A major component of next-generation operational models will include coupling between meteorology and wave models. However, most current ocean wave models use non-physically based source terms, often empirically derived, as forcing functions in the spectral density transport equation. For example, the wind-wave interaction parameterization term often accounts for growth only due to vertical wind shear in the surface layer, neglecting the effects of viscous dissipation in the air and water. Furthermore, most wind-wave interaction schemes are valid only for low wind speeds, neglecting growth due to boundary layer turbulence in windy regimes. Sajjadi (2001) has derived a physically-based parameterization scheme which is not only more accurate in slow wind regimes, but accounts for turbulent interaction in windy conditions.

Typically, meteorology models account for ocean wave interaction implicitly through the Charnock relationship, which is only valid for an ocean consisting of mostly swells (an "old windsea.") When the wave field is dominated by a young windsea during the initial stages of wave growth, the chaotic sea yields larger roughness length values. For the latter situation, the roughness length is related to the wave energy transfer coefficient, thereby enabling feedback from a wave model to an atmospheric model. In turn, the modified stress field will impact the wave model, thereby enabling two-way coupling.

To assess the Sajjadi wind-wave parameterization scheme for turbulent regimes, and its impact on an atmospheric model, results will be presented of a two-way coupled COAMPS-WAVEWATCH simulation for Hurricane Gordon. The sensitivity of the hurricane simulation to different wind-wave parameterization schemes will be presented. A second wind-wave interaction parameterization term, which accounts for wind gusts, will also be presented.

The presentation will end with a discussion of a new software package called the Model Coupling Executable Library (MCEL), which allows the exchange and manipulation of data between two simultaneously running models. This requires writing a C or FORTRAN program which computes the coupling expressions, and then specifying model information into different library calls. In this manner, little rewriting of both model codes is needed, and/or the practice of combining both models into one code for coupling expressions can be avoided.