Using radio frequency signal classification to monitor explosive eruptive activity

Explosive volcanic eruptions often produce a repeatable pattern of electrical activity that can be exploited for volcano hazard monitoring. First, a swarm of small “vent discharges” occurs within the gas thrust region of the plume starting at the onset of an explosion. Vent discharges often persist for several seconds, depending on the duration of the explosion. In addition, vent discharges are known to occur in high-intensity explosions involving the fragmentation of fresh magma. Several seconds after the onset of an explosion, lightning starts to occur throughout the eruption column as charge begins to separate. This chronological sequence of vent discharges followed by lightning has been observed during eruptions from several different volcanoes, including Augustine Volcano, Redoubt Volcano, Eyjafjallajokull, and Sakurajima. In this presentation we demonstrate a proof-of-concept method for an eruption detection algorithm that exploits this common and repeatable pattern. The algorithm leverages a logistic regression classifier to distinguish between radio frequency waveforms of vent discharges and lightning. To demonstrate our method, we use broadband (20-80 MHz) very high frequency (VHF) waveform data of explosive volcanic eruptions from the Minamidake crater of Sakurajima volcano in Japan collected between May 2019 and May 2020. We show that individual VHF impulses produced by vent discharges and lightning can be accurately classified due to differences in the amount of signal clutter surrounding each type of impulse. In particular, we show that impulses from vent discharges are more isolated in time compared to impulses from lightning. The results of the signal classifier are then used to identify the characteristic pattern of volcanic electrical activity to determine if an explosive event has occurred. Implementation of the detection algorithm on an agile and deployable VHF sensor would engender a new method of volcano hazard monitoring, and help facilitate the research necessary to operationalize measurements of volcanic electrical activity in order to inform an eruption response.