The Guy-Greenbrier seismic sequence revisited with deep learning

Yongsoo Park, Mostafa Mousavi, Weiqiang Zhu, William L. Ellsworth, & Gregory C. Beroza

Published August 14, 2019, SCEC Contribution #9651, 2019 SCEC Annual Meeting Poster #077

The emerging deep-learning-based algorithms for earthquake detection and phase arrival time picking have shown to be promising replacements for traditional methods such as STA/LTA and template matching. We revisited the Guy-Greenbrier, Arkansas, seismic sequence to demonstrate the power of these algorithms. The seismicity during the period from June 2010 to October 2011 is thought to be due to a combination of hydraulic stimulation of horizontal production wells and injection of wastewater into deep disposal wells. We used the PhaseNet picker (Zhu and Beroza, 2019) to produce P- and S-wave picks that were associated using a back-projection method. Using this workflow on continuous seismic waveform data, we were able to locate over 80,000 events with minimal human intervention. In doing so, we increased the spatio-temporal resolution of the earthquake catalog substantially compared to the previous studies where different earthquake detection algorithms were used, and this can bring new insights into the relationships between injection and seismicity. Moreover, the increased efficiency and robustness of the detection, picking and location routines have potential to improve existing real-time seismic monitoring techniques for industrial scale applications.

Park, Y., Mousavi, M., Zhu, W., Ellsworth, W. L., & Beroza, G. C. (2019, 08). The Guy-Greenbrier seismic sequence revisited with deep learning. Poster Presentation at 2019 SCEC Annual Meeting.

Related Projects & Working Groups