Splay fault propagation in subduction zones

Different fault geometries in the accretionary wedge activated in 40000 years in the geodynamic seismic cycle model. Each dot is a marker with a slip velocity larger than the velocity threshold (2e-9 m/s). Dots are coloured according to time of activation. Black lines indicate the rock contours.


Numerical methods that span a large range of spatial and temporal scales are required to accurately study the seismicity in subduction zones. Whereas the stresses on the fault need hundreds to thousands of years to build up on a tectonic scale, the consequent earthquake rupture propagation is determined by both these initial fault conditions and the feedback of seismic waves over time scales up to minutes. Spatial scales are also challenging, because the stress state of the fault can be affected by the subducting slab on scales of tens to hundreds of kilometers, while reflecting waves can influence (splay fault) rupture propagation in the hundred-meter-scale close to the tip of the wedge. To accurately model the physics involved over al temporal and spatial scales, we couple a geodynamic seismic cycle (SC) model to a dynamic rupture (DR) model. The SC models have the advantage of solving earthquake cycles in a self-consistent manner concerning stress, strength and fault geometry, but lack a high enough spatial and temporal resolution to resolve wave propagation. In contrast, dynamic rupture models solve for dynamic fault rupture and seismic wave propagation, but their initial conditions cannot be constrained in a self-consistent manner. By coupling these two codes, the advantages of both can be exploited. The initial stresses and geometry from a reference megathrust rupture from the SC model are used as input in the DR model, resulting in the spontaneous nucleation of dynamic rupture. This fully resolved earthquake is qualitatively similar to its unresolved SC equivalent in terms of stress drop and upward rupture propagation. To explore the effects of the differences between the two models on the subsequent rupture behaviour in the DR model, a comparison of models with different initial stress conditions and off-fault plasticity is presented. We exploit the advantages of our coupled model by studying when and how often a rupture favours propagation on the splay fault over the megathrust. This question of rupture path selection is of importance when assessing a region’s tsunami hazard. The SC models show that subduction zones with a larger sediment thickness and with relatively weaker sediments favour splay fault formation with corresponding splay fault ruptures. Our coupled method allows us to verify whether these conclusions and splay fault selection in general, are influenced by seismic waves reflecting strongly within the confined accretionary prism.

Snapshots of a megathrust event in the geodynamic seismic cycle model (top) and dynamic rupture model (bottom) showing the stress change with respect to the start of the event with identical color scales half way through the event. The fault and surface are indicated in black. The time since the start of the event is indicated in the top right corner. Both ruptures propagate upwards and have similar stress drops. There is an 8 order of magnitude difference between the duration of the events of SC and DR.


This project is in collaboration with Alice-Agnes Gabriel, Stephanie Wollherr, and Elizabeth Madden (LMU Munich).

Back to Research