The University of Alaska deployed a high speed (1000 fps) camera at the Wyoming Infrared Observatory to observe sprites over Midwest U.S. thunderstorms as part of the 1999 NASA Sprites Balloon Campaign. Here we report on the velocity of development of downward spatial structures known as tendrils in several sprite events recorded during a large thunderstorm over eastern Nebraska the night of 18 August 1999. Downward tendril development occurred at velocities that varied by more than two orders of magnitude, ranging from ∼105 to ≥3 × 107 m/s. The tendrils progressed through multiple velocity regimes, typically in the order fast-slow or slow-fast-slow. Examples are presented of multiple sets of temporally distinct tendrils that develop from the same sprite event and tendrils that have a horizontal component of expansion.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Sprites are large (∼103 km3) but brief (∼ms) optical flashes that have been reported above very active thunderstorms. They span the full vertical extent of the atmosphere, from above cloud tops to the ionosphere ∼100 km [Sentman and Wescott, 1995]. Takahashi et al.  and Fukunishi et al.  showed with photometers that sprites are initiated near 70 km altitude with downward tendrils moving with velocities on the order of several times 107 m/s, followed by the creation of upward branches. Stanley et al.  and Stanley  used a 1000–3000 frames per second (fps) low light level camera and observed sprite tendril propagation velocities of 106 to 107 m/s, and suggested they were streamers. Raizer et al.  modeled a streamer in the mesosphere, with resultant velocities undergoing a 106 to 107 to 106 m/s (slow-fast-slow) progression over the lifetime of the streamer, with the maximum velocity being attained in ∼4 ms following initiation. In this paper, we show representative examples of sprite tendril behavior from high speed imager observations.
2. Instruments and Data
 In August 1999, the University of Alaska Fairbanks deployed a high speed imager (HSI) to wyoming for ground-based sprite observations as part of the 1999 NASA Sprites Balloon Campaign [Bering et al., 2002]. The HSI consists of a 256 × 256 pixel, 8 bit (256 greylevel) intensified CCD operated at 1 ms frame resolution. The field of view is 6.4° square. The imager is described in more detail in Stenbaek-Nielsen et al. . The HSI was boresighted to an intensified CCD scene camera with a wider field of view, which was operated at a television rate of 60 fields/sec [McHarg et al., 2002]. On the night of 18 August 1999 an active thunderstorm occurred over Nebraska, at a distance of 500–700 km to the southeast of the observing site. Over the course of the night the thunderstorm moved 500–700 km distance from WIRO. At these distances, the 6.4° field-of-view of the HSI corresponded to a transverse extent of 55 to 80 km, which, given typical dimensions of sprites, provided a nearly optimum image frame size. These operating parameters bracket the range of velocities that could be determined from the observations. A brightness feature visible in the upper pixel of one field and propagating to the lowest pixel of the subsequent field corresponds to the maximum resolvable velocity of approximately 8 × 107 m/s, while a brightness feature propagating over 1 pixel (corresponding to approximately 300 m at that range) every ∼10 ms results in the lowest effective detectable velocity of ∼104 m/s. The upper limit of 8 × 107 m/s is below the fastest velocities reported by McHarg et al. . We used range based on the location of the parent lightning discharge as registered by the National Lightning Detection Network [Cummins et al., 1998]. We assumed the sprites occurred directly above the parent discharge, although they may be laterally displaced by up to 50 km from the parent lightning [Wescott et al., 2001]. This assumption introduces an uncertainty in the altitudes of ±5 km and in the velocity calculations of less than 10%.
3. Fast: Order of 107 m/s
 Sprite tendrils develop and propagate with a wide range of velocities. In this section we present examples of fast (107 m/s).
 In Figure 1 two images are shown, 1 ms apart. The first image shows a bright and horizontally extensive halo, together with multiple beads that have persisted from an earlier sprite. A halo is an unstructured emission modeled to be due to quasi-electrostatic fields from lightning [Barrington-Leigh et al., 2001]. The second field 1 ms later shows the same halo and beads, but also a large sprite with extensive downward tendrils. The velocity of propagation, assuming the sprite initiated from beads immediately below the halo in the first image (not visible at this resolution), is at least 3 × 107 m/s, close to the velocity detection limit of the imager. This event was also observed with a multichannel photometer by McHarg et al.  (their Figures 3, 4), yielding similar velocities. The downward velocity of the tendrils is the same order of magnitude as measured by previous researchers.
4. Slow: Order of 105 m/s
4.1. Downward Propagation
 After the initial phase of fast-moving tendrils sprite tendrils always slow down as they continue to propagate. Most often sprite tendril brightness decreases beneath instrument threshold within <5 ms after deceleration. The sprite shown in Figure 2 includes, in contrast, long-lasting tendrils. The sprite itself is shown at top, together with a thin black rectangle, indicating the region of interest. This region was cropped from 25 successive images, and stacked side-by-side, giving effectively the time series of the brightness within it. The contrast-enhanced time series is shown at the bottom of Figure 2. The width of a 1-ms “slice” is shown above the altitude labels of the time series. Visible in the time series are the tendrils coming in from the top (label “a”). At time “b”, the tendrils are bright and emit sufficient light in all directions, that the light scattered by the intervening atmosphere back towards the detector raises significantly the background noise level of the detector. With contrast enhancement to show dim features at a later time, at time “b” the tendrils thus falsely appear to reach the bottom of the image, instead of the actual altitude reached of approximately 40 km. During the next ∼20 ms, a dim brightness is visibly descending downward (“c” and dashed line) into the lowest quarter of the image. This downward moving brightness feature indicates that the tendril, having decayed in brightness and velocity, nevertheless continues its downward motion. The velocity is 4 × 105 m/s downward.
4.2. Horizontal Component of Expansion
 In addition to continuing downward propagation following their fast phase, sprite tendrils may also include a horizontal component of expansion, as is shown in Figure 3. On the left is the sprite marked with a rectangle showing the region of interest. This region was cropped from 18 successive images 1 ms apart, and these were stacked into a top-to-bottom time series as shown to the right. The time series again includes a section with much scattered light (“a”). After this period, the high speed imager was able to pick up the slower and dimmer outward expansion (during “b” with slope equivalent to dashed lines shown below). The speed of this horizontal component of expansion is 3 × 105 m/s towards the left, similar towards the right.
5. Varying Velocities
 The downward propagation velocity often varies in time within individual tendrils as shown in Figure 4. At the top of Figure 4 a sprite is marked with a region indicated by a black rectangle, and at the bottom the time series for that region is shown. The time series shows 50 successive 1 ms images. The cropped region includes the sprite initiation, which is visible in the beginning of the time series (“a”), as two a beads brightened by a passing tendril. The beads are visible as a horizontally dotted line due to its long lifetime. In the first 12 ms, the velocity of the tendril is, on ∼9 × 105 m/s. There is a sudden brightening of the tendril, and an increase in velocity by an order of magnitude, that appears as the first vertically extensive brightness on the 13th field (“b”). The velocity is at least 1.3 × 107 m/s. Following the acceleration of the tendril, there occur upward growing branches observed as a brightness extending vertically into the upper half of this time series. There is enough light scattered to increase the background level of the next two crops (“c”). During this same interval the tendrils slow down to near original level.
Figure 4 also illustrates a common characteristic of many complex sprite events, namely that the tendrils and branches tend to be much brighter but of shorter duration than the beads. The sprite duration, based only on branches and fast tendrils, is less than 5 ms, but beads brightened by passing tendrils and branches (“d”) may persist >20 ms.
6. Multiple Sprite Tendrils Within a Sprite
 The examples presented above include only one set of tendrils per sprite. Other events exhibit more complex tendril structure. Multiple tendrils occurring sequentially result in longer sprite durations, as well as more complex sprite shape. One such example is shown in Figure 5. In the upper right is the sprite image, with two rectangles showing the regions of interest, one oriented vertically to show the downward motion and one horizontally to show the horizontal component of expansion explicitly. Two time series are constructed as before. At the left of Figure 5 the time series is shown for the horizontal rectangle, with 51 successive 1 ms images. At the bottom the time series is shown of vertical rectangle region, covering the same time interval.
 The two time series in Figure 5 show multiple downward-propagating tendrils (arrows in bottom time series) that may also be seen expanding horizontally (arrows in left time series). The horizontal component of expansion has speeds of 4 × 105 and 1.5 × 105 m/s, respectively. The tendrils visible in the first three fields of bottom time series (first arrow) start with velocities of 7 × 106 m/s, then slow immediately to 3×106 m/s and lower. The second wave of tendrils has velocities near this smaller value, then slows to 6 × 105 m/s. The third wave of tendrils, once it leaves the bright region, has velocities around 106 m/s. Other downward propagating tendrils are (barely) visible at later times. The multiple tendrils within the sprite result in emissions which persist much longer than those of the sprites shown previously. Due to the multiple tendrils, the sprite shape does not fit into the basic sprite forms.
 Sprite tendrils have been suggested to be streamers [Raizer et al., 1998; Pasko et al., 1998]. The velocity of 105 m/s in Figure 2 corresponds to the low end of velocity regimes in which streamers are known to exist [Pasko et al., 1998]. The slow streamer, at small overvoltage and a high pressure, appears to be the most likely explanation for the slow downward velocities of the tendrils seen in our data. Distinguishing between the two effects is difficult since they are related - at lower altitudes the breakdown electric field is higher, and a given external electric field will have a smaller overvoltage over the breakdown field.
 The nearly uniform downward speed of sprite tendrils in Figure 2 for intervals exceeding 10 ms creates an intrigue. Raizer et al.  have shown with their constant electric field model that once sprite tendrils reach their maximum velocity approximately 4 ms from their onset, the sprite tendril velocity continuously decreases from that moment on, and the sprite tendrils are practically stopped 4 ms later. Thus, in order to have a near-constant velocity of propagation for the slow downward tendrils shown in Figure 2, if these brightness structures are streamers, the electric field would need to be increasing in order to sustain their propagation. However, the emissions in label c of Figure 2 do not have the characteristic shape expected to be created by streamers. Most of the energy deposition in streamers occurs at the head [Pancheshnyi et al., 2000]. For example, Stanley et al.  observed sprite tendrils with a 0.3–1.0 ms resolution high speed imager and showed that in many of their examples, the sprite tendril brightness extends down to a certain altitude on one field, and continues from that altitude downward on the following field. This indicates that most of the emissions come from a relatively narrow region propagating downward. In contrast, our spatially-extensive emissions behind the front of the slowly-propagating tendrils in Figure 2 would not be expected with a streamer mechanism alone.
 The slow tendrils (both downward and horizontal component) of a sprite always appear very dim, and we observe them as a wavefront rather than as individual structures. At present it is not clear whether the wavefront we observe is real or whether it is only an apparent feature due to insufficient instrumental response. Based on the horizontal component of tendril expansion shown in Figure 3, which occurred in a region where the tendrils initially had nearly exclusively vertical component, we suggest that at least in that case, the wavefront is truly an expanding front and not multiple, dim, horizontally-propagating tendrils.
 The time evolution of tendril velocity, in Figure 4, is consistent with the basic slow-fast-slow characteristics of the model presented by Raizer et al.  (their Figure 3) for some sprites. Other sprites have a fast-slow evolution of sprite tendril velocity instead.
Figure 5 shows a sampling of sprite dynamics observable at 1000 fps imagery. Here, we see that a single sprite may have multiple tendrils with beads, and that upward development of the set of branches follows the development of downward tendrils. The beads do not brighten with each downward tendril, but they tend to persist for longer times than the tendrils. It is presently unknown what initiates successive streamers, or whether a single seed is associated with these events.
 Finally, comparison of tendril velocities with the associated sprite shows that the maximum tendril velocity appears to be correlated to the brightness and the overall extent of the sprite. The brightest sprites tend to have tendril velocities above 107 m/s, while smaller and dimmer sprites tend to have slower tendril velocities (e.g., Figures 1 and 5).
 The detailed view of sprite dynamics observable in 1000 fps imagery has revealed numerous new features that challenge our understanding of the effects of lightning on this region. We have presented the dynamics of tendril propagation in several sprites. The principal conclusions of this work may be summarized as follows:
Very slow downward tendril velocities have been identified in sprites, on the order of 105 m/s.
Horizontal component to propagating sprite tendrils has been identified in some sprites, expanding at ∼105 m/s.
Sprite tendrils do not develop at constant velocity, but rather pass through multiple regimes. Typically they follow a fast-slow pattern. Some sprites follow a slow-fast-slow pattern, similar to that predicted by Raizer et al. .
Sprites may possess multiple tendrils.
The duration of beads within a sprite is typically much longer than that of either tendrils or branches.
 This research was partially supported by NASA Grants NAG5-5019 and NAG5-0131 to the University of Alaska. DRM acknowledges receiving partial support from Geophysical Institute internal research funds and from the University of Alaska Fairbanks Graduate School. We thank the University of Wyoming for permission to use the Wyoming Infrared Observatory at Jelm Mountain, WY.