Characterizing induced earthquake sequences with machine learning
Abstract/Contents
- Abstract
- There has been increasing concern regarding induced earthquakes and their related hazard, especially since the last decade. Although many of the induced earthquake sequences have been extensively investigated by researchers, there are still unresolved questions about their mechanism, which is crucial for hazard assessment. In this thesis, I focused on two induced earthquake sequences: the Timpson induced earthquakes from 2012 to 2013 and the seismicity in the Rangely earthquake control experiment from 1972 to 1974. To better characterize them, I applied various machine learning techniques including both supervised and unsupervised methods. Our results offer new insights on the physical mechanism of the two induced earthquake sequences. Moreover, I developed an image-based, deep-learning workflow to efficiently process Develocorder film seismograms, which demonstrates the potential to conduct seismic analysis on analog data at scale. Chapter 2 focused on the Timpson induced earthquakes, aiming to explain the discrepancy between the predicted fault slip potential in local stress field and the observed seismicity. I applied a deep-learning based phase picker to build an enriched catalog and followed by waveform similarity based earthquake clustering. This revealed the fine structure of the Timpson fault and thus successfully explained the stress and seismicity discrepancy. Chapter 3 described an image-based, machine learning workflow to process Develocorder film seismograms. I demonstrated its performance on one month of the analog seismograms in the Rangely experiment and showed the feasibility of image-based processing for analog data. Chapter 4 further improved this method by training a deep-learning neural network, DevelNet, to perform image-based earthquake detection on Develocorder films. I then applied DevelNet on two years of the film recordings in the Rangely experiment and showed that our reconstructed catalog outperformed the original one by Raleigh et al. (1976). Finally, Chapter 5 built on our reconstructed catalog and performed geomechanical analysis to determine the stress state of the Rangely faults. I further conducted earthquake clustering analysis and found characteristic short-lived sub-sequences in the Rangely earthquakes. Based on the spatiotemporal distribution of the seismicity and its relation to the observed reservoir pressure, I discussed possible mechanisms that may drive the Rangely earthquakes.
Description
| Type of resource | text |
|---|---|
| Form | electronic resource; remote; computer; online resource |
| Extent | 1 online resource. |
| Place | California |
| Place | [Stanford, California] |
| Publisher | [Stanford University] |
| Copyright date | 2021; ©2021 |
| Publication date | 2021; 2021 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Author | Wang, Kaiwen |
|---|---|
| Degree supervisor | Beroza, Gregory C. (Gregory Christian) |
| Thesis advisor | Beroza, Gregory C. (Gregory Christian) |
| Thesis advisor | Ellsworth, William L |
| Thesis advisor | Zoback, Mark D |
| Degree committee member | Ellsworth, William L |
| Degree committee member | Zoback, Mark D |
| Associated with | Stanford University, Department of Geophysics |
Subjects
| Genre | Theses |
|---|---|
| Genre | Text |
Bibliographic information
| Statement of responsibility | Kaiwen Wang. |
|---|---|
| Note | Submitted to the Department of Geophysics. |
| Thesis | Thesis Ph.D. Stanford University 2021. |
| Location | https://purl.stanford.edu/pd778ny1280 |
Access conditions
- Copyright
- © 2021 by Kaiwen Wang
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...