On the effects of geometry in discrete element numerical earthquake simulations

Computer simulation is a widely-used component of earthquake research, but while many computer models of earthquakes exist, there are none that simulate both sub-fault activity and three-dimensional geometry. I develop a computer model of earthquakes that simulates activity on fault systems with three-dimensional geometry by calculating stress transfer between fault elements as a three-dimensional tensor quantity. This model is a discrete, quasi-static, cellular-automaton type model that generates failure cascade sequences of all sizes. The fault is represented as a collection of rectangular sub-faults or “fault patches” that are not constrained to a two-dimensional plane. Stress transfer is calculated as a tensor field originating from point sources in a linear elastic whole-space, though the effects of normal stress on the friction holding the surfaces of a fault element in place are neglected.; I then develop a procedure for studying the effects of geometry on the evolution of synthetic event histories in a computer model by systematically varying the configuration of a z-shaped or “zig-zag” fault and studying the results using scaling, clustering, correlation, and phase dynamic probability change (PDPC) analysis. I also study the effects of roughness and coupling parameters.; I find that, in the absence of normal stress effects, geometry does not act as a barrier to the development and propagation of events, but that differences in the rate of stress accumulation due to tectonic loading forces do; that geometric roughness does not change the dynamics of the system in a qualitative way; and that the PDPC analysis methodology cannot be effectively applied to simulation data of the quality that can be currently generated.