Estimating 1-km-resolution PM2.5 concentrations across China using the space-time random forest approach

Fine particulate matter (PM2.5) is closely related to the atmospheric environment and human life. Satellite-based aerosol optical depth (AOD) products have been widely applied in estimating daily PM2.5 concentrations over large areas using statistical regression models. However, they are often given at coarse spatial resolutions which limit their applications in small or medium scales. This study aims to produce PM2.5 concentrations at a high spatial resolution (1 km) across China based on the newly released Moderate Resolution Imaging Spectroradiometer (MOD1S) Collection 6 Multi-Angle Implementation of Atmospheric Correction (MAIAC) AOD product using a newly developed space-time random forest (STRF) model. Daily PM2.5 concentrations for 2016 were estimated from in-situ surface PM2.5 measurements, and meteorological and ancillary variables. The 10-fold cross-validation (CV) approach and three popular models, including the multiple linear regression model, the geographically weighted regression model, and the two-stage model, are employed for validation and cross comparison. A sample-based CV of the STRF model shows a high and stable accuracy with a coefficient of determination equal to 0.85, a root-mean-square error of 15.57 mu g m(-3), and a mean prediction error of 9.77 mu g m(-3). This finding suggests that the STRF model can predict PM2.5 daily, monthly, and annual concentrations at an unprecedented spatial resolution and accuracy across China. It also appears to have out-performed the above popular models and the previous related studies. In general, the STRF model is robust and can accurately estimate PM2.5 concentrations by taking advantage of the ensemble regression approach, the synergy of space-time information, and the high-resolution, high-quality, and wide-spatial-coverage of the MAIAC AOD product. It may thus also be useful for applications in related air pollution studies, especially those focused on urban areas.