An ecophysiological perspective on likely giant panda habitat responses to climate change

Threatened and endangered species are more vulnerable to climate change due to small population and specific geographical distribution. Therefore, identifying and incorporating the biological processes underlying a species' adaptation to its environment are important for determining whether they can persist insitu. Correlative models are widely used to predict species' distribution changes, but generally fail to capture the buffering capacity of organisms. Giant pandas (Ailuropoda melanoleuca) live in topographically complex mountains and are known to avoid heat stress. Although many studies have found that climate change will lead to severe habitat loss and threaten previous conservation efforts, the mechanisms underlying panda's responses to climate change have not been explored. Here, we present a case study in Daxiangling Mountains, one of the six Mountain Systems that giant panda distributes. We used a mechanistic model, Niche Mapper, to explore what are likely panda habitat response to climate change taking physiological, behavioral and ecological responses into account, through which we map panda's climatic suitable activity area (SAA) for the first time. We combined SAA with bamboo forest distribution to yield highly suitable habitat (HSH) and seasonal suitable habitat (SSH), and their temporal dynamics under climate change were predicted. In general, SAA in the hottest month (July) would reduce 11.7%-52.2% by 2070, which is more moderate than predicted bamboo habitat loss (45.6%-86.9%). Limited by the availability of bamboo and forest, panda's suitable habitat loss increases, and only 15.5%-68.8% of current HSH would remain in 2070. Our method of mechanistic modeling can help to distinguish whether habitat loss is caused by thermal environmental deterioration or food loss under climate change. Furthermore, mechanistic models can produce robust predictions by incorporating ecophysiological feedbacks and minimizing extrapolation into novel environments. We suggest that a mechanistic approach should be incorporated into distribution predictions and conservation planning.