Strategies to improve the groundwater representation of hydrological models, the case of HYPE model

Although the importance of errors introduced by inappropriate numerical schemes is well recognized in the literature, several commonly used hydrological models continue to calculate concurrent fluxes sequentially. Here, two versions of the HYPE model (HYPE and H-HYPE) with different model structures were used in two case study basins to investigate the impacts of sequential calculations on groundwater representation. Implicit Euler (IE) solutions of the continuous state-space formulation for two HYPE versions (IE-HYPE and IE-H-HYPE) were developed to help provide more robust solutions. In contrast to IE-HYPE, results show that HYPE typically simulated no interflow when soil moisture levels exceeded the field capacity. This resulted in an overestimation of groundwater flux by the sequential solution schemes – as compared to literature values from the study basins. Monte-Carlo simulations of the parameter posterior distributions were used to determine the probabilistic distributions of the groundwater contributions to the streamflow. The groundwater contributions simulated by IE-HYPE and IE-H-HYPE (implicit models) show lesser overestimations as compared to the sequential schemes. The discrepancy between the groundwater simulations by the implicit hydrology-alone models and the corresponding literature values can be mainly explained by uncertainty sources other than the numerical errors. Integrating additional data – such as biogeochemical data – can help further reduce hydrological misrepresentations. As a next step, we developed hydro-biogeochemical models for both study cases and jointly inferred the hydrological and biogeochemical parameters. The joint hydro-biogeochemical models provided consistent improvements of the groundwater representations for both study cases – over the corresponding hydrology-alone models – only when the implicit models are used. Overall, the results suggest that a robust numerical scheme, such as the implicit Euler method, combined with a joint-hydro-biogeochemical parameter inference, provides the best groundwater representation for HYPE.