A model ensemble-elasticity-stress test for drought impact on spring discharge and low flow

Increasingly severe drought events demand a basis for prioritizing water uses and protection goals. In addition to classical methods of hydrological design, model experiments can support such aims. They can help investigate the response of hydrological systems to scenarios of exacerbated drought forcing. For known events of the past they might ask: "how much more would water resources be depleted, if forcing conditions had been more severe?". In such a stress-testing approach, this study investigated the sensitivity of low flows and low spring discharges to a range of exacerbated antecedent forcing conditions for a multi-catchment and multi-spring dataset in southwest Germany. Different model realizations for the forcing groundwater recharge, a baseflow separation and various conceptual groundwater storage and release models were combined into a model ensemble and system responses were analyzed in terms of elasticity metrics. Stress was applied as a systematic reduction in groundwater recharge with different magnitudes over different time periods preceding the main event of drought impact. All scenarios caused further reductions in low flow and spring discharge compared to the reference simulations. The presentation elaborates systematic thresholds: for example, the low-flow response of some catchments becomes maximal after a few months, and in others only after two years of stress duration. The experiments illustrate the sensitivities within the study area and allow to expand the derived 'story' as: "in a hydrological systems with (certain, e.g.) hydrogeological characteristics, low flows as in a (certain) memorable summer might be further depleted up to a (certain) maximum additional amount under (certain) drier preceding conditions". However, the importance of a specifically adapted model architecture as well as the estimation of model-related uncertainties becomes apparent from the ensemble experiment. Applied for selected well-validated model structures, the approach can help elucidate and communicate potential limits of drought stress.