This study is to estimate rainfall over oceans around Taiwan using Bayesian approach during the Typhoon season. Simulated brightness temperatures of Tropical Rainfall Measurement Mission (TRMM) Microwave Image (TMI) are obtained through a 3-D microwave radiative transfer model inputting the outputs of the Weather Research and Forecast (WRF) Model. Considering the complicated physical processes between rainfall and microwave measurements, the normalized polarization index P (Petty, 1994) is used in this study and the rainfall algorithm is then established by the Bayesian approach. High resolution Infrared data, collocated within the FOV of the pixel of TMI, are also employed to eliminate the beam-filling problem.

The preliminary result shows the retrieval rainfall pattern is similar to those of GPROF (Goddard Profiling Algorithm) and our previous statistical retrieval result, but the rainfall intensity is greater in this study. The rainfall retrievals have also been validated with rain gauge data. The result also shows that the rainfall seems to be overestimated by the Bayesian approach for weak precipitation system and slightly underestimated for heavy precipitation system. Some evidences show that the amount of snow and ice seems to be overestimated by the current version of WRF model compared with observations. Therefore, more efforts are needed to treat the outputs of the cloud-resolving model and to simulate more reasonable brightness temperatures of TMI in order to obtain acceptable rainfall retrievals.

Key words: Bayesian, normalized polarization, rainfall retrieval, WRF, TMI, GPROF