摘要利用兴义新一代多普勒天气雷达CINRAD/CD复合体扫基数据及气象产品风暴单体识别与跟踪SCIT，对发生在黔西南地区的2007～2008年的冰雹天气过程个例进行验证、分析，分别从识别、跟踪评估、位置预报3个方面进行评估分析，并进行了算法补尝，提出了解决的办法，在降低反射率因子识别的基础上在时间关联中增加一个方向性阈值，提取反射率因子水平结构ZSIGN、反射率因子空间变化率ZH、垂直累积液态含水量VIL、冰雹概率SHI 4个最能描述对流风暴特征的物理量，并对其分配权重，建立概率判别函数P，对2008年3月19日一次超级单体风暴的发展演变过程中P变化进行了分析，对2007～2008年的冰雹天气过程重新进行评估，与算法改进前的评估效果进行对比分析，结果表明，算法改进后识别成功率有明显上升，但对单体的跟踪及位置预报效果不明显。

关键词冰雹云； 单体识别与跟踪；算法评估；检验改进

中图分类号S161；P412.25文献标识码

A文章编号0517-6611（2015）07-199-04

Hail Cell Identification and Tracking Algorithm for Assessment and Improvement

TANG Hao-peng （Meteorological Office of Southwestern Guizhou， Xingyi， Guizhou 562400）

Abstract Using the Xingyi new generation of Doppler weather radar complex scan data and storm cell identification and tracking algorithm， the hail weather courses occurring in southwest Guizhou Province in 2007 and 2008 were assessed from the three aspects： identification， tracking， location prediction. The unfavorable situation for algorithm assessment is made analysis and compensation and proposed solution. On the basis of lower the reflectivity thresholds and adding the direction thresholds in the time association， the four physical alteration were calculated including the level reflectivity gradient， the vertical reflectivity gradient， vertically integrated liquid water content， and the strong hail index， which can best describe the storm cell， then distributed the coefficient respectively. The probability distinction function P was established. Using the probability distinction function P， the supercell on March 19th， 2008， was analyzed， then the hail weather courses in 2007 and 2008 were re-assessed， and the comparative results show that the improved algorithm enhances the identification effect to some extent， but the effect of tracking and location prediction werent obvious.

Key words Hail cloud； Cell identification and tracking； Algorithm evaluation； Inspection and improvement

20世纪以来，国内外雷达气象学者在雷达探测强对流天气系统应用上做了大量的工作，特别是在如何探测强对流天气系统上做了大量的努力，深入地分析了风暴发展、成熟、消亡及其移向移速的物理机制，提出了风暴的一些识别方法。如Rinchart提出应用模式来识别风暴单体及利用相关技术分析法对整幅回波图像进行处理[1]；Austin等提出了三维矩心识别方法[2]，Rosefeld对这一方法进行改进，提出通过对风暴形状的几何推理来处理分裂和合并[3]；Johnson等提出了利用7个反射率因子识别阈值来代替此前唯一的阈值及采取特征核抽取技术并对空间相距较近的多个风暴单体进行合并或删除处理，该算法被WSR-88D的Build 9.0广泛采用，称为B9SI算法[4]。……