融合相似预报方法在陇东南短期强降水预报中的应用

Application of Fusing Analog Method in Short-term Heavy Precipitation Forecast in Southeastern Gansu

黄晓远 ¹李旭 ¹杜梦莹 ¹叶培龙 ²李艳¹

扫码查看

作者信息

1. 兰州大学大气科学学院,甘肃兰州 730000
2. 兰州中心气象台,甘肃兰州 730020
折叠

摘要

基于逐步过滤相似法和自组织映射(SOM)神经网络方法,提出了一种融合相似预报方法.利用ECMWF模式预报产品、ERA5再分析资料和地面气象台站观测数据,使用该方法对2021-2022年陇东南地区开展了时效为72 h的强降水预报试验,并对预报效果进行了检验.结果表明:(1)融合相似预报方法的TS评分处于4.5％～9.1％之间,与ECMWF模式预报结果相比表现出一定的优势.随着预报时效的增长,强降水预报的TS评分呈现减小的趋势,其在08:00(北京时,下同)起报的TS评分相对较高.(2)相比于单独使用逐步过滤相似预报,融合相似预报方法的准确性有所提升,并能在一定程度上降低空报率.其中08:00起报和20:00起报的TS评分提高了 1.31％和0.63％,而FAR同时下降了 2.39％和1.25％.

Abstract

Based on similar forecast method with step-by-step filter and Self-organizing map(SOM)neural net-work,a fusing analog forecast method is proposed.Using ECMWF model forecast products,ERA5 reanalysis data and station data,this method is used to carry out a 72-hour forecast of heavy precipitation in southeastern Gansu from 2021 to 2022,and the forecast effect is tested.The results show that the TS score of the fusing ana-log forecast method ranges from 4.5％to 9.1％,demonstrating a certain advantage compared to the forecast re-sults of the ECMWF model.As the forecast lead time increases,the TS score of the heavy precipitation forecast shows a decreasing trend,with relatively higher TS scores forecasted at 08:00.Compared with the similar fore-cast method with step-by-step filter alone,the accuracy of the fusing analog forecast method is improved,and it can alleviate the problem of high false alarm rate to a certain extent.Specifically,the TS scores forecasted at 08:00 and 20:00 are increased by 1.31％and 0.63％,while the FAR is decreased by 2.39％and 1.25％.

关键词

强降水/短期预报/相似预报/逐步过滤相似/自组织映射(SOM)

Key words

heavy precipitation/short-term forecast/similar forecast/similar forecast method with step-by-step filter/self-organizing map(SOM)

引用本文复制引用

出版年

2025

高原气象

中国科学院寒区旱区环境与工程研究所

高原气象

北大核心

影响因子：2.193

ISSN：1000-0534

段落导航