Abstract
The impact of optimizing the air-sea exchange coefficients and the initial state is examined for a tropical cyclone (TC) with the aim of improving an operational mesoscale four-dimensional variational data assimilation system. In our optimization approach for TC Chaba that approached the Ryukyu Islands in October 2010, the drag coefficient values are adjusted so that they saturate at extreme conditions, while the first guess increases with increasing wind speed. In addition, the latent heat exchange coefficient values are adjusted so that they significantly increase under extreme conditions relative to the first guess. Consequently, the term of the cost function, which represents the discrepancy between the model results and the observational data, decreases by 4.1-22.4% relative to the existing system after some spin-up cycles. The intensity and location of the TC are brought close to those of the corresponding best track produced at the Japan Meteorological Agency. Our optimization approach has the potential to improve the forecast skill.
