INAUGURAL ARTICLE by a Recently Elected Academy Member:The cloud–ionosphere discharge: A newly observed thunderstorm phenomenon

Large-scale systems in nature are often remarkable in that they appear, locally at least, to have reversed the second law of thermodynamics, and have gone toward a state of greater order and decreased entropy. An example is the surface of the sun, in which the photosphere is normally at an average equilibrium temperature of 6000 K, which is equivalent to an energy per particle of less than 1 eV. Yet through the large-scale interaction of ionized matter with regions of enhanced magnetic fields in sunspot groups, thermal energy particles can be accelerated to cosmic ray energies in the GeV (10⁹ eV) range (1). Thunderstorms constitute another more familiar example of a large-scale system that can generate very high electric potentials of a transitory nature in a very low-temperature environment.

Some Basic Principles of Conventional Thunderstorms

The electric fields within thunderclouds have been directly measured by balloon soundings (2) and may reach 150 kV [center dot]

m⁻¹, which corresponds to potential differences within or between clouds in the million-volt range. In the current theory of cloud electrification, when water vapor condenses into droplets, electrical charges are produced and heat is released. The resultant convection carries the smaller, lighter components upward, and the heavier drops fall. This process results in a charge separation, as if the electric polarity of the drops were a function of their size. In the simplest scenario, the clouds where this action occurs, for example the “towering cumulus” thunderheads, develop a charge distribution in which the main component is a dipole, with the cloud tops positive and the bottoms mostly negative [see, for example, Uman (ref. 3, p. 59, figure 3.1)]. The high potentials within thunderclouds are continually dissipated, mostly by lightning discharges within clouds or from cloud to cloud. Cloud–ground (CG) strokes are also common, and are mostly “negative,” that is, they carry the negative charges at the cloud bottoms to earth. This type of lightning has been widely studied because of its potential hazard to objects at ground level, and it is the kind of lightning which is often photographed because the microsecond-duration strokes through relatively clear air between clouds and ground have an enormous optical intensity concentrated in channels less than a meter in diameter. There is a large literature on the subject of conventional lightning—i.e., CG discharges (3).

High-speed photos of CG discharges or “flashes” in a microsecond time frame show that there is an initial “stepped leader” ionized path which makes its way from clouds to ground, at which point a high-current “return stroke” occurs, which is often multiple, with up to 12 successive strokes along the same path in less than a second. The electric current carried by return strokes may reach several hundred kiloamperes in a microsecond interval. Beside the intense flash and the acoustical shock (thunder), the return stroke produces powerful pulses of electromagnetic radiation called “sferics,” often heard as bursts of crashing static on AM radio broadcast channels during nearby lightning storms. Because of the great vertical length of CG lightning strokes (up to several kilometers) they radiate efficiently in the very low frequency (VLF) range below 10 kHz (4, 5). The National Lightning Detection Network (NLDN) uses these emissions and operates radio direction finders across the United States, which locate the CG impacts to an accuracy of a few kilometers. Knowledge of the NLDN CG impact locations is of practical importance—e.g., for public safety—but has also been very useful in locating associated phenomena such as the “CIs” to be described below.

Early Observations of Cloud–Ionosphere Discharges (CIs or “Sprites”)

This paper is concerned with thunderstorm-associated large luminous flashes between cloud tops and the ionosphere—i.e., mesospheric “lightning” very different in character from the classical CG stroke just described. Reports of visual observations date back to the 19th century and refer to luminous flashes appearing above the tops of cumulus thunderclouds. One of these reports was by C. T. R. Wilson, inventor of the cloud chamber that carries his name (6, 7). Air transport pilots, flying above local clouds, were often in a particularly favorable situation for observing distant storms through clear air, and they reported the observation of large flashes above thunderstorms. Many of their recorded comments have been collected and summarized by Vaughan and Vonnegut (8). Finally, on 9 July 1989, while making observations under the “SKYFLASH” (SKF) program, a video image, reproduced in Fig. 1, of a large mesospheric region complex flash above a thunderstorm was obtained by our group from the O’Brien Observatory of the University of Minnesota, located 40 km northeast of the Twin Cities (Minneapolis and St. Paul) area (9). The flash was recorded with an image-intensified CCD type low-light video camera intended for a rocket flight being tested by observing distant lightning on the horizon. The flash, which was observed through clear air, was subsequently associated with an active thunderstorm complex over northern Minnesota, more than 250 km to the north of O’Brien. Near the bottom edge of Fig. 1 in the CG region a typical CG stroke image has been overlaid for showing the striking differences between this mesospheric flash and “ordinary” lightning. For correct proportion to the flash, the CG stroke should be one-third the size shown. Due to large uncertainties in the range of the flash from O’Brien, the actual size could only be estimated at more than 20 km in vertical extent and 10 km in width. Its duration was almost entirely within one video field (16.7 ms). It consisted of two plume- or fountain-like luminous regions, with small secondary features intermingled. The surface brightness was similar to an average auroral display [50 kilorayleighs (kR; 1 R = 10¹⁰/4π quanta [center dot]

m⁻²

s⁻¹

sr⁻¹)]. Observations were continued during the night of 9–10 July 1989 at O’Brien, but this was the only case of this kind recorded.

Figure 1

This first image of a large mesospheric region complex flash above a thunderstorm was recorded on 9 July 1989 from the O’Brien Observatory of the University of Minnesota, located 40 km northeast of the Twin Cities area. The event was recorded (more ...)

Recent Observations of CIs

The published account of this event in Science, by Franz et al. in 1990 (9) eventually sparked much interest in the lightning community, and during the period 1993–1996 mesospheric discharges above thunderclouds were observed by many groups. In this paper such objects are referred to as CIs. Many investigators have called these discharges “sprites,” a meaningless term scientifically, but suggesting a will-of-the-wisp type object. The use of the term “discharge” assumes at the outset that these events are an electrical excitation of the mesospheric atmosphere to emit light as a result of thunderstorm electric fields, a broad widely held concept whose details are under active discussion.

Chronologically, the next recording of CIs after the Franz et al. event was from the payload bay TV camera on space shuttles from flights in 1990 and 1991 as small pillars above cloud flashes, at a large distance from the shuttle near Earth’s limb (10, 11). The summer of 1993 saw a surge of activity, with CI images recorded from a research aircraft by the Sentman group over the central United States (12) and extensive ground-based observations by Lyons (13) from an observing point in the high plains region near Fort Collins, Colorado, known as “Yucca Ridge.” Many CIs were observed above large storm regions called by meteorologists “mesoscale convective complexes” (MCCs) located several hundred kilometers to the east of Yucca Ridge over the central United States. An example of a large CI or “sprite” recorded by W. A. Lyons from the Yucca Ridge station on 6 August 1994 is shown in Fig. 2. Like the Minnesota event, it was recorded with an image-intensified CCD camera through clear air to the storm region, but it was a larger event and extended from about 40 to over 80 km in altitude, and 50 km horizontally. It was located 367 km to the east of Yucca Ridge and was visible to the naked dark-adapted eye as a salmon-colored flash. The top section near 80 km altitude apparently consisted of an unresolved group of vertical filaments, which trailed downwards into streamers reaching toward cloud tops. It was associated with a 128-kA positive CG stroke recorded by the NLDN whose light probably caused the intense cloud light-up visible at the bottom of Fig. 2 by scattering of light in the cloud base.

Figure 2

A giant CI group recorded from the Yucca Ridge Field Station by W. A. Lyons on 6 August 1994. The event was 367 km to the east of Yucca Ridge and extended from about 40 km to over 80 km in altitude, and 50 km horizontally. The top section near 80 km apparently (more ...)

Another important observation was made by Sentman et al. (14) with a color TV camera carried by a jet aircraft, which showed that the CI events were red in the mesosphere and blue at cloud tops (Fig. 3; see also ref. 15). The red color has been identified by Mende et al. (16) as the first positive band lines of molecular nitrogen. This similarity to N₂ lines emitted by the lower borders of polar auroras suggests that the same process of bombardment by energetic electrons which produces the auroral color was active here. However, the mesospheric electron collision frequency is very high compared with the electron gyro frequency, so that the dominance of the magnetic field present at auroral altitudes would not be expected for CIs. The Sentman group also, by triangulation from two aircraft flying near the MCCs, determined the vertical dimensions of CIs directly, without needing the NLDN CG stroke location.

Figure 3

This unusual video image of a CI was obtained by D. D. Sentman, using a color low-light-level TV from an aircraft. It shows that the CI events were red in the mesosphere and shaded into a blue glow around cloud tops. The red color is attributed to the (more ...)

Further observations were also made from the Minnesota O’Brien Observatory in 1993, and on the night of August 9–10 more than 150 CIs were recorded by TV, all located above a giant MCC which spread over Iowa and Missouri, about 450 km from O’Brien. An example of these events is given in Fig. 4, which shows a “carrot” shape CI resembling the 1989 event described above, but about 50 km in vertical extent, extending from 30 to 80 km altitude. All the more intense CI events recorded in this storm were coincident within 1 s with an electromagnetic pulse or “sferic” recorded by the SKF instrumentation and with a CG stroke recorded by the NLDN. An analysis of the 9–10 August 1993 events has been published (17) and will not be discussed in detail here.

Figure 4

“Carrot” CI from the storm of 9–10 August 1993. This CI was about 50 km in vertical extent, extending from 30 to 80 km, and appears to reach almost to the cloud tops, which were illuminated and made visible by the CG stroke. It (more ...)

Observations on 12 July 1994 from the Yucca Ridge Field Site

As experience in recording CIs progressed, it became apparent that although the CCD TV cameras in standard format had ample sensitivity for imaging CIs, the 16.7-ms integration time of the CCD element (i.e., one TV field) was too long to follow the time development of the CI in the mesosphere. In almost all events, the luminosity, when examined field by field, appeared and mostly disappeared in the first TV field after the CG stroke (this was true of the first event imaged, in Minnesota in 1989 (ref. 9; see above). While the SKF photometers had better time resolution (0.1 ms), their response was almost entirely to very intense CG strokes and not to CIs.

The problem was attacked in a preliminary way in a collaborative effort between the Lyons and Winckler groups, in which new observations of CIs were made from the Lyons Yucca Ridge Field Station, near Fort Collins, Colorado (18). One of two TV cameras was equipped with a telescopic lens, and “bore sighted” with a high-speed telescopic photometer which had the same field of view as the TV. The equipment also included a 50-kHz bandwidth sferic detector and a new computerized data system. The telescopic TV and photometer were aimed above the distant clouds to include a mesospheric region where only CIs were expected, and the bright cloud flash due to the initiating CG would be out of the field. On the evening of 12 July 1994, more than 30 CI events originating to the east of Yucca Ridge were recorded, and 5 of these were selected for detailed analysis. The CI designated event 4, which has the form of an “angel,” is shown in Fig. 5. This image is computer processed, but the color is similar to the real color of CI events. The CI has a bright, tall central form, in which detail is lost through halation, flanked with adjacent filaments. It is situated above a large, intense cloud flash, probably as usual from the CG stroke. This event 4 is also shown in Fig. 6 with vertical and horizontal scales annotated on the image, derived under the assumption that the CI was close to the associated CG stroke location determined by the NLDN to be 291 km from Yucca Ridge. The CG stroke azimuth is shown in Fig. 6, about 20 km to the left of the CI center. The telescopic TV view of the upper portion of the CI is shown in Fig. 7, in which the circle defines the field of view of the photometer. When magnified in this manner by the telescopic lens, and with reduced halation in the TV camera, the CI is seen to consist partly, but probably entirely, of vertical striations (see also Figs. 10 and 11 below).

Figure 5

The CI designated event 4, which has the form of an angel. This image is computer processed but the color is similar to the real color of CI events. The CI has a bright, tall central form, in which detail is lost through halation, flanked with adjacent (more ...)

Figure 6

CH01 (wide angle) annotated TV image of the “angel” sprite of event 4. Note that laterally, at least, the associated CG stroke location is about 20 km to the left of the CI center. For details of vertical and horizontal scale annotation, (more ...)

Figure 7

CH02 (telescopic) TV image of the “angel” CI of event 4. The high-speed photometer field of view lies within the circle and includes the top section of the CI. Note the “filaments” of various sizes distributed across the (more ...)

Figure 10

Wide-angle view of a large CI by CH02 (Table 1 event 23). This CI was associated with a +54-kA stroke about 330 km from the observatory in a direction of 203°. As seen by camera CH02, the CI is bright enough to be seen through thin foreground (more ...)

Figure 11

Magnified view by CH01 of the large CI shown in Fig. 10. Note the near-vertically aligned complex filamentary structure, including many bright beads along the filaments. This CI is about 60 km in width.

Both the high-speed photometer and the sferic detector gave data from all five events on the night of 12 July. Fig. 8 shows the record of event 1, which was very similar in image format to event 4 shown in Figs. 6 and 7. The event began with the sferic pulse at about 23 ms (see Fig. 8 x-axis scale) followed by an initial photometer peak at 26 ms, which is interpreted as scattered light from the CG (recall from Fig. 7 that the photometer did not directly view the CG region) and a peak at 29 ms, which is assigned to the CI. The CG flash (or sferic) is coincident within 1 ms with a 50.5-kA positive CG stroke as recorded by the NLDN, which must be responsible for the initial responses. The duration of one TV field has been annotated at the top of Fig. 8, and it can be seen that the entire event, from sferic to the end of the CI, occurs well within one TV field. Thus the time frame important for observing the CI development is substantially shorter than the 16.7-ms time resolution of the cameras.

Figure 8

Yucca Ridge high-speed photometer and sferic traces for the CI of event 1. Note the initial photometer response within 1 ms of the sferic, followed by a small maximum due to the flash from the CG stroke, and finally a large maximum due to the CI.

Another important fact is that all of the CI events in the Winckler–Lyons study (18) were triggered by positive CG strokes. Lyons (19), with a larger data sample, found the same result, as did Winckler for the 21 June 1996 storm described in the next section.

Observations from O’Brien Observatory, 21–22 June 1996

During the years 1993–1996 the SKF system at O’Brien Observatory was set up to record lightning storms within a radius of 1,000 km from O’Brien, and 11 storms were studied. These were for the most part frontal storms that were observed before their arrival or after their passage over the station. Most of these storms produced no observable CIs, but there were two major exceptions: (i) the many CIs observed during the storm of 9–10 August 1993 described above, and (ii) a recent storm on the night of 21–22 June 1996, which produced 38 CIs in the field of view of the TV cameras. Details of the data taken on this night have not previously been published, but will be discussed in some detail here. This storm had two regions of activity, one extending across southwest Minnesota, Iowa, and Nebraska, and the other in western North and South Dakota, as shown in the NLDN stroke map of Fig. 9. Data were recorded for 6 h, from 21:00 Central Standard Time on 20 June to 03:00 on 21 June (03:00 to 09:00 Universal Time, 21 June) 1996, during which 22,111 CG strokes were recorded by the NLDN. Of these fewer than 2,000 were estimated to have been positive. Of the 38 CI events in the field of view of CH02 (see defining lines in Fig. 9) 35 were associated within 1 s with a positive CG stroke (see Table 1) and 3 with negative strokes. The very high flash trigger rate from the storm resulted in a high “dead” time for the SKF equipment, so that only 7 of the 38 CI events had accompanying SKF data which could display the Rayleigh-scattered luminous profile in high time resolution and also record the VLF low and VLF high sferic channels for the event. Also, the observations were impeded by variable high thin clouds between O’Brien and the storm areas, which appear in the TV images shown here. Nevertheless, a variety of CI images were obtained. For example, Fig. 10 shows a large bright CI event from the 12.5-mm normal lens of CH02 and Fig. 11 shows the same event from the 25-mm narrow-field lens of CH01. The CI in Fig. 11 is clearly shown to consist of a collection of more or less parallel near-vertically aligned filamentary luminous columns, with many bright knots distributed along the filament pattern. In many views of CIs these filaments cannot be resolved, because of blooming (or halation) in the TV image, particularly in the altitude range 60–80 km, where the CI pattern is most intense (see, e.g., Figs. 2 and 3).

Figure 9

Lightning strike map for the period 21:00 on 20 June to 03:00 on 21 June Central Standard Time (03:00–09:00 21 June Universal Time). The field of view of the wide-angle camera CH02 is included between the lines and was 36°. It was directed (more ...)

Table 1

Identification of CI (sprite) events in the storm of 20–21 June 1996 observed from O’Brien

Observatory

These figures show wispy streamers extending downward toward cloud tops, which is also a characteristic of intense CIs.

CIs occur in a wide range of sizes. The CI of Figs. 10 and 11 was about 20 km in width and 30 km in vertical height. Fig. 12 shows a very small CI (within the circle) consisting of two isolated vertical filaments near 60 km altitude appearing high above the cloud decks brightly illuminated by the lights of the Twin Cities metropolitan area. This CI was correlated with a +53-kA NLDN event located 223 km from O’Brien at a bearing of 207°. The CI was at 204°, nearly at the NLDN CG stroke azimuth. The bright star Antares (arrow), faintly visible through the clouds, was at 197° and served to calibrate the CI bearing. Using these data, the size of the CI was estimated. The two parts of this CI were each about 1.9 km high and <0.2 km in apparent diameter and were separated by about 0.6 km [another example of a small CI was shown by Winckler (ref. 17, figure 11)]. Such tiny CIs do not have visible streamers and are not very intense, but they may play an important role in revealing a threshhold for CI production.

Figure 12

Example of a very small CI (in the circle) which nevertheless has the correlated attributes of a large CI (Table 1 event 8) correlated with a +53-kA NLDN stroke, and the SKF response in Fig. 13. The CI consists simply of two small vertical filaments. (more ...)

This event was also correlated with one of the nine SKF triggers associated with CIs (see Table 1), as shown in a 0.2-s SKF expanded time base format in Fig. 13. The flash event (Fig. 13 Upper) consists of a sharp pulse superposed on a bright, slow maximum, which saturates the vertical scale. The 120-Hz ripple on the trace of this event is due to the Rayleigh-scattered lights of the Twin Cities metropolitan area. These lights are largely of the gas discharge type, and fire synchronously over the whole area on each half cycle of the 60-Hz alternating current power system. The sferic record (Fig. 13 Lower) from the high-frequency VLF receiver shows a “nest” of sferics in coincidence with the slow flash maximum. It is quite possible that the slow maximum in the flash event could be resolved into a multitude of sharp peaks to match the peaks in the VLF sferic, but as the flash signal was processed with an “RC” filter which had an integrating action over 1 ms, such detail was lost. It is tempting to attribute this SKF response pattern to the positive CG stroke (sharp peak), and to a large number of small cloud discharges which accomplished the charge rearrangement in the cloud (slow peak). The burst of sferics shown in Fig. 13 Lower may be the source of a distinct “pop” heard in the 4-kHz audio track of the video tape, in coincidence with a CI flash event. The cause-and-effect sequence of the sharp and slow maxima is not understood and is under study. “Slow” luminous flash events in lightning possibly similar to this case have been reported by Li et al. (20) during a research program using sounding rockets.

Figure 13

The SKF 0.2-s data plot associated with the very small CI of Fig. 12 shows an optical flash of simple form (Upper) consisting of a sharp peak (<1-ms duration, characteristic of a CG stroke) in the initial part of a large, broad maximum flash whose (more ...)

Summary of CI Characteristics

Morphology.

The evidence suggests that CIs consist of bundles of more or less vertically oriented luminous filaments of less than kilometer width, but often extending from 40 to nearly 90 km vertically. The filaments have a bright section between 60 and 80 km, where TV images are often overexposed and do not resolve the filaments. The bottom ends of bright filaments may show thin tendrils which extend downward toward cloud tops. Very small CIs may consist of one or two less-than-kilometer-diameter filaments several kilometers high, but are situated in the 60- to 70-km “bright” level. Large events may be of 50 km horizontal dimension. They always occur above an active lightning storm, such as a frontal storm or a MCC, particularly in the storm later stages and above the “anvil” or stratiform layers, where positive CGs are more abundant. In certain multiple CI events, the individual CIs appear to “dance” across the cloud tops, in a time frame of a few 0.1-s intervals, as if following a travelling disturbance in the clouds below.

Optical Qualities.

CIs have a surface brightness similar to the aurora borealis—i.e., from 10 to 100 kR. The color is red, identified as N₂ positive bands, like the aurora. The tendrils shade off to blue below 50 km altitude. CIs are difficult to see with the unaided eye, even if dark adapted, because of the short duration (1/60 s) and low brightness, and have an integrated intensity too weak for photography. For a TV image, an image-intensified, or equivalent, camera must be used.

Duration and Triggering.

CI durations are about 3 ms at peak intensity, but may trail off at low level for several 16.7-ms TV fields. CI groups sometimes appear to last over many TV fields, but usually, on careful examination, the group is seen to be “transforming” with individual elements dropping out and being replaced by new members. Apparently, all CIs require a CG stroke trigger, and they occur about 3 ms after this trigger. With few exceptions the trigger CGs are identified by the NLDN as positive strokes, which may be situated up to 50 km lateral distance from the CI central region. Triggers and CIs are sometimes multiple, several occurring within 1 s.

Further Remarks on Triggering.

The trigger requirement contrasts with conventional CG lightning, which discharges spontaneously during a relatively slow build-up of cloud potential. However, the time interval between the trigger CG and the CI (several milliseconds) is too short for major rearrangements of charges in the local clouds. The trigger action must be to release a state of electrical stress or polarization already present in the cloud–mesosphere system. This stressed state must be of a very special nature, as only a small fraction (<10%?) of the total positive CG strokes in a storm trigger CIs. Furthermore, very few of the much more numerous negative CGs trigger CIs even though the peak currents may be similar. It is becoming clear that solving the mystery of positive CG triggering is very essential to understanding the origin of CIs.

Theoretical Models for CIs

Several writers have dealt with “quasi-electrostatic” (QE) thundercloud fields as sources of CIs (21–23). These theories are claimed by the authors to account in a general way for CIs—i.e., their timing, size, and intensity. The theories do not explain in detail such factors as the 60- to 80-km “bright” region, the filamentary structure, the wide size range of CIs, and particularly the strong bias for triggering by positive CG strokes. Theoretically, starting the CI discharge with a positive charge removed from the cloud system is equally well served by removing a negative charge. This follows simply from the fact that the initial discharge breakdown in the mesosphere is not dependent on the direction of the electric field—i.e., whether it points up or down. Furthermore, the spreading of this discharge through plasma columns may be linked to the large gradient of atmospheric density and conductivity and not to the polarity of the electric charges involved. Yet, as repeatedly pointed out in this paper, the observations are highly asymmetric with respect to the polarity of the CG stroke. For example, of the 38 CI events summarized in Table 1, 3 were coincident with negative, whereas 35 were coincident with positive, CG strokes. The cloud situation that produces CIs has been described in many papers by Lyons—for example, figure 18 in ref. 19, where a positive CG stroke produces a downward mesospheric electrical field. But a similar scenario could be devised for a negative CG stroke producing an upward mesospheric field.

There are, however, some theories that might have a “built-in” preference for positive stroke triggers, particularly those that use “runaway electrons” responding to the atmospheric density gradient and a downward mesospheric field (6, 24, 25).