Identifying climate refugia and its potential impact on Tibetan brown bear (Ursus arctos pruinosus) in Sanjiangyuan National Park, China

Abstract Climate change has direct impacts on wildlife and future biodiversity protection efforts. Vulnerability assessment and habitat connectivity analyses are necessary for drafting effective conservation strategies for threatened species such as the Tibetan brown bear (Ursus arctos pruinosus). We used the maximum entropy (MaxEnt) model to assess the current (1950–2000) and future (2041–2060) habitat suitability by combining bioclimatic and environmental variables, and identified potential climate refugia for Tibetan brown bears in Sanjiangyuan National Park, China. Next, we selected Circuit model to simulate potential migration paths based on current and future climatically suitable habitat. Results indicate a total area of potential suitable habitat under the current climate scenario of approximately 31,649.46 km2, of which 28,778.29 km2 would be unsuitable by the 2050s. Potentially suitable habitat under the future climate scenario was projected to cover an area of 23,738.6 km2. Climate refugia occupied 2,871.17 km2, primarily in the midwestern and northeastern regions of Yangtze River Zone, as well as the northern region of Yellow River Zone. The altitude of climate refugia ranged from 4,307 to 5,524 m, with 52.93% lying at altitudes between 4,300 and 4,600 m. Refugia were mainly distributed on bare rock, alpine steppe, and alpine meadow. Corridors linking areas of potentially suitable brown bear habitat and a substantial portion of paths with low‐resistance value were distributed in climate refugia. We recommend various actions to ameliorate the impact of climate change on brown bears, such as protecting climatically suitable habitat, establishing habitat corridors, restructuring conservation areas, and strengthening monitoring efforts.

Researchers have simulated models that minimize extinction risk by identifying species and habitats susceptible to climate change and how wildlife may respond to large scale environmental shifts (Balzotti, Kitchen, & McCarthy, 2016;Foden et al., 2013;Guisan et al., 2013). Species distribution models (SDMs) use environmental variables to explain both current and future distributions Struebig et al., 2015). SDMs have become essential for approaching research challenges in fields such as biogeography, evolution, ecology, and conservation biology (Guisan & Thuiller, 2005).
At present, researchers have used the maximum entropy (MaxEnt) model to assess the habitat suitability for a variety of rare or endangered wildlife around the world Li, Liu, Xue, Zhang, & Li, 2017;Li et al., 2019;Zhang, Jiang, et al., 2019;Zhang, Clauzel, et al., 2019). Furthermore, the Circuit model is commonly used by natural resource managers to predict wildlife dispersal paths and design ecological corridors, which are often used in wildlife management and conservation practice McRae & Beier, 2007;McRae, Shah, & Mohapatra, 2013;Walpole, Bowman, Murray, & Wilson, 2012;Zhang, Clauzel, et al., 2019).
Paleoecological records and observed species migrations indicate that species distributions follow suitable climatic conditions (Parmesan & Yohe, 2003). However, changes in distributions are limited by climate, landscape features, and dispersal potential (Lambers, 2015;Littlefield, McRae, Michalak, Lawler, & Carroll, 2017). Refugia are areas which possess relatively stable climatic conditions with high connectivity between suitable habitat in different climate scenarios (Littlefield et al., 2017;Morán-Ordóñez, Briscoe, & Wintle, 2017). Knowledge of current and future habitat refugia of threatened species, such as the Tibetan brown bear (Ursus arctos pruinosus) is vital in designing conservation plans aimed at promoting long-term species persistence.
The Tibetan brown bear, also known as the Tibetan blue bear, is a rare brown bear subspecies living at high altitudes in close proximity to humans in Asia (Aryal et al., 2012;Aryal, Sathyakumar, & Schwartz, 2010;Xu et al., 2006; Figure 1). The species population estimate is 5,000-6,000 individuals (Wu, 2014). Sanjiangyuan National Park of China provides important habitat and migration corridors for the species. At present, their primary threat is habitat reduction and fragmentation (Aryal et al., 2010;Coulon et al., 2004;Littlefield et al., 2017;McRae & Beier, 2007). A habitat assessment for Tibetan brown bears by Wu (2014) used species distribution data and eco-geographic variables, combined with Generalized Linear Models, to assess species-appropriate habitat in the Suojia region of Sanjiangyuan National Park. However, this study did not consider climate change, leaving a substantial knowledge gap in our understanding of its potential impacts on species distribution. This research constructed a projected distribution model for brown bears based on presence data and related bioclimatic and environmental factors. We used Circuit model to simulate the potential movement paths of brown bears under both current and future climate scenarios. The aims of this work were to (a) project current and future climatically suitable habitat for brown bears, (b) identify climate refugia, and (c) recognize dispersal paths that allow for migration from current to future suitable habitat. Our findings will be incorporated into a brown bear protection plan in the context of global climate change in Sanjiangyuan National Park, China.

| Study area
Sanjiangyuan National Park is China's first pilot national park. It lies in the hinterland of the Tibetan Plateau (between 89°50′ and 99°14′E, 32°22′ and 36°47′N), spanning an area of 123,100 km 2 , 14 times larger than Yellowstone National Park ( Figure 2). The altitude is between 3,500 and 4,800 m. It is a plateau continental climate.
The weather is typically dry and cold with the annual average temperature ranging from −5.6°C to −7.8°C and the annual precipitation consisting predominately of snowfall ranging from 262.2 mm to 772.8 mm. Sanjiangyuan, or Source of Three Rivers, refers to the area's role as the headwaters of China's three largest rivers (Yangtze river, Yellow river and Lancang river). The region has global influence and dictates China's ecosystem (Zhang, Jiang, et al., 2019). A variety F I G U R E 1 Tibetan brown bear (Ursus arctos pruinosus) captured by camera trapping in the Yangtze River Zone of Sanjiangyuan National Park, China of endemic alpine flora and fauna constitutes Sanjiangyuan's excessive biological diversity. As China's first national park, it has become an exhibition area of nature protection and ecological culture heritage on the Tibetan Plateau (Zhang, Jiang, et al., 2019). Unfortunately, it is also one of the most sensitive regions to climate change Wang, Song, & Hu, 2010).

| Data preparation
We collected 528 GPS coordinates of brown bear. Among them, 315 were obtained via ground surveys from 2016 to 2018 (recorded coordinates of brown bear presence, including feces, footprints, hair, and foraging traces), 65 from infrared camera traps and 148 from published literature (Wu, 2014;Xu et al., 2006; Figure 2). To reduce autocorrelation, presence points were filtered by randomly selecting one point in each 1 km 2 grid (Aryal et al., 2016;Li et al., 2017;Zhang, Jiang, et al., 2019).

F I G U R E 2 Location of Sanjiangyuan National Park, China
The 19 bioclimatic factors (at 1 km resolution) undercurrent (average for 1950-2000) and future climatic conditions (average for 2041-2060) were extracted from the WorldClim database (http://www.world clim.org/version1). Future climate data consisted of IPCC and the Coupled Model Intercomparison Project 5 (CMIP5) climate projection (Wu, Chen, & Wen, 2013) (Baek et al., 2013). Under the RCP4.5 scenario, the global average temperature would rise by 0.9-2.0°C by the 2050s, consistent with the expectations of the Paris Agreement (UNFCCC, 2015). Previous studies (Choi, Lee, & Oh, 2011;Ahn et al., 2013;Baek et al., 2013;Hong & Ahn 2015) have found HadGEM2-AO to have satisfactory performance in simulating climate change trends and general patterns of current climate over the East Asia region. In addition, previous studies have used HadGEM2-AO to construct species distribution models under the future climate scenario in China (Li, Li, Xue, et al., 2018;Li et al., 2017Li et al., , 2019. Therefore, we elected to use HadGEM2-AO for predicting future distributions of Tibetan brown bear.
All spatial variables (climate and nonclimate) were resampled to 1 km resolution and unified in a projection coordinate system (WGS_ 1984_UTM_Zone_47N)

| Habitat suitability model
MaxEnt model is considered one of the most efficient tools to predict species distribution with presence-only data, leading to its widespread use (Aryal eT al., 2016;Gomes et al., 2018;Lamsal, Kumar, Aryal, & Atreya, 2018;Ma & Sun, 2018;Phillips, Anderson, & Schapire, 2006). The parameters of MaxEnt model were set to: 25% for random test percentage and 1 regularization multiplier.
We ran 15 replicates and performed a cross validation (Phillips et al., 2006;Vedel-Sørensen, Tovaranonte, Bøcher, Balslev, & Barfod, 2013). Percent contribution was used to estimate the importance of variables. The logistic results of MaxEnt were regarded as the probability of species occurrence, with values ranging from 0 to 1. A threshold value was used to distinguish between suitable and unsuitable regions. The average logistic threshold value of maximum training sensitivity plus specificity (MTSPS) was recommended (Liu, White, & Newell, 2013). Grids with probability values greater than the threshold were deemed suitable habitat. We then withdrew suitable patches with areas <10 km 2 based on the known minimum home range of brown bears (Nagy & Haroldson, 1990). We evaluated MaxEnt model performance by using the area under the receiver operating characteristic curve (AUC). AUC is an independent threshold value to verify the accuracy of model outputs. Values range from 0 to 1, with those closer to 1 indicating a more accurate model (Araujo, Pearson, Thuiller, & Erhard, 2005;Phillips et al., 2006). where A c is the projected area of current suitable habitat; A f is the projected area of future suitable habitat; and A cf is the area of climate refugia.

| Geographical features of climate refugia
Altitude characteristics and typical land use types of climate refugia for brown bears were analyzed by overlaying the climate refugia map with the layers of land use types and altitude in ArcGis 10.1.

| Habitat connectivity analysis
We simulated the potential migration paths for brown bears based on current and future habitat connectivity by using Circuit model (Circuitscape software 4.0; https ://circu itsca pe.org/; Li et al., 2019;McRae & Beier, 2007;McRae et al., 2013;Walpole et al., 2012;Zhang, Clauzel, et al., 2019). The model mode, calculation, and mapping options for Circuitscape were set to: Pairwise mode (run in lowmemory mode), use average conductance instead of resistance for connections between cells, write cumulative and max current maps only, and set focal node currents to zero. We inverted the habitat suitable index (HSI) value to link the suitable habitat of brown bears with low movement resistance and vice versa. We used the functions of negative exponential transformation to convert HSI into resistance values (Keeley, Beier, & Gagnon, 2016):  Figure 3).

| Changes in potential suitable habitat
The average logistic threshold value of MTSPS was 0.3562. Cells with a value of habitat suitability <0.3562 covered an area of 31,649.46 km 2 under the current climate scenario in Sanjiangyuan National Park (Table 2). Potential suitable habitat for brown bears was primarily distributed in the southeastern region of the Yangtze River Zone, northwestern region of the Lancang River Zone, and northern region of the Yellow River Zone (Figure 4a). Under the future climate scenario, the area of potential suitable habitat was projected to be 26,609.77 km 2 (Table 2), a reduction of 5,039.69 km 2 .
Habitat reduction primarily occurred in the Lancang River Zone, the southeastern region of the Yangtze River zone and the northeastern region of the Yellow River Zone (Figure 4b).
We found that potential suitable habitat of brown bears in all regions except the Yellow River Zone (AC = 62.78%) would decrease under the future climate scenario. In the 2050s, the potential suitable habitat area of brown bears in the Yangtze River Zone was reduced by 469.6 km 2 (AC = −2.02%), and the potential suitable habitat area of brown bears in the Lancang River Zone reduced by 6,064.48 km 2 (AC = −100%). Suitable habitat in the Yellow River Zone increased  Figure 4b).

| Vulnerability assessment
Due to climate change, 28,778.29 km 2 (SH c = 90.93%) of current suitable brown bear habitat was predicted to be vulnerable. Areas of potential suitable habitat under the future climate scenario covered 23,738.6 km 2 (SH f = 89.21%) and were mainly distributed in the northwestern and northeastern region of the Yangtze River Zone and northern region of the Yellow River Zone. The area of climate refugia was 2,871.17 km 2 , and aggregated in the midwestern and northeastern regions of the Yangtze River Zone, as well as the northern part of the Yellow River Zone (Table 2, Figure 5).

| Brown bears' potential movement paths
The

| D ISCUSS I ON
The decline in habitat connectivity and quality has led to a highly fragmented distribution of Tibetan brown bears (Aryal et (Aryal et al., 2012;Nawaz et al., 2014;Wu, 2014), have been degraded due to overgrazing in the Sanjiangyuan region (Li, Brierley, Shi, Xie, & Sun, 2012;Zhou, Zhao, Tang, Gu, & Zhou, 2005). More recent implementation of conservation measures in national parks by the Chinese government has resulted in better protection for its wildlife and their habitats ). Yet, climate change remains of great concern, as it may negate current conservation efforts, including those set to protect Tibetan brown bears (Balzotti et al., 2016;Stephens et al., 2017;Su et al., 2018).
Assessing climatically suitable habitat is a key step in developing proactive strategies that reduce the impacts of climate change on the brown bear.

| Habitat analysis in current and future climate scenarios
The Himalaya region encompasses significant habitats for the brown bear (Aryal et al., 2012). However, the distribution range and suitable habitat area of brown bears in the Himalaya region have changed significantly since the 1990s. These changes were primarily caused by habitat fragmentation and loss (Nawaz et al., 2014).
At present, there are large areas of suitable habitat distribution of anticipated to be severe (Su et al., 2018). Possible outcomes which could result without adequate conservation action include brown bear population crashes due to insufficient habitat and movement into residential areas would cause conflicts between humans and brown bears (Dai, Xue, Cheng, et al., 2019).

| The altitude range of climate refugia and its typical land use types
Tibetan brown bears are adapted to high altitudes and can be found above 5,000 m (Wu, 2014). We found the climate refugia of brown bears had an altitude range of 4,307-5,524 m, with over half of climate refugia located between 4,300 and 4,600 m. Tibetan brown bears occupy a large elevational gradient corresponding with daily activity budgets. For examples, brown bears usually move to higher altitudes in the morning (before 10 hr) looking for rocks or dens to rest, and then descend to lower altitudes in the afternoon (after 18 hr) to forage (Wu, 2014).
Rocky outcrops along mountainsides provide brown bears areas of refuge for rest and concealment in the form of natural dens, while alpine steppe and alpine meadow house food sources for brown bears, such as marmots (Marmota himalayana) and pikas (Ochotona curzoniae). Hence, this type of land use is essential for the survival of brown bears and therefore needs to be greatly protected.

| Potential migration paths for brown bears
We simulated potential migration routes for brown bears based on Circuit modeling and found that low-resistance areas were primarily divided into three isolated parts under the current climate scenario: the southeastern region of the Yangtze River Zone, central region of the Lancang River Zone, and north central region of the Yellow River Zone (Figure 7a). Despite a low-resistance region within the Lancang River Zone, brown bears movement to the Yangtze River Zone would be difficult, as the current was low at the border between the Yangtze River Zone and the Lancang River Zone. This may hinder gene flow and dispersal between these populations, ultimately reducing genetic diversity and decreasing species adaptability. The Yellow River Zone was found to be a highcurrent area; nonetheless, because of the Yellow River Zone being geographically distant from the other two zones, along with absence of protected areas to serve as stepping stones, connectivity between brown bear populations in the Yangtze and Lancang River Zones is restricted.
Species ability to track suitable habitat in the future strongly depends on population dynamics and dispersal processes that evolve over time (Early & Sax, 2011). Current high-current regions would dramatically decrease by the 2050s, with potential movement routes substantially shrinking. Emerging high-current regions in the western area of the Yangtze River Zone would be relatively remote from current brown bear populations and hence would likely not facilitate adequate movement to new habitat. More meaningful regions for brown bear migration will depend on the persistence of high-current routes in climate refugia.

| Protecting climatically suitable habitat
Habitat quality and loss directly affect how wild animals exploit the resources available to them (Hiller, Belant, & Beringer, 2015). Loss of brown bear habitat causes shortages in natural food availability, which may increase dependence on food linked to anthropogenic sources, increasing levels of livestock depredation and human-bear conflicts (Dai, Xue, Cheng, et al., 2019;Su et al., 2018). Currently, suitable areas should be strictly protected to avoid loss of habitat and natural food sources with priority given to localities less susceptible to climate change. In these areas, it is paramount that the grassland is preserved via reduction of livestock grazing intensity. Compensation programs aimed at sustainable grazing practices may be an adequate solution for encouraging local communities to play a more active role in conserving their environment and complying with government regulations while still maintaining financial livelihoods.

| Establishing potential corridors
Choosing appropriate regions to establish ecological corridors between isolated habitat patches would be one of the most effective techniques for facilitating dispersal between brown bear populations (Ramiadantsoa, Ovaskainen, Rybicki, & Hanski, 2015). These

| Strengthening monitoring on brown bears
Most master plans for protected areas only address strategies to combat the early impact stages of climate change (Xu et al., 2017).
It is not possible to holistically understand how wildlife will respond to climate change and what management strategies would be most effective (Li, Li, Xue, et al., 2018). Therefore, long-term scientific standardized monitoring should be implemented in Sanjiangyuan National Park to regularly assess changes in the population status and habitat of brown bears. Action plans can then be developed as changes materialize, providing timely and continuous efforts to preserve this at-risk species.

ACK N OWLED G M ENTS
This research was supported by National Key R&D Program of China (2017YFC0506405). We thank Sanjiangyuan National Park Administration and Qinghai Forestry and Grassland Administration for their ground survey support for this research.

CO N FLI C T O F I NTE R E S T
None declared.