金曜日セミナー:(2019年8月30日) Dr. Zachary E. Ross (CalTech)

Earthquakes and the era of artificial intelligence


Zachary E. Ross
Seismological Laboratory, California Institute of Technology

The volume of seismic data recorded around the world is exploding. Standard techniques for earthquake detection routinely miss the smallest earthquakes, yet these hidden events form the vast majority of seismic activity due to power law magnitude scaling. To overcome these challenges we applied a template matching detection technique to the entire continuous waveform archive of the Southern California Seismic Network (2008-2017). This effort resulted in a catalog of 1.8 million earthquakes, a factor of ten increase over the standard catalog. The catalog provides new fundamental observations on foreshocks and nucleation processes, the geometry of fault zones, and the physics of earthquakes. However, there is one aspect about the catalog that is unsatisfactory: since seismograms of previous earthquakes are used to search for new ones, the procedure is biased towards events that have been seen before.

To move beyond these problems, we develop an end-to-end detection pipeline incorporating algorithms from artificial intelligence that can produce seismicity catalogs starting from raw seismic data. First, continuous data is processed with a deep convolutional network designed for generalized detection of body waves. This network operates on data from individual 3-component stations to reduce the total continuous volume to a set of discrete phase detections. The resulting phase detections are then examined by a recurrent neural network that is trained to link together phases that originate from the same earthquake. The training for this network is performed entirely with synthetic data. The final clusters of phase picks can then be used with a hypocenter inversion. We demonstrate the performance of the complete pipeline on datasets from southern California and Japan. The developed framework has the potential to significantly improve earthquake monitoring capabilities.

Main references

  1. Ross, Z. E., Trugman, D. T., Hauksson, E., and Shearer, P. M. (2019). Searching for hidden earthquakes in Southern California, Science 364(6442), 767-771, doi: 10.1126/science.aaw6888

  2. Ross, Z. E., Yue, Y., Meier, M.-A., Hauksson, E., and T. H. Heaton (2019). PhaseLink: A Deep Learning Approach to Seismic Phase Association, J. Geophys. Res.-Solid Earth, 124, doi: 10.1029/2018JB016674.

  3. Ross, Z. E., Meier, M.-A., Hauksson, E., and T. H. Heaton (2018). Generalized Seismic Phase Detection with Deep Learning, Bull. Seismol. Soc. Am., 108 (5A), 2894-2901, doi: 10.1785/0120180080.