基于深度学习的天气雷达回波外推的研究进展

Research Progress of Weather Radar Echo Extrapolation Based on Deep Learning

基于深度学习的天气雷达回波外推预测降水的性能实现了新的进展。综述了常见的循环神经网络、卷积神经网络、深度生成模型和多个网络的组合深度学习模型,并从实际应用角度分析了不同深度神经网络在天气雷达回波外推预测降水量任务中的优点与不足。此外,简要分析了天气雷达回波外推的过程机理,并对最新的研究进展进行了介绍。并分析了天气雷达回波外推面临的一些挑战和潜在的解决方案,为相关研究人员提供有益的参考和借鉴。

New advances have been achieved in the performance of deep learning-based weather radar echo extrapolation for precipitation prediction.This survey reviews recurrent neural networks,convolutional neural networks,deep generative models and combination deep learning models of multiple networks.Moreover,the advantages and shortcomings of different deep neural networks for the task of extrapolating weather radar echoes to predict precipitation are summarizing from the practical application perspective.In addition,the process mechanism of weather radar echo extrapolation is briefly analyzed and the latest research progress is presented.Finally,some challenges and potential solutions of weather radar echo extrapolation are analyzed,which can provide useful reference for relevant researchers.