Alaska Subduction Zone – Tsunamic Structures in a Creeping Section

Australia: The Land Where Time Began

Alaska Subduction Zone – Tsunamic Structures in a Creeping Section

According to Bécel et al. there appear to be structural characteristic configurations associated with subduction zones that are capable of tsunamigenic earthquakes. Included among these structures are heterogeneous plate interfaces, a small wedge of deformed sediment situated at the toe of the plate that is overriding (the frontal prism), and in the crust of the overriding plate splay faults that root within the plate boundary megathrust. In this study Bécel et al. used seismic reflection imaging in to display the presence of these features within the creeping segment of the Alaska subduction zone, the Shumagin Gap. An active crustal-scale normal fault system was identified that dips landwards which resembles that of the 2011 Tohoku-oki earthquake in Japan. The Shumagin Gap was also found to have a small frontal prism, a splay fault in deep water, and that here the plate interface is rough and sedimented thinly. Lateral propagation of rupture from a neighbouring segment into the Shumagin Gap was proposed by Bécel et al. to possibly explain a tsunamigenic earthquake that occurred there in 1788, and that in hazard assessments for the region tsunamigenic potential should be considered. Structural configurations that are similar to those in Tohoku are demonstrated by the results of this study to possibly exist in other subduction zones, including within creeping segments or segments that have no record of megathrust earthquakes. The ability to identify regions that have similar configurations capable of generating large tsunamis globally may improve tsunami forecasting.

Very large tsunamis can be triggered by slip on the shallow portion of the subduction zone plate boundary faults. According to Bécel et al. this can occur when large interplate earthquakes propagate to the trench or during a special class of earthquakes, ‘tsunami earthquakes’, produce larger tsunamis than expected for their moment magnitude (Kanamori, 1972 and are characterised by rupture velocities that are low and of long durations. Subduction zones with a small frontal prism, a thin subducting sediment layer and oceanic crust that is highly faulted are where both types of event primarily occur (Polet & Kanamori, 2,000; Tsuji et al., 2013; Okamura et al., 2008). The 2011 Tohoku-oki earthquake (Tsuji et al., 2013) is the most dramatic recent example of shallow, tsunamic slip occurring; more than 50 m of slip occurred near the trench, which produced a local tsunami run-up of up to 30 m. Another important feature that can be associated with tsunamigenic slip, a localised normal fault, which is landwards dipping, branching from the megathrust, was highlighted by this earthquake. When earthquake rupture propagates to shallow depths extension is expected in the overriding plate (Li et al., 2014; Xu et al., 2016). It is suggested by studies of the Tohoku-oki earthquake that this normal fault may, alternatively, promote shallow slip by decoupling the portion of the wedge that is located seawards of the normal fault (McKenzie & Jackson, 2012; Ide, Baltay & Beroza, 2011). In great earthquakes and/or tsunamigenic earthquakes shallow slip is rare, though it is very destructive, which makes it imperative to extract any information on structures that are associated with these events, e.g. the Tohoku normal fault, which can then be applied to the global subduction zone system. Only a few normal faults that branch from a megathrust have been identified so far (Tsuji et al., 2013; Tsuji et al., 2011; Klaeschen, 1994). Bécel et al. suggest this may be a result of these faults being steep, extending to great depths, and straddle the shoreline in many subduction zones which makes them a challenge to image and also, the significance of these was not recognised until the Tohoku event.

Evidence has been accumulating that tsunamigenic earthquakes can be hosted even by creeping or weakly coupled subduction zone segments, e.g. Java (Bilek & Engdahl, 2007); Nicaragua (Satake, 1994; Correa-Mora et al., 2009). The M8.6 tsunami earthquake of 1957 is an example of such events along the Alaska-Aleutian margin that propagated laterally through a shallow portion of the megathrust, that was weakly coupled, offshore Unalaska Island, which caused a tsunami that was unusually large, 23 m, (Nicolsky et al., 2016), and another example is the M_w8.6 tsunami earthquake in 1946 (Lopez & Okal, 2006) that ruptured the weakly coupled segment offshore Unimak Island. The importance of considering such potentially tsunamigenic structural configurations that are thought to be creeping or with no record of historical megathrust earthquakes is highlighted by these and other events in segments that are weakly coupled.

The Shumagin Gap – current knowledge

In this study Bécel et al. focussed on the Shumagin Gap which is about 200 km wide and that has not ruptured in a great earthquake in 150 years (Davies et al., 1981). It is indicated by data from the Global Positioning System (GPS) that this system is creeping to weakly locked, though on the shallow part of the plate boundary coupling is not well resolved by the shore data (Fournier & Freymueller, 2007). A few moderate earthquakes, M6.5 to M7.5, have occurred at depths greater than 35 km in the area. It has been suggested by some authors that rather than rupturing as a single large earthquake steady creep has released most of the seismic moment and therefore a moderate earthquake of M ⁓ 7 every ⁓40 years, as has been observed over the last century, is sufficient to release the residual slip deficit (Fournier & Freymueller, 2007). Earthquakes in 1788 and possibly 1847 are, however, inferred to have ruptured laterally through part or all of the Shumagin Gap (Davies et al., 1981). During the 1788 earthquake rupture occurred from the neighbouring Semidi segment (Davies et al., 1981), which is strongly coupled at the present. It is suggested by historical records that the 1788 earthquake, in particular, generated a large tsunami (Davies et al., 1981). A recent tsunami scenario for the Alaska Subduction Zone (Kirby et al., 2013) which showed that a large tsunami in this segment could cause devastating consequences to local communities in Alaska as well as around the Pacific has emphasised how important it is to recognise the hazard that is posed by the Shumagin Gap that is coupled weakly.

Shumagin Gap – structural configuration

In this paper Bécel et al. present new observations of the Shumagin Gap in the Alaska Peninsula that were obtained from multichannel seismic (MCS) reflection, wide angle reflection/refraction, and bathymetric data that was acquired using the RV Marcus G. Langseth in the summer of 2011 during the ALEUT (Alaska Langseth Experiment to Understand the megathrust) programme. Acquisition of the MCS data was achieved by using 2 8-km-long seismic streamers and a 6,600 cu. in. tuned airgun array.

It is clearly revealed that the structural configuration in the Shumagin Gap could make it prone to generating transoceanic as well as local tsunamis, and that can explain events in the past that were generated by this segment that has been assumed to have been creeping, Bécel et al. observed a large normal fault that was active that may slip either coseismically which would allow displacement over a large area seawards of the fault (Makenzie & Jackson, 2012) or be triggered by propagation of a shallow earthquake rupture (Xu et al., 2016), as was determined for the Tohoku earthquake (Tsuji et al., 2013; Tsuji et al., 2011), and therefore could promote or result from shallow slip. They also observed a heterogeneous character along the shallow plate boundary, a small frontal prism and a thrust splay fault in deep water, all features that are favourable to tsunamigenesis (Wang & Bilek, 2014).

As was mentioned above, a large normal fault that is landwards-dipping in the overriding plate that bounds the eastern Sanak Basin is the most prominent feature that was imaged. This fault system crosses the upper slope 75 km from the trench, dips ⁓40^o-45^o, cutting the entire crust and connecting to the plate boundary fault at a depth of ⁓35 km, near the intersection of the megathrust with the forearc mantle wedge. At a depth of around ⁓6 km the fault splays into 2 branches, which breaches the seafloor at water depths of ⁓500-700 m. The association of this fault with the Sanak Basin, which dates to the Miocene, is ⁓6 km thick, which suggests that the fault had sustained activity and accommodated significant total normal displacement. An indication of the accumulative slip on the fault is the offset of the acoustic basement which is very significant, ⁓2 km. The strongly coupled Semidi segment, which is further to the northeast, ruptured in a M8.2 earthquake in 1938, on the upper slope the sediment is much thinner, and imaging shows no major deep basin or normal fault system in the overriding plate (21). The 1938 Semidi earthquake epicentre was deep and the earthquake appeared to not reach the trench and therefore only a small tsunami was produced (Estabrook, Jacobs & Sykes, 1994).

As presented here, the new data also demonstrate that the fault that bounds the Sanak Basin and connects to the boundary of the plate has been active recently. The youngest sediments are offset by the fault and at the seafloor scarps about 5 high are observed in the multibeam bathymetry data. Near the surface expression of the fault pockmark features that are observed and the perturbed sediments could suggest the presence of fluid flow that is localised. At the intersection of this fault with the main plate boundary fault and the occurrence of some seismicity also occurring within the upper plate near the fault and further landwards (Abers, 1992).

In the Shumagin Gap the properties of the shallow plate boundary and structures in the overriding plate near the trench are constrained by the seismic reflection data. These features are consistent with settings that are known to generate tsunami earthquakes. The oceanic plate that is incoming has a sediment cover that is thin, 50-800 m, and irregular that is disrupted strongly by the bending fault (Shillington, 2015). It appears that only about 250-500 m of sediment are being subducted, which can only be traced about 10 km from the trench. The megathrust itself is marked further landwards by a reflection that is discontinuous, and clear older thrust faults are imaged in the outer wedge that merge with the top of the oceanic crust. The thrust fault that is the most landward is located 20 km from the trench and merges with the megathrust at a depth of about 12 km. This prominent fault separates a highly deformed outer wedge with thrust faults, which are clearly imaged, from ‘transparent’ basement rocks further landwards that are overridden by tilted slope

Implications of imaged structure for tsunamigenesis

A process for tsunamigenesis that accompanies earthquakes in the Shumagin Gap, that is weakly coupled, could be provided by the structures in the overriding plate that are newly discovered and properties of the plate boundary, when taken together (19,25). The generation of a transoceanic tsunami would be favoured by the deep water splay fault near the trench and or the rough, shallow plate interface (Moore et al., 2007). The propagation of erratic slow rupture could be favoured by the rough surface of the plate interface itself near the trench with sediment that has been trapped in topography that was fault generated at the top of the oceanic crust, as was proposed in 1986 for the Sanriku tsunami earthquake (Tanioka, Ruff & Satake, 1997) or the occurrence of slow slip events (Saffer & Wallace, 2015). Lithified materials, including basement rocks, are present near the trench, and which are implied by the narrow frontal prism and the storage of elastic strain, could therefore permit the storage of elastic strain that could lead to coseismic slip on the shallow part of the subduction zone and that could favour large far-field tsunamis, as has been proposed for other subduction zones (Polet & Kanamori, 2000; Contreras-Reyes, Flueh & Grevemeyer, 2010).

Several studies that were carried out since the Tohoku earthquake have shown that a deep-rooted normal fault situated within the overriding plate may be an indicator of propagation of rupture to the trench (Tsuji et al., 2013; Li et al., 2014; Xu et al., 2016) because large slip on the shallow plate interface would drive extension in the overriding plate (Li et al., 2014; Xu et al., 2016). If a large earthquake activated the normal fault system the decoupling of the wedge seawards of the normal fault could be allowed, which could therefore allow shallow slip and uplift of the seafloor (Makenzie & Jackson, 2012) which would lead to the release of gravitational potential energy over a large area or enabling overshoot (Ide, Baltay & Beroza, 2011; Ito et al., 2011). According to Bécel et al. it appears the normal fault involved in the Tohoku earthquake separated parts of the seafloor that experienced very different amounts of horizontal motion, 5-30 m of horizontal motion landwards of the fault and 58-74 m of horizontal motion seawards of the normal fault, though it appears the normal fault itself slipped by only 1 m during the earthquake (Ito et al., 2011). At steeply tapered outer wedges normal faults in the overriding plate could be activated during large earthquakes by the reduction of basal traction that is associated with drops in stress (Cubas et al., 2013).

According to Bécel et al. a clear boundary in the distribution of seismicity appears to be marked by the normal fault system in the Shumagin Gap, were abundant seismicity associated with plate interface is observed landward of the intersection of this fault with the megathrust, though seawards of the intersection there is less seismic activity. The seismicity change that is observed is robust and is not related to station distribution. Where this fault intersects the megathrust the sharp seismicity change implies that it could be associated with a change in the physical properties of the upper plate or conditions and frictional behaviour of the plate interface. It is suggested by recent studies that changes in frictional parameters are required in order to localise normal faulting along a major fault (Cubas et al., 2013). These studies indicate, more specifically, a lower effective friction on the plate interface updip of the normal fault than downdip in this fault. It is known that a heterogeneous plate interface facilitates activation of branching faults above the megathrust (Wendt, Oglesby & Geist, 2009).

It is also expected that slip on the major fault, which is an actively landward-dipping normal fault system in the Shumagin Gap, is located 75 km from the trench, is also expected to produce vertical displacement of the seafloor near the fault, as well as decoupling the seaward portion of the wedge, and could therefore enhance the local tsunami, though this effect is known to be smaller than the propagation of rupture to the trench (Li et al., 2014) and depends strongly on bathymetric effects.

It is suggested by comparison of the thermal models for the subduction zone off the Alaska Peninsula (Syracuse, van Keken & Abers, 2010) with the structural configuration that was imaged here and the estimated ruptures of earthquakes in the past that temperature is not a primary control on slip behaviour here. The 150^oC isotherm, which was hypothesised by Bécel et al. to represent the updip limit of ruptures in some subduction zones (Hyndman & Wang, 1993), is at a depth of 30 km about 120 km from the trench, which is deeper than estimated great earthquake ruptures in this region (e.g. the 1938 earthquake) (Li et al., 2015; Estabrook, Jacob & Sykes, 1994). This is similar to other subduction zones that are relatively cold, such as Tohoku, where temperature also does not appear to be an important control.

The segment of the Shumagin Gap is made particularly prone to producing tsunamigenic slip because of the structural configuration that has been observed, though the causes of coseismic slip capable of producing tsunamis to occur in a creeping or weakly coupled section is not clear. It has been suggested by earthquakes in the past that ruptures can propagate laterally into the Shumagin Gap from neighbouring segments that are locked. A tsunamigenic earthquake occurred on 7 August 1788 in the Shumagin segment (Davies, Sykes, House & Jacob, 1981). Wave heights of 10s of metres were reported on Sanak and Unga Islands, as based on historical descriptions from Russian settlers, which is similar to the wave height that was recorded for the 1946 Unimak near-field tsunami. It is suggested by Bécel et al. that the shallow plate boundary and/or splay faults in the overriding plate in this section that is creeping and coupled weakly may have moved coseismically when the 1788 earthquake propagated into the shallow Shumagin Gap. It is demonstrated by other past events that asperities can ultimately break through a neighbouring creeping to weakly coupled segment. Significant seismological evidence of the propagation of rupture laterally through creeping segments (35) or updip, through part of the subduction interface that is believed to be coupled weakly (Noda & Lapusta, 2013), has been accumulating over the last 15 years. An example is the 1964 earthquake in Alaska in which 2 large asperities were ruptured (Freymueller et al., 2008), and the Alaska earthquake in 1957 when the weakly coupled Unalaska segment (Nicolsky et al., 2016) ruptured.

Bécel et al. showed that there was a possibility of earthquakes nucleated in the neighbouring locked Semidi segment propagating laterally into the Shumagin Gap where they had shown that there was a structural architecture that was favourable to large local and transoceanic tsunamis, should be taken into account when considering hazards for this region. The propagation of rupture through the creeping to weakly coupled Shumagin Gap and activation of the normal fault system then leading to either gravitational energy release or amplified motion seawards of this fault by overshooting dynamically would result in the tsunamigenic potential for this segment that is predicted by the elastic coupling estimates (McKenzie & Jackson, 2012; Cubas, Avouac, Leroy & Pons, 2013). If this is not taken into account this could lead to an under-estimation of the seismic hazard.

It is demonstrated by this study that normal faults comparable to the one that was involved in the Tohoku earthquake are present in other subduction zones. Bécel et al. suggest such faults may be of under-recognised significance around the world as a result of their significance for tsunamigenesis previously not being appreciated given the rarity of shallow slip in the short historical record, and as imaging seismically requires the use of modern long-offset, MCS data, that is available on only a small number of subduction zones. Also, such structures would not be detected by imaging efforts if they straddled the shoreline.

Also suggested is that creeping regions might have greater potential for tsunamis than has been recognised previously. The identificatio0n and characterisation of active normal faults on a crustal scale in the overriding plate as promotors or indicators of rupture to the trench, and the shallow megathrust configuration is therefore essential to a complete and comprehensive understanding of hazards in the global subduction system.

Sources & Further reading