A new workflow to incorporate prior information in minimum gradient support (MGS) inversion of electrical resistivity and induced polarization data

Abstract The current paradigm for geophysical inversions is to select the simplest solution according to Occam's principle. The implicit assumption usually made is that the parameters of interest have a smooth spatial distribution, which is rarely geologically plausible. An alternative is the Minimum Gradient Support (MGS), a functional that allows to compute a regularized inversion favoring sharp contrasts. However, solutions are highly sensitive to the selection of a variable called the focusing parameter β and the method is not very performant when many structures are present in the subsurface. Thus, we propose a new workflow to apply this functional to real case studies where heterogeneous structures are expected: a smooth solution is first computed and used as a starting model for sharp MGS inversions. Sequentially incorporating additional prior information on resistivity (e.g., from drillings, previous geophysical surveys) is possible and further improves imaging for resistivity and chargeability structures. The new methodology is first tested on a synthetic case and then applied to ERT/IP data collected on a gold deposit. The methodology enables to compute plausible electrical resistivity spatial distributions in accordance with the vast prior geological knowledge, and reveals new insights about the mineralization key characteristics. The choice of β remains challenging and should be automated in future developments.