Summary

Sprites and their parent discharges were analyzed in this study. The 3 dimensional and temporal development of discharges was observed with the Kennedy Space Center Lightning Detection and Ranging (KSC LDAR) instrument. A pattern of increasing horizontal dimensions of discharges which propagated into the stratiform region of MCSs was readily evident at least 20 minutes prior to the production of sprites.

Most of the sprite-producing discharges began as relatively ordinary bilevel intraclouds (ICs) similar to those documented by Shao and Krehbiel (1996). However, one of the sprite-producing discharges began as a more complex hybrid IC and negative cloud-to-ground (-CG) discharge. The negative leaders in the upper level portion of the ICs dropped in altitude from 10-12 km to 7-8 km MSL as they propagated away from the convective region into the stratiform region. The leaders then propagated horizontally through the stratiform region at average speeds of about $1-2\times10^5$ m/s while maintaining an altitude of 7-8 km MS. The altitude of the leaders is in excellent agreement with the altitude of the upper positive charge layer in stratiform regions (MacGorman and Rust, 1998, pg. 267).

The horizontal development of the negative leaders away from the convective region either stopped or was redirected after propagating a distance of about 40-60 km to near the periphery of the 20 dBZ reflectivity region at 7-8 km MSL. The outward propagation was followed by a steady static electric field change which was consistent with the downward propagation of one or more positive leaders. These leaders were poorly located with often not more than one VHF source located above the subsequent strike location.

The sprite-producing $+$CG discharges removed positive charge from 7-8 km MSL altitude, corresponding to the upper positive charge layer in the stratiform region. This altitude range is consistent with multi-station electric field measurements of $+$CG charge heights in MCSs (Krehbiel, 1981). The height of the charge removed by the $+$CG is similar to the $\simeq$7 km MSL altitude of negative charge removed by $-$CGs in the midlatitude summer (Krehbiel, 1986), indicating that neither (stratiform) $+$CGs nor $-$CGs have the ``upper hand'' in the generation of a charge moment from a given amount of charge. The idea that sprite-producing discharges remove charge from the lower positive charge layer near the $0^{\circ}$C in the stratiform region (Williams, 1998) is not supported by these measurements. Rather, negative leaders were observed to propagate into the lower positive charge region primarily during the subsequent IC portion of the discharge after the sprites had faded.

The sprite-producing $+$CGs had average downward ($-$z direction) currents of $\simeq$28-37 kA within the initial 1 ms post-return stroke interval and $\simeq$13-16 kA within the initial 4 ms. The range of charge transfers to ground for these time intervals was $\simeq$28-37 C and $\simeq$52-64 C, respectively. The non-sprite-producing CGs of both polarities had smaller average currents and charge removals for the post-stroke time intervals. These CGs also had smaller horizontal extents of charge removal compared to the sprite-producing CGs.

The charge moment threshold for sprite initiation was found to be $\simeq$300$\cdot$km for delays of up to a few milliseconds after the +CG stroke. This threshold is similar to previous estimates based on ELF measurements (Cummer and Inan, 1997; Huang et al., 1999). Most sprites were located above the approximate periphery of the parent discharge. Since the +CG continuing current is produced by the propagation of negative leaders at the end of the channels, much of the charge removed by the +CG will likely be located at the discharge periphery. Thus, the sprites may be initiating in regions of local quasi-static electric field maxima. However, there were also a few sprites which appeared to be located beyond the discharge periphery where the quasi-static field should have been smaller.

The sprites were only associated with the most recent ($\sim$200-300 ms) portion of the discharge. Also, the +CG location was usually located under the periphery of the sprite cluster in the general direction away from the previous negative leader propagation. These characteristics suggest that the older ($>$300 ms) portion of the discharge had ``cutoff'' (become nonconducting). The cutoff of a channel might have produced enough electrical stress at the conducting end opposite of the negative leader(s) for the generation of a positive leader, as suggested by (Krehbiel, 1981).

Two of the sprite-producing $+$CGs were followed by one or more $-$CG strokes which occurred underneath the region of rapid charge removal. The locations of the $-$CGs for the different discharges were nearly identical, while the $+$CGs struck at different locations. I speculate that these $-$CG flashes were not produced by negative leaders from the horizontally extensive flash, but were instead initiated from the top of one or more tall structures under the influence of the large and rapidly increasing electric fields produced by the $+$CGs.

The initiation altitude of nighttime sprites is 76$\pm$6 km MSL with a maximum density at $\simeq$76-79 km MSL. The initiation altitude is consistent with conventional breakdown predictions, though the measured charge moments were about 1.4-2.3 times less than required to initiate conventional breakdown in a homogeneously stratified atmosphere. Significant spatial variation in conductivity might enhance the electric field in localized regions to the point of breakdown. Such conductivity inhomogeneities might be produced by radiated fields from the lightning discharge (Valdivia et al., 1998).

Most sprites developed bidirectionally upwards and downwards from the point of initiation. The bidirectional development is similar to that of breakdown initiated in long air gaps in which positive and negative streamers propagate in opposite directions away from an electron avalanche region (Dhali and Williams, 1985; Loeb and Meek, 1940). The downward developing positive streamers typically attained peak velocities of $\sim\,1\times10^7$ m/s, consistent with the predictions of Raizer et al. (1998) for the observed charge moment changes in excess of 300 C$\cdot$km.

Parent discharges with unusually large and rapid charge moment changes produced sprite halos, consistent with the theoretical predictions of Pasko et al. (1997b). Angel (jellyfish) sprites initiated from enhanced luminous regions at the base of the sprite halos. Positive streamers propagated downwards from the initiation regions at velocities well in excess of more typical sprite positive streamers initiated by less energetic discharges. The angel sprite positive streamer velocities could exceed $3\times10^7$ m/s (10% of light).

On some sprites, inferred negative streamers were spawned from regions through which the positive streamers propagated previously. The negative streamers propagated and branched upwards, transforming the sprite from a columniform shape (with tendrils) to an upward-V, or ``carrot'', shape. In normal speed video, the apparent vertex of a carrot sprite at $\sim\,$65 km altitude (or lower) can appear to be the initiation point for the upward and downward branching streamers. The high-speed video observations clearly show that the actual initiation altitude is always higher than the apparent vertex of carrot sprites.

A unique sprite ELF signature was used to detect the presence of daytime sprites. During the day, the base of the ionosphere is located at a lower altitude resulting in a higher charge moment threshold for conventional breakdown. The charge moment changes of the parent discharges, in chronological order, was 6100, 4300, and 3900 C$\cdot$km at the onset of the sprite ELF signatures. In contrast, the onset of nighttime sprite ELF signatures occurred at +CG charge moment changes well below 1100 C$\cdot$km (Cummer and Stanley, 1999). The 6100 C$\cdot$km charge moment change of the first parent discharge may have been sufficient for conventional breakdown below the base of the ionosphere at an altitude of $\simeq$54 km, assuming an experimentally measured ion conductivity profile of Holzworth et al. (1985).

The daytime sprites themselves contained unusually large charge moment changes of $\simeq$2800 C$\cdot$km, $\simeq$1200 C$\cdot$km, and $\simeq$910 C$\cdot$km. These charge moments are larger than the largest nighttime sprite charge moment change published to date of $\sim$840 C$\cdot$km (Cummer and Stanley, 1999) (see Section 5.3.3).