Estimating Ground-Level PM2.5 Concentrations from Satellite AOD in Central China Using Seasonal-Differential Geographically and Temporally Weighted Regression Model during 2015-2017

Compared to the research studies concentrated in the highly populated urban areas of China on the estimation of surface PM_2.5 (particulate matter with aerodynamic diameter less than 2.5 µm) from satellite aerosol optical depth (AOD), it remains less intensive over the Central China. In this study, a 10 km ×10 km AOD datasets using the Moderate Resolution Imaging Spectroradiometer (MODIS) Collection 6.1 were employed to estimate the ground PM_2.5 concentrations over central China covering Henan-Hubei-Hunan provinces from January 2015 to December 2017. A seasonal-differential geographically and temporally weighted regression (SD-GTWR) model was developed with the investigation in the base of data available including hourly PM_2.5 concentrations, meteorological fields (relative humidity, wind speed, temperature, pressure), atmospheric boundary layer height (ABLH), and land use variables as explanatory predictors. The results showed good agreements between the satellite-derived and ground-based PM_2.5 concentrations over the study domain. The overall cross-validation (CV) R² is 0.88 and the root mean squared prediction error (RMSE) is 6.22 µg m^-3 for the MODIS derived AOD. The GTWR model with seasonal differential optimization has better fitting accuracy than the traditional GTWR, least square model (OLS), and geographically weighted regression (GWR) methods. The results obtained from this model are useful in developing the techniques for the predicted surface PM_2.5 from satellite AOD and may serve as a reference for Central China.