The Atmospheric Response to Realistic Reduced Summer Arctic Sea Ice Anomalies - Arctic Sea Ice Decline: Observations, Projections, Mechanisms, and Implications - page 89

for the continued ice melt [ Polyakov et al. , 2005], and recent
work shows that heat from the Atlantic layer can penetrate
through the halocline into the upper ocean [ Walsh et al. ,
2007]. Maslanik et al. [2007] demonstrate that even though
the NAO is in the negative phase, the net local atmospheric
circulation in the Arctic is consistent with continued ice re-
duction. Another contributor that could hasten ice retreat in
summer is enhanced moisture in the Arctic leading to in-
creased downward longwave fluxes [ Francis and Hunter ,
2006]. Circulation trends since 1979 were found to be weak
by Deser and Teng [this volume], consistent with the view
that multiple mechanisms have led to recent ice declines.
The summer sea ice is expected to continue its decline based
on the recently documented decreases in winter ice [ Comiso ,
2008] and the warm Atlantic water headed for the Arctic
that is being tracked by various ocean observing programs
[ Polyakov et al. , 2007]. In a warmer climate large decreases
in summer sea ice may become more common, and while
the ice anomalies initially result from both atmospheric and
oceanic forcing, we hypothesize that they can, in turn, mark-
edly alter the air-sea exchanges of heat and moisture to sub-
sequently influence the large-scale climate.
The summer warming and sea ice reductions are corre-
lated with cold season circulation anomalies [ Wallace et al. ,
1996; Rigor et al. , 2002; Maslanik et al. , 2007], which lead
to changes in low-level horizontal temperature advection.
For example, reduced summer sea ice in the Barents-Kara-
Laptev seas is associated with anomalously low pressure
centered in the Arctic the preceding spring (April-June)
[ Deser et al. , 2000]. The general tendency toward lower
pressure in the Arctic [ Walsh et al. , 1996] from 1960 to 2000
is consistent with enhanced penetration of storms into the
Arctic. There has been an increase of warm season cyclone
count and intensity in the Arctic (north of 60°N) since late
1950s [ Serreze et al. , 1997; Zhang et al. , 2004]. Maslanik et
al. [1996] find an increase of cyclone activity over the cen-
tral Arctic Ocean, which advects warm southerly winds into
the Laptev and East Siberian seas as well as transports ice
away from the coast. Ice that is particularly thin as a result
of wintertime circulation patterns can be easily broken down
and transported because of summer storms, further reducing
ice area/concentration.
When high-albedo ice is replaced by low-albedo ocean,
there is significantly more net solar flux at the surface, in-
creasing the heat stored in the upper layer of the ocean. This
heat stored during the summer can then slow the freezeup
the following winter as well as melt ice at the ice/ocean in-
terface. There is still a reasonably strong correlation (+0.6)
between the time series of EOF1 of sea ice concentration
during the summer and that of the following winter [ Deser
et al. , 2000]. By August the sea surface temperature (SST)

can warm in the marginal seas by several degrees [ Steele et
al. , 2008] when ice extent is low. Fluxes of sensible and la-
tent heat into the atmosphere increase with a warmer ocean,
which, we hypothesize, could exert some influence back on
the atmosphere.
Climate models are ideal tools for understanding the influ-
ence of sea ice on the atmosphere because in the observa-
tions, climate anomalies are dominated by the atmospheric
forcing of the ice. Singarayer et al. [2006] ran the Hadley
Centre Atmospheric Model (HadAM3) with climatological
SSTs and observed sea ice concentrations from 1978 to 2000
to investigate the impact of sea ice on the atmospheric cir-
culation. The model surface air temperature (SAT) response
to ice forcing most closely matches the observed SAT vari-
ability over the 1993-1995 period [see Singarayer et al. ,
2006, Figure 4a]. This suggests that sea ice forcing played a
more important role than SST (note that this simulation used
climatological SSTs) in shaping the SAT anomalies. The
observed sea ice anomalies display large interannual vari-
ability in the mid-1990s and reached a low for the decade
in 1995. Singarayer et al. [2006] argue that the ice anoma-
lies were likely large enough that sea ice forcing dominated
the atmospheric response. The SAT response during sum-
mer strengthened and become statistically significant when
observed above-normal SSTs replace climatological values.
Sewall [2005] investigated the response to reduced Arctic
sea ice in the Community Climate System Model, version 3
(CCSM3) plus a suite of coupled Intergovernmental Panel
on Climate Change Fourth Assessment simulations and
found a robust pattern of reduced wintertime precipitation
for the western United States by ~30%.
Magnusdottir et al. [2004] and Deser et al. [2004] inves-
tigated the response to sea ice and SST anomalies during
winter in the North Atlantic using CCm3. The ice anomaly
pattern corresponds to an enhanced observed trend with ice
reductions (increases) east (west) of Greenland. Magnusdot-
tir et al. [2004] found a significant model circulation response
to sea ice that resembled the negative phase of the North
Atlantic Oscillation, which is opposite of the atmospheric
pattern that forced the observed sea ice trend, suggesting that
sea ice has a negative feedback on the atmosphere. There is
growing evidence that a model's internal variability influ-
ences its forced response. To investigate this further, Deser
et al. [2004] decomposed the atmospheric response to sea
ice into the part that projects on the leading mode of model
variability and the residual from this projection. The lead-
ing mode has an equivalent barotropic vertical structure and
resembles the NAO, while the residual is baroclinic. A sub-
sequent study by Deser et al. [2007] examines the transient
response to wintertime sea ice anomalies in the North Atlan-
tic. They analyzed the general circulation model (GCm) out-