Until recently, it has been commonly assumed in seismic hazard and earthquake risk calculations that earthquakes occur randomly in space and time. However, there is clear evidence that earthquakes are clustered in both space and time, says Professor Paul Somerville from Risk Frontiers.
While it is still not possible to predict earthquakes in a deterministic sense (ie by specifying the location, time of occurrence and magnitude within usefully narrow bounds), it is becoming increasingly possible to base probabilistic-based earthquake forecasts on various models of earthquake clustering.
Several investigators have proposed the presence of two temporal clusters of very large earthquakes during the past century (eg, Bufe and Perkins (2005) and Ammon et al. (2011)).
The first cluster occurred in the middle of last century and included the 1952 Mw 9.0 Kamchatka earthquake, the 1960 Mw 9.5 Chile earthquake and the Mw 9.2 Alaska earthquake (Bufe and Perkins 2005).
The second cluster began with the occurrence of the Mw 9.15 Sumatra, Indonesia earthquake of 26 December 2004 and has continued with the Mw 8.8 Maule, Chile earthquake on 27 February 2010 and the Mw 9.0 Tohoku, Japan earthquake on 11 March 2011 (Bufe and Perkins 2011; Ammon et al. 2011). This recent cluster has given rise to debate about whether the observed temporal clustering of these very large earthquakes has some physical cause or has occurred by
Testing the global earthquake catalogue for indications of non-Poissonian attributes has been an area of intense research, especially since the 2011 Tohoku earthquake. The usual approach is to test statistically the hypothesis that the global earthquake catalogue is well-explained by a Poissonian process (eg Michael (2011)).
The Pisagua earthquake
Risk Frontiers analysed one aspect of this problem which has been neglected in literature: the power of such tests to detect non-Poissonian features if they exist; that is, the probability of type II statistical errors (Dimer, 2012).
We showed that the low frequency of large events and the brevity of our earthquake catalogues reduce the power of the statistical tests, hindering an unequivocal answer to this question. We did this by designing a counter example of a stochastic process that was clustered by construction, and showing that the tests incorrectly identified it as Poissonian.
Each of the events in the current cluster of very large earthquakes has posed questions about what might occur next. In the case of Chile, a magnitude 8.2 subduction earthquake, termed the Pisagua earthquake, occurred in northern Chile on 2 April 2014, north of the rupture zone of the Mw 8.8 Maule earthquake. This region had not experienced a major earthquake since 1877, when a magnitude 9.0 earthquake occurred on a much larger rupture plane.
Seismic activity began to increase in the Pisagua region last August, when a swarm of small earthquakes occurred in the area. Another sequence followed over the New Year, and a third cluster in March.
Most researchers were expecting that the next large earthquake in northern Chile would break the entire interface between the Nazca and South American plates in a much larger (magnitude 9.0) event.
The 2014 Pisagua earthquake did not break the part of this subduction zone that has built up the most stress, leading to the expectation that a magnitude 9.0 event may currently have a fairly large probability of occurrence.
The Tohoku earthquake
The 2011 Mw 9.0 Tohoku earthquake generated an aftershock sequence that affected a large part of northern Honshu. This has given rise to widely divergent forecasts of changes in earthquake occurrence probabilities in northern Honshu, Japan.
We have assessed these forecasts as they relate to potential changes in the occurrence probabilities of damaging earthquakes in the Kanto region (Somerville, in press). The Kanto region includes the Greater Tokyo area and encompasses seven prefectures: Gunma, Tochigi, Ibaraki, Saitama, Tokyo, Chiba, and Kanagawa. It is generally agreed that the 2011 Mw 9.0 Tohoku earthquake increased the stress on faults in the southern Kanto district.
Studies on earthquake probability in Kanto region
Toda and Stein (2013) further conclude that the probability of earthquakes on such faults has increased by a factor of 2.5 for the time period 11 March 2013 to 10 March 2018 in the Kanto Corridor beneath Tokyo. However, estimates of earthquake probabilities in a wider region of the Southern Kanto District based on post-Tohoku observed seismicity and standard statistical laws by Nanjo et al. (2013) suggest that any increase in the probability of earthquakes is insignificant in this larger region.
The results of Uchida and Matsuzawa (2013) support the conclusion that fault creep in southern Kanto may be slowly relaxing the stress increase caused by the Tohoku earthquake without causing more large earthquakes.
Stress transfer calculations indicate a large stress transfer to the off Boso segment, in the offshore region just east of Tokyo, as a result of the 2011 Tohoku earthquake. However, the largest aftershock of the Tohoku earthquake, with a magnitude of 7.9, occurred off the Boso Peninsula about 30 minutes after the main shock.
Moreover, Ozawa et al. (2012) used onshore GPS measurements to infer large post-Tohoku creep on the plate interface in the off-Boso region, and Uchida and Matsuzawa (2013) measured similar large creep off the Boso Peninsula. This is consistent with the interpretation of geodetic data by Nishimura et al (2007), who find that the Pacific–Okhotsk subduction zone east of Tokyo including the off Boso segment may have low seismic potential.
Therefore, although there was a large stress transfer to the off Boso Segment as a result of the 2011 Tohoku earthquake, it is possible that some of this stress is being released seismically in slow-slip events, consistent with the pre-Tohoku geodetic data, leading to the possibility that a large earthquake on the off Boso segment has a low probability.
Situation still far from clear
The situation remains uncertain, however, with Ozawa (2014) finding that slow-slip earthquakes on the off Boso segment are occurring more frequently than before. The latest slow-slip event came only 2.2 years after the previous one, whereas the first slow-slip events that were detected, beginning in 1996, were 6.4 years apart.
This may be a sign of increasing tectonic stress in the region, which may foreshadow a coming large event.
Researchers hope that ongoing monitoring of earthquakes and crustal deformation in this region may provide some indication of what may occur next.
Professor Paul Somerville is Chief Geoscientist at Risk Frontiers.