A Bayesian Approach to Microwave Precipitation Profile Retrieval

Abstract

A multichannel passive microwave precipitation retrieval algorithm is developed. Bayes theorem is used to combine statistical information from numerical cloud models with forward radiative transfer modeling. Amultivariate lognormal prior probability distribution contains the covariance information about hydrometeor distributions that resolves the nonuniqueness inherent in the inversion process. Hydrometeor profiles are retrieved by maximizing the posterior probability density for each vector of observations. The hydrometeor profile retrievalmethod is tested with data from the Advanced Microwave Precipitation Radiometer (IO, 19, 37, and 85 GHz) of convection over ocean and land in Florida. The CP-2 multiparameter radar data are used to verify theretrieved profiles. The results show that the method can retrieve approximate hydrometeor profiles, with larger errors over land than water. There is considerably greater accuracy in the retrieval of integrated hydrometeor contents than of profiles. Many of the retrieval errors are traced to problems with the cloud model microphysicalinformation, and future improvements to the algorithm are suggested.