Understanding eruption dynamics: insights from volcanic seismicity in Ecuador

Abstract

Persistently active volcanoes in close proximity to society can pose a huge danger to infrastructure, lives and the economy. Careful monitoring of volcanic seismicity is integral to successful hazard assessment and risk management. Geophysical monitoring at active volcanoes can provide rich datasets to examine internal systems. Specifically, seismic monitoring offers the potential to develop real time analysis and forecasts. The generation of volcanic seismicity has been linked to processes such as magma ascent, degassing and rock fracturing. However, studies are often limited to individual volcanoes or specific episodes of unrest, and so it is difficult to compare interpretations. This aim of this thesis is twofold: to develop methodologies to better quantify and characterise volcanic seismicity, and to use these to provide new understanding of volcanic systems, the hazards they might pose and how we can better forecast and monitor unrest.

First, I present an extensive literature review of our current understanding of volcanic seismicity. As there is no standardised procedure for the analysis of volcanic earthquakes, there are inconsistent uses of techniques and ambiguous terminology. Existing studies also tend to focus on a handful of well monitored volcanoes where dense arrays can be used to calculate source mechanisms and depths to interpret seismic swarms. In order to address this, I develop a thorough signal processing routine which generates a suite of metrics to characterise a single earthquake event. These metrics can be used across a sequence of earthquakes to track changes in the behaviour of seismicity, and distinguish different types of earthquakes. It is developed with poorly monitored volcanoes in mind, as metrics can be determined for signal from a single station, and even a single component instrument. I use parameters in the time domain including amplitude, duration and cross correlation, and compare three different approaches to calculate the quality (Q) factor, in the frequency domain.

I then present two candidate volcanoes to apply the methodology and attempt to describe the internal processes at each. Tungurahua and Cayambe are two relatively understudied volcanoes and yet they are potentially the most dangerous natural hazards in Ecuador. Tungurahua’s most recent eruptive phase (1999-2016) was explosive and persistent. In contrast, Cayambe volcano has not erupted in over 200 years and yet has been seismically restless in recent years. This presents an opportunity to compare the seismicity associated with ongoing, and reawakening volcanic processes.

In chapter 4, I characterise the seismicity atTungurahua between 2012 and the final explosions in 2016. Seismicity at Tungurahua was dominated by long-period (LP) earthquakes, particularly episodes of highly periodic, repeating LP seismicity, known as drumbeats. In this chapter, I show that persistent drumbeats occur in phase with cyclical Vulcanian eruptions. These events are attributed to the initial failure and subsequent resealing of an upper conduit plug. In each explosive episode, the signal metrics are able to distinguish a shift in the signal properties of drumbeat LPs. In chapter 5, I focus specifically on accelerating rates of drumbeat LPs, often considered precursors to eruptions. I use temporal statistics and a Markov chain Monte Carlo (MCMC) approach to model three episodes of drumbeats. In one significant episode, the last ever recorded drumbeats at Tungurahua, I show these events are precursors to a ‘failed’ attempt at an explosion. In chapter 6 I then compare these findings at Tungurahua, with the 2016 seismic crisis at Cayambe. Here I demonstrate the repeating LP seismicity is likely a result of shallow hydrothermal systems, rather than surficial ‘icequakes’ or magmatic ascent. However, swarms of volcano-tectonic events (VTs) in 2016, are likely attributed to stresses on regional faults and ascent of a new pulse of magma. Finally, I begin to explore the complex volcano-tectonic interactions at both Tungurahua and Cayambe.

Where there are high rates of tectonic events globally, and high rates of eruptions, it is important to distinguish causality and coincidence. VT swarms at Cayambe occur two months after the Mw7.8 Pedernales earthquake, 200km west. Using models of static stress change I suggest the crust at Cayambe was subject to a dilational regime, prompting resumed activity in 2016. However, the Pedernales earthquake occurs just two months after the final eruption at Tungurahua and yet does not appear to promote or restrict further explosive activity. This thesis presents case studies of two active volcanoes that are subject to limited seismic monitoring. These methods are not computationally intensive and could be readily adopted into routine volcano monitoring, to further inform hazard assessment. Although Cayambe and Tungurahua are neighbouring volcanoes, comparable in their rheology, they are very different in their current dynamic state, and this is evident in the seismicity. An enhanced understanding of these systems should inform further assessment of seismicity at intermediate-composition, arc volcanoes.