For time-accurate numerical modelling of short waves, among other applications, the following are important: non-hydrostatic effects must be taken into account; computational efficiency (since typical computations consist of long time series on fine grids); and the existence of little or no numerical dispersion and numerical dissipation. WL / Delft Hydraulics is studying whether the extension of Delft3D-FLOW with a non-hydrostatic module similar to the one in non-hydrostatic TRIWAQ (an R&D version developed by Zijlema (RIKZ)) is a good idea. In addition to this, WL / Delft Hydraulics has developed a Boussinesq-type wave model named TRITON which is expected to be launched on the market within a few years. Recent discussions regarding swell propagation in the Amelander Zeegat led to the need to compare TRITON to non-hydrostatic TRIWAQ. For this purpose, we studied a situation in both models which resembles a scenario frequently encountered in coastal areas: wave propagation on a sloping bar. The question addressed in this report is: What is required to compute propagation of waves (phase-resolved) over a large domain (several tens of wavelengths) with an acceptable degree of accuracy? Due to the fact that in non-hydrostatic TRIWAQ some hydrostatic elements are also different from those contained in Delft3D-FLOW, it was considered wise to also perform some Delft3D-FLOW computations to study the wave propagation characteristics of that model (even though it is a hydrostatic model). The results may be summarised as follows:
- TRITON and non-hydrostatic TRIWAQ have similar resolution (spatial and temporal) requirements. Also, the computational performance of both models (two layers in nonhydrostatic TRIWAQ) is roughly the same.
- The standard method applied in Delft3D-FLOW and TRIWAQ to deal with open boundaries (prescribing Riemann or alpha-boundary conditions at the outflow) appears to be insufficient for practical applications. Also partial reflection, typical for short-wave applications, is not yet included in these two models. TRITON does have accurate and robust modelling of open boundaries and partial reflection. Non-hydrostatic TRIWAQ makes use of absorbing boundary conditions at outflow boundaries by means of the sponge layer technique.
- In coastal areas, a suitable wave breaking model is required. If this type of model is not available, the simulated waves become too high and too steep, and lead to all sorts of unwanted numerical effects (numerical breakdown, waves trapped on the finest grid scales, large numerical dispersion). In the event of high waves, a non-hydrostatic model such as TRITON or non-hydrostatic TRIWAQ is required even if the principal wave shows only a slight amount of dispersion. This is caused by the higher harmonics, whose presence is important and which contain a large amount of dispersion. This clearly illustrates the need for non-hydrostatic models. |
