Toward multidisciplinary volcanic eruption models and forecasts in Alaska: contributions from automated seismic and acoustic signal detection

Abstract

Explosive, ash-generating volcanic eruptions pose an increasing threat to a growing, globally connected population. Accurate forecasts of volcanic eruptions remain challenging due to the complex­ity of volcanic systems, but recent multidisciplinary synthesis efforts have proven effective. The National Science Foundation Prediction of and Resilience against Extreme Events (PREEVENTS) project aims to re-analyze and combine multidisciplinary observations at eight Alaska volcanoes to develop eruption forecast models. Leading the seismology and infrasound discipline, this dis­sertation details the development of automated tools capable of producing high-resolution catalogs of volcanic unrest signals from continuous seismic and acoustic data. By leveraging these catalogs and synthesizing insights from other disciplines, we re-examine past eruptions at select Alaska volcanoes and investigate their underlying mechanisms. Chapter 1 provides an overview of volcano monitoring in Alaska, and how different disciplinary insights converge to help us infer pre-, syn- and inter-eruptive processes. Chapter 2 presents an integrated workflow for improving volcanic earth­quake catalogs. Using a combination of standard triggering, cross-correlation clustering, matched- filtering, and earthquake relocation methods, we recover previously undocumented seismic activity and refine interpretations of seismogenic zones at Redoubt and Augustine volcanoes. Chapter 3 introduces the Volcano Infrasound and Seismic Spectrogram Neural Network (VOISS-Net), a ma­chine learning method for detecting and characterizing volcanic tremor and other transient signals. VOISS-Net provides a rapid and consistent means of classifying data in near real time and sum­marizing long-term data. Trained and applied on Pavlof Volcano, VOISS-Net reveals vent-specific seismic tremor profiles. Chapter 4 builds upon this idea of vent-specific unrest, integrating geode­tic, petrologic, gas and thermal remote sensing data. We find that the summit and southeast flank vents at Pavlof Volcano exhibit contrasting eruption dynamics, which we attribute to differences in volatile content, magmatic ascent rate, and conduit geometry. Lastly, Chapter 5 concludes with a discussion of other relevant work and future research directions.