Abstract Using photovoltaic energy to pump water from aquifers is an interesting solution to circumvent the low electricity grid coverage and provide improved domestic water access in off-grid areas in sub-Saharan Africa. When pumping and during the lifetime of a pumping installation, the borehole water level changes, which impacts the amount of energy required to extract water from the aquifer. In order to address alterations in energy requirements, this article develops a data-driven borehole water level model adapted to photovoltaic water pumping systems (PVWPS). The proposed model is applicable to all types of PVWPS and aquifers. It has been validated against experimental data from a pilot PVWPS located in a rural off-grid village in Burkina Faso having achieved more than 97% accuracy. Thanks to this borehole model, we have been able to assess the influence of the variability of groundwater resources on both the performance of PVWPS and on their optimal sizing. We show that the variation of the static water level can require a increase of the peak power of the PV modules of up to 100%. Nonetheless, the effect of the drawdown due to the pumping is negligible. This study can help companies, governments and non-governmental organizations to better take into account the variability and the sustainability of groundwater resources in the optimal sizing and monitoring of PVWPS.