TGFs, Gamma‐Ray Glows, and Direct Lightning Strike Radiation Observed During a Single Flight of a Balloon‐Borne Gamma‐Ray Spectrometer

Terrestrial gamma‐ray flashes (TGFs) and other high‐energy radiation phenomena related to thunderstorms remain incompletely understood since their discovery nearly 30 years ago. Space and ground‐based platforms have provided insights, but limitations exist in temporal resolution, signal‐to‐noise ratios, and proximity to events. This study presents findings from a balloon‐borne gamma‐ray spectrometer flown into a thunderstorm over northern Mississippi on 19 June 2023. The instrument detected three distinct millisecond‐duration TGF events correlated with lightning strikes. Sustained gamma‐ray glows lasting up to 2 minutes were measured and tied to thunderstorm electrostatic fields. Notably, radiation was observed during a direct lightning strike to the payload. This complex 800‐millisecond event reveals insights into radiation from leader propagation and return strokes. These findings from a single balloon flight demonstrate the prevalence of energetic processes within thunderstorms. Finally, the balloon platform offered an exceptional combination of temporal resolution and validation capabilities to advance the understanding of thunderstorm radiation motivating future follow‐up studies.


Introduction
Following their detection in 1994 by the Burst and Transient Source Experiment (Fishman et al., 1994), Terrestrial gamma-ray flashes (TGFs) and high energy radiation in thunderstorms have become areas of considerable interest within the scientific community due to their intriguing yet not fully understood mechanisms.TGFs are associated with thunderstorms and lightning discharges in Earth's atmosphere and are believed to be related to phenomena such as the ejection of electrons with significant energy, a process first theorized by Wilson (1925), which can lead to the production of gamma-rays.The study of TGFs is of great importance for aviation safety (Dwyer et al., 2010), the prediction of severe weather (Chronis et al., 2016), and advances in physics research (Chilingarian et al., 2022).
Almost 30 years since their initial discovery, TGFs and high-energy radiation events continue to be predominantly detected through space-based observation systems.The National Aeronautics and Space Administration's Reuven Ramaty High Energy Solar Spectroscopic Imager (RHESSI), operational from 2002 until its decommissioning in 2018, contributed valuable insights into radiation in thunderstorms.Utilizing RHESSI data, Cummer et al. (2005) proposed a robust correlation between TGFs and lightning events, where there was an association of TGFs with very low frequency radiation.In addition, D. M. Smith et al. (2005) revealed that both the frequency of occurrence and maximum photon energy of TGFs exceed what was previously known for such lightning events.This study also demonstrated that on average, RHESSI detected 10 to 20 TGFs per month, equivalent to approximately 50 per day globally.Similarly, the Astro-Rivelatore Gamma a Immagini Leggero (AGILE) satellite in low Earth orbit above the equator found comparable results, detecting an average of 10 TGFs per month since its launch in 2007 (Fuschino et al., 2011).Another space-based observation system, the Fermi Gamma-Ray Burst Monitor (GBM) (Meegan et al., 2009) detected 12 intense TGFs during its first year in orbit.Briggs et al. (2010) showed that the majority of TGFs detected by GBM had maximum energies around 30-40 MeV.Additionally, Briggs et al. (2010) presented compelling evidence of partially overlapping TGF pulses, showing that TGFs could occur before, during, and after lightning discharge.More recently, since 2018, the Atmosphere-Space Interactions Monitor, installed outside the European space laboratory Columbus on the International Space Station, has provided insights into TGFs.Heumesser et al. (2021) demonstrated that TGF sources at 180-230 nm, 337 nm, and 777 nm bands are estimated to occur at altitudes 1-5 km below the cloud tops.
Aircraft and ground-based platforms have also been used to detect TGFs.In 2009, the Airborne Detector for Energetic Lightning Emissions (Airbone Detector for Energetic Lightning Emissions (ADELE)), equipped with a collection of six gamma-ray detectors, recorded a brief burst of gamma-rays (D.Smith et al., 2011).This detection took place while ADELE was aboard a Gulfstream V jet during a flight near two active thunderstorm cells.Notably, this TGF was linked to a complex intra-cloud flash.More recently, the Airborne Lightning Observatory for FEGS and TGFs campaign, a joint initiative between NASA, the University of Bergen, and other academic institutions, deployed an ER-2 aircraft over tropical thunderstorms in the Gulf of Mexico, Central America, and the Caribbean (Ostgaard, Marisaldi, & Lang, 2023;Ostgaard, Marisaldi, Lang, & Quick, 2023).Results from this campaign may offer valuable insights into TGFs and high-energy radiation phenomena in tropical storms.Ground-based assets, such as lightning arrays (Berge & Celestin, 2019;Hare et al., 2016;Lu et al., 2010;Lyu et al., 2021;Shao et al., 2010), have also contributed information into storm charge structure, various radio frequency emissions, and the temporal nature of radiation events; these surface-based platforms often validate and enhance data collected by space-based observation systems (Bankert et al., 2011;Erdmann et al., 2020;Peterson & Rudlosky, 2018).
Despite the understanding gained from previous studies on TGFs and high energy radiation in thunderstorms, limitations persist with space-based, aircraft, and ground platforms, including limited temporal and spatial resolution and lower signal-to-noise (SNR) ratios due to a larger distance from a radiation event.Atmospheric and signal research on balloon-based platforms has proven to be an effective method for understanding variables close to the source.Previous studies have utilized radiosonde balloons to gain an understanding of thermodynamic properties influencing thunderstorms (Bartos et al., 2022;Coniglio & Parker, 2020;Stolzenburg & Marshall, 2009).Additionally, weather balloon flights have been instrumental in determining the radio refractive index of the atmosphere (Clinger & Straiton, 1960;Hao et al., 2022;Pham & Nguyen, 2022), while smaller pico balloons have proven useful in investigating ionospheric signal propagation over remote regions (McKinney, 2023a(McKinney, , 2023b(McKinney, , 2023c)).Previous work has been done using balloon-based platforms to understand radiation and TGF detection in thunderstorms.During the Strateole-2 intertropical balloon campaign, a gammaray spectrometer capable of measuring energies exceeding 20 MeV was developed to detect gamma-ray glows and TGFs.The payload was flown on a superpressure balloon at altitudes above 18-20 km (Pallu et al., 2021).These flights provided details into lightning flash counts in various via various observations over South America.Using weather balloons, an early paper by Eack et al. (2000) reported a significant, threefold increase in the gamma-ray flux observed as a balloon descended through a thunderstorm anvil, where a strong electric field was suspected to be present.This finding suggested that gamma-ray production in thunderstorms might be more common than previously believed.Subsequently, in a series of balloon flights, Eack and Beasley (2015) incorporated an X-ray spectrometer and an electric field meter to investigate the possibility of strong electric fields accelerating electrons, leading to X-ray emissions.These flights revealed X-ray intensities ranging from 10 to 1,000 times higher than the typical background levels, coinciding with the presence of robust electric fields.Moreover, Eack and Beasley (2015) found evidence of electron production by electric fields in thunderstorms.
Since it began development in 2017, the High Energy Lightning Emission Network (HELEN) has conducted multiple flights of balloon-borne radiation detection payloads into thunderstorms during 2021, 2022, and 2023 (Helmerich, 2022).While the focus of this paper is on data collected during a specific HELEN balloon flight, details about the development, calibration, and validation of the HELEN system can be found in Helmerich (2020).The utilization of balloon-based platforms presents certain challenges, including unstable balloon payload rotation (Stark et al., 2023), unpredictable balloon trajectories due to convection events (Lee & Yee, 2017), and effects from flying in high-noise electric fields (Zeng et al., 2016).However, this study has focused on minimizing these impacts.HELEN's exceptional capability lies in its ability to record high-resolution energy measurements with sub-microsecond UTC precision, while also capturing complete traces of radiation pulses.This feature, unprecedented on light balloon-based platforms, enables particle discrimination and allows for the removal of Electromagnetic Interference (EMI) from triggered events.In addition, the omnidirectionally of the radiation detector reduces the need for a stable balloon platform.The HELEN payloads are also engineered to be compatible with standard-sized latex weather balloons, guaranteeing quick deployment, prompt recycling times, and flexibility in launch locations, which allows for choice of landing sites and dynamic positioning within the storms.As a result, HELEN balloon flights have yielded valuable data sets that offer insights into the formation of lightning and TGFs.The flight on 19 June 2023 marked the seventh mission in the flight campaign of HELEN and was of particular interest due to the payload being directly struck by lightning.Specific analysis of other flights and general statistics gathered from all HELEN flights is left to a future study.
We present the findings from a single HELEN flight conducted on 19 June 2023, within a thunderstorm, launched from Baldwyn, Mississippi, USA.In the course of this flight, the balloon remained inside the storm for approximately 30 min until it was struck by lightning at 12 km above ground level (AGL).This lightning strike caused the balloon to burst, initiating its descent back through the storm.Apart from the lightning strike, this single flight is believed to have detected three TGF events, with one occurrence during the ascent and two during the descent.Additionally, we observed a gamma-ray glow phenomenon during the ascent that persisted for approximately 2 min, and, in the minutes before and after the lighting strike, we observed multiple increases in radiation of various intensity and duration.In the following sections, we will first discuss the defining features of the HELEN payload and the data analysis process.We subsequently introduce the meteorological context of the thunderstorm, detailing both its thermodynamic and dynamic structures.This sets the stage for ensuing discussions and analyses concerning radiation emissions, including those from lightning strikes, TGFs, and observed gamma-ray glows.Additionally, we perform a comparative analysis between HELEN and the data sets from the Lightning Mapper Array (LMA) (Zhu et al., 2020), the Geostationary Lightning Mapper (GLM) (Peterson & Rudlosky, 2018), and NEXRAD WSR-88D radars (Heiss et al., 1990).This comparison aims to validate the measurements obtained during the balloon flight.Lastly, we provide the readers with key takeaways and reflections from this research, highlighting its implications for future lightning and TGF studies.

The HELEN Instrument and Its Components
HELEN payloads are typically launched on a 1,200 g latex weather balloon filled with helium.Figure 1 shows a HELEN payload shortly after launch with parts of the flight train labeled.The total flight train weighs 2,500 g which is 226 g under the limit for payload mass set by the Federal Aviation Administration (FAA) (FAA, 2023).Balloons were filled with helium to provide an ascent rate of 5-6 m/s at launch, but this ascent did not remain constant throughout the flight.Due to convection, the balloons were frequently caught in updrafts, often reaching vertical speeds above 18 m/s.This is especially apparent on the 19 June 2023 flight.The balloon was carried by a vigorous updraft into the center of the storm; we will further show this in the flight data section.
The HELEN payload is made up of six main components.Figure 2 shows the inside of a HELEN payload, while Figure 3 shows a diagram of the radiation detector.The radiation detector features a 12 × 12 × 20 mm LYSO:Ce scintillator crystal, which was selected for its high efficiency within the TGF energy spectrum and short decay time to mitigate pulse pile-up.This scintillator crystal is coupled with a Hamamatsu R6095 photomultiplier tube (PMT) via a coupling compound.The PMT is further connected to a Hamamatsu C14019-02, serving dual functions by supplying high voltage to the PMT and converting the output current from the PMT into a voltage through a built-in trans-impedance amplifier.The pulse processing operation involves a DE-10 Standard Cyclone V FPGA System on a Chip Development Board, equipped with an AD/DA Daughter card that digitizes incoming radiation pulses.The Development Board hosts a Cyclone V FPGA chip coupled with an ARM-based Hard Processor System (HPS).The FPGA enables the programming of logic circuits for real-time data sampling, storage, and transfer, while the HPS, running a Linux operating system, interfaces with the FPGA and a microSD card for data storage.The detector output is split into two channels.Channel A captures the unmodified output, while Channel B captures the output after a reduction in gain by a factor of 12.This configuration allows the full range of the detector to be measured by the Channel B ADC.The FPGA, triggered by a threshold voltage reading, transfers a series of 32 samples along the voltage pulse to the HPS, with a 10 ns interval between samples.This process is executed simultaneously for both Channel A and Channel B, yielding a total of 64 14-bit samples within a 320 ns interval.
The onboard GPS, a CAM-M8Q-0-10 module, provides critical timing data, as well as information and uncertainties for latitude, longitude, altitude, and speed.The GPS sends a pulse-per-second (PPS) signal to the FPGA for accurate timing of radiation events.All GPS NMEA data is logged on the microSD card.
The onboard optical camera of the HELEN payload captures visual data of lightning events during the flight.The camera, a Raspberry Pi-based model, was equipped with a wide-angle lens providing a 160-degree field of view.The camera is positioned on the side of the payload facing directly downward, with a clear unobstructed view below the payload.The camera operated at a frame rate of 30 frames per second, with a resolution of 1080p.Its focal length was 3.15 mm and the lens aperture was f/2.35.
For data processing, the optical brightness time series was calculated by analyzing the video frames captured by the camera.Each frame was converted to a grayscale image, and the average pixel intensity was computed to represent the overall brightness of that frame.This method provided a quantitative measure of the brightness variation over time, correlating with the intensity of lightning events captured in the video.These brightness measurements were then synchronized with the timestamped data from the radiation detector, allowing for a precise correlation between optical and radiation data.The environmental board supports a range of 1 Hz sensors, including an Inertial Measurement Unit, a digital temperature sensor, and four strategically placed thermistors.These components provide measurements of pitch, roll, yaw, acceleration, gyroscopic motion, magnetic field, and temperature at various locations within the payload.The entire payload operates on twelve 18650 batteries, enabling up to 8 hr of uninterrupted data recording.Future upgrades to the HELEN system are expected to lighten the load, potentially by using fewer batteries, and make room for more equipment such as atmospheric and magnetic field sensors.

Data Analysis
The raw radiation and environmental data retrieved from the payload were parsed into a structured format using MATLAB.This transformed the radiation data from its binary form into a comprehensive table of pulses, each including a pulse number, pulse traces for each ADC channel, timing information, and diagnostic data.The environmental data was similarly parsed from its raw text file into a structured table.
To remove noise and pulse pile-up from the radiation pulses, all pulses were grouped via k-means clustering as shown in Figure 4.This process was applied to the entire data set of the flight.The clusters in green were chosen because they were consistent with in-lab observations of pulse shape.Such tests included calibration with gammaray, neutron, and electron sources.The red clusters, likely EMI, were only present during the period of enhanced activity and comprised approximately 25% of pulses during that period.They were excluded from the analysis to better depict the actual high-energy radiation count rates during that time.
For the light curve plots in this paper, pulses were binned into 1-ms bins for a millisecond-based light curve and 1s bins for a second-based light curve.Occasionally, when the count rate exceeds the full pulse saving data rate of the FPGA, only count rate information is captured and the full pulse shape is dropped.In this occurrence, the ratio of valid full pulses to invalid full pulses is applied to the total number of detected pulses when calculating the count rate for that time window.
where the ratio of valid pulses (P v ) to all pulses (P all ) times the count rate of all pulses (R all ) for each bin during the flight is represented by R v .Earth and Space Science

10.1029/2023EA003317
To perform spectral analysis of events (including probable TGFs, gamma-glow, and lightning strikes), valid pulses with full traces were used to determine the spectrum of the event.The peak ADC channel of each of the pulses was calculated and used to create a spectrum in ADC-channel space.The resulting spectrum was then scaled using calibration coefficients, which were determined in flight using the LYSO self-count spectrum in order to correspond to keV units as shown in Figure 5.This in-flight calibration was only possible due to the intrinsic radioactivity of the LYSO scintillator crystal.These linear calibration coefficients were also crosschecked with on-the-ground measures of radiation sources including Cs-137 (Helmerich, 2020).This method of using LYSO's existing background source for calibration of a detector is commonly used in nuclear medicine as LYSO is a crystal commonly used in Positron Emission Tomography scanners (Conti et al., 2010).The calibration coefficients are a simple linear scaling factor and a zero-offset and do not take into account any non-linear response of the detector.Therefore, the spectra in this paper should be taken as a preliminary analysis of the energy deposited in the detector, not necessarily the spectrum of the radiation source itself.The spectrum was then further converted into counts/keV/second.For various periods of interest, the background spectrum of LYSO, as  measured during quiet portions of the flight, was subtracted from the total measured spectrum during that period.This results in the measured spectrum of external sources.
To correlate the radiation data collected by the HELEN payload with external sources, lightning event data from the North Alabama Lightning Mapping Array (LMA) and the GLM were used.These data sets were filtered spatially and temporally to coincide with the start and end of the flight and the 1°latitude by 1°longitude region in which the flight took place.Additionally, radar data from the KGWX radar was used to show the payload's position relative to the storm cell.

Meteorological Features of the 19 June 2023 Thunderstorm
For the launch site, north of Baldwin, Mississippi, USA, at coordinates 34.565 N, 88.630 W, thunderstorm development was forecasted to be favorable.The Storm Prediction Center of the National Weather Service issued  (Brown & Nowotarski, 2019;Davies, 2006), particularly in the southeast region (Lyza et al., 2022).
In relation to the launch area in this study, convection commenced in the late evening around 2 UTC.The significant convective event occurred when a smaller storm, accompanied by leading precipitation, migrated from the Tennessee border into more favorable and undisturbed moisture conditions over northern Mississippi.This mesoscale setup provided an advantageous opportunity for launching a balloon into the developing storm; the leading precipitation from the preceding storm did not heavily burden the updraft in relation to the launch location.Figure 6 illustrates the reflectivity and radial velocity maps obtained from the NEXRAD KGWX radar, along with the launch location denoted by a star.This radar scan was conducted 3 min before the HELEN balloon was launched.The KGWX radar is positioned approximately 80 km to the SSE (south-southeast) from the launch location.As a result, the 0.5 beam height is approximately 1 km above the launch site.The KGWX radar detected high-reflectivity structures with minimal leading precipitation.This resulted in an inflow-dominated regime; radial velocity scans reveal positive velocities in front of the storm, indicating air being brought inward into the main storm and away from the KGWX radar.This increase in inflow resulted in lifting, leading to the formation of a smaller storm ahead of the main line.This can be seen in Figure 6, where elevated reflectivity is evident just northwest of the launch site.The HELEN payload was launched at the leading edge of this small storm cell.The balloon was launched at 04:36:56 UTC time.At launch, there was slight precipitation due to the emerging cell.Figure 8 shows the meteorological data from the flight up to 11.3 km.The emerging cell in Figure 6 caused cold pool propagation, first pushing the balloon southward after launch.After rising 1 km, the balloon was pulled in by the inflow of the main storm, causing the balloon to shift northward.After this northward shift, the balloon's ascent rate began to increase.At 4.1 km, the balloon reached a max ascent rate of 16.7 m/s, about triple the ascent rate of a normal balloon flight.This observation suggests that the HELEN payload effectively utilized the storm's updraft and was positioned optimally to capture radiation events.Previous research has established a connection between a strong updraft and the emergence of well-defined charge structures within storms (F.Wang et al., 2015;Xu et al., 2016).Other meteorological features of the thunderstorm include a freezing height of 6.1 km.Additionally, we observed an inversion created by condensation, which led to the release of heat and subsequently raised the temperature by 1.2°C across a vertical distance of 0.2 km.

Overview of Radiation Events
An overview of the June 19th flight is given in Figure 9.After the launch at 04:36:56 UTC, the balloon began to ascend into the storm.Just 3 min and 17 s after launch at an altitude of 830 m AGL, the payload detected a 2 mslong spike in radiation coupled with an intense optical flash.This strong flash was also observed by the ground crew and prompted the safety decision to only launch a single balloon out of the four available.
Over the course of the following 20 min, the balloon was then drawn into the storm cell.As recorded by the onboard camera, the flash rate in the proximity to the payload increases in both frequency and intensity.The payload quickly ascended through 5 km as it was caught in a >10 m/s updraft.Between 04:52:30 and 04:54:30 the payload detected a gamma-ray glow as it passed through the freezing layer of the storm at an altitude of 6-7 km.
After the glow, the optical flash rate continued to increase as the payload ascended directly through the cell, once again carried by updrafts of >5 m/s.Starting at 05:05:00, a 5-min long period of increased radiation activity was observed with count rates of nearly 10 3 counts/millisecond measured.During this period, the payload was directly struck by lightning at 05:07:08.This led to the complete loss of all external sensors including the optical camera, external temperature thermistor, and radio-based tracking using the Stratotrack.Luckily, most internal sensors and electronics remained functional including all radiation detection hardware, the power control board, the GPS, and the satellite-based SPOT tracker.Some environmental sensors were temporarily disabled (accelerometer, gyroscope, magnetometer, digital internal temperature) and came back online within a few seconds after the strike.All internal thermistor temperature sensors were permanently disabled.
After the strike, the payload descended under the remains of the shredded parachute back into the heart of the storm.During the first 2 minutes of descent, the payload measured numerous spikes (<1 s) and prolonged periods (>1 s) of radiation that were both coincident with and terminated by lightning strikes.The measured radiation returned to normal background levels around 05:09:00 as the payload passed below 10 km in altitude.At 05:15:57 and 05:22:44, the payload detected two more short, millisecond-long bursts of radiation correlated with lightning strikes before landing at 05:26:35 UTC.

Millisecond-Long TGFs
The first of the three stand-alone, short-duration TGFs was detected at 04:40:13.565UTC.As is shown in Figure 10, the TGF is coincident with a sharp increase in camera brightness as measured by HELEN's onboard camera, as well as lightning strike events as detected by GLM and LMA.Of note, the detected radiation increase occurs during the later stages of the lightning strike as indicated by the LMA radio data.The GLM data indicates that it is coincident with the brightest GLM events and hence, is likely coincident with the main return stroke of the lightning strike.While the GLM brightness measurements of events build over the course of 20 ms starting at 04:40:13.550, the radiation increase is concentrated within a 2 ms window between 04:40:13.565and 04:40:13.567,just 1 ms before the GLM brightness data peaks at 04:40:13.568as shown in Figure 11.The burst of radiation is not directly linked with a particular LMA event.The energies detected were in the range of 400-500 keV.
In addition to the strong, multi-messenger temporal correlation, there also appears to be a distinct spatial correlation as compared to other lightning strikes during a similar time window.As shown in Figure 12, the events temporally correlated with the lightning strike occur directly over the location of the payload at the time of the strike.The LMA data suggests that the strike occurred at altitudes above 7 km although visual observations of the strike made by the launch team indicated lower altitude activity was present for the strike.Whether the strike was cloud-to-cloud or cloud-to-ground was unclear.
Two other, weaker TGFs were detected during the second half of the flight at 05:15:57.533UTC and 05:22:44.066UTC.These detections are shown in Figure 13.These occurred in the same timeframe as lightning activity from GLM and LMA, although they did not show the same millisecond-timescale alignment with specific events that the first TGF showed.

Standalone Gamma-Ray Glow at Freezing Level
After the first millisecond TGF, there was a quiet period in the radiation detector lasting 12 min with no notable increases in radiation.During this quiet period, dozens of lightning strikes were captured by HELEN's onboard camera which were coincident with detections by both LMA and GLM.Then at 04:52:37, as shown in Figure 14, a lightning strike occurred that increased the average count rate by 50 counts/second.This 50 counts/second increase lasted for 17 s until a radio-only event without any optical component seems to have increased the count rate by another 50-100 counts/second.After 20 s of fluctuating count rates between 50 and 175 counts/second over the background, a lightning strike occurred at 04:53:24 that spiked the count rate up to 300 counts per second more than the background for approximately 2 s.During this 2 slong spike, no abnormal millisecond-long bursts of radiation were detected.The increased count rate then descended back down to 50-150 counts/second over baseline until a lightning strike at 04:54:30 marked an end to the period of the observed increase in radiation.The spectrum, as shown in Figure 15, shows an increase in mostly low energy radiation of <400 keV.Interestingly, this gamma-ray glow occurs just as the payload crosses the freezing altitude as shown in Figure 16.This is the altitude at which strong electric fields commonly occur within thunderstorms (Gunn, 2004).

Period of Increased Activity
The payload then experienced another quiet period of 10 min from 04:55:00 to 05:05:00, despite being in the heart of the storm.Over the course of the next 5 min, starting at 05:05:00, the payload experienced multiple radiation events of varying duration, magnitude, and correlation with lightning as shown in Figure 17.
With the payload at an altitude of 11.5 km and at 05:05:20 UTC, a lightning strike occurred that increased count rates by approximately 180 counts/second.This count rate decreased over the course of the next 25 s to only 80 counts/second above background.Then, at 05:05:47, a lightning strike brought the count rates back down to baseline.There was a quiet period lasting 30 s until count rates began to rise at 05:06:20 increasing to almost 800 counts/second over the course of the next 12 s before finally being reset to baseline by a lightning strike at 05:06:33.The lightning strike at 05:06:44 marked the last lightning strike detected by the onboard camera before its demise at 05:07:08.The count rate begins rising at 05:06:55 and increases to 3,000 counts/second above the background at 05:07:02.The rate drops back down once again for a few seconds before the payload is directly struck by lightning at 05:07:08 where count rates shoot past 10 5 counts in a single second.More details of the strike are given in Section 3.6.
During the lightning strike, the balloon carrying the payload bursts, which causes the payload to descend back through the storm.Over the following 2 min from 05:07:10 to 05:09:10, the payload experienced upwards of 20 radiation events of increased count rates of more than 10,000 counts/second.Some were second-long sustained increases in count rate while others were millisecond-long bursts of radiation.Most of the events began abruptly and decreased gradually in intensity, while some others gradually rose in intensity and abruptly dropped.Some events had relatively high levels of noise (>50%) as measured by pulse clustering while others had virtually none (<1%).After 05:09:10, the payload does not experience any further increases in radiation except for 2 millisecond-long spikes at 05:15:57 and 05:22:44.

Lightning Strike Details and Discussion
The payload was directly struck by lightning at 05:07:08 UTC.This led to an 800 millisecond-long increase in measured radiation lasting from 05:07:07.7 to 05:07:08.5 shown in Figure 18.Throughout this period there were approximately 30 unique bursts of radiation lasting from 2 to 30 ms.The lightning strike was detected by both the NALMA and by GLM with the events of the two detection networks occurring simultaneously with the increases in measured radiation.Notably, the radiation levels greatly increase during the early portions of the lightning strike when the lightning leaders propagate outwards.This detection occurs simultaneously with the very first LMA events and before any noticeable optical detections by GLM.These sequential bursts of radiation occur with varying separations of 1-20 ms.A large optical signal occurs at 500 ms into the strike,    Furthermore, the spacing of spikes later in the strike from 05:07:08.1 to 05:07:08.5 is consistent with multiple return strokes (X.Wang et al., 2017).
As shown in Figure 20, the location of the strike as determined by LMA readings begins at precisely the location of the payload and propagates upward and eastward as the strike progresses.Further confirmation of the payload being struck by lightning is given in Figure 21.When the payload was recovered later the same day, the first evidence of the payload being struck was noted.Interesting features on the recovered balloon train found by the recovery team included charring of the balloon neck, singeing of the styrofoam in the radiosonde, shredding of the parachute, and melting of the Stratotrack antenna and aluminum housing.
The measured spectrum of the lightning strike is given in Figure 22.While mostly lower energy (<400 keV) radiation was detected, the spectrum does extend into the multiple-MeV range.In addition, a sharp spike of the spectrum occurs from 510 to 518 keV.While this is possibly due to the 511 keV photon emission typical of positron annihilation, the energy resolution of the detector is limited to 8% (Helmerich, 2020).Therefore, the measurement of such a sharp peak cannot be taken as conclusive evidence of a 511 keV line.The full energy versus time scatter plots of events is given in Figure 23.
During three of the spikes, count rates reached above 1,000 counts per millisecond, equating to a 10 6 counts per second fluence during those milliseconds.Approximately 1.2 × 10 5 pulses were detected throughout the entire second-long time period of the strike.This number excludes the pulses that did not meet the clustering criteria.While this tentatively rules out most of the sporadic EMI, it does not rule out the possibility of local discharges creating short bursts of x-rays inside the instrumentation.In addition, despite the EMI shielding of the detector, Gaussian-like changes in electric fields over the course of tens of nanoseconds could liberate electrons from the PMT dynodes causing pulses that may look very similar to radiation pulses.Although this has not been observed in laboratory testing of the instruments, the possibility cannot be ruled out.This number does not include pulse However, several limitations of this preliminary study do exist.First, despite EMI shielding and the use of kmeans clustering to distinguish radiation pulses from electrical interference, not all electrical interference may be accounted for, especially during nearby, high-current events such as a direct lightning strike.During these events, it is probable that no amount of shielding completely reduces the effects of EMI to negligible values.Furthermore, a simple linear map from ADC channel to energy was used instead of a Detector Response Matrix which has been left to a future study.Finally, this single flight can only make limited claims of the general prevalence of high-energy radiation.

Conclusion
This study has presented novel findings from a balloon-borne gamma-ray spectrometer flown into a thunderstorm over northern Mississippi, USA on 19 June 2023.The flight resulted in the detection of multiple high-energy radiation events including terrestrial gamma-ray flashes (TGFs), gamma-ray glows, and direct radiation from lightning strikes.
The balloon-borne gamma-ray spectrometer provided exceptional temporal resolution on radiation events, while also yielding in-situ observations and localization of radiation.Key results include the detection of 3 distinct TGF events on the millisecond scale that were temporally and spatially correlated with lightning strikes recorded by ground-based lightning networks.Additionally, gamma-ray glows were observed, marked by sustained increases in radiation count rates over baseline in the range of 50-300 counts/sec and lasting up to 2 min.The initiation and cessation of these glows were correlated with lightning strikes and were likely tied to the presence of strong electric fields.Furthermore, direct measurement of a lightning strike was achieved while the payload was aloft at 11.9 km altitude.The strike resulted in a complex 800 ms long radiation profile with multiple spikes offering insights into the leader propagation and return strokes phases.The combination of instrumentation on the payload enabled effective validation using lightning networks and radar data.The flight highlighted multiple periods of enhanced radiation activity with varying spikes and glows demonstrating the prevalence of high-energy radiation events in thunderstorms.
This research illustrates the value of balloon-based observatories in furthering our understanding of TGFs, gamma-ray glows, and other high-energy phenomena in thunderstorms.This study has shown that balloon-borne gamma-ray spectrometers can provide unique insights into thunderstorm radiation through high temporal resolution measurements and correlation with ground and space-based systems.Key areas for future work include comparing data across multiple flights.These efforts will provide greater insights into the prevalence, mechanisms, and properties of TGFs, glows, and other energetic radiation produced by thunderstorms.

Figure 1 .
Figure 1.High energy lightning emission network (HELEN) balloon flight a few moments after launch into a thunderstorm over Huntsville, AL, USA, on 25 June 2023.Parts of the HELEN balloon line are labeled.

Figure 2 .
Figure 2. Inside of a high energy lightning emission network payload.(A) LYSO:Ce scintillator crystal coupled with a Hamamatsu R6095 photomultiplier tube for radiation detection, (B) DE-10 Standard Development Board with ADC Card for processing radiation pulse data, (C) Raspberry Pi-based optical camera capturing 1080p video for lighting flashes, (D) 18650 battery power supply, (E) Power distribution board, (F) Environmental data acquisition board(s) recording temperature, GPS timing/location, and payload orientation.

Figure 3 .
Figure 3. Diagram of the radiation detector.(A) Incoming high-energy particle, (B) Scintillator emits light when struck by high-energy particle, (C), Photo-Multiplier Tube (PMT) converts light from scintillator into electrons, producing an output current, (D) Socket converts current from PMT to voltage pulse, (E) ADC converts the analog voltage signal into digital voltage time-series, (F) FPGA reads ADC, calculates radiation timing, and saves timing, full digitized time-series, and debug information to a micro-sd card.

Figure 4 .
Figure 4. K-means clusters of all radiation pulses observed during the flight.Green clusters matched expected pulse profiles and were used in computing light curves and spectrum data; Red clusters were not used.

Figure 5 .
Figure 5. Calibrated spectrum of LYSO measured during the first 1,500 s of the high energy lightning emission network flight.The figure shows the trigger threshold and callouts for LYSO's known energy peaks (Alva-Sánchez et al., 2018).

Figure 6 .
Figure 6.Reflectivity (a) and radial velocity (b) at the 0.5-degree tilt from the NEXRAD KGWX radar.The star represents the high energy lightning emission network (HELEN) launch location in relation to the storm.The KGWX radar is positioned approximately 80 km to the SSE (south-southeast) from the launch location.The scan was completed approximately 5 min before the HELEN balloon launch.

Figure 7 .
Figure 7. Event rate for geostationary lightning mapper and NALMA throughout the flight.Events inside 1°× 1°from 34.6816°N to 88.3512°W (the payload's location at the time of the lightning strike) were included.

Figure 8 .
Figure 8. Meteorological parameters from the balloon flight.Features of note are labeled.

Figure 10 .
Figure 10.The strike at 04:40:13.565produced a sharp, 2 ms-long spike in radiation as measured by the high energy lightning emission network (HELEN) radiation detector.The lightning strike was jointly detected by the HELEN payload camera, geostationary lightning mapper data, and LMA data.

Figure 7
Figure 7 shows the count rate for the storm cell during the duration of the flight.The storm strength remains fairly consistent across the hour of the flight window.The steady increase in the NALMA events can be attributed to the storm moving closer to northern Alabama where NALMA is based.

Figure 11 .
Figure 11.Scatter plot time-series of individual radiation pulse energies for the first TGF event along with geostationary lightning mapper event energies.

Figure 12 .
Figure 12.Depictions of location and altitude for the initial TGF event.(a) displays the nearest KGWX radar returns to the event, incorporating LMA signals within ±60 s and ±1 s of the TGF event, and geostationary lightning mapper (GLM) locations ±60 s of the TGF event using contour kernel density estimate.The number of GLM flashes using a contour estimate are labeled.(b) the KGWX reconstructed RHI to the high energy lightning emission network payload during the radiation occurrence, highlighting the altitudes of LMA signals and the payload itself.The bottom plot is adjusted to the distance and elevation relative to the KGWX radar, where LMA data within a 1-degree FOV of the KGWX beam are plotted.Azimuth angles used in the reconstruction are 0.5°, 0.9°, 1.3°, 1.8°, 2.4°, 4.0°, 6.4°, and 10.0°.The location and altitude of the payload are marked by a white star in both plots.

Figure 13 .
Figure13.At 05:15:57.533,high energy lightning emission network (HELEN) measured a much weaker, 1 ms-long spike in radiation.This TGF was generally correlated with a lightning strike but no geostationary lightning mapper (GLM) or LMA data was simultaneous within the millisecond of emission as in the case of the first, much stronger, TGF.At 05:22:44.066,HELEN measured a moderate strength, 8 ms-long spike in radiation.Similarly, this TGF is correlated with a lightning strike but occurs after the majority of the GLM and LMA events.

Figure 14 .
Figure14.From approximately 04:52:30-04:54:03 UTC an increased count rate of +100 counts/second was detected, consistent with a gamma-ray glow.The glow appears to be initiated by the lightning strike at 04:52:37 and terminated by the lightning strike at 04:54:30.In addition, the lightning strike at 04:53:24 results in a notable second-long increase of +300 counts/second of detected radiation.

Figure 15 .
Figure 15.Resultant energy spectrum of glow calculated by subtracting the background spectrum of LYSO from the spectrum of all radiation measured from 04:52:30 to 04:54:30.

Figure 16 .
Figure 16.Depictions of location and altitude for the Gamma-Ray Glow event.(a) displays the nearest KGWX radar returns to the event, incorporating LMA signals within ±60 s of the Gamma-Ray Glow event, and geostationary lightning mapper (GLM) locations ±60 s of the Gamma-Ray Glow event using contour kernel density estimate.The number of GLM flashes using a contour estimate are labeled.(b) illustrates the KGWX reconstructed RHI to the high energy lightning emission network payload during the radiation occurrence, highlighting the altitudes of LMA signals and the payload itself.The bottom plot is adjusted to the distance and elevation relative to the KGWX radar, where LMA data within a 1-degree FOV of the KGWX beam are plotted.Azimuth angles used in the reconstruction are 0.5°, 0.9°, 1.3°, 1.8°, 2.4°, 4.0°, 6.4°, and 10.0°.The location and altitude of the payload are marked by a white star in both plots.

Figure 17 .
Figure 17.Time series of high energy lightning emission network (HELEN) radiation, HELEN camera brightness, and geostationary lightning mapper data around the 4 min of enhanced radiation activity.During the first 2 min of enhanced activity (05:05:00-05:07:00) there appear to be at least 2 measured gamma-ray glows.The first is bookmarked by two lightning strikes and the second, which builds in intensity, is terminated by a lightning strike.The radiation then builds until the payload is struck directly by lightning at 05:07:08.Numerous spikes in radiation both coincident with and terminated by lightning strikes follow until 05:09:00, at which point the payload descends below 10 km.

Figure 18 .
Figure 18.Two seconds of high energy lightning emission network radiation count rate, geostationary lightning mapper data, and LMA data surrounding the direct strike by lightning.

Figure 19 .
Figure19.Plot of the durations of the spikes of radiation and the gaps between the spikes throughout the lightning strike.

Figure 20 .
Figure 20.Depictions of location and altitude for the lightning strike event.(a) upper plot displays the nearest KGWX radar returns to the event, incorporating LMA signals within ±60 s and ±1 s of the lightning strike, and geostationary lightning mapper (GLM) locations ±60 s of the lightning strike using contour kernel density estimate.The number of GLM flashes using a contour estimate are labeled.(b) illustrates the KGWX reconstructed RHI to the high energy lightning emission network payload during the radiation occurrence, highlighting the altitudes of LMA signals and the payload itself.The bottom plot is adjusted to the distance and elevation relative to the KGWX radar, where LMA data within a 1-degree FOV of the KGWX beam are plotted.Azimuth angles used in the reconstruction are 0.5°, 0.9°, 1.3°, 1.8°, 2.4°, 4.0°, 6.4°, and 10.0°.The location and altitude of the payload are marked by a white star in both plots.

Figure 21 .
Figure 21.Post-flight photos of the lightning-struck payload line showing the condition of the balloon neck, radiosonde, parachute, and stratotrack.

Figure 23 .
Figure 23.Energy versus time scatter plot of radiation events measured by high energy lightning emission network during the direct lightning strike of the payload.