A Geologically Constrained Basis for Global Inversion of CSEM Data Using a Dynamic Number of Parameters

In this paper, we introduce a new parameterization of the model space, where the basis is constrained on a priori information about the geology. The parameterization is able to represent complex model structures using only a few parameters, which can significantly reduce the computational complexity of the inversion problem. This facilitates a global inversion approach, and we consider a simulated annealing optimization. In order to facilitate the search for a minimum-parameter representation, we extended the inversion to be able to optimize for a dynamically varying number of variables. We demonstrate the method by inversion of marine CSEM data from the Troll West Oil Province. The algorithm is able to recover a resistivity profile which agrees with well log data from the area. The dimensionality of the parameter space is reduced by more than an order of magnitude using our approach compared to a layer-based discretization of the resistivity. The physically feasible models obtained are attributed to the constrained basis which makes the inversion very robust.