Clustering of Large Earthquakes Explained by Random Variability?

Sumatra, 2004. Chile, 2010. Japan, 2011. Odds are these names and dates ring a bell. The scars are still too fresh and widespread for them to have slipped from memory. They identify, of course, the three largest recorded earthquakes that have occurred in the last 45 years — with magnitudes of 9.1, 8.8 and 9.0, respectively. Add in the 2005 magnitude 8.6 quake that also hit Sumatra, likely triggered by the 2004 event, and you’re talking about the four largest earthquakes in roughly the last half century all occurring within a few years of each other. It seems strange that they have been so bunched up, doesn’t it? 
 

An aerial view of Minato, Japan taken about a week after the
March 11, 2011 magnitude 9.0 earthquake and resulting tsunami that
devastated large swaths of the Japanese coast. (Credit: Lance Cpl.
Ethan Johnson, U.S. Marine Corps, Creative Commons Attribution
2.0 Generic)

In fact, there was a similar sequence of major earthquakes in the middle of the last century as well: Kamchatka, magnitude 9.0, 1952; Chile, magnitude 9.5, 1960 (the largest known earthquake); and Alaska, magnitude 9.2, 1964. No other 9.0+ events were recorded in the 20th century. Throw in a few more 8.5+ quakes in the same time frame, and you’ve got quite a cluster of destructive temblors occurring over just a decade and a half.

Contrast these turbulent stretches with the decades-long periods before 1950 and from about 1965 until 2004, when the planet was relatively calm, seismically speaking, and it certainly appears that these enormous earthquakes timed so close together are connected by more than coincidence, right? And if so, shouldn’t we be expecting more major quakes in the near future as one popular author suggested following the earthquake off the coast of Japan last spring?

According to a study released earlier this month in Geophysical Research Letters, not necessarily. Odds are, writes the study’s author Andrew Michael, that the apparently meaningful clustering of large earthquakes witnessed recently is actually just statistical coincidence after all.

A leveled coastal village on Sumatra following the December 2004
earthquake and tsunami. (Credit: Photographer's Mate 2nd Class
Philip A. McDaniel, U.S. Navy; photo released to public domain)

Michael, a geophysicist with the U.S. Geological Survey in Menlo Park, Calif., used a series of statistical tests to examine the known record of magnitude 7.0 and greater earthquakes from 1900 through the present. The tests analyzed the record for patterns in inter-event times (the time between earthquakes) and increases in the rate of global seismicity following large quakes (potential triggers for additional earthquakes) that would indicate robust connections among large quakes. However, other than picking out obvious aftershock sequences, known beforehand to be caused by a triggering event, Michael found no evidence for such connections.

He also analyzed simulated earthquake records that had the same overall duration and, on average, the same number of events as the observed record for temporal clustering of large magnitude events. Comparing the results to the clustering in the observed record, he found that no fewer than 30 percent of the simulations showed similar clustering. In other words, the sequences of frequent large earthquakes that Earth has experienced are not as peculiar as one might intuitively suspect.

The upshot of the study, Michael writes, is that the observed record of large (7.0 and greater) earthquakes since 1900 is “well-described by a random process,” and that the apparent clusters of the mid-20th century and the last few years are due to “random variability.” Thus, he writes, the risk of additional high-magnitude earthquakes in the near future is neither increased nor decreased based on the recent spate of sizeable events.

The study’s results follow on and agree with those of another recent study that found no evidence of long-distance triggering among large, remote earthquakes. (Large earthquakes are known to trigger additional earthquakes locally — both large and small — in the form of aftershocks, as well as to trigger small, remote quakes.) Together, these studies represent what seems to be the consensus view among seismologists and geophysicists that scant evidence exists to suggest that the recent clusters are not random or that they have any predictive power for future earthquakes.

There are some who disagree, however. They argue that the probability of rare, massive earthquakes occurring in such close temporal proximity as they have is so small that the observed clusters are almost certainly connected and non-random. Michael disputes this alternate conclusion, suggesting that it is based on limited data and methods that “underestimate the true variability of a Poisson [random] process.” Still, he is careful to remark that his analysis does not rule out the possibility that large earthquake triggering and clustering occurs. But for the time being at least, discerning such triggering is not possible without a longer historical record of large earthquakes and/or more sensitive techniques to test physical hypotheses of long-distance triggering mechanisms.

Given its relevance to earthquake prediction — the elusive Holy Grail of seismology — the question of whether large earthquakes beget other large earthquakes around the planet, and whether this leads to periods of significantly increased global seismicity, is bound to remain a hot topic. Odds are that is.

More:

To listen to an interesting and whimsical discussion about some of the oddities of randomness, check out this Radiolab episode.

Explore the largest recorded earthquakes, courtesy of the USGS and Google Earth (you'll need the Google Earth plug-in). Scroll around the globe and click on the orange (mag 8.5+) and red (mag 9.0+) dots for more information about specific events.

View Larger Map