- © 2016 by the Seismological Society of America
Estimating the hazard resulting from great earthquakes on active subduction megathrusts is plagued with difficulties, yet it is necessary in populated areas proximal to active subduction. In practice, we are increasingly including knowledge of the physical fault system and mechanics into estimations of seismic hazard, aided by rapid advances in understanding the diverse spectrum of megathrust behavior. In general, statistical approaches to estimating hazard are hampered by the paucity of observed data and deterministic approaches involve large uncertainty. In Taiwan and New Zealand, megathrust‐hazard estimates are further complicated by the relatively short history of seismic observations and lack of any historical, local, or regional great earthquakes to use as analogs. Thus, it is useful to consider what insights can be gained from observations of the 2011 Mw 9.0 Tohoku‐Oki earthquake, Japan.
Tectonic similarities in Japan, Taiwan, and New Zealand (Fig. 1), including the complexity of subduction geometry, rapid convergence, and similarity in age and temperature of the incoming plate, make the comparison of these three subduction zones appealing. It is worth noting that unless complicated linking of multiple and disparate faults is invoked, rupture extent in all three regions is clearly limited on one side by local changes in the geometry of the convergent margins. Perhaps the most striking similarity between Tohoku and the southern Hikurangi (Fig. 1) is the distribution of interseismic coupling as determined from the inversion of Global Positioning System (GPS) data (Nishimura et al., 2004; Wallace et al., 2004; Suwa et al., 2006; Hashimoto et al., 2009). Similarities include both the down‐dip extent of locking and strong along‐strike variations.
Although some authors have used plate motion budgets and subduction zone length to argue for Mmax in many or all subduction zones to be Mw 9 or greater (e.g., McCaffrey, 2007, 2008; Kagan and Jackson, 2013), this suggestion has not been widely implemented in national‐hazard models around the Pacific. Driven by the need to represent megathrust hazard in a spatially meaningful way, the source is often considered as a series of regional segments. Mmax is then determined from informed estimates of the likelihood of neighboring sources rupturing in a single event, thereby increasing the effective rupture area. The Ryukyu and Hikurangi margins have structural controls on maximum rupture extent on one side of the system, but the other side (determining maximum length extent) and down‐dip extension of possible rupture are much less constrained. Wang and Bilek (2011) and subsequent studies have presented evidence to suggest that a smooth subduction megathrust ‐ void of widespread asperities ‐ is needed to facilitate rupture in great or giant earthquakes. Although portions of the incoming plate in both Taiwan and New Zealand exhibit greater roughness than in the incoming Pacific plate at the Japan trench, more research is required to understand if and how this idea is appropriate to be applied to seismic‐hazard assessment in the two regions. Mapping recurrent and predicable aseismic slow‐slip, nonvolcanic tremor, and low‐frequency or very low‐frequency earthquakes also provided prospects for defining the region in which the in situ physical conditions would allow for rupture propagation. However, this area of research also requires further investigation prior to application in seismic‐hazard assessment.
Whether or not slow‐slip areas also host seismic rupture is currently the matter of much debate within the slow‐slip community.
Geodetic measurements of interearthquake coupling (slip‐rate deficit) of faults have increasingly been applied to subduction megathrust systems to quantify the ability of the plate interface to accumulate stress. Slow‐slip events are believed to occur at the transition between fault segments that are capable of storing interseismic stresses and areas that freely creep. Hence, the spatial distribution of strong coupling and mapped slow‐slip events has been used to define maximum possible rupture and consequently Mmax using empirical scaling laws extrapolated largely from smaller events (e.g., Stirling et al., 2013). It is often assumed that areas of low coupling or areas that host frequent aseismic or recent significant seismic deformation are barriers to rupture because they lack the elastic strain budget to facilitate rupture propagation. Whether or not slow‐slip areas also host seismic rupture is currently the matter of much debate within the slow‐slip community.
If we assume that only areas with strong coupling (e.g., coupling coefficient >0.5) are capable of sustaining rupture propagation (as was observed in the Tohoku‐Oki earthquake), a conservative maximum rupture size of ∼20,000 km2 yields an Mmax≈8.2 for rupture of the southern Hikurangi margin (Fig. 1b). Accordingly, Mmax in the New Zealand National Seismic Hazard Model (NSHM) was set to 8.2 prior to 2011. Regardless, notwithstanding uncertainty, the spatial extent of the strongly coupled area in the southern Hikurangi (Wallace et al., 2014) is comparable to the spatial area of significant displacement in Tohoku‐Oki (Fig. 1).
Consideration of the strong ground motions recorded in the 2011 Tohoku‐Oki earthquake can aid our understanding of the potential impacts of great earthquakes on Taiwan and New Zealand.
For the southernmost Ryukyu, Hsu et al. (2012) estimated the Mmax of about 8.7 from interseismic GPS observations inland of northeast Taiwan. Using the down‐dip extent of seismicity along the Ryukyu trench, and considering lateral tectonic constraints, a maximum rupture area was defined yielding an Mmax of ∼8.8 with the area–magnitude relationship of Yen and Ma (2011). However, the largest magnitude considered in the Taiwan seismic‐hazard model is 8.0, defined by the historical catalog of earthquakes that only spans the last ∼400 years. Taiwan is located in the western/southernmost end of the Ryukyu trench subduction system where little is known about interseismic coupling, slow‐slip and tremor events, and earthquake potential. Future study from broadband ocean bottom seismometers (Kuo et al., 2014) might be helpful to understand the seismic potential of the western Ryukyu.
Many estimations of Mmax in the rupture region of the Tohoku‐Oki earthquake prior to the 2011 event underestimated the eventual size as suggested by Kagan and Jackson (2013, and references therein), including the official Japanese hazard map (Headquarters for Earthquake Research Promotion, 2005, see Data and Resources), which contained a Tohoku area Mmax of 7.7. Part of the reason for this magnitude underestimation was the belief that the fault segments that eventually ruptured in the Tohoku‐Oki event would not fail in a single earthquake, but in many earthquakes smaller than Mw 8.0, as had been observed in the historical record. In fact, much of the 2011 Mw 9.0 rupture traversed regions of the megathrust that have hosted large (Mw>7) earthquakes in the last century (Hasegawa and Yoshida, 2015). Although it is enticing to place great significance on secular trends in deformation or seismicity that have persisted at most for decades, Tohoku‐Oki has taught us that it is prudent to use caution when interpreting these data for refining estimates of seismic hazard.
Both the Ryukyu and Hikurangi subduction zones have lengths of continuous megathrust larger than the Tohoku‐Oki rupture area, necessitating a reassessment of maximum plausible rupture area and the implications of this for Mmax. If we assume that considerable slip can occur to depths of >40 km as seen in the Tohoku‐Oki earthquake (Fig. 1b), detailed maps of the Hikurangi megathrust allow us to map the down‐dip extent of plausible rupture width, which is on average over 150 km. Assuming rupture could potentially propagate at least the entire along‐strike extent of mapped, spatially connected deep south‐southeast, rupture area for the margin is over 50,000 km, requiring consideration of Mmax=9.3 based on area scaling laws developed for strike‐slip earthquakes (Hanks and Bakun, 2008) or Mmax=8.9 based on length relationships also developed for strike‐slip faults (Wesnousky, 2008). If we further increase lateral estimates of the extent of rupture to consider bridging the gap between recently mapped deep south‐southeast (Wallace and Eberhart‐Philips, 2013), the length of potential rupture is at least 500 km, yielding an Mmax of ∼9.2 based on Wesnousky (2008) length relationships. In the present NSHM, similar lateral and down‐dip limits of maximum rupture have been drivers for the current Mmax of 9 (Stirling et al., 2012). We suggest that, given uncertainties, an Mmax of at least 9 for both Taiwan and New Zealand is appropriate, and a conservative estimate of Mw 9.3 for New Zealand should be considered.
TOHOKU‐OKI GROUND MOTIONS AS ANALOGS
Predicting ground motions of great earthquakes has challenges. There is large uncertainty in numerically simulated ground‐motion models, particularly given the potential complexity and uncertainty associated with the earthquake source. Moreover, great (Mw>8.0) or giant (Mw>9.0) earthquakes differ from Mw<8 events in that variations in source processes and resulting ground motions have greater potential consequences, because the energy released from great earthquakes is significantly larger and covers a greater spatial extent. Given these uncertainties, the application of simulations to seismic‐hazard modeling is not immediately practical in Taiwan and New Zealand. The use of traditional ground‐motion prediction equations (GMPEs) is also limited, owing mainly to a paucity of recordings used in their development. We therefore suggest that even though it is an imperfect analog, consideration of the strong ground motions recorded in the 2011 Tohoku‐Oki earthquake can aid our understanding of the potential impacts of great earthquakes on Taiwan and New Zealand.
Following the recognition that a Tohoku‐Oki style event could happen in either the Ryukyu or Hikurangi subduction zones, it is instructive to use the recorded ground motions from the 2011 Mw 9 event (Aoi et al., 2011) as a simple analog for a gross estimate of ground shaking and impacts of great earthquakes on Taiwan and New Zealand during a similar‐sized event. We recognize that 3D path effects alter ground motions, in some cases reducing amplitude through anelastic attenuation and scattering, or by increasing amplitude through efficient wave guiding. We note that the variability resulting from these effects is likely within the standard deviation of many empirically based GMPEs, which can commonly extend to a factor of 2.
Prior to the Tohoku‐Oki earthquake, the number of high‐quality records of ground shaking during great megathrust earthquakes was severely limited. The extensive national strong‐ and weak‐motion seismic networks in Japan recorded ground motions resulting from the 2011 Mw 9 event with unprecedented spatial resolution. This recorded dataset represents the best analog to date for potential effects of a great or giant earthquake along the Ryukyu or Hikurangi trench. We therefore overlay published maps of recorded ground motions from the Japanese strong‐motion networks (Aoi et al., 2011) on the Ryukyu and Hikurangi subduction margins by aligning the Japan trench to those offshore Taiwan and New Zealand (Fig. 2). We note that due to the comparable geometries of the Japan and Hikurangi megathrusts, the hypocentral depth in these two regions align as well as the trench.
We suggest that Mw 9+ earthquakes are possible on the Ryukyu and Hikurangi megathrusts that would produce damaging ground motions over most of Taiwan and the North Island, New Zealand.
Many researchers have delineated areas of high‐ and low‐frequency radiation from the Tohoku‐Oki earthquake, possibly due to variation in friction up‐dip and down‐dip of the origin (Ide et al., 2011). The great Tohoku‐Oki earthquake could be considered as the nearly simultaneous combination of a deep interface earthquake radiating high‐frequency energy and a low‐frequency energy radiation from a shallow tsunami earthquake resulting from the slow failure of a shallow asperity. Based on rupture analysis of the Mw 9.0 Tohoku‐Oki, the Mw 9.2 Sumatra–Andaman, and the Mw 8.8 Maule great earthquakes, Lay et al. (2012) developed a conceptual model with four distinct deformational regimes on the subduction megathrust. These regimes include, in order of increasing down‐dip location: an area of weak short‐period energy release near the trench that could host tsunami earthquakes, an area of earthquakes hosting large slip with weak short‐period energy release (green patches in Fig. 2), a region with modest slip and strong short‐period energy release (blue patches in Fig. 2), and finally, a down‐dip region that is conditionally stable, hosting slow‐slip events, low‐frequency earthquakes, and tectonic tremor. It is also worth noting that large slip near the trench has previously affected the northern Hikurangi Margin in New Zealand when the area hosted two slow tsunami earthquakes (Mw 6.9–7.1) in 1947 (Power et al., 2016).
The great Mw 9.0 Tohoku‐Oki earthquake was a sobering reminder that earthquake prediction is not currently possible and that estimating seismic hazard in subduction zones remains a nascent science.
In New Zealand, the dominant rupture directivity, driving maximum ground motions, is fortunately directed offshore, as was the case in the Tohoku‐Oki earthquake. However, Figure 2 illustrates that regions of peak ground acceleration (PGA) >0.5g could cover most of the North Island, including the national capital of Wellington, whereas 0.25g stretches nearly to Auckland, with a population of ∼1.4 million people, representing approximately one‐third of New Zealand’s inhabitants. Given the uncertainties in this crude analysis, including maximum depth of rupture and northward extent of rupture, there is undoubtedly the possibility that 0.25g accelerations could affect Auckland in an Mw 9 earthquake. This is significant, as the city, and many others in the North Island, contains concentrations of structures (e.g., unreinforced masonry buildings) that are prone to severe damage when exposed to 0.25g, which would be exacerbated given the significant duration of strong ground motion expected in a megathrust event.
In an equivalent Mw 9 rupturing the western Ryukyu subduction zone, Taipei, a city of almost 3 million inhabitants, is likely to suffer PGA >0.5g. In large offshore megathrust earthquakes, long‐period motions need special consideration, as they are likely to strongly affect Taichung City, directly west of the Ryukyu trench and the third largest city in Taiwan. The resulting ground motions in the rest of the country are less clear. The northward concavity and dip of the western Ryukyu trench are likely to support a wide region of rupture directivity that could span much of the central and southern Taiwan (shown as red region in Fig. 2). However, there are no available data from the equivalent offshore region in the Tohoku‐Oki analog.
STRESS‐INDUCED CHANGES IN SEISMICITY
Large‐scale changes in the upper‐plate stress regime in Japan resulted from the Tohoku‐Oki earthquake. The ∼50 m of slip near the trench during the event suggests near‐total stress drop of the plate interface (Lin et al., 2013). However, many reverse or transpressional faults in the upper plate nearest to the high slip region experienced normal‐faulting aftershocks, suggesting dynamic overshoot (Ide et al., 2011). It is uncertain how forecasting models perform in the case of wide‐ranging displacement overshoot and care should be exercised in forecasting, especially during the early stages of an aftershock sequence. Static stress redistribution and overshoot are particularly interesting in the case of Taiwan due to its complicated tectonics and fault‐margin geometry. The seismicity rates in the Tokyo area (Kanto), nearly 400 km away from the epicenter of the Tohoku‐Oki earthquake, increased as a result of stress loading from the mainshock (Ishibe et al., 2015). Given this, static stress loading from a megathrust earthquake would most likely increase the rates of earthquakes in many of the transpressive northeast/southwest‐faulting regions in Taiwan, as would margin‐parallel normal and right‐lateral faults in the upper plate of the Hikurangi margin on the scale of years to decades.
The great Mw 9.0 Tohoku‐Oki earthquake was a sobering reminder that earthquake prediction is not currently possible and that estimating seismic hazard in subduction zones remains a nascent science (e.g., Morikawa and Fujiwara, 2016). Many aspects of the earthquake, including the rupture area, magnitude, and large displacement near to the trench, were surprising. The complex rupture involving rupture propagation through numerous patches that have experienced recent displacement dictates that current efforts to define earthquake hazard in subduction zones solely through interpretation of the historical earthquake record needs to be carefully scrutinized. Resulting from discussion during the 2015 meeting of the 3rd Annual Joint Workshop of the Japan/New Zealand/Taiwan Seismic Hazard Programs (Gerstenberger and Fry, 2016), we use the Tohoku‐Oki earthquake as a coarse analog to explore potential consequences of great (or giant) events on the Hikurangi and Ryukyu subduction zones. We suggest that Mw 9+ earthquakes are possible on the Ryukyu and Hikurangi megathrusts that would produce damaging ground motions over most of Taiwan and the North Island, New Zealand. Furthermore, given the orientation of crustal faults in both New Zealand and Taiwan, stress changes as a result of megathrust rupture could also produce sequences of upper‐plate fault earthquakes.
Data and Resources
The Headquarters for Earthquake Research Promotion (2005) is available at http://www.j-shis.bosai.go.jp/en/ (last accessed September 2016).
This analysis was a result of discussions during the 3rd Annual Joint Workshop of the Japan/New Zealand/Taiwan Seismic‐Hazard Programs. Portions of this work were supported by the New Zealand Royal Society Marsden Fund and the GNS core funded Rethinking PSHA program. This discussion was benefited from helpful discussions with Anna Kaiser, Matt Gerstenberger, Laura Wallace, Yu‐Ju Wang, and Tetsuya Takeda. The article was also improved through the extensive review of two anonymous reviewers and Editor Zhigang Peng.