TRR 181 Seminar "Development and applications of a 4D variational assimilation scheme for wave models"

A 4D variational assimilation scheme including partition methods for nearshore wave models has been tailored to SWAN model for adjusting both, wave boundary conditions and wind fields. Nonstationary wave boundary conditions were assimilated by providing defined correlations of model inputs in time. Twin experiments covering both stationary and nonstationary wave conditions were carried out to assess the adequacy of the proposed scheme. The value of the cost function declined to less than 5% and root mean square errors (RMSE) of spectrum were reduced to less than 10% under stationary wave conditions. Under nonstationary wave conditions, the RMSE were reduced to less than 10% after 30 iterations in most assimilation windows. Also, the proposed scheme was modified for low spatial observation coverage by assuming a group of 'basic' inputs to contain all errors so as to be applied in the German Bight. The modified scheme using a single observing location performed comparable assimilation effect with the original scheme without the basic inputs assumption using 25 observing locations in twin experiments. For both schemes wave spectrum RMSE were reduced by around 50% throughout the whole computation domain. The practical experiment revealed that the wave spectrum RMSE at the validation buoy declined by more than 60% by means of the modified assimilation scheme with observations from a single in-situ buoy. The results can suggest that the developed 4D variational scheme is capable of improving the model results effectively under wave conditions when nonlinear interactions are insignificant. The modified scheme with ’basic’ inputs can be applied in coastal areas where observations are spatially limited provided sea states are strongly correlated in space.