Three‐dimensional magnetotelluric inversion for arbitrarily anisotropic earth using unstructured tetrahedral discretization

We propose a new three-dimensional anisotropic inversion scheme for magnetotelluric (MT) data. In this method, the earth is discretized into unstructured tetrahedral grids that can fit complex structures well, such as the earth topography and coastline. We use a 3 x 3 tensor to describe the anisotropic conductivity in the governing equation for MT inversions that can be simplified to three principal conductivities and three rotation angles. The edge-based finite-element (FE) method is adopted to do the forward and adjoint modeling to compute full impedance and sensitivity. To make the inversion flexible in the restoration of anisotropic and isotropic structures tendentiously, we add an additional regularization in the objective function, with its weight controlled by a distribution function. Before solving the inverse problem using L-BFGS optimization, we add upper-bound and lower-bound constraints to the parameters to suppress the solution space. Four synthetic experiments are designed to test the resolvability and resolution for the six anisotropic parameters in the inversion. The results show that our inversion method can well recover the resistivity in the principal x and y directions and the strike angle gamma. Finally, we apply our inversion algorithm to the USArray MT data with the aim to reveal anisotropic structure under Oregon State.