“雨燕”中风暴算法与新一代雷达SCIT产品的对比分析  

COMPARISON BETWEEN STORM SERIES ALGORITHMS IN THE SWIFT AND SCIT IN THE NEXRAD

作　　者：庄旭东[1] 胡胜[1] 陈荣 何如意[1] 汪瑛[1] 

机构地区：[1]广州中心气象台,广东广州510080 [2]广州市番禺区气象局,广东广州511490

出　　处：《热带气象学报》2011年第3期299-306,共8页Journal of Tropical Meteorology

基　　金：中国气象局新技术推广项目(CMATG2009YB09)资助

摘　　要：利用2008年广州地区强对流资料,在构建风暴预报位置客观评分模块的基础上,统计分析了三种风暴算法(分别简称为SCIT、雨燕2006、雨燕2008)的预报误差。SCIT在30和60分钟预报时效,分别识别出5 124和2 203个风暴,预报位置绝对平均距离误差为12.9和23.7 km。雨燕2006识别为4 975和2 122个,绝对平均误差为12.2和22.2 km。两者预报误差相当。雨燕2008识别为5 474和2 516个,绝对平均距离误差为8.3和15.6 km,与前两者相比误差减小明显。分析上述三种风暴算法的误差原因:SCIT采用了历史运动矢量预报风暴位置,对于孤立的、分散的单体表现较好,但无法给出大范围回波区域中每个单体随着系统的运动信息。雨燕2006预报方法采用了COTREC矢量,能适用于大范围回波区域,但对于孤立的单体,误差较大。雨燕2008进行了一系列的改进,包括:利用风暴历史运动矢量对COTREC矢量做一致性检验、设计风暴预报位置超出回波范围或COTREC运动矢量接近零时的处理方案等。改进后,无论对于孤立的单体还是大范围回波区域中的风暴系统,都能选择合适的外推方案,因而整体误差较小。First,Storm Cell Identification and Tracking（SCIT） Algorithm in the Next Generation Weather Radar and storm algorithms in the SWIFT,including storm cell identification,storm tracking,storm forecast,are introduced.These storm products were applied in Guangzhou in 2008.For the SCIT product,the mean absolute errors for 30 and 60 minutes storm forecast are about 12.9 and 23.7 km,respectively.The storm samples are 5 124 and 2 203,respectively.With the increase of forecast length,the mean absolute errors of the storm product become larger.For the storm algorithms in the SWIFT,the mean absolute errors for 30 and 60 minutes are about 12.2 and 22.2 kilometers,respectively.Second,the reasons for errors are discussed.Historical movements of storms are used in the SCIT to forecast their future positions.It performs very well for isolated cells and often gives the incorrect storm movements when these storms are in a big echo region.On the other hand,TREC technique is adopted in the storm algorithms in the SWIFT.It is not suitable to forecast isolated storms because of the discontinuity of TREC（Tracking Radar Echoes by Correlation） vectors in isolated echo regions.Finally,improvements of the storm forecast scheme in the SWIFT are newly designed,including consistency check for TREC vector by using storm history vector,and new processing method for the radar detection border.Storm products are recalculated by using the improved scheme.The mean absolute errors for 30 and 60 minutes are about 8.3 and 15.6 km,respectively.Storm samples are 5 474 and 2 516,respectively.Storm tracking also benefits from the improvements of the storm forecast scheme.

关 键 词：雷达 风暴 检验 统计分析 对比 

分 类 号：P446[天文地球—大气科学及气象学]