CaltechAUTHORS
  A Caltech Library Service

Generalized Seismic Phase Detection with Deep Learning

Ross, Zachary E. and Meier, Men-Andrin and Hauksson, Egill and Heaton, Thomas H. (2018) Generalized Seismic Phase Detection with Deep Learning. Bulletin of the Seismological Society of America, 108 (5A). pp. 2894-2901. ISSN 0037-1106. https://resolver.caltech.edu/CaltechAUTHORS:20180821-095421224

[img] PDF - Submitted Version
See Usage Policy.

5Mb
[img] Archive (ZIP) (Figures showing training and validation loss history, example waveforms used, and additional performance statistics) - Supplemental Material
See Usage Policy.

451Kb

Use this Persistent URL to link to this item: https://resolver.caltech.edu/CaltechAUTHORS:20180821-095421224

Abstract

To optimally monitor earthquake‐generating processes, seismologists have sought to lower detection sensitivities ever since instrumental seismic networks were started about a century ago. Recently, it has become possible to search continuous waveform archives for replicas of previously recorded events (i.e., template matching), which has led to at least an order of magnitude increase in the number of detected earthquakes and greatly sharpened our view of geological structures. Earthquake catalogs produced in this fashion, however, are heavily biased in that they are completely blind to events for which no templates are available, such as in previously quiet regions or for very large‐magnitude events. Here, we show that with deep learning, we can overcome such biases without sacrificing detection sensitivity. We trained a convolutional neural network (ConvNet) on the vast hand‐labeled data archives of the Southern California Seismic Network to detect seismic body‐wave phases. We show that the ConvNet is extremely sensitive and robust in detecting phases even when masked by high background noise and when the ConvNet is applied to new data that are not represented in the training set (in particular, very large‐magnitude events). This generalized phase detection framework will significantly improve earthquake monitoring and catalogs, which form the underlying basis for a wide range of basic and applied seismological research.


Item Type:Article
Related URLs:
URLURL TypeDescription
https://doi.org/10.1785/0120180080DOIArticle
https://arxiv.org/abs/1805.01075arXivDiscussion Paper
ORCID:
AuthorORCID
Ross, Zachary E.0000-0002-6343-8400
Meier, Men-Andrin0000-0002-2949-8602
Hauksson, Egill0000-0002-6834-5051
Heaton, Thomas H.0000-0003-3363-2197
Additional Information:© 2018 Seismological Society of America. Manuscript received 15 March 2018; Published Online 21 August 2018. The authors thank an anonymous reviewer and Eric Chael for reviews of the article. This research was supported by grants from the Gordon and Betty Moore Foundation, the Swiss National Science Foundation, and the National Science Foundation (NSF) Geoinformatics program. The authors used waveforms and metadata from Japanese networks and the California Institute of Technology (Caltech)/U.S. Geological Survey (USGS) Southern California Seismic Network (SCSN; doi: 10.7914/SN/CI); stored at the Southern California Earthquake Data Center (doi: 10.7909/C3WD3xH1). They used Tensorflow to train the convolutional network. Data and Resources: In this study, we used seismograms for 273,882 earthquakes (−0.81 < M < 5.7) recorded by the Southern California Seismic Network (SCSN) at 692 broadband and short‐period three‐component stations from 2000 to 2017 (Southern California Earthquake Data Center, 2013). The waveform data were associated with 1.5 million P‐wave picks and 1.5 million S‐wave picks that were manually determined by SCSN analysts. We also used 24 hrs of continuous data at station CI BOM on 26 September 2016, the first day of the 2016 Bombay Beach, California, swarm. Seismicity during this sequence was taken from the SCSN regional catalog. Finally, we used KiK‐net and K‐NET accelerograms of the 2016 Mw 7.0 Kumamoto earthquake at stations within 100‐km distance of the hypocenter (http://www.kyoshin.bosai.go.jp/, last accessed January 2018). The trained model and an accompanying model will be available through the Southern California Earthquake Data Center at scedc.caltech.edu (last accessed January 2018).
Group:Seismological Laboratory
Funders:
Funding AgencyGrant Number
Gordon and Betty Moore FoundationUNSPECIFIED
Swiss National Science Foundation (SNSF)UNSPECIFIED
NSFUNSPECIFIED
Issue or Number:5A
Record Number:CaltechAUTHORS:20180821-095421224
Persistent URL:https://resolver.caltech.edu/CaltechAUTHORS:20180821-095421224
Official Citation:Zachary E. Ross, Men‐Andrin Meier, Egill Hauksson, Thomas H. Heaton; Generalized Seismic Phase Detection with Deep Learning. Bulletin of the Seismological Society of America ; 108 (5A): 2894–2901. doi: https://doi.org/10.1785/0120180080
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:88978
Collection:CaltechAUTHORS
Deposited By: Tony Diaz
Deposited On:21 Aug 2018 18:25
Last Modified:03 Oct 2019 20:12

Repository Staff Only: item control page