Abstract:  The SAL quantitative verification method was used to evaluate precipitation forecast  results of the Grapes-Meso model on the rainfall event occurring on 5 September 2018 in Sichuan Province firstly, in this process, the optimal precipitation threshold scheme was selected through combination of three threshold schemes to identify the precipitation body. On this basis, SAL verification was applied to precipitation forecast of three heavy rainfall events in 2018 and rainfall processes from July to September in 2018 in Sichuan Province, and the verification results were compared with test score (TS) to understand prediction effect of the Grapes-Meso model in flood season in Sichuan Province. The conclusions are as follows: (1) The selected threshold determining method could identify the precipitation body well and the “connected neighborhood method” provided a good support for automatic identification of precipitation individuals. (2) The value of L reflected prediction effect of the model on precipitation forecast to a certain extent. If the value of L and the absolute value of A were both small, the probability of better prediction effect was bigger. (3)The overall prediction effect of the Grapes-Meso model on rainfall processes in Sichuan Province showed that the forecasted rainfall intensity was stronger and rainfall range was larger than actual rainfall, or forecasted precipitation center was smaller than the actual, or both cases were existed at the same time.

Key words: Grapes-Meso model, SAL verification, precipitation forecast 

摘要： 以四川省2018年9月5日降水过程为例，选择最佳阈值确定方法识别降水主体，在此基础上通过对四川省3次大范围强降水以及2018年7—9月降水过程进行SAL检验，并结合TS评分进行对比，分析和评价Grapes-Meso模式在本地汛期的预报效果。结果表明：（1）选取的阈值确定方法可以更好地识别降水主体，并且“连通邻域法”为降水个体的自动识别提供了很好的支撑；（2）L值的大小可以反映模式对降水预报的效果，如果L值较小，且A的绝对值也较小，则模式对降水预报偏好的可能性越大，反之则越小；（3）Grapes-Meso模式对四川省降水预报效果表现为：对降水强度预报较实况偏强，降水范围较实况偏大，或者降水中心值较实况偏小，或者前述两种情况同时存在。

关键词: Grapes-Meso模式, SAL检验, 降水预报