Observations of lightning in relation to transitions in volcanic activity during the 3 June 2018 Fuego Eruption

Satellite and ground-based remote sensing are combined to characterize lightning occurrence during the 3 June 2018 Volcán de Fuego eruption in Guatemala. The combination of the space-based Geostationary Lightning Mapper (GLM) and ground-based Earth Networks Total Lightning Network observed two distinct periods of lightning during this eruption totaling 75 unique lightning flash occurrences over five hours (57 in cloud, 18 cloud-to-ground). The first period of lightning coincided with the rapid growth of the ash cloud, while the second maxima occurred near the time of a deadly pyroclastic density current (PDC) and thunderstorm. Ninety-one percent of the lightning during the event was observed by only one of the lightning sensors, thus showing the importance of combining lightning datasets across multiple frequencies to characterize electrical activity in volcanic eruptions. GLM flashes during the event had a median total optical energy and flash length of 16 fJ, and 12 km, respectively. These median GLM flash energies and lengths observed in the volcanic plume are on the lower end of the flash spectrum because flashes observed in surrounding thunderstorms on 3 June had larger median total optical energy values (130 fJ) and longer median flash lengths (20 km). All 18 cloud-to-ground flashes were negative polarity, supportive of net negative charge within the plume. Mechanisms for the generation of the secondary lightning maxima are discussed based on the presence and potential interaction between ash plume, thunderstorm, and PDC transport during this secondary period of observed lightning.

Therefore, the goal of this study is to characterize lightning observed in a volcanic plume during growth stages of the plume and near the time of a deadly pyroclastic density current (PDC) during the 3 June 2018 Fuego eruption. This is accomplished using a combination of space based Geostationary Lightning Mapper (GLM) 37,38 data, ground based Earth Networks Total Lightning Network (ENTLN) 39 data, and visible imagery from the GOES-16 Advanced Baseline Imager (ABI) 40 . One primary objective is to examine the evolution of the lightning observed throughout the 3 June 2018 eruption to determine how the lightning data aligns with key changes in volcanic activity (e.g., plume growth, PDC generation). Additionally this paper assesses GLM flash lengths and total optical energy of lightning events within the volcanic plume and compares with flash length and flash energy observed in thunderstorms. A final objective is to demonstrate how GLM and ENTLN provide unique information on lightning during an explosive eruption, that when combined, provide a more complete picture on the evolution of lightning during these hazardous volcanic events.

Eruption of Volcán de Fuego, Guatemala
Volcán de Fuego is an active stratovolcano, located in central Guatemala as part of the Fuego-Acatenango volcanic complex [41][42][43] . The Fuego eruption on 3 June 2018 (Table 1) was a relatively short duration explosive eruption that displaced more than 12,000 people and resulted in 165 confirmed deaths and 260 missing, based on reports of the National Coordination for Disaster Reduction of Guatemala (CONRED), Instituto Nacional de Sismología, Vulcanologia, Meteorologia e Hidrología (INSIVUMEH) and the American Red Cross. According to the Global Volcanism Program, the eruption had a volcanic explosivity index (VEI) of 3 out of 8 and the ash plume reached ~ 10 km above the vent. A large portion of the devastation from this eruption was caused by pyroclastic density currents (PDCs), which are a high-temperature mix of gases, ash, lapilli, and bombs. The Fuego PDCs traveled 10 km from the summit at 700 °C with secondary ash plumes reaching 6 km above the ground surface. Volcanic lightning can occur in PDCs 44 , but has not been previously observed by satellites. Increases in lightning due to propagation of PDCs during the 2015 Calbuco eruption in Chile were suggested in an analysis 16 using World Wide Lightning Location Network (WWLLN) 45 data.

Lightning detection
The launch of GOES-16 and GOES-17 GLMs on the GOES-R series of satellites could potentially improve the scientific understanding of lightning in volcanic columns and plumes due to its consistent location in orbit above North, Central, and South America, allowing for the development of more advanced volcano warning systems. The GLM has an 8 km × 8 km spatial resolution at nadir, a 2 ms temporal resolution, and detects lightning in a 1 nm window centered on 777.4 nm 37,38 . The field of view of GLM is ± 54° latitude and GOES-16 is located at longitude of 75.2° W. GLM level-2 data files are generated every 20 s, containing information about flash, group, and event location, along with flash characteristics like flash area and flash total optical energy.
ENTLN flash data are used to illustrate lightning activity from the perspective of a ground based lightning location system during the 3 June 2018 Fuego eruption. ENTLN operates in the electromagnetic spectrum www.nature.com/scientificreports/ between 5 kHz and 10 MHz and differentiates between intra-cloud (IC) and cloud-to-ground lightning (CG) 39 . When compared with other global lightning location systems, ENTLN has the best-reported detection efficiency in the area of the Fuego volcano at the time of the eruption 46 .  (Tables S1 and S2). ENTLN observed 44 lightning flashes during the first period and 10 lightning flashes during the second period. Interestingly, 76% of these flashes remained in the cloud. Only seven of the 75 total flashes observed by the two systems were observed GLM and ENTLN at the same time. ENTLN had 47 flashes that GLM did not observe, while GLM had 21 flashes that ENTLN did not observe. This means 91% of the lightning during the eruption was only seen by one of the two lightning sensors at any given time. GLM measurements indicated a different time for the onset and cessation of lightning in the volcanic column/plume compared to ENTLN, while ENTLN observed lightning within the plume that GLM did not detect. Therefore, the combination of the ground-based and satellite-based lightning observations provided a more complete picture of the electrical evolution of the Fuego eruption.  (Fig. 3c). This was the largest flash observed by GLM during the entire event, with a flash length of ~ 38 km and the brightest total optical energy of 1,857 fJ. Its position further away from the OT feature as compared to the other smaller flashes in the vicinity of the overshooting top matches theories on the relationship between flash size and kinematic processes 12,35,36 , where the largest flashes occur away from the most turbulent regions of a cloud. The mean GLM flash length was 14 km and the average total optical energy was 20 fJ for this growth stage. www.nature.com/scientificreports/ The ENTLN observed lightning within the plume starting at 18:32 UTC, and detections of IC and CG discharges continued through 19:21 UTC (Fig. 2). ENTLN observed a total of 34 IC discharges and 10 CG discharges during this time. All 10 CG discharges were of negative polarity, with a mean peak current of -13.7 kA and maximum negative peak current magnitude of -26.8 kA. Only one single ENTLN discharge corresponded with a GLM flash at 18:33 UTC. This flash was identified as an IC discharge by ENTLN, and GLM indicates the flash had a length of 16 km, and a total optical energy value of 18 fJ.

GLM
The lack of lightning detection by GLM beyond 18:33 UTC was likely due to the optically thick ash cloud expansion by 18:30 UTC (Fig. 3c,d), which extinguished the optical signal from the lightning before it reached the GLM instrument. Both GLM and ENTLN detections were all to the north and northeast of the vent, and within the ash plume during this period (Fig. 1b).
Lightning during the thunderstorm and PDC transport stage. At 21:37 UTC, a second period of lightning activity was observed near the volcano by both ENTLN and GLM (Fig. 2), with a total of 10 flashes and 22 flashes, respectively. Satellite imagery indicates that a small thunderstorm develops over Fuego around 21:30 UTC to 22:30 UTC (Fig. 5). This period is also near the time of the reported PDC transport at 22:00 UTC.
ENTLN observed 10 flashes between 21:37 UTC and 21:43 UTC, with a median distance of 4.3 km from the vent. Importantly, eight of the ten ENTLN lightning flashes during this period were negative CG flashes, with a mean peak current of -27.1 kA and a maximum peak current magnitude of -42.6 kA. Twenty-two flashes were observed by GLM between 21:38 and 22:03 UTC. GLM flash length was less than 21 km with an average of 13 km and flash energies were less than 120 fJ with an average of 28 fJ. All 22 flashes were within 16 km of the vent, with a median distance of 8.4 km.
Sixteen of the GLM flashes were to the north of the vent, while the remaining six of the flashes were along the southern and eastern flanks in the direction of PDC transport (Fig. 1b). The first two of these six GLM flashes during the period are 13 and 16 km away from the vent at 21:38 and 21:40 UTC. The next four flashes in identified to the east-southeast of the volcano gradually increased in distance. The first flash at 21:42 UTC was 3 km from the vent and the last flash to east-southeast of the volcano was 9 km at 21:50 UTC (Table S2). Unfortunately, it is not possible to discern the exact generation source of these lightning flashes because of the co-location of the thunderstorm, the volcanic plume, and the PDC transport. The temporal resolution of the satellite information is too temporally coarse to resolve cloud feature development, there is limited reporting of the PDC event during this second period of lightning, and height information does not exist for the lightning data 11 .    However, thunderstorm development over the volcano during the secondary maxima in lightning activity (Fig. 5) and PDC transport complicates the interpretation of the lightning data because all three features in previous events have been observed to generate lightning [15][16][17]44,55 . The additional six GLM flashes and one ENTLN flash were located east-southeast of the volcano, along the direction of PDC transport, and directly under the thunderstorm. There is not enough information in the lightning data to partition the lightning occurrences observed by GLM and ENTLN between the thunderstorm, reinvigoration of the ash emission, and PDC transport.
It is also worth noting how the thunderstorm cloud arced with time in Fig. 5 during this period. This arc in the clouds aligns with the direction and time of the reported PDC, where the cloud arc moves further to the southeast with time between 21:45 and 22:30 UTC. Note the dark spot immediately southeast of the volcano indicator in Fig. 5b, which is in agreement with the location of the PDC in Fig. 1B. This dark area expands with time through 2230 UTC. Video evidence from the 3 June 2018 eruption shows the PDC propagating down the side of Fuego and extending into the cloud base (https ://www.youtu be.com/watch ?v=1BVl2 GsrQ0 g). The interaction between the PDC and the thunderstorm is difficult to characterize because of a lack of observations and is worth investigation in future studies of the 3 June 2018 event.  www.nature.com/scientificreports/ lightning events near Fuego as the plume became optically thick. However, ENTLN detections increased in frequency and continued to pinpoint the locations of lightning until 19:21 UTC when lightning activity ceased in the plume. 2. A secondary peak in lightning occurrence was observed between 21:30 UTC and 22:05 UTC. Twenty-two GLM flashes and 10 ENTLN flashes were observed during this period. Sixteen GLM flashes and nine ENTLN flashes were collocated with the ash plume during this secondary maxima in activity. The remaining six GLM flashes and one ENTLN flash were collocated with the direction of PDC transport. However, at the same time a small thunderstorm was identified using satellite data in this same area as PDC transport, thus its difficult to tell if these additional flashes were part of the ash cloud, the thunderstorm, or within the PDC itself. Additionally, hypotheses related to the sources of the secondary maxima in lightning near the time of PDC are discussed. However, without additional data, definitive conclusions could not be determined. Additional work is need to understand the dynamics of the event, potentially from a modeling perspective, to better understand the primary drivers of the secondary increase in lightning during the 3 June 2018 Fuego eruption and the physical interaction between the PDC and the thunderstorm.

Data and methods
Spaceborne lightning data. The GLM monitors total lightning during the day and night in the near infrared (777.4 nm band) part of the spectrum 38,39 . The GLM instrument is a 1372 × 1300 pixel charge coupled device that is in the geostationary orbit GOES-East Position of 75°W longitude. The nadir resolution of the instrument is 8 km by 8 km, with a resolution of ~ 9 km by 14 km at the edges of the field of view. GLM data consist of a three-tier hierarchy of events, groups, and flashes 38,39 . A GLM event is defined as the occurrence of a GLM single pixel exceeding the instrument background threshold during a 2 ms period. A GLM group is defined as the grouping of one or more simultaneous GLM events that occur in the same 2 ms period and are adjacent to each other. A GLM flash is defined as a set of GLM groups that are sequentially separated in time and space by no more than 330 ms and 16.5 km, respectively. The reported position of the GLM flash is the space-averaged location of all the GLM groups that make up the flash. Similarly, the GLM group location is the space-averaged location of all the GLM events that make up the GLM group. Data were obtained through the NOAA GOES-16 Validation Campaign web portal run by the Information Technology and Systems Center at the University of Alabama in Huntsville, and are available via Amazon Web Services: https ://aws.amazo n.com/blogs /publi csect or/acces sing-noaas -goes-r-serie s-satel lite-weath er-image ry-data-on-aws/.
Ground-based lightning data. Lightning data from the ENTLN was utilized to quantify lightning activity as sensed by ground-based networks. The ENTLN operates in a frequency range of 5 kHz to 10 MHz and detects rapid changes in the vertical electric field to pinpoint the location of IC and CG flashes 27,39 . In an inter-comparison between all global ground-based lightning location sensors, ENTLN provided the optimal performance in Guatemala when compared to the Lightning Imaging Sensor aboard the Tropical Rainfall Measurement Mission (TRMM) satellite 46,56,57 . Furthermore, the ENTLN data stream presently includes the WWLLN 45,46,58,59 to incorporate lightning data that is not readily observed in specific regions of the world due to the lack of lightning sensors 46 . WWLLN operates at frequencies between 3 and 30 kHz 45 . ENTLN data can be requested through Earth Networks directly and are not included in the supplemental materials because of their proprietary nature: https ://www.earth netwo rks.com/why-us/netwo rks/light ning/.
Polarity information was also utilized from the ENTLN because it provides insight into the overall charge structure of the cloud. Importantly, + IC and -IC flash designations do not imply that the IC flashes contain a polarity because all IC flashes are net charge neutral 60 . The sign designation implies the direction of the flash propagation, where + IC is indicative of upward propagation, where -IC is indicative of downward flash propagation between regions of positive and negative charge within the electrified cloud 49,61 . Geostationary Operational Environmental Satellite (GOES-16) Advanced Baseline Imager (ABI). ABI has a total of 16 different spectral bands and includes two visible channels and ten infrared channels 39 . The aerosol monitoring visible band, ABI channel 1 at 0.47 µm (Fig. 1a), was chosen to observe the difference in reflectance between the ash plume and 'ash-free' convective clouds located near the eruption, using Python 2.7. GOES-16 provides continuous geostationary monitoring and a time of 19:30 UTC was used in this study because it is near the time of the Moderate Resolution Imaging Spectroradiometer (MODIS)-Aqua pass, which was used to get an approximation of the ash plume extent outlined in Fig. 1b. Scientific Reports | (2020) 10:18015 | https://doi.org/10.1038/s41598-020-74576-x www.nature.com/scientificreports/ Esri ArcGIS: GLM and ENTLN data were loaded with Python into ArcGIS and were then displayed over Esri high-resolution world map imagery (Fig. 1b) sourced from: DigitalGlobe | Sources: Esri, DigitalGlobe, GeoEye, Earthstar Geographics, CNES/Airbus DS, USDA, USGS, AeroGRID, IGN, and the GIS User Community | Esri, HERE, Garmin. A Copernicus Emergency Management Service map and vector data were used to obtain PDC extent, ash fall, and populated areas and uploaded into ArcGIS. The location map (Fig. 1a insert) is an Esri National Geographic basemap. Figure 1a insert and 1b were created using ArcGIS® software by Esri. ArcGIS® and ArcMap™ are the intellectual property of Esri and are used herein under license. Copyright © Esri. All rights reserved. For more information about Esri® software, please visit www.esri.com. Data processing. MATLAB R2018b, in combination with Python (2.7 and 3.7) data processing, was used in order to obtain statistically significant relationships between flash length (the square root of flash area 34 ), flash energy, flash rate, and time. Flash length was used instead of flash area in order to relate the data to previous studies (Fig. 4). MATLAB R2018b was also used to calculate the distance in degrees and the azimuth angles in degrees from Fuego using great circle arcs (Fig. 2 and Figs. S1 and S2, Table S1) and to display the data. Both ENTLN (n = 54) and GLM (n = 27) lightning data are plotted as a function of time and distance in degrees from the volcanic vent.