This electronic supplement contains three tables of source information including location, time, and magnitude or yield if available, and additional figures showing example detections using the Chael/Selby method are presented as well with added descriptions.
Table S1. Source information for earthquake events used in this study. Hypocenters were relocated using a 1D velocity model by O’Rourke et al. (2016) and have typical errors of ~1–2 km.
Table S2. Source information for mine blast events used in this study. Blasts with known source information are located at the mine where they were detonated. Blasts with unknown source information were located using a 1D velocity model by O'Rourke et al. (2016) and have larger (~10 km) location errors.
Table S3. Source information for active-source borehole shot events used in this study.
Figure S1. Example of a nondetection using the Chael/Selby method for a mining blast on a station in the Powder River basin at 86 km distance. Event origin is at 60 s. (a) Black trace shows the Hilbert-transformed vertical component, and red trace shows the radial component. Traces are filtered from 0.4 to 0.8 Hz. The expected Rg arrival is between ~115 and 135 s (1.5 and 1.1 km/s, respectively). Retrograde motion would produce peaks that are in phase with each other, whereas prograde motion produces out-of-phase peaks, as observed here. (b) Cross-correlation values for each time step, using a 6-s scanning window. (c) Penalty values based on the mismatch between expected back azimuth and the back azimuth with the best correlation. Good matches have low penalties, and poor matches are highly penalized. (d) The result of dividing the correlation value by the penalty results in peaks when arrivals are correlating at the correct azimuth. The prograde motion observed in the top panel has good correlation values (b), but only at back azimuths 180° off from the known back azimuth, which receives high penalties (c). Figure 6 of the main article illustrates the change from prograde to retrograde as the arrival moves from the basin into the mountains.
Figure S2. Example of multiple detector peaks using the Chael/Selby method for a mining blast on a station in Bighorn Mountains at 170 km distance. Figure panels are the same as those used in Figure S1. Multiple detections like this present problems for identifying the Rg arrival, especially automatically. We performed the process manually, visually confirming that a detection peak matched with an arrival in the expected velocity range. In this example, the high-amplitude signal at ~160 s in the seismogram corresponds to a group velocity of around 1.7 km/s, and the later arrival at ~215 s has a velocity of ~1.1 km/s. The first arrival is likely Rg, because this station is located in the mountain bedrock and we would expect Rg to arrive at a slightly higher group velocity than in the basin, which often had group velocities in the ~1.4 km/s range. The later arrival may be a scattered or multipathed arrival.
An important observation from this example is the presence of not just the two higher-amplitude arrivals but the triggers before and after as well. For example, (d) shows a large detection at ~240 s, but the seismograms show only a very small but well-correlated signal at that time. Similar effects are observed at 140 and 280 s. In fact, the correlation is high throughout the record, possibly due to direct and scattered higher-mode surface waves before the Rg and scattered and multipathed Rg later. Because amplitude does not play into the trigger criteria, all highly correlated signals and noise get equal weight in the detector as long as they correlate at the correct back azimuth.
O’Rourke, C. T., A. F. Sheehan, E. A. Erslev, and M. L. Anderson (2016). Small magnitude earthquakes in north-central Wyoming recorded during the Bighorn Arch Seismic Experiment, Bull. Seismol. Soc. Am. 106, 281–288, doi: 10.1785/0120150114.
[ Back ]
