辽宁区域性大暴雨成因分析及多模式数值预报空间检验

利用三源融合格点降水实况、加密自动站观测资料、雷达基本反射率因子、高分辨率数值预报产品及FNL再分析资料, 对2020年汛期辽宁地区12次区域性暴雨过程进行天气系统分型检验表明, 气旋型暴雨模式的可预报性较低。选取2021年7月12—14日辽宁地区典型气旋型暴雨过程进一步分析, 采用面向对象目标的空间检验方法SAL, 结合传统检验方法, 从结构、强度和位置三个方面定量分析不同模式预报偏差的原因。结果表明: 暴雨落区集中且呈双雨带分布, 局地雨强大, 辽宁东、西部降水成因不同。CMA区域模式较全球模式暴雨TS评分高; SAL空间检验表明, CMA区域模式对于雨带内部结构把握较好, 全球模式结构误差主要来源于降水极值预报不足; 强度检验表明, CMA-MESO3km强度接近实况, EC_THIN次之, CMA_GFS的降水强度预报较差; 各模式暴雨落区基本可信, CMA-MESO3km最优, 暴雨落区的误差主要由于模式预报降水过程主体重心与实况的偏差较大所致。

Cause analysis and spatial test of multi-mode numerical prediction on regional rainstorms in Liaoning province

Based on the actual precipitation of three source fusion grid, the observation data from intensive automatic station, the basic radar reflectivity factors, high-resolution numerical prediction products, and FNL reanalysis data, the synoptic system classification tests on twelve regional rainstorm processes during flood season of 2020 in Liaoning province were carried out, and it was shown that the predictability of the cyclone rainstorm was low.Then the typical cyclonic rainstorm during July 12-14 was selected for further analysis.Using the object-oriented spatial test method, SAL (Structure, Amplitude, Location), combined with the traditional test method, the causes of prediction errors in different models were quantitatively analyzed from three aspects including structure, strength, and position.The results show that the rainstorm area is concentrated and presents double rain belt distribution.Rain intensity in the local area is significant and the reasons for precipitation are different in the areas of east and west of Liaoning province.The TS scores of CMA regional model are higher than those of the global model.SAL space test shows that CMA regional model well represented the internal structure of the rain belt, whereas the structural errors in the global model mainly due to the forecasting weakness of extreme precipitation.The strength tests show that the predicted rainfall intensities appear close to the actual situation in CMA-MESO3km, followed by EC_THIN, and insufficient in CMA_GFS.In general, the rainstorm area in each model is credible with the best result in CMA-MESO3km.The prediction errors in the rainstorm area mainly result from the significant discrepancies between the model prediction focus and the actual situation.