The heartland of the United States lies within a tectonic plate, certain regions of which have experienced large and geologically recent earthquakes. Explanations for those events are still being sought.
Over the past several decades, the earthquake cycle along tectonic-plate boundaries has become increasingly well understood. There is a consensus that geological, geodetic and palaeoseismic data can be combined to establish long-term earthquake probabilities, with a degree of certainty that improves as more and better data become available.
There is no such consensus when it comes to intraplate earthquakes. The reason is that there are no well accepted principles that account for why large earthquakes have occurred where they did in the recent past, where they are likely to occur in the future, or how large they might be. In this context, Calais et al.1 (page 608 of this issue) provide a valuable contribution. Their study region lies in the central United States, around New Madrid, Missouri, which in 1811–12 experienced a sequence of three earthquakes estimated to be of magnitude 7 or larger.
Much still needs to be done to reduce earthquake hazards for those living along active plate boundaries. To recognize that, one needs only to look at the devastating consequences of the 2004 earthquake and tsunami in Sumatra (230,000 dead in 14 countries), or the earthquake in Haiti earlier this year (approximately 200,000 dead and 2 million left homeless). But the situation is even worse in intraplate regions, especially in the developing world. In the past decade alone, tens of thousands of people have died in each of the earthquakes that hit Bhuj, India (2001), and Bam, Iran (2003), as well as in the magnitude 7.9 Wenchuan event that occurred in China in 2008 (Fig. 1). We know that intraplate earthquakes result from plate-driving forces transmitted through plate interiors2,3. But without a better understanding of why intraplate earthquakes occur where they do, the potential for future damaging earthquakes must be considered 'high impact but low probability'. In the developing world, it is unlikely that much will be done to prepare for such events.
The New Madrid seismic zone is the best studied of locations that have been affected by intraplate earthquakes. One of the enigmatic features of this zone is the rate at which large earthquakes occur. Palaeoseismic data4 indicate the occurrence of at least three, and possibly five, large earthquakes (or sequences of such earthquakes) in just the past few thousand years. However, faults seen on seismic reflection profiles show little cumulative deformation over the past few million years5, during which time the regional geological processes have been essentially identical. Hence, the long-term earthquake rate seems to be much lower than that of the past few thousand years.
Moreover, unlike at plate boundaries, where over time the average rate of seismic-strain release in big earthquakes matches the rate at which strain energy accumulates as a result of relative plate motion, analysis of data from the Global Positioning System (GPS) has shown that the rate of strain accumulation in the New Madrid region is quite low6. The occurrence of multiple large events in a relatively short period of time seems to be due to the release of strain energy that accumulated over a very long period of time.
In turning to the new paper by Calais et al.1, I should declare an interest in that the model used is conceptually similar to one proposed by Grollimund and myself a few years ago7. Both studies invoke the consequences of the retreat of glaciers from much of continental North America at the end of the last ice age. And both assume that the brittle crust is in a state of frictional failure equilibrium — that is, even in relatively stable plate interiors, stress levels are close to that at which slip could occur on faults that are appropriately oriented to the current stress field. This allows even a relatively small perturbation of stresses in the lithosphere to induce brittle faulting in the upper crust, and time-dependent flow in the viscous lower crust and upper mantle.
In Calais and colleagues' model1, the perturbation is caused by localized erosion of approximately 12 metres in the past 16,000 years, produced by river incision. This induces upward flexure of the lithosphere in the New Madrid area, 'unclamping' some of the critically stressed faults in the region. In our paper7 we argued that, consequent on the removal of ice-sheet load, seismicity is localized around New Madrid because of anomalously low viscosity in the upper mantle, the result of an ancient, failed rift in the region.
Importantly, both models produce crustal deformation rates that are consistent with the rates observed by GPS measurements in the region; and both predict that the rate of large earthquakes seen over the past few thousand years is likely to continue for thousands of years into the future, because of the long time it takes for the triggered viscous flow in the lower crust and upper mantle to diminish. In other words, seismic hazard in the region remains high. The paper by Calais et al. is valuable both in reinforcing that point and in providing a plausible mechanism that merits further investigation.
It has been argued8 that, as in the New Madrid region, several intraplate fault zones in Australia have exhibited episodes of relatively frequent earthquakes separated by long periods of quiescence. Similar behaviour may characterize earthquakes in the southeastern United States near Charleston, South Carolina9. These regions, as well as others that have been struck by intraplate earthquakes, deserve detailed study, with the aim of revealing what might have triggered the release of strain energy stored in Earth's crust for millions of years.
About this article
Journal of Ocean University of China (2018)