Is the groundwater reservoir linear? Learning from data in hydrological modelling

Although catchment behaviour during recession periods is better identifiable than in other periods, the representation of hydrograph recession is often weak in hydrological simulations. Among, the various aspects that influence model performance during low flows, in this paper we concentrate on those more inherently related to the modelling, such as the development of a suitable model conceptualization, and the choice of an appropriate calibration strategy. In this context we develop a methodology where the calibration procedure is combined with an iterative process of model improvement, to obtain an optimal model configuration that performs well both during low flows and high flows. The methodology starts by calculating a synthetic master recession curve that represents the long-term recession of a given catchment. Subsequently, using a simple reservoir model, we determine the storage-discharge relation that simulates the slow hydrograph component. This relation is determined without making any a-priori assumption on its form and is inferred from discharge data available through an iterative process. Next, high flow related parameters are recalibrated separately, to avoid that the simulation of low discharges is neglected in favour of a higher performance in simulating peak discharges. This methodology is applied on several catchments in Luxembourg, and as a result we determined that in all catchments except one (where human interference is high) within the chosen model structure a linear reservoir describes best the observed groundwater behaviour. This result is used to trigger a discussion as to the general suitability of the of a linear groundwater reservoir in hydrological modelling.