For earthquakes in remote areas, comparisons of low-pass-filtered seismograms to calculated synthetics potentially can provide more accurate depths than are possible by fitting travel-time picks. This is because the waveform matching must fit filtered waveforms for body phases, leaving the source traveling both upward and downward, as well as matching the amplitudes of surface waves.
The waveform-matching algorithm applied in this investigation was developed at St. Louis University (SLU) to routinely determine source parameters for moderate-magnitude (3.8<Mw<5.5) earthquakes in North America (Herrmann et al., 2011). Currently the SLU catalog includes results for more than 850 events that have occurred since 1962.
For the 21 September 2013 13:16 UTC earthquake, observed ground velocities at 25 stations (Fig. S1) were band-pass filtered in the 0.02–0.08 Hz band, using a cascade of a three-pole high-pass filter and three-pole low-pass filters. All waveforms were cut in a time window 30 s before and 180 s after the model predicted P first-arrival time. A time-shifting procedure was used for minor adjustments to account for location, origin time, and velocity model errors.
Initial waveform matches were determined using the western United States (WUS) regional velocity model. Although developed for matching waveforms in the western part of the country, this model has been applied to earthquakes elsewhere in the United States. After an initial search in the 1–29 km depth range did not lead to a definitive source depth, the search was extended to 100 km. The best fit for a shear-dislocation source was for a depth of 76 km, an Mw of 4.77, and strike, dip, and rake of 65°, 65°, and 30° and 321°, 63°, and 152° for the two nodal planes. Figure 5 of the main article displays the fit parameter as a result of the grid search over strike, dip, and rake as a function of source depth. The fits are very good (Fig. S2) and indicate a source depth in the upper mantle.
The 76 km depth for the 21 September 2013 earthquake places it in the half-space of the WUS velocity model. Because that model was constructed by adjusting a regional earthquake location model to provide a fit to observed surface-wave group velocities, important questions arise as to the proper velocities to use in the lower crust and upper mantle and of the effect these changes might have with respect to the source parameters of this Wyoming earthquake.
To address this issue, we determined source parameters using two additional regional velocity models (Fig. S3). The first was a regional velocity model, here named LIN, developed by F.-C. Lin at the University of Utah (Lin et al., 2012). We also created a new model that fit group and phase velocity dispersion of the Love and Rayleigh fundamental modes in the source region, as well as teleseismic P-wave receiver functions at USArray station K19A (see Fig. 2 in the main article). Except for some vertical offsets (which would map into time shifts in the waveform fitting process), the Rayleigh and Love group velocity curves for all three models (Fig. S4) have similar shapes in the critical 10−50 s period range. The K19A model differs significantly in the upper mantle because of the constraints on the long-period dispersion data. The K19A model also does not have a sharp Moho because the receiver functions do not support such a feature, a conclusion similar to that of Gans (2011) and Gilbert (2012). We generated Green’s functions for the LIN and K19A models and reran the grid search for source parameters using the same filter band.
For waveform matches using the LIN and K19A models, the best-fitting focal mechanisms and source depth were only slightly different than determined for the WUS model. In particular, the best fits for LIN and K19A models were for depths of 75 and 72 km, respectively. The shallower depth using the K19A model is attributable to lower model velocities in the upper mantle. All three models constrain the source to be beneath the crust.
Figure S1. Broadband seismograph stations providing data for waveform matching in this study. These stations (red circles) were used to determine source parameters of the 21 September 2013 13:16 UTC earthquake (see Figs. 5 and 7 in the main article and Fig. S2). The yellow circle is the earthquake epicenter.
Figure S2. Observed and predicted ground velocities for the 21 September 2013 13:16 UTC earthquake. The 210 s segments of observed waveforms are plotted in red; blue waveforms are as predicted for the best-fit source at 76 km depth using the WUS model. Z, R, and T are the vertical, radial, and transverse components of ground velocity as measured by the stations indicated on the right. Station abbreviations listed to the right are from the International Seismic Station Registry, which is maintained by the International Seismological Centre.
Figure S3. Shear velocities versus depth for three regional models (WUS, K19A, and LIN) used to calculate synthetic waveforms for the 21 September 2013 earthquake.
Figure S4. Comparison of observed (red dots) Love fundamental-mode group velocity U to predicted values (continuous lines) determined for three different regional models (WUS, K19A, and LIN; see Fig. S3) used to calculate synthetics for the 21 September 2013 earthquake.
Gans, C. R. (2011). Investigations of the crust and upper mantle of modern and ancient subduction zones, using Pn tomography and seismic receiver functions, Ph.D. Dissertation, Tucson campus, University of Arizona.
Gilbert, H. (2012). Crustal structure and signatures of recent tectonism influenced by ancient terranes in the western United States, Geosphere 8, 147–158.
Herrmann, R. B., H. Benz, and C. J. Ammon (2011). Monitoring the earthquake source process in North America, Bull. Seismol. Soc.America 101, 2609–2625.
Lin, F.-C., B. Schmandt, and V. Tsai (2012). Joint inversion of Rayleigh wave phase velocity and ellipticity using USArray: Constraining velocity and density structure in the upper crust, Geophys. Res. Lett. 39, L12303, doi: 10.1029/2012GL052196.
[ Back ]
