Content uploaded by Selvaraj Kandasamy
Author content
All content in this area was uploaded by Selvaraj Kandasamy on Aug 25, 2015
Content may be subject to copyright.
NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience 1
Knowledge of natural climate variability is essential to better con-
strain the uncertainties in projections of twenty-rst-century
climate change1–5. e past 2,000 years (2kyr) have emerged
as a critical interval in this endeavour, with sucient length to char-
acterize natural decadal-to-centennial scale change, known external
climate forcings6 and with distinctive patterns of spatiotemporal tem-
perature variations7. However, reconstructions for the full 2kyr inter-
val are not available for the global ocean, a primary heat reservoir8
and an important regulator of global climate on longer timescales9–11.
Here we present a global ocean sea surface temperature (SST) syn-
thesis (Ocean2k SST synthesis) spanning the Common Era, which
shows a cooling trend that is similar, within uncertainty, to that sim-
ulated by realistically forced climate models for the past millennium.
We use the simulations to identify the climate forcing(s) consistent
with reconstructed SST variations during the pastmillennium.
The Ocean2k SST synthesis data set
e Ocean2k SST synthesis is based on 57 marine-origin, peer-
reviewed and publicly available SST reconstructions spanning some
Robust global ocean cooling trend for the
pre-industrial Common Era
Helen V. McGregor1*, Michael N. Evans2, Hugues Goosse3, Guillaume Leduc4, Belen Martrat5,6,
Jason A. Addison7, P. Graham Mortyn8, Delia W. Oppo9, Marit-Solveig Seidenkrantz10,
Marie-Alexandrine Sicre11, Steven J. Phipps12,13, Kandasamy Selvaraj14, Kaustubh Thirumalai15,
Helena L. Filipsson16 and Vasile Ersek17
The oceans mediate the response of global climate to natural and anthropogenic forcings. Yet for the past 2,000 years — a key
interval for understanding the present and future climate response to these forcings — global sea surface temperature changes
and the underlying driving mechanisms are poorly constrained. Here we present a global synthesis of sea surface tempera-
tures for the Common Era () derived from 57 individual marine reconstructions that meet strict quality control criteria. We
observe a cooling trend from 1 to 1800 that is robust against explicit tests for potential biases in the reconstructions. Between
801 and 1800, the surface cooling trend is qualitatively consistent with an independent synthesis of terrestrial temperature
reconstructions, and with a sea surface temperature composite derived from an ensemble of climate model simulations using
best estimates of past external radiative forcings. Climate simulations using single and cumulative forcings suggest that the
ocean surface cooling trend from 801 to 1800 is not primarily a response to orbital forcing but arises from a high frequency of
explosive volcanism. Our results show that repeated clusters of volcanic eruptions can induce a net negative radiative forcing
that results in a centennial and global scale cooling trend via a decline in mixed-layer oceanic heat content.
or all of the past 2kyr (Fig.1; Methods; Supplementary Section 1;
Supplementary Table S1; Supplementary Fig. S1). e temporal
diversity and spatial distribution of the 57 reconstructions present
methodological challenges for generating a global synthesis product:
• To avoid biases towards reconstructions with greater temporal
resolution, each SST reconstruction was averaged into 200-yr
‘bins’ (that is, 200-yr averages for 1–200 and so on, up to
1801–2000; Methods; Supplementary Section 3). is allows
for typical errors in reservoir-corrected radiocarbon dates, the
most common dating method used for the 57 reconstructions. In
addition, the oceans’ large thermal inertia and integration of short-
term climate variations10 suggest that at 200-yr resolution we may
expect a global SST signature to emerge from the datacomposite.
• Our network includes near-polar to tropical regions (Fig.1) that
record a wide range of SST means and variances (Supplementary
Fig. S1). Consequently, our 200-yr binned reconstructions are
standardized (Methods) prior to compositing. Standardizing is
routinely employed to composite reconstructions with dierent
1School of Earth and Environmental Sciences, Northfields Avenue, University of Wollongong, New South Wales 2522, Australia. 2Department of Geology
and Earth System Science Interdisciplinary Center, University of Maryland, College Park, Maryland 20742, USA. 3Earth and Life Institute, Université de
Louvain, Place Pasteur 3, 1348 Louvain-la-Neuve, Belgium. 4Aix Marseille Université, CNRS, IRD, CEREGE UM34, 13545 Aix-en-Provence Cedex 4, France.
5Department of Environmental Chemistry, Institute of Environmental Assessment and Water Research (IDÆA), Spanish Council for Scientific Research
(CSIC), 08034 Barcelona, Spain. 6Department of Earth Sciences, University of Cambridge, Downing Street, Cambridge CB2 3EQ, UK. 7U.S. Geological Survey,
345 Middlefield Road, MS 910, Menlo Park, California 94025, USA. 8Institute of Environmental Science and Technology (ICTA) and Department of Geography,
Universitat Autonoma de Barcelona, Bellaterra 08193, Spain. 9Department of Geology and Geophysics, Woods Hole Oceanographic Institution, Woods Hole,
Massachusetts 02543, USA. 10Centre for Past Climate Studies and Arctic Research Centre, Department of Geoscience, Aarhus University, Hoegh-Guldbergs
Gade 2, DK-8000 Aarhus C, Denmark. 11Sorbonne Universités (UPMC, Univ. Paris 06)-CNRS-IRD-MNHN, LOCEAN Laboratory, 4 Place Jussieu, F-75005
Paris, France. 12ARC Centre of Excellence for Climate System Science, University of New South Wales, Sydney, New South Wales 2052, Australia. 13Climate
Change Research Centre, University of New South Wales, Sydney, New South Wales 2052, Australia. 14State Key Laboratory of Marine Environmental
Science, Xiamen University, Xiamen 361102, China. 15Institute for Geophysics, Jackson School of Geosciences, University of Texas at Austin, J. J. Pickle
Research Campus, Building 196, 10100 Burnet Road (R2200), Austin, Texas 78758-4445, USA. 16Department of Geology, Lund University, Sölvegatan 12,
SE-223 62 Lund, Sweden. 17Department of Geography, NorthumbriaUniversity, Newcastle upon Tyne NE1 8ST, UK. *e-mail: mcgregor@uow.edu.au
PROGRESS ARTICLE
PUBLISHED ONLINE: 17 AUGUST 2015|DOI: 10.1038/NGEO2510
© 2015 Macmillan Publishers Limited. All rights reserved
2 NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience
interval — there is instead a statistically signicant warming trend
of +0.08s.d. units (100 yr)−1 (see Supplementary Table S13 for full
test statistics). ere is no obvious global scale Medieval Climate
Anomaly2,17, although alternate choices for the centre of each bin
might allow for better denition of the Medieval Climate Anomaly
interval (Supplementary Fig. S5).
Bin-to-bin cooling is especially pronounced for the transition
into the 1201–1400 and 1401–1600 bins (–0.17and –0.18s.d.
units (100 yr)–1 , respectively; Supplementary Table S13), and the
overall coldest 200-yr bins are 1401–1600 and 1601–1800
(–0.70 and –0.71s.d. units, respectively; Supplementary Table S13).
ese coldest bin transitions and individual bins are contemporane-
ous with the onset of the globally coherent Little Ice Age recorded in
many Northern and Southern Hemisphere continental regions7,17,18,
suggesting that there was a global mean ocean SST ngerprint
associated with this interval.
e 1801–2000 bin, which includes the industrial era, has the
widest range of values compared with all other bins (Fig.2a). We
further investigate this result using a subset of 21SST reconstruc-
tions with high-precision ages based on 210Pb dating, layer counting,
or coral band counting (Supplementary Table S1; Supplementary
Fig. S10; Supplementary Section 8), which have approximately
decadal or better sampling resolution and that span the nineteenth
and twentieth centuries. Although assessment of signicance is
limited by the number and resolution of the reconstructions, and
by the small amount of overlap with historical SST estimates, we
nd that the composite of reconstructions from tropical regions are
in qualitative agreement with historical SST warming at the same
locations19 (Supplementary Fig. S10). Upwelling processes recorded
at a number of the sites may also inuence the twentieth-century
composite (Supplementary Sections 1and 8). Further conrmation
and analysis of results for the nineteenth and twentieth centuries
requires extended historical reanalyses9,20, higher spatiotemporal
resolution SST reconstructions from other marine palaeoclimate
marine archives21–23 and climate model simulations10.
regional variability12,13, and: (1) minimizes bias owing to any
individual reconstruction overprinting the average trend14–16;
(2) maximizes our chances of extracting a global signal; and
(3)allows us to compare results across regions and climate zones,
and against terrestrial and model composites. We also scaled
the Ocean2k SST synthesis standardized values back to Celsius
temperature units, and found consistent results (Supplementary
Section 4).
• We test the possibility that the Ocean2k SST synthesis includes
a spatial bias owing to the network’s sparse and hetero-
geneous global distribution. We use six Paleoclimate Modelling
Intercomparison Project Phase III (PMIP3)-compliant climate
model simulations available for the past millennium4 (Methods;
Supplementary Table S4; Supplementary Section 5) and calculate
the median correlation eld of each model grid point with its
model global mean SST for the interval 801–1800. High cor-
relations are found for most locations across the globe (Fig.1),
which suggests that the Ocean2k SST network, on bicentennial
timescales, should contain a common global signal.
• e 57 Ocean2k SST reconstructions are skewed towards ocean
basin margins (Fig. 1), where sedimentation rates are su-
ciently high to provide centennial scale resolution, which could
mean that these locations reect terrestrial rather than marine
climate. However, at 200-yr resolution, model-simulated true
global ocean surface temperatures are similar to simulated area-
weighted or unweighted composites based on our 57 reconstruc-
tion locations (Supplementary Fig. S7). As such, we interpret
the Ocean2k SST data synthesis as representative of global
SSTvariations8.
Robust long-term cooling trend
e global Ocean2k SST synthesis shows a statistically signicant
median cooling trend of −0.65 s.d. units kyr−1 for the past 2kyr, and
the cooling trend is steepest for the 1000–1800 interval (Fig.2a).
A SST synthesis weighted by ocean basin area gives similar results
(Fig. 2a; Supplementary Section 5). However, the marine climate
reconstructions that underpin our synthesis have potential biases
related to reconstruction type, signal seasonality, sample resolu-
tion, age constraints, location, age model and specic proxy-related
issues (Supplementary Table S1; Supplementary Section 6); the
Ocean2k SST synthesis could reect a dominance of some of these
biases rather than a global signal.
We investigate the Ocean2k synthesis data set for these sources
of bias via a series of sensitivity tests (Fig. 2b). Here, we divide
the Ocean2k reconstructions into subpopulations (Methods), and
assess their long-term trends relative to the Ocean2k SST synthe-
sis. e sensitivity tests indicate that the Ocean2k global SST cool-
ing trend is not sensitive to the number of dated levels down-core,
sedimentation rate (growth rate for the coral reconstruction),
SST reconstruction type (for example, alkenone), recording sea-
son for the reconstruction (seasonality), sampling resolution, or
water depth at the coring site (as a measure of coastal proximity;
Supplementary Section 6; Fig.2b). Furthermore, the cooling trend
is observed in most ocean basins, across hemispheres, in tropical
and extratropical latitudes, and in localities characterized as either
upwelling or non-upwelling regions (Fig. 2b). We conclude that
the cooling trend observed in the Ocean2k SST synthesis is robust
despite these uncertainties.
Centennial scale SST variability
e rate of SST cooling recorded by the Ocean2k SST synthesis for
the past 2kyr is variable, as observed in the rate of change between
each 200-yr bin (Fig. 2a; Supplementary Section 7). Analysis of
these SST changes indicates a statistically signicant cooling in all
bin-to-bin transitions for the 1001–1800 period. In contrast, for
the transition to the 1801–2000 bin — the most recent 200-yr
Mean = 0.81
Median = 0.91
1.0
0.0 0.5
Median correlation coecient
PMIP3 simulations:
bcc-csm1-1
CCSM4
CSIRO Mk3L
FGOALS-s2
MPI-ESM
LOVECLIM_PMIP3
2
3
2
2
2
2
2
2
Figure 1 | Correlation map and locations of the 57 reconstructions
in the Ocean2k SST synthesis. Map is based on six climate model
simulations for the interval 801–1800 (Supplementary Table S4). For
each simulation, SST data were composited into 200-yr bins, and the
correlation between grid point SST and model global average SST was
calculated (Supplementary Section 5). The map shows the six-model
median correlation coecient (colours). Numbers indicate where multiple
SST reconstructions are from nearby locations. The correlation map
qualitatively suggests that mean SST at the 57 locations is representative
of global mean SST on centennial timescales.
PROGRESS ARTICLE NATURE GEOSCIENCE DOI: 10.1038/NGEO2510
© 2015 Macmillan Publishers Limited. All rights reserved
NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience 3
Comparisons with simulations and terrestrial data
e long-term cooling trend captured by the Ocean2k SST synthesis
seems to be consistent with a similar cooling trend observed in global
temperature syntheses that include both marine and terrestrial input
reconstructions2,5,24. To compare global surface ocean and global
land temperature trends, we calculate a terrestrial composite of non-
marine reconstructions7,25 (Terrestrial 2k composite; Supplementary
Section 9), ensuring that independent data sets are used to determine
marine and terrestrial signals. e Terrestrial 2k composite shows
a cooling trend qualitatively similar to the Ocean2k SST synthesis
over the 801–1800 interval, the interval with the highest avail-
ability of SST reconstruction data and model simulations (Fig. 3;
Supplementary Table S14). e probability that either the Ocean2k
synthesis or the Terrestrial 2k composite indicates a cooling trend
−3
−2
−1
0
1
2
3
Standardized anomaly
(s.d. units)
100 300 500 700 900 1100 1300 1500 1700
Median 1σ
1900
1.0
0.5
0.0
Median 1σ
1.0
0.5
0.0
Median 1σ
1.0
0.5
0.0
Median 1σ
1.0
0.5
0.0
Median 1σ
1.0
0.5
0.0
Median 1σ
1.0
0.5
0.0
Median 1σ
1.0
0.5
0.0
Median 1σ
1.0
0.5
0.0
Median 1σ
1.0
0.5
0.0
Median 1σ
1.0
0.5
0.0
–1.5
–1
–0.5
0
0.5
1
1.5
Alkenones
Mg/Ca
Others
Mean SST
Warm SST
Cold SST
NH
SH
Depth >500 m
Depth <500 m
>4 14C dates
4 14C dates
100 300 500 700 900 1100 1300 1500 1700 1900
57 27 24 6
57 43 11
3
57 229 16
244
57 2924 4
–1.5
–1
–0.5
0
0.5
1
1.5
57 46 11
57 28 29
57 43 14
57 39 18
55 27 26
a
b
Year (CE)
Year (CE)
Upwelling
Non-upwelling
57 750
Standardized anomaly (s.d. units)
–1.5
–1
–0.5
0
0.5
1
1.5
–1.5
–1
–0.5
0
0.5
1
1.5
–1.5
–1
–0.5
0
0.5
1
1.5
–1.5
–1
–0.5
0
0.5
1
1.5
8 SST pts per bin
<8 SST pts per bin
0.1 cm yr–1
<0.1 cm yr–1
–1.5
–1
–0.5
0
0.5
1
1.5
–1.5
–1
–0.5
0
0.5
1
1.5
Extratropical NH
Tropics
Extratropical SH
–1.5
–1
–0.5
0
0.5
1
1.5
Arctic
Atlantic Indian
Mediterranean
Pacific
Southern
–1.5
–1
–0.5
0
0.5
1
1.5
Proxy type
Response seasonality
Sampling resolution
Chronological control
Sedimentation rate
Water depth
Latitudinal band
Ocean basin
Hemisphere
Upwelling
Figure 2 | Ocean2k global SST composite and sensitivity analyses. a, Standardized Ocean2k SST synthesis. Lines are the N=57 reconstructions,
coloured by ocean basin (see b). Box plots show 25th to 75th percentile range (black box), median (black horizontal lines) and outliers (red crosses) to
approximately 99.3% of the data range (black dashed lines and caps (‘whiskers’)) assuming normally distributed bin contents (Methods). Thick black line
is the median of the ocean basin area-weighted composite (Supplement section 5). b, Standardized 57 Ocean2k reconstructions resampled into various
categories (medians plotted; Methods). Bar graphs give error estimates and number of reconstructions per subcategory (Methods).
PROGRESS ARTICLE
NATURE GEOSCIENCE DOI: 10.1038/NGEO2510
© 2015 Macmillan Publishers Limited. All rights reserved
4 NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience
for the 801–1800 interval is 85% and 73%, respectively (Methods;
Supplementary Table S14). Although some of this agreement pre-
sumably arises from the proximity of the marine margins to land,
the consistency of the Ocean2k SST synthesis and the Terrestrial 2k
composite trends is probably due to a relatively coherent response to
a common forcing on bicentennialtimescales.
We also composite SST from six PMIP3-compliant simulations
driven with realistic natural and anthropogenic forcings, matched to
the same locations, time intervals and seasonality as the 57 Ocean2k
reconstructions (multi-model composite; Fig. 3; Methods). e
multi-model composite is also qualitatively consistent with the
Ocean2k SST synthesis and has a higher (92%) probability of a
cooling trend, albeit dened by a pronounced comparatively cold
1200–1400 bin. e qualitatively similar cooling trends in the
Ocean2k SST synthesis and the multi-model composite are not
an artefact of the standardization method (see Supplementary
Section4) and suggest that we may analyse the simulations to infer
the mechanisms most likely to explain the palaeoclimate data.
External forcing of the global SST cooling
We isolate the forcing, or combination of forcings, most qualita-
tively consistent with the Ocean2k SST cooling trend observed for
the past thousand years (the interval of common overlap) using
two models from our multi-model composite (Fig. 4): SST sim-
ulated by the Commonwealth Scientic and Industrial Research
Organisation (CSIRO) Mk3L model26, run with the cumulative
addition of orbital (O), greenhouse gas (G), solar (S) and volcanic
(V) forcings; and by the LOVECLIM model27, run with individ-
ual forcings as for CSIRO Mk3L, plus land-use forcing (L) and
with all forcings (All). ese simulations were matched to the
57Ocean2k reconstructions (Methods). e CSIRO Mk3L simu-
lations26 suggest that OGS forcings combined give rise to only a
weak and non-signicant cooling trend, and are insucient to
explain the long-term Ocean2k global SST cooling trend (Fig.4a;
Supplementary Table S14). e LOVECLIM simulations27, run
individually with O, G, or S forcings, consistently show that these
forcings do not explain the Ocean2k SST cooling trend (Fig.4b;
Supplementary Table S14).
e modest eect of greenhouse gases in generating a global
ocean surface cooling trend in the model simulations is likely because
greenhouse gas forcing is small prior to 1800 (Supplementary
Fig. S4). Similarly, the magnitude of solar forcing, although uncer-
tain6,28,29, is small and does not generate a signicant long-term cool-
ing trend in the model simulations. Our analysis, however, cannot
rule out regional and global mechanisms in which solar activity
forces climate change on decadal to centennial timescales30,31.
e single and cumulatively forced model simulations suggest
that orbital forcing has only a minor role in generating a global
SST cooling trend for the 801–1800 interval. Orbital forcing is
associated with changes in insolation that are strongly dependent
on the season and latitude32, and over the Pleistocene epoch, orbital
changes forced global climate through amplication mechanisms at
high northern latitudes, including the well-known ice-albedo ampli-
cation33. High northern latitude temperature trends during the
past millennium27,34,35 have also been attributed to orbital forcing,
specically to declining high northern latitude summer insolation,
amplied by feedbacks in the Arctic region and resulting in cool-
ing34,35. However, when integrated over the full calendar year and
spatially across the globe32, the 1–2000 change in orbital radia-
tive forcing at the top of the atmosphere is only +4.4×10−3Wm−2
(ref.34). Consequently, the CSIRO Mk3L and LOVECLIM models
both give weak and non-signicant global ocean SST trends for the
orbital forcing simulations, because the global ocean integrates the
average global orbital forcing.
We nd that volcanic forcing in CSIRO Mk3L, and volcanic and
land-use forcings in LOVECLIM, produce a cooling trend most
consistent with the Ocean2k SST synthesis (Fig. 4). e role of
land-use change in forcing the Ocean2k SST cooling trend in the
LOVECLIM simulation (Fig.4b) arises from the increase in sur-
face albedo owing to deforestation, inducing a net negative radiative
forcing on land. e associated cooling is only partly compensated
for by the reduced latent heat ux from lower summer evapotran-
spiration, resulting in overall cooling on land that is transmitted
globally by the atmospheric circulation27,36,37. However, there are
large uncertainties in our understanding of land-use forcing back in
time, and with the simulated eects of land-use change on radiative
forcing and the hydrological cycle36,38,39.
Volcanic forcing
e inuence of volcanic forcing on driving the SST cooling trend
for the past millennium is surprising, as this forcing, although
large, is relatively episodic40,41. e dominance of volcanic forcing
−3
−2
−1
0
1
2
3
Standardized anomaly
(s.d. units)
a
−3
−2
−1
0
1
2
3
b
−3
−2
−1
0
1
2
3
900 1100 1300 1500 1700
c
Year (CE)
Ocean2k SST synthesis
Terrestrial 2k composite
Multi-model composite
Standardized anomaly
(s.d. units)
Standardized anomaly
(s.d. units)
Figure 3 | Global SST and temperature trends for 801–1800. a, Ocean2k
SST synthesis (Fig. 2a). b, Terrestrial 2k composite (Supplementary
Section 9). c, Multi-model composite, at the locations and periods for
reconstructions in a (Methods; Supplementary Table S4). Box plots show
25th to 75th percentile range (blue boxes), median (red lines) and outliers
(red crosses) to approximately 99.3% of the data range (blue ‘whiskers’)
assuming normally distributed bin contents (Supplementary Table S14).
Distribution (grey lines) and median (black line) of 10,000 Monte Carlo
trend estimates are displayed for each synthesis. Trends are qualitatively
similar between data and model composites.
PROGRESS ARTICLE NATURE GEOSCIENCE DOI: 10.1038/NGEO2510
© 2015 Macmillan Publishers Limited. All rights reserved
NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience 5
over the 801–1800 interval may arise from increased volcanic
event frequency and the occurrence of very large events early in the
past millennium, and/or internal amplication of volcanic forcing
within the climate system6,42–44 (Supplementary Fig. S4). In par-
ticular, large volcanic eruptions between 1150 and 1300 , and
again during the early eenth century, may be responsible for the
observedcooling43,45.
We use an energy balance model46 (EBM; Supplementary
Section 10) to simulate the thermodynamic response of a mixed-
layer ocean to the volcanic radiative forcing (Fig. 4). e EBM
results do not fully match the bin-to-bin changes in the Ocean2k
SST synthesis, probably because internal variability is absent in the
EBM, there are contributions from other forcings and the EBM has
a weak long-term memory (it contains no deep-ocean coupling).
However, we nd that the radiative change from repeated clustering
of volcanic eruptions over the past millennium by itself is sucient
to explain a long-term cooling trend. is suggests that the Ocean2k
SST cooling trend, at least for >200-yr timescales, might simply rep-
resent a rst-order thermodynamic response to stochastic volcanic
forcing that increased in frequency from the early to late past mil-
lennium. A comparison of Northern Hemisphere reconstructed
temperatures with a volcanically forced upwelling–diusion EBM
reached a similar conclusion47.
Nonetheless, several dynamical mechanisms have been proposed
to explain an amplied ocean response to the volcanic forcing during
the past millennium43,48,49. Specically, on decadal timescales the vol-
canic eruptions could have induced a fast cooling in the tropics, lead-
ing to anomalously high pressure over the continents that reduced
Atlantic Ocean trade wind stress curl anomalies49. e subsequent
ocean adjustment propagated the cooling to the high latitudes and,
over time, weakened the Atlantic meridional overturning circulation
(AMOC), resulting in sea ice expansion. On centennial timescales,
eruptions are thought to have reduced downwelling shortwave radi-
ation, increased surface albedo and led to lower elevation snowlines
in regions north of 60°N (refs 43,48). e associated cooling and
increase in sea ice extent further amplied Arctic and North Atlantic
cooling43,48. Open ocean convection may have declined, weaken-
ing the AMOC and its associated ocean heat transport to the high-
latitude North Atlantic48. In turn, the reduced AMOC may have
reduced sea ice melt, permitting sea ice to persist for a century, thus
perpetuating the initial cooling induced by the frequent eruptions
in the late thirteenth century43,48. ese model results suggest that
dynamical mechanisms might transform increases in frequency of
volcanic eruptions into a longer term cooling.
We test for changes in AMOC both in the volcanic-only forc-
ing LOVECLIM simulation and in the orbital–greenhouse
P(m < 0) (%)
O OG OGS OGSV O2k
Forcing
–2
0
50
100
Slope
(s.d. units kyr–1)
a
O G S V LAll O2k
Forcing
b
Negative
slope
Negative
slope
likelihood
–0.4
0.0
200018001600140012001000
800
Year (
CE
)
–10
–5
0
Volcanic forcing
(W m
–2
)
Temperature anomaly (°C)
c
d
Volcanic forcing and EBM
CSIRO Mk3L LOVECLIM
Figure 4 | Common Era Ocean2k SST synthesis and model simulation slopes, volcanic forcing and EBM response. a, CSIRO Mk3L simulations (blue).
b, LOVECLIM simulations (yellow). Monte Carlo slopes (circles) and negative slope probabilities (bars; P=50%: equal probability of positive or negative
slope; Methods). Ocean2k SST synthesis is also shown (O2k; red). Slope ± 2 s.e.m. values are smaller than the symbols. Time series in Supplementary
Fig. S11. c,d, Volcanic forcing6 (c) and EBM (d )temperature response (aqua line). 200-yr bin averages (blue lines) and linear fit (straight line) are shown.
Volcanic forcing contributes to the cooling trend via a thermodynamic response.
PROGRESS ARTICLE
NATURE GEOSCIENCE DOI: 10.1038/NGEO2510
© 2015 Macmillan Publishers Limited. All rights reserved
6 NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience
gas–solar–volcanic forcing CSIRO Mk3L simulation (Supplementary
Section 10). We nd no persistent weakening of the AMOC in
response to the volcanic forcing for the past millennium in either sim-
ulation (Supplementary Figs S12 and S13). Furthermore, although
Northern Hemisphere sea ice expands following the late thirteenth-
century volcanic cluster, the increase is transient and probably repre-
sents a short-term thermodynamic response rather than a dynamical
feedback. Although these results may be model dependent, they are
consistent with analysis of a wider range of models50. Taken together,
our analysis of EBM and coupled model simulations suggests that on
timescales >200yr, ocean dynamics may not be required to translate
volcanic forcing into a long-term cooling trend.
e Ocean2k SST synthesis, built on rigorous quality control of
well-dated SST reconstructions, denes a statistically signicant
global SST cooling trend for the pre-industrial Common Era. State-
of-the-art climate model simulations using realistic forcing show a
qualitatively similar cooling trend. Furthermore, a suite of transient
model simulations using single and cumulative forcings suggests
that the cooling trend does not arise from orbital or solar forcings,
but from the increased frequency and magnitude of explosive vol-
canism, with land-use forcing also a factor. e results point to a
dominant role of volcanic forcing in driving the global SST cooling
trend for the pre-industrial Common Era, with the ocean’s thermal
response integrating the forcing on these longer timescales. Further
improved observational coverage, estimates of radiative forcings and
comparisons with simulations will permit more ne-grained assess-
ments of the mechanisms underlying observed climate change—
both past and future.
Methods
Methods and any associated references are available in the online
version of the paper.
Received 24 October 2014; accepted 17 July 2015;
published online 17 August 2015
References
1. Moberg, A., Sonechkin, D.M., Holmgren, K., Datsenko, N.M. & Karlen, W.
Highly variable Northern Hemisphere temperatures reconstructed from low-
and high-resolution proxy data. Nature 433, 613–617 (2005).
2. Mann, M.E. et al. Global signatures and dynamical origins of the Little Ice Age
and Medieval Climate Anomaly. Science 326, 1256–1260 (2009).
3. Jones, P.D. etal. High-resolution palaeoclimatology of the last millennium:
A review of current status and future prospects. Holocene 19, 3–49 (2009).
4. Braconnot, P. et al. Evaluation of climate models using palaeoclimatic data.
Nature Clim. Change 2, 417–424 (2012).
5. Marcott, S.A., Shakun, J.D., Clark, P.U. & Mix, A.C. A reconstruction
of regional and global temperature for the past 11,300years. Science
339, 1198–1201 (2013).
6. Schmidt, G.A. et al. Climate forcing reconstructions for use in
PMIP simulations of the last millennium (v1.1). Geosci. Model Dev.
5, 185–191 (2012).
7. PAGES 2k Consortium Continental-scale temperature variability during the
past two millennia. Nature Geosci. 6, 339–346 (2013).
8. Gill, A.E. Atmosphere–Ocean Dynamics Vol. 30 (International Geophysics
Series, Academic, 1982).
9. Balmaseda, M.A., Trenberth, K.E. & Källén, E. Distinctive climate
signals in reanalysis of global ocean heat content. Geophys. Res. Lett.
40, 1754–1759 (2013).
10. Rhein, M. et al. in Climate Change 2013: e Physical Science Basis
(edsStocker,T.F. etal.) 255–315 (IPCC, Cambridge Univ. Press, 2013).
11. Kennedy, J.J. A review of uncertainty in insitu measurements and data sets of
sea surface temperature. Rev. Geophys. 52, 1–32 (2014).
12. Lee, T.C. K., Zwiers, F.W. & Tsao, M. Evaluation of proxy-based millennial
reconstruction methods. Clim. Dynam. 31, 263–281 (2008).
13. Mann, M.E. et al. Proxy-based reconstructions of hemispheric and
global surface temperature variations over the past two millennia.
Proc. Natl Acad. Sci. USA 105, 13252–13257 (2008).
14. Evans, M.N., Tolwinski-Ward, S.E., ompson, D.M. & Anchukaitis, K.J.
Applications of proxy system modeling in high resolution paleoclimatology.
Quat. Sci. Rev. 76, 16–28 (2013).
15. Lohmann, G., Pfeier, M., Laepple, T., Leduc, G. & Kim, J.H. A model–data
comparison of the Holocene global sea surface temperature evolution.
Clim.Past 9, 1807–1839 (2013).
16. Laepple, T. & Huybers, P. Global and regional variability in marine surface
temperatures. Geophys. Res. Lett. 41, 2528–2534 (2014).
17. Lamb, H.H. e early medieval warm epoch and its sequel.
Palaeogeogr.Palaeoclimatol. Palaeoecol. 1, 13–37 (1965).
18. Neukom, R. et al. Inter-hemispheric temperature variability over the past
millennium. Nature Clim. Change 4, 362–367 (2014).
19. Kaplan, A. et al. Analyses of global sea surface temperature 1856–1991.
J.Geophys. Res.-Oceans 103, 18567–18589 (1998).
20. Allan, R. et al. e international Atmospheric Circulation
Reconstructions over the Earth (ACRE) initiative. Bull. Am. Meteorol. Soc.
92, 1421–1425 (2011).
21. Emile-Geay, J., Cobb, K.M., Mann, M.E. & Wittenberg, A.T. Estimating
central equatorial Pacic SST variability over the past millennium. Part I:
Methodology and validation. J.Climate 26, 2302–2328 (2013).
22. Tierney, J.E. et al. Tropical sea-surface temperatures for the past four centuries
reconstructed from coral archives. Paleoceanography 30, 226–252 (2015).
23. Evans, M.N., Kaplan, A. & Cane, M.A. Pacic sea surface temperature eld
reconstruction from coral δ18O data using reduced space objective analysis.
Paleoceanography 17, 1007 (2002).
24. Masson-Delmotte, V. et al. in Climate Change 2013: e Physical Science Basis
(eds Stocker, T.F. etal.) 409–415(IPCC, Cambridge Univ. Press, 2013).
25. McKay, N.P. & Kaufman, D.S. An extended Arctic proxy temperature database
for the past 2,000years. Sci. Data 1, 140026 (2014).
26. Phipps, S.J. et al. Paleoclimate data–model comparison and the role of climate
forcings over the past 1500years. J.Climate 26, 6915–6936 (2013).
27. Crespin, E., Goosse, H., Fichefet, T., Mairesse, A. & Sallaz-Damaz, Y. Arctic
climate over the past millennium: Annual and seasonal responses to external
forcings. Holocene 23, 321–329 (2013).
28. Delaygue, G. & Bard, E. An Antarctic view of Beryllium-10and solar activity
for the past millennium. Clim. Dynam. 36, 2201–2218 (2011).
29. Schurer, A.P., Tett, S.F. B. & Hegerl, G.C. Small inuence of solar variability
on climate over the past millennium. Nature Geosci. 7, 104–108 (2014).
30. Bard, E. & Frank, M. Climate change and solar variability: What’s new under
the sun? Earth Planet. Sci. Lett. 248, 1–14 (2006).
31. Rind, D., Lean, J., Lerner, J., Lonergan, P. & Leboissitier, A. Exploring the
stratospheric/tropospheric response to solar forcing. J.Geophys. Res.-Atmos.
113, D24103 (2008).
32. Berger, A.L. Long-term variations of daily insolation and Quaternary climatic
changes. J.Atmos. Sci. 35, 2362–2367 (1978).
33. Hays, J.D., Imbrie, J. & Shackleton, N.J. Variations in the Earth’s orbit:
Pacemaker of the Ice Ages. Science 194, 1121–1132 (1976).
34. Esper, J. et al. Orbital forcing of tree-ring data. Nature Clim. Change
2, 862–866 (2012).
35. Kaufman, D.S. et al. Recent warming reverses long-term Arctic cooling.
Science 325, 1236–1239 (2009).
36. Brovkin, V. et al. Biogeophysical eects of historical land cover changes
simulated by six Earth system models of intermediate complexity.
Clim.Dynam. 26, 587–600 (2006).
37. Goosse, H. et al. e origin of the European “Medieval Warm Period”.
Clim.Past 2, 99–113 (2006).
38. de Noblet-Ducoudré, N. et al. Determining robust impacts of land-use-induced
land cover changes on surface climate over North America and Eurasia: Results
from the rst set of LUCID experiments. J.Climate 25, 3261–3281 (2012).
39. He, F. et al. Simulating global and local surface temperature changes
due to Holocene anthropogenic land cover change. Geophys. Res. Lett.
41, 623–631 (2014).
40. Hansen, J., Lacis, A., Ruedy, R. & Sato, M. Potential climate impact of Mount
Pinatubo eruption. Geophys. Res. Lett. 19, 215–218 (1992).
41. Robock, A. Volcanic eruptions and climate. Rev. Geophys. 38, 191–219 (2000).
42. Stenchikov, G. et al. Volcanic signals in oceans. J.Geophys. Res.-Atmos.
114,D16104 (2009).
43. Miller, G.H. et al. Abrupt onset of the Little Ice Age triggered by
volcanism and sustained by sea-ice/ocean feedbacks. Geophys. Res. Lett.
39, L02708 (2012).
44. Plummer, C.T. et al. An independently dated 2000-yr volcanic record from
Law Dome, East Antarctica, including a new perspective on the dating of the
1450s eruption of Kuwae, Vanuatu. Clim. Past 8, 1929–1940 (2012).
45. Sigl, M. et al. Insights from Antarctica on volcanic forcing during the Common
Era. Nature Clim. Change 4, 693–697 (2014).
46. Goosse, H., Barriat, P.Y., Lefebvre, W., Loutre, M.F. & Zunz, V. Introduction
to Climate Dynamics and Climate Modeling (Université catholique de Louvain,
2015); http://www.climate.be/textbook
47. Crowley, T.J. Causes of climate change over the past 1000years. Science
289,270–277 (2000).
PROGRESS ARTICLE NATURE GEOSCIENCE DOI: 10.1038/NGEO2510
© 2015 Macmillan Publishers Limited. All rights reserved
NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience 7
48. Zhong, Y. et al. Centennial-scale climate change from decadally-paced
explosive volcanism: A coupled sea ice-ocean mechanism. Clim. Dynam.
37,2373–2387 (2011).
49. Mignot, J., Khodri, M., Frankignoul, C. & Servonnat, J. Volcanic impact on the
Atlantic Ocean over the last millennium. Clim. Past 7, 1439–1455 (2011).
50. Ding, Y. et al. Ocean response to volcanic eruptions in Coupled
Model Intercomparison Project 5 simulations. J.Geophys. Res.-Oceans
119, 5622–5637 (2014).
Acknowledgements
We thank the many scientists who made their published data sets available via public
data repositories. T. Kiefer, L. von Gunten and C. Telepski from the IGBP PAGES-IPO
provided organizational and logistical support. V. Masson-Delmotte, C. Giry, S.P. Bryan,
S. Stevenson, D. Colombaroli, B. Horton, J. Tierney and the Ocean2k HR Working Group
are thanked for early input to the project design and methodology. G. Lohmann assisted
with model output. A. Mairesse assisted with model gures. L. Skinner and D.Reynolds
are thanked for discussions on age models. We are grateful to the 75 volunteers who
constructed the Ocean2k metadatabase (see Supplementary Information for full list of
names). We acknowledge the World Climate Research Programme’s Working Group on
Coupled Modelling, which is responsible for CMIP, and we thank the climate modelling
groups (listed in Supplementary Table S4) for producing and making available their
model output. For CMIP, the US Department of Energy’s Program for Climate Model
Diagnosis and Intercomparison provides coordinating support and led development of
soware infrastructure in partnership with the Global Organization for Earth System
Science Portals. We acknowledge support from PAGES, a core project of IGBP nancially
supported by the US and Swiss National Science Foundations (NSFs) and NOAA;
Australian Research Council (ARC) Discovery Project grant DP1092945 (H.V.M., S.J.P.),
ARC Future Fellowship FT140100286 grant (H.V.M.), AINSE Fellowship grant (H.V.M.)
and the research contributes to ARC Australian Laureate Fellowship FL120100050;
US NSF awards NSF/ATM09-02794 (M.N.E.) and NSF/ATM0902715 (M.N.E), and
Royal Society of New Zealand Marsden Fund grant 11-UOA-027 (M.N.E.); F.R.S-FNRS
(Belgium; H.G.); French National Research Agency (ANR) under ISOBIOCLIM grant
(G.L.); European Union’s Seventh Framework Programme (FP7/2007-2013) under
grant agreement number 243908, Past4Future ‘Climate change — Learning from the
past climate’ contribution no. 81 (H.G., B.M., P.G.M., M.-S.S.); CSIC-Ramón y Cajal
post-doctoral programme RYC-2013-14073 (B.M.), Clare Hall College, Cambridge,
Shackleton Fellowship (B.M.) and Red CONSOLIDER GRACCIE CTM2014-59111-
REDC (B.M.); US Geological Survey’s Climate and Land Use Change Research
andDevelopment Program and the Volcano Science Center (J.A.A.); Ralph E. Hall
Endowed Award for Innovative Research (D.W.O.); Danish Council for Independent
Research Natural Science OCEANHEAT project 12-126709/FNU (M.-S.S.); LEFE/INSU/
NAIV project (M.-A.S.); NSF of China grant 41273083 (K.S.) and Shanhai Fund grant
2013SH012 (K.S.); UTIG Ewing-Worzel Fellowship (K.T.); Swedish Research Council
grant 621-2011-5090 (H.L.F.); and from a Marie Curie Intra-European Fellowship for
Career Development (V.E.).
Author contributions
M.N.E., H.V.M., D.W.O., H.G., G.L. and B.M. designed the project with input from J.A.A.,
M.-S.S., M.-A.S., K.S. and V.E.; H.V.M. and G.L. led the synthesis. H.V.M. and B.M.
collated and evaluated the reconstructions, and managed the data with assistance from
J.A.A.; M.N.E. led the analysis with important contributions from H.G., J.A.A., B.M., G.L.,
S.J.P., H.V.M., D.W.O., P.G.M., M.-S.S. and M.-A.S.; H.G. and S.J.P. collated, managed and
analysed the model simulations with input from M.N.E., G.L. and H.V.M.; H.V.M. led the
writing with the assistance of M.N.E., H.G., G.L., B.M., J.A.A., P.G.M., D.W.O., M.-S.S.,
M.-A.S., S.J.P., K.S., K.T., H.L.F and V.E.; all authors reviewed the manuscript.
Additional information
Supplementary information is available in the online version of the paper. Reprints
and permissions information is available online at www.nature.com/reprints.
Correspondence should be addressed to H.V.M.
Competing financial interests
e authors declare no competing nancial interests.
PROGRESS ARTICLE
NATURE GEOSCIENCE DOI: 10.1038/NGEO2510
© 2015 Macmillan Publishers Limited. All rights reserved
NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience
Methods
SST data. e Ocean2k SST synthesis is based on 57 peer-reviewed, publicly avail-
able reconstructions solely from marine archives (Fig.1; Supplementary Table S1)
and listed in the Past Global Changes Ocean2k metadatabase (http://pages-igbp.
org/ini/wg/ocean2k/intro). Reconstructions are expected to record SST and have a
chronology anchored by at least two ages, within error, between 200yr before the
() and present. Only data between the oldest and youngest dates were used
(see Supplementary Section 1for additional details). Ages were converted to the
/ timescale and SST calibrations from the original publications were used.
Models. Climate simulations available from 850 were selected from the
BCC-CSM1-1, CCSM4, FGOALS-s2, LOVECLIM and MPI-ESM models, and
available from 801 from the CSIRO Mk3L model (see Supplementary Section
2for references). e model forcings follow the Coupled Model Intercomparison
Project Phase 5 (CMIP5)/PMIP3 protocol6. For this subset of simulations there is
no signicant dri owing to experimental design24. Model output was truncated
at 1800 to focus on the pre-industrial past millennium. For Fig.4, additional
LOVECLIM and CSIRO Mk3L simulations were used, driven by only selected forc-
ings (Supplementary Section 2; Supplementary Table S4; Supplementary Fig. S11).
EBM details are in Supplementary Section 10. e model simulation time series in
Figs3and 4 are from the same location (or, if a margin site, the nearest straight-
line distance grid box), time interval and season as the 57SST reconstructions. See
Supplementary Section 5for details on Fig.1 construction.
Binning and standardization. Data from each of the 57 reconstructions were
averaged into 200-yr bins, to give 10 bins, centred on 100, 300 and so on, up
to 1900. Each binned series was then standardized by its average and standard
deviation. e Ocean2k synthesis trends are insensitive to bin centre placement
(Supplementary Section 5; Supplementary Fig. S5). e mean of the age distri-
bution of individual data points within each bin is very close to the bin centre
(Supplementary TableS5). e binned and standardized values are then treated
as a sample of the SST population within each bin, with age uncertainty approxi-
mately equal to bin width. e numbers of chronological control points per bin
are in Supplementary Fig. S1. Slope and slope probabilities were estimated using
Monte Carlo simulations (below). Our results are insensitive to standardization
(Supplementary Section 4). All simulated SSTs were binned and composited as
described here.
Sensitivity tests. Sensitivity tests were carried out on the Ocean2k synthesis cool-
ing trend to assess the inuence of (Supplementary Table S1): (1) reconstruction
type, for example, alkenone, foraminiferal Mg/Ca, other (microfossil transfer
functions or modern analogue technique, TEX86 and coral Sr/Ca); (2) response sea-
sonality (mean annual, warm, cool); (3) number of 14C dates/reconstruction (>4or
≤4 14C dates; note, four reconstructions were not 14C dated and are not included in
this test (Supplementary Table S1)); (4) sampling resolution (<8, ≥8 samples per
bin); (5) sedimentation rate (≤0.1cmyr–1, >0.1cmyr–1); (6) water depth (>500m,
<500m); (7) basin (Arctic, Atlantic, Indian, Mediterranean, Pacic, Southern);
(8) latitude (extratropical Northern Hemisphere (>30°N), tropical (30°N–30°S),
extratropical Southern Hemisphere (>30°S)); (9) hemisphere (northern or south-
ern); and (10) locations within pre-dened upwelling zones (Supplementary
Table S3). e sub-data sets were binned, standardized and composited as per the
Ocean2k synthesis. Medians are plotted in Fig.2b. We use water depth as a proxy
for open ocean conditions (Supplementary Section 6). Maps of each sensitivity
analysis (Supplementary Fig. S9) were generated to examine the potential spatial
bias associated with each characteristic.
Each bin value (in each sensitivity test) represents a standardized mean of up
to 57 data points per 200-yr bin, and the full 2-kyr data set includes 10 standard
deviations (one value per bin). erefore, to represent the overall error for each
sensitivity analysis, a median standard deviation for the 10 bins was calculated, and
is plotted in the error histograms (Fig.2b, right).
Slope calculations. Slopes and slope probabilities were estimated from Monte
Carlo simulations. Ten thousand linear least-squares ts were estimated; for each
realization, a single observation from each bin of a given synthesis was randomly
selected (with replacement) to create the t estimate. e median slope of these
ts was then calculated. Slopes were similarly estimated for the Ocean2k synthesis
(Figs2–4), Terrestrial 2k and multi-model composites (Fig.3), and single and
cumulative forcing model SST estimates (Fig.4; Supplementary Fig. S11). Slopes
in Fig.4are plotted with +2s.e.m., assuming a normally distributed sample of the
regression slope.
Data availability. Data URLs for the 57 reconstructions are in Supplementary
Table S2. Ocean2k SST synthesis data matrix and metadata are at http://www.
ncdc.noaa.gov/paleo/study/18718. Model simulation URLs are in Supplementary
Table S4. Terrestrial 2k composite data are at https://www.ncdc.noaa.gov/cdo/
f?p=519:2:0::::P1_study_id:12621 and update 1.1.1 at http://dx.doi.org/10.6084/
m9.gshare.1054736.
Code availability. Compositing code used to generate the Ocean2k SST synthesis
(Fig.2a) is at http://www.ncdc.noaa.gov/paleo/study/18718.
PROGRESS ARTICLE NATURE GEOSCIENCE DOI: 10.1038/NGEO2510
© 2015 Macmillan Publishers Limited. All rights reserved