机器学习在强对流监测预报中的应用进展

Advances in Application of Machine Learning to Severe Convective Weather Monitoring and Forecasting

近年来,机器学习理论和方法应用蓬勃发展,已在强对流天气监测和预报中广泛应用。各类机器学习算法,包括传统机器学习算法(如随机森林、决策树、支持向量机、神经网络等)和深度学习方法,已在强对流监测、短时临近预报、短期预报领域发挥了积极的重要作用,其应用效果往往明显优于依靠统计特征或者主观经验积累的传统方法。机器学习方法能够更有效提取高时空分辨率的中小尺度观测数据的强对流特征,为强对流监测提供更全面、更强大的自动识别和追踪能力;能够有效综合应用多源观测数据、分析数据和数值预报模式数据,为强对流临近预报预警提取更多有效信息;能够有效对数值模式预报进行释用和后处理,提升全球数值模式、高分辨率区域数值模式在强对流天气预报上的应用效果。最后,给出了目前机器学习方法应用中存在的问题和未来工作展望。

In recent years,the theory of machine learning and its applications to severe convective weather has been developed at an unprecedented speed.Various machine learning algorithms,such as random forest,decision tree,support vector machine,neural network and deep learning have played important roles in severe convective weather monitoring,nowcasting,short-term forecasting and short-range forecasting.These algorithms often have better performances than traditional methods.With the help of machine learning,it is easier to extract the mesoscale features of convective systems in high spatio-temporal resolution observation data,resulting in better performances of automatic convective weather identification and tracking and warning.Machine learning is also a good tool to effectively use the multi-source observation data,analyze the observation and numerical weather prediction(NWP)data.In addition,machine learning can also be an effective postprocessing method for NWP.It has been showed that machine learning can extract the features of severe weather occurrence from global or regional NWP data and give a reliable severe weather forecasts.Finally,the issues and outlooks of machine learning application are presented.