Three-dimensional Electric Structure Of Crust In Qiangtang Terrane

Qiangtang Terrane (QT) locates in the northern part of Tibetan Plateau (TP) and signifies both scientifically and economically. With the development of research on TP, more and more geophysical evidences are needed to verify the geological inference and observation. Being a key tool to understand the deep interior of the Earth, magnetotelluric (MT) method has been frequently used to study the electric structure of TP. Three-dimensional MT is the international frontier of geophysics, because 3D MT can avoid hypothesis of 1D or 2D geology and present actual distribution of physical parameters of the Earth. This research, in which we interpret MT data of TP by 3D inversion, is in the world the first case of 3D investigation of electric structure in TP and also the first publication of 3D field data inversion in China.The major contents include:Chapter 1: brief introduction to geology in both TP and QT; significance of research; and development of electric structure study in TP and QT.Chapter 2: introduction to data acquisition and processing of raw data. Details include survey configuration, topography, layout of survey line, acquisition parameters (frequency points), and calculation of impedance tensor.Chapter 3: calculation of apparent resistivity and phase, data analysis, static effect, method used to eliminate correlated noise in data set, and qualitative analysis of processed pseudo-section. To get rid of static shift, we artificially recognize the source of static effect by geological map and satellite image, and then set the filtering parameters for the elimination of correlated noise. 2D filtering imposed on pseudo-section of complex impedance tensor is able to suppress correlated noise existing in our data set, so the inversion program will not be prevented from convergence due to non-Gaussian noise.Chapter 4: major theory of MT. We emphasize the application and algorithm of 3D MT, for example, the forward modeling and inversion. We specifically introduce the Occam's algorithm of 3D MT inversion in data space, which maps inverse problem from traditional model space to data space. Therefore, the total computational load of RAM and CPU time can be reduced greatly, which makes this method more practical. Besides, we describe the job of 3D MT inversion and the determination of inversion parameters. To be robust to local minimum in 3D optimization, we propose an inversion strategy called "MT multiple-coverage" that covers the same parameters underground for many times. This method is a copy of multiple-coverage in seismic exploration. On operation level, we subdivide the whole data set to odd frequency data and even frequency data, which use different initial models to generate different convergence paths so as to explore the model space as widely as possible. In the stage of interpretation, the results from multiple-coverage provide a judgement about the reliability of inverted parameters. We also solve the problem on how to determine the desired RMS misfit in Occam's inversion by excessive iterations and artificial intervention.Chapter 5: display of 3D inversion results in 14 blocks of QT; each we have slices along depth, X and Y direction and preliminary geological interpretation. The inversion results show that there are many low resistive bodies inside the crust of QT. The first type of low resistive anomaly is horizontal moniliform distribution, and the second vertical pillar. They have their own specific dynamics of mechanism. Furthermore, we discover a series of large-scale discontinuities of electric interface, which may be considered as evidence of new subdivision of regional tectonics.Chapter 6: discuss some problems in 3D inversion and give corresponding suggestions. We have found that quality and density of data set is quite important for healthy inversion, in other word, bad data or poorly-covered data may cause ill-conditioned models. We also argue that it is a relatively good method to evaluate the static shift by artificial recognition of static body. Another discovery is that on-diagonal and off-diagonal elements of impedance tensor have different manners of response to the distribution of electric parameters, which has potential to block the convergence of iteration, so we think separate inversions for different modes is much better. We also recommend dynamic parameters in inversion to improve the efficiency. Geologically, we discuss the mechanism of moniliform and pillar low resistive bodies and the tectonic implication of deep and large electric interfaces.Finally, in Chapter 7, we summarize the experiences and knowledge obtained in this research as well as some weakness.