A multi-scale Maxent approach to model habitat suitability for the giant pandas in the Qionglai mountain, China

Wild animals usually respond to different landscape features at different spatial scales. The adoption of multi-scale modeling frameworks in habitat suitability modeling studies have been shown to improve model performance and provide greater insights into relationships between species and habitat components. Although the advantage of multi-scale modeling, the implementation of this framework lagged considerably. In the present study, we used a multi-scale approach to assess the habitat suitability for the globally endangered giant panda (Ailuropoda melanoleuca) in Qionglai mountain range, Sichuan, China in an effort to provide improved species-environment relationships with the aim of informing conservation efforts. The occurrence data collected from the Fourth National Giant Panda Survey and a presence-only, multi-scale Maxent approach were used to model habitat suitability for giant pandas. Our results showed that the optimal scale identified for each environmental variable varied, and most variables were strongly related to giant panda habitat suitability at a relatively fine-scale (≤ 2000 m). Multi-scale models outperformed their analogous single-scale counterparts with respect to discrimination and predictive ability. Additionally, there were significant differences in spatial predictions between multi-scale and single-scale model. This study reveals the multi-scale response of giant pandas to environmental features and confirms the advantage of multi-scale habitat modeling. Therefore, it is necessary and beneficial to take scale dependence into consideration in future habitat suitability modeling for the giant panda.