基于ECMWF集合预报的海河流域降水概率预报应用和检验

徐姝; 熊明明; 陈法敬

doi:10.3969/j.issn.1004-9045.2021.05.009

基于ECMWF集合预报的海河流域降水概率预报应用和检验

Application and verification of probabilistic precipitation forecasting in Haihe River Basin based on ECMWF Ensemble Prediction System

摘要

摘要: 利用海河流域逐日降水历史观测资料、ECMWF集合预报降水预报数据，通过贝叶斯产品处理技术（Bayesian Processor of Output，BPO）对海河流域内289个格点进行BPO建模，将ECMWF集合成员确定性降水预报修订为贝叶斯降水概率预报，结果显示概率密度峰值较确定性预报更加接近实况；再以51个成员的有效信息得分（Informativeness Score，IS）衡量各集合成员的预报能力，融合各成员的概率预报结果，得到代表ECMWF集合预报不确定性的贝叶斯集成降水概率预报。采用RPS和BS评分方法对海河流域2018年6—8月降水概率预报进行检验，结果表明，在海河流域降水预报中基于BPO方法的贝叶斯集成概率预报评分结果优于集合预报的直接概率预报结果，为海河流域降水概率预报业务奠定了基础。

Abstract: Using the historical precipitation observation data in the Haihe River Basin and the ECMWF ensemble prediction, the 289 grid points in the Haihe River Basin are modeled with Bayesian Processor of Output (BPO), which revise the determine precipitation forecast of the ensemble members to the Bayesian precipitation probability distribution and probability density, and obtain an effective informativeness score (IS) representing the forecasting ability of the ensemble members. Based on the IS values of 51 members, the Bayesian probability forecast information of each member is fused to obtain an integrated Bayesian precipitation probability forecast representing the uncertainty of the ECMWF ensemble forecast. The RPS and BS tests are used. The results show that the reliability of the integrated Bayesian precipitation probability forecast in the Haihe River Basin is higher than the direct probability forecast obtained by the integrated forecast.

HTML全文

参考文献(24)

施引文献

资源附件(0)