Assimilation of Dual-polarization Radar Observations Using 1D Bayesian Retrieval for the Extreme Rainfall over Henan on 20 July 2021

Dual-polarization radar observations provide critical microphysical information for understanding and forecasting severe precipitation events. To assess their impact on numerical weather prediction, a dual-polarization radar assimilation framework combining one-dimensional (1D) Bayesian retrieval and three-dimensional variational (3DVAR) assimilation is developed and applied to the extreme rainfall event in Henan, China, on 20 July 2021. Relative humidity (RH) pseudo-observation profiles are first retrieved from horizontal reflectivity (ZH) or from a combination of ZH , differential reflectivity (ZDR), and specific differential phase (KDP), and are then assimilated. RH profiles retrieved using ZH alone are moister than those in the background. The inclusion of ZDR or KDP further enhances humidity, with ZDR producing a larger increase in domain-averaged RH and KDP yielding stronger moistening within convective cores. Cycled data assimilation and forecast experiments demonstrate that assimilating dual-polarization radar data strengthens moisture transport, upward motion, and convective instability, ultimately producing improved precipitation forecasts. Assimilation of ZDR better reproduces the overall rainfall pattern and improves the representation of moderate-intensity precipitation, while KDP assimilation better captures the intensity and location of the heavy rainfall center, yielding a simulated 6-h maximum precipitation of 409.39 mm, the closest to the observed 438.8 mm. Joint assimilation of ZDR and KDP integrates their complementary strengths, producing forecasts whose rainband structure, orientation, and heavy precipitation center location are more closely aligned with observations.