On the cause of enhanced landward motion of the overriding plate after a major subduction earthquake

We focus on the overriding plate at along-trench distances of hundreds of kilometers from the rupture area of a megathrust earthquake. Here, geodetic observations show increased landward velocities following the 2003 Tokachi, 2010 Maule, and 2011 Tohoku earthquakes. This may give the impression of increased plate coupling or of slab acceleration. Is increased coupling or faster slab sinking required, or are these observations a consequence of common earthquake cycle-related processes? We use kinematically driven finite element simulations to understand how velocity changes similar to the observed ones can be produced in the context of viscous relaxation and afterslip, without changes in interseismic locking or slab sinking. The 3D model has repeated, periodic seismic cycles, a uniform cross-section and predefined, fully coupled asperities. The interface properties are based on parameterizations of slip weakening and strengthening, and viscoelastic rheology. During postseismic relaxation, lateral sides of the megathrust event show increased landward as well as trench-parallel velocities. Displacement during the initial relaxation driven by primary afterslip is sensitive to the assumed distribution and size of asperities. The subsequent 3D viscoelastic relaxation of the mantle wedge produces similar velocities but is not sensitive to the presence and location of other asperities. All relaxation-induced velocities in the overriding plate are independent of velocity changes in the slab. Overall, the models indicate that postseismic deformation driven by slip-induced stress changes produce landward velocity changes over similar lateral distances from the megathrust earthquake source as observed and with comparable amplitude. This suggests that transient changes in interplate coupling and slab velocity might not be required to explain the observed enhanced landward motion. The models also make testable predictions that landward displacements will occur simultaneously and thus over the same timescales as viscous relaxation and afterslip, and that earthquakes with long-lived afterslip will have landward velocity changes only above unlocked portions of the plate interface.