Content uploaded by James A. Screen
Author content
All content in this area was uploaded by James A. Screen
Content may be subject to copyright.
LETTERS
The central role of diminishing sea ice in recent Arctic
temperature amplification
James A. Screen
1
& Ian Simmonds
1
The rise in Arctic near-surface air temperatures has been almost
twice as large as the global average in recent decades
1–3
—a feature
known as ‘Arctic amplification’. Increased concentrations of
atmospheric greenhouse gases have driven Arctic and global aver-
age warming
1,4
; however, the underlying causes of Arctic amp-
lification remain uncertain. The roles of reductions in snow and
sea ice cover
5–7
and changes in atmospheric and oceanic
circulation
8–10
, cloud cover and water vapour
11,12
are still matters
of debate. A better understanding of the processes responsible for
the recent amplified warming is essential for assessing the like-
lihood, and impacts, of future rapid Arctic warming and sea ice
loss
13,14
. Here we show that the Arctic warming is strongest at the
surface during most of the year and is primarily consistent with
reductions in sea ice cover. Changes in cloud cover, in contrast,
have not contributed strongly to recent warming. Increases in
atmospheric water vapour content, partly in response to reduced
sea ice cover, may have enhanced warming in the lower part of the
atmosphere during summer and early autumn. We conclude that
diminishing sea ice has had a leading role in recent Arctic temper-
ature amplification. The findings reinforce suggestions that strong
positive ice–temperature feedbacks have emerged in the Arctic
15
,
increasing the chances of further rapid warming and sea ice loss,
and will probably affect polar ecosystems, ice-sheet mass balance
and human activities in the Arctic
2
.
The Arctic region has long been expected to warm strongly as a
result of anthropogenic climate change
1,2
, owing to positive feed-
backs in the Arctic climate system. It is widely accepted that changes
in the surface albedo associated with melting snow and ice enhance
warming in the Arctic
3,15,16
, but other processes may contribute. In
some global climate models, changes in cloud cover and atmospheric
water vapour content are more important for Arctic amplification
than the surface albedo feedback
17–19
. However, the same climate
models significantly underestimate the recent Arctic sea ice decline
5
and surface warming
20
, in part due to unrealistic negative feedbacks
20
.
One reanalysis data set suggests that Arctic warming may have been
enhanced by an increase in the atmospheric poleward transport of
heat and moisture
8
. However, another reanalysis data set reveals a
decrease in poleward heat transport since the early 1980s
21
, which was
a period of rapid sea ice declines
5–7
. Changes in Arctic storm beha-
viour
9
may have also enhanced the warming.
The vertical profile of recent warming can provide insight into its
underlying causes. For instance, retreating snow and sea ice cover is
expected to induce maximum warming at the surface
15,22
, whereas
changes in atmospheric poleward heat transport may cause warming
with large vertical extent
8
. The ERA-40 reanalysis has been used to
show
8
that Arctic warming trends aloft were of equal or greater mag-
nitude than those at the surface, leading to the conclusion that atmo-
spheric circulation changes were a more important cause of recent
Arctic amplification than retreating snow and sea ice cover. However,
notable discrepancies exist between the vertical profiles of warming
in different reanalysis data sets
15
. The findings of ref. 8 have been
contested
15,23–25
, and concerns have been expressed over the validity
of trends in ERA-40 that may reflect inhomogeneities or artefacts in
the reanalysis rather than true climate signals
23,24
.
Here we present results from a new reanalysis data set, ERA-
Interim
26
. Some of the key improvements over the ERA-40 data set
include higher resolution, improved model physics, a better hydro-
logical cycle, four-dimensional variational data assimilation and
variational bias correction of satellite radiance data
26
. The last feature
is of particular relevance for this study because the scarcity of direct
temperature measurements over the Arctic Ocean dictates that the
majority of observations come from satellite radiances. The vari-
ational bias correction of satellite radiance data accounts for biases
that change in time, for instance owing to changes in the observing
network or drift of satellite orbits. ERA-Interim depicts more realistic
Arctic tropospheric temperatures and probably suffers less from
spurious trends than any previous reanalysis data set
26
(Supplementary Information). Furthermore, we build on the results
of ref. 8 by including the post-2001 period, during which sea ice
retreat has accelerated
5–7
.
Arctic amplification is a clear feature of the warming over the
1989–2008 period based on the ERA-Interim reanalysis (Fig. 1).
We diverge considerably from ref. 8 in finding that the maximum
Arctic warming is at the surface and that warming lessens with height
in all seasons except summer. This vertical structure suggests that
changes at the surface, such as decreases in sea ice and snow cover, are
the primary causes of recent Arctic amplification. The trends at the
near-surface (herein the atmospheric levels at 950–1,000 hPa) are 1.6,
0.9, 0.5 and 1.6 uC per decade, averaged over the Arctic (herein lati-
tudes 70–90uN) during winter, spring, summer and autumn,
respectively. The near-surface warming is modest in summer because
energy is used to melt remaining sea ice and warm the upper
ocean
3,15
. The surface amplification, defined here as the ratio of the
near-surface warming to that of the whole tropospheric column
(below 300 hPa), averaged over the Arctic, is greatest in autumn, with
a value of 2.3. The surface amplification is aided by strong low-level
stability that limits vertical mixing. The corresponding values of
surface amplification for winter and spring are 2.1 and 1.8, respect-
ively. We note that amplified Arctic warming, above ,700 hPa, is
confined to winter and is still consistently weaker than the near-
surface warming (Fig. 1a). However, the presence of amplified warm-
ing aloft hints that processes in addition to the increased transfer of
heat from the ocean to the atmosphere resulting from sea ice loss have
had a contributing role in winter.
The surface amplified warming is closely linked to diminishing sea
ice cover over the 1989–2008 period (linear trends of 22.6, 21.4,
25.8 and 27.9% per decade relative to the 1989–2008 means for
winter, spring, summer and autumn, respectively). The components
1
School of Earth Sciences, University of Melbourne, Melbourne, Victoria 3010, Australia.
Vol 464
|
29 April 2010
|
doi:10.1038/nature09051
1334
Macmillan Publishers Limited. All rights reserved
©2010
of the seasonal temperature trends that are linearly congruent with
changes in sea ice (Fig. 2) show remarkable resemblance to the ver-
tical profiles of the total temperature trends (Fig. 1). North of 70uN, a
large portion of each total trend is linked to reduced Arctic sea ice
cover (Fig. 2). The majority of the winter warming is associated with
changes in sea ice cover (Fig. 2a) even though the sea ice declines are
relatively small and the albedo feedback is weak during this season.
Strong winter warming is consistent with the atmospheric response
to reduced sea ice cover
22,27
and reflects the seasonal cycle of ocean–
atmosphere heat fluxes
22
: during summer, the atmosphere loses heat
to the ocean whereas during winter the flux of heat is reversed. Thus,
reduced summer sea ice cover allows for greater warming of the
upper ocean but atmospheric warming is modest (Fig. 2c). The inter-
action is undoubtedly two-way because warmer upper-ocean tem-
peratures will further enhance sea ice loss. The excess heat stored in
the upper ocean is subsequently released to the atmosphere during
winter
20,22
. Reduced winter sea ice cover, in part a response to a
warmer upper ocean and delayed refreezing
6,7
, facilitates a greater
transfer of heat to the atmosphere. The observed thinning of Arctic
sea ice
28,29
, albeit not explicitly represented in ERA-Interim, is also
likely to have enhanced the surface heat fluxes.
Another potential contributor to the surface amplified warming
could be changes in cloud cover. Clouds decrease the incoming
short-wave (solar) radiation. However, this shading effect is partly
offset, or exceeded, by a compensating increase in incoming long-wave
radiation. In the Arctic, this greenhouse effect dominates during
autumn, winter and spring (Fig. 3), in agreement with in situ observa-
tions
30
. In summer, the shading effect dominates in the lower-latitude
regions of the Arctic basin whereas north of 80uN the two competing
effects approximately cancel out (Fig. 3c). Spring is the only season that
exhibits significant trends in Arctic average cloudinessin ERA-Interim,
and these are negative (the ERA-Interim cloud-cover trends are con-
sistent with satellite estimates; see Supplementary Information).
400
600
800
1,000
400
600
800
1,000
400
600
800
1,000
400
600
800
1,000
Level (hPa)
90º 80º 60º 50º 40º70º 90º 80º 60º 50º 40º70º
90º 80º 60º 50º 40º70º 90º 80º 60º 50º 40º70º
b
a
dc
Latitude north
–1.0 –0.5 0.0 0.5 1.0 1.5 2.0 2.5
Temperature trend (ºC per decade)
2
1
0
2
1
0
2
1
0
2
1
0
Figure 1
|
Surface amplification of temperature trends, 1989
–
2008.
Temperature trends averaged around circles of latitude for winter
(December–February; a), spring (March–May; b), summer (June–August;
c) and autumn (September–November; d). The black contours indicate
where trends differ significantly from zero at the 99% (solid lines) and 95%
(dotted lines) confidence levels. The line graphs show trends (same units as
in colour plots) averaged over the lower part of the atmosphere
(950–1,000 hPa; solid lines) and over the entire atmospheric column
(300–1,000 hPa; dotted lines). Red shading indicates that the lower
atmosphere has warmed faster than the atmospheric column as whole. Blue
shading indicates that the lower atmosphere has warmed slower than the
atmospheric column as a whole.
400
600
800
1,000
400
600
800
1,000
400
600
800
1,000
400
600
800
1,000
Level (hPa)
ab
cd
90º 80º 70º 60º 50º 40º 90º 80º 70º 60º 50º 40º
90º 80º 70º 60º 50º 40º 90º 80º 70º 60º 50º 40º
–1.0 –0.5 0.50.0 1.51.0 2.0 2.5
Temperature trend (ºC per decade)
Latitude north
Figure 2
|
Temperature trends linked to changes in sea ice. Temperature
trends over the 1989–2008 period averaged around circles of latitude for
winter (a), spring (b), summer (c) and autumn (d). The trends are derived
from projections of the temperature field on the sea ice time series (Methods
Summary). The black contours indicate where the ice–temperature
regressions differ significantly from zero at the 99% (solid lines) and 95%
(dotted lines) uncertainty levels.
200
0
100
–100
200
0
100
–100
200
40º50º60º70º80º90º
40º50º60º70º80º90º40º50º60º70º80º90º
40º50º60º70º80º90º
0
100
–100
200
0
100
–100
Mean net surface radiation (W m–2)
Latitude north
ab
cd
Figure 3
|
Impacts of cloud-cover changes on the net surface radiation.
Mean net surface radiation (short-wave plus long-wave) over the 1989–2008
period under cloudy-sky (solid lines) and clear-sky (dotted lines) conditions.
Means are averaged around circles of latitude for winter (a), spring
(b), summer (c) and autumn (d). The fluxes are defined as positive in the
downward direction. Red shading indicates that the presence of cloud has a
net warming effect at the surface. Blue shading indicates that the presence of
cloud has a net cooling effect at the surface. The dashed lines show the
approximate edge of the Arctic basin. Symbols show latitudes where
increases (triangles) and decreases (crosses) in total cloud cover significant
at the 99% uncertainty level are found.
NATURE
|
Vol 464
|
29 April 2010 LETTERS
1335
Macmillan Publishers Limited. All rights reserved
©2010
Rather than contribute to the warming, decreased cloud cover would
be expected to promote surface cooling because clouds have a warming
influencein spring (Fig. 3b). It islikely that the temperature responseto
reduced cloud cover is exceeded by warming due to other processes.
The radiativeeffect of cloud-cover changes is small in comparison with
compensating changes inthe temperature and humidity profiles assoc-
iated with varying ice conditions
11
. We find that the large majority of
spring warming occurs in the Siberian sector of the Arctic basin (not
shown), where ice clouds are the predominant cloud type
12
. In ice-
cloud-dominated regions, the radiative effects of changes in cloud
cover are less important than changes in water vapour content
12
.In
short, we find no evidence of changes in cloud cover contributing to
recent near-surface Arctic warming.
A final consideration arises from model simulations which suggest
that changes in atmospheric water vapour content may amplify
Arctic warming
17–19
. Increases in water vapour are expected with
increasing air temperatures and reduced sea ice cover
19,27
. In turn,
water vapour is a powerful greenhouse gas
1
and can lead to further
warming and sea ice loss. In ERA-Interim, specific humidity trends
are found only during the summer and early autumn, and are con-
fined to the lower part of the atmosphere (Fig. 4a). The largest
humidity increases are found in the Arctic basin. An associated
increase in incoming long-wave radiation has probably enhanced
warming in summer and early autumn. It is of further interest to
determine whether these increases in humidity are locally driven or
are a result of increased moisture transport into the Arctic. It is worth
noting that the humidity trends coincide with the months of lowest
sea ice coverage and largest sea ice declines. The pronounced warm-
ing in winter and spring is not accompanied by increases in humidity.
A large portion of each total humidity trend is linked to changes in sea
ice (Fig. 4b) and, furthermore, to significant increases in the surface
latent-heat flux (that is, evaporation) in the Arctic basin (Fig. 4a).
The humidity increases at latitudes 50–65uN show weaker links to sea
ice and are probably influenced by other processes. However, within
the Arctic these lines of evidence support the notion that part of the
humidity increase is driven by enhanced surface moisture fluxes
associated with sea ice reductions.
The evidence from the past two decades, based on ERA-Interim,
reveals that recent reductions in sea ice cover and thickness have been
great enough to enhance Arctic warming strongly during most of the
year. Our results suggest that the majority of the recent Arctic tem-
perature amplification is due to diminishing sea ice cover. The amp-
lification is strongest in the lowermost part of the atmosphere, where
modified surface heat fluxes have their greatest influence. The emer-
gence of strong ice–temperature positive feedbacks increases the like-
lihood of future rapid Arctic warming and sea ice decline.
METHODS SUMMARY
The raw data we used were monthly mean fields from the ERA-Interim
26
reana-
lysis for the period 1989–2008. A discussion of the data quality and comparisons
with the older ERA-40 reanalysis data set are given in the Supplementary
Information. These data were averaged around circles of latitude (at 1.5ureso-
lution). Standard seasonal means were computed and used in Figs 1, 2 and 3 (the
winter mean for 1989 contains no data for December 1988), and June–October
means were used in Fig. 4. We estimated trends using least-squares linear regres-
sion. The statistical significances of the regressions were calculated from a two-
tailed t-test. Changes in sea ice cover were calculated by averaging sea ice
concentrations over the Arctic Ocean (north of 70uN). To construct Fig. 2, we
regressed the temperature field against the index of Arctic-wide sea ice cover.
These regressions were then multiplied by the sea ice time series to give a pro-
jection of the temperature field onto the sea ice time series. The linear trends of
these projections (Fig. 2) represent the temperature trends statistically linked to
changes in sea ice cover. We used the same procedure for specific humidity data
(Fig. 4b).
Caution is required when interpreting regressions between two variables that
both show pronounced trends—as is the case with recent Arctic temperatures
and sea ice cover. It is plausible that two variables linked statistically are phys-
ically independent in reality. To address this possibility, we recalculated the
regressions using detrended data. We found that year-to-year variations in sea
ice cover are linked to approximately the same patterns of temperature and
humidity anomalies as found in the raw data. This gives us further confidence
that the associations revealed here are physically meaningful.
Received 10 November 2009; accepted 12 March 2010.
1. Solomon, S. et al. (eds) Climate Change 2007: The Physical Science Basis
(Cambridge Univ. Press, 2007).
2. Symon, C., Arris, L. & Heal, B. (eds) Arctic Climate Impact Assessment (Cambridge
Univ. Press, 2004).
3. Serreze, M. C. & Francis, J. A. The Arctic amplification debate. Clim. Change 76,
241
–
264 (2006).
4. Gillett, N. P. et al. Attribution of polar warming to human influe nce. Nature Geosci.
1, 750
–
754 (2008).
5. Stroeve, J., Holland, M. M., Meir, W., Scambos, T. & Serreze, M. Arctic sea ice
decline: faster than forecast. Geophys. Res. Lett. 34, doi:10.1029/2007GL029703
(2007).
6. Serreze, M. C., Holland, M. M. & Stroeve, J. Perspectives of the Arctic’s shrinking
ice cover. Science 315, 1533
–
1536 (2007).
7. Comiso, J. C., Parkinson, C. L., Gersten, R. & Stock, L. Accelerated decline in the
Arctic sea ice cover. Geophys. Res. Lett. 35, doi:10.1029/2007GL031972 (2008).
8. Graversen, R. G., Mauritsen, T., Tjernstro
¨m, M., Ka
¨lle
´n, E. & Svensson, G. Vertical
structure of recent Arctic warming. Nature 451, 53
–
56 (2008).
9. Simmonds, I. & Keay, K. Extraordinary September Arctic sea ice reductions and
their relationships with storm behavior over 1979
–
2008. Geophys. Res. Lett. 36,
doi:10.1029/2009GL039810 (2009).
10. Chylek, P., Folland, C. K. & Lesins, G. Dubey, M. K. & Wang, M. Arctic air
temperature change amplification and the Atlantic multidecadal oscillation.
Geophys. Res. Lett. 36, doi:10.1029/2009GL038777 (2009).
11. Schweiger, A. J., Lindsay, R. W., Vavrus, S. & Francis, J. A. Relationships between
Arctic sea ice and clouds during autumn. J. Clim. 21, 4799
–
4810 (2008).
12. Francis, J. A. & Hunter, E. Changes in the fabric of the Arctic’ s greenhouse blanket.
Environ. Res. Lett. 2, doi:10.1088/1748
–
9326/2/4/045011 (2007).
13. Holland, M. M., Bitz, C. M. & Tremblay, B. Future abrupt reductions in the summer
Arctic sea ice. Geophys. Res. Lett. 33, doi:10.1029/2006GL028024 (2006).
14. Boe
´, J., Hall, A. & Qu, X. September sea ice cover in the Arctic Ocean projected to
vanish by 2100. Nature Geosci. 2, 341
–
343 (2009).
15. Serreze, M. C., Barrett, A. P., Stroeve, J. C., Kindig, D. N. & Holland, M. M. The
emergence of surface-based Arctic amplification. Cryosphere 3, 11
–
19 (2009).
16. Holland, M. M. & Bitz, C. M. Polar amplification of climate change in coupled
models. Clim. Dyn. 21, 221
–
232 (2003).
17. Winton, M. Amplified climate change: what does surface albedo feedback have to
do with it? Geophys. Res. Lett. 33, doi:10.1029/2005GL025244 (2006).
18. Lu, J. & Cai, M. Seasonality of polar surface warming amplification in climate
simulations. Geophys. Res. Lett. 36, doi:10.1029/2009GL040133 (2009).
19. Graversen, R. G. & Wang, M. Polar amplification in a coupled model with locked
albedo. Clim. Dyn. 33, 629
–
643 (2009).
20. Boe
´, J., Hall, A. & Qu, X. Current GCMs’ unrealistic negative feedback in the Arctic.
J. Clim. 22, 4682
–
4695 (2009).
21. Smedsrud, L. H., Sorteberg, A. & Kloster, K. Recent and future changes of the
Arctic sea-ice cover. Geophys. Res. Lett. 35, doi:10.1029/2008GL034813 (2008).
90º 80º 70º 60º 50º 40º
1,000
800
600
400
1,000
800
600
400
Level (hPa)
90º 80º 70º 60º 50º 40º
–0.12 –0.06 0.00 0.06 0.12 0.18 0.24 0.30
Specic humidity trend (g kg–1 per decade)
ab
Latitude north
Figure 4
|
Atmospheric moisture trends, 1989
–
2008. Specific humidity
trends averaged around circles of latitude for June–October: total trends
(a); trends that are linked to changes in sea ice (b). The black contours
indicate where trends (a) or humidity–ice regressions (b) differ significantly
from zero at the 99% (solid lines) and 95% (dotted lines) uncertainty levels.
In a, triangles show latitudes where increases in the surface latent-heat flux
significant at the 99% uncertainty level are found.
LETTERS NATURE
|
Vol 464
|
29 April 2010
1336
Macmillan Publishers Limited. All rights reserved
©2010
22. Deser, C., Tomas, R., Alexander, M. & Lawrence, D. The seasonal atmospheric
response to projected Arctic sea ice loss in the late twenty-first century. J. Clim.
23, 333
–
351 (2010).
23. Thorne,P. W. Arctic troposphericwarming amplification?Nature 455, E1
–
E2 (2008).
24. Grant, A. N., Bro
¨nnimann, S. & Haimberger, L. Recent Arctic warming vertical
structure contested. Nature 455, E2
–
E3 (2008).
25. Bitz, C. M. & Fu, Q. Arctic warming aloft is data set dependent. Nature 455, E3
–
E4
(2008).
26. Dee, D. P. & Uppala, S. Variational bias correction of satellite radiance data in the
ERA-Interim reanalysis. Q. J. R. Meteorol. Soc. 135, 1830
–
1841 (2009).
27. Higgins, M. E. & Cassano, J. J. Impacts of reduced sea ice on winter Arctic
atmospheric circulation, precipitation and temperature. J. Geophys. Res. 114,
doi:10.1029/2009JD011884 (2009).
28. Kwok, R. & Rothrock, D. A. Decline in Arctic sea ice thickness from submarine and
ICESat records: 1958
–
2008. Geophys. Res. Lett. 36, doi:10.1029/2009GL039035
(2009).
29. Lindsay, R. W., Zhang, J., Schweiger, A., Steele, M. & Stern, H. Arctic sea ice retreat
in 2007 follows thinning trend. J. Clim. 22, 165
–
176 (2009).
30. Intrieri, J. M. et al. An annual cycle of Arctic surface cloud forcing at SHEBA. J.
Geophys. Res. 107, doi:10.1029/2000JC000439 (2002).
Supplementary Information is linked to the online version of the paper at
www.nature.com/nature.
Acknowledgements We thank N. Gillett and R. Graversen for comments on the
manuscript. The ERA-Interim data were obtained from the European Centre for
Medium-Range Weather Forecasts data server. Parts of this research were
supported by funding from the Australian Research Council.
Author Contributions The analysis was performed and the manuscript written by
J.A.S. Both authors contributed with ideas and discussions.
Author Information Reprints and permissions information is available at
www.nature.com/reprints. The authors declare no competing financial interests.
Readers are welcome to comment on the online version of this article at
www.nature.com/nature. Correspondence and requests for materials should be
addressed to J.A.S. (screenj@unimelb.edu.au).
NATURE
|
Vol 464
|
29 April 2010 LETTERS
1337
Macmillan Publishers Limited. All rights reserved
©2010
SUPPLEMENTARY INFORMATION
1
www.nature.com/nature
doi: 10.1038/nature09051
The central role of diminishing sea ice in recent Arctic temperature
amplification
James A. Screen & Ian Simmonds
SUPPLEMENTARY DISCUSSION
Our results and conclusions are based on the ERA-Interim
reanalysis (see Dee & Uppala1 and references therein). ERA-
Interim is the latest global atmospheric reanalysis produced by
the European Centre for Medium-Range Weather Forecasts
(ECMWF), covering the data-rich period since 1989. It
succeeds the older ERA-40 reanalysis2 which covers the
period 1979-2001 and upon which Graversen et al.3 draw their
conclusions. The ECMWF, applying lessons learned from
ERA-40 and well-documented weakness in other first- and
second-generation reanalyses, have implemented a number of
improvements in ERA-Interim. These include an assimilating
model with higher spectral resolution (T255 from T159),
improved model physics, a more sophisticated hydrological
cycle, and data assimilation based on a 12-hourly four-
dimensional variational analysis (4D-Var) that includes
adaptive estimation of biases in satellite radiance data.
Arguably, the single biggest concern of using reanalyses for
climate monitoring is that changes in the observing system,
combined with the presence of biases in models and
observations, can cause shifts and trends in reanalyses that
interfere with the true climate signal4. ERA-Interim is the first
reanalysis to include an assimilation scheme that adjusts for
biases that change in time, for instance due to changes in the
observing network or the decay and drift of satellite orbits.
The ERA-Interim reanalysis only became available to the
scientific community in 2009. Consequently, the validation
and evaluation of the output is in its early stages, which
imposes some uncertainty in our results. The only previous
assessment of ERA-Interim performance in the Arctic found
that the vertical temperature profile north of 70oN shows
improved fit to radiosonde temperatures in comparison to
ERA-401. In order to further evaluate the performance of
ERA-Interim in the Arctic region, and to reduce the
uncertainties in our conclusions, we have compared the
reanalysis data with observations. We have used monthly
radiosonde temperature anomalies from the United Kingdom
Meteorological Office (UKMO) Hadley Centre atmospheric
temperature analysis (HadAT)5. Since HadAT contains no
surface data, we have used surface temperature observations
from the National Aeronautics and Space Administration
(NASA) Goddard Institute for Space Studies surface
temperature analysis (GISTEMP)6. Data were extracted for all
meteorological stations north of 70oN which had near-
complete records (at least 90% coverage over the period 1989-
2008). Missing data were not interpolated. The selected
stations provide reasonable circumpolar coverage in the
latitudes 70-80oN (locations in Supplementary Fig. 1a). Few
stations exist north of 80oN. We first calculated seasonal
means (no missing months allowed) and then averaged over
all the circumpolar stations. Linear trends and their
uncertainties were calculated as a function of season and
height. The same procedure was applied to the ERA-Interim
temperature fields after sub-sampling at the locations, levels
and times of available observations. The vertical structures of
temperature trends in ERA-Interim and observations are in
close agreement (Supplementary Fig. 1). In all seasons except
summer, both observations and ERA-Interim display strongest
warming at the surface, consistent with the results in the main
material. In contrast, significant discrepancies have been
identified between the vertical profiles of Arctic temperature
trends in ERA-40 and radiosondes7, 8. There are quantitative
differences between the trends in ERA-Interim and
observations, particularly at the 850hPa level during summer
and autumn (Supplementary Fig. 1). However, the magnitudes
of the trends in the reanalysis and in the observations are not
significantly different when their uncertainties are taken into
account. Furthermore, our conclusions are based on the
qualitative vertical structure of the temperature trends and not
on absolute magnitudes.
Supplementary Figure 1 | Comparison of the vertical
structures of temperature trends in ERA-Interim and
observations, 1989-2008. Trends are averaged from
meteorological stations north of 70oN (locations in a) for winter
(December-February; b), spring (March-May; c), summer (June-
August; d) and autumn (September-November; e). In b to e, solid
lines show trends from observations whereas the dotted lines show
trends from ERA-Interim sub-sampled at the locations and times of
available observations. Also shown are the 95% confidence intervals
(grey bands for observations and error bars for ERA-Interim). In a,
the Arctic stations are: 1. Bjornoya, 2. Ostrov Dikson, 3. Tiksi, 4.
Barrow, 5. Resolute, 6. Eureka, 7. Alert, 8. Danmarkshavn, 9. Jan
Mayen.
2
www.nature.com/nature
doi: 10.1038/nature09051 SUPPLEMENTARY INFORMATION
To further test our hypothesis that ERA-Interim offers a
more realistic depiction of Arctic temperature trends than
ERA-40, we have compared the two reanalyses with
observations during the period of overlap, 1989-2001. Using
the same set of observations and sub-sampled ERA-Interim
fields discussed above, plus sub-sampled ERA-40 fields, we
subtracted the seasonal mean temperature anomalies (relative
to the 1989-2001 mean) in the two reanalyses from the
corresponding anomalies in the observations. The root-mean-
square error (RMSE) of these seasonal temperature anomalies
is shown as a function of height in Supplementary Figure 2a.
At all levels, the ERA-Interim temperature anomalies are
considerably closer to observations than are the ERA-40
anomalies. The improved accuracy in ERA-Interim compared
to ERA-40 is most pronounced in the mid- to lower-
troposphere. A similar pattern is revealed with respect to the
seasonal temperature trends (Supplementary Fig. 2b). The
mid- to lower-tropospheric trends in ERA-Interim are more
realistic than those depicted in ERA-40. It is at these levels
that ERA-40 displays amplified warming3 that is not apparent
in ERA-Interim or in the radiosonde data (over the 1989-2008
or 1979-20018 periods).
An additional aspect of ERA-Interim performance we
have tested is its representation of Arctic cloud cover trends.
We have compared the trends in ERA-Interim with those in
the International Satellite Cloud Climatology Project (ISCCP)
D2 data set9. Both the satellite product and ERA-Interim show
significant decreases in spring cloud cover over the 1989-
2008 period (Supplementary Table 1). We note that
conflicting spring trends have been found over different time
periods10 suggesting substantial decadal-scale variability in
Arctic cloud cover. The direction of the winter and summer
trends (downward and upward, respectively) are also in
agreement between ISCCP and ERA-Interim. The autumn
trends differ in sign but are both small and not statistically
different from zero. Thus, ERA-Interim faithfully represents
DJF MAM JJA SON
ERA-Int -0.75 +/- 1.94 -2.44 +/- 0.63 0.89 +/- 0.96 0.26 +/- 0.56
ISCCP -1.16 +/- 1.03 -2.04 +/- 1.24 0.55 +/- 1.60 -0.18 +/- 0.84
the cloud trends shown by satellites, at least in respect to total
cloud cover and over this period.
In summary, ERA-Interim and ERA-40 depict differing
vertical profiles of Arctic temperature trends. Based on the
discussions above we find that ERA-Interim temperature
trends are more reliable than those in ERA-40. ERA-Interim
and radiosondes are in agreement in showing strongest Arctic
warming at the surface. Additionally, differences in the
magnitudes and structures of temperature trends between
ERA-40 and ERA-Interim may arise because the two
reanalyses cover non-identical time periods. Given that we
find strong associations between near-surface warming and
the loss of sea ice cover, and that sea ice retreat has
accelerated in the past decade relative to the preceding two
decades, it follows that the near-surface Arctic warming signal
is more pronounced over the 1989-2008 period than over the
1979-2001 period. The improved representation of Arctic
temperature trends in ERA-Interim, combined with the
emergence (or strengthening) of surface-based temperature
amplification in the past decade11, help reconcile the
differences between our results and those of Graversen et al.3.
1. Dee, D.P. & Uppala, S. Variational bias correction of satellite
radiance data in the ERA-Interim reanalysis. Q. J. R.
Meteorol. Soc. 135, 1830-1841 (2009).
2. Uppala, S.M. et al. The ERA-40 re-analysis. Q. J. R.
Meteorol. Soc. 131, 2961-3012 (2005).
3. Graversen, R.G., Mauritsen, T., Tjernström, M., Källén, E. &
Svensson, G. Vertical structure of recent Arctic warming.
Nature 451, 53-56 (2008).
4. Bengtsson, L., Hagemann, S. & Hodges, K.I. Can climate
trends be calculated from reanalysis data? J. Geophys. Res.
109, doi:10.1029/2004JD004536 (2004).
5. Thorne, P.W. et al. Revisiting radiosonde upper air
temperatures from 1958 to 2002. J. Geophys. Res. 110,
doi:10.1029/2004JD005753 (2005).
6. Hansen, J., Ruedy, R., Glascoe, J. & Sato, M. GISS analysis
of surface temperature change. J Geophys. Res. 104,
doi:10.1029/1999JD900835 (1999).
7. Thorne, P.W., Arctic tropospheric warming amplification?
Nature 455, E1-E2 (2008).
8. Grant, A.N., Brönnimann, S. & Haimberger, L. Recent Arctic
warming vertical structure contested. Nature 455, E2-E3
(2008).
9. Rossow, W. & Schiffer, R. Advances in understanding clouds
from ISCCP. Bull. Amer. Meteor. Soc. 80, 2261-2288 (1999).
10. Schweiger, A.J. Changes in seasonal cloud cover over the
Arctic seas from satellite and surface observations. Geophys.
Res. Lett. 31, doi:10.1029/2004/Gl020067 (2004).
11. Serreze, M.C., Barrett, A.P., Stroeve, J.C., Kindig, D.N. &
Holland, M.M. The emergence of surface-based Arctic
amplification. Cryosphere 3, 11-19 (2009).
Acknowledgements The ERA-Interim and ERA-40 data were
obtained from the ECMWF; HadAT2 data from the UKMO Hadley
Centre; GISTEMP and ISCPP data from the NASA Goddard
Institute for Space Studies.
Supplementary Table 1 | Comparison of Arctic cloud cover
trends from ERA-Interim and satellites, 1989-2008. Area-
weighted Arctic-mean total cloud cover trends (% per decade) and
their 95% confidence limits for the four seasons.
Supplementary Figure 2 | Comparison of the mean errors in
ERA-Interim and ERA-40 relative to observations, 1989-2001.
(a) the root-mean-square error of seasonal temperature anomalies as
a function of height calculated from observations minus reanalyses.
The solid line is for ERA-Interim and the dotted line for ERA-40. (b)
as in a but for the seasonal temperature trends.