Python在短临气象预报检验中的应用  

Application of Python in Test of Nowcasting

作　　者：何佳 惠建忠[1] 何险峰 王曙东[1] 高金兵 HE Jia;HUI Jianzhong;HE Xianfeng;WANG Shudong;GAO Jinbing(Public Meteorological Service Center of CMA,Beijing 100081;Huafeng Meteorological Media Group,Beijing 100081)

机构地区：[1]中国气象局公共气象服务中心,北京100081 [2]华风气象传媒集团有限责任公司,北京100081

出　　处：《气象科技》2021年第5期738-745,共8页Meteorological Science and Technology

基　　金：国家重点研发计划课题“气象预警精准快速发布业务化中试/示范平台技术研发”(2018YFC1507805)资助。

摘　　要：基于机器学习方法的短临多气象要素预报系统(Weather Elements Nowcasting based on machine learning,简称WEN)具有高发布频次、高时间分辨率、基于候和时辰的复杂预报模型等特点。应用多维标签数组、机器学习工具、并行计算框架等Python库,以快速计算为目标,建立以预报模型覆盖时间范围为统计检验时间边界的检验子系统,客观给出预报性能,为模型调优效果评估、产品业务化运行提供依据。Weather Elements Nowcasting based on machine learning(WEN)has the characteristics of high release frequency,high time resolution,and complex forecast model based on climate and time.Using Python libraries such as multidimensional tag array,machine learning tools,parallel computing framework aiming at"fast computing",a testing subsystem is established.It takes the time range covered by the"prediction model"as the statistical test time boundary.It objectively gives the prediction performance,which provides a basis for evaluating model optimization effect and the operation of products.

关 键 词：并行计算 检验评估 标签化多维数组 PYTHON 

分 类 号：P409[天文地球—大气科学及气象学]