Arctic Environmental Change of the Last Four Centuries

Abstract

A compilation of paleoclimate records from lake sediments, trees, glaciers, and marine sediments provides a view of circum-Arctic
environmental variability over the last 400 years. From 1840 to the mid-20th century, the Arctic warmed to the highest temperatures
in four centuries. This warming ended the Little Ice Age in the Arctic and has caused retreats of glaciers, melting of permafrost
and sea ice, and alteration of terrestrial and lake ecosystems. Although warming, particularly after 1920, was likely caused
by increases in atmospheric trace gases, the initiation of the warming in the mid-19th century suggests that increased solar
irradiance, decreased volcanic activity, and feedbacks internal to the climate system played roles.

Full text

Arctic Environmental Change of the
Last Four Centuries
J. Overpeck,* K. Hughen, D. Hardy, R. Bradley, R. Case, M. Douglas, B. Finney, K. Gajewski,
G. Jacoby, A. Jennings, S. Lamoureux, A. Lasca, G. MacDonald, J. Moore,
M. Retelle, S. Smith, A. Wolfe, G. Zielinski
A compilation of paleoclimate records from lake sediments, trees, glaciers, and marine
sediments provides a view of circum-Arctic environmental variability over the last 400
years. From 1840 to the mid-20th century, the Arctic warmed to the highest temperatures
in four centuries. This warming ended the Little Ice Age in the Arctic and has caused
retreats of glaciers, melting of permafrost and sea ice, and alteration of terrestrial and
lake ecosystems. Although warming, particularly after 1920, was likely caused by in-
creases in atmospheric trace gases, the initiation of the warming in the mid-19th century
suggests that increased solar irradiance, decreased volcanic activity, and feedbacks
internal to the climate system played roles.
Global climate change is likely amplified
in the Arctic by several positive feedbacks,
including ice and snow melting that de-
creases surface albedo, atmospheric stability
that traps temperature anomalies near the
surface, and cloud dynamics that magnify
change (1–3). The Arctic, in turn, influ-
ences climate change at lower latitudes
through changes in river runoff and effects
on global thermohaline circulation, impacts
on atmospheric circulation, and modulation
of atmospheric CO
2
and CH
4
concentra-
tions (1, 4). Although the Arctic is now
one of the least disturbed regions on Earth,
it may also be one of the most susceptible to
both natural and human-induced climate
change.
The instrumental record of Arctic cli-
mate change is brief and geographically
sparse. The few records that extend back to
the beginning of the 20th century suggest
that the Arctic has warmed by about 0.6°C
since that time, with peak temperatures
(1.2°C more than in 1910) around 1945
(2). Although some areas of the Arctic
have cooled recently—for example, as a
result of dynamics within the North Atlan-
tic (5)—the observed 20th-century Arctic-
average temperature increase exceeds that
of the hemisphere as a whole. The record is
thus consistent with amplification of cli-
mate change through snow/ice radiation
and other feedbacks (6, 7).
In this article, we use the paleoenviron-
mental record to assess the climate events
of this century from the perspective of the
last four centuries. We build on previous
work (8–10) by compiling a variety of com-
plementary paleoenvironmental indicators
of climate from around the entire Arctic.
This perspective permits the visualization of
natural subdecadal to century-scale climate
variability in the circum-Arctic region and
allows us to examine the role that natural
forcing mechanisms play in driving Arctic
climate. Given the growing focus on Arctic
environmental dynamics (4), we also exam-
ine the centuries-long paleoenvironmental
perspective to determine whether the cli-
mate changes we infer have resulted in
changes to the natural Arctic ecosystem.
The Paleoclimate Perspective
Proxy data from lakes, wetlands, ice cores,
and marine sources demonstrate that Arctic
J. Overpeck is at the National Oceanographic and Atmo-
spheric Administration (NOAA)–National Geophysical
Data Center (NGDC) Paleoclimatology Program, 325
Broadway, Boulder, CO 80303, USA, and Institute of
Arctic and Alpine Research (INSTAAR), University of Col-
orado, Boulder, CO 80309, USA. K. Hughen, A. Jen-
nings, J. Moore, and A. Wolfe are at INSTAAR, University
of Colorado, Boulder, CO 80309, USA. D. Hardy, R.
Bradley, and S. Smith are in the Department of Geo-
sciences, University of Massachusetts, Amherst, MA
01003, USA. R. Case and G. MacDonald are in the De-
partment of Geography, University of California, Los An-
geles, CA 90095, USA. M. Douglas is in the Department
of Geology, University of Toronto, Toronto, Ontario M5S
3B1, Canada. B. Finney is at the Institute of Marine Sci-
ence, University of Alaska, Fairbanks, AK 99775, USA. K.
Gajewski is in the Department of Geography, University of
Ottawa, Ottawa, Ontario K1N 6N5, Canada. G. Jacoby is
at the Lamont-Doherty Earth Observatory, Columbia Uni-
versity, Palisades, NY 10964, USA. S. Lamoureux is in
the Department of Earth and Atmospheric Sciences, Uni-
versity of Alberta, Edmonton, Alberta T6G 2H3, Canada.
A. Lasca and M. Retelle are in the Geology Department,
Bates College, Lewiston, ME 04204, USA. G. Zielinski is
at the Climate Change Research Center, University of
New Hampshire, Durham, NH 03824, USA.
*To whom correspondence should be addressed. E-mail:
jto@ngdc.noaa.gov
Table 1. Paleoclimate site information.
Map no. Site name Proxy type Source
1 Svalbard Ice Core Percent summer melt (69)
2 Tornetrask Tree ring widths (22, 28)
3 Polar Urals Tree ring widths (22, 70)
4 Lena River Tree ring widths This article
5 Kolyma (Jack London Lake) Tree ring widths (71)
6 Site 412 Tree ring widths (8, 9)
7 Arrigetch Peaks Tree ring widths (8, 9)
8 Dune Lake
13
C isotopes This article
9 Sheenjek Tree ring widths (8, 9)
10 TTHH Tree ring widths (8, 9)
11 Eagle Fecal Tree ring widths (8, 9)
12 Franklin Mountains Tree ring widths (8, 9)
13 MacKenzie Mountains Tree ring widths (8, 9)
14 Coppermine Tree ring widths (8, 9)
15 Hornby Cabin Tree ring widths (8, 9)
16 Churchill Tree ring widths (8, 9)
17 Lake DV09 Varve thicknesses (63)
18 Devon Island Ice Core Percent summer melt (27)
19 Lake C2 Varve thicknesses (21)
20 Lake C3 Varve thicknesses (65)
21 Lake Tuborg Varve thicknesses This article
22 Castle Peninsula Tree ring widths (8, 9)
23 Kuujuaq Tree ring widths (8)
24 Upper Soper Lake Dark lamination thicknesses (20)
25 Okak Tree ring widths This article
26 Salt Water Pond Tree ring widths (72)
27 Donard Lake Varve thicknesses This article
28 South Greenland Ice Core Percent summer melt (10, 73)
29 Nansen Fjord Fossil foraminifera (29)
ARTICLE
www.sciencemag.org zSCIENCE zVOL. 278 z14 NOVEMBER 1997 1251

interannual to century-scale environmental
variability of the last 400 years is superim-
posed on longer-term changes of the Holo-
cene (the last 10,000 years). Most of the
Arctic experienced summers warmer (1° to
2°C) than today during the early to middle
Holocene, but the timing of this Milanko-
vitch-driven warmth (that is, warmth re-
sulting from greater summer insolation) dif-
fered geographically because of the effects
of local sea surface temperatures (SSTs) and
land-ice cover (11–13). Decreases in sum-
mer insolation—perhaps coupled with oth-
er, more abrupt changes in climate forc-
ing—led to successively cooler summers in
the late Holocene; this trend culminated in
the Little Ice Age, a period that began
before 1600 (14–17) and evidently ended
sometime in the last century (11, 12, 14).
Our compilation of paleoclimate records
provides a 400-year view of Arctic variabil-
ity that is superimposed on lower frequency
Holocene trends. This view is based (18)on
more than 20 records from the North
American Arctic and substantially fewer
records spread out across the Eurasian Arc-
tic (Table 1 and Figs. 1 and 2). Although
covariation among continental records in
Alaska and western Canada (Fig. 2A) sug-
gests that the Eurasian records may be
broadly representative, a need for caution is
highlighted by the differences among
records from the eastern Canadian Arctic
and Greenland (Fig. 2B), a region marked
by geographic variability even in the instru-
mental period (2). Thus, although our com-
pilation provides a much improved basis for
understanding the patterns and causes of
interannual to century-scale environmental
variability in the circum-Arctic, future
work will be needed to increase data cover-
age, particularly in Eurasia.
The most notable pattern of change re-
vealed by most records across the Arctic is
the near ubiquitous transition from anoma-
lously cold conditions of the 19th century
to peak warm conditions of the 20th cen-
tury. This event can be seen even more
clearly in a new proxy record of average
Arctic temperature change over the last
400 years (19) (Fig. 3). Where quantitative
estimates are available, it appears that the
19th-to 20th-century warming was 1° to
3°C locally (8, 15, 20–25) and averaged
about 1.5°C across the Arctic. We thus
confirm that the 20th-century warming ob-
served in the instrumental record was only
part of a more circum-Arctic climate event
that began in the mid-19th century and
marked the end of the Little Ice Age in the
Arctic (8, 26, 27).
Another striking aspect of Arctic tem-
perature change over the last 400 years is
that most of the Arctic experienced cooling
in the first part of the 19th century, result-
ing in the coldest temperatures of the Little
Ice Age. In addition, although much of the
Arctic was colder than today during parts of
the 17th century, several of the records
show temperatures nearly as warm as today
during the 18th century (9). Arctic climate
change before 1800 was more regional in
nature than after this time. For example,
many sites around the Arctic were warmer
at some time between 1700 and 1820 than
they were later in the 19th century, but the
timing and duration of this warmth varied
from region to region, as did the preceding
17th-century period of generally colder
conditions.
The annually dated record of Arctic cli-
mate variability encompassing the last 1000
years has less spatial coverage than does the
multiproxy record of the last 400 years.
Sediment, ice core, historical, and tree ring
data for this earlier period indicate that
although Arctic summers of the 20th cen-
tury were generally the warmest of the last
400 years, they may not have been the
warmest of the last millennium (21, 28–
30). The few time series of climate change
spanning the last millennium also suggest
that the Arctic was not anomalously warm
throughout the so-called Medieval Warm
Period of the 9th to 14th centuries (31).
Climate Forcing Mechanisms
The paleoclimate record of the Arctic re-
veals large variability over all of the last 400
years. Natural variability in most regions
was as large before the 20th-century buildup
of atmospheric trace gases as afterward
(Figs. 2 and 3). Natural climate forcing
mechanisms must therefore be considered
before attributing some or all of the recent
changes in the Arctic to human influences.
The similarities of our new Arctic climate
reconstruction (Fig. 3) to earlier re-
constructions for the Northern Hemisphere
(8, 10), and the coincidence of high inferred
solar irradiance and warm Arctic tempera-
tures in both the 18th and 20th centuries,
suggest that solar forcing played a role (32–
36). The strong recovery of high tempera-
tures after the 1840s may also have been a
partial response to high solar irradiance at
that time. On the other hand, the lack of a
distinct prolonged cold period associated
with the 17th-to early 18th-century Maun-
der sunspot minimum period (34) argues
against a dominant role for solar forcing over
the past 400 years, as does the observation
that changes in solar irradiance evidently did
not lead climate warming into the 20th
century.
Comparison of our temperature record
with an Arctic volcanic sulfate record re-
cently developed from the Greenland Ice
Lake sediment records
Marine sediment records
Ice core records
Tree ring records
Fig. 1. Map showing locations of proxy climate records used in this study and listed in Table 1. Circles
around groups of sites indicate the groups averaged before calculating the Arctic-wide temperature
average. Isolated single sites not circled were averaged directly into the Arctic-wide series (19).
SCIENCE zVOL. 278 z14 NOVEMBER 1997 zwww.sciencemag.org1252

Sheet Project Two (GISP2) ice core (37,
38) indicates that most peaks in recon-
structed atmospheric volcanic sulfate load-
ing correspond to mean circum-Arctic cool-
ing (Fig. 3). The repeated coincidence of
high sulfate loading with the onset of Arc-
tic cold events suggests that eruptions en-
train positive ocean feedbacks capable of
enhancing and prolonging Arctic cooling.
For example, the anomalous early 19th-
century period of frequent large sulfur-pro-
ducing eruptions seems to have helped pre-
cipitate the drop into the coolest period of
the Arctic Little Ice Age (37, 39). These
findings agree with earlier evidence for vol-
canic forcing of Arctic temperatures over
the last 100 years (40), as well as the more
regional 200-year record of volcano-climate
linkages for sub-Arctic North America (41)
and earlier assertions that volcanic activity
plays a role in modulating the climate of the
Northern Hemisphere in general (10, 42).
It is difficult to attribute the lack of a clear
solar-climate correspondence, particularly
during the period of the Maunder sunspot
minimum (1650 to 1710), to counterbal-
ancing by volcanic forcing. However, it is
quite likely that the period of uncommonly
low volcanic activity from 1935 to 1960
may have contributed to the peak Arctic
summer temperatures at this time (43, 44),
reducing the implied potential sensitivity of
the climate system to solar forcing during
this same period (36).
Variability internal to Earth’s climate
system, particularly in the ocean’s thermo-
haline circulation, also can modulate cli-
mate over decades to centuries (45, 46).
The variability evident in our 400-year
Arctic compilation, although an order of
magnitude smaller in scale than that driven
by internal climate system processes over
the last deglaciation (47), may relate to
decadal-scale fluctuations in the state of the
North Atlantic Oscillation (NAO) or to
the transport of heat northward by thermo-
haline circulation. Such a forcing mecha-
nism would leave a diagnostic pattern most
obvious in, and downwind of, the North
Atlantic (46, 48). The Nansen Fjord record
from southeast Greenland (Fig. 2B), a proxy
for North Atlantic SSTs (49), indicates
that the range of North Atlantic variability
over the past 400 years exceeds that of the
recent instrumental record, and that the
last century was marked by warming in the
North Atlantic. However, comparison of
this long proxy record of North Atlantic
SSTs with terrestrial temperature proxies
from sites across the Arctic (Fig. 2) does not
reveal clear evidence for strong Atlantic
forcing of Arctic climate. The full role of
NAO and thermohaline forcing of Arctic
climate remains to be evaluated as more
multicentury time series from the circum–
North Atlantic region become available.
Half of the post-1840 warming (about
0.75°C) took place from 1840 to 1920,
during a period in which concentrations of
atmospheric CO
2
and CH
4
increased only
by about 20 ppm and 200 parts per billion
by volume (ppbv), respectively (Fig. 3).
Given the current best estimates of climate
sensitivity to trace-gas forcing [1.5° to 4.5°
global mean warming for a 280-ppm in-
crease in CO
2
, and less for a doubling of
CH
4
(6, 46)], trace-gas forcing alone might
be able to explain only a small part (0.1° to
0.4°C) of the pre-1920 warming. The first
half of the unprecedented 19th-to 20th-
century increase in Arctic temperatures ap-
pears to be a natural readjustment as volca-
nic forcing weakened, and irradiance (and,
to a lesser extent, greenhouse gases) in-
creased, between 1840 and 1920. After
1920, both anomalously high solar irradi-
ance and low volcanic aerosol loading likely
AB
1600 1700 1800 1900 2000 1600 1700 1800 1900 2000
Year
Europe Russia Alaska Western Canada
Eastern CanadaCentral Canada Greenland
Fig. 2. (A) Standardized 400-year
proxy climate records reflecting
surface air temperature for sites
from Arctic Europe east to western
Canada (Fig. 1 and Table 1) (18).
Red indicates temperatures greater
than one standard deviation warm-
er than average for the reference
period (1901–1960), whereas dark
blue indicates at least one standard
deviation colder than this average.
(B) Same as (A) but for sites in Can-
ada east to Greenland. All series are
presented as 5-year averages ex-
cept for sites 8 and 29, which are
plotted at their original lower reso-
lution. All time series represent sur-
face air temperature except for site
29, which represents SST. All time
series shown will be available at
http://www.ngdc.noaa.gov/paleo/
paleo.html.
ARTICLE
www.sciencemag.org zSCIENCE zVOL. 278 z14 NOVEMBER 1997 1253

continued to influence Arctic climate, but
exponentially increasing atmospheric trace-
gas concentrations probably played an in-
creasingly dominant role (36). More recent-
ly, the observed slowdown in warming from
1950 to 1970 (2) (Fig. 3) may have been
influenced by the increase in Arctic tropo-
spheric aerosols that occurred after 1950 (6,
50).
The Response to
Arctic Climate Variability
With the emergence of a coherent picture
of Arctic climate change over the last sev-
eral centuries comes the realization that
interannual to century-scale Arctic climate
variability is the norm. This variability has
affected many aspects of the Arctic envi-
ronment. The primary implication is that
today’s Arctic cryosphere (glaciers and fro-
zen ground) and biosphere (terrestrial,
lacustrine, and marine) are not at steady
state; they have changed and will continue
to change in response to an evolving Arctic
climate.
An obvious impact of recent climate
change is the widespread retreat of glaciers
throughout the Arctic over the last century
(51). On Baffin Island, retreats in glacial
equilibrium-line altitudes (52) coincided
with the 19th-to 20th-century warming
observed in sediment, tree ring, and ice core
records. Evidence for similarly large post–
Little Ice Age glacial retreats can also be
found farther east in Greenland, Iceland,
Spitsbergen, and Scandinavia (14, 53), as
well as to the north on Devon Island (27)
and to the west in Alaska (25). Although it
has been suggested that climate warming
may be accompanied by an increase in
snowfall sufficient to expand ice sheets
(54), it appears probable that most Arctic
glaciers will continue to melt if the Arctic
continues to warm, as forecast by model
simulations of the climate response to in-
creasing greenhouse gas concentrations (6,
55). This inference may not apply to the
Greenland Ice Sheet, whose modern mass
balance is poorly known (51).
High-latitude permafrost conditions have
also changed during the last 200 years.
Where available (that is, North America),
ground temperature records reconstructed
from borehole temperature logs support the
notion that large-scale warming has oc-
curred since the 19th century, but these
records also indicate that high-latitude per-
mafrost conditions have changed, and are
likely to continue changing, as the Arctic
warms (24). This future change in frozen
ground will, in turn, likely affect construc-
tion, transportation, hydrology, ecology,
and trace-gas fluxes in the Arctic (4).
The paleoenvironmental record of the
last four centuries supports recent ecologi-
cal studies (56) and indicates that, even if
climate warming takes place at the low end
of the range suggested by climate model
simulations of the next 100 years (6), such
warming is capable of driving changes in
the Arctic biosphere that will likely rival
any changes driven by nonclimate process-
es, including human land use. Tree growth
correlates positively with temperature in
many parts of the Arctic (Fig. 2), just as
Arctic tree growth forms and fire distur-
bance regimes change with climate (57).
Studies of recent seedling establishment
and viability north of the present treeline
(58, 59), together with evidence from an-
nually dated fossil pollen records, suggest
that Arctic plant species abundances and
ranges have varied in response to varying
temperatures (60).
The limnology of Arctic lakes will also
likely continue to respond to climate
change (61). Sediments from shallow ponds
and lakes on Ellesmere (62) (Fig. 4) and
Devon islands (63) show that warming
since the mid-19th century caused acute
shifts in diatom algal floras, including the
recent establishment in some lakes of dia-
tom populations previously too light-limit-
ed by perennial ice for their subsistence.
Freshwater algae are sensitive recorders of
environmental change and may also be har-
bingers of more profound ecological reorga-
nizations, as changes in their populations
may influence the structure of higher tro-
phic levels. The paleoenvironmental record
of the Arctic makes it clear that we should
expect changes to occur in both terrestrial
and aquatic ecosystems if the present Arctic
warming persists.
Finally, our records confirm the hypoth-
esis, originally based on a single high-Arctic
ice core record of summer temperature vari-
ations (27), that the repeated failure of Euro-
peans to find the Northwest Passage during
the 19th century was likely a result of excep-
tionally severe summertime air temperature
and sea ice extent (64). Conditions more
amenable to Arctic sea travel, and to the
discovery of the Northwest Passage, arose
naturally after the turn of the century (27).
Implications for the Future
Our reconstruction of past environmental
change in the Arctic suggests that natural
A
B
C
D
0.5
0
-0.5
-1
-1.5
0.5
0
-0.5
-1
-1.5
0.5
0
-0.5
-1
-1.5
0.5
0
-0.5
-1
-1.5 -50
0
50
100
150
1363
1364
1365
1366
1367
1368
1369
250
260
270
280
290
300
310
320
330
600
800
1000
1200
1400
1600
1600 1650 1700 1750 1800 1850 1900 1950 2000
Temperature anomaly (sigma units)
CH4 (ppbv) CO2 (ppm) Total irradiance (W/m2)SO
42- residual (ppb)
Year
Fig. 3. Comparison of
hypothesized external
climate forcing (colored
lines) and standard-
ized proxy Arctic-wide
summer-weighted annu-
al temperature [gray
lines, plotted as sigma
units (18, 19)] for (A)
atmospheric CH
4
(66),
(B) atmospheric CO
2
(67), (C) solar irradiance
(32), and (D) Greenland
(GISP2) ice core volcanic
sulfate. Eruptions known
to be overrepresented in
the GISP2 record are
marked with an asterisk
(37, 38).
SCIENCE zVOL. 278 z14 NOVEMBER 1997 zwww.sciencemag.org1254

variability is large in this region and is
working together with human forcing
(through increased concentrations of atmo-
spheric trace gases) to drive unprecedented
changes in the Arctic environment. The
complexity of natural and anthropogenic
forcing highlights the probability that as-
sumptions of climate stability, or efforts to
simply extrapolate past patterns of change
into the future, will ultimately fail to antic-
ipate future Arctic climate change and its
impacts. Reliable predictions of future
change will require climate system models
that prove effective in simulating past
changes such as those reconstructed here.
Even as these models are being developed
and tested, however, the Arctic environ-
ment is likely to continue its pace of
change.
REFERENCES AND NOTES
___________________________
1. J. T. Houghton, G. J. Jenkins, J. J. Ephraums, Eds.,
Climate Change: The IPCC Scientific Assessment
(Cambridge Univ. Press, Cambridge, 1990).
2. W. L. Chapman and J. E. Walsh, Bull. Am. Meteorol.
Soc.74, 33 (1993).
3. J. A. Curry et al., J. Clim. 9, 1731 (1996).
4. G. Weller et al., J. Biogeogr. 22, 365 (1995).
5. B. Dickson, Nature 386, 649 (1997).
6. J. T. Houghton et al., Eds., Climate Change 1995
(Cambridge Univ. Press, Cambridge, 1996).
7. P. Y. Groisman et al., Science 263, 198 (1994); P. Y.
Groisman et al., J. Clim. 7, 1633 (1994); M. C. Ser-
reze, J. A. Maslanik, J. R. Key, R. F. Kokaly, Geo-
phys. Res. Lett. 22, 2183 (1995).
8. G. C. Jacoby and R. D. D’Arrigo, Clim. Change 14,
39 (1989).
9. R. D. D’Arrigo and G. C. Jacoby, in Climate Since
A.D. 1500, R. S. Bradley and P. D. Jones, Eds.
(Routledge, London, 1992), pp. 296–311; R. D.
D’Arrigo and G. C. Jacoby, Clim. Change 25, 163
(1993).
10. R. S. Bradley and P. D. Jones, The Holocene 3, 367
(1993).
11. R. S. Bradley, Quat. Sci. Rev.9, 365 (1990).
12. L. D. Williams and R. S. Bradley, in Quaternary Envi-
ronments—Eastern Canadian Arctic, Baffin Bay and
Western Greenland, J. T. Andrews, Ed. (Allen and
Unwin, Boston, 1985), pp. 741–772.
13. H. F. Diaz et al., Arct. Alp. Res. 21, 45 (1989); J. I.
Svendsen and J. Mangerud, Clim. Dyn. 6, 213
(1992); G. M. MacDonald, T. W. D. Edwards, K. A.
Moser, R. Pienitz, J. P. Smol, Nature 361, 243
(1993); H. E. Wright Jr. et al., Eds., Global Climates
Since the Last Glacial Maximum (Univ. of Minnesota
Press, Minneapolis, 1993); P. M. Anderson and L. B.
Brubaker, Quat. Sci. Rev.13, 71 (1994); D. A. Fisher
et al., The Holocene 5, 19 (1995); S. C. Zoltai, Geogr.
Phys. Quat.49, 45 (1995); K. M. Williams et al., ibid.,
p. 13; R. B. Alley and S. Anandakrishnan, Ann. Gla-
ciol. 21, 64 (1995); P. J. H. Richard, Geogr. Phys.
Quat.49, 117 (1995); A. S. Dyke et al., ibid.50, 125
(1996).
14. J. M. Grove, The Little Ice Age (Methuen, London,
1988).
15. P. D. Jones and R. S. Bradley, in Climate Since A.D.
1500, R. S. Bradley and P. D. Jones, Eds. (Rout-
ledge, London, 1992), pp. 649–666.
16. D. Meese et al., Science 266, 1680 (1994).
17. S. R. O’Brien et al., ibid. 270, 1962 (1995).
18. The proxy temperature time series in Fig. 2 were
chosen because they were available in time series
form. Most of the paleoclimate time series have been
calibrated against instrumental data, but the scarcity
of weather stations in the Arctic makes it necessary
to rely on similar records that have been calibrated
directly against instrumental data, or to use theory, in
order to define the nature of the climate signal rep-
resented in a few of the proxy records. For example,
ice cores are commonly far from any instrumental
station, but theory supports the interpretation of ice
core melt layers as a proxy of summer temperature
[whereas ice core stable isotopic variations are more
complex on decadal time scales (10)]. Likewise, the
selection of site characteristics ensures that tree ring
chronologies correlate well with summer or annual
temperatures where they can be compared with
nearby instrumental records, reinforcing the likeli-
hood that tree ring width also correlates well with
summer or annual temperature at sites remote from
any instrumental station (8, 9, 26). Theory and (in
cases where nearby meteorological data are avail-
able) direct correlation with instrumental data sup-
port the assertion that the lake sediment records
included in this study are also useful centuries-long
proxies for early summer temperatures (20, 21, 65)
[D. R. Hardy et al.,J. Paleolimnol. 16, 227 (1996)].
Thus, although most of the time series summarized
here have been calibrated quantitatively against
summer temperature, it is more appropriate to plot
raw proxy data (x) in terms of “sigma units” defined
as normalized deviations (z) from the 1901–1960
mean (m): z5(x–m)/s, where sis the standard
deviation of the raw series for the period 1901–1960.
The reference period was selected to highlight the
20th century in the context of preceding centuries.
Five-year averages are plotted to emphasize the low-
er frequency patterns of variability and to accommo-
date the ;1% temporal error that characterizes the
sediment and ice core–based records. Tree ring
chronologies generally are accurate at annual res-
olution, but they may be biased in Alaska after 1970
by moisture stress or new anthropogenic causes
(58).
19. The 400-year Arctic average summer temperature
series (Fig. 3) was generated by first averaging the
Fig. 4. Lake sediment records
from climatically and limnologi-
cally contrasting regions of Elles-
mere Island (62, 68), each show-
ing abrupt changes in the compo-
sition of fossil diatom assemblag-
es deposited within approxi-
mately the last 150 years. These
biostratigraphic changes are un-
related to differential silica preser-
vation and represent the greatest
floristic shifts of the middle to late
Holocene. Taxonomic diversifica-
tion with greater representations
of littoral and periphytic taxa (Ato
C), increased diatom algal bio-
mass (C), and recent diatom re-
colonization (D) are all consistent
with the abrupt 19th- to 20th-
century shift toward longer sum-
mer growing seasons, reduced
lake-ice severity, and greater hab-
itat availability. The limnological
consequences of the 19th- to
20th-century warming appear to
be unprecedented in the con-
text of pre–18th century natural
variability. Except where noted,
tick marks on horizontal axes
represent 10% relative frequen-
cy intervals.
ARTICLE
www.sciencemag.org zSCIENCE zVOL. 278 z14 NOVEMBER 1997 1255

standardized 5-year–averaged series (Fig. 2) for
each of the groups circled in Fig. 1 (to eliminate site
density bias) and then averaging the group averages
with the other single sites not circled in Fig. 1. Sites 8
and 29 were excluded because of their subdecadal
resolution.
20. K. A. Hughen et al., in preparation.
21. S. F. Lamoureux and R. S. Bradley, J. Paleolimnol.
16, 239 (1996).
22. K. R. Briffa et al., in Climatic Variations and Forcing
Mechanisms of the Last 2000 Years, P. D. Jones,
R. S. Bradley, J. Jouzel, Eds. (Springer-Verlag, Ber-
lin, 1996), pp. 9–41.
23. H. Garfinkel and L. B. Brubaker, Nature 286, 872
(1980).
24. Borehole climate reconstructions do not have the
high temporal resolution of the records shown in Fig.
2, but they provide a clear confirmation of the recent
warming trend seen in the annually dated proxies, as
well as useful quantitative estimates of the magni-
tude of warming [A. H. Lachenbruch et al., J. Geo-
phys. Res.87, 9301 (1982); A. H. Lachenbruch and
B. V. Marshall, Science 234, 689 (1986); S. Kakuta,
Palaeogeogr. Palaeoclimatol. Palaeoecol. 98, 225
(1992); H. Beltrami and A. E. Taylor, Global Planet.
Change 11, 127 (1995)].
25. Other quantitative records include: P. E. Calkin,
Quat. Sci. Rev.7, 159 (1988); iiii and G. C.
Wiles, in International Conference on the Role of Po-
lar Regions in Global Change, G. Weller, Ed. (Univer-
sity of Alaska, Fairbanks, AK, 1991), pp. 617–625;
C. Caseldine and J. Sto¨tter, The Holocene 3, 357
(1993); J. M. Szeicz and G. M. MacDonald, Quat.
Res.44, 257 (1995); L. H. Evison et al., The Holo-
cene 6, 17 (1996).
26. G. C. Jacoby et al., Science 273, 771 (1996).
27. R. M. Koerner, ibid. 196, 15 (1977); iiiiand D. A.
Fisher, Nature 343, 630 (1990).
28. K. R. Briffa et al., Clim. Dyn. 7, 111 (1992).
29. A. E. Jennings and N. J. Weiner, The Holocene 6,
179 (1996).
30. W. Dansgaard et al., Nature 255, 24 (1975); A. E. J.
Ogilvie, Acta Archaeol. 61, 233 (1990).
31. M. K. Hughes and H. F. Diaz, Clim. Change 26, 109
(1994).
32. J. Lean et al., Geophys. Res. Lett. 22, 3195 (1995).
33. T. J. Crowley and K.-Y. Kim, ibid. 23, 359 (1996).
34. In general, the reconstructed record of total solar
irradiance (32) agrees well with independent records
of solar activity inferred from
14
C and
10
Be cosmo-
genic isotopes (35). All three records agree that the
lowest total solar irradiance of the last 400 years
occurred in the interval 1650–1710. This period was
characterized by Arctic temperatures that, although
lower than at present, were not anomalously cold
relative to the interval 1600–1650.
35. M. Stuiver and T. F. Braziunas, The Holocene 3, 289
(1993); J. Beer et al., in The Sun as a Variable Star,
J. M. Pap, C. Fro¨lich, H. S. Hudson, S. K. Solanki,
Eds. (Cambridge Univ. Press, Cambridge, 1994),
pp. 291–300.
36. Lean et al.(32) carried out a correlation analysis
between reconstructed solar irradiance (the same
used in our Fig. 3) and the Northern Hemisphere
temperature reconstruction of Bradley and Jones
(10). For the period 1610 to 1800, they found the
solar temperature correlation to be 0.86. Assuming
that the period 1610 to 1800 was influenced by nat-
ural climate forcing only, they used regression to
define the solar-temperature relation and extrapolate
it to the present, with the resulting inference that up
to 50% of the Northern Hemisphere warming be-
tween 1860 and 1970 is attributable to solar forcing.
When we repeated the analysis using Arctic temper-
ature, we found that the 1610 to 1800 correlation
between solar irradiance and Arctic temperature
(Fig. 3) is less high at 0.47, whereas the correlation
between the volcanic activity (38) and Arctic temper-
ature (Fig. 3) over the same period is higher (–0.21,
calculated without any lags) than was suggested by
the comparison with the Northern Hemisphere
record (32). Extrapolation of the 1610 to 1800 rela-
tions to the present suggests that increasing atmo-
spheric trace-gas concentrations have been the
dominant influence on Arctic temperatures after
1920, but that solar and volcanic forcing also play a
role.
37. G. A. Zielinski et al., Science 264, 948 (1994).
38. Almost all sulfate peaks in recent polar ice core
records [in this case, the Greenland GISP2 ice core
volcanic sulfate residuals (37)] correspond to known
volcanic eruptions, and the few that are not known
from the historical record (such as the large event in
1809) are known to be real because they can also be
seen in other ice core records, including some
events recorded in the Southern Hemisphere [E.
Mosley-Thompson et al., Quat. Sci. Rev.12, 419
(1993)]. All major peaks in the GISP2 ice core sulfate
record, except the peak in 1912, correspond to a
cool event in our proxy summer temperature record
for the Arctic (Fig. 3). However, the sulfate peaks
associated with the Icelandic Laki (1783) and Alas-
kan Katmai/Novarupta (1912) eruptions, both rela-
tively proximal to Greenland, have been shown to be
overestimates of the true stratospheric volcanic op-
tical depth loading generated by these eruptions (39,
44).
39. G. A. Zielinski, J. Geophys. Res.100, 20937 (1995).
40. S. Self, M. R. Rampino, J. J. Barbera, J. Volcanol.
Geotherm. Res. 11, 41 (1981); W. B. Lyons et al.,
Ann. Glaciol. 14, 176 (1990).
41. K. R. Briffa et al., J. Geophys. Res.99, 25,835
(1994).
42. S. C. Porter, Quat. Res. 26, 27 (1986).
43. R. S. Bradley and P. D. Jones, in Climate Since A.D.
1500, R. S. Bradley and P. D. Jones, Eds. (Rout-
ledge, London, 1992), pp. 606–622; M. Sato, J. E.
Hansen, M. P. McCormick, J. B. Pollack, J. Geo-
phys. Res. 98, 22987 (1993); A. Robock and M. P.
Free, ibid.100, 11549 (1995).
44. R. B. Stothers, J. Geophys. Res. 101, 3901 (1996).
45. T. F. Stocker and L. A. Mysak, Clim. Change 20, 227
(1992).
46. D. Rind and J. T. Overpeck, Quat. Sci. Rev.12, 357
(1993).
47. G. Bond et al., Nature 365 143, (1993).
48. H. van Loon and J. C. Rogers, Mon. Weather Rev.
106, 296 (1978); J. C. Rogers and H. van Loon, ibid.
107, 509 (1979); J. W. Hurrell, Science 269, 676
(1995).
49. As shown by Jennings and Weiner (29), the Nansen
Fjord record covaries closely with the record of
Icelandic sea ice severity as reconstructed from
historical accounts. These records also show
strong covariation with the one other available cen-
turies-long decadal-resolution North Atlantic SST
record—that of L. D. Keigwin [Science 274, 1504
(1996)]. Thus, we consider the Nansen record to be
an appropriate proxy for the time-dependent vari-
ations in high-latitude North Atlantic SST over the
last 400 years.
50. P. A. Mayewski et al., Science 232, 975 (1986); P. A.
Mayewski et al., Nature 346, 554 (1990); P. A.
Mayewski et al., Atmos. Environ. 27, 2915 (1993).
51. J. A. Dowdeswell, Quat. Res.48, 1 (1997).
52. G. H. Miller, ibid.3, 561 (1973); P. T. Davis, in Qua-
ternary Environments—Eastern Canadian Arctic,
Baffin Bay and Western Greenland, J. T. Andrews,
Ed. (Allen & Unwin, Boston, 1985), pp. 682–718.
53. A. Weidick et al., Glacial Inventory and Atlas of West
Greenland, Grøndlands Geologiske Undersøgelse
Repport 158 (Greenland Geological Survey, Copen-
hagen, 1992); C. Caseldine and J. Sto¨ tter, The Ho-
locene 3, 357 (1993); A. Werner, ibid., p. 128.
54. G. H. Miller and A. de Vernal, Nature 355, 244
(1992).
55. S. L. Thompson and D. Pollard, J. Clim. 10, 871
(1997).
56. F. S. Chapin III et al., Ecology 76, 694 (1995).
57. S. Payette, L. Filion, A. Delwaide, C. Be´ gin, Nature
341, 429 (1989); Y. Bergeron and S. Archambault,
The Holocene 3, 255 (1993).
58. G. C. Jacoby and R. D. D’Arrigo, Global Biogeo-
chem. Cycles 9, 227 (1995).
59. D. J. Cooper, Arctic 39, 247 (1986).
60. J. T. Overpeck, in Climatic Variations and Forcing
Mechanisms of the Last 2000 Years, P. D. Jones,
R. S. Bradley, J. Jouzel, Eds. (Springer-Verlag, Ber-
lin, 1996), pp. 479–497.
61. D. R. Hardy and R. S. Bradley, Geosci. Can. 23, 217
(1997).
62. M. S. V. Douglas, J. P. Smol, W. Blake Jr., Science
226, 416 (1994); N. C. Doubleday, M. S. V. Douglas,
J. P. Smol, Sci. Total Environ. 160–161, 661 (1995);
A. P. Wolfe, Geol. Surv. Can. Mem., in press.
63. K. Gajewski, P. B. Hamilton, R. McNeely, J. Pale-
olimnol. 17, 215 (1997).
64. B. T. Alt, R. M. Koerner, D. A. Fisher, J. C. Bourgeois,
in The Franklin Era in Canadian Arctic History,P.D.
Sutherland, Ed. (National Museum of Canada, Otta-
wa, 1985), pp. 69–92; M. Dunbar, in Climatic
Change in Canada 5, Syllogeus 55, C. R. Harington,
Ed. (National Museum of Canada, Ottawa, 1985),
pp. 107–119; J. Newell, in Climatic Change in Can-
ada 3, Syllogeus 49, C. R. Harington, Ed. (National
Museum of Canada, Ottawa, 1983), pp. 108–129.
65. A. N. Lasca, thesis, Bates College, Lewiston, ME
(1997).
66. T. Nikazawa et al,Geophys. Res. Lett. 20, 943
(1993).
67. D. M. Etheridge et al., J. Geophys. Res.101, 4115
(1996).
68. The inflection between unsupported and back-
ground (in situ production)
210
Pb activities in the Col
Pond and Elison Lake cores indicates a date of about
1850, which coincides with the onset of major dia-
tom shifts in these cores. The Solstice Lake chronol-
ogy is based on a linear interpolation between calen-
dar age–calibrated radiocarbon ages. This model
places the major diatom changes within the last 120
years. Although the Lower Dumbell Lake core has no
radiometric chronology, it also suggests that sub-
stantial floristic change occurred within the last 100
to 150 years.
69. A. Tarussov, in Climate Since A.D. 1500, R. S. Brad-
ley and P. D. Jones, Eds. (Routledge, London,
1992), pp. 505–516.
70. K. R. Briffa et al., Nature 376, 156 (1995).
71. C. J. Earle et al., Arct. Alp. Res. 26, 60 (1994).
72. R. D. D’Arrigo et al., Can. J. For. Res.26, 143 (1996).
73. T. Kameda et al., in Proceedings of the International
Symposium on the Little Ice Age Climate, T. Mikami,
Ed. ( Tokyo Metropolitan University, Tokyo, 1992), p.
101.
74. This work is a Paleoclimates of Arctic Lakes and
Estuaries (PALE) contribution to the NSF Arctic
System Science (ARCSS) Program. It is also a
product of the NSF- and NOAA-sponsored Analy-
sis of Rapid and Recent Climatic Change (ARRCC)
and other non-PALE (for example, dendroclimatic)
projects, and is also a result of research funded by
the Natural Sciences and Engineering Research
Council of Canada (NSERC), including the Paleo-
ecological Analysis of Circumarctic Treeline
(PACT ) program. The Polar Continental Shelf
Project provided logistical field support. We thank
our colleagues for discussion of our work at
ARCSS workshops, as well as J. Cole, J. Lean, A.
Ogilvie, and two anonymous reviewers for critical
comments on the manuscript.-
SCIENCE zVOL. 278 z14 NOVEMBER 1997 zwww.sciencemag.org1256