摘要
将一种暴雨云团的多尺度识别方法——层级聚类法,应用于β中尺度对流系统识别及临近预报中。该方法的基本思路是:将笛卡尔坐标下的雷达反射率因子进行聚类,得到比较详细的较小尺度的暴雨云团,然后设定阈值,将云团之间差异小于阈值的进行合并,可以得到较大尺度的云团,逐步放宽合并阈值,可得到更大尺度的云团。选取广州雷达2005年3月的飑线过程和温州2005年9月的台风过程对这种方法的识别结果进行了详细说明,结果表明:该方法能够识别不同系统的β中尺度对流云团,并能识别出其中反射率较强的γ中尺度云团,识别结果合理。采用这种方法识别不同尺度的暴雨云团,有利于跟踪、预报造成中国暴雨主要原因的β中尺度系统,也可兼顾β中尺度系统中的γ中尺度对流单体。根据预报时效的不同,可以选择不同的云团识别尺度。
Heavy rainfall in the Meiyu front and typhoon precipitation are mainly caused by meso-β Mesoscale Convective Systems （MCS）. In order to identify the convective systems with different spatial scales, a multi-scale cloud cluster identification algorithm called hierarchical K-Means clustering method is developed in the paper. The algorithm is clustering radar reflectivity data in Cartesian coordinate using K-Means cluster to classify all grids according to a criterion, and the detailed and smaller clusters are identified in the first step. The coarser cloud clusters are formed by merging the clusters with differences less than the threshold. The algorithm and its application in nowcasting are described in detail for the squall line heavy rainfall and the typhoon observed by Guangzhou and Wenzhou radars. The main conclusions are gotten as follows： （1） The clustering algorithms, which are widely used in the field of market analysis and medicinal practice, are successfully used in meso-β-scale and meso-γ-scale convective cells identification, and the results are reasonable for the two cases. The multiscale algorithm is helpful for identifying, tracking and forecasting meso-β-scale and meso-γ-scale systems. （2） The convective systems with different scales can be tracked and extrapolated with different forecast time. The further study to improve the forecast and the evolvement of storm with the algorithm still need to be done.
出处
《大气科学》
CSCD
北大核心
2007年第3期400-409,共10页
Chinese Journal of Atmospheric Sciences
基金
国家自然科学基金资助项目40375008
国家重点基础研究发展规划项目2004CB418305
关键词
多尺度识别
层级聚类
跟踪
预报
multi-scale, hierarchical clustering, tracking, forecasting
