High Production of Gigantic Jets by a Thunderstorm Over Indian Ocean

Thirteen gigantic jets (GJ) were observed on 12 February 2020, with a sensitive video camera located at Maïdo Observatory, La Réunion Island in Indian Ocean. They were produced within 68 min by nearby cells embedded in a convective system almost 500 km from Maïdo. The video imagery combined with the lightning activity from GLD360, the cloud top temperature (CTT), the ELF radiations, and the reanalysis of several meteorological parameters allow us to analyze their characteristics and their conditions of production. The altitudes for 12 GJ events are estimated between 85 and 89 km. All jets are of negative polarity and most of them preceded by a positive stroke/pulse in the discharge event. They are produced in sequences of a few minutes, during short pulses of convection within cells in phase of development and associated with dominant positive cloud‐to‐ground lightning flashes. The most luminous GJ produced the strongest Current Moment (CM) maximum, close to 280 kA km, and the largest Charge Moment Change (CMC). The CMCs associated with the GJ events range from about 1,000 C km to close to 5,500 C km, especially thanks to the CM during the trailing phase. Several GJ events exhibit a double structure with two jets slightly shifted in space, most of them occurring within the same field of the video imagery. The environment of this exceptional storm in terms of GJ production, exhibits extreme values of various parameters known to be favorable for GJ production, especially related to the warm cloud depth.

1 hr and in a restricted area of about 590 km 2 over Indian Ocean 500 km from La Réunion island • The jets were produced during short pulses of convection in cells with dominant positive cloud-to-ground lightning flashes • The rapid surge in the current moment waveform corresponds with the established fully developed jet (FDJ) and a second peak with a slower charge transfer after the FDJ triangulated and one was a multiple jet) during a storm activity between 24° and 26°N over Sea, east of Taiwan.However, van der Velde et al. ( 2022) recorded 70 cases of GJ occurring in Colombia (tropical region), between 2016 and 2022 for most of them, and found a majority over land, and Singh et al. (2017) reported 4 GJ over India, far from the coast.Other cases, most of the time with one GJ event, were reported outside the tropics, mostly in or around the southern United States, but sometimes related to tropical airmass (Boggs et al., 2022;Cummer et al., 2009;Lazarus et al., 2015;Meyer et al., 2013;van der Velde et al., 2007).Different from tropical GJs, shallow winter thunderstorms in mid-latitude regions are sometimes capable of producing GJ as well (van der Velde et al., 2010).
In terms of characteristics of electrical discharge, they correspond to the negative polarity insofar as they connect the main negative charge of the cloud to the ionosphere (negative charge moving upwards) according to the associated radio signals in the ULF/ELF range recorded simultaneously to optical observations (Cummer et al., 2009;Huang et al., 2012;Krehbiel et al., 2008;Lu et al., 2011;Soula et al., 2011;van der Velde et al., 2019).The discharge starts with intracloud activity which is clearly visible when the GJ events are optically observed at short distance (Soula et al., 2011, <50 km) or when radio signals are recorded and show signatures associated with the initiating lightning (Cummer et al., 2009;Huang et al., 2012;van der Velde et al., 2019).Furthermore, the observations allow to identify the different GJ development stages, that is, the leading jet (LJ), the fully developed jet (FDJ), and the trailing jet (TJ) (Soula et al., 2011;Su et al., 2003;van der Velde et al., 2019).A GJ can develop from an unbalanced thunderstorm charge structure (in this case, a large central negative region and a weaker positive region above) which allows leaders to escape from the cloud (Krehbiel et al., 2008).Thus, the event starts with intracloud lightning processes that neutralize a part of the positive region, which may help to reinforce the imbalance and to allow the jet to emerge from the cloud (Lu et al., 2011).These conditions can be created when the cloud exhibits an overshooting convective top and the charge structure allows the lightning channel to grow more vertically out of the cloud (Boggs et al., 2018;Lu et al., 2011).Furthermore, when the cloud is analyzed in terms of vertical structure, the GJs match with the coldest cloud tops as noted by Lazarus et al. (2015) or with the maximum of flash density (Soula et al., 2011).
According to the low number of GJ cases observed, compared to other types of TLEs observed in the same conditions (Chen et al., 2008), the requirements for their development apparently are uncommon.According to some studies which analyzed the meteorological conditions, for example, Lazarus et al. (2015), GJs could benefit from the weakening of the positive charge region in the upper part of the cloud by mixing and divergence close to the top of the cloud.Indeed, by using dual Doppler polarization radar data for the tropical storm which produced their four GJs, they could characterize a speed shear layer located near the storm equilibrium level.They observed a tilted structure of the storm and the strongest turbulence near the cloud top when the GJs were produced.Boggs et al. (2018) presented meteorological observations to identify a probable thundercloud charge structure favorable to GJ production.The charge structure exhibits a narrow upper charge region which could be the result of an intense convective pulse producing strong storm top divergence and turbulence.These characteristics at the cloud top region can be inferred from large values of storm top radial velocity differentials and spectrum width.van der Velde et al. ( 2022) published a study based on a comparison of meteorological conditions in which storms produced GJs (70 GJ events in 48 nights) and a larger number of other situations without any GJ production (83 cases).They noted a difference in thermodynamic aspects between the two categories of conditions, pointing out colder temperatures at low level and warmer ones at mid-levels for the GJ conditions.Besides that, they noted a wide range of wind shear in the upper levels for both GJ and null storms, which did not support the hypothesis of a weakened upper positive cloud charge due to mixing by strong vertical wind shear at the cloud top.
The present study is devoted to an exceptional storm system that produced 13 GJs optically observed from La Réunion Island in Southwest Indian Ocean during the night of 12 and 13 February 2020.The meteorological conditions, the lightning activity and the storm structure are analyzed.

Optical Observations
The optical observations are provided by a video system installed at the Maïdo observatory (55.38°E; 21.08°S; 2,200 m) on the Reunion Island (Figure 1).This instrumentation comprises the highly sensitive Watec 902H camera previously described in Soula et al. ( 2017) with a 12 mm lens which has a field of view (FOV) of 31° 10.1029/2023JD039486 3 of 16 horizontally.The triggering and recording of the videos were handled by the UFOCapture V2 software, making videos of about 1 s with a rate of 25 frames per second, or 50 interlaced fields per second corresponding to a time resolution of 20 ms.The GPS time is integrated in each field of the video imagery thanks to a Video Time-Inserter TIM-10-Alexander Meier Elektronik.The camera is remotely oriented in elevation and azimuth thanks a pan/tilt platform and a dashboard built in the labview environment.
During the night of 12-13 February 2020, the camera pointed in a constant azimuth of about 295° (west-northwestward) toward active thunderstorms at distances of around 500 km (Figure 1).When a TLE is identified in the video imagery, the azimuth of the line of sight of the event from the camera location is determined by using the software "Cartes du Ciel" (SkyCharts) by matching visible stars in the images with the software star catalog, given the observation time, the camera location and the FOV.Furthermore, this software allows us to obtain the elevation of the top of the luminous event.
By considering the distance of the lightning activity detected at the exact time of the TLE (within ±1 s) and this elevation, we determine the altitude of the TLE top based on the great circle geometry on the spherical geoid of the Earth.With an uncertainty between 5′ and 10′ for the elevation, that about the elevation at 480 km is about 1 km.In terms of luminosity and structure, the description is qualitative.Given that the storm activity associated with the jets observed was at about 500 km distance (Figure 1), the lower part of the GJ was not detected as optical emissions are reduced by the long path through the atmosphere.

Lightning Data
The lightning activity produced by the storm system is provided by the Vaisala Global Lightning Detection Network GLD360 (Said & Murphy, 2016;Said et al., 2010).The data contains time, location, peak current and type (CG or IC) for each detected event which can be CG strokes or IC pulses.We use essentially discharges flagged as CG strokes to characterize the overall lightning activity of the cells in terms of flash rate.CG flashes are made with series of CG strokes, associated together when they follow each other with time intervals of less than 0.5 s and distances of less than 10 km (Soula et al., 2017).The flash associated with a jet event, when it is detected, is characterized with the times of the beginning and the end by considering both IC and CG lightning discharges.

ELF Measurements
The Current Moment (CM) waveform (CMW) and Charge Moment Change (CMC) were obtained from measurements of an ELF receiver system in the Bieszczady mountains in Poland (49.2°N, 22.5°E) 7,980 km from the storm.It measures the magnetic field component with two antennas aligned in the geographic north-south and east-west directions and in the frequency range 0.02 Hz-1.1 kHz.The receiver features a Bessel antialiasing filter with a bandwidth of 900 Hz.The sampling frequency is 3 kHz.The CMW and the CMC were reconstructed using the method of Mlynarczyk et al. (2015) that accounts for the dependence with the frequency of the signal attenuation and the propagation velocity in the ELF range.

Cloud Top Temperature
The cloud structure is estimated from the cloud top temperatures (CTT), obtained from the Spinning Enhanced Visible and InfraRed Imager (SEVIRI) radiometer on board the Meteosat Second Generation (MSG) operated by the European Organization for the Exploitation of Meteorological Satellites.The SEVIRI scans the Earth disk within about 12 min, from east to west according to the satellite rotation and from south to north due to the rotation of a scan mirror (Aminou et al., 1997).It provides images in 12 spectral bands every 15 min and we use the thermal infrared (IR) band at ∼11-13 μm.The spatial resolution for the thermal channel is 0.027°, which corresponds to 3 km at the subsatellite point and about 3.5 and 4.5 km in latitude and longitude, respectively, for the study area.The study area is therefore scanned four times in one hour, around 3, 18, 33, and 48 min of each hour.The accuracy of the temperature values obtained from this radiometer depends on several parameters, such as the geographic location on the Earth, the type of clouds, and the time of observation during the day.In the study by Taylor et al. (2017), that compared SEVIRI CTTs from the new CLAAS-2 (CLoud property dAtAset using SEVIRI, Edition 2) data set against Cloud-Aerosol Lidar with Orthogonal Polarization data, the uncertainty was smaller over ocean than over land.In the region of the study area, this uncertainty can be considered at less than 2 K.The parallax errors are considered in both latitude and longitude on the charts of CTT, where information on lightning activity, jet lines of sight and geographic information are superimposed.It is estimated to be about 0.05° in latitude and 0.23° in longitude for cloud tops at ∼15 km altitude.

Atmospheric Parameters
We use meteorological data from the ERA5 reanalysis (Hersbach et al., 2020)

Optical Observations
On 12 February 2020, between 21:02 and 22:10 UTC, that is, within a time interval of 68 min, 13 GJs were observed with the optical camera installed at Maïdo observatory.The identification of these TLEs as GJs leaves no doubt, according to several criteria among the duration, the shape, the presence of visible TJ, the ELF signature, the absence of a strong positive stroke detected a few milliseconds before and the development phase of the storm which caused them.Table 1 shows the characteristics of the jet events, the lightning flash closest in time (±1 s) within the thunder cell which produced them and the CTT of this cell (Figure 1).The jets are grouped according to the cell at which they are associated, with a different background color in the table.Thus, the three first jets were detected within 5 min above a cell with a minimum CTT of −82°C, the following four within 8 min above a cell with a minimum CTT of −77°C, five others within about 22 min above a cell with a minimum CTT of −82°C and a last one above a cell with a minimum CTT of −82°C.All jets were detected at distances between 460 and 499 km and they reach altitudes between 85 and 89 km.It is remarkable that the altitude calculated for the 12 cases is relatively constant, despite the large observation distance.
Figure 2 shows images for 8 jet events where the time indicated corresponds with that in Table 1 for the FDJ occurrence.Some events are clearly composed of two structures of jet, as #2, #3, #5, and #13, both becoming visible together in the same field of the video imagery, except for #5 where the second jet body (that on the right side) becomes visible one field (20 ms) after the first one.For each of these four cases, one of the two jet bodies is less luminous and it is not possible to see if the two bodies are already separated at the cloud top level.Indeed, due to the large distance, the complete vertical extent and duration of the GJ events may not be detectable.Thus, the time of the jet events indicated in the first column of Table 1 does not correspond to the real beginning of the upward luminous part of the event.The nearest lightning flash detected by GLD started before the GJ in each case, except for the jet #7 where the flash discharge was detected later.For two events (#2 and #13) no lightning flash was detected.For a large majority of the jets with the lightning flash detected (9 out of 11) a positive discharge was detected several hundreds of milliseconds earlier (between 282 and 771 ms).In one case (#7) no discharge was detected closely before the jet (the first positive discharge in the GLD data was detected 389 ms after the visible jet).In another case (#11) a negative discharge was detected 472 ms before.There is therefore a strong tendency to have positive discharges before the jets.We believe the lightning flash activity was not fully reported and discharges could be missed by GLD360.as it is shown from the video imagery when they are observed at a small distance (e.g.Liu et al., 2015;Soula et al., 2011).

Thunderstorm and Lightning Activity
As indicated in Figure 1, several storms produced lightning activity during the time of the jet observation.Thus, the lightning activity has been separately considered within three areas as indicated by Storms 1, 2, and 3 in Figure 1.The names used for this separation correspond more precisely to groups of cells in the same area, according to the short lifetime and the development of several cells in terms of storm structure.Figure 3 shows the time series of the CG lightning activity as indicated by GLD360 (panel a) and the CTT (panel b) for the cells producing the jets and embedded in Storm 2. Indeed, all jets were produced within the area of Storm 2. However, we have an uncertainty about the identification of the type of the discharges in the database for this region of the world, especially when powerful positive IC (+IC) discharges occur within a positive cloud dipole that can be classified as +CG.We will talk about positive discharges (+CG/+IC) in the following, especially when they are associated with a GJ event.In panel a, it is possible to consider the time with a good resolution: 0.1 µs for the lightning strokes (orange dots), 5 min for the CG flash rates (blue and red curves for −CG and 7 of 16 +CG/+IC flashes, respectively) and 1 ms for the jet events (diamonds).The four groups of jets shown in Table 1 are also clearly visible on this graph, and each corresponds to a phase of increase/decrease (sometimes short) of the +CG/+IC flash rate, roughly in the intervals  Finally, the lines of sight of the 13 jets from the camera are reported on CTT maps in Figure 4 at the times of the SEVIRI scans over the area of the storm system.They are reported with white lines, while the +CG/+IC and -CG strokes, detected during a time interval between 5 and 10 min around the jet times, are plotted with red and pink symbols, respectively.The lines of sight of the jets show the thunderstorm cells where the jets are produced, with the same groups identified in Table 1 and Figure 3a: the first 3 jets in Figure 4a in a line of sight of about 308° were associated with the same cell; 1 jet in Figure 4b and 4 in Figure 4c in a line of sight of about 304° with another cell; 1 jet in Figure 4d, 2 jets in Figure 4e  For comparison, the CG lightning activity in the same time interval of time (20:00-24:00 UTC) is considered for Storms 1 and 3 in Figure 5. Storm 1 was the first to produce lightning flashes as it is indicated in Figure 5a since the CG (and some IC) flashes were detected a few minutes after 20:00 UTC.Storm 1 is not in the FOV of the camera as we can see in Figure 4, since it is on the right of the black line which corresponds with the right limit of the FOV (see in Figure 1).Then, Storm 2 started to produce lightning flashes from 20:19 UTC (Figure 3a) and Storm 3 later from 21:11 UTC (Figure 5b).
In the case of Storm 1 the positive flashes dominate during the whole lightning activity with a maximum rate of 10 flashes min −1 between 20:45 and 20:50 UTC, while the negative CG flash rate reaches a maximum of 3.4 flashes min −1 between 20:55 and 21:00 UTC.Storm 2 produces more positive flashes during the first period of lightning activity, especially between 21:00 and 22:15 UTC which corresponds to the production of the jets.Furthermore, during this period of lightning activity for Storm 2, the CG flash rate was very variable, reaching successive maxima of 2.4, 4.6, and 5 flashes min −1 , associated with short periods of cell developments (Figure 4).Then, the negative CG flash rate dominates the positive one.Storm 3 is dominated by negative CG flashes with a maximum flash rate of 3.6 min −1 between 22:00 and 22:05 UTC, while the positive flash rate is maximum with 1.4 flash min −1 between 21:55 and 22:00 UTC.Since the active region of Storm 1 was not in the FOV of the camera, we cannot know if it produced jets or other TLEs during the observation time.On the contrary, Storm 3 was in the FOV and no TLE was detected from this storm.For the CTT minimum values, in Storm 3 it was about −75°C around 22:00 UTC, which was about the tropopause temperature on that day at the storm location.Storm 2 differs from Storm 3 because of the much colder cloud top (CTT was about −82°C at 21:00 and 22:00 UTC and then much lower, up to −88°C at 22:45 UTC).Storm 2 produced also larger CG flash rates and for a longer duration (Figure 3a).Indeed, the maximum rates were 20 and 14.4 flashes min −1 for −CG and +CG/+IC, respectively, between 23:05 and 23:10 UTC and between 23:25 and 23:30 UTC.

ELF Observations
All 13 jets were recorded by the broadband ELF station located about 8,000 km from the storm.Due to a very low attenuation of electromagnetic waves in the ELF range, the recorded signal had a high signal-to-noise ratio and all the waveforms were easily identified.All signatures associated with the 13 jet events correspond with discharges of negative polarity (negative charge moving upwards).Figure 6 shows the CM waveform (CMW) associated with the jets shown in Figure 2. Figure 6 also shows the CMC associated with the jets, obtained by integration of the CMW until the current settles down to zero.The timing in the plots is relative to the jet time shown in Table 1, which corresponds to the beginning of the first video field in which the jet was captured.The 10.1029/2023JD039486 9 of 16 timing in all the plots in Figure 6 was reconstructed at the source, that is at the location of the storm associated with the jets.The method takes the dependency of the propagation velocity of ELF waves on the frequency into account, based on an analytical model developed by Kulak and Mlynarczyk (2013).Since the distance from the ELF station was long, we validated the accuracy of the inferred timing using large discharges from the same storm.The reconstructed CMW allows us to obtain the jet timing with a better accuracy than the video recordings (which is limited to the video field timing resolution).
We start the analysis of the CMW with jet #3, which has the largest CMC and is the brightest.The first video field in which the jet was captured coincides with a large impulse in the CMW at close to 280 kA km (see Figure 6).About 7 ms after the impulse the current starts to rise again and forms a distinct second peak at a little lower value.Afterward, the current is slowly decreasing and settles to zero ∼50 ms after the second peak.From the video recording we know that the FDJ and the TJ stages are both short and very bright.This may be led to the double-peaked structure.The double-peaked structure is similar to that of GJs observed in India (Singh et al., 2017) and Taiwan (Peng et al., 2018).Those jets also had a similar short lifetime.Singh et al. (2017) interpreted the first peak as associated with establishment of the lower portion of the jet and the second peak with a streamer-type flash from the tip of the stem to the ionosphere.Another reason for the second peak could be a very particular structure of jet #3.We can clearly see in Figure 2 that it looks as if it is composed of two distinct jets, one much less bright than the other, similar to the double-peaked case of Peng et al. (2018).Due to the video field rate limitation we can only tell that they occur within the same 20 ms.Looking at the CMW of jet #3 in Figure 6, we can also see that the signal associated with the jet is followed by slow oscillations that began at t = 100 ms and last for about 300 ms.We associate these oscillations with the round-the-world propagation: the wave that reached the receiver by the longer path from the other side of the globe, and the waves that after reaching the receiver propagated around the world and were recorded once again.Due to a strong dispersion and a much higher attenuation at higher frequencies most details in the waveform are lost.The delay between the first impulse and the first minimum in the CMW is ∼105 ms.A very similar delay can be observed in the CMW of other jets analyzed in this study.It matches well the expected propagation delay between the direct path and the longer path, which have the length of 8,000 and 32,000 km, respectively.The inferred average propagation velocity v is equal 0.76 of the speed of light, which is consistent with the model (v/c = 0.76, hence c/v = 1.3, see Figure 6 in Kulak and Mlynarczyk (2013)).A very low attenuation in the ELF range allows us to record high quality waveforms even for events that occur far away from the receiver, but on the other hand, in case of powerful discharges we also record the round-the-world wave which could sometimes be an issue.The larger the distance from the receiver, the smaller the difference in amplitude and propagation delay between the direct propagation path and the longer path from the other side of the globe.In case of the present jets this is not an issue.
The CMW of jet #1 is similar to jet #3 at a lower amplitude (Figure 6).A large impulsive increase of the CMW reaches 70 kA km and is followed by a second smaller peak.In case of this jet the second peak is less sharp, and the current is decreasing more slowly afterward.Near t = 100 ms, the wave that propagated by the longer path reached the receiver.We know from the video recording that the TJ is visible until t = 200 ms.Therefore, the last phase of the TJ overlaps with the round the world wave.The current is already very small, so the total CMC is only a little underestimated.Jet #1 is also less powerful than jet #3.We can see that the amplitude of the CMW is significantly smaller and, as a result, the noise is clearly visible in the waveform.This is also the case of jets #2, #5, and #6, all three featuring a very similar waveform.A weaker signal leads to a smaller CMC and is also associated with a smaller luminosity of these jets.Jet #5 is composed of two bodies, the first alone and less bright in a first field of the video imagery and both in the following field.The CMW exhibits a small increase that peaks during the first field and a larger one during the following field, that is, a peak for each jet body (Figure 6).Jet #7 has a very particular waveform, different from other jets from this storm.A steep rise in current at the beginning does not lead to a distinct maximum but the current keeps increasing slowly, reaching the highest value after about 50 ms, and then slowly decreases.
Jets #10 and #13 have similar waveforms to jet #2, but a much higher amplitude.Their CMW also have an additional interesting feature: a small rise in current before the largest impulse.Its timing coincides with the video field before the beginning of the jet.For jet #10 this rise in current peaks 9 ms before the first field with the jet visible, while in the case of the double jet #13 it peaks only 2 ms before the visible jet.Looking carefully at the CMW of jet #3, we can see a similar rise in current about 6 ms before the main impulse.We hypothesize that it could be associated with in-cloud activity preceding the jet (see increase in cloud brightness recorded by a high-speed camera and by the Geostationary Lightning Mapper shown in Figure 6 of van der Velde et al. ( 2019)).

Case of the 13 GJs, 12 February 2020
The storms that produced the series of GJ were associated with tropical depression Francisco, which was a Moderate Tropical Storm from 5 to 7 February but weakened and drifted westward toward Madagascar.On 12 February the residual depression passed to the northwest of Réunion island, where it started to reactivate and deepen during the night of these events (Figure 7).This figure shows the location of the pressure troughs (low values of geopotential in blue) at 20:00 UTC, approximatively at the beginning of the storm development, especially in the upper right corner of the study area (white frame).The region of low pressure induces lifting of low-level air and therefore storm developments.On the other hand, the horizontal pressure gradient near the surface is not large, causing only a moderate wind.On 14 February, Francisco was declared a Moderate Tropical Storm again just before making landfall on Madagascar a day later.The storms producing the gigantic jets were located just at the southwest side of the center of the circulation, which was situated at −17.7°N and 52.3°E at 21:00 UTC on 12 February and moving westward.The system was surrounded by very dry air in mid-levels directly to the north.
We extracted 3 vertical profiles from ECMWF at locations where CAPE and CIN were optimal for convection.The locations (Table 2) represent the environment to the south and east of the storm at 20:00 UTC, and a bit further to the east at 21:00 UTC, because between these times the CAPE/CIN situation became less favorable, as the reanalysis likely reduced it as result of the presence of convection.The highest CAPE and lowest CIN values were present in a large area to the northeast of the surface low pressure center, with dry air in the mid-levels, whereas the area south and west of the low (first profile in Table 2) shows more humidity in mid-levels but some what lower CAPE.
Table 2 shows various parameters discussed in the statistical study of the meteorological environment of GJ compared against those of null cases by van der Velde et al. (2022).Null cases are nights without GJ observed, while active thunderstorms with cloud-free skies above were present in view of the camera for several hours.Several of these parameters included are not typically used in severe convection studies but have been identified by the authors as those with the largest effect size (i.e. the difference of the means of gigantic jet and null case populations normalized by their pooled standard deviation, known as Cohen's d), even if the understanding of their physical implications for electric charge structures and gigantic jet production is still limited.We use the mean values in that study as a reference.
First, we see that CAPE is moderate.The 1,000-2,000 J/kg range is typical for gigantic jet cases, as is the resulting theoretical convective overshoot (Maximum Parcel Level-Equilibrium Level).The low-level relative humidity is very high (RH at 925 hPa >92%) which is above average for GJ.Mid-level relative humidity in the two profiles east of the storm are drier than usual, with a remarkably low value of just 4%-7% around the 450 hPa  2011), with 5 gigantic jets near Réunion island.The two values of CAPE correspond to the 1,000-950 hPa mixed layer and most unstable parcels, respectively.An explanation of the parameters is provided in Section 3.4 and in van der Velde et al. (2022).The subscript values in the parameter names without indicated units are in hPa.

Table 2
Selection of Parameters Derived From ERA5 Proximity Profiles and Their Values level.The downdraft buoyancy measured in θ e values at 950 hPa using the saturated adiabat from the mid-levels (DD Δθ ES 950 in Table 2) is similar to gigantic jet environments in Colombia (15-25 K).The most significantly different parameters between gigantic jet and null cases in the study of van der Velde et al. ( 2022) are those related with the Warm Cloud Depth (WCD) and adiabatic water vapor content.The altitude of the −10°C isotherm is situated at 7,000 m for the present case, which is 250-300 m higher than the Colombia GJ cases.Similarly, the WCD between Lifting Condensation Level (LCL) and the −10°C level, as well as the associated difference in adiabatic water content between these levels are greater than in the Colombia events.Two ratios found by van der Velde et al. ( 2022) to perform well in the discrimination between GJ and null events, WCDRAT (1.24-1.29)and WCCRAT (3.9-4.2) show very strong values in the present study compared to the average Colombia GJ case (1.15-1.17 and 3.5-3.6,respectively).WCDRAT is the ratio between aforementioned WCD and the depth of the layer between −10 and −50°C isotherms (DZ −10 −50°) .This latter parameter is related to the upper level ambient temperature lapse rates and buoyancy.Lower values (increased mid-upper level instability) were found in gigantic jet environments than in the environment of other storms (van der Velde et al., 2022).The values for our present case are 200-300 m lower than the average for Colombia GJ cases.The WCCRAT*EL (equilibrium level height) was found by van der Velde et al. ( 2022) to perform better still, as storm environments with a high WCCRAT but too low EL were not conducive to GJ.Here we observe values of 55-59, while mean values for GJ in Colombia were ∼50 and null cases ∼45.The parameters CON −10°C and CON −10 −30°C in Table 2 are the components of WCCRAT.CON −10°C is the difference in adiabatic water vapor content in the parcel between the LCL and the −10°C level and is about 1 g kg −1 higher here than the mean of the Colombian GJ cases (13.8 vs. 12.8 g kg −1 ) indicating potential for higher liquid water content across the warm cloud levels.
Finally, 1,000-600 hPa mean wind shear was very weak, like in Colombia GJ cases, while shear between other levels, including the storm top (SH 150-100 hPa level) was weak as well: just 4-8 m s −1 .Not only is this in agreement with the general Colombia GJ population, but especially in multiple-GJ events (2-4 events) the storm top shear was weak in comparison with null cases.

Case of 5 GJ Near Réunion Island, 7 March 2010
In the last column of Table 2 we have added the same parameters for the case of Soula et al. (2011) a storm just east of Reunion island that produced 5 GJs between 17:00 and 19:00 UTC on 7 March 2010.The values for practically all parameters were very similar to the ones for the first profile selected for the present case.The largest difference is that the −10°C isotherm altitude is 160-260 m lower but is offset by a lower LCL so that the WCD results almost the same.We found in ERA5 10.8 m s −1 shear near the cloud top caused by ∼10 m s −1 westerly winds at 125 hPa level and weak winds in variable directions above and below this level (see open research link).A jet stream with maximum speeds of 35 m s −1 was described by Lazarus et al. (2015) for this case but lies hundreds of kilometers away to the south in ERA5, with no more than 20 knots winds at any level in the surroundings of the storm.The most-unstable CAPE value listed in Table 2 of Lazarus et al. (2015) (4,672 J kg −1 ) is not confirmed in ERA5 (1,389 J kg −1 ).

Discussion
The case of storm activity analyzed in this study appears as exceptional in terms of production of GJ compared to other cases in the literature.Indeed, 13 events were observed during a period of a little more than one hour and within an area of about 5° 30′ in angular opening and a distance difference of about 40 km (see the frame in Figure 4a for this area).This area can be estimated as a section of a circular crown of area S formed from two disks with radius of d 2 and d 1 , the maximum and minimum distances of the jets, respectively.With an angular opening of 5° 30′ and the values found for the distances (Table 1), the area is estimated at 587 km 2 which corresponds to a square 24 km of side.This gives an idea of the area occupied by the region of the storm in question.Thus, the rate of GJ events produced in such a reduced area is therefore much larger than in the documented cases in the literature.Indeed, the largest number reported up to now for one storm was 14 events (Huang et al., 2012).In this case, the authors presented the 14 jet events produced in a small area of the Typhoon Lionrock south of Taiwan in parallel of one hundred of TLEs and during a period of about 3 hr.They did not provide the exact time of each, but according to the information given in the paper they were produced between around 16:32 UTC for the first ones and around 19:32 UTC for the last ones, and the area concerned was about 50 × 50 km.The distance of these events was about 200-250 km.Another case of storm producing a few jet events was analyzed in Peng et al. (2018).In this case, a total of 9 jets was reported over about 5 hr and 200 km of distance.Six jets were triangulated thanks to several cameras, one of which was a multiple-jet event.Other cases of observations report much less events for one storm, as for example, Su et al. (2003) with 5 jet events, Soula et al. ( 2011) also with 5 jet events in one hour, Boggs et al. (2019) with 11 jet events within 5 hr, and even less for other studies (Boggs et al., 2022;Liu et al., 2015).All jet events were visually found to reach between 85 and 89 km of altitude which is well in the range found by Peng et al. (2018) who could triangulate several jet events and therefore better determine the distance for an accurate evaluation of the top height.
The CG lightning activity, during the period and within the area of jet production, is clearly dominated by positive CG flashes (including probably some +IC flashes) with short increases of their rate of production.For example, the first three jets were observed within only 5 min (between 21:02 and 21:07 UTC), after a rapid decrease of the CTT (−5.5°C between 20:48 and 21:03 UTC) and during the first period of lightning CG flashes largely dominated by the positive ones.The time resolution of the CTT (15 min) does not allow us to be accurate on the cloud rise rate corresponding with this temperature drop, we can just consider this period as a surge of convection which has been already associated with GJ production in the literature, especially in Meyer et al. (2013) and Lazarus et al. (2015).We must note that the cell area with cold top is small (about 90 km 2 for temperature between −82 and −78°C at 21:03 UTC, according to Figure 3b) which means the minimum temperature can be underestimated at the resolution of the SEVIRI radiometer onboard MSG which is about 3.5 × 4.5 km.This first convective sequence was short since the CG flash rate recovered zero and the CTT increased to −77°C, taking the form of a pulse, at the new value of the CTT 15 min later.The tropopause temperature was −74°C above the area, which indicates an overshoot of about 800 m for this first sequence, since the CTT was −82°C, with a small area at this value of CTT.The four following jets (#4-7) were produced 14 min later during a new convective sequence, with the CG flash rate still dominated by the positive polarity and a larger area with cold top temperature, before a new decrease of the CG flash rate.The other jets were produced more spread out over time, with still dominant positive CG flashes and decreasing CTT.After the last jet, the system was growing larger with increasing CG flash rate and dominant negative polarity.These observations could suggest a rapid spreading of the upper positive charge layer and a negative core at high elevation, as in Boggs et al. (2018).
In this study, we find jets produced during a period during which convective cells develop and the dominant flash polarity detected by GLD360 is positive.In one previous study by van der Velde et al. ( 2007) the dominant polarity was positive, even after removing low peak current flashes which are typically intracloud (Fleenor et al., 2009).It may be that the stronger positive strokes are intracloud flashes as well, as often is the case for compact intracloud flashes known as narrow bipolar events (Nag & Rakov, 2008).In that case, the detections suggest strong +IC flash initiations occurring in a stronger electric field region between the central negative and upper positive charge region.Gigantic jets may develop as a second stage of an intracloud discharge, as shown by Lu et al. (2011) and Boggs et al. (2022) using VHF Lightning Mapping Arrays.The initial upper level negative leader may neutralize the positive charge and pave the way for the GJ development (as a negative leader escaping the upper cloud charge region) by creating an imbalance in the vertical charge distribution (Krehbiel et al., 2008) available to the remainder of the discharge.
The ELF-derived CM waveforms provide interpretation of the charge transfer related with the jets.Previously, Huang et al. (2012) showed that the contact of the GJ with the ionosphere had an associated surge in the CM.We found some jets producing double CM peaks similarly as by Singh et al. (2017) who interpreted that the first peak was due to the establishment of the lower portion of the GJ at ∼50 km and the second peak to be related with the discharge occurring later from this altitude to the ionosphere.But the high-speed video observations of GJs by van de Velde et al. (2019) showed that the CM surge occurs when the GJ reaches the ionosphere and not with the establishment of a stem at ∼50 km before the completion of the FDJ.Our observations show that the rapid CM surge occurs within the video field where the FDJ is established reaching the ionosphere (e.g. Figure 8), which supports the interpretation by Huang et al. (2012) and van der Velde et al. ( 2019).This surge is not always in the form of a prominent peak, this is the case of the jet #7 (Figure 8c) where the surge was step-like rather than an impulse.After the rapid CM surge occurring with the establishing of the FDJ, a second peak can be found (e.g.jets #1,2,3,5,6,7 in Figure 6).The larger broadness of this second peak suggests that it is associated with a slower charge transfer in the form of continuing current (e.g., Huang et al., 2012).The cases of the jets #1 and #7 displayed in Figure 8, show that the broad CM peak occurs during the brightest fields in the FDJ stage.In the case of the jet #1 (Figure 8a), this extends for the duration of two fields (2 and 3) whereas in the jet #7 (Figure 8c), the peak is broader consistently with a larger duration of the FDJ (fields 2 to 5).In addition, in some jets (#1 and #3) the secondary broad peak is superimposed with a surge.In the case of the jets #1 and #3, the FDJ enhances its brightness at the time of the superimposed surge (field 3 in jet #1 and field 2 in jet #3).The hypothesis that the slow charge transfer is associated with the FDJ is clearly supported by the jet #13 (Figure 8d).In this case, the FDJ was first identified in fields 1 and 2 at the time of the first peak in the CM waveform.The FDJ disappeared in the image after the field 3 for ∼140 ms until the time of field 11 when it became visible again.This new appearance was accompanied by a new slow front peak in the CM.Finally, the peak at 280 kA km of the charge moment surge of jet #3 is much higher than the ones reported before by Huang et al. (2012), Singh et al. (2017) and similar to the one reported by van de Velde et al. ( 2010) for a positive GJ.So, negative GJ can produce CM changes as high as positive polarity ones.Considering a discharge channel of 75 km this would correspond to a peak current of 3.7 kA.
The analysis of the meteorological environment of the parent storm using ERA5 data confirms that parameters related to the −10°C level and related warm/cold cloud parameters, as found by van der Velde et al. ( 2022) are enhanced well above the average for GJ, in fact, the parameters WCD10, WCCRAT, and WCDRAT of this case were just above the maximum observed in that study.Low-level wind shear and downdraft buoyancy also were weak.The 1,000-600 hPa mean shear values of 2 m s −1 observed in two ERA5 proximity soundings are below the minimum reported in the data of van der Velde et al. (2022).No evidence for strong vertical wind shear in the upper levels or top of the storm was found, in fact, the values were low, similar to cases of multiple GJ reported by van der Velde et al. (2022).Altogether, it appears that the larger number of GJ produced is commensurate with the warm/cold cloud parameters: the warmer the profile, given a moderate amount of CAPE and high EL, the more jets can be produced.van der Velde et al. ( 2022) detailed a possible mechanism by which the microphysics may lead to temporarily enhanced negative charging, during which the upper positive charge region may also weaken, resulting in the large negative imbalance in charge required for GJ (Krehbiel et al., 2008).Here, we can add that the typical negative charge altitude shifts higher in the cloud as a result of the −10°C isotherm altitude.This may reduce the amount of negative cloud-to-ground discharges under the cloud.At the same time, it may increase the bolt-from-the-blue type from the side of the cloud (described in Krehbiel et al. (2008)).The upper level cloud top expansion could be another factor in diluting the upper positive charge layer density (Soula et al., 2011).

Conclusion
A gigantic jet-prolific storm produced 13 events in about one hour and in a restricted area of about 590 km 2 over Indian Ocean northwest of French La Réunion Island.The altitude of the GJ events ranges from 85 to 89 km and their visible lower part at this distance ranges between roughly 45 and 60 km because of the presence of low clouds between the storm and the camera.
All GJ events produced a signature in a broadband ELF station located about 8,000 km from the storm and were characterized by a negative polarity (negative charge upwards).The CMC values associated with the GJ events range from about 1000 C km to close to 5500 C km, especially thanks to the CM during the TJ.The brightest GJ event produced the strongest CM maximum (close to 280 kA km) and the largest CMC.The rapid surge in the CMW clearly corresponds to the established FDJ and the second peak is associated with a slower charge transfer in the form of continuing current after the FDJ.At least four GJ events exhibit a double structure with two jet bodies slightly shifted in space and for three of them occurring in the same video field.For one, the two jet bodies occur separately within two successive video fields and are associated with distinct peaks of the CMW.
The jets were produced in sequences of a few minutes, during short pulses of convection within cells in phase of development and characterized by dominant positive cloud-to-ground lightning flashes.Most jets were preceded by a positive polarity stroke (or positive intracloud pulse) as detected by GLD360.The other cell in the camera FOV exhibits different lightning activity with dominant CG negative flashes and less cold cloud tops and did not produce visible jets.
The ERA5 reanalysis of meteorological parameters confirms that this exceptionally GJ-productive case occurred in an environment with WCD and related parameters stronger than any of the cases with fewer jets in Colombia.This suggests that gigantic jet production is mainly driven by a microphysical condition allowed by this specific type of tropical environment (van der Velde et al., 2022).

Figure 1 .
Figure 1.Cloud top temperature from Spinning Enhanced Visible and InfraRed Imager/Meteosat Second Generation on 12 February 2020 at 21:03 UTC in the region of the storm producing the jets.The white dashed circle with a 500-km radius is centered on the camera location (white triangle) in La Réunion island.The black lines show the field of view of the camera and the white line corresponds to the line of sight of the first jet detected at 21:02 UTC.The pink circles and the red plus symbols represent the −CG and +CG strokes, respectively detected between 21:00 and 21:05 UTC.Three storms are considered in the study.The white dashed-line frame will be used for the zoom of the study area in Figure 4.

Figure 2 .
Figure 2. Pictures of eight gigantic jets issued from the video imagery at the full development stage.Each image corresponds with the right side (55%) of one field.The time indicated for each case (hour:minute) is in UTC.

Figure 3 .
Figure 3. Description of the activity for the storm producing 13 gigantic jets (Storm 2): (a) Time series of the lightning activity in terms of CG flash rates (blue and red curves for negative and positive, respectively) in (5 min) −1 , peak current for all CG strokes (dots), jets (diamonds).Some +IC strokes (flashes) can be misclassified in +CG strokes (flashes).(b) Time series of the minimum cloud top temperature (blue curve) and the area of the cloud top for different ranges of temperature (histograms).
(21:00-21:15 UTC), (21:20-21:38 UTC), (21:38-21:47 UTC), (21:47-22:02 UTC), while a last jet was produced at 22:10 UTC.The jet production always occurs during a period when positive lightning flash activity dominates over the -CG activity in terms of rate.In panel b, the time resolution is that of the SEVIRI data at 15 min which does not allow to separate the jet events, which are then associated to the closest CTT values in time.The parameters related to the CTT are its minimum value (blue curve) and the area corresponding to different interval values from −70°C to less than −86°C with 2°C of width (histograms).
and 1 jet in Figure 4f in a line of sight of about 308° with a new cell; 1 last jet in Figure 4f in a line of sight of 304° with another cell.These cells have short development and activity durations, low CG flash rates according to Figure 3a: 0.6 and 2.4 flashes min −1 for -CG and +CG/+IC, respectively, for the first cell at 21:03 UTC, 0.6 and 4.6 flashes min −1 for −CG and +CG/+IC, respectively, from the second cell at 21:33 UTC.The CG flash rates increase rapidly after the jet period, that is, after about 22:20 UTC and the −CG flashes dominate after 22:32 UTC to reach about 20 flashes min −1 at 23:07 UTC.

Figure 5 .
Figure 5.Time series of the CG lightning activity in terms of CG flash rates (blue and red curves for negative and positive, respectively) in (5 min) −1 , peak current for all CG strokes (orange dots): (a) for Storm 1 and (b) for Storm 3. Some +IC strokes (flashes) can be misclassified in +CG strokes (flashes).

Figure 6 .
Figure 6.The current moment waveform and charge moment change for the jets shown in Figure 2.

Figure 7 .
Figure 7. Geopotential (color) and horizontal wind (arrows) at the level 500 hPa in a large area in Western Indian Ocean, at 20:00 UTC on 12 February 2020.The white square corresponds with the area shown in Figure 4 and the white arrow indicates Réunion island.
by the European Centre for Medium-Range Weather Forecasts (ECMWF) on pressure levels obtained from the Copernicus Climate Change Service (C3S) climate data store.Vertical profiles are extracted at a location corresponding with the vicinity of the storm cells that produced the GJs, and various parameters are calculated.The statistical study of GJ and non-GJ cases in Colombia by van der Velde et al. (2022), serves as a reference for our present case for the interpretation of the values of parameters of interest.As in that study, the time and location of the profile is chosen where Convective Available Potential Energy (CAPE) is maximized and especially Convective Inhibition (CIN) minimized closest to the actual storm location, to avoid areas and times where the profile was stabilized by deep convection in the model.

Table 1 Jets
The jets are usually preceded by lightning activity 10.1029/2023JD039486 and Lightning Information: Column 1, Number; Column 2, Time When the Jet Is Visible in the Video Imagery; Note.Values printed in bold are those more in agreement with the typical values for gigantic jet cases than for null cases in northern Colombia according to van der Velde et al. (2022).The fourth column (different shade) contains the values of the parameters for the case ofSoula et al. (