530
WEATHER AND FORECASTING
VOLUME 18
NOTES AND CORRESPONDENCE
Refined Supercell and Tornado Forecast Parameters
ERIK N. RASMUSSEN
Cooperative Institute for Mesoscale Meteorological Studies, National Severe Storms Laboratory, and
University of Oklahoma, Norman, Oklahoma
12 February 2002 and 3 January 2003
ABSTRACT
This note updates a previous study that utilized a baseline climatology of soundings associated with large
hail, significant tornadoes, and 10 or more cloud-to-ground lightning flashes from 1992. Expanding on the earlier
analysis, it is shown that three modified forecast parameters have more value in distinguishing between environments that favor significant tornadoes and those that favor large hail but no significant tornadoes, in the
climatological data. These parameters are storm-relative helicity in the 0–1-km layer adjacent to the ground,
energy–helicity index computed from this measure of helicity, and the convective available potential energy that
accrues from the surface to 3 km above ground level. In addition, this note provides caveats regarding the
interpretation of the climatological findings.
1. Introduction
A previous paper (Rasmussen and Blanchard 1998,
hereinafter RB98) established a baseline climatology of
commonly used severe storm forecast parameters. The
methodology of climatological analysis can be found in
that paper. In brief, every lightning flash and severe
storm event (hail report .5.1 cm and/or tornado with
damage rated F2 or greater) from 1992 was associated
with a 0000 UTC sounding using inflow and proximity
criteria.
Research conducted prior to 1998 indicated that shear
in the near-ground layer might be critically important
for distinguishing tornadic from nontornadic supercells
(e.g., Markowski et al. 1998b,d; Wicker 1996). Based
on these findings from observational studies and numerical simulations, some preliminary calculations were
made using the 1992 baseline climatology. These calculations revealed that helicity concentration near the
ground tended to favor significant tornadoes (Markowski et al. 1998c). Thompson, Edwards, and collaborators at the Storm Prediction Center have subsequently utilized the Rapid Update Cycle, Version 2 (RUC2), operational model (Thompson and Edwards 2000;
Thompson et al. 2002c) to evaluate and confirm very
effectively the importance of the near-ground layer in
distinguishing between tornadic (significant and weak)
Corresponding author address: Dr. Erik Rasmussen, PO Box 267,
Mesa, CO 81643-0267.
E-mail: rasmussen@nssl.noaa.gov
q 2003 American Meteorological Society
and nontornadic supercells (Edwards and Thompson
2000; Thompson et al. 2002a,b). In a sense, some of
the findings reported in this paper are no longer novel,
but they are still useful in that they confirm the emerging
consensus of the importance of forecast parameters that
focus on the near-ground layer. It should be noted that
other multiyear climatological studies have been conducted to attempt to confirm or to refute these recent
findings (Craven et al. 2002).
Several modified forecast parameters show more utility for distinguishing among the categories of RB98 and
are reported in section 2. These parameters are ranked
and discussed further in section 3, and comments are
included about some recently discovered potential biases in the earlier findings. It is hoped that the physical
importance of these forecast parameters may be more
fully understood through numerical simulations, once
some of the current deficiencies in cloud models have
been overcome (Straka and Rasmussen 1998).
2. New forecast parameters
a. SRH in the 0–1-km AGL layer
The de facto standard method of computing stormrelative helicity (SRH; Davies-Jones et al. 1990) is to
integrate through the 0 –3-km above-ground-level
(AGL) layer. Recent findings on the importance of
boundaries (Markowski et al. 1998a) and the character
of the near-ground shear in numerically simulated
storms (M. Weisman 1998, personal communication;
JUNE 2003
NOTES AND CORRESPONDENCE
531
FIG. 2. Scatter diagram of SRH 0–1 vs SRH 2–3 for soundings representing the three classes described for Fig. 1. Labeled lines have
constant ratio of SRH 0–1 to SRH 2–3 .
FIG. 1. Box-and-whiskers graph of 0–1-km AGL SRH for soundings associated with significant tornadoes (TOR; on right), soundings
with hail .5.1 cm in diameter but without significant tornadoes (SUP;
in middle), and nonsevere thunderstorms (ORD; on left). Gray boxes
denote 25th–75th percentiles, with heavy horizontal bar at the median
value. Vertical lines (whiskers) extend to the 10th and 90th percentiles.
Wicker 1996) suggest that perhaps the shallower nearground layers are of more importance in affecting the
development of supercells and tornadoes (through lowlevel mesocyclone genesis). The data in the 1992 baseline climatology substantiate these findings: they indicate that the 0–1-km AGL SRH (SRH 0–1 ) is a better
forecast parameter for distinguishing among the TOR
and SUP1 classes than the commonly used 0–3-km SRH
(SRH 0–3 ), although SRH 0–3 is of more utility in distinguishing between SUP and ORD. The box-and-whiskers
diagram for SRH 0–1 is shown in Fig. 1 for direct comparison with Fig. 4 of RB98. Note that Edwards and
Thompson (2000) have found a similar signal with
RUC-2-derived SRH 0–1 in their analysis of a set of radar-identified supercells.
The distributions of SRH 0–3 published in RB98 are
similar in appearance to those shown in Fig. 1. Thus,
it is legitimate to ask whether there is any additional
information in SRH 0–1 or whether large SRH 0–1 is simply a contribution to large SRH through the deeper layer.
In examining the vertical distribution of SRH it is found
that it varies among the categories and that SRH 0–1 is
1
RB98 defined the categories as follows: TOR were soundings
associated with at least one tornado of F2 or greater damage intensity,
SUP were soundings associated with at least one report of hail .5.1
cm in diameter but no tornadoes of F2 or greater damage intensity,
and ORD soundings were associated with at least 10 cloud-to-ground
lightning flashes but no severe weather reports of any type.
not simply related to the deeper layer measure. One
aspect of this variation is illustrated by Fig. 2, which is
a scatter diagram of SRH 0–1 against that in the 2–3-km
AGL layer (SRH 2–3 ). Category TOR soundings in which
the SRH 2–3 is larger than SRH 0–1 (SRH 0–1 /SRH 2–3 ,
1.0) are seldom found. TOR soundings preferentially
have SRH 0–1 that is much larger than SRH 2–3 . The current investigation makes it clear that SRH 0–1 is of more
value in distinguishing between TOR and SUP than is
the quantity customarily computed. The vertical distribution of SRH also has been examined by Edwards and
Thompson (2000), who report a similar finding.
b. Modified EHI
In RB98, it was verified that combinations of convective available potential energy (CAPE; here computed as
in RB98) and low-level shear were generally better forecast parameters than measures of shear or CAPE alone.
Herein, CAPE is combined with SRH 0–1 in a modified
version of the energy–helicity index (EHI; Hart and Korotky 1991). This modified version (EHI 0–1 ) is identical
to the traditional form except for the substitution of the
SRH 0–1 . The scatter diagram of this parameter space is
given in Fig. 3 (for direct comparison with Fig. 11 of
RB98), and the box-and-whiskers diagram is shown in
Fig. 4 (for direct comparison with Fig. 12 of RB98). This
formulation of EHI is substantially better at distinguishing between the TOR and SUP classes than is any parameter in RB98 (refer to Table 1). Only 25% of the SUP
soundings had EHI 0–1 . 0.5, whereas nearly 2/3 of the
TOR soundings had values this large. The improvement
of EHI 0–1 over the conventional EHI can only be due to
the improvement of the SRH 0–1 over its 0–3-km counterpart. As shown in Table 1, this version of the EHI is
somewhat poorer in the SUP/ORD forecast discrimination than is the traditional EHI. This finding is similar
to that reported by Markowski et al. (1998d) and Edwards
and Thompson (2000).
532
WEATHER AND FORECASTING
VOLUME 18
FIG. 3. As in Fig. 2, but for the CAPE vs SRH parameter space of EHI 0–1 . Labeled curves are
lines of constant EHI.
c. CAPE below 3 km AGL
RB98 established that lifting-condensation-level
(LCL) height was one of the most important parameters
distinguishing between the TOR and SUP categories
(see their Table 3). There are various possible explanations, including the fact that low LCLs are associated
with humid near-ground air (ergo reduced low-level
evaporative cooling and downdraft production) as well
as the possibility of relatively larger CAPE (and thus
perhaps vertical stretching) occurring closer to the surface. The latter explanation is investigated further after
being first explored by McCaul (1991) and McCaul and
Weisman (1996). Using the baseline climatology, the 0–
3-km AGL CAPE (CAPE 0–3 ) was investigated and was
found to have a relatively strong ability to discriminate
between the TOR and SUP classes while having very
little value in the SUP/ORD forecast (Fig. 5 and Tables
1 and 2). This finding, although requiring further investigation, strongly suggests that the development of
tornadoes does require some near-ground CAPE, whereas this is not so important for storms that produce large
hail but not significant tornadoes. This may be because
near-ground CAPE is associated with more effective
interaction between low-level shear and low-level updraft, thereby augmenting ascent and stretching through
the nonhydrostatic pressure field (Rotunno and Klemp
1982). In RB98, the CAPE in the lowest 3 km above
the level of free convection was considered but found
to be of little forecast utility.
TABLE 1. Parameters ranked according to Heidke’s skill score in
the TOR vs SUP forecast when HSS . 0.29. VGP is the vorticity
generation parameter (RB98); CIN is convective inhibition.
FIG. 4. As in Fig. 1, but for EHI 0–1 .
Parameter
Optimal HSS
Optimal value
EHI 0–1
SRH 0–1
LCL
CAPE 0–3
CIN
VGP
0.355
0.349
0.348
0.336
0.297
0.295
0.725
120 m 2 s22
800 m
69 J kg21
16 J kg21
0.31
JUNE 2003
533
NOTES AND CORRESPONDENCE
TABLE 2. Parameters ranked according to Heidke’s skill score in
the (SUP 1 TOR) vs ORD forecast when HSS . 0.15. BL–6 km
shear is the magnitude of the shear vector between the boundary layer
and 6 km AGL.
FIG. 5. As in Fig. 1, but for CAPE 0–3 .
3. Discussion and conclusions
In RB98, forecast parameters were ranked according
to Heidke’s skill score (HSS; Doswell et al. 1990). Herein, the parameters (RB98 and the modified parameters)
are reranked and presented in Tables 1 and 2. From Table
1 it can be seen that EHI 0–1 has assumed the first position
and that SRH 0–1 has assumed the second position in the
TOR/SUP forecast utility rank. The optimal parameter
value is found by seeking the value producing the largest
HSS. The relatively large HSS for CAPE 0–3 may be one
indication that low-level stretching is important in the
TOR/SUP discrimination.
The general purpose of this note was to update RB98,
and attention is now focused on the findings concerning
the LCL that were reported therein. In conversations
concerning the importance of the LCL in forecasting
significant tornadoes, some interesting questions have
been posed that motivated an examination of the geographical distributions of SUP and TOR events as well
as LCL heights. Assume that the RB98 interpretation
of the importance of LCL height is correct. That is, low
LCLs favor formation of significant tornadoes because
they somehow alter the rear-flank downdraft and outflow
character of supercells. If this hypothesis is correct, it
should apply regardless of geographical location. It was
Parameter
Optimal HSS
Optimal value
EHI
SRH 0–1
EHI 0–1
VGP
SRH
Mean shear
BL–6 km shear
CAPE
0.386
0.356
0.355
0.315
0.268
0.227
0.189
0.179
1.40
180
0.72
0.280
230 m 2 s22
0.009 63 s21
20.5 m s21
2100 J kg21
found that, in the 1992 dataset, SUP events rarely occurred east of the Mississippi River (Fig. 6) but are
relatively much more common in the High Plains. Thus,
it is impossible to test the hypothesis in the eastern
United States, and the null hypothesis (LCL does not
have value at predicting SUP vs TOR) is permitted in
that region. Further, when the LCL heights associated
with all SUP and TOR soundings combined are examined geographically (Fig. 7), it can be seen that, when
severe storms occur, the LCL is climatologically large
in the High Plains owing to the typically drier boundary
layer. It could be argued that the RB98 result simply
showed that large LCL heights and storms with large
hail occur together in the High Plains, but the reasons
have little to do with the occurrence or nonoccurrence
of significant tornadoes. In fact, recent work by Thompson et al. (2002c) has demonstrated that about 90% of
the storms that produce large hail (RB98 2-in. criterion)
are supercells, but nearly 2/3 of radar-identified supercells do not produce hail this large. This, too, argues
that it is possible that the RB98 finding regarding low
LCL height had more to do with environments that support large hail occurrence at the ground than promoting
the occurrence of significant tornadoes.
This geographical analysis points the way to some
FIG. 6. Geographical distribution of TOR (squares) and SUP (diamonds) soundings. Symbol size varies linearly with number of
soundings, with the largest symbol on the graph representing 12
soundings.
534
WEATHER AND FORECASTING
VOLUME 18
utility. Among these is a modified measure of SRH from
the 0–1-km layer, and the associated EHI based on this
same layer. Further, it was shown that CAPE that accrues
below 3 km AGL might be important in discriminating
between environments supportive of significant tornadoes and those that are not. Information was also presented that points toward the need for further research
into the possible bias in the significant tornado climatology that might be present because tornadoes from
high-based storms appear weaker and occur in geographical regions in which damage is much less likely
to occur.
FIG. 7. Distribution of average LCL height for all SUP and TOR
soundings at each sounding site. The symbol sizes vary linearly from
zero to the maximum average LCL height observed (3477 m). Gray
symbols represent sites with #median number of soundings (three).
badly needed additional research to add more knowledge regarding the use of the LCL as a tornado forecast
parameter. In particular, a large climatological dataset
of identified supercells, their tornadic character, and
low-level humidity must be assembled to further test
low-level humidity as a forecast tool. The work of
Thompson et al. (2002b) makes much progress in this
direction, because it uses radar supercell identification
instead of the RB98 large-hail proxy. Their work supports the finding that low LCL height favors significant
tornadoes.
However, the problem of tornado intensity assessment
in the high-LCL environment remains unsolved. When
assessing tornado character, it will be necessary to devise a means to determine tornado strength rather than
relying on the assumption that tornadoes occurring with
higher-based storms are small and weak simply because
of their less menacing aspect ratios, or because they did
not do damage in areas where no structures were present
to be damaged.
Further research also is required as to what layer is
most important in terms of SRH computation for the
tornado forecasting problem. The 0–1-km AGL layer
was identified herein, but it is possible that a more shallow layer is actually key. However, reported National
Weather Service soundings do not have sufficient vertical resolution to quantify near-ground SRH adequately.
This note demonstrates that near-ground CAPE is important in distinguishing between soundings associated
with significant tornadoes and those associated with
large hail but not significant tornadoes. Whether this
CAPE is physically important, or is more indirectly related to physically important quantities such as (possibly) low-level humidity, remains to be determined.
Further, the optimum near-ground layer for the computation of CAPE has yet to be determined.
In summary, this note has demonstrated that certain
forecasting parameters can be modified to improve their
Acknowledgments. Partial support for this research
was provided by the National Science Foundation
through Grants ATM-9617318 and ATM-0003869. Additional support was provided by NOAA through a Presidential Early Career Award for Scientists and Engineers. Helpful insights were provided by Drs. Harold
Brooks of NSSL, Charles Doswell of CIMMS, Jerry
Straka of the University of Oklahoma, and Paul Markowski of The Pennsylvania State University. The examination of the geographical distribution of the TOR
and SUP events, as well as LCL heights, was motivated
by the interesting research conducted by Mr. Kyle Beatty
at the University of Oklahoma. Richard Thompson and
Roger Edwards of the NOAA Storm Prediction Center
provided much valuable information regarding their recent research using RUC-2 soundings and radar-identified supercells.
REFERENCES
Craven, J. P., H. E. Brooks, and J. A. Hart, 2002: Baseline climatology
of sounding derived parameters associated with deep, moist convection. Preprints, 21st Conf. on Severe Local Storms, San Antonio, TX, Amer. Meteor. Soc., 643–646.
Davies-Jones, R. P., D. Burgess, and M. Foster, 1990: Test of helicity
as a tornado forecast parameter. Preprints, 16th Conf. on Severe
Local Storms, Kananaskis Park, AB, Canada, Amer. Meteor.
Soc., 588–592.
Doswell, C. A., III, R. Davies-Jones, and D. L. Keller, 1990: On
summary measures of skill in rare event forecasting based on
contingency tables. Wea. Forecasting, 5, 576–585.
Edwards, R., and R. L. Thompson, 2000: RUC-2 supercell proximity
soundings, Part II: An independent assessment of supercell forecast parameters. Preprints, 20th Conf. on Severe Local Storms,
Orlando, FL, Amer. Meteor. Soc., 435–438.
Hart, J. A., and W. Korotky, 1991: The SHARP workstation v1.50
users guide. National Weather Service, NOAA, U.S. Department
of Commerce, 30 pp. [Available from NWS Eastern Region
Headquarters, 630 Johnson Ave., Bohemia, NY 11716.]
Markowski, P. M., E. N. Rasmussen, and J. M. Straka, 1998a: The
occurrence of tornadoes in supercells interacting with boundaries
during VORTEX-95. Wea. Forecasting, 13, 852–859.
——, ——, ——, and D. C. Dowell, 1998b: Observations of lowlevel baroclinity generated by anvil shadows. Mon. Wea. Rev.,
126, 2942–2958.
——, J. M. Straka, and E. N. Rasmussen, 1998c: A preliminary
investigation of the importance of helicity ‘location’ in the hodograph. Preprints, 19th Conf. on Severe Local Storms, Minneapolis, MN, Amer. Meteor. Soc., 230–233.
——, ——, ——, and D. O. Blanchard, 1998d: Variability of storm-
JUNE 2003
NOTES AND CORRESPONDENCE
relative helicity during VORTEX. Mon. Wea. Rev., 126, 2959–
2971.
McCaul, E. W., Jr., 1991: Buoyancy and shear characteristics of hurricane–tornado environments. Mon. Wea. Rev., 119, 1954–1978.
——, and M. L. Weisman, 1996: Simulations of shallow supercell
storms in landfalling hurricane environments. Mon. Wea. Rev.,
124, 408–429.
Rasmussen, E. N., and D. O. Blanchard, 1998: A baseline climatology
of sounding-derived supercell and tornado forecast parameters.
Wea. Forecasting, 13, 1148–1164.
Rotunno, R., and J. B. Klemp, 1982: The influence of the shearinduced pressure gradient on thunderstorm motion. Mon. Wea.
Rev., 110, 136–151.
Straka, J. M., and E. N. Rasmussen, 1998: Thirty years of cloud
modeling: Does the emperor wear clothes? Preprints, 19th Conf.
on Severe Local Storms, Minneapolis, MN, Amer. Meteor. Soc.,
342–347.
Thompson, R. L., and R. Edwards, 2000: A comparison of Rapid
535
Update Cycle 2 (RUC-2) model soundings with observed soundings in supercell environments. Preprints, 20th Conf. on Severe
Local Storms, Orlando, FL, Amer. Meteor. Soc., 551–554.
——, ——, and J. A. Hart, 2002a: Evaluation and interpretation of
the supercell composite and significant tornado parameters at the
Storm Prediction Center. Preprints, 21st Conf. on Severe Local
Storms, San Antonio, TX, Amer. Meteor. Soc., J11–J14.
——, ——, and ——, 2002b: An assessment of supercell and tornado
forecast parameters with RUC-2 model close proximity soundings. Preprints, 21st Conf. on Severe Local Storms, San Antonio,
TX, Amer. Meteor. Soc., 595–598.
——, ——, ——, and K. L. Elmore, 2002c: RUC-2 model analysis
soundings as a surrogate for observed soundings in supercell
environments. Preprints, 21st Conf. on Severe Local Storms, San
Antonio, TX, Amer. Meteor. Soc., 591–594.
Wicker, L. J., 1996: The role of near surface wind shear on low-level
mesocyclone generation and tornadoes. Preprints, 18th Conf. on
Severe Local Storms, San Francisco, CA, Amer. Meteor. Soc.,
115–119.