干旱气象 ›› 2026, Vol. 44 ›› Issue (2): 325-337.DOI: 10.11755/j.issn.1006-7639-2026-02-0325

• 技术报告 • 上一篇    下一篇

基于邻域最优百分位的逐时降水预报订正方法

王春晓1,2(), 许平平3, 苏爱芳1,2(), 栗晗1,2, 吴稀稀4, 董俊玲1,2, 李朝兴1,2, 崔丽曼1,2,5   

  1. 1 中国气象局河南农业气象保障与应用技术重点实验室河南 郑州 450003
    2 河南省气象台河南 郑州 450003
    3 中国西昌卫射中心四川 西昌 615000
    4 洛阳市气象局河南 洛阳 471003
    5 中国气象局水文气象重点开放实验室北京 100081
  • 收稿日期:2025-05-08 修回日期:2025-08-11 出版日期:2026-05-20 发布日期:2026-05-18
  • 通讯作者: 苏爱芳(1971—),女,博士,正高级工程师,主要从事暴雨和强对流天气预报技术研发。E-mail:610061618@qq.com
  • 作者简介:王春晓(1990—),女,博士,高级工程师,主要从事天气预报技术研发。E-mail:wangchx1990@126.com
  • 基金资助:
    中国气象局水文气象重点开放实验室重点项目(23SWQXZ003);河南省科技研发计划联合基金(应用攻关类);河南省科技研发计划联合基金(232103810097);河南省重点研发专项项目(251111322500);河南省农业气象保障与应用技术重点实验室应用技术研究基金(KB202501)

A correction method for hourly precipitation forecast based on the optimal neighborhood percentiles

WANG Chunxiao1,2(), XU Pingping3, SU Aifang1,2(), LI Han1,2, WU Xixi4, DONG Junling1,2, LI Chaoxing1,2, CUI Liman1,2,5   

  1. 1 CMA Henan Key Laboratory of Agrometeorological Support and Applied TechniqueZhengzhou 450003, China
    2 Henan Meteorological ObservatoryZhengzhou 450003, China
    3 Xichang Satellite Launch CenterXichang 615000, Sichuan, China
    4 Luoyang Meteorological ServiceLuoyang 471003, Henan, China
    5 China Meteorological Administration Hydro-Meteorology Key and Open LaboratoryBeijing 100081, China
  • Received:2025-05-08 Revised:2025-08-11 Online:2026-05-20 Published:2026-05-18

摘要:

如何提高小时降水预报准确率是目前业务中亟需解决的难点问题。本文结合集合预报和邻域窗提出邻域最优百分位法(Optimal Neighborhood Percentile,ONP),实现了传统最优百分位法在确定性模式逐时降水预报分级订正中的应用。利用河南省区域自动气象站降水数据,对中国气象局中尺度天气数值预报系统(China Meteorological Administration Meso-scale Numerical Prediction Model,CMA-MESO)、中国气象局上海区域中尺度数值预报系统(CMA Shanghai 9 km Model,CMA-SH9)和中国气象局北京快速更新数值预报系统(CMA Beijing Model,CMA-BJ)的逐时降水预报进行订正。2024年4—9月业务应用结果显示,ONP通过减小降水空报提高晴雨准确率(Probability of Correction,PC)和0.1 mm以上降水TS评分(Threat Score,TS),通过增加降水命中提高2 mm以上降水TS评分。ONP在提高PC和不同降水量级TS评分方面的表现明显优于最优TS评分法(Optimal Threat Score,OTS)和频率匹配法(Frequency-Matching Method,FMM)。基于CMA-MESO模式的订正结果显示,ONP晴雨预报技巧达10.0%以上,远高于OTS(0.4%);ONP订正后的20 mm以上降水TS技巧分别为OTS和FMM的1.80、2.25倍。ONP存在大量级(20 mm)以上降水过报现象,需进一步调整最优百分位选取指标加以抑制。

关键词: 小时降水预报, 邻域最优百分位法, TS评分, 晴雨准确率

Abstract:

How to improve the accuracy of hourly precipitation forecasting is a problem that needs to be solved urgently. Combining the ensemble forecast and the neighborhood window, the Optimal Neighborhood Percentile (ONP) method is proposed, which realizes the application of the traditional optimal percentile method in the correction of hourly precipitation forecast from deterministic forecasting model. Based on the hourly precipitation data from regional automatic stations in Henan Province, the hourly precipitation forecasts of China Meteorological Administration (CMA) mesoscale numerical prediction model (CMA-MESO), CMA Shanghai 9 km model (CMA-SH9) and CMA Beijing model (CMA-BJ) are corrected by using the ONP method. The application results from April to September 2024 show that the ONP method can improve the probability of correction (PC) and the threat score (TS) of forecasting of precipitation above 0.1 mm by reducing the false alarms of precipitation, and improve the TS of forecasting of precipitation above 2 mm by increasing the hit rate of precipitation. The performance of the ONP method is better than that of Optimal Threat Score (OTS) and Frequency-Matching Method(FMM). The results of correction based on CMA-MESO show that the skill of the PC by using the ONP method is over 10%, which is higher than that of the OTS method (0.4%). The TS accuracy of the ONP for precipitation above 20 mm is 1.80 and 2.25 times of the OTS and FMM, respectively. The ONP method can lead to a relatively large wet bias for heavy rain (especially precipitation above 20 mm), and it needs further adjust the selection rules to reduce bias.

Key words: hourly precipitation forecast, Optimal Neighborhood Percentile (ONP), Threat Score (TS), probability of correction (PC)

中图分类号: