基于聚类分析的冰雹预测研究

Prediction of Hail Based on Clustering Analysis

冰雹是我国频发的气象灾害之一,冰雹云的预测是对强对流天气的重点监测对象,文中所采用的信息源图像为雷达回波反射率图像。主要从冰雹云的形态特征、基本反射率因子这两个方面进行分析计算,运用MATLAB程序提取出可能出现降雹的有效区域,进行聚类分析判别,识别出雹云单体所在的有效区域。通过对训练样本的聚类归纳总结,基于对降雹云单体的形态特征、反射强度研究的一般规律,建立判别函数。在此模型基础之上,计算出反射强度值大于45 dBZ可能出现降雹区域的概率,从中获取一个参数值,当概率大于这个预定的参数值时,则可能出现降雹天气。通过仿真实验结果表明,此模型对冰雹云的预测有良好的判别效果,可以应用到实际生活中去,做到提前预警,减少经济损失。

Hail is one of the most frequent meteorological disasters in China. Hail cloud prediction is the key monitoring ob- ject for the strong convective weather information. Information source images used in this paper is the radar reflectivity image. By calculating respectively from the morphological characteristics of hail cloud and basic radar reflectivity factor, and by using MATLAB program to extract the possible effective area of hail, the paper can identify the effective area of hail cloud monomer, and establishes discriminant functions through the summarizing of clustering the training samples, based on the general rules of the reflection intensi- ty and the morphological characteristics of hail cloud monomer. On this basis of the model, the probability of hail areas which reflec- tive intensity value is greater than 45 dBZ can be figured ont. When the parameters of the probability are greater than this predeter- mined value, it is likely to have hail weather. The simulation results show that the prediction model of hail cloud has a good recogni- tion efi^ct, which can be used in the real life for early warning and reducing economic losses.