2006 2008 2010 2012 2014 Years lapsed from 2015 Years f. Russia (gray dots) and Mongolia (black dots) g. Mongolia (herder agreement = black dots, NDVI anomaly = gray dots) Figure 5. Temporal patterns of herder perceptions of grassland quality and satellite-derived measures. (a - c): NDVI anomaly in June-August and herder’s perception of summer pasture quality in the 5 km buffer zones of herder camps measured over 2006-2016 in in Altai Mountain range, Mongolia, Russia, and China. (d - e): Linear regressions describing relationships between (d) herder perception of pasture quality and NDVI anomaly in June-August in the 5 km buffer zones of herder camps in Mongolia, Russia and China in 2006-2016; and (e) Gini-Simpson Index (measure of herders’ agreement) and years lapsed from 2016. (f - g): Number of herders in Russia (gray) and Mongolia (black) reporting “medium” summer pasture quality in 2006-2015 (f); Dynamics of Gini-Simpson Index as a measure of herder agreement (black) and NDVI anomaly in Mongolia (grey) between 2006-2015 (g).

Discussion By evaluating relationships between herder TEK and satellite-derived measures of rangeland condition over time and space we sought to understand the degree of correspondence between these two approaches and hence opportunity for integration of herders’ TEK and remote sensing derived indices of rangeland quality. Our results demonstrate a relatively high level of consistency between herder-derived estimates of pasture forage values and monthly Terra MODIS NDVI datasets, at least in western Mongolia (Fig. 3 a), which supports our first hypothesis that satellite indices and herders perceptions reflect in a similar manner spatial variation in grassland productivity. NDVI is a complex index in terms of its relationship to ecological parameters measured on the ground and of relevance to herders. Relative composition of grass and forbs in the vegetation cover apparently had no influence on the herders’ assessment of pasture forage value, whereas forage value strongly associated with percentage of vegetation cover (68% of overall variance explained) and less so with vegetation height (30%). We did not integrate plant species composition and relative palatability known to be used by Mongolian herders to estimate pasture forage quality (Fernandez-Gimenez 2000) because these parameters are unlikely to be reflected in variation in the MODIS vegetation index. Mongolian herders do rely heavily on vegetation cover to assess pasture quality (Fernandez-Gimenez 1997, 2000) with vegetation cover known to drive variation in NDVI (e.g., Purevdorj et al., 1998; Liu et al. 2005; Guo et al. 2007; Cui et al. 2012). In addition to vegetation cover, vegetation height is also used in the herders’ estimates of pasture forage value (Fernandez-Gimenez 1997, 2000). Both cover and height were positively correlated with monthly MODIS NDVI data, however, magnitude of these correlations was much stronger for percentage of vegetation cover (>69% of variance explained) than for vegetation height (~12%). This was expected as optical sensors, such as MODIS, return a signal from the surface skin with limited ability to measure heights. Similarly, Kong et al. (2015) also found that assessment of rangeland conditions by local herders in the rangelands of South African Kalahari Duneveld was consistent with on-the-ground measurements of vegetation and bare ground cover whereas Landsat-derived NDVI, Soil Adjusted Vegetation Index (SAVI) and tasseled cap greenness poorly correlated with both herders’ assessment and on-the-ground measurements due to soil parameters in that influenced performance of the vegetation indices (Kong et al. 2015). Other cases of spatial correlations between TEK and remotely sensed indices for spatial assessment and mapping of ecosystems have been described. Naidoo and Hill (2006) found that vegetation classification of Mbaracayu Forest Reserve in Paraguay by Ache indigenous tribe was consistent with a supervised classification of Landsat 7 TM imagery of the reserve. Lauera and Aswani (2008) also described a successful case of integration of indigenous ecological knowledge of fishers from Solomon Islands and remote sensing analysis to interpret Landsat-7 ETM multi-spectral satellite image and produce accurate broad-scale marine habitat maps useful for management. In a related study, Pitt et al. (2012) reported on high value of local ecological knowledge (LEK) as additional tool for remote sensing to identify isolated wetlands in South Carolina’s Piedmont and Blue Ridge regions: LEK yielded discovery of 5-6x more isolated wetlands than aerial photography analysis. Our analysis also supported the hypothesis of temporal correlation herder’s perception of summer pasture quality with satellite-derived indices for the same pastures, albeit with some important caveats. Although herder perception of summer pasture quality in Mongolia, Russia and China together reflected NDVI anomalies around their summer camps in 2006-2015, these correlations

32 were weak (R²adj = 0.12) (Table 2, A). When considered by country, herder perception of grassland quality over time was not related to NDVI anomaly in Russia and China, whereas in Mongolia it was marginally so (P = 0.056) and explained >30% of herders’ perception variance. Notably, negative peaks of herders’ perception of 2010 both in Mongolia and China and another strong drop in herder recollection of pasture quality in China in 2015 (Fig. 5, a and c) were not associated with notable changes in grassland productivity as reflected by NDVI, but did coincide with dzud (heavy snowfall associated with unusually low temperature leading to massive livestock die-offs). Dzud events occurred during winters 2009-2010 and 2015-2016 both in western Mongolia and Altai prefecture of Xinjiang and resulted in total ~10 million livestock died and significant economic losses for herders (Shang et al., 2012; Seaniger, 2016). No dramatic livestock decline was reported for the Russian part of our study area in 2006-2016. We conclude that herders’ perception of pasture quality in Mongolia and China were influenced by and perhaps conflated with effects of dzuds. A similar disproportionate influence of extreme climate events with dramatic impact on their livelihood than real changes of climate conditions on rangeland assessments was found for herder communities on Qinghai–Tibet Plateau (He and Richards, 2015). The perhaps counterintuitive pattern of herders’ agreement on the pasture quality increasing with more years lapsed (Table 2, E and Fig. 5, e) can be explained less by similarity of historical conditions than by herders tending to rank pasture quality as “medium” with increasing years lapsed (Fig. 5, f). We propose that herders likely recall pasture conditions with less specificity as period of recall increases such that they collectively converge upon a “medium” assessment, a pattern expressed in our data (Fig. 5, e). Mongolian herders tended to demonstrate higher agreement on years with “good” pasture quality and high mean NDVI values and have more disagreement about “bad” years with relatively low mean NDVI (Fig. 5, g). Thus, it seems years 2007, 2012, and 2013 with high vegetation productivity triggered widespread agreement by majority of herders as “good” years, however, relatively low production years (2008-2010) were perceived quite differently among herder groups. Regional-scale factors, particularly grassland productivity and its variability, also seemed to affect degree of correlation of herder perception with satellite-derived NDVI values. More specifically, mean NDVI on summer pastures and its temporal variation were inversely related among countries such that the least productive areas (Mongolia) also had the highest temporal variation in productivity while the most productive areas (Kazakhstan) had one of the lowest temporal variation (Fig. 2). Remarkably, no more than 6% of interviewed herders in Kazakhstan could recall pasture conditions back to 2006 whereas most herders in Russian and Chinese Altai reported on pasture conditions back to 2006; however, the perception of herders in Russia and China of pasture conditions was not related to NDVI anomaly. Only in Mongolian Altai with its low pasture productivity and high productivity variance NDVI anomaly was consistently related to herder assessment and agreement on pasture quality in different years. We conclude that stronger influence of vegetation productivity on socio-economic conditions of the herders’ livelihoods (mainly number of livestock they depend on), which is the case in Mongolia, drives more detailed and consistent recollection of actual pasture conditions whereas in areas where dependence of livestock is less (Russia, Kazakhstan, and China) herder recall is correspondingly weaker.

Table 3. Summary of rangeland conditions across the Altai Mountain study area including mean NDVI, NDVI CV, percentage of herders that could report on pasture conditions back to 2006, and significance of NDVI influence on herder perception of pasture quality and level of agreement in Mongolia, Russia, China, and Kazakhstan over previous decade. Country Mean NDVI NDVI CV % of herders Significance of Significance of (2006-2016) (2006-2016) that could NDVI influence NDVI influence report back to on herder on herder 2006 perception of agreement on pasture quality pasture quality Mongolia >30% variance explained, 0.314 0.112 98% significant marginally significant Russia 0.418 0.082 86% insignificant insignificant China 0.521 0.034 98% insignificant insignificant Kazakhstan 0.717 0.039 6% - -

Other studies have demonstrated varying strengths of herder recall associated with regional factors. Egeru et al. (2015) in savanna grasslands of Uganda found similar, but much stronger positive correlations (R² = 0.79) between herders’ perception of forage availability and monthly long-term NDVI values derived from National Oceanic and Atmospheric Administration-Advanced Very High Resolution Radiometer (NOAA-AVHRR). Fernandez-Gimenez et al. (2015) reported about agreement of herder perception of declining trend in rangeland production and AVHRR-NDVI trend in Selenge and Tuv Aimags of Mongolia in 1993-2013. Klein et al. (2014) discovered that long-term NDVI trend (1982-2010) toward delayed rangeland green-up on the Tibetan Plateau corresponds with herders’ local observations of shifts in grassland conditions. The primary conclusion of our study is that herder-derived estimates of pasture forage value at the peak of growing season can be used as additional tool to complement widely used satellite-based methods for assessment of grasslands conditions at local level along with satellite-derived vegetation indices at least in western Mongolia. Western Mongolia, where we observed a high level of spatial and temporal correlations with MODIS NDVI and herder perceptions of rangeland quality, offers the best opportunity. Mongolian herders’ decision about pasture quality were influenced greatly by vegetation cover and to a lesser degree by vegetation height. Furthermore, as was reported by Fernandez-Gimenez (1997 and 2000), Mongolian herders generally rely on vegetation cover in their traditional assessment of pasture quality; therefore, this indicator can be better understood and accepted by local communities than other grassland parameters. Given strong correlations of vegetation cover and MODIS NDVI and EVI (Paltsyn et al., 2017) this simple and consistent with traditional herders’ knowledge indicator can be basic for participatory rangeland monitoring and management system in western Mongolia. Elsewhere in the Altai region, herders’ perceptions of summer pasture quality are positively and significantly correlated with mean MODIS NDVI value for the pastures in June-August for the same year but the relationships are weak and strongly affected by extreme climate events (e.g., dzuds) not always associated with inter-annual changes in pasture productivity. Generally speaking, we determined that satellite- derived estimates conform with herder’s perceptions but can be obtained at much broader spatial scale more efficiently. Similarly, herders’ TEK broadly validates the utility of satellite-derived indices for informing herders of trends and patterns in grassland productivity in the context of

34 parameters of most interest to them (e.g., vegetation cover), thereby possibly facilitating their acceptance of broad-scale, satellite-based rangeland monitoring and management. By extension, given the strength of the spatial relationships we observed in western Mongolia between NDVI variation and on-the-ground TEK estimates, one pragmatic means of increasing herder acceptance of insights from satellite-derived indices of rangeland dynamics would be mapping of herder- derived estimates of pasture forage value projected on the base of the MODIS NDVI values using regressions we have established between these two parameters. To conclude, we provide further support to Egeru et al. (2015), Fernandez-Gimenez et al. (2015), and Klein et al. (2014) regarding the consistency and complementarity of TEK of local herders on pasture and rangeland conditions with satellite-derived vegetation indexes and advocate for integration of local TEK in regional rangeland monitoring and management programes along with remote sensing techniques for a more efficient, applicable and acceptable approach to rangeland monitoring.

Acknowledgments This study was supported by the United State Agency for International Development USAID Climate Change Resilient Development Program (grant # CCRDCS0007) and NASA’s Land Cover Land Use Change Program (grant # NNX15AD42G). We are grateful to Munkhtogtokh Ochirjav (WWF-Mongolia), Atay Ayatkhan (Protected Area Administration of Mongol Altai, Mongolia), Leonid Bilagasov (Gorno-Altaisk State University, Russia), Yury Robertus (Altai Institute of Ecology, Russia), Andrey Chelyshev and Alia Gabdullina (Katon-Karagay National Park, Kazakhstan) for the project support and organization of interviews of local herders. We are also thankful to our field technicians, Khairat, Bagdat, Jukhan, and Totembek and Hamish Gibbs for their valuable help estimating grassland forage value and other parameters in Western Mongolia in July-August 2013.

References Addison, J., Friedel, M., Brown, C., Davies, J., Waldron, S (2012) A critical review of degradation assumptions applied to Mongolia’s Gobi Desert. Rangeland Journal. 34, 125–137. Bailagasov, L.V. (2011) The use of pastures in light of the establishment of the national park on the ridge Sailugem (Altai Republic). Steppe Bulletine 31: 11-18. Booth, D.T., and P.T. Tueller (2003) Rangeland Monitoring Using Remote Sensing. Arid Land Research and Management 17: 455–467. Brown, L. H. (1971) The biology of pastoral man as a factor in conservation. Biology Conservation 3: 93–100. doi:10.1016/0006-3207(71)90007-3. Cui, X., Z.G. Guo, T.G. Liang, Y.Y. Shen, X.Y. Liu, and Y. Liu (2012) Classification management for grassland using MODIS data: a case study in the Gannan region, China. International Journal of Remote Sensing 33 (10): 3156–3175. De Haan, C., H Steinfeld, and H. Blackburn (1997) Livestock and the Environment: Finding a Balance. Brussels: Commission on of the European Communities, Food and Agricultural Organization of the United Nations, and the World Bank. Egeru, A., O. Wasonga, J., Yazan, E. Mburu, M.G.J. Majaliwa, L. MacOpiyo, and Y. Bamutaze (2015) Drivers of forage availability: An integration of remote sensing and traditional

ecological knowledge in Karamoja sub-region, Uganda. Pastoralism: Research, Policy and Practice 5: 1-18. ESRI (2014) ArcGIS Desktop: Release 10.2.2. Redlands, California: Environmental Systems Research Institute. Fernandez-Gimenez, M. (2000) The role of Mongolian nomadic pastoralists’ ecological knowledge in rangeland management. Journal of Ecological Applications 10: 1318-1326. doi:10.1890/1051-0761(2000)010[1318:TROMNP]2.0.CO;2. Fernandez-Gimenez, M.E. (1997) Landscapes, livestock, and livelihoods: social, ecological, and land-use change among the nomadic pastoralists of Mongolia. PhD Dissertation,University of California, Berkeley, CA. Fernandez-Gimenez, M.E., J.P. Angerer, A.M. Allegretti, S.R. Fassnacht, A. Byamba, J. Chantsallkham, R. Reid, and N.B.H. Venable (2015) Integrating Herder Observations, Meteorological Data and Remote Sensing to Understand Climate Change Patterns and Impacts across an Eco-Climatic Gradient in Mongolia. Proceedings of the Trans- disciplinary Research Conference: Building Resilience of Mongolian Rangelands. Ulaanbaatar, Mongolia. 228-234. Ghorbani, M., H. Azarnivand, A.A. Mehrabi, M. Jafari, H. Nayebi, and K. Seeland (2013) The Role of Indigenous Ecological Knowledge in Managing Rangelands Sustainably in Northern Iran. Ecology and Society 18 (2): 15. doi:10.5751/ES-05414-180215. Guo, N., C.M.A. Lanzhou, X. Wang, D. Cai, and J. Yang (2007) Comparison and evaluation between MODIS vegetation indices in Northwest China. Geoscience and Remote Sensing Symposium, 2007. IGARSS 2007. IEEE International. Barcelona: IEEE. 3366 - 3369. He, S., Richards, K. (2015) Impact of meadow degradation on soil water status and pasture management—a case study in Tibet. Land Degradation and Development 26: 468–479. Hellier, A., A.C. Newton, and S.O. Gaona (1999). "Use of indigenous knowledge for rapidly assessing trends in biodiversity: a case study from Chiapas, Mexico." Biodiversity and Conservation 8: 869-889. Huete, A., K. Didan, T. Miura, E.P. Rodriguez, X. Gao, and L.G. Ferreira (2002) Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sensing of Environment 83: 195-213. Homewood, K., and W.A. Rodgers (1989) Pastoralism, conservation and the overgrazing controversy. In Conservation in Africa: people, policies, and practice , by D. Anderson and R. Grove, 111-128. Cambridge, UK: Cambridge University Press. Huntington, H., T. Callaghan, S. Fox, and I. Krupnik (2004) Matching Traditional and Scientific Observations to Detect Environmental Change: A Discussion on Arctic Terrestrial Ecosystems. JSTOR 18-23. http://www.jstor.org/stable/25094583. Jost, L. (2006) Entropy and diversity. Oikos 113: 363-375. Kendrick, A., and M. Manseau (2008) Representing traditional knowledge: Resource management and inuit knowledge of barren-ground caribou. Society and Natural Resources 21: 404– 418. Klein, J.A., K.A. Hopping, E.T. Yeh, Y. Nyima, R.B. Boone, and K.A. Galvin (2014) Unexpected climate impacts on the Tibetan Plateau: Local and scientific findings of delayed summer. Global Environmental Change 28: 141-152. Kong, T.M., S.E. Marsh, A.F. van Rooyen, K Kellner, and B.J. Orr (2015) Assessing rangeland condition in the Kalahari Duneveld through local ecological knowledge of livestock

farmers and remotely sensed data. Journal of Arid Environments (The University of Arizona) 113: 77-86. http://hdl.handle.net/10150/268615. Lauera, M., and S. Aswani (2008) Integrating indigenous ecological knowledge and multi-spectral image classification for marine habitat mapping in Oceania. Ocean & Coastal Management 51 (6): 495–504. Liu, Y.S., Y.C. Hu, and L.Y. Peng (2005) Accurate Quantification of Grassland Cover Density in an Alpine Meadow Soil Based on Remote Sensing and GPS. Pedosphere 15 (6): 778-783. Maynard, N.G., P. Burgess, A. Oskal, J.M. Turi, S.D. Mathiesen, I.G.E. Gaup, B. Yurchak, V. Etylin, and J. Gebelein. n.d. Eurasian Reindeer Pastoralism in a Changing Climate: Indigenous Knowledge & NASA Remote Sensing. Accessed January 15, 2016. http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/20080041555.pdf. Mills, D., R. Blech, B. Gillam, M. Martin, G. Fithardinge, J. Davies, S. Campbell, and L. Woodhams (2002) Rangelands: people, perceptions and perspectives. In Global rangelands: progress and prospects, by A.C. Grice and K.C. Hodgkinson, 43–54. Wallingford, Oxfordshire, UK: CAB International. Ministry of Food and Agriculture (2015) National Report on the Rangeland Health of Mongolia. Ulaanbaatar, Mongolia Naidoo, R., and K. Hill (2006) Emergence of Indigenous Vegetation Classifications Through Integration of Traditional Ecological Knowledge and Remote Sensing Analyses. Environmental Management 38 (3): 377-387. Niamir-Fuller, M. (1995) Indigenous systems of natural resource management among pastoralists of arid and semi-arid Africa. In The cultural dimension of development, by D.M. Warren and L.J. Slikkerveer, 245-257. London, UK: Intermediate Technology Publications. Oba, G. (2012) Harnessing pastoralists’ indigenous knowledge for rangeland management: Three African case studies. Pastoralism: Research, Policy and Practice 2 (1). http://pastoralismjournal.springeropen.com/articles/10.1186/2041-7136-2-1. Oba, G., and D.G. Kotile (2001) Assessment of landscape level degradation in southern Ethiopia: pastoralists versus ecologists. Land Degradation & Development 12: 461–475. Olson, J.S., J.A. Watts, and L. J. Allison (1983) Carbon in Live Vegetation of Major World Ecosystems. Report ORNL-5862, Tennessee: Oak Ridge National Laboratory. Paltsyn, M.Y., S.V. Spitsyn, A.N. Kuksin, and S.V. Istomov (2012) Conservation of snow leopard in Russia. Krasnoyarsk, Russia: WWF Russia. http://wwf.ru/resources/publ/book/eng/599. Paltsyn, M.Yu., J.P. Gibbs, L.V. Iegorova, G. Mountrakis (2017) Estimation and Prediction of Grassland Cover in Western Mongolia Using MODIS-Derived Vegetation Indices. Rangeland Ecology & Management. Rangeland Ecology & Management 70 (6): 723-729 Pitt, A.L., R.F. Baldwin, D.J. Lipscomb, B.L. Brown, J.E. Hawley, C.M. Allard-Keese, and P.B. Leonard (2012) The missing wetlands: using local ecological knowledge to find cryptic ecosystems. Biodiversity and Conservation 21 (1): 51-63. Polfus, J.L., K. Heinemeyer, and M. Hebblemewhite (2014) Comparing traditional ecological knowledge and western science woodland caribou habitat models. The Journal of Wildlife Management 78 (1): 112-121. Purevdorj, TS., R. Tateishi, T. Ishiyama, and Y. Honda (1998) Relationships between percent vegetation cover and vegetation indices. International Journal of Remote Sensing 19 (18): 3519-3535. doi:10.1080/014311698213795.

Roba, H.G., and O. Gufu (2009) Efficacy of integrating herder knowledge and ecological methods for monitoring rangelands degradation in Northern Kenya. Human Ecology 37: 589-612. Robbins, P. (2003) Beyond Ground Truth: GIS and the Environmental Knowledge of Herders, Professional Foresters, and Other Traditional Communities. Human Ecology 31 (2): 233- 253. Seaniger, C-A. (2016) Dzud may affect up to 150,000 herders. UB Post. December 23 2016. http://theubpost.mn/2016/12/23/dzud-may-affect-up-to-150000-herders/ Shung, Z.H., Gibb, M.J., Long, R.J. (2012) Effect of snow disasters on livestock farming in some rangeland regions of China and mitigation strategies – a review. The Rangeland Journal, 34: 89-100. Selemani, I.S., L.O. Eik, O. Holand, T. Ådnøy, E. Mtengeti, and D. Mushi. (2012) The role of indigenous knowledge and perceptions of pastoral communities on traditional grazing management in north-western Tanzania. African Journal of Agricultural Research 7 (40): 5537-5547. Spooner, B. (1973) The cultural ecology of pastoral nomads. Addison-Wesley Module in Anthropology No. 45 ( Addison-Wesley Publishing Company, Reading, Massachusetts, USA. ). Sulieman, H.M., and A.G.M. Ahmed (2013) Monitoring changes in pastoral resources in eastern Sudan: A synthesis of remote sensing and local knowledge. Pastoralism: Research, Policy and Practice 3 (22). doi:10.1186/2041-7136-3-22. UNESCO (1992-2016) UNESCO World Heritage Convention: World Heritage List . Accessed June 16, 2016. http://whc.unesco.org/en/list. White, R.P., S. Murray, and M. Rohweder (2000) Pilot Analysis of Global Ecosystems: Grassland Ecosystems. Technical Report, Washington DC: World Resources Institute. WWF (2012) Altai-Sayan Ecoregional Conservation Strategy. WWF.

CHAPTER 3: FORECASTING OF SNOW LEOPARD (PANTHERA UNCIA) AND ALTAI ARGALI (OVIS AMMON AMMON) HABITAT DYNAMICS IN THE ALTAI-SAYAN ECOREGION UNDER A CHANGING CLIMATE

Abstract: Snow leopard (Panthera uncia) and argali (Ovis ammon) are examples of species facing considerable risks associated with climate change due to their high dependence on alpine ecosystems. Herein we project snow leopard and Altai argali (Ovis ammon ammon) distribution in the Altai-Sayan Ecoregion, Central Asia, a global priority region for biodiversity conservation and the northernmost population segments of both snow leopard and argali. Our models predict that 39-90% of snow leopard and 69-100% of Altai argali habitat in the ecoregion may be lost by 2060-2080 as a result of predicted expansion of the forest in the alpine zone of the mountains in the north of the species range and increasing aridity of these species’ habitats in the southern part of the ecoregion. As climate-smart measures for conservation of the species in the Altai-Sayan Ecoregion for snow leopards we recommend 1) maintaining key species sub-populations located mainly in Mongolia and transboundary zone of Mongolia and Russia that are likely to shrink but still persist and 2) focus more conservation efforts on the snow leopard sub-population in Sengelen Ridge that is likely to sustain significant habitat area and even extend its range in the context of climate change if current population declines due to poaching are addressed. For argali populations, current Altai argali range in the future is more likely to be more suitable for Gobi argali due to climate-induced conversion of alpine ecosystems to dry steppe and semi- desert biomes; therefore, maintaining Gobi argali sub-populations in the Gobi Altai along with habitat connectivity between Gobi Altai and southern part of Mongolian Altai should be a priority for facilitating this Gobi argali colonization.

Introduction

One of the major current concerns in biodiversity conservation is global climate change, which can directly and indirectly affect species’ habitat and patterns of habitat use and, ultimately, affect survival, abundance, and geographical distribution (Pounds et al. 2006; Forrest et al. 2012; Maharaj & New 2013; Fun et al., 2014). Average global temperatures have increased by 0.74º C over the last hundred years and are projected to increase by 2.0–8.5ºC by 2070–2099 (IPCC 2007, 2014). Climate change has already begun to affect species and ecosystems, for example by shifting some species’ ranges pole-ward and up-slope, affecting seasonal activities, migration patterns, abundances and species interactions, and advancing of timing of phenological events (Parmesan & Yohe 2003; Hickling et al. 2005; IPCC 2014). On the global scale up to 30% of plant and animal species may be at the risk of extinction as the result of climate change impact (IPCC 2007, 2014). Understanding potential climate change impacts on species distribution and abundance is critical for development of appropriate strategies for biodiversity conservation (Williams et al. 2005; Mawdsley et al. 2009; Gillson et al. 2013; Nakao et al. 2013).

Snow leopard (Panthera uncia) and argali (Ovis ammon) are examples of species facing multiple threats from human-made environmental changes, including climate change. The range of snow leopards (listed as Vulnerable, IUCN 2017) considerably coincides with alpine zone between glaciers and forests in the mountains of the Central Asia (Li et al. 2016). The distribution of Altai argali (Ovis ammon ammon) (listed as Near Threatened, IUCN 2017) overlaps with that of snow leopards but is mainly limited to alpine grassy tundra and mountain steppe landscapes in the Altai Mountains in the elevation range 2,000 – 3,700 m above sea level (Fedosenko 2000). Strong dependence on the alpine habitat at the top of the mountains makes both species vulnerable to global warming and projected shrinking of the alpine zone (IPCC 2014).

Snow leopard and argali are used as flagship and umbrella species for biodiversity conservation throughout Central Asia, including Altai and Sayan Mountains (UNDP/GEF 2005, 2006; WWF

2012; UNEP/CMS 2014; Snow Leopard Working Secretariat 2013). The Altai-Sayan Ecoregion is an area of particular concern for both species. This large mountainous area extends over 1,000,000 km² and is located at the junction of Russia, Mongolia, China and Kazakhstan in the most northern extent of snow leopard and argali populations (Fedosenko 2000; McCarthy and Mallon 2016). The Altai-Sayan snow leopard population is 650-950 individuals (Baidavletov, 1999; McCarthy, 2000; Paltsyn et al., 2012; WWF 2012) comprising 14-23% of the total worldwide snow leopard population, which is estimated to be 3,920-6,590 animals (McCarthy et al. 2003; Snow Leopard Working Secretariat 2013; McCarthy and Mallon 2016). Current Altai argali distribution is limited to Altai-Sayan Ecoregion only with total population estimates of 4,200-6,700 (Reading et al. 1997, 1999; Baidavletov 1999; Fedosenko 2000; Amgalanbaatar and Reading 2000; Harris et al. 2009; Paltsyn et al., 2011; WWF 2012; Spitsyn 2014; Munkhtogtogh 2014).

Snow leopard and Altai argali in the Altai-Sayan Ecoregion are threatened by various anthropogenic pressures, whose combined effects will likely intensify due to climate change (Batima 2006; Kokorin et al. 2011). The threats include mining, linear infrastructure development (roads, rail roads, and gas pipelines), overgrazing, and wildlife poaching (Paltsyn et al., 2011, 2012; WWF 2012). A climate change assessment for the Russian and Mongolian part of the Altai- Sayan Ecoregion (Kokorin et al. 2011; Batima, Batnasan & Lehner 2004; Dagvadorjet al. 2009) indicates that warming is already taking place: the average rate of warming in 1940-2008 in the Russian part of the Altai-Sayan Ecoregion was 1.85° C (Kokorin et al. 2011). Across Mongolia an increase in annual mean air temperature of 1.80° C during 1940-2003 has been measured (Batima et al. 2006). Annual average precipitation increased by 42 mm in the Russian part of Altai-Sayan in 1966-2008 (Kokorin et al. 2011), and by 20-36 mm in Mongolian part in 1961-2008 (Batima et al. 2006; Dagvadorj et al. 2009). Similarly mean annual air temperature increased by 1.24° C and annual precipitation increased by 29.45 mm from 1962 to 2011 in Xinjiang province of China (includes south-western part of Altai-Sayan Ecoregion) (Fang et al. 2013). Additionally, forecasts for Altai-Sayan predict an increase of the summer air temperatures by 3-8° C and winter temperatures by 2-5° C, and annual precipitation - by 60-100 mm by 2080-2090 (Kokorin et al. 2011). For western Mongolia similar increase in air temperature is projected, but precipitation is predicted to decrease by 10% by 2090 (Batima et al. 2006; Dagvadorj et al. 2009).

The climate change impacts could directly affect the alpine habitats upon which snow leopard and Altai argali depend. More specifically, positive correlations have been reported between precipitation and Leaf Area Index in Xinjiang province of China in 1981-2011 with the index apparently increasing over time in Chinese segment of Altai-Sayan Ecoregion (Fang et al. 2013). Recent studies in Mongolia indicate that net primary production is likely to decline in most areas in Mongolia in the nearest 50-80 years, except high mountainous regions like Altai-Sayan and Khangay, there NPP is likely to increase by 25-75% (Batima et al. 2006; Angerer et al. 2008). Thus, further increase in precipitation and temperature as predicted for Altai-Sayan (Kokorin et al. 2011) may well produce increased vegetation production in the mountain steppe and tundra and projected expansion of forests and shrub in the current alpine grasslands (Tchebakova et al. 2009), that is, the habitat of snow leopard and Altai argali. Modeling of forest shift due to climate change in the Himalaya provides similar predictions: about 30% of snow leopard habitat there may be lost due to an upward shifting treeline and consequent shrinking of the alpine zone by the end of 21st century (Forrest et al. 2012). At the same time a general increase in precipitation in the ecoregion

40 is projected for winter period in the Altai region (Batima et al. 2006; Dagvadorj et al. 2009; Kokorin et al. 2011), potentially generating deeper snow cover, especially in the northern and north-western parts of snow leopard and Altai argali range in Altai-Sayan, limiting winter habitats for argali and other wild ungulates and possibly increasing wildlife-herder conflicts. In some areas of the ecoregion, the effect of livestock grazing on wildlife could be magnified by degradation of mountain pastures due to intensive mining development (WWF 2012). At lower elevations, particularly in the more arid southern zones of the ecoregion, projected expansion of the desert and semi-desert (Tchebakova et al. 2009) is likely to further constrain snow leopard and Altai argali habitat.

Projections of availability of suitable habitat for sentinel species in the context of climate change can provide helpful guidance for long-term conservation planning at regional level (Araújo et al. 2004; Hole et al. 2009; Luo et al. 2015). Existing conservation strategies for snow leopard and argali only consider climate change as one of the many threats for these species and declare importance of addressing climate change but they do not take in account projected changes in the species habitat at different climate change scenarios (Paltsyn et al., 2011; Paltsyn et al., 2015; WWF 2012; CMS 2013; Snow Leopard Working Secretariat 2013). Herein we built snow leopard and Altai argali distribution models for the Altai-Sayan Ecoregion based on current species occurrence records and a set of bioclimatic variables to project likely changes in distribution of these species’ by 2060-2080 in the context of predicted climate change. Our main hypothesis was that snow leopard and Altai argali habitat in the ecoregion will shrink and move upslope as a result of projected expansion of forest boundary and increasing depth of snow cover on the north and expansion of deserts and semidesert zone on the south in the species habitat (Batima et al. 2006; Angerer et al. 2008; Dagvadorj et al. 2009; Tchebakova et al. 2009; Kokorin et al. 2011). The goal of this research is to provide conservation managers and planners with insight on possible changes in the species habitats to recommend appropriate “climate-smart” (Gillson et al. 2013) conservation approaches for these species in the Altai-Sayan, a globally significant ecoregion.

Materials and Methods

Study Area

Altai-Sayan Ecoregion (Fig. 1) is one of the 200 priority areas delineated by World Wide Fund for Nature (WWF) for conservation of the Earth’s biodiversity (Olson & Dinerstein, 2002). Wide elevation range (200 - 4580 m above sea level) and climate diversity of this mountain region define its high landscape variability (Kupriyanov et al. 2003). The Ecoregion represents a mix of different landscapes such as conifer forests (taiga) and forest-steppe occupying (39%), steppe (24%), alpine meadows and mountain tundra (including bare rocks and glaciers, 17%), semi-deserts and deserts (6%), and wetlands and lakes (4%). High landscape diversity is the basis for the region’s rich species diversity encompassing about 10,000 of known species of plants, animals and fungi, including 3,500 of vascular plant and 680 vertebrate animal species (Kupriyanov et al. 2003; UNDP/GEF 2007). Global importance of the Altai-Sayan Ecoregion was underlined by two nominations of the UNESCO World Nature Heritage sites – Golden Mountains of Altai and Uvs Nuur Basin (UNESCO World Heritage Convention, 2013). Conservation of this area offers an extraordinary opportunity to preserve and manage sustainably over 1,000,000 km² of practically pristine transboundary landscape in the area of the junction of Russia, Mongolia, China and

Kazakhstan (WWF 2012). The total area of actual and potential habitats of snow leopard and Altai argali in the Ecoregion is 140,000 – 150,000 km, or 14-15% of the entire area (Paltsyn et al. 2011, 2012; WWF 2012).

Study approach

We built snow leopard and Altai argali distribution models using a maximum entropy approach (Maxent) (Phillips et al. 2004; Phillipset al. 2006) for the Altai-Sayan Ecoregion based on current species occurrence records and a set of bioclimatic variables to project likely changes in species distribution by 2060-2080 in the context of climate change and recommend appropriate climate- smart conservation approaches for the species in the area. Maxent is machine-learning algorithm based on maximum entropy principle that uses presence-only occurrence data and environmental variables to predict environmental suitability (niche) for a particular species (Phillips et al. 2006). Maxent uses the occurrence records as training and test datasets to predict environmental suitability for a species based on an assessment among different combinations of environmental variables. By maximizing the entropy of the probability distribution, it matches the expected values of environmental variables under the estimated distribution to their empirical averages (Phillips et al. 2006).

Training and test data: Snow leopard and Altai argali occurrence records

The majority (~80%) of the species occurrence points for Altai argali and snow leopard were collected from the transboundary zone of Russia and Mongolia where regular species surveys have been organized since 1998. For snow leopards, we used 2,886 occurrence records gathered within the Altai-Sayan Ecoregion collected in 1998–2016 by WWF Russia and Mongolia, Severtsov’s Institute of Ecology and Evolution of the Russian Academy of Science, Biology Institute of Mongolian Academy of Science, Katon-Karagai National Park in Kazakhstan, and the Snow Leopard Network during sign and camera-trapping surveys. For Altai argali, a total of 2,985 occurrence records from 1996-2014 were assembled from annual argali surveys by WWF Russia and Mongolia (WWF, unpublished); Altaisky State Nature Reserve archives (Nature Chronicles 993-2003); nationwide ungulate survey 2009 report (Harris, Wingard and Lhagvasuren 2009); location records in Uvs Aimag, collected in 2005-2009 from radio-collared argali (Chimeddorj et al. 2013); location records in Khukh-Serkh National Park, collected from radio-collared argali by Dr. B. Rosenbaum (Rosenbaum, unpublished data); and Mongolian Red List of Mammals (Clark et al. 2006) (See Supplementary Materials, Fig 1) .

Training and test data: Forest and desert cover data

As training and test data for projections of the forest cover we used percentage of the tree cover dataset MODIS Vegetation Continuous Fields (VCF) for the region in 2016 where all pixels with >=20% of tree cover classified as “forest” and converted in the random point layer (n = 3,000) in ArcGIS Pro (Create Random Points function, ESRI 2017). The 20% tree cover threshold was selected based on the modeling of Altai argali and snow leopard habitat in Russia and adjacent Mongolia with probability of occurrence of both species negligible in the areas with higher percentage of tree cover (Paltsyn, unpublished data). Similarly, for projections of deserts we extracted all desert landscapes from digital landscape map for the Altai-Sayan Ecoregion

(Samoilova 2001) and converted them to random point layer (n = 400) as training data for model construction.

Environmental variables

For snow leopard and Altai argali, we assembled environmental variables known to be important drivers of snow leopard and Altai argali habitat: ruggedness, slope, aspect (Aryal et al. 2016; Li et al. 2016; Fedosenko 2000; Paltsyn et al. 2011), bio10 (mean temperature of the warmest quarter), bio11 (mean temperature of the coldest quarter), bio12 (annual precipitation) (Li et al. 2016; Luo et al. 2015; Ye et al. 2018), and bio19 (precipitation of the coldest quarter, as a proxy of snow cover depth, critical for both – argali and snow leopard) (Fedosenko 2000; Paltsyn et al., 2012; Fox, Raghunandan, and Chundavat 2016). For projections of the forest cover, we focused on aspect, bio10, bio11, bio12, and bio19 variables, because these variables have been identified as important drivers of forest distribution in the region (Tchebakova et al. 2009; Kharuk et al. 2010; Kokorin et al. 2011; Forrest et al. 2012; Körner 2012). Similarly, for deserts we focused on bio10, bio11, bio12, and bio19 variables (Tchebakova et al. 2009; Kokorin et al. 2011).

In terms of deriving topographical variables, ruggedness (3x3 pixels), slope, and aspect variables were developed from SRTM data (Shuttle Radar Topography Mission, 30×30 m) resampled to 1 km resolution in ArcGIS Pro (ESRI 2017). Aspect data were reclassified in 4 classes: north (0-45º and 315-360º), east (>45º and < 135º), south (>135º and < 225º), and west (>225 and <315º). Bio10, bio11, bio12, and bio19 30-second resolution data for current period and 2070 (2060-2080) projections were downloaded from the WordClim web-site http://worldclim.org and resampled to 1 km resolution (Table 1).

For projections of the species occurrence and likely extent of forest and desert ecosystem cover in 2070 we used outputs from three global circulation models (GCM): CCSM4, HadGEM2-ES, and MIROC-ESM for four Representative Concentration Pathways (RCP) scenarios RCP2.6, RCP4.5, RCP6.0 and RCP8.5 adopted by the Inter-Governmental Panel on Climate Change (IPCC), representing low-high greenhouse gas emission rates (Stocker et al., 2013; IPCC 2007, 2014). We did not use elevation in our projections because it does not have direct impact on the species and forest cover, but rather affects them indirectly via temperature (Aryal et al., 2016). Including such conflated variables may negatively impact predictive power of species distribution models (Hof et al., 2012).

Table 2. Topographic and climate variables used for distribution projections for snow leopard, Altai argali, forest and desert/semidesert cover in the Altai-Sayan Ecoregion under future climate scenarios

Variable Type Units Source

Ruggedness 3x3 Continuous Terrain Derived from CGIAR-CSI Shuttle Ruggedness Index Radar Topography Mission digital elevation data (Jarvis et al. 2008): Slope Continuous Degrees http://www.cgiar-csi.org/data/srtm- Aspect Categorical N, E, S, W 90m-digital-elevation-database-v4-1

Mean temperature of the Continuous º C warmest quarter (Bio10), current and 2060-2080

Mean temperature of Continuous º C coldest quarter (Bio11), current and 2060-2080

Annual precipitation Continuous mm (Bio12), current and 2060-2080 WordClim 1.4 database (Hijmans et al. 2005): http://worldclim.org Precipitation of the Continuous mm coldest quarter (Bio19), current and 2060-2080

Species Distribution Modeling Operations

Modeling of current and projected (2060-2080) snow leopard and Altai argali habitat in the Altai- Sayan Ecoregion was performed in the Maxent software (Version 3.3.3k), program for maximum entropy modelling of species’ geographic distributions (Phillipset al. 2004; Phillips et al. 2006). We selected Maxent to build the species distribution projections under a changing climate because it is one of the most popular, robust, effective, and easy to use modeling methods that uses presence only data and ranked among the best modeling tools based on its predictive performance, including bioclimatic modeling (Elith et al. 2006, 2011; Merow et al. 2013; Shabani et al. 2016). Following Li et al. 2016 we used all default settings for MaxEnt and only selected hinge feature for modeling. Given ~80% of all occurrence locations for snow leopard and Altai argali were concentrated in transboundary zone of Russia and Mongolia we performed correction of the sampling bias by selecting for modeling only random species locations with a distance > 4 km between the locations for snow leopard and > 10 km – for Altai argali (so called “spatial filtering,” Kramer-Schadt et al. 2013). After spatial filtering a total of 361 locations were available for snow leopard and 85 for Altai argali as training and test data. To evaluate model performance, we used cross-validation with 5 iterations as well as area under the ROC curve. To extract species “habitat” and forest and desert/semidesert cover from the averaged models we used an equal training sensitivity and specificity logistic threshold (Liu et al. 2013; Li et al. 2016).

Results Models of current snow leopard and Altai argali habitat distributions Species distribution models performed well for both snow leopard (AUC=0.896±0.013) and Altai argali (0.942±0.004) (Fig. 1). Ruggedness was the most important contributor to the snow leopard distribution model (33.2%), followed by annual precipitation (bio12) (20.2%) and total precipitation of the coldest quarter (bio19) (19.4%). The Altai argali distribution was most influenced (35.1%) by annual precipitation (bio12), mean temperature of the warmest quarter (Bio10) (30.6%) and mean temperature of the coldest quarter (bio11) (18.9%). Estimated area of

44 current potential habitat in the Ecoregion was 182,002 km² for snow leopard and 104,222 km² for argali (Fig. 1). Habitat distribution models for forest cover had satisfactory prediction ability with AUC = 0.622±0.004 and performed well for deserts with AUC = 0.955±0.004. Annual precipitation (Bio12) made the highest contribution for determining distribution of these ecosystem types (75.2% for forest cover and 83.6% for deserts) followed by mean temperature of the warmest quarter (18.3% for forest and 14.2% for deserts). Snow leopard Altai argali

a. Current b. Current

c. CCSM4 RCP 2.6 (2060-2080) d. CCSM4 RCP 2.6 (2060-2080)

e. CCSM4 RCP 8.5 (2060-2080) f. CCSM4 RCP 8.5 (2060-2080) Figure 1. Predicted snow leopard and Altai argali habitat (yellow), forest cover (green) and desert cover (pink) models for current conditions (a and b) and RCP 2.6 and 8.5 scenarios (c-f) for CCSM4 GCM in the Altai-Sayan Ecoregion.

Projected changes in the snow leopard and Altai argali habitat in 2060-2080 For snow leopards, mean temperature of the warmest quarter in their current range is projected to increase by 44-55% (from 11.2 up to 16.1-17.3 ºC) for RCP 8.5 scenario (Supplementary Materials, Fig. 5a). Mean temperature of the coldest quarter is likely to increase by 18-36% (from -22.5 up to -18.4 to -17.3 ºC) for RCP 8.5 scenarios (Supplementary Materials, Fig. 5c). Total annual precipitation is projected to increase by 3-13% (from 257 up to 266-291 mm) (Supplementary Materials, Fig. 5e) with significant increase of the coldest quarter precipitation by 32-48% (from 13.9 up to 18.3-20.6 mm) (Appendix 1, Fig. 5g). Similar changes are predicted for Altai argali habitat: mean temperature of the warmest quarter will increase by 46-56% (from 10.8 to 15.8-16.8 ºC), mean temperature of the coldest quarter – by 21-41% (from -20.0 to -15.9 – -11.8 ºC), total annual precipitation – by 2-10% (from 197 to 201- 2017 mm), and cold quarter precipitation – by 31-46% (from 10.1 to 13.2-14.7 mm) for RCP 8.5 scenario (Appendix 1, Fig. 5 b, d, f, h). These climate changes are predicted to generate a substantial increase in the area suitable for forest, which was projected to increase by 82-195% in the current snow leopard habitat and by 13-370% in the current Altai argali habitat (Fig. 2 a and b). Area of deserts is projected to increase also in the southern part of the ecoregion by 56-123% in the current snow leopard habitat and by 2,829- 18,422% in the Altai argali habitat (Fig. 2 c and d). In relationship to elevation, the upper forest boundary in the snow leopard and argali habitat is projected to increase from current maximum of 2,995 m to 3,418 – 4,106 m and desert upper boundary – form current maximum of 2,299 to 2,344- 2,681 m above sea level by 2060-2080 (Appendix 1, Fig. 6). No substantial changes in snow leopard distribution in relation to elevation is predicted by 2060-2080 whereas a substantial elevation shift is projected for Altai argali habitat from the current mean value of 2,543 m up to 2,778-3,876 above sea level (Appendix 1, Fig. 6).

In terms of snow leopard habitat extent, total area of snow leopard suitable range is projected to decline by 39-90% and Altai argali by 69-100% by 2060-2080 (Fig. 2 c and d, and Fig. 3). Number of habitat patches may slightly increase (3-8%) for snow leopard and Altai argali at CCSM4 RCP 2.6 scenario but is projected to decline by 8-53% for snow leopard and by 14-100% for Altai argali under other climate change scenarios (Fig. 2 e and f). Average size of a habitat patch is projected to decline for both species at all climate change scenarios by 41-79% for snow leopard and 72- 100% for Altai argali (Fig. 2 g and h). Notably, for snow leopards despite an overall projected decline, 1,751- 24,331 km² of new habitat are predicted to emerge in the eastern part of the Altai- Sayan Ecoregion on Sangilen Ridge, Western Hovsgol Mountains and Khan Khukhiyn Ridge (Fig. 3). No new habitat is predicted to emerge for argali.

Snow leopard Altai argali ² 70,000 ² 11,000

60,000 9,000

50,000 7,000

40,000 5,000

30,000 3,000 Area km forest, covered Area by 20,000 km forest, covered Area by 1,000 Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Climate change scenarios Climate change scenarios

a. Forest cover b. Forest cover ² ² 30000 12000 10000 25000 8000

20000 6000 4000 15000 2000

10000 desert km covered Area by 0 Area covered by desert, km desert, by covered Area Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Climate change scenarios Climate change scenarios c. Desert cover d. Desert cover

100,000

170,000

² ² 80,000 130,000 60,000 90,000 40,000

50,000 20,000

Area km of Area habitat, Area km of Area habitat,

10,000 0 Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Climate change scenarios Climate change scenarios

e. Habitat area f. Habitat area 3,500 400 3,000 300 2,500 200 2,000

1,500 100

0 Number of of patcheshabitatNumber Number of of patcheshabitatNumber 1,000 Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Climate change scenarios Climate change scenarios g. Number of habitat patches h. Number of habitat patches 60 250 50 200 40 150

30 100

20 50

Area km² of patch,habitatArea a 10 0 Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 km² of patch,habitatArea a Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Climate change scenarios Climate change scenarios i. Average size of a habitat patch j. Average size of a habitat patch Figure 2. Projected changes of area covered by forest in current habitat, area of habitat, number and average size of habitat patches for snow leopard (a, c, e, and g) and Altai argali (b, d, f, and h) in the Altai-Sayan Ecoregion in 2060-2080 predicted by CCSM4 (green), HadGEM2-ES (blue), and MIROC-ESM (red) GCMs under RCP 2.6-8.5 scenarios.

Snow leopard Altai argali

120000 Loss 120000 Gain Loss 90000 90000 No changes No changes 60000 60000

30000

Area km² of Area habitat, 30000 Area km² of Area habitat,

0 0 RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Climate change scenarios Climate change scenarios

a. Habitat area by CCSM4 GCM b. Habitat area by CCSM4 GCM

c. RCP 2.6 d. RCP 2.6

e. RCP 8.5 f. RCP 8.5 Figure 3. Projected changes in snow leopard and Altai habitat in the Altai-Sayan Ecoregion in 2060-2080 predicted by CCSM4 GCM under RCP 2.6 and 8.5 scenarios (yellow – no changes, red – habitat loss, green – habitat gain).

Discussion Our projections suggest a likelihood of strong contraction of snow leopard habitat and severe reduction of Altai argali range in the Altai-Sayan Ecoregion by the end of 21st century associated with changing climate: the snow leopard may lose 39-90% of its current range whereas Altai argali 69-100%. Species ranges are projected to contract to their current range’s centers from all sides albeit more from north and south for argali, which manifested a clear upslope contraction associated with an increase of mean annual temperature and precipitation. Number of habitat patches is projected to decline by 8-53% for snow leopard and 14-100% for argali and their average size by 41-79% for snow leopard and 72-100% for argali implying smaller subpopulation sizes and habitat connectivity for both species. These predictions are comparable to projections of Ye at al. (2018) for Chinese part of the Altai- Sayan Ecoregion who forecasted up to 50% of snow leopard and 75% of Altai argali habitat loss by 2050. In contrast, Li et al. (2016) predicted an increase of snow leopard habitat area in Russia and Mongolia (located mainly in the Altai-Sayan Ecoregion) by 21-67% and 8-29% respectively by 2060-2080 with significant shift northward and upslope whereas we project decline of snow leopard habitat both in Russia and Mongolia. Li et al. (2016) also projected new snow leopard habitat emerging mainly in the Eastern Sayan Mountains, Western Hovsgol Mountains, Sengelen Ridge and Khan-Khuhiyn Ridge in the boundaries of the Altai-Sayan Ecoregion (Li et al. 2016). Our projections of snow leopard habitat are in agreement with projections by Li et al. (2016) only for the last three mountain regions at the border of Russia and Mongolia with overall area of new habitat of 1,751- 24,331 km² but predict 80-90% decline of the area of existing snow leopard habitat in the Eastern Sayan Mountains located further on the north in Russia (Fig. 3). We contend that emergence of new snow leopard habitat in the Eastern Sayan Mountains in Russia is unlikely due to projected shift of the upper forest boundary up to the top of the mountains (Tchebakova et al. 2009; Kokorin et al. 2011). Similar loss of snow leopard habitat due to shifting treeline and consequent shrinking of the alpine zone by the end of the 21st century was projected by Forrest et al. (2012) and Li et al. (2016) for Himalaya. Current snow leopard and Altai argali distribution models were consistent with expert knowledge and natural history information on the species distribution in the Altai-Sayan Ecoregion (Fedosenko 2000; Maroney and Paltsyn 2003; Harris, Wingard and Lhagvasuren 2009; Paltsyn et al. 2011, 2012; Munkhtsog et al., 2016). Precipitation variables exhibited higher percentage contribution toward current snow leopard distribution than temperature variables. The importance of precipitation for snow leopard distribution can be explained by wide elevation range of snow leopard habitat (540 – 4,200 m above sea level) in the Altai-Sayan Ecoregion (McCarthy and Chapron 2003; Paltsyn et al. 2015; Munkhtsog et al. 2016) and wide variety of habitats – from mountain steppe and forest-steppe to alpine zone with tundra, alpine meadow, and bare rock biomes (Paltsyn et al. 2015; Munkhtsog et al. 2016). However, snow leopards avoid relatively moist areas with percentage of forest cover more than 20-50% and deep snow in winter (Paltsyn et al. 2015). Current Altai argali distribution was explained nearly equally by temperature variables and precipitation variables. In contrast to snow leopards, Altai argali occupy a much narrower range

50 of habitat in the Altai-Sayan Ecoregion, that is, mainly alpine grassy tundra and alpine steppe (or tundra-steppe) undulating landscapes in the elevation range of 2,000 – 3,900 m above sea level (Fedosenko 2000; Abaturov et al. 2004; Paltsyn et al. 2011). Also, Altai argali are very sensitive to the depth of snow cover and cannot graze with a snow cover greater than 0.3-0.4 m (Sopin 1975; Fedosenko 2000; Paltsyn 2011). Altai argali are well adapted to low temperatures, however, long periods of extreme cold (below -30ºC) as well summer temperature higher than 23-24 ºC can negatively affect animal’s metabolism and survival (Fedosenko 2000; Abaturov et al. 2004). Altai argali avoid forests and shrubs as well as dry steppe and semi-desert landscapes located in the lower parts of mountains and valleys (Tsalkin1951; Fedosenko 2000; Abaturov et al. 2004). In terms of the key changes in the current snow leopard and Altai argali habitat that may lead to significant contraction of the species habitats, mean temperature in the current snow leopard and Altai argali habitat is projected to increase by 2.1-6.1 ºC (+19-56%) in summer and by 1.6-8.0 ºC (+7-41%) in winter in average. At the same time annual precipitation is likely to show a smaller increase: by 6-33 mm (+2-14%) including 1.5-7.0 mm in winter (+11-48%). However, climate change patterns are predicted to be different in the northern and southern parts of these species’ ranges. In the northern part of the current snow leopard habitat (Russia) the mean temperature of the warmest season is projected to increase from 9.9 to 14.6 ºC and annual precipitation – from 365 to 380 mm at the RCP 8.5 scenario in 2060-2080. Our models of forest cover predict that major loss of the snow leopard habitat in the north may occur as a result of upper forest boundary shift higher in the alpine zone by 500-1,000 m (Fig. 1). This projection is also confirmed by identification of the areas in the Altai-Sayan Ecoregion with current mean temperature of the warmest quarter 14.6 ºC (±1SD) and annual precipitation of 380 mm (±1SD) that are projected in the northern part of the snow leopard habitat in 2060-2080: all of them are 80-90% Siberian pine (Pinus sibirica), Siberian larch (Larix sibirica), Scot’s pine (Pinus sylvestris) or mixed forests on the ecoregional landscape map developed by Samoilova (2001). Our predictions are also consistent with climate and vegetation change models constructed by Tchebakova et al. (2009) for the Altai-Sayan Ecoregion whose models project that current tundra and alpine meadows zone in the Russian part of the Ecoregion may be completely occupied by montane conifer forest (taiga) by 2050-2080. An upper forest boundary upward shift up to 50-100 m and forest expansion in tundra and mountain steppe was recorded in South Siberia and Altai-Sayan in the last and this century (Mikhailov et al. 1992; Modina et al. 2002; Kharuk et al. 2009, 2010). The upward migration rate of the current tree line was about 0.8 m/year during the last century and about 2.3 m/year over the past three decades (Kharuk et al. 2010). However, migration rates of conifer tree species given paleoecological evidence may be as high as 400–1000 m/year (King and Herstrom 1997). Similar modeling of forest upslope shift due to climate change in the Himalaya indicates that about 30% of snow leopard habitat there may be lost as a result of forest expansion in alpine zone (Forrest et al. 2012). Another factor that can negatively affect snow leopard and Altai argali distribution in the northern part of the Altai-Sayan range is projected increase of precipitation in winter by 30- 48% (from mean 24 to 32-40 mm at the RCP 8.5 scenario) leading to increase of snow cover depth in the species habitat mainly in the Russian, Chinese, and Kazakhstan parts of the Ecoregion.

In contrast, in the southern part of current snow leopard and Altai argali habitat the mean temperature of the warmest season is projected to increase from 14.0 to 19.0 ºC and annual precipitation – from 148 to 156 mm at the RCP 8.5 scenario. Our projections predict that area of deserts in the current snow leopard habitat will increase by 56-120% and in the current Altai argali habitat - 29-185-fold! Areas with current mean temperature of the warmest quarter 19.0 ºC (±1SD) and annual precipitation of 156 mm (±1SD) in the Altai-Sayan Ecoregion that are projected in the southern part of current snow leopard and Altai argali habitat in 2060-2080 are mainly semi-deserts and deserts on the ecoregional landscape map developed by Samoilova (2001) (Fig. 8). It is notable that climate of the current desert zone in the Altai-Sayan Ecoregion in 2060-2080 is likely to be similar to the current climate of dried Aral Sea in Uzbekistan, hot mountain ranges of central Iran, and even some deserts of Yemen (mean temperature of the warmest quarter – 24.6º C and annual precipitation – 110 mm). Thus, it is likely that current snow leopard and Altai argali habitat in the southern part of the Ecoregion (mainly mountain steppes and tundra) will become much dryer and hotter turning into dry steppe and desert unsuitable for Altai argali (Tsalkin 1951; Fedosenko 2000; Abaturov et al. 2004) and marginally suitable for snow leopard (Schaller et al. 1988; Fox and Chundawat 2016). This scenario is supported by climate and vegetation change projections Tchebakova et al. (2009) for 2050-2080 predicting increase of dry steppe and desert zone in the southern portion of the Altai-Sayan Ecoregion from current 29.9% up to 48% of the ecoregional area. In aggregate, habitat availability for snow leopard and Altai argali in the Altai-Sayan Ecoregion is likely to decrease considerably as a result of pressure of increasing forest cover and snow depth on the north and expansion of dry steppe and deserts on the south as we initially hypothesized. The situation seems most critical for Altai argali (Ovis ammon ammon) that is projected to lack any suitable habitat and hence likely face extinction as an argali sub-species by 2060-2080. However, another argali sub-species – Gobi argali (Ovis ammon darwini) currently inhabiting arid mountains of Gobi Altai and isolated mountain ranges of Gobi Desert (Fedosenko 2000; Amgalanbaatar et al. 2002; Clark et al. 2006) could potentially occupy former Altai argali habitat in the Altai-Sayan Ecoregion. The climate change situation is less dramatic for snow leopard that despite habitat contraction is likely to persist in the current habitat types as well as in the arid dry steppe and deserts mountain biomes given that snow leopard are known to occur in the arid mountain ranges of Gobi and Taklamakan deserts (Shaller et al. 1988; Wang and Schaller 1996; Fox and Chundawat 2016) and recent disappearance of the species from mountain outcrops in the deserts of Inner Mongolia (Riordan and Shi 2016). Moreover, our projections predict emergence of new snow leopard habitat at the border of Russia and Mongolia in Sengelen, Western Khovsgol and Khan-Khukhiyn Mountains in the lower parts of the mountains at the average elevation of 1790 m above sea level (~435 m lower than current mean elevation of the snow leopard habitat in the Altai-Sayan Ecoregion). These areas are projected to have mean temperature of the warmest quarter equal to 14.9 ºC and total annual precipitation of 299 mm in average in 2060-2080 that indicates current climate of steppe and forest-steppe biomes in the Altai-Sayan Ecoregion (Samoilova 2001). These habitat is likely to emerge as a result of lower forest boundary shift higher in the mountains predicted by the climate and vegetation models of Tchebakova et al (2009).

The range changes predicted for snow leopards and Altai argali also suggest potential for an increase in human-wildlife conflicts in the region. Predicted reduction of area of alpine tundra and meadows may result in decreasing availability of summer pastures for herders and wild ungulates in the Altai-Sayan Ecoregion. This situation could lead to increased number of livestock in the habitats of Altai argali and snow leopard in the Altai-Sayan Ecoregion by the middle-end of 21st century and increased competition between wild ungulates and livestock for summer habitat and likely higher level of snow leopard – human conflicts as a result of increased snow leopard predation on livestock. Increased snow fall in winter is likely to result in deeper snow cover in northern and north-western parts of snow leopard and Altai argali range in the Altai-Sayan and decreased availability of winter habitats for snow leopard, argali, and other wild ungulates. As a result, snow leopard and wild ungulates will need make seasonal migrations downslope in the areas occupied by herders and livestock as winter pastures, leading to potential increase of human- wildlife conflicts again. Another important consequence of the climate change that can directly affect snow leopards and Altai argali is increasing frequency and magnitude of extreme climate events. Catastrophic environmental phenomena such as harsh winters, cold and long springs, and droughts can directly and indirectly influence snow leopard and Altai argali in the Altai-Sayan. For example, dramatic declines in Altai argali number and other mountain ungulates due to extreme weather events in Altai-Sayan were recorded by various researchers in 19th-20th centuries (Kolosov 1938; Tsalkin 1951; Sobansky 1992; Fedosenko 2000). Ungulate population declines results in the reduction in prey species for snow leopard as well. The frequency and severity of extreme weather events such as droughts and dzuds (sudden heavy snowfall, long-lasting or frequent snowfall, extremely low temperatures, or drifting windstorms that reduce or prevent animals from feeding) increased by 50% for last 60 years in Mongolia (Batima et al. 2006; Dagvadorj, Natsagdorj, Dorjpurev, & Namkhainyam 2009). The frequency of extreme weather events in Altai-Sayan is predicted to increase by 100% in the nearest 80-100 years due to climate change (Batima et al. 2006; Kokorin et al. 2011).

Given the uncertainties of species distribution modeling in the context of climate change such as we have undertaken, we do not recommend applying our models as direct and site-specific guidance for development of climate-smart conservation measures for snow leopard and Altai argali in the Altai-Sayan ecoregion. The projections provided are more appropriate for generating a general understanding of likely consequences of climate change for these species in this region in the next 60-80 years. The uncertainty makes the process of conservation planning in an era of climate change much more difficult and challenging. Complex models developed for prediction of the response of different species to climate change now incorporate different processes, including dispersal, physiology, population dynamics, competition, habitat change, and adaptation (Bomhard et al. 2005; Williams et al. 2005; Jackson & Sax 2009; Midgley et al. 2010; Huntley et al. 2010; Hampe 2011; Dawson et al. 2011; Pagel & Schurr 2012; Beale & Lennon 2012; Nakao et al. 2013). Despite complexity and sophistication of the models, accumulating cases of novel species responses to climate change demonstrate that our current predictive tools do not always integrate many important drivers of species distributions (Chen et al. 2011; Burrows et al. 2011; Schweiger et al. 2012). Also to date there is little research on what the most effective management interventions in the context of climate change are likely to be (Mawdsley et al. 2009; Gillson et al. 2013). Great majority of current conservation strategies incorporating climate change have focused

53 on accommodating species range changes by maximizing habitat connectivity and future climate space at higher latitudes and elevations (Hannah 2011; Gillson et al. 2013). However, this approach does not look adequate to deal with complexities of species responses to climate change (Rowland et al. 2011; Weeks et al. 2011; Thomas et al. 2011; Gillson et al. 2013) and probably will require input from more integrated approaches based on wide understanding of species responses, survival probabilities and range determinants (Gillson et al. 2013).

Conservation implications

The most obvious conservation recommendations for snow leopard conservation in the Altai- Sayan Ecoregion in the context of climate change may be maintaining of the key species sub- populations located mainly in Mongolia and transboundary zone of Mongolia and Russia, that are likely to shrink but still persist even at the RCP 8.5 scenario assuming habitat connectivity remains between them. More conservation efforts should be focused on snow leopard sub-population in Sengelen Ridge currently declining due to poaching (Paltsyn et al. 2015) that is likely to be located in a significant area of the snow leopard habitat that may be sustained in the face of climate change and perhaps even expand. Currently the Sengelen Ridge snow leopard sub-population receives almost no attention of on-going conservation program and projects in the Russian part of Altai- Sayan Ecoregion and may be lost within a decade.

In case of Altai argali it is difficult to suggest meaningful management response given the fact that almost all species habitat are projected to disappear in the Altai-Sayan at relatively severe climate change scenarios (e.g., RCP 6.0 and 8.5). However, key known Altai argali sub-populations in transboundary area of Russia and Mongolia, Turgen Mountains and central Mongolian Altai Ridge are likely to persist at the relatively mild climate change scenarios (RCP 2.6 and 4.5) and should be a continuing long-term conservation focus of on-going government and non-government conservation programs (Paltsyn et al. 2011, 2015; WWF 2012). Special attention should be given to conserve Gobi argali populations in the Gobi Altai and maintain connectivity between Gobi Altai and southern part of Mongolian Altai as a migration route for Gobi argali to the north in the former habitat of Altai argali that are likely to become suitable for the Gobi sub-species as a result of possible climate induced conversion of alpine ecosystems in the dry steppe and semi-desert biomes.

Acknowledgments This study was supported by the United State Agency for International Development USAID Climate Change Resilient Development Program (grant # CCRDCS0007), NASA’s Land Cover Land Use Change Program (grant # NNX15AD42G), World Wide Fund for Nature (WWF) (grants # WWF706/9Z1428(FY16-17)/1/1/GLM and WWF837/9Z1428(FY16-17)/1/1/GLM).

References Abaturov, B.D., Anchiforov, P.S., Ogureeva, G.N., Paltsyn, M.Yu., Spitsyn, S. V., Subbotin, A.Ye. 2004. Distribution of Altai Wild Sheep (Ovis ammon ammon L.) in the Altai Mountains in Russia Dependent on Character of Vegetation Cover. Zoological Journal 83(2): 241-251[in Russian]. Altaisky State Nature Reserve. 1996-1999. Nature Chronicles of Altaisky State Nature Reserve. Annual Biodiversity Monitoring Report, Yailu. Amgalanbaatar, S., and R.P. Reading. 2000. Altai argali. In Endangered Animals: Conflicting Issues, by R.P. Reading and B. Miller, 5-9. Westport: Greenwood Press. Amgalanbaatar, S., R.P. Reading, B. Lhagvasuren, and N. Batsukh. 2002. "Argali sheep (Ovis ammon) trophy hunting in Mongolia." Pirineos 157: 129–150. Angerer, J., Han G., I. Fujisaki, and K. Havstad. 2008. "Climate Change and Ecosystems of Asia With Emphasis on Inner Mongolia and Mongolia. ." Rangelands 30 (3): 46-51. Araújo, M.B., Cabeza, M., Thuiller, W., Hannah, L., and Williams, P.H. 2004. Would climate change drive species out of reserves? An assessment of existing reserve‐selection methods. Global Change Biology 10(9): 1618-1626. Aryal, A., Shrestha, U.B., Ji, W., Ale, S.B., Shrestha, S., Ingty, T., Maraseni, T., Cockfield, G and D. Raubenheimer. 2016. Predicting the distributions of predator (snow leopard) and prey (blue sheep) under climate change in the Himalaya. Ecology and Evolution 6(12): 4065– 4075. Baidavletov, R. Zh. 1999. Current status of Altai argali (Ovis ammon ammon) in Kazakhstan part of Altai. Survey Report, WWF Project «Ensuring long-term conservation of the Altay- Sayan Ecoregion». Batima, P. 2006. Climate Change Vulnerability and Adaptation in the Livestock Sector of Mongolia. Final Report Submitted to Assessments of Impacts and Adaptations to Climate Change (AIACC), Project No. AS 06. Batima, P., N. Batnasan, and B. Lehner. 2004. Freshwater systems of the Great Lakes Basin, Mongolia. Opportunities and challenges in the face of climate change. Ulaanbataar: WWF Mongolia Program Office. Beale, C.M., and J.J. Lennon. 2012. "Incorporating uncertainty in predictive species distribution modelling ." Philosophical Transactions of the Royal Society B: Biological Sciences 367 (1586): 247-258. Bomhard, B., D.M. Richardson, J.S. Donaldson, G.O. Hughes, G.F. Midgley, D.C. Raimondo, A.G. Rebelo, M. Rouget, and W. Thuiller. 2005. "Potential impacts of future land use and climate change on the Red List status of the Proteaceae in the Cape Floristic Region, South Africa." Global Change Biology 11 (9): 1452-1468. Burrows, M.T., D.S. Schoeman, L.B. Buckley, P. Moore, E. S. Poloczanska, K.M. Brander, C. Brown, et al. 2011. "The Pace of Shifting Climate in Marine and Terrestrial Ecosystems." Science 334 (6056): 652-655. doi:10.1126/science.1210288. Chen, I.C., J.K. Hill, R. Ohlemüller, D.B. Roy, and C.D. Thomas. 2011. "Rapid range shifts of species associated with high levels of climate warming." Science 333 (6045): 1024-1026.

Chimeddorj, B., B. Buuveibaatar, Yo. Onon, O. Munkhtogtokh, and R. Reading. 2013. "Identifying Potential Conservation Corridors Along the Mongolia-Russia Border Using Resource Selection Functions: A Case Study on Argali Sheep." Mongolian Journal of Biological Sciences 11: 45-53. Clark, E. L., Munkhbat, J., Dulamtseren, S., Baillie, J. E. M., Batsaikhan, N., Samiya, R. and Stubbe, M. (compilers and editors). 2006. Mongolian Red List of Mammals. Regional Red List Series Vol. 1. Zoological Society of London, London. Dagvadorj, D., L. Natsagdorj, J. Dorjpurev, and B. Namkhainyam. 2009. Mongolia Assessment Report on Climate Change 2009. Ulaanbaatar: Ministery of Environment, Nature&Tourism, Mongolia. Dawson, T.P., S.T. Jackson, J.I. House, I.C. Prentice, and G.M. Mace. 2011. "Beyond Predictions: Biodiversity Conservation in a Changing Climate." Science 332 (6025): 53-58. doi:10.1126/science.1200303. Elith, J., C. H. Graham, R. P. Anderson, M. Dudı´k, S. Ferrier, A. Guisan, R. J. Hijmans, et al. 2006. "Novel methods improve prediction of species’ distributions from occurrence data." Ecography 29: 129-151. Elith, J., Phillips, S., Hastie, T., Dudik, M., Chee, Y.E. and Yates, C.J. 2011. A statistical explanation of MaxEnt for ecologists. Diversity and Distributions 17: 43–57. ESRI. 2017. ArcGIS Pro: Release 2.1.0. Redlands, California: Environmental Systems Research Institute. Fang, S., J. Yan, M. Che, Y. Zhu, Z. Liu, H. Pei, H. Zhang, G. Xu, and Lin X. 2013. "Climate change and the ecological responses in Xinjiang, China: model simulations and data analyses." Quaternary International 311: 108-116. Fedosenko, A.K. 2000. Arkhar (Ovis ammon) in Russia and adjacent countries. Moscow. Forrest, J.L., E. Wikramanayake, R. Shrestha, G. Areendran, K. Gyeltshen, A. Maheshwari, S. Mazumdar, R. Naidoo, G. Jung Thapa, and K. Thapa. 2012. "Conservation and climate change: Assessing the vulnerability of snow leopard habitat to treeline shift in the Himalaya." Biological Conservation 150 (1): 129-135. doi:dx.doi.org/10.1016/j.bio. Fox, J.L. and Chundawat, R.S. 2016. What is a Snow Leopard? Behavior and Ecology. In: McCarthy, T.M., and D. Mallon (Editors). 2016. Snow Leopards. Biodiversity of the World: Conservation from Genes to Landscapes. Elsevier. Gillson, L., T.P. Dawson, S. Jack, and M.A. McGeoch. 2013. "Accommodating climate change contingencies in conservation strategy." Trends in Ecology & Evolution 28 (3): 135-142. Hampe, A. 2011. "Plants on the move: the role of seed dispersal and initial population establishment for climate-driven range expansions." Acta Oecologica 37: 666–673. Hannah, L. 2011. "Climate Change, Connectivity, and Conservation Success." Conservation Biology (6) 25: 1139–1142. doi:10.1111/j.1523-1739.2011.01788.x. Hannah, L., G. F. Midgley, and D. Millar. 2002. "Climate change-integrated conservation strategies." Global Ecology and Biogeography 11: 485–495. doi:10.1046/j.1466- 822X.2002.00306.x.

Harris, R.B., G. Wingard, and B. Lkhagvasuren. 2009. "National assessment of mountain ungulates in Mongolia." Research Report to Mongolian Academy of Science, Ulaanbaatar, 64. Hickling, R., D. B. Roy, J. K. Hill, and C. D. Thomas. 2005. A northward shift of range margins in British Odonata. Global Change Biology 11:502–506. Hijmans, R.J., S.E. Cameron, J.L. Parra, P.G. Jones and A. Jarvis. 2005. Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology 25: 1965-1978. Hof, A. R., Jansson, R., and C. Nilsson. 2012. The usefulness of elevation as a predictor variable in species distribution modelling. Ecol. Model. 246:86–90. Hole, D.G., Willis, S.G., Pain, D.J., Fishpool, L.D., Butchart, S.H.M., Collingham, Y.C., Rahbek, C., Huntley, B. 2009. Projected impacts of climate change on a continentwide protected area network. Ecology Letters12: 420–431. Huntley, B., P. Barnard, R. Altwegg, L. Chambers, B.W.T. Coetzee, L. Gibson, P.A.R. Hockey, et al. 2010. "Beyond bioclimatic envelopes: dynamic species' range and abundance modelling in the context of of climatic change." Ecography 33 (3): 621-626. IPCC. 2007. "The physical science basis. Chap. 10." In Climate change 2007 , by S. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis, K. B. Averyt, M. Tignor and H. L. (Eds.) Miller. Cambridge : Cambridge University Press. IPCC. 2014. Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Core Writing Team, R.K. Pachauri and L.A. Meyer (eds.)]. IPCC, Geneva, Switzerland, 151 pp. IUCN. 2017. The IUCN Red List of Threatened Species. Accessed December 19, 2017. www.iucnredlist.org. Jackson, S.T., and D.F. Sax. 2009. "Balancing biodiversity in a changing environment: extinction debt, immigration credit and species turnover." Trends in Ecology and Evolution 25: 153– 160. Jarvis, A., H.I. Reuter, A. Nelson, E. Guevara. 2008. Hole-filled SRTM for the globe Version 4, available from the CGIAR-CSI SRTM 90m Database (http://srtm.csi.cgiar.org). Kharuk, V.I., K.J. Ranson, S.T. Im, and M.L. Dvinskaya. 2009. "Response of Pinus sibirica and Larix sibirica to climate change in southern Siberian alpine forest–tundra ecotone." Scandinavian Journal of Forest Research 24 (2): 130-139. doi:10.1080/02827580902845823. Kharuk, V.I., S.T. Im, M.L. Dvinskaya, and K..J. Ranson. 2010. "Climate-induced mountain treeline evolution in southern Siberia ." Scandinavian Journal of Forest Research 25 (5): 446- 454. doi:10.1080/02827581.2010.509329. King, G.A. and Herstrom, A.A. 1997. Holocene tree migration rates objectively determined from fossil pollen data. In: Huntley, B. et al. (Editors). Past and Future Rapid Environmental Changes: The Spatial and Evolutionary Responses of Terrestrial Biota. New York. Springer. 91–101.

Kokorin, A. O. (Editor). 2011. Climate change and its impact on ecosystems, population and economy of the Russian portion of the Altai-Sayan Ecoregion. Assessment Report, Moscow: WWF-Russia. Kolosov, A.M. 1938. History of fauna research in Altai Mountains . Vol. 1, in Researches of Altaisky Nature Reserve, by Altaisky State Nature Reserve, 327-366. Moscow. Körner, C., 2012. Alpine Treelines: Functional Ecology of the Global High Elevation Tree Limits. Springer Science & Business Media, Basel. Kramer-Schadt, S., Niedballa, J., Pilgrim, J.D., Schroder, B., Lindenborn, J., Reinfelder, V., Stillfried, M., Heckmann, I., Scharf, A.K., Augeri, D.M. Cheyne, S.M., Hearn, A.J., Ross, J., Macdonald, D.W., Mathai, J., Eaton, J., Marshall, A.J., Semiadi, G., Rustam, R., Bernard, H., Alfred, R., Samejima, H., Duckworth, J. W., Breitenmoser-Wuersten, C., Belant, J.L., Hofer, H., and A. Wilting. 2013. The importance of correcting for sampling bias in MaxEnt species distribution models. Diversity and Distributions 19: 1366–1379. Kupriyanov, A.N. 2003. Biodiversity of the Altai-Sayan Ecoregion. Kemerovo: WWF-Russia. Luo, Zh., Jiang, Zh., and Tang, S. 2015. Impacts of climate change on distributions and diversity of ungulates on the Tibetan Plateau. Ecological Applications 25(1): 24–38. Li, J., McCarthy, T.M., Wang, H., Weckworth, B.V., Schaller, G.B., Mishra, C., Lu, Z., Beissinger, S.R. 2016. Climate refugia of snow leopards in High Asia. Biological Conservation 203:188-196. Liu, C., White, M., Newell, G. 2013. Selecting thresholds for the prediction of species occurrence with presence-only data. Journal of Biogeography 40: 778–789.

Maharaj, S.S., and M. New. 2013. "Modelling individual and collective species responses to climate change within Small Island States ." Biological Conservation 167: 283-291. doi:10.1016/j.biocon.2013.08.027. Margules, C. R., and R. L. Pressey. 2000. "Systematic conservation planning ." Nature 405: 243- 253. doi:10.1038/35012251. Maroney, R.L., and M.Yu. Paltsyn. 2003. "Altai argali status and distribution in western Mongolia and the Altai-Sayan." IUCN Caprinae News 4–7. Mawdsley, J.R., R. O’Malley, and D.S. Ojima. 2009. "A review of climate-change adaptation strategies for wildlife management and biodiversity conservation." Conservation Biology 23 (5): 1080–1089. doi:10.1111/j.1523-1739.2009.01264.x. McCarthy, T.M. 2000. Ecology and conservation of snow leopards, Gobi brown bears and wild Bactrian camels in Mongolia. Ph.D. Dissertation, Amherst: University of Massachusetts, 133. McCarthy, T.M., and G. (Editors) Chapron. 2003. Snow Leopard Survival Strategy. International Snow Leopard Trust nd Snow Leopard Network: Seattle. Merow, C., M.J. Smith, and J.A. Silander. 2013. "A practical guide to MaxEnt for modeling species’ distributions: what it does, and why inputs and settings matter." Ecography 36: 1058–1069.

Midgley, G.F., I.D. Davies, C.H. Albert, R. Altwegg, L. Hannah, G.O. Hughes, L.R. O'Halloran, C. Seo, J.H. Thorne, and W. Thuiller. 2010. "BioMove - an integrated platform simulating the dynamic response of species to environmental change." Ecography 33 (3): 612-616. Mikhailov, N.N., K.V. Chistyakov, and M.I. Amosov. 1992. Geoecology of mountain depressions . Leningrad: Leningrad University. Modina, T.D., S.S. Drachev, and M.G. Sukhova. 2002. "On the issue of global climate change on the territory of the Altai Mountains ." Materials from the regional scientific-practical conference “Readings in memory of M.V. Tronov”. Tomsk. 161. Munkhtsog, B., Purevjav, L., McCarthy, T., Bayrakçısmith, R. 2016. Chapter 39: Northern Range: Mongolia. In: McCarthy, T.M., and D. Mallon (Editors). 2016. Snow Leopards. Biodiversity of the World: Conservation from Genes to Landscapes. Elsevier. Nakao, K., M. Higa, I. Tsuyama, T. Matsui, M. Horikawa, and N. Tanaka. 2013. "Spatial conservation planning under climate change: Using species distribution modeling to assess priority for adaptive management of Fagus crenata in Japan." Journal for Nature Conservation 21 (6): 406-413. doi:10.1016/j.jnc.2013.06.003. Olson, D. M., and E. Dinerstein. 2002. "The Global 200: Priority Ecoregions for Global Conservation." Annals of the Missouri Botanical Garden 89: 199-224. Pagel, J., and F.M. Schurr. 2012. "Forecasting species ranges by statistical estimation of ecological niches and spatial population dynamics." Global Ecology and Biogeography 21 (2): 293- 304. Paltsyn, M.Yu., S.V. Spitsyn, A.N. Kuksin, B. Lkhagvasuren, Yo. Onon, and O. Munkhtogtogh. 2011. Conservation of Altai argali in transboundary area of Russia and Mongolia . Krasnoyarsk: UNDP/GEF Project "Biodiversity Conservation in the Russian portion of the Altai-Sayan Ecoregion". Paltsyn, M.Yu., S.V. Spitsyn, Istomov S.V., and Kuksin A.N. 2012. Conservation of Snow Leopard in Russia. Krasnoyarsk: World Wide Fund for Nature (WWF-Russia). Paltsyn M.Yu., Spitsyn,S.V., Kuksin, A.N., Isomov S.V, Poyarkov, A.N., Rozhnov, V.V. 2015. Strategy for conservation of Snow Leopard in the Russian Federation. Approved by the Ministry of Natural Resources and Environment of the Russian Federation, Decree of August 18, 2014, # 23-r. https://wwf.ru/resources/publ/book/eng/1006 Parmesan, C., and G. Yohe. 2003. "A globally coherent fingerprint of climate change impacts across natural systems." Nature 421: 37–42. Phillips, S.J., and M. Dudik. 2008. "Modeling of species distributions with Maxent: new extensions and a comprehensive evaluation." Ecography 31: 161-175. Phillips, S.J., R.P. Anderson, and R.E. Schapire. 2006. "Maximum entropy modeling of species geographic distributions." Ecological Modeling 190: 231-259. Pounds, A.J., M.R. Bustamante, L.A. Coloma, J.A. Consuegra, M.P.L. Fogden, P.N. Foster, E. La Marca, et al. 2006. "Widespread amphibian extinctions from epidemic disease driven by global warming." Nature 439 : 161–167. Pressey, R., M. Cabeza, M. Watts, Cowling R.M., and K.A. Wilson. 2007. "Conservation planning in a changing world." Trends in Ecology & Evolution 22 (11): 583-592.

Reading, R.P., S. Amgalanbaatar, H. Mix, and B. Lhagvasuren. 1997. "Argali (Ovis ammon) surveys in Mongolia’s South Gobi." Oryx 31: 285-294. Reading, R.P., S. Amgalanbaatar, and Mix H. 1998. "Recent Conservation Activities for Argali (Ovis ammon) in Mongolia – Part 1." Caprinae News 1-3. Reading, R.P., S. Amgalanbaatar, and H. Mix. 1999. "Recent Conservation Activities for Argali (Ovis ammon) in Mongolia – Part 2." Caprinae News 1-4. Riordan, P., and Shi, K. 2016. China: Current State of Snow Leopard Conservation in China. In: McCarthy, T.M., and D. Mallon (Editors). 2016. Snow Leopards. Biodiversity of the World: Conservation from Genes to Landscapes. Elsevier. Rowland, E.L., J.E. Davison, and L.J. Graumlich. 2011. "Approaches to Evaluating Climate Change Impacts on Species: A Guide to Initiating the Adaptation Planning Process." Environmental Management 47 (3): 322-337. Samoilova, G.S. 2001. The GIS Landscape Map of the Altai-Sayan Ecoregion. WWF Russia (CD- R). Schweiger, Oliver, Risto K. Heikkinen, Alexander Harpke, Thomas Hickle, Stefan Klotz, Otakar Kudrna, Ingolf Kühn, Juha Pöyry, and Josef Settele. 2012. "Increasing range mismatching of interacting species under global change is related to their ecological characteristics." Global Ecology and Biogeography 21 (1): 88–99. Shabani, F., Kumar, L., and Ahmadi, M. 2016. A comparison of absolute performance of different correlative and mechanistic species distribution models in an independent area. Ecology and Evolution 6(16): 5973–5986. Shackleton, D.M. (Editor). 1997. Wild Sheep and Goats and their Relatives: Status Survey and Conservation Action Plan for Caprinae. Gland, Switzerland: IUCN. Snow Leopard Working Secretariat. 2013. Global Snow Leopard and Ecosystem Protection Program. Bishkek, Kyrgyz Republic. Sobansky, G.G. 1992. Ungulates of Altai Mountains. Novosibirsk: Science. Sopin, L.V. 1975. "Wild sheep of South Siberia ." Doctor of Science Dissertation, Irkutsk . Stocker, T., Qin, D., Plattner, G.K., Tignor, M., Allen, S.K., Boschung, J., Nauels, A., Xia, Y., Bex,V., and Midgley, P.M. 2013. Climate Change 2013: The Physical Science Basis. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

Tchebakova, N.M., T.A. Blyakharchuk, and E.I. Parfenova. 2009. "Reconstruction and prediction of climate and vegetation change in the Holocene in the Altai-Sayan Mts, Central Asia." Environmental Research Letters. doi:10.1088/1748–9326/4/4/045025. Thomas, C.D., J.K. Hill, B.J. Anderson, S. Bailey, C.M. Beale, R.B. Bradbury, C.R. Bulman, et al. 2011. "A framework for assessing threats and benefits to species responding to climate change." Methods in Ecology and Evolution 2 (2): 125–142. doi:10.1111/j.2041- 210X.2010.00065.x. Tsalkin, V.I. 1951. Mountain sheep of Europe and Asia. Moscow: Moscow Society of Nature Explorers.

UNDP/GEF. 2005. "Biodiversity Conservation in the Russian Portion of the Altai-Sayan Ecoregion – Phase I." Full size project document. UNDP/GEF. 2007. Biodiversity of the Altai-Sayan Ecoregion. Web database . Accessed December 19, 2013. http://www.bioaltai-sayan.ru/regnum/eng/index.htm. UNDP/GEF. 2006. "Community-based Conservation of Biological Diversity in the Mountain Landscapes of Mongolia’s Altai-Sayan Ecoregion." Full size project document. UNEP/CMS. 2014. International Single Species Action Plan for the Conservation of the Argali. CMS Technical Series No. XX. April 2014. UNESCO World Heritage Convention. 2013. World Heritage List. Accessed December 15, 2013. http://whc.unesco.org/en/list. Wang, X.M., and Schaller, G.B., 1996. Status of large mammals in western Inner Mongolia, China. J. East China Normal Univ. Nat. Sci. 12, 93–104. Weeks, A.R., C.M. Sgro, A.G. Young, R. Frankham, N.J. Mitchell, K.A. Miller, M. Byrne, et al. 2011. "Assessing the benefits and risks of translocations in changing environments: a genetic perspective." Evolutionary Applications 4 (6): 709–725. doi:10.1111/j.1752- 4571.2011.00192.x. Williams, P., L. Hannah, S. Andelman, G. Midgley, M. Araújo, G. Hughes, L. Manne, E. Martinez- Meyer, and R. Pearson. 2005. "Planning for Climate Change: Identifying Minimum- Dispersal Corridors for the Cape Proteaceae." Conservation Biology 19: 1063–1074. doi:10.1111/j.1523-1739.2005.00080.x. World Wide Fund for Nature. 2012. Altai-Sayan Ecoregional Conservation Strategy. WWF. Ye, X., Yu, X., Yu, C., Tayibazhaer, A., Xu, F., Skidmore, A.K., Wang T. 2018. Impacts of future climate and land cover changes on threatened mammals in the semi-arid Chinese Altai Mountains. Science of the Total Environment 612: 775-787.

CHAPTER 4: TIGER RE-ESTABLISHMENT POTENTIAL TO FORMER CASPIAN TIGER (PANTHERA TIGRIS VIRGATA) RANGE IN CENTRAL ASIA

Abstract: Caspian tigers (Panthera tigris virgata), a now extinct subspecies genetically similar to the Amur tiger (P. t. altaica), occurred until the mid-1900s from modern day Turkey and Iran east through Central Asia into northwest China. A literature analysis we conducted revealed that Caspian tigers occupied ca. 800 000-900 000 km² historically, mostly within isolated patches of tugay- and reed-dominated riparian ecosystems at densities up to 2-3 tigers/100 km². Herein we explored options to restore tigers to Central Asia using Amur tiger as an “analog” form. Spatial analyses based on remote sensing data indicated that options for Amur tiger introduction are limited in Central Asia but at least two habitat patches remain potentially suitable for tiger re-establishment, both in Kazakhstan, with a total area of <20 000 km². The most promising site—the Ili river delta and adjacent southern coast of Balkhash Lake—hosts ca. 7 000 km² of suitable habitat that our tiger-prey population models suggest could support a population of 64-98 tigers within 50 years if 40-55 tigers are translocated and current Ili river flow regimes are maintained. Re-establishment of tigers in Central Asia may yet be tenable if concerns of local communities in the Ili-Balkhash region are carefully addressed, prey population restoration precedes tiger introduction, Ili river water supplies remain stable, and the Amur tiger’s phenotype proves adaptable to the arid conditions of the introduction site.

Introduction

An estimated 100 000 tigers belonging to nine subspecies roamed the Earth in the early 1900s (Jackson 1993) but by 2000 only 3 200-3 600 remained, with four subspecies extinct (Global Tiger Recovery Program 2010; Seidensticker 2010). One extinct subspecies – the Caspian tiger (Panthera tigris virgata) – once populated the largest geographical range of any tiger subspecies: from modern day Turkey through much of Central Asia to northwestern China (Geptner and Sludski 1972). By the late 1800’s Caspian tigers still inhabited modern day Afghanistan, Armenia, Azerbaijan, northwest China, Georgia, Iran, Iraq, Kazakhstan, Kyrgyzstan, Tajikistan, Turkey, Turkmenistan, and Uzbekistan (Geptner and Sludskiy 1972) but had completely disappeared by the 1940s-1960s leading to the subspecies’ current designation under Category 1 “Extinct” on the IUCN Red List (Nowell and Jackson 1996). Considering that by the early 2000s only 13 countries still supported tigers, the extinction of Caspian tiger reduced aggregate tiger range countries almost by half.

Several factors led to the extinction of Caspian tigers. In the first half of 1900s widespread extermination of Caspian tigers was accomplished mainly through poisoning and trapping promoted by bounties paid in the former Soviet Union until the 1930s (Geptner and Sludskiy 1972). Numerous irrigation and agriculture projects in Central Asia during the Soviet era destroyed the tugay woodlands (a riparian and coastal ecosystem comprised of trees, shrubs, and wetlands) and reed thickets earlier found alongside all major rivers and lakes that were critical tiger habitats. The tiger’s main prey, wild boar (Sus scrofa) and Bukhara deer (Cervus elaphus bactrianus), also disappeared along with riparian habitats (Sludskiy 1953; Geptner and Sludskiy 1972). In the former Soviet Union a complete ban on tiger hunting was introduced in 1947; however, occasional killings led to complete disappearance of tigers from Central Asia by 1950s (Geptner and Sludskiy 1972; Sludskiy 1973).

Despite the grim history of extinction of the Caspian tiger, two factors have combined to raise the possibility of tiger restoration to Central Asia. First, the breakup of the Soviet Union and introduction of market economies in newly established states lead to the recovery of tiger habitats in some areas as many state-sponsored agricultural programs along the rivers have been abandoned (WWF 2014). Second, tiger phylogenetics (Driscoll et. al. 2009) recently revealed that Caspian tigers were extremely close relatives to the still extant Amur tiger (P. t. altaica) which might therefore serve as suitable “analog” form for restoration of tigers to Central Asia. Modest recovery of Caspian tiger habitats along with new perspectives on tiger phylogenetics triggered wider discussions about possible tiger introduction to Central Asia (Driscoll et al. 2011, 2012; WWF 2014). Herein we explore the biological feasibility of bringing tigers back to Central Asia based on: (1) detailed analysis of historical Caspian tiger distribution and biology, (2) exploration of habitat suitability for the tiger restoration in the region, and (3) modeling of tiger and prey relationships to evaluate options for restoration. We conclude that although options for tiger restoration remain extremely limited in Central Asia, successful introduction of Amur subspecies to the Ili-Balkhash region of Kazakhstan is biologically tenable.

Materials and Methods Analysis of historic Caspian tiger distribution in Central Asia An assessment of former geographic range of Caspian tigers was implemented via analysis of scientific publications and technical reports containing historical records from the 1800s and 1900s on tiger distribution, sightings, habitat, prey populations, and years since extinction in Central Asia and adjacent countries: Flerov (1935), Sludskiy (1953, 1966, 1973), Chernyshev (1958), Shukurov (1958), Vereshchagin (1959), Stroganov (1961), Burchak Abramovich and Mamedov (1965), Niethammer (1966), Mambetzhumaev and Palvanijazov (1968), Geptner (1969), Geptner and Sludski (1972), Baytop (1974), Kock (1990), Nowell and Jackson (1996), Hajiyev (2000), Can (2004), Abdukadir and Breitenmoser (2008), Jackson and Nowell (2011). All records collected on tiger sightings were mapped as point data and combined maps of tiger distribution in Central Asia published by Geptner and Sludskiy (1972) to extrapolate historic range. Integrating the mapping exercise with the literature analysis enabled identification of preferred habitats of Caspian tigers in Central Asia as well as estimation of population density where tigers occurred.

Evaluation of site suitability for tiger restoration We performed a spatial analysis to identify sites for tiger introduction that might be currently available given the following criteria for designating critical tiger habitats. The first was availability of large (>5 000 km²) and continuous areas of tugay forest and reed thickets, which were the primary habitat of Caspian tigers originally (Flerov 1935; Sludskiy 1953, 1966, 1973; Chernyshev 1958; Shukurov 1958; Vereshchagin 1959; Stroganov 1961; Burchak Abramovich and Mamedov 1965; Niethammer 1966; Mambetzhumaev and Palvanijazov 1968; Geptner 1969; Geptner and Sludski 1972; Baytop 1974; Kock 1990; Hajiyev 2000; Can 2004; Abdukadir and Breitenmoser 2008). This area threshold was based on an area large enough to support a population

63 of at least 100-150 tigers at the historical density of 2-3 tigers/100 km² (Chernyshev 1958; Geptner 1969; Geptner and Sludski 1972). The second was human population density (< 3 persons/km²), a threshold well below that typically used for defining “wilderness” (<5 people/km²) (Mittermeier et al. 2003). The third was high current or historical population densities of wild ungulates as tiger prey for which wild boar, Bukhara deer, and roe deer (Capreolus capreolus) are the primary species and of which 10-30 ungulates/km² is apparently required to support tiger populations (Geptner 1969; Geptner and Sludski 1972). Based on these criteria we performed a four-stage selection process by first extracting all tugay forest and reed thicket ecosystems in Central Asia (territories of Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan, and Uzbekistan) based on the spatial layer “Ecosystems of Central Asia” (amalgamating all 21 types tugay/reed ecosystems provided in GEF/UNEP/WWF 2005). Second, we subset continuous areas of tugay forest and reed thickets > 5 000 km² and, third, filtered areas of the habitat with low human population density based on the 2.5 arc-minute Population Density Grid 2000 (CIESIN - Columbia University and CIAT 2005). Last, the areas subset was evaluated based on Landsat 8 images for August 2014 (source: USGS Earth Explorer http://earthexplorer.usgs.gov/) to identify current land use. We supplemented these analyses with reports from ungulate field surveys (Jungius et al. 2009; Lukarevskiy and Baidavletov 2010) to evaluate current population density of ungulates in the sites selected.

Modeling of tiger and tiger prey habitat and population dynamics at a high priority introduction site We modeled habitat dynamics for ungulate prey (wild boar, Bukhara deer, and roe deer) and for tigers as well as prey and predator population growth for the priority area highlighted in our critical habitat analysis (see Results) and also previously suggested (Lukarevskiy and Baidavletov 2010; Institute of Geography of Kazakhstan Academy of Sciences 2013; WWF 2014): the Ili River delta and adjacent southern coast of Balkhash Lake. Our modeling explored over the next 50 years the time required for restoration of the tigers’ prey base, number and introduction timing of tigers required for their successful restoration, and overall tiger population carrying capacity. Habitat modeling for all species was based on Landsat 5 TM images for July-August 1989 (period of the lowest water level in Balkhash Lake in 1950-2015) and July-August 2010 (current high water level) (source: USGS Earth Explorer http://earthexplorer.usgs.gov/) classified into four general habitat classes: water; desert and semi-desert; dry bush, meadow and steppe; and tugay woodlands and reed thickets using the ArcMap 10.2.2’s (ESRI 2014) Image Classification function. Resulting habitat maps were used to model three management scenarios of tiger introduction program in the area: No Fire Management (area of high quality tiger habitat constant over next 50 years), Fire Management (area of high quality tiger habitat gradually increasing by 25-27% over next 50 years under fire and grazing control), and Water Scarcity (area of high quality tiger habitat gradually decreasing by 70% over next 50 years as a result of increased water consumption from Ili river and decreasing level of Balkhash Lake) (Paltsyn et al. 2015, see Appendix 2).

Based on our historical analysis we assumed that only tugay woodlands and reed thickets represent suitable, high quality habitat for tiger and two of its key prey species (wild boar and Bukhara deer) (Geptner 1969; Geptner and Sludski 1972; WWF 2014). Dry bush, meadow and steppe were classified as marginal habitat for these species, but optimal for roe deer, another tiger prey species (Lukarevsky and Baidavletov 2010; WWF 2014). Burned tugay woods and reed thickets were considered to represent marginal habitats for all species, due to their intensive use as pastures and low cover. Water, deserts and semi-deserts were designated as non-habitat (for all species). We used population densities of tiger prey species derived from Geptner et al. (1961), Danilkin (2006), Pereladova (2013), and WWF (2014) to estimate carrying capacity (K) for the prey species (Table 1). Table 1. Estimated maximum population densities of wild boar, Bukhara deer and roe deer in the Central Asia biomes

Maximal population density (ind./km²) Species High quality tugay Low quality tugay Dry bush, meadow and woodlands and reed woodlands and reed steppe thickets thickets (burned areas) Wild boar 6.0* 0.5 0.5* Bukhara deer 10.0** 0.5 1.0** Roe deer 0.5*** 0.5 1.0*** *- population density of wild boars can achieve up to 5-30 animals/km² in tugay and reed ecosystems of Central Asia, but does not usually exceed 0.5-1.0 animals/km² in arid ecosystems (Geptner et al. 1969; Danilkin 2006; WWF 2014) ** - Pereladova (2013) *** - WWF (2014)

To further evaluate feasibility of tiger introduction scenarios for the priority site in the Ili-Balkhash region (see Results) the aggregate tugay and reed thicket ecosystem was divided in three Tiger Management Units (TMUs): Ili Delta, Balkhash, and Karatal because their spatial separation. Given that the tiger introduction program may start only in one TMU due to limited source of tigers and Bukhara deer for translocation to restore populations, we built prey species and tiger population projections for each TMU separately. To describe population growth of the prey species we integrated vital rates of each prey species in a model of density-dependent population (logistic) growth typically used in wildlife population projection scenarios (Caughley 1980):

Nt+1 = Nt + rNt(1- Nt/K), where

Nt is abundance of population in year t;

Nt+1 is abundance of population the following year (t+1); r is instantaneous per capita growth rate; and

K is population carrying capacity. Vital rates for the prey species (mean r +1 SD and population abundance at Year 0) were calculated from available time series for these species (Ignatova et al. 2004; Zaumyslova 2005; Mayer 2009; Nilsen et al. 2009; Melis et al. 2010; Pereladova 2013) and adopted from survey reports

(Lukarevskiy and Baidavletov 2010; WWF 2014) (Table 2). Environmental stochasticity was incorporated via selection of lognormal random deviates for mean r (+1 SD) at each time step. Each modeled scenario was iterated 1,000 times to calculate mean prey species abundance (and its 95% confidence interval) at years 5, 10, 15, …, 50 of the introduction program.

Table 2. Estimated values of N0 and r = for wild boar, Bukhara deer, and roe deer in the Ili- Balkhash region

Species r mean SD Population abundance (Year 0)

Wild boar 0.349* 0.503 2 697**

Bukhara deer 0.144*** 0.111 30**** Roe deer 0.020***** 0.143 1 270**

*- calculated as average from time-series for wild boar populations in the Russian Far East, Belovezhskay Puscha, Crimea, and Europe (Zaumyslova 2005; Mayer 2009). No time-series for tugay forest and reed thickets were found. ** - our assumption based on the data of Lukarevskiy and Baidavletov (2010) and WWF (2014). *** - calculated as average using time-series from Badai-Tuagai Nature Reserve, Uzbekistan, and Karachingil Game Management Area, Kazakhstan, in 1999-2011(Pereladova 2013). **** - 30 Bukhara deer will be brought to the reintroduction site from Karachingil Game Management Area annually in the first 10 years of the program. ***** - calculated as average from time-series from Russian Far East, Norway, and France (Ignatova et al. 2004; Nilsen et al. 2009; Melis et al. 2010). No time-series for tugay forest and reed thicket were found.

We then predicted tiger population growth by integrating an age-structured tiger population matrix model (following Tian et al. 2011) with projected prey species abundance inherited from the prey models. The model was female-only and assumed that density-dependence affected only reproduction but not survival rates based on the following assumptions: Amur tigers (P. t. altaica) from Russian Far East are used for introduction in the Ili-Balkhash region; habitat carrying capacity for tigers is limited by prey species population abundance only (Karanth et al. 2004); only sub-adult and adult tigers kill their own prey; number of individuals of a particular prey species killed by a tiger is directly proportional to the ratio of a given species abundance to all prey species abundance; only sub-adult and adult tiger abundance influence carrying capacity; female/male ratio of tiger population is constant and similar to that of the Amur tiger (Carroll and Miquelle 2006; Miquelle et al. 2009); and tiger vital rates in the introduction area are comparable to those of the Amur tiger in its native range (Tian et al. 2011). We used following formula to calculate carrying capacity for tigers in the project site:

Kt = Nprey t/450, where

Kt is carrying capacity for tigers in the project site in the year t,

Nprey t is abundance of all prey species (wild boar, Bukhara deer and roe deer) in the project site,

450 is the mean number of ungulates of different body mass required to support one tiger (Karanth et al. 2004; Ministry of Natural Resources and Environment of Russian Federation 2010).

The effect of tiger predation on a particular ungulate species was calculated using following formula:

Nt+1 = Nt + rNt(1- Nt/K) – Ns+a t * (Nt / Np t)*R s+a , where

Nt is abundance of an ungulate species population in year t ;

Nt+1 is abundance of an ungulate species population next year (t+1); r is the instantaneous per capita growth rate of an ungulate species; K is ungulate species population carrying capacity;

Ns+a t – number of sub-adult and adult tigers in year t;

Np t – abundance of all ungulate species in year t;

R s+a – number of ungulates killed by one sub-adult or adult tiger annually (constant value equal 60: Karanth et al. 2004; Ministry of Natural Resources and Environment of Russian Federation 2010). Poaching for tiger (including retaliatory killing) and prey species was not addressed in the models assuming that since 2017-2018 100% of tugay and reed ecosystems in the Ili-Balkhash region will be integrated into the planned Ili-Balkhash Nature Reserve (IUCN Category II) with trained staff of ca. 100 inspectors and strong regime of protection (see Government of the Republic of Kazakhstan 2010; WWF 2014).

Results

Historic Caspian tiger distribution in Central Asia and adjacent countries

Historic records demonstrate that in 10-12th centuries AD the range of Caspian tigers extended from Eastern Europe to the west to western Siberia and northwestern China to the east and from south Russia to the north to the Middle East on the south (Geptner 1949; Geptner 1969; Geptner and Sludski 1972, Fig. 1). According to our range reconstruction, the species once occupied ca. 800,000-900,000 km² and had a metapopulation structure based around large patches of habitat (3,000 – 250,000 km²) isolated or semi-isolated by steppe, desert, and high mountains (Fig. 1). During 1920-1970 Caspian tigers vanished from 11 modern day countries (Table 3).

Figure 1. Documented sightings of Caspian tigers in 10-20th centuries and reconstructed 19-20th century range based on review of historic accounts (see text).

Table 3. Estimated year of Caspian tiger extinction in Central Asian and adjacent countries Country Last tiger presence recorded (source)

Afghanistan 1950-1960s (Niethammer 1966; Abdukadir and Breitenmoser 2008) Armenia 1948 (Vereshchagin 1959) Azerbaijan 1964 (Baytop 1974) China 1920s (Abdukadir and Breitenmoser 2008) Georgia 1936 (Vereshchagin 1959) Iran 1966 (Niethammer 1966) Iraq 1887 (Kock 1990) Kazakhstan 1948 (Geptner and Sludskiy 1972) Kyrgyzstan 1890 (Sludskiy 1966) Tajikistan 1971 (Geptner and Sludskiy 1972) Turkey 1970-1973 (Baytop 1974; Can 2004) Turkmenistan 1954 (Shukurov 1958) Uzbekistan 1968 (Mambetzhumaev 1968)

In the tugay woodlands and reed thickets along larger rivers and lakes that tigers frequented, wild boar and Bukhara deer were evidently the main prey species for Caspian tiger in Central Asia; however, tigers occasionally hunted saiga antelopes (Saiga tatarica), porcupines (Hystrix indica), and kulans (Equus hemionus kulan) (Geptner and Sludsky 1972). Highest densities of tigers in Central Asia were recorded for tugay and reed thicket ecosystems in the watersheds of Amu Darya, Syr Darya, and Ili rivers (Geptner 1969; Geptner and Sludski 1972). Sludsky (1953) reported that about 50 tigers were killed annually in the Amu Darya watershed in 1840-1850s: approximately 1 tiger was killed per 100 km² of tugay ecosystems every year. Chernyshev (1958) reported about 12-15 tigers inhabiting 490 km² of Tigrovaya Balka Nature Reserve in Tajikistan in 1930s. In 1948

68 he counted tracks of three different tigers on a 100 km long route in tugay woodlands along Vakhsh river (Chernyshev 1958). Based on these historic records we approximate population density of Caspian tiger in tugay and reed ecosystems as at least 2-3 animals/100 km². From this density estimate we extrapolate that 5000 km² of continuous area of tugay and reed ecosystem would be minimal habitat patch to support viable tiger population of 100-150 individuals.

Evaluation of site suitability for tiger introduction Our spatial analysis revealed that current distribution of tugay and reed ecosystems in Central Asia does not exceed 70 000 km² and is generally limited to lower valleys and deltas of Ural, Amu Darya, Syr Darya, Chu, and Ili rivers and coasts of Balkhash, Alakol, and Zaisan lakes (Fig. 2). All these areas, excluding the Ural River, originally had high population densities of Caspian tigers (2-3 animals/100 km²) (Chernyshev 1958; Geptner 1969; Geptner and Sludski 1972), wild boars (up to 30 animals/km²) (Geptner et al. 1961; Danilkin 2006), and Bukhara deer (up to 10 animals/km²) (Pereladova 2013). We could identify only two sites that qualified as tugay and reed ecosystems larger than or minimum criterion for 5 000 km² and with low human population density (<3 persons/km²) throughout all of Central Asia: lower valley and delta of Syr Darya River (~8 700 km²) and Ili River delta with adjacent area of Balkhash Lake coast (~7 000 km²), Kazakhstan (Fig. 2). However, analysis of land use patterns for both sites with Landsat 8 2014 images revealed that 50-60% of Syr Darya site is used for agriculture, whereas Ili-Balkhash region remains in native tugay and reed thicket cover. In contrast, the Ili-Balkhash region has low human population density (< 2 persons/km²) and relatively low livestock numbers (total 120,000-150,000 animals) in the targeted habitats (Institute of Geography of Kazakhstan Academy of Sciences 2013). Field surveys (a combination of traditional winter transect survey and interviewing of local hunters) in the Ili River delta reported a low population of wild boar (<=3 000 individuals) and roe deer (< 1 500 individuals), absence of Bukhara deer, widespread poaching, and intensive burning of tugay woodlands and reed thickets by local people to convert them to pastures for livestock (Lukarevskiy and Baidavletov 2010).

Figure 2. Present distribution of tugay and reed ecosystems in Central Asia. Two potential sites for tiger introduction, both in Kazakhstan, are Syr Darya River lower valley and delta (A) and Ili River delta with adjacent area of Balkhash Lake coast (B).

Modeling of tiger population dynamics at a high priority introduction site Our habitat projections (Figs. 3, 4) projected carrying capacity for prey species in the Ili-Balkhash region would remain stable at the level of ~100 000 animals under the No Fire Management scenario, gradually increase up to ~ 124 000 under the Fire Management scenario, or gradually decline to ~ 37 000 animals under the Water Scarcity scenario within 50 years from program initiation (Fig. 4). Without tiger predation total ungulate populations in the Ili delta TMU may reach 40 000-85 000 animals by Year 50 (Fig. 5). These ungulates’ abundance could potentially support 94 -187 tigers if tiger predation is not taken into account (Fig. 5). The Balkhash and Karatal TMUs could support no more than 35 tigers each by Year 50 (Fig. 5). In the conditions of increasing water scarcity Ili delta TMU will be likely to host an ungulate population adequate to support about 49 – 66 tigers by Year 50 (without effect of tiger predation) (Fig. 5) whereas Balkhash and Karatal TMUs are likely to have enough prey to support no more than 9 and 12 tigers, respectively (Fig. 5).

In terms of scheduling introduction efforts, years 3-10 of the program are projected as the earliest option in terms of prey availability to bring a first group of 5 tigers (3 females and 2 males) to the Ili delta TMU; however, Year 15-20 would be the preferred and conservative option as doing so accommodates the slowest prey population recovery scenario. For the same reason Year 20-25 is preferred to bring first 5 tigers to Balkhash or Karatal TMUs (Fig. 5), even though years 5 – 20 might be possible assuming more rapid prey population growth. Introduction of eight groups of five tigers each (40 tigers in total) to the Ili delta TMU in Years 20-48 was projected to generate a population of 64-70 tigers by Year 50 (Fig. 6a & c). Release of 50 tigers (10 groups by 5 tigers) in the TMU in Years 15-42 would be likely to achieve 77-88 individuals by Year 50, but population decline to 21-22 individuals was also likely as a result of subsequent prey depletion (Fig. 6, b). Similarly, introduction of 55 tigers (11 groups) in the TMU in Years 15-45 may result in 92-98 individuals by Year 50 with probability of population decline to 60 individuals due to prey depletion (Fig. 6, d). Only 17-18 tigers were projected for Balkhash or Karatal TMUs by the Year 50 if two groups of five tigers were introduced in each TMU in Year 20-28 (Fig. 6, g and j). Introduction of three groups (15 tigers) in the TMUs may result in 23-25 tigers each by Year 50 if the areas have sufficient prey base (Fig. 6, h and k, see also Paltsyn et al. 2015 for details, Appendix 2). The largest, self-sustaining population of tigers (estimated carrying capacity including effect of tiger predation on prey populations) that could be achieved in Ili delta TMU in the next 50-80 years was projected at 80-84 tigers (adults, sub-adults, juveniles and cubs) under No Fire Management Scenario and 90-94 tigers under Fire Management Scenario. Estimated carrying capacity of Balkhash and Karatal TMU would not apparently exceed 24-26 tigers each. All three TMUs in the long-term may have a total tiger population of 121-129 individuals if no fire management and grazing restrictions are implemented in the area, and 137-145 animals if habitats are improved under Fire Management Scenario. Under the scenario of progressive water level decline the Ili delta TMU would likely to sustain a small population of 14-47 tigers by Year 50 if no more than 25 tigers are introduced to the area in Year 15-27 (Fig. 6, e and f); however, Balkhash and Karatal tiger populations are likely to go below quasi-extinction threshold (10 tigers) under the same scenario (Fig. 6, i and l).

Figure 3. Projected dynamics of tugay woodland and reed thicket ecosystem in the Ili-Balkhash region under three management scenario: A - No Fire Management, B – Fire Management Years 30-50, and C – Water scarcity Year 20 (see text and Paltsyn et al. (2015) for details). Tiger Management Units (TMU) planned for introduction: 1 – Balkhash, 2 – Ili delta, 3 – Karatal.

Figure 4. Dynamics of tiger habitat area (a-c) and prey carrying capacity (d-f) predicted in the Ili-Balkhash region over the nearest 50 years under three management scenarios.

Figure 5. Projections for prey populations for wild boar, Bukhara deer, and roe deer and carrying capacity for tigers in three TMUs of Ili-Balkhash region under 3 management scenarios. Prey population and tiger carrying capacity were calculated without factoring in effect of tiger predation.

Figure 6. Population projections for tigers at different scenarios and schemes of introduction in 3 TMUs in the Ili-Balkhash region (see text for details).

Discussion Despite the Caspian tiger’s extensive geographic range historically throughout Central Asia, Middle East and northwest China, these tigers apparently occurred in patches of habitat (riparian and mountain forest ecosystems) separated by large areas of steppe and desert (Fig. 1). Narrow stripes of tugay and reed ecosystems in river valleys generally < 5 000 km² in area apparently represented optimal habitat for tigers where their population density was highest (> 2-3 tigers per 100 km²) (Sludsky 1953; Chernyshov 1958; Geptner and Sludsky 1972). According to Geptner and Sludsky (1972), tigers could cross tens and even hundreds kilometers of steppe and deserts and moved between the Amu Darya and Syr Darya river valleys as well as between other tugay and reed ecosystem complexes in Central Asia. Thus, a patch-based metapopulation organization with occasional migration linkages likely defined Caspian tiger population structure and should be considered as the conceptual basis for any future species introduction program. It also argues that Caspian tigers formerly occurred in semi-isolated, relatively small sub-populations due to the fragmented nature of their preferred habitat—tugay woodlands and reed thickets along larger rivers and lakes—and could do so again under an introduction program focused on a large but isolated habitat remnant such as the Ili River delta and adjoining area. Today we estimate that no more than 70 000 km² of tugay woodlands and reed ecosystems remain in Central Asia. More than 70% of this remaining area has relatively high human population density (5-50 people/km²) and is at least partly converted to agriculture. Thus, no more than 15 000 - 20 000 km² of the habitat located in the Ili-Balkhash region, middle and lower parts of Syr Darya River valley, middle part of Chu River valley, and delta of Amu Darya River could be potentially suitable for tigers. This said, extent of tugay woodlands and reed thickets may well increase by another 20-30% (5 000-7 000 km²) in the next 20-40 years as a result of recovery of the ecosystems in the Amu Darya River delta and establishment of tugay and wetlands on adjacent former bottom of the Aral Sea assuming water flow remains sufficient (Jungius et al. 2010). Our analysis suggests that the contention of Driscoll et al. (2012) that 1 000 000 km² of habitat are available for reintroduction in former Caspian tiger range in Central Asia is overly optimistic. Currently only the Ili River delta and southern shore of Balkhash Lake provide continuous suitable tiger habitat with sufficient total area (~ 7,000 km²) that can be considered as a potential focus for tiger restoration program for Central Asia. Our results are also consistent with results of previous reports (Jungius 2010; Lukarevsky and Baidavletov 2010) that identify the Ili-Balkhash region as the optimal place for tiger reintroduction program in Central Asia, however, we found that the actual area of available habitat is smaller than identified by Jungius 2010 (10 000 km²). At present Ili River delta and adjacent shore of Balkhash Lake have a limited prey base (3 500- 4 200 of wild boar and roe deer) sufficient to support no more than 5-8 tigers (Lukarevskiy and Baidavletov 2010); however, intensive law enforcement along with strict control of wild fires, livestock grazing management and reintroduction of Bukhara deer may increase total number of ungulates up to 25 000-35 000 individuals in the Ili delta the nearest 15-20 years (Fig. 5). If this level of prey-population restoration is achieved, total tiger population in the Ili delta could reach 80-94 tigers in the next 60-80 years assuming high prey carrying capacity of the habitat. Our study suggests that the entire Ili-Balkhash region could potentially support a population of 121-145 tigers

(population density 1.7-2.1 tigers/100 km²) within 100-150 years from the inception of an introduction program. An expectation that the program area will experience very low poaching pressure is reflected by the Kazakhstan Government’s current proposal to establish the Ili- Balkhash Nature Reserve (IUCN Category II, 1 000 000 ha) that will cover 100% of tugay and reed ecosystems in the region and include ca. 100 inspector staff as well as the government commitment to establish and support mobile anti-poaching brigades to ensure successful tiger introduction (Government of the Republic of Kazakhstan 2010; WWF 2014). It is unfortunate that as far as is known no Caspian tigers remain extent, including in captivity (Jackson and Nowell 2011) as a possible source of tigers for the reintroduction. Given that the Amur tiger (P. t. altaica) is almost genetically identical to Caspian tiger (P. t. virgata) (Driscoll et al. 2009), this subspecies could be considered as an analog and source population for tiger introduction in the Ili-Balkhash region. Amur tiger is apparently the only tiger subspecies that has significantly increased in number during the last 65 years (Global Tiger Recovery Program 2010) and now consists from 523-540 individuals (WWF 2016). Translocation of 25-50 tigers from the Russian Far East into the Ili river delta is likely to be sufficient to eventually establish a wild population of 45-90 tigers in 50 years of the program and would not threaten the source population (see Paltsyn et al. 2015 for details) assuming poaching and habitat destruction impacts on the Amur population continue to be successfully mitigated. As a source of translocated tigers, every year Amur tigers associated with conflicts as well as orphaned cubs are removed from the wild (Goodrich and Miquelle 2005) and could be considered candidates. Moreover, a new population established in the Ili River delta would also provide an insurance population for the Far Eastern taxon. Adaptive capacity of Amur tiger to cope with environmental conditions of Central Asia (arid continental climate and different type of habitat) is a chief concern. Notably, relocation of five South China tigers (P. t. amoyensis) to climatically novel, semi-wild conditions in South Africa was followed by their successful reproduction and subsequent growth of this small group to 20 individuals (Save China's Tigers 2016). Mountain lions (Puma concolor) provide another example in which eight mountain lions (P. c. stanleyana) were translocated from western Texas to recover highly endangered Florida panther (P. c. coryi) in 1995 (Johnson et al. 2010). Moreover, the Everglades subpopulation of the Florida panther may well be descended in part from released captive panthers of South American origin (Roelke, Martenson and O'Brien 1993). A more complicated yet still successful reintroduction of Canada lynx (Lynx canadensis) from Canada to the Colorado Plateau (southern extreme of the species historic range) is another case study (Devineau et al. 2010). Together these case studies suggest high adaptive potential of big cats to novel environments. We know of no large cat translocation programs that failed strictly due to maladaptation of source population to environment of release. A major risk for this introduction program is increasing water consumption in the upper reaches of Ili River in Xinjiang, China: in 1995 - 2003 total area of irrigated lands in the Chinese part of Ili river basin increased by 364% (from 156,000 to 568,400 ha) (Bragin 2006). Water consumption higher that 10-15% of annual flow of Ili river may result in the catastrophic decrease of water level, increased salinity in Balkhash lake and degradation of tugay and reed ecosystems (Bragin

2006). We project that progressive water level decline over the nearest 50 years would undercut an otherwise biologically feasible tiger introduction program; therefore, successful restoration of tiger populations is contingent on consistent river flow which may require a special agreement on future water consumption be reached between China and Kazakhstan. Reintroduction of large carnivores such as tigers has an extremely high risk of failure if interests of local communities are not carefully considered and thoughtfully addressed by the program (Breitenmoser et al. 2001). Despite low human population density and 80% coverage of tugay and reed ecosystems by Wildlife Refuges (IUCN Category IV), Ili-Balkhash region remains in use for livestock grazing (albeit with livestock density decreased by 70% since the collapse of Soviet Union), fishing and hunting by local residents (for details of current landuse see: WWF 2014); therefore, informed consent and collaboration from local communities must be obtained before initiating any tiger introduction program. Most importantly special measures to decrease risks of tiger predation on livestock and provide people safety must be planned in detail (IUCN/SSC 2013). Currently the tiger introduction program has strong support from Kazakhstan Government (Massimov 2010) as well as local communities due to the prospect of opportunities to improve their socio-economic situation in the region as a result of development of wildlife tourism, small business and associated employment opportunities at Ili-Balkhash Nature Reserve. Education and direct involvement by local peoples in the program as well as positive media coverage and functional compensation program for livestock owners has been critical for restoration of Florida panther population in the USA (Clucas et al. 2008). Restoration of wolf populations in Idaho, USA, was managed by indigenous communities and generated great cultural value for Nez Perce Indian tribe (Breitenmoser et al. 2001). While always complex, these case studies provide models for introducing tigers to Kazakhstan in a manner acceptable to local populations.

Conclusions Our assessment suggests that tiger introduction program in Central Asia could be considered a viable option and significant contribution to the Global Tiger Recovery Program (Global Tiger Recovery Program 2010) by following these steps (Wolf et al. 1998; Breitenmoser et al. 2001): the focus of restoration is on the core of the species’ historical range, that is, the Ili-Balkhash region is the central part of Caspian tiger range; habitat quality and quantity (~7 000 km²) are improved with prey restoration program well before any release of tigers; appropriate and effective programs are in place well in advance of tiger translocation to ensure social acceptability of the program; current water flow in the Ili River is assured through international agreement; some 25-50 Amur tigers are released to the introduction site over the next 50 years once prey populations are restored; careful monitoring of all project outcomes is undertaken to inform and adapt project operation.

Acknowledgments

This study was supported by the World Wide Fund for Nature (WWF Russia) (grant # WWF566) and NASA’s Land Cover Land Use Change Program (grant # NNX15AD42G).

References Abdukadir, A., and U. Breitenmoser. 2008. "The Last Tigers in Xinjiang." Cat News 47: 26-27. Baytop, T. 1974. "La présence du Vrai tigre, Panthera tigris (Linné, 1758) en Turquie." Säugetlerkdi 22: 254-256. Bragin, E.A. 2006. Review of Balkhash Lake conditions and politics in the field of water resource management in Ile-Balkhash watershed. Technical Report, WWF-Russia. Breitenmoser, U., C. Breitenmoser-Würsten, L.N. Carbyn, and S.M. Funk. 2001. "Assessment of carnivore reintroductions." In Carnivore Conservation, by J.L. Gittleman, S.M. Funk, D.W. Macdonald and R.K. Wayne, 241–281. Cambridge, U.K.: Cambridge University Press. Burchak Abramovich, N.I., and A.A. Mamedov. 1965. "On the issue of protection of some large predators Transcaucasia (tiger, leopard, and hyena) ." Trudy of the 3rd South Caucasus Conference on Protection of Nature. Tbilisi. 114-127. [in Russian] Can, Ö.E. 2004. Status, Conservation and Management of Large Carnivores in Turkey. Convention on the conservation of European wildlife and natural habitats. Standing Committee 24th meeting Report, Ankara: WWF-Turkey, 15-17. Carroll, C., and D.G. Miquelle. 2006. "Spatial viability analysis of Amur tiger Panthera tigris altaica in the Russian Far East: the role of protected areas and landscape matrix in population persistence." Journal of Applied Ecology 43: 1056–1068. Caughley, G. 1980. Analysis of Vertebrate Populations. New York: Wiley. Center for International Earth Science Information Network - CIESIN - Columbia University; Centro Internacional de Agricultura Tropical - CIAT. 2005. Gridded Population of the World, Version 3 (GPWv3): Population Density Grid. Palisades, NY: NASA Socioeconomic Data and Applications Center (SEDAC). Accessed February 24, 2016. http://dx.doi.org/10.7927/H4XK8CG2. Chernyshev, V.I. 1958. "Fauna and ecology of mammals of riparian forests of Tajikistan." Proceedings of the Institute of zoology and parasitology of the Academy of Sciences of Tajik SSR 85. [in Russian] Clucas, B., K. McHugh, and T. Caro. 2008. "Flagship species on covers of US conservation and nature magazines." Biodiversity and Conservation 17 (6): 1517-1528. Danilkin, A.A. 2006. "Wild Ungulates and Game Management". GEOS. Moscow: 366. [in Russian] Devineau, O., T. Shenk, G. White, P. Doherty, P. Lukacs, and R. Kahn. 2010. "Evaluating the lynx reintroduction programme in Colorado: patterns in mortality." Journal of Applied Ecology 47: 524-531. Driscoll, C. A., S. Luo, D. Macdonald, E. Dinerstein, I. Chestin, O. Pereladova, and S.J. O’Brien. 2011. "Restoring tigers to the Caspian region." Science 333 (6044): 822-823. Driscoll, C.A., I. Chestin, H. Jungius, O. Pereladova, Y. Darman, E. Dinerstein, J. Seidensticker, et al. 2012. "A postulate for tiger recovery: the case of the Caspian Tiger." Journal of Threatened Taxa 4 (6): 2637–2643. Driscoll, C.A., N. Yamaguchi, G.K. Bar-Gal, A.L. Roca, S. Luo, D.W. Macdonald, and S.J. O’Brien. 2009. "Mitochondrial phylogeography illuminates the origin of the extinct Caspian Tiger and its relationship to the Amur Tiger." PLoS ONE 4(1): e4125. doi:10.1371/journal.pone.0004125.

ESRI. 2014. "ArcGIS Desktop: Release 10.2.2." Redlands, California: Environmental Systems Research Institute. Flerov, K.K. 1935. "Prey mammals (Fissipedia) of Tajikistan." In Animals of Tajikistan, by B.S. Vinogradov, E.N. Pavlovskiy and K.K. Flerov, 131-200. Moscow: Academy of Sciences of USSR. [in Russian] GEF/UNEP/WWF. 2005. Econet Central Asia GIS Database. Moscow. https://www.wwf.ru/about/where_we_work/asia/closed/econet/maps. Geptner, V.G. 1949. "On the fierce beast from Sermon to the children by Monomakh." Okhota i okhotnichye khozyaistvo 5: 43-45. [in Russian] Geptner, V.G. 1969. "Tiger." In Mammals of the Soviet Union. Cats and Hyenas, by V.G. Geptner, N.P. Naumov and A.G. Bannikov. Moscow. [in Russian] Geptner, V.G., Nasimovich, A.A., Bannikov, A.G. 1961. "Wild Boar". In Mammals of the Soviet Union, Even-toed and Odd-toed Ungulates, by by V.G. Geptner and N.P. Naumov. Moscow: 45-46 [in Russian] Geptner, V.G., and A.A. Sludski. 1972. "Mammals of the Soviet Union." In Mammals of the Soviet Union, Carnivora (Hyaenas and Cats), by V.G. Geptner and N.P. Naumov. Moscow: Smithsonian Institution and the National Science Foundation, Washington, D.C. Global Tiger Recovery Program. 2010. Accessed August 2015, 2015. http://www.globaltigerinitiative.org. Goodrich, J.M., and D.G. Miquelle. 2005. "Translocation of Probelm Amur Tigers to Alleviate Tiger-Human Conflicts." Oryx 39: 1-4. Government of the Republic of Kazakhstan. 2010. Decree of the Government of the Republic of Kazakhstan #924 dated on September 10 2010 "On approval of Zhasyl Damu (Green Economy) Programme for 2010-2014". [in Russian] http://online.zakon.kz/document/?doc_id=30818245#pos=1;-264 Hajiyev, D.V. 2000. "Carnivorous." In The fauna of Azerbaijan, 552-588. Baku : Elm. [in Russian] Ignatova, N.K., N.K. Khristoforova, and N.A. Chaus. 2004. "Population dynamics of wild boar and roe deer in Wildlife Refuges and game management areas of South-West of Primorsky Kray." Electronic journal "Studied in Russia" 1149-1161. http://zhurnal.ape.relarn.ru/articles/2004/107.html . Institute of Geography of Kazakhstan Academy of Sciences. 2013. Preparation of socio-economic component of long-term Program for tiger reintroduction in the Ili River delta and southern shore of Balkhash Lake. Research Report, WWF-Russia. Jackson, P., 1993. The status of the tiger in 1993. Report to the CITES Animal Committee. CITES, Geneva (August 1993). IUCN/SSC. 2013. IUCN Guidelines for Reintroduction and Other Conservation Translocations. Version 1.0. Gland, Switzerland: IUCN/SSC Reintroduction Specialist Group. Johnson, W.E., D.P. Onorato, M.E. Roelke, E.D. Land, M. Cunningham, R.C. Belden, R. McBride, et al. 2010. "Genetic Restoration of Florida Panther." Science 329: 1641-1645. Jungius, H. 2010. Feasibility study on the possible restoration of the Caspian tiger in the Central Asia. Survey Report, WWF Russia. Karanth, K.U., J.D. Nichols, N. S. Kumar, and W.A., Hines, J.E. Link. 2004. "Tigers and their prey: Predicting carnivore densities from prey abundance." PNAS 101 (14): 4854–4858.

Kock, D. 1990. "Historical record of a tiger, Panthera tigris (Linnaeus, 1758) in Iraq." Zoology in the Middle East 4: 11-15. Lukarevskiy, V.S., and R.D. Baidavletov. 2010. About field survey of Ili River valley as a potentila site for tiger reintroduction. Moscow: WWF Russia, Survey Report. http://wwf.ru/about/where_we_work/asia/tiger/eng Mambetzhumaev, A.M., and M. Palvanijazov. 1968. "Ecology, distribution and practical importance of some cat’s species (Carnivore, Felidae) in Karakalpak ASSR." Zoological Journal 47 (3): 423-431. [in Russian] Massimov, K. 2010. Greetings of the Prime-Minister of the Republic of Kazakhstan to participants of the 6th Ministerial Conference on Environment and Development of Asia – Pacific Region. September 27-October 2 2010. Astana, Kazakhstan. http://wwf.ru/about/where_we_work/asia/tiger/eng Mayer, J. J. 2009. "Wild Pig Population Biology." In WildPigs: Biology, Damage, Control Techniques and Management, by J.J Mayer and I. L. Brisbin, 157-192. Aiken, South Carolina: Savannah River National Laboratory. Melis, C., M. Basille, I. Herfindal, J.D.C. Linnell, J. Odden, J-M. Gaillard, K-A. Hogda, and R. Andersen. 2010. "Roe deer population growth and lynx predation along a gradien of environmental productivity and climate in Norway." Ecoscience 17 (2): 166-174. Mittermeier, R. A., C. G. Mittermeier, T. M. Brooks, J. D. Pilgrim, W. R. Konstant, G. A. B. da Fonseca, and C. Kormos. 2003. "Wilderness and biodiversity conservation". PNAS 100 (18): 10309–10313. Ministry of Natural Resources and Environment of Russian Federation. 2010. Strategy for Amur Tiger Conservation in the Russian Federation. Moscow: WWF Russia. Miquelle, D.G., Y.M. Dunishenko, D.A. Zvyagintsev, A.A. Darensky, A.M. Golyb, V.V. Dolinin, V.G. Shvetx, et al. 2009. "A Monitoring Program for the Amur Tiger Twelve-year Report: 1998-2009." Survey Report, Vladivostok. Niethammer, J. 1966. "Zur Ernahrung des Sumpfluchses (Felis chaus Guldenstaedt, 1776) in Afghanistan." Zeitschr f. Saugentierk 31 (5): 393-394. Nilsen, E.B, Gaillard J-M., R. Andersen, J. Odden, D. Delorme, G. van Laere, and J.D.C. Linnell. 2009. "A slow life in hell or a fast life in heaven: demographic analyses of contrasting roe deer populations." Journal of Animal Ecology 78: 585–594. Nowell, K., and P. Jackson. 1996. Wild Cats. Status Survey and Conservation Action Plan. Gland, Switzerland: IUCN/SSC Cat Specialist Group. Paltsyn, M. Yu., J.P. Gibbs, L. Iegorova, and O. B. Preladova. 2015. Modeling of tiger and its prey populations in Balkhash Lake Region as the basis for adaptive management of the Tiger Reintroduction Program in Kazakhstan. WWF-Russia. http://wwf.ru/about/where_we_work/asia/tiger/eng Pereladova, O.B. 2013. "Restoration of Bukhara deer (Cervus elaphus bactrianus Lydd.) in Central Asia in 2000-2011." IUCN Deer Specialist Group News 19-30. Roelke, M.E., J.S. Martenson, and S.J. O'Brien. 1993. "The consequences of demographic reduction and genetic depletion in the endangered Florida panther." Current Biology 3 (6): 340–350. n.d. Save China's Tigers. Accessed April 28, 2016. http://savechinastigers.org/.

Seidensticker, J. 2010. "Saving wild tigers: A case study in biodiversity loss and challenges to be met for recovery beyond 2010." Integrative Zoology 5: 285-299. doi:10.1111/j.1749- 4877.2010.00214.x. Shukurov, G.S. 1958. "About Tiger in Kopetdag." Matherials of the Turkmenian Academy of Sciences 6: 110-111. [in Russian] Sludskiy, A.A. 1973. "Distribution and abundance of wild cats in the USSR." Game animals of Kazakhstan. Works of the Institute of Zoology of Kazakhstan 34: 5-106. [in Russian] Sludskiy, A.A. 1966. Global distribution and abundance of tigers. Vol. 26, in Game animals of Kazakhstan. Works of the Institute of the Zoology of Kazakhstan, 212-261. Almaty. [in Russian] Sludskiy, A.A. 1953. "Tiger in the USSR." Matherials of the Academy of Science of the Kazakh SSR, Biology Seria 8: 8-73. [in Russian] Stroganov, S.U. 1961. "Tiger in Central Asia." Journal of Hunting and Game Management 12: 22-24. [in Russian] Tian, Y., J. Wu, A.T. Smith, T. Wang, X. Kou, and J. Ge. 2011. "Population viability of Siberian tiger in changing landscape: going, going and gone?" Ecological Modeling 222: 3166- 3180. Vereshchagin, N.K. 1959. Mammals of Caucasus. History of the fauna. Moscow: Materials of the Academy of Sciences of the USSR. [in Russian] Wolf, C.M., T. Garland, and B. Griffith. 1998. "Predictors of avian and mammalian translocation success: reanalysis with phylogenetically independent contrasts." Biological Conservation 86 : 243-255. WWF. 2014. "Tiger Reintroduction Program in Kazakhstan." http://wwf.ru/about/where_we_work/asia/tiger/eng WWF. 2016. WWF Russia Annual Report 2015. Moscow: WWF Russia. Zaumyslova, O. Yu. 2005. "Ecology of wild boar in Sikhote-Alin Biosphere Reserve." In Tigers of Sikhote-Alin Nature Reserve: Ecology and Conservation. Vladivostok.

CHAPTER 5: SNOW LEOPARD CONSERVATION IN RUSSIA: LESSONS LEARNED OVER 15 YEARS

Abstract: Snow leopard (Panthera uncia) in Russia represents the most northern segment of the species’ range; total area of potential snow leopard habitat in Russia is about 60,000 km² of which only about 1/3 is occupied by likely no more than 70-90 snow leopards (1-2% of the species’ total wild population). Since 1998 the snow leopard has been a flagship species for biodiversity conservation programs in the Altai-Sayan Mountains generally focused on mitigation of two major threats for this species in Russia: poaching and retaliatory killing. Primary lessons learned during the implementation of two snow leopard-herder conflict mitigation and three anti-poaching projects were as follows: background analysis and results-based planning with involvement of local stakeholders is critical for project success; protection of livestock corrals from snow leopard is the easiest and the most effective way to decrease level of herder- snow leopard conflicts in the Russian part of the species’ range; anti-poaching activities for snow leopard protection should focus primarily on permanent snare removal in key snow leopard habitats until other conservation mechanisms are found; to be effective livelihood alternatives to poaching should be clearly linked to activities that actually increase snow leopard populations; financial incentives for snow leopard conservation can be an effective community-based tool for protecting this endangered species; on-line awareness campaigns can help to change government regulations in favor of snow leopard conservation. Most importantly, snow leopard conservation programs should continually incorporate learning from previous projects and share those lessons widely in order to be more effective both in Russia and across the species’ range.

Introduction Snow leopard (Panthera uncia) in Russia represents the most northern segment of the species’ range, where its current distribution is generally limited to the Altai-Sayan Ecoregion. The region is characterized by mountainous country spread over 1,000,000 km² located at the very center of Eurasia at the junction of Russia, Mongolia, China, and Kazakhstan (Poyarkov et al., 2002; Paltsyn et al., 2012; Ministry of Natural Resources and Environment of the Russian Federation, 2014). The total area of potential snow leopard habitat in Russia is about 60,000 km² (Poyarkov et al., 2002). Areas currently inhabited by snow leopards are much smaller, however, and do not exceed 20,000 km². These areas are home to likely no more than 70-90 snow leopards (1-2% of the species’ total wild population) (Paltsyn et al., 2012). The few stable snow leopard populations in Russia occupy no more than 12,000 km² in the mountains of Altai, Tyva, and Western and Eastern Sayan (Figure 1). Nearly all persisting snow leopard populations in Russia are associated with transboundary zones at the borders of Russia with Mongolia, China, and Kazakhstan. Details of current snow leopard distribution in Russia are presented by Paltsyn et al. (2012) and Ministry of Natural Resources and Environment of the Russian Federation (2014). Population genetics studies suggest that there is high demographic connectivity of the small and geographically isolated Russian snow leopard populations with a larger population nucleus of this species in western Mongolia (Rozhnov et al, 2011; Galsandorj et al., 2011).

Figure 1. Snow leopard habitat and known populations in Russia. Locations of snow leopard populations: 1 – South Altai Ridge 2 – Argut River Watershed 3 – Chikhachev Ridge and Mongun-Taiga Massif 4 – Tsagan-Shibetu and Shapshal Ridges 5 – Sayano-Shushensky Nature Reserve and its buffer zone 6 – Sengelen Ridge 7 – Tunkinsky Ridge

Since 1998 snow leopard has been a flagship species for the World Wide Fund for Nature (WWF) Biodiversity Conservation Program in the Altai-Sayan Ecoregion (WWF, 2012) and United Nations Development Program/Global Environmental Facility (UNDP/GEF) Project “Biodiversity Conservation in the Russian portion of the Altai-Sayan Ecoregion” (UNDP/GEF, 2005). To this end considerable financial resources (as well as other project budgets) have been dedicated to snow leopard conservation in Russia aimed at mitigation of two major threats for this species: poaching and retaliatory killing. About 90% of project efforts were implemented in key snow leopard habitats: Argut River watershed, Chikhachev Ridge, Tsagan-Shibetu and Shapshal Ridges, and Sayano-Shushensky Nature Reserve and its buffer zone (Figure 1). Over the last 15 years we have gained considerable experience in snow leopard conservation and we describe it here in the form of “lessons learned” from project implementation in these areas. In this chapter we emphasize lessons learned because sharing experiences and insights is an extremely valuable tool for advancing snow leopard conservation both regionally and range-wide yet, paradoxically, it is widely recognized that conservation practitioners generally do not share

84 their experiences in published form (Redford & Taber, 2000; Pullin et al. 2003; Sutherland et al. 2004; Sunderland et al., 2009). This results in a huge volume of collective conservation experience being lost for future projects. Moreover, unexpected outcomes and project failures regularly occur in snow leopard and other endangered species conservation projects given the complex socio- economic and uncertainty issues associated with the species’ conservation. Yet these are very rarely reported despite their high practical value for others working in similar situations (Sunderland et al. 2009). For these reasons we highlight a number of lessons we learned in Russia from both project successes and failures while working to mitigate snow leopard poaching and retaliatory killing between 2000-2014.

Snow leopard-herder conflict mitigation projects Insurance of livestock against snow leopard depredation in western Tyva. This project was implemented by WWF-Russia in cooperation with RESO-Garantia insurance company in three districts of Tyva Republic (Mongun-Taiga, Bai-Taiga, and Ovur) between 2000- 2003. The overall project goal was to protect the snow leopard population in western Tyva from retaliatory killings by decreasing conflicts between herders and snow leopards. This was attempted through a livestock insurance program and concurrent public awareness campaign. Local herders living in snow leopard habitats were provided the opportunity to insure their livestock against snow leopard predation. WWF paid full premium for insured livestock to RESO-Garantia. Every herder willing to insure their livestock received a simple camera with a 12-exposure roll of film to take pictures of livestock killed or wounded by snow leopards as well as special forms to record details of any attack. Each herder participating in the program was expected to report any cases of snow leopard depredation to the project coordinator as soon as possible. The project coordinator and expert zoologist were expected to investigate on-site every reported case of livestock depredation to make a determination if the livestock was killed by a snow leopard or not. Every case approved by the coordinator and expert had to be covered by RESO-Garantia in full. In turn, herders participating in the program had to tolerate snow leopards killing their livestock and not pursue and kill any of the cats that may have been involved (Chimed Ochir, 2003; Kuksin, 2003). As a result of this program a total of 54,841 head of livestock were insured in 2000, 21,105 in 2001, and 24,048 in 2002 (Kynyraa, 2001; Kuksin, 2003). Twenty reports of snow leopard attacks on livestock were collected in 2000-2003 claiming 233 livestock kills. Only seven cases were investigated and four cases (56 heads of livestock killed) were covered by insurance (Kuksin, 2003). In July 2003 the project was terminated. Factors that contributed to project success: • The project was the first initiative in Russia to mitigate conflicts between snow leopard and local herders. It was implemented at the right time – after collapse of the Soviet system of collective livestock management – when herders were left without any support or attention from the Russian government and relied only on their livestock to survive. Any livestock losses, including loss to wild predators, are very economically damaging to herders; therefore, the insurance program was accepted by local communities with great enthusiasm. People were excited that WWF understood their urgent needs and that the organization was willing to help them in a harsh economic situation after a decade of being ignored by their own government.

• Before the project, Tyvan herders like those elsewhere regarded snow leopards only as a pest and few if any knew the species to be globally endangered and in need of protection (Chimed Ochir, 2003; Jackson and Wangchuk, 2001; Kynyraa, 2001; Kuksin, 2003). This represented the first time in Tyvan history that the issues of snow leopard predation on livestock and necessity of leopard conservation were raised at the regional and national level and discussed in the mass media. Thanks to an information campaign, most Tyvan herders became willing to tolerate livestock losses due to snow leopard predation if these losses were compensated by insurance. • The project allowed collection of valuable information about locations and intensity of snow leopard predation on livestock as well as leopard distribution in western Tyva. These data helped design other snow leopard conservation projects in Tyva Republic. • The fact that four local herders received compensation for livestock killed by snow leopards inspired other herders to tolerate snow leopard attacks for two or three more years after the project was terminated. Herders continued to report snow leopard attacks until 2008 because they still hoped to receive compensation and still desisted from killing snow leopards involved in livestock depredation.

Factors contributing to project failure: • The insurance program was planned “top down”: the project concept was developed by WWF-Russia and RESO-Garantia without consultations with key stakeholders, especially local herders of western Tyva (Chimed Ochir, 2003). Moreover, the project did not have clear goal, objective, or success indicators, and therefore it was difficult to measure whether the insurance program reduced the level of retaliatory killing of snow leopards in the project area. • The project was based only on limited data about snow leopard distribution and predation on livestock collected in 1998-1999 by external experts during two or three short visits to western Tyva. Obviously, these data were insufficient to clearly define the project area. Thus, Ovur District, where not a single snow leopard attack on livestock was documented in 1998-2003 (Kuksin, 2003), was included in the project area in 2000-2002. Poor understanding of the spatial pattern of snow leopard attacks resulted in a huge project area (about 8,000 km²) spread over a remote mountain area mainly in Mongun-Taiga and Bai-Taiga Districts resulting in a very high number of insured livestock (up to 55,000 head in 2000) and considerable premiums paid to RESO-Garantia by WWF (e.g., total insurance premium of $US 61,866 in 2000) (Kynyraa, 2001). • Small technical issues generated problems. Specifically, herders were not trained in the proper use of cameras to record pictures of livestock killed by snow leopards and wounds on the carcasses of dead animals. Most images were blurred and inappropriate as evidence. Also, the herders had only one roll of 12-exposure film each to document losses, too little to collect the information needed. Moreover, most herders had only limited knowledge of the Russian language and could not record data on snow leopard attacks correctly on the forms provided (Kuksin, 2003). • Many herders with insured livestock lived in remote mountainous areas and had no means of remote communication. Therefore, it took them at least one full day on horseback to reach the district center and report snow leopard attacks to the project

coordinator, who lived 500 km away from the project area in the capital of Tyva Republic. At best, it required 2-3 days for the project coordinator and expert to reach the site of a snow leopard attack and make an expert assessment. In many cases 3-4 days was sufficient to enable scavengers (foxes, wolverines, mustelids, and domestic dogs) to damage animal carcasses and destroy all signs of snow leopard presence near dead livestock. No local experts were trained or hired by the project to help with documentation or to offer needed expertise. As a result only 7 cases from 20 reported were actually investigated by experts and only 4 cases were approved for insurance payments. • RESO-Garantia was only able to pay insurance coverage to herders via their bank accounts. This was impractical for local people because there were no bank branches in Mongun-Taiga District center. To receive coverage a herder had to travel 500-600 km to Kyzyl (capital of Tyva Republic), open a bank account, and then return yet again to actually withdraw money from the account. The considerable time and money required for impoverished herders to obtain compensation made the project’s utility dubious.

Protection of livestock corrals from snow leopards in western Tyva. After the conclusion of the livestock insurance project, WWF and UNDP-GEF Project “Biodiversity Conservation in the Russian portion of the Altai-Sayan Ecoregion” (2006-2011) invested significant funds in field surveys of snow leopard populations and analyzing the nature of snow leopard predation on livestock in western Tyva. We mapped all points of snow leopard attacks on livestock reported by local herders as well as nearly all herder camps in snow leopard habitats on Chikhachev, Mongun-Taiga, Shapshal and Tsagan-Shibetu Ridges. We learned that 57% of all attacks occurred on Tsagan-Shibetu and 36% on Shapshal Ridges. Also 94% of all cases of livestock depredation (n=104) occurred in open pastures and only 6% in corrals (n=6). However, just a few snow leopard attacks in corrals were responsible for 56% (n=260) of all livestock killed by the leopards in western Tyva in 2000-2011 (Paltsyn et al., 2012), a pattern similar to the situation in Ladakh, India (Jackson and Wangchuk 2001). Whereas a snow leopard only kills 1-3 animals in a single attack in open landscape, in a corral a snow leopard is capable of killing and wounding dozens of sheep and goats (up to 80 head), resulting in considerable losses for the herder (Paltsyn et al., 2012). Corrals also become traps for the snow leopard itself, as the cat is often not able to jump out again through the corral roof and is killed by the herder. Between 2000 and 2008 we learned of 6 cases of snow leopards being killed by herders during livestock attacks, 4 of which took place inside corrals. Tyvan herders usually tolerated small losses of livestock, but after catastrophic losses inside corrals the snow leopard involved was killed in the most cases. As a result of these investigations we developed the idea of covering corral ventilation holes in the roof and its windows with metal mesh to prevent snow leopards from entering. Doing so would prevent mass kills of sheep and goats that were often fatal for the leopards themselves. In October 2007 all winter corrals in 21 herder camps in Tsagan-Shibetu Ridge (a hot spot of snow leopard predation on livestock) were strengthened. In 2009 more corrals were improved in 25 other herder camps in snow leopard habitats in Shapshal, Mongun-Taiga and Chikhachev Ridges. More than 70 Tyvan herders participated in the project and learned about this method for corral protection from leopard attacks (Kuksin, Kuksina, 2009). Since 2008 not one instance has been recorded of

87 snow leopard attack on livestock inside corrals in western Tyva. This simple low-cost effort eliminated 50-60% of livestock loss due to snow leopard and decreased the loss to more or less tolerable level (no more than 20 heads of livestock killed annually) during the period of 2009-2014 (Figure 2). Also, this simple measure decreased snow leopard mortality simply because no unprotected corrals – deadly traps for snow leopard themselves – remained in the area. Moreover, the project was strongly supported by herders.

160 140 120 100 80 60 snow leopards snow 40 20

Number of livestock killed by killed Number by livestock of 1 2 0 1998 2000 2002 2004 2006 2008 2010 2012 2014

Figure 2. Number of livestock killed by snow leopards in western Tyva (Chikhachev Ridge, Mongun-Taiga Massif, Tsagan-Shibetu and Shapshal Ridges), Russia, in 2000-2013 based on reports by local herders. Three big spikes in 2002, 2006 and 2008 indicate massive killing of sheep and goats inside corrals. Green arrows: 1 – 21 corrals are reinforced with metal mesh in snow leopard habitat on Tsagan-Shibetu Ridge in 2007; 2 – 25 corrals are protected against snow leopards on Chikhachev, Mongun-Taiga and Shapshal Ridges in 2008. Project success factors: • The project was based on substantial information collected in 2000-2007 about snow leopard distribution, location and intensity of livestock predation events, and an understanding of grazing patterns in western Tyva. • The project had clearly defined goals, objectives, and measurable results. More specifically, the project output (46 protected corrals in the key snow leopard habitats in western Tyva) manifested as significant project outcomes: 50-60% decrease in the number of livestock killed by snow leopards and at least 50% decrease in mortality of snow leopards due to livestock predation in western Tyva. The expected project impact – increase of snow leopard population in western Tyva - was measurable as well, as local snow leopard populations on Chikhachev, Mongun-Taiga, and Tsagan-Shibetu Ridges have been monitored nearly every year since 2004. The snow leopard population in southwestern Tyva slightly increased since 2004. Thus, in 2004 density of snow leopard tracks on Tsagan-Shibetu Ridge was 3.6 tracks/100 km of routes walked and no cub tracks were found (Paltsyn, 2004); in 2010 – 4.6 tracks/100 km walked and 2 cub tracks (Kuksin, 2010); and in 2011 – 5.6 tracks/100 km walked and 3 cub tracks (Kuksin, 2011). On

Mongun-Taiga Massif only one sighting of snow leopard track was recorded in 2003- 2009 (Spitsyn, 2006), in 2010 – tracks of two snow leopards were found in this area (Paltsyn et al., 2012), and in 2011 – photographs of two snow leopards were taken during a camera-trap survey by the Severtsov Institute of Ecology and Evolution (Poyarkov, unpublished data). Of course, we cannot attribute this obvious increase of the snow leopard population only to the effect of reinforced corrals, but the management intervention likely contributed positively to snow leopard population trajectory. • Local herder communities and authorities had great interest in the project and were actively involved in project planning and implementation. The initiative was discussed and agreed upon with local authorities and herders. Herders actively participated in the transportation of metal mesh rolls and reinforcing of corrals, contributing horses, vehicles, and labor to the project. Local community capacity building to decrease livestock loss from snow leopards was part of the initiative. Lastly, every herder who participated in the project signed an agreement with Ubsunurskaya Kotlovina Nature Reserve pledging to protect snow leopards, and in 2012 some of the herders were involved in snow leopard camera-trapping. • The project was cheap and very simple logistically.

Issues unsolved by the project: • The project did not suggest any solutions for the protection of free-ranging livestock in mountain pastures. Thus, Tyvan herders still lose livestock from snow leopard attacks in western Tyva. The amount of loss is more or less tolerable (no more than 20 head of livestock annually 2009-2014), but it could increase with growing snow leopard numbers in the area in the future. An attempt to protect pastured livestock using electric fences in 2009 was unsuccessful due to considerable difficulties in setting up and maintaining long electric fences on the rugged and uneven terrain (Kuksin, 2009).

Lessons learned from anti-poaching projects Protecting snow leopards from poaching in the Argut River watershed, Altai Republic (2005- 2008). In December 2005 WWF-Russia and Game Management Department (GMD) of Russian Federal Agricultural Control Agency (Rosselkhoznadzor) launched the first anti-poaching campaign aimed at protection of the snow leopard population in Argut River watershed and the surrounding area. According to the 2002 “Strategy for Snow Leopard Conservation in Russia” (Poyarkov et al., 2002) this area was believed to have the largest snow leopard population in the country, totaling 30-40 leopards. WWF provided funding for 4 patrol groups of GMD (12 inspectors) for fuel, car parts, and field equipment to organize regular raids and anti-poaching patrols in Shavla Wildlife Refuge, a key protected area that encompasses key snow leopard habitats in the Argut River watershed. The total area is about 328,811 ha of rugged and remote mountainous terrain (Government of the Altai Republic, 2002). In 2006-2007 GMD groups patrolled the area every month (a level of protection 4-6 times higher than in 2004-2005) with active participation of the police. Intensive patrolling resulted in 30 poachers, mainly local hunters, being fined, and 25 firearms and 100 snares were confiscated in the refuge (in 2004-2005 only 5 poachers were

89 discovered and punished in the same area). At the end of 2007 the project was halted, because RosSelkhozNadzor lost its authority to fight poaching, and these rights were transferred to the newly established regional Wildlife Protection Committee of Altai Republic. It appears that the project contributed to a temporary and significant decrease in poaching in snow leopard habitats in the lower and middle parts of Argut River valley: in February 2004 we found 35-40 fresh signs of poaching in the area (snares, traps, skins of killed animals, etc.) but in March 2007 only 2 examples were found in the same area (Paltsyn, 2004; Paltsyn, 2007).

Factors contributing to project success: • GMD of Altai Republic was a federal agency with an adequate number of highly professional staff (about 35 inspectors) and relatively good salaries and per diem compensation. Every patrol group participating in the project had its own vehicle for raids. All they needed to work effectively was fuel, parts for vehicles, and equipment (supplies not provided sufficiently by the federal government). • Formal competition was established among inspector groups for better results measured in numbers of citations and confiscated firearms. The best groups received annual financial bonuses from RosSelkhozNadzor. These incentives stimulated the brigades to work hard to find more poachers and confiscate more firearms.

Project shortcomings: • The project was based on the false assumption that in 2005-2007 the Argut snow leopard population still had 30-40 individuals generally concentrated in the middle part of Argut River watershed where there is a high population density of Siberian ibex (one of the snow leopard’s main prey species) of up to 20 animals per 1 km² (Paltsyn et al., 2012). Detailed sign and camera-trap surveys in 2004-2011 did not find any signs of permanent snow leopard presence in that part of the watershed: camera-trapping in 2010-2011 proved that all signs and tracks of big cats found in the area belonged to Eurasian lynx (Lynx lynx) (Paltsyn et al., 2012). The only obvious presence of snow leopards was discovered in 2004 and 2012 in a very remote part of the upper Argut near Mount Belukha, where 3-5 snow leopards survived (Spitsyn, 2012). Thus, anti-poaching raids were conducted generally in an area where no or very few snow leopards remained by 2005-2007, and no patrolling occurred in the upper Argut watershed that still supported leopards. Therefore, despite considerable overall anti-poaching efforts, the project was of little help to the remaining Argut snow leopard population although it likely generally benefited area wildlife and perhaps will help snow leopards recolonize the middle-Argut in the future (see below). • About 90% of all cases of poaching discovered by anti-poaching brigades in Shavla Wildlife Refuge in 2005-2007 were located in easily accessible areas near roads or significant trails, but not in the remote habitat areas so important for snow leopards. Anti- poaching activities in remote habitat requires a lot of time and labour and usually results in a maximum of 1-3 poachers interdicted and fined and often none at all for 2 weeks of hard work, whereas patrolling relatively accessible areas may result in 10-12 discovered poaching cases just within a week. Thus, brigades achieve better results in anti-poaching activities (and competitions) by patrolling accessible places. In the view of the donor

agency, the more poachers found, the better. But in emphasizing this approach, anti- poaching brigades missed high concentrations of snares in more remote areas where snow leopards might still live. Also in such cases inspectors miss snare-poachers – a small segment of the poacher population that presents the most danger to snow leopards. From the inspector’s viewpoint it makes little sense to climb mountain slopes and remove snares because snare removal data was not tracked; instead they were paid on the basis of numbers of poachers fined and numbers of firearms confiscated. As a result, anti-poaching activities with seemingly good results offered little real benefit to snow leopards. • Conservationists in Altai-Sayan persistently underestimated snare poaching until it recently became evident that the Argut snow leopard population was devastated by snares and that the other significant population in the Sayano-Shushensky Nature Reserve decreased similarly for the same reason (Paltsyn et al., 2012). Thus, snare removal – a critical aspect of snow leopard conservation – was not part of the 2005-2007 anti-poaching campaign.

Snare removal campaigns in Argut River Watershed and Sayano-Shushensky Nature Reserve in 2008-2014. Snare poaching is the most serious and widespread threat contributing to the decline of at least 5 of 7 known stable snow leopard populations in the Altai-Sayan (Figure 3) (Paltsyn et al., 2012). Local residents, mainly herders and hunters who overwinter in snow leopard habitat, are the main poachers, killing cats and other species in snares. High prices for derivative products of snow leopards, musk deer, and other species are the main drivers of snaring. Animal parts are one of the very few income sources for local residents living in remote villages and herder camps near snow leopard habitat, places where unemployment levels can reach 80-90% (Paltsyn et al., 2012).

Figure 3. Snow leopard female with a wire snare around its neck in Sayano-Shushensky Nature Reserve. The cat was caught in a snare in spring 2013 when regular snare removal patrolling was terminated by the Reserve’s new administration. This is the final picture of this snow leopard recorded. In fall 2013 the cat and its three cubs disappeared. Photo by S. Istomov. Snare removal activities in snow leopard habitats in the Argut River watershed and Sayano- Shushensky Nature Reserve started in 2008-2009. Based on lessons learned from anti-poaching projects in 2005-2007 the snare removal campaign used different tactics than previously: • The geographic focus of the campaign was very narrow: 600 km² of optimal snow leopard habitat in Argut River watershed in 2008-2009 and 500 km² in Sayano-Shushensky Nature Reserve and its buffer zone. In 2012 after the locations of remaining snow leopards were identified in the Argut River watershed, the area of snare removal activities in this region shrank to just 300 km². So, only habitats currently inhabited by snow leopards were patrolled in order to increase the effectiveness of anti-poaching activities. • Patrol groups in both areas had not only inspectors but also included snow leopard researchers and local people as collaborators (in Argut) who knew the area very well and had good prior knowledge about snaring. • The groups were trained in snare removal techniques and focused on snare poachers instead of pursuing all types of illegal hunters. • Success was measured in meaningful ways: the number of snares found and removed and the number of snare poachers fined and held criminally liable, instead of the total number of poachers fined and weapons confiscated. • During snare removal, field groups remained in the habitat for at least two weeks and also collected data on snow leopard presence and number and prey species abundance – important ancillary data for measuring conservation success. Camera traps were deployed to monitor the species after the team left the field.

• Snare removal activities occurred during the key snaring period – November-March – when snaring for musk-deer, snow leopards, lynx, wolves, and other wildlife is most frequent (frozen rivers increase access and pelage is most valuable). Raid intensity increased from 2 patrols per snaring season in 2008-2009 to 6 in Argut and 4 per season in Sayano-Shushensky NR in 2010-2014.

As a result of the intensive snare removal campaign in Sayano-Shushensky Nature Reserve and its buffer zone, the number of snares in the snow leopard habitats decreased from more than 800 in 2008 to only 5 in 2012 (Sashko, 2012). Snow leopard numbers in this area were relatively stable in 2008-2012 and estimated at 7-9 individuals (Istomov, 2008; 2010; 2012). But in 2013-2014 snare removal activities were ceased completely in the Reserve when the new administration fired nearly all inspectors who had participated in these important activities and hired new inspectors without snare removal experience. Thus, in 2013-2014 no raids took place within snow leopard habitat. The absence of snare removal over one winter resulted in a dramatic increase in snare numbers to several hundred and an abrupt decrease in the local snow leopard population to only 2-3 individuals by spring 2014. In the Argut area the number of snares in the patrolled snow leopard habitats decreased from more than 600 in 2008 to 0 in winter 2013-2014. Since 2014, inspectors of the newly established Sailyugem National Park joined the snare removal patrols, removed snares, and interdicted poachers from the Argut cluster of the Park (about 800 km²), adjacent to our project area. Two snow leopards – Kryuk (male) and Vita (female) – found in Argut in 2012 successfully survived until the present and even produced 2 cubs in 2013. One more snow leopard was camera-trapped, and tracks of 2-3 other leopards were found in the upper part of Argut valley in 2013-2014. Therefore, no less than 7-8 snow leopards now dwell in the Argut River watershed and give hope for the restoration of a healthy population of 25-30 individuals within 10 years. The success of the Argut snare removal project was achieved not only by regular patrols in the limited project area, but was also due to the following factors:

• The project has been a long-term campaign with adequate and stable funding provided by WWF, UNDP/GEF Project, Panthera, US Fish and Wildlife Service, Weeden Foundation, Disney Worldwide Conservation Fund, The Altai Project, Snow Leopard Conservancy, State University of New York College of Environmental Science and Forestry, Snow Leopard Network Conservation Grants Program, M-Video Corporation, and private Russian donors. • In addition to snare removal work, our group deployed camera-traps to secure images of snow leopards. Local hunters did not know the locations of the cameras and many of them thought that the cameras had been hidden along trails to take pictures of poachers instead. Therefore, some poachers stopped snaring in the project area in a misguided attempt to avoid prosecution. Ironically, in winter 2014 camera-traps were actually used by Sailyugem National Park inspectors to identify and prosecute a poacher snaring within the Park. • To increase effectiveness of anti-poaching raids, a satellite-based poacher detection system developed by Wildlife Intel (2014) was deployed in two Argut cabins used by local hunters as bases for snaring. The system was triggered by heat sensor as soon as a fire was lighted

in a cabin’s stove. In response, an alert message was sent in real-time, through satellite messaging, to mount a response from patrol group of the Wildlife Protection Committee of Altai Republic (Wildlife Intel, 2014). In winter 2012-2013 the system revealed 3-4 visits of poachers to one of the cabins, but unfortunately none of the system alerts was responded to by the patrol group due to the very limited staff and resources of the Wildlife Protection Committee. Nevertheless, the systems helped to collect data about the frequency of poacher visits to the middle part of Argut River watershed. This information was used to direct anti- poaching raids by Sailyugem National Park staff in 2014. • Two professional snow leopard poachers living in the project area were contracted by the project team to protect and monitor snow leopards in the upper part of Argut River watershed in 2013. These local people voluntarily agreed to cooperate with our team on restoration of the Argut snow leopard population and to give up snaring. They signed a contract and received several camera-traps with the task of monitoring snow leopards and removing snares. As a result of their work, 3 more previously unknown snow leopards were identified and the key Argut leopard habitat remained snare-free in 2013-2014. The men completely lost their illegal income from snaring; in compensation they began receiving incentives for snow leopard conservation as well collecting horse rental fees from our monitoring team and per diem expenses for their participation in fieldwork. Currently their legal annual income is about twice their previous illegal income from snaring. To make this income possible WWF-Russia organized a special fundraising campaign and collected funding from private and corporate donors amounting to about $US 40,000 in 2014. Additionally support from Disney Worldwide Conservation Fund (via The Altai Project) has provided further incentives to local people for saving leopards (Arkhar NGO, 2014). Ex-poachers will receive their annual fee for conservation of snow leopards (about $US3,000 each annually) as long as they prove with camera-trap photographs that snow leopards still persist in the area they are responsible for watching over. At the same time these two ex-poachers were presented to the general public as protectors of Argut snow leopards and one of them, Mr. Mergen Markov, received a Disney Conservation Hero Award in 2014. These men are very proud of their new profile and livelihood obtained by helping to restore the snow leopard population in their homeland in the Argut watershed.

We believe that the ex-poacher/donor agreement we implemented offers a very promising and ethical tool for conservation and restoration of critical populations of other endangered species. In some sense it is similar to a payment for ecosystem services. It provides local and indigenous people living in remote areas and depending upon poaching and snaring a more attractive source of income and an immediate and profitable alternative to poaching. This alternative pathway allows people to use all their traditional knowledge and skills as hunters, does not require them to switch to new unusual or unknown activities to generate income, does not require the considerable time for development and promotion that small business would; and finally this tool can be used in very remote, small, and disadvantaged communities where small business development is simply not feasible. Moreover, in this case, the income of local ex-poachers is directly connected to the well-being of snow leopard and other endangered species in the project area. Camera trapping, with individual identification of snow leopards, offers a sound tool for verifying the program’s conservation targets. Of course, we do not suggest ex-poacher/donor agreements as an alternative to all alternative income generation and community-based wildlife management mechanisms, but they can be effective as a crisis tool where immediate action is required to save

94 an endangered population from extinction before other income alternatives for local community can be developed.

Development of alternative to poaching incomes for local communities in Altai Republic in 2010- 2014. Beginning in 2010, WWF and Citi Foundation implemented a small business development project for local communities living near snow leopard habitat and Protected Areas in Altai. The main goal of the project has been to decrease poaching by providing local communities with alternative sources of income like tourism, souvenir, and felt production as well as small-scale agriculture. To achieve this goal the project trained more than 1,000 local people in basic small business skills and provided financial support totaling $US 170,000 to 129 local entrepreneurs for launching small family enterprises in 2010-2014 (WWF-Russia, 2014). As a result of the project 230 new seasonal and permanent jobs for local and indigenous people were established in Altai (WWF-Russia, 2014). The project has had a noticeable positive socio-economic effect on local communities, but its value for snow leopard conservation is questionable due to following factors: • The project mainly targeted tourist enterprises (homestays, summer camps, souvenir production, excursions, and horse riding for visitors), but these are mainly summer activities in Altai. Therefore, in winter, when poaching is most viable, local people have no occupation or income and may thus continue their snaring and other poaching activities. • Small and very remote communities, like Argut village, are not promising for the development of ecotourism or souvenir production because they are rarely visited by any tourists and have no facilities for visitors. It may take many years to develop any kind of viable tourism business in such remote places. • None of the known snow leopard poachers have been involved in the project. Snaring poachers live in remote places, are not aware of alternative income projects, and typically have no interest in such projects as a source of income. Often they follow a traditional herder and hunter lifestyle and earn greater income from illegal wildlife trade thanks to very low levels of government control of poaching in Altai. Finally, locals may not be eager to switch livelihoods, especially when there are few comparable economic incentives to do so. • There is no direct link between support for project enterprises that benefit local people and the well-being of snow leopards. For example, tourism in mountain areas of Russia does not need the presence of snow leopards as a guarantee for income generation. Presence of snow leopards certainly augments tourist experience but is not a requisite. Thus, local communities do not consider snow leopards as a necessary economic resource for income generation nor do snow leopards necessarily benefit from such projects within areas they occupy.

On-line public awareness campaign to prohibit legal snare poaching in snow leopard habitat. Before 2013 all snare hunting was completely illegal in Russia. But in December 2013 Ministry of Natural Resources and Environment (MNRE) of the Russian Federation adopted decree #581

“On changes in hunting regulations” that allowed snaring for wolves and hare in some regions of Russia including Krasnoyarsky Krai, where the most northern Russian snow leopard population survives in and around Sayano-Shushensky Nature Reserve. This legal amendment practically legalized snaring in critical snow leopard habitat in Krasnoyarsky Krai adjacent with Sayano- Shushensky Nature Reserve. Snow leopard expert A. Poyarkov wrote a petition in opposition with this decree to MNRE and Russian President Vladimir Putin with an appeal to completely ban snaring in the habitats of snow leopard and Amur tiger. The petition was published on the public web-site www.change.ru and was signed by 88,000 Russian citizens in just two months. In August 2014 in response to the petition MNRE prohibited snaring in the habitats of snow leopard, Amur tiger, Siberian musk deer, and roe deer in Krasnoyarsky, Khabarovsky and Zabaikalsky Krais (Ministry of Natural Resources and Environment, 2014). This positive experience clearly demonstrates that public awareness campaigns can be very effective for improvement of legislation in the field of snow leopard and other endangered species protection. On the other hand, given extremely limited funding and low level of wildlife law enforcement in Russia we cannot assume these legislative changes will result in any positive outcomes for snow leopards and other species vulnerable to snaring.

Conclusions Conservation of snow leopard in Russia and other parts of Central Asia is as challenging as it is everywhere but particularly so due to remoteness and limited accessibility of the snow leopards’ habitats, extreme poverty of local communities, the low economic development of most snow leopard range countries, limited funding options for the conservation of this species, and lack of government enforcement of existing wildlife protection laws (Snow Leopard Secretariat, 2013). Often snow leopard conservationists must act despite very limited background information on snow leopard distribution and number, intensity of different threats, and the local socio-economic and political situation. In such conditions success is impossible without trial and error and lessons taken from the shortcomings and even outright failures of other projects. Lessons learned from conservation projects in different parts of the snow leopard’s range can have tremendous value for the rapid improvement of overall effectiveness in snow leopard protection. The value of learning conservation lessons increases greatly within the context of poorly understood environmental and socio-economic change caused by global climate change (Cox and Stephenson, 2007). Therefore, we highlight again that planning and implementation of snow leopard and any other endangered species program should incorporate learning from previous projects and ongoing activities both regionally and across the species’ range. Such strategic planning facilitates conservation programs by clarifying their long-term vision and questioning key assumptions (hypotheses), developing effective threat-specific objectives and tightly-targeted activities, and then measuring their success (Pressey et al., 2007; WWF, 2012a). This then enables such programs to adapt over time based on lessons learned – to practice adaptive management in the face of uncertainty of natural systems and processes – and thereby make conservation more effective and efficient in the long run (Ludwig et al., 1993; Parma et al., 1998; Gunderson, 2008; Allen et al., 2011). We hope that our lessons learned in Russia over the last 15 years and shared herein will encourage other snow leopard conservation practitioners to disseminate their valuable experiences

96 broadly to enable collective work toward more effective conservation measures to secure a sustainable future for this charismatic wild cat.

References:

Allen, C.R., J.J. Fontaine, K.L. Pope, and A.S. Garmestani. 2011. Adaptive management for a turbulent future. Journal of Environmental Management 92 (5), 1339-1345. doi:dx.doi.org/10.1016/j.jenvman.2010.11.019. Arkhar NGO, 2014. “Sponsor a Snow Leopard!” Program. Retrieved from http://www.arkhar.org/our-programs/snow-leopard/ Chimed Ochir, B., 2003. Livestock insurance campaign against snow leopard predation on livestock in the Republic of Tyva. Evaluation Report. WWF-Russia. Cox, P., Stephenson, D., 2007. A changing climate for prediction. Science 317, 207-208. Galsandorj, N., Munkhtsog, B., Janecka, J., 2011. Population structure and genetic diversity of snow leopards in Mongolia and implications for conservation. Final Report, Snow Leopard Conservation Grant Program. Government of the Altai Republic, 2002. Постановление Правительства Республики Алтай «О восстановлении Шавлинского государственного биологического заказника» № 242 от 22.08.2002 г. [Order of the Government of Altai Republic #242 “On re-establishment of Shavla Wildlife Refuge” dated on August 22nd 2002] Gunderson, L., 2008. Adaptive Management and Integrative Assessments. In Encyclopedia of Ecology, by S.E. Jørgensen and B.D. (editors) Fath, 55-59. Oxford: Academic Press. doi:dx.doi.org/10.1016/B978-008045405-4.00635-2. Istomov, S.V., 2008. Отчет рабочей группы о результатах полевых работ “Оценка численности снежного барса в Саяно-Шушенском заповеднике и его охранной зоны” [Field survey report “Evaluating the snow leopard population in Sayano-Shushensky Nature Reserve and its buffer zone”]. UNDP/GEF in the Altai-Sayan Ecoregion. Archive of the Association of Nature Reserves and National Parks in the Altai-Sayan Ecoregion. Istomov, S.V., 2010. Отчет “Оценка численности снежного барса в Саяно-Шушенском заповеднике и его охранной зоны” [Report “Results of the winter snow leopard survey in Sayano-Shushensky Nature Reserve and its buffer zone in 2010”]. WWF Archive. Istomov, S.V., 2012. Отчет “Обследование местообитаний снежного барса в южной части Саяно-Шушенского заповедника и его охранной зоны” [“Study of snow leopard habitats in southern Sayano-Shushensky Nature Reserve and its buffer zone”]. WWF Archive. Jackson, R. and R. Wangchuk. 2001. Linking snow leopard conservation and people-wildlife conflict resolution: grassroots measures to protect the endangered snow leopard from herder retribution. Endangered Species UPDATE 18(4), 138-141. Kuksin, A.N., 2003. Отчет о результатах проекта по страхованию скота от нападений ирбиса в Западной Туве [Report on the results of a project to protect livestock from snow leopard attacks in western Tuva]. WWF Archive. Kuksin, A.N., 2009. Окончательный отчет по проекту «Организация семинаров для скотоводов Западной и Юго-Восточной Тувы по укреплению загонов и использованию электроизгородей для защиты скота от нападения снежного барса» [Final report on the project “Organization of seminars for herders of western and south-

eastern Tuva on improvement of corrals and use of electric fences to protect livestock from snow leopard attacks]. UNDP/GEF Project “Biodiversity Conservation in the Russian portion of Altai-Sayan Ecoregion. Kuksin, A.N., Kuksina, D.K., 2009. Укрепление кошар от проникновения ирбиса [Protection of livestock corrals from snow leopards]. UNDP/GEF Project “Biodiversity Conservation in the Russian portion of Altai-Sayan Ecoregion. Krasnoyarsk. Kuksin, A.N., 2010. Ubsunurskaya Kotlovina Nature Reserve Report on a WWF grant “Разработка рекомендаций для сохранения снежного барса на хр. Цаган-Шибету, Юго-Западная Тува” [Developing recommendations for snow leopard conservation on Tsagan-Shibetu Ridge in southwestern Tuva]. WWF Archive. Kuksin, A.N., 2011. Ubsunurskaya Kotlovina Nature Reserve report on the WWF grant Разработка рекомендаций по сохранению ирбиса в южной части Шапшальского хребта и на хр. Цаган-Щибету [“Development of recommendations for snow leopard conservation on southern Shapshalsky Ridge and Tsagan-Shibetu Ridge”]. WWF Archive. Kynyraa, M.M., 2001. Oтчет о деятельности проекта ВВФ RU 0074.01 по добровольному страхованию домашних животных от гибели или вынужденного убоя явившихся результатом нападения ирбиса (снежного барса) [Report on activities of WWF RU 0074.01 project in the framework of voluntary insurance of livestock from snow leopard predation]. WWF Archive. Ludwig, D.R., R. Hilborn, Walters, C.J., 1993.Uncertainty, resource exploitation, and conservation: lessons from history. Science 260, 17-36. Ministry of Natural Resources and Environment of the Russian Federation, 2014. Стратегия сохранения снежного барса в Российской Федерации [Strategy for snow leopard conservation in the Russian Federation]. Approved by the Ministry of Natural Resources and Environment of the Russian Federation, Decree of August 18, 2014, # 23-r. Retrieved from http://wwf.ru/resources/news/article/12782 Ministry of Natural Resources and Environment, 2014a. Минприроды России инициирует введение запрета петельного лова волка в районах обитания снежного барса и амурского тигра [Ministry of Natural Resources and Environment of Russia will initiate a ban of wolf snaring in the habitats of snow leopard and Amur tiger]. Retrieved from http://www.mnr.gov.ru/news/detail.php?ID=134933 Paltsyn, M.Y., 2004. Отчет Алтайской рабочей группы о результатах полевых работ в бассейне р. Аргут в феврале 2004 г.[Altai working group report on results of field survey in the Argut River Watershed, February 2004]. WWF Archive. Paltsyn, M.Y., 2004. Akhar NGO report “Организация антибраконьерского рейда и сбора информации для выявления основных локалитетов снежного барса (Uncia uncia Schreber) в пределах хр. Цаган-Шибэту. Разработка конкретных проектов по сохранению ирбиса в Юго-Западной Туве” [Work in western Tuva as part of the ‘Organizing antipoaching patrols and information gathering to identify the main snow leopard (Uncia uncia Schreber) localities near Tsagan-Shibetu Ridge area’ project. Developing concrete snow leopard conservation projects in southwestern Tuva]. WWF Archive. Paltsyn, M.Yu., S.V. Spitsyn, Kuksin A.N., Istomov S.V., 2012. Conservation of snow leopard in Russia. WWF, Krasnoyarsk. Retrieved from http://wwf.ru/resources/publ/book/eng/599

Parma, A., P. Amarasekare, M. Mangel, J. Moore, W. W. Murdoch, E. Noonburg, Pascual, M. A., 1998. What can adaptive management do for our fish, forests, food, and biodiversity? Integrative Biology 1, 16–26. Poyarkov, A.D., Lukarevsky, V.S., Subbotin, A.E., Zavatsky, B.P., Prokofyev, S.M., Kelberg, G.V., Malkov, N.P., 2002. Стратегия сохранения снежного барса (ирбиса) в России [Strategy for snow leopard conservation in Russia]. WWF. Moscow. Retrieved from http://wwf.ru/resources/publ/book/eng/7 Pressey, R., M. Cabeza, M. Watts, Cowling R.M., Wilson, K.A., 2007. Conservation planning in a changing world. Trends in Ecology & Evolution 22 (11), 583-592. Pullin, A. S., Knight, T.M., Stone, D. A., Charman, K., 2003. Do conservation managers use scientiﬁc evidence to support their decision making? Biological Conservation, 119, 245–252. Redford, K., Taber, A., 2000. Writing the wrongs: Developing a safe-fail culture in conservation. Conservation Biology, 14, 1567–1568. Rozhnov V.V., Zvychainaya E.Y., Kuksin A.N., Poyarkov A.D. 2011. Неинвазивный молекулярно-генетический анализ в исследовании экологии снежного барса: проблемы и перспективы. [Noninvasive Molecular Genetic Analysis in Studying the Ecology of Snow Leopard: Problems and Prospects. Russian Journal of Ecology, 42, 439- 444. DOI 10.1134/S1067413611060142. Sashko, V.K., 2012. Отчет о выполнении гранта WWF245/9Z1428 «Предотвращение нелегального петлевого промысла в местообитаниях западносаянской группировки снежного барса силами объединенной оперативной группы «Ирбис»» [Grant report WWF245/9Z1428 “Prevention of illegal snare poaching in the habitat of Western Sayan snow leopard population by joint Irbis anti-poaching brigade”]. WWF Archive. Snow Leopard Secretariat, 2013. Global Snow Leopard and Ecosystem Protection Program . Bishkek, Kyrgyz Republic. Spitsyn, S.V., 2006. “Отчет о научно-исследовательской работе по теме ‘Ведение мониторинга видов животных, внесенных in Красную книгу РФ на территории Республики Тыва. Снежный барс. [“Report on scientific research on the topic of ‘Monitoring species listed in the Russian Federation Red Book in Tuva Republic’. Snow Leopard”]. Ubsunurskaya Kotlovina Nature Reserve Archive. Spitsyn, S.V., 2012. Отчет по гранту WWF 9Z1428/RU007408 «Разработка комплекса рекомендаций по восстановлению популяционной группировки ирбиса в бассейне р. Аргут» [Report on the grant WWF 9Z1428/RU007408 “Development of recommendations for restoration of snow leopard sub-population in the Argut River watershed’]. WWF Archive. Sunderland, T., Sunderland-Groves, J., Shanley, P., & Campbell, B., 2009. Bridging the gap: How can information access and exchange between conservation biologists and field practitioners be improved for better conservation outcomes. Biotropica, 41(5), 549–554. Retrieved from http://onlinelibrary.wiley.com/doi/10.1111/j.1744-7429.2009.00557.x/pdf. Sutherland, W., Pullin, A.S., Dolman, P.M., Knight, T. M., 2004. The need for evidence-based conservation. Trends in Ecology&Evolution, 19, 305–308. UNDP/GEF, 2005. Biodiversity Conservation in the Russian Portion of the Altai-Sayan Ecoregion – Phase I. Full size project document. Retrieved from http://www.thegef.org/gef/project_detail?projID=1177 Wildlife Intel, 2014. Cutting-edge technologies for monitoring trespass in remote areas. Retrieved from http://wildlifeintel.com/index.php/how-it-works/how-it-works

WWF, 2012. Altai-Sayan Ecoregional Conservation Strategy. Retrieved from http://wwf.ru/resources/publ/book/eng/843 WWF, 2012a. WWF Standards of Conservation Project and Programme Management (PPMS). Practical Guide. WWF-Russia, 2014. Sustainable Small Business Development in the Altai helps to conserve the nature values. Retrieved from http://wwf.ru/resources/news/article/eng/12620

100

CONCLUSIONS

In five chapters of this thesis I attempted to demonstrate how science can be applied to benefit regional natural resource management and endangered species conservation programs in Central Asia directly addressing sustainable management or conservation needs and providing relatively simple tools and guidance for field managers in locally acceptable form. My objective was to demonstrate how science can be directly integrated in the regional programs and projects on all stages, including design, implementation, monitoring and evaluation and can be easily used by the project field staff. Remote-sensing is one particularly important tool. Monthly Aqua and Terra NDVI and EVI datasets that are readily available and that do not require complex processing can be applied by rangeland managers and decision makers of western Mongolia as a tool for monitoring and predicting of summer pasture conditions (on the base of percentage of vegetation cover) at the regional level. Vegetation cover is the most frequently used indicator in remote sensing to measure grassland production and level of pasture degradation (Liu et al. 2004; Li et al. 2010; Cui et al. 2012) because it strongly reflects the ecological value of grasslands and has strong positive correlation with the above ground biomass in rangelands (Guo et al. 2006; Li et al. 2006; Al-Bakri and Abu-Zanat 2007; Cui et al. 2012; Eckert and Engesser 2013; Yan and Lu 2015). Vegetation cover demonstrates the most reliable response to grazing and precipitation in Mongolian rangelands (Fernandez-Gimenez and Allen-Diaz 1999) and is one of the key traditional indicators of pasture quality used by Mongolian herders (Fernandez-Gimenez 1997; 2000). Also as I demonstrated in the Chapter 2, herder-derived estimates of pasture forage value at the peak of growing season in western Mongolia is highly correlated with vegetation cover and NDVI and can be used as additional tool to complement widely used satellite-based methods for assessment of grasslands conditions at local level along with satellite-derived vegetation indices. Together these studies suggest that decision makers in western Mongolia responsible for environmental monitoring and grassland management could benefit from incorporating simple vegetation cover analysis based on MODIS VIs in their decision-making process that will be consistent with traditional herders’ knowledge and practices. Species distribution modeling represents another such tool. Snow leopard and Altai argali habitat projections under changing climate in the Altai-Sayan Ecoregion provide valuable insights for development of climate-smart conservation strategies for both species that are considered as flagships for the Altai-Sayan (WWF 2012). Such outputs are readily available for integration in the World Wide Fund for Nature’s (WWF) Altai-Sayan Ecoregional Strategy (WWF 2012), government species conservation strategies for snow leopard and Altai argali (Paltsyn et al., 2011; 2015) and regional climate adaptation plans. The most obvious conservation recommendations for snow leopard conservation in the Altai-Sayan Ecoregion in the context of climate change may be maintaining of the key species sub-populations located mainly in Mongolia and transboundary zone of Mongolia and Russia, that are likely to shrink but still persist even at the RCP 8.5 scenario assuming habitat connectivity remains between them. More conservation efforts should be focused on the snow leopard sub-population in Sengelen Ridge currently declining due to poaching

101

(Paltsyn et al. 2015) that is likely to be located in a significant area of the snow leopard habitat that may be sustained in the face of climate change and perhaps even expand. Species distribution models predict alarming outcomes for Altai argali. Nevertheless, key known Altai argali sub-populations in transboundary area of Russia and Mongolia, Turgen Mountains and central Mongolian Altai Ridge are likely to persist at the relatively mild climate change scenarios (RCP 2.6 and 4.5) and should be a continuing long-term conservation focus of on-going government and non-government conservation programs (Paltsyn et al. 2011, 2015; WWF 2012). Special attention should be given to conserve Gobi argali populations in the Gobi Altai and maintain connectivity between Gobi Altai and southern part of Mongolian Altai as a migration route for Gobi argali to the north in the former habitat of Altai argali that are likely to become suitable for the Gobi sub-species as a result of possible climate induced conversion of alpine ecosystems in the dry steppe and semi-desert biomes. Combining many widely available geographic data layers with the relatively simple practice of Maxent projections of target species habitat can be easily incorporated in the WWF’s Standards of Conservation Program and Project Management, or PPMS (WWF 2012) and climate-smart conservation planning guide “Climate Adaptation: Mainstreaming in existing Conservation Plans” (Morrison and Lombana 2011). Doing so represents a scientifically robust alternative to “ecological drawings” currently recommended by the last document as a way to assess likely climate change consequences for species and ecosystems. Integrated population and habitat modeling is yet another tool for bringing applied science to conservation decision making in central Asia. Modeling of tiger habitat and population under different management scenarios provides robust basis for adaptive management of the WWF’s Tiger Reintroduction Program (WWF 2014) using Amur tiger as analog species to replace extinct Caspian tiger subspecies in Central Asia. Spatial analyses based on remote sensing data indicated that options for Amur tiger introduction are limited in Central Asia but at least two habitat patches remain potentially suitable for tiger re-establishment, both in Kazakhstan, with a total area of <20,000 km². The most promising site—the Ili river delta and adjacent southern coast of Balkhash Lake—hosts ca. 7,000 km² of suitable habitat that our tiger-prey population models suggest could support a population of 64-98 tigers within 50 years if 40-55 tigers are translocated and current Ili river flow regimes are maintained. Re-establishment of tigers in Central Asia may yet be tenable if concerns of local communities in the Ili-Balkhash region are carefully addressed, prey population restoration precedes tiger introduction, Ili river water supplies remain stable, and the Amur tiger’s phenotype proves adaptable to the arid conditions of the introduction site. Publication of this thesis chapter as an article in Biological Conservation actually catalyzed the start of the tiger restoration program in 2017: the Memorandum of Understanding between WWF and Government of Kazakhstan for realization of the tiger restoration program was signed in September 2017 after the article publication in the Biological Conservation https://new.wwf.ru/en/resources/news/bioraznoobrazie/wwf-i-minselkhoz-obsudili-pervye- prakticheskie-shagi-po-realizatsii-programmy-reintroduktsii-tigra-v/ . Establishment of the Ili- Balkhash Nature Reserve (IUCN Category II, 1 000 000 ha) that will cover 100% of tugay and reed ecosystems in the Ili-Balkhash region to ensure restoration of tiger habitat and prey base is planned as the first step of the program.

102

The real challenge for conservation is simultaneously integrating scientific perspectives with on- the-ground practices often working with among the poorest people on Earth. My final chapter represents such an integration of robust science and practical conservation experience to inform adaptive management of an endangered species conservation projects. Conservation of snow leopard and many other endangered species is challenging due to remoteness and limited accessibility of the habitats, extreme poverty of local communities, limited economic development of most snow leopard range countries, limited funding options for the conservation of this species, and lack of government enforcement of existing wildlife protection laws (Snow Leopard Secretariat, 2013). Often species conservationists must act despite very limited background information on species distribution and number, intensity of different threats, and the local socio- economic and political situation. In such conditions success is impossible without adaptive management approach and lessons taken from the shortcomings and even outright failures of other projects. Lessons learned from conservation projects in different parts of the world can have tremendous value for the rapid improvement of overall effectiveness of conservation practice. The value of learning conservation lessons increases greatly within the context of poorly understood environmental and socio- economic change caused by global climate change (Cox and Stephenson, 2007). Therefore, I highlight that planning and implementation of any endangered species program should systematically incorporate learning from previous projects and ongoing activities both regionally and across the species’ range. Such strategic planning facilitates conservation programs by clarifying their long-term vision and questioning key assumptions (hypotheses), developing effective threat-specific objectives and tightly-targeted activities, and then measuring their success (Pressey et al., 2007; WWF, 2012a). This then enables such programs to adapt over time based on lessons learned – to practice adaptive management in the face of uncertainty of natural systems and processes – and thereby make conservation more effective and efficient in the long run (Ludwig et al., 1993; Parma et al., 1998; Gunderson, 2008; Allen et al., 2011). It is no secret that global conservation is chronically underfunded and losing the battle for biodiversity: according to the last WWF’s report populations of vertebrate animals have been reduced to a quarter of their abundance over the last 40 years (WWF 2016). Our current conservation approaches are often not effective enough to make a difference and achieve ambitious global conservation goals, often because of the “great divide” between conservation biologists and conservation managers (Redford & Taber 2000; Pullin et al. 2003; Sutherland et al. 2004; Sunderland et al. 2009; Shackleton et al. 2009). And yet success is possible, as indicated by snow leopard conservation programs outlined herein, and, we hope climate-smart adaption and species translocation programs also outline here that are in too early stages to fully evaluate as yet. It is my clear opinion that the key solution of the problem is to incorporate advanced science, creativity, innovations and effective learning system in the global conservation practice and to increase investments in the global conservation via mutually beneficial collaboration with research centers, business, multilateral agencies, governments, and public. Learning is the most difficult, but the most important, part of the conservation process. Learning brings difference to conservation as currently practiced. And this difference brings success.

103

APPENDICES

APPENDIX 1. SUPPLEMENTARY MATERIALS FOR CHAPTER 3

a. Snow leopard b. Altai argali Figure 1. Snow Leopard and Altai argali occurrence records collected in the Altai-Sayan Ecoregion in 1998-2016. Snow leopard Altai argali

g. Current h. Current

104

i. CCSM4 RCP 2.6 j. CCSM4 RCP 2.6

k. CCSM4 RCP 4.5 l. CCSM4 RCP 4.5

m. CCSM4 RCP 6.0 n. CCSM4 RCP 6.0

105

o. CCSM4 RCP 8.5 p. CCSM4 RCP 8.5 Figure 2. Predicted snow leopard and Altai argali habitat (yellow), forest cover (green) and desert cover (pink) models for current conditions (a and b) and different RCP scenarios (c-j) for CCSM4 GCM in the Altai-Sayan Ecoregion.

Snow leopard Altai argali

a. Current b. Current

c. HadGEM2-ES RCP 2.6 d. HadGEM2-ES RCP 2.6

106

e. HadGEM2-ES RCP 4.5 f. HadGEM2-ES RCP 4.5

g. HadGEM2-ES RCP 6.0 h. HadGEM2-ES RCP 6.0

i. HadGEM2-ES RCP 8.5 j. HadGEM2-ES RCP 8.5 Figure 3. Predicted snow leopard and Altai argali habitat (yellow), forest cover (green) and desert cover (pink) models for current conditions (a and b) and different RCP scenarios (c-j) for HadGEM2-ES GCM in the Altai-Sayan Ecoregion.

107

Snow leopard Altai argali

a. Current b. Current

c. MIROC-ESM RCP 2.6 d. MIROC-ESM RCP 2.6

e. MIROC-ESM RCP 4.5 f. MIROC-ESM RCP 4.5

108

g. MIROC-ESM RCP 6.0 h. MIROC-ESM RCP 6.0

i. MIROC-ESM RCP 8.5 j. MIROC-ESM RCP 8.5 Figure 4. Predicted snow leopard and Altai argali habitat (yellow), forest cover (green) and desert cover (pink) models for current conditions (a and b) and different RCP scenarios (c-j) for MIROC- ESM GCM in the Altai-Sayan Ecoregion.

Snow leopard Altai argali

18 18

16 16

14 14 Bio10,ºC 12 Bio10,ºC 12

10 10 Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Climate change scenarios Climate change scenarios a. Mean temperature of the warmest b. Mean temperature of the warmest quarter quarter

109

-15 -11 -13 -17 -15 -19

-17

Bio11,ºC Bio11,ºC -21 -19 -23 -21 Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Climate change scenarios Climate change scenarios c. Mean temperature of the coldest d. Mean temperature of the coldest quarter quarter 300 230 290 220 280 210

270 Bio12, mm 260 Bio12, mm 200

250 190 Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Climate change scenarios Climate change scenarios e. Mean annual precipitation f. Mean annual precipitation 21 15

19 14 13 17

12 Bio19, mm 15 Bio19, mm 11

13 10 Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Climate change scenarios Climate change scenarios g. Mean precipitation of the coldest h. Mean precipitation of the coldest quarter quarter Figure 5. Changes of the mean temperature of the warmest quarter (Bio10), mean temperature of the coldest quarter (Bio11), annual precipitation (Bio12), and precipitation in the coldest quarter (Bio19) in the current snow leopard (a, c, e, and g) and Altai argali (b, d, f, and h) habitats in 2060- 2080 predicted by CCSM4 (green), HadGEM2-ES (blue), and MIROC-ESM (red) GCMs under RCP 2.6-8.5 scenarios for the Altai-Sayan Ecoregion.

110

4500 2800

4000 2600

3500 2400

3000 2200

distribution, m.a.s.l distribution, distribution, m.a.s.l. distribution,

Maximal the of elevation forest Maximal 2500 2000 Maximal the of elevation desert Maximal Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Climate change scenarios Climate change scenarios a. Forest upper boundary b. Desert upper boundary

4400 max max 4000 3400

3000 mean 2400 mean

2000 Elevation, m.a.s.l Elevation, Elevation, m.a.s.l Elevation, 1400 min min 1000 400 Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Current RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Climate cahnge scenarios Climate change scenarios c. Altai argali elevation range d. Snow leopard elevation range Figure 6. Projected changes in the forest upper boundary (a) and elevation range of Altai argali (b) and snow leopard (c) in 2060-2080 predicted by CCSM4 (green), HadGEM2-ES (blue), and MIROC-ESM (red) GCMs under RCP 2.6-8.5 scenarios.

Snow leopard Altai argali

120000 Loss 120000 Gain Loss 90000 90000 No changes No changes 60000 60000

30000

Area km² of Area habitat, 30000 Area km² of Area habitat,

0 0 RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Climate change scenarios Climate change scenarios

g. Habitat area by CCSM4 GCM h. Habitat area by CCSM4 GCM

111

i. RCP 2.6 j. RCP 2.6

k. RCP 4.5 l. RCP 4.5

m. RCP 6.0 n. RCP 6.0

112

o. RCP 8.5 p. RCP 8.5 Figure 7. Projected changes in snow leopard and Altai habitat in the Altai-Sayan Ecoregion in 2060-2080 predicted by CCSM4 GCM under different RCP scenarios (yellow – no changes, red – habitat loss, green – habitat gain). Snow leopard Altai argali 150000

120000 120000

90000 90000

60000 60000 Area km² of Area habitat, Area km² of Area habitat, 30000 30000

0 0 RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Climate change scenarios Climate change scenarios a. Habitat area by HadGEM2-ES GCM b. Habitat area by HadGEM2-ES GCM

c. RCP 2.6 d. RCP 2.6

113

e. RCP 4.5 f. RCP 4.5

g. RCP 6.0 h. RCP 6.0

i. RCP 8.5 j. RCP 8.5 Figure 8. Projected changes in snow leopard and Altai habitat in the Altai-Sayan Ecoregion in 2060-2080 predicted by HadGEM2-ES GCM under different RCP scenarios (yellow – no changes, red – habitat loss, green – habitat gain).

114

Snow leopard Altai argali

150000 120000 120000 90000 90000 60000 60000 30000

30000

Area km² of Area habitat, Area km² of Area habitat, 0 0 RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 RCP 2.6 RCP 4.5 RCP 6.0 RCP 8.5 Climate change scenarios Climate change scenarios a. Habitat area by MIROC-ESM GCM b. Habitat area by MIROC-ESM GCM

c. RCP 2.6 d. RCP 2.6

e. RCP 4.5 f. RCP 4.5

115

g. RCP 6.0 h. RCP 6.0

i. RCP 8.5 j. RCP 8.5 Figure 9. Projected changes in snow leopard and Altai habitat in the Altai-Sayan Ecoregion in 2060-2080 predicted by MIROC-ESM GCM under different RCP scenarios (yellow – no changes, red – habitat loss, green – habitat gain).

116

APPENDIX 2: REPORT MODELING OF TIGER AND ITS PREY POPULATIONS IN THE BALKHASH LAKE REGION AS THE BASIS FOR ADAPTIVE MANAGEMENT OF THE TIGER REINTRODUCTION PROGRAM IN KAZAKHSTAN

By Mikhail Paltsyn, James P. Gibbs, Liza Iegorova, Olga Pereladova

INTRODUCTION

The southern shore of Lake Balkhash in Kazakhstan as well as the area surrounding and to the east of the Ili River delta has been identified by WWF as a potential site for a tiger restoration program1. This area still has vast tugay woodlands and reed thickets that were populated by Caspian tigers (Panthera tigris virgata) up to the middle of 20th century2. These riparian ecosystems still cover more than 7,000 km² on the southern shore of Balkhash Lake and may represent suitable habitat for tiger introduction. Currently about 25% of the ecosystems are degraded due to intense annual fires and livestock grazing, but may be restored via fire management, grazing restrictions and support of optimal water regime in Ili River. Tugay woodlands and reed thickets can contain high population densities of wild boar (Sus scrofa) and Bukhara deer (Cervus elaphus bactrianus), main prey species for tiger in Central Asia. Roe deer (Capreolus pygargus) is another traditional prey species for tiger in the area, although its population density is relatively low. Circumstantial historical evidence collected from hunting records and encounters indicate that the Caspian tiger population density in Central Asia was much higher than that of the Amur tiger and was similar to tiger densities in India3. Therefore, restoration of the tiger population in the Balkhash region could make a considerable contribution to the Global Tiger Recovery Program. This report provides details of projection modeling of the area of potential tiger habitat – tugay woodlands and reed thickets - as well as population growth models for tiger and its prey species (wild boar, Bukhara deer, and roe deer) on the southern shore of Balkhash Lake and Ili River delta over the next 50 years given three scenarios of possible habitat change and management options for tiger introduction program. The objectives of this modeling exercise are as follows: - To identify possible changes in the distribution and area of potential tiger habitat (tugay woodlands and thicket); - To define time-to-recovery for prey species (wild boar, Bukhara deer, and roe deer) and predict population dynamics given three scenarios of possible habitat changes; - To estimate optimal timing and regimes for tiger introduction given temporal change in the availability of their prey;

1 WWF 2014. Tiger Reintroduction Program in Kazakhstan 2 Sludskiy A.A. 1953. Tiger in the USSR. News of the Academy of Sciences of Kaz. SSR, ser. boil., №8, p.18-73, (in Russ) 3 WWF 2014. Tiger Reintroduction Program in Kazakhstan

117

- To evaluate potential number of tigers that could populate the project area and number of tigers needed for introduction; - To evaluate impact on source populations of tigers and Bukhara deer due to removal of individuals for introduction.

Data and methods

Habitat and Scenarios Landsat 5 TM imageries for July-August 1989 and July-August 2010 were used as basis for habitat mapping in the project area (source: USGS Earth Explorer http://earthexplorer.usgs.gov/). These imageries were mosaiced and classified into 30 classes using Band 4 (Near-infrared), Band 1(blue) and Band 2 (green) with the ArcMap 10.2.2. Image Classification function. After that the imageries were reclassified in four general habitat classes: 1 – water; 2 – desert and semi-desert; 3 – dry bush (including saxaul forest), meadow and steppe; 4 – tugay woodlands and reed thickets. Map of ecosystems of Ili River delta produced by TERRA Center, Kazakhstan, was used for reference during habitat classification process. We assumed that habitat map prepared on the base of Landsat 5 TM 2010 classification adequately represents current distribution and area of different habitat types at water level in Balkhash Lake about 343 m above sea level.4 The habitat map developed on the base of Landsat 5 TM 1989 imagery was assumed to represent general habitat distribution in the conditions of water scarcity (in 1987-1989 the water level in the lake was 340-341 m above sea level). Also, we assumed that the habitats can potentially come back to the conditions of 1987-1989 in about 20 years from now if water deficiency has the same effects on habitat as in 1969-1988 (the last historic period of water deficiency in the Ili River) (Figure 1)5.

4 Bragin E.A. 2006. Review of Balkhash Lake conditions and politics in the field of water resource management in Ile-Balkhash watershed. Technical Report. WWF-Russia. (in Russ.) 5 Ibid

118

Figure 1. Dynamic of water level in Balkhash Lake in 1880-2005 (Bragin 2006).6 We used MODIS MCD45A1 data (Burned Area Monthly L3 Global 500m of LP DAAS https://lpdaac.usgs.gov/products/modis_products_table/mcd45a1 ) to map area and intensity of fires in tugay forest and reed thickets. Data used were from for April-July 2000-2014 - period of wild fires in the project area. Fires are one of the main causes of degradation of tugay and reed thicket ecosystems7. According to the data of the TERRA Center and the Institute of Geography8 reed thickets require about one year to restore after one-time burning and tugay woodlands can restore themselves in 4-5 years after burning in conditions of periodic flooding (at least once in 3-4 years). Tugay forest and reed thickets represent one type of land cover on our habitat maps, therefore, for the purpose of modeling we assumed that the areas that were burned once requires at least 3 years for recovering in average, areas burned twice – at least 6 years to recover, areas burned thrice – 9 years, and so on given optimal regime of flooding and current water level in Balkhash Lake, and absence of intensive grazing. Thus, all areas of tugay forest and reed thicket burned in 2000-2014 were classified as degraded ecosystems. The map representing current conditions of tiger and its prey habitat is shown on the Figure 2 A. This map was used as starting point for three scenarios considered in our modeling exercise. For the purpose of modeling all areas including potential tiger habitats were divided in three Tiger Management Units. We assumed that only tugay woodlands and reed thickets represent optimal habitat for tiger, wild boar and Bukhara deer9. Dry bush (including saxaul forest), meadow and steppe were classified as marginal habitat for these species, but optimal – for roe deer10. Burned

6 Bragin E.A. 2006. Review of Balkhash Lake conditions and politics in the field of water resource management in Ile-Balkhash watershed. Technical Report. WWF-Russia. (in Russ.) 7Institute of Geography 2013. Preparation of socio-economic component of long-term Program for tiger reintroduction in the Ili River delta and southern shore of Balkhash Lake. Research Report. (in Russ.) 8 Ibid 9 WWF 2014. Tiger Reintroduction Program in Kazakhstan 10 Ibid

119 tugay woods and reed thickets also represent marginal habitats in our modeling, due to their intensive use as pastures and low quality of cover. Water, deserts and semi-deserts were designated as non-habitat for all four species listed above. Maximal population densities for the prey species in different habitats are shown in the Table 1. Table 1. Estimated maximum population densities of wild boar, Bukhara deer and roe deer in Central Asia biomes. Maximal population density (ind./km²) Species High quality tugay Low quality tugay Dry bush (including woodlands and reed woodlands and reed saxaul forest), thickets thickets (burned meadow and steppe areas) Wild boar 6.011 0.5 0.5 Bukhara deer 10.012 0.5 1.013 Roe deer 0.5 0.5 1.014

11 population density of wild boars can achieve up to 5-30 animals/km² in tugay and reed ecosystems of Central Asia (Geptner et al. 1969; Danilkin 2006; WWF 2014) 12 Pereladova O.B. 2013. Restoration of Bukhara deer (Cervus elaphus bactrianus Lydd.) in Central Asia in 2000- 2011. IUCN Deer Specialist Group News. March 2013. Pp. 19-30. 13 Ibid 14 WWF 2014. Tiger Reintroduction Program in Kazakhstan

120

Figure 2. Projected dynamics of tugay woodland and reed thicket ecosystem in the Ili-Balkhash region under three management scenario: A - No Fire Management, B – Fire Management Years 30-50, and C – Water scarcity Year 20. Tiger Management Units (TMU) planned for introduction: 1 – Balkhash, 2 – Ili delta, 3 – Karatal. We assumed that habitat quality for tiger and its prey species can be gradually improved through a process of active fire management and restrictions on livestock grazing, and that increase the area of high quality tugay forest and reed thickets by 25-27% given flooding regime is constant and similar to the period of 2010-2013 (Figure 2 B). Also we considered possibility that water volume in Ili River and Balkhash Lake may drop to the level of 1987-1989 as a result of increased water consumption in western China in the nearest 20 years. In this situation tugay forest and reed thicket may shrink considerably (Landsat 5 TM 1989 imagery) in 20 years from now (Figure 2 C) and may continue shrinking further at the same rate. Maps shown on Figure 2 A-C were used for developing of three scenarios of future habitat conditions for tiger and its prey species on the southern shore of Balkhash Lake. Scenarios

121

Three scenarios were used for our modeling exercise: a. No fire management (area and quality of optimal tiger and prey species habitat is constant in the nearest 50 years and equal to the area of the habitat in 2010); b. Fire management (area and quality of optimal habitat gradually increase due to fire control and grazing restrictions); c. Water scarcity (area and quality of optimal habitats gradually decrease as a result of low water volume in Ili River).

No fire management scenario This scenario makes the following assumptions over the next 50 years: • No fire management is implemented in the project area. Thus, area and frequency of fires in tugay woodlands and reed thicket remains constant and follows the pattern of fires in 2000-2014 (Figure 2 A); • Habitat area and quality is constant and do not change (Figure 2 A); therefore, carrying capacity for wild boar, Bukhara deer and roe deer is a constant value (Table 2); • Water volume in Ili River remains constant at the level of 2010-2013; • Harvesting of prey species is completely stopped in the project area; • 30 Bukhara deer are brought annually in each TMU during first 10 years of the program; • Intensive management is used to achieve maximum population densities of wild boar and Bukhara deer in the TMUs (see Table 1). Table 2. Area and carrying capacity of habitat in three TMUs at No Fire Management Scenario (Figure 2)

TMU Habitat Area, Carrying capacity, K km² Wild boar Bukhara Roe deer All prey species Deer

High quality 909.6 tugay woodlands and reed thickets Low quality 27.264 tugay 5580 9328 687 15595 woodlands and Balkhash reed thickets (burned areas) Dry bush (including 218.3 saxaul forest),

meadow and steppe High quality 3696.3 23887 39599 4484 67970 tugay

122

woodlands and reed thickets Low quality tugay 1565.3 woodlands and

Ili River reed thickets Delta (burned areas) Dry bush (including 1853.4 saxaul forest),

meadow and steppe Karatal High quality 690.0 tugay woodlands and reed thickets Low quality 182.0 tugay 4580 7689 1134 13402 woodlands and

reed thickets (burned areas) Dry bush (including 698.0 saxaul forest),

meadow and steppe Estimated carrying capacity of all 34047 56576 6305 96968 TMUs:

Fire Management Scenario The Fire Management scenario is based on the following assumptions: • All fires are prevented in tugay woodlands and reed thickets from the first year of the program; • Strict restrictions on livestock grazing are implemented starting the first year of the program (number of livestock gradually decreases in the project area); • Area of high quality tugay woodlands and reed thicket gradually increases (gradual move from Figure 2 A to Figure 2 B). Habitat carrying capacity for wild boar and Bukhara deer gradually increases with increase of the area of restored tugay and reed ecosystems (Figures 3-5, Table 3); • Water volume in Ili River remains constant at the level of 2010-2013; • Harvesting of prey species is completely stopped in the project area; • 30 Bukhara deer are brought annually in each TMU during first 10 years of the program; • Intensive management is used to generate maximum population densities of wild boar and Bukhara deer in the TMUs (see Table 1).

123

12000

10000

8000

6000 Wild Boar Bukhara deer

4000 Roe deer Carryingcapacity, K

2000

0 0 10 20 30 40 50 60 Years

Figure 3. Estimated carrying capacity of habitat in Balkhash TMU under Fire Management Scenario.

60000

50000

40000

30000 Wild boar Bukhara deer 20000

Roe deer Carryingcapacity, K 10000

0 0 10 20 30 40 50 60 Years

Figure 4. Estimated carrying capacity of habitat in Ili River Delta TMU under Fire Management Scenario.

124

10000 9000 8000 7000 6000 5000 Wild boar 4000 Bukhara deer 3000 Roe deer

Carryingcapacity, K 2000 1000 0 0 10 20 30 40 50 60 Years

Figure 5. Estimated carrying capacity of habitat in Karatal TMU under Fire Management Scenario. Table 3. Estimated carrying capacity of habitat in three TMUs before and after 30 years of fire management

Area of high quality tugay woodland and Carrying capacity, K reed thicket, TMU km² Before After Wild Wild Bukhara Bukhara Roe Roe All prey, All prey, mgmt mgmt boar, boar, deer, deer, deer, deer, before after before after before after before after mgmt mgmt mgmt mgmt mgmt mgmt mgmt mgmt Balkhash 909.6 936.8 5580 5730 9328 9586 687 687 15595 16003

Ili River Delta 3696.3 5261.6 23887 32496 39599 54470 4484 4484 67970 91450

Karatal 690.0 872.0 4580 5580 7689 9418 1134 1134 13403 16132

All TMUs 5295.9 7070.4 34047 43806 56616 73474 6305 6305 96968 123585

Thus, as a result of fire management carrying capacity of the habitat can be potentially increased by 27%.

Water Scarcity Scenario This scenario made the following assumptions over the next 50 years: • Water volume in Ili River and water level in Balkhash Lake gradually decrease with the same rate as in 1969-1989 (Figure 1); • No fire management is implemented in the project area. Thus, area and frequency of fires in tugay woodlands and reed thicket slowly increases (Figure 2 C);

125

• Area and quality of tugay woodlands and reed thickets gradually decrease due to water scarcity with the same rate as in 1969-1989 and come to the conditions shown on Figure 4 (1989) after 20 years from now. The habitat continues to deteriorate and shrink with the same rate through the nearest 50 years; • Carrying capacity of the habitat decreases with the same rate as high quality tugay woodlands and reed thickets deteriorates and shrink (Figure 6-8, Table 4); • Harvesting of prey species is completely stopped in the project area; • 30 Bukhara deer are brought annually in each TMU during first 10 years of the program; • Intensive management is used to generate maximum population densities of wild boar and Bukhara deer in the TMUs (see Table 1).

10000 9000 8000 7000 6000 5000 Wild boar 4000 Bukhara deer

3000 Roe deer Carryingcapacity, K 2000 1000 0 0 10 20 30 40 50 60 Year

Figure 6. Estimated carrying capacity of habitat in Balkhash TMU under Water Scarcity Scenario

126

45000 40000 35000 30000 25000 Wild boar 20000 Bukhara deer 15000

Roe deer Carryingcapacity, K 10000 5000 0 0 10 20 30 40 50 60 Years

Figure 7. Estimated carrying capacity of habitat in Ili River Delta TMU under Water Scarcity Scenario

9000 8000 7000 6000 5000 Wild boar 4000 Bukhara deer 3000

Roe deer Carryingcapacity, K 2000 1000 0 0 10 20 30 40 50 60 Year

Figure 8. Estimated carrying capacity of habitat in Karatal TMU under Water Scarcity Scenario Table 4. Possible change of habitat carrying capacity in three TMUs under Water Scarcity Scenario

Habitat/Species Current Year 10 Year 20 Year 30 Year 40 Year 50 (Figure 2) (Figure 4) Balkhash TMU

Area of high quality tugay woodland and 909.6 752.8 596.0 439.2 282.4 125.6 reed thicket, km² Wild boar, carrying capacity 5580 4714 3752 2790 1828 866 Bukhara deer, carrying capacity 9328 7884 6279 4675 3070 1466

127

Roe deer, carrying capacity 687 687 687 687 687 687 All prey, carrying capacity 15595 13285 10718 8152 5585 3019 Ili River Delta TMU

Area of high quality tugay woodland and 3696.3 3190.5 2685.0 2179.5 1674 1168.5 reed thicket, km² Wild boar, carrying capacity 23887 21189 18192 15194 12196 9199 Bukhara deer, carrying capacity 39599 35113 30129 25145 20161 15177 Roe deer, carrying capacity 4484 4484 4484 4484 4484 4484 All prey, carrying capacity 67970 60786 52805 44823 36841 28860 Karatal TMU

Area of high quality tugay woodland and 690.0 595.0 500.0 405 310 215 reed thicket, km² Wild boar, carrying capacity 4580 4004 3365 2725 2086 1446 Bukhara deer, carrying capacity 7689 6750 5706 4662 3619 2576 Roe deer, carrying capacity 1134 1134 1134 1134 1134 1134 All prey, carrying capacity 13403 11888 10205 8521 6839 5156 All TMUs

All prey, carrying capacity 96968 85959 73728 61496 49265 37035

Under the Water Scarcity Scenario habitat carrying capacity for tiger’s prey species may decrease by 25% in 20 years and by 62% in 50 years.

Prey species population growth models

Three ungulate species – wild boar, Bukhara deer and roe deer – were assumed to be main preys for tigers after introduction in all three TMUs. To describe population growth of the prey species we implemented a model of density-dependent population (logistic) growth typically used in wildlife population projection scenarios such as this:

Nt+1 = Nt + rNt(1- Nt/K), where

Nt is abundance of a population in year t ;

Nt+1 is abundance of a population next year (t+1); r is instantaneous per capita growth rate; and

K is population carrying capacity.

Population abundance for wild boar, Bukhara deer and roe deer in TMUs in the first year of the program (N0) as well as r values for these species are shown in the Table 5. Carrying capacity for prey species at different scenarios are listed in the Tables 2-4.

128

Table 5. Estimated values of N0 and r =for wild boar, Bukhara deer, and roe deer in three TMUs

Population abundance (Year 0) Balkhash Ili River Karatal Species r mean SD Delta Wild boar 0.34915 0.503 46016 188517 35218 Bukhara 0.14419 0.111 3020 30 30 deer Roe deer 0.02021 0.143 20022 90023 17024

Environmental stochasticity was added to the deterministic population models via lognormal distribution of r values25. For each scenario a total of 1000 iterations were performed to calculate mean prey species abundance (and its 95% confidence interval) at years 5, 10, 15, 20, 25, 30, 35, 40, 45, and 50. At present the Bukhara deer population from Karachingil Game Management Area is targeted for use as an only source of deer for introduction in the TMUs. Currently this population is at carrying capacity and has 400-500 animals in the area of 200 km² 26. We used same population model as above for Bukhara deer to define how many deer can be harvested in Karachingil for introduction in the first 5-10 years of the program. The goal of our management for Karachingil population was to avoid population decline below half of the population carrying capacity (approximately 225 animals). Newly established populations in one of the TMUs can be used as an additional source of deer for other two TMUs after 10-15 years of the program.

Tiger and prey models

15 Calculated as average from time-series for wild boar populations in the Russian Far East (Zaumyslova O.Yu. 2005), Belovezhskay Puscha, Crimea, and Europe (Mayer J.J. 2009). No time-series for tugay forest and reed thicket were found. 16 Our assumption based on Lukarevsky and Baydavletov 2010; WWF 2014 17 Ibid 18 Ibid 19 Calculated as average using time-series from Badai-Tuagai Nature Reserve and Karachingil Game Management Area in 1999-2011(Pereladova O.B. 2013. Restoration of Bukhara deer (Cervus elaphus bactrianus Lydd.) in Central Asia in 2000-2011. IUCN Deer Specialist Group News. March 2013. Pp. 19-30.) 20 30 deer will be brought to the selected TMU annually in the first 10 years of the program. 21 Calculated as average from time-series from Russian Far East, Norway, and France ((Ignatova et al., 2004; Nilsen et al., 2009; Melis et al., 2010). No time-series for tugay forest and reed thicket were found. 22 Our assumption based on Lukarevsky and Baydavletov 2010; WWF 2014 23 Ibid 24 Ibid 25 Hood, G. M. 2010. PopTools version 3.2.5. Available on the internet. URL http://www.poptools.org 26 R. Baidavletov, personal communication

129

We used age class matrix for tiger females to build a model of tiger population growth in the project area. We assumed that density-dependence in our models affects only reproduction, but not survival rates. For our modeling exercise we used following assumptions: • tugay forest and reed thickets represent optimal habitat for tiger in the project area27; • dry bush (including saxaul forest), meadow and steppe were assumed to be marginal habitat where tigers can come occasionally for a short period of time; • habitat carrying capacity for tigers is limited by prey species population density and abundance only28; • only sub-adult and adult tigers kill their own prey; • number of particular prey species killed by a tiger directly depends on the ratio of this species abundance to all prey species abundance in the project area; • only sub-adult and adult tiger abundance was used to calculate carrying capacity depending on prey abundance; • female/male ratio of tiger population is constant and similar to that of the Amur tiger; • tiger vital rates in the project area are the same as for Amur tiger. Parameters used for tiger-prey population models are shown in the Table 6. Table 6. Vital rates and other parameters used for tiger-prey population models

Age Survival Fecundity31 Female/male ratio32 Average number of Average prey abundance stages29 rates30 ungulates killed by 1 tiger necessary to support 1 annually tiger Cub (0-1) 0.6 0 0 - Juvenile 0.8 0 0 (1-2) 1.533 Sub-Adult 0.6 0 6034 (2-3) 45035

Adult (3+) 0.8 0.75

27 Sludskiy A.A. 1953. Tiger in the USSR. News of the Academy of Sciences of Kaz. SSR, ser. boil., №8, p.18-73, (in Russ). WWF 2014. Tiger Reintroduction Program in Kazakhstan 28 Karanth K.U., Nichols J.D., Kumar N. S., Link W.A, and J.E. Hines. 2004. Tigers and their prey: Predicting carnivore densities from prey abundance. PNAS. 101(14). Pp. 4854–4858. 29 Tian Yu. Et al. 2011. Population viability of Siberian tiger in changing landscape: going, going and gone? Ecological Modeling. 222. Pp. 3166-3180 30 Ibid 31 Ibid 32 Ibid 33 The actual female/male ratio for adult Amur tigers is between 5/3 and 6/5 based on field surveys (Carroll and Miquelle, 2006; Miquelle et al., 2006; Tian et al. 2011). 34 According to the Strategy for Amur tiger conservation in Russia (2010) one tiger annually kill 50-70 ungulates. Karanth et al. 2004 applied the average kill rate of 50 ungulates/tigers per year consistently observed in field studies of tigers. 35 According to the Strategy for Amur tiger conservation in Russia (2010) one tiger needs population of 400-500 ungulates to support itself with sufficient prey. According to Karanth et al. 2004 average tiger/ungulate ratio should be about 1/500.

130

We used following formula to calculate carrying capacity for tigers in TMUs:

Kt = Nprey t/450, where

Kt is carrying capacity for tigers in a TMU in the year t,

Nprey t is abundance of all prey species (wild boar, Bukhara deer and roe deer) in the TMU, 450 is the mean number of ungulates required to support one tiger. The effect of tiger predation on a particular ungulate species in the TMUs was calculated using the following formula:

Nt+1 = Nt + rNt(1- Nt/K) – Ns+a t * (Nt / Np t)*R s+a , where

Nt is abundance of an ungulate species population in year t ;

Nt+1 is abundance of an ungulate species population next year (t+1); r is the instantaneous per capita growth rate of an ungulate species; K is ungulate species population carrying capacity;

Ns+a t – number of sub-adult and adult tigers in year t;

Np t – abundance of all ungulate species in year t;

R s+a – number of ungulates killed by one tiger annually (constant value equal 60). Environmental stochasticity was added to the tiger models as changing carrying capacity determined by variation of prey species numbers. No demographic stochasticity was considered in our modeling. Tiger population modeling was implemented on the outputs of the ungulate species population growth models36. We explored different schemes of tiger introduction (bringing different number of tigers in different years to TMUs at different scenarios) to find an acceptable one. The criterion for acceptability was that the lower level of 95% confidence interval would not cross quasi-extinction threshold (10 tigers) considerably. Amur tiger population in the Russian Far East is currently considered as the most probable source of tigers for introduction in Ili-Balkhash TMUs. According to the data of last tiger counts in 2005 the population has 428-502 tigers (465 in average)37. The population has following sex and age structure: 25% - adult males, 39% adult females, 22% –cubs up to 1.5 years old, 6% - juveniles and sub-adults (1.5-3 years old)38. The same matrix model for females as above was used to identify number of tigers that can be used for introduction program without considerable impact for Russian Far East population (Table 6). For the model we assumed that carrying capacity for

36 Hood, G. M. 2010. PopTools version 3.2.5. Available on the internet. URL http://www.poptools.org 37 Strategy for conservation of the Amur Tiger in the Russian Federation. Approved by the Ministry of Natural Resources and Environment of the Russian Federation, Decree of July 2, 2010, # 25-r. 38 Ibid

131 the Amur tiger in the Russian Far East in current conditions is 510-550 (530 in average)39. Following population vector for tiger population in the Russian Far East was used for the year 1 of modeling: Age stage Abundance (females only) Cub (0-1) 51 Juvenile (1-2) 30 Sub-Adult (2-3) 17 Adult (3+) 181 N total 279

5 tigers (3 females and 2 males) were removed every 3, 4 or five years from the source population for different number of years according to various schemes of tiger introduction in the TMUs on the shore of Balkhash Lake. For this model we assumed that the source population will have same survival and reproduction rates for the nearest 50 years and will not be considerably affected by poaching, habitat destruction and prey depletion. No environmental and demographic stochasticity was considered for Amur tiger source population model.

Model limitations Our models have following limitations that should be recognized and considered for their implementation as a tool for adaptive management of the tiger introduction program: • Limited data were available wild boar and roe deer current abundance upon which to base the model estimates of prey population dynamics; field population surveys are necessary to evaluate current population status and dynamics of these species in the Ili-Balkhash Basin. • No time-series of population numbers for wild boar and roe deer for tugay forest and reed thickets in Kazakhstan or other regions of Central Asia were available. Thus, the r mean values were calculated using time-series for these species populations from other areas and habitat types. As a result of this approach annual population growth of wild boar and roe deer have very high variability that effect the precision of the models. The models could be improved a great deal if time-series for wild boar and roe deer populations in the tugay woodlands and reed thickets become available. These data may be available from Nature Chronicles of Central Asian Nature Reserves (Zapovedniks). • Our models are based on the highest population densities for wild boar, roe deer and Bukhara deer that could be achieved in the tugay woodlands and reed thickets. It may not be a real option for a 7,000 km² of such habitat in the Ili-Balkhash Basin. Thus, additional

39 Goal 2020 of the Strategy for conservation of the Amur Tiger in the Russian Federation. Approved by the Ministry of Natural Resources and Environment of the Russian Federation, Decree of July 2, 2010, # 25-r.

132

scenarios with average population densities of this species should be considered to make the models to be an effective tool for adaptive management of the program. • Vital rates of tigers in Ili-Balkhash Basin might be very different from vital rates of tigers in the Russian Far East used in our tiger population models. • It is extremely difficult to predict actual water balance in the Ili River and Balkhash Lake over the long term. Thus, our assumptions for the Water Scarcity Scenario based on the satellite imageries of 1989 and 2010 could be far from perfect. • Our models for wild boar, Bukhara deer, roe deer and tigers do not consider effect of demographic and genetic stochasticity on the species population growth (although these are likely to be inconsequential for two reasons: prey population sizes are generally large and tiger populations can persist even at very low densities given the intrinsic mobility of tigers). • Model for source population of Amur tiger in the Russian Far East did not consider demographic, environmental and genetic stochasticities that can make population growth much slower and less predictable. • No environmental autocorrelation, catastrophes or bonanzas were taken in account in our models because these phenomena and their effects of prey populations (should they even occur) are difficult to estimate at this time.

133

RESULTS

Population projections for tiger prey species at different scenarios In all depictions of likely future population scenarios the solid line = mean population abundance and dashed lines = upper and lower limits of 95% confidence interval. BALKHASH TMU No Fire Management Scenario

Wild boar Bukhara deer 6000 10000 9000 5000 8000 4000 7000 6000

3000 5000 Number

Number 4000 2000 3000 1000 2000 1000 0 0 0 20 40 60 0 20 40 60 Year Year

Roe deer All prey 700 18000

600 16000 14000 500 12000 400 10000

300 8000 Number Number 6000 200 4000 100 2000 0 0 0 20 40 60 0 20 40 60 Year Year

Years 10-15 of the program is likely to be the earliest year to start tiger introduction in Balkhash TMU (N mean = 3779-5560, enough prey base for first five tigers (3 females and 2 males)), but at that time high probability exists that the prey base could be still to low (Lower CL = 902-1450). In the worst case introduction can be started at year 20 when there would be minimally 5 tigers.

134

Prey Abundance/ Prey Year 20 Tiger CC Year Prey Year 5040 Tiger CC Year Tiger carrying 20 5041 capacity (CC) Nmean 7145 15 14710 32 Upper CL 8907 19 15320 34 Lower CL 2623 5 13463 29

Fire Management Scenario Wild boar Bukhara deer 10000 7000 9000 6000 8000 7000 5000 6000 4000 5000

3000 Number 4000

3000 Number 2000 2000 1000 1000 0 0 0 20 40 60 0 20 40 60 -1000 Year Year

Roe deer All prey 700 18000 600 16000 14000 500 12000 400 10000

300 8000

Number Number 200 6000 4000 100 2000 0 0 0 20 40 60 0 20 40 60 Year Year

Fire management will not improve prey species carrying capacity considerably in Balkhash TMU due to very limited area and frequency of fires in this territory. Still the earliest time to

40 Prey abundance without tiger predation 41 Carrying capacity without effect of tiger predation

135

bring fist 5 tigers (3 females and 2 males) to the TMU will be year 10-15. At Year 20 even in the worst case scenario it will be appropriate to start introduction.

Prey Abundance/ Prey Year 20 Tiger CC Year Prey Year 5042 Tiger CC Year Tiger carrying 20 5043 capacity (CC) Nmean 6778 15 15030 33 Upper CL 9033 20 15671 34 Lower CL 2576 5 13667 30

Water Scarcity Scenario Wild boar Bukhara deer 6000 5000

5000 4000 4000 3000 3000

2000 Number Number 2000 1000 1000

0 0 0 20 40 60 0 20 40 60 Year Year

Roe deer All prey 600 9000 8000 500 7000 400 6000 5000 300

4000 Number 200 Number 3000 2000 100 1000 0 0 0 20 40 60 0 20 40 60 Year Year

The introduction can be started at the same period as in the two previous scenarios, but since year 30 carrying capacity for prey species is likely to drop considerably if the water volume in

42 Prey abundance without tiger predation 43 Carrying capacity without effect of tiger predation

136

Ili River will continue to go down. In this case Balkhash TMU will be unlikely to support even smallest viable tiger population (at least 10 animals) by year 50.

Prey Abundance/ Prey Year 20 Tiger CC Year Prey Year 5044 Tiger CC Year Tiger carrying 20 5045 capacity (CC) Nmean 5652 12 2869 6 Upper CL 6955 15 4061 9 Lower CL 2940 6 2579 5

ILI RIVER DELTA TMU No Fire Management Scenario Wild boar Bukhara deer 30000 40000 35000 25000 30000 20000 25000

15000 20000 Number 10000 Number 15000 10000 5000 5000 0 0 0 20 40 60 0 20 40 60 Year Year

Roe deer All prey 70000 60000 50000 40000

30000 Number 20000 10000 0 0 20 40 60 Year

44 Prey abundance without tiger predation 45 Carrying capacity without effect of tiger predation

137

4000 3500 3000 2500 2000

Number 1500 1000 500 0 0 20 40 60 Year

In Ili River Delta TMU introduction can be started at the Year 5-10 of the program as the earliest in the best case (N mean = 6582-14123), but still some probability exists that the prey base will be lower than necessary (Lower CL = 1595-2298). But by Year 15 there would likely be enough prey for the first 5 tigers even in the worst case of prey species population growth. The area has much larger capacity for tiger population, than Balkhash TMU.

Prey Abundance/ Prey Year 15 Tiger CC Year Prey Year 5046 Tiger CC Year Tiger carrying 15 5047 capacity (CC) Nmean 20426 45 55643 123 Upper CL 27145 60 63355 140 Lower CL 3430 7 42316 94

46 Prey abundance without tiger predation 47 Carrying capacity without effect of tiger predation

138

Fire Management Scenario Wild boar Bukhara deer 35000 60000

30000 50000 25000 40000 20000 30000

15000

Number Number 20000 10000

5000 10000

0 0 0 20 40 60 0 20 40 60 Year Year

Roe deer All prey 4000 90000 3500 80000 3000 70000 60000 2500 50000 2000

40000

Number Number 1500 30000 1000 20000 500 10000 0 0 0 20 40 60 0 20 40 60 Year Year

139

Fire management and grazing restrictions can improve tugay forest and red thicket quality and increase carrying capacity for prey species and tigers by 29-30% in the nearest 30 years. Still the earliest time for introduction is Year 5-10 and the latest is Year 15.

Prey Abundance/ Prey Year 15 Tiger CC Year Prey Year 5048 Tiger CC Year Tiger carrying 15 5049 capacity (CC) Nmean 24081 53 71943 159 Upper CL 33993 75 84532 187 Lower CL 3681 8 53960 120

Water Scarcity Scenario Wild boar Bukhara deer 25000 20000

20000 15000 15000

10000 Number

Number 10000 5000 5000

0 0 0 20 40 60 0 20 40 60 Year Year

Roe deer All prey 3500 40000 3000 30000 2500

2000 20000

1500 Number Number 1000 10000 500 0 0 0 20 40 60 0 20 40 60 Year Year

48 Prey abundance without tiger predation 49 Carrying capacity without effect of tiger predation

140

Despite severe decline in habitat area and carrying capacity (37-40%) in the conditions of water deficit Ili River Delta TMU is likely to be able to support a relatively stable tiger population (no less than 30-50 animals) by the year 50.

Prey Abundance/ Prey Year 15 Tiger CC Year Prey Year 5050 Tiger CC Year Tiger carrying 15 5051 capacity (CC) Nmean 17821 39 26783 59 Upper CL 23394 51 29840 66 Lower CL 3245 7 22455 50

KARATAL TMU No Fire Management Scenario Wild boar Bukhara deer 5000 10000

4000 8000

3000 6000 Number Number 2000 4000

1000 2000

0 0 0 20 40 60 0 20 40 60 Year Year

Roe deer All prey

50 Prey abundance without tiger predation 51 Carrying capacity without effect of tiger predation

141

1000 14000 12000 800 10000 600 8000

6000 Number Number 400 4000 200 2000 0 0 0 20 40 60 0 20 40 60 Year Year

Tiger introduction in Karatal TMU is likely to be safely started at the year 20 according to the model above even at the slowest population growth of prey species. Year 10-15 is the earliest possible option.

Prey Abundance/ Prey Year 20 Tiger CC Year Prey Year 5052 Tiger CC Year Tiger carrying 20 5053 capacity (CC) Nmean 6127 13 12251 27 Upper CL 7730 17 12909 28 Lower CL 2811 6 11416 25

Fire Management Scenario Wild boar Bukhara deer 6000 12000 5000 10000 4000 8000

3000 6000 Number Number 2000 4000 1000 2000 0 0 0 20 40 60 0 20 40 60 Year Year

52 Prey abundance without tiger predation 53 Carrying capacity without effect of tiger predation

142

Roe deer All prey 1000 20000

800 15000 600

10000 Number Number 400 5000 200

0 0 0 20 40 60 0 20 40 60 Year Year

Fire management and grazing restrictions can improve carrying capacity of Karatal TMU for tiger and prey species by 20-23%.

Prey Abundance/ Prey Year 20 Tiger CC Year Prey Year 5054 Tiger CC Year Tiger carrying 20 5055 capacity (CC) Nmean 7197 15 15186 33 Upper CL 9078 20 15792 35 Lower CL 3170 7 14360 31

Water Scarcity Scenario Wild boar Bukhara deer 5000 5000

4000 4000

3000 3000 Number Number 2000 2000

1000 1000

0 0 0 20 40 60 0 20 40 60 Year Year

54 Prey abundance without tiger predation 55 Carrying capacity without effect of tiger predation

143

Roe deer All prey 700 8000 600 7000 500 6000 5000 400 4000

300 Number Number 3000 200 2000 100 1000 0 0 0 20 40 60 0 20 40 60 Year Year

Prey Abundance/ Prey Year 20 Tiger CC Year Prey Year 5056 Tiger CC Year Tiger carrying 20 5057 capacity (CC) Nmean 6365 14 4846 10 Upper CL 9078 20 5362 12 Lower CL 2643 5 4404 9

In case of progressive water deficit in Ili River for nearest 50 years, Karatal TMU will be unlikely to support viable tiger population by the year 50.

Population projections for tigers at different scenarios and schemes of introduction

In all graphical depictions of tiger population projections below black solid line = mean population abundance, black dashed lines = upper and lower limits of 95% confidence interval, and horizontal red line = quasi-extinction threshold (10 individuals).

BALKHASH TMU No Fire Management Scenario

56 Prey abundance without tiger predation

57 Carrying capacity without effect of tiger predation

144

20 18 16 14 12 10

8 Number 6 4 2 0 10 15 20 25 30 35 40 45 50 55 Year

Scheme 1. 3 tiger females and 2 males relocated to the TMU on the Year 10, 3 more females and 2 more males added on the Year 15. Nmean that is likely to be achieved by the year 50 using this scheme is 13-14 tigers only. This scheme is not considered as optimal, because considerable part of 95% confidence interval is located below quasi-extinction threshold, thus high probability exist the population could go extinct due to rapid prey depletion.

20 18 16 14 12 10

8 Number 6 4 2 0 15 20 25 30 35 40 45 50 55 Year

Scheme 2. 3 tiger females and 2 males are relocated to the TMU on the Year 20, 3 more females and 2 more males added on the Year 23. This scheme allows to achieve Nmean equal to 17-18 tigers by the year 50, but still part of the confidence interval falls below quasi-extinction threshold, though probability of extinction is much lower than in the previous case.

145

20 18 16 14 12 10

Number 8 6 4 2 0 15 20 25 30 35 40 45 50 55 Year

Scheme 3. 3 tiger females and 2 males are relocated to the TMU on the Year 20, 3 more females and 2 more males added on the Year 25. This scheme allows to achieve Nmean equal to 17 tigers by the year 50 and has relatively low probability of the population falling below quasi- extinction threshold.

10 Number 5

0 20 25 30 35 40 45 50 55 Year

Scheme 4. 3 tiger females and 2 males are relocated to the TMU on the Year 20, 3 more females and 2 more males added on the Year 28. This scheme allows to achieve Nmean equal to 17-18 tigers by the year 50 and has confidence interval all above the quasi-extinction threshold since the year 27.

146

10 Number

0 20 25 30 35 40 45 50 55 Year

Scheme 5. 3 tiger females and 2 males are relocated to the TMU on the Year 20, 3 more females and 2 more males added on the Year 28, 2 females and 1 male added at the year 33. This scheme allows to achieve a Nmean equal to 22-23 tigers by the year 50, but population has high probability to again fall below quasi-extinction threshold after year 45.

Number 10

0 20 25 30 35 40 45 50 55 Year

Scheme 6. 3 tiger females and 2 males are relocated to the TMU on the Year 20, 3 more females and 2 more males added on the Year 28 and 33. This scheme allows reaching Nmean equal to 23 tigers by the year 50, but population has a probability to fall below quasi-extinction threshold after year 45 due to prey depletion. The maximum carrying capacity of Balkhash TMU for tigers is no more than 24-26 individuals according to the model (includes effect of tiger predation on prey); to achieve this population some 10-15 tigers (no more than 9 females and 6 males) should be introduced in the TMU in years 20-33.

Fire Management Scenario

147

For Balkhash TMU this scenario provided very similar results to the previous scenario; therefore we do not consider its details further.

Water Scarcity Scenario

18 16 14 12 10

8 Number 6 4 2 0 15 20 25 30 35 40 45 50 55 Year

Scheme 7. 3 tiger females and 2 males are relocated to the TMU on the Year 20, 3 more females and 2 more males added on the Year 25. In this case the tiger population can drop below quasi-extinction threshold by the year 45 and therefore cannot be considered as viable. Balkhash TMU evidently would not provide safe shelter even for small viable tiger population in conditions of water deficit in Ili River.

148

ILI RIVER DELTA TMU No Fire Management Scenario

20 Number

0 15 20 25 30 35 40 45 50 55 Year

Scheme 8. 3 tiger females and 2 males are relocated to the TMU on the Year 20, 3 more females and 2 more males added on the Year 25, 30, 35, and 40 (25 tigers introduced in the TMU in Years 20-40). The total tiger population that is likely to be achieved by year 50 is 43-45 individuals.

80 70 60 50 40

Number 30 20 10 0 15 20 25 30 35 40 45 50 55 Year

Scheme 9. 3 tiger females and 2 males are relocated to the TMU on the Year 20, 3 more females and 2 more males added on the Year 24, 28, 32, 36, 40, 44, and 48 (40 tigers introduced in the TMU in Years 20-48). The total tiger population that is likely to be achieved by year 50 is 64-69 individuals.

149

100 90 80 70 60 50

Number 40 30 20 10 0 15 20 25 30 35 40 45 50 55 Year

Scheme 10. 3 tiger females and 2 males are relocated to the TMU on the Year 15, 3 more females and 2 more males added on the Year 18, 21, 24, 27, 30, 33, 36, 39, 42 (50 tigers introduced in the TMU in Years 15-42). The total tiger population that is likely to be achieved by year 50 is 77-88 individuals, but it is likely that the population will drop down to 21-22 individuals as a result of prey depletion. Maximal population of tigers that could be achieved in Ili River Delta TMU in the nearest 60-80 years at this scenario is 80-84 individuals (actual estimated carrying capacity of the TMU including effect of tiger predation).

Fire Management Scenario

80 70 60 50 40

Number 30 20 10 0 15 20 25 30 35 40 45 50 55 Year

Scheme 11. 3 tiger females and 2 males are relocated to the TMU on the Year 20, 3 more females and 2 more males added on the Year 24, 28, 32, 36, 40, 44, and 48 (40 tigers introduced in the TMU in Years 20-48). The total tiger population that is likely to be achieved by year 50 is 66-70 individuals.

150

120

100

60 Number 40

0 15 20 25 30 35 40 45 50 55 Year

Scheme 12. 3 tiger females and 2 males are relocated to the TMU on the Year 15, 3 more females and 2 more males added on the Year 18, 21, 24, 27, 30, 33, 36, 39, 42 and 45 (55 tigers introduced in the TMU in Years 15-45). The total tiger population that is likely to be achieved by year 50 is 92-98 individuals, but it is likely that the population will drop down to 60 individuals as a result of prey depletion in this case. Maximal sustainable tiger population that is likely to be achieved at this scenario is 90-94 individuals (actual carrying capacity of the TMU improved by fire management and grazing restrictions). Water Scarcity Scenario

80 70 60 50 40

Number 30 20 10 0 15 20 25 30 35 40 45 50 55 Year

Scheme 13. 3 tiger females and 2 males are relocated to the TMU on the Year 15, 3 more females and 2 more males added on the Year 18, 21, 24, 27, 30, 33, 36, 39, 42 (50 tigers

151 introduced in the TMU in Years 15-42). In this case the tiger population could drop down to 5- 32 (mean = 18) individuals by Year 50 as result of habitat deterioration and prey depletion.

30 Number 20

0 15 20 25 30 35 40 45 50 55 Year

Scheme 14. 3 tiger females and 2 males are relocated to the TMU on the Year 15, 3 more females and 2 more males added on the Year 18, 21, 24, 27, 30, and 33 (35 tigers introduced in the TMU in Years 15-33). In this case the tiger population could drop down to about 37 by Year 50 as a result of habitat deterioration and prey depletion.

30 Number 20

0 15 20 25 30 35 40 45 50 55 Year

Scheme 15. 3 tiger females and 2 males are relocated to the TMU on the Year 15, 3 more females and 2 more males added on the Year 18, 21, 24, 27, (25 tigers introduced in the TMU in Years 15-27). In this case tiger population is likely to survive even the period of continuous water scarcity in the nearest 50 years.

KARATAL TMU

152

No Fire Management Scenario

20 18 16 14 12 10

Number 8 6 4 2 0 15 20 25 30 35 40 45 50 55 Year

Scheme 16. 3 tiger females and 2 males are relocated to the TMU on the Year 20, 3 more females and 2 more males added on the Year 28 (10 tigers introduced in the TMU in Years 20-28). Population of 17-18 tigers may be achieved using this option by Year 50. Probability that population will fall below quasi-extinction threshold is low.

15 Number 10

0 15 20 25 30 35 40 45 50 55 Year

Scheme 17. 3 tiger females and 2 males are relocated to the TMU on the Year 20, 3 more females and 2 more males added on the Year 25, and 35 (15 tigers introduced in the TMU in Years 20-35). Tiger population in 21-25 individuals can be achieved by year 50, but probability exists that population will go below quasi-extinction threshold after year 45 due to prey depletion. According to the model 21-25 the tiger population is above the TMU carrying capacity and population will obviously decline after Year 50. Karatal TMU can support maximal sustainable tiger population of 17-19 individuals for long-term at this scenario

153

Fire Management Scenario This scenario is similar to the previous one, but carrying capacity of the TMU could increase up to 23-25 tigers as a result of fire management and grazing restrictions. Thus, three groups of 5 tigers could be brought into the area with interval of 8-10 years.

Water Scarcity Scenario

Number 5

0 15 20 25 30 35 40 45 50 55 Year

Scheme 18. 3 tiger females and 2 males are relocated to the TMU on the Year 20, 3 more females and 2 more males added on the Year 25 (10 tigers introduced in the TMU in Years 20-25). In the conditions of continuous water scarcity tiger population in Karatal TMU is likely to fall below the quasi-extinction threshold by year 45-50.

Population projections for Bukhara deer source population at Karachingil Game Management Area Black solid line = mean population abundance, black dashed lines = upper and lower limits of 95% confidence interval, horizontal red line = half of the source population carrying capacity (225 individuals).

154

500 450 400 350 300 250

Number 200 150 100 50 0 0 10 20 30 40 50 60 Year

Scheme 19. 30 deer are removed annually for 10 years for introduction in one of the TMUs. It is very low probability that the population will fall below half of its carrying capacity. Thus, this scheme will likely be sustainable.

500 450 400 350 300 250

Number 200 150 100 50 0 0 10 20 30 40 50 60 Year

Scheme 20. 35 deer are removed annually for 9 years for introduction in one of the TMUs. There is strong probability that the population will fall below half of its carrying capacity in the area of less sustainable management. This scheme still could be sustainable but should be implemented with caution.

155

500 450 400 350 300 250

Number 200 150 100 50 0 0 10 20 30 40 50 60 Year

Scheme 21. 40 deer are removed annually for 8 years for introduction in one of the TMUs. This scheme has high probability of the source population decline below the half of its carrying capacity. It is better to avoid this pattern of deer removal for introduction.

Population projections for source population of Amur tiger in the Russian Far East

530

520 Stable population, no tiger removal 510 3 F +2M removal Year 20, 28 and 33 500 3 F +2M removal Year 20, 25 and 35

Number 490 3 F +2M removal Year 20, 25, 30, 35, and 40 480 3 F +2M removal Year 15, 18, 21, 24, 27, 30, 33, 36, 39, 42 470 3 F +2M removal Year 15, 18, 21, 24, 27, 30, 33, 36, 39, 42 and 45 460 10 20 30 40 50 60 Year

156

Scheme 22. Population projections for Amur tiger population in the Russian Far East at different schemes of tiger group removal for introduction (3 females and 2 males in each group). If poaching, habitat degradation and prey depletion do not influence Russian Amur tiger population it is likely to continue to grow despite any reasonable off-take of individuals for introduction in the Ili-Balkhash watershed. Effect of tiger removal on the Amur population is summarized in the table below. First two options (removal of 5 tigers in three groups) are acceptable if the introduction program starts with Balkhash or Karatal TMU: they have very minor effect on Amur population (only 2 animal difference with baseline option by Year 50). The last three options consider tiger introduction if the program starts with Ile River Delta TMU: 25-55 animals are removed in the period of 20-30 years. Last two options have noticeable impact on the source population and should be carefully considered with different positions including political context (conservation goals for Amur tiger population). Introduction of 25 tigers in the Ili River Delta TMU will be enough to achieve population of 45-50 tigers (viable population for a short- term) in 50-60 years of the program. Of course, poaching and habitat destruction impact on the Amur population should be carefully considered and monitored at the start of introduction program before any removal of tigers from Russian Far East.

Options of tiger removal Amur tiger population abundance Year 10 Year Year 30 Year 40 Year 50 20 No removal (baseline) 470 492 508 520 526

3 F+2M removal Year 20, 28 and 33 470 487 503 515 524 (Balkhash or Karatal TMU) 3 F+2M removal Year 20, 25 and 35 470 487 504 515 524 (Balkhash or Karatal TMU) 3 F+2M removal Year 20, 25, 30, 35, 470 487 499 508 522 and 40 (Ili Delta TMU) 3 F+2M removal Year 15, 18, 21, 24, 470 485 490 500 514 27, 30, 33, 36, 39, 42 (Ili Delta TMU) 3 F+2M removal Year 15, 18, 21, 24, 470 485 490 500 512 27, 30, 33, 36, 39, 42 and 45 (Ili Delta TMU)

CONCLUSIONS

157

• Beginning of introduction program. Years 10-15 of the program may be considered as the earliest possible option for tiger introduction in Balkhash and Karatal TMUs (first group of 3 females and 2 males) in the best circumstances (high population growth of wild boar first of all). High probability exists that the prey base could be still too low in the TMUs at that time to initiate tiger introduction. In the worst case, introduction in these TMUs can be started at years 20-25 when there could be sufficient prey abundance to support at least 5 tigers.

If the program starts with Ile River Delta TMU first group of 5 tigers may be brought on Year 5-10 of the program as the earliest possible option. As in the case of Balkhash and Karatal TMUs it is probably that prey base may be still low at Year 5-10. If true, Year 15 will likely to have sufficient prey base to start introduction.

• Number of tigers required for introduction. At least two groups of 5 tigers (3 females and 2 males) may be introduced in Balkhash and Karatal TMUs with an interval of 5-10 years between releases of the groups (Schemes 3, 4, and 16). Maximum 3 groups of tigers (15 tigers in total) might be possible to bring to Balkhash TMU (Scheme 6) in years 20-33 with relatively low probability that the population falls below quasi-extinction threshold due to prey depletion in the nearest 50 years. Three groups of tigers (15 tigers total) will likely to be possible to introduce to Karatal TMU if habitat is improved with fire management and grazing restrictions.

Ili River Delta TMU has much greater potential for introduction than other TMUs. Up to 40 tigers may be brought in the area in the Years 15-48 without risk of prey base depletion under No Fire Management Scenario (Scheme 9). Though, introduction of 50 tigers in the TMU at this scenario may be risky, because the population may exceed its carrying capacity and fall down (Scheme 10). Under Fire Management Scenario introduction of 50- 55 tigers in the TMU is possible, thought brining 55 tigers in the area may lead to decrease of prey abundance in the year 50 and following decline of tiger population.

• Maximal population of tigers that is likely to be achieved by the Year 50 of the program. Tiger abundance that is possible to achieve in Balkhash or Karatal TMUs by the Year 50 is 17-18 individuals if two groups of 5 tigers are introduced in each TMU (Schemes 3, 4, and 16). Introduction of three groups (15 tigers) in Balkhash TMU may results in 23-25 tigers by Year 50 if the area has sufficient prey base (Scheme 6). Under Fire Management Scenario introduction of 3 groups (15 tigers) in Karatal TMU will likely to result in 22-24 tigers by the same year. Introduction of 5 groups of tigers (25 tigers total) in Ile River Delta TMU may result in tiger population of 43-45 individuals by Year 50 (Scheme 8). 40 introduced tigers may provide maximal population of 64-69 individuals (Scheme 9) under No Fire Management Scenario and 66-70 individuals (Scheme 11) under Fire Management Scenario by Year 50. Release of 55 tigers in Ile

158

River Delta TMU under Fire Management Scenario is likely to achieve 92-98 individuals (population carrying capacity) by the Year 50 (Scheme 12). • Carrying capacity of TMUs. Carrying capacity of Balkhash TMU under No Fire Management Scenario and Fire Management Scenario does not exceed 24-26 tigers (adults, sub-adults, juveniles and cubs). Karatal TMU is not able to support tiger population more than 17-19 tigers under No Fire Management Scenario and 23-25 tigers under Fire Management Scenario. Ile River Delta TMU has much higher carrying capacity: 80-84 tigers under No Fire Management Scenario and 90-94 tigers under Fire Management Scenario.

All three TMUs may have entire tiger population of 121-129 individuals if no fire management and grazing restrictions are implemented in the area, and 137-145 animals if habitat are improved under Fire Management Scenario.

• Sustainability of tiger populations in the TMU. Tiger populations in Balkhash and Karatal TMU alone cannot be considered as sustainable in the long-term due to their small sizes. Water Scarcity Scenario clearly demonstrates that this population may go extinct in the nearest 50 years if water level in Balkhash Lake drops down more than by 2-3 meters (Schemes 7 and 18). Ili River Delta TMU is likely to be much more sustainable under Water Scarcity Scenario and may sustain 18-43 tigers after progressive water level decline in Ili River for the nearest 50 years (Schemes 13-15). Thus, tiger introduction in Balkhash and Karatal TMU may be considered as an option only in the case if Ili River Delta TMU is populated with tigers. Even we recommend to start introduction program with Ili River Delta TMU first: in this case tiger population in Ili River Delta may be used as a source population for tiger introduction in Balkhash and Karatal TMUs.

• Optimal number of Bukhara deer removed from Karachingil Game Management Area (GMA) for introduction in the TMUs. Removal of 30 deer annually for first 10 years of the program from Karachingil GMA for introduction in one of the TMUs may be considered as the most sustainable option for Karatchingil Bukhara deer population (Scheme 19). At the same time this option will allow to establish sustainable deer population (450-800 individuals) in one of the TMU that can be used as a source population for other TMUs. Removal 35 deer for first 9 years of the program is less sustainable, because the source population will have high probability to decrease below the half of the population carrying capacity (225 individuals) (Scheme 20). Removal of 40 deer for 8 years from Karachingil GMA is even less sustainable due to much higher probability of driving the source population below the half of its carrying capacity (Scheme 21).

• Number of tigers that may be removed from Amur tiger population in the Russian Far East for introduction. If poaching, habitat degradation and prey depletion do not

159

influence Russian Amur tiger population in the nearest 50 years it is likely to continue to grow despite removal of individuals for introduction in the Ili-Balkhash watershed (Scheme 22). Even removal of 11 groups of 5 tigers (3 females and 2 males each) – 55 tigers in total – in Year 15-45 of the program will not stop the population from growing under these ideal conditions of no poaching, prey depletion or habitat loss. Introduction of 25 tigers from outside into the Ili River Delta TMU will be enough to achieve population of 45-50 tigers (viable population for a short-term) in 50-60 years of the program. Of course, poaching and habitat destruction impact on the Amur population should be carefully considered and monitored at the start of introduction program before any removal of tigers from the Russian Far East is considered.

CITATIONS

Bragin, E.A. 2006. Review of Balkhash Lake conditions and politics in the field of water resource management in Ile-Balkhash watershed. Technical Report, WWF-Russia. Carroll, C., and D.G. Miquelle. 2006. "Spatial viability analysis of Amur tiger Panthera tigris altaica in the Russian Far East: the role of protected areas and landscape matrix in population persistence." Journal of Applied Ecology 43: 1056–1068. Danilkin, A.A. 2006. "Wild Ungulates and Game Management". GEOS. Moscow: 366. [in Russian] Flerov, K.K. 1935. "Prey mammals (Fissipedia) of Tajikistan." In Animals of Tajikistan, by B.S. Vinogradov, E.N. Pavlovskiy and K.K. Flerov, 131-200. Moscow: Academy of Sciences of USSR. Geptner, V.G. 1969. "Tiger." In Mammals of the Soviet Union. Cats and Hyenas, by V.G. Geptner, N.P. Naumov and A.G. Bannikov. Moscow. Geptner, V.G., Nasimovich, A.A., Bannikov, A.G. 1969. "Wild Boar". In Mammals of the Soviet Union, Even-toed and Odd-toed Ungulates, by by V.G. Geptner and N.P. Naumov. Moscow: 45-46 [in Russian]

160

Geptner, V.G., and A.A. Sludski. 1972. "Mammals of the Soviet Union." In Mammals of the Soviet Union, Carnivora (Hyaenas and Cats), by V.G. Geptner and N.P. Naumov. Moscow: Smithsonian Institution and the National Science Foundation, Washington, D.C. Global Tiger Recovery Program. 2010. Accessed August 2015, 2015. http://www.globaltigerinitiative.org. Goodrich, J.M., and D.G. Miquelle. 2005. "Translocation of Problem Amur Tigers to Alleviate Tiger-Human Conflicts." Oryx 39: 1-4. Hood, G.M. 2010. PopTool version 3.2.5. http://www.poptools.org. Ignatova, N.K., N.K. Khristoforova, and N.A. Chaus. 2004. "Population dynamics of wild boar and roe deer in Wildlife Refuges and game management areas of South-West of Primorsky Kray." Electronic journal "Studied in Russia" 1149-1161. http://zhurnal.ape.relarn.ru/articles/2004/107.html . Institute of Geography of Kazakhstan Academy of Sciences. 2013. Preparation of socio-economic component of long-term Program for tiger reintroduction in the Ili River delta and southern shore of Balkhash Lake. Research Report, WWF-Russia. Jungius, H., Yu. Chikin, O. Tsaruk, and O. Pereladova. 2009. Pre-feasibility study on the possible restoration of the Caspian tiger in the Amu Darya delta . Survey Report, WWF Russia. Karanth, K.U., J.D. Nichols, N. S. Kumar, and W.A., Hines, J.E. Link. 2004. "Tigers and their prey: Predicting carnivore densities from prey abundance." PNAS 101 (14): 4854–4858. Lukarevskiy, V.S., and R.D. Baidavletov. 2010. About field survey of Ili River valley as a potentila site for tiger reintroduction. Moscow: WWF Russia, Survey Report. Mayer, J. J. 2009. "Wild Pig Population Biology." In WildPigs: Biology, Damage, Control Techniques and Management, by J.J Mayer and I. L. Brisbin, 157-192. Aiken, South Carolina: Savannah River National Laboratory. Melis, C., M. Basille, I. Herfindal, J.D.C. Linnell, J. Odden, J-M. Gaillard, K-A. Hogda, and R. Andersen. 2010. "Roe deer population growth and lynx predation along a gradien of environmental productivity and climate in Norway." Ecoscience 17 (2): 166-174. Ministry of Natural Resources and Environment of Russian Federation. 2010. Strategy for Amur Tiger Conservation in the Russian Federation. Moscow: WWF Russia. Miquelle, D.G., Y.M. Dunishenko, D.A. Zvyagintsev, A.A. Darensky, A.M. Golyb, V.V. Dolinin, V.G. Shvetx, et al. 2009. "A Monitoring Program for the Amur Tiger Twelve-year Report: 1998-2009." Survey Report, Vladivostok. Miquelle D.G., Goodrich J.M., Kerley L.L., Pikunov D.G., Dunishenko Yu.M., Aramiliev V.V., Smirnov E.N., NIkolaev I.G., Salkina G.P., Zhang E., Seryodkin I.V., Carrol C., Gaponov V.V., Fomenko P.V., Kostyria A.V., Murzin A.A., Quigley H.B., Homocker M.G. 2010. Science-based conservation of Amur tigers in Russian far East and Northern China // Tigers of the World: the science, politics, and conservation of Panthera tigris, 2nd edition/ Elsevier Limited, Oxford, United Kingdom, p.399-419. Nilsen, E.B, Gaillard J-M., R. Andersen, J. Odden, D. Delorme, G. van Laere, and J.D.C. Linnell. 2009. "A slow life in hell or a fast life in heaven: demographic analyses of contrasting roe deer populations." Journal of Animal Ecology 78: 585–594. Pereladova, O.B. 2013. "Restoration of Bukhara deer (Cervus elaphus bactrianus Lydd.) in Central Asia in 2000-2011." IUCN Deer Specialist Group News 19-30. R Core Team. 2015. R: A language and environment for statistical computing . Vienna, Austria: R Foundation for Statistical Computing.

161

Sludskiy, A.A. 1953. "Tiger in the USSR." Matherials of the Academy of Science of the Kazakh SSR, Biology Seria 8: 8-73. Sludskiy, A.A. 1953. "Tiger in the USSR." News of the Academy of Sciences of Kazakhstan SSR 8: 18-73. Stroganov, S.U. 1961. "Tiger in Central Asia." Journal of Hunting and Game Management 12: 22-24. Tian, Y., J. Wu, A.T. Smith, T. Wang, X. Kou, and J. Ge. 2011. "Population viability of Siberian tiger in changing landscape: going, going and gone?" Ecological Modeling 222: 3166- 3180. WWF. 2014. "Tiger Reintroduction Program in Kazakhstan." Zaumyslova, O. Yu. 2005. "Ecology of wild boar in Sikhote-Alin Biosphere Reserve." In Tigers of Sikhote-Alin Nature Reserve: Ecology and Conservation. Vladivostok.

162

VITA

Name: Mikhail Yu. Paltsyn Date and Place of Birth: January 6th1971, Barnaul, Russia Education: Name and Location Dates Degree College: Lomonosov’s Moscow 1988-1993 Specialist State University, Moscow, Russia Graduate School: State University of New 2013-2018 Doctor of Philosophy York College of Environmental Science and Forestry, Syracuse, NY

Employment (selected): Employer Dates Title United Nations Development 2015-2018 International Consultant for Programme development and evaluation of UNDP/GEF projects in 15 countries SUNY-ESF 2014-2016 Research Assistant World Wide Fund for Nature 2009-2014 Senior Coordinator for WWF Altai-Sayan Programme United Nations Development 2006-2011 Endangered Species Expert Programme Group Leader for the UNDP/GEF Project “Biodiversity conservation in the Russian portion of Altai- Sayan Ecoregion” Arkhar NGO 2003-2007 Director Altaisky State Nature Reserve 2001-2004 Senior Researcher Katunsky Biosphere Reserve 1999-2001 Science Director Altaisky State Nature Reserve 1993-1999 Researcher and Ranger

163