1. Discriminating Types of Volcanic Electrical Activity: Toward an Eruption Detection Algorithm.
- Author
-
Behnke, S. A., Edens, H. E., Theiler, J., Swanson, D. J., Senay, S., Van Eaton, A. R., Iguchi, M., and Miki, D.
- Subjects
VOLCANIC eruptions ,EXPLOSIVE volcanic eruptions ,ELECTRIC field strength ,EXPLOSIONS ,RADIO frequency ,LOGISTIC regression analysis ,ALGORITHMS - Abstract
We present a method for classification of the two distinct types of electrical activity that occur during an explosive volcanic eruption: vent discharges and lightning. Vent discharges occur at the onset of an explosion and create a distinctive radio frequency signature called continual radio frequency. Seconds to minutes after the onset of the eruption, lightning begins to occur throughout the eruption column. We use logistic regression to classify a radio frequency impulse as being part of either a lightning flash or a period of continual radio frequency. The classifier uses the number of peaks in the amplitude envelope from 1 ms windows before and after the impulsive very high frequency waveform, with an average accuracy of 97.9%. We propose that this method could be used in an algorithm to determine when explosive eruptions occurred by identification of the distinctive signatures of vent discharges and lightning. Plain Language Summary: We present a method for automated identification of two distinct types of electrical activity from explosive volcanic eruptions. Explosive eruptions produce lightning, just like thunderstorms. In addition, they also produce small (<4 m) spark‐like electrical discharges at the vent of a volcano, which are called vent discharges. These vent discharges occur for relatively long durations compared to the duration of a typical lightning flash (seconds vs. hundreds of milliseconds) and are thus easily distinguishable in very high frequency (30–300 MHz) electric field measurements. We use logistic regression to classify an electric field impulse as either being part of a lightning flash or a vent discharge. The classifier uses the number of peaks in the electric field signal in 1 ms time windows before and after an electric field impulse. The accuracy of the classifier is 97.9%. We explain that the classifier could be used on a low‐power lightning sensor to automatically identify that an explosive eruption had occurred. We discuss how this capability would enable a new era of volcanic lightning monitoring that would allow for new research into understanding the physical mechanisms of vent discharges to learn how they can be used during the response to an eruption. Key Points: We created a logistic regression model that classifies electrical activity as either vent discharges or lightning with an accuracy of 97.9%The logistic regression model uses the number of waveform peaks in 1 ms time windows preceding and succeeding an RF impulseThe model could be used to identify the characteristic pattern of electrical activity produced by explosive eruptions [ABSTRACT FROM AUTHOR]
- Published
- 2022
- Full Text
- View/download PDF