Tectonic plates move at a steady velocity of several centimeters per year, which is intermittently interrupted by great megathrust earthquakes with rapid slip up to tens of meters. These large rupture events are driven by the slow tectonic loading and can substantially alter the deformation rate and state of stress in adjacent regions. With advances in computational algorithms, we develop cross-scale finite-element models that self-consistently integrate long-term motion of entire plates and the intervening space-time evolution associated with great earthquakes. The objectives are two-fold: firstly, to gain insight into the occurrence and magnitude of great earthquakes and their relationship with large-scale tectonic processes, and secondly, to constrain rheological properties of the solid Earth with multi-scale geophysical observations. We begin with formulating a generic subduction model that simultaneously resolves the dynamics of plate motions and megathrust seismic cycles (Chapter 2). Driven by internal buoyancy forces and governed by a nonlinear visco-elasto-plastic rheology, the predicted plate convergence and seismic cycle behavior align with observations in subduction zones. Using an efficient 2.5-dimensional approach, we show that the along-strike resistance arising from slip variations plays a key role in modulating the earthquake magnitude. In Chapter 3, we vary the rupture dimensions, rheological parameters and subduction characteristics to examine their influence on the plate velocity and coseismic slip. In Chapter 4, we build a three-dimensional model tailored to the Chilean Subduction Zone, and reproduce both the long-term motion of the Nazca Plate and post-seismic deformation in the adjacent non-ruptured segment after the 2010 Maule earthquake (Mw = 8.8). Combining these multi-scale geodetic observations as constraints significantly improves the uniqueness of inferred mantle viscosity structure. Following the previous work, we implement a margin-resolving global model of plate motions and earthquakes using the highly scalable finite-element code Rhea (Chapter 5). Constrained by both background plate motions and transient earthquake-related deformations, the model quantifies the sensitivity of different geodetic observations to the rheology of the megathrust, lithosphere and underlying mantle. In Chapter 6, we incorporate a true free surface and link plate motions, surface topography and off-megathrust stress state to Earth's nonlinear rheology within a unified cross-scale model. Our work demonstrates the potential for assimilating multi-scale geophysical observations in unified, physics-based models to better characterize Earth's internal structure and assess seismic hazards.
Jiaqi Fang (Thu,) studied this question.