Towards understanding the influence of seasons on low groundwater periods based on explainable machine learning

Abstract. Seasons are known to have a major influence on groundwater recharge and therefore groundwater levels, however, underlying relationships are complex and partly unknown. The goal of this study is to investigate the influence of the seasons on groundwater levels (GWL), especially on low-water periods. For this purpose, we train artificial neural networks and apply layer-wise relevance propagation to understand what relationships are learned by the models to simulate GWLs. We find that the learned relationships are plausible and thus consistent with our understanding of the major physical processes. Our results show that the models learn summer as the key season for periods of low GWL in fall, a connection to the preceding winter is usually only subordinate. Specifically, dry summers show strong influence on low-water periods and generate a water deficit, that (preceding) wet winters cannot compensate. Temperature is, thus an important proxy for evapotranspiration in summer and overall identified as the more important variable, but only on average. Single precipitation events show by far the highest influences on GWL and summer precipitation seems to mainly control the severeness of low GWL periods in fall, while higher summer temperatures do not systematically cause more severe low-water periods.