Résumé
Electrical Resistivity Tomography (ERT) is increasingly used to study subsurface hydrological processes; it shows promising potential for estimating soil water content, a key but challenging property to quantify. However, the complexity of converting the resistivity signal to water content led us to develop approaches to increase the estimate robustness while facilitating uncertainty evaluation. In this paper, we propose an innovative method, called the Ensemble Approach ERT (EA-ERT) that builds an ensemble model of electrical resistivity calibrated from field data and then converts it to a spatial distribution of water content. This approach combines time-lapse ERT data with point-based in-situ soil water content measurements. It makes it possible to (i) circumvent inversion parameter choice by evaluating the performance of multiple models, (ii) estimate uncertainty in the final model by calculating the coefficient of variation among the models that make up the ensemble, and (iii) convert electrical resistivity models to water content. The method was tested at two dissimilar field sites in southern France and on a synthetic case. For each site, an ensemble model, built from multiple inversions, was selected and converted to soil water content. The calculated values showed a strong fit (r ≥ 0.8), with minor differences (RMSE ≤ 3.24% vol.) compared to in-situ measurements. Areas of high uncertainty were identified, providing information that complements more traditional indicators provided by the inversion code. EA-ERT provides a robust and automatable method to convert ERT data to related parameters, contributing to improved monitoring and more complete understanding of subsurface processes.