黄兴友,马玉蓉,胡苏蔓. 2021. 基于深度学习的天气雷达回波序列外推及效果分析[J]. 气象学报, (0):-, doi:10.11676/qxxb2021.041
基于深度学习的天气雷达回波序列外推及效果分析
Extrapolation and effect analysis of weather radar echo sequence based on deep learning
投稿时间:2021-01-12  修订日期:2021-03-10
DOI:10.11676/qxxb2021.041
中文关键词:  雷达临近预报,循环神经网络,卷积神经网络,损失函数,深度学习
英文关键词:Nowcasting,Recurrent neural network,Convolutional neural network,Loss function,Deep learning
基金项目:国家重点研发计划
作者单位邮编
黄兴友 南京信息工程大学气象灾害预报预警与评估协同创新中心 210044
马玉蓉 南京信息工程大学气象灾害预报预警与评估协同创新中心 210044
胡苏蔓 南京信息工程大学气象灾害预报预警与评估协同创新中心 210044
摘要点击次数: 234
全文下载次数: 277
中文摘要:
      天气雷达资料是进行强对流天气临近预报的主要参考数据。针对传统雷达回波外推方法中存在资料信息利用率不足和外推时效有限的问题,[目的]运用神经网络进行雷达回波的外推,预测神经网络模型给出雷达回波外推2h的预报结果。[资料和方法]回波外推问题的关键是回波时空序列预测问题,该网络具有解决时间记忆问题的长短时记忆单元(Long Short-Term Memory,LSTM)和提取空间特征的卷积模块。应用福建、江苏和河南多年的雷达资料构造训练和测试数据集。为消除降水的不平衡性和提高对强回波的预报准确率,网络采用带权重的损失函数进行训练。对光流法和预测神经网络进行测试集检验以及个例分析,[结果]结果表明,在相同外推时效和检验反射率阈值的情况下,预测神经网络的CSI、POD均高于光流法,FAR低于光流法。不同降水类型中,预测神经网络的SSIM值(structural similarity)高于光流法,且层状云降水的SSIM值大于对流云降水。[结论]因此,预测神经网络对强回波的预报能力高于光流法;在预报时效性上,预测神经网络模型具有一定的优越性;预测神经网络对层状云降水预报的准确率要高于对流云降水。
英文摘要:
      Weather radar data is the main reference for nowcasting of severe convective weather. In view of the problems of insufficient data utilization and limited extrapolation time in the radar echo extrapolation methods widely used in China, the neural network is applied to radar echo extrapolation. The predictive neural network model gives the results of radar echo extrapolation for two hours. The essence of radar echo extrapolation problem is the spatiotemporal sequence prediction problem. The network has long and short time memory unit (Long Short-Term Memory,LSTM) to solve the time memory problem and convolutional layers to extract spatial features. The training and testing data sets were constructed using radar data from Fujian, Jiangsu and Henan for several years. Considering the frequencies of different rainfall levels are highly imbalanced and to improve the prediction accuracy of strong echoes, the network is trained by weighted loss function. The test set and individual case evaluation show that the CSI、POD of predictive neural network is higher than that of optical flow method and FAR lower than that of optical flow method under the same extrapolation aging and reflectivity threshold.Among different precipitation types, the SSIM(structural similarity) value of the predictive neural network is higher than that of the optical flow method, and the SSIM value of the stratiform-cloud precipitation is larger than that of the convective-cloud precipitation. Therefore, the predictive neural network has a higher ability to predict strong echoes than optical flow. In terms of the timeliness of forecasting, the predictive neural network model has certain advantages, and it is more accurate in forecasting stratiform-cloud precipitation than in convective-cloud precipitation.
查看全文   查看/发表评论  下载PDF阅读器
分享按钮