CIP: Confidence intervals for SWAN wave forecasts : surrogate North Sea wave model to apply wind ensembles

Generally, the performance of the SWAN wave models used at Rijkswaterstaat Operational Systems (RWsOS) is well known and model bias is within a few percent. In contrast the spread can be large in certain cases, in particular for low frequency waves He10. Therefore, we aim to model uncertainty to anticipate deviations in the wave forecast from actual wave conditions. An important constraint for RWsOS is that the future model is computationally feasible. The approach for this report is twofold. The uncertainty was first modelled and compared with observations to verify the uncertainty model forecast. Secondly, a fast data-driven surrogate model TurboSWAN was trained on the ensemble wave forecast input and output and compared to the physical wave model SWAN results.