Distributed numerical models, considered as optimal tools for groundwater resources management, have always been constrained by availability of spatio-temporal input data. This problem is particularly distinct in arid and semi-arid developing countries, characterized by large spatio-temporal variability of water fluxes but scarce ground-based monitoring networks. That problem can be mitigated by remote sensing (RS) methods, which nowadays are applicable for modelling not only surface-water but also groundwater resources, through rapidly increasing applications of integrated hydrological models (IHMs). This study shows implementation of various RS products in the IHM of the Central Kalahari Basin (~200 Mm2) multi-layered aquifer system, characterized by semi-arid climate and thick unsaturated zone, both enhancing evapotranspiration. The MODFLOW-NWT model with UZF1 package, accounting for variably saturated flow, was set up and calibrated in transient conditions throughout 13.5 years using borehole hydraulic heads as state variables and RS-based daily rainfall and potential evapotranspiration as driving forces. Other RS input data included: digital-elevation-model, land-use/land-cover and soils datasets. The model characterized spatio-temporal water flux dynamics, providing 13-year (2002–2014) daily and annual water balances, thereby evaluating groundwater-resource dynamics and replenishment. The balances showed the dominant role of evapotranspiration in restricting gross recharge to only a few mm yr−1 and typically negative net recharge (median, −1.5 mm yr−1), varying from −3.6 (2013) to +3.0 (2006) mm yr−1 (rainfall of 287 and 664 mm yr−1 respectively) and implying systematic water-table decline. The rainfall, surface morphology, unsaturated zone thickness and vegetation type/density were primary determinants of the spatio-temporal net recharge distribution.