Abstract: A new convective–stratiform separation technique based on the hourly precipitation obtained from CMPAS (China Meteorological Administration multisource precipitation analysis system) and radar reflectivity mosaics data obtained from CMARMOS (China Meteorological Administration radar mosaic operation system) is presented in this paper. The technique, which is based on fuzzy logic, is developed to improve the cloud analysis scheme in the Gridpoint Statistical Interpolation (GSI) assimilation system (referred to as the CUST scheme). The improved scheme was tested in a severe Mei-yu rain that occurred on 19 June 2019, in Zhejiang Province. Several hourly-cycle assimilation experiments were performed using the WRF (weather research and forecasting) model and the GSI assimilation system to analyze the impact of the new scheme on the precipitation simulation, and the scheme was compared with other schemes. The results showed the following: (1) The new convective–stratiform separation technique accurately separated the convective–stratiform cloud, and it can be used as a discriminating factor to improve the cloud analysis scheme in the GSI assimilation system. (2) The CUST scheme adopted the convective cloud analysis scheme in the convective region and the stratiform cloud analysis scheme in the non-convective region, which reduced the false alarm rate in the simple convective cloud scheme and the underestimation in the simple stratiform cloud scheme; this effectively improved the simulation of short-term precipitation. (3) The CUST scheme showed significant improvement in the initial stage of the model (within 6 hours or even 3 hours), and the improvement of small-level rain was greater than that of heavy rain. (4) Compared with the hybrid cloud analysis scheme (referred to as the CSW scheme) based on the convective scale velocity determined by the surface sensible heat and latent heat fluxes, the CUST scheme showed a more reasonable result for the convective–stratiform cloud partition and precipitation simulation, which indicates a good application prospect.