Abstract: The polarimetric variables from dual-polarization radar can more accurately describe the microphysical states in a convective storm. These microphysical states are closely related to the occurrence and development of severe convection systems.Therefore, it is great practical significance to improve the numerical forecast of severe convection weather by using radar dualpolarization measurements. Based on the quality control of polarimetric radar measurements (including non-standard blockage removal, hydrometeor classification, polarization threshold test, and clutter removal and smoothing), the S-band dual-polarization radar data are assimilated by the improved ARPS cloud analysis system.The improved ARPS cloud analysis system makes the polarization assimilation mainly by the hydrometeor classification algorithm and the radar operator based on T-matrix method, based on the principles of fuzzy logic, polarimetric radar measurements (including horizontal reflectivity (Zh), differential reflectivity (Z_DR), specific differential phase (K_DP), correlation coefficient (ρ_HV) and background temperature) are used to classify hydrometeors and determine the ratios among the three main hydrometeor types (rain, snow, hail/graupel). And then the mixing ratio of each type is analyzed using the reflectivity operator which based on T-maxtrix method to comply with the Zh observation.The S-band dual-polarization radar data of Guangzhou, Qingyuan, Shaoguan and Yangjiang are assimilated in Typhoon Hato (No.1713). The results show that the initial hydrometeors distribution is improved when the polarimetric variables are assimilated.The water-vapor mixing ratio is decreased, and a dry tongue appears around the typhoon at 700 hPa. Moreover, there are less false alarms for heavy rainfall. Significant positive impacts on the short-term forecast of heavy rainfall are obtained in the typhoon case.