Performance evaluation of Wavewatch III in the Mediterranean Sea

For this study, we analyzed the performance of the wave model Wavewatch III\uae forced by a limited area atmospheric model in the Mediterranean Sea. The simulation results have been compared to buoy measurements through single point statistical indicators such as normalized bias and symmetrically normalized root mean square error. A performance evaluation of the growth-dissipation source terms and their reference parameterizations was carried out on seventeen case studies corresponding to storms in the Northern Tyrrhenian Sea and off the Mediterranean Spanish coast. The source terms introduced by Ardhuin et al. (2010) proved to be the best overall choice, although they led to an overestimation in the significant wave height under calm conditions and to an underestimation under severe conditions. A sensitivity analysis in the parameter space was performed within the neighborhood of the reference parameterization of Ardhuin et al. (2010), and a calibration was carried out to reduce the overall positive bias in the significant wave height. Furthermore, to investigate the effect from the wind forcing resolution, wind data with different resolutions was used in a sensitivity analysis. Because mesoscale features are relevant to the overall Mediterranean wave dynamics, we carried out a further investigation into the impact of the resolution on a different set of ten case studies characterized by strong mesoscale patterns. A comparison of the simulations with the measurements using single point statistical indicators shows that the high resolution results are affected by the so-called double penalty effect, although in some cases, they apparently provide a better qualitative description of the event. Finally, a hindcast covering 32 years (from 1979 to 2010) was developed using a reference parameterization from Ardhuin et al. (2010) and its calibrated variant. An analysis of the performance of the calibrated parameterization on the hindcast dataset reveals that it performs better than the reference parameterization over a wide range of wave heights, in seas that range from calm to moderate, whereas it increases the tendency to underestimate the significant wave height under severe conditions