Design water levels are a basic concept in flood risk management practice. These water levels with a specified return period are used for the design of dikes and other flood protection measures along the river Meuse. Current practice is that these water levels are determined by hydrodynamic simulation of a standard design hydrograph at the upstream gauging station Borgharen. The peak discharge of this standard design hydrograph is based on a frequency analysis and its shape is determined by scaling and averaging all flood hydrographs in the dataset. Then it is assumed that the simulated water levels have the same return period as the peak discharge at Borgharen. Until now, this method was not validated. The aim of this thesis is to investigate the influence of hydrograph shape on design water levels on the river Meuse, and to evaluate current and alternative methods to take this shape into account. The five alternative methods are (1) hydrodynamic simulation of all floods in the dataset and apply frequency analysis afterwards on the simulated water levels (the reference), (2) the standard method extended with dependence between the hydrograph shape and the peak discharge, (3) vertical averaging, and two probabilistic methods which combine hydrograph shape statistics at Borgharen with a transformation function that relates local water levels to these hydrograph shape statistics. Within the probabilistic methods one can distinguish explicit (4), which expresses the statistics in probability distribution functions, and implicit (5), which does not use these functions. The influence of hydrograph shape variables on the downstream water level was investigated by means of a correlation analysis. To evaluate the different methods, all methods were applied to a GRADE dataset of 50,000 years of generated discharge at Borgharen, and the resulting design water levels compared to the reference. In addition to the ability to estimate design water levels, the methods were evaluated on the ability to estimate the design water level reduction of a retention basin (Lob van Gennep). Peak discharge combined with peak curvature were found to be good predictors of the downstream water levels, and were used in the probabilistic methods. The evaluation of methods shows that the currently used standard hydrograph method overestimates the design water levels up to 37 cm with respect to the reference. The present research shows that the current method to determine design water levels can be improved significantly. A simple improvement is to use the vertically averaged design hydrograph with a modified selection interval. More advanced probabilistic methods also improve the estimates, and are potentially valuable in case of retention basins, of which the effectiveness is sensitive to the hydrograph shape.