Three dimensional magnetotelluric inversion using heterogeneous smoothness-constraint least-squares method

Abstract

In the presentation, we show a rapid algorithm for the electromagnetic inversion that allows to bypass updating of the sensitivity matrix. In the procedure, a sensitivity matrix for a homogeneous half-space is used in all iterations. Instead of the fixing sensitivity matrix, the Laplace filter coefficients in each element are updated based on the resistivity smoothness calculated by reconstructed structure derived from the former iteration. This substitution is not only expected to reduce the computing time required for the large-scale inversion such as for the three dimensional surveys, but also expected to allow the reconstruction for sharp boundary structure. We applied the algorithm for the three dimensional magnetotelluric inversion to confirm the effectiveness. In the presentation, we show the calculation results using several synthetic examples with different structures and conditions.