3D forward modeling of controlled-source electromagnetic data based on the radiation boundary method

Abstract

I have developed an efficient 3D forward modeling algorithm based on radiation boundary conditions for controlled-source electromagnetic data. The proposed algorithm derives computational efficiency from a stretch-free discretization, air-free computational domain, and a better initial guess for an iterative solver. A technique for estimation of optimum grid stretching for multifrequency modeling of electromagnetic (EM) data is developed. This technique is similar to the L-curve method used for the estimation of the trade-off parameter in inversion. Using wavenumber-domain analysis, it is illustrated that, as one moves away from the source, the EM field varies smoothly even in the case of a complex model. A two-step modeling algorithm based on radiation boundary conditions is developed by exploiting the smoothness of the EM field. The first step involves a coarse-grid finite-difference modeling and computation of a radiation boundary field vector. In the second step, a relatively fine grid modeling is performed with radiation boundary conditions. The fine-grid discretization does not include the stretched grid and air medium. An initial solution derived from coarse-grid modeling is used for fine-grid modeling. Numerical experiments demonstrate that the developed algorithm is one order faster than the finite-difference modeling algorithm in most of the cases presented.