Three-dimensional joint inversion of EarthScope magnetotelluric and magnetovariational data.

摘要

The magnetotelluric (MT) method provides a useful and economical way to study the deep geoelectrical structure of the earth. The traditional MT method analyzes impedance and apparent resistivity, both of which are based on the linear relationship between the electric and magnetic fields. However, MT data can be distorted by near-surface inhomogeneities, which can dramatically change the electric field and cause static shift over the entire frequency range of the apparent resistivity curves. It is shown that the magnetovariational (MV) method, which utilizes magnetic tipper data, is less affected by near-surface inhomogeneities. In this thesis, a three-dimensional (3D) joint inversion technique using both impedance and magnetic tipper data is developed and used in the analysis of synthetic models, and later is applied for study of the deep structure underlying Yellowstone National Park over a large MT survey area.;Nowadays implementation of 3D EM interpretation over large areas is still challenging due to the limitation of processing speed and the requirement of large matrix storage. In this thesis, large-scale 3D inversion based on the nonlinear conjugate gradient algorithm and the contraction integral equation forward modeling method is applied to EarthScope MT and MV data in Yellowstone. In order to improve efficiency and reduce memory requirement, the joint inversion algorithm is parallelized and the receiver footprint approach is used. The MT and MV joint inversion of EarthScope data recovered a conductive upper-mantle plume dipping westward and extending to a depth of more than 300 km. This plume is over 100 km deep at the center of Yellowstone National Park.