Statistical Methods and Representation of Extreme Discharge — Findings from a Distributed Hydrological Modelling Study

In the process of developing a physically-based fully distributed hydrological model for the Grote Nete catchment in Belgium, it was desired to improve on the modelling of the extreme flows. To understand the relationship between river network representation and simulation output quality, a set of simulations was designed to investigate firstly, whether the representation of the Grote Laak River may be responsible for the difficulty in representing discharge conditions at Vorst; secondly, whether improved modeling of the flow hydrograph at Vorst would result in improved modelling of the peak flows at Varendonk; and thirdly, to obtain an understanding of the importance of river network modeling in a local versus a regional context. This paper presents the results of the multi-site, multi-criteria calibration protocol that was used in the model development process. This included both traditional goodness-of-fit statistics, and graphical comparisons of indicator variables. While the high discharge values were generally well simulated, some extreme high values were underestimated while some extreme low values were overestimated. This observation was investigated. The paper presents the findings of these analyses, including the results of studies into the importance of representation of river networks. The model has been successfully applied to understand the potential changes to discharge that would result from specific catchment-scale anthropogenic changes.