Revisiting hydrological modelling strategies for low flow simulations in the French Moselle catchment in response to recent hydrological droughts

In recent years, the French part of the Moselle catchment (10 770 km²) has experienced successive severe hydrological droughts. These droughts are characterised by an unprecedented pattern of slow and prolonged periods of recession, reflecting a shift from the typical drought behaviour observed in the past. This difference from past droughts presents a challenge and requires a new approach to adapt hydrological modelling methods for low flow simulations in the context of climate change. A seven free parameters model originally developed for low flow simulation called PRESAGES is considered in this study. Three reservoirs (a production reservoir, a delayed discharge reservoir and a groundwater reservoir) are implemented to manage both internal and external flows and model states. We observed that the current version produces poor simulations in our study area on recent observed discharge values particularly during long recession phases over the period 2010 to 2020. This period has indeed been characterised by severe low flow events with frequent and persistent recessions. In order to improve the efficiency of our model, we propose a non-linear depletion approach, specifically the Wittenberg model, instead of the conventional linear depletion law used in the current version of the model. The reason for this modification is that the linear depletion law oversimplifies the physical processes governing the depletion of groundwater reservoirs by ignoring inherent non-linearities. As a consequence, flow simulations no longer represent the reality of observed events. Another aspect addressed to enhance the model performance is the search of an objective function that, in conjunction with the new model structure, takes account a severity increasing over time. To this end, different mathematical transformations of the single objective function based on the Kling-Gupta efficiency criterion (KGE) were compared to calibrate the PRESAGES model. Benefits from this new modelling strategy was assessed using numerical criteria and visual inspection of hydrographs. The results indicate that employing the power of -0.5 as the objective function transformation to calibrate the model coupled with the non-linear depletion law for the groundwater reservoir makes the model more competitive. Model’s ability to simulate and capture prolonged and slow phases of recession was particularly improved. These results demonstrate the necessity to cope with the progressive aridification of the study area in modelling strategies of water resources.