Of aftershocks and tsunamis

There’s been another bout of seismic activity in Indonesia, with three earthquakes of magnitude 7 or greater, with the largest registering as an 8.4, and dozens of smaller quakes (many of which were still over magnitude 5.0), occurring since last Wednesday. The media coverage of these events prompted GrrlScientist (who keeps tabs on the goings on in her birds’ home region) to e-mail me a couple of excellent questions, which with her permission I’m answering publicly:

I heard this morning that the same area of Sumatra has been hit with three earthquakes within the past three days. What I want to know is: how do the scientists know these are three separate earthquakes instead of aftershocks of one quake? Is it because of the sheer size of each of these quakes? Also: are aftershocks always smaller than the initial earthquake?

To answer this it is necessary to first answer the question, what is an aftershock? An earthquake rarely occurs in isolation, because rupturing and displacement on one section of a fault will redistribute the stresses on adjacent sections, and even on other nearby faults, which may well trigger further earthquakes. So you often get days, weeks or even months of enhanced seismic activity – numerous earthquakes of various sizes – within a particular region. The biggest earthquake in such a sequence is referred to as the main shock; smaller earthquakes that precede it are referred to as foreshocks, and those which follow as aftershocks. By definition, a foreshock or aftershock is always smaller than the main shock.

eqseq.png


So you can’t look at a particular earthquake on its own and label it as an aftershock; it is only within the context of the whole sequence that it is part of – if it occurs after the biggest earthquake in that sequence – that that label can be applied. Geography also needs to be considered: to count as aftershocks, the later earthquakes must be close enough to the initial rupture to make a direct influence likely. In the figure below, I’ve plotted the seismic activity between last Wednesday and Sunday, and it’s clear that all of the earthquakes – including the two later magnitude 7+ quakes – cluster in a particular region of the plate boundary.

indoquakemap.png

1. Magnitude 8.4, 12 Sept 11:10 UTC, Depth 34 km.
2. Magnitude 7.9, 12 Sept 23:49 UTC, Depth 30 km.
3. Magnitude 7.0, 13 Sept 03:35 UTC, Depth 12 km.
Based on this, there is no real reason to view the two largest events after the main shock as ‘separate earthquakes'; they can quite reasonably be viewed as aftershocks of the initial magnitude 8.4. Nevertheless, as Grrl has commented, they do seem to have been singled out in the media, possibly because they are the ones that potentially represent a significant tsunami risk. Which leads on to the second question:

Why do some earthquakes that are underwater (or near the coast) generate tsunamis while others do not? Why do some quakes generate HUGE tsunamis while others generate only small waves, as one of these quakes apparently did?

The most important factor here is how deep the earthquake is beneath the Earth’s surface. To generate a substantial tsunami, an earthquake has to cause large (metre-scale) vertical movements of the sea-bed, which pushes around the ocean water above it. This requires that the rupture occurs at quite a shallow depth; although seismic waves can propagate around the world, the actual displacement associated with a large earthquake quickly decreases as you move away from its source. I’ve tried to illustrate this idea in the figure below, which shows movements of the sea-floor above a subduction boundary associated with a shallow rupture (marked in red in the lower left figure) and a deeper rupture (lower right figure). In both cases, the ruptures are of about the same length, so their maximum displacements, and the size of the earthquake they generate, are also about the same. However, their different depths result in substantially different effects on the ocean floor. The shallow rupture propagates all the way to the surface, displacing the sea bottom by a large amount and generating a tsunami; in contrast, the greater distance between the deeper rupture and the surface results in much smaller ocean floor displacement, generating no, or a very small, tsunami.

eqtsunami.png

The three magnitude 7+ earthquakes last week were caused by ruptures at a depths of 38, 30 and 12 km respectively, the Boxing Day 2004 earthquake was located at a depth of 7 km. Fortunately, therefore, the recent large earthquakes occurred at too great a depth on the subduction boundary to cause very large displacements at the surface.
Hopefully this helps to clear up the issues raised by Grrls’ questions. I’ll also take this opportunity to encourage anyone with geological questions to comment somewhere or e-mail me – I may not always use them as the basis for a post, but I will always try to answer them.

Categories: basics, earthquakes, geohazards

Comments (11)

  1. Kim says:

    I think that one reason why the M 7.9 earthquake has been singled out is that it looks like it occurred on a different fault.
    It’s very close to the coast of Sumatra, and it’s at a depth of 9.5 km. The M 8.4 earthquake occurred at 34 km depth, further from the coast. Because the subduction zone dips (slopes down, for non-geologists) towards Sumatra, any earthquake on the same fault as the M 8.4 earthquake should be deeper, not shallower, than the M 8.4 earthquake.
    For geologists: I’m guessing that the M 7.9 earthquake occurred on a thrust fault in the accretionary wedge, rather than on the subduction zone proper. Does that sound right to anyone who knows the area better?
    One of the aftershocks appears to have been on the strike-slip Sumatra fault, too. (There’s no focal mechanism, so I can’t be sure – I’m guessing from the map location.)
    There’s another complication, beyond typical aftershocks, though. The Sumatra region looks like a good candidate for triggered earthquakes – earthquakes that occur because movement on a nearby fault brought other faults closer to failure. I’m about halfway through reading a 2005 Annual Review of Earth and Planetary Sciences article about earthquake triggering. It doesn’t talk about the great Boxing Day earthquake, but it talks about other M 9 subduction zone earthquakes and their effects on surrounding faults.

  2. Chris Rowan says:

    According to the USGS the 7.9 was located at 30 km depth – it was the 7.0 which was much shallower…
    It seems to me that there’s a rather fuzzy line between ‘aftershocks’ and ‘triggered earthquakes’ if you’re within a few rupture lengths of the initial shock. Although maybe you can cast some light on this.

  3. Kim says:

    That’s weird – the depth on the summary page for the M 7.9 quake was changed between the time I commented and now. (I guess they’re recalculating it or something?) Check out the summary page – there’s a map that has been up since Wednesday or Thursday of last week, with focal mechanisms and basic information – that’s where I first noticed the 10 km depth. (I get the impression that depth is hard to determine.)
    When I finish reading the article, I’ll blog about aftershocks, earthquakes triggered by stress changes, and earthquakes triggered by passage of seismic waves. It’s pretty interesting, and it seems like the conventional wisdom has changed somewhat since I was a student. (Which makes it an exciting time to be following earthquake science. :D ) But I think you’re right – the concept of “aftershock” is kind of fuzzy. I guess the article I’m reading is mostly helping me understand the mechanisms by which earthquakes are related to one another (whether they are foreshocks or aftershocks or earthquakes that don’t quite fit into either concept).

  4. Kim says:

    I just remembered something about earthquake depth. The programs that locate earthquakes automatically set the depth for shallow earthquakes, and the depths are different for oceanic vs continental earthquakes. I forget the exact numbers that they use – it may be 10 km for oceanic crust and 35 km for continental crust, but I seem to remember that they used something slightly different so that the numbers stood out.

  5. Harold Asmis says:

    Geologists writing about earthquakes…. GAG! :)
    I think somebody is taking those depths as fine as the head of pin (10.000000 km). I’m waiting to see the area plots of those earthquakes. There is some thought that they are forming a classic ‘donut’ (Yummm!) around a big hunk of fault real estate. I wrote a modest contribution on this.

  6. Bob O'H says:

    Yes, Harold. Shocking, isn’t it?
    Bob

  7. hc says:

    You have to be very careful with the depths as it is the hardest parameter to determine. The three examples above all have their depth fixed before the inversion program is run.
    Also take care to compare like with like – the USGS method is not the same as the CMT method and results are likely to be different.
    30km is typically used for a notional ‘base of the crust’ as the old location programs used to work with a datum level at the top of a uniform mantle.

  8. Chris Rowan says:

    Do they really fix the depths? I remember from practicals that there’s a trade off between the depth and the seismic moment when fitting synthetic sesimograms, which will lead to differences in both depending on the method you use. And presumably there’s a standard depth to start your inversion from. But fixing it seems a little unwise. Unless they just fix a depth for speed’s sake…
    Anyway, the fact that there wasn’t any sizeable tsunami is a reasonable indication that the recent big quakes were fairly deep, even if the given depths are probably +/- 10 km.
    Oh, and Harold was too modest to link to his post (I think it’s that one anyway).

  9. hc says:

    Location programmes can have a hard time with depth determination particularly with poor/noisy data (particularly for rapid determinations with limited data). They are just inversion programmes and it is common that a solution for the depth parameter is not stable (even suggesting that the ‘quakes are above ground) so in these cases it is sensible to fix the depth at a geologically sensible value and get ‘good’ estimates for the other parameters.
    I would treat all rapid determinations of source parameters with a pinch of salt, CMT solutions doubly so. They are good for generalisations and systematic averages, but in my experience going back to the primary data (i.e. the seismograms) will yield some distinct inconsistencies (e.g. many will not match P-wave first motion patterns).
    And yes, the fact that there were no tsunamis does suggest that they were ‘deep’.

  10. Rich Briggs says:

    Chris, nice blog and interesting post. I just stumbled across this and thought I’d add a few thoughts – sorry if it’s a stale thread by now.
    When earthquakes get this large, it’s probably most helpful to visualize slip occurring over very large areas of the fault, rather than as point sources as the comments on hypocentral depths imply. On your figure, the M 8.4 and 7.9 certainly broke patches much larger than the large red dots (while the 7.0 probably fits inside the orange dot). It doesn’t make much sense to talk about ‘the depth’ of great earthquakes – for example, the 2004 Boxing Day rupture was as much as 1600 km long and ruptured nearly the entire seismogenic width of the fault for much of that distance, from 35-40 km depth to only a few km (and possibly right to the sea floor). The spatial relationships between the September events are quite a bit more complicated than the epicenter dots would imply – the large ruptures abut and overlap to some degree, while the smaller aftershocks are very small indeed in terms of the area of fault slip. And, the maximum slip does not necessarily occur at the hypocenter.
    Also, it’s easy to get sidetracked by discussions of what constitutes an aftershock/foreshock. In terms of seismic hazard, this is clearly an important question (and a source of endless fascination to folks who do statistics on earthquake catalogs). But when it comes to understanding moment accumulation and release in a system like this, it may be more fruitful to dwell on how and why the rupture patches communicate with one another. The 7.9 was an aftershock, but also was triggered on an adjacent, unbroken (since 1833) section of the megathrust. A 7.9 is a big earthquake (check out the USA top 20 list for perspective) and the shaking and local tsunami damage alone from this event make it worthy of study and discussion. I’m reminded of the classic(ally bad) made-for-TV movie 10.5, where some earthquakes are blown off because they are ‘only’ aftershocks. A more interesting question might be why did the 7.4 wait 12 hours after the 8.4 to go off? Why did the 8.4 happen several years later, and well to the south, of the 2005 M 8.6, leaving an unruptured patch in between the two? Why did the patch that ruptured in 2005 resist rupture for 3 months after the 2004 rupture slammed into it? How much slip has been spent and how much is still in the bank? As Kim says, this is a terrifically interesting topic.
    Finally, tsunami creation is not determined simply by rupture initiation depth. What matters is the area, height, and rate of seafloor displacement (as you say), and this turns out to be a very tricky thing to predict as it hinges on the slip distribution and dynamic behavior of the rupture. The M 8.4 and 7.9 did create locally destructive tsunamis, but nothing monstrous and basin-wide, thank goodness. But I’m not sure hypocentral depth alone explains that.
    Sorry so long, and thanks,
    Rich

  11. James says:

    Interesting article. You talk early on about after- (or fore-) shocks going on for days, weeks and months. For how long does a quake region tend to be extra-vulnerable after an initial major quake? Do most major aftershocks occur within a week or is there evidence for expecting more over three and six months?