LETTERS
PUBLISHED ONLINE: 28 APRIL 2013 | DOI: 10.1038/NCLIMATE1884
Springtime atmospheric energy transport and the
control of Arctic summer sea-ice extent
Marie-Luise Kapsch*, Rune Grand Graversen and Michael Tjernström
The summer sea-ice extent in the Arctic has decreased in recent
decades, a feature that has become one of the most distinct
signals of the continuing climate change1–4 . However, the interannual variability is large—the ice extent by the end of the
summer varies by several million square kilometres from year
to year5 . The underlying processes driving this year-to-year
variability are not well understood. Here we demonstrate that
the greenhouse effect associated with clouds and water vapour
in spring is crucial for the development of the sea ice during the
subsequent months. In years where the end-of-summer sea-ice
extent is well below normal, a significantly enhanced transport
of humid air is evident during spring into the region where the
ice retreat is encountered. This enhanced transport of humid
air leads to an anomalous convergence of humidity, and to an
increase of the cloudiness. The increase of the cloudiness and
humidity results in an enhancement of the greenhouse effect.
As a result, downward long-wave radiation at the surface is
larger than usual in spring, which enhances the ice melt. In
addition, the increase of clouds causes an increase of the
reflection of incoming solar radiation. This leads to the counterintuitive effect: for years with little sea ice in September, the
downwelling short-wave radiation at the surface is smaller than
usual. That is, the downwelling short-wave radiation is not
responsible for the initiation of the ice anomaly but acts as an
amplifying feedback once the melt is started.
The sea-ice extent in the Arctic has been steadily decreasing
during the satellite remote-sensing era, 1979 to present (Fig. 1a).
The highest rate of retreat is found in September5 , which coincides
with the month of the annual cycle that has the lowest ice extent.
Factors that are believed to cause the ice retreat are, among
others: changes in surface air temperature6–8 , ice circulation in
response to winds/pressure patterns7–11 , and ocean currents8 , as
well as changes in radiative fluxes (for example, due to changes
in cloud cover)7,10,12–15 , and ocean conditions (for example, ocean
warming16 ). However, large interannual variability is superimposed
onto the declining trend (Fig. 1a). The year-to-year deviation of
the ice extent in September relative to the trend line varies by,
on average, ±0.5 × 106 km2 , but can reach 1.75 × 106 km2 , which
is around 25% of the mean September extent for 1979–2010.
The magnitude of the variability shows considerable regional
differences: a comparison of years with an anomalously large
September sea-ice extent (HIYs—high ice years) with years showing
an anomalously small ice extent (LIYs—low ice years) reveals that
the variability is most pronounced in the Arctic Ocean north
of Siberia (Fig. 1b,c). Significant ice-concentration anomalies of
∼±30% are observed for LIYs and HIYs in this area, which is
chosen as the study area for the following analyses. In 2007 and
2012—the years showing the first and second lowest Arctic ice
extent since the satellite observations began—a large part of this
area became entirely ice free17–19 .
What are the processes causing the year-to-year ice variability?
The processes can be divided into two types, dynamical and
thermodynamical. The former includes processes that transport ice
away from a given location and pack it elsewhere, or export it from
the Arctic to southerly latitudes where it eventually melts. The latter
type includes processes causing ice melt, which are associated with
alterations of the energy exchange between the ice and the ocean
below or the atmosphere above. These two types of process are not
independent, because weather patterns that transport the ice may
also bring warm and humid air as well as warm surface water into
the ice-retreat region. On the basis of correlations between lower
atmospheric winds and September sea-ice extent it has been shown
that 50% of the ice variability can be associated with the winds20 .
Again, the linkage between the two is probably due to a combination
of the direct dynamical forcing by the winds and the winds bringing
warm and humid air in over the ice, which increases the energy flux
to the surface10 . Here we focus on the thermodynamical component
and quantify the reduction in ice extent that this component can be
accounted for. Across the surface the energy balance is given by
Fsrf = SWN + LWN + SH + LH
(1)
where SWN = SWSD−SWSU and LWN = LWSD−LWSU indicate
net short-wave (SWN) and long-wave (LWN) radiation, defined
as downward (SD) minus upward (SU) radiation, and SH and LH
are the turbulent fluxes of sensible and latent heat, respectively.
All terms in equation (1) are here defined positive downward.
Using data from the ERA-Interim reanalysis21 we investigated this
surface energy balance.
The amount of long-wave radiation from the atmosphere to the
surface during spring plays an important role for the September
sea-ice concentration (SIC) in years that show an unusual small
ice extent by the end of the summer (LIYs): during late spring
(April–May) anomalies, based on detrended data, of the net
long-wave radiation plus turbulent fluxes result in a significant
larger-than-average energy flux to the surface over the area where
the September sea-ice anomaly is encountered (Table 1, Fig. 2a
and Supplementary Fig. S6). This alteration of the surface flux is
linked to significant larger-than-average atmospheric content of
cloud water and water vapour (Fig. 3a, Table 1 and Supplementary
Fig. S7). These cloud and water-vapour anomalies lead to an
enhancement of the atmospheric opacity and thus of the greenhouse
effect. This causes a significant increase of long-wave radiation—
but a significant decrease of short-wave radiation—downward at
the surface (Fig. 2b, Table 1 and Supplementary Fig. S6). The
anomaly of net long-wave radiation plus turbulent fluxes stays
significantly positive throughout the late spring and summer
months (April–August) and accounts for most of the enhanced
energy flux to the surface during the melting season (Fig. 2a
and Table 1). In mid May, when the ice anomaly begins to
Department of Meteorology and Bolin Centre for Climate Research, Stockholm University, 10691 Stockholm, Sweden. *e-mail: [email protected].
NATURE CLIMATE CHANGE | ADVANCE ONLINE PUBLICATION | www.nature.com/natureclimatechange
© 2013 Macmillan Publishers Limited. All rights reserved.
1
LETTERS
NATURE CLIMATE CHANGE DOI: 10.1038/NCLIMATE1884
a
a
10
10
Flux anomaly (W m¬2)
LWN + LH + SH
7
6
SWN
SIC
5
5
0
0
¬5
¬5
¬10
Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Sep. Oct. Nov. Dec.
5
b
1988
1992
1996
Year
2000
2004
Flux anomaly (W m¬2)
1984
2008
13
E
5°
13
5°
W
b
30
LWSD
SWSD
LH
SH
5
5
0
0
¬5
¬5
Sea-ice anomaly (%)
1980
¬10
10
10
SIC
4
Sea-ice anomaly (%)
Sea-ice extent (106 km2)
8
¬10
¬10
Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Sep. Oct. Nov. Dec.
20
°W
45
0
5°
E
5°
13
13
W
c
SIC anomaly (%)
45
°E
10
¬10
°W
45
45
°E
¬20
¬30
Figure 1 | Arctic sea-ice extent and ice-concentration anomalies for
September, 1979–2010. a, Black lines are Arctic sea-ice extent for
September and the linear trend line. Red, dashed lines indicate the ±0.5
s.d. used to split the time series into years with a little (LIYs) and large
(HIYs) September sea-ice extent. b,c, SIC anomalies for LIYs (b) and HIYs
(c). Significant anomalies, encapsulated by the grey line (95% significance
level—based on monthly averages), are mainly evident in the Arctic Ocean,
off the Siberian coast. The black outline marks the study area.
appear, and the surface albedo therefore becomes anomalously
low, the net short-wave-radiation anomaly becomes positive. The
net short-wave radiation contributes to the enhanced energy
flux to the surface during the rest of the melting season. These
findings lead to the conclusion that enhanced long-wave radiation
associated with positive humidity and cloud anomalies during
2
Figure 2 | Radiative and turbulent flux anomalies at the surface for LIYs.
The black line shows the SIC (right-hand axis). a, The net long-wave
radiation plus the turbulent fluxes (latent and sensible; red) and the net
short-wave radiation (green). b, The radiative fluxes are split into their
components but only downwelling long-wave (red) and short-wave (green)
radiation are shown together with the latent (dark blue) and sensible (light
blue) heat flux. All time series are based on daily anomalies and averaged
over the area indicated by the black outline in Fig. 1b. All data are detrended
before calculating the anomalies and a 30-day running mean filter is
applied.
spring plays a significant role in initiating the summer ice melt,
whereas short-wave-radiation anomalies act as an amplifying
feedback once the melt has started. This conclusion holds also for
composites of the earlier and later years of the LIYs, if 2007 is
excluded from the LIY composite, for a much larger area covering
most of the Arctic Ocean, and for different reanalysis data (see
Supplementary Discussion).
The positive anomalies of cloudiness and humidity in late spring
of LIYs cannot be explained by the ice anomaly because this
appears first around mid May. Instead, these cloud-water and
humidity anomalies are most likely due to the variability of the
atmospheric circulation: splitting the atmospheric energy-transport
convergence into its dry-static and latent components reveals that
over the ice-retreat area, the latent heat-transport convergence,
which is essentially the convergence of water vapour, is significantly
larger than average during April–May (Fig. 3b, Table 1 and
Supplementary Fig. S7). The convergence of atmospheric water in
spring (March–April) can be estimated roughly by taking the mean
latent energy transport convergence (LTC) as 2 W m−2 (Table 1)
and applying LTC × t /L, where t is set to 60 days and L is the
latent heat of evaporation (2.5 × 106 J kg−1 ). Hereby it is found
that the latent energy transport provides the atmosphere above the
ice-retreat area with ∼4 kg m−2 of water in April–May, which is
much more than the amount associated with the humidity and
cloudiness anomalies (∼0.3 kg m−2 ; Table 1) during this period.
The convergence of the dry-static transport during late spring
of LIYs is positive but not statistically significant (Fig. 3b, Table 1
and Supplementary Fig. S7). However, during the summer season
June–August, the convergence is significantly larger than average
and provides the atmosphere over the ice-retreat area with an extra
∼3.5 W m−2 . This contributes to retaining the anomaly of the net
long-wave radiation plus turbulent fluxes around ∼2 W m−2 , after
the cloud anomaly has become small during summer (Fig. 3a).
NATURE CLIMATE CHANGE | ADVANCE ONLINE PUBLICATION | www.nature.com/natureclimatechange
© 2013 Macmillan Publishers Limited. All rights reserved.
NATURE CLIMATE CHANGE DOI: 10.1038/NCLIMATE1884
LETTERS
Table 1 | Statistical significance of the anomalies of surface fluxes, atmospheric water contents and atmospheric energy transport
convergence for LIYs.
Time period
Variable
Average value
Statistical significance (%)
01 April–31 May
SIC
LWN + SH + LH
LWN
SWN
LWSD
SWSD
SH
LH
Conv. of dry-static energy
Conv. of latent energy
Total column water vapour
Total column cloud water
−0.49%
1.73 W m−2
1.87 W m−2
−0.55 W m−2
5.76 W m−2
−4.96 W m−2
0.15 W m−2
−0.29 W m−2
−0.14 W m−2
2.18 W m−2
0.27 kg m−2
4.89 × 10−3 kg m−2
65
99
99
74
100
100
39
95
4
98
99
99
01 April– 31 Aug
SIC
LWN + SH + LH
LWN
SWN
LWSD
SWSD
SH
LH
Conv. of dry-static energy
Conv. latent energy
Total column water vapour
Total column cloud water
−1.91%
2.01 W m−2
1.43 W m−2
0.83 W m−2
3.43 W m−2
−2.25 W m−2
0.19 W m−2
0.40 W m−2
2.06 W m−2
0.52 W m−2
0.42 kg m−2
0.97 × 10−3 kg m−2
70
99
99
85
100
95
68
89
87
45
100
47
01 June–31 Aug
SIC
LWN + SH + LH
LWN
SWN
LWSD
SWSD
SH
LH
Conv. of dry-static energy
Conv. of latent energy
Total column water vapour
Total column cloud water
−2.84%
2.20 W m−2
1.14 W m−2
1.74 W m−2
1.87 W m−2
−0.45 W m−2
0.21 W m−2
0.85 W m−2
3.51 W m−2
−0.58 W m−2
0.52 kg m−2
−1.63 × 10−3 kg m−2
97
91
85
97
97
31
64
95
97
35
100
65
All fields are averaged over the area indicated by the black box in Fig. 1b and over selected periods. Significance is based a Monte Carlo approach (10,000 iterations; see Methods). Significance values are
given by the nearest integer. LWSD/SWSD: downwelling long-wave/short-wave radiation; LWN/SWN: net long-wave/short-wave radiation; SH/LH: sensible/latent heat flux.
For LIYs, the anomaly of long-wave radiation plus turbulent
fluxes prevails significantly positive during the period April–August,
leading to a total energy surplus of ∼2 W m−2 over the ice-retreat
area (Table 1). Further, the net short-wave radiation contributes
with an additional amount of ∼1 W m−2 . As a result, the extra
energy gained by the surface due to these anomalies can melt on
average ∼13 cm of ice over the area. This value is found using
1Fsrf × t /Lf /ρice , where 1Fsrf = 3 W m−2 , t is set to 150 days,
Lf = 334 × 103 J kg−1 is latent heat of fusion, and ρice = 900 kg m−3
is sea-ice density. Assuming an average ice thickness of 0.5–1 m
by the end of the melt season the surface fluxes can melt 13–26%
of the total ice amount within the study area. The assumption
of the average ice thickness in this area is supported by recent
observations22 and model results23 . If it is further assumed that all
thickness categories are equally represented, the melted ice fraction
is 7–13% (see Supplementary Fig. S8 for details), which can be
compared to the ice-concentration anomaly for LIYs (Fig. 2) of
around 8%. Hence, with these assumptions, the surface fluxes
provide enough energy to reduce the ice concentration by the
observed amount. The results indicate that the thermodynamical
processes examined here can explain the ice retreat during the
LIYs. However, other processes may also contribute, for instance
wind forcing11,20 and ocean currents24 . As mentioned earlier,
thermodynamical and dynamical processes are probably linked,
because positive anomalies of energy convergence are associated
with cyclone and frontal activities and enhanced winds.
For the HIYs, the case is essentially the opposite: convergence
anomalies of both latent and dry-static energy are negative during
late spring and summer (Supplementary Fig. S10). This is consistent
with negative water-vapour and cloud anomalies that act to decrease
the atmospheric opacity and weaken the greenhouse effect (Supplementary Fig. S10a). The sum of the net short-wave and long-wave
radiation plus turbulent fluxes becomes negative and the energy
flux to the surface is reduced by ∼3 W m−2 during April–August
(Supplementary Fig. S9 and Table S1). However, most of these
anomalies for HIYs are not statistically significant. An exception
is the atmospheric humidity, which is significantly smaller than average, both during late spring and throughout the summer season.
NATURE CLIMATE CHANGE | ADVANCE ONLINE PUBLICATION | www.nature.com/natureclimatechange
© 2013 Macmillan Publishers Limited. All rights reserved.
3
NATURE CLIMATE CHANGE DOI: 10.1038/NCLIMATE1884
LETTERS
a
5
0
0
¬5
¬5
Water vapour (10¬1)
Cloud water (10¬3)
Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Sep. Oct. Nov. Dec.
10
LWN + LH + SH
Jdry =
10
SIC
5
5
0
0
¬5
¬5
Conv. of latent energy
Conv. of dry-static energy
¬10
¬10
Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Sep. Oct. Nov. Dec.
Figure 3 | Atmospheric water content and energy convergence anomalies
for LIYs. a,b, Anomalies of SIC (black, right-hand axis) and net long-wave
radiation plus turbulent fluxes (red) are shown together with anomalies of
total column cloud water (liquid plus solid; blue) and water vapour (green)
(a), as well as the convergence of dry-static (green) and latent atmospheric
energy transport (blue; b).
The sea-ice anomalies have a clear but lagged effect on
the atmosphere, when the autumn freeze-up sets in by mid
September: in LIYs, cold air over open water induces enhanced
long-wave radiation plus turbulent fluxes from the surface
to the atmosphere over the ice-retreat region (Fig. 2). As a
result, the atmospheric energy transport convergence is reduced
(Fig. 3b). The opposite pattern of both the surface fluxes and the
atmospheric transport is encountered during autumn for HIYs
(Supplementary Figs S9 and S10).
For years when the Arctic SIC becomes anomalously low in
the end of summer, two processes seem important: an enhanced
atmospheric convergence of moisture during spring over the iceretreat area, leading to an increase of the greenhouse effect, due to
positive anomalies of the water vapour and cloudiness, and to an
increase of the long-wave radiation to the surface; and an enhanced
atmospheric convergence of dry-static energy during summer over
the ice-retreat area, which acts to increase the energy-flux to the
surface from long-wave radiation and sensible heat flux. When the
ice anomaly begins to appear, the net short-wave radiation anomaly
becomes positive. Hence, short-wave radiation is having little effect
when the ice anomaly is initiated, but acts as an amplifying feedback
in response to the melt. We emphasize again that although the
energy transport plays a major role for the September sea ice during
LIYs, wind20 and ocean-current24 anomalies may also be important.
Methods
The ice extent (Fig. 1a) is obtained from the Scanning Multichannel Microwave
Radiometer (SMMR; October 1979–August 1987) and the Special Sensor
Microwave/Imager (SSM/I; July 1987 to present) onboard the Nimbus-7 satellite
and Defense Meteorological Satellite Program, respectively. The data are provided
by the National Snow and Ice Data Center25 (NSIDC). Years with an anomalous
small/large September sea-ice extent are defined as years where the sea-ice extent
deviates with more than ±0.5 s.d. from the linear trend line taken over 1979–2010
(Fig. 1). This definition results in 10 LIYs and 12 HIYs. The linear trend is estimated
using a least-squares linear regression.
Various atmospheric fields are taken from the ERA-Interim reanalysis from the
European Center of Medium Range Weather Forecast21,26 (ECMWF). A reanalysis
product, which is blended model results and observations, provides a high degree
4
of consistency among the variables. At present, ERA-Interim is arguably among
the best data sets for the Arctic (Supplementary Section S3). For the time series
of radiative and turbulent fluxes (Figs 2 and 3 and Supplementary Figs S1, S3, S9
and S10), 24-h forecast accumulations, initiated at 00 utc, are used. For all other
variables 6-hourly analysis are averaged to daily values (Fig. 3 and Supplementary
Figs S6 and S7). The energy transport is estimated with a 6-hour resolution on
model hybrid levels and vertically integrated from the top to the bottom of the
atmosphere27 . A barotropic mass-transport correction is applied28 . The dry-static
and latent energy transport are defined as
¬10
Sea-ice anomaly (%)
Flux anomaly (W m¬2)
10
SIC
5
¬10
b
LWN + LH + SH
Sea-ice anomaly (%)
Flux anomaly (W m¬2)/
water anomaly (kg m¬2)
10
1
g
1
∂p
v ·v + cp T + gz
dη
2
∂η
0
Z
1 1
∂p
Jlatent =
vLq dη
g 0
∂η
Z
1
v
respectively, where g is gravity, v(u,v) is the horizontal wind with v as the
northward and u as the eastward component, cp is the specific heat capacity
of moist air at constant pressure, T is temperature, z is geopotential height,
p is pressure, L is the specific heat of condensation, q is the specific humidity
and η is the vertical hybrid coordinate used in the ERA-Interim reanalysis. All
data used have a 0.5◦ × 0.5◦ horizontal resolution. Ice concentrations are also
taken from ERA-Interim (except for Fig. 1a). Note that the ERA-Interim ice
concentrations are neither modelled nor assimilated but taken directly from
the satellite observations. Two days with obviously erroneous SICs are removed
from the time series.
Anomalies of all fields are estimated for the LIYs and the HIYs. As in the
procedure for the sea-ice extent, these anomalies are calculated relative to the linear
trend over 1979–2010 for each grid point. By estimating the anomalies relative
to the linear trend rather than the mean, the focus is held on the year-to-year
variability rather than the long-term change. This climatology consisting of a linear
trend is referred to as average in the text.
Statistical significance is tested using a Monte Carlo approach29 . At least
1,000 artificial LIY/HIY composites are randomly generated and compared to
the original LIY/HIY composite. Hereby the null hypothesis that the anomaly of
the original composite is not different from zero is tested. If the null hypothesis
can be rejected the anomaly values of the composite are significantly different
from zero. For example, the original composite is significant on a 95% level if
less than 5% of the absolute values of the artificial composites are larger than
the anomalies of the original composite. The advantage of the Monte Carlo
method is that the data are not assumed to follow any particular statistical
distribution. This is in contrast to, for example, a Student t -test, which assumes
that data follow a normal distribution. Thus, robust results are expected, even for a
relatively small sample size.
Received 17 September 2012; accepted 26 March 2013;
published online 28 April 2013
References
1. ACIA Impacts of a Warming Arctic: Arctic Climate Impact Assessment
(Cambridge Univ. Press, 2004).
2. Stroeve, J. C. et al. Tracking the Arctic’s shrinking ice cover: Another extreme
September minimum in 2004. Geophys. Res. Lett. 32, L04501 (2005).
3. IPCC Climate Change 2007: The Scientific Basis (eds Solomon, S. et al.)
(Cambridge Univ. Press, 2007).
4. Serreze, M. C. & Barry, R. G. Processes and impacts of Arctic amplification: A
research synthesis. Glob. Planet. Change 77, 85–96 (2011).
5. Serreze, M. C., Holland, M. M. & Stroeve, J. Perspectives on the Arctic’s
shrinking sea-ice cover. Science 315, 1533–1536 (2007).
6. Holland, M. M. & Bitz, C.M. Polar amplification of climate in coupled models.
Clim. Dynam. 21, 221–232 (2003).
7. Lindsay, R. & Zhang, J. The thinning of Arctic Sea Ice, 1988–2003: Have we
passed a tipping point? J. Clim. 18, 4879–4894 (2005).
8. Comiso, J. C., Parkinson, C. L., Gersten, R. & Stock, L. Accelerated decline in
the Arctic sea ice cover. Geophys. Res. Lett. 35, L01703 (2008).
9. Overland, J. E., Wang, M. & Salo, S. The recent Arctic warm period. Tellus 60A,
589–597 (2008).
10. Graversen, R. G., Mauritsen, T., Drijfhout, S., Tjernström, M. & Mårtensson, S.
Warm winds from the Pacific caused extensive Arctic sea-ice melt in summer
2007. Clim. Dynam. 36, 2103–2112 (2010).
11. Ogi, M. & Wallace, J. M. The role of summer surface wind anomalies in
the summer Arctic sea ice extent in 2010 and 2011. Geophys. Res. Lett. 39,
L09704 (2012).
12. Maksimovich, E. & Vihma, T. The effect of heat fluxes on interannual variability
in the spring onset of snow melt in the central Arctic Ocean. J. Geophys. Res.
117, C07012 (2012).
13. Francis, J. A. & Hunter, E. New insight into the disappearing arctic sea ice.
EOS Trans. Am. Geophys. Union 87, 509–511 (2006).
NATURE CLIMATE CHANGE | ADVANCE ONLINE PUBLICATION | www.nature.com/natureclimatechange
© 2013 Macmillan Publishers Limited. All rights reserved.
NATURE CLIMATE CHANGE DOI: 10.1038/NCLIMATE1884
LETTERS
14. Kay, J. E., L’Ecuyer, T., Gettelman, A., Stephens, G. & O’Dell, C. The
contribution of cloud and radiation anomalies to the 2007 Arctic sea ice extent
minimum. Geophys. Res. Lett. 35, L08503 (2008).
15. 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).
16. Polyakov, I. V. et al. Arctic Ocean warming contributes to reduced polar ice cap.
J. Phys. Oceanogr. 40, 2743–2756 (2010).
17. Maslanik, J. A. et al. A younger, thinner Arctic ice cover: Increased potential
for rapid, extensive sea-ice loss. Geophys. Res. Lett. 34, L24501 (2007).
18. Perovich, D. K., Richter-Menge, J. A., Jones, K. F. & Light, B. Sunlight,
water, and ice: Extreme Arctic sea ice melt during the summer of 2007.
Geophys. Res. Lett. 35, L11501 (2008).
19. 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).
20. Ogi, M., Yamazaki, K. & Wallace, J. M. Influence of winter and summer
surface wind anomalies on summer Arctic sea ice extent. Geophys. Res. Lett. 37,
L07701 (2010).
21. Dee, D. P. et al. The ERA-Interim reanalysis: Configuration and performance
of the data assimilation system. Q. J. R. Meteorol. Soc. 137, 553–597 (2011).
22. Kwok, R. et al. Thinning and volume loss of the Arctic Ocean sea ice cover:
2003–2008. J. Geophys. Res. 114, C07005 (2009).
23. Zhang, J., Lindsay, R., Steele, M. & Schweiger, A. What drove the dramatic
retreat of arctic sea ice during summer 2007? Geophys. Res. Lett. 35,
L11505 (2008).
24. Polyakov, I. V. et al. One more step toward a warmer Arctic. Geophys. Res. Lett.
32, L17605 (2005).
25. Fetterer, F., Knowles, K., Meier, W. & Savoie, M. Sea Ice Index (National Snow
and Ice Data Center, 2002, updated 2009).
26. Simmons, A., Uppala, S., Dee, D. & Kobayashi, S. ERA-Interim: New ECMWF
reanalysis products from 1989 onwards. ECMWF Newslett. 110, 25–35 (2006).
27. Graversen, R. G. Do changes in the midlatitude circulation have any impact on
the Arctic surface air temperature trend? J. Clim. 19, 5422–5438 (2006).
28. Trenberth, K. E. Climate diagnostics from global analysis: Conservation of
mass in ECMWF analysis. J. Clim. 4, 707–721 (1991).
29. Wilks, D. S. Statistical Methods in the Atmospheric Sciences (Academic Press,
1995).
Acknowledgements
This work is part of the ADSIMNOR programme, funded by a grant from the Swedish
research council Formas. The ECMWF ERA-Interim reanalysis data are obtained from
the ECMWF data server and the sea-ice extent from the NSIDC.
Author contributions
The original idea for the paper was suggested by R.G.G. and discussed and developed by
all authors. The data analysis was carried out by M-L.K., who also prepared the figures.
M-L.K. and R.G.G. wrote the manuscript and M.T. provided feedback. All authors
contributed to the discussion and interpretation of the results.
Additional information
Supplementary information is available in the online version of the paper. Reprints and
permissions information is available online at www.nature.com/reprints. Correspondence
and requests for materials should be addressed to M-L.K.
Competing financial interests
The authors declare no competing financial interests.
NATURE CLIMATE CHANGE | ADVANCE ONLINE PUBLICATION | www.nature.com/natureclimatechange
© 2013 Macmillan Publishers Limited. All rights reserved.
5