Towards continuum models of lateral rupture propagation in a segmented megathrust

Abstract:

At subduction megathrusts, propagation of large ruptures may be confined between the up-dip and down-dip limits of the seismogenic zone. This opens a primary role for lateral rupture dimensions to control the magnitude and severity of megathrust earthquakes. The goal of this study is to improve our understanding of the ways in which the inherent variability of the subduction interface may influence the degree of interseismic locking, and the propensity of a rupture to propagate over regions of variable slip potential. We focus in particular on the roughness of the incoming seafloor, which we expect to be of primary importance.

The global absence of a historic record sufficiently long to base risk assessment on, makes us rely on numerical modelling as a way to extend our understanding of the spatio-temporal occurrence of earthquakes. However, the complex interaction of the subduction stress environment, the variability of the subduction interface, and the structure and deformation of the crustal wedge has made it very difficult to construct comprehensive numerical models of megathrust segmentation. We intend to develop and exploit the power of a plastic 3D continuum representation of the subduction megathrust, as well as off-megathrust faulting to model the long-term tectonic build-up of stresses, and their sudden seismic release.

The sheer size of the 3D problem, and the time scales covering those of tectonics as well as seismology, force us to explore efficient and accurate physical and numerical techniques. So far, we have focused our efforts on developing a staggered grid finite difference code that makes use of the PETSc library for massively parallel computing. The code incorporates a newly developed automatic discretization algorithm, which enables it to handle a wide variety of equations with relative ease. What remains now is combining the physics that act on the different spatial and temporal scales. To this end we explore new constitutive models that enable adaptive time stepping, artificial damping of waves, and allow us to make a stable transition from implicit to explicit solution methods. We will present preliminary 2D modelling results of rough seafloor subduction indicating that seamounts may cause down-dip segmentation of the megathrust, and outline our strategy for building a comprehensive 3D setup.