Repeating Earthquakes During Multiple Phases of Unrest and Eruption at Mount Agung, Bali, Indonesia, 2017

Abstract

In 2017, Mount Agung produced a small (VEI 2) eruption that was preceded by an energetic volcano-tectonic (VT) swarm (>800 earthquakes per day up to M4.9) and two months of declining activity. The period of decreased seismic activity complicated forecasting efforts for scientists monitoring the volcano. We examine the time history of earthquake families at Mount Agung in search of additional insight into the temporal changes in the shallow crust prior to eruption. Specifically, we analyze the period of declining seismic activity about five weeks prior to the eruption when forecasting uncertainty was greatest. We use REDPy (Hotovec-Ellis and Jeffries, 2016) to build a catalog of 6,508 earthquakes from 18 October 2017–15 February 2018 and group them into families of repeating earthquakes based on waveform similarity using a cross-correlation coefficient threshold of 0.8. We show that the evolution of earthquake families provides evidence that Mount Agung was progressing toward eruption even though overall earthquake rates and seismic-energy-release declined. We find that earthquake families that dominated seismicity during the beginning of the crisis ceased near the onset of tremor on 12 November 2017. Then, earthquake families took on characteristics commonly observed during effusive phases of eruptions on 15 November—a full six days before the first phreatomagmatic eruption on 21 November 2017 and a full ten days before the actual onset of lava effusion on 25 November 2017. We interpret the transitions in seismicity as the manifestation of a three-phase physical model including an Intrusion Phase, a Transition Phase, and a Eruptive Phase. During the Intrusion Phase, seismicity was dominated by VT earthquakes with a relatively high percentage of repeaters (59%) grouped into numerous (65) simultaneous families. During the Eruptive Phase, seismicity included both VT and low frequency earthquakes that grouped into relatively long-lived families despite a low overall percentage of repeaters (14%). The Transition Phase exhibited characteristics of earthquake families between the Intrusion Phase and Eruptive Phase. We conclude that the time history of earthquake families provides insight into the evolution of the stress distribution in the volcanic edifice, the development of the volcanic conduit, and seismogenesis of magma effusion. Finally, we discuss the role that repeating earthquakes could play in real-time monitoring at restless volcanoes. Our work suggests eruption forecasts can be improved by incorporating automatic processing codes to assist seismologists during sustained periods of high earthquake rates, even at sparsely monitored volcanoes.