When hindcasting wave fields of storm events with state-of-the-art wave models, the quality of the results strongly depends on the meteorological forcing dataset. The wave model will inherit the uncertainty of the atmospheric data, and additional discretization errors will be introduced due to a limited spatial and temporal resolution of the forcing data. In this study, we apply an atmospheric downscaling to (i)?add regional details to the wind field, (ii)?increase the temporal resolution of the wind fields, (iii)?provide a more detailed representation of transient events such as storms and (iv)?generate ensembles with perturbed atmospheric conditions, which allows for a flow-dependent and spatio-temporally variable uncertainty estimation. We test different strategies to generate an ensemble hindcast of a relatively strong storm event in February 2002 in the Baltic Sea. The Weather Research and Forecasting (WRF) model used for this purpose is driven by the ECMWF ERA5 reanalysis, and wind fields are passed to the third-generation wave model WAVEWATCH III?. A combination of initial conditions from the ERA5 ensemble of data assimilations and stochastic perturbations during runtime is identified as the most promising strategy. The final aim of the ensemble approach is to quantify the hindcast error, but this approach can also be used to generate alternative representations of historical extreme events to sample the recent climate and to increase the sample size for statistical studies, such as for civil engineering applications for coastal protection studies.
展开▼
