Introduction
Along with expectations for a warming planet, the Arctic sea ice cover
continues to decrease. During the last two decades the Arctic sea ice
cover has become both thinner and smaller in extent
. In September 2012 the lowest observed
September sea ice extent (SIE) since the satellite record started in
1979 occurred. The 2012 minimum was 18 % lower than in 2007,
and 49 % below the 1979–2000 average minimum
. A number of processes have been suggested to
explain parts of the sea ice loss, such as differences in surface air
temperatures, the poleward advection of heat in both the ocean and
atmosphere, greenhouse gas forcing, and cloud cover.
Changes in ice export is also considered an important contributor to
variability in the Arctic sea ice cover
. Historically, about 10 % of
the Arctic sea ice area is exported through Fram Strait annually, and
the ice export through the other Arctic gateways are on an order of
magnitude smaller . Because quite thick ice is lost
this way a larger than normal export decreases the
remaining mean thickness in addition to the sea ice covered area
inside the Arctic Basin.
Recent studies on sea ice motion have shown that sea ice drift speed
is increasing in the Arctic Basin , and
in Fram Strait . Positive trends were
also found in cumulative Fram Strait ice area export by
(4 % per decade in the period 1950–2000)
and (5 % per decade in the period
1957–2010). Contrary to these studies, found a small
negative trend in cumulative Fram Strait ice area export in the
similar period 1982–2009, but with positive trends for both annual
(October–September) and summer (June–September) from
2001–2009. did not observe any significant change
in Fram Strait ice volume export for the period 2003–2008.
The Arctic seasonal maximum sea ice cover occurs in late February or
early March . Ice export through Fram Strait between
March and August could therefore influence the following September
SIE, because sea ice is mostly melting within the Arctic Basin during
these months . However, such an
influence has not been found so far. Our new timeseries shows such an
influence, and suggests that high accuracy monitoring of the Fram
Strait ice area export should be continued and would contribute to
seasonal forecasts of the next September SIE. This paper focuses on
the continued high level of area export and the relation between
spring and summer sea ice export and the following September Arctic
SIE.
Data and methods
In this study we used observed sea ice drift speed onwards from
February 2004 and updated through December 2013. The drift is
calculated by recognising displacement vectors manually on Envisat
Advanced Synthetic Aperture Radar (ASAR) WideSwath and Radarsat-2
ScanSAR (from 2012) images captured 3 days apart
. Displacement vectors that cross
79∘ N were linearly interpolated to bins (1∘
longitude, each 21 km) from 15∘ W to
5∘ E (Fig. ). For most 3 days image
pairs, displacement vectors with an accuracy of about
±3 km is found with a spacing of 30–50 km.
This accuracy of about ±3 km per vector is considered
sufficient for most analysis. This is because subsequent averaging
or addition in time/space of many unbiased vectors will generally
result in ice mean speed and ice export values with a significantly
better accuracy than 10 %. We only use the monthly mean
cross-strait value of ice drift speed here, being the
spatial-temporal mean southward speed of all ice crossing
79∘ N (Fig. ) between the fast ice edge and
the pack ice edge at 50 % concentration. On the western
side of the strait a linear interpolation from zero motion in the
stable fast ice to the first measured motion vector is made. It is
assumed that ice displacement to the east of the last measured
vector is constant near the ice edge. The mean speed value results
from the averaging of about 50 individual, unbiased displacement
vectors, thus the calculated mean speed value should be
considerably more accurate than 10 %.
Ice area export over consecutive 3-day-periods along 79∘ N
is a product of sea ice drift data and passive microwave sea ice
concentration data . We use cumulative (monthly
and yearly) ice area export onwards from 2004. From here on ice
area export will be referred to as ice export.
Observed monthly mean Sea Level Pressure (mSLP) values were used
onwards from 1979. The cross-strait difference along
78∘ N was then calculated between 18∘ W and
15∘ E. Monthly mSLP observations was obtained from
Longyearbyen (Fig. , Svalbard Airport, Norwegian
Meteorological Institute, http://eklima.met.no). On the
Greenland side surface pressure at 18∘ W was constructed
using weighted average of monthly mSLP from two nearby stations;
Danmarkshavn and Nord (Fig. , Danish Meteorological
Institute, ). mSLP from Danmarkshavn and
Nord correlated well (r=0.94) and allowed calculations of
cross-strait geostrophic winds following , and
correlations and linear regressions between this wind and observed
sea ice drift speed or ice export.
We used National Snow and Ice Data Center ice extent since 1979
. The ice extent is defined as the area
covered by ice greater than 15 % ice concentration, and
has been derived from Scanning Multichannel Microwave Radiometer
(SMMR) and Special Sensor Microwave/Imager (SSM/I) brightness
temperatures using the NASA Team sea ice algorithm.
A coupled climate model (Geophysical Fluid Dynamics Laboratory
Coupled Model version 2.1 (GFDL), ) is used
to further investigate the coupling between sea ice export and
September Arctic SIE. In the ocean the model has 50 vertical levels
and 1∘ zonal resolution. Its atmosphere component has 24
vertical levels, with horizontal resolution of 2∘ latitude × 2.5∘ longitude. We mostly utilize
a 3600 year control simulations were the model is forced by
constant radiative conditions representative of 1860. This produces
a stable multi-millennium control integration without flux
adjustments, that simulate most features of the relationship
between atmospheric/oceanic forcing and Arctic sea ice cover
realistically .
Results and discussion
The most dominant forces acting on sea ice are the wind drag and
the ocean drag . On short time scales (from
days to months), geostrophic winds alone explain above
70 % of the variance of the ice speed in the Arctic Ocean
, and winds are also the dominant force acting
on sea ice in Fram Strait . Here ice drift is in
the south-southwesterly direction, and a mean southward speed of
16 cms-1 has been reported based on moorings at
5∘ W in the period 1997–2000 .
Figure shows that the temporal mean speed is quite
constant across the strait eastward of 5∘ W, and that the
speed decreases westward towards the Greenland coast. Velocities
are clearly strongest during winter, with mean speeds above
20 cms-1 decreasing to less than 10 cms-1
during summer in the outer part (Fig. ) The ice export
occurs mostly between 5∘ W and the Greenwich Meridian. The
export is limited on the western side by the decreasing ice speed,
reaching zero at 16∘ W were stationary land fast ice is
usually found, and by zero concentration on the eastern side
varying from 5∘ W to 5∘ E (Fig. 1 in
). Export estimates from passive satellites
and those based on mSLP difference and SAR
tracking are largely consistent for the 80's and 90's
.
Atmospheric forcing
From a linear regression between monthly mean ice drift speed and
geostrophic wind we find that the ice in Fram Strait drifts at
a speed being 1.6 % of the geostrophic wind speed
(Eq. ). The constant contribution resulting from the
linear regression, which is the speed of the ice given no local
wind forcing, is 6.5 cms-1 (Eq. ). Both
terms are comparable to previous studies
. The standard error
of the regression was 3.5 cms-1.
Vice=0.016×Vg+0.065[ms-1]
The constant contribution of 6.5 cms-1 represents the
mean (not wind-driven) ocean current. A nonlinear component of the
ice drift including forces from variations in ocean currents or
internal ice stress may also represent parts of this constant
. The correlation between monthly mean ice
speed and geostrophic wind was good, r=0.76 with the
95 % confidence interval [0.67 0.83]. We also correlated
the monthly mSLP with the maximum monthly sea ice drift speed along
79∘ N, which resulted in similar correlations. Analyzing
the east-west variations showed that 33.0 % of the
1∘ longitude bins reached the monthly maximum speed. This
means that maximum speed and mean speed relate to the wind forcing
in the same way, and that the east-west averaging contains the
variability. The correlation between Fram Strait cross-strait mSLP
difference and ice export was r=0.73.
Ice drift seasonality
The annual cycle of mSLP-based ice speed (and export) is similar to
earlier estimates , with higher speed during
winter, and weaker during summer. Annual mean speed is close to
12 cms-1 (Fig. ). This mean value
is a spatially avaraged value between 15 and
5∘ W, and a temporal average for the years 1979–2013.
Based on the NCEP reanalysis data found
significantly lower mean ice speed in the 1960's and 1970's.
The previous assumption of a constant contribution throughout the
year of 6.5 cms-1 is challenged by our new results. We
discovered a seasonal difference between the observed ice speed and
the estimated mSLP-based speed (Fig. ). The
difference suggest that the ice speed should be
3.0 cms-1 stronger during winter (December–April),
and 2.8 cms-1 weaker during summer (June–October).
This is inconsistent with vartiations in the internal ice stress;
ice is thicker and more dense during winter, resulting in a larger
ice stress and thereby weaker ice drift speed for a similar wind.
An increase in the mean ocean current below the ice during winter,
which in this case is the East Greenland Current (EGC), is
consistent with the generally stronger winds in the North Atlantic
region during winter. This suggest that the EGC is responding to
the larger scale windforcing as well as to the local winds.
Generally the entire circulation along the continental slope of the
Arctic Basin – Nordic Seas is driven by the wind-stress curl north
of the Greenland-Scotland ridge . Two recent
studies confirm that the EGC is stronger during winter and is
responding to the large-scale wind stress curl in the Nordic
Seas. It is thus likely that this is causing the additional winter
export (Fig. ). De Steur et al. (2014) analysed
mooring data along 79∘ N between 1997 and 2009, and found
that surface currents were below 5 cms-1 during summer
and 10–15 cms-1 during winter, also varying in the
east-west direction. found a maximum in the
flow in January and a minimum in July for the years 1992–2009
based on altimetry at 60∘ N, and that the vertical
distribution remained constant over this time period.
Assuming a stronger EGC during winter, and weaker during summer,
not related to the local winds, we re-calculated the timeseries
using this seasonal difference, and the corrected ice speed fitted
the observations in a much better way
(Fig. ). The same correction was made for the
calculated ice export, so the summer values were decreased and the
winter values increased accordingly (not shown). The seasonal
correction improved the correlations between observed and
mSLP-based ice drift (r=0.88 for ice speed and r=0.87 for ice
export). This means that our new timeseries explains close to
80 % of the observed ice drift variability. The constant
contribution from the EGC during summer makes up
∼34 % of the mean monthly summer ice export of
42.1 × 103 km2, and the contribution during winter makes up ∼55 %
of the mean winter export of 110.5×103 km2 (not
shown).
Annual mean ice export and trends: 1979–2013
Using this newly confirmed seasonal variation explained by the
EGC, in addition to the local winds calculated from observed mSLP,
we calculate the monthly mean ice export prior to 2004 when SAR
images became available. Monthly mean Fram Strait ice export
(Fig. ) was produced by using the mSLP-based
ice export including the seasonal correction from 1979–2003, and
the observed ice export from 2004–2013. Previous studies
used reanalysis mSLP
data, and our present study should therefore represent a more
accurate estimation. We discovered systematic differences between
the NCEP Fram Strait reanalysis mSLP data and observed mSLP in the
recent years that we are not able to explain (not shown), and
decided therefore to base the new ice export estimates solely on
observed mSLP. Prior to 2004 we do not utilize observations of
cross-strait variations in the width of the ice covered area, ice
speed or ice concentrations, but base our ice export values solely
on the regression equation found between observed mSLP from
Longyearbyen, station Nord and Danmarkshavn (Fig. )
and observed SAR ice export.
Figure shows an increase for both annual
and seasonal values. Hereafter we define annual ice export values
as winter centered averages (September–August), winter export
as averages from September through February, and spring export as
averages from March through August. The long-term annual mean ice
export is close to
880 × 103 km2year-1, thus higher than
previous studies suggest
. Fluctuations from year
to year are large, but the ice export has remained higher than the
long-term mean since 2006 (Fig. ). The
annual mean export has been larger than
1000 × 103 km2year-1 since 2010.
A high positive trend of +7.0 % per decade is found for
annual values from 1979 to 2013 (Fig. ).
This trend is consistent with increasing ice drift speed observed
in the Arctic Ocean . The trend is
largely produced by higher ice export during spring. The winter
ice export trend is +3.4 % per decade, while the spring
export trend is +13.9 % per decade over the
35 year long period (Fig. ).
We find comparable trends for the entire period (1979–2013) and
for those based on the observed mSLP only (1979–2003). The
trends are somewhat lower in the early period, but remain positive
(annual, winter, spring values are 3.5, 10.5 and 1.3 %
per decade, not shown). This indicates that the overall ice export
trend is related to the increase in mSLP across Fram Strait, and
has gradually increased since 1979. The increase in export is due
to stronger geostrophic winds, and for spring it is generated by
an increase of 0.6 hPa per decade in the mSLP on
Greenland. The mSLP on the Svalbard side has a slighly lower,
negative trend. In the 1980's the spring export was approximately
half of the winter export. The robust trend in spring export has
resulted in a smaller seasonal difference in recent years, and
since 2011 the two are close in magnitude
(Fig. ).
We note that other results are
consistent with ours for the 80's and 90's but differ in the
period since 2004. This is when we have SAR based ice velocity
data available, and we believe that the cause of the differences
lay in that the passive satellite observations are to coarse to
resolve the high speed export events during winter
.
Recent annual export
The annual mean ice area export since 2010 is
1100 × 103 km2 (Fig. ).
This high export means an overall increase in how much of the
Arctic Basin ice cover is “lost” each year. The Arctic Basin
covers an area of ∼7.8 million km2, and has been fully
ice covered from November through May since 1979 until today. The
annual ice export during the 1980's
(∼800 × 103 km2) was 10 % of this
winter ice covered area. However, during the last three years the
annual ice export has increased to 14 % of this area,
meaning a 40 % increase in the relative sea ice area
export, or the large-scale divergence of the Arctic Basin sea ice
cover. Further, an estimate of the new annual mean ice covered
area inside the Arctic Basin is 7.0 million km2. This is
based on reduced monthly mean ice extent between June and October.
Using this updated annual mean Arctic Basin ice cover of
7.0 million km2, the Fram Strait ice export has been
16 % of the area since 2010. This is a 60 %
increase in relative sea ice area export since the 1980's, and is
clearly connected with the overall Arctic sea ice decline. During
winter the open water anomalies created within the basin are
quickly refrozen, and one major effect of the increased ice export
since 1979 has therefore been to contribute towards the overall
observed thinning since the 1990's .
Ice export and September Arctic sea ice cover
A larger than normal ice export decreases the sea ice cover
directly inside the Arctic Basin by transporting sea ice out.
While the open water anomalies inside the basin re-freeze during
winter, they may be further enhanced by positive feedback
mechanisms during summer. This transition from winter and
re-freezing, to summer and positive feedback, occurs gradually
later in the year as one moves north, but melting will prevail
over most of the Arctic Basin onwards from May .
In this section we will discuss the direct effect of ice export
anomalies onwards from March. Altough there may be some
re-freezing in March, April and May, the newly formed ice will be
thin, have a thin snow cover, and therefore easily melt and
deform. We therefore first summarize the export anomalies onwards
from March, and note that the Arctic seasonal maximum SIE occurs
in late February or early March. We thus summarize open water
areas created by export onwards from March, and assume that they
will contribute directly to open water areas in September the same
year. The areas may not refreeze due to solar radiation and
warmer air temperatures, or they may form thin ice that melts
later in the summer. The indirect effect and the ice-albedo
feedback will be discussed in the next section.
Previous examination of the contribution of summer
(June–September) ice export to the loss of multiyear ice cover
from 2005 to 2008 indicated a small contribution.
Our results show something different. The annual spring ice
export has been about 500 × 103 km2 since 2008
(Fig. ). This is a 100 % increase
compared to the spring export in the early 1980's, so in recent
years an additional ice export of ∼250 × 103 km2 has occurred during spring. Over
the same timeframe the September SIE has decreased ∼2500 × 103 km2, so the increased spring export
can directly explain about 10 % of the loss
(Fig. ).
The low September SIEs in 2007 and 2012 thus seems related to high
spring export these years (Fig. ). Likewise
did the export decrease in 2013, and the September 2013 extent
recovered. Overall the two series correlate well, with r=-0.57
for the un-trended values, and with r=-0.38 when the trends are
removed. We return to the correlation values in
Sect. . Not all high export events are followed by
an anomalously low September SIE, for example in 1990 and 2000.
One suggestion to explain the missing response for a high spring
export and normal September SIE lies in the thickness variability
and its interaction with the regional wind forcing within the
Arctic Basin. The thinner ice cover in more recent years will more
easily deform and compact given a convergent wind field, while in
the past the thicker ice cover may resist such wind forcing. The
thickness in Fram Strait has thinned with about 1 m since
1992, and the average age of the exported ice has decreased from 3
to 2 years, but there are large year-to-year fluctuations
.
One suggested mechanism for the rapid decline in summer Arctic SIE
is that a larger winter export could create more even and thin
first year ice. This more even first year ice may have a larger
fraction of melt ponds during summer. found
a strong correlation between simulated spring melt pond fraction
and September Arctic SIE. We find that the correlation between
winter ice export and the following September SIE is quite low.
For un-trended values r=-0.28, and for de-trended values
r=-0.18. The low increase in winter ice export over the last
35 years (3.4 % per decade) also suggest that
summer ice loss and September SIE is not particularly sensitive to
winter sea ice export.
Role of positive feedbacks
In addition to the direct contribution of about 10 % on
the September SIE from spring export through Fram Strait, the loss
may be further enhanced by positive feedback mechanisms during
spring and summer. Of these the ice- albedo feedback
, is the best known, but a thinner ice cover
will also have a smaller resistance towards wind forcing
and more easily deform. Both amplifies
anomalies in thinner ice cover and leads to more open water areas.
There are significant trends for the melt onset inside the Arctic
Basin, and melt starts about 10 days earlier today than
around 1980 . Melting generally starts in late
May, but this does not mean that export anomalies have no effect
until then. As noted by are early formations of
open water areas important as it boosts the ice-albedo feedback.
A positive export anomaly will lead to a positive solar heating
anomaly because open water is created. We estimate the loss of sea
ice due to the combined effect of increased ice export and the
related ice-albedo feedback below. Weather the heat anomaly just
prevents further growth or leads to direct melting is not
specified.
The additional solar heating of the upper ocean due to increased
ice export can be estimated from the incident solar irradiance, the
change in surface albedo, and the change in open water area
. To estimate the additional solar heating we
used the change in ice export from the beginning of this study
period (1979–1981 average) to the most recent period (2011–2013
average). Between these two periods the additional open water area
from increased ice export has increased ∼63, 105, 140, 155,
196, and 214 × 103 km2 from March through
August. The monthly mean solar radiation is ∼33, 142, 257,
302, 233 and 133 Wm-2, from March through August,
observed at Russian North Pole drift stations (Table 1,
). The relevant change in surface albedo can be
estimated as 0.53, i.e. the difference between a (melting) sea ice
albedo of 0.6, and an open water albedo of 0.07
.
The additional open water areas between March and August lead to an
additional solar heating of ∼2.44×1020 J.
This extra heat is enough to melt or prevent growth of
0.800 × 103 km3 of ice. Choosing a mean ice
thickness of 1.5 m for the ice
(melted or prevented from forming) suggests an effected ice area of
531 × 103 km2. This is twice the size of
the original spring export anomaly, and increases the effect of the
spring export anomaly to almost 30 %. This value appears
when using a recent September Arctic SIE value of
∼5000×103 km2
(Fig. ). If only the direct effect is used
until June, and then the feedback is applied onwards the total
effect still explains about 20 %.
Overall effect of spring export
The above physically based estimates stated a direct effect of
about 10 % rising to 30 % when including the ice
albedo feedback. This was however only accounting for the trends,
or long term changes, while there is additional year to year
variability of both September SIE and the spring export.
Because we expect the ice export to both influence the trend and
the variability, we do not remove the trend initially. The
overall correlation between annual spring ice export and the
following September mean SIE (1979–2013) is r=-0.57, with the
95 % confidence interval [-0.76 -0.29]. This means
that the ice export explains 31 % of the variance in the
mean September SIE. For the recent ten years (2004–2013) when our
ice export values are directly observed by SAR and the Arctic sea
ice cover has been thinner and more responsive the correlation is
r=-0.76 (95 % confidence interval [-0.94
-0.26]). This means that 57 % of the variance can be
explained for this period. This indicate a growing influence of
the spring sea ice export in recent years.
Removing the linear trends results in a correlation r=-0.38,
with the 95 % confidence interval [-0.63 -0.05].
This indicates that ∼14 % of the variance in the
de-trended September SIE can be explained by the de-trended spring
ice export. This shows clearly that the ice export influences the
trend in September SIE, but in addition it also influences the
year-to-year variability (Fig. ). Again is
there higher correlations for the more recent 2004–2013 period,
with r=-0.57.
From a linear regression between annual spring Fram Strait ice
export and mean September Arctic SIE (Fig. ), we
find that the September SIE is -7.3 times the spring ice export
the same year. For an export anomaly of
100 × 103 km2, the response is thus a loss
of September SIE of around
700 × 103 km2. In the above calculations we
were able to explain a loss of ∼300 × 103 km2 including the ice albedo
feedback, so clearly there are also other contributors to the
overall loss of September SIE. The slope of the curve is set
mainly by the recent years when September SIE is below
6000 × 103 km2 (Fig. ), when
we also found the higher correlation and variance explained.
We are lacking updated pressure observations from Greenland to
extend our time-series consistently through 2014. However we note
that the September minimum increased from 2013 to 2014, consistent
with the 2014 spring export being 15 % lower than the
spring export in 2013 (not shown). There is thus significant
year-to-year variability and the spring export seems to have an
increasingly important role to play in explaining this
variability. The positive trend for earlier onset of melt
is also consistent with the higher correlation
between the spring export and the September minimum for the last
decade.
The total loss of September SIE, and also of Arctic sea ice loss
for other months of the year, is driven by a number of factors.
The sea ice export is only one of these factors, but one that
seems to be largely overlooked until today. The major contributor
to the overall loss is probably increased longwave radiation
related directly to increased green house gases
. In addition has increased Atlantic and
Pacific ocean heat transport into the Arctic ocean likely
contributed , and atmospheric heat transport
probably played a role in the overall thinning between the 1960's
and the 1980's . To further investigate the
proposed link between spring sea ice export and the following
September SIE we utilize fully coupled climate model simulations
below.
Coupled climate model simulations
The 3600 year control simulation from the Geophysical
Fluid Dynamics Laboratory (GFDL) Coupled Model version 2.1
presented in was further investigated here.
This control simulation represents pre-industrial levels of
greenhouse gases, and should represent the natural variability of
forcing and response for the Arctic sea ice in a good way. The
simulated long-term climatology of the annual mean Fram Strait ice
area export is about
1457 × 103 km2year-1. This ice export
is higher than that of the long-term mean presented here
(880 × 103 km2year-1), but this is
a common problem for many climate models .
The standard deviation of simulated unfiltered spring Fram Strait
ice area export is
327 × 103 km2year-1, while the
observed standard deviation is
83 × 103 km2year-1. This difference is
related to the much higher mean state in the model. The standard
deviation of simulated 30 year low-pass filtered spring
Fram Strait ice area export is
69 × 103 km2year-1.
The main predictors of low-frequency variability of summer Arctic
SIE was identified by as northward Atlantic heat
transport, Pacific heat transport, and the spring Arctic Dipole
(AD). Note that the time series used were 30 y low-pass filtered
to extract the low-frequency variability. The ocean heat
transport was calculated across the Arctic circle, and the
simulation showed similar correlation between Arctic SIE and ocean
heat transport on the Atlantic and Pacific side (r=-0.50 and
r=-0.51, both at 2 y lead).
The AD index is defined as the second leading mode (PC2) of spring (April–July) SLP anomalies within the Arctic Circle.
Here a positive AD is defined as having a positive SLP anomaly over Greenland and a negative SLP anomaly over the Kara and Laptev Seas,
which is efficient in causing enhanced transpolar ice drift. The AD influence on September Arctic SIE is strongest in spring,
and we would like to investigate this influence further.
Anomalous spring AD were significantly anticorrelated with the
September Arctic SIE anomalies in the control simulation (r=-0.37 at 1 y
lead for 30 year low-pass filtered anomalies, and r=-0.41 for unfiltered anomalies),
and we believe that one of the main mechanisms
for the AD's influence on September SIE is through the Fram strait ice
export, which has increased significantly the last decade (Fig. ).
It is this pressure gradient that has increased and lead to stronger southerly winds in
the Fram Strait. So an higher AD value leads to more ice being exported out of the Arctic Basin through Fram Strait.
This relation between AD and ice cover anomalies has been noted both in observations and models ,
and is suggested to play a large role for future Arctic SIE variability .
The simulated response is stronger in spring and summer and much weaker during autumn and winter ,
consistent with the stronger correlation we found between September Arictic SIE and spring ice export than for the winter ice export.
Using the 3600 year long control simulation we find that the
simulated spring Fram Strait ice export is indeed significantly
correlated with the AD index, for both unfiltered (r=0.63,
Fig. ) and 30 year low-pass filtered
(r=0.59, not shown) time series. This confirms this expected link
between the overall atmospheric circulation and the Fram Strait
export, also over much longer times than the last three decades when
observations are available. In the simulations a larger correlation is
found between the AD index and other months of spring export than
March–August quoted above. For April–August r=0.72, and
April-July (r=0.71), this is reflecting the AD index defined for the
April–July period as well.
For the unfiltered 3600 year simulation, spring Fram Strait
ice area export also has an anti-correlation with the September Arctic
SIE (r=-0.34, Fig. ). This is similar
to the correlation between AD and the September SIE of r=-0.41
mentioned above, suggesting that a substantial part of the coupling
between the AD and the September SIE can be explained by the Fram
Strait spring ice export. This ∼10 % of the total
simulated variance in September SIE that can be explained by the
spring export, is consistent with what we found for the observations
when we de-trended the September SIE (r=-0.38). For the
30 year low-pass filtered time series, this correlation is
lower (r=-0.15). This indicates that the spring export influences
the September SIE mostly on yearly to decadal time scales, and that
the lower frequencies (>30 years) is dominated by the ocean
heat transport. Similar correlations to r=-0.34 appear between the
September SIE and the spring ice export when the export is evaluated
over other months than March–August. Using April–August we found
r=-0.31, and April–July r=-0.32, indicating that the main part of
the signal is carried by the months of April–July.
Given that the control run only simulates natural climate variability
we also investigated the observed AD anomalies since 1979 from the
NCEP/NCAR reanalysis SLP dataset. As shown by is
the observed September Arctic SIE anomaly anticorrelated with the the
observed AD anomalies (r=-0.53). What is further added here is an
explanation on how this link is physically working. The correlation
between September SIE and the spring export was r=-0.57
(Fig. ), so it provides a physical explanation for
the AD correlation. The correlation is also slightly higher between
the observed spring export and the September SIE, than for the
observed AD and September SIE, but the difference is not significant.
So if the goal is to explain September SIE variability the spring
export is a more physical and stronger link than AD. Consistent with
this view is the observed spring export and AD index also highly
correlated (r=0.47, Fig. ). What further
now has become clear is that the AD spring export link also exist in
the control simulation representing natural climate variability over
3600 years. As noted above was the correlation r=0.63
between AD and spring export there.
We also examined a forced historical simulation for the 20th century
combined with a forced 21th century projection under the CMIP3 A1B
scenario using the GFDL model. Such a simulation is forced with
changes in all external forcings such as anthropogenic greenhouse
gases and aerosols. In these results we find no significant trend in
simulated spring Fram Strait ice area export between 1979 and 2013.
This suggests that the observed increase in spring Fram Strait ice
area export since 1979 (Fig. ) is not due to
changes in anthropogenic external forcing, but due to internal natural
variability.
Concluding remarks
A new and updated timeseries of Fram Strait ice area export from
1979–2013 was presented in this study. The new timeseries was
constructed using high resolution radar satellite imagery of sea
ice drift across 79∘ N from 2004–2013, regressed on the
observed cross-strait surface pressure difference back to 1979.
Stronger geostrophic winds, largely due to an observed increase in
the surface pressure on Greenland, has created a high positive
trend of 7 % per decade for annual mean ice area export
since 1979 (Fig. ). The trend is mostly
explained by the high trends for spring and summer months, when
ice export (March–August) has a robust trend of 13.9 %
per decade.
The pressure difference from observed sea level pressure across
the Fram Strait on Svalbard and Greenland directly explained
53 % of the variance in the observed ice export. The best
fit between ice drift and geostrophic winds resulted in a seasonal
difference of ∼3 cms-1, suggesting that the ice
drift is higher during winter, and lower during summer, than can
be explained by local winds. The most likely explanation for this
is a seasonal variation in the underlying East Greenland Current,
driven by the large-scale wind forcing that are generally stronger
during winter than summer. The seasonal cycle has also been
confirmed by . The ice export based on observed
sea level pressure including a seasonal variation in the
underlying current explains almost 80 % of the observed
ice export variance.
The variability in sea ice area export is directly influencing the
September Arctic sea ice cover. For de-trended time series
1979–2013, and for simulations over 3600 years with the
GFDL coupled climate model, the Fram Strait spring ice area export
explains about 10 % of the September SIE variability.
This is explaining the previous noted link between the September
SIE and the Arctic Dipole , that show
anti-correlations at the same level (r=-0.41) for the natural
climate simulations. This influence is of the same order as the
direct response of the increased spring ice area export, where an
additional export of 250 × 103 km2 between
March and August in recent years, compares to about 10 %
of the total loss of September SIE since 1980
(Fig. ).
The sea ice area export influence on the September SIE also seems to have increased in recent years reflecting
a thinner and more mobile sea ice cover.
This influence is on the order of 30 % and appears from the plain correlation (r=-0.57), between not de-trended values
of Fram Strait spring ice area export and the September SIE between 1979 and 2013.
This link is also explaining the earlier noted correlations between the
observed Arctic Dipole anomalies and the September SIE (r=-0.53,
). One of the physical links is between Fram Strait ice export and September SIE, but this correlation
and the one between September SIE and the Arctic Dipole, and the correlation
between the Arctic Dipole and the Fram Strait export, are all on the level of r=0.5.
The increased influence can be explained by positive feedback mechanisms, such as the ice-albedo feedback and
increased deformation of thinner ice, that further enhance such summer anomalies in SIE.
Accounting for the ice-albedo feedback increases the observed annual increase
in spring area export of 250×103 to
750×103 km2, which is about 30 % of the observed September SIE loss.
The last 10 years the Arctic sea ice cover has decreased quite
rapidly, and the contributions from natural and greenhouse gas forcing
are still under debate. In our historical simulations we found no
trend in Fram Strait sea ice export, and we are not aware of results
suggesting a likely systematic change in the Arctic large-scale
circulation. We therefore find that the observed increase in ice
export documented here is caused by natural climate variability, and
that there is potential for a partial recovery of the Arctic September
SIE in the future when, or if, the spring ice export decreases. The
Arctic ice cover is now thinner and more mobile than before, and in
the present state it seems to have an increased coupling to the Fram
Strait ice area export. In the natural climate state this influence
used to be on the order of 10 %, but during the last three
decades the influence has increased to around 30 %.