ERT Data Assimilation to Characterize Aquifer Hydraulic Conductivity Heterogeneity through a Heat-tracing Experiment

Geothermal Energy Systems such as heat pump relying on aquifers uses renewable sources of energy that are accessible in urban areas. It is necessary to characterize the subsurface hydraulic properties prior to the installation of such systems. In this context, heat tracing experiment is a typical field test that can help with characterization of the subsurface. During a heat tracing experiment, monitoring with downhole temperature sensors, water-level pressure transducers and electrical resistivity tomography (ERT) can be used to help to characterize the hydrogeological properties. Previous monitoring tools have shortcomings such as low-resolution data and over-smoothing, thus they fail to reproduce the heterogeneity of hydrogeological properties. Ensemble Kalman filter (EnKF) is a promising tool that can overcome the over-smoothing problem to replicate the hydrogeological property heterogeneity. In this work, we proposed a new procedure to assimilate time-lapse cross-borehole ERT data into a numerical model of groundwater flow and heat transfer, where the ground water is extracted, and heated water is reinjected into an unconfined sandy-gravel aquifer. The finite element model (FEFLOW 7.3) of groundwater flow and heat transfer is integrated with petrophysical relationship and electrical forward modeling (Resipy) to estimate cross-borehole ERT measurements. Then, the estimated apparent resistivity is assimilated to update the hydraulic conductivity model using EnKF. The results of the application of the proposed approach to a experimental site located in Quebec City (Canada) demonstrate that the heterogeneity of K is correctly reproduce since the updated K model is reasonably consistent with the lithological log. In addition, the proposed approach was able to replicate the cross-borehole ERT field and temperature measurements. The comparison between prior and posterior distribution of K with slug test results shows that the EnKF made the final (assimilated) distribution of K move toward K values inferred with slug tests.