Hydrological Modelling of the 2021 Mega-Flood in the East of Belgium

Abstract

This study presents an event-based hydrological analysis of the two most severely impacted catchments in Belgium during the July 2021 floods. The Vesdre catchment experienced the highest rainfall volumes, but no gauging station recorded the entire flood wave. Therefore, this catchment was modelled using a new computationally efficient, gridded, runoff model developed in-house. This approach allowed hindcasting of the event and reconstruction of missing hydrographs. When coupled with 2D hydraulic simulations, the computed flood extent agrees well with field observations.In the Amblève catchment, the flood wave could be measured at several gauging stations. Therefore, additional hydrological modelling approaches were tested, including an improved version of the gridded model and two conceptual lumped models (GR4H and VHM). All models successfully reproduced the flood event, but their transferability in time proved challenging: models calibrated on past floods significantly underestimated the 2021 hydrograph peaks, while those calibrated on the 2021 event showed inconsistent performance when applied to other, less extreme, historical events.The heterogeneity of model parameters and hydrological responses highlighted the difficulties in transferring models between different catchments or sub-catchments. The accuracy of rainfall input data was also found to be a key factor influencing hydrograph quality, further emphasising the challenges of reliable flood simulation.