This paper describes a precipitation-retrieval algorithm for the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) that was developed under the Global Satellite Mapping of Precipitation project (GSMaP) by improving the authors' previous algorithm. The basic idea of the GSMaP algorithm is to find the optimal precipitation for which the brightness temperatures (TBs) calculated by the radiative-transfer model (RTM) fit best with the observed TBs. The main improvements of the GSMaP algorithm over the authors' previous work are as follows: (1) use of precipitation-related variable models (precipitation profiles, drop-size distribution, etc.) and precipitation detection and inhomogeneity estimation methods based on TRMM observation studies; (2) use of scattering signals of the TMI Polarization-Corrected Temperature (PCT) at 37 and 85 GHz (PCT37, PCT85) and scattering-signal correction for tall precipitation (thickness between precipitation top level and freezing level (Dtop) larger than 6 km) over land and coastal areas.
In order to validate the GSMaP algorithm, we compared its retrievals from TMI TBs in 1998 with the TRMM Precipitation Radar (PR) and Goddard Profiling Algorithm (GPROF) retrievals (2A12 version 61). The results show that (1) over land and coastal areas, the GSMaP retrievals agreed better with PR than GPROF for tall precipitation (Dtop > 4 km) weaker than 10 mm h(-1), while both GSMaP and GPROF underestimated PR precipitation rates for precipitation heavier than 10 mm h(-1); (2) over ocean, the GSMaP retrievals agreed better with PR than GPROF for precipitation heavier than 10 mm h(-1), while GSMaP slightly overestimated precipitation weaker than 10 mm h(-1) compared to PR and GPROF; (3) The GSMaP algorithm failed to detect some precipitation areas weaker than 2 mm h(-1) over sub-tropical oceans.
Experimental algorithms with different precipitation-related variable models and retrieval methods using scattering signals were applied to TMI TBs in July 1998 to examine the effect of the above improvements to the GSMaP algorithm. The results show that the improvement of the precipitation profile alleviated the underestimation of precipitation heavier than 10 mm h(-1) over land and coastal areas, that the combined use of new physical-related variable models alleviated the underestimation of precipitation heavier than 10 mm h(-1) over ocean, and that the use of PCT37 and scattering-signal correction reduced the overestimation of tall precipitation (Dtop > 4 km) weaker than 10 mm h(-1) over land and coastal areas.