Three-dimensional Magnetotelluric Inversion With Adaptively Refined Inversion Grids And Topography

The Magnetotelluric method is one of the most important geophysical methods to study the electrical structure of the Earth by using time-varying natural electromagnetic field based on the electrical differences of rocks and ores.Which has been widely used in mineral resource explorations,volcanic and earthquake disaster researches,and deep earth structure and dynamic evolution studies,due to its characteristics of large detection depth,high sensitivity to good conductors and portable instruments.With the improving efficiency of data acquisition,as well as the implementation of multiple deep earth exploration projects,the amount of three-dimensional(3D)magnetotelluric field data is becoming larger and larger.Developing fast and high reliable 3D magnetotelluric data inversion with topography has become an urgent research problem and research hotspot.At present,the magnetotelluric field data acquisition area is gradually transitioning to the complex geological environment such as mountainous areas,which requires the 3D inversion software to have the ability to handle complex terrain.In order to meet the different requirements of the forward and inversion grids for grid density,the inversion algorithm and software are required to have the ability to decouple the forward and inversion grids and accurately map parameters between the forward and inversion grids.In addition,in a complex 3D geological environment,the number of unknowns of underground conductivity can easily far exceed the number of MT observation data,which makes the inversion a serious underdetermined problem and leads to the low reliability of 3D MT inversion results.According to this,the existing 3D magnetotelluric inversion software generally has three problems: difficulty in handling terrain,uncoupling of forward and inversion grids,and low reliability of inversion results.In order to solve these three problems,this study proposed several inversion mesh adaptive refinement strategies,and developed a new generation of magnetotelluric3 D adaptive inversion software that can finely process terrain effects and have higher reliability of inversion results in combination with unstructured grids.The main innovations and contents of this study are listed as follows:(1)An inversion mesh adjustment indicator based on the gradient of the misfit term of objective function(referred to as data fit gradient)is proposed.Besides,an adaptive inversion gird refinement strategy driven by the data fit gradient and the model parameter gradient is developed.Firstly,in the frame of classical MT inversion theory,the feasibility,advantages and disadvantages of using model parameter gradient,data fit gradient,model resolution matrix,and sensitivity matrix for error analysis of inversion results are analyzed.By employing and combining the above mesh refinement indicators,a total of seven adaptive inversion gird refinement strategies are tested with synthetic one-dimensional MT data.The results demonstrate that,the adaptive inversion gird refinement strategy driven by the data fit gradient and the model parameter gradient can obtain better inversion results with less number of inversion unknows.Tests with field data in Dachaidan,Qinghai further verify the performance of our adaptive refinement strategy.(2)A new 3D MT adaptive inversion algorithm for complex undulating topographies is developed.By adopting the adaptively refined inversion grids driven by the data fit gradient and the model parameter gradient,unstructured nested forward and inversion grids,Gauss-Newton and L-BFGS optimazation algorithms,a joint inversion algorithm for MT impedance and tipper data with complex topographies is presented.Firstly,the 3D MT forward modeling algorithm is verified using benchmark test models.Secondly,by considering the different requirements of forward and inversion meshes,a nested forward and inversion grid generation method is proposed.This nested inversion method can allow perfect mapping of inversion parameters between forward and inversion grids and can avoid the errors caused by imperfect mapping.Then,based on the verified adaptive inversion grid refinement strategy and Gauss-Newton and L-BFGS algorithms,a 3D MT adaptive inversion algorithm and software for both MT impedance and tipper data with complex topographies is developed.Finally,tests on synthetic models demonstrate that our algorithm can deal with complex topographies and geological environments.Besides,the adaptive inversion strategy can significantly improve the accuracy and convergence rate of inversion with less number of inversion unknowns,computation time and memory consumption.(3)The 3D subsurface conductivity structures of the Cascadia region of the United States and the North China are obtained.The Cascadia in United States and the North China are important regions for understanding the North American lithosphere and the destruction of North China Craton,respectively.The existing inversion strategies in the above two regions(using WSINV3 DMT and Mod EM software that widely used by international counterparts)havn’t considered the influence of topography,and the inversion of Cascadia’s field data havn’t used tipper vector data,which reduced the reliability of the inversion results.To solve this problem,the adaptive inversion algorithm developed in this study is used to obtain the corresponding new three-dimensional conductivity model by using the full tensor impedance and tipper vector data combined with the real topography of the corresponding area.The inversion results in this study are in good agreement with the existing results in the reflection of the main anomalies.In addition,because the adaptive inversion strategy requires fewer unknowns,the reliability of the inversion results is higher,and there are no false anomalies that may be caused by the anisotropy in the previous results and large scale high conductivity abnormal bodies.The inversion and interpretation of the field data fully verify the correctness and practicability of the adaptive inversion algorithm proposed in this study.