INTERNATIONAL JOURNAL OF CLIMATOLOGY

Int. J. Climatol. 19: 607–617 (1999)

A GCM INVESTIGATION OF GLOBAL WARMING IMPACTS

RELEVANT TO TROPICAL CYCLONE GENESIS
LEONARD M. DRUYANa,*, PATRICK LONERGANb and TIMOTHY EICHLERc
a
Center for Climate Systems Research, Columbia Uni6ersity and NASA/Goddard Institute for Space Studies, 2880 Broadway,
New York, NY 10025, USA
b
Science Systems and Applications, Inc. and NASA/Goddard Institute for Space Studies, New York, USA
c
Geological Science Department, Columbia Uni6ersity, New York, USA

Recei6ed 8 April 1998

Re6ised 21 October 1998
Accepted 11 No6ember 1998

ABSTRACT
Two approaches that consider how greenhouse warming might impact the frequency of tropical cyclone (TC) genesis
are explored. Results are based on GCM experiments with the q-flux version global climate model of the
NASA/Goddard Institute for Space Studies (GISS); one set representing contemporary atmospheric concentrations of
CO2, contrasting with the second set representing the global climate in double CO2 equilibrium. July – September
means of climate parameters relevant to TC genesis are computed from the simulations and combined to formulate
a seasonal genesis parameter (SGP), as suggested in an empirical study by Gray (in Shaw, D.B. (ed.), Meteorology
O6er the Tropical Oceans, 1979, pp. 155–218). The spatial distribution of the July – September SGP based on the
control simulations is compared with the observed distribution and results using other models. The corresponding
spatial distribution of the July–September SGP derived from the double CO2 simulations, when compared with the
control results, projects a 50% increase in the genesis frequency of TC over the western North Atlantic/Gulf of
Mexico basin, but 100–200% increases over the North Pacific Ocean. The increases, most of which are attributable
to enhanced ocean temperatures, may be exaggerated, suggesting that the original SGP formulation requires tuning
or other revisions. For example, it is noted that SGP computed from the NCEP 1982 – 1994 re-analysis climatology
do not accurately reflect the known spatial distributions of TC genesis frequency. The second approach detects
easterly waves over the eastern North Atlantic Ocean by spectral analysis of vorticity and wind component time
trends, comparing wave activity in the control and double CO2 simulations. Results indicate a southward shift in
future trajectories of easterly waves over West Africa and significant increases in their average amplitude as they cross
the African coast and begin to traverse the Eastern Atlantic along 14°N. Copyright © 1999 Royal Meteorological
Society.
KEY WORDS: global warming; climate change; tropical storm frequency; GISS GCM

1. INTRODUCTION

Tropical cyclones (TC) become named tropical storms when the sustained winds associated with
deepening pressure minima in the tropics first exceed 17 m s − 1. The extreme wind and storm surge
property damage and related loss of life accompanying landfall of TC have motivated a keen interest in
forecasting their genesis, intensity and their trajectories. Moreover, several sectors of government and
industry have turned to the scientific community for guidance regarding possible changes in the
climatology of TC spurred by anticipated future increases in surface air temperatures and sea surface
temperatures (SST).

* Correspondence to: Center for Climate Systems Research, Columbia University, 2880 Broadway, New York, NY 10025, USA.

Contract/grant sponsor: Risk Prediction Initiative of the Bermuda Biological Station for Research; Contract/grant number: RPI
96-008
Contract/grant sponsor: National Science Foundation; Contract/grant number: ATM-9725142

CCC 0899–8418/99/060607 – 11$17.50

Copyright © 1999 Royal Meteorological Society
608 L.M. DRUYAN ET AL.

The 1995 Intergovernmental Panel on Climate Change report concludes that ‘ . . . it is not possible to
say whether the frequency, area of occurrence, time of occurrence, mean intensity or maximum intensity
of tropical cyclones will change . . . ’ (Kattenberg et al., 1996), as the global climate warms from the
radiative forcing of increasing atmospheric concentrations of greenhouse gases. More recently, Hender-
son-Sellers et al. (1998) stated that while global frequencies of TC may not change, regional frequencies
of TC genesis ‘ . . . could change substantially in either direction.’
Recognizing how important it is for risk assessment to anticipate future trends in tropical storm
behavior (because of the tremendous destructive potential of TC), the authors have addressed this
question as it applies to the northern hemisphere summer season, using simulations of the 9-layer general
circulation model of the NASA/Goddard Institute for Space Studies (GISS GCM) (Hansen et al., 1996).
In considering how greenhouse warming might impact the frequency of TC genesis, an approach
previously undertaken with the CSIRO9 GCM (Ryan et al., 1992; Watterson et al., 1995) was pursued in
order to gauge the importance of model dependence on the results. In this context, the authors evaluate
how each of several relevant modeled and analyzed atmospheric parameters impacts estimates of TC
frequency and how these components of the climate system change in a double CO2 equilibrium
simulation. Additionally, they examine the impact of the double CO2 environment on the generation of
easterly wave disturbances near the Atlantic coast of West Africa.
The results do not offer definitive answers. They represent an exploration of two approaches to the
problem. First, previous research which discusses the association of TC with seasonal climatic means as
well as the application of GCMs for projecting climate change impacts on TC behavior are reviewed. For
the purposes of this paper, tropical cyclones, named tropical storms, hurricanes and typhoons collectively,
will be referred to as TC.

2. BACKGROUND

Gray (1979) related spatial distributions of seasonal and yearly averages of six climatological indexes to
the spatial distribution of named tropical storms’ genesis frequency. In particular, he showed that the
spatial distribution of 1958 – 1977 means of a ‘genesis parameter’ (GP) were highly correlated to the
observed distribution of TC genesis frequency per 20 years within each 5° square area. The dependence
of TC genesis frequency on these particular indexes, discussed below, has more recently been sustained by
authoritative sources (Lighthill et al., 1994; McBride, 1995).
Given the prominence of ocean temperature in the aforementioned discussions, anticipated warming of
the tropical oceans suggests that future tropical cyclone activity could be different than in the current
climate. However, other environmental factors are also relevant and must be considered. GCMs are an
established tool for most projections of climate change, but the horizontal resolution of most GCMs used
in such research is too coarse to simulate the dynamics of a deepening TC (Lighthill et al., 1994).
Accordingly, there is a reluctance to estimate how expected future climate changes will affect the
occurrence of TC.
Bengtsson et al. (1995) identified TC vortices in simulations of the ECHAM3 global general circulation
model run at T106 (1.1°) horizontal resolution. Although the 5 year simulation did not include
interannually varying SSTs, it nevertheless produced a variable number of TC during each season, with
realistic frequencies and geographical distributions. The same model run, at a more conventional
horizontal resolution (T42), also created ‘hurricane-type’ vortices, but with more diffuse gradients and
wider horizontal dimensions. They were not able to compute a meaningful model analog for the index of
Gray (1979) from their simulations because its variability in the simulations was apparently too great. On
the other hand, SST\ 26°C proved to be a necessary condition for their modeled TC analogs. Seasonal
TC frequency in the GCM appeared to be sensitive to the modeled strength of the Hadley circulation,
with more TC during seasons of stronger implied large-scale uplift over tropical oceans.
In the sequel investigation, Bengtsson et al. (1996) monitored similar tropical vortices in a parallel 5
year double CO2 simulation with the same model. SST anomalies for the lower boundary condition came

Copyright © 1999 Royal Meteorological Society Int. J. Climatol. 19: 607 – 617 (1999)
 A GCM INVESTIGATION OF GLOBAL WARMING IMPACTS 609

from Year 60 of a transient run of the fully coupled ECHAM3 ocean–atmosphere GCM (Cubasch et al.,
1992) run at a much coarser horizontal resolution (T21) with monotonic increases in CO2. At Year 60 in
the coupled model run, the global mean surface air temperature had increased by only 1°C and the largest
SST increases were 0.5 – 1.5°C. They found that the global distribution of simulated TC agreed with that
of the present climate, but compared with the control, the number of TC was significantly reduced,
particularly in the southern hemisphere. Changes were attributed to a weakening of the Hadley circulation
and stronger subtropical upper air westerlies, which increased vertical wind shears over TC genesis areas.
The implications of this study have been debated (Landsea, 1997; Henderson-Sellers et al., 1998) and are
discussed in the ‘Discussion and conclusion’ section.

3. GCM-DERIVED GENESIS PARAMETER

The authors suggest that GCM simulations are suitable for analysis of the meteorological/climatological
conditions favorable for TC genesis, such as those discussed by Gray (1979), Lighthill et al. (1994) and
McBride (1995). Deducing the evolution of climatic conditions that are correlated with TC frequency
from simulations with a range of climate forcings implies projections of TC frequency compatible with
each scenario. Gray’s ‘genesis parameter’ (GP) has six components:
(i) ocean thermal energy, proportional to SST minus 26°C;
(ii) moist stability, proportional to the absolute value of the lapse rate of equivalent potential tempera-
ture between the surface and 500 mb;
(iii) relative humidity, proportional to the mean relative humidity between 700–500 mb, scaled between
0–1 for values ranging 40 – 70%;
(iv) vertical wind shear, proportional to the inverse of the absolute value of the wind shear between
950–200 mb;
(v) vorticity of the near surface circulation;
(vi) Coriolis parameter.
The first three of these comprise a ‘thermal potential’, while the second three combine to form a ‘dynamic
potential’. Gray used seasonal climatologies to compute spatial distributions of indexes based on each of
the six components. The indexes were tuned so that their multiplication yielded a seasonal genesis
parameter (SGP) indicating the frequency of TC genesis for each 5° square per 20 years. Thus, each
component serves as a necessary condition for TC genesis, since SGP will not exceed zero unless all of the
individual components are positive. The spatial distribution of SGP based on Gray’s data from
1958–1977 is remarkably similar to the distribution of observed TC frequency for that period.
Ryan et al. (1992) computed a yearly GP (YGP) for a 15 year double CO2 equilibrium simulation with
the CSIRO9 GCM, with 9-level vertical and R21 (3.2°× 5.6°) horizontal resolution. SST evolution was
computed for a slab ocean with prescribed ocean heat convergence. The global mean surface temperature
increase for this 2 × CO2 run was a rather large 4.8°C compared with the 15 year 1 × CO2 control run,
whose climatology was also used to generate a YGP distribution. YGP based on the control simulation
was globally 37% higher than Gray’s YGP total, the excess being attributed in large measure to
overestimates of ocean thermal energy. The spatial distribution of YGP computed from the control
climatology was overall, reasonable, although several discrepancies from observations were noted. For
example, the model indicated YGP between 1–5 (TC per 5° square per 20 years) in regions where TC
genesis is not observed, such as along the tropical coastlines of the South Atlantic. Results for the western
North Pacific and Australia vicinity were skillful, but YGP was underestimated in the eastern North
Pacific and the North Atlantic maximum was displaced westward into the Caribbean. The global sum of
predicted TC genesis for the 2× CO2 experiment was about three times the control value. Maxima more
than doubled in the western North Pacific and Australia vicinity. The area of predicted genesis in the
North Atlantic spread eastward to Africa compared with the control results. These increases in YGP due
to global warming resulted from the model’s predicted increases in SST, which increased the ocean
Copyright © 1999 Royal Meteorological Society Int. J. Climatol. 19: 607 – 617 (1999)
610 L.M. DRUYAN ET AL.

thermal energy and destabilized the lower troposphere. No significant changes in vorticity or vertical wind
shear were evident between the experiment and control runs.
YGP was also computed from simulations with the ARPEGE GCM (Centre National Recherches
Météorologiques, France), a control with Atmospheric Model Intercomparison Project (AMIP) SST and
two 2× CO2 experiments (Royer et al., 1998). The model has 30 vertical levels and an equivalent
horizontal resolution of about 2.8°. SST for the 2× CO2 were simulated by two different coupled
atmosphere–ocean models that were forced by transient increases of atmospheric CO2 concentration.
Although the global mean increase of surface air temperature for the 2×CO2 simulations was less than
2°C, a near doubling of YGP in the northern hemisphere resulted, influenced mostly by higher ocean
thermal energy and hardly at all by changes in dynamic potential.
Watterson et al. (1995) showed that YGP derived from the European Center for Medium-Range
Weather Forecasts (ECMWF) gridded observational analyses were extremely sensitive to modifications in
the ECMWF GCM. Indeed, the average yearly global number of TC predicted by 1985–1989 data (74)
was less than half the number for 1980 – 1984 (184). In particular, the humidity term during the later years
was on average about one-third of its value up until 1984.
Watterson et al. (1995) also computed seasonal genesis parameters (SGP) from ECMWF gridded
observational analyses, 1985 – 1989, as well as from the seasonal climatologies of 10 year CSIRO9 model
simulations (3.2°× 5.6° horizontal resolution) forced by observed AMIP SST. The ECMWF SGP for
July–September showed a 0.57 correlation over ocean points 40°S–40°N with the record of TC frequency
distribution observed 1967 – 1986, while the correlation coefficient between model results and the same
observations was only 0.44. In particular, model results underestimated TC activity in the tropical North
Pacific by about 60%. In the North Atlantic, the model SGP was only slightly high, but the mid-ocean
maximum was not diagnosed, the higher frequencies being confined to the Caribbean sector.
The central North Pacific had more TC activity in June–September 1982 than in the following year.
Watterson et al. (1995) forced the CSIRO9 model with SST for those seasons and computed the
corresponding SGP that successfully reflected the observed relative surplus of TC activity in JAS 1982.
They found that both the dynamic and thermal components were important in establishing the
interannual differences in SGP. Time series of the JAS SGP, 1979–1988, from model runs were correlated
with corresponding observations best in the central North Pacific (0.70), moderately in the North Atlantic
(0.56) and more poorly in the western (0.23) and eastern (0.41) North Pacific. They also found that the
modeled SGP time series for January – March was not a useful indicator of observed TC interannual
variability (mostly in the southern hemisphere).
In a more recent study, January spatial distributions of SGP were computed from simulations with a
limited area model run at 125 km horizontal resolution over a domain centered on Australia (Walsh and
Watterson, 1997). These SGP arrays were found to bear some relationship with the spatial distributions
of explicitly modeled TC genesis frequency, but the correlation was limited, only 0.26 for the strongest
tropical cyclone-like vortices. For example, the model generated a rather high frequency of TC over the
South China Sea where SST did not generally exceed the 26°C threshold that makes SGP non-zero.
Moreover, SGP patterns based on the modeled climate were better correlated with observed TC patterns
than the explicitly modeled frequency of tropical cyclone-like vortices.

4. GISS GCM STUDY

The authors used the 9-layer general circulation model of the NASA/Goddard Institute for Space Studies
(GISS GCM) (Hansen et al., 1996), running on a Cartesian grid with 4° latitude by 5° longitude
horizontal resolution. A control model climate for July–September (JAS) was computed from the last 30
years of a 50 year simulation. The moist convection parameterization of the GISS GCM (Del Genio and
Yao, 1993), which is very relevant to this application, specifies a vertical mass flux proportional to the
moist static instability. Cumulus mass fluxes are constrained to relax the atmosphere to a neutrally stable
state at the cloud base.

Copyright © 1999 Royal Meteorological Society Int. J. Climatol. 19: 607 – 617 (1999)
 A GCM INVESTIGATION OF GLOBAL WARMING IMPACTS 611

Figure 1. Distribution of July–September mean SSTs (°C) computed by the GISS GCM ‘q-flux’ ocean formulation for years 21 – 35
of the control (1×CO2) simulation

Figure 2. Pressure – longitude cross-section of atmospheric temperature impacts, 2 ×CO2 minus 1 ×CO2, averaged over 14 – 18°N
for multi-year July – September means (°C)

This simulation was made with a mixed layer ocean lower boundary (Russell et al., 1985) in which
vertical and horizontal heat transports are prescribed to be consistent with the seasonal march of
climatological SST. The mixed layer ocean formulation allows SST evolution to be driven by simulated
energy balances at the ocean – atmosphere boundary, resulting in realistic equilibrium SST distributions,
as for example the July – September mean shown in Figure 1. The 30 year ‘q-flux’ run serves as the control
for a 20 year double CO2 (2 × CO2) equilibrium experiment (the last 20 years of a 50 year simulation).
Characteristics of this 2× CO2 simulation were previously described by Rind (1998). Atmospheric
warming is represented by an increase of 3.5°C in the annual global mean surface temperature compared
with the control climate, while this increase was 3.7°C for July–September. Figure 2 shows a height–lon-
gitude cross-section of July – September temperature impacts along 14–18°N. Corresponding SST in-
creases for the 2× CO2 simulation ranged between 2 and 3°C (Figure 3). The global mean annual increase
in SST was 2.3°C and the global mean increase in precipitable water was 6.3 mm or 27% (Rind, 1998).
In an effort somewhat parallel to the studies of Ryan et al. (1992) and Watterson et al. (1995), the
authors computed the July – September SGP from the 20 year 2 × CO2 equilibrium experiment and the 30
year control. They offer here brief highlights that can be compared with the previously published results

Figure 3. Impact of 2 ×CO2 on SSTs (°C, 2 ×CO2 minus 1 ×CO2)

Copyright © 1999 Royal Meteorological Society Int. J. Climatol. 19: 607 – 617 (1999)
612 L.M. DRUYAN ET AL.

using the CSIRO9 GCM. The ineffectiveness of this approach for diagnosing TC genesis over the central
tropical North Atlantic Ocean (discussed below) provides impetus for the study discussed in Section 5
which examines the impact of the 2× CO2 regime on easterly wave generation.

4.1. SGP in control simulation

SGP based on the GISS GCM control simulation indicate a maximum in the Gulf of Mexico that
slightly overestimates observed TC genesis values given by Gray (1979), as did Gray’s SGP values based
on 1958–1977 climatological data (Figure 24 in Gray, 1979).
However, a second observed maximum in the central tropical North Atlantic is not predicted by the
model climatology, due in large measure to underestimates of near-surface vorticity in the Inter-tropical
Convergence Zone (ITCZ), GCM wind shears that are slightly higher than observed and model relative
humidities that seem to be some 10 – 15% too low along 10°N compared with Gray’s climatology (against
which the formulation for SGP was tuned; Gray, 1979). The authors found that NCEP reanalysis relative
humidities in a 1982 – 1994 climatology validate the GISS GCM values quite well, so it is not clear
whether the SGP formulation lacks tuning or whether climatological relative humidities are just uncertain.
In addition, the GCM simulated instability parameter is considerably weaker than that given by Gray
(1979) over most of the Atlantic. Watterson et al. (1995) reported wide interannual fluctuations in both
the humidity and instability terms based on ECMWF gridded analyses.
In summary, the GCM climatology ‘predicts’ a reasonable pattern of TC genesis over the Gulf of
Mexico and Caribbean Sea. However, over the central Tropical Atlantic, modeled vorticity is too weak,
humidity too low and wind shear excessive, leading to gross underestimates of TC genesis based on the
original SGP formulation. Of course, there is significant decadal variability of observed TC genesis
frequency (Landsea and Gray, 1992) that may be reflected by large variability in climate fields.
Calculations of TC genesis frequencies based on control simulations with the ARPEGE GCM incurred
similar underestimates over the central Tropical Atlantic, even when Gray’s thermal potential was
replaced by a newly formulated convective potential (Royer et al., 1998).
The spatial pattern of SGP based on the control simulation over the eastern quarter of the tropical
North Pacific Ocean was highly correlated (R= 0.81) with the observed pattern. However, the values of
SGP implied by the GCM climatology were somewhat underestimated. Over the tropical Western Pacific,
the July–September SGP derived from model variables bears a close resemblance to observed 1958–1977
frequencies of TC genesis, except that the modeled maximum is displaced to the northeast, probably
because of a similar displacement of the model’s representation of the ITCZ.

4.2. Double CO2 equilibrium impact on SGP

The authors next examine how the climate parameters relevant to TC genesis frequency change in the
2×CO2 simulation. There was no significant impact of 2× CO2 on July–September mean lower
tropospheric vorticity or wind shear in the Gulf of Mexico or the western Tropical Atlantic. However,
SST increased by about 2.5°C (Figure 3), causing a concomitant decrease in vertical thermal stability.
In an effort to calibrate the modeled values of SGP, a least squares linear regression relationship
between the simulated (control) and observed SGP patterns over the western section of the Atlantic basin
(95°–75°W) was computed. The resulting relationship was then used to scale the results of the 2× CO2
simulation. An additional regression for control results over the Tropical Eastern Pacific provided similar
scaling for projected 2×CO2 SGP in that region. No calibration was necessary for results over the
Tropical Western Pacific. The July – September distribution of these scaled SGP based on the 2× CO2
imply increases of 2 – 3 TC per 5° square per 20 years (about a 50% increase over the control values) in
the Gulf of Mexico and the western Tropical North Atlantic along 25°N. Over the North Pacific, 2× CO2
implied increases in TC frequency of 100 – 200%, most of which is related to model projected increases of
2 –3°C in SST. However, modest increases from 2× CO2 in the July–September mean near-surface
vorticity over the north-western Pacific Ocean also favored higher frequencies of TC genesis.
These projections of large increases in TC genesis frequency are probably gross overestimates because
the present SGP formulation may not be realistic for future climates. A qualitative analysis of 2× CO2

Copyright © 1999 Royal Meteorological Society Int. J. Climatol. 19: 607 – 617 (1999)
 A GCM INVESTIGATION OF GLOBAL WARMING IMPACTS 613

impacts on relevant climate components may, therefore, be more meaningful than the absolute changes in
SGP itself. This is discussed in more detail in Section 6.

5. SIMULATED EASTERLY WAVES IN THE EASTERN NORTH ATLANTIC

Motivated by the poor representation of tropical storm genesis in the eastern North Atlantic Ocean by
SGP, even for the current climate, the authors suggest another approach for evaluating how climate change
will impact the frequencies of tropical storm genesis in this region. Their analysis examines the impact of
2× CO2 on simulated easterly waves. The summary of the results is preceded by a background discussion.
Landsea and Gray (1992) relate hurricane activity in the North Atlantic with Sahel summer rainfall,
since African wave disturbances moving through the Sahel and out over the adjacent coastal waters are
often precursors of North Atlantic tropical cyclones. Reed et al. (1988) detected African and North
Atlantic wave tracks in ECMWF analyses by examining distributions of 3–5 day period filtered 700 mb
vorticity variances. Variance maxima indicate regions of large fluctuations of cyclonic and anticyclonic
vorticity, hence trajectories of transient cyclonic disturbances. Druyan and Hall (1994, 1996) described
African waves in the GISS GCM and showed the spatial distribution of transient wave spectral amplitudes
based on simulated 780 mb meridional wind time series. African waves in the GISS GCM generally had
longer periods than their observed counterparts.
Encouraged by these diagnostics for detecting easterly waves over West Africa and the North Atlantic,
the authors made relevant spectral analyses of the GCM-simulated vorticity and mid-tropospheric
meridional wind over the North Atlantic Ocean. Since this analysis is based on time series with daily
temporal resolution, archived monthly means from the GCM experiments described above could not be
used. They therefore repeated the July – October portion of the control and 2 × CO2 simulations, each from
eight arbitrarily different initial conditions, saving GCM climate data at 6 h intervals over the tropics.
Results based on means from the eight-run ensembles are discussed below for the spectral band of 3–6
day periods. The authors ascertained, however, that the major conclusions are not particularly sensitive
to small adjustments in the spectral interval. Figure 4A shows the variance of modeled vorticity at 780 mb
for 3–6 day period filtered time series for July–October in the control run. The maximum over West
Africa, representing the area of wave genesis, extends westward out over the Atlantic along 20–25°N,
similar to the ECMWF pattern analyzed by Reed et al. (1988). The corresponding distribution for the
2× CO2 simulation indicates a more narrow swath that is displaced southward to about 14°N, with higher
values of the variance from West Africa westward across most of the North Atlantic (Figure 4B). Positive
differences between the 2×CO2 and control simulations (Figure 4C) are generally 2–6-times the combined
standard deviations of the two eight-run ensembles, indicating statistical significance over a broad region
of the eastern tropical Atlantic Ocean, where so-called Cape Verde hurricanes form. Such increases in the
vorticity variance imply greater amplitudes of the average disturbance intensity and/or greater numbers of
transient wave traversals over the eastern North Atlantic in the warmer climate. Inspection of a number
of 3–6 day band pass filtered vorticity time series, one of which is shown in Figure 5, indicates that easterly
waves in the 2×CO2 have considerably larger amplitudes than in the control, but occur with about the
same frequency.
A similar conclusion can be deduced from examination of the spatial distribution of 3–6 day spectral
amplitudes based on July – October time series of the 780 mb meridional wind from the two sets of
simulations. Power spectra of this variable were previously used to monitor African wave disturbances
simulated by the GISS GCM (Druyan and Hall, 1994, 1996). Figure 6A shows a maximum over West
Africa, roughly corresponding to the location of the vorticity variance maximum, while Figure 6B shows
the organization of a stronger maximum at 14–18°N along the African Atlantic coast for the simulated
warmer climate. The swath of maximum 2×CO2 minus control differences (Figure 6C), statistically
significant along the West Africa coast and over the Cape Verde region of the North Atlantic, also implies
that the warmer climate experiences stronger 3–6 day period waves that move westward off the coast of
West Africa out into the North Atlantic. The center of maximum differences is farther north than the
Copyright © 1999 Royal Meteorological Society Int. J. Climatol. 19: 607 – 617 (1999)
614 L.M. DRUYAN ET AL.

Figure 4. A. Variance of modeled 780 mb vorticity (10 − 12 s − 2) for 3 – 6 day period band pass filtered time series for July – October
averaged over eight control (1 × CO2) simulations. B. Same as A, but for eight 2× CO2 simulations. C. Impacts of 2 ×CO2 on
vorticity variance (B − A). Whitened area highlights region for which differences are significant at the 95% or higher confidence level

corresponding maximum in Figure 4C because the strongest meridional winds associated with such
disturbances are often north of the vorticity maximum.

6. DISCUSSION AND CONCLUSION

Model projections of SST change were of considerable importance to the evaluations of possible increases
in TC activity over northern hemisphere oceans for the 2× CO2 climate. SST warming predicted by the

Figure 5. July – October time series of 780 mb vorticity for single simulations of 1 ×CO2 and 2 ×CO2 at 14°N, 25°W, band pass
filtered for 3–6 day periods. Vertical lines are placed at 4 day intervals

Copyright © 1999 Royal Meteorological Society Int. J. Climatol. 19: 607 – 617 (1999)
 A GCM INVESTIGATION OF GLOBAL WARMING IMPACTS 615

Figure 6. A. Spatial distribution of 3–6 day period spectral amplitudes (m s − 1) of 780 mb meridional wind component (V780)
July – October time series averaged from eight control (1 ×CO2) simulations. B. Same as A, but for eight 2 ×CO2 simulations. C.
Impacts of 2 ×CO2 on V780 spectral amplitudes (B − A). Whitened area highlights region for which differences are significant at the
95% or higher confidence level

GISS GCM 2 ×CO2 simulation (mostly 2 – 3°C) may be overestimated because the ‘q-flux’ modeling
approach does not account for negative feedbacks in ocean circulation that could mitigate the warming.
Still, even projections of smaller SST increases from coupled atmosphere–ocean models in another study
implied a near doubling of future TC genesis frequency (Royer et al., 1998).
The more than doubling of control frequencies of the seasonal TC genesis parameter over the North
Pacific based on the GISS GCM 2×CO2 simulation recalls similar results from the comparable CSIRO9
GCM (Ryan et al., 1992). Moreover, the authors also found, as they did, that most of the projected
increases in TC genesis frequency can be attributed to the predicted ocean warming and not to changes
in vorticity, wind shear or humidity. The exception is that for the GISS 2× CO2 scenario there were
positive contributions from increases in near surface vorticity east of the Philippines, a major breeding
ground for TC associated with a rather large displacement of the ITCZ from the equator. Perhaps,
therefore, the future climate will experience more frequent TC activity in this region partly owing to
stronger convergence (hence vorticity) into that segment of the ITCZ.
The current analysis found that the vorticity of easterly wave disturbances in the 2×CO2 simulations
was consistently and significantly stronger than in the control. This suggests that a warmer climate might
generate stronger incipient disturbances over West Africa and the adjacent North Atlantic, no doubt in
part owing to elevated SST. All other things being equal, the greater vorticity of each disturbance could
presage a more efficient intensification to tropical storm strength than in the current climate. In addition,
vorticity variance maxima, indicating preferred trajectories of African waves in the 2× CO2, were
simulated along 14°N. This represents a southward displacement of easterly waves’ tracks compared with
the control simulation.
A recent assessment (Henderson-Sellers et al., 1998) states that the environmental parameters that
appear important for TC genesis in the current climate will likely change to as yet undetermined
thresholds. The 26°C SST threshold adopted by Gray (1979) identifies areas of high ocean thermal energy,
but it also demarcates areas of vertical thermal instability that are prone to deep convection (Royer et al.,
Copyright © 1999 Royal Meteorological Society Int. J. Climatol. 19: 607 – 617 (1999)
616 L.M. DRUYAN ET AL.

1998; Evans, personal communication). Since the mid-troposphere will also be warmer in future climates,
instability could require higher SST thresholds as necessary conditions for TC genesis. Using the 26°C
threshold, therefore, probably also contributes to exaggerated estimates of TC genesis frequency implied
by SGP for both the current and the Ryan et al. (1992) 2 × CO2 scenarios. Accordingly, Royer et al.
(1998) found that increases in TC genesis frequency over the northern hemisphere for their 2× CO2
experiments were smaller when they replaced Gray’s thermal potential (including the 26°C threshold) with
a convective potential. Nevertheless, since the vertical moist stability component of SGP monitors the
impact of changes in the (equivalent potential) temperature lapse rate, even in its present formulation,
SGP cannot be large wherever the vertical thermal structure has been stabilized. It may, therefore, still be
relevant to refer to an as yet undetermined SST threshold to represent the minimum reservoir of ocean
thermal energy needed to power the storms.
The authors examined the relative warming at each vertical level experienced by their 2× CO2
simulation. Figure 2 shows a cross-section of 2×CO2 minus control temperature changes along 14–18°N,
traversing the centers of maximum observed TC genesis frequency in the North Atlantic and North
Pacific Oceans. Relative to the control, temperatures in the lowest model layer have warmed by about
3°C, compared with warming of about 4°C at 500 mb. This differential warming of the mean temperature
structure stabilizes the lower troposphere and undoubtedly inhibits dry convection. Rind (1998) found a
slight stabilization in the annual/global mean tropospheric lapse rate from − 6.01°C km − 1 to
−5.87°C km − 1 between the control and 2× CO2 simulations. However, the frequency of TC genesis is
more likely sensitive to the vertical lapse rate of the equi6alent potential temperature (Gray, 1979), which
accounts for the enhancement of convection by condensational heating. Accordingly, large increases in
near-surface specific humidity accompanying a future warmer climate make the lower troposphere
conditionally more unstable. Indeed, for this reason, the authors found the vertical stability component of
the SGP to be significantly more fa6orable for TC genesis in the 2× CO2 over the tropical oceans, despite
the simulated moderation in the vertical temperature lapse rate.
Henderson-Sellers et al. (1998) also questioned whether the lowered frequencies of TC reported by
Bengtsson et al. (1996) for the double CO2 climate might have been different if the SST had also been
predicted at the finer horizontal resolution. Landsea (1997) found it inconsistent that the hydrological
cycle speeded up with increasing atmospheric CO2 concentration using the low resolution coupled model
which provided the SST boundary conditions for the Bengtsson et al. (1996) simulation, whereas the
hydrological cycle slowed relative to the control in the high resolution experiment itself. Moreover, their
approach of counting explicitly simulated vortices also introduces uncertainties because, even at 1.1°, the
model’s horizontal resolution may be inadequate to mimic actual genesis processes that have been
observed to occur on scales of 50 km (Emanuel, personal communication). Finally, the SST forcing in this
global warming experiment (increases of 0.5 – 1.5°C) was rather small compared with the ocean warming
in the current GISS GCM double CO2 simulation. One reason is that the SST in the ECHAM3
experiment were not double CO2 equilibrium values, but rather values reached during the 60th year of a
scenario with monotonic increases in atmospheric CO2 concentrations. A similar experiment with the
GISS coupled atmosphere – ocean model (Russell et al., 1995) predicts SST increases by Years 64–75
(2× CO2) mostly between 1 – 1.5°C in the tropics, but with substantial areas warming to more than 1.5°C
in key hurricane genesis regions in the North Atlantic and the East North Pacific.
The authors computed the July – September SGP based on NCEP reanalysis data for 1982–1994. NCEP
reanalysis data are of course quite different and presumably more realistic than the distributions analyzed
by Gray (1979) for the years 1958 – 1977. SGP values are considerably lower based on these NCEP data
than SGP given by Gray (1979) in the North Atlantic and North Pacific Oceans. Seasonal mean values
of the relevant climate parameters from the NCEP spatial distributions were in many cases less favorable
to TC genesis according to the original criteria, as for example, the mean relative humidity from 700–500
mb. Over many key TC genesis areas, NCEP analyzed relative humidity is 15–20% lower than the values
Gray used to derive his SGP. This suggests that the index could be better tuned to reflect even the current
definitions of environmental seasonal means. A TC genesis index that could be derived from realistic
simulations of future climate scenarios should therefore
Copyright © 1999 Royal Meteorological Society Int. J. Climatol. 19: 607 – 617 (1999)
 A GCM INVESTIGATION OF GLOBAL WARMING IMPACTS 617

(i) have a more robust relationship with the observed variability of TC genesis frequency and
(ii) be as relevant to an overall warmer climate as it is to the present one.
Explicit simulation of more realistic tropical vortices should steadily improve as technology brings ever
increasing computing speed and memory, allowing computations to be made at higher and higher
resolution. Numerical simulation experiments with fewer flaws than heretofore may soon offer more
satisfying evidence regarding the impact of climate change on tropical cyclones. The present study can
serve as a benchmark against which future results should be compared.

ACKNOWLEDGEMENTS

This research was partially funded by Risk Prediction Initiative of the Bermuda Biological Station for
Research under Grant RPI 96-008 and the National Science Foundation under Grant ATM-9725142. The
support of the NASA Climate and Earth Observing System Programs is also gratefully acknowledged.

REFERENCES
Bengtsson, L., Botzet, M. and Esch, M. 1995. ‘Hurricane-type vortices in a general circulation model’, Tellus, 47A, 175 – 196.
Bengtsson, L., Botzet, M. and Esch, M., 1996. ‘Will greenhouse gas-induced warming over the next 50 years lead to higher frequency
and greater intensity of hurricanes?’, Tellus, 48A, 57–73.
Cubasch, U., Hasselmann, K., Hock, H., Maier-Reimer, E., Mikolajewicz, U., Santer, B. and Sausen, R. 1992. ‘Time-dependent
greenhouse warming computations with a coupled ocean – atmosphere model’, Clim. Dyn., 8, 55 – 69.
Del Genio, A. and Yao, M.-S. 1993. ‘Efficient cumulus parameterization for long-tern climate studies: The GISS scheme’, Am.
Meteorol. Soc. Monogr., 46, 181–184.
Druyan, L. and Hall, T. 1994. ‘Studies of African wave disturbances with the GISS GCM’, J. Clim., 7, 261 – 276.
Druyan, L. and Hall, T. 1996. ‘The sensitivity of African wave disturbances to remote forcing’, J. Appl. Meteorol., 35, 1100 – 1110.
Gray, W. 1979. ‘Hurricanes: their formation, structure and likely role in the tropical circulation’, in Shaw, D.B. (ed.), Meteorology
O6er the Tropical Oceans, Royal Meteorological Society, London, pp. 155 – 218.
Hansen, J., Sato, M., Ruedy, R., Lacis, A., Asamoah, K., Borenstein, S., Brown, E., Cairns, B., Caliri, G., Campbell, M., Curran,
B., de Castro, S., Druyan, L., Fox, M., Johnson, C., Lerner, J., McCormick, M., Miller, R., Minnis, P., Morrison, A., Pandolfo,
L., Ramberran, I., Zaucker, F., Robinson, M., Russell, P., Shah, K., Stone, P., Tegen, I., Thomason, L., Wilder, J. and Wilson,
H. 1996. ‘A Pinatubo climate modeling investigation’, in Fiocco, G., Fua, D. and Visconti, G. (eds), The Mount Pinatubo
Eruption: Effects on the Atmosphere and Climate, NATO ASI Series Vol. I 42, Springer, Heidelberg, pp. 233 – 272.
Henderson-Sellers, A., Zhang, H., Berz, G., Emanuel, K., Gray, W., Landsea, C., Holland, G., Lighthill, J., Shieh, S.-L., Webster,
P. and McGuffie, K. 1998. ‘Tropical cyclones and global climate change: A post-IPCC assessment’, Bull. AMS, 79, 19 – 38.
Kattenberg, A., Giorgi, F., Grassl, H., Meehl, G., Mitchell, J., Stouffer, R., Tokioka, T., Weaver, A. and Wigley, T. 1996. ‘Climate
models, projections of future climate’, in Houghton, J., Filho, L., Callander, B., Harris, N., Kattenberg, A. and Maskell, K. (eds),
Climate Change 1995, Intergovernmental Panel on Climate Change, Cambridge University Press, Cambridge and World
Meteorological Organization, Geneva, Chapter 6, pp. 289 – 357.
Landsea, C. 1997. ‘Comments on ‘‘Will greenhouse gas-induced warming over the next 50 years lead to higher frequency and greater
intensity of hurricanes?’’ ’, Tellus, 49A, 622–623.
Landsea, C. and Gray, W. 1992. ‘The strong association between western Sahel monsoon rainfall and intense Atlantic hurricanes’,
J. Clim., 5, 435 – 453.
Lighthill, J., Holland, G., Gray, W., Landsea, C., Craig, G., Evans, J., Kurihara, Y. and Guard, C. 1994. ‘Global climate change
and tropical cyclones’, Bull. AMS, 75, 2147–2157.
McBride, J. 1995. ‘Tropical cyclone formation’, in Elsberry, R. (ed.), Global Perspecti6es on Tropical Cyclones, TSP-38/WMO TD
c693, World Meteorological Organization, Geneva, Chapter 3, pp. 63 – 105.
Reed, R., Klinker, E. and Hollingsworth, A. 1988. ‘The structure and characteristics of African easterly wave disturbances as
determined from the ECMWF operational analysis/forecast system’, Meteorol. Atmos. Phys., 38, 22 – 33.
Rind, D. 1998. ‘Latitudinal temperature gradients and climate change’, JGR, 103, 5943 – 5971.
Royer, J.-F., Chauvin, F., Timbal, B., Araspin, P. and Grimal, D. 1998. ‘A GCM study of the impact of greenhouse gas increase
on the frequency of occurrence of tropical cyclones’, Clim. Change, 38, 307 – 343.
Russell, G., Miller, J. and Tsang, L.-C. 1985. ‘Seasonal oceanic heat transports computed from an atmospheric model’, Dyn. Atmos.
Oceans, 9, 253 – 271.
Russell, G., Miller, J. and Rind, D. 1995. ‘A coupled atmosphere – ocean model for transient climate change studies’, Atmos.–
Oceans, 33, 687 – 730.
Ryan, B., Watterson, I. and Evans, J. 1992. ‘Tropical cyclone frequencies inferred from Gray’s yearly genesis parameter: validation
of GCM tropical climates’, Geophys. Res. Let., 19, 1831 – 1834.
Walsh, K. and Watterson, I. 1997. ‘Tropical cyclone-like vortices in a limited area model: comparison with observed climatology’,
J. Clim., 10, 2240 –2259.
Watterson, I., Evans, J. and Ryan, B. 1995. ‘Seasonal and interannual variability of tropical cyclogenesis: diagnostics from
large-scale fields’, J. Clim., 8, 3052–3066.

Copyright © 1999 Royal Meteorological Society Int. J. Climatol. 19: 607 – 617 (1999)