Lightning initiation in the anvils of two supercell storms
K. M. Kuhlman,
1,2
D. R. MacGorman,
1
M. I. Biggerstaff,
2
and P. R. Krehbiel
3
Received 13 November 2008; revised 9 January 2009; accepted 4 February 2009; published 1 April 2009.
[1] Previous studies of lightning in anvil clouds have
reported that flashes began in or near the storm core and
propagated downwind into the anvil. It had been thought
that flashes could not be initiated far downwind in the anvil,
because anvil charge was thought to be produced mainly
in the storm’s deep updraft and to decrease with distance
into the anvil. Here we report observations of the in-cloud
development of lightning flashes in the anvils of two
supercell storms, including the first observations of flashes
that began in the anvil 30–100 km from the cores of
the storms and propagated upwind back toward the cores.
Interaction between charge regions in the two converging
anvils of adjoining storms appeared to cause some of the
distant flash initiations, but a local charging mechanism in
the anvil likely also contributed to the flash initiations.
All flashes that struck ground beneath the distant anvil
transferred negative charge to ground instead of the positive
charge usually transferred to ground there, an apparent
consequence of the parent storm having an inverted-polarity
electrical structure. Citation: Kuhlman, K. M., D. R.
MacGorman, M. I. Biggerstaff, and P. R. Krehbiel (2009),
Lightning initiation in the anvils of two supercell storms,
Geophys. Res. Lett., 36, L07802, doi:10.1029/2008GL036650.
1. Introduction
[2] Substantial advances have been made in our under-
standing of lightning in deep convective regions and in the
widespread stratiform precipitation that trails large storm
systems [e.g., MacGorman and Rust, 1998; Lang et al.,
2004; Bruning et al., 2007; MacGorman et al., 2008].
However, lightning in anvil clouds has received far less
attention. Though large electric fields are observed in
anvils [e.g., Marshall et al., 1989], charge in the anvil
has been viewed as originating primarily in deep convec-
tion and flowing passively into the anvil, with some
contributions from environmental currents to anvil cloud
boundaries and complications from lightning [e.g., Byrne
et al., 1989; Dye et al., 2007]. As will be discussed,
however, this paper and Dye and Willett [2007] infer from
new observations that significant charging probably also
occurs in the anvil, an inference likely having important
implications for our understanding of anvil kinematics and
microphysics.
[3] Most previous studies of anvil lightning analyzed
only cloud-to-ground (CG) lightning channels below
clouds, as observations of in-cloud lightning structure have
been scarce. CG studies found that anvil flashes striking
ground near the storm core can lower either positive charge
(+CG flashes) or negative charge (
CG flashes) to ground,
but those striking ground >30 km from the core are almost
always +CG flashes [Rust et al., 1981].
[4] The scarcity of in-cloud lightning observations in
anvils has been overcome recently by the advent of auto-
mated systems for mapping very high frequency (VHF)
electromagnetic radiation sources [Thomas et al., 2004]. A
few subsequent studies [e.g., Wiens et al., 2005; Dye and
Willett, 2007; Tessendorf et al., 2007] have reported the
in-cloud development and structure of anvil lightning. These
flashes all began in or near the storm core and propagated into
the anvil, but the number of storms analyzed by these studies
is small.
[5] Here we report the lightning and electrical structure
observed in the anvils of two supercell storms on 29–30
May 2004. These observations provide the first documen-
tation of flashes beginning in an anvil several tens of
kilometers from the storm core, instead of beginning in or
near the storm core and propagating into the anvil. Besides
having implications concerning anvil processes, these
observations provide insight into situations under which
distant anvil CG flashes are
CG flashes, instead of the
+CG flashes usually observed there.
2. Observations From 29–30 May 2004
[6] The Oklahoma Lightning Mapping Array (OKLMA)
provided the in-cloud lightning data we have analyzed.
Several authors [e.g., Thomas et al., 2004; MacGorman et
al., 2008] describe this system, its mapping accuracy, and
the lightning analysis techniques we use to infer the polarity
of storm charge. The OKLMA tends to map much smaller
densities of VHF sources in negative regions than in positive
regions, because channels propagating into negative charge
typically produce much weaker VHF signals than produced
by channels propagating into positive charge [e.g., Rison et
al., 1999]. When one interprets relative densities, the small
densities typically observed in negative regions and the much
larger densities observed in positive regions should be given
roughly the same weight. Cloud-to-ground strike data from
the National Lightning Detection Network [Cummins et al.,
1998] was used to identify which flashes struck ground and to
determine their polarity.
[7] Storm winds and precipitation were measured by two
mobile, C-band Doppler radars [Biggerstaff et al., 2005],
which were 30–40 km apart and took 2–3 minutes to
complete a synchronized sector volume scan of the storm.
Radar data were interpolated to a grid having 1-km horizontal
spacing and 0.5-km vertical spacing by using a modified
GEOPHYSICAL RESEARCH LETTERS, VOL. 36, L07802, doi:10.1029/2008GL036650, 2009
Click
Here
for
Full
Article
1
Cooperative Institute for Mesoscale Meteorological Studies, University
of Oklahoma and NOAA National Severe Storms Laboratory, Norman,
Oklahoma, USA.
2
School of Meteorology, University of Oklahoma, Norman, Oklahoma,
USA.
3
New Mexico Institute of Mining and Technology, Socorro, New
Mexico, USA.
Copyright 2009 by the American Geophysical Union.
0094-8276/09/2008GL036650$05.00
L07802 1of5

Barnes weighting scheme [Trapp and Doswell, 2000] in the
REORDER software package [Oye et al., 1995]. The
CEDRIC package [Miller and Fredrick, 1998] was used to
complete a dual-Doppler wind synthesis.
[8] The two storms we analyzed produced large hail
and extremely large flash rates (typically >200 min

1
,up
to roughly 500 min

1
) as they moved across Oklahoma
on 29–30 May 2004. Most lightning occurred in the deep
convective cores of the storms, but many flashes occurred in
the large anvils. Both storms strengthened rapidly as they
progressed into an environment of increasing moisture in the
boundary layer and substantial (>4000 J kg

1
) convective
available potential energy. The northern storm was initiated
by a dryline near the Oklahoma-Texas border around
2030 UTC and was already a supercell storm when our
analysis period started. The southern storm began around
2120 UTC, but overtook the northern storm and produced
multiple tornadoes. Our analysis began at 2255 UTC, when
both storms were in the region of 3D lightning mapping
coverage and the anvil of the southern storm was starting
to overlap the anvil of the northern storm. Before 2255 UTC,
all anvil lightning was associated solely with the northern
storm.Most anvil flashes began in or near themain core of the
northern storm and propagated into the anvil, as seen in
previous studies [e.g.,Marshall et al., 1989; Tessendorf et al.,
2007], but some flashes were initiated in the anvil of the
northern storm >30 km from the storm core. Distances from
the core are from the eastward boundary of 30 dBZ reflec-
tivity at 6–8 km mean sea level (MSL).
[9] Between 2255 and 2325 UTC, lightning continued to
be initiated in the distant anvil of the northern storm, but
some of these flashes also began to involve part of the
southern storm’s anvil. Lightning initiations migrated south
from a region well inside the northern anvil to a region in
which the northern and southern anvils merged and over-
lapped (e.g., Figure 1). Of the ten flashes initiated >30 km
from the core during this period, five were
CG flashes and
five did not strike ground. The flashes propagated back
toward the storm core through a midlevel layer of positive
charge in each anvil, in reflectivity >20 dBZ (Figure 1).
Channels propagating through the upper negative charge
(the polarity of charge lowered to ground) were in the
northern anvil only. Channels from flashes initiated in the
distant anvil approached, but did not overlap flashes prop-
agating out from the core.
[10] The flash in Figure 1 was the last to be initiated in the
northern anvil >30 km from the storm core. The estimated
distribution of the charge involved in all anvil lightning
from 2320 to 2330 UTC (which included Figure 1’s flash) is
shown in Figures 2a, 2b, and 3a. During this period, the
southern storm was just starting to become a supercell
storm. By 2330 UTC, supercell characteristics, including a
rotating updraft, were evident in radar data. The northern
storm was weakening rapidly from its supercell stage,
because the southern storm had started interfering with
low-level inflow into the northern storm’s updraft.
[11] The polarity of electrical structure inferred from
lightning in the convective cores of both storms 2320–
2330 UTC (an uppermost negative charge in a vertical
tripolar or dipolar arrangement) (Figures 2a, 2b, and 3a)
could be described as inverted from the usual polarity [e.g.,
Rust et al., 2005]. This structure clearly extended into the
anvil of the northern storm. Midlevel positive charge also
extended into the anvil of the southern storm, but little or no
lightning involved upper negative charge in the southern
anvil. As seen most clearly in the side-by-side distributions
for the anvils shown in Figure 3a, the upper part of the
positive charge in the anvil of the southern storm was at the
same altitude as, and adjacent to, the upper negative charge
in the anvil of the northern storm. (Ingesting colder, drier air
Figure 1. Mapped VHF sources at all altitudes superimposed on reflectivity and synthesized horizontal winds (vectors)
from the SMART-R. Yellow squares mark flash initiation; blue circle, connection to ground. (a-c) LMA sources from a
single flash at 2321:45.2–46.9 UTC, SMART-R volume scan beginning at 2321 UTC. (a) Vertical profile along line A,
includes LMA points north of line A. (b) Vertical profile along line B, includes LMA points south of line A. (c) Z = 8.3 km
AGL. (d-e) LMA sources for the 2-minute period beginning 0030:30 UTC, SMART-R volume scan beginning at
0038 UTC. (d) All mapped LMA, initiation, and -CG locations in bottom panel superimposed on vertical cross section of
reflectivity along line C. (e) Z = 7.8 km AGL.
L07802 KUHLMAN ET AL.: LIGHTNING INITIATION IN SUPERCELL ANVILS L07802
2of5