This work is part of the AUVERWATCH project (AUVERgne WATer CHemistry), which aims to better characterise some Auvergne water bodies, specifically the alluvial hydrosystem of Allier River (France). Alluvial aquifers constitute worldwide a productive water resource, superficial and easily exploitable. In France, 45% of the groundwater use comes from these aquifers. The study site is a wellfield that withdraws 8.5 million m3 of water annually from an alluvial aquifer to produce domestic water for 80% of the local population.

Coastal groundwater is a vulnerable resource, estimated to sustain the water needs of about 40% of the world’s population. The Roussillon aquifer is a regional aquifer near Perpignan (southern France). It covers over 800 km2 of land and is used for irrigation, drinking water, and industrial purposes. The aquifer has experienced significant piezometric lowering in the last decades, weakening the regional resource.

Groundwater modelling at the mine sites involves assumptions from the geological model, mining stages, parametrization, and fractures, among others. Modelling work mainly focuses on calibrating against historical measurements before operations (pre-mining) or afterwards (transient calibration). Calibration is carried out mainly with gradient-based algorithms. However, the majorlimitation is the number of model runs, since the number of parameters can easily reach hundreds or more. PEST has become the common tool for parameter estimation.

The joint application of water supply system security, groundwater modelling, and multicriteria analysis (MCA) indicated the potential of Managed Aquifer Recharge (MAR) to increase water supply security in Eastern Botswana substantially. Botswana faces increased water stress due to decreased water availability as climate change exacerbates variability in rainfall and increases evaporation losses and water demand.

Groundwater systems are complex and subject to climate change, abstraction, and land use stresses, making quantifying their impacts on aquifers difficult. Groundwater models aim to balance abstraction and aquifer sustainability by simulating the responses of an aquifer to hydrological stresses through groundwater levels. However, these models require extensive spatial data on geological and hydrological properties, which can be challenging to obtain. To address this issue, data-driven machine learning models are used to predict and optimize groundwater levels using available data.