To facilitate identification of conditions that lead to the dynamic triggering of seismic events as catalogs of these events keep growing, we applied a machine-learning algorithm (decision tree) to a published data set of known instances of dynamically triggered seismic tremor in central California. To investigate the possible universality of our findings and to further test the algorithm, we also applied it to new observations, presented here, of potentially dynamically triggered seismic activity in three intraplate regions: Raton Basin (CO), Yellowstone, and central Utah. We report potential tremor or local earthquake signals from here during the propagation of surface waves from the 2012 Mw 8.6 Sumatra earthquake. These surface waves also triggered seismic activity along the western boundary of the North American plate and did not trigger seismic activity in the central and eastern USA. We report additional potential dynamic triggering in the three aforementioned intraplate regions from an investigation of seismograms from 37 additional large earthquakes, recorded between 2004 to 2017. Our findings show that transient stresses generated by surface waves from large earthquakes and arriving from favorable directions generally lead to triggered tremor in seismically, volcanically, and hydrothermally active regions like central California and possibly Yellowstone. These stresses do not appear to be decisive factors for the potentially dynamically triggered local earthquakes reported for the Raton Basin and central Utah, while surface waves’ incidence angles do appear to be important there.

We present new, densely sampled shear-wave splitting results from southern Minnesota and adjacent areas of neighboring states, sampling the southwestern limit of the Archean Superior Province and straddling the Proterozoic Mid-Continent Rift (MCR). The new measurements include data from the Earthscope Transportable Array (TA) as well as the Superior Province Rifting Earthscope Experiment (SPREE), yielding 99 new station-averaged measurements. The split times show a consistent decrease from 1.1 s in the NE to 0.2 s in the SW, with the lowest values being associated with the Minnesota River Valley Terrane (MRVT). From modelling and other geophysical constraints, we interpret the split time variations to represent variations in fabric strength within a thick lithosphere, rather than lithospheric thinning or a multi-layered effect, and propose that the weak fabric of the MRVT is associated with a different mechanism of formation than elsewhere in the Superior Province. The fast directions we measure range from NNE-SSW to E-W and vary on a shorter length scale than the split times, with a pattern of NE-SW splits that closely follows the axis of the MCR. We interpret this as a perturbation of the net fast direction due to anisotropy in an underplate along the rift.

A comprehensive North American upper mantle seismic-tomographic model, NA13, was inferred from a combination of several decades of seismic data from early North American seismic models and USArray data. This data-driven modeling inverted a combination of regional waveform misfits, teleseismic S delay times, and constraints on Moho depths. The joint inversion model combines the contrasting and overlapping resolving power of the different data sets, and demonstrates enhanced resolution over regional models created with a single geophysical dataset. Resolution in the upper mantle is achieved on the scale of ~100km, although travel time delay studies suggest that anomalies in parts of the western United States are smaller. In NA13, velocities beneath the Yellowstone plume are low enough to require the presence of partial melt. NA13 also models a variety of smaller scale velocity variations beneath the Central and Eastern United States that reveal remnant velocity anomalies in the upper mantle from Mesozoic and Cenozoic tectonics.

Dynamically triggered earthquakes and tremor generate weak seismic signals whose detection, identification, and authentication call for a laborious analysis. Citizen science project Earthquake Detective leverages the eyes and ears of volunteers to detect and classify weak signals in seismograms from potentially dynamically triggered (PDT) events. Here, we present the Earthquake Detective data set - A crowd-sourced set of labels on PDT earthquakes and tremor. We apply Machine Learning to classify these PDT seismic events and explore the challenges faced in segregating such weak signals. The algorithm confirms that machine learning can detect signals from small earthquakes, and newly demonstrates that this specific algorithm can also detect signals from PDT tremor. The data set and code are available online.

To facilitate identification of conditions that lead to the dynamic triggering of seismic events as catalogs of these events keep growing, we applied a machine-learning algorithm (decision tree) to a published data set of known instances of dynamically triggered seismic tremor in central California. To investigate the possible universality of our findings and to further test the algorithm, we also applied it to new observations, presented here, of potentially dynamically triggered seismic activity in three intraplate regions: Raton Basin (CO), Yellowstone, and central Utah. We report potential tremor or local earthquake signals from here during the propagation of surface waves from the 2012 M8.6 Sumatra earthquake. These surface waves also triggered seismic activity along the western boundary of the North American plate and did not trigger seismic activity in the central and eastern USA. We report additional potential dynamic triggering in the three aforementioned intraplate regions from an investigation of seismograms from 37 additional large earthquakes, recorded between 2004 to 2017. Our findings show that transient stresses generated by surface waves from large earthquakes and arriving from favorable directions generally lead to triggered tremor in seismically, volcanically, and hydrothermally active regions like central California and possibly Yellowstone. These stresses do not appear to be decisive factors for the potentially dynamically triggered local earthquakes reported for the Raton Basin and central Utah, while surface waves’ incidence angles do appear to be important there.

We present shear-wave splitting analyses of SKS and SKKS waves recorded at sixteen Superior Province Rifting Earthscope Experiment (SPREE) seismic stations on the north shore of Lake Superior, as well as fifteen selected Earthscope Transportable Array instruments south of the lake. These instruments bracket the Mid-Continent Rift (MCR) and sample the Superior, Penokean, Yavapai and Mazatzal tectonic provinces. The data set can be explained by a single layer of anisotropic fabric, which we interpret to be dominated by a lithospheric contribution. The fast S polarization directions are consistently ENE-WSW, but the split time varies greatly across the study area, showing strong anisotropy (up to 1.48 s) in the western Superior, moderate anisotropy in the eastern Superior, and moderate to low anisotropy in the terranes south of Lake Superior. We locate two localized zones of very low split time (less than 0.6 s) adjacent to the MCR: one in the Nipigon Embayment, an MCR-related magmatic feature immediately north of Lake Superior, and the other adjacent to the eastern end of the lake, at the southern end of the Kapuskasing Structural Zone (KSZ). Both low-splitting zones are adjacent to sharp bends in the MCR axis. We interpret these two zones, along with a low-velocity linear feature imaged by a previous tomographic study beneath Minnesota and the Dakotas, as failed lithospheric branches of the MCR. Given that all three of these branches failed to propagate into the Superior Province lithosphere, we propose that the sharp bend of the MCR through Lake Superior is a consequence of the high mechanical strength of the Superior lithosphere ca. 1.1 Ga.