2018: The Year of the Earthquake Apocalypse?

You may have noticed recent headlines making some scary apocalyptic 2018 earthquake predictions. Scientific thinking generally evolves slowly and thoughtfully. A headline that proclaims a revolutionary or shocking change in our understanding of earthquakes is likely overstating the certainty of a hypothesis, or may be a misinterpretation that is sensationalized by the media. In such cases, it is always best to go back to the original source of information.

These headlines are based on an article published back in August in Geophysical Research Letters entitled Do Weak Global Stresses Synchronize Earthquakes?

This article and its authors have indicated:

• Worldwide there is a temporal pattern to large earthquake (M7+) occurrence.
• The shift in earthquake occurrence seems to correlate with variations of a few milliseconds in the length of the day on Earth.
• The theory proposes that shifts in mass are changing the speed of the Earth’s rotation (like an ice skater unfolding her arms) and are related to the build-up in stress that precedes earthquakes.
• The most recent change in rotation began in 2011, so the authors of the paper propose we are now at the start of a new higher period of M7+ earthquake activity.
• The increase in seismicity will be in the equatorial regions where there are large populations at risk.

Let’s examine these new assertions and see whether we should be concerned.

Starting in about 1900, the worldwide coverage of seismic monitoring instrumentation has been sufficient that all earthquakes in the M7+ range should have been observed, wherever they were located.

For some time, it has been recognized that there have been “temporal groupings” of these large M7+ earthquakes with four peaks in 1910, 1943, 1970, and 2011, implying a cycle with an average length of 32 years — ranging from 27 to 41 years. These temporal groupings of earthquakes, however, are not generally spatial clusters, so any linkages between the events have been hard to identify. The August 2017 article is proposing that these temporal variations are due to very small changes in the rotation speed of the Earth. This correlation is a hypothesis and only time (i.e., decades) will tell if the evidence supports there being a physical connection. The observed correlation does not necessarily indicate causation.

Given the most recent peak in seismicity was in 2011, you might think we should now be on a downward trend. However, the Earth’s rotation has continued to increase its rate of slowing over the past few years, so if there is a correlation between slowing and seismicity then, according to the hypothesis, we may instead be heading into another period of higher M7+ earthquake activity. This is what has spawned the sensational headlines. One of the authors of the paper, Dr. Roger Bilham, was quoted in the Observer newspaper as saying “We could easily have 20 a year starting in 2018.” However, the fact that the most recent 2011 peak in seismicity does not correspond to the peak in slowing may undermine this argument.

Examination of the earthquake record over the last two decades indicates that a typical year has seen between 15 and 16 events with magnitudes equal to or greater than M7. The figure below shows the count of events of M7+ observed globally by year since 2000. Just in the last eight years, we have seen one year with 24, one year with 20 and two with 19. Twenty events in 2018 is not substantially different than what we have been experiencing recently, so it should not be an immediate cause for alarm. It is interesting to note how few earthquakes of this magnitude range have occurred in 2017 so far — only seven events in total (although at the time of writing we still have five weeks to top up that number). From the 2017 record, this does not look like the start of a boom in activity.

The authors also propose that we could see 25 to 30 events per year in the next few years. This seems to be something of an extrapolation as no multi-year average since 1900 has reached more than about 21 events per year. Of course, if there were more events than the long-term average it comes down to whether those events impact exposures or populations are risk. Based on the physical process they are proposing, the authors of the paper claim that these small changes in the rotation of the Earth would produce more events near the equator. Many of the news articles commenting on the paper have pointed out that there are significant populations at risk to earthquakes in the equatorial regions.

Looking back at the figure showing the number of events per year, while there have been 268 M7+ events since 2000, only about a dozen of these (around five percent) have been significant loss events to the insurance sector. To be a significant loss event, an M7+ event should be shallow and close to an exposure concentration. About 25 percent of these larger events occur at depths greater than 50 kilometers (31 miles) and in some case as deep as several hundreds of kilometers. The deeper the earthquake, the less likely it is to be impactful. Additionally, 71 percent of the Earth’s surface is water, so a high percentage of the events are offshore or at the mid-ocean ridges, so are essentially harmless even though their magnitudes are significant. While we do see about one significant earthquake insurance loss event per year, an increase in the event count would not necessarily translate directly to changes to the expected losses.

In conclusion, the original article in Geophysical Research Letters is interesting from an earthquake process point of view. We will not know if the authors’ proposed mechanism for temporal grouping of M7+ events in the last 100 years is correct until there is more data. The authors have proposed a hypothesis. Only data from the next few years and indeed the next decades might confirm whether evidence across multiple years for an increased rate of slowing of the Earth’s rotation can become the basis for a prediction around global large magnitude earthquake activity. Additionally, further research will be required to understand the physical processes behind the temporal linkage between the two datasets.