Big data for small earthquakes : detecting earthquakes over a seismic network with waveform similarity search
Abstract/Contents
- Abstract
- Earthquake detection, the problem of extracting weak earthquake signals from continuous waveform data recorded by networks of seismic sensors, is a critical and challenging task in seismology. To overcome the limitations of traditional energy-based detectors, waveform cross-correlation is now widely used to detect weak earthquake signals with waveforms similar to those of known events. Earthquake catalogs are often limited or incomplete, so we require a more general detector capable of finding events with unknown sources. Because earthquakes occur infrequently, detection algorithms must be capable of processing months to years of continuous data dominated by noise. To address these challenges, we have introduced Fingerprint and Similarity Thresholding (FAST), a new detection method that uses locality-sensitive hashing to perform a computationally efficient similarity search to identify all similar waveforms in the continuous data. FAST is the first detector based on waveform similarity that is capable of identifying events with unknown sources in long-duration data sets. In this work, I introduce three new extensions to the FAST earthquake detection method that help FAST achieve both high sensitivity and a low false detection rate when applied to years of continuous data. First, I present an improved feature extraction process that produces discriminative features ("waveform fingerprints") for earthquake waveforms. The proposed fingerprinting scheme is designed to detect low signal-to-noise earthquake waveforms using similarity search in imbalanced data sets dominated by noise. Next, I introduce a new analysis pipeline that extends a single-station FAST detector over a seismic network. The method leverages the pairwise event detection model of FAST to formulate network detection as a constraint on inter-event times of single-station detections. This approach is robust to missing data, effective in sparse networks with large station spacing, and scalable to large data sets while maintaining a low false detection rate. Finally, I present a partially informed variant of FAST that prioritizes data more likely to contain earthquake signals in similarity search. This approach makes more efficient use of available computational resources, including reducing runtime and memory usage, with a negligible effect on detection performance. I demonstrate the performance of these methods on real waveform data from northern California and Iquique, Chile.
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 | 2018; ©2018 |
| Publication date | 2018; 2018 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Author | Bergen, Karianne Jodine |
|---|---|
| Degree supervisor | Beroza, Gregory C. (Gregory Christian) |
| Degree supervisor | Biondi, Biondo, 1959- |
| Thesis advisor | Beroza, Gregory C. (Gregory Christian) |
| Thesis advisor | Biondi, Biondo, 1959- |
| Thesis advisor | Ellsworth, William L |
| Degree committee member | Ellsworth, William L |
| Associated with | Stanford University, Institute for Computational and Mathematical Engineering. |
Subjects
| Genre | Theses |
|---|---|
| Genre | Text |
Bibliographic information
| Statement of responsibility | Karianne Jodine Bergen. |
|---|---|
| Note | Submitted to the Institute for Computational and Mathematical Engineering. |
| Thesis | Thesis Ph.D. Stanford University 2018. |
| Location | electronic resource |
Access conditions
- Copyright
- © 2018 by Karianne Jodine Bergen
- 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...