convert

An Observational Examination of Long-Lived Supercells.

Part I: Characteristics, Evolution, and Demise

ATTHEW

J. B

UNKERS

NOAA/National Weather Service, Rapid City, South Dakota

ARK

R. H

JELMFELT AND

AUL

L. S

MITH

South Dakota School of Mines and Technology, Rapid City, South Dakota

(Manuscript received 5 July 2005, in final form 17 December 2005)

ABSTRACT

Observations of supercells and their longevity across the central and eastern United States are examined,

with the primary focus on understanding the properties of long-lived supercells (defined as supercells lasting

ⱖ

4 h). A total of 224 long-lived supercells, occurring in 184 separate events, are investigated. These

properties are compared with those of short-lived supercells (lifetimes

ⱕ

2 h) to determine the salient

differences between the two classifications. A key finding is that long-lived supercells are considerably more
isolated and discrete than short-lived supercells; as a result, the demise of a long-lived supercell (i.e., the end
of the supercell phase) is often signaled by a weakening of the storm’s circulation and/or a rapid dissipation
of the thunderstorm. In contrast, short-lived supercells commonly experience a demise linked to storm
mergers and convective transitions (e.g., evolution to a bow echo). Also noteworthy, 36% of the long-lived
supercell events were associated with strong or violent tornadoes (F2–F5), compared with only 8% for the
short-lived supercell events. Evolutionary characteristics of long-lived supercells vary geographically across
the United States, with the largest contrasts between the north-central United States and the Southeast. For
example, 86% of the long-lived supercells across the north-central United States were isolated for most of
their lifetime, whereas only 35% of those in the Southeast displayed this characteristic. Not surprisingly, the
convective mode was discrete for 70% of the long-lived supercell events across the north-central United
States, compared with 39% for the Southeast.

1. Introduction

a. Importance of long-lived supercells

Supercell thunderstorms, which were originally de-

fined by Browning (1962, 1964), represent the most or-
ganized, most severe, and longest-lived form of iso-
lated, deep moist convection. Their updraft cores can
be largely undiluted (i.e., conserved equivalent poten-
tial temperature from cloud base), with vertical veloc-
ities approaching 50 m s

⫺

in the strongest storms (e.g.,

Musil et al. 1986; see their Fig. 9). Exemplifying their
severity, the 5 May 1995 supercell that struck the Dal-
las–Fort Worth, Texas, area was the single costliest
nontornadic thunderstorm in U.S. history, producing
large hail, damaging winds, and flash flooding with eco-

nomiclosses near $2 billion (NOAA 1995; Calianese et
al. 2002). Moreover, some of the longest-lived super-
cells can produce an extensive swath of severe weather,
and on occasion they generate a devastating combina-
tion of large hail and damaging winds (e.g., Nelson
1987; Klimowski et al. 1998; Glass and Britt 2002). It is
also noteworthy that long-lived supercells

have been

Corresponding author address:

Dr. Matthew J. Bunkers, Na-

tional Weather Service, 300 E. Signal Dr., Rapid City, SD 57701.
E-mail: matthew.bunkers@noaa.gov

Long-lived supercells are hereafter arbitrarily defined as

supercells

that persist in a quasi-steady manner for

ⱖ

4 h, moder-

ate-lived supercells are defined as supercells whose lifetime is
between 2 and 4 h, and short-lived supercells are defined as su-
percells with lifetimes

ⱕ

2 h. The goal of the present research is to

better understand the supercell longevity spectrum, and these
definitions were chosen with this in mind. Note that it is possible
to have a

long-track

supercell by virtue of its speed, yet it might

not meet the above definition of long lived. A temporal definition
of supercell longevity was chosen because time is inherent to the
definition of a supercell, whereas the distance a supercell travels
is not germane to its definition.

OLUME

W E A T H E R A N D F O R E C A S T I N G

CTOBER

2006

673

WAF949

observed during many tornado outbreaks in the United
States, including the 3 May 1999 outbreak in central
Oklahoma (Thompson and Edwards 2000).

Not surprisingly, part of the supercell definition in-

cludes persistence, namely, that the storm’s midlevel
circulation (i.e., mesocyclone or mesoanticyclone) per-
sists at least on the order of tens of minutes (Moller et
al. 1994), representing the upper bound of the time it
takes for a single thermal to be processed by an updraft
(Doswell 2001). Operational experience suggests the
typical lifetime of a supercell is 1–2 h, which is consis-
tent with Burgess et al. (1982), who found the average
lifetime of single-core mesocyclones to be 90 min. In
some cases, a supercell may persist in a quasi-steady
manner for several hours, and in even fewer cases, the
lifetime of a supercell may extend beyond 4–5 h (e.g.,
Paul 1973; Browning and Foote 1976; Warner 1976).

b. Motivation for this study

Considerable research has been directed toward the

understanding of supercell dynamics since the 1940s
(Doswell 2001), and much of this effort has focused on
tornadic versus nontornadic supercells (e.g., Johns et al.
1993; Brooks et al. 1994a,b; Rasmussen and Blanchard
1998; Markowski et al. 2003; Thompson et al. 2003;
among many others). Conversely, comparatively little
has been done to elucidate supercell longevity and, ul-
timately, supercell demise.

Most of what has been

done to study long-lived supercells involves case studies
(e.g., Browning and Foote 1976; Glass and Britt 2002).
Indeed, very little information exists on the general
properties of long-lived supercells, and a formal defini-
tion of long-lived supercells does not exist (i.e., only
nebulous definitions appear in the literature).

There are several reasons why it would be advanta-

geous to anticipate long-lived supercell occurrence in
an operational setting. First, the duration of severe
weather warnings sometimes might be lengthened, be-
yond what is commonly used, if an operational fore-
caster could correctly anticipate long-lived supercells
(e.g., a 60-min warning versus a 45-min warning). One
potential benefit of this is increased lead time for tor-
nado and severe thunderstorm warnings. Second, lon-
gevity predictions would be useful when generating
short-term forecasts for locations (or counties) down-

stream of an existing supercell. Third, a forecast of su-
percell longevity may be a beneficial factor in making
operational staffing decisions—both for the number of
personnel needed and their duration of duty. Fourth,
information about anticipated supercell longevity may
assist operational forecasters when determining wheth-
er to issue or extend severe weather watches, which
typically have a valid time of at least 4 h (corresponding
to the long-lived supercell definition). Finally, knowl-
edge of thunderstorm demise mechanisms, which di-
rectly affect thunderstorm longevity, may be especially
beneficial in an operational setting where several su-
percells are present, thereby enabling warning forecast-
ers to effectively subdivide warning operations.

Considering the preceding discussion, the general

characteristics, evolution, and demise of long-lived su-
percells across the United States are documented in the
current study. Short-lived supercells are also examined
to determine how their properties differ from their
longer-lived counterparts. After defining supercells, su-
percell events, and data collection methods (section 2),
the results are presented for the long- versus short-lived
supercell events, as well as for various geographical re-
gions of the United States (section 3). In section 4 the
results are summarized, with a focus on operational
forecasting, although a more detailed analysis of long-
lived supercell environments is presented in Bunkers et
al. (2006, hereafter Part II).

2. Data and methods

a. Defining and identifying supercells and supercell

events

The feature that distinguishes supercells from other

thunderstorms is their persistent, rotating updraft in the
midlevels (approximately 2–8 km AGL) of the storm
(Browning 1964; Weisman and Klemp 1984; Doswell
and Burgess 1993; Moller et al. 1994). Although a mul-
ticellular structure can be imposed upon a supercell
(e.g., Foote and Wade 1982; Foote and Frank 1983;
Nelson 1987; Doswell et al. 1990), a quasi-steady rotat-
ing updraft in the midlevels of a thunderstorm, lasting
at least on the order of tens of minutes, defines it as a
supercell. It is not always easy to classify a thunder-
storm as a supercell because of the subjectivity in de-
ciding how “strong and steady” the rotating updraft
needs to be (e.g., Browning 1977; Moller et al. 1994);
however, supercells become relatively easy to identify
given ample operational experience.

Supercells in the present study were identified con-

sistent with the procedures in Thompson (1998) and
Klimowski et al. (2003). A combination of radar reflec-

The word “demise” is used throughout this study to indicate

the end of the supercell phase (i.e., the end of the midlevel cir-
culation), and not the end of the thunderstorm. Accordingly, a
supercell can experience its demise, but via merger and/or evolu-
tion to another form (e.g., a bow echo) the remaining nonsuper-
cellular thunderstorm may persist well beyond this time.

674

W E A T H E R A N D F O R E C A S T I N G

OLUME

tivity characteristics was sufficient to identify most su-
percells, especially during the middle portions of their
lifetimes. The characteristics included hook echoes,
midlevel overhang or bounded weak echo regions
(BWERs), strong reflectivity gradients on the inflow
side of the storm, quasi steadiness of storm reflectivity,
and deviant motion (Marwitz 1972; Browning 1977;
Lemon 1980; Bunkers et al. 2000). These were chosen
because they can be used to infer updraft rotation and
strength–persistence, which can be particularly helpful
when velocity data are missing (e.g., at relatively far
ranges from the Doppler radar). Furthermore, motion
transverse to the vertical wind shear is a fundamental
property of supercells (Bunkers et al. 2000; Zeitler and
Bunkers 2005), which is a direct result of the contribu-
tion of updraft rotation to updraft strength (Weisman
and Rotunno 2000; see their Fig. 14). Therefore, by
definition storm motion must change when an ordinary
cell becomes a supercell (and vice versa), at times pro-
ducing a “reverse S shaped” storm track (Fankhauser
1971; see his Fig. 3); this can be an invaluable tool when
identifying supercells.

In cases where supercell identification was not clear

based on the reflectivity signatures (e.g., when storms
were not isolated), storm-relative velocity data were
consulted to verify the existence of mesocyclonic or
mesoanticyclonic rotation within the storm. Moreover,
storm-relative velocity data were used, when available,
during the beginning (ending) stages of a supercell’s
lifetime to help determine the onset (cessation) of ro-
tation (and also demise); a minimum azimuthal velocity
difference of 20 m s

⫺

across a distance of less than

10 km was required (similar to Thompson et al. 2003).
If a supercell weakened (i.e., the parent thunderstorm
lost its supercell characteristics), and then the same
thunderstorm subsequently reacquired supercell status,
it was treated as two separate supercells.

As noted in the introduction, a supercell was consid-

ered long lived if its individual lifetime, as a supercell,
was

ⱖ

4 h. By way of distinction, a long-lived supercell

event was defined as any convective event

in which at

least one long-lived supercell was observed (this in-
cluded both right and left movers). Accordingly, this
means not all supercells in a long-lived event had to
persist for

ⱖ

4 h. Thus, there were one or more long-

lived supercells per event (the average is 1.2; see section

3b below)—along with some shorter-lived supercells at
times. When more than one long-lived supercell event
occurred on the same day, the events had a separation
of at least 185 km (100 n mi), and most often they were
in different states with few intervening thunderstorms.
Of the 184 long-lived supercell events obtained for the
current study, there were two (three) long-lived events
per day on 23 (5) separate occasions, and four long-
lived events occurred on only one day.

A slightly different approach was taken for defining

short-lived supercell events. These were determined by
averaging the lifetimes of the supercells for a given con-
vective event, and if the average supercell lifetime was

ⱕ

2 h (consistent with the definition of a short-lived

supercell), then the event was considered to be short
lived. Note that some short-lived supercell events in-
cluded individual supercells with lifetimes

⬎

2 h, but not

ⱖ

4 h (in which case it would have been a long-lived

event), provided the average supercell lifetime did not
exceed 2 h. In addition, a number of short-lived super-
cell events lasted well over 2 h, but again, the average
lifetime of the individual supercells per event was

ⱕ

2 h.

This approach was chosen so as not to bias the moder-
ate-lived supercell events (defined below) with several
short-lived supercells. Of the 119 short-lived supercell
events obtained for the current study, there were two
(three) short-lived events per day on 10 (3) separate
occasions.

On days when a large number of supercells occurred

across a fairly broad region, it was possible to have a
combination of long- and short-lived supercell events.
For example, the northern end of a series of convective
clusters may have consisted of one or more long-lived
supercells, thus classifying it as a long-lived supercell
event. Farther south, presumably in a different environ-
ment, a short-lived supercell event may have occurred.
Accordingly, there were 21 days in which both long-
and short-lived supercell events occurred. The average
distance between the nearest supercells in the long-
and short-lived events in these situations was 740 km
(400 n mi), with only three events as close together as
185–278 km (100–150 n mi; refer to footnote 3).

As is typical with such classification-type studies, cer-

tain convective events containing supercells were diffi-
cult to place in either of the two above categories. For
example, there were convective events with supercells
displaying average lifetimes between 2 and 4 h, but in
which no long-lived supercells occurred. These were
subsequently classified as moderate-lived supercell
events, with a total of 137 such events acquired. This
portion of the database will be used in Part II of this
study and is not discussed further here.

Based on operational experience, a convective event is de-

fined as an area of thunderstorms that is separated from any other
area of thunderstorms by at least 185 km (100 n mi) at its edges.
If areas of thunderstorms were closer together than this, they were
considered part of the same event. The duration of the parent
mesoscale convective system defined the temporal constraint.

CTOBER

2006

B U N K E R S E T A L .

675

b. Supercell data sources and radar data

Several sources were used to gather supercell data.

Initially, a literature search produced 17 long-lived su-
percell events (Hamilton 1958; Fujita and Grandoso
1968; Fujita et al. 1970; Paul 1973; Browning and Foote
1976; Warner 1976; Fujita 1974; Wade 1982; Curran and
Rust 1992; MacGorman and Burgess 1994; NOAA
1994; Pence and Peters 2000; Rasmussen et al. 2000;
Thompson and Edwards 2000; Doswell et al. 2002;
Knupp et al. 2003). Clear graphical or written evidence
of supercells that persisted for

ⱖ

4 h was required in

these documents. For example, the Lahoma, Oklaho-
ma, supercell (Conway et al. 1996) was only considered
a moderate-lived event (a conservative estimate) be-
cause it was unclear whether or not the supercell per-
sisted for

ⱖ

4 h

before

it evolved into a bow echo. If

there was any doubt whether or not a supercell’s life-
time was

ⱖ

4 h, the case was not considered long lived,

even though the supercell may have become a bow
echo with a total

thunderstorm

lifetime (supercell plus

bow echo) much greater than 4 h.

Bunkers (2002, plus references therein) and Kli-

mowski et al. (2003) provided a significant source of
supercell information. These data were reviewed for
both long- and short-lived supercell events. To augment
these data, and especially to obtain commensurate
coverage for the remainder of the United States, radar
data were also collected for supercell events observed
across the United States from 2001 to 2004. Last, the
National Oceanic and Atmospheric Administration
(NOAA)

Storm Data

resource was interrogated—using

the software of Hart and Janish (2005)—to graphically
examine spatial and temporal characteristics of severe
storm reports from 1996 to 2000 across the United
States. If there was a temporally consistent and quasi-
systematic progression of severe storm reports on a
given day (indicating the possibility of a long-lived su-
percell; e.g., Fig. 1), the relevant radar data were ac-
quired. These data were subsequently examined for the
presence of long-lived supercells, and some short- and
moderate-lived supercells were also discovered via this
process.

Radar data consisted of a myriad of Weather Sur-

veillance Radar-1988 Doppler (WSR-88D) sources.

1) When available, WSR-88D level IV data (Crum et

al. 1993) from the National Weather Service (NWS)
Advanced Weather Interactive Processing System
were used. These data were confined mostly to the
northern Great Plains.

2) In cases where level IV data were not easily acces-

sible, the NWS’s Next Generation Weather Radar
(NEXRAD) Information Dissemination Service

. 1.

Storm Data

reports for the 24-h periods beginning

at 1200 UTC on (a) 24 Jul 2000, (b) 1 May 2002, and (c) 5 Apr
2003. Hail reports are indicated with gray circles, wind reports
are indicated with dark plus signs, and tornado reports are indi-
cated with black dots or lines. Arrows point to the beginning–
ending locations of the long-lived supercells responsible for the
storm reports. All images are at the same scale. In (a) 34%
(100%) of the total (significant) severe weather reports across
the United States originated from this one long-lived supercell,
in (b) 60% (91%) of the total (significant) severe weather reports
across the United States originated from these two long-
lived supercells, and in (c) 77% (100%) of the total (signifi-
cant) severe weather reports across the United States origi-
nated from these three long-lived supercells (two of the tracks
overlap).

676

W E A T H E R A N D F O R E C A S T I N G

OLUME

(NIDS) data were acquired from the Cooperative
Program for Operational Meteorology, Education,
and Training (COMET) archive. These NIDS data
consisted of a 0.5°-elevation mosaic, derived from
the WSR-88D sites across the continental United
States, and possessed a temporal (spatial) resolution
of at least 10 min (2 km).

3) If the NIDS data were not available from COMET,

the following Internet resources were examined for
archived WSR-88D data:
(a) http://www.spc.noaa.gov/exper/archive/events/
(b) http://locust.mmm.ucar.edu/case-selection/
(c) http://www.rap.ucar.edu/weather/radar/

4) Last, WSR-88D archive level II data (Crum et al.

1993) were obtained from the National Climatic
Data Center for the remaining events in which the
above radar data could not be obtained.

For most of the events, radar data were available

from at least two of the above sources. Using these data
and procedures, supercell lifetimes were approximated
to the nearest 15 min.

c. Properties recorded for the supercell events

Properties recorded for the supercell events were

based on operational experience, and information col-
lected for each of the long-lived supercell events was
slightly more detailed than what was collected for the
short-lived supercell events (Table 1). For example, the
beginning time (when the storm became a supercell),
ending time (when the storm ceased being a supercell),
and reflectivity-based centroid track of each long-lived
supercell were recorded, but only the average supercell
lifetime and general location were needed for the short-
lived supercell events. Furthermore, the occurrence of
supercell cycling (defined below) was noted only for the
long-lived supercells, and the observed storm motion
was also calculated over a longer duration for the long-
lived supercells (versus the short-lived supercells; see
Table 1).

Tornado and severe storm reports for the supercell

events were obtained from

Storm Data

as described in

section 2b (see above). There are several deficiencies
within the

Storm Data

database (Doswell and Burgess

1988; Weiss et al. 2002; Brooks et al. 2003), including
(i) an underreporting of weak tornadoes (F0 and F1),
(ii) an increase in the number of “low end” severe re-
ports [i.e., hail diameter 1.9–2.5 cm (0.75–1.0 in.)
and/or measured or estimated wind gusts 26–27 m s

⫺

(58–60 mi h

⫺

)] in the last few decades of the 1900s, and

(iii) an overall increase in the total number of severe
reports since the mid-1900s. However, the reporting
of strong and violent tornadoes (F2–F5; also referred to
as significant tornadoes) has been more consistent
throughout the period of record. As a result, the main
items of interest from

Storm Data

are (i) whether F2–

F5 tornadoes were reported (versus F0 and F1 torna-
does), (ii) whether any severe hail or wind was reported
(as opposed to the total number of reports), and (iii)
whether significant severe hail or wind was reported
[i.e., hail diameter

ⱖ

5.1 cm (2.0 in.) and/or measured or

estimated wind gusts

ⱖ

33.5 m s

⫺

(75 mi h

⫺

); Hales

(1988)]. Other more scarce information about storm
reports was derived from the meteorological litera-
ture—mostly from the 1940s to 1980s (e.g., as listed in
Wade 1982).

The convective mode of each long- and short-lived

supercell event was classified into one of three catego-
ries: linear, discrete, or mixed. These partitions are con-
sistent with Dial and Racy (2004). In the linear mode,
supercells were embedded in, or attached to, one or
more line segments (i.e., length to width ratio

ⱖ

5 to 1),

each contained within a common 35-dB

echo. These

were usually manifest in the form of one large squall
line, or else a few smaller bowing or linear segments,
with

⬍

10% discrete convection. Taken by itself, the

squall line in the northwestern half of Fig. 2 would be
considered a linear convective mode. On the other
hand, the discrete convective mode consisted of sepa-
rate identifiable cells that were distinct from one an-

ABLE

1. Properties recorded for the long- and short-lived supercell events. Refer to section 2c for definitions of these properties.

Long-lived supercell events

Short-lived supercell events

Beginning–ending times and location (including duration and track)

Avg lifetime and location of supercells (per event)

No. of long-lived supercells

—

Obs motion (2-h average) during most steady part of motion

Obs motion (1-h average) during most steady part of motion

Occurrence of tornadoes and other severe weather

Convective mode

Isolated vs nonisolated

Occurrence of storm mergers

Occurrence of supercell cycling

—

Storm demise mechanism

CTOBER

2006

B U N K E R S E T A L .

677

other (e.g., the cells in the southeastern half of Fig. 2).
In some cases the discrete mode consisted of supercells
occurring along a common line, but the 45-dB

echo of

the storms did not overlap, and the cells were clearly
identifiable. If a supercell event could not be classified
into the linear or discrete convective mode categories,
it was considered to be mixed mode. In addition, if
discrete supercells occurred ahead of, or close to, a line
of thunderstorms, the event was also classified as mixed
mode (e.g., Fig. 2). Last, if an event began as discrete
supercells (which was common), but then transitioned
into a linear mode about halfway through the event, the
convective event was also classified as mixed mode.

A supercell was considered isolated if it was sepa-

rated from nearby cells by at least one storm diameter
for

⬎

75% of its lifetime. This diameter was taken as the

average diameter of the 35-dB

echo. For example, the

discrete supercells ahead of the line in Fig. 2 (located in
central Tennessee) were not considered isolated at that
point in time, although the lone storm in southeastern
Tennessee was classified as isolated. The primary goal
of this classification was to identify supercells whose
inflow source regions presumably were not cut off, or
affected, by nearby thunderstorms. By way of compari-
son, Bluestein and Parker (1993) used 100 km (54 n mi)
as a threshold separation for defining isolated storms
related to severe convective development along the
dryline (note the scale on the bottom of Fig. 2). They
also defined “companion storms” as cells that stayed

“close” to another storm for at least 20 min, but did not
merge with it. Beyond this, quantitative definitions of
“isolated” in the meteorological literature are generally
lacking, with the phrase “relatively isolated” commonly
used. As will be seen in section 3b, long-lived supercells
can be isolated but still experience mergers, especially
when the mergers are nondestructive and relatively
brief.

Given the importance of storm mergers in modulat-

ing convective evolution, this information was noted for
both the long- and short-lived supercell events. These
mergers can be either destructive or reinforcing (e.g.,
Lindsey and Bunkers 2005). To eliminate the smallest
“feeder” cells (Browning 1977) from consideration, a
merger between a supercell and another storm (i.e., the
blending of the storm cores) was only documented if
the reflectivity of the second storm was

⬎

35 dB

Supercells often go through a time evolution involv-

ing a cyclic process (Burgess et al. 1982; Foote and
Frank 1983; Doswell et al. 1990; Adlerman et al. 1999;
Dowell and Bluestein 2002). In the present study, su-
percells that followed a cyclic mesocyclogenesis (or me-
soanticyclogenesis) process, such as described in Bur-
gess et al. (1982), were treated as one supercell. How-
ever, if a supercell developed separate from an existing
supercell, even though in close proximity, it was treated
as a new supercell. For example, if the outflow of a
supercell surged ahead of the storm and initiated a new
storm (i.e., a separate area of reflectivity on radar) that

. 2. The 0.5°-elevation mosaic reflectivity image centered on TN at 0000 UTC 11 Nov

2002, illustrating a mixed convective mode. The references to isolated, nonisolated, discrete,
and linear storms are discussed in section 2b.

678

W E A T H E R A N D F O R E C A S T I N G

OLUME

subsequently developed a persistent rotating updraft,
then two separate supercells were identified. This pro-
cess is akin to discrete propagation due to daughter
cells (Browning 1977) along a storm’s gust front, and
can result in a multicell–supercell hybrid as described in
Weaver and Nelson (1982) and Foote and Frank
(1983). This information on supercell cycling was re-
corded only for the long-lived supercells.

An important component of supercell longevity is the

way in which a supercell reaches its demise. Based on a
literature search and operational experience, the most
common mechanisms responsible for the

end of the su-

percell phase

are (i) the supercell travels—spatially or

temporally—into an environment that is too stable to
support deep moist convection (e.g., Maddox et al.
1980; Doswell et al. 2002); (ii) the supercell travels into
a thermodynamically and/or kinematically different en-
vironment that favors a transition into another convec-
tive mode, such as a bow echo—perhaps related to an
enhanced surface gust front (e.g., Johns and Leftwich
1988; Brooks et al. 1993; Conway et al. 1996); (iii) the
supercell merges or interacts with other thunderstorms,
which can either destroy its circulation or cause it to
evolve into another convective mode (e.g., Calianese et
al. 2002; Knupp et al. 2003); (iv) the supercell has its
supply of buoyant and moist inflow cut off by other
storms or by its own downdraft (e.g., Grasso 2000;
Doswell et al. 2002); and (v) the supercell interacts de-
structively with gravity waves, an occurrence that may
be quite rare (e.g., Barnes and Nelson 1978; Warner
1976). Although this list is not exhaustive, it is believed
to represent most causes of supercell demise. There-
fore, to assist in the understanding of supercell longev-
ity, an effort was made to document the basis for de-
mise of both long- and short-lived supercells.

3. Results

a. Distribution of long-lived supercells

We emphasize that the present study does not at-

tempt to depict a precise geographical climatology of
long-lived supercells. As noted in section 2b, supercell
data were obtained from a wide variety of sources, and
were not systematically gathered so as to faithfully rep-
resent each region of the United States over a continu-
ous time interval. For example, there is a bias of long-
lived supercell tracks across the northern high plains
and Oklahoma (Fig. 3), which is due, in part, to the
availability of data from these areas.

Keeping the above caveat in mind, it is nevertheless

apparent that most long-lived supercells occur across
the central and eastern United States (Fig. 3). By way
of contrast, only one long-lived supercell (which barely

reached the 4-h definition) was documented across the
western United States—along the Snake River valley of
southeastern Idaho. The Teton–Yellowstone F4 tor-
nadic thunderstorm (Fujita 1989) may have been an-
other long-lived supercell in the West, but insufficient
evidence precluded classification of the parent thunder-
storm. Other than these cases, Tolleson (1996) pre-
sented a case of a moderate-lived supercell (

⬃

3 h) over

the northwestern United States. It is plausible that
long-lived supercells are quite rare across the West be-
cause of the rugged terrain and relatively limited low-
level moisture. Another possibility is that sparse radar
and observer coverage over parts of the western United
States has prevented the identification and documenta-
tion of long-lived supercells in this region.

Other locations where long-lived supercells appear to

be relatively infrequent are the far northeastern and
southeastern United States (Fig. 3). It is difficult to
determine if the reasons for this are meteorological, or
if they are an artifact of the data collection process. For
example, the frequent lack of strong deep-layer shear
across Florida may be a hindrance to the development
of long-lived supercells in this region. On the other
hand, long-lived supercells may not be documented
across Florida and adjacent waters because the radar
coverage does not extend far enough to encompass the
complete supercell lifetime. At any rate, limited radar
coverage near the northern and southern borders of the
United States, along with a lack of other documenta-
tion for Mexican and Canadian storms, likely reduced
the number of long-lived supercell detections in these
areas.

In contrast to the aforementioned areas, the fre-

quency of long-lived supercells appears relatively en-
hanced east of the Front Range of the Rocky Moun-
tains. This is an area where thunderstorms commonly
initiate, and subsequently track across the central
plains. It is also tempting to view the area from south-
ern Illinois through Kentucky, Tennessee, Virginia, and
North Carolina as a region where long-lived supercell
occurrence is above average. However, confirmation of
this observation awaits a detailed climatology (i.e., con-
tinuous observations over a period of many years).

The long-lived supercells tracked mostly toward the

northeast, east, or southeast (Fig. 3). There were re-
gional variations, however, with northeastward-moving
supercells common in the Southeast, and southeast-
ward-moving supercells relatively more frequent in the
Northeast. Additionally, long-lived supercells tended
toward a southeast direction east of the Front Range of
the Rockies, and some supercells tracked nearly due
south across the central United States. The fact that
there were at least 10 supercells with a strong south-

CTOBER

2006

B U N K E R S E T A L .

679

ward component to their motion (over the central
United States), and only a few with a strong northward
component (excluding left movers), is intriguing, and
perhaps related to the low-level moisture source from
the Gulf of Mexico. Rarely did supercells track very far
over the ocean after leaving the eastern United States,
likely because of the stabilizing effect of the relatively
cold ocean surface on the lower atmosphere. Last, the
motion of the supercells was generally constant
throughout their lifetime, but some gradually curving
tracks were observed.

There were only four long-lived left-moving super-

cells in the dataset; all occurred across the south-central
United States (SCNTRL) during April–June, and they
also resulted in significant severe hail. Three of the four
left movers were isolated, and all were associated with
a discrete convective mode. One of the left-moving su-
percells traveled straight eastward, while the other
three had a motion toward the north or northeast; each
of the four left movers had a lifetime at or just slightly

greater than 4 h. All four of the long-lived left-moving
supercells resulted from thunderstorm splits [one of
these left movers was elevated as its inflow originated
above a surface-based stable layer (Rochette et al.
1999)], and their right-moving counterparts were also
long lived. Not surprisingly, three of the hodographs
associated with these left-moving supercells displayed
unidirectional shear above 1 km AGL, and the fourth
hodograph had a shear profile that was unidirectional
throughout the lowest 8 km AGL.

Although Fig. 3 cannot be interpreted as a true spa-

tial climatology of long-lived supercells across the
United States, a rough temporal climatology can be
constructed using the current dataset. Not surprisingly,
the diurnal and monthly frequencies of long-lived su-
percells follow the cycle of daytime heating (Fig. 4a), as
well as the typical increase in thunderstorm activity
throughout the spring and summer months (Fig. 4b).
The peak frequency of occurrence at 0000–0100 UTC is
about 25 times the minimum from 1000 to 1600 UTC,

. 3. Tracks of the 224 long-lived supercells in the present study (composing 184 long-

lived supercell events). The tips of the arrows indicate where the supercells reached their
demise. Four arbitrary regions of the central and eastern United States are delineated with the
thick dashed lines (NCNTRL, Northeast, SCNTRL, and Southeast). The first number in
parentheses corresponds to the number of long-lived

supercell events

per region, and the

second number represents the number of long-lived

supercells

per region. Supercell tracks

were assigned to the region that contained the greatest length of the track. Refer to section
2 for a discussion of supercell events.

680

W E A T H E R A N D F O R E C A S T I N G

OLUME

and furthermore, 95% of the occurrences of individual
long-lived supercells fell between 1800 and 0800 UTC
(Fig. 4a). (Note, however, that the maximum in Fig. 4a
is smoothed slightly because the data were not normal-
ized to local time.) The average initiation time of a
long-lived supercell was 2200 UTC, with a mean lon-
gevity of 5.5 h.

Even though long-lived supercells were relatively

rare in most of the winter and fall seasons, a small
secondary peak was noted in November and December
(Fig. 4b). One gains a different perspective of this sec-
ond-order cold-season maximum when inspecting the
monthly distribution geographically. Over the north-
central United States (NCNTRL), 94% of the long-
lived supercell events occurred between May and Au-
gust, with none recorded in the cold season. Values
were similar in the Northeast, with 86% of the events
from April through June, and the rest in August and

September. Conversely, across the Southeast 56% of
the long-lived events were observed from October
through March, with none in the summer months. Fi-
nally, the SCNTRL was somewhere in between the
other regions, with 24% of the events occurring from
October through March, and 48% in the summer
months.

b. Comparison of properties between long- and

short-lived supercell events

Severe weather of some variety [i.e., hail diameter

ⱖ

1.9 cm (0.75 in.), measured or estimated wind gusts

ⱖ

26 m s

⫺

(58 mi h

⫺

), or wind damage; tornadoes are

treated separately below] was observed with all of the
long-lived supercell events, and with 92% of the short-
lived supercell events (Table 2). This is consistent with
previous studies, which report that 90% or greater of
supercells are severe (e.g., Burgess and Lemon 1991;
Bunkers 2002). Indeed, it was extremely rare for any of
the supercell events not to be associated with severe
hail; just 1%–2% of the events produced severe wind
only. Many of the long-lived supercells produced severe
hail along a substantial portion of their path (74% were
associated with significant severe hail; Table 2), and
particularly damaging storms resulted when this was
accompanied by severe wind (e.g., Klimowski et al.
1998; Parker et al. 2005; Segele et al. 2005). In addition,
a substantial portion of the severe storm reports on
some days was generated by one (e.g., 24 July 2000—
34% of total), two (e.g., 1 May 2002—60% of total), or
three (e.g., 5 April 2003—77% of total) long-lived su-
percells (Fig. 1), illustrating the significant forecasting
and societal impacts from these storms.

It is important to note that just because a supercell is

short lived does not mean the severe weather threat is
diminished. In 38% of these short-lived events, the su-
percells were observed to evolve into a squall line or
bow-echo system (Table 2), often producing a swath of
severe winds

after

the supercell phase had ended and

thus prolonging the convective event (e.g., Tuttle and
Carbone 2004; also recall footnote 2). Indeed, signifi-
cant severe winds were associated with 31% of the
long- and short-lived supercell events (Table 2). Addi-
tionally, there was a tendency for the short-lived super-
cells, along with the associated nonsupercells, to be
more numerous at any given location over the course of
an event (versus the case for the long-lived supercells),
thus increasing the local probability of severe storm
reports.

The production of tornadoes by the two supercell

classes revealed a greater difference than what was ob-
served for the severe hail and wind. For example, the
percentage of long-lived supercell events that produced

. 4. (a) Hourly and (b) monthly frequencies of occurrence

for the long-lived supercell events. In (a), one count is plotted for
each hour a long-lived supercell existed.

CTOBER

2006

B U N K E R S E T A L .

681

strong and violent tornadoes (F2–F5) was over four
times the value for short-lived supercell events (36%
versus 8%; Table 2). Moreover, only about one-third of
the long-lived supercell events did not result in a tor-
nado of any rating, while this increased to nearly two-
thirds for the short-lived supercell events. Perhaps this
higher percentage of tornado occurrence with the long-
lived supercells is due, in part, to the greater likelihood
of these storms interacting with a favorable environ-
ment—via their longevity. This difference may also be
related to the increased probability of receiving a local
storm report from the long-lived supercells simply be-
cause they persisted for a long time. Indeed, the aver-
age lifetime of the long-lived supercell events was 3.9
times that of the short-lived supercell events. There-
fore, the fact that about twice as many of the long-lived
supercell events were tornadic should not necessarily
be surprising.

In contrast to the F0–F1 tornadoes (see Table 2), the

production of strong and violent tornadoes by the long-
lived supercells does appear noteworthy. For the 184
long-lived supercell events, 66 (or 36%) produced tor-
nadoes of

ⱖ

F2 rating, but for the 119 short-lived super-

cell events, only 9 (or 8%) produced tornadoes of

ⱖ

rating. To compare the number of hours of strong and

violent tornado production between the long- and
short-lived events for equivalent numbers of total storm
hours, the ratio of total long-lived supercell lifetime to
total short-lived supercell lifetime in the dataset was
computed. Because the average lifetime of the super-
cells per long-lived event was 5.5 h, with 1.2 long-lived
supercells per event, there were approximately 1214 h
of long-lived supercell lifetime (184 events

⫻

1.2 super-

cells per event

⫻

5.5 hours per supercell). The average

lifetime of the supercells per short-lived event was 1.4
h, with an average of 2–3 short-lived supercells per
event; thus, there was a range of 333–500 h of short-
lived supercell lifetime (119 events

⫻

2–3 supercells per

event

⫻

1.4 hours per supercell). Therefore, the data-

base contains 2.4–3.6 times the number of hours of
long-lived supercell occurrence versus short-lived su-
percell occurrence. Taking this range of values times
the number of F2–F5 short-lived events (nine) yields
22–33 per 1214 hours (or one event per 36.8–55.2
hours). The number of F2–F5 long-lived events (66) is
2–3 times this (or one event per 18.4 hours). Therefore,
on a normalized basis, the long-lived supercell events
produced about 2–3 times the number of strong and
violent tornadoes per hour when compared with the
short-lived supercell events, suggesting there may be a

ABLE

2. Comparison of properties between the long- and short-lived supercell events. Refer to section 3b for a discussion of these

variations.

Long-lived supercell events

Short-lived supercell events

Tornado reports

36%

ⱖ

34% F0–F1

30% F0–F1

30% nontornadic

62% nontornadic

Hail/wind reports

100% severe

92% severe

84% hail and wind

67% hail and wind

15% hail only

23% hail only

1% wind only

2% wind only

24% significant hail and wind

19% significant hail and wind

74% significant hail

42% significant hail

31% significant wind

Convective mode

68% discrete

37% discrete

28% mixed (discrete and linear)

50% mixed (discrete and linear)

4% linear

13% linear

Supercell demise

63% weakened or dissipated

38% weakened or dissipated

6% experienced a destructive merger

20% experienced a destructive merger

21% evolved to another mode

7% merged and then evolved

17% merged and then evolved

3% had inflow source region cut off

4% had inflow source region cut off

Isolated and mergers

79% isolated (vs 21% not)

30% isolated (vs 70% not)

47% with mergers (vs 53% without)

74% with mergers (vs 26% without)

35% of the isolated also had mergers

17% of the isolated also had mergers

89% of those not isolated also had mergers

99% of those not isolated also had mergers

682

W E A T H E R A N D F O R E C A S T I N G

OLUME

connection between supercell longevity and the pro-
duction of F2–F5 tornadoes.

Many of the remaining supercell properties—

convective mode, demise, degree of isolation (i.e., the
percent isolated), and mergers—appear to be interre-
lated. Of these, the most discriminating variable be-
tween the two supercell classes is the degree of isolation
of the supercells; 79% of the long-lived supercell events
were classified as involving isolated storms, but only
30% of the short-lived supercell events fell into this
category (Table 2). On some occasions there were no
other storms within 100–200 km of a long-lived super-
cell. Furthermore, in a few instances there were two or
three distinct long-lived supercells in relatively close
proximity (but still defined as isolated) for a majority of
their lifetimes (e.g., Fig. 1). When the long-lived super-
cells were isolated (nonisolated), their average lifetime
was 5.6 h (5.1 h). Consistent with the degree of isolation
of the supercells, only 47% of the long-lived supercell
events displayed storm mergers, versus 74% for the
short-lived supercell events. This result may seem coun-
terintuitive since one might expect the longest-lived su-
percells to have the greatest chance of experiencing
mergers, simply because of their longevity. However,
this finding is consistent with the above observations
that supercells were more numerous in the short-lived
events (see previous paragraph), indicating coverage is
a nontrivial influence on longevity. Simply put, if a su-
percell is relatively isolated for most of its lifetime, no
matter how long it lives, storm mergers are not likely
(refer to the bottom of Table 2).

With this understanding of the degree of isolation for

the two supercell classes, it is not surprising that 68% of
the long-lived supercell events exhibited a discrete con-
vective mode, in contrast to just 37% for the short-lived
supercell events (Table 2). The discrete-mode long-
lived supercells had the greatest average lifetime at
5.7 h, compared with 5.4 h for the mixed mode and
4.5 h for the linear mode. The most common convective
mode for the short-lived supercell events was mixed
(50%), whereby the supercells commonly began in a
discrete fashion, but grew upscale into a quasi-linear
MCS by about midway through the supercell event.
Moreover, only 7% of these mixed-mode short-lived
supercell events were classified as involving isolated
storms. The linear convective mode was least common
for both of the supercell classes, but still over three
times as likely for the short-lived supercell events ver-
sus the long-lived events. Finally, isolated supercells
were not observed within the linear convective mode
for either supercell class, whereas 92% (70%) of the
long-lived (short lived) supercell events exhibiting a
discrete convective mode were isolated.

Supercell demise was partitioned into five categories

as described in section 2c [Table 2; 1) weakened or
dissipated; 2) destructive merger; 3) evolved to other
convective mode; 4) merger, then evolved to other con-
vective mode; and 5) inflow source region cut off]. The
results of this partitioning fit together with the findings
on the degree of isolation of the supercells, as well as
storm mergers. For example, 63% of the long-lived su-
percell events involved a demise characterized by a
weakening of the supercell circulation (but the parent
thunderstorm was still present), or else simply by storm
dissipation whereby the reflectivity quickly decreased
below 35 dB

(Table 2); isolated storms are most likely

to experience such a demise. By way of comparison,
only 13% of the long-lived supercell events involved a
demise characterized by storm mergers.

On the other

hand, just 38% of the short-lived supercell events were
typified by supercells that weakened or dissipated,
while 37% involved a supercell demise distinguished by
storm mergers; this is consistent with storms that are
relatively numerous.

Another demise mechanism that was equally com-

mon between the two supercell classes was evolution to
another convective mode [e.g., supercell to bow echo;
Johns and Leftwich (1988); Moller et al. (1994)], which
subsequently may have persisted for some time after
the supercell phase ended. This evolution was some-
times triggered by a storm merger, especially for the
short-lived supercells (Table 2; Klimowski et al. 2004).
Finally, although the percentage of supercell events
identified as having their inflow source region cut off
was

⬍

5% for both classes (an admittedly conservative

result), this value may have been higher if more de-
tailed radar data (e.g., 1-km resolution), as well as a
denser network of radar data, had been available for all
events.

In summary, long-lived supercells, when compared

with short-lived supercells, have a much greater ten-
dency to be isolated and discrete, thereby limiting ad-
jacent storm interactions and potentially destructive
mergers. Because of this, the production of F2–F5 tor-
nadoes appears to be enhanced for long-lived supercells
[although other factors such as low-level shear and
cloud-base height are equally or even more important
for tornadogenesis; Thompson et al. (2003)]. Conse-
quently, the less isolated supercells are, the greater the
likelihood they will experience destructive mergers and

Note that the percentage of supercell events with storm merg-

ers exceeds the percentage of supercell events reaching their de-
mise via storm mergers. This is because not all storm mergers
were destructive (i.e., they did not all result in the end of the
supercell phase).

CTOBER

2006

B U N K E R S E T A L .

683

be short lived, potentially curtailing their production of
F2–F5 tornadoes.

c. Regional variations for the long-lived supercells

Several differences were noted between long- and

short-lived supercells in section 3b. In the present sec-
tion, differences in long-lived supercell properties
among four regions (see Fig. 3) of the central and east-
ern United States are discussed. Because of the ex-
tremely small sample of long-lived supercell events
across the western United States (one), this discussion
of regional variations does not include that area.

Considering this somewhat arbitrary geographical di-

vision, the largest overall contrast in long-lived super-
cell properties is evident between the NCNTRL and
the Southeast. Most noteworthy are the following:
(i) long-lived supercells were more than twice as likely
to be isolated in the NCNTRL versus the Southeast;
(ii) the convective mode was dominated by discrete su-
percells over the NCNTRL, versus a mixed mode in the
Southeast; (iii) on an event basis, significant severe hail
was 1.5 times more common in the NCNTRL and sig-
nificant severe wind was 1.6 times more common in
the Southeast; and (iv) significant tornadoes were
four times less likely to occur with supercells in the

NCNTRL than in the Southeast (Fig. 5). In fact, the
NCNTRL stands out as the only region where the oc-
currence of tornadoes with long-lived supercell events
(58% tornadic) was less than average for the entire
long-lived dataset (70% tornadic), whereas the other
three regions displayed above average percentages for
the tornadic long-lived events (78%–82% tornadic).
The long-lived supercells in the NCNTRL were also
2–4 times more likely to be associated with hail reports
only—versus hail and wind reports—when compared
with the other three regions.

Although the storm environmental conditions will be

discussed in Part II, a few things can be said at this point
to help explain the above differences. For one, low-
level moisture is considerably more abundant in the
Southeast versus the NCNTRL. As a result, deep
moistconvection should lead to storms that are more
closely spaced in the Southeast, and more isolated in
the NCNTRL. In turn, this would translate into rela-
tively more destructive mergers and mixed convective
modes over the Southeast. Recall from section 3a that
most of the long-lived supercell events across the
Southeast occurred during the cold season, suggesting a
propensity for strong atmospheric forcing, which in
turn would favor a mixed or linear convective mode.
Other nonmeteorological factors, such as differences in

. 5. Properties of all 184 long-lived supercell events (left column), along with the prop-

erties of four arbitrarily defined regions of the central and eastern United States. Refer to
section 3c for a discussion of these variations.

684

W E A T H E R A N D F O R E C A S T I N G

OLUME

population and tree densities, may partially explain the
fewer wind and tornado reports over the NCNTRL
(e.g., Klimowski et al. 2003).

Some other regional differences are also of interest.

For example, the Southeast is distinguished from the
other three regions by way of destructive mergers (29%
versus 4% for the other three regions), as well as a
greater percentage of mergers that produce an evolu-
tion to another convective mode (Fig. 5). This may very
well be related to the relatively small percentage of
isolated supercells in the Southeast (35% versus 79%–
86% for the others), and the fact that 61% of the events
in the Southeast consisted of mixed or linear convective
modes (versus 18%–37% for the others). Another ob-
servation is that just under one-third of the long-lived
supercell events over the central United States reached
their demise by way of evolution into another convec-
tive mode (26%–31%); however, this was quite a rare
occurrence over the eastern United States (only 4% of
the events). The drier low-level environment over the
central United States, compared with the eastern
United States, may foster this type of transition via
enhancement of evaporatively cooled downdrafts. In
support of this, Klimowski et al. (2003) found 29% of
the high-wind-producing supercells studied across the
northern high plains evolved into a bow echo. It is also
noteworthy that mergers were relatively more common
over the eastern United States (62%–65%) when com-
pared with the central United States (42%–43%).

Like the NCNTRL, the Northeast appears to have

some interesting properties that set it apart from the
other three regions, especially from the southern areas.
First, discrete-mode long-lived supercells were most
common in the Northeast (82%), ranging from 19%
to 43% more frequent than the values across the
SCNTRL and Southeast, and 12% more than for the
NCNTRL (Fig. 5). Second, for 88% of the events, long-
lived supercells simply weakened over the Northeast,
which is 27%–35% greater than for the other regions.
Some of this weakening occurred as the supercells trav-
eled off the coast and over the Atlantic Ocean. Finally,
it is difficult to explain why mergers were most common
over the Northeast, yet 79% of the long-lived supercell
events were still classified as isolated. Apparently
mergers are not as destructive in the Northeast when
compared with the other regions, which is consistent
with the predominantly discrete convective mode
(82%). Note that these results for the Northeast and
Southeast may possibly be affected by the small sample
size (only 22 and 18 events, respectively); care should
be used when making use of them.

Last, anywhere from 39% to 70% of the supercells

experienced a cycling or regeneration, in which a new

mesocyclone developed within the same storm, or the
storm experienced some noncontinuous development
(Foote and Frank 1983; Browning 1977; Burgess et al.
1982). This appears to be a relatively common feature
of these longest-lived storms. Given the difficulty in
observing this transition within the present dataset (i.e.,
some of the radar data were at 2-km resolution), com-
parisons of storm cycling among the four regions should
be made with caution. However, it can be concluded
that in general, long-lived supercells are likely multiple-
core mesocyclones (or mesoanticyclones).

4. Conclusions and summary

It is proposed that long-lived supercells be defined as

supercells that persist in a quasi-steady manner for at
least 4 h. Moreover, an upper limit for short-lived su-
percell lifetime is suggested at 2 h. Using these defini-
tions and a dataset of supercells across mainly the cen-
tral and eastern United States, the radar-observed
properties and severe weather production of long-lived
supercells were compared with those of short-lived su-
percells. In addition, geographical variations of long-
lived supercell properties across the United States were
also examined. Based on these analyses, the primary
conclusions of this study are as follows.

•

Long-lived supercells produce notably more F2–F5
tornadoes when compared with short-lived super-
cells, and a single long-lived supercell can also pro-
duce a substantial amount of nontornadic severe
weather.

•

Long-lived supercells are frequently isolated (which
helps them live longer), and therefore experience
storm mergers relatively infrequently. Short-lived su-
percells are not isolated very often and, thus, experi-
ence storm mergers more frequently.

•

Long-lived supercells are most often observed when
there is a discrete convective mode, but short-lived
supercells are mostly likely when a mixed convective
mode is present.

•

Long-lived supercells typically reach their demise
(i.e., the end of the supercell phase) when the parent
thunderstorm dissipates or its circulation weakens.
Short-lived supercells are most likely to experience a
destructive (i.e., circulation disrupting) merger or
else evolve into another convective mode at their de-
mise.

•

Long-lived supercells are rare in the western United
States, and long-lived left-moving supercells are rare
everywhere in the United States.

•

Long-lived supercells have the highest probability of
producing F2–F5 tornadoes in the Southeast (

⬎

60%)

CTOBER

2006

B U N K E R S E T A L .

685

when compared with the NCNTRL, Northeast, and
SCNTRL.

•

Long-lived supercells in the Southeast are not iso-
lated as often as in other parts of the United States,
and they commonly occur within a mixed convective
mode.

•

Most of the long-lived supercell events over the

NCNTRL occur between May and August, but more
than half of the long-lived events over the Southeast
occur from October through March (with none in the
summer months).

When ruminating on these results, it is important to

remember the definitions in section 2a. First, note that
long-lived supercell events can, and sometimes do, con-
tain short-lived supercells. Second, some short-lived
events have supercells with lifetimes between 2 and 4 h.
Therefore, there can be a long-lived supercell event
that has predominately short-lived supercells, while at
other times a short-lived supercell event may contain a
few moderate-lived supercells. Because the goal of this
study is to better understand the supercell longevity
spectrum, and also given the relatively large dataset, it
is believed that these “blurred” boundaries among
long-, moderate-, and short-lived supercell events do
not pose a serious problem.

With this word of caution as a backdrop, it is clear

that a fundamental property of long-lived supercells is
their generally isolated nature. Furthermore, if long-
lived supercells are going to occur on any given day,
then the probability of significant tornadoes is en-
hanced. Along these lines, a single long-lived supercell
may produce a majority of the severe storm reports
during a convective event. Therefore, if supercell lon-
gevity can be anticipated in an operational setting, one
of the potential benefits is improved severe local storm
watches and warnings. Short-term (0–12 h) forecasts
can also be improved with proper knowledge of long-
lived supercell occurrence.

The differences between long- and short-lived super-

cells appear significant enough to warrant further in-
vestigation into the environmental conditions support-
ing their existence. For example, why are long-lived
supercells more isolated and discrete than short-lived
supercells? What is the reason why long-lived super-
cells produce significantly more F2–F5 tornadoes than
short-lived supercells? Can regional differences in en-
vironmental variables help to explain variations in con-
vective mode and tornado production? We will address
these and other questions in Part II of this work.

Acknowledgments.

We greatly appreciate the efforts

from the following: (i) Lee Czepyha (NWS RAP),

Jason Grzywacz (NWS DDC), Jeffrey Johnson (NWS
RAP), and Brian Klimowski (NWS FGZ) for their as-
sistance in the data collection process; (ii) Dolores
Kiessling (and the COMET program) for providing a
considerable amount of radar data; (iii) the NOAA
Central Library who supplied many of the references
pertinent to this study; (iv) Don Burgess for his insight
on cyclic mesocyclogenesis and the nature of isolated
versus nonisolated supercells; (v) Chuck Doswell, Rich-
ard Grumm, Ray Wolf, and Jon Zeitler for providing
information on some historical long-lived supercell
events; and (vi) Andrew Detwiler, Roger Edwards,
Dennis Todey, and one anonymous reviewer who
helped clarify the manuscript significantly. Finally, the
first author thanks David Carpenter, meteorologist-in-
charge at the NWS in Rapid City, South Dakota, for
supporting this work.

REFERENCES

Adlerman, E. J., K. K. Droegemeier, and R. Davies-Jones, 1999:

A numerical simulation of cyclic mesocyclogenesis.

J. Atmos.

Sci.,

56,

2045–2069.

Barnes, S. L., and S. P. Nelson, 1978: Oklahoma thunderstorms on

29–30 April 1970. Part IV: Study of a dissipating severe
storm.

Mon. Wea. Rev.,

106,

704–712.

Bluestein, H. B., and S. S. Parker, 1993: Modes of isolated, severe

convective storm formation along the dryline.

Mon. Wea.

Rev.,

121,

1354–1372.

Brooks, H. E., C. A. Doswell III, and R. Davies-Jones, 1993: En-

vironmental helicity and the maintenance and evolution of
low-level mesocyclones.

The Tornado: Its Structure, Dynam-

ics, Prediction, and Hazards, Geophys. Monogr.,

Vol. 79,

Amer. Geophys. Union, 97–104.

——, ——, and J. Cooper, 1994a: On the environments of tornadic

and nontornadic mesocyclones.

Wea. Forecasting,

606–618.

——, ——, and R. B. Wilhelmson, 1994b: The role of midtropo-

spheric winds in the evolution and maintenance of low-level
mesocyclones.

Mon. Wea. Rev.,

122,

126–136.

——, ——, and M. P. Kay, 2003: Climatological estimates of local

daily tornado probability for the United States.

Wea. Fore-

casting,

18,

626–640.

Browning, K. A., 1962: Cellular structure of convective storms.

Meteor. Mag.,

91,

341–350.

——, 1964: Airflow and precipitation trajectories within severe

local storms which travel to the right of the winds.

J. Atmos.

Sci.,

21,

634–639.

——, 1977: The structure and mechanisms of hailstorms.

Hail: A

Review of Hail Science and Hail Suppression, Meteor.
Monogr.,

No. 38, Amer. Meteor. Soc., 1–43.

——, and G. B. Foote, 1976: Airflow and hail growth in supercell

storms and some implications for hail suppression.

Quart. J.

Roy. Meteor. Soc.,

102,

499–533.

Bunkers, M. J., 2002: Vertical wind shear associated with left-

moving supercells.

Wea. Forecasting,

17,

845–855.

——, B. A. Klimowski, J. W. Zeitler, R. L. Thompson, and M. L.

Weisman, 2000: Predicting supercell motion using a new
hodograph technique.

Wea. Forecasting,

15,

61–79.

——, J. S. Johnson, L. J. Czepyha, J. M. Grzywacz, B. A. Kli-

mowski, and M. R. Hjelmfelt, 2006: An observational exami-

686

W E A T H E R A N D F O R E C A S T I N G

OLUME

nation of long-lived supercells. Part II: Environmental con-
ditions and forecasting.

Wea. Forecasting,

21,

689–714.

Burgess, D. W., and L. R. Lemon, 1991: Characteristics of meso-

cyclones detected during a NEXRAD test. Preprints,

25th

Int. Conf. on Radar Meteorology,

Vol. 1, Paris, France, Amer.

Meteor. Soc., 39–42.

——, V. T. Wood, and R. A. Brown, 1982: Mesocyclone evolution

statistics. Preprints,

12th Conf. on Severe Local Storms,

San

Antonio, TX, Amer. Meteor. Soc., 422–424.

Calianese, E. J., J. K. Jordan, E. B. Curran, A. R. Moller, and G.

Woodall, 2002: The Mayfest high-precipitation supercell of 5
May 1995—A case study. Preprints,

21st Conf. on Severe Lo-

cal Storms,

San Antonio, TX, Amer. Meteor. Soc., 105–108.

Conway, J. W., H. E. Brooks, and K. D. Hondl, 1996: The 17 Au-

gust 1994 Lahoma, OK supercell: Issues of tornadogenesis
and bow echo formation. Preprints,

18th Conf. on Severe Lo-

cal Storms,

San Francisco, CA, Amer. Meteor. Soc., 52–56.

Crum, T. D., R. L. Alberty, and D. W. Burgess, 1993: Recording,

archiving, and using WSR-88D data.

Bull. Amer. Meteor.

Soc.,

74,

645–653.

Curran, E. B., and W. D. Rust, 1992: Positive ground flashes pro-

duced by low-precipitation thunderstorms in Oklahoma on
26 April 1984.

Mon. Wea. Rev.,

120,

544–553.

Dial, G. L., and J. P. Racy, 2004: Forecasting short term convec-

tive mode and evolution for severe storms initiated along
synoptic boundaries. Preprints,

22d Conf. on Severe Local

Storms,

Hyannis, MA, Amer. Meteor. Soc., CD-ROM,

11A.2.

Doswell, C. A., III, Ed., 2001:

Severe Convective Storms. Meteor.

Monogr.,

No. 50, Amer. Meteor. Soc., 561 pp.

——, and D. W. Burgess, 1988: On some issues of United States

tornado climatology.

Mon. Wea. Rev.,

116,

495–501.

——, and ——, 1993: Tornadoes and tornadic storms: A review of

conceptual models.

The Tornado: Its Structure, Dynamics,

Prediction, and Hazards, Geophys. Monogr.,

Vol. 79, Amer.

Geophys. Union, 161–172.

——, A. R. Moller, and R. Przybylinski, 1990: A unified set of

conceptual models for variations on the supercell theme. Pre-
prints,

16th Conf. on Severe Local Storms,

Kananaskis Park,

Alberta, Canada, Amer. Meteor. Soc., 40–45.

——, D. V. Baker, and C. A. Liles, 2002: Recognition of negative

mesoscale factors for severe-weather potential: A case study.

Wea. Forecasting,

17,

937–954.

Dowell, D. C., and H. B. Bluestein, 2002: The 8 June 1995

McLean, Texas, storm. Part I: Observations of cyclic torna-
dogenesis.

Mon. Wea. Rev.,

130,

2626–2648.

Fankhauser, J. C., 1971: Thunderstorm–environment interactions

determined from aircraft and radar observations.

Mon. Wea.

Rev.,

99,

171–192.

Foote, G. B., and C. G. Wade, 1982: Case study of a hailstorm in

Colorado. Part I: Radar echo structure and evolution.

J. At-

mos. Sci.,

39,

2828–2846.

——, and H. W. Frank, 1983: Case study of a hailstorm in Colo-

rado. Part III: Airflow from triple-Doppler measurements.

Atmos. Sci.,

40,

686–707.

Fujita, T. T., 1974: Jumbo tornado outbreak of 3 April 1974.

Weatherwise,

(3), 116–126.

——, 1989: The Teton–Yellowstone tornado of 21 July 1987.

Mon.

Wea. Rev.,

117,

1913–1940.

——, and H. Grandoso, 1968: Split of a thunderstorm into anti-

cyclonic and cyclonic storms and their motion as determined
from numerical model experiments.

J. Atmos. Sci.,

25,

416–

439.

——, D. L. Bradbury, and C. F. Van Thullenar, 1970: Palm Sun-

day tornadoes of April 11, 1965.

Mon. Wea. Rev.,

98,

29–69.

Glass, F. H., and M. F. Britt, 2002: The historic Missouri–Illinois

high precipitation supercell of 10 April 2001. Preprints,

21st

Conf. on Severe Local Storms,

San Antonio, TX, Amer. Me-

teor. Soc., 99–104.

Grasso, L. D., 2000: The dissipation of a left-moving cell in a

severe storm environment.

Mon. Wea. Rev.,

128,

2797–2815.

Hales, J. E., Jr., 1988: Improving the watch/warning program

through use of significant event data. Preprints,

15th Conf. on

Severe Local Storms,

Baltimore, MD, Amer. Meteor. Soc.,

165–168.

Hamilton, J. W., 1958: Some features in the use of radar in fore-

casts and warnings of heavy hail, damaging winds, and/or
tornadoes.

Proc. Seventh Weather Radar Conf.,

Miami Beach,

FL, Amer. Meteor. Soc., H18–H25.

Hart, J. A., and P. R. Janish, cited 2005: SeverePlot: Historical

severe weather report database version 2.5. Storm Prediction
Center, Norman, OK. [Available online at http://www.spc.
noaa.gov/software/svrplot2/.]

Johns, R. H., and P. W. Leftwich Jr., 1988: The severe thunder-

storm outbreak of July 28-29 1986. . .A case exhibiting both
isolated supercells and a derecho producing convective sys-
tem. Preprints,

15th Conf. on Severe Local Storms,

Baltimore,

MD, Amer. Meteor. Soc., 448–451.

——, J. M. Davies, and P. W. Leftwich, 1993: Some wind and

instability parameters associated with strong and violent tor-
nadoes. 2. Variations in the combinations of wind and insta-
bility parameters.

The Tornado: Its Structure, Dynamics, Pre-

diction, and Hazards, Geophys. Monogr.,

Vol. 79, Amer. Geo-

phys. Union, 583–590.

Klimowski, B. A., M. R. Hjelmfelt, M. J. Bunkers, D. Sedlacek,

and L. R. Johnson, 1998: Hailstorm damage observed from
the

GOES-8

satellite: The 5–6 July 1996 Butte–Meade storm.

Mon. Wea. Rev.,

126,

831–834.

——, M. J. Bunkers, M. R. Hjelmfelt, and J. N. Covert, 2003: Se-

vere convective windstorms over the northern high plains of
the United States.

Wea. Forecasting,

18,

502–519.

——, M. R. Hjelmfelt, and M. J. Bunkers, 2004: Radar observa-

tions of the early evolution of bow echoes.

Wea. Forecasting,

19,

727–734.

Knupp, K. R., S. Paech, and S. Goodman, 2003: Variations in

cloud-to-ground lightning characteristics among three adja-
cent tornadic supercell storms over the Tennessee Valley re-
gion.

Mon. Wea. Rev.,

131,

172–188.

Lemon, L. R., 1980: Severe thunderstorm radar identification

techniques and warning criteria. NOAA Tech. Memo. NWS
NSSFC-3, 60 pp. [Available from NOAA Central Library,
1315 East–West Highway, Silver Spring, MD 20910.]

Lindsey, D. T., and M. J. Bunkers, 2005: Observations of a severe,

left-moving supercell on 4 May 2003.

Wea. Forecasting,

20,

15–22.

MacGorman, D. R., and D. W. Burgess, 1994: Positive cloud-to-

ground lightning in tornadic storms and hailstorms.

Mon.

Wea. Rev.,

122,

1671–1697.

Maddox, R. A., L. R. Hoxit, and C. F. Chappell, 1980: A study of

tornadic thunderstorm interactions with thermal boundaries.

Mon. Wea. Rev.,

108,

322–336.

Markowski, P. M., C. Hannon, J. Frame, E. Lancaster, A. Pietry-

cha, R. Edwards, and R. L. Thompson, 2003: Characteristics
of vertical wind profiles near supercells obtained from the
Rapid Update Cycle.

Wea. Forecasting,

18,

1262–1272.

CTOBER

2006

B U N K E R S E T A L .

687

Marwitz, J. D., 1972: The structure and motion of severe hail-

storms. Part I: Supercell storms.

J. Appl. Meteor.,

11,

166–179.

Moller, A. R., C. A. Doswell III, M. P. Foster, and G. R. Woodall,

1994: The operational recognition of supercell thunderstorm
environments and storm structures.

Wea. Forecasting,

327–

347.

Musil, D. J., A. J. Heymsfield, and P. L. Smith, 1986: Microphysi-

cal characteristics of a well-developed weak echo region in a
high plains supercell thunderstorm.

J. Climate Appl. Meteor.,

25,

1037–1051.

Nelson, S. P., 1987: The hybrid multicellular–supercellular

storm—An efficient hail producer. Part II: General charac-
teristics and implications for hail growth.

J. Atmos. Sci.,

44,

2060–2073.

NOAA, 1994: Southeastern United States Palm Sunday Tornado

Outbreak of March 27, 1994. U.S. Dept. of Commerce, 54 pp.
[Available from NOAA/National Weather Service, Warning
and Forecast Branch, 1325 East–West Highway, Silver
Spring, MD 20910.]

——, 1995: The Fort Worth–Dallas hailstorm/flash flood May 5,

1995. Natural Disaster Survey Rep., NWS Southern Region,
Fort Worth, TX, 39 pp. [Available from NOAA/National
Weather Service, Warning and Forecast Branch, 1325 East–
West Highway, Silver Spring, MD 20910.]

Parker, M. D., I. C. Ratcliffe, and G. M. Henebry, 2005: The July

2003 Dakota hailswaths: Creation, characteristics, and pos-
sible impacts.

Mon. Wea. Rev.,

133,

1241–1260.

Paul, A. H., 1973: The heavy hail of 23–24 July 1971 on the west-

ern prairies of Canada.

Weather,

28,

463–471.

Pence, K. J., and B. E. Peters, 2000: The tornadic supercell of 8

April 1998 across Alabama and Georgia. Preprints,

20th

Conf. on Severe Local Storms,

Orlando, FL, Amer. Meteor.

Soc., 206–209.

Rasmussen, E. N., and D. O. Blanchard, 1998: A baseline clima-

tology of sounding-derived supercell and tornado forecast
parameters.

Wea. Forecasting,

13,

1148–1164.

——, S. Richardson, J. M. Straka, P. M. Markowski, and D. O.

Blanchard, 2000: The association of significant tornadoes
with a baroclinic boundary on 2 June 1995.

Mon. Wea. Rev.,

128,

174–191.

Rochette, S. M., J. T. Moore, and P. S. Market, 1999: The impor-

tance of parcel choice in elevated CAPE computations.

Natl.

Wea. Dig.,

(4), 20–32.

Segele, Z. T., D. J. Stensrud, I. C. Ratcliffe, and G. M. Henebry,

2005: Influence of a hailstreak on boundary layer evolution.

Mon. Wea. Rev.,

133,

942–960.

Thompson, R. L., 1998: Eta Model storm-relative winds associ-

ated with tornadic and nontornadic supercells.

Wea. Forecast-

ing,

13,

125–137.

——, and R. Edwards, 2000: An overview of environmental con-

ditions and forecast implications of the 3 May 1999 tornado
outbreak.

Wea. Forecasting,

15,

682–699.

——, ——, J. A. Hart, K. L. Elmore, and P. Markowski, 2003:

Close proximity soundings within supercell environments ob-
tained from the Rapid Update Cycle.

Wea. Forecasting,

18,

1243–1261.

Tolleson, P., 1996: Oregon supercell of July 9, 1995 as seen by the

KRTX WSR-88D. NOAA/NWS Western Region Tech. At-
tachment 96-01, 3 pp. [Available from NWS Western Region,
125 South State St., Salt Lake City, UT 84103.]

Tuttle, J. D., and R. E. Carbone, 2004: Coherent regeneration and

the role of water vapor and shear in a long-lived convective
episode.

Mon. Wea. Rev.,

132,

192–208.

Wade, C. G., 1982: A preliminary study of an intense thunder-

storm which moved across the CCOPE research network in
southeastern Montana. Preprints,

Ninth Conf. on Weather

Analysis and Forecasting,

Seattle, WA, Amer. Meteor. Soc.,

388–395.

Warner, C., 1976: Wave patterns with an Alberta hailstorm.

Bull.

Amer. Meteor. Soc.,

57,

780–787.

Weaver, J. F., and S. P. Nelson, 1982: Multiscale aspects of thun-

derstorm gust fronts and their effects on subsequent storm
development.

Mon. Wea. Rev.,

110,

707–718.

Weisman, M. L., and J. B. Klemp, 1984: The structure and classi-

fication of numerically simulated convective storms in direc-
tionally varying wind shears.

Mon. Wea. Rev.,

112,

2479–2498.

——, and R. Rotunno, 2000: The use of vertical wind shear versus

helicity in interpreting supercell dynamics.

J. Atmos. Sci.,

57,

1452–1472.

Weiss, S. J., J. A. Hart, and P. R. Janish, 2002: An examination of

severe thunderstorm wind report climatology: 1970–1999.
Preprints,

21st Conf. on Severe Local Storms,

San Antonio,

TX, Amer. Meteor. Soc., 446–449.

Zeitler, J. W., and M. J. Bunkers, 2005: Operational forecasting of

supercell motion: Review and case studies using multiple
datasets.

Natl. Wea. Dig.,

(1), 81–97.

688

W E A T H E R A N D F O R E C A S T I N G

OLUME