多普勒天气雷达冰雹探测算法评估及检验改进

利用新一代多普勒天气雷达资料和WSR-88D提供的冰雹指数算法，对2005年1月至2007年8月发生在贵州省黔西南地区的20个冰雹个例进行验证。用此算法对20个风暴日的样本计算了WT（警报阈值）、H（相对雷达的高度）、M（漏报率）、FA（虚警率）、POD／FAR／CSI（探测概率／误报率／临界成功指数）等多个函数，并将这些数据与强冰雹指数（SHI）作对比分析，将SHI作为冰雹尺寸的预报因子进行独立评估，对实际观测到的冰雹尺寸与模式预报尺寸进行比较。在考虑了本地环境、气候特征的前提下对误警率较高的情况进行了算法补尝，并针对误警率较高的现象提出解决办法：①输入当天的正确0℃／-20℃高度，②提高冰雹探测反射率阈值。用改进方法对2007年发生的9次冰雹天气过程进行对比检验。结果表明，误报率有所降低，预报冰雹尺寸更接近实际探测尺寸。

Assessment and Improvement of CINRAD/CD Hail Detection Algorithm

By means of the CINRAD/CD and WSR-88D hail indexes, the hail data from January 2005 to August 2007 is analyzed. The hail indexes, such as warning threshold, relative radar height, missing report rate, false alarm rate, POD/FAR/CSI （Probability of Detection/False Acceptance Rate/Critical Success Index）, are calculated and compared with SHI （Severe Hail Index）, which is used as a prediction factor of the hail size, and the actual and predicted hail sizes are compared. Considering the local environment and climate characteristics, attempts to improve FAR are made （inputting the correct heights of 0℃ （-20℃）, and increasing the reflectivity）, and a method for reducing FAR is discussed. Using the method, the hail events from January 2007 to August 2007 are analyzed and tested, and the results show that FARs are reduced, and the forecast sizes of hailstones are closer to the actual sizes.