Using geographically weighted regression to explore the effects of environmental heterogeneity on the space use by giant pandas in Qinling Mountains

The effects of environmental heterogeneity on the distribution of wildlife are obviously uneven over space. Traditional approaches, such as classical linear regression model, are unable to accurately depict the spatial variations in species-environment relationship. Geographically weighted regression(GWR) is a newly proposed spatial regression method that shows promise in detecting spatial variability of environmental relationship through embedding spatial structure into linear regression model. Taking giant pandas(Ailuropoda melanoleuca) in Qinling Mountains as an example, we used the GWR method to analyze the potential relationship between the spatial distribution of giant pandas and environmental heterogeneity. We also compared the results of GWR with those of the classical global ordinary least squares regression(OLS). The results show that AIC, R~2 and adjust R~2 of GWR model were significantly better than those of OLS model. Local regression coefficients of GWR model can reveal the complex spatial relationship between the spatial distribution of giant pandas and environmental variables, and provide more effective theoretical support for the scientific protection of species. Therefore, the GWR is an effective tool for exploring the spatial heterogeneity of species-environment relationship, which would have a wide application prospect in the research of species habitat selection and utilization.