Inversion of magnetometer array data by thin-sheet modelling

Abstract

A conjugate gradient relaxation method is employed in seeking an iterative solution to a particular 3-D electromagnetic inverse problem. The problem, which is non-linear, is that of modelling observed magnetic fluctuation response functions by a thin-sheet earth structure. The method optimizes data misfit and model roughness, and starting from a base model designed on known geology produces a series of improved models. Established thin-sheet algorithms are employed to compute the model responses and sensitivity, and utilization of a parallel computing system enables reasonable execution times. The method is demonstrated by inverting data generated for a synthetic example. Gaussian noise is added to the synthetic data. The inversion scheme has been applied to geomagnetic induction data which have been recorded widely on the Australian continent. Highly conductive structures in models given by the inversion method confirm regional conductivity anomalies indicated by induction arrow patterns, and give more quantitative information about them. The method appears to work well in areas of strong conductivity contrast such as coastlines.