Abstract: On 2 September 2017, an Mw 5.3 normal-faulting earthquake occurred in the Intermountain Seismic Belt about 15 km east of Soda Springs, ID, near Sulphur Peak. It was widely felt throughout southeastern Idaho, northern Utah, and western Wyoming, but caused little damage. It appears to be co-located with two earlier seismic sequences, one in 1960 and the other in 1982. To achieve high-accuracy locations, the University of Utah and the USGS partnered with the Idaho Geological Survey to deploy eight telemetered seismographs in the epicentral region. Using absolute and differential arrival times we determined high-accuracy locations for a catalog of ~1,100 aftershocks, complete to ML 2.5. Initial locations were determined with HYPOINVERSE using a local 1D velocity model, and further refined using joint hypocentral decomposition (MLOC) and cluster-based relative relocation (GrowClust). For 76 of the largest events in the sequence, we inverted regional distance waveforms for moment tensors. The small uncertainties in absolute hypocenters imply that two first-order observations about the 2017 Sulphur Peak sequence are robust. First, the events are to the east of the west-dipping Eastern Bear Lake Fault, perhaps in a complex fracture zone within the footwall. Second, the events are confined to the upper crust, with a maximum depth of ~10 km. Like previous seismicity in this region—such as the M4.7 Soda Springs sequence of 1982 and the M5.7 Draney Peak sequence of 1994—the 2017 Sulphur Peak sequence had an energetic aftershock sequence with the cumulative aftershock moment 0.7–1.4 times as large as the mainshock moment. The unusually high productivity of the sequence is evident from the fact that 17 aftershocks had magnitudes larger than the upper bound expected from Bath’s Law, and 16 of the 17 occurred within ten days of the mainshock. Following Reasenberg and Jones [1994], the expected number of such aftershocks is 0–4, and the probability of observing 16 is only 2.3×10-12.
Media has not been submitted for this Presentation
Abstract: The Alpine Fault is inferred to be late in its typical ~300-year cycle of M 8 earthquakes based on paleoseismological evidence. The central Southern Alps exhibit the highest deformation and uplift rates along the Alpine Fault. However, levels of contemporary earthquake activity (M≥2) are low when compared to adjacent areas. Compiling a micro-earthquake catalogue for the central Southern Alps may help better understand the distribution of deformation in a region prior to the next major earthquake. We have created an almost decade long earthquake catalogue containing 9,111 earthquakes spanning the time period between late 2008 and early 2017. The earthquake locations are based on high-quality picks (83,138 P- and 67,119 S-phase picks of which approximatelly 35% are manual). To quantify the spatial distribution of the seismic moment release, we have derived a new local magnitude scale based on Mw values. Magnitudes range between ML -1.5 and 4.6 and have a magnitude of completeness of 1.1 with a b-value of 0.85. The seismogenic cut-off depths vary from 8 to 20 km and show both strike parallel and strike perpendicular variations that are strongly related to lateral changes in thermal structure and strain rate.
Media has not been submitted for this Presentation
Abstract: Rapid and precise determination of earthquake magnitude is of critical importance in earthquake early warning systems. A single station magnitude estimate that relies on a single feature of the waveform recorded near the epicenter, may have large uncertainties and may lead to false and missed alarms. In this study we show that, with state-of-the-art machine learning tools, a single station magnitude estimate can be much more accurate than those obtained with traditional methods. We trained an ensemble of convolutional networks using a large amount of historic data recorded in Japan and California to infer the magnitude directly from P waves of 3 s duration. The test results show that the networks estimate 94% of magnitudes in Japan within 1-unit difference from catalog ones and 99% in California. The trained networks require minimal computational effort when operating online and can be conveniently transferred to other regions even with a short history of seismic instrumentation.
Media has not been submitted for this Presentation
Abstract: The characterization of high-frequency (>10 Hz) ground motions for hard rock sites in Eastern North America is a critical seismic response issue for major infrastructure, particularly nuclear power plants. The diminution of amplitudes with increasing frequency is modeled using kappa (Anderson and Hough, 1984 BSSA), which is a measure of the slope of amplitude decay at high frequencies in the spectral domain. This study examines kappa and its variability for 7 hard rock sites near Charlevoix, Quebec (Canada), using hundreds of recording from earthquakes of M>3 within 150 km. Kappa values are compared to bedrock velocities measured at the recording stations to gain insight into the relationship between kappa and physical rock properties. We also examine whether there is evidence of source or path effects on kappa.
Media has not been submitted for this Presentation
Abstract: The seismic and acoustic activity of the Colima Volcano is studied and monitored by the Seismic Network of the State of Colima (RESCO), which belongs to the University of Colima. Nine broadband stations around the volcano record seismic activity in real time. Additionally, an acoustic sensor installed by the UNAM near the seismic station of Montegrande recorded the most important explosions in January and February 2017. The seismic-acoustic monitoring includes the location of volcano-tectonic events (VTs), explosions, rockfalls, count of events automatically with Hidden Markov Models, evaluation of the seismic and acoustic energy of the explosions, RSEM, etc. In recent years (2013-present), the activity of the volcano has been intense, lava domes growths, lava flows, lava dome collapses, as well as moderate to large explosions. Particularly the explosions in 2017, in their acoustic signals had a characteristic waveform of ‘N’. The seismic energy observed in the explosions had values above 1e + 8 Joules, while that the acoustic energy only presented 5 explosions that exceeded this value. With these values the ratio between acoustic and seismic energy remained mostly less than 1. This indicates that the seismic-acoustic source is probably generated in a long and narrow conduit, likewise, the volcanic columns were charged with ash according to the models of Johnson and Aster (2005). Similarly, the time between the arrival of the seismic wave and the acoustic wave is observed between 15.6 to 19.6 sec, which tells us that there are a small variation in the depth of the source. These explosions are lower in the seismic energy released compared to the 2005 activity.
Media has not been submitted for this Presentation
Abstract: Stress drop is an important ground-motion parameter as it controls the relative amount of high-frequency energy. We use the Empirical Green’s function (EGF) method to determine the stress drops of three earthquakes in Alberta, Canada, linked to hydraulic fracturing, with moment magnitudes (M) greater than 4. The EGF method is based on taking the spectral ratio of the ground motions from the M>4 earthquakes to those of smaller collocated earthquakes; path and site effects cancel in the spectral division, leaving the source spectral ratio. To ensure the smaller earthquakes have similar focal mechanisms to the targeted earthquakes, cross-correlation is performed. The source spectra can then be fit to Brune’s source model, and the corner frequencies and stress drops estimated for the three M>4 events. Results from this study will be used to compare source properties of induced events in western Canada to those in other regions. The results will also be used to compare classic stress drop estimates with stress parameter values inferred from stochastic models of high-frequency ground motions.
Media has not been submitted for this Presentation
