Attribution of cyclogenesis region sea surface temperature change to anthropogenic influence
ABSTRACT 1] Previous research has identified links between tropical cyclone activity and sea surface temperatures in the tropical cyclogenesis regions of the North Atlantic and Western North Pacific. Other work has demonstrated that warming in these regions is inconsistent with simulated internal variability. After evaluating the variability of a suite of climate models on a range of timescales, we use detection and attribution methods and a suite of 20th century simulations including anthropogenic and natural forcing to identify a significant response to external forcing in both regions during the June– November hurricane season over the 20th century. We then use separate simulations of the response to natural and anthropogenic forcing to identify anthropogenic influence independently of natural influence in both the Atlantic and Pacific Cyclogenesis Regions.
Article: Incorporating model quality information in climate change detection and attribution studies[show abstract] [hide abstract]
ABSTRACT: In a recent multimodel detection and attribution (D&A) study using the pooled results from 22 different climate models, the simulated “fingerprint” pattern of anthropogenically caused changes in water vapor was identifiable with high statistical confidence in satellite data. Each model received equal weight in the D&A analysis, despite large differences in the skill with which they simulate key aspects of observed climate. Here, we examine whether water vapor D&A results are sensitive to model quality. The “top 10” and “bottom 10” models are selected with three different sets of skill measures and two different ranking approaches. The entire D&A analysis is then repeated with each of these different sets of more or less skillful models. Our performance metrics include the ability to simulate the mean state, the annual cycle, and the variability associated with El Niño. We find that estimates of an anthropogenic water vapor fingerprint are insensitive to current model uncertainties, and are governed by basic physical processes that are well-represented in climate models. Because the fingerprint is both robust to current model uncertainties and dissimilar to the dominant noise patterns, our ability to identify an anthropogenic influence on observed multidecadal changes in water vapor is not affected by “screening” based on model quality.Proceedings of the National Academy of Sciences 08/2009; 106(35):14778-14783. · 9.68 Impact Factor
Attribution of cyclogenesis region sea surface
temperature change to anthropogenic influence
N. P. Gillett,1P. A. Stott,2and B. D. Santer3
Received 15 February 2008; revised 28 March 2008; accepted 10 April 2008; published 13 May 2008.
cyclone activity and sea surface temperatures in the tropical
cyclogenesis regions of the North Atlantic and Western
North Pacific. Other work has demonstrated that warming in
these regions is inconsistent with simulated internal
variability. After evaluating the variability of a suite of
climate models on a range of timescales, we use detection
and attribution methods and a suite of 20th century
simulations including anthropogenic and natural forcing to
identify a significant response to external forcing in both
regions during the June–November hurricane season over
the 20th century. We then use separate simulations of the
response to natural and anthropogenic forcing to identify
anthropogenic influence independently of natural influence
in both the Atlantic and Pacific Cyclogenesis Regions.
Citation: Gillett, N. P., P. A. Stott, and B. D. Santer (2008),
Attribution of cyclogenesis region sea surface temperature change
to anthropogenic influence, Geophys. Res. Lett., 35, L09707,
Previous research has identified links between tropical
 Over recent decades there has been an increase in
the frequency of the most intense category four and five
tropical cyclones according to Webster et al.  and
an increase in the Potential Destructiveness Index (PDI),
in the Western North Pacific and North Atlantic, indicating
increased duration and intensity of tropical cyclones
[Emanuel, 2005, 2007; Trenberth et al., 2007]. Tropical
cylcone activity is strongly correlated with sea surface
temperatures (SSTs) in the Atlantic Cyclogenesis Region
(ACR) [Emanuel, 2005; Elsner et al., 2006; Holland and
Webster, 2007; Emanuel, 2007; Saunders and Lea, 2008]
and more weakly correlated in the Pacific Cyclogenesis
Region (PCR) [Chan and Liu, 2004; Emanuel, 2005,
2007]. Future changes in wind shear and atmospheric
stability may act to decrease tropical cyclone intensity
[Vecchi and Soden, 2007; Trenberth et al., 2007, Box
3.5], but high resolution models tend to show that the
effect of warming SSTs dominates, giving an increase in
tropical cyclone intensity in the future [Meehl et al.,
2007, section 10.3.6.3].
 SSTs in both the tropical North Atlantic and tropical
Western North Pacific have warmed over recent decades
[Emanuel, 2005; Trenberth et al., 2007]. Warming in the
tropical North Atlantic has been associated with the Atlantic
Multidecadal Oscillation (AMO) [Goldenberg et al., 2001],
a mode of climate variability associated with variations in
the strength of the thermohaline circulation [Trenberth et
al., 2007, section 3.6.6]. However the contribution of the
AMO to the North Atlantic warming trend is subject to
debate, and Trenberth and Shea  suggest an alternate
definition of the AMO in which the global mean tempera-
ture is subtracted from Atlantic SSTs. Warming of the ACR
is thus interpreted as being associated with global warming
[Trenberth and Shea, 2006; Mann and Emanuel, 2006].
Such analyses led Hegerl et al.  to conclude that
‘‘increasing greenhouse gas concentrations have likely
contributed to a warming of SSTs’’ in this region.
 Nonetheless, even if warming in the ACR is consistent
with larger-scale global warming, the question remains of
in this region, and if so, whether anthropogenic influence is
identifiable independently of natural climate influences.
Santer et al.  address the first of these questions by
comparing annual mean observed SST data over the ACR
and PCR with simulated internal variability in 22 CMIP3
coupled climate models. They conclude that observed trends
over the 20th century are inconsistent with simulated internal
variability over both regions. However, while interannual
variability was found to be realistic in both the PCR and
ACR, most models were found to exhibit lower decadal
variability in the ACR compared to observations, which
Santer et al.  suggest may be partly explainable by
 point out that coupled climate models are able to
simulate the AMO, but they do not comment on whether its
amplitude or timescale is generally realistic. Santer et al.
 go on to demonstrate that observed warming is
consistent in magnitude with the simulated response to
external forcings in 20th century simulations from the
CMIP3 models including both natural and anthropogenic
forcings. However, while this result is suggestive of an
anthropogenic influence on SSTs in these regions, it does
not conclusively demonstrate the presence of an anthro-
pogenic response. We build on this work by separating
anthropogenic and natural influences, by focusing on the
hurricane season alone, and by considering the temporal
evolution of cyclogenesis region SSTs.
2.Observational and Model Data
 We use three globally-complete gridded observational
SST data sets, the Hadley Centre Sea Ice and SST data set
[Rayner et al., 2003] (HadISST), the National Oceanic and
Atmospheric Administration Extended Reconstructed SST
GEOPHYSICAL RESEARCH LETTERS, VOL. 35, L09707, doi:10.1029/2008GL033670, 2008
1Climatic Research Unit, School of Environmental Sciences, University
of East Anglia, Norwich, UK.
2Met Office Hadley Centre, Exeter, UK.
3Program for Climate Model Diagnosis and Intercomparison, Lawrence
Livermore National Laboratory, Livermore, California, USA.
Copyright 2008 by the American Geophysical Union.
1 of 5
data set [Smith and Reynolds, 2004] (ERSST), and Kaplan
SST V2 [Kaplan et al., 1998] (Kaplan). We also use the
non-interpolated HadSST2 data [Rayner et al., 2006] in
detection and attribution analyses. Following Santer et al.
 we compare observed SSTs with simulated SSTs
from 22 coupled ocean-atmosphere models with data
available on the PCMDI CMIP3 archive. We use control
simulations with constant external forcing and 20th century
simulations which at a minimum contain specified green-
house gas and sulphate aerosol changes, and in some cases
other forcings including stratospheric ozone depletion,
volcanic aerosol and solar irradiance changes. We divide
the simulations into those including volcanic forcing (V)
and those excluding volcanic forcing (No-V), since a
response to volcanic forcing is evident in observations of
ACR and PCR SST [Santer et al., 2006]. In all cases, we
average SSTs over the June–November hurricane season,
and take spatial means over the ACR (6?N–18?N,
20?W–60?W) and PCR (5?N–15?N, 130?E–180?E)
(Figure 1) [Emanuel, 2005]. Since many simulations finish
in 1999 and start in 1900, we base our analysis on the
1900–1999 period, although observations over the period
1870–2007 are shown in Figure 2 for comparison.
 Figures 2a and 2b show time series of the simu-
lated and observed ACR and PCR 5-yr smoothed mean
temperature anomalies, along with approximations of the 5th
and 95th percentiles of the simulated variability. Awarming
trend is clearly visible over the 20th century in both regions,
and differences between the three data sets are small
compared to the variability. While the PCR shows a
relatively smooth warming trend, the ACR shows a strong
warming in the 1920–1940 period and again in the 1990s.
Both periods of warming have been associated in part with
the AMO [Trenberth et al., 2007, section 3.6.6]. In both
regions the V and No-V models appear to simulate the
warming trend reasonably well: The response to volcanoes,
particularly for Pinatubo (1991) and El Chicho ´n (1982)
[Santer et al., 2006]. Figure 1a shows that the ACR has
mean response of the CMIP3 models (Figure 1b), although
observed grid box temperature trends in the ACR and PCR
are within the 5th–95th percentile range of trends simu-
lated across the 72-member CMIP3 ensemble, indicating
that the enhanced warming of the ACR and reduced
warming of the PCR are likely due to internal variability
[Vecchi and Soden, 2007].
 Power spectra of simulated and observed ACR and
PCR SSTs, based on annual June–November averages over
the period 1900–1999 and using a Tukey-Hanning window
with a width of 30 years are shown in Figures 2c and 2d.
On average the 20th century CMIP3 simulations tend to
overestimate variability in the PCR on all the timescales
considered [Santer et al., 2006], and this difference in
variability is significant compared with the ERSST and
HadISST data sets. In the ACR interannual variability is
significantly less in the V models than in ERSST, but
differences between the V models and other observational
data sets or between ERSST and the No-V models are not
significant. There are no significant differences in decadal
variability between models and observations in the ACR.
 To test for the presence of an externally-forced signal
in the cyclogenesis region SSTs, we use a detection and
attribution analysis. After taking five-year means of simu-
lated and observed 1900–1999 temperature averaged over
the ACR and PCR and sampling the simulated 5-yr means
where observations are present, we use a total least squares
optimal fingerprinting method [Allen and Stott, 2003;
Hegerl et al., 2007, section 9.A.1] to regress the observed
SST changes onto the multi-model mean simulated 20th
century response. We denote this response ALL here,
although the forcings included in each simulation varied.
Simulated and observed SST time series were projected onto
control simulations, and uncertainties in the regression
coefficients were estimated using the latter halves of the
controls, following the approach of Gillett et al. .
Figure 3a shows the estimated regression coefficients for
the ALL response in the ACR and PCR, and using
HadSST2, ERSST and Kaplan SSTs. The presence of
Figure 1. Linear trends in June–November 1900–1999 SST from (a) ERSST observations and (b) 72 twentieth century
all-forcings simulations from 22 CMIP3 coupled climate models in K/century. Black boxes in the Atlantic and Pacific mark
the ACR and PCR, respectively.
GILLETT ET AL.: ATTRIBUTION OF CYCLOGENESIS REGION SST
2 of 5
external forcing is detected in both regions using all three
data sets: None of the 5–95% uncertainty ranges overlap
with zero. All the uncertainty ranges are consistent with one,
except for HadSST2 and ERSST in the PCR. The large
difference between the HadSST2 regression coefficient and
to limited spatial sampling at the beginning of the record. A
residual consistency test [Allen and Tett, 1999] indicated that
residuals in the regression were not significantly different
from those expected based on control variability for the
Kaplan data set, but were inconsistent for the ERSST data
set over the ACR, and for the HadSST2 data set over
both regions. This may in part relate to incomplete spatial
sampling in the HadSST2 data set. When the standard
deviation of the control was inflated by 50% the response
to external forcing was still detected and residuals were
consistent with control variability in all cases. When the
analysis was repeated over the period 1870–1999 using a
Figure 2. Time series of observed and simulated June-November SSTs averaged over the (a) ACR (6?N–18?N,
20?W–60?W) and (b) PCR (5?N–15?N, 130?E–180?E) smoothed with a running 5-yr mean and expressed as anomalies
relative to 1900–1999. Black lines show observed SSTs from HadISST, ERSST and Kaplan data sets. Solid blue lines
show the ensemble mean of 50 20th century simulations including volcanic forcing (from CCSM3, GFDL-CM2.0,
GFDL-CM2.1, GISS-EH, GISS-ER, MIROC3.2(hires), MIRCO3.2(medres), MIUB/ECHO-G, MRI-CGCM2.3.2, PCM,
UKMO-HadCM3, and UKMO-HadGEM1), and solid red lines show the ensemble mean of 22 20th century
simulations without volcanic forcing (from BCCR-BCM2.0, CGCM3.1(T47), ECHAM5/MPI-OM, CGCM3.1(T63),
CNRM-CM3, CSIRO-Mk3.0, FGOALS-g1.0, GISS-AOM, INM-CM3.0, and IPSL-CM4). For each year dotted red lines
show the 3rd and 48th warmest V simulation, and dotted blue lines the 2nd and 21st warmest No-V simulation,
approximately representing the 5th and 95th percentiles. Both sets of simulations also include greenhouse gas and direct
sulphate aerosol forcing, and other forcings such as the indirect sulphate aerosol effect and solar forcing in some cases.
The corresponding power spectra of (c) ACR and (d) PCR SSTs are shown below, based on annual June-November
means, calculated with a Tukey-Hanning window with a width of 30 years.
GILLETT ET AL.: ATTRIBUTION OF CYCLOGENESIS REGION SST
3 of 5
subset of 16 models with simulations starting at or before
1870, external influence was detected over the ACR with
all three data sets, and over the PCR with HadSST2 and
 To separate the effects of natural and anthropogenic
forcing we use two sets of ensembles of simulations from
the Parallel Climate Model (PCM) and third Hadley Centre
Coupled Model (HadCM3), one with natural forcing only
and one with anthropogenic forcings only (greenhouse gas,
sulphate aerosol, and stratospheric ozone changes, denoted
ANT). Figures 3b and 3c show regression coefficients from
two-way regressions of ACR and PCR SSTs onto two-
model-mean ANT and NAT responses, using each of the
three data sets HadSST2, ERSST and Kaplan, and the
same 5-yr mean diagnostic and 15 EOF truncation as for
the single pattern analysis. In all cases an anthropogenic
response is separately detectable, and in all cases the
residual test [Allen and Tett, 1999] indicated residual
variability consistent with simulated internal variability.
IntheACR anatural response isalsodetected,althoughNAT
NAT response is underestimated by these models. The
detectable natural response is likely dominated by volcanic
aerosol, with the anthropogenic response dominated by
greenhouse gas-induced warming, partly compensated
by sulphate-induced cooling [Santer et al., 2006]. An
attributable warming calculation using ERSST [Tett et al.,
2002] indicates that 0.53 K (5–95% uncertainty range of
0.21–0.93 K) of the observed 0.69 K warming over the
ACR and 0.32 K (0.09–0.56 K) of the observed 0.41 K
warming over the PCR is attributable to anthropogenic
influence. The best estimates therefore indicate that
warming over both regions is mainly anthropogenic,
although the uncertainty ranges are relatively large. We
could not extend the multi-model analysis back to 1870
because the PCM simulations started in 1900. A single-
model analysis with HadCM3 over the period 1870–1999
did not yield robust separate detection of anthropogenic
influence over either region, likely because of a reduced
signal-to-noise ratio due to our use of a single model.
 Sea surface temperature changes in the Atlantic and
Pacific cyclogenesis regions are of particular interest
Figure 3. (a) Dimensionless regression coefficients from
a single-pattern optimal detection analysis of observed
ACR and PCR temperatures against the ensemble mean
temperature simulated in 72 20th century simulations from
22 CMIP3 models. Dimensionless regression coefficients
from two-pattern optimal detection analyses of observed
to natural (NAT) and anthropogenic (ANT) forcings
simulated by HadCM3 and PCM. All calculations are based
on 5-yr June–November mean SSTs over each region.
Kaplan (blue). Uncertainty bars show 5–95% uncertainty
ranges derived from control variability. In Figures 3b and 3c,
curves enclose 90% of the estimated joint distribution of the
regression coefficients for each dataset.
GILLETT ET AL.: ATTRIBUTION OF CYCLOGENESIS REGION SST
4 of 5
because of their links to tropical cyclone intensity and
duration [Emanuel, 2005; Elsner et al., 2006; Saunders and
SST trends in these regions are inconsistent with internal
variability as simulated by a suite of 22 CMIP3 coupled
climate models. We build on this work by considering SSTs
during the June–November hurricane season alone, by
considering temporal evolution of 5-yr means of SSTs rather
than just trends, and by separately identifying natural and
anthropogenic influence. Using the 22 CMIP3 models, we
and PCR. The attribution analysis is dependent on simulated
internal variability: A residual consistency test indicates that
model variability may be underestimated over both regions
compared to HadSST2 data, but is consistent with Kaplan
data. When model variability was inflated by 50%, external
influence was still detected over both cyclogenesis
regions, and residuals were consistent with model variability
in all cases. We find that over both the ACR and PCR,
anthropogenic influence is detectable independently of
natural climate influences. Since Santer et al.  show
that greenhouse gases are the only forcing which gives rise to
a strong simulated warming over this period, our results
indicate that greenhouse gas increases are indeed likely the
dominant cause of the ACR warming, consistent with the
suggestion of Hegerl et al. . Moreover these results
disagree with the suggestion that warming in the Atlantic
Cyclogenesis Region is driven primarily by internal
multidecadal variability [Goldenberg et al., 2001].
groups for providing their data, PCMDI for archiving the data, and the JSC/
CLIVARWGCM for organizing the data analysis. We thank Kerry Emanuel
(MIT) for useful advice, Myles Allen (University of Oxford) for advice and
his detection and attribution code, and Nikos Christidis (Hadley Centre,
Met Office) and Michael Wehner (Lawrence Berkeley National Laboratory)
for assistance with the provision of model data. NPG acknowledges support
from the Leverhulme Trust and the Climate Change Detection and
Attribution Project, jointly funded by NOAA’s Office of Global Programs
and the US Department of Energy. PAS was supported by the Joint Defra
and MoD Programme, (Defra) GA01101 (MoD) CBC/2B/0417_Annex C5.
We acknowledge the international modeling
Allen, M. R., and P. A. Stott (2003), Estimating signal amplitudes in
optimal fingerprinting, Part I: Theory, Clim. Dyn., 21, 447–491.
Allen, M. R., and S. F. B. Tett (1999), Checking for model consistency in
optimal fingerprinting, Clim. Dyn., 15, 419–434.
Chan, J. C. L., and K. S. Liu (2004), Global warming and western North
Pacific typhoon activity from an observational perspective, J. Clim., 17,
Elsner, J. B., A. A. Tsonis, and T. H. Jagger (2006), High-frequency
variability in hurricane power dissipation and its relationship to global
temperature, Bull. Am. Meteorol. Soc., 87, 763–768.
Emanuel, K. A. (2005), Increasing destructiveness of tropical cyclones over
the past 30 years, Nature, 436, 686–688.
Emanuel, K. A. (2007), Environmental factors affecting tropical cyclone
power dissipation, J. Clim., 20, 5497–5509.
Gillett, N. P., F. W. Zwiers, A. J. Weaver, G. C. Hegerl, M. R. Allen,
and P. A. Stott (2002), Detecting anthropogenic influence with a multi-
model ensemble, Geophys. Res. Lett., 29(20), 1970, doi:10.1029/
Goldenberg, S. B., C. W. Landsea, A. M. Mestas-Nun ˜ez, and W. M. Gray
(2001), The recent increase in Atlantic hurricane activity: Causes and
implications, Science, 293, 474–479.
Hegerl, G. C., et al. (2007), Understanding and attributing climate change,
in Climate Change 2007: The Physical Science Basis, edited by S.
Solomon et al., chap. 9, 663–745, Cambridge Univ. Press, Cambridge,
Holland, G. J., and P. J. Webster (2007), Heightened tropical cyclone
activity in the North Atlantic: Natural variability or climate trend?,
Philos. Trans. R. Soc. London, 365, 2695–2716, doi:10.1098/
Kaplan, A., M. A. Cane, Y. Kushnir, A. C. Clement, M. B. Blumenthal, and
B. Rajagopalan (1998), Analyses of global sea surface temperature
1856–1991, J. Geophys. Res., 103, 18,567–18,589.
Mann, M. E., and K. A. Emanuel (2006), Atlantic hurricane trends linked to
climate change, Eos Trans. AGU, 87, 233–241.
Meehl, G. A., et al. (2007), Global climate projections, in Climate Change
2007: The Physical Science Basis, edited by S. Solomon et al., chap. 10,
747–845, Cambridge Univ. Press, Cambridge, U.K.
Randall, D. A., et al. (2007), Climate models and their evaluation, in
Climate Change 2007: The Physical Science Basis, edited by S. Solomon
et al., chap. 8, 589–662, Cambridge Univ. Press, Cambridge, U.K.
Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V. Alexander,
and D. P. Rowell (2003), Global analyses of sea surface temperature,
sea ice, and night marine air temperature since the late nineteenth
century, J. Geophys. Res., 108(D14), 4407, doi:10.1029/2002JD002670.
Rayner, N. A., P. Brohan, D. E. Parker, C. K. Folland, J. J. Kennedy,
M. Vanicek, T. J. Ansell, and S. F. B. Tett (2006), Improved analyses of
the mid-nineteenth century:TheHadSST2dataset, J.Clim.,19,446–469.
Santer, B. D., et al. (2006), Forced and unforced ocean temperature changes
in Atlantic and Pacific tropical cyclonogenesis regions, Proc. Natl. Acad.
Sci. U.S.A., 103, 13,905–13,910.
Saunders, M. A., and A. S. Lea (2008), Large contribution of sea surface
warming to recent increase in Atlantic hurricane activity, Nature, 451,
Smith, T. M., and R. W. Reynolds (2004), Improved extendedreconstruction
of SST (1854–1997), J. Clim., 17, 2466–2477.
Tett, S. F. B., et al. (2002), Estimation of natural and anthropogenic
contributions to twentieth century temperature change, J. Geophys.
Res., 107(D16), 4306, doi:10.1029/2000JD000028.
Trenberth, K. E., and D. J. Shea (2006), Atlantic hurricanes and natural
variability in 2005, Geophys. Res. Lett., 33, L12704, doi:10.1029/
Trenberth, K. E. et al. (2007), Observations: Surface and atmospheric
climate change, in Climate Change 2007: The Physical Science
Basis, edited by S. Solomon et al., chap. 3, 235–336, Cambridge
Univ. Press, Cambridge, U.K.
Vecchi, G. A., and B. J. Soden (2007), Effect of remote sea surface
temperature change on tropical cyclone potential intensity, Nature,
Webster, P. J., G. J. Holland, J. A. Curry, and H.-R. Chang (2005),
Changes in tropical cyclone number, duration, and intensity in a
warming environment, Science, 309, 1844–1846.
? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?
N. P. Gillett, Climatic Research Unit, School of Environmental Sciences,
University of East Anglia, Norwich NR4 7TJ, UK. (email@example.com)
B. D. Santer, Program for Climate Model Diagnosis and Intercomparison,
Lawrence Livermore National Laboratory, P.O. Box 808, Mail Stop L-103,
Livermore, CA 94550, USA.
P. A. Stott, Met Office Hadley Centre, Fitzroy Road, Exeter EX1 3PB,
GILLETT ET AL.: ATTRIBUTION OF CYCLOGENESIS REGION SST
5 of 5