A geographically weighted regression model augmented by Geodetector analysis and principal component analysis for the spatial distribution of PM2.5

This study develops an augmented geographically weighted regression (GWR) model to analyze the spatial distribution of PM2.5 concentrations through the incorporation of Geodetector analysis and principal component analysis (PCA). The modeling approach we propose allows an effective identification of important PM2.5 drivers and their spatial variation. Geodetector analysis is used to select predictor variables that truly affect the dependent variable, and PCA is adopted to eliminate multicollinearity among the variables. The spatial distribution of PM2.5 concentrations within the Pearl River Delta region, China, is analyzed using the augmented GWR model. The augmented GWR model has an obvious advantage of parsimony. Moreover, it significantly outperforms the traditional regression model.