Three-Dimensional Joint Inversion Of Gravity Data And Magnetotelluric Sounding Data

The non-uniqueness is the main problem that affects the reliability and resolution of geophysical inversions.Thus,how to reduce the non-uniqueness in the inversions is the key in all kinds of geophysical exploration.Gravity exploration and magnetotelluric sounding are both widely-used exploration methods in geophysics.The low cost,high efficiency,simple construction,and large detection depth make these two methods become important means in geological survey,deep exploration and also in mineral exploration and engineering environmental investigations.However,gravity data has less sensitivity to distinguish anomalies at different depths,which induces the poor resolution in the vertical direction;the electromagnetic signal of magnetotelluric is weak and noise is mixed into the low-frequency signal.All these can lead to serious non-uniqueness in three-dimensional inversions.The subsurface structure always has differences on physical properties,which means that the variation of different geophysical fields in the same area may be caused by the same subsurface structures.This sets up the the precondition for joint inversions of different geophysical data.Compared with single data inversion,a joint inversion can combine the advantages of different exploration methods,complement each other and increase the constraints of the objective functional.Thus,the conditional number of inversion can be reduced and the non-uniqueness will be improved.In actual field surveys,in order to reduce costs,improve work efficiency and reliability of data interpretation,multiple exploration methods are often carried out simultaneously.Therefore,the joint inversion,as an effective way to improve non-uniqueness,has secured data support and wide application prospect.In this thesis,based on the similar structure of subsurface anomalies and the works of the predecessors,I studied on the joint inversion of magnetotelluric sounding data and gravity data,and proposed a joint inversion algorithm based on Pearson correlation constraints of subregions.In this method,the subsurface area is divided into a series of overlapping three-dimensional subregions with fixed size,and I mearsure the linear correlation and variation of two physical attributes by calculating the Pearson correlation coefficient in each subregion.By setting the standard deviation threshold,I can constrain the larger fluctuating subregions in the joint inversion procedure,and enhance the linear correlations of subregions to reduce the non-uniqueness of inversion.In this thesis,I systematically discribled the construction of joint constraints based on subregion Pearson correlation coefficients,threshold setting of standard deviation and gradient calculation.The mechanism of the method is explained from the point view of product of standardized vetcors.The method does not need to compute the difference in three directions of models that can improve the stability of joint inversion and regional agreements through the threshold of standard deviation.To vertify the validity of the method,both single data inversions and joint data inversions based on cross-gradient and Pearson correlations of subregions are implemented on three designed theoretical models.The inverse results show that,compared with the single inversion,the depth resolution of gravity inversion can be significantly improved by using the method in this thesis,and compared with cross-gradient joint inversion,the regional agreements between the models from my inversion results also can be improved.Besides,the influence of different subregion sizes in joint inversion based on Pearson correlation of subregions are analyzed and discussed.In this thesis,the gravity forward modeling based on unstructured grids is also illustrated,and an improved wavelet compression technique is used to compress the gravity sensitivity matrix to significantly reduce the memory requirement in gravity forward modeling and inversions.In magnetotelluric sounding forward modeling,the finite difference algorithm base on staggered grids is used to calculate the electric field,and parallel computation is used to accelerate the forward modeling.The strategy of alternate iteration is adopted in my joint inversion procedure to avoid balancing different data weightings in the same objective functional and improve the iteration speed as magnitude of different data misfit may be not on the same order.In the single and joint inversions of magnetotelluric sounding and gravity,I use the conjugate gradient algorithm to optimize the objective functional.Parameters involved in the inversion processes are set by a unified approach,so that I can ensure the reliability and validity of the comparisons and analyses in this thesis.Finally,to verify the effectiveness and practicability of the joint inversion algorithm in this thesis,the single inversions,the cross-gradient joint inversion and the joint inversion algorithm are tested on the gravity field survey data and magnetotelluric sounding data acquired in Mount Isa,Queensland,Australia.The inversion results are compared with regional geological background and previous geophysical works.The results show that the method in this thesis has obvious improvement on the non-uniqueness problem and the resolution of inversion.