基于伯努利-伽马-高斯模型的ECMWF逐日降水预报订正研究  

Study on the calibration of ECMWF daily precipitation forecast based on Bernoulli-Gamma-Gaussian model

作　　者：简志健 赵铜铁钢 黎学优 蔡华阳 段凯 王海龙 刘智勇 JIAN Zhijian;ZHAO Tongtiegang;LI Xueyou;CAI Huayang;DUAN Kai;WANG Hailong;LIU Zhiyon(School of Civi Engineering,Sun Yat-sen University,Zhuhai 519082,Guangdong,China;Southern Marine Science and Engineering Guangdong Laboratory(Zhuhai),Zhuhai 519080,Guangdong,China)

机构地区：[1]中山大学土木工程学院,广东珠海519082 [2]南方海洋科学与工程广东省实验室(珠海),广东珠海519080

出　　处：《水利水电技术（中英文）》2023年第8期67-78,共12页Water Resources and Hydropower Engineering

基　　金：国家重点研发计划(2021YFC3001000);国家自然科学基金项目(51979295);广东省珠江人才计划团队项目(2019ZT08G090)。

摘　　要：【目的】针对欧洲中期天气预报中心ECMWF降水原始预报存在的系统误差及随机误差,应用伯努利-伽马-高斯模型进行误差订正。【方法】以我国80个二级流域为单位评估检验模型订正效果,重点对我国珠江流域对应的二级子流域进行分析。使用排序概率技巧得分CRPSS等对订正前后预报进行评价。【结果】在预报技巧方面,经订正后珠江10个流域都得到了改善;平均而言珠江流域经订正后CRPSS提高了30,其中红柳江流域提升最大,订正后精度技巧得分提升了近80。【结论】伯努利-伽马-高斯模型能有效订正原始预报的系统偏差,提高预报的准确性;不同时期订正效果有所区别,对出现极值降水较少的偏枯期和特枯期,订正效果有更好的表现;累积降水预报订正后几乎所有流域都优于多年气候态的精度;订正的效果和相关系数大小有关,相关系数表现好的流域有更高的预测技巧得分,经订正后有利于后续水文集合预报应用。[Objective]Precipitation is a meteorological element concerned by the public and government departments related to industry and agriculture.Bernoulli-gamma-gaussian model is devised to correct the systematic and random errors in the raw forecasts of European for Medium-Range Weather Forecasts Integrated Forecasting System.[Methods]Eighty secondary river catchments in China were taken as units to evaluate the revised effect of the model,with emphasis on the analysis of the corresponding river catchments in the Pearl River Basin.The ranking probability skill score CRPSS was used to evaluate the forecast before and after calibration.[Results]In terms of forecasting skills,all the 10 basins of the Pearl River are improved after calibration.On average,the CRPS score of the Pearl River Basin increase by 30 after calibration,among which the Hongliujiang River Basin improve the most,and the accuracy skill score increase by nearly 80 after calibration.[Conclusion]The result show that Bernoulli-Gamma-Gaussian model can effectively correct biases of the raw precipitation forecast and improve the accuracy of the forecast,yielding forecast skill for daily and accumulated precipitation.The predictive performance after calibration is different in different cases,and the performance is better in the dry and extremely dry cases with less extreme precipitation.The accuracy of cumulative precipitation forecast is better than that of climatology in almost all catchments after the calibration.The effect of the calibration is related to the value of the correlation coefficient,and the catchments with higher correlation coefficient has better prediction skill score,which is beneficial to the subsequent application of hydrological ensemble prediction after the calibration.

关 键 词：逐日降水 次季节 集合预报 预报订正 预报检验 降水 不确定性 水文预报精度 

分 类 号：P338[天文地球—水文科学] P339[水利工程—水文学及水资源]