Ocean is a primary energy source of tropical cyclones. Therefore, good parameterizations for air-sea interactions and ocean are essential to accurate simulations of tropical cyclone by numerical models. Supported by KAKUSHIN Program of MEXT, the Japan Meteorological Agency (JMA) has developed a high-resolution global atmospheric model (JMA-GSM) for purposes of the numerical weather prediction (NWP) and the future extreme weather projection. JMA-GSM has a horizontal resolution of approximately 20km and 60 vertical layers, and it was in operation for NWP at JMA on November 2007. In this study, revised atmospheric boundary layer scheme and parameterization for air-sea fluxes in JMA-GSM are examined on the simulation of tropical cyclones focusing on their intensities. Impacts of coupling with an oceanic mixed-layer model are also investigated.

The improved atmospheric boundary layer scheme, including non-local turbulent mixing and revised air-sea flux, brings a clear modification of tropical cyclone intensity in the case of typhoon Krosa-2007. However, their impact on the track prediction is quite less. On surface drag over ocean, some observational studies pointed out that under extreme wind conditions the Charnock's relation may overestimate sea surface roughness length and wind stress (e.g., Powell et al., 2003). Introducing a modified sea surface roughness parameterization based on Makin (2005), the model simulates somewhat stronger surface winds in the case of Krosa, and while its impact on the intensity is relatively small. In order to further improve intensity prediction of tropical cyclones, simple ocean models, such as a layered reduced gravity model and a one-dimensional ocean model, have been developed to be coupled with JMA-GSM. For some tropical cyclone cases (e.g., Ioke-2006), coupling with an oceanic mixed-layer model modifies a prediction of cyclone intensity through consuming heat energy of the ocean. It is thus suggested that oceanic effects from the heat content are dominative for cyclone intensity predictions rather than modifications in the surface drag.