Machine learning insights into groundwater demand under changing surface water availability: Murray-Darling Basin, Australia

In the Murray–Darling Basin (MDB) of Australia, climate change is leading to shifts in the availability of surface water which is driving an increased reliance on groundwater resources. Projected declines in rainfall and rising climate variability are expected to amplify this trend, heightening the role of groundwater in supplementing water demand. Quantifying this change is important for ensuring water resource resilience and sustainability into the future. This study explores hydroclimatic conditions associated with periods of elevated groundwater use and evaluates how future reductions in surface water reliability may influence extraction patterns. Relationships between surface water and groundwater dependence are analysed and groundwater requirements under a range of hypothetical future scenarios are simulated. The deep learning-based stress-testing framework used here accounts for simultaneous changes in important surface water components amid the inherent uncertainty of future conditions. Results show groundwater demand could increase by up to 16 % under plausible future reductions in rainfall and surface water storage, compared with modelled predictions based on 2010–2020 data. The study demonstrates the utility of machine learning for scenario testing under uncertainty and at multiple-aquifer scale. Findings emphasize the interconnected nature of surface and groundwater systems in the MDB and highlight the importance of conjunctive water management strategies to ensure long-term water security under changing climate conditions. • Machine learning reveals groundwater use sensitivity to hydroclimatic change. • Low rainfall and storage are linked to increased groundwater use in the MDB. • Groundwater demand response to shifting surface water varies across the region. • Findings support conjunctive water management under future climate variability.