Tsunamigenic earthquakes from tectonics to dynamic rupture

Earthquakes on large thrust faults in subduction zones can cause tsunamis with devastating consequences. Therefore, it is important to understand how these events nucleate, propagate, and generate tsunamis. As it is difficult to achieve this with the limited number of direct measurements of the conditions in subduction zones, many studies have attempted to use observations of earthquake and tsunami occurrence, or numerical modelling techniques to shed light on the tsunamigenic earthquake process. The aim of this thesis is to increase our understanding of tsunamigenic earthquakes and to expand the current numerical modelling state-of-the-art to facilitate research into tsunamigenic earthquakes that takes into account geodynamics, dynamic rupture, and tsunamis. First, I present a data-driven study into the occurrence of tsunamigenic earthquakes. Using a newly compiled database including geodynamic, earthquake, and tsunami characteristics, I find that tsunamigenic earthquakes may be more prone to occur at subduction zones where plates subduct relatively fast at a sediment-starved, erosional margin. In addition, I present several modelling advances in this dissertation to study tsunamigenic earthquakes from geodynamic to dynamic rupture timescales. Taking these timescales into account is important, because the large-scale tectonics determine the stress heterogeneity and build-up in a subduction zone. The latter is released during earthquakes. I present a two-dimensional modelling framework in which I couple a geodynamic seismic cycle (SC) model to a dynamic rupture (DR) model and a tsunami propagation and inundation (TS) model. Using the output of the SC model as input for the DR model reduces the amount of assumptions that enter the DR model. Indeed, the resulting DR model is physically consistent and contains spontaneous nucleation and rupture arrest due to the tectonic stresses. I also present extensions of this modelling framework by including off-fault plasticity in the DR model and by coupling the SC model to a three-dimensional DR model and a two-dimensional TS model. I find that material heterogeneity in subduction zones leads to complicated rupture dynamics, with crack- and pulse-like behaviour, and supershear rupture speeds. The accretionary wedge, which consists of weak sediments, traps seismic waves. This leads to the continuous reactivation of fault slip on the shallow part of the megathrust, resulting in increased shallow slip. Splay faults in the accretionary wedge can be activated by the main megathrust rupture when they are favourably orientated with respect to the stresses. They can also be activated by dynamic stress changes caused by wave reflections from the surface. Splay fault rupture localises the vertical surface displacements in high-amplitude peaks. Therefore, the resulting waves have larger amplitudes and smaller wavelengths than waves resulting from a pure megathrust rupture. The innovative, interdisciplinary research presented in this thesis is a step forward in our understanding of and ability to model tsunamigenic earthquakes. Future studies can use this a building block to, hopefully one day, solve the enigma of tsunamigenic earthquakes.