We developed a fully finite-element (FE) based minimum structure inversion for magnetotelluric (MT) data in three dimensions. Both data misfit and regularization functionals are treated using the relevant FE approximations on unstructured meshes. The inversion algorithm is coded based on the Gauss-Newton (GN) iterative optimization scheme with model perturbation being sought in each solution step. The synthetic MT modelling subset of the inversion here is implemented using an E-field edge-based FE method. The coefficient matrix of the discretized forward solver is directly factorized and stored on multiple parallelized processors that an efficient sensitivity matrix-vector product was performed within GN iterative steps. We developed a novel FE approximation of the regularization functional that quantifies the conductivity gradient across irregularly-shaped neighboring cells. Preliminary results of the method and the developed algorithm is demonstrated for a test model from literature.
Presentation Date: Tuesday, October 13, 2020
Session Start Time: 8:30 AM
Presentation Time: 9:45 AM
Presentation Type: Oral