Vers une meilleure compréhension des propriétés de résistivité électrique time-lapse associées aux contaminants organiques et aux processus de bioremédiation en sous-sol.

Management of contaminated sites represents a major problem for our society. Geophysical methods arise progressively as non-conventional techniques that should allow decreasing the uncertainty linked to the local nature of punctual drilling/sampling measurements classically used to identify, characterize and monitor such sites. However, the use of geophysics for this purpose is relatively recent and still requires an improvement of geophysical imaging and a better understanding of the impact of contaminants and remediation processes on measured properties to be fully effective. The main objective of this thesis was thus to improve our knowledge on these two aspects. In order to assess the reliability of geophysical imaging, and electrical resistivity tomography (ERT) in particular, we first compare quantitatively different image appraisal indicators. The latter are developed to detect artefacts, estimate depth of investigation, address parameters resolution and appraise ERT-derived geometry. Numerical benchmarks are created representing different geological situations in terms of heterogeneity and scale. On the basis of this comparison, we propose a methodology and guidelines to appraise both qualitatively and quantitatively field ERT images. We show the successful applications on real data coming from the contaminated sites we investigated. It notably allows us to exclude from our interpretation zones of the electrical images that are not considered as sufficiently reliable. To enhance electrical imaging, we investigate three different approaches to incorporate prior information into the ERT inverse problem, namely reference model, structural constraint and regularized geostatistical inversion that we notably apply on real data coming from two contaminated sites. The results are benchmarked against the standard smoothness constraint inversion. Results with real data show that adding prior information in the inversion process always lead to a modification of the solution at least in zones of low sensitivity (allowing notably to better image contaminant plumes at depth). However, the choice of the constraint to apply is highly dependent on the type and amount of information available. Therefore, we provide guidelines that should help the practitioner to include their prior information directly into the inversion process through an appropriate way. To understand the temporal geoelectrical signature of organic contaminants and bioremediation processes, we monitor a site contaminated with hydrocarbons and subjected to stimulated bioremediation. We first show that the most contaminated areas above the groundwater table level are associated to very low resistivities. We then show that during biostimulation (promoting aerobic degradation) and natural attenuation, observed resistivity variations (up to 140%) are mostly located in the saturated zone of the contaminated area. They follow a seasonal trend suggesting a temperature dependence not observed in an uncontaminated zone of the site. However, in the contaminated area, changes largely exceed the expected variations due only to the temperature. We therefore investigate systematically different hypotheses that may explain such changes. Among those hypotheses, we show that microbial activity is a factor that may potentially influence the electrical signature of a contaminated soil and may contribute to the observed resistivity changes. In order to further study the electrical response associated to bacterial activity during bioremediation processes, we monitor a tank experiment that contains soils contaminated with hydrocarbons subjected first to biostimulation and then to bioaugmentation (with an inoculum of Rhodococcus erythropolis T902.1). Whereas no particular electrical signature is observed during the biostimulation phase, we observe a correlation between the evolution of bulk resistivity changes and the specific oil-degrading flora after bioaugmentation that cannot be attributed to fluid resistivity changes. This suggests a direct impact of microbial growth/activity on electrical properties through the modification of surface and/or local electrolytic conduction mechanisms. These latter results open up new perspectives for future experiments that should involve spectral induced polarization measurements allowing a better discrimination between the two conduction mechanisms