Groundwater resources constitute a primary water source for most nations in the world. Its importance lies in supplying water for drinking, for agricultural and industrial activities, and for healthy ecosystems. However, sustainability of groundwater systems is difficult to maintain as sustainability does not rely only on the actual physical product of the aquifers (e.g., recharge), but there are other factors in play including climatic stresses and continuous anthropogenic use. Further, there is an intrinsic twin relationship between aquifer sustainability and water balance. If the underground system is at equilibrium, the amount of groundwater recharge and infiltration must equal the out flows including water consumption. Therefore, understanding the implications of climate change and discharging fluxes (e.g., pumping of water) is crucial in developing the mechanisms for good water resource management. Significantly, plenty of studies have investigated the impact of climatic variations on groundwater resources and a variety of hydrological models have been developed. Of these hydrological models, process-based models are of great interest because of their capability to educe and extrapolate beyond known cases when the direct measurements are not possible and to provide an easy analysis of multiple scenarios. Although the process-based models are the most common, they are computationally demanding; data-driven models have recently gained popularity in the scientific community due to their speed of performance and low cost.

Gaps in knowledge were addressed by implementing a systematic literature review which confirmed the potential negative impacts of climate change on groundwater recharge, regardless of the assumed emission scenarios and the length of projection. In addition, the Middle East and North Africa (MENA) and Latin America regions are among the most vulnerable to CC, as predictions showed a decline in the level of groundwater in conjunction with over-pumping, and less studies were observed. It was also found that the source of uncertainty, whether resulting from atmospheric system, hydrological models, or climate scenarios constitute a direct obstacle in simulating climate data and thus, in turn, leads to a lack of accuracy in hydrological models' prediction results. Finally, process-based models, in particular MODFLOW, are more commonly used than data-driven or conventional water balance models.

Thus, the objectives of this thesis are: (a) to model a groundwater aquifer system through constructing a physical process-based model; (b) to utilise the specified replenishment in safe yield re-estimation under current and proposed future climate conditions; (c) and to emulate the groundwater net recharge using machine learning models. Such objectives were achieved by: (1) undertaking a systematic literature review to identify gaps in the knowledge of groundwater modelling and to pinpoint the most affected areas around the world ; (2) conducting a preliminary data analysis to find possible trends and correlations between variables; (3) subsequently, choosing the best hydrological model to perform simulations; and (4) finally, predicting groundwater net recharge using the selected ML models. [...]