A Bayesian approach to electrical resistance tomography data inversion

Electrical resistance tomography can be an effective tool to delineate contaminated subsurface zones and to monitor subsurface remediation processes, provided accurate electrical resistivity images are achieved. Solutions from traditional regularization inverse methods are often unsatisfactory due to deleterious effects of the regularization term. In this work, datasets are obtained from multiple ERT configurations and assimilated in a Bayesian framework. The parameters are inverted by maximizing the a posteriori density. A modified total variation function is used to form the prior estimate. A new approach, successive partial variation relaxation, is developed to successfully minimize the deleterious effect of the subjective (prior) information.