Validation of Ice Cloud Microphysical Properties Retrieval Using a Markov Chain Monte Carlo Algorithm

Abstract The Markov chain Monte Carlo (MCMC) algorithm has been used for retrieving ice cloud microphysical properties in this paper. The retrieval data include effective radius (re), ice water content (IWC), number density (N0), and distribution width parameter (w). First we examine the algorithm feasibility using simulated W‐band radar reflectivity. The retrievals locate in a solution space around the exact results. Two methods are applied to produce a unique solution with maximum likelihood function, that is, the mean value of all acceptable sets of parameters in the chain and the first convergence sample. The results of the second method have a better performance compared with that of the first method. The algorithm has better performance when one particle size distribution (PSD) parameter is known than that when three PSD parameters are unknown. Then we apply the algorithm for CloudSat Cloud Profiling Radar (CPR) observations. Four habits, including column, plate, bullet rosette and sphere, are considered to analyze the influence on the retrieval accuracy. The averaged deviation of res and IWCs obtained from MCMC algorithm among different habit assumptions has a maximum of about 10% and 5%, respectively. It appears clearly that the MCMC algorithm retrieval for short hexagonal column (HEXS) are the closest to that of the official products (2B‐CWC‐RO), while thick hexagonal plate (HEXF) has the largest bias. Retrieved re, IWC, N0, and w from radar reflectivity are in reasonable agreement with 2B‐CWC‐RO product, which indicate that the MCMC algorithm can produce reliable cloud properties from radar observations.