No abstract

No abstract

The 13 November 2016 Mw 7.8 Kaikoura, New Zealand, earthquake was investigated using teleseismic P waves. Backprojection of high‐frequency P waves from two regional arrays shows unilateral rupture of at least two southwest‐northeast striking faults with an average rupture speed of 1.4–1.6 km/s and total duration of ~100 s. Guided by these backprojection results, 33 globally distributed low‐frequency P waves were inverted for a finite fault model (FFM) of slip. The FFM showed evidence of several subevents; however, it lacked significant moment release near the epicenter, where a large burst of high‐frequency energy was observed. A local strong‐motion network recorded strong shaking near the epicenter; hence, for this earthquake the distribution of backprojection energy is superior to the FFM as a guide of strong shaking. For future large earthquakes that occur in regions without strong‐motion networks, initial shaking estimates could benefit from backprojection constraints.

On the evening of 10 April 2013 (MDT) a massive landslide occurred at the Bingham Canyon copper mine near Salt Lake City, Utah, USA. The northeastern wall of the 970-m-deep pit collapsed in two distinct episodes that were each sudden, lasting ~90 seconds, but separated in time by ~1.5 hours. In total, ~65 million cubic meters of material was deposited, making the cumulative event likely the largest non-volcanic landslide to have occurred in North America in modern times. Fortunately, there were no fatalities or injuries. Because of extensive geotechnical surveillance, mine operators were aware of the instability and had previously evacuated the area. The Bingham Canyon mine is located within a dense regional network of seismometers and infrasound sensors, making the 10 April landslide one of the best recorded in history. Seismograms show a complex mixture of short-and long-period energy that is visible throughout the network (6-400 km). Local magnitudes (M L) for the two slides, which are based on the amplitudes of short-period waves, were estimated at 2.5 and 2.4, while magnitudes based on the duration of seismic energy (m d) were much larger (>3.5). This magnitude discrepancy, and in particular the relative enhancement of longperiod energy, is characteristic of landslide seismic sources. Interestingly, in the six days following the landslide, 16 additional seismic events were detected and located in the mine area. Seismograms for these events have impulsive arrivals characteristic of tectonic earthquakes. Hence, it appears that in this case the common geological sequence of events was inverted: Instead of a large earthquake triggering landslides, it was a landslide that triggered several small earthquakes.

No abstract

A large and tragic underground collapse occurred in the Crandall Canyon coal mine in eastcentral Utah on 6 Aug 2007, causing the loss of six miners and attracting national attention. This collapse was accompanied by a local magnitude (M L) 3.9 seismic event having a location and origin time coincident with the collapse, within current uncertainty limits. Two lines of evidence indicate that most of the seismic wave energy of this event was generated by the mine collapse rather than a naturally-occurring earthquake: (1) the observation that all of the observed P-wave first motion directions are down and (2) the results of a moment tensor inversion by Ford et al. (2008). We propose one possible model for the collapse that has dimensions of 920 m E-W by 220 m N-S and an average roof-floor closure of 0.3 m. This model is consistent with the seismic moment, volumetric constraints on the amount of closure, available underground observations, and our best location for the M L 3.9 epicenter. This epicenter is near the western end of our proposed collapse area, suggesting that the collapse propagated mostly eastward from its initiation point. Our locations for the M L 3.9 event and for other seismic events that occurred in the area before and after it were greatly improved by the use of a double difference method and data from a 5-station temporary network that the University of Utah deployed near the mine beginning on 8 Aug. The Crandall Canyon Mine is in an area of Utah where there is abundant mining-induced seismicity, including events with both collapse and shear-slip sources. Prior to the 6 Aug 2 collapse, and within a 3 km radius of it, there were 28 seismic events during 2007 that were large enough to be detected and located as part of the routine data processing for the University of Utah regional seismic network: 8 in the 2.5-week period prior to the collapse (M L ≤ 1.9) and 15 during an earlier period of activity in late February and early March (M L ≤ 1.8). These events occurred primarily in areas where there was concurrent or recent mining activity. By the end of August, the 6 Aug collapse had been followed by 37 locatable seismic events of M L ≤ 2.2, which clustered near the eastern and inferred western ends of the collapse area. One of these "aftershocks" (M L 1.6) occurred in conjunction with the violent burst of coal from the mine walls on 17 Aug (UTC) that killed three rescuers and injured six others. The aftershocks have an exponential frequency-magnitude distribution with a lower ratio between the frequencies of smaller-and larger-magnitude events (lower b-value) than for the prior events in the area. Aftershock rates generally decreased with time through August. However, there was a noteworthy 5.8-day hiatus in activity, above a completeness threshold of coda magnitude (M C) 1.6, that began 37 hours after the collapse.

Long‐term seismic monitoring networks are well positioned to leverage advances in machine learning because of the abundance of labeled training data that curated event catalogs provide. We explore the use of convolutional and recurrent neural networks to accomplish discrimination of explosive and tectonic sources for local distances. Using a 5‐year event catalog generated by the University of Utah Seismograph Stations, we train models to produce automated event labels using 90‐s event spectrograms from three‐component and single‐channel sensors. Both network architectures are able to replicate analyst labels above 98%. Most commonly, model error is the result of label error (70% of cases). Accounting for mislabeled events (~1% of the catalog) model accuracy for both models increases to above 99%. Classification accuracy remains above 98% for shallow tectonic events, indicating that spectral characteristics controlled by event depth do not play a dominant role in event discrimination.

S U M M A R YWe report on a unique set of infrasound observations from a single earthquake, the 2011 January 3 Circleville earthquake (M w 4.7, depth of 8 km), which was recorded by nine infrasound arrays in Utah. Based on an analysis of the signal arrival times and backazimuths at each array, we find that the infrasound arrivals at six arrays can be associated to the same source and that the source location is consistent with the earthquake epicentre. Results of propagation modelling indicate that the lack of associated arrivals at the remaining three arrays is due to path effects. Based on these findings we form the working hypothesis that the infrasound is generated by body waves causing the epicentral region to pump the atmosphere, akin to a baffled piston. To test this hypothesis, we have developed a numerical seismoacoustic model to simulate the generation of epicentral infrasound from earthquakes. We model the generation of seismic waves using a 3-D finite difference algorithm that accounts for the earthquake moment tensor, source time function, depth and local geology. The resultant acceleration-time histories on a 2-D grid at the surface then provide the initial conditions for modelling the near-field infrasonic pressure wave using the Rayleigh integral. Finally, we propagate the near-field source pressure through the Ground-to-Space atmospheric model using a time-domain Parabolic Equation technique. By comparing the resultant predictions with the six epicentral infrasound observations from the 2011 January 3, Circleville earthquake, we show that the observations agree well with our predictions. The predicted and observed amplitudes are within a factor of 2 (on average, the synthetic amplitudes are a factor of 1.6 larger than the observed amplitudes). In addition, arrivals are predicted at all six arrays where signals are observed, and importantly not predicted at the remaining three arrays. Durations are typically predicted to within a factor of 2, and in some cases much better. These results suggest that measured infrasound from the Circleville earthquake is consistent with the generation of infrasound from body waves in the epicentral region.