Improving recharge-controlled groundwater level behavior in a transient data-driven 3D model

Data-driven models are powerful tools for analyzing the evolution of groundwater flow and thermal field in response to hydrometeorological forcing. However, they usually come with uncertainties in flux boundary conditions and in the distribution of rock properties. To overcome this, we coupled a subsurface 3D model of Brandenburg (NE Germany) with the distributed hydrologic model mHM to simulate a 60-year-long monthly time series of regional groundwater dynamics. Recharge fluxes, derived from mHM and assigned to the top of the saturated subsurface model, allowed us to reproduce magnitudes of seasonal groundwater level fluctuations as observed in shallow monitoring wells (0-5 m). However, approximating the multi-annual periodicity that is pronounced in deeper wells (10-30 m) and the long-term decline in groundwater levels recorded in parts of Brandenburg has proven to be more challenging. This highlights the need to consider damping the infiltration signal in order to better approximate the delayed response of the subsurface to the imposed precipitation pulses, as well as additional sinks contributing to the loss of groundwater storage. To this purpose, we analyzed the frequency of groundwater level fluctuations in >100 observation wells as a function of the unsaturated zone thickness and compared them against the results obtained from a 1D analytical model solution. The established relationship of recharge damping with depth was then utilized to correct the flux boundary conditions. This, along with optimization of river network density and aquifer storativity, resulted in an improved match in modeled versus monitored hydraulic heads. This enables further use of the coupled groundwater and surface-water model for ongoing forecasting studies of the thermo-hydraulic evolution of the aquifer system with respect to climate scenarios.