Content uploaded by Hong Yan
Author content
All content in this area was uploaded by Hong Yan on Dec 02, 2017
Content may be subject to copyright.
LETTERS
PUBLISHED ONLINE: 14 AUGUST 2011 | DOI: 10.1038/NGEO1231
A record of the Southern Oscillation Index for the
past 2,000 years from precipitation proxies
Hong Yan1, Liguang Sun1*, Yuhong Wang1,2, Wen Huang3, Shican Qiu4and Chengyun Yang4
The El Niño-Southern Oscillation (ENSO) is a coupled
ocean–atmosphere climate phenomenon in the tropical Pacific
Ocean. The interannual climate variations have been shown
to modify both the Hadley and Walker meridional and zonal
atmospheric circulations, with strong impacts on global
climate1–3. Proxy-based reconstructions of the Southern Os-
cillation Index on a multi-decadal scale have shown that the
strength and frequency of El Niño occurrences have varied
over the past millennium4–7. Here we compile reconstructions
of precipitation8–15 from regions that experience substantial
ENSO variability to extend the multidecadal-scale South-
ern Oscillation Index to include the past 2,000 years. We
find that the Medieval Warm Period (∼AD 800–1300) was
characterized by a negative index, which indicates more El
Niño-dominated conditions, whereas during the Little Ice Age
(∼AD 1400–1850) more La Niña-dominated conditions pre-
vailed. The Southern Oscillation Index we derive is significantly
correlated with reconstructions of solar irradiance and mean
Northern Hemisphere temperature fluctuations.
The Southern Oscillation (SO) is principally a seesaw trend in
atmospheric mass involving coherent exchanges of air between
the eastern and western Pacific (Supplementary Fig. S1; ref. 2).
The traditional Southern Oscillation index (SOI) is defined as
the difference between the sea level pressure (SLP) of antiphase
oscillatory behaviour at Tahiti, in the eastern Pacific, and Darwin, in
the western Pacific16. Owing to the close link between SLP and local
convection and precipitation, the rainfalls in the western Pacific and
the eastern and mid Pacific are closely related to the SOI. For ex-
ample, Trenberth and Caron (2000; ref. 2) examined the statistical
relationship between the SOI and tropical Pacific precipitation and
showed that the SOI is persistently and positively correlated with
the precipitation over the Indo-Pacific warm pool and negatively
correlated with the precipitation over the eastern and mid-tropical
Pacific (Fig. 1, Areas in the equatorial Pacific where precipitation
was significantly (above 95% confidence level) positively correlated
with the SOI are marked in yellow and red (named positive area, or
PA) and areas with significant negative correlation with the SOI are
marked in green and blue (named negative area, or NA); ref. 2).
The influence of the SO on equatorial Pacific precipitation can
be explained by the variation of the Pacific Walker Circulation3.
During an El Niño event, when the SOI is low, the difference of
SLP between the eastern and western Pacific decreases. As a conse-
quence, the Walker Circulation is weakened, and the ascending limb
in the western Pacific switches to the mid-Pacific, leading to aridity
in the western Pacific and humidity in the mid-Pacific. At the same
time, the eastern tropical Pacific experiences a large increase, up to
an order of magnitude, in precipitation owing to the weakened and
transferred descending limb of the Walker Circulation.
1Institute of Polar Environment, School of Earth and Space Science, University of Science and Technology of China, Hefei, Anhui 230026, China, 2Institute
of Oceanology, Chinese Academy of Sciences, Qingdao 266071, China, 3Department of Dairy Science, University of Wisconsin, Madison, Wisconsin
53706, USA, 4School of Earth and Space Science, University of Science and Technology of China, Hefei, Anhui 230026, China. *e-mail: slg@ustc.edu.cn.
,
100 E 180 140 W 60 W
30 N
EQ
30 S
W1
W2 W3 M1 E1 E2
E3
¬0.55 0.55¬0.35 ¬0.15 0.15 0.35
Figure 1 |Locations of hydrological records. Correlations of monthly
anomalies of precipitation (NCEP reanalysis2) with the SOI from January
1979 to December 2010. Areas in equatorial Pacific with positive and
significant correlation (above 95% confidence level) are marked in yellow
and red, and areas with negative and significant correlation were marked in
green and blue. Locations of the rainfall records in the equatorial Pacific
(W1 (refs 8,11), W2 (refs 10,12), W3 (ref. 9), M1 (ref. 15), E1 (ref. 13), E2
(ref. 14) and E3 (ref. 22)) are also indicated. Locations that were
drier/wetter during the Little Ice Age than during the Medieval Warm
Period are marked in red/blue.
Based on such strong and persistent correlation between the SOI
and precipitation, in this study, we proposed a novel and SOI-like
index, SOIpr, as the difference between normalized annual rainfalls
in the PA of tropical western Pacific and the NA of the equatorial
eastern and mid-Pacific (SOIpr =wpZP−wnZN; where ZPand ZN
are the normalized Z-scores of the precipitation in the PA and NA,
respectively. wpand wnare the optimal weights of ZPand ZNand
wp+wn=1. See Methods and Supplementary Fig. S4 for details). To
evaluate the feasibility of using SOIpr as a SOI proxy, we calculated
SOIpr from ad 1951 to 1997 using the instrumental precipitation
data from Galapagos of the eastern Pacific and Indonesia of
the western Pacific (Supplementary Figs S3–S5). We observed a
significant positive correlation between SOIpr and the instrumental
sea level pressure-based SOI (Supplementary Fig. S5, r=0.68,
p<0.0001, neff =40.6 (effective number of independent values; see
Methods for details)), indicating that SOIpr is indeed a good proxy
for the SOI.
Recently, several rainfall reconstructions for the PA of the
western Pacific8–12 and the NA of the eastern13,14 and mid
Pacific15 have been published (Fig. 1). The records from the PA
(refs 8–12) and NA (refs 13–15) contained substantial multi-
decadal variability, characterized by an anti-phase oscillatory
behaviour over the last two millennia (Supplementary Fig. S6),
and allowed us to reconstruct SOIpr for the past two millennia.
Considering these resolution and time span of these records
(Supplementary Table S1 and Fig. S6), in this study we chose the
NATURE GEOSCIENCE |ADVANCE ONLINE PUBLICATION |www.nature.com/naturegeoscience 1
© 2011 Macmillan Publishers Limited. All rights reserved.
LETTERS NATURE GEOSCIENCE DOI: 10.1038/NGEO1231
precipitation records of Indonesia10 and Galapagos13 to calculate
SOIpr for the past two millennia. The Indonesian historic rainfalls
(ad 25–1955) were derived from a salinity reconstruction based on
planktonic-foraminifera δ18O and the Mg/Ca ratio (ref. 10). The
δ18O of foraminiferal calcite (δ18 Oc) reflects the combined effects of
calcification temperature (T) and seawater oxygen isotope (δ18Ow);
the latter varies as a function of salinity. The Mg/Ca ratio in the
shells is primarily temperature dependent. By measuring both δ18Oc
and the Mg/Ca ratio on the same samples, the salinity component in
the δ18O signal can be extracted10 . The salinity component indicates
the precipitation-caused freshening trend of the surface water in the
Indo-Pacific Warm Pool and is thus an indicator of precipitation.
The rainfall history of the Galapagos (7241 bc–ad 2004) is derived
from a lake level reconstruction which is based on the grain size
data from the Lago El Junco sediment core13. As suggested in the
hydro-climatic model simulations, El Junco lake sediment grain size
responds sensitively to the precipitation changes associated with the
Pacific Walker Circulation (PWC) and El Niño events, increasing
during wet El Niño events (weak PWC) and decreasing during the
intervening dry periods13.
The reconstructed precipitation in the Indo-Pacific has an
average resolution of 10 years, and that in Galapagos of 11 years.
We adjusted the two records to one year resolution using linear
interpolation, normalized them, and calculated SOIpr for the time
period from ad 50 to ad 1955 (Fig. 2 and Supplementary Fig. S7).
The calculated SOIpr for the past 100 years reveals a multi-
decadal variability similar to that seen in the smoothed instrumental
SOI time series (Supplementary Fig. S8). Smoothed over 11 years
(the averaged resolution of the used palaeo-precipitation record is
11 years), SOIpr and SOI from ad 1867 to ad 1955 have a significant
correlation (Supplementary Fig. S8, r=0.69, p<0.01, neff =14.6).
This result is consistent with our analysis above using instrumental
rainfall records, confirming that SOIpr is a reliable proxy for the
SOI. We also compared our SOIpr with some high resolution ENSO
reconstructions over the past 350 years4,6,7 and found that the
multi-decadal variations of SOIpr and these ENSO reconstructions
are consistent, but the long term trends are not. For example, the
boreal cold-season Niño-3 index, mainly derived from temperature
records, showed an obvious trend towards a La Niña-like state over
the past 350 years7, whereas our SOIpr, derived from hydrological
records, indicated an opposite trend toward a El Niño-like state.
Two other ENSO reconstructions4,6, based on multiple proxies
including temperature records and hydrological records, suggested
no obvious trend over the past 350 years.
The reconstructed SOIpr has five distinct centennial-scale phases
over the past 2,000 years (Fig. 2): persistent negative values during
ad 50–500, ad 1000–1400 and ad 1850–1955, interrupted by
positive values during ad 500–1000 and ad 1400–1850. For the past
millennium, the SOIpr shows three phases associated with the solar
irradiance and background climate state (Fig. 3). Negative SOIpr
values (indicating a weak PWC) are associated with higher solar
irradiance and global mean temperature during the Medieval Warm
Period (MWP) and the so-called Modern Warm Period. On the
other hand, positive SOIpr values (indicating an enhanced PWC)
are concurrent with lower solar irradiance and cooler global mean
temperature during the Little Ice Age (LIA).
The positive SOIpr and enhanced PWC during the relatively
cool LIA period (Fig. 2) suggest a more La Niña-like mean state
than that during the MWP, contradicting current mainstream
theory, mainly from sea surface temperature (SST) reconstruction,
about the tropical Pacific ENSO variability or mean state changes
over the past millennium (Supplementary Fig. S10). Coral-based
SST reconstruction from Palmary Island in the central equatorial
Pacific17 implied a more El Niño-like mean state during the
LIA than that in MWP (Supplementary Fig. S10). The SST
reconstructions in the western10,12, eastern18,19 and mid17 tropical
AD
Galapagos
Indo–Pacific
RWP DACP MWP LIA
SOIpr
Z–scores
Z–scores
Z–scores
0 200 400 600 800 1000 1200 1400 1600 1800 2000
¬3
¬2
¬1
0
1
2
3
¬2
¬1
0
1
2
¬2
¬1
0
1
2
3
Figure 2 |SOIpr reconstruction. SOI proxy SOIpr as the difference between
the reconstructed precipitation records from Indonesia, in the western
Pacific (top; ref. 10), and the Galapagos, in the eastern Pacific (middle)13.
All records were normalized to a standard Z-score before taking the
difference (see Methods for detail). High precipitation and positive SOIpr
are shown in red and low precipitation and negative SOIpr are shown in
blue. Time periods: RWP–Roman Warm Period (AD 50–400),
MWP–Medieval Warm Period (AD 1000–1300), DACP–Dark Ages Cold
Period (AD 500–900), LIA–Little Ice Age (A D 1400–1850).
Pacific suggested20,21 an LIA marked by a relatively warming period
in the eastern and central tropical Pacific, a substantial cooling in the
western equatorial ocean, and a more El Niño-like state (than that
of the MWP; Supplementary Fig. S10). In contrast with SST-based
reconstructions, most of the hydrological reconstructions from
the tropical Pacific suggested a more La Niña-like mean state
during the LIA than during the MWP (Fig. 1 and Supplementary
Fig. S10). These include records from the Indo-Pacific8–12, central
tropical Pacific15 and eastern equatorial Pacific13,14 that suggested
wetter, drier and drier conditions, respectively, during the LIA
than during the MWP (Fig. 1 and Supplementary Fig. S10). On
the other hand, a record of flood deposits near 13◦S, on the Peru
Margin, indicating wet conditions during the LIA (ref. 22), is also
consistent with a more La Niña-like mean state during the LIA. This
site experiences dry conditions during El Niño events (Fig. 1 and
Supplementary Fig. S2).
Theoretical models and computer simulations also gave contra-
dictory results for the mean state in the LIA and the MWP. The
main external forcing difference between the LIA and the MWP is
believed to be the minimum solar irradiance (Fig. 3) from about
ad 1400 to 1850 (ref. 23) and the near lowest surface temperature
in the Northern Hemisphere (NH) and many other places globally
over the past millennium24. Some models have been proposed for
the response of the tropical Pacific ENSO to decreased solar forcing
and mean global temperature from the MWP to the LIA. The ‘ocean
dynamical thermostat’ mechanism25 predicted a more La Niña-like
state in the MWP than in the LIA. According to this model,
a positive solar forcing and increasing mean global temperature
during the MWP would result in a large zonal temperature gradient
across the equatorial Pacific. In the western tropical Pacific, rising
atmospheric temperature will warm the sea surface, but in the east-
ern equatorial Pacific, surface warming is restrained owing to the
cooling from upwelling. The increased zonal SST gradient enhances
the equatorial trade winds, further drives cooling by upwelling,
and increases the SST gradient. This model prediction is supported
by the SST-based reconstructions17,20,21 over the past millennium.
The most recent coupled general circulation models (CGCM),
however, project a weakening of the atmospheric overturning
2NATURE GEOSCIENCE |ADVANCE ONLINE PUBLICATION |www.nature.com/naturegeoscience
© 2011 Macmillan Publishers Limited. All rights reserved.
NATURE GEOSCIENCE DOI: 10.1038/NGEO1231 LETTERS
AD
EASM
1000 1200 1400 1600 1800 2000
NAOms
Net radiative forcing
MWP LIA
NH–T
SOIpr
Z–scores
Z–scores
Z–scores
Z–scores
Z–scores
¬3
¬2
¬1
0
1
2
3
4
¬2
¬1
0
1
2
¬4
¬3
¬2
¬1
0
1
2
3
¬3
¬2
¬1
0
1
2
3
¬3
¬2
¬1
0
1
2
3
Figure 3 |Linkage between SOIpr, solar irradiance and Northern
Hemisphere climate. Comparison of long-term proxy records SOIpr,
Northern Hemisphere temperature (NH-T; ref. 24), NAOms (ref. 28),
EASM (ref. 29), and solar irradiance23 over the past millennium. All time
series were 30-years smoothed and normalized to a standard Z-score. The
correlations among them are presented in Supplementary Table S2.
circulation (especially for the Walker Circulation) as the climate
warms, driven by changes in the atmospheric hydrologic cycle26,27.
The water vapour in the lower troposphere increases by roughly
7% per ◦C warming27, whereas precipitation increases more slowly,
approximately by 2% per ◦C warming27. The increased moisture has
to be transported from the atmospheric boundary layer to the free
troposphere, weakening the boundary layer/troposphere mass ex-
change and tropical overturning circulations26,27. This mechanism
predicted a less La Niña-like state in the MWP than in the LIA, in
agreement with our SOIpr reconstruction and other hydrological
records from the tropical Pacific8–15,22.
The tropical Pacific is important for global climate change1,3,9,
and the interaction between tropical Pacific hydrology and
Northern Hemisphere climate is an important aspect of the global
climate. In this study, we examined and found a significant link
between the reconstructed SOIpr and the Northern Hemisphere
temperature for the past millennium (Fig. 3; ref. 24). Negative
SOIpr, indicative of a weak PWC, corresponded to the warm climate
in the Northern Hemisphere during the Medieval Warm Period and
the last century. Positive SOIpr occurred during the relatively cold
LIA. The significant correlation (r=−0.54, p<0.05, neff =11.94,
ad 1000–1955) suggested a dynamic link between the hydrology
in the tropical Pacific and the Northern Hemisphere climate,
and this coupling may be established by the interplay between
the PWC and the Northern Atlantic Oscillation (NAO; ref. 28),
monsoonal circulations29 and other mechanisms (see Fig. 3 and
Supplementary Discussion).
Nonetheless, further research is also needed to reconcile the
contradiction between SST and hydrologic reconstructions, to
examine whether and how the centennial-scale variations of the
Pacific Walker Circulation are related to those of the ENSO during
the last millennium, and to study the relationships between the
tropical Pacific ENSO and mid-high latitude climate systems over
the past millennium and their mechanisms.
Methods
SOIpr calculation. To calculate SOIpr, we first normalized the instrumental and
reconstructed precipitation records to the standard Z-score: Z =(X−V)/SD; here
X is original value; V and SD are the averaged value and standard deviation of the
time series. SOIpr =wpZP−wnZN; here ZPand ZNare the normalized Z-scores
of the precipitation in the PA of the western Pacific and in the NA of the eastern
and mid Pacific, respectively. wpand wnare the optimal weights of ZPand ZN
and wp+wn=1. The values of wpand wnare chosen to optimize the correlation
between SOIpr and the instrumental SOI (Supplementary Fig. S4).
Error estimation of SOIpr.There are two kinds of uncertainties in the reconstructed
SOIpr: non-systematic and systematic. Non-systematic means those errors of local
nature, due to dating uncertainties and index measurement errors. For example,
the real age of a sample could be either earlier or later than the dated one. The
maximum dating uncertainties for the Galapagos record are ±100 years from
ad 50 to 1000 and ±60 from ad 1000 to 1890, based on a basal accelerator mass
spectrometry (AMS) AMS14C date assessment, and ±5 years in recent 110 years,
based on a basal Pb-Cs date assessment13. The index measurement errors for
the Galapagos record are not available, and we use the mean instrumental error
(±1%) instead. The dating uncertainties for the Indonesia record are ±40 years
from ad 50 to 1500 (AMS14C), ±90 years (AMS14 C) from ad 1500 to ad 1900,
and ±5 years (Pb-Cs) in the last century10. The index measurement errors for
Indonesia are available. We used Monte Carlo simulations to estimate this kind of
uncertainty. In each Monte Carlo simulation, to simulate the index measurement
noise, we generated and added Gaussian noise to the proxy records using the
available standard deviation and the Box–Muller algorithm30. The simulation of
dating uncertainties was based on the consideration that the effect of the dating
uncertainty is roughly equivalent to temporally shifting the time series by the
dating error. For each time point, we calculated the minimum and maximum
index values in the window of the dating error and assigned the index value to
an evenly distributed random number between the minimum and maximum
values. After application of these noises, both records were normalized to the
standard Z-score; the difference between the two records was recalculated and
denoted as SOIpr (see Methods above). We repeated the Monte Carlo simulation
Ntimes, obtained the distribution of ∼SOIpr −SOIpr, and calculated the standard
deviation for each time point. A Student ttest showed no significant differences
in the calculated standard deviations between N=100 and N=1,000 (p=0.69)
and greater, so we chose 1,000 for N. We estimated this uncertainty of SOIpr as
two standard deviation (95%) confidence intervals and the result is presented in
Supplementary Fig. S7.
The systematic uncertainty arises from the reservoir effect correction, and it
affects the dated ages of all subsamples of a time series globally. For example, an
over-correction of 10 years will shift all sample ages to be 10 years younger than
their actual ages. As SOIpr is the difference between two time series, this kind of
uncertainty could have a significant impact. We also used Monte Carlo simulations
to estimate this uncertainty. The reservoir age for the Indonesia record is 475±80
years10, and the estimated reservoir age for the Galapagos record is 0 ±4 years13. In
each Monte Carlo simulation, we randomly added from −80 to 80 years and from
−4 to 4 years, respectively, to the dated ages for the IPWP and Galapagos records
before calculating SOIpr. The other procedures are the same as above. The result of
the error estimation from both the systematic uncertainty and the non-systematic
uncertainty (combined) is also presented in Supplementary Fig. S7.
Correlation analysis. For two time series, Xand Y, the Pearson correlation
coefficient rxy was calculated as
rxy =Pn
i=1(xi−x)(yi−y)
(n−1)sxsy
where nis the number of samples, xand yare the sample means of Xand Y, and Sx
and Syare the sample standard deviation of Xand Y.
For two time series (Xand Y) with smoothing, we have to consider and
adjust the autocorrelation in Xand Yby using the effective sample size or
effective number of independent values. Following Trenberth (1984; ref 16),
and Bretherton et al. (1999; ref. 31), we first calculated τ, the time between
independent values (or the time to obtain a new degree of freedom) according to
the following equation32:
τ=1+2
(n−1)
X
l=1
rxl ryl
where rxl and ryl are the autocorrelation at lag lfor Xand Y. The effective number
of independent values was calculated as neff =n/τ , and the student t-value for
assessing significance was calculated as
t=rxy √neff −2
p(1−r2
xy )
NATURE GEOSCIENCE |ADVANCE ONLINE PUBLICATION |www.nature.com/naturegeoscience 3
© 2011 Macmillan Publishers Limited. All rights reserved.
LETTERS NATURE GEOSCIENCE DOI: 10.1038/NGEO1231
Received 17 May 2011; accepted 13 July 2011; published online
14 August 2011
References
1. Tsonis, A. A., Hunt, A. G. & Elsner, J. B. On the relation between ENSO and
global climate change. Meteorol. Atmos. Phys. 84, 229–242 (2003).
2. Trenberth, K. E. & Caron, J. M. The Southern Oscillation revisited: Sea
level pressures, surface temperatures, and precipitation. J. Clim. 13,
4358–4365 (2000).
3. Wang, C. Atmospheric circulation cells associated with the El Niño–Southern
oscillation. J. Clim. 15, 399–419 (2002).
4. McGregor, S., Timmermann, A. & Timm, O. A unified proxy for ENSO and
PDO variability since 1650. Clim. Past 6, 1–17 (2010).
5. Braganza, K., Gergis, J., Power, S., Risbey, J. & Fowler, A. A multiproxy index
of the El Nino–Southern Oscillation, AD 1525–1982. J. Geophys. Res. 114,
D05106 (2009).
6. Stahle, D. et al. Experimental dendroclimatic reconstruction of the Southern
Oscillation. Bull. Am. Meteorol. Soc. 79, 2137–2151 (1998).
7. Mann, M. et al. Global temperature patterns in past centuries: An interactive
presentation. Earth Interact. 4, 1–1 (2000).
8. Yan, H. et al. South China Sea hydrological changes and Pacific Walker
Circulation variations over the last millennium. Nature Commun. 2,
293 (2011).
9. Tierney, J., Oppo, D., Rosenthal, Y., Russell, J. & Linsley, B. Coordinated
hydrological regimes in the Indo-Pacific region during the past two millennia.
Paleoceanography 25, PA1102 (2010).
10. Oppo, D. W., Rosenthal, Y. & Linsley, B. K. 2,000-year-long temperature and
hydrology reconstructions from the Indo-Pacific warm pool. Nature 460,
1113–1116 (2009).
11. Liu, X. D. et al. A 1,100-year palaeoenvironmental record inferred from stable
isotope and trace element compositions of ostracode and plant caryopses in
sediments of Cattle Pond, Dongdao Island, South China Sea. J. Paleolimnol. 40,
987–1002 (2008).
12. Newton, A., Thunell, R. & Stott, L. Climate and hydrographic variability in the
Indo-Pacific Warm Pool during the last millennium. Geophys. Res. Lett. 33,
L19710 (2006).
13. Conroy, J. L., Overpeck, J. T., Cole, J. E., Shanahan, T. M. & Steinitz-Kannan,
M. Holocene changes in eastern tropical Pacific climate inferred from a
Galapagos lake sediment record. Quat. Sci. Rev. 27, 1166–1180 (2008).
14. Moy, C. M., Seltzer, G. O., Rodbell, D. T. & Anderson, D. M. Variability
of El Nino/Southern Oscillation activity at millennial timescales during the
Holocene epoch. Nature 420, 162–165 (2002).
15. Sachs, J. P. et al. Southward movement of the Pacific intertropical convergence
zone AD 1400–1850. Nature Geosci. 2, 519–525 (2009).
16. Trenberth, K. Signal versus noise in the Southern Oscillation. Mon. Weather Rev.
112, 326–332 (1984).
17. Cobb, K. M., Charles, C. D., Cheng, H. & Edwards, R. L. El Nino/Southern
Oscillation and tropical Pacific climate during the last millennium. Nature 424,
271–276 (2003).
18. Conroy, J. L. et al. Unprecedented recent warming of surface temperatures in
the eastern tropical Pacific Ocean. Nature Geosci. 2, 46–50 (2009).
19. Kennett, D. J. & Kennett, J. P. Competitive and cooperative responses
to climatic instability in coastal southern California. Am. Antiquity 65,
379–395 (2000).
20. Conroy, J. L., Overpeck, J. T. & Cole, J. E. El Nino/Southern Oscillation and
changes in the zonal gradient of tropical Pacific sea surface temperature over
the last 1.2 ka. PAGES News 18, 32–34 (2010).
21. Graham, N. E. et al. Tropical Pacific—mid-latitude teleconnections in medieval
times. Clim. Change 83, 241–285 (2007).
22. Rein, B. et al. El Nino variability off Peru during the last 20,000 years.
Paleoceanography 20, PA4003 (2005).
23. Bard, E., Yiou, Raisbeck & Jouzel, G. Solar irradiance during the last 1200 years
based on cosmogenic nuclides. Tellus 52, 985–992 (2000).
24. Mann, M. et al. Proxy-based reconstructions of hemispheric and global surface
temperature variations over the past two millennia. Proc. Natl Acad. Sci. 105,
13252 (2008).
25. Mann, M., Cane, M., Zebiak, S. & Clement, A. Volcanic and solar
forcing of the tropical Pacific over the past 1000 years. J. Clim. 18,
447–456 (2005).
26. Vecchi, G. A. et al. Weakening of tropical Pacific atmospheric circulation due
to anthropogenic forcing. Nature 441, 73–76 (2006).
27. Held, I. & Soden, B. Robust responses of the hydrological cycle to global
warming. J. Clim. 19, 5686–5699 (2006).
28. Trouet, V. et al. Persistent positive north atlantic oscillation mode dominated
the medieval climate anomaly. Science 324, 78–80 (2009).
29. Zhang, P. Z. et al. A test of climate, sun, and culture relationships from an
1810-Year Chinese cave record. Science 322, 940–942 (2008).
30. Box, G. E. P & Muller, M. E. A note on the generation of random normal
deviates. Ann. Math. Stat. 29, 610–611 (1958).
31. Bretherton, C. S., Widmann, M., Dymnikov, V. P., Wallace, J. M. & Blade, I.
The effective number of spatial degrees of freedom of a time-varying field.
J. Clim. 12, 1990–2009 (1999).
32. Box, G. E. P., Jenkins, G. M. & Reinsel, G. C. Time Series Analysis: Forecasting
and Control Vol. 16 (Holden-Day San Francisco, 1976).
Acknowledgements
Financial support for this research was provided by the Natural Science Foundation of
China (NSFC) (40730107) and the Major State Basic Research Development Program of
China (973 Program) (No.2010CB428902).
Author contributions
H.Y., L.S. and Y.W. designed the study and wrote the paper; Y.W., W.H., S.Q. and C.Y.
contributed to the statistical analysis and improving the English; all authors discussed the
results and implications and commented on the manuscript at all stages.
Additional information
The authors declare no competing financial interests. Supplementary information
accompanies this paper on www.nature.com/naturegeoscience. Reprints and permissions
information is available online at http://www.nature.com/reprints. Correspondence and
requests for materials should be addressed to L.S.
4NATURE GEOSCIENCE |ADVANCE ONLINE PUBLICATION |www.nature.com/naturegeoscience
© 2011 Macmillan Publishers Limited. All rights reserved.