Viscoelastic subduction modelling to understand megathrust earthquake potential

The largest earthquakes, of magnitudes up to 9.5, occur at subduction plate boundaries. At these boundaries, one plate dives below the other as the two plates converge, and the rubbing of the plates at their contact can lead to very large earthquakes, which together release over 80% of the global budget of earthquake energy. These so-called megathrust earthquakes and their associated hazards, including ocean-wide tsunamis, have caused 100s of thousands of deaths, as in Sumatra in 2004 and in Japan in 2011. A long-standing question is why only some parts of subduction boundaries have a record of very large earthquakes, while others do not appear capable of hosting major earthquakes.

Earthquakes occur when elastic stresses, which accumulate at the locked megathrust as the balance of forces at the subduction boundary continuously drives convergence, are suddenly released when frictional fault strength is exceeded by the stress. It is agreed that variations in earthquake potential reflect large-scale differences in elastic loading of the megathrust. Variable loading has been attributed to subduction parameters such as plate convergence velocities, strength of the upper plate, or thickness of sediments on the interface between the plates. However, earthquake repeat times are generally much longer than the duration of our instrumental catalogues, and as a result, statistical correlations of subduction parameters with maximum earthquake size remain inconclusive. An interplay between plate and interface properties probably determines stress build up, and therefore a physical modelling approach is needed to understand how different factors contribute to megathrust earthquake potential.

Local-scale subduction models of visco-elastic plates, tailored to a specific setting by prescribing geometry and convergence velocities, have helped to understand consequences of the earthquake cycle, including surface deformation and fault slip patterns that determine tsunami potential. Such models do not provide insight in how subduction parameters control large-scale and long-term differences in stress loading. Other, larger-scale, models let plate geometry and motions develop dynamically and have helped to understand the force balance and long-term stresses. However, these models usually approximate plates as viscous and neglect elastic stresses. Only now have modelling capabilities matured sufficiently to make running systematic sets of large-scale models of visco-elastic subducting plates feasible. In a 2D pilot study by our team, we derived relationships from models that let us estimate the elastic component of plate bending at actual subduction zones. We found that higher estimates of elastic bending correlated with higher observed relative numbers of large earthquakes (compared to smaller events). This illustrates the promise of such large-scale visco-elastic subduction modelling for understanding megathrust seismic potential.

In the here-proposed project, we will use a state-of-the-art plate-modelling platform (Underworld) that will allow us to, for the first time, run a systematic set of 3D visco-elastic subduction models to characterise the variation of elastic energy storage in the subduction system. By determining the response of plate bending (downdip and along-strike) and upper-plate stress to variations in properties of the subducting plate, upper plate and coupling strength between them, we will test what combination of these properties can explain observed relations between subduction parameters and maximum earthquake size. The relations derived from our models will provide a novel, physics-based, method to estimate of the potential of subduction segments around the globe to host very large earthquakes, including at boundaries without a historic earthquake catalogue.