5. Conclusion
In nature, subduction zone faults are more complicated than what is depicted in our numerical models (fault roughness, multiple asperities, off-fault inelastic deformation etc.). In particular, our choice of type of initial stress distribution (i.e. an ‘asperity’ model) only examines a small part of the potential stress models and it might not be compatible with stress profiles derived from the modelling of the whole seismic cycle. Therefore, future work should focus on the application of the model to a much wider range of heterogeneous initial stress conditions.
Nevertheless, our model based on lithological and depth dependent friction law tuned to the 2011 Tohoku fault region allows us to better understand and reproduce to the first order the different types of (tsunamigenic) earthquakes. Consistently with geophysical observations, the numerical simulations have shown that events with a number of characteristics resembling tsunami earthquakes were generated either near to or below the accretionary wedge. Their rupture area was constrained to remain there due to the fault strength and breakdown energy increasing with depth. We also found that standard thrust earthquakes, with relatively larger stress drops, shorter durations and faster rupture velocities occurred in crystalline rock where both the energy release rate and fault resistance are high. Finally, if the rupture initiated at the bottom of or just below the rock-clay transition and propagated towards the surface and into a zone characterised by low fault strength and frictional resistance, this leads to the production of great thrust earthquakes.