Verification of Cloud Dynamics and Radiation Database (CDRD) passive microwave precipitation retrieval algorithm using TRMM satellite radar and radiometer measurements over Southern Mediterranean Basin

A verification study is conducted for the Cloud Dynamics and Radiation Database (CDRD) rainfall algorithm over the southern Mediterranean Basin (25-36 °N/25° W-45° E), based on TRMM Microwave Imager (TMI) passive microwave (PMW) brightness temperature (TB) measurements. The CDRD's required synthetic database is generated from a high resolution, nonhydrostatic cloud resolving model equipped with a precision RTE model. This dual modeling system is used to simulate numerous and precipitating storms over Europe and across the Mediterranean Basin as the database's foundation. Extracted from the simulations for the database are the a priori knowledge variables involving many thousands of synthetic precipitating microphysical-meteorological profiles and their concomitant synthetic multi-frequency PMW TBs. This information provides the probability framework for the CDRD algorithm's Bayesian inversion process and thus the means to relate measured satellite TBs to the most probable rainfall solutions. As the newest algorithm feature, the CDRD scheme makes use of meteorological parameters (tags) of observed precipitation systems, obtained from global NWP model synthesized initialization data, for both precipitation regime recognition and Bayesian retrieval constraints. These two steps effectively reduce non-uniqueness problems in obtaining retrieval solutions. Comparison of TMI-based CDRD retrievals to coincident TRMM Precipitation Radar (PR) retrievals, used as ground truth, provides the means to assess the algorithm's performance and the twofold use of meteorological tag information for control and constraint. PR-retrieved cloud property parameters (e.g., freezing level) are also compared to NWP model tags to further verify algorithm design performance. The verification procedure utilizes a full year (2010) of TMI and PR measurements, indicating CDRD over-ocean and over-land results are within 0.10 and 0.17 mm hr-1 of the PR means, with accompanying 1.24 and 2.94 mm hr-1 rms errors. The CDRD algorithm is also compared to the TRMM TMI facility algorithm (2al2-v7), exhibiting slight improvement over water (in correlation coefficient) and significantly improvement over land. Error statistics indicate CDRD vis-à-vis 2al2 over-land improvements of 78% and 14% in regards to mean error and rms error, respectively. © 2012 IEEE.