Optical emissions associated with narrow bipolar events in radio signals from thunderstorm clouds penetrating into the stratosphere

Narrow bipolar events (NBEs) are signatures in radio signals from thunderstorms observed by ground-based receivers. They are electromagnetic waves radiated by impulsive currents of electrical discharges. They come with two polarities, the positive that brings negative charge aloft, and the negative, that brings negative charge towards the earth. The sources of negative NBEs are at the very top of thunderclouds, and positive NBEs are at the upper levels, but inside the clouds. NBEs may occur at the onset of lightning, but the discharge process is not well understood. Here, we present spectral measurements by the Atmosphere-Space Interactions Monitor (ASIM) on the International Space Station that are associated with nine negative and three positive NBEs observed by a ground ‐ based array of receivers at the closest distance of about 100 km. We found that both negative and positive NBEs are associated with emissions at 337 nm with weak or no detectable emissions at 777.4 nm, suggesting that NBEs are associated with fast streamer breakdown. The rise times of the emissions for negative NBEs are about 10 µs, consistent with source locations at cloud tops where photons undergo little scattering by cloud particles, and for positive NBEs are ~ 1 ms, consistent with locations deeper in the clouds. For negative NBEs, the amplitude of the emissions is almost linearly correlated with the peak current of the associated NBEs. Our ndings suggest that ground-based observations of radio signals provide a new means to measure the occurrences and strength of cloud-top discharges with implications for studies of perturbations of greenhouse gas concentrations at the tropopause. with an unprecedented time resolution of 10 µs. Our analyses, for the rst time, show the distinct optical signals of negative NBEs and the relationship between the blue emissions and E-eld signal of negative NBEs. This indicates the direct link of blue emissions to NBEs and conrms the suspicion of Neubert et al 21 that the microsecond blue emissions could be the optical equivalent of negative NBEs.


Introduction
Narrow bipolar events (NBEs) are one special class of intra-cloud (IC) fast discharges that have been considered as one of the most intriguing phenomena receiving enormous interest from many lightning researchers 1,2,3,4 . They are characterized with short-duration (typically 10-30 μs) electromagnetic waveforms and strong radiation in the high-frequency (HF) and very high-frequency (VHF) bands 5,6 .
NBEs often appear isolated from other IC discharges 1,7 , but in many cases they also occur as the initial event of ordinary bi-level IC lightning ashes 4,8,9 . The experimental and modeling work suggests that they are formed by streamers inside thunderclouds 10,11 .
In the normally electri ed thunderstorms, NBEs of positive and negative polarities are distinctly segregated into thundercloud regions centered at different altitudes 2,3,5,12 . The positive NBEs usually occur between the main negative and the upper positive charge layer of thunderclouds 2,6,13 . The negative NBEs are far from being as frequent as their positive counterpart, whereas they are usually produced at much higher altitude between the upper positive and the negative screening layer with thundercloud tops typically reaching over 16 km 14,15 , and thus serve as a proxy to monitor the global deep convection. Additionally, they have been thought to trigger the blue emissions 15,16 , one kind of electrical discharges shooting upward from the thundercloud top into the stratosphere 17,18,19,20,21 , which can perturb the concentrations of greenhouse gases near the tropopause 22,23,24 .
For decades the luminous features of NBEs were a mystery and somewhat under debate. NBEs have been thought to be relatively "dark" (non-luminous) compared to other fast lightning discharges 25,26,27 . Jacobson et al 28 compared 193 positive and 24 negative NBEs from the Los Alamos Sferic Array (LASA) with the optical observations from FORTE. It was found that NBEs of both polarities had dim optical emissions in the range of 400-1100 nm (roughly from violet to far infrared). Recently, the Modular Multi-spectral Imaging Array (MMIA) of the Atmosphere-Space Interaction Monitor (ASIM) aboard the International Space Station (ISS) provided the multi-spectral observations of thunderstorm discharges with the highest spatial (~400 m/pixel) and temporal (10 µs) resolution so far 29 . Positive NBEs inside thunderclouds were associated with the 337 nm emissions rather than the 777.4 nm emissions from ASIM. It is suggested that positive NBEs are corona-like discharges formed by many cold streamers 30 . However, this leads to a question as how is the optical signature of negative NBEs occurring near the thundercloud top 21 . The optical features of NBEs still merit verification.
Here we present the rst observations of NBEs with both polarities by ground-based lightning detection array at the closest distance of around 100 km and the associated optical blue emissions by ASIM with an unprecedented time resolution of 10 µs. Our analyses, for the rst time, show the distinct optical signals of negative NBEs and the relationship between the blue emissions and Eeld signal of negative NBEs. This indicates the direct link of blue emissions to NBEs and con rms the suspicion of Neubert et al 21 that the microsecond blue emissions could be the optical equivalent of negative NBEs.

Results
Meteorological condition of the parent thunderstorm. On the evening of 7 August 2019, during a short time period between 1305:56 and 1306:32 UTC, MMIA recorded an outbreak of 12 blue emissions over a compact thunderstorm near the coastline of South China. We compared the trigger time of blue emissions with the data recorded by a ground-based very-low-frequency/low-frequency (VLF/LF) network of multiple stations (detailed information of time match is provided in Methods and Table S1). It is found that nine of these 12 blue emissions are associated with negative NBEs, and the other three are with positive NBEs. To the best of our knowledge, it is the rst ever report of NBEs with both polarities produced in one thunderstorm observed from space and ground-based Eeld sensors at a closest distance of about 100 km. Among these events, the height of six negative NBEs and one positive NBE can be determined through the ionospheric re ection pair 6,31 . The source altitude of negative NBEs was located at about 18 km (above mean sea level, MSL) near the cloud top, while the positive NBE occurred at an altitude of about 14 km inside the thundercloud, which is consistent with previous studies on the height distribution of NBEs 8,12,15 , indicating that the higher negative NBEs are inferred to initiate between upper positive charge region and screening negative charge layer 12,14 . Additionally, this positive NBE also gave off weak 777.4 nm emission. The context of these events plotted in Fig. 2c shows that the main pulse of NBE signals is followed by some extra sferic-radiating activity, indicating that the NBE was not the only discharging activity. Thus, it is likely that the 777.4 nm emission is equally related to the additional discharging activity but not related to the NBE. Jacobson et al 28 examined the optical emission of 193 positive and 24 negative NBEs based on the measurements of FORTE and LASA. It was found that no obvious optical emissions were associated with NBEs except for two positive NBEs. They examined the context of these two NBE events and found that their main sferic pulses were both preceded by some earlier sferic activity. Therefore, the 777.4 nm optical emission is likely related to additional sferic activity accompanying the NBE itself. The remaining two positive NBEs shown in Fig. S2 generally exhibit similar optical features that include a distinct 337 nm emission and no 777.4 nm signal above the noise level. The distant stations failed to record the signal of two positive NBEs due to the relatively small amplitude, and their height is also not calculatable from the Eeld signal recorded at the GZ station due to the weak ionospheric re ection.
Optical and electrical observations of negative NBEs. Fig. 3 compares the blue emissions recorded by MMIA at 1306:20.968870 UTC and the associated negative NBE waveform measured at ve VLF/LF stations located at range of 105 km to 1300 km. The waveforms recorded at four relatively distant stations (>800 km) exhibit the obvious ionospoheric re ection that can be used to determine a source height of about 18 km (MSL). It can be seen that this negative NBE was also associated with the 337 nm emission, but there was no signi cative 777.4 nm signal. The emissions at 777.4 nm are from atomic oxygen and are one of the major lines of the lightning leader spectrum, suggesting that negative NBEs are associated with fast streamer breakdown, similar to the positive NBE. An evaluation, based on optical radiation transfer, of the streamers of these NBEs is reported in a complementary publication, which presents the streamer-like structure of negative NBEs that typically involves around 10 9 streamer initiation events 33 . Here it is emphasized that the negative NBE produced the distinct 337 nm optical signature with rise time of <0.05 ms, showing the much sharper optical emission of this negative NBE than positive NBEs. In addition, we can see that the VLF/LF signal of negative NBEs corresponds to the rise stage of 337 nm emissions. For the remaining eight negative NBEs shown in Fig. S2, our analyses generally obtain the similar optical feature, namely the rise time of the 337 nm emissions for the negative NBEs is much shorter than that of positive NBEs.
In order to characterize the waveform parameters of the optical signals, we de ne the rise time, duration, and signal-to-noise ratio (SNR) (de nition of rise time, duration, and SNR is given in Fig. S1 of the Supplementary information). Table 1 summarized the detailed parameters of optical signature and VLF/LF sferics for these NBEs. The observed negative NBEs have a wide range of strength. Previous observations show that the fast breakdown of NBEs occurs with an extremely wide range of strength, both in VHF and VLF/LF bands, while still initiating ordinary IC discharges 4,10 . We can see that the optical rise time of negative NBEs is in the range of 0.03 ms-0.08 ms, which is much shorter than the positive counterparts (>0.2 ms in all three cases). For the whole optical duration, there is also a considerable difference that in all cases positive NBEs endure longer than the negative ones. The reason why negative NBEs appear much sharper than positive NBEs in the optical emission is discussed below.
Comparison between blue emission and E-signal of NBEs. As the negative NBEs in our obervations occur close to the cloud top, its optical radiation is less affected by the thundercloud scattering 34, 35 . This allows us to compare by means of radio observations. As shown in Fig.  3, we see that the VLF/LF signal of negative NBEs corresponds to the onset of 337 nm blue emission, but the Eeld signal of NBEs in the VLF/LF band is much shorter than their optical blue emission. The Eeld signal of NBEs is usually shorter than 0.03 ms, while the optical duration of the associated blue emissions is usually longer than 3 ms. This suggests that the lifetime of current source is much shorter than the optical pulses.
To characterize the source current further from optical measurements and compare the VLF/LF signal of negative NBEs with the 337 nm emission in association, we implement the rst-and second-order differential on the original 337 nm optical signal shown in Fig. 4a. Since the sampling rate of the optical signal is 100 kHz, its maximum bandwidth is 50 kHz. To eliminate the in uence of bandwidth, we apply a 50-kHz lowpass lter to the VLF/LF signal. It is seen that the second-order differential of 337 nm emission is very similar to the 50-kHz lowpass ltered waveform of negative NBEs. The second-order differential of 337 nm emission can be divided into two parts. The rst part is the main narrow bipolar waveform, which resembles the sferic waveform of negative NBEs in the VLF/LF band; the second part is the one ensuing small oscillation pulses, and these small pulses are also discernible on the VLF/LF signal owing to the close distance (105 km) of the GZ station (the VLF/LF waveform for other cases is given in Fig. S2-S10). These small pulses after the main bipolar pulse of the VLF/LF signal suggest that there still exists some weak current pulses after the fast breakdown 4 , which corresponds to the slow descent stage of 337 nm blue emissions. The E-field changes observation by Karunarathne et al 36 suggest that NBE has static offset in its bipolar pulse and was also followed by a slow electrostatic change lasting about several ms, which is similar to our optical observation. Therefore, the duration of NBEs could be as long as several ms as observed by the optical detector, but on the VLF/LF waveform, the subsequent small oscillation pulses may be attenuated due to distance.
The question is why the second-order differential of the optical signal exhibits a similarity with the lowpass-ltered VLF/LF waveform. The luminosity is roughly proportional to the electrical current intensity (I) 37 . The VLF/LF signal at the GZ station is the time derivative of the vertical E-field (d /dt). The measurement distance of about 105 km for the VLF/LF signal might be very critical for nding this waveform similarity. As the Eeld is also proportional to the temporal derivative of I at such a distance 38,39 , it is reasonable that the second-order differential of the optical signal is similar to the lowpass-ltered VLF/LF signal of negative NBEs.

Discussions
Soler et al 30 reported seven positive NBEs located at heights of 8 to 15 km (MSL) in the thundercloud that were all associated with the 337 nm blue emissions, and there was no 777.4 nm emission based on the multi-spectral data of MMIA. They suggested that blue flashes are from cold ionization waves (associated with streamers) and the positive NBEs are corona discharges formed by many streamers. For our observations, both positive and negative NBEs produced by the same thunderstorm were observed by ASIM and the ground-based sferic array. It is found that negative NBEs are also associated with 337 nm, without 777.4 nm, either, suggesting that both polarities of NBEs are streamer discharges, which con rms previous results that fast breakdown is formed by streamers, as inferred by Liu et al 11 .
The optical rise time of positive and negative NBEs exhibits a signi cant difference; that is, the rise time of optical emissions from negative NBEs is much shorter than that of positive NBEs. It is speculated that there might be two reasons causing this distinct difference: First, the source position in the thundercloud and the cloud morphology could cause the light scattering, which would lead to the optical attenuation 34, 35 . Light et al 34 simulated the optical waveform of lightning for the satellite observations and discussed the effects caused by the optical depth, the position of light emission, and the size and sharpe of thundercloud. They set the source position at varying heights below the thundercloud, and found that the rise time at the higher altitude is generally sharper than that of the event inside the thundercloud. Light et al 40 also showed that the peak amplitude of lightning event could be attenuated by up to 2-3 orders of magnitude. The height of negative NBEs is usually close to the cloud top (~18 km), so the optical emissions generated by the discharge is not seriously blocked by the thundercloud; the positive NBEs typically occur in the middle of hosting thunderclouds, and their optical emission is anticipated to be delayed, attenuated and prolonged before being observed by the space-born detector.
Second, in addition to the scattering by thunderclouds, the distinct difference in the rise time may also be caused by the difference in the discharging channel and the speed of current propagation involved in positive and negative NBEs. Rison et al 4 found that the breakdown speed does not vary with altitude. The speed is essentially the same for IC discharges (probably positive NBEs) initiating at altitudes of 8 to 10 km, and also for the screening discharges (probably negative NBEs) at altitudes of 14 to 15 km. This invariance may be due to that the discharge process obeys the similarity law related to air density 41,42 . Thus, this optical difference is likely due to the shorter channel length and higher altitude of negative NBEs.
Space-borne optical observations provide an effective tool for characterizing the global lightning activity. However, due to the lack of comparison with the ground-based sferic data, it is di cult to determine the speci c lightning type from the optical signal. Our ndings that the distinct optical signals of NBEs with different polarities would provide a new tool to identify and distinguish NBEs of both polarities based on a space-borne optical observation platform and further provide the possibility to evaluate the global NBE occurrence.
Furthermore, we compare the peak optical brightness of 337 nm and peak current estimated from the Eeld for both positive and negative NBEs. In order to eliminate the in uence of distance, each peak of 337 nm emission was normalized to 450 km. Fig. 4b shows the relationship between the peak brightness and peak current of NBEs. Although the number of samples is relatively limited, there exists a clear linear correlation between the peak 337.4 nm emission and peak current of the associated NBEs, except for one positive NBE with a high peak current (about +80 kA). The deviation is likely caused by the attenuation of thunderclouds. This correlation is also observed in "normal" lightning 43,44 . Kikuchi et al 44 examined the relationship between the photometer (PH4) at wavelengths of 599-900 nm and electromagnetic waveforms in the LF band for 11 negative cloud-to-ground (CG) strokes. They also found a linear correlation between the absolute optical intensity and the peak current.
Negative NBEs have been thought to be the initial event of blue emissions due to the relatively low time resolution of previous optical observations 15,16 . The optical waveform obtained with an unprecedented 10-µs resolution and the peak correlation analyzed above present direct evidence that the blue emissions are the optical signature produced by negative NBEs. Blue emissions usually occur in the strong thunderstorms with overshooting tops penetrating into the lower stratosphere 20,21,45,46 . Such discharges (that can be precursors of blue jets) can affect the exchange of greenhouse gases between the troposphere and stratosphere through the production of nitrous oxide, and the depletion of stratospheric ozone 24,44,48,49 . However, this region of atmosphere is difficult to access experimentally. Our results present the direct link between the blue emissions and Eeld signal of negative NBEs. On that perspective, our ndings suggest that ground-based observations of radio signals provide a new solution to measure the occurrences and strength of cloud-top discharges with implications for studying the perturbations of greenhouse gas concentrations near the tropopause.

Methods
Optical measurements. The optical emission data analyzed in this study are obtained from the MMIA instrument of ASIM aboard the International Space Station (ISS) at an altitude of about 400 km since April 2, 2018. So far it is the space-borne platform at the lowest altitude to study lightning activities in tropospheric thunderstorms and their effects in the near-Earth space 29 . MMIA consists of two parts, including a pair of ltered cameras (operating at 12 frames per second) and three photometers 50  Radar and Cloud top temperature measurements. The IR brightness temperature at 10.4 μm of the parent thunderstorm is derived from the latest-generation geostationary meteorological satellite Himawari-8 32 . The target observation provides an image of the thunderstorm every 10 minutes. The spatial resolution of IR data is 2 km×2 km. Base re ectivity is derived from a 2.88-GHz S-band Doppler radar with a detection range of about 340 km. The radar was operated with a 6-min cycle producing polar volumes, and the distance between the radar and the thunderstorm is typically about 100 km. The height of tropopause is estimated from the atmospheric sounding data (http://weather.uwyo.edu/upperair/sounding.html).
Time match. The time accuracy of ASIM is within ~ 20 ms due to the time shift. The time provided by ground-based lightning array is assumed to be the correct time of lightning events. In order to match the event between the two systems, rst, we identify all the blue emissions from ASIM during the over ight of the thunderstorm, and the occurrence time of blue emissions can be obtained by subtracting the propagation delay. Then, we search the lightning event from the ground-based lightning array during the time window of about 30 ms. It is noted that the NBEs especially those of negative polarity are isolated with other discharges as seen from the context of sferic waveform, hence, it can exclude optical emissions from other discharges.