Research On Forward Modeling And Inversion Of Tensor CSAMT In Three-dimensional Axial Anisotropic Media

Geophysics exploration has developed from initial directly qualitative explanation to relatively quantitative analysis since its inception. Behind this, the increasing development of forward modeling and inversion technology play a very important role.First, for the forward modeling technology, it has progressed to three-dimensional (3D)finite difference from the integral equation method of the last century. And the application of adaptive finite element method starts to show a huge advantage of considering topography. About the inversion technology, although the problem of trapping into the local minimum exists, the lots of practical works prove that these methods can relatively meet the exploration demand well. Besides 3D inversion, the occurrence of four-dimensional exploration which has been adopted in parts of industry area opens a new chapter for geophysics.However, in practical application, there are still a lot of factors strongly influence electromagnetic observation data. The first factor is intense terrain and the second possible reason is serious anisotropy. Except these two problems, some areas even exist the serious influence of permeability?polarization and some other parameters. The current electromagnetic exploration is trying its best to study them so that we can obtain a more precise and reliable interpretation.The focus of this paper is to bring resistivity anisotropy parameter into Controlled-Source Audio-Frequency Magnetotelluric Methods(CSAMT). In the strict sense,anisotropy parameter is arbitrary anisotropy which the resistivity of arbitrary direction is different. Although this way better reflects practical conditions, it cannot be inverted in the current inversion study field. So in this paper,we simplify arbitrary anisotropy to axial anisotropy which to some extent reduces the inversion reliability. But this way also reduces the multiplicity of inversion solution. From this view, our methods can obtain a reliable optimization results in the current system of theories. Under this research idea, this paper adopts 3D stagger-grid finite difference as forward operator.Finite difference is easy to be realized and is highly efficient which is widely used in current geophysical field.After we realizing CSAMT 3D axial anisotropy forward, we do a lot of forward modeling tests research which shows that adopting tensor source as transmitter and its multi-component data can distinguish anisotropy as soon as possible. Furtherly, this paper adopts Limited-memory Broyden-Fletcher-Goldfarb-Shanno (LBFGS) and non-linear conjugate gradient method(NLCG) as inversion method to study. We firstly completed tensor CSAMT 3D isotropic inversion research. The research results show that neglecting source effect and directly using magnetotelluric method to invert will make the deep inversion results unreliable. Meanwhile the inversion results of tensor CSAMT also show a higher resolution than those of scalar CSAMT. Besides,this paper also researches the comparison of effect and efficiency between LBFGS and NLCG. In the last, the theoretical tests show that neglecting anisotropy and directly do isotropy inversion will bring serious fault which will seriously influence final interpretation.