Modelling the Mixing Function to Constrain Coseismic Hydrochemical Effects: An Example from the French Pyrénées

Groundwater coseismic transient anomalies are evidenced and characterized by modelling the mixing function F characteristic of the groundwater dynamics in the Ogeu (western French Pyrénées) seismic context. Investigations of water-rock interactions at Ogeu indicate that these mineral waters from sedimentary environments result from the mixing of deep waters with evaporitic signature with surficial karstic waters. A 3-year hydrochemical monitoring of Ogeu springwater evidences that using arbitrary thresholds constituted by the mean ± 1 or 2σ, as often performed in such studies, is not a suitable approach to characterize transient anomalies. Instead, we have used a mixing function F calculated with chemical elements, which display a conservative behavior not controlled by the precipitation of a mineral phase. F is processed with seismic energy release (E_s) and effective rainfalls (R). Linear impulse responses of F to E_s and R have been calculated. Rapid responses (10 days) to rainwater inputs are evidenced, consisting in the recharge of the shallow karstic reservoir by fresh water. Complex impulse response of F to microseismic activity is also evidenced. It consists in a 2-phase hydrologic signal, with an inflow of saline water in the shallow reservoir with a response delay of 10 days, followed by an inflow of karstic water with a response delay of 70 days, the amount being higher than the saline inflow. Such a process probably results from changes in volumetric strain with subsequent microfracturation transient episodes allowing short inflow of deep salted water in the aquifer. This study demonstrates that groundwater systems in such environments are unstable systems that are highly sensitive to both rainfall inputs and microseismic activity. Impulse responses calculation of F to E_s is shown to be a powerful tool to identify transient anomalies. Similar processing is suggested to be potentially efficient to detect precursors of earthquakes when long time-series (5 years at least) are available in areas with high seismicity.