汉江流域多模式面雨量预报效果评估及集成应用

Evaluation of multi-model surface rainfall forecasting effect and integrated application in Hanjiang River Basin

为探明不同数值模式预报产品对汉江流域降水的预报效果,利用欧洲中期天气预报中心(ECMWF)、德国气象局(GERMAN)、美国国家环境预报中心(NCEP)、日本气象厅(JMA)与中国国家气象局的GRAPES数值模式,分析了汉江流域面雨量的分布特征,并利用TS评分检验了多种数值模式对汉江流域不同分区、季节、量级的面雨量预报效果,提出了多模式降水预报集成应用方案。结果表明:汉江流域各分区中,皇庄以下区间出现大雨及以上量级强降水次数最多且强度最大,白河-丹江口区间的次数最少且强度最弱。各模式中,GERMAN模式对汉江流域面雨量的预报效果整体最好,其次为ECMWF模式,再次为NCEP模式,而JMA和GRAPES模式相对较差。不同季节中,模式预报在主汛期的降水预报效果整体最好,其次为秋汛期,枯季相对较差;其中,秋汛期的空报率和漏报率相对较低,主汛期的漏报率相对较高。采用动态加权的方法对多模式预报产品进行集成,发现集成预报可以综合各模式预报信息,其预报效果在各模式中位于前列。

To evaluate the forecasting results of different numerical forecasting products for precipitation in Hanjiang River Basin,this paper uses numerical modes from the European Centre for Medium-Range Weather Forecasts(ECMWF),the German Weather Service(GERMAN),the National Centers for Environmental Prediction(NCEP),the Japan Meteorological Agency(JMA),and the GRAPES modes from the China Meteorological Administration(CMA)to analyze the distribution characteristics of surface precipitation in Hanjiang River Basin,and uses TS rating method to evaluate the forecasting results of various numerical modes on the surface precipitation in different sub-basins,seasons and intensities in the Hanjiang River Basin,then proposes an integrated application scheme for multi-model precipitation forecasting.The results show that among the various sub-basins of Hanjiang River Basin,the sub-basin below Huangzhuang Station has the most frequent and intensity of heavy precipitation,and the sub-basin between the Baihe Station and Danjiangkou Reservoir has the least frequent and intensity of heavy precipitation.Among these modes,the GERMAN mode had the best forecasting result of surface precipitation compared to others,followed by the ECMWF mode and the NCEP mode,while the JMA and GRAPES mode's results are relatively poor.Among the different seasons,the precipitation forecasting results of all models are the best in the main flood season,followed by the autumn flood season,while the dry season had the worst result.However,rate of false-alarm and false-negative in the autumn flood season are relatively low,and the rate of false-negative in the main flood season is relatively high.It is also found that using dynamic weighted method to integrate multi-model forecasting products can combine forecasting information from other modes,and the forecasting results usually are at the top among all modes.