网格降水预报客观检验订正方法研究进展

Research Progress on Objective Verification and Correction Methods of Grid Precipitation Forecast

降水在网格要素预报中最为关键和重要,降水的预报表现直接体现了网格化要素的预报能力和水平。首先回顾了数值天气预报中针对降水的各种检验方法,主要包括:基于二分法列联表的经典检验,基于属性和尺度特征的空间检验,集合预报检验以及针对极端稀有事件的检验技术。其次归纳了降水统计后处理订正技术:一是基于模式输出降水产品的直接统计后处理和在模式输出各种要素基础上客观诊断的间接后处理;二是针对集合预报的参数化和非参数化降水以及和预报变量结构相关联的后处理订正方法;三是气象工作者基于人工智能和大数据技术对降水订正的探索和尝试。最后,讨论了各种检验方法的优缺点、客观性、适用性和业务使用中需要关注的问题,对网格降水后处理的订正方法进行了总结和讨论。

Precipitation is the most critical and important in grid element prediction.The prediction performance of precipitation directly reflects the prediction ability and level of grid elements.Firstly,various test methods for precipitation in numerical weather forecast are reviewed,mainly including:classical test based on dichotomy contingency table,spatial test based on attribute and scale characteristics,ensemble forecast verification and verification for extremely rare events.Secondly,the post-processing techniques of precipitation statistics are summarized:first,the direct statistical post-processing of precipitation products based on model output and the indirect post-processing of objective diagnosis based on various elements of model output;Second,for parameterized and nonparametric precipitation of ensemble prediction and post-processing correction method associated with the structure of prediction variables;The third is the exploration and attempt of meteorologists on precipitation correction based on artificial intelligence and big data technology.Finally,the advantages and disadvantages,objectivity,applicability and problems needing attention in the operation of various test methods are discussed,and the revised methods of grid precipitation post-processing are summarized and discussed.