主办单位:中国气象局沈阳大气环境研究所
国际刊号:ISSN 1673-503X
国内刊号:CN 21-1531/P

气象与环境学报 ›› 2021, Vol. 37 ›› Issue (4): 26-32.doi: 10.3969/j.issn.1673-503X.2021.04.004

• 论文 • 上一篇    下一篇

多源降水预报集成技术应用研究

唐娴1,2(),周荣卫1,2,何晓凤1,2,王津宇3,任晓晨4   

  1. 1. 华风气象传媒集团有限责任公司, 北京 100081
    2. 北京玖天气象科技有限公司, 北京 100081
    3. 国网河南省电力公司电力科学研究院, 河南 郑州 450052
    4. 中国人民解放军96813部队, 安徽 黄山 245000
  • 收稿日期:2020-08-07 出版日期:2021-08-30 发布日期:2021-09-10
  • 作者简介:唐娴, 女, 1988年生, 工程师, 主要从事数值模拟及预报解释应用、专业气象预测服务, E-mail: tangxian@jtmet.com
  • 基金资助:
    国家重点研发计划(2018YFC1507801);中国电力建设股份有限公司项目(DJ-ZDZX-2016-02)

Research on application of multi-source precipitation forecast integration technology

Xian TANG1,2(),Rong-wei ZHOU1,2,Xiao-feng HE1,2,Jing-yu WANG3,Xiao-chen REN4   

  1. 1. Huafeng Meteorological Media Group, Beijing 100081, China
    2. Beijing Jiutian Meteorological Technology Co., Ltd., Beijing 100081, China
    3. Electric Power Research Institute of State Grid He'nan Electric Power Company, Zhengzhou 450052, China
    4. 96813 Troops, PLA, Huangshan 245000, China
  • Received:2020-08-07 Online:2021-08-30 Published:2021-09-10

摘要:

基于欧洲中期天气预报中心全球高分辨率预报模式ECMWF、中国自主研发的新一代业务化区域数值模式GRAPES_Meso和WRF中国全国区域预报模式的降水预报结果进行未来3 d降水集成预报。以中国地面-卫星-雷达三源融合逐时降水格点产品(CMPA-Hourly,V2.0)作为"观测值"进行建模,采用消除偏差多模式平均法和基于无偏平均绝对误差集成法对中国大陆地区进行降水集成预报,同时采用2800个国家自动气象站降水观测数据对降水集成预报效果进行检验。结果表明:基于无偏平均绝对误差的降水集成方法能够综合每个模式成员降水预报场的优势,提供一种更为稳定可靠且具有更高分辨率的优质精细化降水预报产品;其在试验期间对中国大陆地区汛期的降水预报ETS评分,优于消除偏差多模式平均降水集成预报和最优单模式降水预报,BIAS评分更接近于1,与"实况"的距平相关系数也明显提高,是对降水大值捕捉能力较优的一种集成方法,尤其对中雨到暴雨量级预报的改进较好。

关键词: 多源, 消除偏差多模式平均法, 无偏平均绝对误差

Abstract:

The integrated forecasting of 3-d precipitation was conducted based on precipitation forecast results of the global high-resolution European Centre for Medium-Range Weather Forecasts (ECMWF) model, the Chinese new generation of operational Globe/Range Assimilation and Prediction Enhance System (GRAPES_Meso), and the Weather Research Forecast (WRF) model. Taking the three-source (ground-satellite-radar) hourly precipitation grid product in China (CMPA-Hourly V2.0) as the "observed value", we adopted a simple bias-removed ensemble mean (ENSM) method and a multi-model ensemble (MME) method to conduct integrated precipitation forecast in the mainland of China. The performance of integrated precipitation forecasts was evaluated using precipitation observation data from 2800 national automatic weather stations. The results indicated that the MME method can integrate the advantages of each model member's precipitation forecast field, and provide a more stable, reliable, and high-quality refined precipitation forecast product with a higher resolution. During the test period, the EST score of the precipitation forecast based on the MME method during the flood season in the mainland China is better than that using the ENSM method and using the optimal single-mode precipitation forecast. The BIAS score is closer to 1, and the anomaly correlation coefficients between the forecasts using the MME method and the observed values increases. The MME method has a better ability to capture large values of precipitation, especially for improving the prediction of grades above moderate precipitation.

Key words: Multi-source, Bias-removed ensemble mean method, Unbiased mean absolute error

中图分类号: