Protecting rare and endangered species under climate change on the Qinghai Plateau, China

Abstract Climate change‐induced species range shift may pose severe challenges to species conservation. The Qinghai‐Tibet Plateau is the highest and biggest plateau, and also one of the most sensitive areas to global warming in the world, which provides important shelters for a unique assemblage of species. Here, ecological niche‐based model was employed to project the potential distributions of 59 key rare and endangered species under three climate change scenarios (RCP2.6, RCP4.5 and RCP8.5) in Qinghai Province. I assessed the potential impacts of climate change on these key species (habitats, species richness and turnover) and effectiveness of nature reserves (NRs) in protecting these species. The results revealed that that climate change would shrink the geographic ranges of about a third studied species and expand the habitats for two thirds of these species, which would thus alter the conservation value of some local areas and conservation effectiveness of some NRs in Qinghai Province. Some regions require special attention as they are expected to experience significant changes in species turnover, species richness or newly colonized species in the future, including Haidong, Haibei and Haixi junctions, the southwestern Yushu, Qinghai Nuomuhong Provincial NR, Qinghai Qaidam and Haloxylon Forest NR. The Haidong and the eastern part of Haibei, are projected to have high species richness and conservation value in both current and future, but they are currently not protected, and thus require extra protection in the future. The results could provide the first basis on the high latitude region to formulate biodiversity conservation strategies on climate change adaptation.

Climate change is driving shifts in species ranges (Parmesan et al., 1999) and the redistributions of species pose a huge challenge for the static boundary of current protected area (PA) networks (Chen, Hill, Ohlemüller, Roy, & Thomas, 2011;D'Amen et al., 2011;Zomer, Xu, & Wang, 2015). In order to adapt to climate change, most species will adopt countermeasures for migration, and move out of the PAs, which will counteract the conservation effectiveness of these PAs (Hannah et al, 2007;Hole et al., 2011). Meanwhile, climate change could bring new species into PAs, which will affect the conservation goal of the existing PAs as well as their management effectiveness (Araújo & Rahbek, 2006, Dawson, Jackson, House, Prentice, & Mace, 2011. Therefore, it is crucial to project the impacts of climate change on species habitat and turnover across time and space, which can greatly help conservation managers with development of adaptation strategies aimed at improving the effectiveness of PA networks and reducing the extinct risk of these key endangered species under the rapid climate change (Araújo, Cabeza, Thuiller, Hannah, & Williams, 2004;Dawson et al., 2011;Li, Xu, Wong, Qiu, Sheng, et al., 2015).
The impacts of climate change on species distributions also referred to as "species distribution models", have been generally assessed through ecological niche models. These niche-based models project species distributions by analyzing the relationships between species distributions and a number of environmental variables (Synes & Osborne, 2011;Virkkala & Lehikoinen, 2017). Although these relatively simple models may under-represent complex natural systems by neglecting competitive interactions, species plasticity, adaptation and time-lag (Davis, Jenkinson, Lawton, Shorrocks, & Wood, 1998;Pearson & Dawson, 2003), with a good understanding of the modeling techniques, and appropriate model validation and testing, they can be regarded as the primary tools for projecting species habitats and extinction risk, evaluating conservation priorities and assessing reserve designs (Akçakaya, Shm, Mace, Stuart, & Hiltontaylor, 2006;Duckett, Wilson, & Stow, 2013;Gallagher, Hughes, & Leishman, 2013). Therefore, they play a critical role in supporting spatial conservation planning, especially when conservation biologists are often pressed to make recommendations about conserving biodiversity based on limited species distribution data under climate change (Addison et al., 2013;Guisan et al., 2013).
Qinghai Province is situated in the northeast of the Qinghai-Tibet Plateau. As a traditionally sparsely inhabited region with a variety of different climatic zones and natural habitats, it provides important habitats for the Tibetan antelope (Pantholops hodgsonii), snow leopard (Panthera uncia), Procapra przewalskii and other key rare and endangered animals. It is the highest and biggest Plateau, one of the most sensitive regions to climate change in the world (Li, Powers, Xu, Zheng, & Zhao, 2018). Climate change will lead to higher temperatures and more precipitation in most areas in the year 2061-2080 under three RCPs from Global Circulation Model-HadGEM2-ES compared with the current climate condition. (Table 1), which could bring severe challenges for the regional biodiversity conservation (Chen et al., 2013;Duo, 2013). However, it has not been clear how climate change might affect the conservation of key rare and endangered species in Qinghai province. For the first time, this paper aims: (a) to project the potential climate change impacts on the habitats of the key rare and endangered species, species richness, and species turnover in Qinghai province; (b) to assess the efficacy of the existing nature reserves (NRs) for protecting these key species under future climate change; and (c) to comprehensively propose the adaptation strategies of biodiversity conservation in Qinghai province.

| Study area and species
Qinghai Province, the "water tower" of China and Asia, is located on the northeast part of Qing-Tibetan Plateau. It covers an area of over 720,000 square kilometers, one thirteenth of China's total area. Yangtze River, Yellow River and Lancang River, China's three major rivers all start in Qinghai province (Fang, 2013). Qinghai Province is administratively divided into eight prefecture-level divisions: two prefecture-level cities (Xining and Haidong) and six autonomous prefectures (Hainan, Haibei, Huangnan, Yushu, Guoluo, and Haixi).
The elevation in this Province ranges from 1,664 to 6,619 m, and its average elevation is over 3,000 m above sea level ( Figure 1). Most of the area is situated over 4,000 m above sea level-including the Qilian, Kunlun, Tanggula and other high mountain ranges.
In this study, I integrated species distribution data from two sources to achieve maximum representation of biodiversity and compensate for limitations in data availability on the high latitude region: (a) China key rare and endangered species database collected by The Nature Conservancy's China biodiversity blueprint project. This database has been successfully used to predict climate  (Proosdij, Sosef, Wieringa, & Raes, 2016). Therefore, I excluded these species with <15 presence points from the two databases, and obtained species presence data for 59 key rare and endangered animal species, which represents the indicator species of biodiversity conservation in Qinghai Province (see details in Li et al., 2018). Among these species, there were 39 species with more than 100 presence points, and 51 species with over 50 presence points. The minimum number of occurrences was 21.  Li et al., 2018;Li, Xu, Wong, Qiu, Sheng, et al., 2015). Slope and aspect were derived from a DEM with a resolution of 90 m, which was obtained from USGS. Global human footprint index at 1 km resolution was collected as human interference data, which integrates disturbance variables such as land-use change, infrastructure and population density (Sanderson, Jaiteh, Levy, Redford, & Wannebo, 2002). Because reliable future projection of human footprint index is not available, and including static variables in models alongside dynamic variables can improve model performance (Li, Xu, Wong, Qiu, Sheng, et al., 2015), I kept these variables static in our projections.

| Species distribution modeling and testing
The maximum entropy approach (Phillips, Anderson, & Schapire, 2006) was employed to project habitat suitability for 59 rare and endangered species on Qinghai Plateau, which has shown to be one of the best performing models in predicting species distributions with presence-only data (Elith et al., 2006;Hijmans & Graham, 2006), and it has been extensively applied to project species range and of five GCMs were used to produce probability outputs for each scenario. I performed 10 replications and a maximum of 500 iterations for each species, using a cross-validation procedure where I divided our dataset using 75% of the data for model calibration and retaining 25% of the data for evaluation. I calculated the average predicted probability of occurrence across the five GCMs for each grid as our ensemble forecast (Hole et al., 2009;Marmion, Parviainen, Luoto, Heikkinen, & Thuiller, 2009). Subsequently, I applied the Maximum Training Sensitivity Plus Specificity as the threshold to define the presence-absence distribution of species habitats, as this method has been found to be a robust approach (Fajardo, Lessmann, Bonaccorso, Devenish, & Muñoz, 2014;Liu, Berry, Dawson, & Pearson, 2005). The Areas under the Operating Characteristic Curve (AUC), a widely-used approach, was adopted to evaluate the model performance of our species models. As AUC is not appropriate to evaluate the accuracy of binary predictions, I also used true skill statistic (TSS) as suggested by recent studies (Li & Guo, 2013;Lobo, Jiménez-Valverde, & Real, 2008) to assess the accuracy of the studied species models.
The TSS takes into account both omission and commission errors, and success as a result of random guessing and ranges from −1 to +1, where +1 indicates perfect agreement and values of zero or less indicate a performance no better than random. It is a simple and intuitive measure for the performance of species distribution models when predictions are expressed as presence-absence maps (Allouche, Tsoar, & Kadmon, 2006;Mainali et al., 2016). I used both presence and background data for calculating AUC and TSS. The true absence data was unavailable in our study, so I used the randomly extracted background points within the whole study area for ROC analysis and calculate AUC and TSS (Please see Supporting Information Table S1).

| Assessment of climate change impacts in species habitats and assemblages
To estimate the sensitivity to climate change at the species level, I intersected the current and future habitat distribution maps to calculate the potential changes of species habitats. This allowed us to identify areas of the habitat range that are projected to be lost, gain or remain under future climate scenarios. Secondly, two indicators were chosen to evaluate the impact of climate change on species assemblages, including species richness and species turnover. Species richness was generated by calculating the number of species present in each 1-km 2 grid cell across the entire study region based on the binary distribution maps produced for the 59 species. Additionally, species turnover was also calculated (Broennimann et al., 2006) from a modification of the "classical" species turnover (beta-diversity) indicator (Lennon, Koleff, Greenwood, & Gaston, 2001;Whittaker, 1960). This index was measured in geographic space using a defined spatial neighborhood according to Equation 1. This index is usually used to measure the intensity of species change in a region with a range from 0 to 100 (Ramirez-Villegas et al., 2014).
It has a lower limit of zero when "species gain" and the "species loss" are both zero (this is generally not possible to happen), and 100 represents a complete species change from one period to another (i.e., the species gain or loss equals to the initial species richness and there is no loss or gain, respectively).

| Accuracy of species distribution models
All models for the 59 rare and endangered species achieved good or excellent performance, with high-average AUC scores and low omission rates (OR) at the 10% cumulative threshold value, indicating that these models had a high level of accuracy. According to the AUC model assessment criteria: 0.9-1.0 is excellent; 0.8-0.9, good; 0.7-0.8, general; 0.6-0.7, poor; 0.5-0.6, poor (Swets, 1988

| The impacts of climate change on spatial pattern of species richness
The spatial pattern of current species richness shows a general reduction from low altitude in southeast to northwest high altitude ( Figure 3a). The maximum value of species richness with 48 is found to occur in the east, north-east, and south of Qinghai Province.
Under future climate scenarios, species richness is predicted to increase in most areas by 2070, but the spatial pattern was similar to

| Conservation effectiveness and species turnover of NR network
Climate change-induced species range shifts would alter the conservation effectiveness of NR network in protecting these endan-

| D ISCUSS I ON
This study adopted niche-based models to project potential distribution pattern of the key rare and endangered species in Qinghai Province, and explored the adaptation conservation strategies under climate change. The results suggest that climate change would lead to the expansion of most rare and endangered species habitats, and the shrinking of a few species habitats. This is different from most of the existed research results about the impact of climate change on species habitats in other regions (Warren et al, 2013). There are two possible reasons for this simulation results: (a) Qinghai province is located at a high altitude, and climate change would lead to more climatically suitable habitats in this region; (b) It is assumed in the process of simulation that these animal species could freely migrate to any new climatically suitable areas.
However, due to the impact of natural and man-made barriers on species dispersal and migration, as well as the destruction and restrictions of habitat conditions outside NRs, most species would possibly confront habitat contraction under future climate change.
Limited dispersal scenario for species was not considered in this study. Universal dispersal means that species could disperse to any suitable places for population persistence in future, while limited dispersal means that species only can inhabit only places that are modeled to be suitable both in the present and in future (Li, Xu, Wong, Qiu, Sheng, et al., 2015;Thomas et al., 2004). Therefore, universal dispersal scenario was regarded as providing more useful information for implementing human-assisted adaptation to climate change in future conservation planning.
Under the assumption of species universal dispersal, the spatial distribution pattern of species richness and turnover would experience great changes due to the shifts of species habitats, although the conservation effectiveness of NR network as a whole would not First, there is a lack of sufficient information on biodiversity in Qinghai province. In particular, the habitat characteristics and threat factors of many endangered species are not completely clear. The ongoing investigation of animal and plant resources in Qinghai Province will help to understand the current status of key protected species, including population size, habitat distribution and threat level, which could further improve the accuracy of future assessment. Moreover, it is difficult for the current equilibrium simulation method to take into account any biotic interactions, such as competition with other species or other individuals, predation, and changes in food availability, which may lead to some uncertainty. Furthermore, the species distribution models cannot account for acclimation and adaptation of different species to future climate change. In reality, evolution and adaptation of species could be rapid and potentially help them counter stressful conditions or realize ecological opportunities arising from climate change. Thus, a future direction for improving predictive accuracy should incorporate evolutionary considerations and interspecific relationship of species, into predictive modeling or exploring the climate adaptive capacity of species to climate change.
In spite of this, this study employed widely-used method to conduct the first preliminary assessment of climate change impacts on rare and endangered species in Qinghai Province, which can provide a key reference for adapting biodiversity conservation strategies to climate change on the Qinghai Plateau.

This study was funded by National Key Research and Development
Project "Climate Change Impact and Adaptation in Major Economies

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

AUTH O R CO NTR I B UTI O N S
R.L. conceived and completed this study.

DATA ACCE SS I B I LIT Y
Data will be available at the external repository Dryad.