Quantitative exploration of the mechanisms behind the urban thermal environment in Beijing

The driving mechanism behind the formation of urban thermal environments is the result of a combination of factors. Beijing was chosen as the study area, and the technique of principal component analysis (PCA) was used. A spatial regression method was also applied for quantitative explanation of the thermal mechanism. Multiple Landsat thematic mapper images were used to quantify potential causing factors. Considering the eigenvalues of each factor and its relationship with land surface temperature, the first three principal components (PCs) are regarded as the main causative factors explaining the mechanism as independent variables. The first three PCs mainly reflect urban construction, road density and the normalized difference vegetation index (NDVI), respectively. Ordinary least squares, spatial lag and spatial error regression models were established separately for the relationships between the first three PCs and land surface temperature (LST). In the two spatial regression models, z-statistics for both the spatial lag parameter (ρ) and spatial residual parameter (λ) are significant, indicating the necessity of using spatial regression to replace the OLS regression model, as well as indicating that the spatial error regression model is superior to the spatial lag regression model. Overall, the normalized difference built-up index (NDBI) and road density are the most significant positive contributions to LST.