An investigation of anthropogenic influences on hydrologic connectivity using model stress tests

Abstract. Human influences threaten environmental flows directly or indirectly through groundwater abstraction. In alluvial geological settings, these may affect the contributions from groundwater-sustaining streamflow during dry summer months. The Dreisam River valley in southwest Germany represents a typical case where recurrent hydrological drought events between 2015 and 2022 have led to interruptions of longitudinal connectivity in the stream network. When and where vertical connectivity changes and where the streambed dries out have therefore become important questions. To help answer them, zero water level (ZWL) occurrences were previously measured at 20 locations in the river network during the drought of 2020, but they revealed high variability. This study therefore aimed to develop a methodology that allows the connectivity to be assessed along the entire stream network, i.e. by employing a numerical groundwater model to obtain the spatial distribution of the exchange flow between groundwater and surface water along the river. A reference model simulation for the period 2010–2022 assumed near-natural conditions. Stress test scenario model runs then imposed either an altered recharge regime or a set of introduced groundwater abstraction wells or both. To gain confidence in the model, ZWL patterns are compared to observations of dry riverbed locations in 2020, and the model generally reproduces the observed relative drying. Modelled exchange flows of the stress tests were then compared against the reference simulation. A set of specific metrics combining longitudinal and vertical connectivity is introduced for this task. The results of the stress test model runs show stronger changes of vertical connectivity in response to groundwater (GW) abstraction than to the imposed recharge stress. Reaches are identified where the effects of the stresses are particularly strong. Nevertheless, these results have to be interpreted within the limits of model realism and uncertainty. For more model realism, a number of improvements will be needed such as a higher-resolution parametrization of the riverbed's hydraulic conductivities and better coupling to contributions from hillslopes; for a quantification of the uncertainties, a systematic sensitivity analysis would be required. The study introduces a framework for modelling stress tests and metrics for surface water–groundwater interaction that can be transferred to other cases. It also suggests that even if not all influences can be modelled, the approach may help inform a resilient management of water resources under multiple stresses.