This dissertation presents a series of algorithms and plans for implementation of a proto- type earthquake early warning system by merging two seismic observation tools: strong motion accelerometers and Global Navigation Satellite Systems (GNSS). The seismogeodetic approach, that which optimally merges these disparate data types, allows a reliable estimation of the earth- quake source for early warning procedures from event detection to magnitude estimation and higher order products. We address implementation of this system in a real-time, automated environment, designed to operate without manual oversight. We assess the capabilities of low- cost micro-electro-mechanical systems (MEMS) accelerometers merged with observatory-grade GNSS for seismic P-wave detection, and suggest statistical methods to determine whether P-wave detections are consistent between stations, which allows for the automated removal of poor quality detections to avoid propagating these errors to higher order warning products.
In line with these goals, we demonstrate how seismogeodetic observations contribute to our view of initial rupture dynamics of large earthquakes. We address an open question in seismology regarding the deterministic nature of earthquakes, assessing how early it is possible to fully characterize the source parameters of large, damaging earthquakes. Our seismogeodetic observations suggest that final earthquake magnitude is not discernable from the first few seconds of observation and therefore the earthquake rupture process is not strongly deterministic. We further investigate the complete temporal evolution of seismic moment release to identify the earliest magnitude-dependent features in our observational dataset. We create synthetic rupture models to identify the physical basis responsible for the timing of observed magnitude-dependent qualities. Our findings suggest that earthquake magnitude can be estimated prior to rupture completion, consistent with a weakly deterministic rupture process. These results provide new insights into the best practices for early warning and rapid response, and suggest limitations on the timeliness of earthquake source characterization.
Finally, we turn to the built environment to demonstrate the applicability of these multi- instrument sensors to long-term structural health and seismic monitoring, to bolster earthquake and other natural hazard response practices. We describe efforts to characterize local buildings by creating a baseline model for the healthy structure and propose methods by which structural health can be evaluated over time or following a major seismic event to assess building safety without the need for manual inspection. We again assess the capabilities of low-cost MEMS accelerometers and describe the limitations of this lower-quality instrumentation for structural monitoring applications.
