Abstract: 3-D clustering technique is the fundamental part of storm identification. One shortcoming of the clustering technique is that the connection area between adjacent storms may cause the problem of false merger. To surmount this difficulty, a special erosion operation is used. Instead of the constant convolution mask, it uses a dynamic convolution mask which depends on the reflectivity distribution of the storm boundary. The results of the experiment show that this method can successfully recognize false merger. These advantages result in better forecast evaluation scores.

Key words: Doppler radar, storm identification, nowcasting, severe weather warning

摘要： 为了解决聚类过程中常碰到的虚假合并问题，针对对流风暴的特点，使用了一种特殊的腐蚀方法，即在执行腐蚀操作时考虑了风暴边缘的反射率因子的强弱分布，进行动态的卷积模板的设计，而不是始终使用一个固定的卷积模板。试验及评价的结果表明，这种方法能够成功地识别出风暴体的虚假合并，进而提高风暴体的正确识别率和预报评价指标。

关键词: 多普勒雷达, 风暴体识别, 临近预报, 强天气预警