Doctor of Philosophy

dissertationSeismic catalogs are one of the most important products of seismic network operations, allowing quantitative assessment of event statistics for assessing stress state, and stress transfer and release in the crust. As seismic networks change in scope, and the problems addressed through the use of catalogs expand, new strategies for event detection and catalog building are needed. Here we present new methods for detecting and locating small earthquakes from seismic networks with large station spacing (50-70 km). We use our newly developed method to demonstrate some of the limitations in recovering complete faulting histories with cataloged earthquakes. Our work suggests that small earthquakes, even when they have waveforms similar to those of larger earthquakes, contain valuable information for interpreting local fault structures. We also develop automated strategies for source discrimination. Long-term earthquake catalogs are manually curated to discriminate tectonic events from other event types. The separation of events by source type is important because different sources reflect different physical processes, some of which are anthropogenically driven. We demonstrate that various deep learning architectures are able to replicate analyst decisions (above 99%) and can identify analyst errors in existing catalogs