Résumé : [en] In-situ bioremediation is as a green and cheap process to clean soils from pollution compared to other techniques which often imply the excavation of soils. Amongst the bacteria used, Rhodococcus erythropolis appears as one of the best candidates for bioaugmentation. In fact, this species forms biofilms and produces biosurfactants to solubilize hydrocarbons, which are consequently more available for this bacterium and the endogenous oil-degrading flora. Moreover, its large genome allows the degradation of various persistent pollutants, such as polyaromatic hydrocarbons or sulfur-containing hydrocarbons. In addition to these benefits, our strain Rhodococcus erythropolis T902.1, isolated from a dried polluted soil, resists to desiccation during industrial process or drought, and maintains its biodegradation capabilities.To test this strain in field conditions, a bioaugmentation experiment at a pilot scale was initiated in partnership with the Department ArGEnCo, Applied Geophysics of the University of Liège. The pilot contains 2 m3 of sand, in which a vertical lens of highly polluted clayey soil (7200 mg of hydrocarbons/g of dry weight) was inserted. During the first three months, 75% of the hydrocarbons content was degraded, whereas a previous biostimulation experiment with KNO3 and H2O2 did not lead to any depletion of the pollutant. This degradation was correlated with the increase of total and specific microorganisms (by a factor 13 and 10 respectively) and the almost complete NO3- consumption (from 50 to nearly 0 mg/L). Furthermore, electrical resistivity tomography images of the contaminated lens also depicted a switch in the bulk conductivity values that does not correspond to the trend followed by the aqueous conductivity. It could be explained by the implementation of the injected bacteria and their production of hydrophobic biosurfactants desorbing hydrocarbons from soil particles. This assumption is strengthened by the fact that low concentrations of hydrocarbons were detected in piezometers downstream of the contaminated area.Further experiments will be carried out at a smaller scale to validate this hypothesis. On the one hand, we are currently designing a protocol to follow the biofilm formation by Rhodococcus erythropolis T902.1 with spectral induced polarization (SIP) signature in sand columns of 1.5 L. On the other hand, the analysis of biosurfactants will be performed in liquid cultures containing diesel oil, to characterize the hydrophobicity developed by the strain in presence of a common but complex pollutant.To conclude, all these characteristics showed by Rhodococcus erythropolis T902.1 make it an ideal candidate for the production of a bioremediation starter to quickly treat hydrocarbons-polluted soils. . Furthermore, the better comprehension of geophysical signatures associated with such a process may lead in the future to use them as a low-cost monitoring tool for a better visualization of active remediation zones.