Rectilinear grids which are commonly used for the inversion of magnetotelluric (MT) data lack the flexibility required for representing arbitrary structures and for local refinement of the mesh. This abstract reports preliminary results of the inversion of MT data using unstructured tetrahedral grids. A minimum-structure procedure with an iterative Gauss-Newton algorithm for optimization is used. The sensitivity matrix-vector products that are required by the iterative solver are calculated using pseudo-forward problems to reduce the required computation memory. Forward problems are formulated using an edge-based finite-element technique and a sparse direct solver is used for the solutions. This solver allows saving and reusing the factorization of the matrices which greatly reduces the computation time. An example is presented which shows the capability of the algorithm to recover an anomalous region in a halfspace. The data that is inverted is the full-tensor impedance at the observation locations. Three frequencies are used for this example and computations are performed in parallel using MPI processes. The recovered model indicates the correct position of the synthetic model and reproduces the synthetic data very well.
Presentation Date: Thursday, October 20, 2016
Start Time: 10:10:00 AM
Location: 141
Presentation Type: ORAL
