Mixed earthquake signals in the South Sandwich Islands

Aftershocks of the multiple large-magnitude earthquakes that occurred recently in the South Sandwich Islands suggest shallow rupture to the trench.

By Judith Hubbard, Ph.D., Earth Observatory of Singapore, Nanyang Technological University, Singapore (@JudithGeology)

Citation: Hubbard, J., 2021, Mixed earthquake signals in the South Sandwich Islands, Temblor, http://doi.org/10.32858/temblor.202

On Aug. 12, 2021, the U.S. Geological Survey reported that a magnitude-7.5 earthquake had occurred in the South Sandwich Islands in the south Atlantic Ocean, more than 2,500 kilometers east of the southern tip of South America (Figure 1). This event led to a small flurry of interest from earthquake scientists, but was generally considered to be low risk with respect to both ground shaking and tsunami hazards because of both its distance from population centers and depth of 60 kilometers.

Map of South Sandwich Islands region
Figure 1: (Top) Regional setting of the Sandwich Plate, where the August 12, 2021 magnitude-7.5 and magnitude-8.1 earthquakes occurred. Arrows show relative plate velocities at block boundaries. White box indicates extents of maps in Figures 2 and 3. Lower inset shows map location on a lower hemisphere projection of the globe. Map produced using tectoplot (Bradley, 2021) with assistance from Bryan Low Kai Sheng.


Consternation among the earthquake community grew as the aftershocks started to roll in, illuminating a 400-kilometer-long length of the subduction zone — much larger than the typical seismic region for a magnitude-7.5 earthquake. More than a day later, the USGS reported that in fact, the magnitude-7.5 earthquake had been hiding the signal of a much larger mainshock — a magnitude-8.1 event — that initiated about 2.5 minutes after the first quake. This second event released nearly eight times the energy of the first, but the seismic waves arrived at global seismic stations while they were still shaking from the first event, preventing automatic earthquake detection algorithms from working as intended.

Although this magnitude-8.1 earthquake still did not cause damage, a tsunami should have been a much greater concern. However, by the time this earthquake was discovered, the tsunami had already propagated through both the Atlantic and Indian Oceans, a process that took nearly a day. The tsunami was detected at tide gauges as far away as the Azores Islands in the North Atlantic (20 centimeters, 14 hours after the rupture) and eastern Madagascar (6 centimeters, 14 hours after the rupture) (personal communication with Amir Salaree). The tsunami was impressive for its wide yet nondestructive range.

The overlapping seismic waves have made it difficult for seismologists to accurately assess the earthquake. Most critically, the reported depth and mechanism of the event(s) vary widely according to different organizations that monitor earthquakes. Understanding this variability and why it matters requires an examination of the tectonic and geographic setting.

Tectonic setting

The earthquake occurred on the 700-kilometer-long bow-shaped South Sandwich Subduction Zone (Figures 1 and 2). Here, the South American Plate is subducting westward at a rate of about seven centimeters per year below the Sandwich Plate. The Sandwich Plate is a small piece of young oceanic crust, forming only 300 kilometers to the west at the East Scotia Ridge. On its southern edge, the Sandwich Plate borders the Antarctic Plate across a short transform fault. The plate is marked by a series of regular islands, formed along the volcanic arc of the subduction zone; these volcanic islands (the South Sandwich Islands) give the region its name. They are remote and inhospitable, with no native population, and parts of the islands are permanently covered with ice. This setting helps to explain the lack of nearby seismic or geodetic observations: There are no nearby seismic or GNSS stations and very limited potential for satellite data due to the lack of exposed land.

Map of ocean area with ocean floor topography
Figure 2: (Top) Map of the tectonic setting with variously reported focal mechanisms (beachballs), which illustrate either the interpreted initiation point of the earthquake (origin method) or the interpreted center of moment release (centroid method). Colored contours represent depth of slab2.0 (Hayes, 2018). (Bottom) The five profiles are aligned along the 40-km depth contour of the slab, and show that the reported locations broadly vary from above to below the slab, and from deep to shallow. Map produced using tectoplot (Bradley, 2021).


Between two plates or within the slab?

At the subduction zone, the South American Plate dives beneath the Sandwich Plate. This creates both a fault at the boundary between the two plates — a megathrust — and a region at depth below the Sandwich Plate, where the subducting slab is sinking into the mantle. In a region like this, we expect various kinds of earthquakes, representing both slip along the megathrust and deformation of the subducting slab as it bends, heats and sinks.

A global map of subduction zones has been generated on the basis of both earthquakes and tomographic images of the mantle, which map the temperature of the mantle (subducting slabs are colder than the surrounding mantle rock). This map, called slab2.0, provides a baseline for understanding the relationship of new earthquakes to the overall setting (Hayes, 2018).

At the location of the magnitude-8.1 South Sandwich Islands earthquake (reported by the USGS), the slab2.0 puts the megathrust fault’s location about 25 kilometers below sea level. USGS reports an origin depth of 48.3 kilometers for the earthquake, which would place it below the megathrust fault and deeper within the slab, making it an intraslab (within slab) earthquake (Figure 2). At magnitude-8.1, the fault area must be hundreds of kilometers wide, raising the question: How would such a large fault form within the slab? Large earthquakes are not unusual on the megathrust because it is a major plate boundary, but the slab is a piece of relatively stable crust, and forming a fault of that scale is no minor task. Such earthquakes are rare, although not unheard of; a similar-scale earthquake broke the slab below Mexico in 2017 (Ye et al., 2017).

More careful investigation of the seismic records reveals that the already complex story is even more complicated. In addition to reporting the location where the earthquake started (the hypocenter), the USGS also reports depths for the event based on alternate methods that utilize the direction of movement of the rocks on either side of the fault. USGS reported highly variable depths for this event, ranging from 26.5 kilometers based on one method (centroid) to 70.5 kilometers based on another method (W-phase), which is preferred by USGS. As a further complication, other automatic earthquake catalogs report depths of 20 kilometers (reported by GCMT) and 10 kilometers (reported by GEOFON) (Figure 2). As of this writing, several weeks after the event, USGS notes on its event page that the location, depth, mechanism, and magnitude are all preliminary pending further investigation.

It therefore remains unclear if the mainshock was a megathrust event, accommodating slip between the South American and Sandwich plates, or an intraslab event, associated with break-up of the subducting slab.

Ways forward

The orientation of the fault as reported by seismic methods might provide a clue. For a megathrust event, the fault orientation should match the megathrust, whereas for an intraslab earthquake, a steeper rupture surface would be more common. However, reported fault orientations are also variable: GCMT and GFZ report that the fault is tilted either 14 degrees or 11 degrees from horizontal, respectively, generally matching the megathrust, whereas USGS reports variable dips, either 17 degrees in the opposite direction, or 43-54 degrees, depending on the method.

Further study of the seismic waves might allow seismologists to disentangle the mainshock from the foreshock to gain more clarity. For instance, by using the waves from the magnitude-7.5 foreshock that arrived prior to those of the mainshock, it may be possible to map out the slip associated with only the foreshock. That could provide a better understanding of the foreshock’s slip pattern, which could then be used to calculate how that event’s seismic signals should have been detected around the world. This calculated signal should match the early signal detected by seismometers globally, but would not match the later signal, which includes seismic waves from both the foreshock and the mainshock. By subtracting the calculated signal from the measured signal collected at each seismic station, it may be possible to reveal the signal produced exclusively by the mainshock (personal Communication with Stephen Hernandez).

Chain of islands surrounded by white ice
The South Sandwich Islands are inaccessible by ship part of the year due to sea ice, as shown by this satellite image. The inhospitable islands are a challenging place to make seismic measurements. Credit: LANCE/EOSDIS Rapid Response, NASA


Another potential lead could come from the tsunami: The size and pattern of a tsunami is sensitive to the depth and orientation of an earthquake’s slip. However, the distance of the event from tide gauges makes this challenging because tsunami height and time also depend on the detailed bathymetry between the source and the recorder. The bathymetry around the remote South Sandwich Islands is, unfortunately, not well constrained.

What more creative approaches might illuminate this event? One possibility is to examine the frequencies of the shaking: Intraslab events tend to produce higher-frequency shaking than megathrust events (e.g., Garcia et al., 2004; Van Daele et al., 2019). In addition, shallow ruptures of megathrusts sometimes have a characteristic pattern of slow rupture velocities and can be associated with larger-than-expected tsunamis (“tsunami earthquakes,” Kanamori, 1972). Searching for these kinds of characteristics might help with classifying the earthquake by type. Unfortunately, this approach is also based on the seismic signals, and therefore remains challenging.

Satellite image of cloud cover with a few tiny black dots
The South Sandwich Islands are often hidden by cloud cover, making satellite observations of ground movement nearly impossible. Credit: Jeff Schmaltz, LANCE/EOSDIS Rapid Response, NASA


Another possibility would be to use satellite observations of the region, but this won’t work either. The South Sandwich Islands likely moved as the blocks on either side of the fault shifted. A horizontal shift would be impossible to detect; horizontal shifts are measured by the “stretching” or “squeezing” of one part of the land surface relative to another, and the islands are so small that this kind of signal is below the resolution of the satellites. Vertical movement — likely subsidence — that could potentially impact the coastlines of the islands would be easier to detect. But the southern islands, which are closer to the earthquake, are largely glaciated, and their coastlines are obscured by both land and sea ice. It may be possible to calculate vertical change on the small lengths of exposed coastline to the north, but these point measurements would not be sufficient to extract earthquake parameters.

Another satellite-based approach would be to look at maps of gravity over time (personal communication with Soli Garcia). The GRACE-FO (Gravity Recovery and Climate Experiment Follow-On) satellite circles Earth, measuring the gravitational pull in different locations. Earth’s gravity field varies over time due to changes in ice sheets, water distribution, magma movement, and earthquakes. The South Sandwich Islands earthquake involved a block movement of several meters and must have modified the gravitational field in the region. However, these signals are broad, with spatial footprints of about 400 kilometers. Therefore, although the earthquake would certainly generate a detectable signal, it might not be possible to use GRACE-FO data to invert for details like depth or fault orientation (personal communication with Rishav Mallick).

“Fingerprinting” the earthquake using its aftershocks

There is an additional dataset that seems to be telling us what might have happened: the aftershocks. The magnitude-7.5 and magnitude-8.1 earthquakes triggered widespread aftershocks, reaching about 600 kilometers along strike (nearly the entire length of the subduction zone). They even triggered seismicity on faults outside the subduction zone, including strike-slip earthquakes on the southern transform fault, where the Sandwich Plate is sliding past the Antarctic Plate, and normal-mechanism earthquakes (up to magnitude-7.1) in the outer rise — a region of the subducting plate outboard of the trench, where the plate bends and fractures as it approaches the subduction zone (Figure 3). While aftershocks are expected after every large earthquake, this pattern of seismicity is unusual. It may be possible to “fingerprint” the mainshock based on how it modified the stress state in the crust and triggered seismicity.

Map of ocean area with ocean floor topography. Cross sections below.
Figure 3: (Top) Map of the tectonic setting with the earthquake sequence. The magnitude-7.5 foreshock, magnitude-8.1 mainshock, and aftershocks are shown, with color indicating reported depth. (Bottom) Profiles A and B illustrate the depth distribution of seismicity; background highlights subduction of oceanic lithosphere. Seismicity in the outer rise may be a clue to the slip pattern in this otherwise enigmatic event. Colored contours represent depth of slab2.0 (Hayes, 2018). Map produced using tectoplot (Bradley, 2021).


Notably, intraslab earthquakes tend to produce relatively few aftershocks (e.g., Gomberg and Bodin, 2021). Given the productivity of the aftershock sequence in the South Sandwich Islands, it seems unlikely to have been an intraslab event, despite the reported depths by USGS. The aftershocks suggest that the magnitude-8.1 event ruptured the megathrust, and therefore that the hypocenter depth reported by the USGS is wrong.

We can extend this fingerprinting further. Megathrust earthquakes can either occur on a confined patch of the plate boundary at depth or can rupture the shallow region all the way to the trench. Trench rupture carries a characteristic signal of seismic activation in the outer rise (Sladen and Trevisan, 2018). Notably, this signal seems to be largely absent in deeper ruptures. The South Sandwich Islands earthquake caused significant aftershock activity in the outer rise, which spread across 250 kilometers (Figure 3). This seismicity suggests not only that the earthquake ruptured all the way to the trench, but also that this shallow slip region was at least 250 kilometers long. The density of seismicity has been shown to correlate with the amount of slip at the trench (Sladen and Trevisan, 2018), indicating that it may even be possible to map out the relative amount of slip at the trench using this dataset.

This approach is qualitative, but the seismic, tsunami and gravity data can be used to test the hypothesis of trench rupture, looking for higher tsunami generation near the trench, slow rupture propagation or other signals associated with this type of earthquake.

The importance of shallow rupture to the trench

The idea of trench rupture at subduction zones is becoming more accepted, but examples of large earthquakes with this pattern remain rare (e.g., Hubbard et al., 2015). Because trenches in subduction zones are below water, it has been difficult to detect what is going on near the trench. A number of studies previously assumed that these sections of subduction zone faults were creeping during the long period between earthquakes, and therefore would not slip during a sudden earthquake. However, more recent research has shown that these regions are within a “stress shadow” cast by the deeper part of the subduction zone, and generally cannot slip until an earthquake occurs in the deeper region (Almeida et al., 2018). New methods are being developed to assess this hazard (Lindsey et al., 2021).

These theoretical studies are useful, but observations of trench rupture are critical to expand our understanding. Thus, identifying this event as a trench rupture earthquake — if indeed it was — could help cast light on this process. Alternatively, if the event was in fact a deeper rupture, either on the megathrust or within the slab, we may need to reevaluate our understanding of aftershock generation.

The mixed seismic signals from these earthquakes will remain an issue and highlight a potential weakness in existing earthquake response systems, which currently rely on rapid detection and characterization of earthquake signals. Fortunately, this issue should not generally impact regions with reasonable instrumentation — like most populated areas — and therefore remains a concern primarily for earthquake scientists interested in the physics of earthquakes, rather than the population at large.


Almeida, R., Lindsey, E. O., Bradley, K., Hubbard, J., Mallick, R., Hill, E. M. (2018). Can the up-dip limit of frictional locking on megathrusts be detected geodetically? Quantifying the effect of stress shadows on near-trench coupling. Geophysical Research Letters 45 (10), p. 4754-4763, https://doi.org/10.1029/2018GL077785.

Bradley, K. (2021). Tectoplot, github, https://github.com/kyleedwardbradley/tectoplot.

García, D. Singh, S. K. Herráiz, M. Pacheco, J. F. and Ordaz, M. 2004. Inslab earthquakes of central Mexico: Q, source spectra and stress drop, Bull. Seism. Soc. 94, 789–802.

Gomberg, J. and Bodin, P. (2021). The productivity of Cascadia aftershock sequences. Bulletin of the Seismological Society of America, 111 (3), p. 1494-1507, https://doi.org/10.1785/0120200344.

Hayes, G., 2018, Slab2 – A Comprehensive Subduction Zone Geometry Model: U.S. Geological Survey data release, https://doi.org/10.5066/F7PV6JNV.

Hubbard, J., Barbot, S., Hill, E.M., Tapponnier, P. (2015). Coseismic slip on shallow décollement megathrusts: Implications for seismic and tsunami hazard. Earth-Science Reviews 141, p. 45-55, https://doi.org/10.1016/j.earscirev.2014.11.003.

Kanamori, H. (1972). Mechanism of tsunami earthquakes. Physics of the Earth and Planetary Interiors, 6 (5), p. 346-359, https://doi.org/10.1016/0031-9201(72)90058-1.

Lindsey, E. O., Mallick, R., Hubbard, J., Bradley, K. E., Almeida, R. V., Moore, J. D. P., Bürgmann, R., Hill, E. M. (2021). Slip rate deficit and earthquake potential on shallow megathrusts. Nature Geoscience, 14, p. 321-326, https://doi.org/10.1038/s41561-021-00736-x.

Sladen, A and Trevisan, J. (2018). Shallow megathrust earthquake ruptures betrayed by their outer-trench aftershocks signature. Earth and Planetary Science Letters 483, p. 105-113, https://doi.org/10.1016/j.epsl.2017.12.006.

Van Daele, M., Araya-Cornejo, C., Pille, T., Vanneste, K., Moernaut, J., Schmidt, S., Kempf, P., Meyer, I., Cisternas, M. (2019) Distinguishing intraplate from megathrust earthquakes using lacustrine turbidites. Geology, 47 (2) https://doi.org/10.1130/G45662.1.

Ye, L., Lay, T., Bai, Y., Cheung, K. F., Kanamori, H. (2017). The 2017 Mw 8.2 Chiapas, Mexico, earthquake: energetic slab detachment. Geophysical Research Letters, 44 (11), p. 11824-11832, https://doi.org/10.1002/2017GL076085.