Estimating Regional PM2.5 Concentrations in China Using a Global-Local Regression Model Considering Global Spatial Autocorrelation and Local Spatial Heterogeneity

Linear regression models are commonly used for estimating ground PM2.5 concentrations, but the global spatial autocorrelation and local heterogeneity of distribution either ignored or only partially considered in concentrations. Therefore, taking both into consideration, a global-local (GLR) model is proposed concentrations Yangtze River Delta (YRD) Beijing, Tianjin, Hebei (BTH) regions China based on aerosol optical depth data, meteorological remote sensing pollution source data. Considering autocorrelation, GLR extracts factors by eigenvector filtering (ESF) method, combines fraction them that passes further with geographically weighted (GWR) method to address heterogeneity. Comprehensive results show outperforms ordinary GWR ESF models, has best performance at monthly, seasonal, annual levels. The average adjusted R2 monthly YRD region (the BTH region) 0.620 (0.853), which 8.0% 7.4% (6.8% 7.0%) higher than respectively. cross-validation root mean square error 7.024 ?g/m3 region, 9.499 lower models. can effectively heterogeneity, overcome shortcoming overfocuses features disadvantage poor model. Overall, good temporal applicability promising