Natural Variability of Greenland Climate, Vegetation, and Ice Volume During the Past Million Years

Abstract

The response of the Greenland ice sheet to global warming is a source of concern notably because of its potential contribution to changes in the sea level. We demonstrated the natural vulnerability of the ice sheet by using pollen records from marine sediment off southwest Greenland that indicate important changes of the vegetation in Greenland over the past million years. The vegetation that developed over southern Greenland during the last interglacial period is consistent with model experiments, suggesting a reduced volume of the Greenland ice sheet. Abundant spruce pollen indicates that boreal coniferous forest developed some 400,000 years ago during the "warm" interval of marine isotope stage 11, providing a time frame for the development and decline of boreal ecosystems over a nearly ice-free Greenland.

Full text

DOI: 10.1126/science.1153929
, 1622 (2008); 320Science
et al.Anne de Vernal,
and Ice Volume During the Past Million Years
Natural Variability of Greenland Climate, Vegetation,
www.sciencemag.org (this information is current as of June 20, 2008 ):
The following resources related to this article are available online at
http://www.sciencemag.org/cgi/content/full/320/5883/1622
version of this article at:
including high-resolution figures, can be found in the onlineUpdated information and services,
http://www.sciencemag.org/cgi/content/full/320/5883/1622/DC1
can be found at: Supporting Online Material
found at:
can berelated to this articleA list of selected additional articles on the Science Web sites
http://www.sciencemag.org/cgi/content/full/320/5883/1622#related-content
http://www.sciencemag.org/cgi/content/full/320/5883/1622#otherarticles
, 9 of which can be accessed for free: cites 27 articlesThis article
http://www.sciencemag.org/about/permissions.dtl
in whole or in part can be found at: this article
permission to reproduce of this article or about obtaining reprintsInformation about obtaining
registered trademark of AAAS.
is aScience2008 by the American Association for the Advancement of Science; all rights reserved. The title
CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005.
(print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theScience
on June 20, 2008 www.sciencemag.orgDownloaded from

olivines in the light isotopes of Fe (and the heavy
isotopes of Mg) (16, 20).
The extent of equilibrium isotope fractionation
is mainly controlled by the relative mass dif-
ference between the isotopes, and more fraction-
ation happens in isotopes with a larger relative
mass difference (14, 24).IftheFeisotopicvaria-
tion in the lava lake was produced by equilibrium
isot o pe fract io na ti on , Mg isotopes should show
more significant fractionation than Fe isotopes
because of their larger relative mass difference.
Furthermore, kinetic isotope fractionation driven
by th ermal and chemical diffusion should also
result in larger fractionation in Mg isotopes as
compared with that in Fe isotopes (16, 17, 20).
The absence of Mg isotope fractionation in Kilauea
Iki lavas may result from the low-precision iso-
topic analysis of Mg relative to Fe (e.g., 0.1
versus 0.04), which prevents the detection of
Mg isotopic variation. More likely, the presence
of Fe isotope fractionation and the absence of Mg
isotope fractionation may reflect the influence of
Fe oxidation states on kinetic or equilibrium iso-
tope fractionation (as compared with those of Mg,
two oxidation states of Fe exist in terrestrial mag-
matic systems) (5, 25).
Our study suggests that, unlike Li and Mg
isotopes (2, 3), Fe isotopes fractionate during ba-
saltic differentiation at both whole-rock and crys-
tal scales. Mineral compositions should therefore
be used to help interpret whole-rock basalt Fe
isotopic data. The elevated d
56
Fe of crustal igne-
ous rocks, which is more evolved than that in
basalts, could be explained by fractional crystal-
lization (10).
References and Notes
1. F. Poitrasson, A. N. Halliday, D. C. Lee, S. Levasseur,
N. Teutsch, Earth Planet. Sci. Lett. 223, 253 (2004).
2. F.-Z. Teng, M. Wadhwa, R. T. Helz, Earth Planet. Sci. Lett.
261, 84 (2007).
3. P. B. Tomascak, F. Tera, R. T. Helz, R. J. Walker, Geochim.
Cosmochim. Acta 63, 907 (1999).
4. S. Weyer, D. A. Ionov, Earth Planet. Sci. Lett. 259, 119
(2007).
5. H. M. Williams et al., Earth Planet. Sci. Lett. 235, 435
(2005).
6. B. L. Beard et al., Chem. Geol. 195, 87 (2003).
7. J. A. Schuessler, R. Schoenberg, H. Behrens,
F. von Blanckenburg, Geochim. Cosmochim. Acta 71,
417 (2007).
8. A. Shahar, C. E. Manning, E. D. Young, Earth Planet. Sci.
Lett. 268, 330 (2008).
9. R. Schoenberg, F. von Blanckenburg, Earth Planet. Sci.
Lett. 252, 342 (2006).
10. F. Poitrasson, R. Freydier, Chem. Geol. 222, 132 (2005).
11. R. T. Helz, in Magmatic Processes: Physicochemical
Principles, B. O. Mysen, Ed. (Geochemical Society,
University Park, PA, 1987), vol. 1, pp. 241–258.
12. Materials, methods, data, and modeling details are
available as supporting material on Science Online.
13. V. B. Polyakov, R. N. Clayton, J. Horita, S. D. Mineev,
Geochim. Cosmochim. Acta 71, 3833 (2007).
14. E. A. Schauble, in Geochemistry of Non-Traditional Stable
Isotopes, C. M. Johnson, B. L. Beard, F. Albarede, Eds.
(Mineralogical Society of America, Washington, DC,
2004), vol. 55, pp. 65–111.
15. N. Dauphas, O. Rouxel, Mass Spectrom. Rev. 25, 515
(2006).
16. F. M. Richter, Geochim. Cosmochim. Acta 71, A839 (2007).
17. F. Huang, C. C. Lundstrom, A. J. Ianno, Geochim.
Cosmochim. Acta 71, A422 (2007).
18. R. T. Helz, H. Kirschenbaum, J. W. Marinenko, Geol. Soc.
Am. Bull. 101, 578 (1989).
19. A. D. Anbar, J. E. Roe, J. Barling, K. H. Nealson, Science
288, 126 (2000).
20. F. M. Richter, E. B. Watson, R. A. Mendybaev,
F.-Z. Teng, P. E. Janney, Geochim. Cosmochim. Acta 72,
206 (2008).
21. R.T.Helz,C.R.Thornber,Bull. Volcanol. 49, 651 (1987).
22. A. Jambon, Geochim. Cosmochim. Acta 44, 1373 (1980).
23. R. T. Helz, U.S. Geol. Surv. Prof. Pap. 1350, 691 (1987).
24. H. C. Urey, J. Chem. Soc. (London) 1947, 562 (1947).
25. H. M. Williams et al., Science 304, 1656 (2004).
26. D. H. Richter, J. P. Eaton, K. J. Murata, W. U. Ault,
H. L. Krivoy, U.S. Geol. Surv. Prof. Pap. 537-E, 1 (1970).
27. R. T. Helz, H. Kirschenbaum, J. W. Marinenko, R. Qian,
U.S. Geol. Surv. Open-File Rep. 94-684, 1 (1994).
28. Discussions with S. Huang, A. T. Anderson Jr., F. M. Richter,
M.Wadhwa,P.B.Tomascak,R.J.Walker,andA.Pourmand
are appreciated. We thank three anonymous reviewers for
constructive comments. This work was supported by a
Packard fellowship, the France Chicago Center, and NASA
through grant NNG06GG75G to N.D.
Supporting Online Material
www.sciencemag.org/cgi/content/full/320/5883/1620/DC1
SOM Text S1 to S5
Fig. S1
Tables S1 to S4
References
29 February 2008; accepted 12 May 2008
10.1126/science.1157166
Natural Variability of Greenland Climate,
Vegetation, and Ice Volume During
the Past Million Years
Anne de Vernal* and Claude Hillaire-Marcel
The response of the Greenland ice sheet to global warming is a source of concern notably because of its
potential contribution to changes in the sea level. We demonstrated the natural vulnerability of the ice
sheet by using pollen records from marine sediment off southwest Greenland that indicate important
changes of the vegetation in Greenland over the past million years. The vegetation that developed over
southern Greenland during the last interglacial period is consistent with model experiments, suggesting a
reduced volume of the Greenland ice sheet. Abundant spruce pollen indicates that boreal coniferous forest
developed some 400,000 years ago during the “warm” interval of marine isotope stage 11, providing a
time frame for the development and decline of boreal ecosystems over a nearly ice-free Greenland.
T
he potential for sea-level rise, caused by
melting of the Greenland ice-sheet as sur-
face air temperature increases, is consid-
erable (1). Although there is evidence that the
velocity of ice streams flowing into the ocean and
therateofthinningoftheicehaveincreasedre-
cently (2, 3), large uncertainties remain about the
long-term stability of the ice sheet. The climate
Fig. 4. Modeling of Fe
isotopic variations dur-
ing magmatic differen-
tiation in Kilauea Iki
lava lake (12). Solid
lines represent calcu-
lated Fe isotopic com-
positions of residual
melts during fractional
crystallization by assum-
ing a Rayleigh distilla-
tion process with average
crystal-melt fractionation
factors (Dd
56
Fe
crystal-melt
=
d
56
Fe
crystal
– d
56
Fe
melt
)of
–0.1, –0.2, and –0.3‰.
Dashed horizontal lines
represent calculated mix-
ing lines between the
most magnesian melt
from the 1959 eruption
(23)andthemostmag-
nesianolivines[(MgO=46.6±1wt%andd
56
Fe = 0, –0.1, and –0.2‰ (black squares)]. The blue star
represents the most magnesian melt (MgO = 10.7 wt %; assumed d
56
Fe = 0.11‰). The green bars
represent the ranges of measured d
56
Fe and estimated MgO in olivine grains from two drill core samples
(MgO = 33.6 to 39.8 wt % and 41.9 to 42.7 wt %; table S3). Sample crystallization sequences are the same
as those in Fig. 2. Error bars indicate 95% CI of the mean.
Melt + olivines
01020304050
–0.4
–0.2
0.0
0.2
0.4
MgO (wt%)
Fractional
crystallization
–0.2
–0.3
–0.1
δ
56
Fe = –0.41 to + 0.01
KI75-1-139.3
KI81-5-254.5
δ
56
Fe = –1.10 to + 0.09
δ
56
Fe (‰)
20 JUNE 2008 VOL 320 SCIENCE www.sciencemag.org1622
REPORTS
on June 20, 2008 www.sciencemag.orgDownloaded from

and ice volume of Greenland seem to have varied
considerably in the recent geological past, as shown
by paleoecological data indicating a warmer re-
gional climate and reduced ice volume during the
last interglacial period (4) and by biogenic remains
of coniferous trees from forests that grew on Green-
land during Pliocene and early-to-mid-Pleistocene
times some hundred thousands to million years ago
(5, 6). However, although the climate and ice sheet
history of Greenland during the last climatic cycle
are well documented by isotope and geochemical
records from ice cores, which reveal high sensitiv-
ity to sea-surface conditions over the northern
North Atlantic Ocean (7), very little is known
about conditions preceding the onset of the last
glaciation because of the lack of continuous direct
records. On one hand, glacial activity on Green-
land over millions of years is evidenced by ice-
rafted debris in marine cores from continental
margins (8), but the precise size of the Greenland
ice sheet and its relative stability over time remain
unknown. On the other hand, sedimentary out-
crops from the Greenland coasts and near-shore
marine sediment cores suggest the recurrence of
relatively warm climatic conditions during the past
(5, 6, 9), but the duration and timing of these
phases remain uncertain. We used the pollen con-
tent of sediment cores from the Ocean Drilling
Program (ODP) site 646 on the continental rise,
off southern Greenland (Figs. 1 and 2), as an in-
dependent proxy for assessing the dominant type
of vegetation and the timing of the last forested
episodes. The stratigraphy of the cores was estab-
lished from d
18
O in foraminifer shells (10), which
permits correlation with the stack marine isotope
stratigraphy of Lisiecki and Raymo (11)andthe
setting of a time scale (12) (fig. S1).
One difficulty in interpreting pollen assem-
blages from marine sediments is the identification
ofthevegetationsourceareabecausethepollenis
necessarily exotic and derives from more or less
long-distance transport. T wo main transport mech-
anisms have to be taken into account: atmospheric
transport by winds, and hydrodynamic transport
through runof f, rivers, and marine currents. The
long-distan ce atmospher ic tran sport of pollen re-
sults in low concentrations with distorted assem-
blages characterized by an overrepresentation of
Pinus pollen grains that show exceptional aerial
dispersion properties (13). Along continental mar -
gins, detailed studies have shown that most of the
pollen in marine sediments is due to fluvial inputs
from adjacent lands, therefore allowing direct com-
parison with terrestrial palynostratigraphy (14). In
the Labrador Sea, pollen analyses along a near-
shore to offshore transect showed that atmospheric
transport is accompanied with an asymptotic de-
crease in the concentration of pollen from the coast-
line and an increase in the relative proportion of
1
GEOTOP Geochemistry and Geodynamics Research Center–
Université du Québec à Montréal, Case Postale 8888,
succursale Centre-Ville, Montréal, Québec H3C 3P8, Canada.
*To whom correspondence should be addressed. E-mail:
devernal.anne@uqam.ca
Fig. 1. Location of ODP
site 646 (58°12.56 N,
48°22.15 W; water
depth 3460 m) in the
northern North Atlantic
and of other coring sites
referred to in the text:
HU-90-013-013 (58°12.59
N, 48°22.40 W; water
depth 3379 m); ODP
site 647 (53.19.9 N,
45°14.7 W, water depth
3862 m); and HU-84-
030-003 (53.19.8 N,
45°14.7 W, water depth
3771 m). The Dye 3 cor-
ing site, where spruce
DNA was found, is indi-
cated by a blue square
(6). The white arrows
correspond to the mean
surface vector wind from
June to September based
on the 1968-to-1996 clima-
tology [see the National
Centers for Environmen-
tal Prediction/National
Center for Atmospheric
Research reanalysis (www.
cdc.noaa.gov/cgi-bin/Composites/comp.p/)], available from the Earth System Research Laboratory, Physical
Science Division, of the National Oceanic and Atmospheric Administration (www.esrl.noaa.gov/psd/). The thin
andthickwhitearrowscorrespondtowindspeedslowerandhigherthan2ms
–1
, respectively. The blue
arrows schematically illustrate the surface ocean circulation pattern along the Greenland coast, in the
Labrador Sea, and in Baffin Bay. The dashed green line corresponds to the present-day northern limit of the
potential natural tree line or cold evergreen needle-leaf forest in Biome models (27).
Fig. 2. Stratigraphy and chronology of the upper 76 m at ODP site 646 (58°12.56N, 48°22.15 W; water
depth 3460 m) based on
18
OmeasurementsinNeogloboquadrina pachyderma (10) and correlation with the
stack curve LR04 of Lisiecki and Raymo (11). The abundance of pollen grains and spores of pteridophytes is
expressed in concentration per cm
3
of sediments. Sedimentation rates are uniform and average 7.8 cm per ka,
which permits the assumption that pollen concentrations are approximately proportional to fluxes (fig. S1)
(12). The vertical gray bands correspond to modern values of concentrations, and the horizontal green bands
correspond to phases with concentrations at least twice that of those recorded during the late Holocene.
www.sciencemag.org SCIENCE VOL 320 20 JUNE 2008
1623
REPORTS
on June 20, 2008 www.sciencemag.orgDownloaded from

Pinus (15). Pliocene to recent pollen contents at
offshore sites from the northwest North Atlantic
(ODP site 647, core sample HU84-030-003) (Fig.
1), where mostly wind inputs can be recorded, are
characterized by pollen fluxes lower than 0.5 grain
cm
−2
year
−1
and largely dominated by Pinus (16).
The analyses of Arctic snow, Greenland ice, and
pollen traps along the southern Greenland edge
show an exotic component from boreal forests of
North America, but illustrate that long-distance
atmospheric transport is responsible for low inputs
(17, 18). Therefore, the large-amplitude variations
in pollen content from the southern Greenland
margin records at ODP site 646, with fluxes
well above “modern” or Holocene (that is, the past
1 1, 500 years) values, can be interpreted as reflecting
changes in hydrodynamic inputs from a relatively
proximal source-vegeta tion located on southern
Greenland (12). Furthe r evidence for the promi-
nence of proximal sources during interglacials is
provided by the comparison of total pollen content
to long-distance transported grains of Pinus (12).
The pollen record of the last million years at
ODP site 646 shows important variations both in
concentrations [thus fluxes, because sedimentation
rates remained fairly constant in the study se-
quence (12)] and dominance of the main taxa
(Figs.2and3andfigs.S2andS3)(12). Pollen
concentrations vary by orders of magnitude, from
less than 10 grains cm
−3
to more tha n 10
3
grains
cm
−3
. In general, low concentrations are recorded
during glacial stages. Minimum values (close to
zero) characterize the marine isotope stage (MIS)
6, indicating very low fluxes from both long-
distance and proximal sources, which is consistent
with extensive development of the Laur entide and
Greenland ice sheets (9). Higher concentrations
are seen in interglacial sediments. The Holocene
is characterized by concentrations of about 100
grains cm
−3
. The assemblages include inputs from
the boreal forest of southeastern Canada linked to
predominant southwest-northeast summer winds,
but show components (12) that are from more-
proximal shrub-tundra vegetation.
Earlier interglacial stages record much higher
pollen concentrations than the Holocene. Those of
MIS 5e are five times higher, and the concen-
trations of pteridophytes spores are also higher .
The assemblages are characterized by dominant
Alnus and abundant spores of Osmunda (Fig. 3B
and fig. S2). In core sample HU-90-013-013 col-
lected near ODP site 646 (19), more detailed anal-
yses of MIS 5e document the pollen succession
(Fig. 4). A rapid increase of Alnus occurred during
an early phase of MIS 5e characterized by high
summer sea-surface temperatures, which suggests
rapid development of shrub tundra after the ice
retreat (9). The subsequent increase of Osmunda
represents a unique event in the last million years.
It coincides with maximum sea-surface temper-
atures in winter and suggests the development of
dense fern vegetation over southern Greenland un-
der climatic conditions not unlike those of the
modern boreal forest, given the present distri-
bution of the genus. Osmunda expanded possibly
in a la r g e geog r ap h i c a l do m a in, because spores
were identified at the base of ice cores drilled in
the Agassiz ice cap (20). Toward t h e end of MIS
5e, pollen and spore influxes decreased concomi-
tantly with the augmentation of herb percentages.
This event corresponds to the fi r st ste p to w a r d
higher d
18
OvaluesinGlobigerina bulloides and
Neogloboquadrina pachyderma. It suggests a
change to the herb tundra resulting from regional
cooling at the onset of ice growth.
MIS 7, the penultimate interglacial period, dif-
fered from MIS 5e in many respects (Fig. 3C).
Sea-surface temperatures never reached those of
MIS 5e, and the pollen and spore content of sed-
iment remained lower . Its pollen assemblages are
characterized by dominant herb taxa (notably
Poaceae and Cyperaceae), suggesting the devel-
opment of tundra along southern Greenland coasts.
The MIS 1 1 interglacial is different than others
because of its near 50,000-year duration [374 to
424 thousand years ago (ka) (11)]. At site 646,
MIS 11 is also unique because of pollen concen-
trations one order of magnitude higher than those
of the Holocene, the dominance of Picea spp., and
the occurrence of Abies pollen grains (Figs. 2 and
3D and fig. S2). The dominance of Picea from the
beginning to the end of the interglacial period sug-
gests the presence of forest vegetation throughout
the entire interval, at least over southern Greenland.
The base of MIS 1 1 is marked by higher pro-
Fig. 3. (A to E)Stratig-
raphy, sea-surface temper-
atures, and concentration
of dominant pollen and
spore taxa during the
present interglacial pe-
riod [MIS 1 (A)], the last
interglacial period [MIS
5e (B)], MIS 7 (C), MIS
11 (D), and MIS 13 (E).
Sea-surface temperatures
are estimated from dino-
cysts [the thin lines are
best estimates from five
modern analogs and the
thick lines correspond to
three-point running av-
erag e s (28)]. Dinocyst as-
semblages were reported
by de Vernal and Mudie
(29). The modern sea-
surface temperatures at
ODPsite646are3.9±
0.7°C and 7.3 ± 1.1°C
in winter and summer, re-
spectively [data from the
World Ocean Atlas, 2001
(30)]. Among pollen assem-
blages, herb taxa include
mostly Pocacea, Cyperace-
ae, and Asteraceae. Shrub
pollen is dominated by
Alnus and Betula with the
occasional occurrence of
Salix and Ericaceae (fig. S2).
Pinus has been excluded
because of its overrepresen-
tation due to long-distance
atmospheric transport.
20 JUNE 2008 VOL 320 SCIENCE www.sciencemag.org1624
REPORTS
on June 20, 2008 www.sciencemag.orgDownloaded from

portions of shrub and herb pollen, indicating more
open vegetation and a cooler climate, but Picea
was probably already present regionally , taking
into account the fact that its concentrations reached
hundreds of grains per cm
3
. The covariance of d
18
O
in planktic foraminifers and of Picea concentra-
tio ns suggests a synchronous ice retreat and early
forest development, a maximum of Picea concen-
tration when maximum sea-surface temperatures
occurred, and a concomitant glacial onset and
forest decline at the end of MIS 11. The devel-
opment of spruce forest over Greenland probably
indicates relatively mild conditions (6)anda
substantial reduction of the Greenland ice sheet
during the long MIS 11 interglacial period. How-
ever , precise paleoclimatic and paleoecological in-
ferences from pollen or DNA are uncertain without
knowing the species of conifer trees. The identifi-
cation of Picea pollen grains down to the species
level is difficult because of uniform morphologica l
characteristics of the genus. Nevertheless, detailed
microscopic examination suggests the occurrence
of several species, among which Picea abies
dominated (fig. S4). In northern Europe and
Fennoscandia, Picea abies is a common conifer
tree of montane and boreal environments that
often occurs at the tree limit and acted as a pi-
oneer along emerging postglacial coasts (21).
Growth of Picea abies is fostered by high July
temperatures and cool and snowy winters, but
has a low temperature threshold (2.6°C) for the
initiation of bud and stem growth. Picea abies
has adapted to survive severe climate; it can
persist for hundreds of years by vegetative prop-
agation. Therefore, its development, at least over
southern Greenland during MIS 11, does not nec-
essarily imply a zonal climate that was warmer
than at present, because the northern tree limit
and the Picea abies timberline occur now near
the polar circle in Europe. However, it certainly
indicates ice-free conditions over a large area of
Greenland, and thus a much-reduced ice-sheet
volume, otherwise katabatic winds (22)would
have restricted any forest development.
Before MIS 11, the pollen content was rarely
less abundant than it was during the Holocene,
thus suggesting vegetation that was generally as
extensive as it is at present. Pollen was particu-
larly abundant during MIS 13 (Figs. 2 and 3E),
but Picea concentrations did not reach values as
high as those during MIS 11, thus suggesting
shrub-tundra–type vegetation.
In conclusion, although the pollen record
from site 646 does not provide a direct picture of
climate changes over Greenland, it yields impor-
tant information that helps link fragmentary
terrestrial records into a continuous sequence.
Furthermore, the pollen record is as a proxy for
the ice volume of Greenland in two ways. First, it
provides information on pollen production, and
thus on the vegetation density on adjacent land,
which implies ice-free conditions. Second, it
depends on the distance to site 646 from the
source vegetation, which has been shorter during
ice-free episodes in southern Greenland because
of low relative sea levels that are a result of iso-
static adjustment. A substantially reduced Green-
land ice volume seems to have characterized MIS
5e, 1 1, and 13, as well as the Pliocene (23), indi-
cating a long-term sensitivity of the Greenland
ice sheet to warm temperatures. Among warm
climate intervals of the last million years, MIS 11
stands out in terms of forest vegetation spreading
over southern Greenland. Thus, if the melting of
Greenland and other Arctic ice caps are assumed
to have contributed to the equivalent of a 2.2- to
3.4-m-higher sea level during MIS 5e (24), we
may assume that they contributed some more
during MIS 11. The actual volume of the ice-
sheet decline during these episodes is difficult to
estimate, but it did occur under natural forcing
with an atmospheric partial pressure of CO
2
≈280
parts per million by volume (25). During MIS 5e,
particularly high summer insolation probably con-
tributed to the Greenland ice melt (4), whereas
the long duration of MIS 1 1 might explain the
retreat of the ice sheet under an insolation pattern
that is similar to that of the Holocene (26). The
data presented here provide evidence of the
vulnerability of the Greenland ice sheet to natural
forcing and should increase concerns about its
fate during the anticipated global warming.
References and Notes
1. J. A. Dowdeswell, Science 311, 963 (2006).
2. E. Rignot, P. Kanagaratnam, Science 311, 986
(2006).
3. S. B. Luthcke et al., Science 314, 1286 (2006).
4. B. L. Otto-Bliesner et al., Science 311, 1751 (2006).
5. O. Bennike et al., Palaeogeogr. Palaeoclimatol.
Palaeoecol. 186, 1 (2002).
6. E. Willerslev et al., Science 317, 111 (2007).
7. K. Andersen et al., Nature 431, 147 (2004).
8. H. C. Larsen et al., Science 264, 952 (1994).
9. S. Funder et al., Quat. Sci. Rev. 17, 77 (1998).
10. A. E. Aksu, C. Hillare-Marcel, P. Mudie, Proceedings of the
Ocean Drilling Program 105B, 689 (1989).
11. L. E. Lisiecki, M. E. Raymo, Paleoceanography 20,
PA1003 10.1029/2004PA001071 (2005).
12. Material and methods are available as supporting
material on Science Online.
13. L. E. Heusser, W. L. Balsam, Quat. Res. 7, 45 (1977).
14. M. F. Sánchez Goñi, F. Eynaud, J. L. K. Andersen,
N. J. Shackleton, Earth Planet. Sci. Lett. 171, 123 (1999).
15. A. Rochon, A. de Vernal, Can. J. Earth Sci. 31, 15 (1994).
16. C. Hillaire-Marcel, A. de Vernal, Géographi e physique et
Quaternaire 43, 263 (1989).
17. J. C. Bourgeois, K. Gajewski, R. M. Koerner, J. Geophys. Res.
106, (D6), 5255 (2001).
18. D. D. Rousseau et al., Rev. Palaeobot. Palynol. 141, 277
(2006).
19. C. Hillaire-Marcel, A. de Vernal, G. Bilodeau, A. J. Weaver,
Nature 410, 1073 (2001).
20. R. M. Koerner, J. Bourgeois, D. Fisher, Ann. Glaciol. 10,
85 (1988).
21. T. Giesecke, K. D. Bennett, J. Biogeogr. 31, 1523 (20 04).
22. A. G. Meesters, N. J. Bink, E. A. C. Henneken, H. F. Vugts, F.
Cannemeijer, Boundary-Layer Meteorol. 85, 475 (1997).
23. A. de Vernal, P. J. Mudie, Proceedings of the Ocean
Drilling Program 105B, 401 (1989).
24. J. T. Overpeck et al., Science 311, 1747 (2006).
25. U. Siegenthaler et al., Science 310, 1313 (2005).
26. A. Berger, M.-F. Loutre, Science 297, 1287 (2002).
27. Kaplan, J.O. et al., Journal of Geophysical Research 108
D19, 8171, doi:10.1029/2002JD002559 (2003).
28. A. de Vernal et al., Quat. Sci. Rev. 24, 897 (2005).
29. A. de Vernal, P. J. Mudie, in Neogene and Quaternary
Dinoflagellate Cysts and Acritarchs, M. J. Head, J. H. Wrenn,
Eds. (American Association of Stratigraphic Palynologists
Foundation, College Station, TX, 1992), p. 329.
30. National Oceanographic Data Center, World Ocean
Database 2001, Scientific Data Sets, Observed and Standard
Level Oceanographic Data (CD-ROM) (National Oceanic and
Atmospheric Administration, Silver Spring, MD, 2001).
31. This study is a contribution of the Polar Climate Stability
Network supported by the Canadian Foundation of
Climate and Atmospheric Science. We also acknowledge
financial support from the Natural Sciences and
Engineering Resear ch Council of Canada and the Fonds
Québécois de Recherche sur les Sciences de la Nature et
les Technologies.
Supporting Online Material
www.sciencemag.org/cgi/content/full/320/5883/1622/DC1
Materials and Methods
Figs. S1 to S4
References
10 December 2007; accepted 9 May 2008
10.1126/science.1153929
Fig. 4. Close-up on the stratigraphy of the last
interglacial period (MIS 5e) from core sample HU-90-
013-013, collected near ODP site 646. Shown are the
isotope stratigraphy based on Globigerina bulloides
and Neogloboquadrina pachyderma (19), the sea-
surface temperatures estimated from dinocysts (28),
and the concentration and percentages of the
dominant pollen and spore taxa. The percentages
of Osmunda were calculated from the pollen sum,
excluding spores.
www.sciencemag.org SCIENCE VOL 320 20 JUNE 2008 1625
REPORTS
on June 20, 2008 www.sciencemag.orgDownloaded from