Pulses of Rapid Ventilation in the North Atlantic Surface Ocean During the Past Century

The Suess Effect is a gradual decrease of the 14C/12C ratio in tree rings from 1870 to 1950 that is attributed mostly to the burning of 14C-free fossil fuels for energy. In contrast, pre-nuclear bomb high-precision measurements of 14C/12C ratios in banded corals from the North Atlantic Ocean do not exhibit the expected Suess Effect. Instead, large biennial-to-decadal shifts in the 14C/12C ratios appear throughout the coral records. This pattern indicates that rapid pulses of increased mixing between surface and subsurface waters occurred in the North Atlantic Ocean during the past century.

the right on the Sr-Ca peak.This sample was determi ned to be pure aragonite by XRD (7) .The dashed li ne in Fig. 5 is the Fourier transform of the Sr K-edge EXAFS spectrum fo r coral A P-14.A linear combination of -60% Sr substituted in aragonite and -40% Sr in a SrC0 3 phase approximates (::!:1 5%) rhe EXAFS of AP-14 and all of the MA corals.The remaining two AP corals did not show as large a SrC0 3 shoulder on rhe doubler ar -4 A, suggesting a more homogeneous d istribution of their Sr in che aragonite structure.
Plummer and Busenberg (6) showed rhac there was a wide miscibility gap in the aragon ire-SrC0 3 solid solution from a SrC0 3 mole fraction of 0.0058 ::!: 0.0003 to 0.875 ::!: 0.0 11 at 25°C.A ragonite with more than -5000 ppm of Sr substituted for Ca is therefore metastable.For the twophase corals, our estimate of 4500 ppm of Sr at the Ca structural sites in aragonite (60% of 7500 ppm) and 3000 ppm in the SrC0 3 (40% of 7500 ppm) agrees with thermodynamic analysis of the laboracory aragonite-SrC03 syscem (6).Most scleractinian corals grow between 25° and 29°C, a temperature range close co the 25°C of the synrheric system (6) .The rwo AP corals exhibiting smaller SrC0 3 shoulders apparently formed as metastable h igh-Sr arago nire.le seems reasonable that the SrC0 3 domains are primary features char formed during biomineralizarion .The domains are present in specimens that were collected live, and solid-stare diffusion or dissolucionreprecipicacion is difficult or slow in seawater at ambient temperatures.The size and distribution of the SrC0 3 domains is not known , and neither is the mechanism by which corals sequester more Sr in aragonite skeletons than expected at equili bri um wi th seawater, although Plummer and Busenberg (6) suggest that precipitation kinetics cause nonequilibrium incorporation of Sr in many marine carbonates.
These findings indicate chat Sr does nor substitute solely for Ca at its structural sires in MA and A P, species char have nor been directly cali brared for Sr/Ca SST paleorhermomerry.Further invescigadon of the SrC0 3 in these and ocher species will advance our understanding of the mechanisms of Sr incorporation into scleractinian coral.
jor aragonite peaks were present and identified.No other peaks were encountered, implying a purity in excess of 98% aragonite.Strontium and calcium were determined in these samples by direct current plasma atomic emission spectroscopy.
8. The materials for XAS analysis were ground to fine powders of 300 to 400 mesh size.The standards were mixed 1n Duco cement, and spread as a thin layer on Al foil.Beam line 4-1 at SSRL was used with a Si(l 11) (for Ca) and Si(220) (for Sr) monochromator.The stored electron beam energy was 3 GeV a1 a current of 50 to 90 mA.The monochromator was detuned to reduce harmonics, and the slits were set at 1 mm by 20 mm to give -1-and 4-eV resolution at the Ca and Sr K-edges, respectively.The data were gathered in fluorescence mode for the corals and in absorption mode for the standards.The Ca K-edge experiment was perlormed in the electron yield mode, as noted in the data where appropriate.9.The Fourier transform peaks were identified with the use of the software

R. M. Druffel
The Suess Effect is a gradual decrease of the 14 C/ 12 C ratio in tree rings from 1870 to 1950 that is attributed mostly to the burning of 1 4 C-free fossil fuels for energy.In contrast, pre-nuclear bomb high-precision measurements of 14 C/ 12 C ratios in banded corals from the North Atlantic Ocean do not exhibit the expected Suess Effect.Instead, large biennial-to-decadal shifts in the 14 C/ 1 2 C ratios appear throughout the coral records.This pattern indicates that rapid pulses of increased mixing between surface and subsurface waters occurred in the North Atlantic Ocean during the past century.
Since the late 19th century, humans have burned substantial amounts of foss il fuels for energy.As a result of chis combustion, the concentration of C0 2 in the armosphere has increased by at least 25% (I).T his increase represents about half of the C0 2 that has been emirted into the armosphere; most of the ocher half is in the oceans.The decline in with the same 6 14 C signature as that in DIC in the surround ing seawater (5,6).T he 8 18 0 (7) values in banded corals have been shown to correlate inversely with sea surface temperature (SST ) (8)(9)(10) and to correlate d irectly with the salinity and 8 18 0 of seawater ( 11 , 12).Here I report highprecision 6 14  Bermuda and Pickles Reef coral, and 5-year bands we re cut from T he Rocks coral (13).
Most of the samples were analyzed for 6. 14 C wich the use of conventional gas counting ( 13); a few samples were analyzed with acce leracor mass spectrometry (AMS) at the University of Arizon a ( 14).The 6. 14 C measurements fro m conventional counting have l a uncertainties (counting statistics and laboratory errors) of 2.5 co 3.0 per m i l.U n certai nties for t he AMS measurements were 2.8 to 3.9 per mi l (15)....  A leasr-squares fit of~ 14 C versus 8 18 0 revealed an inverse correlation (linear correlation coefficient r = -0.31;n = 30), which is significant at the 90% confidence level (a = 0.90).An inverse correlation (r = -0.68, a = 0.95; n = 9) was also found between annual 8 18 0 and ~14 C values lagged by I year for the post-1971 bands from this coral (17).This association suggests that low SSTs or high salinities (indicated by high 8 18 0) or both accompany periods of low ~ 14 C values in the Sargasso Sea and are controlled by atmospheric forcing of the ocean surface.The 8 18 0 results (lagged by 2 years) for the Florida coral ( J 9) also showed an inverse correlation with ~ 14 C data ( 1• = -0.45, a = 0.98; n = 28) .This association and the inverse correlations between average annual wind speed and average annual SST, and between annual average wind speed and annual water mass renewal rate at Bermuda ( 17), suggest that a decrease in the Bermuda coral ~14 C value is caused by increased mixing between surface and subsurface waters ( :S;400 m) during the formation of subtropical-mode water (18°C and 36.5 per mil salinity) (20) in late winter in the northern half of the Sargasso Sea.The 8 13 C (7) values for the Bermuda coral decreased by about 0.4 per mil from 1885 to 1955 (Fig. lB); this may represent evidence of excess C0 2 (from fossil fue l and biomass burning with a 8 13 C of -26 to -28 per mil) in the surface ocean.Combined with an additional decrease of 0.4 per mil from 1955 to 1983 (17), values decreased overall by 0.8 per mil over a 100-year period.The 8 13 C values of DIC in surface waters of the Pacific decreased by about 0.4 per mil from 1970 to 1990 (2 l ).Similarly, DIC 8 13 C in surface waters near Bermuda decreased by 0.018 ::!::: 0.005 per mil year-1 from 1984 ro 1993 (adjusted for interannual variation) (22).This value is close to the average decrease observed from 1955-1983 in the Bermuda coral of 0.014 ::!::: 0.003 per mil year-1 (17) .In a separate Bermuda coral, a decrease in 8 13 C of 0.5 per mil from 1880-1950 (0.007 per mil year-1 ) (23) was measured, which is similar to the decrease in the Bermuda coral that I observed of 0.006 per mil year-1 (Fig. IB).In contrast, the 8 13 C values of the Florida coral remained constant over the past 100 years (19).Coral 8 13 C should be used cautiously as a proxy for water mass DIC 8 13 C values, because physiological processes in the cora l polyp involve large isotopic fractionations (JO, 24,25).
The lack of a Suess Effect implies that some process other than gas exchange with the atmosphere is controlling the ~ 14 C values of the surface waters in the North Atlantic.Exchange of C0 2 between atmosphere and surface ocean has too long a time constant (5 to 10 years) to be responsible for the huge shifts in ~ 14 C values.Lateral advection of water is unlikely to have caused large 6 14 C shifts, because the ~ 14 C signature of the Gulf Stream is similar to that of Bermuda.One process that could account for the shifts is changes in vertical circulation.
To evaluate this effect, I used a multibox, isopycna l mixing model to estimate the ventilation rate (the water mass renewal rate, W) of the upper 700 m.This model is similar to that used to evaluate the W; in the Sargasso Sea from 1950 t0 1983 with post-bomb ~ 14 C values ( 17).The model has an atmospheric box (A), and annual ~ 14 C values of atmospheric C0 2 are set equal to values reported for tree rings (3).The rate constant of the output of 14 C0 2 from the sea sU1face to the air, k 11 is a function of the mixed layer depth Z ( 135 m), the LC0 2 in surface waters (2._12 mol of C0 2 m-3 ), and the C0 2 gas exchange rate 1 ( 18.9 mol of C0 2 m-2 year-1 ), which is a function of the gas exchange piston velocity VP (in meters per day), according to VP X 0.019 mM X 365 days year-1 (year-1 ) Z x :LC01 (1) where VP = WS X 0.9995 -3.47 (the average annual wind speed WS is 6.2 m s-1 ), the C0 2 concentration in seawater is 0.019 mM, and I = V 1 , x 0.019 mM x 365 days year-1 (26,27).The input rate of 14 C0 2 from the air to the sea surface, 1<_ 1 , depends on the partial pressure of C0 2 in the surface ocean (pC0 2 .) and that in the air (pC0 23 ) and on the 14 C concentration in the atmosphere at time t [A(t)], which includes the Suess Effect (3 ), and in surface waters [B(t)I where 0.983 is the isotopic fractionat ion factor when C0 2 is transferred from the atmosphere to surface waters, and pC0 2 .f The three surface ocean boxes are B (Bermuda), S (slope water entrained in tbe Sargasso Sea), and GS (Gulf Stream); 15% of the water in B is replaced with water from S and GS each year (with a ratio of 3/17 for S/GS) (28), and the same amount is lost from the sides of B to conserve mass.
where ~r = 0 .2year (O.l year in earlier model) and S; is W/W 1 , the relative change of the ventilation rates on each isopycnal (S 1 through S 7 are 1.0, 0.8, 0.65, 0.49, 0.36, 0.24, and 0.15) (31).When the forward model (Eq.3) was solved using a steadystate, pre-1 900~1 4 C value of-43.4 per mil, a constant annual value for W 1 (t), of 0.08 year-1 was obtained (32).Results of the forward calculation revealed that a constant W 1 (t) of 0.08 year-1 was roo low during several periods when the surface ~ 14 C values in the Bennuda corals were lower than about -48 per mil; they are 1889-1890, 1901-1903, 1907-1912, 1925-1926, 1929-1930, 1935-1938, and 1945-1948.Numerous variables were tested in the forward model to determine their effects on the B(t) record.An instantaneous reduction in the mixed layer depth from 135 to 50 m caused a 7-per-mil rise in ~ 14 C of B(t) in 2 years.A lso, if all mixing with subsurface isopycnals had ceased in the forward model, then the ~ 14 C of the surface B(t) would have increased 6 per mil in 4 years.Likewise, if wl had increased instantaneously by a factor of 6, then a 5-per-mil decrease of~ 14 C in surface box B(t) would have occurred in 1 year.
I then ran an inverse calculation to determine the W 1 record that was needed to reproduce the Bermuda   1901-1903, 1908-1912, 1926-1927, 1930-1931, 1935-37, and 1945-1948.When the Bermuda D. 14 C measurements were :::=::-50 per mil, W 1 values were > 1, which means that the water in box B was being replaced completely more than once per year (Fig. IC).The periodic pulses of high W 1 values in Fig. lC suggest that subsurface depths in the ocean were ventilated faster during these times.The W 1 values calculated for the early 1960s were also high (17), which suggests that these pulses of high ventilation occurred throughout the entire past century.Conversely, high D. 14   1915-1916, 1921-1924, 193 2-1933, and 1939-1941.Sensitivity analyses were conducted on the inverse model to ascertain the change required in each of the factors influencing the W 1 record.A change in the average annual WS by ± 12% (from 5.5 m s-1 to 7 .0m s-1 ) altered the C0 2 exchange rate I from 14.1to24.5 mol 111-2 year-1 .The peaks and valleys in the W 1 record occurred at the same time, however, and the W 1 values changed by <30%.The influence of the Gulf Stream D. 14 C time history was tested by running the inverse model with a constant GS value ( -4 7.4 per mil for 1880-1940).The change in the W 1 record was <5%.An average D. 14 C value (-43.4 per mil for 1885-1955) was used for B(t) in place of the actual record, and the resultant W 1 record was 0.08 ± 0.04 year-1 over the entire 1885-1955 period.These tests show that the Bermuda u 14 C variability is responsible for the large peaks in the W 1 record (Fig. lC).
The decadal time-scale decreases of the winter SST anomaly east of Newfoundland (33) match the pulses of high W 1 obtained here (Fig. lD).A cross-spectral analysis of these two records (between 1902 and 1953) reveals that the 9-year period is coherent to the 95% confidence level (the SST anomaly lags W 1 by 2 years).This decade timescale variability is superimposed on a gradual warming trend of SST from the 1920s to 1940 in the North Atlantic (33,34) and a decrease in the W 1 values (see Fig. lD).The record of the North Atlantic Oscillarion (NAO), which is the difference in winter sea-level air pressures between the high at Lisbon, Portugal, and the low at Stykkisholmur, Iceland, displays decadal shifts that are linked to changing wind patterns over the North Atlantic (35).A cross-spectral ana lysis of NAO values (35) and annual W 1 values between 1885 and 1983 ( 17) reveals rhat the 6-year period is coherent to the 95% confidence level (no lag).These correlations suggest a link between the pulses of high w]I which represent increased subtropical-mode water ventilation, and both low NAO values and lower SST values in the area northwest of the Sargasso Sea [and likely off the east coast of the United States (36)).
The absence of a predictable Suess Effect in the surface waters of the North Atlantic demonstrates rhe importance of circulation as opposed to gas exchange on the 14 C history of surface waters.The large range of D. 14 C values present in the North Atlantic during the pasr century shows char the vertical mixing in the upper few hundred meters of the water column can be rapid and variable and that it may be related to changes in North Atlantic surface properties and rhe NAO in general.These dramatic shifts of surface ocean A 14 C signatures represent the uncertainty (±200 apparent 14 C years) that needs to be incorporated into a reservoir age correction for the North Atlantic surface ocean [400 years (37)].
C measurements of banded corals from three sites in the North Atl~m tic Ocean and use the results ro calculate the water mass renewal (or ventilation ) rate in the North A tlantic over the past 100 years.Cores of Dit>loria sr:iigosa were collected from surface waters (at a depth of 5 m) near North Roc k (32°29' N, 64°41 'W; 14 km north of Bermuda), and Momasn•ea annularis corals were cored from Pickles Reef off Key Largo, 100 km south of Miami, and from The Rocks Reef (24°57' N, 80°33'W), 1 km south of Plantation Key in Florida.Biennial (2-year) bands were cut from rhe 1454 SCIENCE • VOL.275 • 7 MARCH 1997 • http://www.sciencemag.org

The 6 .
14 C values for pre-1955 coral bands from Bermuda and Florida varied widely through out the records (-34 to -64 per mil) (Fig.l A).The depletion of 6.14 C in rhe surface ocean as compared with 6. 14 C in acmospheric C0 2 (-2 co -25 per m il (3)1.reservoir effect (J 6), was maintained by m ixing of subsu1face waters with lower 14 C into rhe surface ocean.Bermuda 6. 14 C values before 1940 had a slightly higher average value (-43.9 ::!: 4.8 per mil; n = 33) chan chose from Florida (-47.4 :!: 4.5 per mi l; n = 25).T here were three times when the 6. 14 C values of rhe Florida coral were significantly lower than those of the Bermuda coral: 1890-1898, 1919-1924, and 1949-1952.Between 1947 and 1953, 6. 14 C values at th e two Florida sites averaged -58.2 ± 4.4 per mil (n = 5), which is indicative of a late Suess Effect (-1 1 per mil) and is comparable to a modelcalculated trend (see Fig. l A).Only two 6. 14 C values for the Bermuda coral were low (average -52.4 :!: 0.8 per m il between 1945

Fig. 1 .
Fig. 1. (A) A 14 C of biennial coral bands from Bermuda and the two Florida Keys sites [Pickles Reef (PR) and The Rocks ( TR3)) in the North Atlantic Ocean.Duplicate analyses are plotted as the weighted average A 14 C value.The uncertainty of year assignments for these corals is about ± 1 to 2 years by 1885 (73).The model calculations of A 1 •c in the mixed layer are alsoshown and were done with the use of bidecadal tree-ring (atmospheric) A14 C values and a box-diffusion model of the oceans, assuming constant atmospheric C0 2 concentration, constant biospheric reservoir sizes, and constant gas exchange rates(37).(BJ 313 C and 3 18 0 results from biennial coral samples from the Bermuda coral.Solid circles, 318 0; open circles, 313 C. Line indicates least-squares fit of 3' 3 C.Samples were crushed in methanol and roasted at 375°C in vacuo just before analysis.Samples were analyzed with a light-isotope ratio mass spectrometer (Micromass 602E) al Woods Hole Oceanographic Institution in the laboratory of L. Keigwin and have an uncertainty of ±0.1 per mil for both 313 Cand 3 18 0 .(C) Results of the inverse model that show W 1 in the Sargasso Sea between the surface box Band the shallowest subsurface box of cr 0 = 26.4 for the time period1885-1955.

Pulses of Rapid Ventilation in the North Atlantic Surface Ocean During the Past Century Ellen
Shells for EXAFS-lnteractive (B.Rupp, Lawrence Livermore National Laboratory) to calculate bond distances and coordination numbers.Unit cell parameters are available in R. W. Wyckoff, (29)observed coral ~ 14 C values for B and GS are the Bermuda and Florida (Pickles Reef) records, respectively (Fig.lA), and the pre-bomb ~ 14 C level for Sis set tO -90 per mil(29).The seven subsurface boxes D 1 contain 14 C concen- Fig. l C) contained seven periods when W 1 was greater than twice the steady-state value of 0.08 year-1 • These periods coincided with the low u 14 C values in the Bermuda coral record.They were 1889,