Issues

Front: Machine learning (ML) is a collection of algorithms and statistical models that enable computers to extract relevant patterns and information from large datasets. Seismologists have usedML algorithms for decades to analyze seismic signals, but in just the past few years research activity aboutML applications in seismology has surged, driven by the increasing size of seismic datasets, improvements in computational power, new algorithms and architectures (e.g., deep neural networks), and the availability of easy-to-use open-source ML frameworks. In this issue of SRL, the Focus Section on Machine Learning in Seismology presents 16 original articles covering a range of ML applications. This illustration, which is based on figures in Nakano et al. (this issue), shows an example of seismic data flow and architecture for an ML neural network application for the study of tectonic tremor.

Back: When the M_w 7.8 San Francisco earthquake struck on 18 April 1906, a narrow-gauge locomotive and train that had pulled into a siding to refuel at Point Reyes Station toppled due to ground motion caused by the earthquake. This photo shows two people and a canine at the site of the upset locomotive, with Point Reyes Station and its damaged buildings in the distance (U.S. Geological Survey Photographic Library). Veeraraghavan et al. (this issue) mathematically modeled the tipping of the train to calculate a lower limit on the earthquake’s ground motion at the site. Such analyses provide important additional data points for scientists who are working to anticipate the ground motions that will result from future large earthquakes.