19. On the Origin of Mega-thrust Earthquakes

Although there are far more very large earthquakes (M ≥8.0) that have occurred along the plate interface of various subduction zones which can be considered as mega-thrust events, in this study, we have restricted our definition of mega-thrust earthquakes to those that have magnitudes equal to or larger than 8.5. Among the 17 earthquakes since 1900 (Fig. 19.1), based on the data from USGS (USGS 2014), we consider 15 of them as mega-thrust events since the 1950 Assam earthquake have occurred in a different tectonic setting with continent-continent collision, and the 2012 Sumatra earthquake was the largest ever recorded strike-slip faulting event which occurred along one of the fractures zones offshore northern Sumatra. The remaining 15 events have all occurred along the various subduction zones in the Pacific and Indian Oceans. Their space/time correlations indicate that the largest of these earthquakes cluster in time. This is clearly shown in Fig. 19.1, with the two clusters in the time-periods 1950–1964 and 2004-present, separated by a quiescence period of 39 years. The most striking feature of these two clusters is that in the first cluster there were three mega-thrust events with M ≥9 and there were two M ≥9.0 in the second. Although it is tempting to suggest duration of approximately 10–15 years for these clusters with a rough repeat time of 40 years, statistically such conclusions are not warranted. This is mainly due to the fact that the total time of observation during the instrumental period is far too small and the two temporal clusters within 114 years cannot be generalized unless we have longer time series available. In spite of increasing evidence for mega-thrust events in the pre-historic period, assessing the occurrence of mega-thrust events during the historic and pre-historic period has some obvious limitations. In general the uncertainties of the source parameters increase significantly backwards in time. This however, should not undermine the importance of paleoseismological data which has proven useful in cases such as the subduction zone mega-thrust paleo-earthquakes of NW-US (1700, Cascadia earthquake; Satake et al. 1996) and in NE-Japan (869, Jogan earthquake; Minoura et al. 2001).

It is clear that the occurrence of these mega-thrust earthquakes is governed by global tectonics and the total seismic moment-budget associated with the plate convergence rates in the subduction zones (e.g. Pacheco and Sykes 1992; McCaffrey 2007). Nevertheless, their occurrence in time and space is highly dependent on the history of deformation in individual subduction zones and their internal segmentation within the arc. Despite this, there are some common characteristic that can be attributed to the mega-thrust earthquakes. These can be summarized as follows:

Based on some of these common features there have been recent attempts to classify the different subduction zones and the associated mega-thrust earthquakes (e.g. Koyama et al. 2013). A simple classification based on three criteria, along-strike segment boundary, along-dip segmentation and the direction of collision (orthogonal or oblique), although useful to sort out some basic differences, still lacks the necessary details and hence forces one to think in terms of these end members only. However, understanding the subduction zone deformation requires a holistic approach to all controlling factors (Fig. 19.2).

A complete deformation cycle in a subduction zone starts with the inter-seismic period of strain accumulation due to the plate convergence, which in cases where there is strong coupling along the plate interface, results in internal deformation both in the overriding and subducting plates. In the overriding plate uplift occurs along the coastal areas of the island arc and inland regions whereas subsidence is seen towards the trench in the ocean-ward side. Both of these effects are the consequence of compressional forces due to locking of the plate interface. Observation of the sea-level changes in Sumatra and the response of the coral micro-atoll growth have demonstrated these long-term effects of overriding plate deformation (e.g. Zachariasen et al. 2000; Sieh et al. 2008). Similarly, in the subducting plate in the outer-rise region, compressional deformation occurs during the inter-seismic period, coupled with the down dip extension at depth giving rise to normal faulting deep intraplate events. Once the plate interface is ruptured through a mega-thrust earthquake (co-seismic deformation) the relaxation period following this favors the reversal of the forces acting in the same regions both in the overriding and subducting plates. Subsidence along the shore and inland regions accompanied by the uplift along the trench are typical for the overriding plate deformation. In the subducting plate the same structures that were reactivated as reverse faults now act as normal faults due to extension in the relaxation period (post-seismic deformation). There is off course processes both prior to the rupture of the plate interface (foreshock activity) and immediately after the mega-thrust earthquakes (aftershocks) which is part of the total deformation cycle. There are few examples that captures this total deformation cycle such as the triple earthquakes that have occurred along the central Kurile subduction zone in 1963 (M = 7.7), 2006 (M = 8.3) and 2007 (M = 8.1) (Raeesi and Atakan 2009).

Understanding the total deformation cycle in subduction zones and the processes associated with it requires multidisciplinary approaches including a variety of observational data combined with analog and numerical modelling (Funiciello et al. 2013). In recent years indeed a wealth of observational data became available. These include,

Synthesizing such a variety of data brings along some methodological challenges as well. In the first place, it is necessary to realize the importance as well as the limitations of each data set before applying an appropriate method. Detailed studies of co-seismic slip-inversions through various data sets for the 2011 Tohoku-Oki, Japan mega-thrust earthquake, illustrate this problem very clearly. Following the earthquake of March 11, 2011 in Japan, there has been a number of co-seismic slip inversions published using tele-seismic data (e.g. Ammon et al. 2011; Ishii 2011; Lay et al. 2011; Koper et al. 2011; Wang and Mori 2011), strong-motion data (e.g. Ide et al. 2011; Suzuki et al. 2011), GPS data (e.g. Linuma 2011; Miyazaki et al. 2011; Ozawa et al. 2011; Pollitz et al. 2011) as well as tsunami data (e.g. Fujii et al. 2011; Saito et al. 2011). In addition to these there has also been joint inversions of seismological (teleseismic and strong-motion) and geodetic data (e.g. Koketsu et al. 2011; Yokoto et al. 2011; Yoshida et al. 2011; Kubo and Kakehi 2013). Common for all these inversion results is the shallow asperity with a large slip. In general there is a good agreement on the location of the shallow asperity among the various studies (seismological, GPS and tsunami wave data), where the maximum slip exceeds 40 m. When it comes to the details of the rupture there are significant differences in these inversion results. The main conclusion here is that slip inversions are non-unique and there is strong need for independent data which may help calibrating these. In other words, identifying the location of the strong asperities by multidisciplinary data sets seems critical. Independent evidence for the shallow asperity and the observed large slip came from the sea-bottom GPS measurements (Sato et al. 2011) and shallow seismic data (Kodaira et al. 2012) combined with cores from the borehole drilled at the tip of the sedimentary wedge (Chester et al. 2013).

The down-dip extent of the fault rupture is on the other hand, debated and some of the studies conclude that rupture propagated to the bottom of the contact zone. A number of inversions based on teleseismic data from large and dens arrays (US-array and the Stations from Europe) have revealed a strong short period radiation at deeper part of the rupture plane (e.g. Ide et al. 2011; Ishii 2011; Koper et al. 2011; Meng et al. 2011; Wang and Mori 2011). It is now understood that the slip associated with the shallow asperity was slow and lacking short-period radiation, whereas the deeper asperities produced strong short-period energy (Koper et al. 2011).

Apart from that, arguably, it can be said that the joint inversions smear out the slip distribution and a lot of details such as the short-period radiation at depth is not resolved (Meng et al. 2011). As such the common understanding that the joint inversions are better than individual data sets is questionable. The rupture complexity with a dynamic variation at various frequencies is better resolved by individual analysis of different data sets that are sensitive to these frequencies. The results from these individual studies, when combined together in a synthesis, seem to be a far better tool than the joint inversion results.

Mega-thrust earthquakes along subduction zones are mainly controlled by the plate coupling along the interface. Some critical issues related to the degree of coupling are, the location of the strong and weakly coupled zones (asperities and their origin), role of sediments and fluids in coupling, down-dip limit of the coupled zone as well as coupling in the shallow zone close to trench. Regarding the latter the 2011 Tohoku-Oki earthquake has surprised many. Contrary to the common belief that the shallow part of the coupling along the trench is usually weak controlled by the loose sediments of the accretionary prism accompanied by the fluid interaction reducing the friction, more than 40 m of slip is observed along the trench. This very high slip along the trench was also crucial in the development of the following large tsunami wave.

The strongly coupled shallow asperity along the trench was manifested by the various co-seismic slip inversions as discussed earlier. It is also firmly confirmed by the offshore GPS measurements where significant slip was observed (Sato et al. 2011). The maximum horizontal slip measured was as high as 24 m almost 100 km away from the trench. The vertical uplift was as high as 3 m in the same area. The slip was also observed at the very tip of the trench through high resolution seismic data (Kodaira et al. 2012). Later, Chester et al. (2013) have shown the actual plate interface cutting through the contact between the pelagic sediments of the subducting plate and the sediments of the accretionary prism representing the overriding plate.

The rupture process had however started already with the onset of increased earthquake activity just at the periphery of this strong asperity at depth some weeks before the main rupture which culminated in a magnitude 7.5 earthquake at the deeper end of the shallow asperity on March 9, 2011. The static stress transfer from this event was probably the triggering mechanism for the main rupture on March 11, 2011. Such foreshock activity is not unique for the Tohoku-Oki earthquake, similar significant foreshock activity was previously documented in other plate interface thrust events (e.g. Raeesi and Atakan 2009) and more recently during the 2014 Iquique earthquake in northern Chile (Hayes et al. 2014).

Post-seismic slip is usually associated with extensive aftershock activity following the mega-thrust events. This was the case for the Tohoki-Oki earthquake where hundreds of aftershocks were registered in the following weeks after the main shock (Nishimura et al. 2011). The most striking feature of the aftershock sequences was their spatial concentration in areas outside the main asperities that had ruptured during the co-seismic slip. The largest of these aftershocks had a magnitude of 7.9 at the southernmost part of the plate interface off Boso, close to the Sagami trough in the south. In addition to the intensive aftershock activity along the plate interface, there has been also triggered seismic activity both in the overriding plate (Kato et al. 2011) and the subducting slab in the outer rise area such as the M = 7.7 earthquake (Lay et al. 2011). Such outer rise normal faulting events can be very large as was the case for the 1933 (M = 8.4) event further to the north. These events are the manifestation of the total deformation associated with the stress transfer from the main shock (Toda et al. 2011)

Physical conditions leading to the deformation in subduction zones depend on a variety of factors including:

Although the total deformation is controlled by these factors, the physical and the morphological irregularities of the oceanic plate converging to the trench will in time have long term consequences in terms of the segmentation of the plate interface. Iquique ridge entering into the subduction zone in the border area between northern Chile and southern Peru is a good example for this (e.g. Pritchard and Simons 2006; Contreras-Reyes and Carrizo 2011; Métois et al. 2013). The strong coupling along this zone has previously been modelled (e.g. Métois et al. 2013; Chlieh et al. 2014) and is expected to produce large mega-thrust earthquakes probably larger than the recent Iquique event of 2014 (M = 8.2). Other sea-floor irregularities such as sea-mounts, fracture zones and ridges play thus an important role in the overall segmentation of the plate interface in various subduction zones.

Once the segmentation of the interface is understood, the next critical issue is to find the location of the asperities. Inevitably, slip inversion of earthquakes constitutes an important contribution in this sense. However, there is a need for additional independent data to calibrate and verify the slip inversions as well as to find out more about the location of asperities in subduction zones where there are no recent large mega-thrust events in the latest instrumental period. One promising recent development is the use of satellite gravity data, GRACE in resolving the co-seismic gravity changes due to mega-thrust events (e.g. Tanaka and Heki 2014; Han et al. 2014). These new data opens new possibilities for detecting the location of asperities, because the repetitive slip along the same asperities of the plate interface causes mass dislocations. In the long-term, cumulative mass dislocations in the same part of the overriding plate will lead to permanent density changes. The accumulated density changes then leave an imprint on the overriding plate due to gravity (buoyancy forces) that change the degree of coupling along the plate interface. Cumulative effect of these variations should therefore be detectable as subtle deviatoric gravity changes parallel to the trench. These strongly coupled areas constitute the asperities that will slip in future large mega-thrust earthquakes. Mapping asperities by gravity data was first introduced by Song and Simons (2003), where trench parallel topography and gravity anomalies in the circum-Pacific region have revealed the strongly coupled areas along the plate interface. This was later modified (Raeesi and Atakan 2009; Raeesi 2009) to include also trench parallel Bouger anomaly.

Mapping asperities along the plate interface using these new techniques, if combined with detailed monitoring of seismological as well as geodetic changes in time with the recent observations regarding the short-term precursory phenomena such as total electron content (TEC) in the ionosphere (e.g. Liu et al. 2011; Tsugawa et al. 2011), may provide important opportunities to understand the deformation processes before the occurrence of the mega-thrust earthquakes.

In order to understand better the complex processes leading to mega-thrust earthquakes and the total deformation in subduction zones, future studies should focus on identifying the gaps for mega-thrust earthquakes as well as identifying the precursory phenomena in both long- and short-term. Following is a short list of research areas that may be helpful in this sense:

Identifying Gaps for Megathrust-Earthquakes

Identifying Precursory Phenomena

In the long-term:

In the short-term: