利用贝叶斯方法改进华南地区冰雹识别效果  

USING BAYESIAN METHOD TO IMPROVE HAIL IDENTIFICATION IN SOUTH CHINA

作　　者：李博勇 胡志群[2] 郑佳锋[1] 陈超[3] LI Bo-yong;HU Zhi-qun;ZHENG Jia-feng;CHEN Chao(School of Atmospheric Sciences,Chengdu University of Information Technology,Chengdu 610225,China;State Key Laboratory of Severe Weather,Chinese Academy of Meteorological Sciences,Beijing 100081,China;Guangdong Meteorological Observatory,Guangzhou 510641,China)

机构地区：[1]成都信息工程大学大气科学学院,四川成都610225 [2]中国气象科学研究院灾害天气国家重点实验室,北京100081 [3]广东省气象台,广东广州510641

出　　处：《热带气象学报》2021年第1期112-125,共14页Journal of Tropical Meteorology

基　　金：国家重点研发计划(2019YFC1510304);灾害天气国家重点实验室开放课题(2020LASW-B02);广东省重点领域研发计划(2020B1111200001)共同资助。

摘　　要：使用2019年广东S波段双偏振雷达观测的冰雹和非冰雹数据,统计得到冰雹和非冰雹的雷达反射率Z、差分反射率ZDR和相关系数CC先验概率密度分布,采用贝叶斯方法,根据雷达参量在冰雹和非冰雹条件下的概率以及冰雹和非冰雹的先验概率来确定某一距离库上所测到的(Z、ZDR、CC)所代表冰雹和非冰雹的概率,并用两个个例,比较分析了WSR-88D冰雹识别算法和贝叶斯方法对冰雹识别的效果,分析表明,两种方法都能较准确地识别出冰雹云,但是贝叶斯方法识别范围较大,这可能与华南地区多为雨夹雹有关。This paper uses the hail and non-hail data observed by the Guangdong S-band dual-polarization radar in 2019 to statistically obtain the priori probability density distribution of hail and non-hail radar reflectivity Z, differential reflectivity ZDRand correlation coefficient CC. Based on the probability of radar parameters under hail and non-hail conditions and the prior probability of hail and non-hail, the present study uses the Bayesian method to determine the probability of hail and non-hail represented by(Z, ZDR,CC) measured at a certain range bin. Two cases are also used to compare and analyze the effects of WSR-88 D hail recognition algorithm and the Bayesian method on hail recognition. The analysis shows that both methods can identify hail clouds more accurately, but the Bayesian method has a larger recognition range,which may be due to the rain mixed up with hail in southern China.

关 键 词：偏振雷达 贝叶斯方法 冰雹 相态识别 

分 类 号：P412.25[天文地球—大气科学及气象学]