Upload
cecilia-m
View
215
Download
0
Embed Size (px)
Citation preview
Can North Atlantic Sea Ice Anomalies Account for Dansgaard–OeschgerClimate Signals?*
CAMILLE LI
Bjerknes Centre for Climate Research, and Department of Earth Science, University of Bergen, Bergen, Norway
DAVID S. BATTISTI
Department of Atmospheric Sciences, University of Washington, Seattle, Washington, and Geophysical Institute,
University of Bergen, Bergen, Norway
CECILIA M. BITZ
Department of Atmospheric Sciences, University of Washington, Seattle, Washington
(Manuscript received 2 September 2009, in final form 17 May 2010)
ABSTRACT
North Atlantic sea ice anomalies are thought to play an important role in the abrupt Dansgaard–Oeschger
(D–O) cycles of the last glacial period. This model study investigates the impacts of changes in North Atlantic
sea ice extent in glacial climates to help provide geographical constraints on their involvement in D–O cycles.
Based on a coupled climate model simulation of the Last Glacial Maximum (21 ka), the Nordic seas and
western North Atlantic (broadly, south of Greenland) are identified as two plausible regions for large and
persistent displacements of the sea ice edge in the glacial North Atlantic. Sea ice retreat scenarios targeting
these regions are designed to represent ice cover changes associated with the cold-to-warm (stadial-to-
interstadial) transitions of D–O cycles. The atmospheric responses to sea ice retreat in the Nordic seas and in
the western North Atlantic are tested individually and together using an atmospheric general circulation
model. The Nordic seas ice retreat causes 108C of winter warming and a 50% increase in snow accumulation
at Greenland Summit; concomitant ice retreat in the western North Atlantic has little additional effect. The
results suggest that displacements of the winter sea ice edge in the Nordic seas are important for creating the
observed climate signals associated with D–O cycles in the Greenland ice cores.
1. Introduction
Sea ice is an important element in the glacial climate
system because of its pivotal role in the surface heat,
moisture, and momentum budgets of the polar regions.
The presence of sea ice lowers surface temperature by
insulating the atmosphere from the ocean heat reservoir
and by increasing surface albedo. Sea ice also affects the
vertical structure of the upper ocean, alters deep-water
formation sites, and regulates the availability of moisture
for building continental ice sheets (Hebbeln et al. 1994;
Rind et al. 1995).
Sea ice is thought to be a principal player in Dansgaard–
Oeschger (D–O) cycles, the millennial-scale climate cy-
cles that are prominent features of North Atlantic (NAtl)
proxy records throughout much of the last ice age
(Dansgaard et al. 1993; Bond et al. 1993; Taylor et al.
1993; Grootes and Stuiver 1997; Sachs and Lehman
1999; Andersen et al. 2004). The interstadial phase of the
D–O cycle is a warm period in the mid- to high-latitude
Atlantic region, whereas the stadial phase is a cold pe-
riod. A remarkable feature of the D–O cycle is the abrupt
transition between stadial (cold) and insterstadial (warm)
conditions, which manifest in the Greenland ice cores
by an annual mean warming of up to 168C in a matter of
decades (Lang et al. 1999; Landais et al. 2004a,b; Huber
et al. 2006; Sanchez Goni et al. 2008) and a 50%–100%
* Bjerknes Centre for Climate Research Publication Number
A292.
Corresponding author address: Camille Li, Bjerknes Centre for
Climate Research, Allegaten 55, 5007 Bergen, Norway.
E-mail: [email protected]
15 OCTOBER 2010 L I E T A L . 5457
DOI: 10.1175/2010JCLI3409.1
� 2010 American Meteorological Society
increase in snowfall (Cuffey and Clow 1997; Andersen
et al. 2006).
Displacements of the sea ice edge have been proposed
as a key ingredient of D–O cycles (Broecker 2000; Gildor
and Tziperman 2003) because of their strong influence
on regional temperature and their potential to occur rap-
idly in response to relatively weak forcing via processes
such as the ice–albedo feedback (Curry et al. 1995). Re-
constructing past sea ice variability is difficult, but some
observational evidence exists to support the idea that
changing sea ice extent plays an important role in D–O-
related abrupt climate changes. For example, deuterium
excess measurements in the Greenland ice cores exhibit
rapid transitions associated with D–O cycles, and these
transitions have been interpreted as north/south shifts of
the moisture source for Greenland snow because of the
retreat or advance of North Atlantic sea ice (Masson-
Delmotte et al. 2005; Jouzel et al. 2007). Furthermore, a
benthic d18O record from a high-resolution marine core
on the Norwegian slope indicates more sea ice forma-
tion during cold stadial phases of D–O cycles than dur-
ing warm interstadials (Dokken et al. 2010, manuscript
submitted to Nat. Geosci., hereafter DOK).
The exact nature of sea ice’s involvement in D–O
cycles is the subject of much current research. One im-
portant question is what mechanisms drive these dis-
placements of the sea ice edge. Wind- and ocean-driven
Arctic sea ice variability in today’s climate exists on times
scales from weeks to decades (e.g., Fang and Wallace
1994; Venegas and Mysak 2000; Deser et al. 2002), but
the millennial pacing of D–O cycles requires long-lived
changes in sea ice cover. One hypothesis attributes the
cycles to an internal oscillation of the ocean thermoha-
line circulation (OTC) that causes abrupt shifts in ocean
heat transport into the North Atlantic, thereby altering
sea ice and regional climate (Broecker et al. 1985). A
recent multiproxy analysis of a high-resolution marine
core on the Norwegian slope suggests an alternative hy-
pothesis: a series of interactions between the ocean and
the marine and terrestrial cryosphere alternately build
and erode a fresh surface halocline in the Nordic seas
(NS) over millennial times scales, promoting the pres-
ence or absence of sea ice cover and a toggling of climate
between cold stadials and warm interstadials (DOK).
During stadials, the ocean continues to supply heat from
more southerly latitudes; however, this heat is stored
beneath the surface halocline, where it remains inacces-
sible to the atmosphere. The subsurface ocean expe-
riences a warming, also seen in other North Atlantic
marine cores and in climate models under certain fresh-
water hosing scenarios (Knutti et al. 2004; Rasmussen
and Thomsen 2004; Ruhlemann et al. 2004; Cheng et al.
2007; Mignot et al. 2007). The halocline is maintained
throughout the stadial owing in large part to high rates
of sea ice formation, a process that rejects brine in dense,
sinking plumes and encourages stratification (DOK).
Eventually, the subsurface becomes warm enough to
destabilize the water column. The surface halocline is
eroded, and the release of stored heat in combination
with other hypothesized changes in heat flux conver-
gence in the high latitudes produce warm, ice-free ocean
conditions through the duration of the interstadial.
We aim to assess what type of sea ice anomalies can
create the signatures of cold stadials and warm inter-
stadials recorded in the paleoclimate archives. Previous
modeling experiments using atmospheric general circu-
lation models (GCMs) suggest that winter ice edge dis-
placements produce temperature and accumulation
changes consistent with observations from the Green-
land ice cores, whereas summer ice edge displacements
produce too much of an accumulation effect and not
enough of a temperature effect (Li et al. 2005). The Li
et al. (2005) study supports other observational and
modeling studies that identify abrupt North Atlantic
climate events as primarily a wintertime phenomenon
(Renssen and Isarin 2001; Denton et al. 2005), but sev-
eral caveats concerning the modeling results must be
mentioned. First, the ocean boundary conditions used in
the study were based on a glacial sea surface tem-
perature (SST) and sea ice reconstruction [the revised
Climate: Long-Range Investigation, Mapping and Pre-
diction (CLIMAP) reconstruction; Crowley 2000] that is
now known to contain substantial errors, particularly in
regions such as the tropical Pacific and the North Atlantic
(see synthesis in Li and Battisti 2008). Second, the pre-
scribed sea ice edge displacement patterns were some-
what arbitrary and highly simplified.
Li et al. (2005) established seasonality constraints on
D–O-related displacements of the sea ice edge, pointing
to large winter changes and small summer changes as
a necessary condition. The goal of the present work is to
develop some intuition for the geographical nature of
these sea ice anomalies. Because there are still very few
reconstructions of past sea ice conditions, we adopt a
modeling approach using a combination of coupled and
atmosphere-only simulations to extend the work of Li
et al. (2005). Coupled model simulations are examined
to establish robust characteristics of the North Atlantic
sea ice cover in glacial times. Sea ice retreat scenarios
are created based on these glacial sea ice distributions,
making the results of this study more applicable to D–O
events than those of Li et al. (2005), in which the retreat
scenarios were defined in a somewhat ad hoc fashion.
Although geographically more realistic, the scenarios re-
main idealized in the sense that, where sea ice retreats,
the sea surface temperatures are set to the freezing point,
5458 J O U R N A L O F C L I M A T E VOLUME 23
and sea surface conditions elsewhere (outside of the North
Atlantic) are not altered. We focus on two issues in par-
ticular: 1) What are the characteristics of the North Atlantic
sea ice cover in glacial climates, and in which regions is it
susceptible to exhibiting large anomalies? 2) Is it possible to
place geographical constraints on D–O-related displace-
ments of the ice edge in these regions? In answering the
second question, we wish to evaluate where sea ice
anomalies must occur to have a substantial climate impact
consistent with the D–O signals recorded in Greenland
rather than to establish a precise D–O retreat scenario
that produces an exact match to the ice core records.
Section 2 contains a description of the simulations and
model datasets. In section 3, we analyze coupled model
simulations of the Last Glacial Maximum (LGM) to
identify the Nordic seas and western North Atlantic
(WAtl) as regions of interest for North Atlantic win-
ter sea ice variability. Section 4 examines surface cli-
mate and circulation changes associated with sea ice
perturbations in these two regions using a series of
atmosphere-only experiments. Section 5 presents a dis-
cussion of pending issues related to the response of the
atmosphere to North Atlantic sea ice perturbations and
implications for D–O cycles. Finally, concluding re-
marks are presented in section 6.
2. Data and methods
We use sea ice data from coupled climate model sim-
ulations of LGM (21 ka) climate from the second phase
of the Paleoclimate Modelling Intercomparison Project
phase 2 (PMIP2). The models included in the analysis
are listed in Table 1. For the LGM, the models are forced
in compliance with PMIP2 protocol: 21-ka insolation,
atmospheric greenhouse gas concentrations based on ice
core measurements (Fluckiger et al. 1999; Dallenbach
et al. 2000; Monnin et al. 2001), atmospheric aerosols
at preindustrial (PI) values, and land ice and coastlines
from the ICE-5G reconstruction (Peltier 2004; see data-
set online at http://www.atmosp.physics.utoronto.ca/
;peltier/data.php) corresponding to 120-m sea level
depression at LGM. The PMIP2 control simulations for
PI climate are also used for comparison. General features
of the simulated climates are presented in Braconnot
et al. (2007). The LGM simulations exhibit good agree-
ment, with key changes indicated by proxy data such as
a weakening of the monsoons, cooling of the tropics, cooling
of the North Atlantic region and Eurasian continent, and
a southeastward extension of the North Atlantic storm
track (Kageyama et al. 2006; Braconnot et al. 2007; Laıne
et al. 2009; Otto-Bliesner et al. 2009), although cooling in
polar regions is underestimated (Masson-Delmotte et al.
2006). Complete documentation of the models and ex-
perimental setup is available at the PMIP2 Web site
(available online at http://pmip2.lsce.ipsl.fr).
One hundred years of monthly mean sea ice data
from one of the models, the Community Climate Sys-
tem Model, version 3 (CCSM3), developed at the Na-
tional Center for Atmospheric Research (NCAR), are
used to better understand the distribution and natural
variability of North Atlantic sea ice at the LGM. The
model and its performance in a present-day setup is
documented in Collins et al. (2006a), and the LGM sim-
ulation is described in Otto-Bliesner et al. (2006). Briefly,
the relevant configuration of CCSM3 comprises the
primitive equation Community Atmosphere Model, ver-
sion 3 (CAM3), at T42 horizontal resolution with 26 hy-
brid coordinate vertical levels (Collins et al. 2006b); a land
model with land cover and plant functional types, prog-
nostic soil and snow temperature, and a river-routing
scheme (Dickinson et al. 2006); the NCAR implementa-
tion of the Parallel Ocean Program (POP) on a 320 3 384
displaced pole grid (nominal horizontal resolution of 18)
with 40 vertical levels (Smith and Gent 2002); and a
dynamic–thermodynamic sea ice model on the same grid
as the ocean model (Briegleb et al. 2004). The model has
TABLE 1. PMIP2 coupled models. Please note the following for sea ice: elastic–viscous–plastic dynamics (EVP); free drift dynamics
(FD; neglects internal ice force); ice thickness distribution (ITD); viscous–plastic dynamics (VP); the number of layers refers to tem-
perature layers in ice/snow; and heat reservoir for solar radiation (Semtner 1976) or brine pockets (Bitz and Lipscomb 1999). The
sponsoring institutions of the models are the following: CCSM3, National Center for Atmospheric Research, United States; CNRM-
CM33, Meteo-France/Centre National de Recherches Meteorologiques, France; ECBILTCLIO, Royal Netherlands Meteorological
Institute (KNMI), Netherlands; HadCM3M2, UK Met Office, United Kingdom; IPSL CM4, Institut Pierre Simon Laplace, France; and
MIROC3.2, Center for Climate System Research (University of Tokyo), National Institute for Environmental Studies, and Frontier
Research Center for Global Change (JAMSTEC), Japan.
Model Atmosphere Ocean Sea ice
CCSM3 T42 (;2.88 3 2.88) 3 L18 0.38–18 3 18 3 L40 ITD, EVP, 4 layers, brine pockets
CNRM-CM33 T63 (;1.98 3 1.98) 3 L45 0.58–28 3 28 3 L31 ITD, EVP, 4 layers
ECBILTCLIO T21 (;5.68 3 5.68) 3 L3 38 3 38 3 L20 VP, 3 layers
HadCM3M2 2.58 3 3.758 3 L19 1.258 3 1.258 3 L20 FD, leads, 0 layers
IPSL CM4 2.58 3 3.758 3 L19 28 3 28 3 L31 VP, 3 layers, heat reservoir
MIROC3.2 T42 (;2.88 3 2.88) 3 L20 0.58–1.48 3 1.48 3 L43 EVP, 0 layers
15 OCTOBER 2010 L I E T A L . 5459
also been used in a range of paleoclimate studies, in
particular to examine the climate response to freshwater
hosing in glacial and deglacial climates compared to the
present climate (Cheng et al. 2007; Bitz et al. 2007; Liu
et al. 2009; Otto-Bliesner and Brady 2010).
The climatological sea ice cover in the CCSM3 LGM
simulation compares well with reconstructions from this
time period (Li and Battisti 2008), and so it is used as
the starting point in designing sea ice scenarios repre-
senting the stadial and interstadial phases of the D–O
cycle. CAM3, the atmospheric component of CCSM3, is
forced with these scenarios to investigate the climate
response to sea ice perturbations in key regions of the
North Atlantic. The CAM3 experiments are initialized
in January with output from the CCSM3 LGM simula-
tion, and results are based on 30 yr of monthly mean
output after discarding 2 yr of spinup. The National Snow
and Ice Data Center’s sea ice index (detrended) gener-
ated from passive microwave satellite data (Fetterer et al.
2009) is used to calculate Northern Hemisphere (NH)
sea ice extent and variability in the 1978–2008 observing
period for comparison with model simulations.
3. Coupled model simulations of NorthernHemisphere LGM sea ice
The objective of this section is to identify the regions
where we expect large excursions of the winter sea ice
edge in glacial climates. The analysis is based mainly on
CCSM3 coupled model output because there are very
few long, proxy-based reconstructions of glacial sea ice
extent. We first provide an overview of Northern Hemi-
sphere sea ice in all available LGM simulations from
PMIP2, and then we focus on a more detailed analysis
of the CCSM3 simulation.
a. PMIP2 model intercomparison
The PMIP2 model participants listed in Table 1 sim-
ulate a range of sea ice thickness and concentration dis-
tributions in the LGM climate (Fig. 1). In the Northern
Hemisphere, the higher-resolution climate models pro-
duce annual mean ice areas from over 6.4 3 1012 m2 to
10.4 3 1012 m2, with mean Arctic ice thicknesses from
5.2 to 15.7 m (Table 2). The two models with the greatest
mean ice thickness—the Centre National de Recherches
Meteorologiques Coupled Global Climate Model, ver-
sion 3.3 (CNRM-CM33) and ECBILTCLIO, the latter
of which is the only low-resolution intermediate com-
plexity model of the group—produce on the order of
double the volume of Northern Hemisphere sea ice
compared to the other models and exhibit relatively
weak seasonal cycles in ice area and volume (Fig. 1g and
Table 2).
Because the reconstruction of past sea ice extent re-
mains a difficult and unresolved challenge (Kucera et al.
2005), it is unclear how best to evaluate the reliability of
these simulation results. All the higher-resolution models
perform quite well in simulating the mean Arctic sea ice
cover over the 1981–2000 satellite observing period
(Arzel et al. 2006); compared to the National Snow and
Ice Data Center dataset (Comiso 2002), the errors in
March and September sea ice extent are ,10%, with
the exception of certain cases such as CCSM3 in March
(116%) and the Met Office Hadley Centre Coupled
Model, version 3, with the MOSES2 land surface scheme
(HadCM3M2) in September (218%). Northern Hemi-
sphere sea ice area in the PMIP2 control (PI) simulations
ranges from 7.1 to 14.7 3 1012 m2 (Table 2) and also
compares reasonably well with estimates from the 1978–
2008 observing period (approximately 12 3 1012 m2, not
detrended). However, good validation with observations
does not necessarily indicate that these models have
a better representation of sea ice processes, or that they
will produce a more realistic evolution of sea ice in past or
future scenarios.
There are several observational constraints that
help to discriminate among the LGM simulations of
these models. CNRM-CM33 and ECBILTCLIO ap-
pear to be the least reasonable models: they have ex-
tremely thick ice in the Arctic (15.7 and 9.5 m annual
mean) and perennial sea ice in the Nordic seas. The
latter is not in agreement with marine proxy data,
which suggest the presence of at least a seasonally open
ocean in the Nordic seas at the LGM (Meland et al. 2005;
de Vernal et al. 2005; Kucera et al. 2005). Consistent with
the proxy data, the summer Nordic seas are entirely ice
free in two of the models, CCSM3 and HadCM3M2,
whereas L’Institut Pierre-Simon Laplace Coupled Model,
version 4 (IPSL CM4) and the Model for Interdisci-
plinary Research on Climate 3.2 (MIROC3.2) show
a strip of open ocean confined to the Norwegian coast.
All the models simulate sea ice concentrations close
to 100% year-round in the Arctic Basin, but they
show considerable differences in the low-concentration
(,50%) regions of the western North Atlantic and
western North Pacific. Consistent with proxy recon-
structions, CCSM3 and HadCM3M2 exhibit the most
winter sea ice in these regions, particularly in the
Atlantic off the coast of North America, where one
would expect favorable conditions for sea ice formation
because of freshwater input and cold-air advection off
the margin of the Laurentide ice sheet. [Note, however,
that a model bias could also be contributing to the
western North Atlantic ice cover in CCSM3, which ex-
hibits overly extensive sea ice in the Labrador Sea for
the present climate (Jochum et al. 2008).] Hence, we
5460 J O U R N A L O F C L I M A T E VOLUME 23
accept the CCSM3 and HadCM3M2 as plausible rep-
resentations of the sea ice extent at LGM, and we fur-
ther explore the former model to help guide the design
of sea ice retreat scenarios that might be expected dur-
ing glacial times.
b. Maintenance and variability of North Atlantic seaice in CCSM3
We turn our attention to North Atlantic sea ice as
simulated by CCSM3 to gain insight into its glacial dis-
tribution, and to identify regions where it can be sensitive
to forcing and hence likely to undergo large excursions.
The processes controlling the hypothesized D–O sea
ice anomalies are obviously not captured in this type of
equilibrium coupled simulation, but we use the simu-
lated climatology to guide the design of the sea ice per-
turbation experiments in the next section. These analyses
also serve to illustrate that the North Atlantic sea ice
cover can exhibit substantially altered behavior in a dif-
ferent climate.
Annual mean sea ice thickness and concentration dis-
tributions are shown in Fig. 1a. Of particular interest is
FIG. 1. Sea ice thickness and extent at LGM simulated by PMIP2 models. (a)–(f) Color shading shows annual mean sea ice thickness (m).
The thick black lines show March ice extent, and the thin black lines show September ice extent (both 50% ice concentration). Model
descriptions appear in Table 1. (g) Seasonal cycle of North Atlantic sea ice area (1012 m2) for each model. (left to right) For each
month, the bars represent models (a)–(f) and are color coded to match the model names in the map panels.
15 OCTOBER 2010 L I E T A L . 5461
the seasonal sea ice cover in the Nordic seas and western
North Atlantic. In winter, both regions exhibit extensive
sea ice; in summer, ice retreats back toward the east
coasts of Greenland and North America, leaving the
Nordic seas and western North Atlantic ice free.
Figure 2 shows the movement, growth, and melt of
North Atlantic sea ice as January–March (JFM) seasonal
averages. These months were chosen to represent the
heart of the Northern Hemisphere winter growth sea-
son: sea ice growth peaks in February, whereas the net
gain in sea ice volume peaks in December/January. The
growth term includes congelation (basal) and frazil (sur-
face ocean layer) ice growth and snow ice formation by
flooding of snow-covered ice. The melt term includes
basal and lateral melt, determined largely by the ocean–
sea ice heat flux, and top melt (negligible during winter) is
determined largely by atmosphere–sea ice heat flux.
As in today’s climate, growth occurs in the Arctic
Basin and along high-latitude coastlines, where persis-
tent offshore winds drive sea ice away from the land to
produce polynya-like conditions. High growth rates in
the Nordic seas are consistent with the presence of win-
ter sea ice cover there, and thermodynamic processes
account for over 90% of the ice volume tendency over
the JFM season. Dynamics plays a more important role
for the western North Atlantic ice tongue extending off
the coast of North America. Growth rates are relatively
small in the interior of the ice tongue (Labrador Sea and
central North Atlantic), and melt dominates at the
southern and eastern edges. Considerable ice growth is
seen in Baffin Bay and Davis Strait, and ice velocities
indicate advection out from Davis Strait and south with
the East Greenland Current to feed the ice tongue. In
fact, the ice volume tendency due to thermodynamic
processes is on average negative over the JFM season,
and it is the large positive dynamical tendency that main-
tains the sea ice cover in the western North Atlantic.
Winter melt takes place along the ice edge and in
several locations in the interior of the Labrador Sea and
the Nordic seas (Fig. 2, bottom panel). This melt occurs
primarily at the base of the sea ice, although there is a
substantial contribution from lateral melt along the south-
ern portion of the East Greenland Current and at the
southeastern edge of the ice tongue (not shown). Basal
and lateral melt are controlled by ocean–sea ice heat flux,
hence winter melt is attributed mainly to ocean heat
transport at the LGM, much as was found for the present
climate in an earlier version (CCSM2) of the NCAR cou-
pled model (Bitz et al. 2005). However, as noted by Bitz
et al. (2005), large offshore ice growth and ice advection are
important influences on the position of the ice edge in
certain regions (such as the glacial western North Atlantic).
In addition to the sea ice cover being more extensive,
other seasonal aspects are altered in the LGM simulation
compared to the present day. For example, interannual
variability in sea ice extent is mostly a wintertime phe-
nomenon in the LGM simulation (Fig. 3, bottom),
whereas both ice area and sea ice volume exhibit sub-
stantial summer variability in observations (Deser et al.
2000) and in the simulations of present-day climate (Fig. 3,
top). The interannual sea ice variability shown here arises
from internal atmosphere–ocean processes in the coupled
model, and it is unlikely that these are the same processes
controlling the long-lived ice edge displacements in the
proposed D–O mechanism. It is nevertheless informa-
tive that shifts in the seasonality of variability are possi-
ble in different climates; and despite the disparate time
scales, the dominance of winter variability is consistent with
TABLE 2. PMIP2 coupled model control (preindustrial) and LGM sea ice conditions for the NH, NAtl, and Arctic (defined as poleward of
808N). The annual mean and seasonal range are shown for each variable. Model descriptions appear in Table 1.
NH area (1012 m2) NH volume (1013 m3) NAtl area (1012 m2) Arctic thickness (m)
Model Mean Range Mean Range Mean Range Mean Range
Control
CCSM3 13.4 7.9–18.5 4.7 3.7–5.5 3.7 1.9–5.3 4.5 4.1–5.2
CNRM-CM33 10.7 8.7–12.8 2.2 1.5–2.8 2.1 1.0–3.0 2.1 1.7–2.6
ECBILTCLIO 14.7 10.4–18.0 4.4 3.8–4.9 4.5 2.8–5.6 3.5 3.0–4.2
HadCM3M2 10.2 3.6–16.4 2.1 0.8–3.5 2.4 0.5–4.4 2.2 1.5–2.7
IPSL CM4 10.2 5.8–13.7 3.0 2.2–3.7 3.0 1.2–4.7 3.3 2.5–4.5
MIROC3.2 7.1 4.3–10.0 1.3 0.7–2.0 1.9 0.7–3.1 2.1 1.5–2.8
LGM
CCSM3 10.4 6.7–15.1 4.3 3.9–4.8 4.4 2.4–6.7 5.5 5.3–6.0
CNRM-CM33 10.0 8.2–13.2 12.3 11.9–12.6 3.8 3.6–3.9 15.7 15.6–15.9
ECBILTCLIO 13.3 11.2–15.6 8.9 8.6–9.0 6.8 5.9–7.3 9.5 9.5–9.5
HadCM3M2 9.4 6.4–14.2 4.2 3.4–5.2 3.7 2.2–5.8 5.6 5.2–6.0
IPSL CM4 8.1 5.6–10.3 4.8 4.4–5.1 2.9 1.4–4.1 6.9 6.0–8.7
MIROC3.2 6.4 4.9–8.0 2.6 2.3–3.0 2.1 1.5–2.6 5.2 4.8–5.5
5462 J O U R N A L O F C L I M A T E VOLUME 23
paleoclimate evidence and modeling results (Renssen
and Isarin 2001; Denton et al. 2005; Li et al. 2005), in-
dicating that winter season ice anomalies were instru-
mental in producing abrupt climate change in the North
Atlantic region during the last glaciation and the early
Holocene.
Geometry is a chief cause for this shift in seasonality.
The maps in Fig. 4 show that variations in sea ice con-
centration occur in the marginal ice zones, the part of the
ice cover that is affected by ocean processes such as swells
and waves. Variability is small close to the pole, where
relatively thick ice is more or less constantly present. If the
sea ice edge is long (e.g., located far from the coast with
excursions into open ocean), the marginal ice zone is ex-
tensive and there is more opportunity for differences in ice
extent from one year to the next. The LGM simulation has
very little marginal ice zone in summer compared to
winter and consequently very little summer variability in
sea ice area. In today’s climate, summer sea ice is re-
stricted mostly to the Arctic Ocean, but there is a consid-
erable amount of marginal ice because much of the ice
cover at the periphery of the Arctic Basin is relatively thin.
Like seasonality, the spatial signature of sea ice vari-
ability can also change in different climates. Today, sea
ice variability has a dipole pattern with opposing centers
of action in the Davis Strait–Labrador Sea region and
the Nordic seas from weekly (Fang and Wallace 1994) to
seasonal (Walsh and Johnson 1979) time scales. The
temporal variability of this dipole pattern is coupled to
the North Atlantic Oscillation (NAO), the leading mode
of Northern Hemisphere sea level pressure variability
(Rogers and van Loon 1979; Fang and Wallace 1994). In
the LGM simulation, sea ice variability in the Nordic
seas and western North Atlantic is not strongly related.
The simultaneous correlation of winter sea ice area in
the Nordic seas and western North Atlantic regions is
less than 0.3 (,10% shared variance)—similar corre-
lation values between the Nordic seas and western
North Atlantic are found with the latter leading by up to
2 yr. Figure 4 shows composites of JFM sea ice con-
centration for the 25 yr (i.e., 25% of the total 100 yr)
with the least (category: 2) and most (category: 1) sea
ice cover in the two regions. The bar codes indicate that,
FIG. 2. Sea ice velocity, growth rate, and melt rate in CCSM3
LGM simulation: (top) speed (cm s21) in color shading, with ice
flow ($2 cm s21) by arrows; (middle) growth rate (cm day21) in-
cluding congelation and frazil growth, and snow–ice formation;
and (bottom) melt rate (cm day21) including basal, lateral, and top
melt. In all panels, the black contour shows JFM sea ice extent
(15% ice concentration).
15 OCTOBER 2010 L I E T A L . 5463
one-third of the time, the regions experience anoma-
lously low or high sea ice cover simultaneously; the other
two-thirds of the time, the anomalously low or high ice
conditions in the regions share no particular relationship.
The maps in Fig. 3 reinforce the point, with the NS2 and
NS1 composites showing weak concentration anoma-
lies in the western North Atlantic, and the WAtl2 and
WAtl1 composites showing weak concentration
anomalies in the Nordic seas.
In summary, there are two main masses of sea ice in
the glacial North Atlantic. The Nordic seas ice distri-
bution is governed primarily by in situ growth and melt
in the CCSM3 simulation of LGM climate, whereas the
western North Atlantic ice tongue owes its existence in
large part to the import of ice formed in Baffin Bay and
Davis Strait and is advected southeastward into the open
ocean. Variability in the LGM sea ice extent is mainly a
wintertime phenomenon, and it is concentrated in these
two regions, albeit uncorrelated in time.
4. Climate impacts of North Atlantic sea iceperturbations
In this section, we construct scenarios for hypothetical
sea ice anomalies associated with D–O cycles and examine
their climate impacts in CAM3, the atmospheric compo-
nent of CCSM3. The scenarios are based on CCSM3’s
simulated LGM sea ice cover, with alterations to the two
regions that are the most likely to exhibit large displace-
ments of the ice edge—the Nordic seas and western North
FIG. 3. Seasonal cycle and variability of Northern Hemisphere sea ice area in CCSM3 (top) present day (PD) and (bottom) LGM
simulations. Plots show the seasonal cycle of simulated Northern Hemisphere ice area, normalized by and expressed as a fraction of the
annual mean (purple) and of the marginal ice zone (15%–50% ice concentration) area (1013 m2; dashed blue). Shading indicates the in-
terdecile range (10th–90th percentiles; light green) and interquartile range (25th–75th percentiles; dark green) of ice area, normalized by and
expressed as a fraction of the monthly mean. The PD panel also shows observational estimates of Northern Hemisphere ice area (thick gray)
and its interdecile range (thin gray) from the National Snow and Ice Data Center’s sea ice index 1978–2008 (detrended), both standardized as
described earlier. Maps show the variance of seasonally averaged sea ice concentration in JFM and July–September (JAS) (units of sea ice
fraction squared). Black contours mark the 15% (thin) and 50% (thick) ice concentrations that define the marginal ice zone.
5464 J O U R N A L O F C L I M A T E VOLUME 23
Atlantic—during winter months to represent possible
stadial and interstadial conditions. The use of prescribed
SST and sea ice conditions is a common modeling strategy
for testing the response of the atmosphere to changes in
the surface ocean. In this case, the changes (sea ice only)
are attributed to a mechanism for D–O cycles (DOK) in-
volving ocean processes that are not represented by the
model (i.e., halocline formation, brine rejection, and
freshwater input from land ice). Thus, our experiments
provide insight into the climate impacts of sea ice anom-
alies in two regions of the North Atlantic but not into what
creates and maintains the sea ice anomalies themselves.
We first construct a stadial (cold) glacial state with
extensive sea ice in the North Atlantic and all other
boundary conditions (insolation, greenhouse gases, ice
sheets, and land–ocean geometry) corresponding to 21 ka,
as specified in PMIP2. The stadial simulation (Fig. 5a)
has identical sea surface conditions to the CCSM3 simu-
lation in the western North Atlantic ice tongue region
year-round but higher winter sea ice concentrations in and
south of the Nordic seas. We then perturb this state by
removing sea ice in the Nordic seas and the western North
Atlantic during the winter (December–May) months such
that there is a smooth transition from the November ice
cover to the April ice cover (Fig. 5e). Sea surface con-
ditions for the scenarios are shown in Fig. 5: scenario
lowiceNS represents a retreat of sea ice in the Nordic seas,
scenario lowiceWAtl represents a retreat of sea ice in the
western North Atlantic, and scenario lowice represents
a retreat of sea ice in both regions.
Before examining the temperature response to the sea
ice scenarios, we note that the CCSM3’s standard LGM
FIG. 4. Composites of winters (JFM average) with restricted and extensive sea ice cover in WAtl and NS in CCSM3
LGM simulation. Maps show averages of the 25 yr (i.e., 25% of the total 100 yr) with the least (WAtl2 and NS2) and
most (WAtl1 and NS1) JFM sea ice area in each region, expressed as concentration anomalies (%) from the
seasonal average. Black contours show the climatological JFM sea ice extent (15% ice concentration). The bar codes
provide a schematic indication of the specific years that are included in the (top) 25th and (bottom) 75th percentiles.
The vertical axis represents time through the 100-yr simulation, with time running from bottom to top. Gray bars
mark years used in the (left side of bar code) WAtl composites and (right side) NS composites; common years appear
as long black bars.
15 OCTOBER 2010 L I E T A L . 5465
simulation exhibits approximately half the observed an-
nual mean LGM-to-present-day temperature difference
at Greenland Summit. In fact, this weak LGM Summit
cooling is a general feature of the PMIP2 models. In
a five-model intercomparison study, the model-to-data
discrepancy in Greenland cooling was found to range
from 68 to 168C, with the CCSM3 performing better than
all but HadCM3M2 (Masson-Delmotte et al. 2006).
Masson-Delmotte et al. (2006) examine some possible
explanations for the model-to-data discrepancy and
find that the lowering of the Summit because of topo-
graphic smoothing (e.g., in CCSM3, the model Summit
at LGM is approximately 250 m lower than today’s Sum-
mit elevation of 3200 m) and shifts in the seasonality
of precipitation produce small—and, in some cases,
unhelpful—adjustments to the simulated Summit
temperatures. They speculate that the discrepancy could
be linked to missing factors such as the lack of dust
changes, vegetation feedbacks, soil feedbacks, and the
associated impacts on atmospheric heat transport to
Greenland.
The model’s response to sea ice retreat will be com-
pared to the Greenland ice core records of D–O warm-
ings. A comprehensive review of ice core parameters
that vary in concert with D–O cycles can be found in
Wolff et al. (2010). The three proxy records of particular
interest here are the following: 1) abrupt annual mean
temperature changes derived from fractionated gases
trapped in the ice (Severinghaus et al. 1998), 2) annual
accumulation changes derived from ice layer thickness
corrected by ice flow models (Alley et al. 1993), and
3) an accumulation-weighted temperature signal derived
FIG. 5. Sea ice and SST boundary conditions for CAM3 sea ice experiments. These scenarios are based on the SST
and sea ice from the coupled CCSM3 LGM run but with some additional fabricated alterations to North Atlantic sea
ice extent in the winter months as described in the text. (a)–(d) Colors show annual mean SSTs, the thick black lines
show March ice extent, and the thin black lines show September ice extent (both 50% ice concentration). (e) Seasonal
cycle of North Atlantic sea ice area (1012 m2) for each scenario. (left to right) For each month, the bars represent
scenarios (a)–(d) and are color coded to match the scenario names in the map panels.
5466 J O U R N A L O F C L I M A T E VOLUME 23
from the isotopic composition of the ice itself (Jouzel
et al. 1997). This last dataset is one of the best known ice
core proxies, but also perhaps one of the most difficult to
interpret, as it is affected by a host of other factors such
as changes in moisture source, moisture transport path,
seasonality of snowfall, and the strength of the atmo-
spheric inversion (Cuffey et al. 1994; Johnsen et al. 1995;
Masson-Delmotte et al. 2005). As noted in the introduc-
tion, the goal of these scenario experiments is to de-
termine where sea ice anomalies must occur to have a
substantial climate impact that is consistent with these
ice core records; it is not to establish one definitive retreat
scenario that produces an exact match to the ice core
records.
We focus on the large-scale features of the atmo-
spheric response to sea ice retreat with some attention
given to the response at Greenland Summit, defined as
the two highest model grid boxes in Greenland (with
topographic smoothing, the mean elevation is 2957 m
above sea level compared to today’s elevation of 3200 m).
Comparisons are also made at several other locations on
Greenland (see details in Table 3 caption), where there
is a clear signature of D–O cycles in existing ice cores.
While the D–O-related isotope signal at these sites am-
plifies from south to north, differences in moisture source
are thought to be partly responsible, and thus the spatial
pattern of the associated temperature signal remains
unclear (Johnsen et al. 2001; Wolff et al. 2010).
In the sea ice retreat scenarios, there is in general an
increase in mean surface temperature and a reduction in
interannual variability over the North Atlantic (Figs. 6b–d).
The warming is quite localized to the region where sea ice
is removed, is relatively shallow (up to about 700 hPa),
and decays from south to north over Greenland (Table 3).
Temperature is most sensitive to sea ice changes in the
Nordic seas. Removing sea ice in this region alone
(scenario lowiceNS) causes over 108C of wintertime
warming and 58C of annual mean warming in the Summit
region (Table 3)—sea ice retreat in the western North
Atlantic alone (scenario lowiceWAtl) has less than half
the effect on Summit temperature and accumulation
as sea ice retreat in the Nordic seas alone (scenario
lowiceNS). If the Nordic seas are already ice free, then
the additional retreat of western Atlantic sea ice has
little effect, producing a statistically significant but small
warming, amounting to less than half a degree Celsius
even in the winter season (cf. scenarios lowiceNS and
lowice).
Sea ice retreat also produces an increase in snow ac-
cumulation in Greenland (Table 3). Net accumulation
(accumulation minus ablation) is calculated from the
snow hydrology component of the land model (Figs. 6e–h).
A proper assessment of the ice sheet mass balance
requires that the atmospheric fields be downscaled to
adequately high spatial resolutions and then used to
force an ice sheet model that includes a good repre-
sentation of ice dynamics. However, the accumulation
field can give a sense of the expected mass balance
changes when sea ice retreats. As with temperature, ac-
cumulation is more affected by retreat of sea ice in the
Nordic seas alone (50% increase in the annual mean in
the Summit region) than in the western North Atlantic
alone (11% increase in the annual mean in the Summit
region); and once the Nordic seas ice cover has been
removed (scenario lowiceNS), the further removal of
western North Atlantic sea ice (scenario lowice) produces
no additional accumulation increase (Table 3). The ac-
cumulation changes associated with the Nordic seas re-
treat are on the low end of observational estimates, which
show a 50%–100% increase at Summit and the North
Greenland Ice Core Project (NorthGRIP) site during
D–O warmings (Cuffey and Clow 1997; Andersen et al.
2006).
TABLE 3. Temperature and net accumulation on Greenland in
CAM3 sea ice experiments. Winter season temperature TDJF (8C),
annual mean temperature TANN (8C), and net annual accumulation
accANN (mm yr21) averaged over two model grid points near the
core sites of Greenland Summit (mean coordinates 71.28N, 38.58W,
2957 m), Dye 3 (mean coordinates 65.68N, 43.68W, 2333 m),
NorthGRIP (mean coordinates 75.38N, 42.28W, 2612 m), and
Camp Century (mean coordinates 76.78N, 61.98W, 815 m). The
columns marked D show the difference between each scenario and
the stadial case. The 95% confidence interval is #0.028C for tem-
perature and #0.8 mm yr21 for accumulation, making all differ-
ences significant at the 0.05 level.
TDJF TANN accANN
Scenario (8C)
D
(8C) (8C)
D
(8C) (mm yr21)
D
(%)
Greenland Summit
Stadial 252.26 237.65 71.8
LowiceWAtl 247.78 4.47 235.53 2.12 80.0 11
LowiceNS 241.87 10.38 232.61 5.04 107.7 50
Lowice 241.44 10.81 232.51 5.14 108.6 51
Dye 3
Stadial 243.65 228.39 293.9
LowiceWAtl 240.29 3.36 226.84 1.54 337.0 15
LowiceNS 229.03 14.63 221.58 6.81 430.8 47
Lowice 228.84 14.82 221.50 6.89 441.1 50
NorthGRIP
Stadial 252.60 238.17 46.5
LowiceWAtl 249.91 2.68 236.78 1.39 51.6 11
LowiceNS 243.45 9.15 233.53 4.65 59.0 27
Lowice 243.04 9.56 233.36 4.81 58.1 25
Camp Century
Stadial 252.18 234.80 80.5
LowiceWAtl 250.21 1.98 233.86 0.94 88.9 10
LowiceNS 246.76 5.43 232.08 2.72 96.7 20
Lowice 246.98 5.20 232.02 2.80 94.0 17
15 OCTOBER 2010 L I E T A L . 5467
The Greenland warming in these sea ice retreat sce-
narios is somewhat less than estimates of up to 168C
from proxy data (Lang et al. 1999; Landais et al. 2004a,b;
Huber et al. 2006; Sanchez Goni et al. 2008) and up to
78C in the previous model study of Li et al. (2005). We
note, however, that the 168C warming pertains to one of
the largest warmings recorded in a D–O cycle (event 19),
and that the range of abrupt warmings spans from 78 to
168C (see Wolff et al. 2010, and references therein). As
for the comparison to Li et al. (2005), there are several
key differences in experimental setup that contribute to
the larger warming response in that study. First, Li et al.
(2005) used an old version of the NCAR Community
Climate Model, version 3 (CCM3) that allows for only
100% ice concentration or 0% ice concentration in each
grid box, whereas the new version (CAM3) allows for
partial sea ice cover over a grid box such that sea ice
retreats from non-100% cover in most places. Second, the
old experiments included increases in SSTs at locations
where sea ice is removed, whereas the current experi-
ments keep SSTs at the freezing point. If we increase
SSTs above the freezing point where sea ice is removed,
more atmospheric warming is achieved (not shown).
Practical limits on how much SSTs may rise are ultimately
set by the ocean’s heat supply to the region. In the pro-
posed D–O mechanism, the stadial-to-interstadial change
in sea ice is part of a coupled, transient response of the
climate system associated with the disappearance of the
Nordic seas halocline and the opening of the subsurface
ocean heat reservoir to the atmosphere. Benthic fora-
minifera data show evidence of subsurface heat storage
during stadials (possible warming of up to 58–68C in sev-
eral hundred meters of the upper ocean), although there
remains some debate over the exact interpretation of
the benthics (DOK). The existence of subsurface warm-
ing when there is extensive sea ice in the North Atlantic
is also demonstrated in coupled climate model simula-
tions, which produce a maximum warming of several
FIG. 6. Surface temperature and net accumulation in CAM3 sea ice experiments. (a) 2-m temperature for the stadial (cold) simulation
and (b)–(d) temperature change in the sea ice retreat scenarios. (a)–(d) Annual mean temperature at Greenland Summit is indicated at
the bottom left. (e) Net accumulation over the ice sheets for the stadial simulation and (f)–(h) the net accumulation change with sea ice
retreat. The thick black contour shows March sea ice extent, the thin black contour shows September sea ice extent (50% ice concen-
tration), and the location of Greenland Summit (71.168N, 38.458W, 2957 m) is marked by a dot.
5468 J O U R N A L O F C L I M A T E VOLUME 23
degrees to 88C at depths of up to 1500 m in the Nordic
seas, with an additional weaker warming extending be-
low the surface cooling into the tropical and south trop-
ical Atlantic (Knutti et al. 2004; Cheng et al. 2007; Mignot
et al. 2007).
The weaker response upon removal of sea ice in the
western North Atlantic compared to the Nordic seas can
be viewed in terms of the circulation changes and local
diabatic changes produced in each case. Sea ice retreat
in the Nordic seas (scenario lowiceNS) causes a decrease
in sea level pressure along the northern flank of the
Icelandic low, implying a strengthening of the easterly
flow over the Nordic seas (Fig. 7). This is consistent with
Byrkjedal et al. (2006), who tested slightly different
glacial ocean boundary conditions and find an increase
in warm easterly flows at Summit when the Nordic seas
are ice free. These sea level pressure changes are ac-
companied by a local increase in tropospheric eddy ki-
netic energy consistent with more synoptic activity (not
shown). When sea ice retreats in the western North
Atlantic (scenario lowiceWAtl) only the western flank
of the Icelandic low deepens, producing low-level cir-
culation changes that do not greatly impact the Summit.
Modeling studies investigating the atmospheric re-
sponse to sea ice variability in today’s climate and to
future Arctic sea ice decline have found a large-scale at-
mospheric response that projects strongly onto the NAO,
the leading pattern of Northern Hemisphere atmospheric
variability (Alexander et al. 2004; Magnusdottir et al.
2004; Seierstad and Bader 2009; Honda et al. 2009). At
the LGM, the leading pattern of atmospheric vari-
ability exhibits an NAO-like dipole structure and ap-
pears to be governed by the same physics as its modern
counterpart, but its spatial signature is distorted and
shifted (Pausata et al. 2009). The sea level pressure re-
sponse to sea ice retreat in the Nordic seas (scenario
lowiceNS) projects onto this altered leading pattern of
variability [cf. Fig. 2 of Pausata et al. (2009) with Fig. 7b].
Sea ice retreat in the western North Atlantic (scenario
lowiceWAtl) has a weaker sea level pressure response
signature (Fig. 7a) that bears no similarity to the leading
pattern of variability. Despite differences in the mean
climate states, these results are consistent with the study
of Magnusdottir et al. (2004), who find that atmospheric
circulation is more sensitive to sea ice anomalies in the
Barents Sea than those in the Labrador Sea in today’s
climate. Aloft, both glacial retreat scenarios produce
wavelike patterns in geopotential height, reminiscent
of Fig. 11 in Byrkjedal et al. (2006), suggesting a linear
response to midlatitude thermal anomalies on a sphere
(Hoskins and Karoly 1981).
In terms of local diabatic changes, the effect of sea ice
retreat at Greenland Summit is related to (i) whether
the atmosphere feels the influence of a cold ice-covered
surface, warm ice-free ocean, or something in between
(partially ice free) at its bottom boundary, and (ii) how
the atmosphere distributes the (changed) heat flux when
the bottom boundary changes. First, much of the west-
ern North Atlantic ice cover exists in low concentrations
(indicated by the area between the 15% and 50% ice
concentration contours in Fig. 8), allowing some heat ex-
change between the ocean and atmosphere even when sea
ice is present. Hence, the change in ocean–atmosphere
heat flux from the stadial scenario to the retreat scenarios
is moderate in the case of western North Atlantic retreat
compared to Nordic seas retreat. Second, temperature
advection is largest where the mean near-surface winds
have a large component perpendicular to the sea ice
edge (i.e., perpendicular to the strongest temperature
gradients). These locations are primarily associated with
the Nordic seas rather than the western North Atlantic
(Fig. 9, top). When sea ice retreats, the convergence of
heat due to changes in temperature advection increases
over southeastern Greenland (Fig. 9, bottom), and this
occurs whether the western North Atlantic ice tongue
remains or retreats (Figs. 9d,f).
FIG. 7. Sea level pressure in CAM3 sea ice experiments. DJF sea
level pressure (5-hPa contours, black #1030 hPa, gray $1035 hPa)
for sea ice retreat scenarios (a) lowiceWAtl and (b) lowiceNS.
Color shading shows the difference compared to sea level pressure
for stadial sea ice conditions.
15 OCTOBER 2010 L I E T A L . 5469
To summarize, the scenario experiments presented
here indicate that temperature in the vicinity of the
Greenland Summit region is more sensitive to sea ice re-
treat in the Nordic seas than in the western North Atlantic.
5. Discussion
It is well known that variability in the large-scale at-
mospheric circulation causes sea ice variability over a
wide range of time scales in observational studies (Fang
and Wallace 1994). For example, a weak negative feed-
back of winter sea ice on the NAO has been detected in
modeling experiments (Magnusdottir et al. 2004) and
observations (Strong et al. 2009), operating on lags of up
to several weeks. The question of whether sea ice has
a similar restoring negative feedback or rather an am-
plifying positive feedback on the atmosphere in glacial
climates could be crucial to evaluating the role of sea ice
in D–O cycles.
As we have already stated, sea ice anomalies in the
D–O mechanism proposed by DOK are maintained by
interactions between different components of the glacial
climate system, most importantly the ocean and the ma-
rine and terrestrial cryosphere. There are a number of
ways in which the atmosphere might reinforce the North
Atlantic climate response to stadial or interstadial sea
ice conditions. Modeling studies suggest that extratropical
thermal forcing produces atmosphere–ocean changes that
extend to the tropics. Specifically, high-latitude cooling—
either due to Atlantic meridional overturning circulation
(AMOC) shutdown (Vellinga and Wood 2002; Zhang
and Delworth 2005; Broccoli et al. 2006) or forced ex-
clusively in the high latitudes by the presence of sea ice
or land ice sheets (Chiang et al. 2003; Chiang and Bitz
2005)—has been shown to induce a southward shift of
the Atlantic intertropical convergence zone (ITCZ), which
in turn can strengthen the winter Hadley cell as well as the
(thermally driven) subtropical component of the jet (Lee
and Kim 2003). The result would be a strong, steady, and
relatively southward-lying jet that transports less heat into
the North Atlantic region. Conversely, upon retreat of sea
ice, there would be a northward migration of the ITCZ,
a weakening of the Hadley cell, and a shift toward a more
poleward-lying eddy-driven jet. The boost in atmospheric
heat transport and mechanical stirring would further
break up the North Atlantic ice cover and thus provide
a positive feedback to the ice retreat in the Nordic seas.
Such processes are beyond the scope of the model-
ing strategy adopted here, in which atmosphere–ocean
feedbacks are not represented. However, the uncoupled
FIG. 8. Surface heat flux in CAM3 sea ice experiments. Annual mean ocean–atmosphere heat flux (W m22) for
(a) stadial sea ice conditions, (b) sea ice retreat scenario lowice, and (c) the stadial–lowice difference. Black contours
show the 15% (thin) and 50% (thick) annual mean sea ice concentration.
5470 J O U R N A L O F C L I M A T E VOLUME 23
atmospheric GCM experiments we have performed in-
dicate that, in glacial climates, sea ice retreat produces
a more variable jet while sea ice advance produces a
steadier jet. For example, in retreat scenario lowice, there
is an increase in the interannual variability of 300-hPa
winds of about 110% in winter [December–February
(DJF)] despite the fact that the climatological mean wind
field is not affected much (Fig. 10, top panels). The pat-
tern of jet variability, with maxima straddling the jet axis,
indicates more frequent north–south excursions of the jet
over the Atlantic sector when it is ice free. In addition, the
leading mode of interannual jet variability from EOF
analysis of 300-hPa wind explains slightly more of the total
variance and has larger amplitude (Fig. 10, bottom panels).
This result is consistent with the modeling study of
Byrkjedal et al. (2006), who find that reducing sea ice
in the glacial Nordic seas leads to increased sea level
pressure variability.
6. Concluding remarks
The climate impacts of sea ice anomalies in the glacial
North Atlantic have been investigated using coupled
climate model simulations of Last Glacial Maximum
(LGM) climate as well as a series of sea ice perturba-
tion experiments performed with an atmospheric gen-
eral circulation model. The main findings of this work
are summarized as follows:
d Coupled climate models simulate a range of sea ice
distributions for the LGM climate (Fig. 1 and Table 2).
Of the six PMIP2 models examined here, all exhibit
a perennially ice-covered Arctic. Only CCSM3 and
HadCM3M2 simulate the seasonally open Nordic seas
inferred from paleoclimate proxy data (Meland et al.
2005; de Vernal et al. 2005; Kucera et al. 2005).d In the CCSM3 simulation of LGM climate, two main
masses of sea ice exist: one in the Nordic seas, where
the ice distribution is governed primarily by in situ
growth and melt, and one in the western North At-
lantic, where the ice distribution owes its existence in
large part to the import of ice formed in Baffin Bay
and Davis Strait (Fig. 2). Other seasonal and spatial
aspects of the sea ice cover are altered relative to the
present-day CCSM3 simulation, such as the appear-
ance of a pronounced wintertime maximum in inter-
annual variability (Fig. 3).d Surface climate signals recorded at Greenland Sum-
mit are influenced more strongly by Nordic seas ice
anomalies than by western North Atlantic ice anom-
alies (Fig. 6 and Table 3).
Collectively, our experiments suggest that the abrupt
retreat of Nordic seas ice is an effective way to produce
large, abrupt warmings and accumulation increases on
Greenland, consistent with those observed during tran-
sitions from the stadial to intersadial phase of D–O cy-
cles (although the simulated response is not as strong as
the largest transitions in the proxy records). Our results
are also likely to be relevant for understanding a different
class of the near-global abrupt climate changes known as
Heinrich events. Heinrich events involve massive iceberg
FIG. 9. Temperature advection by surface winds for (left) stadial sea ice conditions, (middle) sea ice retreat scenario lowice, and (right)
sea ice retreat scenario lowiceNS. (a),(c),(e) DJF advection of 2-m temperature Ts by surface winds us with 308C day21 reference length
shown by (top right) the bold horizontal line. (b),(d),(f) Convergence of DJF surface temperature advection field. Contours are drawn at
0.58C day21 m21 intervals with black denoting convergence and blue denoting divergence; the zero contour has been omitted. The thick
purple contours show March sea ice extent, and the thin purple contours show September sea ice extent (50% ice concentration) for the
scenario.
15 OCTOBER 2010 L I E T A L . 5471
discharges from the North American ice sheet, and they
produce prominent and systematic climate changes in-
cluding a cooling of the Northern Hemisphere Atlantic
region, a strengthening of winter monsoon winds, and
shifts in ocean circulation (see Hemming 2004, and ref-
erences therein). Heinrich events occur during the stadial
phases of certain selected D–O cycles, but these stadials
are not distinguishable as more extreme than others
in the Greenland ice cores. An explanation for this
discrepancy could be that the icebergs associated with
Heinrich events primarily affect the North Atlantic ice-
rafting zone, a belt of latitudes (408–558N) situated well
south of the Nordic seas. If Greenland is indeed most
sensitive to the Nordic seas, it is possible that surface
ocean variability in other regions of the North Atlantic
could impact global climate despite remaining largely
undetected in Greenland ice cores.
Acknowledgments. We wish to thank T. Dokken,
S. Hemming, and K. Nisancioglu for insightful discus-
sions; M. Michelsen and J.-Y. Peterschmitt for assistance
with model simulations and data; and N. G. Kvamstø,
B. Tremblay, and three anonymous reviewers for help-
ful comments on the manuscript. We acknowledge the
international modeling groups and the Laboratoire des
Sciences du Climat et de l’Environnement (LSCE) for
providing and archiving the PMIP2 model data. The
PMIP2/MOTIF Data Archive is supported by CEA,
CNRS, the EU MOTIF project (EVK2-CT-2002-00153)
and the Programme National d’Etude de la Dynamique
du Climat (PNEDC). This research is part of the Nor-
wegian Research Council ARCTREC project and was
funded by the Comer Fellowship Program (CL, DSB)
and National Science Foundation Grant ATM-0502204
(DSB, CMB).
REFERENCES
Alexander, M. A., U. S. Bhatt, J. E. Walsh, M. S. Timlin, J. S. Miller,
and J. D. Scott, 2004: The atmospheric response to realistic
Arctic sea ice anomalies in an AGCM during winter. J. Cli-
mate, 17, 890–905.
Alley, R., and Coauthors, 1993: Abrupt increase in Greenland snow
accumulation at the end of the Younger Dryas event. Nature,
362, 527–529.
Andersen, K. K., and Coauthors, 2004: High-resolution record of
Northern Hemisphere climate extending into the last inter-
glacial period. Nature, 431, 147–151.
——, and Coauthors, 2006: The Greenland ice core chronology
2005, 15–42 ka. Part 1: Constructing the time scale. Quat. Sci.
Rev., 25, 3246–3257.
Arzel, O., T. Fichefet, and H. Goosse, 2006: Sea ice evolution over
the 20th and 21st centuries as simulated by current AOGCMs.
Ocean Modell., 12, 401–405.
Bitz, C. M., and W. H. Lipscomb, 1999: An energy-conserving
thermodynamic model of sea ice. J. Geophys. Res., 104, 15 669–
15 677.
FIG. 10. Atlantic jet position and variance in CAM3 sea ice experiments. Black contours show DJF 300-hPa zonal
wind (10 m s21 contours starting at 30 m s21). (top) Interannual variance of the wind field, with (upper left) area-
averaged variance over the North Atlantic sector indicated. (bottom) Leading pattern of 300-hPa wind variability
over the North Atlantic sector (m s21, per std dev of PC1), with (upper left) percent of total variance and corre-
sponding area-averaged variance explained by EOF1 indicated. The thick purple contours show March sea ice extent,
and the thin purple contours show September sea ice extent (50% ice concentration) for the scenario.
5472 J O U R N A L O F C L I M A T E VOLUME 23
——, M. M. Holland, E. C. Hunke, and R. E. Moritz, 2005:
Maintenance of the sea ice edge. J. Climate, 18, 2903–2921.
——, J. C. H. Chiang, W. Cheng, and J. J. Barsugli, 2007: Rates of
thermohaline recovery from freshwater pulses in modern, last
glacial maximum, and greenhouse warming climates. Geophys.
Res. Lett., 34, L07708, doi:10.1029/2006GL029237.
Bond, G., W. Broecker, S. Johnsen, J. McManus, L. Labeyrie,
J. Jouzel, and G. Bonani, 1993: Correlations between climate
records from North Atlantic sediments and Greenland ice.
Nature, 265, 143–147.
Braconnot, P., and Coauthors, 2007: Results of PMIP2 coupled
simulations of the mid-Holocene and Last Glacial Maximum—
Part 1: Experiments and large-scale features. Climate Past, 3,
261–277.
Briegleb, B., C. Bitz, E. Hunke, W. Lipscomb, M. Holland,
J. Schramm, and R. Moritz, 2004: Scientific description of
the sea ice component in the Community Climate System
Model, version 3. NCAR Tech. Note NCAR/TN-4631STR,
70 pp.
Broccoli, A. J., K. A. Dahl, and R. J. Stouffer, 2006: The response
of the ITCZ to Northern Hemisphere cooling. Geophys. Res.
Lett., 33, L01702, doi:10.1029/2005GL024546.
Broecker, W. S., 2000: Abrupt climate change: Causal constraints
provided by the paleoclimate record. Earth Sci. Rev., 51, 137–154.
——, D. M. Peteet, and D. Rind, 1985: Does the ocean–atmosphere
system have more than one stable mode of operation? Nature,
315, 21–26.
Byrkjedal, Ø., N. G. Kvamstø, M. Meland, and E. Jansen, 2006:
Simulated climate response in the Last Glacial Maximum to
changed sea-ice conditions in the Nordic seas. Climate Dyn.,
26, 473–487.
Cheng, W., C. M. Bitz, and J. C. H. Chiang, 2007: Adjustment of the
global climate to an abrupt slowdown of the Atlantic meridi-
onal overturning circulation. Ocean Circulation: Mechanisms
and Impacts, A. Schmittner et al., Eds., Geophys. Monogr.,
Vol. 173, Amer. Geophys. Union, 295–313.
Chiang, J. C., and C. M. Bitz, 2005: The influence of high-latitude
ice on the position of the marine intertropical convergence
zone. Climate Dyn., 25, 477–496.
——, M. Biasutti, and D. S. Battisti, 2003: Sensitivity of the Atlantic
Intertropical Convergence Zone to Last Glacial Maximum
boundary conditions. Paleoceanography, 18, 1094, doi:10.1029/
2003PA000916.
Collins, W. D., and Coauthors, 2006a: The Community Climate
System Model version 3 (CCSM3). J. Climate, 19, 2122–2143.
——, and Coauthors, 2006b: The formulation and atmospheric
simulation of the Community Atmosphere Model version 3
(CAM3). J. Climate, 19, 2144–2161.
Comiso, J. C., cited 2002: Bootstrap sea ice concentrations from
Nimbus-7 SMMR and DMSP SSM/I. National Snow and Ice
Data Center. [Available online at http://nsidc.org/data/nsidc-
0079.html.]
Crowley, T. J., 2000: CLIMAP SSTs re-revisited. Climate Dyn., 16,
241–255.
Cuffey, K., and G. Clow, 1997: Temperature, accumulation, and ice
sheet elevation in central Greenland through the last deglacial
transition. J. Geophys. Res., 102, 26 383–26 396.
——, R. Alley, P. Grootes, J. Bolzan, and S. Anandakrishnan, 1994:
Calibration of the d18O isotopic paleothermometer for cen-
tral Greenland, using borehole temperatures. J. Glaciol., 40,
341–349.
Curry, J., J. L. Schramm, and E. E. Ebert, 1995: Sea ice–albedo
climate feedback mechanism. J. Climate, 8, 240–247.
Dallenbach, A., T. Blunier, J. Fluckiger, B. Stauffer, J. Chappellaz,
and D. Raynaud, 2000: Changes in the atmospheric CH4
gradient between Greenland and Antarctica during the Last
Glacial Maximum and the transition to the Holocene. Geophys.
Res. Lett., 27, 1005–1008.
Dansgaard, W., and Coauthors, 1993: Evidence for general in-
stability of past climate from a 250-kyr ice-core record. Nature,
264, 218–220.
Denton, G. H., R. B. Alley, G. C. Comer, and W. S. Broecker, 2005:
The role of seasonality in abrupt climate change. Quat. Sci.
Rev., 24, 1159–1182.
Deser, C., J. E. Walsh, and M. S. Timlin, 2000: Arctic sea ice var-
iability in the context of recent atmospheric circulation trends.
J. Climate, 13, 617–633.
——, M. Holland, G. Reverdin, and M. Timlin, 2002: Decadal
variations in Labrador Sea ice cover and North Atlantic sea
surface temperatures. J. Geophys. Res., 107, 3035, doi:10.1029/
2000JC000683.
de Vernal, A., and Coauthors, 2005: Reconstruction of sea-
surface conditions at middle to high latitudes of the
Northern Hemisphere during the Last Glacial Maximum
(LGM) based on dinoflagellate cyst assemblages. Quat. Sci.
Rev., 24, 897–924.
Dickinson, R. E., K. W. Oleson, G. B. Bonan, F. Hoffman,
P. Thornton, M. Vertenstein, Z.-L. Yang, and X. Zeng, 2006:
The Community Land Model and its climate statistics as a
component of the Community Climate System Model. J. Cli-
mate, 19, 2302–2324.
Fang, Z., and J. M. Wallace, 1994: Arctic sea ice variability on a
timescale of weeks and its relation to atmospheric forcing.
J. Climate, 7, 1897–1913.
Fetterer, F., K. Knowles, W. Meier, and M. Savoie, cited 2009:
Sea ice index. National Snow and Ice Data Center.
[Available online at http://nsidc.org/data/seaice_index/index.
html.]
Fluckiger, J., A. Dallenbach, B. Stauffer, T. Stocker, D. Raynaud,
and J.-M. Barnola, 1999: Variations of the atmospheric N2O
concentration during abrupt climatic changes. Science, 285,
227–230.
Gildor, H., and E. Tziperman, 2003: Sea-ice switches and abrupt
climate change. Philos. Trans. Roy. Soc. London, A361, 1935–
1942.
Grootes, P., and M. Stuiver, 1997: Oxygen 18/16 variability in
Greenland snow and ice with 1023- to 105-year time resolu-
tion. J. Geophys. Res., 102, 26 455–26 470.
Hebbeln, D., T. Dokken, E. Andersen, M. Held, and A. Elverhoi,
1994: Moisture supply for northern ice-sheet growth during
the Last Glacial Maximum. Nature, 370, 357–360.
Hemming, S. R., 2004: Heinrich events: Massive late Pleistocene
detritus layers of the North Atlantic and their global climate
imprint. Rev. Geophys., 42, RG1005, doi:10.1029/2003RG000128.
Honda, M., J. Inoue, and S. Yamane, 2009: Influence of low Arctic
sea-ice minima on anomalously cold Eurasian winters. Geo-
phys. Res. Lett., 36, L08707, doi:10.1029/2008GL037079.
Hoskins, B. J., and D. J. Karoly, 1981: The steady linear response
of a spherical atmosphere to thermal and orographic forcing.
J. Atmos. Sci., 38, 1179–1196.
Huber, C., and Coauthors, 2006: Isotope calibrated Greenland
temperature record over marine isotope stage 3 and its relation
to CH4. Earth Planet. Sci. Lett., 243, 504–519.
Jochum, M., G. Danabasoglu, M. Holland, Y.-O. Kwon, and
W. Large, 2008: Ocean viscosity and climate. J. Geophys. Res.,
113, C06017, doi:10.1029/2007JC004515.
15 OCTOBER 2010 L I E T A L . 5473
Johnsen, S., D. Dahl-Jensen, W. Dansgaard, and N. Gundestrup,
1995: Greenland paleotemperatures derived from GRIP
borehole temperature and ice core isotope profiles. Tellus,
47B, 624.
——, and Coauthors, 2001: Oxygen isotope and palaeotemperature
records from six Greenland ice-core stations: Camp Century,
Dye-3, GISP2, Renland, and NorthGRIP. J. Quat. Sci., 16,
299–307.
Jouzel, J., and Coauthors, 1997: Validity of the temperature re-
construction from water isotopes in ice cores. J. Geophys. Res.,
102, 26 471–26 487.
——, and Coauthors, 2007: The GRIP deuterium-excess record.
Quat. Sci. Rev., 26, 1–17.
Kageyama, M., and Coauthors, 2006: Last Glacial Maximum
temperatures over the North Atlantic, Europe, and western
Siberia: A comparison between PMIP models, MARGO sea-
surface temperatures, and pollen-based reconstructions. Quat.
Sci. Rev., 25, 2082–2102.
Knutti, R., J. Fluckiger, T. Stocker, and A. Timmerman, 2004:
Strong hemispheric coupling of glacial climate through fresh-
water discharge and ocean circulation. Nature, 430, 851–856.
Kucera, M., and Coauthors, 2005: Reconstruction of sea-surface
temperatures from assemblages of planktonic foraminifera:
Multi-technique approach based on geographically constrained
calibration sets and its application to glacial Atlantic and Pacific
Oceans. Quat. Sci. Rev., 24, 951–998.
Laıne, A., and Coauthors, 2009: Northern hemisphere storm tracks
during the last glacial maximum in PMIP2 ocean–atmosphere
coupled models: Energetic study, seasonal cycle and pre-
cipitation. Climate Dyn., 32, 593–614.
Landais, A., and Coauthors, 2004a: A continuous record of tem-
perature evolution over a sequence of Dansgaard–Oeschger
events during marine isotopic stage 4 (76 to 62 kyr BP). Geo-
phys. Res. Lett., 31, L22211, doi:10.1029/2004GL021193.
——, and Coauthors, 2004b: Quantification of rapid temperature
change during DO event 12 and phasing with methane in-
ferred from air isotopic measurements. Earth Planet. Sci. Lett.,
225, 221–232.
Lang, C., M. Leuenberger, J. Schwander, and S. Johnsen, 1999:
168C rapid temperature variation in Central Greenland 70 000
years ago. Science, 286, 934–937.
Lee, S., and H.-K. Kim, 2003: The dynamical relationship be-
tween subtropical and eddy-driven jets. J. Atmos. Sci., 60,
1490–1503.
Li, C., and D. S. Battisti, 2008: Reduced Atlantic storminess during
Last Glacial Maximum: Evidence from a coupled climate
model. J. Climate, 21, 3561–3579.
——, ——, D. P. Schrag, and E. Tziperman, 2005: Abrupt climate
shifts in Greenland due to displacements of the sea ice edge.
Geophys. Res. Lett., 32, L19702, doi:10.1029/2005GL023492.
Liu, Z., and Coauthors, 2009: Transient simulation of last de-
glaciation with a new mechanism for Bølling–Allerød warm-
ing. Science, 325, 310–314.
Magnusdottir, G., C. Deser, and R. Saravanan, 2004: The effects of
North Atlantic SST and sea ice anomalies on the winter cir-
culation in CCM3. Part I: Main features and storm-track
characteristics of the response. J. Climate, 17, 857–876.
Masson-Delmotte, V., and Coauthors, 2005: GRIP deuterium ex-
cess reveals rapid and orbital-scale changes in Greenland
moisture origin. Science, 309, 118–121.
——, and Coauthors, 2006: Past and future polar amplification of
climate change: Climate model intercomparisons and ice-core
constraints. Climate Dyn., 27, 437–440.
Meland, M. Y., E. Jansen, and H. Elderfield, 2005: Constraints on
SST estimates for the northern North Atlantic/Nordic seas
during LGM. Quat. Sci. Rev., 24, 835–852.
Mignot, J., A. Ganapolski, and A. Levermann, 2007: Atlantic
subsurface temperature: Response to a shutdown of the over-
turning circulation and consequences for its recovery. J. Cli-
mate, 20, 4884–4898.
Monnin, E., A. Indermuhle, A. Dallenbach, J. Fluckiger, B. Stauffer,
T. Stocker, D. Raynaud, and J.-M. Barnola, 2001: Atmospheric
CO2 concentrations over the last glacial termination. Science,
291, 112–114.
Otto-Bliesner, B., and E. Brady, 2010: The sensitivity of the cli-
mate response to the magnitude and location of freshwater
forcing: Last glacial maximum experiments. Quat. Sci. Rev.,
29, 56–73.
——, ——, G. Clauzet, R. Tomas, S. Levis, and Z. Kothavala, 2006:
Last Glacial Maximum and Holocene climate in CCSM3.
J. Climate, 19, 2526–2544.
——, and Coauthors, 2009: A comparison of PMIP2 model simu-
lations and the MARGO proxy reconstruction for tropical sea
surface temperatures at last glacial maximum. Climate Dyn.,
32, 799–815.
Pausata, F. S. R., C. Li, J. J. Wettstein, K. H. Nisancioglu, and
D. S. Battisti, 2009: Changes in atmospheric variability in a
glacial climate and the impacts on proxy data: A model inter-
comparison. Climate Past, 5, 489–502.
Peltier, W. R., 2004: Global glacial isostasy and the surface of the
ice-age Earth. Annu. Rev. Earth Planet. Sci., 32, 111–149.
Rasmussen, T., and E. Thomsen, 2004: The role of the North
Atlantic Drift in the millennial timescale glacial climate fluc-
tuations. Palaeogeogr., Palaeoclimatol., Palaeoecol., 210,
101–116.
Renssen, H., and R. Isarin, 2001: The two major warming phases of
the last deglaciation at ;14.7 and ;11.5 ka cal BP in Europe:
Climate reconstructions and AGCM experiments. Global
Planet. Change, 30, 117–153.
Rind, D., R. Healy, C. Parkinson, and D. Martinson, 1995: The role
of sea ice in 2 3 CO2 climate model sensitivity. Part 1: The
total influence of sea ice thickness and extent. J. Climate, 8,
449–463.
Rogers, J., and H. van Loon, 1979: The seesaw in winter temper-
atures between Greenland and Northern Europe. Part II: Some
oceanic and atmospheric effects in middle and high latitudes.
Mon. Wea. Rev., 107, 509–519.
Ruhlemann, C., S. Mulitza, G. Lohmann, A. Paul, M. Prange, and
G. Wefer, 2004: Intermediate depth warming in the tropical At-
lantic related to weakened thermohaline circulation: Combining
paleoclimate data and modeling results for the last deglaciation.
Paleoceanography, 19, PA1025, doi:10.1029/2003PA000948.
Sachs, J. P., and S. J. Lehman, 1999: Subtropical North Atlantic
temperatures 60 000 to 30 000 years ago. Science, 286, 756–
759.
Sanchez Goni, M. F., A. Landais, W. J. Fletcher, F. Naughton,
S. Desprat, and J. Duprat, 2008: Contrasting impacts of
Dansgaard–Oeschger events over a western European lat-
itudinal transect modulated by orbital parameters. Quat. Sci.
Rev., 27, 1136–1151.
Seierstad, I. A., and J. Bader, 2009: Impact of a projected future
Arctic sea ice reduction on extratropical storminess and the
NAO. Climate Dyn., 33, 937–943.
Semtner, A. J., 1976: A model for the thermodynamic growth of sea
ice in numerical investigations of climate. J. Phys. Oceanogr.,
6, 379–389.
5474 J O U R N A L O F C L I M A T E VOLUME 23
Severinghaus, J. P., T. Sowers, E. J. Brook, R. B. Alley, and
M. L. Bender, 1998: Timing of abrupt climate change at the
end of the Younger Dryas interval from thermally fraction-
ated gases in polar ice. Nature, 391, 141–146.
Smith, R., and P. Gent, 2002: Reference manual for the Parallel
Ocean Program (POP), ocean component of the Community
Climate System Model (CCSM2.0 and 3.0). Los Alamos Na-
tional Laboratory Tech. Rep. LA-UR-02-2484, 74 pp.
Strong, C., G. Magnusdottir, and H. Stern, 2009: Observed feed-
back between winter sea ice and the North Atlantic Oscilla-
tion. J. Climate, 22, 6021–6032.
Taylor, K. C., and Coauthors, 1993: Electrical conductivity mea-
surements from the GISP2 and GRIP Greenland ice cores.
Nature, 366, 549–552.
Vellinga, M., and R. A. Wood, 2002: Global climatic impacts of
a collapse of the Atlantic thermohaline circulation. Climatic
Change, 54, 251–267.
Venegas, S. A., and L. A. Mysak, 2000: Is there a dominant timescale of
natural climate variability in the Arctic? J. Climate, 13, 3412–3434.
Walsh, J. E., and C. M. Johnson, 1979: An analysis of Arctic sea ice
fluctuations. J. Phys. Oceanogr., 9, 580–591.
Wolff, E., J. Chappellaz, T. Blunier, S. Rasmussen, and A. Svensson,
2010: Millennial-scale variability during the last glacial: The ice
core record. Quat. Sci. Rev., doi:10.1016/j.quascirev.2009.10.013,
in press.
Zhang, R., and T. Delworth, 2005: Simulated tropical response to
a substantial weakening of the Atlantic thermohaline circu-
lation. J. Climate, 18, 1853–1860.
15 OCTOBER 2010 L I E T A L . 5475