Exciting news! We're transitioning to the Statewide California Earthquake Center. Our new website is under construction, but we'll continue using this website for SCEC business in the meantime. We're also archiving the Southern Center site to preserve its rich history. A new and improved platform is coming soon!

A focal mechanism catalog for Southern California derived with deep learning algorithms

Yifang Cheng, Zachary E. Ross, Egill Hauksson, & Yehuda Ben-Zion

Published July 26, 2019, SCEC Contribution #9273, 2019 SCEC Annual Meeting Poster #084

We present an updated focal mechanism catalog for 16,341 events that occurred in southern California during 2017. The approach uses S/P amplitude ratios and first-motion P wave polarities to determine focal mechanisms with the HASH method of Hardebeck and Shearer (2002, 2003). In addition to the broadband channels used by Yang et al. (2012), three-component extremely short period channels (EH-) are also used to obtain P- and S-wave amplitudes. Moreover, we utilize a convolutional-neural-network (CNN) phase picker (Ross et al., 2018b) and polarity picker (Ross et al., 2018a), which are trained on millions of manually labelled seismograms from Southern California, to detect more phase arrivals and polarities and further improve the focal mechanism catalog. For all events, we detect additional P- and S-wave arrivals using the CNN phase picker to obtain more S/P amplitude ratios. For 14,569 small magnitude events (M<1.7), extra CNN P-wave first-motion polarities are also added into focal mechanism calculations. Compared with 4,728 focal mechanisms in Yang et al. (2012), 7,839 (66% more) quality A-D focal mechanisms are obtained, with >90% quality A and B common events having rotation angles less than 35 degree, which is the uncertainty value used to define quality B focal mechanism. The number of quality A, B, C, and D focal mechanisms increased by 28%, 46%, 81% and 69%, respectively. We also apply CNN pickers to 66 co-located events with similar waveforms to test the accuracy improvement. The solved focal mechanisms are more consistent after implementing CNN pickers. The results indicate that CNN pickers can significantly improve the quality and quantity of obtained focal mechanisms. The procedure will be applied to the rest of the catalog events in southern California to improve the focal mechanisms data base and related analysis results.

Cheng, Y., Ross, Z. E., Hauksson, E., & Ben-Zion, Y. (2019, 07). A focal mechanism catalog for Southern California derived with deep learning algorithms. Poster Presentation at 2019 SCEC Annual Meeting.

Related Projects & Working Groups