ATMOSPHERIC SCIENCE LETTERS
Atmos. Sci. Let. (2007)
Published online in Wiley InterScience
(www.interscience.wiley.com) DOI: 10.1002/asl.159
Assessment of the severe weather environment in North
America simulated by a global climate model
Patrick T. Marsh,1* Harold E. Brooks2 and David J. Karoly3
1University of Oklahoma, 120 David L. Boren Blvd, Norman, OK 73071, USA
2National Severe Storms Laboratory, 120 David L. Boren Blvd, Norman, OK 73071, USA
3School of Earth Sciences, University of Melbourne, VIC 3010, Australia
*Correspondence to:
Patrick T. Marsh, University of
Oklahoma, 120 David L. Boren
Blvd, Norman, OK 73071, USA.
E-mail: patrick.marsh@ou.edu
Received: 21 May 2007
Revised: 6 September 2007
Accepted: 10 September 2007
Abstract
Annual and seasonal cycles of convectively important atmospheric parameters for North
America have been computed using the Community Climate System Model version 3
(CCSM3) Global Climate Model using a decade of CCSM3 data. Results for the spatial and
temporal distributions of environments conducive to severe convective weather qualitatively
agree with observational estimates from NCAR/NCEP global reanalyses, although the model
underestimates the frequency of occurrence of severe weather environments. This result
demonstrates the possibility for future studies aimed at determining possible changes in
the distribution of severe weather environments associated with global climate change.
Copyright  2007 Royal Meteorological Society
Keywords: climatologies; severe convection; climate variability
1. Introduction
The value of climate change predictions hinges on the
understanding of current climatologies and the correct
simulation of these climatologies by climate models.
Severe convective weather events (thunderstorms, hail,
tornadoes, etc.) are relatively rare, but high-impact
atmospheric phenomena at any location due to their
very small temporal and spatial scales. Consequently,
assessing climatologies of actual severe convective
weather events is difficult. Inconsistencies in reporting
criteria and improvements in the technology used to
observe severe weather make the problem of develop-
ing reliable long-term climatologies of severe weather
events nearly impossible (Brooks and Doswell, 2001;
Doswell et al., 2005).
Brown and Murphy (1996) and Brooks et al. (2003)
proposed the use of environmental conditions as
covariates for the occurrence of weather events that
could not be accurately quantified. In each of these
studies, extreme values of the covariates were closely
related to the average occurrence of the weather event
in question. Environmental conditions conducive to
the occurrence of severe weather can be quantified
from meteorological soundings in terms of the con-
vective available potential energy (CAPE) and vertical
shear of the horizontal wind. In the context of estab-
lishing climatologies of severe convective weather
events, the problem is altered from trying to assess an
inherently limited database of observed severe convec-
tive weather events to trying to establish a correlation
between better observed environmental conditions and
the original events in question.
It has been shown that most convective parameters
derived from the NCAR/NCEP global reanalysis
(Kalnay et al., 1996, hereafter simply referred to
as reanalysis) are qualitatively similar to convective
parameters derived from observed soundings (Lee,
2002). Niall and Walsh (2005) showed that CAPE cal-
culated from reanalysis showed a strong relationship to
incidents of hail. Brooks et al. (2003) calculated CAPE
values using the mixed layer within the lowest 100 hPa
of the atmosphere and shear values over the 0–6 km
range (hereafter referred to as deep-layer shear). They
concluded that the higher the CAPE and shear, the
greater the probability became that the environmental
conditions would be associated with severe convec-
tive weather. This is consistent with the results of
Rasmussen and Blanchard (1998) who used observed
environmental parameters from neighboring meteoro-
logical soundings. One caveat mentioned by Brooks
et al. (2003) is that the difference in tornadic and
nontornadic environments is harder to differentiate in
the reanalysis environments than observed environ-
ments. This result is also consistent with Rasmussen
and Blanchard (1998).
Brooks et al. (2007) explored the annual cycle of
the severe convective weather environments from the
reanalysis. Qualitatively, it was found that the reanal-
ysis data capture the spatial and temporal variability
observed in the United States (Brooks et al., 2003;
Doswell et al., 2005). Brooks et al. (2003) concluded
that the central part of the United States was the region
with the most frequent number of days with favor-
able severe convective weather environments. This
ability of the reanalysis data to recreate severe con-
vective weather environments, at least qualitatively,
points toward the possibility that atmospheric models
of similar resolution are capable of doing the same.
Copyright  2007 Royal Meteorological Society

P. T. Marsh, H. E. Brooks and D. J. Karoly
Typically, assessments of possible climate change
impacts resulting from changes in anthropogenic forc-
ing are based or derived from global climate models.
Previously, analyses derived from global climate mod-
els have had too coarse a resolution – on the order of
greater than 200 km – to simulate severe convective
weather environments. The resolution of these global
climate models is even too coarse to adequately sim-
ulate mesoscale environments.
The Intergovernmental Panel on Climate Change
(IPCC et al., 2001) has been unable to reach conclu-
sions on climate change projections for severe weather
due to anthropogenic forcing. The IPCC Third Assess-
ment’s chapter on model projections of future climate
change, specifically the subsection on precipitation and
convection, states.
‘Although of great importance to society for their
potential for causing destruction, as well as their
human and economic impacts, there is little guidance
from AOGCMs concerning the future behavior of
tornadoes, hail or lightning. This is because these
phenomena are not explicitly resolved in AOGCMs,
and any studies that have been done have had to rely
on empirical relationships between model features and
the phenomenon of interest . . . However, there have
been no recent studies examining this problem with the
current generation of global climate models. Due to the
fact that these severe weather phenomena are sub-grid
scale (even more so than discussed below for tropical
cyclones), and that second and third order linkages
between model output and empirical relationships for
limited regions must be used to derive results, we
cannot reach any definitive conclusions concerning
possible future increases in hail and lightning, and
there is no information from AOGCMs concerning
future changes in tornado activity.’ (Cubasch et al.,
2001).
Recent improvements in global climate model res-
olution, and the declining cost of data storage, have
enabled the archiving of high-resolution model out-
put. Currently, several runs of the high-resolution
NCAR Community Climate System Model version 3
(CCSM3) are archived at 6 h resolution, 26 vertical
levels, and approximately, a 1.4◦ horizontal resolution.
Even with the advances made in global climate
models, these models are still incapable of resolving
actual severe weather events, as these events occur at
scales that are still well below the horizontal resolution
within the models. As a result, assessing the distribu-
tion of severe weather within a global climate model
is still limited to assessing environments associated
with severe convective weather. Here, an assessment
is presented of how the CCSM3 global climate model
represents the severe weather environment and, in turn,
if the severe weather environments of modern global
climate models can be used as a covariate for esti-
mating future distributions of observed severe weather
events.
Preliminary results are presented from an investiga-
tion of the ability of the CCSM3 to simulate severe
convective weather environments of North America.
The model severe weather environments are compared
with the severe weather environments from reanaly-
sis data discussed in Brooks et al. (2003). This serves
as the basis for future analyses aimed at describing
changes in the severe convective weather environment
under different future climate change scenarios.
2. Description of the climate model
The CCSM3 is a coupled global climate model con-
sisting of atmosphere, land surface, sea-ice, and ocean
components (Collins et al., 2006). Each component
is a model in itself joined together through a flux
coupler. The archived high-resolution runs include a
control run (no changes in external climate forcing),
a 20th century simulation (containing the observed
changes of greenhouse gases, sulphate aerosols, vol-
canic aerosols, and solar irradiance from the 20th
century), and 21st century ‘future scenarios’ (contain-
ing estimated changes in greenhouse gas concentration
and aerosol concentrations). For this particular study,
20 years (1980–1999) of a simulation initialized in
1870 and run through the 20th century (b30.030e) was
chosen in an effort to assess how well climate models
can simulate current severe weather environments.
The atmospheric portion of the CCSM3, the Com-
munity Atmospheric Model (CAM3), is a spec-
tral model with 85-wavenumber triangular truncation
(approximately 1.4◦ at the equator) in the horizontal
with 26 terrain-following hybrid levels in the verti-
cal (Collins et al., 2006). Specifically, CAM3 vertical
resolution contains four levels below 850 hPa and 13
levels above 200 hPa (topmost being 2.2 hPa). CAM3
output fields are archived every 6 h. Fields used in
calculating CAPE are the 3-dimensional fields of tem-
perature (T), mixing ratio (Q), geopotential height (Z3)
and pressure (P). Additionally, surface geopotential
and surface pressure are necessary. It should be noted
that in calculating the CAPE fields, the model data
were used on its vertical grid and not interpolated.
Additionally, the CAPE was computed by taking the
potential temperature and specific humidity of the low-
est model layer and averaged over a depth of 500 m.
The resulting sum of the positive buoyancy of an air
parcel raised from the surface layer from the lever of
free convection (LFC) to the equilibrium level (EL) is
defined as CAPE.
The NCAR/NCEP Global Reanalysis (Kalnay et al.,
1996) data resolution is approximately 1.9◦ in both
latitude and longitude, and 28 levels in the vertical,
which is roughly equal to the resolution of the
CCSM3. Brooks et al. (2003) demonstrated that the
reanalysis data reproduce CAPE and deep-layer shear
environments reasonably accurately. Thus, evaluating
the CCSM3 against the reanalysis data was a natural
choice. It is important to note that the reanalysis
data are more robust as calculations were made using
Copyright  2007 Royal Meteorological Society Atmos. Sci. Let. (2007)
DOI: 10.1002/asl