Data-driven selection of the minimum-gradient support parameter in time-lapse focused electric imaging

Abstract

We have considered the problem of the choice of the minimum-gradient support (MGS) parameter in focused inversion for time-lapse (TL) electric resistivity tomography. Most existing approaches have relied either on an arbitrary choice of this parameter or one based on the prior information, such as the expected contrast in the TL image. We have decided to select the MGS parameter using a line search based on the value of the TL data root-mean-square misfit at the first iteration of the nonlinear inversion procedure. The latter was based on a Gauss-Newton scheme minimizing a regularized objective function in which the regularization functional was defined by the MGS functional. The regularization parameter was optimized to achieve a certain target level, following the Occam principles. We have validated our approach on a synthetic benchmark using a complex and heterogeneous model and determined its effectiveness on electric tomography TL data collected during a salt tracer experiment in fractured limestone. Our results have determined that the approach was successful in retrieving the focused anomaly and did not rely on prior information.