Cyclic variations of sulfate and boron concentrations and isotopes in deep groundwaters in the Aquitaine Basin, France

Concentrations and isotope contents of major and trace elements are principal factors in determining the origin of the chemical composition of groundwater. This paper focuses specifically on the use of sulfur and boron isotopes to characterize the origin of cyclic variations in the deep Eocene aquifer in the Aquitaine sedimentary basin (southwest of France). It is part of a multi-layer system mainly composed of sands and sandstone deposits. Groundwater contained in this deep reservoir is known to present stable chemical compositions, allowing its use for various purposes like drinking water, geothermal energy, thermal activity and agricultural irrigation. However, among the dozens of wells exploiting this aquifer and despite the reservoir's substantial depth, variations in sulfate concentration have been identified in a limited area of the reservoir. These fluctuations are cyclic and they seem to be correlated with water level variations due to gas storage activities nearby in the same aquifer. Regular water samplings and analyses of major and trace elements and their isotopes have identified that bore concentration variations are correlated with sulfate variations. A geochemical modelling approach based on water mixes elucidates the causes of these variations in chemical composition, especially the boron and sulfate concentrations and their respective δ¹¹B and δ³⁴SSO₄ values. From these numerical results, we identify that different sources explain the variations of boron concentration, one part coming from the silicates alteration (present in sub-layers of the exploited aquifer) and the other part coming from the evaporites alteration (present in the underlying molasse unit). These results also confirm the existence of mass and potentially water transfers between the different sub-layers of the reservoir and with the underlying molasse aquitard, implying new constraints for the future hydrogeologic modelling.