A precipitation retrieval using TMI during the Baui Period of 1988

A new over-ocean precipitation retrieval algorithm is presented in this study using multi-channel brightness temperatures(TB's) of TRMM Microwave Imager (TMI) measurements. The basic idea of this algorithm is to find the optimal precipitation field which gives radiative transfer model calculated TB's that fit best with the TMI TB's at 10, 19, and 85 GHz. The algorithm utilizes vertically-polarized TB's in order to minimize the sea-surface wind effects. The radiative transfer model of Liu (1998) is used for computing field-of-views(FOV) averaged TB's from precipitation fields. For the radiative transfer calculation, spatial inhomogeneity of precipitation and freezing level height are estimated from TMI TB's. The agreement between the calculated TB's and the observed TB's is expressed as a cost function which is a sum of square of the statistically-normalized differences between the above 2 TB fields. The optimal precipitation field with 10 km resolution is obtained by solving the gradient equation of the cost function.

Precipitation retrievals using this algorithm were validated using TRMM Precipitation Radar (PR) data at spatial resolution of 50 km around western part of Japan during the rainy season (June-July) of 1998. The results indicate : 1) The precipitation retrievals at the 50 km resolution agreed well with PR precipitation, even for heavy precipitation areas. 2) Spatial precipitation inhomogeneity estimated by this algorithm were highly correlated with Spatial precipitation inhomogeneity analyzed from PR data when 50 km average of precipitation (PR50) is over 2 mm/hr. 3) Good agreement was found between precipitation coverage estimates by this algorithm and PR rain flag data averaged over 50 km, except for a very weak precipitation ( PR50 < 0.5 mm/hr).

Some experimental retrievals were executed to examine impacts of the parameter estimation (spatial inhomogeneity of precipitation and freezing level height) on precipitation retrievals. The results indicate : 1) Use of the estimated spatial inhomogeneity alleviated underestimation due to beam-filling error in a heavy precipitation range. 2) Use of the estimated freezing level height improved the accuracy of precipitation retrievals for weak precipitation.