Three-dimensional Magnetotelluric Inversion For Resistivity And Magnetic Susceptibility Using Cross-gradient Constraints And Prior Information From Magnetic Data

The magnetotelluric method,which belongs to the geophysical electromagnetic methods and uses the natural sources,has been widely applied in various geophysical exploration fields.In recent decades,rapid improvements have been attained in threedimensional magnetotelluric forward modeling and inversion.However,most of the work are focusing on the main electrical parameter,namely the resistivity of the materials.The other electrical parameters,such as magnetic susceptibility,are ignored in three-dimensional magnetotelluric forward modeling and inversions.Instead,the magnetic susceptibility is considered to be zero.Sometimes,it is questionable to accept the assumption when dealing with strongly-magnetic materials,such as magnetite.By studying three-dimensional magnetotelluric forward modeling and inversion in relation to both resistivity and magnetic susceptibility,not only can the theory be supplemented and improved,but also the information about the magnetic susceptibility of the subsurface can be extracted effectively.To reduce the non-uniqueness and ambiguity of the geophysical inverse problems,we usually conduct joint interpretation of multiple geophysical data.Also,the magnetic method can be used to detect the abnormal of the magnetic susceptibility.So,by joint interpretation of the magnetotelluric and magnetic data,the resistivity and magnetic susceptibility of the subsurface structure could be properly recovered.The main goal of this thesis is to study the three-dimensional magnetotelluric forward modeling and inversion accounting for magnetic susceptibility.Three-dimensional magnetotelluric forward modeling was implemented based on the staggered-grid finite difference method.The modeling results were verified by comparison with that from calculations of Comsol Multiphysics software.For the three-dimensional magnetotelluric inverse problem,the objective functional included the cross-gradient constraints and would be optimized by the L-BFGS method.For the magnetic modeling,the analytic solution of the vertical cuboid was used.Through the magnetic inversion based on the L-BFGS-B method,the magnetic susceptibilities of the materials could be bounded.And the susceptibility model from magnetic inversions would be incorporated as the prior model into the three-dimensional magnetotelluric inversions,which serves as a new kind of constraints.The results from the synthetic inversion examples showed that the distribution of the resistivity model could be recovered by the resistivity-only-type magnetotelluric inversion even for the magnetotelluric data generated by a magnetic anomaly.But the information of the magnetic susceptibility was wasted.If the simultaneous inversion for resistivity and magnetic susceptibility from magnetotelluric data were conducted with the help of crossgradient constraints,the magnetic susceptibility could be recovered.It should be noted that spurious features could be seen for the magnetic susceptibility if the magnetotelluric data to be inverted was generated by conductive but non-magnetic anomalies.The synthetic test of the magnetic inversion showed,the L-BFGS-B method is so promising to be spread into the family of the magnetic inversion algorithms.When the inverted susceptibility model from the magnetic data was incorporated as the prior model into the three-dimensional magnetotelluric inversion,the spurious features mentioned above could be eliminated and the recovery of the resistivity structures were locally improved.