To improve the knowledge of the regionally important Continental Terminal 3 (CT3) aquifer in south-western Niger, fifteen magnetic resonance soundings (MRS) were carried out in December 2005 in the vicinity of wells and boreholes. The output MRS geophysical parameters, i.e. water content and decay constants versus depth, were compared to hydrogeological characteristics, i.e. water table depth, total porosity, specific yield and transmissivity estimated from direct measurements, pumping tests and transient groundwater modelling. The MRS-determined parameters were then used to estimate the rates of groundwater recharge. Contained in poorly consolidated Tertiary sandstones, the CT3 aquifer's water table has continuously risen by 4 m in total over the past four decades. Additionally, a significant portion of this increase has occurred in the past decade alone, with an annual rise now ranging between 0.1 and 0.3 m depending on the monitored well. Increase in groundwater recharge due to land clearance and deforestation explains this situation. According to previous estimations, the pre-clearing recharge ranged from 1 to 5 mm per year in 1950–60 s, while more recent recharge rates (1990s–2000s) range from 20 to 50 mm per year. These recharge values are directly affected by estimated aquifer specific yield value, while the spatial variation of rates of water table rise can be attributed to large scale hydrodynamic heterogeneities in the aquifer. However, few field measurements were available to confirm these assumptions. The main results of this study are: (1) The water table depth and aquifer transmissivity are estimated from MRS output parameters with an average accuracy of ± 10% and ± 9% respectively. (2) The MRS-determined water content is linked to both the total porosity and the specific yield of the aquifer, but no quantitative formulation can be proposed as yet. (3) Using the average MRS-determined water content over the investigated area, i.e. 13%, the groundwater recharge rates can be estimated to be ~ 2 mm per year in the 1950–1960s (pre-clearing period), and ~ 23 mm per year for the last decade. (4) The variations in specific yield and transmissivity cannot explain by themselves the spatial variability of the rise of the water table. (5) The ranges in transmissivity and water content obtained from MRS are more realistic than the groundwater modelling outputs. Therefore, MRS could be used to better constrain the aquifer parameters in groundwater modelling with a dense site network. Finally, this work illustrates how MRS can successfully improve characterisation and transient multi-year groundwater balance of commonly found sedimentary aquifers, particularly when integrated with well observations and pumping tests.