|Ontwikkelingen in het spectraal modelleren van windgolven in kustgebieden = Advances in the spectral modelling of wind waves in the nearshore|
Van der Westhuysen, A. (2007). Ontwikkelingen in het spectraal modelleren van windgolven in kustgebieden = Advances in the spectral modelling of wind waves in the nearshore. PhD Thesis. Technische Universiteit Delft: Delft. ISBN 978-90-9022235-6. viii, 207 pp.
Water waves > Surface water waves > Wind waves
|Auteur|| || Top |
SWAN is a numerical model for the simulation of wind-generated waves in coastal regions. Application of SWAN to numerous idealised and field situations has proved the robustness and general good performance of the model, especially with respect to the prediction of wave height in both deep and shallow water. However, some inaccuracies and spurious behaviour remain. A characteristic inaccuracy of SWAN is a persistent underprediction of the wave period measures. These results are also affected by the presence of swell, which can cause a strong overprediction of wind sea growth in combined swell and wind sea conditions. In the surf zone, the reproduction of spectra tends to be rather inaccurate, leading to large scatter in the results of period measures. In addition to these concerns about model physics, some aspects of the numerical implementation of SWAN may negatively influence model outcomes. In particular, the so-called action limiter (used to enhance numerical stability) may affect model outcomes during stationary simulation.
The aim of the present study is to determine whether the general accuracy of SWAN can be improved by implementing alternative expressions for source terms in deep and shallow water, and by improving the numerical solution of the action balance equation. Two numerical aspects concerning stationary simulation are considered, namely the influence of the action limiter on model outcomes, and the criteria that determine when the iterative solution procedure should be ended. To this end, the outcomes of simulations using a number of numerical techniques for the integration of the source terms were compared, and the iteration behaviour of the model was studied. Concerning model physics, it was first investigated to what extent the accuracy of SWAN in deep water and water of finite depth can be improved by replacing the DIA quadruplet source term by an exact method that computes the full set of quadruplet interactions. Subsequently, it was investigated whether the accuracy of SWAN could be improved by using a whitecapping expression that, for breaking waves, is dependent on variables that are local in the frequency spectrum, as opposed to spectrally averaged. To improve model performance in the surf zone, a twoequation triad interaction model that includes all resonant and near-resonant interactions was implemented, for the reduced case of parallel depth contours. The performance of SWAN using this alterative triad source term was compared with the performance when using the LTA.
A major finding concerning the numerics in SWAN is that the action limiter does not significantly influence converged model results. A second important finding concerning stationary simulation is that the default convergence criteria are insufficient and can yield unconverged results that differ significantly from fully converged ones. Alternative, stricter convergence criteria are proposed. Concerning deep water source terms, it was found that replacing the DIA with an exact method for quadruplet interaction improves model performance somewhat, but is prohibitively expensive for operational applications. Model accuracy in deep and intermediate water depths is improved by using a whitecapping dissipation expression that, for breaking waves, is dependent on variables that are local in frequency space. The improvement in model performance is particularly evident under combined swell and wind sea conditions, where the default expression fails because of dependencies on mean spectral variables. Concerning shallow water source terms, it was found that the details of frequency spectra can be significantly improved through the use of a triad interaction source term that takes all sum and difference interactions into account. However, this model variant is significantly more computationally intensive than the default version (LTA).