The seismic potential of the shallow portions of the northern Cascadia and the North Sumatra subduction zones: insights from laboratory friction experiments

Abstract

Many regions that are prone to experience strong earthquakes and tsunamis are densely populated, such as the coastlines of the Pacific Ocean and some of the Indian Ocean. These regions are subduction zone settings, where one tectonic plate subducts beneath another, which produces a gigantic fault – a megathrust fault. Subduction zone earthquakes largely occur on such megathrust faults. They have cost an incredible number of lives, and future events pose a constant threat to many more. Especially those megathrust earthquakes that nucleate in or propagate to very shallow depths can cause large damage and tsunamis. In general, the seismicity in the shallow portion of subduction zone megathrusts is low, but recent events such as the 2004 Aceh-Andaman earthquake and tsunami offshore North Sumatra have tragically shown the potential of shallow seismicity. Despite extensive investigations of multiple geoscientific disciplines, the shallow extent of earthquake rupture and slip of subduction zones around the world is still poorly constrained. Reasons for this lie in the challenging nature of such investigations, because the shallow extent of subduction zone earthquakes lies at sea and well below the ocean floor. Limited knowledge of this shallow earthquake extent reduces the chance of meaningful earthquake and tsunami hazard assessment and thus damage mitigation. Because earthquakes are friction phenomena, a large body of work in earthquake research is based on laboratory friction experiments. Early friction experiments have shown that repetitive frictionally unstable stick-slip sliding on artificial faults in the laboratory represents the small-scale equivalent of earthquakes on faults in nature. Friction on a fault evolves with velocity, slip, and time (rate- and state-dependent friction) and thus can lead to unstable sliding. Unstable sliding includes periods of fault locking and accumulation of elastic energy, with intermittent periods of fault rupture and slip, which releases the stored energy. The depth interval on the megathrust fault that is capable of unstable frictional sliding and thus earthquake nucleation is called the seismogenic zone. Crucial to estimating the extent of the seismogenic zone is knowledge of the variation of the velocity-dependent frictional behavior with depth. Especially the velocity-dependent frictional behavior at plate tectonic rate has shown to be crucial. This information can be derived from laboratory friction experiments and application of so-called rate- and state-friction laws. Ideally, such experiments should be conducted on fault-zone material. However, such material is difficult to obtain and its availability is very limited. Subduction zone input materials, which are the marine sedimentary column on the subducting plate, are less difficult to recover and hold important information on where a megathrust forms or what intrinsic frictional behavior the fault-forming material has. Measurements on input material are therefore a valuable alternative to measurements on fault zone material. This thesis presents the results of laboratory friction experiments at room temperature, under relatively low pressure, and driven at velocities starting from plate rate. These experiments were designed to investigate the frictional behavior of subduction zone input sediments and its implications on the fault slip behavior and seismic potential of the shallow portions of two subduction zones. The first is the northern Cascadia subduction zone, located along the West coast of North America, where a major earthquake is about to be due. The second is the North Sumatra subduction zone, a region of the Sunda subduction zone and the location of recent destructive earthquakes and tsunamis. At northern Cascadia, the megathrust has so far not been sampled. Based on measurements of frictional strength contrasts in the input sedimenatry column, we propose that the megathrust fault will likely form in a weak illite-rich hemipelagic clay near the top of the oceanic basement. Because this inference is in good agreement with interpretations of seismic imaging, we focused on the frictional behavior of this specific material. The absence of shallow non-destructive slow slip events at northern Cascadia has recently been interpreted to result from a megathrust that is locked and potentially seismogenic all the way to the trench. In contrast, the results presented in this work indicate that the shallow part of the megathrust is not capable of producing slow slip events nor capable of locking and thus likely not seismogenic. However, our friction data also indicate low resistance to a propagating earthquake nucleating at greater depth. This low resistance is evident from substantially elevated pore pressure, low frictional strength, and low cohesion. Therefore, the northern Cascadia subduction zone holds the potential of shallow earthquake slip and tsunamigenesis. At North Sumatra, seismic slip during the 2004 Aceh-Andaman subduction zone earthquake was unexpectedly shallow and resulted in a devastating tsunami. Recent work suggested that the cause is a very shallow seismogenic zone that may be created by diagenetic strengthening of fault-forming input sediments prior to subduction. This thesis presents the results of laboratory friction experiments designed to test this hypothesis. We showed that input sediments to the North Sumatra subduction zone exhibit pronounced frictional instability, offering evidence for a frictionally unstable and thus seismogenic shallow megathrust and thus an explanation for shallow earthquake slip in the 2004 event. However, our measurements indicate that the shallow megathrust is not seated in frictionally strong, but in very weak sediments. The combination of weak and unstable sediments is striking because a large number of previous friction studies have established that weak materials under low temperature and pressure conditions are generally associated with stable frictional sliding. This relationship offers an explanation for the observed general lack of seismicity in the shallow portion of subduction zone megathrusts, where unconsolidated, clay-rich, weak materials are typically encountered. We proposed that threshold amounts of dispersed hydrous amorphous silica in otherwise weak and clay-rich sediments are responsible for an unstable sliding character, which can explain the shallow seismicity at North Sumatra. To test the hypothesis that small amounts of hydrous amorphous silica induce unstable sliding behavior, we designed friction experiments on artificial mixtures of weak shale and biogenic opal, a type of hydrous amorphous silica. These experiments revealed pronounced potentially unstable behavior in mixtures with ≥ 30 % opal that had low frictional strength. Based on our results, we proposed that potential unstable sliding at low frictional strength can be explained by the viscous behavior of frictional contacts of hydrous amorphous silica. This highlights the necessity to reevaluate the strength-stability relationship. Our findings support the hypothesis on the role of hydrous amorphous silica in unstable sliding behavior, which has important implications for the potential of shallow seismogenesis at other subduction zones where input sediments contain critical amounts of hydrous amorphous silica. This thesis demonstrates that the northern Cascadia and the North Sumatra subduction zone have very different intrinsic frictional fault slip behavior despite very similar extrinsic properties and attributes, such as temperature or pressure. Thus, intrinsic factors are found to be crucial to the estimation of the slip behavior of shallow megathrust faults, such as a mineral composition of fault material with threshold amounts of hydrous amorphous silica. Hydrous amorphous silica-bearing sediments could form megathrust faults due to intrinsically low strength and potential of overpressure. The shallow portion of megathrust faults formed in such sediments may thus be able to host large and slow earthquakes. This could for instance be the case in the northern Barbados subduction zone, a setting that similar to the North Sumatra subduction zone has been shown to have a porous, overpressured décollement and predécollement consisting of material that contains elevated amounts of hydrous amorphous silica. Thus, this thesis raises the possibility that subduction zones with a shallow seismogenic zone may be more common than predicted by the seismogenic zone model. This inference implies that earthquake and tsunami hazards could be highly underestimated at some subduction zone settings.