Particle Identification for Polarimetric Weather Radar Using A Bayesian Method

The distribution and structure of particle types in clouds have research andapplication significances for cloud microphysical studies, quantitative precipitation estimation, and severe weather forecasting. In literature, particle identification often adopts a fuzzy-logic based method, however, the construction of membership functions in this method heavily relies on empirical relations. We have developedanother particle identification approach for polarimetric weather radar by using a Bayesian method.The new approach has be applied to (1) investigating the relationship between polarimetric radar measurements and particle types based on the clustering techniques, and derive the probability density functions for particle types; (2) modellingthe prior probability using contextual information based on Markov random field, the stratiform and convective classification, and the location of melting layer; and(3) analyzing the microphysical characteristics of typical precipitating clouds. This method can provide accurate particle types for the study of cloud and precipitation formation processfor the weather cases of China,it can also improve the understanding of quantitative precipitation estimation and numerical weather prediction.