Coupling catchment runoff models to groundwater flow models in a multi-model ensemble approach for improved prediction of groundwater recharge, hydraulic heads and river discharge

Abstract

Groundwater recharge plays an important role in groundwater resource sustainability. Many approaches are available for quantifying recharge, including the use of numerical hydrological models. Various studies have assessed the efficacy of integrating recharge information from rainfall-runoff models into groundwater models. These studies, however, have been mostly restricted to a single hydrological model and were not able to evaluate the impact on simulating groundwater conditions caused by different recharge inputs. Furthermore, hydrological recharge models used in these studies are often fully distributed and physically based, resulted in high complexity in model structure, and, hence, posed equifinality challenges in calibration. On the other hand, these models may be more reliable for impact studies, which often involve extrapolation beyond the range of available calibration datasets. This paper proposes a modelling approach that replaces the groundwater reservoir in conceptual rainfall-runoff models by a more detailed and more physically based groundwater component. The approach enables assessment of the impact on groundwater conditions by different recharge models and to improve the model performance on simulating baseflow. The methodology starts with three well-calibrated lumped conceptual models, which were then disaggregated into spatially distributed codes to provide spatial–temporal information on groundwater recharge. The approach was tested for three conceptual models and applied to a catchment in Belgium. Model results were found to be consistent with respect to observations when evaluated for seasonal variations of groundwater heads and for cumulative groundwater discharge. However, evaluation of river hydrograph shapes and water-table variations revealed distinct results produced by the different recharge models.