Numerical Simulations of Slow Slip Events and Stress Build-up in Seismic Cycles of Large Inter-Plate Earthquakes

Recent studies have revealed the occurrence of low-frequency tremor and slow slip events (SSEs) in several subduction zones. In the Nankai subduction zone, SSEs are classified into two types: short- and long-term SSEs, according to the duration of slipping and the recurrence interval. Short-term SSEs have duration of several days and an interval of several months, and occur along the belt-like distribution of tremor. On the other hand, long-term SSEs have duration of more than several months and an interval of several years, and occur only in the Tokai region and the Bungo channel region. As shown in Hirose and Obara (2005), the SSEs are interpreted as an episodic slip event at the deeper extent of the strongly coupled plate interface (i.e., locked region), where large inter-plate earthquakes occur at the interval of 100-200 years. This suggests that the repeating SSEs could cause stress accumulation at the bottom of the locked region. In this study, we numerically simulate short- and long-term SSEs in seismic cycles to evaluate the interaction between the behavior of SSEs and the stress build-up on a plate interface. In our numerical model, we consider a subducting plate interface with a dip angle of 15 degrees in a 2D and 3D semi-infinite elastic medium. The plate interface is divided into rectangular cells. Temporal evolution of the slip velocity on the cells is calculated incorporating the frictional stress on each cell and the elastic interactions between cells. We adopted a rate- and state-dependent friction law with cut off velocity. In our simulation, two different distributions of pore pressure are assumed to reproduce short- and long-term SSEs. In the short-term SSE model, pore pressure increases from the depth of 24 km and becomes close to the lithostatic pressure below 28 km. In the long-term SSE model, pore pressure is kept slightly lower at the depth of 27-30 km than that in the short-term SSE model. In a 2D model, we calculated the case of short- and long-term SSE model, separately. In a 3D model, we assumed short-term SSE model with a patch region of long-term SSE model which has 30 km-width in horizontal direction. In the 2D model, SSEs repeatedly occur at the interval of 7-9 years and 3-7 months for the long- and short-term SSE model, respectively. Both models show that the recurrence interval of SSEs becomes shorter as the next large earthquake approaches. In the 3D model, long- and short-term SSEs are successfully reproduced within a single model. The interval of short- and long-term SSEs also becomes shorter as a large earthquake approaches. The bottom of the locked region gradually migrates upward in a seismic cycle, reflecting the stress build-up by the SSEs and stable sliding. Finally, a large earthquake nucleates at the bottom of the locked region. Our numerical results suggest that SSEs may work as an indicator of stress build-up prior to a large earthquake, although further investigations of SSEs are essential to evaluate whether our models are valid in such a region.