Towards semi-distributed modelling for low-flow simulation: comparing four hydrological models in the French Meuse catchment

Droughts are a growing concern for water managers, as climate change is expected to intensify both their severity and frequency. Accurately forecasting these events is crucial to mitigate their impacts. Semi-distributed hydrological modelling, by dividing the catchment into interconnected hydrological units, provides flow estimation at gauged and ungauged locations and explicitly accounts for some physical and climatic spatial variability across the catchment.In this study, four semi-distributed models (GRSD, MORDOR-TS, PRESAGES, RAMEAU) are implemented in the Meuse River catchment at Chooz, an area characterized by contrasting geological, topographic, and meteorological conditions. The models differ in their structural assumptions, notably regarding groundwater exchanges beyond topographic boundaries and the potential use of piezometric data during calibration.Following a joint calibration exercise, the four models provide consistent results on streamflow across the entire catchment and comparable performance at the outlet at Chooz in comparison to their lumped-model counterparts. Similar biases are observed among the models, which may reflect common limitations in their assumptions or uncertainties in flow measurements and meteorological data. The case study of the 2022 low-flow event highlights variability in simulated low flows, linked in particular to the choice of model and to the climate data used for calibration. Gauging measurements taken during low-flow periods would help strengthen these results. Future work should focus on improving the understanding and the representation of groundwater flows in semi-distributed hydrological models.