资源环境科技发展态势分析平台

Fengyun-4A (FY-4A) is a new generation geostationary satellite that provides high temporal resolution atmospheric observations of China and the adjacent regions. This study proposed to assimilate FY-4A observations via a combined utilization of a 3D variational method and a random forest approach. The ground-level PM_2.5 concentrations were estimated as an intermediate variable and subsequently assimilated based on the Gridpoint Statistical Interpolation (GSI) system. Four parallel experiments were conducted to verify the proposed method, including a control experiment and three data assimilation experiments that assimilated satellite observations and ground observations alone and simultaneously. Results show that the proposed approach improved PM_2.5 predictions for most sites, especially in the highly polluted Beijing-Tianjin-Hebei and Yangtze River Delta regions. Assimilating satellite-based PM_2.5 estimations showed an advantage over the assimilation of ground PM_2.5 measurements in places where local and upstream regional PM_2.5 monitoring sites are sparse. Simultaneous assimilation of the satellite-based PM_2.5 estimations and ground observations can further improve the accuracy of the PM_2.5 predictions in most places.