A M~8.3 Chile earthquake has become more likely

By David Jacobson and Ross Stein, Temblor

See earthquakes in Chile

Click for Spanish version

Updated: 3 p.m. (PDT) 1 May 2017

The city of Santiago, Chile is an area susceptible to large earthquakes. Due to the recent swarm of quakes, a M=8.3-8.4 earthquake in the region is now more likely.


Since 22 April 2017, an earthquake swarm has been in progress offshore the port city of Valparaíso, which has been widely felt in Chile’s capital, Santiago. This sequence began with a M=6.0, and was followed two days later on 24 April, by a M=6.9 earthquake in the same swarm cluster. Today there were M=5.8 and M=5.9 earthquakes. What are the implications of the swarm, in light of historical earthquakes in the region, for what may happen next?


Geological Disneyland

Large megathrust earthquakes are one of three dramatic consequences of the ‘Nazca’ oceanic plate being shoved under the South American continent at about 80 mm/yr (3 in/yr). The second consequence of this ‘subduction’ is the compression and uplift of the Andes, and the third is the string of Pacific rim volcanoes. The Chilean coastline has suffered repeated great earthquakes over the 500 years of its recorded history. The largest earthquake ever recorded, a M=9.5 in 1960, occurred 570 km (350 mi) south of Santiago in southern Chile.


How big could the next shock be?

This week’s seismic swarm struck in the center of a 300-km-long span of the megthrust hemmed in by two recent great quakes. In 2010, the M=8.8 Maule earthquake ruptured the region to the south, and in 2015, the M=8.3 Illapel earthquake ruptured to the north (Melgar et al., 2016). The last large earthquakes to have occurred within this span were in 1971 (M=7.9) and 1985 (M=8.0), and so 50-60 years have since elapsed. Because the long term subduction, or slip rate on the megathrust, is about 80 mm/yr, a deficit of roughly 4-5 m of slip has accumulated. If a 300 km x 100 km section of the megathrust suddenly slipped 4-5 m (12-16 ft), that would yield a M~8.3 shock.

This Temblor map shows the location of the recent earthquake swarm (with the M=6.9 highlighted). Additionally, the rupture zones in the 2010 and 2015 earthquakes are shown in purple. These quakes, plus the recent seismicity have stressed 75% of an area capable of rupturing in a M=8.3-8.4 earthquake.


The megathrust is stuck and so great quakes must eventually strike

The velocities of GPS stations in Chile reveal that the megathrust surface offshore this 300-km-long span is stuck (Klein et al., 2017): The fault is not slipping freely, and so the accumulating stress must ultimately be released by a great quake or a large series of smaller ones. On the other hand, the swarm we witnessed this week is probably the manifestation of creep on at least one spot on the fault. So, its behavior is mixed, but dominated by great quakes.


The recent shocks ratcheted up the stress on the Valparaíso section

One can calculate the stress imparted by the 2010, 2015, and 2017 earthquakes to the unbroken span (We use Coulomb 3.4: Toda et al., 2011). The stress transfer is not an observation, but instead a calculation based on a series of reasonable assumptions, so one must take it with a grain of salt. Nevertheless, we find that roughly 3/4 of this span has experienced an estimated stress increase of at least 0.5 bars (shown as dark red in the figure). In general, stress increases of much less than this (0.1 bars) are associated with seismicity increases. One way to think about this result is that earthquake stress drops average 30 bars, and so a 0.5-bar increase represents a 1.5% jump.

So, if we think of great shocks here having about 50-75 year inter-event times, then the next shock has been advanced by several years. But the effect of stress imparted to rupture zones tends to fade with time, and so the impact of this week’s shock is large despite its small size, whereas the effect of the 2010 Maule shock might be smaller today despite its large size. The Global Earthquake Activity Rate (GEAR) model used by Temblor gives a 1% chance per year of a M~8 shock off Valparaíso, and we estimate that the combined effect of the recent shocks might double or triple this to 2-3% per year. So, the best we can say is that it’s a large increase on a small likelihood.


Could the M=6.9 prove to be a foreshock of a M~8?

It’s possible, but not obvious. Worldwide, no more than 5-10% of mainshocks are preceded by anything that a seismologist—even in retrospect—would want to call a foreshock. But the 2014 M=8.2 Iquique earthquake (Hayes et al., 2014) in northern Chile offers a stunning counter-example, and so we cannot exclude it down the road at Valparaíso. One can see from the ‘time series’ below that a M=6.7 quake struck 15 days before the Iquique mainshock, and it was followed a week later by many M=5-6 shocks that migrated northward. Ultimately, about half the future rupture zone lit up before the mainshock. In the case of Valparaíso, so far no more than about one fifth of the span has been filled by recent seismicity. But now, take a look at the time series for the 2010 M=8.8 Maule earthquake: No foreshocks whatsoever, and this is far more common.

These time series (From Hayes et al., 2014) show the earthquake sequence behavior in the M=8.2 Iquique earthquake and the M=8.8 Maule earthquake. The Iquique earthquake was preceded 15 days earlier by a M=6.7 shock, while the Maule quake showed no seismicity prior to the mainshock.


Today’s M=5.8 shock has extended the swarm nest to the south (red shocks). The aftershocks today’s M=5.8 are also more frequent than for the preceding M=6.0 on April 23, and for M=6.7 on April 25. The center of the swarm now seems less active than before, perhaps because the M=6.7 slippage transferring stress to a ‘halo’ of stress around the M=6.7 rupture.


Reading the tea leaves

So, if the pattern of Iquique were followed, in the next few weeks we could see an expansion of the swarm zone leading to a larger rupture. But if seismicity follows a far more common pattern, then the M~8.3 is out there, somewhere in the future.



Melgar, D., W. Fan, S. Riquelme, J. Geng, C. Liang, M. Fuentes, G. Vargas, R. M. Allen, P. M. Shearer, and E. J. Fielding (2016), Slip segmentation and slow rupture to the trench during the 2015, Mw8.3 Illapel, Chile earthquake, Geophys. Res. Lett., 43, doi:10.1002/ 2015GL067369.

Klein, E., Vigny, C., Fleitout, L., Grandin, R., Jolivet, R., Rivera, E., Métois, M. (2017), A comprehensive analysis of the Illapel 2015 Mw 8.3 Earthquake from GPS and InSAR data, in press, Earth and Planetary Science Letters.

Gavin P. Hayes, Matthew W. Herman, William D. Barnhart, Kevin P. Furlong, Sebastian Riquelme, Harley M. Benz, Eric Bergman, Sergio Barrientos, Paul S. Earle & Sergey Samsonov (2014), Continuing megathrust earthquake potential in Chile, after the 2014 Iquique earthquake, Nature, 512, doi:10.1038/nature13677.

Shinji Toda, Ross S. Stein, Volkan Sevilgen, and Jian Lin, Coulomb 3.3 Graphic-Rich Deformation and Stress-Change Software for Earthquake, Tectonic, and Volcano Research and Teaching—User Guide, U.S. Geological Survey Open-File Report 2011–1060 [https://pubs.usgs.gov/of/2011/1060/]

  • Gonzalo Rosselló

    “The last large earthquakes to have occurred within this span were in 1971 (M=7.9) and 1985 (M=8.0), and so 50-60 years have since elapsed. Because the long term subduction, or slip rate on the megathrust, is about 80 mm/yr, a deficit of roughly 4-5 m of slip has accumulated.”
    The calculation is wrong… from ’85 to ’17, that’s only 32 years @ 80mm/yr that gives only 3.0m.

    • Dal Stanley

      I am sort of leaning the same way. I first guessed that the ’85 M8 was recent enough that the M6.9 on 4/24 will not be followed by a M8+ real soon. And I agree with David and Ross’s statement about the M8.8 Maule lack of foreshocks:
      ‘No foreshocks whatsoever, and this is far more common’.
      However, there were extensive foreshocks for the Iquique in March before the event, including a M6.9 two weeks before the M8.2. Here is a plot of the two weeks before the M8.2

      • Nicholas Bremner

        Currently living right in the red zone, so I hope you guys are right.

    • Ross Stein

      Correct. But a M=7.8 is only 1/4 the size of the M=8.0, and so it’s elapsed time is less important to the calculation. Another way of looking at is this: If in the north, the future slip were only 3 m, the earthquake covering the entire 300 km span would still be M=8.3, perhaps M=8.2 at smallest.

  • Your assessment that foreshocks occur only 5-10% of the time before larger earthquakes is shortchanging their importance. The successful prediction of a dangerous earthquake in China back in 1975 prompted a study by Lucy Jones and Peter Molnar in Nature (vol. 262, No 5570, pp 677-679, 19 August 1976) related to the frequency of foreshocks. What they found was that for earthquakes of at least magnitude 7, instrumentally detectable foreshocks were seen 44% of the time from 1950-1973. Also, there were records kept of humanly detectable foreshock activity 21% of the time for such quakes inside China from 1900-1949 (there were no doubt more, but they just didn’t make it into the historical record).

    • Ross Stein

      I do not think Lucy Jones would today argue for a 44% foreshock rate, and I certainly would not.

      • My impression, after pondering the disparity, is that let us say that on average, 1 in 3 quakes of 6.8 magnitude or larger are preceded by a foreshock which is on average 1.2 magnitude less than the main shock.

        But looking at it from the opposite perspective, perhaps only 5-10% of 5.6-6.7 magnitude quakes are followed by an earthquake that is 1.2 magnitude larger (on average). The reason being that much more numerous smaller quakes would often be aftershocks of prior bigger quakes or in zones where the largest quake possible would be in the upper area of that 5.6-6.7 magnitude range.

        The 44% rate includes those tremors that would not be perceived by human beings, but again it relates to bigger quakes preceded by smaller ones and not smaller quakes which are followed by larger ones. With the much greater frequency of smaller quakes, there would be a much greater chance of one of them preceding the larger ones than the other way around. Thus, rather than state that larger quakes are preceded by foreshocks only 5-10% of the time, my impression is that it would be more accurate to state that 5-10% of all smaller quakes are foreshocks of larger ones.

      • Fernando Meneses

        I have a question. Can a M6.9-earthquake be a foreshock to major event that’s to hit “in the months” to come or normally foreshocks occur only days or weeks before the mainshock? For example, could that M6.9 offshore Valparaíso still be a foreshock after nearly a month today? I know noboby can say it “is” a foreshock for certain, but in your view based on empirical observation/research over the years, can it be possible and/or also likely? I’d really appreciate your answers as I’m very keen on this topic, especially because I live in Central Chile (I have no fear of earthquakes though :-D) and I just found this website today and it’s incredibly helpful and informative. It’s great!
        Thank you very much in advance.

        • Ross Stein

          Yes, a M=6.9 event could be a foreshock to a M=8 event several weeks to months later. For example, on October 23, 2002, there was a M=6.7 earthquake on the Denali fault. On November 3, a M=7.9 event struck nearby, rupturing 340 km of the Denali and Totchunda faults.

  • Christophe Vigny

    Hi Ross,
    thanks for your interest in Emilie Klein’s recent work

    this swarm, I’m under the impression that the localization and area of
    the swarm coincides extremely well with the locked patch visible around
    33°S in Marianne Métois’ last paper (Métois et al., Pageophy, 2016) –
    figure 3.

    This patch is “surrounded” by weakly coupled areas so that I
    would guess it is creep (possibly related to post-seismic visco-elastic relaxation,
    see Klein et al., GJI, 2016,) in those areas that triggered the rupture
    of this moderate size asperity…

    If correct, I guess this sequence should end there without further noise (appart aftershocks of course)…. this is an educated guess, not a prediction… I lack a slip model for the main shock and GPS data to asses whether the deformation is purely co-seismic or partly a-seismic. Work in progress…

    • Ross Stein

      This is very thoughtful comment is by a leading French tectonic geodecist. My only reply would be that the resolution of these offshore stuck patches is quite limited. Also, if a locked patch were to rupture because of stress transferred by the surrounding creeping areas, what is to prevent further widespread creep piling up at the next locked patch?

  • Dal Stanley

    Looking at the zone south of Valparaiso, the M8.8 in 2010 is a somewhat unique/prominent event in which there were no foreshocks. However, we should possibly factor in the nature of the 1960 M9.5 just south of the M8.8 in which there were large foreshocks with a M8.1 occurring on May 21, the day before. On May 22, before the M9.5 there were M7.1, M6.8, and a M7.8 15 minutes before the M9.5. Here is the location of the M9.5 compared to the M8.8 copied from the Berkeley Blog
    And and even better plot of the regional Chilean events offshore comes from the 2013 paper by Hayes et al.
    Figure Caption:
    Article Reference:

  • Néstor Espinoza

    Hi David, Ross. A somewhat technical question from a non-expert on seismology: what are the errorbars on those rates that you cite from the GEAR model? A quick read of the GEAR paper gave me the impression that the model doesn’t perform that well for events M7.0+, so how do we know for certain that the increase from 1% per year to 2-3% per year is actually real and not due to, e.g., low number statistics both in terms of number of events and time-scales used for training the model?