Two-dimensional Anisotropic Magnetotelluric Inversion And Applications In The Northern Tibet

Electric anisotropy could distort the magnetotelluric(MT)response and produce some special phenomena,such as phase rolling out of quadrant,the inconsistence of the analysis results of different dimensionality tools,the macro-anisotropy or unrealistic structures in the isotropic inversion results,and so on.To a certain extent,these special phenomena could be used as the footprints of the existence of anisotropy.However,it should be noted that,on the one hand,these foot prints only indicate a possible case for the presence of anisotropic structures,and they may also be induced by some special isotropic structures.On the other hand,due to the influence of complex underground structural environment and the degree of anisotropy,those footprints may not be observed even if anisotropic structures exist.Therefore,it has long been a difficulty in MT sounding to distinguish and interpret the data affected by anisotropy.Most anisotropic MT studies are based on qualitative analysis methods,such as dimensionality tools and forward modeling,to construct and interpret anisotropic structures,and it is very important to develop anisotropic inversion methods for reasonably and comprehensively describing the electric anisotropic structures underground.In addition,in the northern part of the Qinghai-Tibet Plateau studied in this paper,there is abundant subsurface anisotropy background.It is crucial to develop a reliable anisotropic inversion algorithm to enhance the interpretation of the measured magnetotelluric data and get a more comprehensive understanding of the structure under the northern Qinghai-Tibet Plateau.In this paper,the MT anisotropic inversion algorithm is realized and applied to the interpretation of MT data collected in the northern Tibet.(1)Two-dimensional MT forward algorithm considering arbitrary anisotropy.The forward algorithm is the foundation of the inversion.The finite difference method is used to realize the two-dimensional arbitrary anisotropic forward algorithm.The reliability of the forward algorithm is examined with three numerical examples compared with three othe forward methods.Compared with previous anisotropic MT researches,the effects of full anisotropic parameters on MT forward responses are firstly systematacially and comprehensively analyzed in this paper.A conclusion is highly drawn from those chacracteristics of effects: the vertical components of resistivity tensor weakly contribute to the MT impedance responses.This result could not only interpret the corresponding phenomena in previous studies,but also give a guidance to anisotropic inversion: the anisotropic parameters related to the components of resistivity tensor could be hardly identified by inversion.(2)Two-dimensional anisotropic MT inversion algorithm.The limited memory quasi-Newton method(L-BFGS-B)is introduced to realize the two-dimensional arbitrary anisotropic inversion algorithm.Based on the feature of the research problem in this paper,the standard L-BFGS-B method is optimized by using a more accurate form of Hessian approximation in which only the data misfit Hessian is approximated and the accurate part of regularization Hessian matrices are remained.Two methods of using the regularization factor during inversion are introduced,one is the simplest and practical fixed regularization parameters,the other is the relaxation regularization parameter approach.According to the numerical calculation experiences,the strategy of selecting regularization parameters method is put forward:in the anisotropic inversion,the fixed regular factor is first used for test calculation,and if the satisfactory results cannot be obtained,the relaxation regularization parameter approach is called.Both of these two methods need to carry out a large number of tests with different regularization parameter combinations.Generally,in theoretical model examples and isotropic cases,good results could be obtained by using the fixed regularization parameters method,while the more complex anisotropic inversion of real data requires the relaxation regularization factor method to get a satisfactory level of data misfit.Additionally,the features of L-BFGS-B method are used to expand the functionality of the inversion algorithm,which can handle a variety of simplified anisotropic scenarios,including isotropic one.The validity and stability of the inversion algorithm are verified by numerical examples.The analysis result of inversion examples is consistent with the forward.The azimuthal anisotropic MT inversion is the most stable pattern.Considering the complexity of anisotropic inversion,a strategy for inverting real data is proposed:(1)Analyze the anisotropic background in the survey area;(2)To search for the footprints that may be caused by the electrical anisotropy in the previous regular work;(3)Anisotropic inversion of the relevant profile data;(4)Referring to the main structure characteristics of the previous isotropic model,the robustness of local anisotropic anomalies was analyzed with different inversion parameters to finally determine the anisotropic anomalies.Thereinto,the anisotropic area imaged with less robustness would not be confirmed.The key of this strategy is that the imaged model,which could be consistent with the main structural features of previous isotropic model,contain the robust anisotropic anomaly with different inversion configurations.This conservative inversion strategy could futherest ensure the stability of inverted anisotropic anomaly,especially under the situation that the non-uniqueness of anisotropic inversion is very serious.(3)Applications of anisotropic inversion algorithm in the northern Tibetan Plateau.The anisotropic inversion is performed to three MT profiles(L15,L11 and L14)cross the transition zone between the east Kunlun orogenic belt and the Qaidam Basin.The low resistivity body at the lower crust and uppermost mantle show an anisotropic feature.The orientation of the horizontal lower principal resistivity is 25°north to west.From this position to the south,where the L11 crosses the Qiman Tagh range,a significant anisotropic anomaly is imaged in the middle and lower crust,and the orientation of the horizontal lower principal resistivity is almost parallel to the tectonic strike of Qiman Tagh range.No obvious anisotropic anomaly was found in the area covered by L14.(4)Tectonic cause and significance of the anisotropic structure in the surveying area.A novel interpretation that electric anisotropy could be used to indicate the orientation of shear action and analyze the deep shear stress environment.This is also what isotropic model could not be utilized to obtain.Combining with previous published work about the cause of the electric anisotropic structure,it is considered that the observed anisotropy anomalies are generated by the weak material flow at lower crust from Songpan-Ganzi terrain interacted with the shear movement.The deep partially molten rock shows a different connectivity along different direcitions under a shear action.High connectivity is generated parallel to the direction of shear action,showing low resistivity,while relatively low connectivity is generated perpendicular to the direction of shear action,producing high resistivity.And then,electrical anisotropy is formed.The shear motion orientation indicated by anisotropic anomaly is used to analyze the deep shear environment in the surveying area.The large scale strike-slip fault would affect the underground stress background in the surrounding area,and the influence of the left-lateral strike-slip Altyn Tagh fault could transmit to the lower crust and upper mantle at the intersection of the northern Qiman Tagh Mountains and profile L15.In the southern Qiman Tagh range,where the profile L11 intersects,the middle and lower crustal layers are almost no longer affected by the Altyn Tagh fault.Besides,the Kunlun left-lateral strike-slip fault basically does not affect the stress environment in this region.