Use High-performance Computing Platform Study 3D Seismic Propagation Modelling And Anisotropy Inversion

With the development of computer technology, single-core supercomputer has gradually stepped out of people’s sight and replaced by a growing number of low-cost parallel computers. Parallel computing technology has brought enormous changes to our life, industry and technology with its effectiveness and efficiency. Geophysical industry has been dependent on the development of large-scale computing for many years. Two issues which are the most typical in seismology are selected in this thesis and the cluster of the China Earthquake Administration Lanzhou Institute of Seismology is used to carry out forward and inversion research work about seismic waves. The first part of this paper is about seismic forward modelling using spectral element method(SEM). Disaster assessment after the earthquake has always been an issue of concern for Earthquake Administration; it is very difficult to get a peak acceleration distribution when the station is sparse. Beginning with the introduction of the theory of SEM(use 1D example), we use SPECFEM program code by Princeton University to model Minxian-Zhangxian earthquake. This example indicates that using SEM to model earthquake is an effective method to get peak acceleration distribution; it is also useful before earthquake happens, which may contribute to urban planning. The second part is the main part of this paper which is about seismic anisotropy. Upper-mantle anisotropy has been studied for many years, while research of the crust anisotropy was faced with the inferior situation. On the one hand, the crust anisotropy is extremely complex, and the transplant of general research methods of mantle anisotropy to the study of crust anisotropy will encounter difficulties. This paper compares three receiver function methods to study crustal anisotropy in detail. Accounting that weighted stack method and shear wave split method both have limitation when dealing with a multi-layer with arbitrary axis anisotropy problem; we mainly use waveform inversion method. The influence of each anisotropy parameter was studied and verified by numerical experiments. The result indicates that the second inversion can improve the accuracy. The code of Raysum and Neighborhood method was rebuilt and H-k constraint was added to the code in order to get a rational result.At the end of this paper, 15-years’ data of 16 station locating in northeaster edge of Tibetan Plateau was used to study the dynamic reason for the uplift of the Tibet Plateau. After studying receiver functions, we find that there indeed exists low shear wave velocity in the middle and lower crust, and this result is more consistent with channel-flow hypothesis. The inversion result indicates that the upper crust anisotropy is fit for cracks-cause hypothesis. If channel-flow hypothesis is true, we can successfully explain our result of the fast axis distribution of the middle and lower crust anisotropy(in the northeast edge, channel-flow was blocked in the middle crust by the neighbor plate, and overflow in the lower crust). This research is still preliminary; and the dynamic mechanism of Tibet Plateau is a complex issue, which needs integrated research with other geophysical methods. My work is just to provide some criterion through anisotropy study.