Waveform-similarity link analysis for low frequency earthquake detection

Placeholder Show Content


Slow slip events were first detected by GPS as a reversal of motion of the Earth's surface due to episodic slip on the deep extension of the Cascadia subduction zone, down dip of the seismogenic zone. They were initially thought to be silent, but later related to a seismic signal resembling coherent noise across a network. This signal was called non-volcanic tremor and the association of the tremor and slow slip sources came to be called Episodic Tremor and Slip (ETS). Since its discovery, this type of event has been found in many tectonic regimes and other seismic signals, such as low frequency earthquakes (LFEs), have been associated with them, as a manifestation of the same process at depth. LFEs were originally found in Japan within tectonic tremor. Tectonic tremor was later described as a superposition of LFEs, making it possible to locate the tremor with greatly improved accuracy. LFEs were used as templates to find other LFEs within the tremor using a match filter technique. An autocorrelation method was later developed to find the repeating signals where templates were not known. Both of these methods are based on detection in a pair wise manner but do not exploit the fact that LFE signals repeat many times. I have derived a method based on autocorrelation that exploits the repetitive nature of the LFEs. I apply Google's PageRank algorithm to the window pairs found as matches above a statistically significant threshold. This algorithm ranks windows based on the links to other windows, and considers both direct and indirect links, takes into account the complex hierarchical relationships between windows and assigns a ranking to each window based on all links. I use the PageRank results to create robust templates by stacking windows linked directly and indirectly to the highest ranked window. I validate this method against data from the April 2006 Shikoku, Japan tremor episode and find that stacks created using the PageRank algorithm match known LFEs from the JMA catalog that locate to the same area of LFEs from the 2006 event. Using these templates we find similar detections to Shelly et al. (2007b) for the same time periods, but also find detections for weaker segments of tremor not previously reported. I also show that the PageRank algorithm can help differentiate between tremor and noise by using the histogram of its the probabilities where tremor data show far greater numbers of highly ranked values and noise data shows relatively fewer numbers of high values. I have applied the PageRank algorithm to two new data sets. The first is from Northern New Zealand, where tremor is not easily observed due to attenuation and a limited seismic coverage. I create stacks using the PageRank algorithm for different tremor bursts during the 2010 Gisborne slow slip event. I find that LFE templates from different small tremor bursts have a high correlation suggesting a similar source for both. Using these templates, I detect many LFEs within the tremor bursts as well as outside of the tremor, suggesting that the tremor may not be easily observed in this data set. The second data set I applied my method to is tremor in southern Taiwan. Here tremor has been found to be triggered by surface waves and also occurs spontaneously. It is composed of LFEs, which were found manually and later used as templates. I applied the PageRank algorithm both during triggered and ambient tremor. I found a repeating signal in both data sets and find that the stacks created from data during the triggered tremor closely match the stacks created from ambient tremor suggesting a similar location for both. Finally, as a different focus, I use estimates of attenuation terms from Ground Motion Prediction Equations calculated from tremor in Cascadia to map the spatial variability of the amplitude attenuation term across the northern Cascadia subduction zone.


Type of resource text
Form electronic; electronic resource; remote
Extent 1 online resource.
Publication date 2015
Issuance monographic
Language English


Associated with Aguiar Moya, Ana Cristina
Associated with Stanford University, Department of Geophysics.
Primary advisor Beroza, Gregory C. (Gregory Christian)
Thesis advisor Beroza, Gregory C. (Gregory Christian)
Thesis advisor Lawrence, Jesse
Thesis advisor Segall, Paul, 1954-
Advisor Lawrence, Jesse
Advisor Segall, Paul, 1954-


Genre Theses

Bibliographic information

Statement of responsibility Ana Cristina Aguiar Moya.
Note Submitted to the Department of Geophysics.
Thesis Thesis (Ph.D.)--Stanford University, 2015.
Location electronic resource

Access conditions

© 2015 by Ana Cristina Aguiar Moya
This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).

Also listed in

Loading usage metrics...