Comparing large-scale 3D Gauss–Newton and BFGS CSEM inversions

Abstract

Electromagnetic signals are exponentially attenuated in conductive media. Thus, marine controlled-source electromagnetic (CSEM) data where the source and the receivers are located in the water column has exponentially low sensitivity towards the deep stratigraphy, compared to the shallow stratigraphy. In addition, CSEM inversions are also highly non-linear and ill-posed. It is therefore often difficult to achieve good inversion results for the deeper part of the subsurface using gradient based inversion methods. In this abstract, we describe a large-scale 3-dimensional anisotropic Gauss-Newton (3DGN) CSEM inversion implementation and discuss its advantages over gradient based algorithms. We also show, by synthetic and real data case studies, the large improvements in the 3DGN inversion results compared to those from the Broyden–Fletcher–Goldfarb–Shanno (BFGS) algorithm. Presentation Date: Tuesday, October 18, 2016 Start Time: 3:45:00 PM Location: 174 Presentation Type: ORAL