A Bayesian approach for hydrometeor classification of polarimetric weather radar variables

In this paper, hydrometeor type classification is studied using the observations of CP-2 polarimetric weather radar located in Brisbane, Australia. The problem is formulated in a Bayesian classification framework, where total ten bulk hydrometeor types are considered. The conditional measurement distribution which describe the probabilities of radar measurements corresponding to hydrometeor types is approximated by a multivariate Gaussian distribution with parameters characterized by the scattering properties of hydrometeors. Locations and boundaries of the melting layers are estimated using reflectivity, differential reflectivity and correlation coefficient. They are then incorporated into the classification process together with convection and stratiform classification. The proposed Bayesian classification algorithm is tested using the CP-2 polarimetric radar data over 100 scan volumes and results show the consistency with cloud microphysical models.