Item

Abstract

Subduction zone megathrust earthquakes may grow into giant M9 events by unzipping the plate boundary fault along a length larger than its seismogenic thickness. Existing hypotheses on a physical control of how big an earthquake might grow in specific settings have been falsified by nature: The giant 2004 Sumatra and 2011 Japan earthquakes occurred where slab age, subduction velocity and interplate roughness were regarded indicative of not generating giant events. Similarly, geometric features or creeping zones have been shown to not necessarily form a barrier to rupture propagation. The question of which subduction zone is prone to giant earthquakes is therefore replaced by the question of how likely a megathrust earthquake may grow into a giant event in any given subduction zone. Assuming that subduction zone type plate boundary faults are segmented into locked/velocity weakening and creeping/velocity strengthening regions (potential asperities and barriers, respectively), we here argue that giant earthquakes occur more likely when neighboring fault segments are sufficiently synchronized in their seismic cycles to allow a concerted, multi-segment failure. The Japan 2011 earthquake serves as an example of such synchronized failure of several segments which failed independently in M8 events in the historic past. We investigate the process of fault synchronization in the presence of stress transfer using analytical, statistical and analogue models of subduction zones and define a proxy of synchronization: the "degree of phase locking". The simulation results show that the latter correlates with static stress transfer between the megathrust fault segments which is a nonlinear function of asperity distance and vertical offset. Accordingly, over few tens of simulated seismic cycles the segments can become transiently phase locked (i.e. generate highly synchronized failure of neighboring segments) for stress transfer as little as 1/1000 of coseimic stress drop. These observations are typical of "weakly coupled" oscillators elsewhere in physics and the values of "stress coupling" realized in the models correspond to natural asperity spacing of few tens of kilometers as is typically observed in subduction zones. We therefore argue that while giant earthquakes may theoretically occur in every subduction zone worldwide, they are more likely in those zones where asperities are narrowly spaced (< 100 km) and seismic cycles highly synchronized. Both states are potentially observable in nature by means of geodetic and paleoseismological methods.