Abstract:The standard precipitation verificaiton method and the MODE （Method for Object-Based Diagnostic Evaluation） spatial method are applied to evaluate the performance of the CMA-MESO and CMA-SH9 models in predicting precipitation in eastern China in 2021 in this article. The results show that the two models have relatively high prediction skills for the second and third seasons of 2021, while the prediction skills for the first and fourth seasons are relatively low. The regional numerical models have good application potential in warm season precipitation forecasting. Based on the ETS and BIAS of the four seasons, the overall precipitation prediction skills of the CMA-MESO model in the third season are higher than those of the CMA-SH9 model, while in other seaons, the CMA-SH9 model has relatively higher prediction skills. Both models show a higher BIAS and a higher false alarm ratio in each season. Improving the shortcomings of these two aspects is an important means to enhance the precipitaiton prediction skills of regional models. The spatial verification results of torrential rain for four seasons show that CMA-MESO and CMA-SH9 models have relatively better forecasting abilities for the second and third seasons, and both of them tend to overestimate the object area of torrential rain. The CMA-SH9 model tends to overestimate the object quantile intensities of torrential rainfall in four seasons compared to observations, whereas the CMA-MESO model shows closer agreement with observations, exhibiting an overestimation only in the first season. Spatial verification of the MODE for the July 2021 severe torrential rain in Henan and Typhoon In-Fa reveals that both the CMA-MESO and CMA-SH9 models exhibit a tendency to overestimate the affected area for intense precipitation exceeding heavy torrential rain levels. Nonetheless, these models continue to underpredict the maxima of rainfall, with the CMA-SH9 model outputs more closely aligned with the observed extreme values.