Center for Earth Observing and Space Research/School of Computational Sciences, George Mason University, Fairfax, Virginia
Department of Atmospheric Science, Colorado State University, Fort Collins, Colorado
A methodology is described to construct fully parametric rainfall retrieval algorithms for a variety of passive microwave sensors that exist today and are planned for the future. The Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) is used to retrieve nonraining geophysical parameters. The method then blends these background geophysical parameters with three-dimensional precipitation fields obtained by matching the TRMM precipitation radar (PR) reflectivity profiles with cloud-resolving model simulations to produce a consistent three-dimensional atmospheric description. Based upon this common description, radiative transfer simulations corresponding to specific microwave sensors are then employed to compute radiances from clear and rainy scenes, as might be seen by any specified microwave radiometer. Last, a Bayesian retrieval methodology is used in conjunction with this database to derive the most likely surface rainfall as well as its vertical structure. By avoiding any dependencies on specific channels or channel combinations, the technique can readily be adapted to different sensor configurations. The algorithm performance is tested for a variety of sensor designs using synthetic retrievals to demonstrate its capability for consistent rainfall estimates. Whereas actual retrievals would be sensitive to the details of the a priori database construction, results from this study indicate that even modest radiometers can retrieve unbiased rainfall rates when constrained by an a priori database constructed from the TRMM satellite. Random errors are correlated to unobserved variations in the vertical and horizontal structure of the precipitation and, thus, depend upon sensor design specifications. The fidelity of these synthetic retrievals is briefly examined by comparing the simulated brightness temperature (Tb) generated in this study with direct observations by the TRMM TMI. Good physical consistency between the simulated and TRMM observed Tbs is found in precipitating regions for frequencies at which emission processes dominate the radiometric signal. The consistency is poor for higher-frequency microwave channels for which ice scattering is important. Greater consistency between the computed and observed Tbs should be sought before replacing current operational algorithms with the parametric equivalent.