气象2023，Vol.49Issue(12)：1481-1494,14.DOI:10.7519/j.issn.1000-0526.2023.083101
CAST-LSTM:一种用于雷达回波外推的时空LSTM模型
CAST-LSTM:A Spatio-Temporal LSTM Model for Radar Echo Extrapolation
摘要
Abstract
The forecast results of radar echo extrapolation algorithm based on recurrent neural network are gradually blurred and distorted with time,and it is difficult to forecast the severe echo area.To solve the above problems,this paper proposes a spatio-temporal long short-term memory network model based on context fusion and attention mechanism.The method fully extracts the short-term context information of different scales of radar image through the context fusion module.The attention module broadens the time perception domain of the prediction unit,so that the model perceives more time dynamics.Taking the weather radar data of Jiangsu Province from April to September in 2019-2021 as a sample,the spatio-tem-poral long short-term memory network based on context fusion and attention mechanism achieves better prediction performance through experimental comparison and analysis.Under the conditions of 60 min extrapolation and the thresholds of 10,20 and 40 dBz,the critical success index(CSI)and heidke skill score(HSS)reach 0.7611,0.5326,0.2369 and 0.7335,0.5735,0.3075,respectively,which effectively improved the prediction accuracy.关键词
雷达回波外推/深度学习/降水预报/长短期记忆Key words
radar echo extrapolation/deep learning/precipitation forecast/long short-term memory(LSTM)分类
大气科学引用本文复制引用
渠海峰,何光鑫,康志明,程勇,王军,庄潇然,李远禄..CAST-LSTM:一种用于雷达回波外推的时空LSTM模型[J].气象,2023,49(12):1481-1494,14.基金项目
山东省自然科学基金项目(ZR2022MD072、ZR2021QD028、ZR2020MD052、ZR2020MD053)、中国气象局创新发展专项(CXFZ2022J034)、山东省气象局重点科研项目(2021sdqxz05、2021sdqxz09)共同资助 （ZR2022MD072、ZR2021QD028、ZR2020MD052、ZR2020MD053）
