Comparison of quality of determination of fault plane orientations (focal mechanisms) between automated (CNN) and manual data sets (SCSN). Correct polarity determinations (whether the earth is moving up or down at a particular spot) are essential to determine the focal mechanisms, and CNN yields many more focal mechanisms with polarities that fit the solutions (low percentage of misfits). In simple terms: while good arrival time picks have become fairly easy to determine, polarities have been much harder; this study shows that the polarities of the machines are better than of the humans. Credit: Ross et al., 2018, Figure 5
Source: Journal of Geophysical Research: Solid Earth

It is laborious and time consuming to determine arrival times and polarities for large seismic network data, but that data is necessary to determine the earthquake locations and orientations of the fault planes, which can impact on the energy propagation. Ross et al. [2018] show that by training a computer using deep learning methods, it is possible to accomplish these tasks automatically and accurately. Getting better results from a machine than people will allow seismic networks to determine much better locations and focal mechanisms, which can be used for detailed studies of faulting and stress. Many seismic networks have become automated, but those networks often have their location qualities suffer and the resulting automatic catalogues become worse for seismic research than the previous, manual ones, were. This technique could potentially revolutionize network seismology by making research quality data available much more quickly.

Citation: Ross, Z. E., Meier, M.‐A., & Hauksson, E. [2018]. P wave arrival picking and first‐motion polarity determination with deep learning. Journal of Geophysical Research: Solid Earth, 123.

—Martha K. Savage, Editor, JGR: Solid Earth

Text © 2018. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.