A Physical Validation Approach for Precipitation

This proposal consists of two separate but complementary components. The first objective deals with the rather routine but necessary task of verifying that the operational AMSR rainfall product is as good as possible within the limitations of extremely sparse validation data outside the tropics. This will be accomplished through continuous and careful comparisons to existing missions such as TRMM, AMSR on ADEOS-II, and ground validation estimates being carried out over the Baltic Sea region and Eureka, California. The latter two sites are part of the AMSR extra-tropical validation activities. While straightforward, this activity requires considerable effort to prepare datasets, carry out the intercomparisons, report the results to the AMSR science team and the broader community and correct any algorithm deficiencies should these be encountered. This effort is briefly described under ``Routine Validation'' in the proposal. The second, and scientifically more challenging task, is described under ``Physical Validation''. Here, data collected through the AMSR field experiment carried out in Wakasa Bay, Japan, in Jan-Feb 2003, is used to address the question of how to validate rainfall globally with only sparse or sometimes non-existent ground validation data (e.g., most of the world's oceans). Physical validation, as defined by the AMSR rainfall team, distinguishes between (a) algorithm validation -- verifying that the algorithm is properly formulated (i.e., that it properly distinguishes assumed parameters from retrieved ones), and provides correct results when any assumed parameters are specified from external sources, and (b) algorithm performance -- determining the uncertainties introduced by imperfect knowledge of any assumed parameters. While the Wakasa experiment was designed to address both objectives, only the first is dealt with in this proposal. For this objective, we intend to use data from the entire suite of radars and radiometers collected by the P3 aircraft (high resolution radar data at 10, 35 and 94 GHz plus Doppler velocities as well as 17 radiometer channels between 10.7 and 340 GHz) to address a very simple question: Can we, within the framework of the satellite retrieval assumptions, construct a hydrometeor profile that is consistent with all the observed radar and radiometer observations for the variety of meteorological conditions observed during the experiment? In answering this question, we hope to learn whether or not we fully understand the forward radiative calculations sufficiently to have properly formulated the inverse problem. Despite the rather long history of satellite rainfall measurements, this question has not been answered to date. Yet, it is crucial to formulating algorithms and accompanying error models that are needed to assign uncertainties in regions of the world where ground-based data simply does not exist. If successful, it then becomes straightforward to replace any assumptions in the operational algorithm with retrieved hydrometeors from the field campaign and test the fidelity of the algorithm itself. A final part, but one that is described only very briefly, is to extend the above study to focus on ice concentrations in the regions surrounding the precipitation where cloud retrievals generally do poorly and rainfall retrievals make no estimate at all. The intent is to bridge the two retrievals in a physically consistent formulation.