Earthquake source process analysis codes incorporeting the effect of the heterogeneous structure at the subduction zone.

点击翻译按钮获取中文摘要

(1)Far field body wave codeA new, very effective 2.5D plane wave propagation FDM code based on Takenaka & Kennet (1996) algorithm has been developed by co-working with them (Takenaka and Kennett), It is much faster than the previous one (more than 10 times faster). We expect it is applied to various problems including our own research.2)Near field complete seismic waveform codeWe have verified the accuracy for the fluid-solid boundaries for finite-difference method and have found a proper grid arrangement to improve the accuracy. The resultant code generates waveforms that matches well with the observed waveforms for the seismo-acoustic environment.3)Moment tensor inversion by using near-field full waveformsWe have developed a method based on the reciprocal theorem. This method can generate Green's functions for a station and various source locations in a single calculation. It is considered that this approach will become one of the standard method. We tested the fundamental aspects of … More this method by applying our code for a shallow earthquake at Japan Trench and obtained a very reasonable inversion result. This ensures the applicability of the approach.4)Far fried body waveform analysis codeWe have developed two codes : one is to determine the source mechanism and the source position and the other is to determine the distribution of the moment release on a fault, In the latter code, we have formulated a non-linear inversion procedure to avoid a negative moment. We have applied our new method to the analysis of 1992 Nicaraguan Tsunami earthquake for which a very shallow moment release is suggested by tsunami waveform analysis. Our result based on the seismic waveforms suggest a relatively deep moment release.5)Parallel codeWe have used the standard mechanism for inter-process communications on UNIX OS to implement a parallel functionality in our near-field waveform code. For a dual processor calculation, we observed a 1.5 times acceralation against a single processor calculation. Our code has not achieved the optimum parallelism and further study will be required. Less