Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
 Continuous magnetotelluric (MT) measurements were conducted at Sakurajima volcano, Japan, revealing syn-eruption electric pulses (and sometimes accompanying geomagnetic pulses). Movies of the eruptions, recorded with timing provided by a GPS clock, show a large number of volcanic lightning flashes. Some MT pulses occurred simultaneously with lightning flashes. Pulses were observed more than 10 seconds after the onset of the eruption, and tend to occur during eruptions that emit volcanic ash to high altitudes. Pulses were more common during mild eruptions rather than during Vulcanian eruptions. The observations suggest that the dominant mechanism of volcanic lighting is similar to that of lightning in thunderstorms, in that it requires the collision of particles and subsequent separation of positive and negative charge.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 One useful approach in this regard is to characterize the features of volcanic lightning based on long-term observations. In this study, we analyze lightning at Sakurajima volcano, Japan, recorded by ground-based observations. The features of the electromagnetic signal are investigated with the aim of determining the mechanism of lightning formation.
 Continuous magnetotelluric (MT) observations were carried out at Sakurajima volcano in southern Kyushu, Japan. Sakurajima is an andesite–dacitic volcano characterized by Vulcanian eruptions (∼8000 explosions in the past 50 years), mild eruptions, and continuous ash emissions. Since 1955, eruptions have occurred at the summit crater (Minami-dake); however, Showa crater, located 500 m east of Minami-dake, started to erupt in June 2006 after lying dormant for 58 years [Yokoo, 2009].
 In 2008, we established two MT sites at Sakurajima (Figure 1) with the aim of detecting temporal changes in subsurface electric resistivity related to volcanic activity (K. Aizawa et al., Temporal changes in electrical resistivity at Sakurajima volcano from continuous magnetotelluric observations: Possible indication of lateral magma degassing, submitted to Journal of Volcanology and Geothermal Research, 2010). Naturally occurring geomagnetic fields and earth electric currents were measured using Phoenix MTU5 systems with a GPS clock. Geomagnetic fields were measured by induction coils, and electric currents (voltage difference) were measured by wires to which electrodes were attached at each end. In this study, we analyze continuous 15 Hz data from which high-frequency variations have been sharply decreased using an anti-aliasing filter of the data logger. At a site located 100 m from KURMT observation site, a GPS-clock-synchronized movie, whose one frame image correspond to 1/30 seconds, was continuously recorded on HD/DVD video decks [Yokoo et al., 2009].
3. MT Pulses
 We analyzed 94 eruptions at Showa crater from May 2008 to July 2009, during which time both MT data and movies were recorded. We applied a Fourier transform to the raw 24-bit MT time series, and incorporated the frequency response of the MTU5 system. Next, the time series (in physical units of mV/m and nT) were obtained by an inverse Fourier transform. DC and low-frequency variations were then removed by applying a moving average with a 2 sec time window.
Figure 2 shows an example of the MT data and snapshots of a movie of an eruption. Lightning flashes are observed, synchronous with a pulse (hereafter termed an MT pulse) in the MT record (Figure 2d), indicating the MT pulse is generated by the lightning. A close examination of the pulses reveals that they occur at one or two sampling points in the 15 Hz time series. The vertical magnetic field (Bz) is insensitive to the lightning.
 First, we assessed the morphology of lightning captured in the movies. In the case of nighttime eruptions, the lightning is easy to detect, although its shape is not resolved (Figures 2a–2c). For daytime eruptions, the shape of lightning can be determined, but it is rarely visible (we captured only 12 lightning strikes in 77 daytime eruptions). We made the following observations with regard to simultaneous pulses and lightning strikes: (1) in the nighttime movies, the lightning flashes are relatively bright as compared to those without MT pulse, (2) in the daytime movies, the lightning appears to occur as cloud-to-ground (CG) discharge rather than intra-cloud (IC), (3) lightning flash are recognized in 1 frame image (∼1/30 s) in the daytime movies, while in 3 frames images (∼3/30 s) in the nighttime movies (the difference is due to the automatically changes of camera sensitivity.)
 Next, we evaluated the origin time, amplitude, and arrival directions of pulses. In this study, an MT pulse is defined in the case that the amplitude of Ex or Ey exceeds 0.002 (mV/m). Because MT data collected at KURMT are of relatively low quality, we only used HARMT data in defining the origin time of each pulse. According to the above criteria, pulses were identified in 36 of 94 eruptions, amounting to a total of 105 pulses.
Figure 3 shows a rose diagram of the arrival directions of MT pulses derived from the amplitude ratios (Ex/Ey and Bx/By). Long axes in each diagram reveal that electric current flows approximately along the direction of the crater, whereas the geomagnetic field is mainly oriented perpendicular to the electric current.
4. Mechanism of MT Pulse
4.1. Macroscopic View
 Lightning generates electromagnetic waves in a wide frequency range primarily in the ELF and VLF bands [Rakov and Uman, 2003]. Consequently, electromagnetic induction in the instruments and cables [McNutt and Davis, 2000] may be proposed as the cause of observed pulses. However, this is unlikely in the present study, as Bz is insensitive to lightning (Figure 2) and there exists a dominant arrival direction (Figure 3). Alternatively, it is possible that only the electric pulses are caused by induction acting to the earth, whereas geomagnetic pulses are radiated by the lightning current in the air.
 An alternative mechanism may be formulated based on the observation that only CG lightning accompanies the pulses. CG lightning directly injects electric current into the earth, generating a voltage difference at any two arbitrary points. This phenomenon is known as a “step voltage” [Kitagawa et al., 2005]. According to the step voltage model for a simple 1D uniform earth, the relationship between E and B can be expressed based on the assumption of magneto-static approximation associated with a vertical electric current [Rakov and Uman, 2003], as follows:
where I is the electric current of CG lightning, r is the distance between the point struck by lightning and the observation site, Hm is the vertical length of the lightning, and R(Hm) is the distance between the lighting top and the observation site. At HARMT, the average amplitudes of the MT pulses in the E and B fields in the 15 Hz sampling records are approximately 10−5 (V/m) and 2 · 10−11 (T), respectively. Substituting r and R(Hm) = 3000 (m), Hm = 150 (m), and observed E(r) and B(r), ρ is calculated to be approximately 100 (ohm-m), which appears to be a reasonable value as an average of Sakurajima.
 In equation (2), I is estimated to be 6 (Ampere), which is too small relative to the typical value of 10 k∼100 k (Ampere) observed for return strokes generated by thunderstorms, which last tens to hundreds (micro-seconds). 6 (Ampere) is also small relative to the typical value of tens to hundreds (Ampere) observed for continuing currents generated by thunderstorms, which continues tens to hundreds (milli-seconds) [Rakov and Uman, 2003]. However, it should be noted that the anti-aliasing filter of MTU5 cannot completely decrease the amplitude of short duration pulse. For example, the amplitude of pulse lasting for 100 (micro-seconds) is decreased to approximately 1/1000 in the 15 Hz sampling records. Therefore, the 6 (Ampere) could be a 6 (kilo-ampere) if the lightning current last only 100 (micro-seconds).
4.2. Microscopic View of Particle Electrification
 Various electrification mechanisms have been suggested for within volcanic plumes, including ice–hail (frozen volcanic gas) collisions, as observed in thunderstorms, groundwater–magma interaction within the volcanic edifice, collisions among ash particles, and fragmentation of magma (see the review by Mather and Harrison , James et al. ). Among these mechanisms, ice–hail collision is unlikely to occur at Sakurajima because of the warm weather at the site (annual range of daily average temperature of 10–30°C).
Figure 4 shows the relationship between the occurrence of pulses and the initial velocity of rising cloud, maximum cloud height, and maximum infrasound at Arimura (see Figure 1 for the location). We found no significant correlation between the occurrence of MT pulses and background meteorology, such as temperature, atmospheric pressure, humidity, and wind speed. The pulses tend to occur during eruptions that emit volcanic ash to high altitudes. The initial velocity appears to have little influence on the MT pulses. Interestingly, pulses were not observed during Vulcanian “explosions” associated with strong infrasound waves. Indeed, nighttime movies reveal that visible lightning flashes tend to occur during mild eruptions rather than during Vulcanian explosions (Movie 1). The water content of magma may influence the generation of volcanic lightning [Williams and McNutt, 2005], just as it controls the explosivity of eruptions [e.g., Eichelberger et al., 1986]. Therefore, it is intriguing to consider that the relationship between MT pulses and infrasound strength (Figure 4c) could be explained in terms of the water content in magma. However, this possibility is unlikely at Sakurajima, where a gas pocket that developed beneath the crater is the most likely control on the style of eruption [Iguchi et al., 2008; Yokoo et al., 2009].
Figure 5 shows the temporal distribution of pulses throughout the observation period, relative to the time since eruption onset (defined as the time when the ash plume emerged at the crater rim). Following the step voltage model, the arrival direction of a pulse (Figure 3) corresponds to the polarity of an electric charge that reached the ground. In the case that CG lightning provide negative charge to the ground, it is conventionally referred to as negative lightning. We distinguished positive and negative pulses by the arrival direction of the electric field at HARMT. Ten percent of the recorded pulses are not shown because they indicated a direction oriented away from the crater, which we are unable to explain, although it may reflect the non-concentricity of lightning current in the earth. Figure 5 suggests that both positive and negative charges exist in the volcanic cloud, and that negative lightning is dominant (70% of cases). These results imply that a large negative charge is located in the lower part of the volcanic cloud at Sakurajima. This interpretation is consistent with the results of measurements of the vertical electric field at Sakurajima [Lane and Gilbert, 1992; Miura et al., 2002]. Hoblitt  showed that the polarity progresses from one sign to the other during the evolution of the large eruptions at Redoubt volcano. In contrast, the polarity seems to change randomly in Sakurajima.
 An important finding is that the MT pulses rarely occur within 10 seconds of eruption onset, suggesting that a preparation process, such as charge generation and/or separation, is required for CG lightning and corresponding MT pulses. We speculate that the continuous collision of particles during plume growth is the dominant mechanism in producing CG lightning and the corresponding MT pulses. If the groundwater–magma interaction and/or magma fragmentation are assumed to be the electrification mechanism, the particles are charged prior to the emergence of ash plume, and MT pulses should also be observed within 10 seconds after the eruption onset. The electrification by continuous collision of particles may be supported by the relationship between explosivity and occurrence of pulse (Figure 4). In the case that fragmentation of magma at eruption onset is a dominant mechanism of electrification, MT pulses should also be observed during explosive eruptions. The results of laboratory experiments on particle collisions suggest that particle size controls the sign of the generated electric charge [e.g., Lacks and Levandovsky, 2007]. In the case that this size-dependence is dominant at Sakurajima, as suggested by Miura et al. , the quiet 10-second period may represent the time during which particle separation (including gas and aerosol) by gravity has yet to fully develop.
 We thank NHK for permitting the use of eruption movies. Infrasound data were provided by MLIT. The careful review by two anonymous reviewers improved the manuscript. A. Y. acknowledges the assistance of a JSPS Research Fellowship.