Reconstruction of 3D precipitation measurements from FY-3G MWRI-RM imaging and sounding channels

Abstract

The Fengyun-3G satellite (FY-3G), China's first precipitation measurement satellite, was launched on 17 April 2023. FY-3G carries the advanced Multi-Channel Microwave Radiance Imager-Rainfall Measurement (MWRI-RM) system, which, compared to the previous GPM/GMI, includes more sounding channels. Additionally, a Ka/Ku-band dual-frequency Precipitation Measurement Radar (PMR) on board FY-3G provides 3D observations of severe precipitation systems. Due to the high cost and hardware limitations of precipitation radars, most precipitation-affected satellite observations rely on passive data. Deep learning methods have become effective tools to bridge these two types of observations. In this study, we propose a deep convolutional neural network (CNN) to reconstruct PMR-Ku reflectivity profiles based on MWRI-RM multi-channel radiances across different precipitation scenarios and analyze the effects of dual oxygen absorption sounding channels and polarization differences (PDs) on reconstruction outcomes. Experimental results demonstrate that incorporating dual oxygen absorption channels substantially improves reflectivity reconstruction accuracy, particularly over land (root mean square error reduction: 5.43 % ± 1.56 %) and coastal regions (5.47 % ± 1.17 %). In contrast, polarization differences provided only marginal improvements (0 %-1.6 % RMSE reduction), with statistical significance constrained within the uncertainty bounds of model training variability. Validation against extreme precipitation events - including Typhoon Khanun (2023) and the "07.2023"(for July 2023) Beijing-Tianjin-Hebei (BTH) extreme rainfall event - confirms the operational value of dual oxygen channels in resolving three-dimensional precipitation structures. These enhancements not only improved accuracy but also enabled a more comprehensive three-dimensional representation of precipitation systems.