Assimilation of TRMM and SSM/I observations in a global spectral model: A case study of Nov–98 tropical cyclone

Medium range prediction over the tropics is known to suffer from the problem of "spin-up" which refers to unrealistic simulation of rainfall patterns during the initial hours of model prediction. To counter this problem, Krishnamurti et. al (1991) suggested a unique approach of " physical initialization " in which the model initial fields of humidity and divergence are adjusted in the initial fields in such a way that the predicted rainfall amount and diabatic heating are consistent with the respective observed patterns. However, the success of this approach depends on the availability of accurate observations of rainfall. With the use of passive microwave techniques, physically direct and reasonably accurate observations measurements of rainfall and precipitable water (PW) can be made in "all weather conditions" . Special Sensor Microwave/Imager (SSM/I) sensor onboard US defense satellite DMSP has been providing these parameters very successfully for the past decade. TRMM Microwave Instrument (TMI), a sensor designed on the lines of SSM/I , and is a part of ambitious TRMM (Tropical Rainfall Measuring Mission) , is focused at providing high resolution observations at tropical regions.

In the present study we have made an attempt to assimilate the precipitation and PW observations from TMI and SSM/I in a global spectral model at T-126 resolution. The period selected for the experiments was May 13 to 21 , 1999, which covers the development, intensification, and the landfall phases of a tropical cyclone over the Arabian Sea . ECMWF analysis was taken as the initial state for these experiments. Rainfall assimilation was carried out using physical initialization approach, while, the PW fields were assimilated by using satellite observed PW fields as additional constraint during the physical initialization procedure. Our approach is partially based on the technique suggested by Kuo et al (1994). During the assimilation, more weight was given to the TMI observations at the points where both TMI and SSM/I observations were present. Our preliminary results suggest that the assimilation of PW in addition to rainfall gives an improvement in predicted rainfall distribution and cyclone track, compared to the experiments where only rainfall was assimilated.