Decadal Sea Surface Temperature Variability in the Subtropical South Pacific from 1726 to 1997 A.D.

Abstract

We present a 271-year record of Sr/Ca variability in a coral from Rarotonga in the South Pacific gyre. Calibration with monthly sea surface temperature (SST) from satellite and ship measurements made in a grid measuring 1 degrees by 1 degrees over the period from 1981 to 1997 indicates that this Sr/Ca record is an excellent proxy for SST. Comparison with SST from ship measurements made since 1950 in a grid measuring 5 degrees by 5 degrees also shows that the Sr/Ca data accurately record decadal changes in SST. The entire Sr/Ca record back to 1726 shows a distinct pattern of decadal variability, with repeated decadal and interdecadal SST regime shifts greater than 0. 75 degrees C. Comparison with decadal climate variability in the North Pacific, as represented by the Pacific Decadal Oscillation index (1900-1997), indicates that several of the largest decadal-scale SST variations at Rarotonga are coherent with SST regime shifts in the North Pacific. This hemispheric symmetry suggests that tropical forcing may be an important factor in at least some of the decadal variability observed in the Pacific Ocean.

Full text

⬍100 ppm show an almost constant N abundance
signature, irrespective of the concentrations, with
␦
15
N⬇–15 ⫾15‰. There are two major sources for
the analytical blank: outgassing of the sample chamber
and sample surface contamination. Only the outgassing
N is corrected here. This correction represents ⬍1% of
the N concentrations and only a few ‰ of the extreme
␦
15
N values measured in the first few tens of nm of the
grains. At ⬎150-nm depth of the grain, blank (outgas-
sing) N often accounts for ⬃50% of the detected N,
although the exact fraction varies, depending on the
analytical condition, such as the sputtering rate or the
rastering size. (For the correction for the outgassing N,
we adopt the minimum amount among the possible
range to avoid overcorrection of the blank.) Degrees of
surface contamination are largely variable, depending
on grains. The surface contamination is observed prom-
inently at ⬍50-nm depth. Its concentration at the most
surface of the sample (when the primary ion beam
crosses the border of the gold film and the sample)
normally ranges from 100 to 5000 ppm. No systematic
difference is observed in the contamination level be-
tween the grains in samples 71501 and 79035 or be-
tween ilmenite and silicate grains.
20. The ␦D values of sample 71501 corrected for cosmo-
genic D (␦D
trapped
), assuming a maximum exposure
age of 200 Ma (11) and using a D production rate [L.
Merlivat, M. Lelu, G. Nief, E. Roth, Proc. Lunar Sci.
Conf. VII, 649 (1974)], are available at Science Online
(18). The amount of cosmogenic D relative to im-
planted H in grain 1 from sample 79035 is negligible.
21. The best estimate for the minimum ␦
15
N value ob-
served in grain 1 from sample 79035 (–240 ⫾25‰)
is obtained from the weighted average, taken from
␦
15
N values as low as –200‰. Though lower values
are observed, e.g., –290 ⫾100‰ (18) at 120-nm
depth (Fig. 1A), the differences from the best esti-
mate value are not significant. The observed ␦
15
N
value (–240‰) can be regarded as the upper limit for
the SW ␦
15
N value, because the observed values are
not perfectly free of blank N, especially the surface
contamination N.
22. R. Wieler, H. Baur, Astrophys. J. 453, 987 (1995).
23. L. P. Keller, D. S. McKay, Geochim. Cosmochim. Acta
61, 2331 (1997).
24. J.-P. Benkert, H. Baur, P. Signer, R. Wieler, J. Geophys.
Res. 98, 13147 (1993).
25. P. Bochsler, R. Kallenbach, Meteoritics 29, 653 (1994).
26. F. Humbert, thesis, Universite´ Henri Poincare´, Nancy
1, France (1998).
27. J. Geiss, G. Gloeckler, R. von Steiger, Philos. Trans. R.
Soc. London A 349, 213 (1994).
28. R. Kallenbach et al.,Astrophys. J. 507, L185 (1998).
29. T. Fouchet et al.,Icarus 143, 223 (2000).
30. E. Zinner, Annu. Rev. Earth Planet. Sci. 26, 147
(1998).
31. M. B. McElroy, Y. L. Yung, A. O. Nier, Science 194,70
(1976).
32. Assuming that N is lost from a N-bearing host phase
with a rate proportional to the inverse square root of
the atomic mass (m
⫺1/2
), 4.5 or 18 orders of mag-
nitude depletion of N concentration is required to
enrich
15
N by 40 or 300%, respectively, from the
initial isotopic composition. To explain by such frac-
tionation the N isotopic composition of CI chon-
drites, which is enriched in
15
N by 40% relative to the
solar value, the initial N concentration must be about
three orders of magnitude overabundant in compar-
ison to the solar composition (because the N/major
solid elements ratio in CI chondrites is 1.7 orders of
magnitude lower than the solar value), which is high-
ly unrealistic.
33. N. G. Adams, D. Smith, Astrophys. J. 247, L123
(1981).
34. Ratios averaged over wide depth ranges, rather than
H and N isotopic compositions observed at respec-
tive depths, are plotted in Fig. 2. This is because the
latter may not always reflect the original mixing
proportion of the SW and nonsolar components
when they were acquired, because H in the present
grain is not located at the original site but is diffused
over a long range within the grains.
35. B. Marty, L. Zimmermann, Geochim. Cosmochim.
Acta 63, 3619 (1999).
36. D. C. Jewitt, H. E. Matthews, T. Owen, R. Meier,
Science 278, 90 (1997).
37. Samples were provided by NASA. We thank R. Wieler
for constructive comments and suggestions, H. Fukui
and O. Ohtaka for preparing the synthetic glass stan-
dards, and N. Shimobayashi and M. Kitamura for
nondestructive elemental mapping of lunar grains.
K.H. thanks members of CRPG-CNRS for their hospi-
tality during his stay. This study was supported by
the Japanese Ministry of Education, Science, Sports
and Culture and by CNRS through a “Poste Rouge”
fellowship by grants from Institut National des Sci-
ences de l’Univers–Programme, National de Plan-
e´tologie and from Re´gion Lorraine (K.H.). This work is
CRPG-CNRS contribution 1488.
3 August 2000; accepted 29 September 2000
Decadal Sea Surface
Temperature Variability in the
Subtropical South Pacific from
1726 to 1997 A.D.
Braddock K. Linsley,
1
Gerard M. Wellington,
2
Daniel P. Schrag
3
We present a 271-year record of Sr/Ca variability in a coral from Rarotonga in
the South Pacific gyre. Calibration with monthly sea surface temperature (SST)
from satellite and ship measurements made in a grid measuring 1° by 1° over
the period from 1981 to 1997 indicates that this Sr/Ca record is an excellent
proxy for SST. Comparison with SST from ship measurements made since 1950
in a grid measuring 5° by 5° also shows that the Sr/Ca data accurately record
decadal changes in SST. The entire Sr/Ca record back to 1726 shows a distinct
pattern of decadal variability, with repeated decadal and interdecadal SST
regime shifts greater than 0.75°C. Comparison with decadal climate variability
in the North Pacific, as represented by the Pacific Decadal Oscillation index
(1900–1997), indicates that several of the largest decadal-scale SST variations
at Rarotonga are coherent with SST regime shifts in the North Pacific. This
hemispheric symmetry suggests that tropical forcing may be an important
factor in at least some of the decadal variability observed in the Pacific Ocean.
It is now recognized that significant tropical
and subtropical Pacific ocean-atmosphere vari-
ability occurs on decadal-to-interdecadal time
scales. However, in order to evaluate decadal-
scale climate variability over time, climate
records long enough to capture multiple dec-
adal periods are needed, and these records are
limited (1,2).In the North Pacific, sufficient
instrumental climatic data exist to identify a
pattern of irregular decadal-to-interdecadal
ocean-atmosphere climate variability over the
past ⬃100 years (2–7).The time history of the
leading eigenvector of North Pacific SST back
to 1900 A.D. has been termed the Pacific Dec-
adal Oscillation (PDO) by Mantua et al.,(6).
The PDO is a recurring pattern of ocean-atmo-
sphere variability in which the central gyre
cools at the same time as the eastern margin
warms, or vice versa. Alternating phases of the
PDO can last for two to three decades, with
reversals being noted in 1924/25, 1946/47, and
1976/77 (6).Before 1900, little is known about
decadal variability in the North Pacific. In the
South Pacific, oceanographic data are extreme-
ly sparse. However, the limited data indicate
that the subtropical South Pacific may also play
a role in decadal-scale oceanographic variabil-
ity in the Pacific (8–11).For example, the South
Pacific is currently the dominant (50 to 75%)
source region for isopycnal water transport to
the equatorial thermocline (10,11).This is due
in part to the partial blocking effect of the
surface ocean beneath the Intertropical Conver-
gence Zone (ITCZ) in the North Pacific and the
more limited influence of the South Pacific
Convergence Zone (SPCZ) on South Pacific
isopycnal equatorward flow (8,11).
The spatial pattern of decadal variability in
Pacific SSTs is similar to that associated with
El Nin˜o –Southern Oscillation (ENSO), but
with lower amplitude in the tropics and higher
amplitude outside the tropics (3,12). Some
studies suggest that this decadal variability may
originate in the subtropics of the North Pacific
Ocean through unstable ocean-atmosphere in-
teractions (2,13), whereas other studies suggest
that tropical ENSO forcing plays a key role (3,
12,14,15).However, several key questions
regarding the nature of Pacific decadal variabil-
ity remain, including the recurrence period of
decadal changes in SST; whether the decadal-
scale SST variability in the subtropical South
Pacific is in phase with the North Pacific; and
1
Department of Earth and Atmospheric Sciences, ES
351, University at Albany–State University of New
York, Albany, NY 12222, USA.
2
Department of Biology
and Biochemistry, University of Houston, Houston, TX
77204, USA.
3
Laboratory for Geochemical Oceanog-
raphy, Department of Earth and Planetary Sciences,
Harvard University, Cambridge, MA 02138, USA.
REPORTS
www.sciencemag.org SCIENCE VOL 290 10 NOVEMBER 2000 1145

whether the decadal SST variability is periodic
or more concentrated during certain times, such
as in the late 19th and 20th centuries.
To examine decadal variability in the South
Pacific, we produced a proxy record of SST
from a coral growing at Rarotonga. The island
of Rarotonga is located at 21.5°S and 159.5°W
in the Cook Islands in the region of the eastern
SPCZ. Corals growing at Rarotonga are ex-
posed to open ocean conditions of the westward
flowing South Equatorial Current. SST in the
2°-by-2° grid surrounding Rarotonga averages
25.7°C, with a consistent 4° to 5°C seasonal
SST range, with maximum water temperatures
occurring in February-March of each year (16 ).
This region has also been identified as a “center
of action” for monitoring changes in the South-
ern Oscillation (17 ).During ENSO “warm
mode” (El Nin˜o) events, the SPCZ moves to the
northeast, joining the ITCZ in the central-west-
ern tropical Pacific. This condition leads to cool-
er and drier than average conditions in the re-
gion. During ENSO “cool modes” (or La Nin˜a
conditions), the situation reverses, SST rises,
and the SPCZ intensifies.
In April 1997, 3.5 m of continuous coral
core (representing 271 years of growth) (18)
was collected by hydraulic drill from a massive
colony of Porites lutea in 18.3 m of water on the
southwest side of Rarotonga at 21°14⬘11⬙S,
159°49⬘59⬙W. We measured Sr/Ca on 1-mm-
interval samples spanning the entire core and
measured oxygen isotopes (␦
18
O) over the in-
tervals from 1950 to 1997 (at 1 mm resolution)
and 1726 to 1770 (at 2 mm resolution) for
calibration purposes (19–21).Over the interval
from 1981 to 1997, both Sr/Ca and ␦
18
O were
compared to instrumental SST from the Inte-
grated Global Ocean Service System Products
(IGOSS) data (16 ) in the 1°-by-1° grid centered
at 22°S and 160°W. The Sr/Ca data show a
remarkable coherence with SST over seasonal
and interannual time scales, with a seasonal
amplitude consistent with the average annual
SST cycle of 4° to 5°C (Fig. 1). For ␦
18
O(22),
the seasonal range of ⬃0.9 to 1.0 per mil (‰)
matches the expected range if water temperature
was the dominant influence on coral ␦
18
O(23).
However, there are also intervals where ␦
18
O
does not track SST very well. Linear least-
squares regression analysis indicates that the
variability in Sr/Ca explains significantly more
of the variance in SST (r
2
⫽0.75) than does
coral ␦
18
O(r
2
⫽0.54) (24 ). This is not surpris-
ing, because we expect coral skeletal ␦
18
Otobe
affected by both SST and ␦
18
O
seawater
, whereas
skeletal Sr/Ca has been shown to vary with SST
alone (25).
The derived relation between SST and Ra-
rotonga coral Sr/Ca is [SST (°C) ⫽140.55 –
12.15(Sr/Ca ⫻1000)]. The slope of this regres-
sion equation is similar to that determined for
coral calibrations at other locations (25–28).
Considering that IGOSS SST values represent a
1°-by-1° grid centered at 22°S, 160°W, where-
as the Sr/Ca data represent a single point from
18 m depth at Rarotonga, the high degree of
correlation obtained is remarkable and demon-
strates the utility of Sr/Ca as a proxy for SST at
this location.
To determine whether interannual and dec-
adal Sr/Ca variations are also the result of SST
changes, we compared Sr/Ca-derived SST
anomalies to instrumental SST anomalies for
the period from 1950 to 1997 (Fig. 2). Both
interannual and decadal changes in coral Sr/Ca-
derived SST anomalies are significantly corre-
lated with instrumental SST anomalies, particu-
larly back to 1960, which includes the most
continuous part of the instrumental SST record
(29). Most El Nin˜o events are recorded as cool
anomalies at Rarotonga but with reduced ampli-
tudes as compared to the tropical Nin˜o3/4 region
(30). Although not shown here, it is noteworthy
that decadal changes in coral ␦
18
O at Rarotonga
do not closely track SST and coral Sr/Ca, pos-
sibly indicating the effects of decadal changes in
the ␦
18
O composition of seawater and/or salinity
on coral ␦
18
O at this site.
The complete Sr/Ca-derived SST record
back to 1726 shows variability over a range of
time scales (Fig. 3). SSTs have varied between
23° and 24°C in the winter to 27° to 28°C in the
summer from 1997 back to 1765, with pro-
nounced interannual and decadal variations.
From 1726 to 1765, mean annual SST was ⬃1°
to 1.5°C higher, with the same seasonal ampli-
tude. Because this excursion is so unusual, we
tested the reliability of the Sr/Ca data by mea-
suring ␦
18
O over the same interval. The ␦
18
O
data show a 0.3‰ excursion equivalent to the
SST change indicated by the Sr/Ca data. This
suggests that a large SST shift in the central
South Pacific gyre did occur at this time, al-
though this conclusion should be confirmed
with additional coral records. A mid-1700s
warm interval has not been found in other
Fig. 1. Comparison of monthly IGOSS SST (16) for the grid including Rarotonga (1° by 1°; centered at
22°S, 160°W ) and near-monthly Rarotonga coral Sr/Ca, spanning the interval from 1981 to 1997. The
linear least-squares correlation between SST data and Rarotonga Sr/Ca has an r
2
⫽0.75 (24).
Fig. 2. Comparison of Rarotonga Sr/Ca-calculated SST anomalies (deseasonalized) with instrumen-
tal SST from a grid measuring 2° by 2° (CAC SST anomalies) and 5° by 5° (OS SST anomalies) for
the Rarotonga region (29).
REPORTS
10 NOVEMBER 2000 VOL 290 SCIENCE www.sciencemag.org1146

paleoclimatic records from the subtropical
South Pacific, but none of these records are
from the central gyre region. After the abrupt
1.5°C cooling from 1764 to 1766, the Sr/Ca
series records a gradual 0.75°C cooling to
⬃1910 and then gradual 0.5°C warming to the
present.
In comparison to equatorial areas of the
central Pacific, decadal and interdecadal vari-
ance in Rarotonga coral Sr/Ca is relatively large
in relation to variance attributed to interannual
variability (31). This observation is in agree-
ment with what is known about SST variability
in the North and South Pacific gyres, based on
historical measurements (3,12).As recorded at
Rarotonga, the mean recurrence period of the
irregular, broad-band, decadal SST variability
is near 14 years, with a total of 11 decadal
intervals in which mean annual water tempera-
tures cooled by ⬎0.75°C: 1976–77, 1956 –58,
1922–1940, 1903– 06, 1871–75, 1841– 44,
1829 –37, 1820 –23, 1788 –92, 1764 – 66, and
1756 –1759 (Fig. 4). A decadal mode of vari-
ability is also found in the multicentury ␦
18
O
coral record from New Caledonia, located in
the western South Pacific south of the SPCZ
(32). However, this decadal variability is not
coherent with the Rarotonga Sr/Ca record. Per-
haps this is due to the different climate regime
at New Caledonia as compared to the central
gyre location of Rarotonga. The abrupt 0.6°C
cooling event that occurred in 1815, with a
gradual warming until 1819, may be related to
the Tambora eruption in April 1815 (33–35).
The stratospheric dust veil produced by this
eruption is known to have lowered tempera-
tures in Europe and North America for several
years after the eruption. The 0.5°C cooling
observed at Rarotonga is also in agreement with
reports of a comparable cooling in the tropics
(35) and as found in other coral records (36,
37 ).
To examine whether the decadal-scale SST
changes inferred from the Rarotonga coral are
consistent with what is known about decadal-
scale variability in the subtropical North Pacif-
ic, in Fig. 4 we compare the Rarotonga Sr/Ca-
derived SST record to the PDO index. Mantua
et al. (6) formulated the PDO index so that
when the central North Pacific is cooler than
average and the Gulf of Alaska and the waters
along the Pacific Coast of North America are
warmer than average, the index is positive.
These periods tend to correspond with times of
increased frequencies of ENSO warm phases.
Increased frequencies of La Nin˜a years corre-
spond with the negative phase of the index,
when the central North Pacific is warmer than
average, and the coastal waters of the NE Pa-
cific are cooler than average. Over the past
century, several of the most pronounced dec-
adal changes in the PDO index (1976 –77,
1956 –58, 1945– 47) are evident in the Raro-
tonga Sr/Ca record as well as other more subtle
changes in Rarotonga SST (see 1915– 40).
Times of disagreement around 1910 and the
early 1960s may be related to the fact that we
are comparing a specific point in the South
Pacific with a North Pacific–wide index. Cross-
spectral analysis of Rarotonga Sr/Ca and the
PDO indicates that the Rarotonga Sr/Ca record
is moderately coherent with the PDO index at a
decadal period centered near 15 years and at an
interannual ENSO period near 6 years (80%
confidence level) (38). The degree of correla-
tion is hampered by the short section of overlap
(97 years) in relation to the long period of the
decadal-scale changes. Longer records from the
North Pacific would allow a more rigorous
evaluation of the extent of coherence of decadal
SST variability about the equator.
The fact that several of the largest decadal
changes observed in the past 100 years are in
phase in the North and South Pacific gyres
suggests that the origin of the decadal variabil-
ity during these times of coherent behavior is
likely to lie in the tropics, which is consistent
with some previous suggestions (3,12,14,15).
The specific mechanism could be the export
Fig. 3. (Upper curve)
Near-monthly changes
in calculated SST span-
ning 1726 to 1997, us-
ing Rarotonga coral Sr/
Ca and the regression
relationship [SST ⫽
140.55 ⫺12.15(Sr/
Ca ⫻1000)] derived
using data shown in
Fig. 1. (Lower curve)
␦
18
O measurements
spanning the interval
from 1726 to 1770.
Fig. 4. Comparison of 8-year low-pass–filtered versions of Rarotonga Sr/Ca-calculated SST with the
mean removed (solid line) and the PDO index (dashed line). As defined by Mantua et al. (6), a
positive phase of the PDO index corresponds to a El Nin˜o mode, and a negative phase corresponds
toaLaNin˜a mode. Solid arrows denote decadal cooling shifts of ⬎0.75°C at Rarotonga, and the
open arrow indicates a cooling trend, possibly related to the Tambora eruption in April 1815.
REPORTS
www.sciencemag.org SCIENCE VOL 290 10 NOVEMBER 2000 1147

and subtropical amplification of a decadal
mode of ENSO from the tropics to the subtrop-
ics. However, as discussed in (12), different
spatial patterns of midlatitude atmospheric cir-
culation anomalies for ENSO-band and decadal
variability suggest that different tropical forcing
mechanisms are involved on these two time
scales. Thus, the specific mechanism for gener-
ating subtropical decadal SST variability in the
Pacific gyres appears to be complex and may
involve tropical-subtropical ocean-atmosphere
interactions other than ENSO.
References and Notes
1. K. E. Trenberth, Bull. Am. Meteorol. Soc. 71, 988
(1990).
2. M. Latif, T. P. Barnett, Science 266, 634 (1994).
3. Y. Zhang, J. M. Wallace, D. S. Battisti, J. Clim. 10, 1004
(1997).
4. C. Deser, M. A. Alexander, M. S. Timlin, J. Clim. 9,
1840 (1996).
5. S. Minobe, Res. Lett. 24, 638 (1997).
6. N. J. Mantua, S. R. Hare, Y. Zhang, J. M. Wallace, R. C.
Francis, Bull. Am. Meteorol. Soc. 78, 1069 (1997).
7. A. Gershunov, T. P. Barnett, D. R. Cayan, Eos Trans.
AGU 80, 25 (1999).
8. W. B. White, D. R. Cayan, J. Geophys. Res. 103, 21335
(1998).
9. R. G. Peterson, W. B. White, J. Geophys. Res. 103,
24573 (1998).
10. M. Tsuchiya et al.,Prog. Oceanogr. 23, 101 (1989).
11. G. C. Johnson, M. J. McPhaden, J. Phys. Oceanogr. 29,
3073 (1999).
12. R. D. Garreaud, D. S. Battisti, J. Clim. 12, 2113 (1999).
13. N. E. Graham, Clim. Dyn. 10, 135 (1994).
14. G. A. Jacobs et al.,Nature 370, 360 (1994).
15. K. E. Trenberth, J. W. Hurrell, Clim. Dyn. 9, 303 (1994).
16. R. W. Reynolds, T. M. Smith, J. Clim. 7, 929 (1994).
17. K. E. Trenberth, in Teleconnections Linking Worldwide
Climate Anomalies, M. H. Glantz, R. W. Katz, N. Nicholls,
Eds. (Cambridge Univ. Press, 1991), pp. 13–42.
18. We cut 7-mm-thick slabs of coral, which were cleaned
in deionized water in an ultrasonic bath. Dry slabs were
sampled with a low-speed microdrill along tracks par-
allel to corallite traces, as identified in x-ray positives,
with a round diamond drill bit 1 mm in diameter.
19. We used an inductively coupled plasma atomic emis-
sion spectrophotometer at Harvard University to
measure coral skeletal Sr/Ca, following a technique
described in detail by D. P. Schrag [Paleoceanography
14, 97 (1999)]. The external precision was better
than 0.15% (relative standard deviation), based on
analyses of replicate samples.
20. We measured oxygen isotopes on a gas source mass
spectrometer with an individual acid reaction vessel
system, following procedures outlined by B. K. Lins-
ley, L. Ren, R. B. Dunbar, and S. S. Howe [Paleocean-
ography 15, 322 (2000)]. External precision was bet-
ter than 0.04‰ for ␦
18
O, based on analyses of
replicate samples.
21. To construct the chronology, we tied the annual
minima in Sr/Ca and ␦
18
O to February (on average
the warmest month) and maximum Sr/Ca and ␦
18
O
to August/September (on average the coolest
months). We also assumed that density bands in this
coral skeleton are deposited annually.
22. B. K. Linsley, G. M. Wellington, D. P. Schrag, data not
shown.
23. G. M. Wellington, R. B. Dunbar, G. Merlen, Paleocean-
ography 11, 467 (1996).
24. Three-month running averages result in correlations
to IGOSS SST of r
2
⫽0.82 for Sr/Ca and r
2
⫽0.59 for
␦
18
O . We believe that the 3-month running average
regression results may be more representative of the
actual relation with SST, because the age model for
the coral record may have uncertainties of 1 to 2
months.
25. J. W. Beck et al.,Science 257, 644 (1992).
26. S. de Villiers, G. T. Shen, B. K. Nelson, Geochim.
Cosmochim. Acta 58, 197 (1994).
27. C.-C. Shen et al.,Geochim. Cosmochim. Acta. 60,
3849 (1996).
28. C. Alibert, M. T. McMulloch, Paleoceanography 12,
345 (1997).
29. Evaluation of the Comprehensive Ocean Atmosphere
Data Set SST data for the 2°-by-2° region around
Rarotonga reveals that near-continuous monthly
data exist only back to 1960. The Global Ocean
Surface Temperature Atlas data for the 5°-by-5°
region around Rarotonga only contain near-continu-
ous SST anomaly data back to 1950. Both SST data-
bases contain virtually no measurements before
1930. Thus, we compared interannual and decadal
changes in Sr/Ca to measured SST anomalies only
back to 1950. The Climate Analysis Center (CAC) SST
anomalies are from (16). The optimally smoothed
(OS) SST anomaly data are from Kaplan et al. [J.
Geophys. Res.103, 18567 (1998)]. Correlations be-
tween annually averaged coral Sr/Ca and annually
averaged OS SST and CAC SST anomalies have r
2
values of 0.45 and 0.37, respectively. Although these
are lower than the correlation to 1°-by-1° IGOSS
SST, they are both significant at the 99% level.
30. The Nin˜o3/4 SST record reflects SST anomalies in the
central equatorial Pacific region bounded by 5°S–5°N
and 120°–170°W.
31. This observation is based on Singular Spectrum Anal-
ysis (SSA) of the Sr/Ca Series (1726–1997), using
software written by E. Cook (Lamont-Doherty Earth
Observatory). After bandpass filtering to remove the
annual cycle and periods greater than 50 years, 37%
of the variance is attributed to decadal and interdec-
adal variability and 43% of the variance to ENSO
band (3- to 7-year) variability (n⫽3244 and window
length ⫽120 months). In comparison, SST variability
in the equatorial Nino3/4 region contains only 18%
of variance in the decadal and interdecadal bands and
64% in the interannual band [based on the same SSA
analysis of Kaplan OS SST anomaly data (29) for the
Nino 3/4 region].
32. T. M. Quinn et al.,Paleoceanography 13, 412 (1998).
33. M. R. Rampino, S. Self, Quat. Res. 18, 127 (1982).
34. R. B. Stothers, Science 224, 1191 (1984).
35. H. H. Lamb, Philos. Trans. R. Soc. London Ser. A 266,
425 (1970).
36. T. J. Crowley, T. M. Quinn, F. W. Taylor, C. Henin, P.
Joannot, Paleoceanography 12, 633 (1997).
37. J. E. Cole, R. B. Dunbar, T. R. McClanahan, N. Muthiga,
Science 287, 617 (2000).
38. Cross-spectral analysis was done on 2-year low-
bandpass–filtered versions of Rarotonga Sr/Ca and
the PDO with the ARAND software package, which is
maintained and distributed by P. Howell of Brown
University.
39. We thank O. Hoegh-Guldberg and J. Caselle for assist-
ance with field sampling and E. Goddard and S. Howe
for analytical assistance. Comments from two anony-
mous reviewers were also greatly appreciated. Support-
ed by NSF grant ATM-9901649 and NOAA grant
NA96GP0406 to B.K.L., NSF grant ATM-9619035 and
NOAA grant NA96GP0470 to G.M.W., and NSF grants
OCE-9733688 and OCE-9819257 to D.P.S.
8 June 2000; accepted 10 October 2000
Contributions of Land-Use
History to Carbon
Accumulation in U.S. Forests
John P. Caspersen,
1
* Stephen W. Pacala,
1
Jennifer C. Jenkins,
2
George C. Hurtt,
3
Paul R. Moorcroft,
1
Richard A. Birdsey
4
Carbon accumulation in forests has been attributed to historical changes in land
use and the enhancement of tree growth by CO
2
fertilization, N deposition, and
climate change. The relative contribution of land use and growth enhancement is
estimated by using inventory data from five states spanning a latitudinal gradient
in the eastern United States. Land use is the dominant factor governing the rate
of carbon accumulation in these states, with growth enhancement contributing far
less than previously reported. The estimated fraction of aboveground net ecosys-
tem production due to growth enhancement is 2.0 ⫾4.4%, with the remainder due
to land use.
Although mid-latitude forests of the northern
hemisphere are known to provide a large sink
for atmospheric CO
2
(1–3), considerable un-
certainty remains about the cause of the sink.
Nitrogen deposition, CO
2
fertilization, and cli-
mate change have been shown to enhance tree
growth in forest ecosystems (4), but historical
changes in land use also provide an alternative
explanation for the sink, particularly the re-
growth of forests after agricultural abandon-
ment, reduced harvesting, and fire suppression
(5). Assessing the relative contribution of land
use and growth enhancement is critical for plan-
ning strategies to mitigate the accumulation of
CO
2
in the atmosphere (6). If forests are simply
regrowing in response to changes in land use,
then the sink can be expected to saturate as
forests regain their former biomass. However, if
tree growth has been enhanced, then the future
storage potential of forests is much less certain.
Estimates of the fraction of the forest sink
due to regrowth versus enhancement vary wide-
ly, but growth enhancement has been consis-
tently estimated to be large. In the United
1
Department of Ecology and Evolutionary Biology,
Princeton University, Princeton, NJ 08540, USA.
2
Northeastern Research Station, USDA Forest Service,
Post Office Box 968, Burlington, VT 05402, USA.
3
Complex Systems Research Center, Institute for the
Study of Earth, Oceans, and Space, University of New
Hampshire, Durham, NH 03824, USA.
4
Northeastern
Research Station, USDA Forest Service, 11 Campus
Boulevard, Suite 200, Newtown Square, PA 19073,
USA.
*To whom correspondence should be addressed. E-
mail: jpc@eno.princeton.edu
REPORTS
10 NOVEMBER 2000 VOL 290 SCIENCE www.sciencemag.org1148