Several criteria used for the identification of elves in narrow
field-of-view photometers are discussed in Chapter 4. One additional
criterion not mentioned there is the fast relaxation time scale
(100
s) which is often a characteristic of optical pulses due
to elves. Such fast relaxation is not typical for scattered light
from lightning [Thomason and Krider, 1982; Guo and Krider, 1982], and
observations outlined below show that it is also generally not
observed for sprites.
On the night of 19 July 1998 a large mesoscale convective system over northwestern Mexico produced exceptionally bright sprites. Measurements were made from Langmuir Laboratory using the Fly's Eye camera, optical array, and VLF receiver, as well as an image-intensified telescopic video system described by Gerken et al. [2000]. In addition, many sprites were bright enough to be visible to the unaided and unadapted eye.
Determination of total sprite luminosity lifetimes has generally been challenging [Winckler et al., 1996; Rairden and Mende, 1995]. Video recordings give generally poor time resolution and some systems, such as the image intensifier of the Fly's Eye video, exhibit a phosphor persistence following intensely bright signals. This ``afterglow'' may last for several frames, making the instrument unreliable for quantifying long sprite durations. On the other hand, photometers designed with high time resolution are not optimized for the measurement of slowly varying, weak signals, especially in the near infrared region of the spectrum [Winckler et al., 1996]. In the Fly's Eye photometers, the slow glow of sprites often appears to decay gradually into the background photometer signal level. Extra bright sprites facilitate the measurement of these longer timescales using the Fly's Eye.
Sprites are known sometimes to occur well after (up to tens of milliseconds) an associated lightning return stroke [Bell et al., 1998]. It has been proposed that this property may be due to slowly-varying currents, possibly undetectable by ELF radio measurements, which may be flowing along the ionized return stroke channel or possibly horizontally within the thundercloud [Cummer and Stanley, 1999; Bell et al., 1998; Cummer et al., 1998b]. In some cases, a series of (positive) cloud-to-ground discharges may occur sequentially over a large horizontal distance within a fraction of a second (spider lightning), suggesting the existence of an expansive travelling network of intracloud currents [Lyons, 1996]. These events are typically accompanied by a series of sprites mirroring the propagation of the lightning below (``dancing sprites''). In such cases several sprites can occur with continuous luminosity over a large fraction of a second and may appear to be associated with several lightning strokes. This paradigm was typical for the sprites observed on 19 July 1998.
In the following sections, several notable features of sprites are discussed in the context of the observations carried out on 19 July 1998. The extra signal available on this night may have highlighted some hard-to-observe but common features of sprites, or the observations may correspond only to the special case of unusually intense ionization and emissions. Several studies have suggested that the degree of ionization in sprites can vary greatly and somewhat independently of the intensity of luminous emissions [Armstrong et al., 1998b; Armstrong et al., 2000; Heavner et al., 1998].
Figure 5.15 shows a sample photometric record of a
bright sprite, and illustrates the existence of more than one time scale
in sprites. The inset image shows an intense sprite halo and bright
patches near its lower boundary at 75 km which appear to have
initiated downward streamers in a manner similar
to that described by
Stanley et al. [1999]. This event is accompanied by the photometric
signature of elves (Section 4.1) in the full array of
photometers (not shown). In photometer 4 (shown) the initial optical
pulse due to elves becomes very bright and is protracted for
2 ms.
This brightness is likely to be due to the sprite halo evident in the
video image. Approximately 6 ms after the event onset a pulse with
characteristic rise and fall times both of
2 ms appears and then
relaxes into
50 ms of less intense glow.
This example highlights several features of sprites which were frequently observed on 19 July 1998 and in the course of the annual sprite campaigns conducted by the author. Many events exhibit a bright peak which is often only a few ms in duration and tends to grow and decay with similar timescales. In addition, overall photometric durations much larger than 10 ms were found to be normal on this day, in contrast to the observations of Winckler et al. [1996].
Cummer and Stanley [1999] found that the peak in optical intensity of sprites occurred after the propagation of streamers to their lowest altitudes was complete. The same phenomenology is observed for the event shown in Figure 5.2 and may be analogous to the luminous return stroke of lightning following the connection of a leader channel to ground. In this analogy, the slower sprite glow evident in Figure 5.15 may correspond to lesser excitation of the channel during the analogue of the continuing current phase in lightning.
Measurements from the Fly's Eye's video camera, ELF/VLF sferic receiver, and
three photometers, as well as from an ultra low frequency (ULF, 30 Hz)
search coil, are shown in Figures
5.16 and 5.17.
Several notable features are apparent in the event shown in
Figure 5.16. Two cloud-to-ground discharges cause sprites
exhibiting both short, bright features and a longer dimmer luminosity
which is not well resolved by the photometers. It is likely that this
sprite sustained some luminosity during the entire time between the
two lightning discharges. The ULF magnetic
field indicates the existence of a vertical
current flowing continuously for 140 ms during this period.
Figure 5.17 shows the unusual fact that the brightening
of the sprite in photometer 11 appears to anticipate the onset of the
second cloud-to-ground discharge.
The dashed lines superposed on the photometer traces in Figure 5.17 show curves of the form
An electric field imposed on a conductive medium by a rapid
rearrangement of charges is expected to decay exponentially in time
(see Section 1.3, page ).
Indeed, the typical timescales
for the observed decay are comparable
to expected electric relaxation times
at the
observed altitudes. However, according to
Figure 2.4 on page
the optical emissions should
not relax exponentially in such a case because of their highly
nonlinear dependence on the electric field strength.
On the other hand, the observed exponential relaxation would be obtained if we adopt the ad hoc assumption that the electric field remains constant in time. Such an assumption was first proposed in March 1999 [Victor Pasko, private communication].
For an altitude where quenching is insignificant and with the assumptions used for equation (2.18), the optical emission rate is
It may be cautioned that a wide variety of physical systems may be well approximated by exponential behavior, sometimes due to statistical or geometric reasons rather than those relating to local physics. For instance, in the case of elves the temporal structure of optical emissions is locally determined by temporal properties of the causative lightning pulse, and at a ground observer site is determined largely by geometrical considerations. These geometrical effects can lead to an apparently closely-exponential relaxation of luminosity from the ``back'' part of elves both in theory and observations for the case of a photometer with a field-of-view as large as that of P11 in the Fly's Eye.
Nevertheless, the exponential decay feature is found in a majority of the bright
sprites observed between 04:00 and 06:00 UT on 19 July 1998 and often
with a more exact fit than the cases shown in
Figure 5.17.
Figure 5.18 shows values of the relaxation time constant
determined for peaks observed in 27 events exhibiting good to excellent
closeness of fit between equation (5.1) and the data in narrow
field-of-view photometers 1 to 9 (red filter) and 10 (blue filter).
The altitudes corresponding to the narrow fields-of-view for these
observations were primarily in the range 60 km to 85 km, with
considerable uncertainty (
km) based on the possible range to
the sprites, as explained on page
(Section 4.1.4).
![]() |
The curve fitting is done by a nonlinear least squares algorithm for
periods chosen by hand to correspond well to a decaying exponential
form. In some cases the initial period following a bright peak relaxes
faster than the exponential fit, and not all of it is included.
Instead, whenever possible, the fit period includes many times the
duration of so as to appropriately fix the value of
in
equation (5.1) to the background luminosity. The quality, or
closeness, of fit is then assessed by comparing the values of
from data and fit using the linear correlation parameter
given by Bevington and Robinson [1992, p. 199].
Also shown for reference are some time constants determined with the
same algorithm and associated with optical pulses from the same storm
but which were determined to be due to elves. The apparent close fit
in these cases, however, is less significant since the parameter
is barely resolved by the sample period of the photometer.
Nevertheless, the values of
given for elves in Figure 5.18 do
give an indication of the time scales for the optical signals due to
elves viewed with a narrow field-of-view. The sample period of the data is shown by a
dashed vertical line.
It is apparent that while the instrument and the fitting method are
capable of resolving decay constants well below 100 s, and while
the measured variation in
extends over nearly two orders of magnitude
for sprites, a lower limit of
200
s exists among the
observed sprite cases.
Two more dashed vertical lines show the fastest rate constant expected
at two different altitudes for
in the region where dissociative attachment dominates over
ionization. As shown in Figure 2.4 on page
, this rate
is reached at
and is also the fastest optical
relaxation that is predicted for a constant electric field, according
to equation (5.2). Figure 5.19
reproduces some time scales previously shown in Figure 1.1
as frequencies. Included is the variation of
with
altitude. The suggestion that the observed optical relaxation rate
may be bounded by the maximum rate of
supports the ad hoc assumption of an essentially constant electric field during
these times.
![]() |
It is remarkable that a constant electric field should arise so
frequently in a dynamically driven conducting medium. One likely
scenario is the existence of a constant source current term in the
troposphere over a time long compared to the local relaxation time
. As an example, if a thundercloud vertical charge moment change of 1000
is required for
to exceed
at some altitude, then if
1 ms, a steady-state electric field of
could be sustained only by a current moment of 1000
. This value
is comparable to the peak current flowing in a powerful return stroke.
This relationship between current and electric field may be quantified
as follows.
Because of the finite atmospheric conductivity, an infinitesimal charge
moment change
makes a contribution
to
the instantaneous electric field at time
which decays with time
constant
.
That is,
![]() |
(5.4) |
Theoretical studies [e.g., Pasko et al., 1997b] suggest that sprite breakdown
occurs after the integrated charge moment change surpasses a value
needed to exceed the breakdown electric field. Judging from
Figure 5.19 this ``integration'' may occur over times
of up to 5 ms at 75 km altitude, in accordance with
equation (5.5). However, once the breakdown threshold is
reached the electron density may increase rapidly (for instance
through streamer breakdown) on timescales much faster than
and as a result the conductivity may increase
drastically and the value of
could be reduced to much less
than that shown in Figure 5.19 for the ambient
electron density. As a result, the electric field would rapidly decay
(with time scale
) to the steady state value
given in
(5.6) and thereafter the electron density and optical
emissions would decay exponentially (with time scale
) for any
case in which
. This entire sequence of events could occur
with no temporal variation in the tropospheric source term
.
With this view, the initial non-exponential decrease following an
optical peak before a closely-exponential form is observed may
correspond to the establishment of a steady-state electric field
. A complication to the interpretation of this sequence of
events when streamers are involved is the difficulty of carrying out a
theoretical
calculation of the ionization level left behind a propagating
streamer, where the electric field is expected to be quite low
[Bazelyan and Raizer, 1998]. In light of the work of Cummer and Stanley [1999], the
sequence of events described above might occur after the initial propagation of
streamers is complete and may apply to the reexcitation of the remnant
channels.
In any case, the measurement of exponential optical relaxation
constants may constitute a significant new method for remotely sensing
the local electric field within a sprite. For a given altitude
observed within a narrow field-of-view, and with the assumption that
, the observed relaxation time constant
determines
which
in turn prescribes
. In addition, according to the interpretation
presented here, this observation gives non-spectral evidence of
significant ionization changes. However, when taken alone it is
likely not useful for measuring absolute electron densities.
On the other hand, it does suggest in accordance with equation (5.3) that the free electron population
likely becomes almost completely
depleted in these regions where the electric field remains constant
(and below
) for durations several times
.
The ULF magnetic field data shown in
Figure 5.16 indicate that an essentially time invariant
continuing current in lightning is a realistic possibility,
even over many milliseconds. Recent unpublished work by Steven Cummer
and Martin Füllekrug has used such ULF data to infer the vertical
source lightning currents with a method similar to that previously
used for ELF recordings [Cummer and Inan, 2000; Cummer and Stanley, 1999; Barrington-Leigh et al., 1999a; Cummer and Inan, 1997; Cummer et al., 1998b]. The inferred vertical current
moments were as high as 40
for 160 ms and may account for
sprite breakdown in long-delayed sprites even without appealing to
unmeasured horizontal charge motion, an idea
which was invoked to explain previous ELF/VLF results.
On the other hand, the occurrence of a sprite just preceding the
return stroke in Figure 5.17 suggests that a large
(horizontal) charge motion within the cloud may have both led to a
sprite and been involved in the initiation of the return stroke,
indicating that horizontal currents may indeed be sufficient to
initiate sprites without vertical (return stroke) charge motion.
According to NLDN, the second return stroke was in a new location from
the previous one. The time scale for stepped leader breakdown is
10 ms altogether, and
1 ms for the final leader pulse
preceding the return stroke [Uman, 1987, p. 14-16], implying that
the channel taken by the second return stroke in Figure 5.16
must have been developing well before the onset of the bright optical
signature. An alternative interpretation for the peculiar observation
of a sprite immediately preceding the second return stroke is that the
timing of the sprite onset with respect to the return stroke is
coincidental. However, given the 31 km proximity of the two lightning
strokes as reported by NLDN, it is likely that they were closely coupled
electrically through intracloud currents.
For reasons discussed previously (Section 3.1), it is difficult to determine experimentally the relative contributions of vertical and horizontal charge motions in the production of mesospheric electric fields. The results reported above give evidence of sustained source currents of one kind or another without discriminating between them.
In the opposite extreme of velocity resolution, normal speed video at
the high spatial resolution of a telescopic imager may be used to
determine how slowly optical structures may propagate. With its 17 ms
resolution and 0.7 vertical field-of-view corresponding to
240 pixels, the telescopic imager described by Gerken et al. [2000] may
in principal resolve vertical motion as slow as
m/s and
as fast as
10
m/s at a range of 500 km. Not surprisingly,
optical structures in sprites are regularly seen which appear
completely stationary on this time scale [Gerken et al., 2000]. Here we
report those which appear to be coherent and in motion.
![]() |
Figure a shows a portion of an integrated series
of video fields from the telescopic imager. Several events from 19
July 1998 such as this one were analyzed by tracking the motion of
bright features. The measured velocities are shown in
Figure
b as a function of time after the first
appearance of luminosity in each event. In a number of events,
several features were tracked and are plotted with the same color.
The velocities are primarily in the range of 10
to
m/s, reflecting the resolution of the instrument, and in
a number of cases luminous regions propagate steadily for 100 ms or
longer.
A common approximation made in the theoretical modeling of streamers
is to assume a streamer propagation velocity high compared with the
mean electron drift velocity
[e.g., Dhali and Williams, 1985].
Raizer et al. [1998] conclude that streamers should not propagate when
the streamer velocity approaches
, and predict a lower velocity
limit of
m/s. Our observations show propagation of some
form, presumably in the presence of steady electric fields over tens
of milliseconds, which differs from these predictions by nearly two
orders of magnitude. These features, as well as static luminous
regions, require further study.