Morphological impact of the Sinterklaas storm at Het Zwin

Abstract

A very large storm hit the coasts of The Netherlands-and most of north-west Europe-during the evening of the 5th of December 2013. The Sinterklaas storm, as it was later known, induced extremely high water levels and substantial dune erosion all along the Dutch coastline. In this thesis the morphological impact of the Sinterklaas storm at one particular location-Het Zwin-will be analysed from measurements and simulated with the process-based model XBeach. Het Zwin is a relatively small natural reserve shared between The Netherlands and Belgium. It is a brackish reservoir dominated by the tide, composed mainly by tidal flats protected from the sea by beach dunes and separated from the hinterland by clay dikes. Actually it is hard to classify the Zwin; it is not really an estuary, given its small dimensions, nor a tidal inlet since no lagoon is present behind the dunes. Probably it is best defined by the Dutch word "slufter". Two different storm regimes were observed at Het Zwin during the Sinterklaas storm: dune erosion (collision), and storm overwash, the latter confined at a precise location where the Dutch dunes were lower and narrower. The storm impact was deduced from lidar measurements of the terrain elevation prior and after the event, and the hydrodynamic conditions of the storm were obtained from wave buoys and tide gauges deployed and maintained by both countries. These data were used as inputs for a numerical hind-cast of the storm impact. Calibration of the model considered sensitive parameters that were either meaningful physical inputs, such as the bottom friction, or numerical proxies for physical processes, such as the critical slope for avalanching. As suggested in previous studies, the collision regime was found to be dependent on the onshore transport induced by short-waves, whereas the overwash regime and the washover fan were determined by the bottom friction at the higher parts of the dunes. The best fit was obtained with parameters facua = 0.10, and C = 25, respectively. Additionally, the critical slope for avalanching which produced the best fit, both in sand loss and profile shape, was found to be wetslp = 0.20, slightly lower than the recommended default value of 0.30. However, it is argued that the actual value to be used in other studies might be a function of the grid size, since it determines the ability of the model to assess the terrain slope. Morphodynamic numerical modelling is normally a costly task. For this particular site of study three factors induced very large simulation times, which rendered the modelling with XBeach unworkable: 1. A rather large domain is needed to incorporate the refraction of the wave data. 2. A very fine grid is required in order to reproduce the flooding and drying processes of the overwash and (in a lesser way) the avalanching of the dunes' face. 3. The duration of the storm, of a few days, is rather large compared with the physical events being modelled, which have a time scale of a fraction of seconds. A central issue in this thesis is the optimization of the runtime of XBeach (and by extension, any other process-based numerical model). Two techniques are discussed and tested, which allowed the hind-cast to be computed in just over a day: 1. The use of curvilinear over rectangular grids, in order to focus computational effort. The same precision can be obtained using fewer grid cells. 2. The computation of the hydrodynamic forcing by waves and water levels only once, and later feed those results to a smaller dedicated grid to perform the morphodynamic computations. This is a straightforward offline coupling scheme. The use of these techniques significantly reduce the computation effort without compromising the accuracy nor the reliability of the results.