降雨短时临近预报技术研究进展

Advances on rainfall nowcasting techniques

降雨预报在农业、水资源管理、城市规划和自然灾害预警等方面有着至关重要的作用,短时临近预报可以为实时决策提供更为有效的参考信息。对目前降雨短时临近预报的主要研究方法进行了总结,阐述了基于探测数据外推、数值模式、统计学习3种常用方法的最新进展。基于探测数据外推的算法对于短时间和小范围内的降雨预报效果较好。数值模式强调物理过程,能全面分析和模拟大气环流场和降雨系统演变过程。机器学习技术的发展推动了基于统计学习的算法在降雨短时临近预报中的应用,具有广阔的应用前景。

Rainfall forecasting is of great significance in agriculture,water resource management,urban planning,and natural disaster early warning.Rainfall nowcasting can provide more effective information for real-time decision.This review summarized the main research methods of rainfall nowcasting:extrapolation from observations,numerical weather prediction models,statistical learning methods,and elaborated on the progress.Extrapolation from observations performs better for cases with short lead time and small scale.Numerical weather prediction models emphasize physical processes and can comprehensively analyze and simulate the evolution of atmospheric circulation and rainfall systems.The development of machine learning technique has promoted the application of algorithms based on statistical learning in short-term rainfall prediction,which has broad application prospects.