Weakening of tropical Pacific atmospheric circulation due to anthropogenic forcing

Abstract

Since the mid-nineteenth century the Earth's surface has warmed, and models indicate that human activities have caused part of the warming by altering the radiative balance of the atmosphere. Simple theories suggest that global warming will reduce the strength of the mean tropical atmospheric circulation. An important aspect of this tropical circulation is a large-scale zonal (east-west) overturning of air across the equatorial Pacific Ocean--driven by convection to the west and subsidence to the east--known as the Walker circulation. Here we explore changes in tropical Pacific circulation since the mid-nineteenth century using observations and a suite of global climate model experiments. Observed Indo-Pacific sea level pressure reveals a weakening of the Walker circulation. The size of this trend is consistent with theoretical predictions, is accurately reproduced by climate model simulations and, within the climate models, is largely due to anthropogenic forcing. The climate model indicates that the weakened surface winds have altered the thermal structure and circulation of the tropical Pacific Ocean. These results support model projections of further weakening of tropical atmospheric circulation during the twenty-first century.

Full text

© 2006 Nature Publishing Group
© 2006 Nature Publishing Group
Weakening of tropical Pacific atmospheric
circulation due to anthropogenic forcing
Gabriel A. Vecchi
1
, Brian J. Soden
2
, Andrew T. Wittenberg
1
, Isaac M. Held
1
, Ants Leetmaa
1
& Matthew J. Harrison
1
Since the mid-nineteenth century the Earth’s surface has
warmed
1–3
, and models indicate that human activities have caused
part of the warming by altering the radiative balance of the
atmosphere
1,3
. Simple theories suggest that global warming will
reduce the strength of the mean tropical atmospheric circula-
tion
4,5
. An important aspect of this tropical circulation is a large-
scale zonal (east–west) overturning of air across the equatorial
Pacific Ocean
—
driven by convection to the west and subsidence to
the east
—
known as the Walker circulation
6
. Here we explore
changes in tropical Pacific circulation since the mid-nineteenth
century using observations and a suite of global climate model
experiments. Observed Indo-Pacific sea level pressure reveals a
weakening of the Walker circulation. The size of this trend is
consistent with theoretical predictions, is accurately reproduced
by climate model simulations and, within the climate models, is
largely due to anthropogenic forcing. The climate model indicates
that the weakened surface winds have altered the thermal struc-
ture and circulation of the tropical Pacific Ocean. These results
support model projections of further weakening of tropical
atmospheric circulation during the twenty-first century
4,5,7
.
The Walker circulation is fundamental to climate throughout the
globe: its variations are closely linked to those of the El Nin
˜
o/
Southern Oscillation
6
and monsoonal circulations over adjacent
continents
8
, and variations in its intensity and structure affect climate
across the planet
8,9
. The strength of equatorial Pacific zonal wind-
stress, associated with the Walker circulation, is critical to equatorial
Pacific Ocean circulation
10
and to biogeochemical processes
11
.
Observational and modelling evidence indicates that since the
mid-nineteenth century tropical sea surface temperatures (SSTs)
have warmed by 0.5–0.6 8C (refs 1–3). Climate models predict a
weakening of the atmospheric convective overturning in response to
surface warming driven by increases in greenhouse gases
4,5,7
. One
expects this weakening to be manifest, in part, as a reduction in the
zonal overturning of tropical air
4
, a large component of which is the
Pacific Walker circulation
6
.
The weakening of tropical atmospheric overturning circulations in
response to warming can be understood in terms of the energy and
mass balance of the ascending branch of these circulations. As the
surface warms, the water vapour concentration in the lower tropo-
sphere increases by roughly 7% per 8C of surface warming
12,13
,
consistent with the Clausius–Clapeyron equation and fixed relative
humidity. However, the rate of precipitation (which on long time-
scales is limited by the rate of change of the radiative cooling of the
troposphere) increases more slowly
—
approximately 2% per 8C
warming
1,14,15
. The global-mean rate of precipitation must be
balanced by the moisture transport from the atmospheric boundary
layer to the free troposphere, which is a product of the boundary layer
water vapour concentration and the exchange of air between the
boundary layer and free troposphere. Thus, the differential rate of
response to surface warming of water vapour and precipitation
LETTERS
Figure 1 | Spatial pattern of observed and modelled sea level pressure linear
trends.
Linear trend of sea level pressure (SLP) from: a, Kaplan SLP
reconstruction
29
(1861–1992), and ensemble-mean of GCM experiments
(1861–1992) as follows; b, all-forcing (five-member mean), c, natural forcing
(three-member mean) and d, anthropogenic forcing (three-member mean).
The trend averaged over the domain 158 S–158 N, 08–3608 is removed
from each panel. Dashed rectangles indicate the regions used to define the
large-scale Indo-Pacific SLP gradient index (DSLP).
1
NOAA/Geophysical Fluid Dynamics Laboratory, Princeton, New Jersey 08540-6649, USA.
2
Rosenstiel School for Marine and Atmospheric Sciences, University of Miami,
Miami, Florida 33149-1098, USA.
Vol 441|4 May 2006|doi:10.1038/nature04744
73

© 2006 Nature Publishing Group
© 2006 Nature Publishing Group
implies a weakening of the boundary layer/troposphere mass
exchange of ,5% per 8C warming
4
. Closely related arguments have
been provided for a weakening of tropical overturning circulations
based on the energy balance of the subsiding branch of these
circulations
5
. On the basis of observed tropical warming since the
mid-nineteenth century (0.5–0.6 8C), these theoretical arguments
predict a 2.5–3% reduction in the strength of tropical atmospheric
overturning circulation. Is there evidence of such a slowdown in the
observational record?
We use historical observations of sea level pressure (SLP) to assess
observed changes in the Walker circulation over the tropical Pacific;
an index of the large-scale tropical Indo-Pacific SLP gradient (DSLP)
serves as a proxy for the mean intensity of the Pacific Walker
circulation (see Methods). Ensembles of global climate model
(GCM) experiments
—
with different radiative forcings
—
serve to
explore the origin of the observed circulation changes, and allow
for the estimation of the statistical significance of the observed
changes. The model used here is the US National Oceanic and
Atmospheric Administration (NOAA) Geophysical Fluid Dynamics
Laboratory (GFDL) CM2.1 GCM
16–18
, with three historical inte-
gration sets over the period 1861–2000: (1) a five-member ensemble
including estimates of natural (solar variations, volcanoes) and
anthropogenic (well-mixed greenhouse gases, ozone, direct aerosol
forcing and land use) sources of climate change; (2) a three-member
ensemble applying only natural forcing; and (3) a three-member
ensemble applying only anthropogenic forcing (see Methods and
Supplementary Information).
There have been spatially coherent patterns in observed trends of
tropical SLP since the mid-nineteenth century (Fig. 1a), and similar
patterns are evident in the five-member ensemble-mean of the
historically forced GCM integrations (Fig. 1b). These trends indicate
a reduced zonal SLP gradient, and thus a weakened zonal circulation,
because the climatological pattern in the tropical Indo-Pacific has
SLP larger in the east than in the west. The naturally forced GCM
experiments are unable to recover these observed patterns in the SLP
trends (Fig. 1c); however, the GCM recovers many of the principal
observed SLP trend patterns using only anthropogenic forcing
(Fig. 1d).
Trends computed from observed DSLP are inconsistent with
those expected from the variability in the pre-industrial control
simulation, and they are inconsistent with the trends from the
‘natural-forcing’ GCM ensemble experiment (Fig. 2). However,
the trends in observed DSLP fall within the range of trends from
the ‘all-forcing’ GCM ensemble, and within that of the ‘anthropo-
genic-forcing’ GCM ensemble (Fig. 2). Thus, within the framework
of this GCM, a significant part of the observed reduction of
DSLP since the mid-nineteenth century resulted from anthropogenic
forcing. The trend computed from observed DSLP is also signifi-
cantly distinct from that expected from the pre-industrial control
experiments of all other models in the Intergovernmental Panel on
Climate Change 4th Assessment Report (IPCC/AR4) archive (Sup-
plementary Fig. 3). Most other models in the IPCC/AR4 archive also
show a weakening of the equatorial Pacific pressure gradient,
although CM2.1 shows the largest reduction (Supplementary Fig. 4.)
Though we use linear trends to summarize the changes in tropical
SLP since the mid-nineteenth century, these changes have not
appeared as a smooth reduction (Fig. 3). Both the observational
record and individual GCM ensemble members exhibit substantial
decadal variability (Supplementary Fig. 1), arising from processes
internal to the coupled system. The GCM decadal variability in DSLP
is comparable to that in observations, even though the GCM
interannual variability is overly energetic owing to a simulated
El Nin
˜
o variability that has too large an amplitude
17
. The strong
decadal variability complicates the detection of a relatively small
forced change even in multi-decadal records of DSLP; on the basis
of the 2,000-year control integration, a record shorter than
100–120 years is insufficient to detect the forced linear trend in
DSLP from the GCM at P ¼ 0.05 (Supplementary Fig. 2).
Figure 2 | Summary of the linear trends in SLP gradient across the
Indo-Pacific (DSLP) from observations and the various GCM historical
radiative forcing experiments.
Circles indicate the trend value from each
observational data set: K, Kaplan (1854–1992)
29
; H, Hadley Centre
(1871–1998)
28
; and B, a blend of Hadley
28
and Kaplan
29
, extended into 2005
using the NCEP gridded ship data (1854–2005)
27
. Model trends are
computed over the period 1861–2000. Confidence intervals are computed
from a 2,000-year control experiment, at the two-sided P ¼ 0.05 level.
Figure 3 | Observed and modelled evolution of DSLP since the nineteenth
century.
Five-year running-mean DSLP from: a, observations (black from
Kaplan
29
, and blue from Hadley Centre
28
; the record is extended through to
2005 with NCEP ship-based observations
27
(dashed line)); and b, GCM
historical integrations (five-member ensemble mean in black, an illustrative
ensemble member (number 3) in blue, dashed lines indicate linear trend of
ensemble members 1, 2, 4 and 5). In both panels, the linear trends in DSLP
are shown as thick lines, with shadings corresponding to each time-series.
LETTERS NATURE|Vol 441|4 May 2006
74

© 2006 Nature Publishing Group
© 2006 Nature Publishing Group
In the 1970s there was a rapid reduction in observed DSLP; such
rapid changes are also evident in other decades and in individual
ensemble members. Fifty-year DSLP trends ending in the 1990s are
significantly larger than those computed over the entire record,
suggesting an amplifying trend in DSLP. However, in recent
years the statistical significance of the amplification of the trend
disappears; though nominally larger than the long-term trend, the
1956–2005 trend is not significantly different from the long-term
trend at P ¼ 0.05. Further work is required to determine if the larger
recent trend since the 1950s is forced or a result of internal variability.
We estimate zonal wind stress across the equatorial Pacific using
DSLP, following the method of ref. 19 (Fig. 4). Linear relationships
with D SLP capture most of the interannual and longer timescale
variability in equatorial Pacific zonal-mean easterlies (,t
x
.; zonal
wind-stress averaged 1208 E–708 W, 58 S–58 N; easterlies are winds
from the east). DSLP-reconstructed ,t
x
. recovers the principal
extremes and transitions of the observed and modelled ,t
x
.,
including the long-term weakening of ,t
x
. in the model, and the
various El Nin
˜
o and La Nin
˜
a events observed in the recent decades
(Fig. 4, upper panels). The trend of the linear fit of ,t
x
. to observed
DSLP represents a ,t
x
. reduction of ,7% since 1860; as surface
stress is roughly proportional to the square of wind speed, this
suggests a mean reduction in equatorial Pacific zonal wind of ,3.5%,
roughly consistent with the theoretical prediction.
Equatorial Pacific ,t
x
. is of critical importance to the large-scale
oceanic circulation in the equatorial Pacific
9
. The GCM experiments
indicate that since the mid-nineteenth century, the weakening of
equatorial ,t
x
. has resulted in a weakening of surface equatorial
currents, a vertical shift in sub-surface currents, and a reduction in
the intensity and depth of equatorial upwelling (Supplementary
Fig. 5). By bringing nutrient-rich waters close to the surface,
equatorial upwelling exerts a strong control on biological activity
in the tropical Pacific
10
; its weakening and shoaling suggest a possible
reduction of biological productivity under global warming.
The GCM experiments indicate a substantial shoaling of the
western equatorial Pacific thermocline depth (Z
tc
) since the mid-
nineteenth century (Fig. 4, lower panels); the thermocline is the zone
of rapid temperature change in a typical vertical oceanic temperature
profile between the warm well-mixed surface layer and cold abyssal
waters. Reduction of ,t
x
. affects both the east–west tilt of the
equatorial Pacific thermocline and its mean depth
20
(Supplementary
Fig. 6); these changes could affect the character of El Nin
˜
o varia-
bility
21
. On seasonal to interannual timescales (timescales too short
for the equatorial thermocline to come to equilibrium with the
winds), a reduction in ,t
x
. has a strong impact on both western
and eastern equatorial Pacific Z
tc
. However, on timescales longer
than that of equatorial adjustment, the impact of reductions in
,t
x
. is felt almost entirely by the western equatorial Pacific Z
tc
(Fig. 4). Because Z
tc
and SST are tightly coupled only in the eastern
equatorial Pacific
22
, these long-term thermocline depth changes in
the GCM are unlikely to affect SST directly. The observational record
of equatorial Pacific thermocline slope and depth over the past
50 years is consistent with the recent reduction in the strength of
the easterlies and the low-frequency relationship between Z
tc
and
,t
x
. in the GCM.
Atmosphere–ocean conditions in the equatorial Pacific have
changed since the mid-nineteenth century: there has been a signifi-
cant slowdown of atmospheric circulation, which models indicate
has driven a response in ocean circulation. The observed trend in the
Pacific surface zonal SLP gradient is unlikely to be due to natural
variability. However, much of the long-term trend is reproduced in
model simulations that account for human impacts on the radiative
budget of the planet, and is consistent with the changes expected
from simple thermodynamic arguments
4
. The agreement between
Figure 4 | Observed and modelled equatorial Pacific zonal-mean zonal
wind-stress anomaly, <t
x
>, and equatorial thermocline depth anomaly,
Z
tc
. Upper panels: model/observed ,t
x
. and reconstruction using linear
relation to DSLP; dashed line shows (1854–2005) trend in ,t
x
.
reconstructed using blended Kaplan
29
/Hadley
28
/NCEP
27
DSLP. Lower
panels: Z
tc
in the western (black line, 28 S–28 N, 1408 E–1808 E) and eastern
(blue line, 28 S–28 N, 1308 W–908 W) equatorial Pacific. Left panels:
ensemble-mean all-forcing CM2.1 GCM experiment, showing five-year
running mean. Right panels: five-year running mean (thick lines) and
annual-mean (thin lines) observational estimates. Observed stress is from
European Centre for Medium Range Weather Forecasting Reanalysis 40,
observed Z
tc
is from GFDL ocean data assimilation
30
. Z
tc
is the location of
the maximum vertical temperature gradient. Note different scales in each
panel.
NATURE|Vol 441|4 May 2006 LETTERS
75

© 2006 Nature Publishing Group
© 2006 Nature Publishing Group
the theoretical, observed and model-produced changes in strength
of atmospheric circulation suggests increased confidence in the
model-projected reduction in the strength of tropical circulation
during the twenty-first century
4,5,7
; on the basis of climate model
simulations, this weakening may be of the order of 10% by the end of
the twenty-first century
4,7
.
METHODS
SLP data sets and calculations. Direct assessment of long-term changes in the
strength of tropical circulation is problematic, because there have been changes
in the methods used to make wind measurements at sea
23,24
. Trends in tropical
winds over recent decades are ambiguous, with some studies showing a
strengthening
25
and others a weakening
26
. SLP offers a proxy to recover changes
in wind velocity, because of the dynamical connection between large-scale zonal
gradients of SLP and zonal-mean zonal wind-stress
19
. Further, SLP measure-
ments at sea have always been made using instruments, in contrast with
measurements of surface wind, which have been instrument-based only in
recent decades
23,24
. Equatorial Pacific wind-stress estimated from the zonal
gradients of SLP has shown a reduction over the period 1920–90
19
. Ship-based
SLP measurements
27
have been used to build global gridded data sets of monthly
SLP
28,29
using knowledge of its global co-variability, and allow the exploration of
changes from the mid-nineteenth century.
In this study two global reconstruction data sets of monthly SLP are used: (1)
the Kaplan SLP data set
29
version 1 spanning the period 1854–1992, with data
only over the ocean, and (2) the Hadley Centre SLP reconstruction
28
version 1
spanning the period 1871–1998, with data over both land and ocean. These data
sets are extended into recent years using the gridded monthly ship-based
SLP observations available from NOAA’s National Center for Environmental
Prediction (NCEP)
27
. For the analyses presented here, the monthly long-term
climatology of each data set is removed to compute SLP anomalies.
Using the GCM and observed SLP data, a large-scale tropical Indo-Pacific SLP
gradient (DSLP) is computed from the difference in SLP averaged over the
central/east Pacific (1608 W–808 W, 5 8 S–58 N) and over the Indian Ocean/west
Pacific (808 E–1608 E, 58 S–58 N). The index is computed with SLP anomalies
from monthly climatology; positive values indicate a strengthened Indo-Pacific
SLP gradient. A least-squares linear trend in DSLP is used as a concise metric for
the long-term changes in the strength of zonal circulation. As shown in ref. 19 a
large-scale SLP gradient, like DSLP, provides a useful proxy for the mean
intensity of Pacific zonal surface winds. For the observational record of surface
winds available to us (1957–2003) and for the entire climate model record
(1861–2000), DSLP provides a useful proxy for the strength of the mean zonal
circulation over the equatorial Pacific Ocean (Fig. 4).
Climate models. Global climate models (GCMs) simulate the variations of, and
interactions between, various elements of the climate system (ocean, atmos-
phere, cryosphere and land) forced by radiatively active naturally occurring and
anthropogenic gases and aerosols. Internal climate variability can be isolated
from that forced by changes to atmospheric composition through ensemble
experiments, in which the same model physics and forcing fields are applied to
different initial conditions. The model used here is the NOAA GFDL CM2.1
GCM
16–18
, which uses estimated radiative forcings over the period 1861–2000.
The principal historical integration set is a five-member ensemble including
estimates of natural (solar variations, volcanoes), and anthropogenic (well-
mixed greenhouse gases, ozone, direct aerosol forcing and land use) sources of
climate change. Two additional sets of experiments isolate the effects of each set
of forcing elements, by applying only natural or anthropogenic forcing; each
consists of a three-member ensemble. Statistical significance estimates are
computed from a 2,000-year control integration with invariant radiative
conditions from the 1860s (see Supplementary Information).
Received 27 October 2005; accepted 22 March 2006.
1. Houghton, J., et al. Climate Change 2001: The Scientific Basis (Cambridge Univ.
Press, Cambridge, UK, 2001).
2. Rayner, N. A. et al. 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 (2003).
3. Knutson, T. R. et al. Assessment of twentieth-century regional surface
temperature trends using the GFDL CM2 coupled models. J. Clim. (in the press).
4. Held, I. M. & Soden, B. J. Robust responses of the hydrological cycle to global
warming. J. Clim. (in the press).
5. Knutson, T. R. & Manabe, S. Time-mean response over the tropical Pacific to
increased CO
2
in a coupled ocean-atmosphere model. J. Clim. 8, 2181–-2199
(1995).
6. Julian, P. R. & Chervin, R. M. A study of the Southern Oscillation and the
Walker Circulation. Mon. Weath. Rev. 106, 1433–-1451 (1978).
7. Tanaka, H. L., Ishizki, N. & Kitoh, A. Trend and interannual variability of
Walker, monsoon and Hadley circulations defined by velocity potential in the
upper troposphere. Tellus A 56, 250–-269 (2004).
8. Webster, P. J. et al. Monsoons: Processes, predictability, and the prospects for
prediction. J. Geophys. Res. 103(C7), 14451–-14510 (1998).
9. Deser, C. & Wallace, J. M. Large-scale atmospheric circulation features of
warm and cold episodes in the tropical Pacific. J. Clim. 3, 1254–-1281 (1990).
10. Cane, M. A. & Sarachik, E. S. Forced baroclinic ocean motions, Part II: The
linear equatorial bounded case. J. Mar. Res. 35, 395–-432 (1977).
11. Barber, R. T. & Chavez, F. P. Biological consequences of El Nin
˜
o. Science 222,
1203–-1210 (1983).
12. Trenberth, K. E., Fasullo, J. & Smith, L. Trends and variability in column
integrated atmospheric water vapor. Clim. Dyn. 24, doi:10.1007/s00382–-005–-
0017–-4 (2005).
13. Soden, B. J., Jackson, D. L., Ramaswamy, V., Schwarzkopf, M. D. & Huang, X.
The radiative signature of upper tropospheric moistening. Science 310,
841–-844, doi:10.1126/science.1115602 (2005).
14. Boer, G. J. Climate change and the regulation of the surface moisture and
energy budgets. Clim. Dyn. 8, 225–-239 (1993).
15. Allen, M. R. & Ingram, W. J. Constraints on future changes in the hydrological
cycle. Nature 419, 224–-228 (2002).
16. Delworth, T. L. et al. GFDL’s CM2 global coupled climate models–-Part 1:
Formulation and simulation characteristics. J. Clim. 19(5), 643–-674 (2006).
17. Wittenberg, A. T., Rosati, A., Lau, N.-C. & Ploshay, J. J. GFDL’s CM2 global
coupled climate models–-Part 3: Tropical Pacific climate and ENSO. J. Clim.
19(5), 698–-722 (2006).
18. Stouffer, R. et al. GFDL’s CM2 global coupled climate models–-Part 4: Idealized
climate response. J. Clim. 19(5), 723–-740 (2006).
19. Clarke, A. J. & Lebedev, A. Long-term changes in equatorial Pacific trade
winds. J. Clim. 9, 1020–-1029 (1996).
20. Jin, F.-F. An equatorial ocean recharge paradigm for ENSO. Part I: Conceptual
model. J. Atmos. Sci. 54, 811–-829 (1997).
21. Fedorov, A. V. & Philander, S. G. Is El Nin
˜
o changing? Science 288, 1997–-2002
(2000).
22. Harrison, D. E. & Vecchi, G. A. El Nin
˜
o and La Nin
˜
a–-Equatorial Pacific
thermocline and sea surface temperature anomalies, 1986–-98. Geophys. Res.
Lett. 28, 1051–-1054 (2001).
23. Cardone, V. J., Greenwood, J. G. & Cane, M. A. On trends in historical marine
wind data. J. Clim. 3, 113–-127 (1990).
24. Ramage, C. S. Can shipboard measurements reveal secular changes in tropical
air-sea heat flux? J. Clim. Appl. Meteorol. 23, 187–-193 (1984).
25. Whysall, K. D. B., Cooper, N. S. & Bigg, G. R. Long-term changes in tropical
Pacific surface wind field. Nature 327, 216–-219 (1987).
26. Harrison, D. E. Post World War II trends in tropical Pacific surface trades.
J. Clim. 2, 1561–-1563 (1989).
27. Worley, S. J., Woodruff, S. D., Reynolds, R. W., Lubker, S. J. & Lot, N. ICOADS
release 2.1 data and products. Int. J. Climatol. 25, 823–-842 (2005).
28. Basnett, T. & Parker, D. Development of the Global Mean Sea Level Pressure Data
Set GMSLP2 (Climate Research Technical Note 79, Hadley Centre, Met Office,
Exeter, UK, 1997).
29. Kaplan, A., Kushnir, Y. & Cane, M. A. Reduced space optimal interpolation of
historical marine sea level pressure. J. Clim. 13, 2987–-3002 (2000).
30. Derber, J. & Rosati, A. A global oceanic data assimilation system. J. Phys.
Oceanogr. 19, 1333–-1347 (1989).
Supplementary Information is linked to the online version of the paper at
www.nature.com/nature.
Acknowledgements G.A.V. was supported by the Visiting Scientist Program at
the NOAA/GFDL administered by UCAR. We are grateful to the model
development teams at GFDL, and thank A. E. Johansson, M. P. Vecchi,
T. Knutson, T. Delworth and J. Russell for comments and suggestions.
Author Information Reprints and permissions information is available at
npg.nature.com/reprintsandpermissions. The authors declare no competing
financial interests. Correspondence and requests for materials should be
addressed to G.A.V. (Gabriel.A.Vecchi@noaa.gov).
LETTERS NATURE|Vol 441|4 May 2006
76