Observations of transient luminous events (TLEs) associated with negative cloud to ground (−CG) lightning strokes



[1] A balloon campaign was conducted in summer, 1999, to measure the stratospheric electromagnetic fields associated with sprites. Ground observations for detection of sprites included low light level TV (LLTV) observations from three sites. Flight 1 flew from Palestine, Texas at 01:14:31 UTC to 09:45:00 UTC on 07/06/1999. Flight 3 of the campaign flew from Ottumwa, Iowa at 00:39:32 UTC to 11:12:00 UTC on 08/21/99. During flight 3, 26 sprite halos associated with positive cloud-to-ground (+CG) strokes and 17 −CG sprite halos were observed. Of these, 22 +CG and 12 −CG sprite halos were observed by the ground observatories. Seven of the +CG and all 17 −CG halos were not followed by sprites. Next the balloon data were examined during and after the times of the recorded NLDN strokes during 4.1 hours of data. An additional 88 −CG TLEs were found in the flight 3 data and 56 TLEs (7 +CG, 49 −CG) were found in the flight 1 data. It appears that −CG TLEs, mostly spriteless halos, occurred 5–7 times more often than the +CG TLEs. The halo appears to be a fundamental mesospheric response to lightning.

1. Introduction

[2] Interest in sprites investigation was stimulated by an initial report of video observations of flashes in the ionosphere above active thunderstorms [Franz et al., 1990]. Video cameras and photometers have been used to study these phenomena from the ground and airplanes. These events are now thought to be commonplace. At least five distinct transient luminous event (TLE) phenomena have been observed for which names have been coined: red sprites, blue jets, elves, halos and trolls. These seemingly whimsical names attempt to avoid identifying causative physical mechanisms. The term sprite refers to spatially extended visible light emissions occurring in the mesosphere and ionosphere above active thunderstorms [Sentman and Wescott, 1993; Lyons, 1994; Sentman et al., 1995]. The sprite is a mesospheric and ionospheric phenomenon that typically occurs in the 40–95 km altitude range, with average maximum altitude of 88 ± 5 km and maximum brightness at 66 km [Sentman et al., 1995]. The upper portion is red; with wispy blue tendrils sometimes extending down to below 40 km. Sprites have a horizontal extent of 10's of km. Sprite appearances range from simple columnar forms to complex dendritic structures. The brightest components of sprites exist for several ms, with afterglow persisting for 100 ms in some cases. Sprites are most often produced in association with positive cloud-to-ground (CG) lightning strokes with unusually large moment changes of at least 600 C-km [Hu et al., 2002; Lyons et al., 2003] typically found within the trailing stratiform region of large mesoscale convection systems (MCS) [Boccippio et al., 1995]. The term elve refers to a fast (∼100 μs) brightening of the ionosphere [Fukunishi et al., 1996]. Halos are amorphous glows that occur at ∼77 km altitude, are delayed after the initiating CG stroke by 2–6 ms and last for a few ms [Barrington-Leigh et al., 2001; Wescott et al., 2001].

[3] The identification of sprite halos poses several questions [Wescott et al., 2001]: Why are some halos not followed by a sprite? Do any sprites occur without a preceding halo? Why do some large sprites have a nearly simultaneous halo? The triangulation of halos showed they seem to be centered over the causative flash whereas discrete sprites are scattered up to 50 km from the flash. Wescott et al. [2001] speculated about answers to these questions, which involve the triggering process of sprites. In the quasi-electrostatic (QE) model explanation for sprites, the magnitude of the charge transfer from cloud to ground and the time the electric field is present in the sprite region are important, whether the field is upward or downward. It must take a minimum electric field to excite the N2 molecules near 77 km to produce the uniform glow centered over the lightning that is the sprite halo [Barrington-Leigh et al., 2001; Wescott et al., 2001]. What happens next can potentially be divided into several cases:

[4] 1. If the electric field is large enough to excite N2 via impact excitation by electrons accelerated in the electric field, but not so large as to produce streamers, which appear as c-sprites, and there is no random ionizing perturbation, nothing more happens, and sensitive instruments detect only a halo.

[5] 2. If the electric field is not large enough to produce streamers by itself, but a random or storm produced ionizing trigger occurs in the region near 70 to 80 km while the electric field is large enough, streamers appear and develop into sprites, which are located where the ionizing event occurred, not centered over the lightning flash.

[6] 3. If the electric field is much larger, the process develops into a large sprite without a trigger, guided by preexisting irregularities in the region. Sprite activity may form such irregularities [Stenbaek-Nielsen et al., 2000].

[7] Wescott et al. [2001] suggest that the random trigger could be a micrometeor of sufficient mass to cause an ion trail in the atmosphere. This picture leads to two predictions. Spriteless halos are presumed to be the initial response of the mesosphere. This concept suggests that spriteless halos should be the most common type of TLE and that spriteless halos should occur frequently in association with negative CG strokes. We will investigate these predictions in this paper. The properties of halo associated strokes will be reported later.

2. Instrumentation

[8] This project involved making simultaneous observations of the sprites with from a balloon launched at Ottumwa, Iowa and three ground stations: Yucca Ridge Field Station (YRFS), Ft. Collins, Colorado; Wyoming Infrared Observatory (WIRO) on Jelm Mtn., Wyoming; and Bear Mtn. Fire Lookout in South Dakota.

2.1. Balloon Payload

[9] The balloon payload contained instruments that measured the vector electric and magnetic fields, X-ray counting rate, light emissions from the events, vertical current density, conductivity, temperature and balloon location. Descriptions of the instruments have been published [Bering et al., 2002a, 2002b]; more detail can be found at http://www.uh.edu/research/spg/Sprites99.html. The light emissions from the events were detected by a broadband, rapid response all sky photometer that had a measured bandwidth of 25 kHz. The phototube was 28 mm in diameter with a 28 mm diameter, hemispherical lens glued directly to the front surface and no filter. The tube was recessed in a well looking up at a ring of sky and a conical mirror that blocked reflections of CG flashes from the balloon and increased light collection from distant TLEs. Telemetry timing was provided by an on-board GPS receiver.

2.2. Ground-Based Optical Observations

[10] We made low light level video observations of the sprites, jets, and elves from three sites located east of the Rocky Mountains. Simultaneous multiple site video recordings permitted the triangulation of some events [Wescott et al., 2001; Bering et al., 2002a, 2002b].

2.3. Flight Operations and Weather Conditions

[11] The 1999 Sprite Balloon Campaign consisted of three high altitude balloon flights, one from Palestine, Texas, and two from Ottumwa, Iowa. Flight 1 flew from 01:14:31 UTC on 6 July 1999 to cut down at 09:45:00 UTC. Flight 2 was not useful since the storms faded after launch. Flight 3 flew from 00:39:32 UTC to 11:12:00 UTC on 21 August 1999. The first flight of the program, on July 6, was an engineering test. However, it will be shown that we also made TLE observations. There was a thunderstorm in NE New Mexico and the Texas panhandle at the time of the flight [Bering et al., 2002b]. During flight 3, there were two thunderstorms which produced observable TLEs, in South Dakota and Kansas [Bering et al., 2002a, 2002b]. The first of these was active in eastern South Dakota during the first part of the night. It produced about half of the observed TLEs, including all of the events that could be triangulated. Bear Mt. clouded over at about 0830 UTC, and YRFS was cloud covered from 0830 to 0900 UTC. From 0830 UTC, all activity was observed in southern Kansas and northern Oklahoma by only WIRO.

3. Results

[12] Events were identified using a three step process. The ground station video tapes and logs were reviewed. Since the video frame rate is not fast enough to distinguish between halos and elves, amorphous blobs were initially classified as “halo/elve(?)” events. The burst memory data from flight 3 were reviewed for evidence of TLE activity in the uplooking photometer. The polarity of the associated ΔEz pulses was also noted. The second stage of the process involved examining the balloon and high speed imager (HSI) data at the time of the events in normal speed video data and reexaming the video and HSI data at the time of the burst memory events. The availability of 1 ms timing in the balloon and HSI data enabled resolving the halo/elve ambiguity. In some cases, including the event presented by Bering et al. [2002b], the sprite halo at the start of the event could only be seen in HSI data, since the halo was obscured in the normal speed video by the sprite.

[13] The results of this study are summarized in Table 1. There were 26 sprites recorded on the ground, with 23 recorded by the balloon. Fifteen of these events were associated with NLDN CG strokes, while 11 were not so associated. There were 26 +CG halos found in this process, 22 by the ground based video and 22 by the balloon, with 18 events seen by both ground and balloon detectors. Seven of these +CG halos were not followed by sprites. Five of the 26 sprites were not preceded by observed sprite halos. In 3 of these cases, HSI data were not available, which means that halos could have gone undetected. In one of the other two cases, there was a TLE observed by the balloon 180 ms before the sprite, which may have been the precursor halo. In the other case, the field of view of the HSI is too narrow to provide conclusive evidence of the absence of a halo, since the brightest sprite was off the left edge of the HSI image. At this stage of the study, 17 TLEs had been identified in association with −CG strokes, 12 of which were confirmed as halos by video data. All of the 5 events seen only from the balloon were delayed enough with respect to the causative spheric to be identified as halos rather than elves. None of these were followed by sprites. Three of the −CG halos were only found in the LLTV data by using the balloon event times as guides. The difficulty in detecting them at first stems from the fact that they were dim, fast, low contrast events (see below). The remaining 5 unknown balloon TLEs are classified as halos owing to the fact that normal speed TV can fail to detect a 1 ms low contrast amorphous blob, whereas a sprite is much harder to miss and should have been seen. The last step reviewed 100 ms detailed plots of the balloon data at the time of all reported NLDN strokes during 150 min from flight 3 and 96 min from flight 1. This search found another 7 +CG and 49 −CG TLEs during flight 1 and 88 −CG TLEs during flight 3.

Table 1. Sprite and Halo Observations
SourceTime Tag+CG−CG
Total 33155
Ground 2212
 Gnd or Burst2217
 NLDN Flt. 1749
 NLDN Flt. 3 88
Both Gnd and Balloon1812
No Sprite (Video Available)717
Total 250
Ground 250
Balloon 220
No Halo 50

3.1. Negative Halo at 0746:35.732 (Flight 3)

[14] The event at 0746:35.732 UTC is an example of a −CG halo, shown in Figure 1. The balloon was about halfway along the flight track and the stroke was in northwest Iowa. This event was found in the YRFS Xybion LLTV record after it was discovered in the balloon data in a 17 ms video frame beginning at 0746:35.720 (Figure 2). The halo image subtends 24° at 952 km, which implies a width of 396 km or four times wider than some but not all previous reports [Barrington-Leigh et al., 2001; Yair, 2003]. On the balloon, maximum light intensity was observed at 0746:35.732. The NLDN observed a 248 kA −CG stroke at 0746:35.728, at a distance of 365 km from the balloon. The dashed line in Figure 1 indicates the expected arrival time of the stroke signal at the balloon. The VLF sferic associated with the CG stroke appears as a burst of spikes in all 6 components of the field for ∼1 ms immediately following the dashed line.

Figure 1.

Burst memory data showing a −CG halo event digitized at 50 kHz. From top to bottom, the figure plots photometer anode current, Bz, Bϕ, Br, Ez, Eϕ, and Er, in stroke centered cylindrical coordinates as functions of Universal Time. The dashed line indicates the retarded time of the CG stroke.

Figure 2.

Yucca Ridge LLTV frame beginning at 0746:35.720. Halo is circled in the lower right. The halo image subtends 24° at a range of 952 km.

[15] There were significant differences between the electromagnetic signature of this event (halo) and those of previously published events [Bering et al., 2002a, 2002b]. First, the sign of all the halo associated perturbations was reversed compared to +CG sprite perturbations [Bering et al., 2002a]. Further, the electric field perturbation starts just ∼80 μs before the halo light pulse seen on the balloon and has longer rise and fall times than a positive CG sprite. The delayed light and ELF pulses are inconsistent with the EMP model of an elve, but consistent with an event that had significant current flow in the mesosphere [Pasko et al., 1998; Stanley et al., 2000], since delayed ELF pulses like this are only produced by mesospheric currents [Bering et al., 2002a]. The duration of the light pulse is well resolved by the photometer. In fact the anode protection circuit may contribute to underestimating it. We have one true elve in burst memory. For that event, our photometer observes the delay to be 180 μs and the light pulse width to be 80 μs which shows that the photometer can resolve the duration difference between an elve and a halo. The absence of a detectable ELF pulse in the electric field at the time of the CG stroke is consistent with previously published +CG sprite observations [Bering et al., 2002a], and is very poorly understood.

4. Discussion

[16] The difficulty in interpreting these results arises from the fact that the balloon borne all sky photometer is just that, a photometer, with no imaging capability. Thus, strictly speaking, all of the TLE events detected by the balloon instruments that were not observed by at least one ground based video system must be classified as unidentified TLEs. It is productive, therefore, to begin by considering the events that were observed from the ground. There were a total of 34 sprite halos observed by one or more video cameras during flight 3. Of these, 12 were associated with −CG strokes. The event shown in Figures 1 and 2 is one of these events. The 120 mV/m EZ pulse shown in Figure 1 indicates that the sprite had a current moment of −44 kA km with an estimated energy deposition of only 570 kJ [Bering et al., 2002a]. This moment corresponds to case 1 or 2 in the introduction, but was not large enough to expect an untriggered sprite. Several of these events were only found in the video data by using balloon event times as a reference. Of the halos seen on the ground 7 +CG and all 12 −CG events were not followed by a sprite. In the case of events that were not observed with the HSI, the 1 ms resolution PCM data was sufficient to permit halo/elve discrimination if the delay between CG stroke time and light emission exceeds 2 ms, which it almost always did. One of the 12 −CG halos was observed with the HSI along with an elve and was clearly a halo.

[17] These results have implications for the identification of TLEs observed only on the balloon. It is known that sprite occurrence in association with −CG strokes is extremely rare [Barrington-Leigh and Inan, 1999; Barrington-Leigh et al., 2001]. On this basis, one may infer that a TLE occurring within 200 μs of −CG was an elve and that a TLE delayed a few ms from the −CG was a spriteless halo. This approach gives a total of 154 −CG halos during the total of 4.1 hours of data examined in detail from both flights. The seven +CG events observed during Flight 1 are more problematic. If the statistics were the same as during Flight 3, 5 or 6 were sprites. However, the storm was a relatively small frontal storm, a situation not known for sprite production, which suggests it is more conservative to assume that all or most of these events were also halos, which is how they have been classified in Table 1. Observational selection owing to the faint, fast and amorphous character of the events and extreme range (∼700 km) are major factors in accounting for the absence of corresponding −CG spriteless halos in the video data. Even with a known event time, it was usually impossible to identify signatures in the video data. However, fast ground-based photometer data do show a large number of 1–2 ms pulses that are not accompanied by corresponding video observations (Geoff McHarg, private communication, 2003). In conclusion, most of the predictions of the concept that the sprite halo is a fundamental mesospheric response to tropospheric lightning appear to be right. Negative haloes not only occur, they are more common than positive ones in absolute numbers. Even for positive events, spriteless halos are fairly common, whereas the few apparently haloless sprites that were observed are not convincing evidence that the opposite occurs often, if at all.


[18] The authors are indebted to Danny Ball, Glenn Rosenberger and the staff of the National Scientific Balloon Facility. Tom Nelson served as technical project manager at Yucca Ridge Field Station during the campaign. This research was supported by NASA grant NAG 5-5126 and NSF Grant 0221512 to FMA Research.