Development of innovative methodology for seismological analysis with applications to the studies of tectonic structures and natural hazards

点击翻译按钮获取中文摘要

Seismology is a branch of observational science that depends critically on innovative analysis of seismic data. High-resolution results come from accurate and comprehensive analysis of available data with minimum artifacts. One of the most effective ways to make scientific breakthroughs is to develop innovative methodologies capable of extracting subtle, yet important information that were underexplored in the past. The purpose of this proposal is to seek modest funds for partial support of two graduate students to conduct innovative development of seismological methodologies that will significantly improve the quality of seismic data analysis. Specifically, the proposed research consists of three (3) major activities. The first one is to develop an artificial intelligence (AI)-based method to accurately detect weak earthquake signals. Recent development of convolutional neural networks has greatly improved the ability and efficiency of extracting systematic information from large datasets. We plan to focus our research on the issue of induced seismicity that are possibly related to hydraulic fracturing and/or wastewater disposal in western Canada. The ultimate goal is to establish a physical model that can not only explain the observed pattern of injection-induced seismicity but also provide a certain level of predictability on the spatiotemporal distribution of induced events based on the identified factors and variables. The second activity is to develop an earthquake location algorithm that is automatic, comprehensive, accurate and versatile. Determination of precise source locations is a very time-consuming process that requires skillful analysts with consistent performance of seismic phase-picking. Existing automatic earthquake location algorithms all have specific advantages and drawbacks and the final results still require human verification. We propose to take a totally new cocktail approach by taking advantage of the merit of each algorithm to achieve the optimal performance. The newly developed method will be capable of generating the most accurate and comprehensive seismic catalogue without any human intervention. Finally, we plan to develop a systematic method to accurately determine the shallow and deep velocity structures simultaneously, especially for regions without frequent local seismicity such as the continental interior of North America. Surface wave tomography is an ideal tool if the dispersion curve spans a broad period range. However, the dispersion curve derived from conventional processing of ambient seismic noise (between 2 and 50 s) is not always consistent with that from earthquake surface waves (between 20 and 150 s). We propose to study this inconsistency in detail and develop a totally automatic algorithm to process seismic data such that a broadband (from 1 s to >100 s) dispersion curve can be reliably constructed and used for velocity structure inversion at all depths.