Electronic Supplement to
Triggered Seismic Events along the Eastern Denali Fault in Northwest Canada Following the 2012 M_w 7.8 Haida Gwaii, 2013 M_w 7.5 Craig, and Two M_w>8.5 Teleseismic Earthquakes

by Chastity Aiken, Jessica P. Zimmerman, Zhigang Peng, and Jacob I. Walter

Auxiliary Analysis, Observations, and Data

This electronic supplement describes the Rayleigh-wave polarization detection used for locating the short-duration events triggered by the Haida Gwaii, British Columbia, earthquake. It contains ten figures, three tables, and two animations (referenced here and/or in the main text).

Rayleigh-Wave Detection Analysis

We use a Rayleigh-wave polarization analysis (e.g., Stachnik et al., 2012) to determine the nature of the triggered waves. It utilizes the Rayleigh-wave property of retrograde elliptical particle motion (e.g., Lay and Wallace, 1995) when the horizontal components are aligned parallel to the seismic source. In our case of an unknown receiver-to-source direction, Hilbert-transformed vertical component seismograms are correlated with rotated horizontal seismograms to directions parallel (radial) and perpendicular (transverse) to a hypothetical seismic source. This horizontal rotation is based on a right-handed coordinate system, with vertical up, radial pointing to the source, and the right face of the transverse component along the positive radial component. Here, we test all possible source–receiver azimuths (j = 1:360). The normalized correlation coefficient reaches its maximum for the azimuth that is plausibly the actual receiver-to-source azimuth (Stachnik et al., 2012). Ideally, the coefficient is nearly 1 for surface waves; the method provides a means of evaluating whether ambiguous seismic signals are indeed surface waves and quantifies their approximate azimuth relative to the station.

We examine the continuous time series around the triggered events at station YUK3. We apply a 5 Hz high-pass filter to each station’s seismograms and perform the analysis described above in windows of 2 s length that progress through the data with a 1 s time step (effectively a 50% overlap). At each time step, we record the maximum correlation coefficient and the corresponding back azimuth. When the correlation coefficient crosses a threshold of 0.45, we mark a red dot on the bottom panel (Figure S5). The moving-window polarization analysis for YUK3 indicates a southerly source during the discrete signals, with 180°–200° directional azimuth (from north). The collapse of azimuths near a single value also suggests a repeating or nearby source for the triggered events. The same analysis on the YUK2 seismograms was inconclusive, perhaps due to complex site effects at the installation site.

Because the repeating events arrive at YUK3 first, we can make a crude estimation of the source location(s) relative to the two stations using simple geometry and the observed 180°–200° directional azimuth from the Rayleigh-wave polarization analysis (Figure S6c). Based on the repeating event arrival times, the distance between station YUK2 and the source must be longer than the distance between station YUK3 and the same source by a distance Δx, such that

in which x, x₁, and x₂ are the longitudinal coordinates and y, y₁, and y₂ are the latitudinal coordinates of the source, station YUK2, and station YUK3, respectively. The distance Δx can be approximated using the ~3.5 s time delay (Δt) between the stations (moveout) and an assumed Rayleigh-wave nominal group velocity (v) of 3.1 km/s, because the polarization analysis suggests the observed short-duration signals are mostly Rayleigh waves. This gives Δx = v − (Δt) = 11 km. We then performed a grid search for all possible (x, y) pairs that result in Δx = 11 km and use our range of back azimuths (180°–200°) from Rayleigh-wave polarization analysis to find all possible short-duration event locations (Figure S6a).

Given the paucity of stations in the area, it is impossible to ascertain an accurate location for the triggered short-duration events. However, our analysis suggests that the source is south-southwest of station YUK3. The Klutlan Glacier terminus is near this point. The waveforms have similar characteristics to glacial tectonic sources (i.e., icequakes) detected in Antarctica that were also triggered by Rayleigh waves of distant earthquakes (Peng et al., 2014). Furthermore, icequakes are routinely recorded in this area, though these have not been studied in detail (West, 2014). In addition to a possible glacial source, significant ambient tectonic seismicity follows the trace of the Toschunda–Duke River fault. Thus, the data may indicate two possible sources: small, shallow, tectonic earthquakes and/or icequake events. Deducing which type of events actually occurred would require more stations in the region.

---

Tables

Table S1. Moderate-size events occurring on or near the eastern Denali Fault since 1995. Focal mechanisms are from the Geological Survey of Canada centroid moment tensor (CMT) catalog (Kao et al., 2012) and the Global (formerly Harvard) CMT catalog (Dwiewonski et al., 1981; Ekström et al., 2012).

Table S2. Event information of examined earthquakes. Included are date, time, location, magnitude, epicentral distance, back azimuth, and peak ground velocities for all events. The events that triggered tremor near the eastern Denali fault are indicated.

Table S3. Velocity model for the northern Canadian Cordillera (from Meighan et al., 2013).

---

Figures

Figure S1. Spectrogram of tremor triggered by the Craig earthquake as recorded by station HYT. (a) Comparison between the broadband and high-pass filtered >20 Hz vertical component five hours before and after the mainshock origin time. Red lines delineate the times around the teleseismic waves of the mainshock. (b) Instrument-corrected broadband transverse and vertical components. (c) High-pass-filtered >20 Hz vertical component. (d) Spectrogram. The dashed line indicates the 20 Hz corner frequency used in the high-pass filter to remove the P and coda waves from the distant mainshock.

Figure S2. Evidence that no dominant tremor signals triggered by the 2012 Indian Ocean earthquake (EQ) can be seen at higher frequency bands. The spectrogram of the transverse component at station HYT with its broadband velocity transverse component is plotted on top for reference. The Love wave of the Indian Ocean mainshock, as well as the P and S waves of the M_w 3.9 Nenana earthquake, are marked.

Figure S3. S-wave attenuation of the M_w 3.9 Nenana earthquake. S-wave amplitudes are measured from high-pass-filtered 5 Hz north (N) and east (E) components at multiple stations along the event’s ray path to the study region. Network and station names are marked by distance with solid black lines. The blue lines are empirical attenuation relationships accounting for event magnitude and distance, with the dashed lines marking 95% confidence. The empirical attenuation relationship and constants follow those of van der Elst and Brodsky (2010).

Figure S4. Evidence that the M_w 8.2 Indian Ocean aftershock did not trigger tremor on the eastern Denali fault. (a) 1–10 Hz band-pass filtered waveforms at multiple stations. The epicentral distance from the station to the tremor source is indicated. (b) Transverse- and vertical-component velocity (V) and displacement (D) waveforms recorded at the HYT station. In (b), the black vertical bar indicates the scale for broadband velocity waveforms.

Figure S5. Rayleigh wave polarization analysis for station YUK3. (a) High-passed (5 Hz) vertical component (blue), east component (green), and north component (red) seismograms. (b) Maximum correlation coefficient at the corresponding azimuth in (c). (c) Red filled circles indicate sample points when the correlation coefficient is greater than 0.45. Higher correlation coefficients occur during the triggered bursts and indicate a consistent source azimuth.

Figure S6. Locations of stations and seismicity, with a schematic of source determination. (a) Satellite map, including YUK2 and YUK3 stations and ANSS seismicity (>m_L 1.0 during 2010–2013). The yellow area marks possible locations of short-duration events. (b) Longitude versus depth plot, indicating relatively shallow depths for seismicity. The linear trend of seismicity indicates the approximate trace of the Duke River fault. (c) Steps in determining approximate location of the source.

Figure S7. Lack of seismic activity prior to the arrival of surface waves from the Haida Gwaii earthquake. Each trace is an average 3-component, high-pass 20 Hz-filtered envelope (log₁₀ scale), with the exception of station YUK4. As stated in the main text, station YUK4 has significant instrumental noise on the horizontal components. The dashed line marks the origin time of the Haida Gwaii mainshock.

Figure S8. Static stress changes caused by the Haida Gwaii earthquake. Faults and plate boundaries are shown as black lines. Gray triangles indicate AK network stations, and black triangles indicate the CN network stations. The main reference station HYT is marked. We utilized source information for the mainshock from Lay et al. (2013). Receiver information is based on a local earthquake from Kao et al. (2012); the focal mechanism is marked. We show the static stress changes at 25 km depth; this depth is the assumed tremor nucleation depth. There is a change of −0.03 kPa static stress at the HYT station.

Figure S9. Static stress changes caused by the Craig, Alaska, earthquake. Symbols and notation are the same as in Figure S8. We utilized source information for the mainshock from Lay et al. (2013). There is a change of 0.01 kPa in static stress at the HYT station.

Figure S10. Evidence that the 27 February 2010 M_w 8.8 Maule, Chile, earthquake did not trigger near the aftershock zone of the 2002 Denali fault earthquake. Symbols and notations are the same as in Figure 4, with the exception of (a) being 5 Hz high-pass filtered waveforms.

---

Animations

Animation S1 [h.264-encoded MPEG-4 video; ~185 KB]. Tremor around the eastern Denali fault, Yukon, Canada, recorded at station HYT triggered by the 28 October 2012 M_w 7.8 Haida Gwaii earthquake. (Top) Broadband transverse-component seismogram recorded at station HYT. (Middle) A 20 Hz high-pass-filtered transverse-component seismogram showing the P wave of the distant mainshock and triggered tremor signals. (Bottom) Spectrogram of the transverse-component seismogram. A 5 Hz high-pass filter is applied to remove long-period signals before computing the spectrogram (Peng et al., 2011). The sound is generated by speeding up the seismic data by 100 times (Kilb et al., 2012).

Animation S2 [h.264-encoded MPEG-4 video; ~193 KB]. High-frequency bursts in Yukon, Canada, recorded at station YUK2, triggered by the 28 October 2012 M_w 7.8 Haida Gwaii earthquake. The descriptions are the same as in Animation S1.

---

[ Back ]