Spatially varying relationships between surface urban heat islands and driving factors across cities in China

Global regression models, such as Ordinary Least Squares, are generally used to explore driving factors of surface urban heat island (SUHI) effects across large cities on a national level, but issues of spatial non-stationarity or local variations have rarely been taken into account. Our study quantifies SUHI effects for 274 cities in China with MODIS LST products and explores spatially varying relationships between SUHI intensity (SUHII) and their driving factors using geographically weighted regression (GWR). The results show that GWR models have stronger explanatory power and lower spatial autocorrelations of residuals compared with ordinary least square models; the application of GWR models finds that the relationships between SUHII and the driving factors vary across China. Spatially varying coefficients from GWR models could contribute to the development of local-specific urban planning or policies in different regions. The findings from our investigation suggest that GWR has the potential to serve as a useful tool for environmental investigations on a national scale