Precipitation remains one of the largest sources of uncertainty in weather and climate models, and the launch of China's Fengyun-3G (FY-3G) satellite promised to narrow that gap with its advanced MWRI-RM radiometer. Yet the mission lacked a dedicated algorithm to turn its 26-channel passive microwave observations into an operational precipitation product. That gap has now been filled by researchers from the University of Science and Technology of China, the National Satellite Meteorological Center, the Meteorological Observation Centre of the China Meteorological Administration, and collaborators, who developed the first systematic retrieval algorithm for FY-3G's MWRI-RM. The work was Published (DOI: 10.34133/remotesensing.1057) on August 11, 2026, in the Journal of Remote Sensing.
The algorithm couples Bayesian retrieval with a physically optimized prior database to produce a multiple-hydrometeor profile (MHP) product that describes surface rainfall and vertical cloud and ice structure. In comparisons with two established Global Precipitation Monitoring products—GPROF and CMB—the MHP estimates showed relative mean biases generally within 14% over ocean and land. Calibration tests further demonstrated strong agreement between simulated and observed brightness temperatures, with Pearson correlations above 0.94 under clear skies and RMSEs below 2.5 K over land.
What sets the new method apart is its treatment of cloud liquid water. The MHP product exhibited a markedly stronger relationship between cloud liquid water path and precipitation intensity than either comparison product, indicating a more physically realistic cloud–rain linkage. In a typhoon case, all three products captured the eyewall and three spiral rainbands, but MHP achieved a cloud-liquid-water-to-rainfall spatial correlation of 0.84, versus 0.72 for GPROF and 0.52 for CMB. In a land rainfall case, the correlations were 0.84, 0.76, and 0.47, respectively. Validation against 2,486 rain gauges showed that extending accumulation from 10 to 60 minutes increased correlation from 0.455 to 0.542 and reduced RMSE from 1.743 to 1.341 mm/h.
These improvements matter because passive microwave satellites can monitor precipitation over oceans, mountains, and regions poorly covered by ground instruments, while providing wider coverage than radar. An operational product that retrieves rainfall together with cloud and frozen-water structure could strengthen typhoon and flood monitoring, support cloud–precipitation research, and improve microphysical parameterization in climate models. The authors note that the MHP product "exhibits good consistency with GPROF and CMB" and that its precipitation-related variables can reveal storm microphysics. Remaining biases include heavy-rain underestimation and uncertainties related to surface type and ice-particle assumptions, with oceanic rainwater path underestimated by 27% versus GPROF and 19% versus CMB.
The team built the retrieval around a prior database linking hydrometeor profiles and atmospheric states with microwave brightness temperatures, combining GMI observations, GPROF profiles, ERA5 variables, and radiative-transfer calculations. A one-dimensional variational method adjusted cloud liquid water, rainwater, ice water, and surface emissivity to improve physical consistency. Separate retrievals were applied for ocean, coast, low-elevation land, and high-elevation regions. The authors propose transferring the prior database and retrieval architecture to microwave imagers on other Fengyun-3 satellites and future missions, and see potential for extending the framework beyond FY-3G. Such transferability could broaden the impact of the work, particularly where ground networks are sparse and accurate precipitation data are most needed.
