Abstract:The CMA Earth System Modeling and Prediction Centre is currently developing a regional ensemble prediction system with a horizontal resolution of 3 km. In view of the intricate terrain and meteorological conditions in Southwest China, coupled with the limited efficacy of the precipitation forecast products from the CMA regional model in this particular area, it is intended to place a significant emphasis on the impact of enhancing the horizontal resolution of the CMA regional ensemble prediction system on the precipitation forecast performance for the southwestern region during the research and development phase. Considering the aforementioned, this study, grounded in the CMA regional ensemble prediction system, has formulated two ensemble forecast tests with horizontal resolutions of 3 km and 10 km, respectively. Focusing on the recurrent precipitation events in the western Sichuan Basin during August 2020, a 25-day continuous test was executed. The experimental results were subsequently evaluated and comparatively analyzed. The findings reveal that the augmentation of horizontal resolution contributes significantly to the escalation of perturbation energy within the vertical layers of the troposphere, and it facilitates a more precise delineation of the forecast uncertainties associated with mesoscale and smaller-scale waves. Evaluation based on isobaric surfaces and ground meteorological parameters demonstrates that increasing the horizontal resolution helps to amplify the ensemble spread, improve the probabilistic forecasting skill for wind field, temperature and precipitation, and elevate the precision of precipitation forecases. Nonetheless, the 3 km ensemble prediction system still faces the problem of insufficient ensemble spread. The analysis of a heavy rainfall event shows that increasing the horizontal resolution facilitates more detailed simulation of multi-scale terrain and more accurate depiction of water vapor and dynamic structure in the rainstorm area, which in turn elevates the forecast performance for this event.