Flood risk assessment and the design of protection measures often require the estimation of high water levels of a given probability of exceedance, i.e. design flood levels. The common approach for the estimation of design flood levels has typically three main steps. First, direct measurements of annual maximum water levels in a river cross-section are converted into annual maximum flows by using a rating curve. Second, a probability distribution function is fitted to these annual maximum flows to derive discharges of a desired probability of exceedance, i.e. design peak flows. Third, a hydraulic model is applied to derive the corresponding design flood levels. Each of the three steps is associated with significant uncertainties, affecting the accuracy of estimated design flood levels. In this study, we compare this common approach with an alternative one based on the statistical analysis of time series of annual maximum water levels. The rationale behind this study is that high water levels are directly measured and often come along with higher precision and accuracy than peak flows. While the direct use of high water levels is typical in coastal flood hazard and risk assessment, the potential of this approach in the context of river flooding has not been sufficiently explored. In this study we compare the common approach with an alternative approach based on statistical analysis of annual maximum water levels.