Effects of the groundwater conceptualization in PCR-GLOBWB 2 and Wflow. A test case for the Rhine-Meuse catchment

Abstract

The water cycle is represented as a closed-loop where several physical processes interact with each other. To simulate these processes we rely on models that often are not capable of solving the whole system and incur in simplifications. The most common is splitting the surface hydrology from the subsurface (often dealt by hydrogeology). In reality, knowing how both parts interact is frequently needed to represent for example river discharge. This makes it increasingly necessary to couple the surface and subsurface models as extreme conditions become more common. Recent developments, like the BMI specification, allow for tight model coupling through the use of a standardized procedure. To explore these capabilities this work explores the coupling of two distributed hydrological models (Wflow-SBM and PCR-GLOBWB) to a full-fledged groundwater model using Modflow 6 and its new python API. The focus is on evaluating the discharge prediction capabilities of each model, particularly low flows. We test this for the Rhine-Meuse catchment using two different coupling schemes. In one which we call offline, the hydrological model map flows to the groundwater model on each timestep. The second scheme, called online, works by allowing flow exchanges that work both ways, this means the groundwater model states are now capable of influencing on every time step the behavior of the hydrological model. The results are evaluated using a series of statistically and hydrologically relevant indicators. The development of the scripted coupled models was successful for PCR-GLOBWB and Wflow but conceptual deficiencies were found for the Wflow-online coupling scheme related to the lack of interflow generation. PCR-GLOBWB results show that online coupling allows for better discharge predictions and more homogeneous results than its non-coupled version and those of Wflow. The latter though shows outstanding performance in rain-dominated catchments with runtimes considerably lower than PCR-GLOBWB. It was also possible to prove that low discharge performance can be evaluated.