Comparison study of a second-generation and of a third-generation wave prediction model in the context of the SEMAPHORE experiment

Numerical wave prediction models presently used in the meteorological insti tutes are still of two types: the so-called second-generation and third-gen eration models. In this paper, the authors present a comparison of the perf ormance of a second-generation model-the VAG model from Meteo-France-and of the third-generation WAM model. These two models have been run with simila r characteristics (same wind input, same resolution). Simple tests show the differences between the behaviors of VAG and WAM in typical situations (co nstant wind, rotating wind). Hindcasts have been performed in the general c ontext of the SEMAPHORE experiment. A one month hindcast over the North Atl antic domain has been run by driving both models with the same wind fields. A comparison between the model output and the available observations, incl uding significant wave height from satellite measurements, is presented. Th e results show that VAG and WAM results are in a general good agreement wit h the observations, but also that WAM results are a little better than VAG when the satellite data are taken as a reference. A modification of VAG is then proposed, which allows the performances of VAG to be closer to those o f WAM. This study shows that (i) the second-generation VAG model is nearly as good in predicting wave heights as the third-generation model WAM in spi te of its poor representation of the nonlinear interactions and (ii) VAG ha s been improved when introducing the growth and dissipation terms of WAM in stead of parameterizations taken from Golding.