Seismic precursors to volcanic explosions during the 2006 eruption of Augustine Volcano: Chapter 2 in The 2006 eruption of Augustine Volcano, Alaska

The 2006 eruption of Augustine Volcano, Alaska, generated more than 3,500 earthquakes in a month-long time frame bracketing the most explosive period of activity. We examine two quantitative tools that, in retrospective analysis, were excellent indicators of imminent eruption. The first tool, referred to as the frequency index (FI), is based on a simple ratio of high- and low-frequency energy in an earthquake seismogram. It is a metric that allows us to quantify the differences between the canonical high-frequency, hybrid, and low-frequency volcanic earthquakes. FI values greater than -0.4 indicate earthquakes classically referred to as high-frequency or volcano-tectonic events. FI values less than -1.3 correspond to events usually referred to as low-frequency earthquakes. Because the FI is based on a ratio and not a spectral peak, hybrid earthquakes are successfully assigned FI values intermediate to these two classes. In this eruption, we find a remarkable correlation between events with FI less than -1.8 and explosive eruptions. The second tool we examine is based on repeating seismic waveforms identified through waveform cross-correlation. Although the vast majority of earthquakes during this eruption have unique waveforms, subsets of events exhibiting a high degree of similarity occur and are closely tied to explosive eruption events. Of the 13 large explosion events, seven were preceded by clusters of highly similar earthquakes. We apply the FI and correlation tools together to identify changes in high- and low-frequency earthquake occurrences and examine their relations to the precursory, explosive, and continuous phases of the eruption. We find that earthquakes that have low FI values and earthquakes exhibiting high degrees of similarity occur almost exclusively within hours of explosive eruptions and postulate that they occur as a result of the final ascent of magma in the volcanic edifice. Because neither of these methods requires analyst-reviewed earthquake locations, we believe that they have considerable potential as automated real-time volcano monitoring tools.