Limited understanding of aquifers’ responses to global warming and human activities challenges scientists and water professionals. This study, an early attempt to simulate natural groundwater recharge in the Chiba watershed, assesses the impact of human activities and climate change using Google Earth Engine and machine learning. With a Kappa coefficient of about 90%, the study produced reliable results. From 1985 to 2021, the SWAT model effectively replicated hydrological dynamics. Findings show that a 12% increase in agricultural land and a 2% decrease in precipitation result in a 16% rise in evapotranspiration and a 33% decline in natural recharge. Hydrological processes are sensitive to precipitation and land use changes. Spatial distribution of annual recharge indicates low groundwater recharge with upstream-downstream variance. Landsat images and machine learning enhance land use/land cover classification in Tunisia’s semiarid context. This research calls for deeper investigations into groundwater levels for comprehensive groundwater resource management and sustainability.