Performance of a Fast Approximate Solver in Identification of Electric Conductivity Changes from Time-lapse CSEM

Rigorous integral-equation (IE) forward modelling leads to a computationally intensive dense-matrix problem. Simplified IE (SIE) modelling is an approximate forward modelling method replacing the dense-matrix problem by a sparse-matrix problem. The accuracy and computational efficiency of SIE in a single forward solve has been demonstrated earlier. In an inversion setting, the forward model is used repeatedly and for several different purposes. In the present paper, we investigate the accuracy and computational performance of SIE in an inversion setting, by using it for forward and sensitivity calculations. It is found that SIE is sufficiently accurate while the computational gain wrt. IE is huge.