The integration of shared earth modeling and robust 3D CSEM modeling and inversion is the key to deriving a reliable quantitative interpretation from marine controlledsource electromagnetic (CSEM) data. Workflows should make use of all available subsurface data and enable the interpreter to select the most geologically relevant resistivity model from the multitude of models that satisfy the same CSEM data. To this end, we present our implementation of an iterative migration method for CSEM data, equivalent to rigorous inversion. Our iterative migration method is based on the 3D integral equation method with inhomogeneous background conductivity and focusing regularization with a priori terms. Here, we will show that focusing stabilizers recover more geologically realistic models with sharper geoelectric contrasts and boundaries than traditional smooth stabilizers. Additionally, we will show that focusing stabilizers have better convergence properties than smooth stabilizers. Our method is implemented in a fully parallelized code, which makes it practical to run large-scale 3D iterative migration on multi-component, multi-frequency and multi-line CSEM surveys for models with millions of cells. We present a suite of interpretations obtained from different migration scenarios for a 3D CSEM feasibility study computed from a detailed model of the Shtokman gas field in the Russian sector of the Barents Sea.
The premise for the various marine CSEM methods is sensitivity to the lateral extents and thicknesses of resistive bodies embedded in conductive hosts. For this reason, CSEM methods have been applied to de-risking exploration and appraisal with direct hydrocarbon indication (Hesthammer et al. 2010). Methods for interpreting CSEM data are complicated by the very small responses of hydrocarbon-bearing reservoir units when compared to the total fields. Quantitative interpretation of CSEM data is inherently reliant on iterative inversion methods since the data cannot simply be separated or transformed with linear operators as per seismic methods. Best practice is to run multiple 3D inversion scenarios in order to enable interpreters to vary their inversion parameters so as to explore alternative resistivity models that satisfy the data, and select the most geologically plausible ones for subsequent interpretation. This practice identifies any artifacts that may arise from interpreting a single resistivity model. Alternative models may also be used to reveal what additional data, if any, are needed to further constrain the interpretation. Generation of these alternative models requires rigorous but fast 3D inversion methods. Rigorous inversion methods are not the most practical, as the sensitivity matrix needs to be constructed and stored at each iteration for the many transmitter-receiver combinations in a CSEM survey. In terms of 3D CSEM inversion, geological prejudice is introduced via regularization; whether that is an a priori model, data or model weights, model bounds and/or by the choice of stabilizing functional. Resistivity models are often obtained from regularization with smooth stabilizing functional; the first or second derivatives of the resistivity distribution are minimized, resulting in smooth distributions of the resistivity. This type of smooth solution allegedly satisfies Occam’s razor since it is claimed to produce the most “simplistic” model for the data.
