Imaging the rupture process of the 10 January 2018 M_W7.5 Swan island, Honduras earthquake

1.   Introduction

On 10 January 2018 (UTC), a M_W7.5 earthquake occurred in the east of Great Swan island, Honduras (location 17.483°N, 83.520°W, depth 19.0 km), according to the U.S. Geological Survey (USGS). Nine other earthquakes of M6 or larger have occurred within 400 km of the January 10, 2018 event over the preceding century.

Previous strong earthquakes in this region include a destructive M7.5 Guatemala earthquake on 4 February 1976, which resulted in more than 23,000 fatalities and occurred 650 km west-southwest of the 2018 event. In May 2009, a M7.3 earthquake occurred approximately 300 km west of the 2018 event. The 2009 earthquake (which was much closer to land than the 2018 event) resulted in 7 fatalities, 40 injuries and 130 buildings being damaged or destroyed (according to the USGS).

The dominant tectonic element in the source areas is the Cayman Trough, an 80-km-wide, 1000-km-long and 5-km-deep ocean basin formed at the north-south-trending Mid-Cayman spreading center (Brinka et al., 2002; Mann et al., 1991) (Figure 1). The Cayman Trough originated as a pull-apart basin in early to middle Eocene time and is currently spreading along the Mid-Cayman spreading center at about 15 mm/a (Mann et al., 1991; Rosencrantz and Sclater, 1986). The Cayman Trough is a region of oceanic crust, and is the transform plate boundary between the North American and Caribbean plates. It was formed at the Mid-Cayman Rise (MCR) oceanic spreading center. The linear ridges and rough topography of the central Cayman Trough are typical of active spreading centers (Brinka et al., 2002). The trough is bounded on the south by the Nicaragua Rise, a shallow submerged platform and on the north by the Cayman Ridge, a narrow sliver that was probably part of the Nicaragua Rise before the opening of the trough (Brinka et al., 2002).

Figure  1.  Seismicity in the source region of the 2018 M_W7.5 Swan island, Honduras earthquake. The red star represents the location of the epicenter determined by the USGS. The open and yellow circles show earthquakes that occurred before the mainshock and within three months following the mainshock, respectively. The focal mechanism is determined by the GCMT

DownLoad: Full-Size Img PowerPoint

The south wall of the Cayman Trough includes and is defined by active strike-slip (transform) faulting, to the east of the spreading axis as well as to the west (Rosencrantz and Mann, 1991). The Cayman Trough contains a ~100-km-long slow-spreading ridge between two offset transform faults, the Swan islands fault and the Oriente fault (Brinka et al., 2002; Holcombe et al., 1973). To the west of the MCR, the Cayman Trough is bounded by the left-lateral strike-slip Swan island fault. To the east of the MCR, the Cayman Trough is bounded on the north by the Oriente fault. On the basis of bathymetric and seismic information, Holcombe et al. (1973) showed that the rugged topography central to the Trough was symmetrical to a north-south trending, deep, flat-floored valley at longitude 81°40’W and that this valley marked the location of a spreading center connecting the two transform faults (Rosencrantz and Sclater, 1986). The western fault, known as the Swan islands transform fault, connects the southern end of the Cayman spreading axis with the Motagua fault in Central America (Holcombe et al., 1973; Rosencrantz and Mann, 1991). The three earthquakes mentioned above occurred on the east segment of the left-lateral strike-slip Swan fault, which is moving with a speed of 18.4 mm/a (Rogers and Mann, 2007).

This earthquake is an erect left-lateral strike-slip faulting, with a magnitude of M_W7.5 and a focal depth of about 10 km (according to the USGS). The rupture does not show obvious directivity and the slip distribution on the fault plane is rather concentrated (Wang et al., 2018). The ruptured slip is compact, with a maximum slip of 1.9 m.

Here we apply a back-projection method to trace the rupture migration of the 2018 M_W7.5 earthquake using seismic data recorded in Alaska, US. The results show that at least three faults were ruptured during the earthquake. The rupture speed of the ruptured fault with large energy radiation and longest source length, shows a fast rupture speed of ~5 km/s, which is faster than local shear wave velocity of ~4 km/s based on the Crust 1.0 (http://igppweb.ucsd.edu/~gabi/rem.html, accessed February 2020).

2.   Data and method

Follow Wang et al. (2016), we back-project the source process of the 2018 Honduras earthquake, and estimate the rupture speed. Since the success of exploring source extent and duration of the 2004 M9.0 Sumatra earthquake (Ishii et al., 2005; Kruger and Ohrnberger, 2005), the back-projection method has been widely used in probing source characteristics and resolving complex source patterns (Meng et al., 2011; Wang et al., 2012; Yao et al., 2012; Fan and Shearer, 2017).

We utilize data recorded at 330 seismic stations in Alaska to back-project the source process of the 2018 M_W7.5 Honduras earthquake. The seismic stations were located at distances of 45–78 degrees to the epicenter, with azimuths of 321–340 degrees (Figure 2a). The observed waveforms are aligned with P arrivals (Figure 2b), showing high similarities among each other. The high similarities of the observed waveforms suggest fewer site effects among stations, and ensure high accuracy of back-projection results. Since all the waveforms are similar, the differences in travel times caused by source migrations will be accurately determined by comparing the stacked energies of the tested grids in the source region. This is the key why regional arrays usually get better results than those determined by global arrays in back-projection. Due to large differences of structure heterogeneity in ray paths and directivity effect, global observations show considerable differences at high-frequency bands (say higher than 0.5 Hz).

Figure  2.  (a) Locations of seismic stations (red triangle) in Alaska used for back-projection. Solid and dashed lines show the epicentral distances and strikes of the fault planes determined by the GCMT. Black star shows the epicenter determined by the USGS. (b) Representative raw waveforms recorded at seismic stations that are shown in (a). The waveforms are aligned by P onset (blue line)

DownLoad: Full-Size Img PowerPoint

The observed waveforms are first band pass filtered in the band of 0.01–0.2 Hz using a two-pole Butterworth filter. We cross correlate a time window of 20 s (starting from the P onset) with a offset of 5 s, and apply these obtained station corrections to the waveforms. Then we cross correlate a second time window (6 s) for the higher frequency bands of 0.5 Hz to 2.0 Hz with a small offset (0.5 s). We estimate the station corrections in two steps to avoid cycle slip in the high band passed waveforms that are utilized in back-projection. We set up a grid of 100×40 points at a focal depth of 19 km according to the USGS, and use a shifting time window of 10 s that is offset by 1 s. The velocity structure model used in calculating travel times between grid points and seismic stations is IASPEI 1991 (Kennett and Engdahl, 1991).

3.   Results

Our results show that the earthquake initiated from the Swan island fault, then unilaterally propagated to the east (Figure 3). About 10 s after the origin time, the rupture propagated along an SW fault that is obliquely crossing the general striking direction of the Swan island fault for another 10 s. 20 s after the origin time, the rupture migrated to the east with a duration of ~20 s. 40 s after the origin time, the stacked energies decreased, and the locations became scattered, suggesting that the end of the main rupture and/or lower resolution in back-projection in these time windows (Figure 4a). The total source duration is around 40 s (Figure 4b).

Figure  3.  Snap shots of the rupture propagation. Colors show values of the squared amplitude stacks at each grid point. The initial arrivals were constrained to originate from the epicenter indicated by the red star (determined by the USGS). The circles, which sizes indicate the sizes of the earthquakes, represent the earthquakes that occurred in this region since 1950

DownLoad: Full-Size Img PowerPoint

Figure  4.  (a) Timing (color of circles) and amplitude (size of circles) for the stack with the maximum correlation at each time step (1 s) shown in map view. Black star and triangles indicate the reference points used to estimate rupture speeds in (d). Open and gray filled circles represent the historical seismicity and aftershocks in three months, respectively. (b) The normalized value for the maximum amplitudes in each time window as a function of time. (c) Sketch of the ruptured faults. (d) Rupture speeds of the earthquake in each stage. The distances are estimated from the epicenter, reference points for the stage 1 and stages 2 and 3, respectively

DownLoad: Full-Size Img PowerPoint

Because the rupture was not simply over a straight fault, the rupture speed of this earthquake cannot be estimated by a straight distance to the epicenter for the entire process. At least three faults were ruptured during this earthquake. Thus, we set three reference points at the starting place of three ruptured faults, to better estimate the rupture speeds of the earthquake at the three stages (Figure 4c).

In the first 10 s, the rupture speed is as low as 1 km/s, similar to the initial rupture speeds of many other large earthquakes (Wang et al., 2016). In the second stage, that is, 10–20 s after the origin time, the rupture speed is around 2–3 km/s depending on assumed fault orientation and rupture starting points (Figure 4d). Because of the relatively short length of the rupture fault (~20 km) in this stage, and considering a 10–20 km uncertainty in locations of the local maximums, the uncertainty of the estimated rupture speed in this stage is relatively large. About 20 s after the origin time, the rupture propagated to the west with a fast rupture speed of 5.3 km/s. The radiated energy reached a peak at this stage. Therefore, the observed rupture locations are the most trustable. Considering that the shear wave velocity is 4.05 km/s at the focal depth (19 km) according to the Crust 1.0 (Laske et al., 2013), the rupture speed is faster than local shear wave speed, supporting a previous study that suggests that supershear rupture is common in large strike-slip earthquakes (Wang et al., 2016).

The complicated rupture pattern is evidenced by the global bathymetry and topography of SRTM 15+ (Tozer et al., 2019) and the original satellite scenes of Google Earth. In both the SRTM 15+ and the Google Earth, several NW faults that are obliquely crossing the Swan island fault are obvious in the source region. The SW fault in stage 3 is also well outlined by the bathymetric offsets (Figure 5). Although we couldn’t verify the ruptured faults on site, the bathymetric data supports the complicated rupture pattern we determined from the back-projection.

Figure  5.  Bathymetry and topography of SRTM 15+ in the source region (Tozer et al., 2019). Red lines sketch the rupture pattern of the 2018 M_W7.5 Honduras earthquake

DownLoad: Full-Size Img PowerPoint

4.   Discussion

Our results show a rupture model for the 2018 M_W 7.5 Honduras earthquake with at least three ruptured faults. The stages 1 and 3 are along the general strike direction of the Swan island fault, and the stage 2 is obliquely crossing the Swan island fault. The strike direction of the stage 2 is similar to the layout of the aftershocks that occurred in the east of the mainshock epicenter. As more seismic stations have been deployed and new methods are developed, such complexities in large earthquakes are more and more frequently observed in recent cases, such as the 2012 M_W8.6 Off-Sumatra (Satriano et al., 2012; Wang et al., 2012; Yue et al., 2012; Wei et al., 2013) and the 2016 M_W7.8 New Zealand earthquakes (Hamling et al., 2017; Kaiser et al., 2017; Zhang et al., 2017; Wang et al., 2018). Even for a M_W5.5 earthquake, with dense seismic observations, we observed rupture over several faults during the earthquake sequence (Li et al., 2019).

There were few oceanic strike-slip earthquakes identified as supershear rupture speed, mostly due to the space distributions of seismic observations in ocean. Only the 2012 M_W8.6 and M_W8.2 Off-Sumatra earthquakes and the 5 January 2013 Craig, Alaska earthquake (M_W7.5) showed evidences of supershear ruptures (Wang et al., 2012; Meng et al., 2012; Yue et al., 2013). The 2012 M_W8.6 Indo-Australia intraplate oceanic earthquake was the largest strike-slip event yet observed. The rupture process of the Indo-Australia event was very complex, involving rupture of no less than four segments. The 5 January 2013 M_W7.5 Craig, Alaska earthquake was the first oceanic interplate event reported involved supershear rupture, and it was the first offshore event for which multiple data set independently indicated supershear rupture. Since those earthquakes were with large magnitudes and were ruptured in relatively long faults, resolving their rupture speeds is relatively easier. With more and more dense regional seismic arrays and technological progress, we might be able to reveal more details of rupture patterns of oceanic earthquakes with lower magnitudes in the near future.

More detailed rupture patterns help better estimate rupture parameters, such as rupture speed and source extent. Rupture speed, as a key parameter that controls earthquake strong motion, is difficult to estimate due to lack of near field observations. However, knowing rupture speed is important in rupture dynamics and earthquake engineering. When rupture speed is as slow as several kilometers per hour (slow earthquakes), they don’t cause any shaking damages (Figure 6). Tsunami earthquakes usually have rupture speeds of ~1 km/s, causing small to moderate shaking damages. When the rupture speed reaches a higher level, saying faster than local shear wave velocity, it will generate Mach waves, creating widespread and strong damages if there is a large population around the source area. Such supershear ruptures were only observed in long and straight faults such as the Kunlun fault and the Denali fault, partly because long distance and simple fault pattern make the rupture easily estimated. For strike-slip earthquakes with complex rupture patterns, correctly estimating rupture speed remains challenging. Therefore, many rupture speeds of moderate to large earthquakes derived from simple fault patterns exist large uncertainties, and worth further investigations.

Figure  6.  Schematic diagram of the relationship between rupture speed and intensity of shaking damages of earthquakes

DownLoad: Full-Size Img PowerPoint

5.   Conclusions

We analyze the rupture process of the 1 January 2018 M_W7.5 Honduras earthquake by back-projecting the seismic data recorded at Alaska stations. Our results show that at least three faults were ruptured during the earthquake. The earthquake started from an SW striking fault in the first 10 s, then jumped to a NW striking fault for another 10 s. 20 s after the origin time, the rupture propagated along an SW striking fault for 20 s. The rupture speeds for the first two stages were 1 km/s and 1–3 km/s, respectively. The longest segment, stage 3, showed a clear and fast supershear rupture speed, as fast as 5.3 km/s. There are many oceanic strike-slip earthquakes, only a few of them have been verified as ruptured supershear. This study, together with recent literature, suggests that supershear rupture may commonly occur in strike-slip earthquakes.

Acknowledgments

This work is supported by the National Key R&D Program of China (No. 2018YFC0603500), Programme on Global Change and Air-Sea Interaction (GASI-GEOGE-02), NSFC (Nos. 41474050, 41425012, 41874062 and 41922025), 111 project (No. BP0719022), and the Fundamental Research Funds for the Central Universities, China University of Geosciences (Wuhan) CUG170602 (D.W.). The moment tensor solutions of the mainshock are available at U.S. Geological Survey (USGS, https://earthquake.usgs.gov/earthquakes/eventpage/us1000c2zy/executive, accessed February 2020), Global Centroid Moment Tensor (GCMT, https://www.globalcmt.org/CMTsearch.html, accessed February 2020). Figures are generated using the Generic Mapping Tools (Wessel et al., 2013).