Investigating the Origin of Continual Radio Frequency Impulses During Explosive Volcanic Eruptions
S. A. Behnke, H. E. Edens, R. J. Thomas, C. M. Smith, S. R. McNutt, A. R. Van Eaton, C. Cimarelli and V. Cigala, 2018
Map showing instrument sites for the Sakurajima volcanic lightning study. Lightning Mapping Array sensor sites are indicated by circles. The sites located on Sakurajima are Sabo North (SN), Sabo South (SS), Kurokami (K), Okuyama (O), and Arimura (A). The sites located on the mainland are Taniguchi (TN), Nakanochaya (N), Astronomy (AS), Forest (F), and Tarumizu (T). All nonautonomous sensors were located at Kurokami. Infrasound and seismic sensors were located at Kurokami and near the Sabo North site. The autonomous slow antenna was also located near the Sabo North Lightning Mapping Array site.
Volcanic lightning studies have revealed that there is a relatively long-lasting source of very high frequency radiation associated with the onset of explosive volcanic eruptions that is distinct from radiation produced by lightning. This very high frequency signal is referred to as “continual radio frequency (CRF)” due to its long-lasting nature. The discharge mechanism producing this signal was previously hypothesized to be caused by numerous, small (10 – 100 m) leader-forming discharges near the vent of the volcano. To test this hypothesis, a multiparametric data set of electrical and volcanic activity occurring during explosive eruptions of Sakurajima Volcano in Japan was collected from May to June 2015. Our observations show that a single CRF impulse has a duration on the order of 160 ns (giving an upper limit on discharge length of 10 m) and is distinct from near-vent lightning discharges that were on the order of 30 m in length. CRF impulses did not produce discernible electric field changes and occurred in the absence of a net static electric field. Lightning mapping data and infrared video observations of the eruption column showed that CRF impulses originated from the gas thrust region of the column. These observations indicate that CRF impulses are not produced by small, leader-forming discharges but rather are more similar to a streamer discharge, likely on the order of a few meters in length.
Plain Language Summary
This paper investigates the origin of a radio frequency (RF) signal previously termed “continual RF” that has been detected from the ash plumes of explosive volcanic eruptions. Continual RF had been hypothesized to be caused by many very small lightning-like discharges on the order of 100 m in length occurring at the vent of the volcano. To test this hypothesis, we conducted a multiparametric observational experiment at Sakurajima Volcano in Japan. We measured continual RF during eruptions of Sakurajima in 2015 and found that continual RF is a collection of short duration RF impulses, on the order of 160 ns in duration. These RF impulses are distinct from those typically produced by lightning. We also observed small, 30-m lightning discharges occurring near the volcanic vent that were distinct from the source of continual RF. Our observations show that continual RF is caused by an electrical discharge that is more simple than even a very small (∼100 m) lightning discharge. We conclude that continual RF is caused by numerous, small, electrical discharges similar to streamer discharges, on the order of a few meters in length.
Comparison of visible light and infrared images. (a) Close-up view of Showa crater from Kurokami. The arrow indicates the eastern crater rim. (b) Infrared image of Showa crater, with arrow pointing to the eastern crater rim. Bright foreground features are trees that were located near the camera. (c) Infrared image taken during eruption. (d) Composite of images shown in (b) and (c). Color scales on (b)–(d) indicate temperature, and scaling is dynamic.
Volcanic plumes are electrically charged and often produce lightning, which has been documented by many studies (Aizawa et al., 2016; Anderson et al., 1965; Arason et al., 2011; Behnke & McNutt, 2014; Behnke et al., 2012, 2013, 2014; Bennett et al., 2010; Brook et al., 1974; Cimarelli et al., 2016; Hoblitt, 1994; McNutt & Davis, 2000; McNutt & Williams, 2010; McNutt et al., 2010; Miura et al., 2002; Thomas et al., 2007, 2010; Van Eaton et al., 2016). One useful means of studying lightning is through measurements of very high frequency (VHF) radiation, of which lightning and other processes of electrical breakdown are copious producers. Previous studies of lightning in volcanic eruptions using VHF antennas have detected a source of VHF radiation that is distinct from lightning and is associated with the explosive ejection of ash (Behnke et al., 2012, 2013, 2014; Thomas et al., 2007, 2010). This radiation source is referred to as continual (or continuous) radio frequency (CRF) and manifests as relatively high rates of VHF radiation impulses (several thousands to over 10,000 impulses per second; Behnke et al., 2013) that occur over durations on the order of seconds or longer. By contrast, lightning is a discrete phenomenon, occurring on timescales on the order of tens to hundreds of milliseconds; thus, the VHF emissions from lightning are similarly short lived. The relatively long duration is one of the key characteristics that distinguish CRF radiation sources from those produced by lightning.
Observations of CRF have been made by Lightning Mapping Array (LMA) sensors (Rison et al., 1999; Thomas et al., 2004) during eruptions of Augustine Volcano (2006; Alaska, United States; Thomas et al., 2007, 2010), Redoubt Volcano (2009; Alaska, United States; Behnke et al., 2012, 2013), and Eyjafjallajökull (2010; Iceland; Behnke et al., 2014). Based on data collected during the Augustine eruption, it was hypothesized that CRF was caused by numerous, small (10 – 100 m), leader-forming discharges occurring at or near the vent of the volcano (termed “vent discharges”), which implied that the volcanic ejecta were charged upon eruption. The basis of this hypothesis was that the CRF radiation sources were relatively powerful, similar to the VHF emissions detected from discrete lightning discharges. Thus, it seemed most plausible that a similar process was producing the CRF, just on a smaller scale and at high rates over a long duration. The follow-on study at Redoubt confirmed that CRF occurred simultaneously with the onset of an explosive eruption and originated at low altitude, likely near the vent in the ash column. The data obtained at Eyjafjallajökull showed that there was a variation in the intensity of CRF among different eruptive events at the range of the VHF antennas. No further insight about the mechanism-producing CRF was gained from these subsequent studies.
It is significant that the overall, macroscopic form of the CRF signal (the collection of impulses at high rates over durations of seconds or tens of seconds) is conspicuous and distinct from VHF radiation produced by lightning (bursts of impulses lasting several hundreds of milliseconds). Since this CRF signal is not known to occur during meteorological thunderstorms, detection of CRF would be an unambiguous indicator of explosive volcanic activity. Thus, knowledge of the source mechanism would be useful for the application of VHF observations to volcano monitoring. Further, since the ash-charging mechanisms are processes inherent to magmatic fragmentation (e.g., fractoemission; Dickinson et al., 1988; James et al., 2000), turbulent ash flow, and eruption column dynamics (e.g., collisional charging; Cimarelli et al., 2014; Houghton et al., 2013; Mendez Harper & Dufek, 2016), study of the electrical activity may add to the understanding of explosive eruptions. CRF is also interesting from a lightning science perspective, as it provides insight into electrical discharge processes that produce VHF radiation.
This paper reports on the results of a multiparametric field campaign at Sakurajima Volcano in Japan that was designed in part to test the hypothesis that CRF is caused by small, leader-forming electrical discharges. In that regard, the main goals of the campaign were to see if (1) there was a detectable electric field change typical of leader stepping associated with CRF and (2) if there was any substantial visible light produced by CRF. The point of the second goal was to reconcile numerous visual observations and photographs of lightning near the vents of many volcanoes during explosive eruptions with VHF observations of electrical activity.
Figure 1 shows a map of the locations of the instruments that were installed around Sakurajima Volcano in May of 2015. Some of these instruments, like the LMA (indicated by circles), ran autonomously and collected data continuously over several months. Other instruments, such as high-speed video cameras, were manually triggered during an approximately 10-day period of 24-hr observations. These observations took place at the Sakurajima Volcano Observatory branch at Kurokami, which is located on the east side of Sakurajima Volcano and has an unobstructed view to Sakurajima’s contemporaneously active Showa crater. In this section we describe the instrumentation that provided data for the analysis in this paper; however, we note that the experiment also involved seismic measurements, which will be reported on in separate papers.
A nine-station VHF LMA was deployed around the volcano such that each station had a line-of-sight view to Showa crater. Each lightning mapping station detects peaks in the logarithmically detected VHF waveform (66 to 72 MHz bandwidth), which is sampled at 25 MS/s. Global Positioning System (GPS) timing is used to time tag the peaks. The data from the sensors are then processed using time-of-arrival methods, which gives the locations of the VHF sources (the peaks in the waveform). The overall timing accuracy for the LMA system is about 30 ns (Thomas et al., 2004). In this study, the LMA was used both for locating sources of VHF radiation produced by electrical activity and for analysis of CRF by making use of the unprocessed data recorded by the individual stations. To maximize the location accuracy in the vertical dimension, LMA stations were deployed at locations of varying altitude (Figure 1); the altitude of the sites located on the mainland ranged between 60 and 630 m, while the sites on Sakurajima ranged in altitude between 40 and 105 m. Note that these and all altitudes reported in this paper are GPS altitudes. The LMA was operating in 10-μs mode, meaning at most one peak is detected in each successive 10-μs time window.
A tenth lightning mapping station was located at the Kurokami observatory for the purpose of directly record- ing the log-RF waveform, which is normally peak detected by the LMA. This is achieved by splitting the output from the LMA’s log amplifier and digitizing the log-detected waveform separately. In this configuration the LMA still also peak detects the waveform; therefore, this station also contributed to the determination of source locations. The waveform was digitized at 25 MS/s, and the manually triggered records were 6.7-s long (all pretrigger). Timing was acquired by digitization of the GPS pulse per second.
In addition to the log-RF antenna, a set of instruments that measure electric field changes, which are referred to as slow and fast antennas, were also located at the Kurokami observatory. Data collection was triggered manually. These data were digitized at 5 MS/s, and data records of 6.7 s (also all pretrigger) were recorded. The slow and fast antennas had time constants of 1 s and 0.1 ms, respectively. Continuous electric field change data were collected by an autonomous slow antenna, which was located near the Sabo North LMA station (SN in Figure 1). This instrument had a time constant of 15 s and a sample rate of 50 kS/s. A GPS receiver provided timing. The amplitudes of the slow and electric field change data represent the electric field but are presented in units of the instrument output voltage. The instrument is linear in response; thus, the electric field change is related to the output voltage by a calibration factor. We present the waveform amplitudes in arbitrary units.
Two high-speed video cameras were operated at the Kurokami observatory. The first was a Phantom v7.3 manufactured by Vision Research. This camera has a CMOS sensor of 800×600 pixels, GPS timing, and 16 GB of RAM. A frame rate of 6,400 frames per second (fps) and a lens of 105-mm focal length were used. The strategy for this camera was to collect the first few seconds following the onset of an explosive event in the hope of recording optical emission from CRF. The strategy for the second camera, a Phantom v711, also manufactured by Vision Research, was to trigger on lightning. This camera has a CMOS sensor of 1, 280 × 800 pixels, 32-GB RAM, and GPS timing. Frame rates of 3,000 (for night use) and 5,000 (day use) fps were used. The camera was fitted with a lens of 100-mm focal length.
Observations of the eruption columns were made in visible and infrared wavelengths from the Kurokami observatory. A FLIR SC600 (640×480 pixels) was used for the infrared observations, and the data were recorded by two different means. Radiometric data were recorded on a PC using a manual trigger to start and stop recordings with a frame rate of 30 fps. In addition, a continuous stream of the FLIR images was recorded on a digital video recorder (DVR) via the camera’s composite video output at 29.9 fps. A GPS video time inserter (model IOTA-VTI v3) was used to time tag the composite video output. The DVR used was a Lorex ECO4, which had a recording resolution of 960 × 480 pixels and a 1-TB hard disk drive. The time-tagged data were used to obtain time series of the eruption column height using photogrammetric methods with MATLAB software developed by Valade et al. (2014). The timing for the radiometric data was acquired by manually matching frames between the GPS time-tagged composite video data and the radiometric data. Figure 2 shows the field of view of the infrared camera and provides context for the infrared images shown in section 2 by comparing a visible light photograph to infrared images before and during eruption.
In the visible spectrum, video observations were made with a high-sensitivity camera (Watec 910 HX/RC, NTSC model). This camera has a resolution of 768 × 494 pixels and minimum sensitivity of 0.0000025 lux (0.0001 lux corresponds to the light level of the background sky on a moonless night). An auto-iris lens with adjustable 4- to 8-mm focal length was used. The images from this camera were also time tagged with a video time inserter and recorded continuously on the same DVR as described above. The camera ran continuously throughout the field campaign and helped to identify when explosive events occurred.
Infrasound data were collected from two sites: the Kurokami observatory and near the Sabo North LMA station. InfraBSU sensors (manufactured at Boise State University) were used at both sites. The sensors had a range of ±125 Pa and a low cutoff frequency of 0.048 Hz. Three sensors were used per array; one array was located at each site. Data were digitized by a Nanometrics Centaur digitizer, which has a 24-bit Analog-to-Digital Converter. Sample rates of 100 S/s were used. Seismic sensors were also hosted at both the Kurokami and Sabo North sites.