Using earthquakes to find earthquakes: Earthquake catalogs elucidate the behavior of faults and allow for rough estimates of when large earthquakes might occur. Cataloging small earthquakes is challenging because the small signal is often indistinguishable from noise. Ross et al. used a template-matching algorithm to find almost two million tiny earthquakes previously missed by other earthquake-logging techniques in Southern California (see the Perspective by Brodsky). This more-complete catalog can be used to better understand faults, earthquake reoccurrence, and other geophysical processes. Science , this issue p. 767 ; see also p. 736
Accurate earthquake locations are essential for providing reliable hazard assessments, understanding the physical mechanisms driving extended earthquake sequences, and interpreting fault structure. Techniques based on waveform cross correlation can significantly improve the precision of the relative locations of event pairs observed at a set of common stations. Here we describe GrowClust, an open‐source, relative relocation algorithm that can provide robust relocation results for earthquake sequences over a wide range of spatial and temporal scales. The method uses input differential travel times, cross‐correlation values, and reference starting locations, and applies a hybrid, hierarchical clustering algorithm to simultaneously group and relocate events within similar event clusters. The method is computationally efficient and numerically stable in its capacity to process large data sets and naturally applies greater weight to more similar event pairs. Additionally, it outputs location error estimates that can be used to help interpret the reliability and resolution of relocation results. As an example, we apply the GrowClust method to the recent Spanish Springs and Sheldon, Nevada, earthquake swarms. These sequences highlight the future potential for applying the GrowClust relocation method on a much larger scale within the region, where existing relocation results are sparse but vital for understanding the seismotectonics and seismic hazard of Nevada and eastern California.
Seismic swarms show the structure: Faults responsible for earthquakes are idealized into two dimensions, despite fault zones being complicated, three-dimensional structures. Ross et al. used machine learning to find 22,000 seismic events near Cahuilla, California, during a seismic swarm. They used the locations and sizes of these events to show how the complex structure of the fault interacted with natural fluid injections from below. The authors' methods highlight the complexities of one fault and suggest a way to characterize other faults around the world. Science , this issue p. 1357
We investigate the physics of laboratory earthquake precursors in a biaxial shear configuration. We conduct laboratory experiments at room temperature and humidity in which we shear layers of glass beads under applied normal loads of 2–8 MPa and with shearing rates of 5–10 µm/s. We show that above ~ 3 MPa load, acoustic emission (AE), and shear microfailure (microslip) precursors exhibit an exponential increase in rate of occurrence, culminating in stick‐slip failure. Precursors take place where the material is in a critical state—still modestly dilating, yet while the macroscopic frictional strength is no longer increasing.
16 April 2021 in Journal of Geophysical Research: Solid Earth
Detecting earthquake arrivals within seismic time series can be a challenging task. Visual, human detection has long been considered the gold standard but requires intensive manual labor that scales poorly to large data sets. In recent years, automatic detection methods based on machine learning have been developed to improve the accuracy and efficiency. However, the accuracy of those methods relies on access to a sufficient amount of high‐quality labeled training data, often tens of thousands of records or more. We aim to resolve this dilemma by answering two questions: (1) provided with a limited amount of reliable labeled data, can we use them to generate additional, realistic synthetic waveform data? and (2) can we use those synthetic data to further enrich the training set through data augmentation, thereby enhancing detection algorithms? To address these questions, we use a generative adversarial network (GAN), a type of machine learning model which has shown supreme capability in generating high‐quality synthetic samples in multiple domains. Once trained, our GAN model is capable of producing realistic seismic waveforms of multiple labels (noise and event classes). Applied to real Earth seismic data sets in Oklahoma, we show that data augmentation from our GAN‐generated synthetic waveforms can be used to improve earthquake detection algorithms in instances when only small amounts of labeled training data are available.