Introduction

In the North Atlantic, ocean thermocline waters along the Gulf Stream (GS) are reintroduced into the surface layer, and waters within and south of the GS are subducted into the thermocline in what is commonly referred to as Sub-Tropical Mode Water or 18° Water (EDW) formation. This process leads to vigorous mixing of waters above and within the thermocline in large volumes1 and is known to occur between 33°–40°N and 30°–75°W (ref. 2). By redistributing nutrients in the upper layers, EDW formation is believed to have a role in regulating the biogeochemical processes of the nutrient-limited subtropical gyre3,4 by supplying nutrients to the surface. In addition, EDW formation contributes as much as 3–10% of the global net annual draw-down of anthropogenic CO2 on intermediate time scales (10s–100s of years)5; thus, changes to the rates of formation could alter the oceanic carbon sink. However, our understanding of the long-term variability of these ocean dynamics is still limited6.

In the North Atlantic, EDW formation occurs as a result of vigorous surface ocean convection caused by rapid heat release from the ocean to the atmosphere, and perhaps also by local frontal instabilities within the GS6. The rate of mode water formation appears closely related to the phase of the North Atlantic Oscillation (NAO)2,7,8,9, the dominant mode of atmospheric variability in the region that is generally defined as the winter (December–March) atmospheric pressure difference between Iceland and the Azores10. In a negative phase of the NAO, the pressure difference weakens, slowing the speed of the atmospheric jet stream across the North Atlantic and increasing storminess off eastern North America. The increased storminess increases the air–sea heat exchange, thus increasing the rate of formation of EDW.

As new EDW is formed during the winter, the colder, denser waters at the surface mix down into the thermocline that includes 'older' (lower 14C) mode waters that have re-circulated around the subtropical gyre. When surface waters warm in spring, the mixed layer shoals and the water column re-stratifies, but some of the deeper 'older' thermocline water remains in the surface, effectively increasing the surface radiocarbon age. Thus, Δ14C (the per mil deviation of 14C/12C from that in nineteenth century wood) in the surface ocean can be used as a measure of ventilation (vertical mixing), because it represents the combination of dissolved inorganic carbon exchanging with atmospheric CO2 and surface water mixing with deeper layers that have been isolated from atmospheric CO2.

Previously in this region, past changes in ventilation were reconstructed back to 1880 from biennial, high-precision Δ14C measurements of a Bermuda coral taken near North Rock along the northern reef11 (Fig. 1). The study revealed significant variability in Δ14C (−50 to −35‰; Fig. 2) correlated to the NAO, with lower surface Δ14C occurring during periods of negative NAO phase associated with enhanced EDW ventilation. The record also demonstrated no 14C Suess Effect, and a significant 13C Suess Effect (Fig. 2), which is the decrease in δ13C in the environment beginning in the late 1800s, and originating mostly from the combustion of fossil fuel and biomass whose δ13C values are much lower than that in the atmosphere. These combined results indicate that waters surrounding Bermuda are influenced by both vertical ocean mixing and atmospheric exchange. In this study, we aimed to extend the record into the end of the Little Ice Age to evaluate changes in mixing rates at times of cooler mean hemispheric temperature and lower NAO variability. We report biennial, high precision Δ14C time-series measurements from an annually banded coral from 1780 to 1940, collected from the Sargasso Sea along the south shore of Bermuda (64°W, 32°N; Fig. 1). The Δ14C time history of atmospheric CO2 is well documented12 as is the Δ14C time history of GS water11 that supplies surface water to the Sargasso Sea, allowing us to evaluate past changes in ventilation.

Figure 1: Map of Bermuda within the North Atlantic subtropical gyre.
figure 1

Approximate locations of sampling sites North Rock and John Smith's Bay are indicated. The island of Bermuda is characterised by a large lagoon surrounded by a barrier reef to the north and a narrow shelf with barrier reef to the south. Inset shows Bermuda relative to an approximate region of 18° mode water (EDW) formation and the eastern coast of the United States. Map based upon figures in Palter et al.4 and Bates et al.13

Full size image
Figure 2: Carbon isotope records from Bermuda and regional climate records.
figure 2

(a) Records of coral δ13C values from the north (dashed, Druffel11) and south (solid, this study) shore of Bermuda. (b) Coral Δ14C records for the north (dashed, Druffel11) and south (solid, this study) shore for the entire 200 years. Note increased variability in the north-shore coral. Measurement error of ±2.5‰ (1σ). (c) Instrumental record of the NAO annually (shaded) and 5-year running mean (solid) from Hurrel10 including the 0-index line (dashed). (d) A multi-proxy record of Northern hemisphere surface temperature anomalies (STA) including mean annual (shaded) and 5-year running mean (solid) from Jones et al.16 Shaded area represents the Dalton Minimum in solar activity.

Full size image

Results

Carbon isotope records from Bermuda

The coral data show similar decadal to centennial scale behaviour as the atmospheric record of Δ14C (Fig. 3). Biennial Δ14C coral measurements from 1780 to 1925 have no apparent trend; though measurements from 1926 to 1940 have a substantial decreasing trend with a low value of −50‰ (range before 1926 is −44 to −35‰), which we attribute to the Suess Effect. In addition, the coral Δ14C record has a significant linear correlation to the record of atmospheric Δ14C with a 10-year lag14, matching the time scale over which the surface ocean and atmosphere CO2 mix, (r=0.62, P≪0.0001, n=72). Cross-spectral analysis of the two records confirms that they are significantly correlated for frequencies lower than 0.12 cycles per year (8-year period and longer) at a 95% confidence interval (Fig. 3b,c). The agreement between the coral and atmosphere Δ14C records implies that surface waters south of Bermuda are influenced predominantly by atmospheric exchange.

Figure 3: Influences on and variability of oceanic Δ14C.
figure 3

(a) Record of atmospheric Δ14C (Stuiver and Quay12; shaded) and the south-shore coral Δ14C (solid). The correlation coefficient between the two records is 0.62 (P≪0.0001), with the coral record lagging the atmospheric record by 10 years. (b) Cross-correlation analysis of the south-shore coral and atmospheric Δ14C reconstructions (solid). Confidence levels (80%) are shown by the dashed line and 95% confidence level by dotted line. (c) Phase lag between the south-shore coral and atmospheric Δ14C reconstructions. (d) A 5-box Hanning filter was applied to the Δ14C records to isolate all frequencies higher than 10 years per cycle. The high-frequency record is plotted versus time for this study (solid) and Druffel11 (shaded). Measurement error of ±2.5‰ is shown for each data point. (e) The absolute (Abs.) value of the difference between each sequential (Seq.) sample pair in the high-frequency (HF) record is plotted versus time for this study (solid dots) and Druffel11 (shaded crosses). Propagated measurement error of ~3.5‰ (1σ) is shown by the solid dashed line and 2σ (~7.0‰) is shown by the shaded dashed line. Relatively few points show statistical independence (that is, above the measurement error) particularly for this study. The 5-year running mean of the instrumental record of the NAO index is shown in red10.

Full size image

Similarly, the coral δ13C data from this study exhibit a 13C Suess Effect with values decreasing 0.013‰ per year (±0.002) over 1885–1978 (Fig. 2), in agreement with an average of coral records from around the globe15. The coral δ13C data show a gradual decline early in this period, with a much sharper decline from the 1960s onward, consistent with many corals in the Atlantic Basin15 (Fig. 2).

Δ14C variability North and South of Bermuda

This study from south Bermuda shows less Δ14C variability than the North Rock coral record previously reported11, with values from 1880 to 1910 ranging from −44 to −35‰ versus −50 to −35‰ for the southern and northern corals, respectively (Fig. 2). The south-shore coral has approximately 50‰ less variability (results of F-test for variance, P=0.06, n=27) than the North Rock coral11, which shows abrupt negative excursions interpreted as times of increased ventilation and injection of older thermocline waters into the surface layer. The contrasting signals between the corals indicate that the two locations are experiencing different apparent rates of ocean mixing.

Previous studies document that south-shore corals more accurately represent Hydrostation S, a 50-year hydrographic station located 25 km southeast of Bermuda (Fig. 1), than do corals from North Rock17,18. Differences between the north and south have been found in sea surface temperature (SST) records generated from coral Sr/Ca that are attributed to differences in coastal circulation19. Studies from the north reef have found agreement between seasonal variations in coral Sr/Ca and SST data from Hydrostation S; however, there is no relationship between North Rock Sr/Ca and SST at interannual timescales (that is, comparing data for wintertime only)19,20. This indicates that the Sr/Ca to SST correlation is driven by the seasonal cycle that occurs each year in both records rather than the North Rock Sr/Ca accurately capturing inter-annual ocean conditions in the southern and central Sargasso Sea. In contrast, the Sr/Ca record from the south-shore coral used here shows strong relationships to Hydrostation S SST at both annual (seasonal cycle) and interannual (wintertime) timescales17,18 (Fig. 4). In addition, a strong correlation has been established between south-shore coral wintertime Sr/Ca and the NAO at varying frequencies21. These results indicate that the south-shore coral more accurately reflects conditions at Hydrostation S, and thus the open ocean conditions in the southern and central Sargasso Sea. The difference between the response of Δ14C to NAO variability at the two locations likely lies in enhanced local upwelling along the north reef because of the trade winds, which shift north to Bermuda during periods of negative NAO. Simultaneously, water formed due to enhanced EDW formation to the north of Bermuda will be advected southward, forming a large wedge of significantly 'older' water just below the seasonal thermocline around Bermuda. These subsurface waters are therefore upwelled along the north but not the south shore, reducing only north Bermuda Δ14C during negative NAO phases. This Δ14C signal will be limited to the waters north of Bermuda as a function of local topography, rather than occurring throughout the southern and central Sargasso Sea.

Figure 4: South-shore coral Sr/Ca and Hydrostation S SST from 1976 to 1997.
figure 4

The data is shown at both monthly for the full seasonal cycle (top two panels) and winter average (December–March) (bottom two panels) timescales and is shown versus time and linearly. (a) Monthly Sr/Ca (orange) and Hydrostation S SST (blue) versus time, (b) monthly Sr/Ca and Hydrostation S SST linearly regressed (r2=0.86, P=0.0001). (c) wintertime Sr/Ca (orange) and Hydrostation S SST (blue) versus time, (d) winter average Sr/Ca and Hydrostation S SST linearly regressed (r2=0.51, P=0.0004). Figure adapted from Goodkin et al.17

Full size image

Discussion

The agreement with atmospheric Δ14C and Hydrostation S, and the lack of negative excursions in Δ14C in the south-shore record brings into question whether there has been a detectable change to the mixing rate of surface and thermocline waters over the past 200 years. To investigate this, we employed a 5-box Hanning filter to remove all frequencies lower than 0.10 cycles per year (>10-year period), when the Δ14C is clearly driven by atmospheric exchange. With few exceptions, the difference between individual pairs of sequential biennial samples does not exceed the measurement error, indicating that high-frequency changes in Δ14C due to variable mixing rates, if present, are likely small and not well resolved from the coral data (Fig. 3d,e). That is, a measurable rate of change to vertical mixing is not present within the precision limits of the 14C data. Between 1900 and 1920, there is a period when both this new coral record from the south shore and the earlier record from North Rock exhibit multiple sample pairs with significant changes to Δ14C above measurement error. These data correspond to a time of extended positive NAO with a short, but sharp, negative excursion around 1915, when changes to NAO-driven mode water formation rates would be high. However, cross-correlation analysis between the entire record of the raw Δ14C values presented here and the instrumental record of the NAO10 shows no significant correlation at high or low frequencies. Low Δ14C variability along the south shore is also seen during the end of the Little Ice Age from 1800 to 1870, when SSTs are low22 and NAO variability is weak21. In summary, surface waters to the south of the Bermuda exhibit a significant atmospheric 14C imprint and minimal changes through time in the mixing rate with older subsurface waters, despite significant changes to NAO behaviour21 and mean hemispheric climate16 (Fig. 2c,d).

Time scales for horizontal circulation in the Sargasso Sea, the tight recirculation component of the anticyclonic subtropical gyre, are believed to be on the order of 5 years1,23, indicating that surface waters are relatively well mixed over the sampling resolution of the coral records (2 years). Using transient tracers, Jenkins24 determined that advection along isopycnals contributes significantly to thermocline ventilation within the Sargasso Sea. Studies evaluating water mass renewal rates in the Sargasso Sea yield tracer-based renewal time scales for the upper thermocline from a few years to a few decades, with slower renewal rates for the deeper isopycnal layers11,23,25,26.

To calculate the water mass renewal rate within the Sargasso Sea, we employ a 5-box model based upon the model previously presented in Druffel11 and similar to Druffel25. The five boxes are: the surface Sargasso Sea as represented by Bermuda (B), the atmosphere (A), slope water (S), GS water and subsurface water (D; Fig. 5). The rate of change of the surface reservoir for 14C at B involves the balance of the air–sea gas exchange, and lateral and vertical seawater flux, which can be described as:

Figure 5: Schematic showing the interactions of the 5-box model.
figure 5

Boxes include the atmosphere (A), Surface Sargasso Sea (B), Subsurface water (D), the Gulf Stream (GS) and Slope water (S).

Full size image

where dB/dt is the time rate of change in Sargasso Sea surface layer 14C, lateralnetflux is the combined net exchange with slope water (S) and the GS, and verticalnetflux is the net exchange due to exchange with the thermocline (D). This can be written more explicitly as:

where 0.983 is the isotopic fractionation factor when 14CO2 is transferred from the atmosphere to surface waters27, and A, B, GS, S and D are Δ14C values for the respective reservoirs based on 14C/12C ratio measured on environmental proxies. Measured coral Δ14C from Bermuda (this study) and from Florida11 (Pickles Reef) are used for B and GS, respectively. The pre-bomb Δ14C for S is set to −71‰28,29,30,31 (http://calib.qub.ac.uk/marine). It is important to note that both the GS and S water are sources of low Δ14C to B before bomb testing, but GS during the peak years of the bomb era (1960s) is found to be higher than B25. The atmospheric box's (A) Δ14C values are set equal to tree-ring values12. The Δ14C for D is −50.6‰, a weighted average of pre-bomb 14C data over the thermocline32. The exchange rates between each box and the surface box taken with respect to the volume of the surface box are k−1,GS,S,D; kd is unknown; kGS and kS are set at 0.1275 and 0.0225 per year, respectively26. Atmospheric flux is described as:

where CO2 (aq) is dissolved CO2 (0.019 mol m−3), ΣCO2 is the total dissolved inorganic carbon (or total CO2) in surface waters (2.12 mol m−3), Z is the annual-average mixed layer depth of the surface ocean (taken as 135 m in this region), and Vp is a function of wind speed (WS) such that:

from Jenkins33 and Roether34 with an average annual wind speed of 6.2 m s−1. Assuming a pre-industrial steady state for equation (2) allows us to solve for kd:

Average values for 1890–1910 14C levels in the model reservoirs (R/R0, ratio of sample 14C/12C to reference R0) are B=0.9599 (−40.1‰), A=0.9959 (−4.1‰), GS=0.9512 (−48.8‰), S=0.9290 (−71.0‰) and D=0.9494 (−50.6‰). Therefore, we find kd to be 0.16 per year or 16% of the surface box volume per year. An error estimate for the exchange rate kd of ±0.10 (1 sigma) is found by error propagation using standard methods35. For B, A and GS, we use the s.e. calculated of measurements made between 1890 and 1910. As B and A are known to co-vary, the error was propagated using a combined error for (A/0.983-B). For S and D, we assume an s.e. to be the average of that for GS and B (0.8‰). The error estimate for kd is likely smaller than the true error, because it neglects error terms in the exchange rates.

The volume of the Sargasso Sea surface layer is approximately 9.5×1014 m3 assuming a mixed layer depth11 of 135 m. If the ventilation rate is 16% per year, then the annual amount of water exchanged would be 1.6×1014 m3 per year or approximately 5 Sv of continuous flow over a period of 1 year (Svy; 1 Sv=106 m3 s−1; 1 Sv flow for 1 year (Svy) is equal to 3.154×1013 m3). These estimates are consistent with recent estimates of net annual EDW formation of 4.6 Svy36.

The limited interannual variability in the south-shore coral Δ14C record can be used to place a bound on water mass renewal rate variability as well. With a measurement error of ±2.5‰ (1σ), the difference between two adjacent samples would need to exceed ±3.5‰ (1σ; assuming independent data error summed in quadrature) to clearly identify a geophysical signal. Relative to surface water Δ14C, the depletion of preindustrial thermocline Δ14C is estimated to be ~10‰, as thermocline Δ14C is estimated to be −51‰32 and the south-shore coral Δ14C is on average −40‰. For biennial sampling, this translates into an upper bound on observable changes in water mass renewal of ~0.2 per year or 20%. To detect such a change, the background replacement rate of 16% per year would have to double. Sustained exchange perturbations over multi-annual to decadal time scales are constrained within a correspondingly smaller range of water mass renewal rates. The strong coherence and similar amplitude of atmosphere and surface water Δ14C variations on decadal time scales leaves little room for substantial low-frequency variability in surface water–thermocline exchange, assuming that exchange with other low Δ14C reservoirs, such as the GS and slope water, do not co-vary and cancel out surface water Δ14C variations.

Despite previous findings11, variations in atmospheric Δ14C appear to be the controlling factor for Δ14C of surface waters in the central and southern Sargasso Sea. The south-shore coral, representative of conditions in the Hydrostation S region, demonstrates that ventilation rates in the central and southern Sargasso Sea have not been highly variable over the past 200 years. Therefore, although ventilation is certainly supplying low Δ14C water to the surface, the amount supplied per year has not changed within a factor of 2 (upper bound) during that time. Although the magnitude of EDW formation in the northern Sargasso Sea is known to co-vary with intensity and phase of the NAO10, we show that ventilation in the central and southern Sargasso Sea has not varied, neither during periods of large changes in NAO intensity nor during the colder conditions at the end of the LIA. Therefore, if a significant portion of anthropogenic carbon is being absorbed into the ocean interior in this region, it does not appear to be returning to the surface at a variable rate with time. Overall, the physical and biogeochemical processes currently established in the southern and central Sargasso Sea5, including rates of heat exchange, carbon sequestration and primary production due to nutrients supplied by ventilation, appear to be more stable over longer time scales than previously estimated, even in the face of changing climatic conditions.

Methods

Coral sub-sampling and analysis

Corals of Diploria labyrinthiformis were collected from surface waters (at a depth of 16 m) south of Johns Smith Bay, Bermuda (Fig. 1). Biennial (2-year) samples were identified from X-radiographs and cut using a diamond wire saw. The Δ14C measurements were conducted using accelerator mass spectrometry at the University of California, Irvine. The accelerator mass spectrometry measurements have an uncertainty of ±2.5‰. The δ13C measurements were made on parallel samples using a Finnigan MAT 253 coupled to a Kiel carbonate device at Woods Hole Oceanographic Institution, with an uncertainty of ±0.03‰. The Δ14C values were then used to compare water mass renewal rates in the Sargasso Sea, using an inverse model11,25.

Additional information

How to cite this article: Goodkin, N. F. et al. Two centuries of limited variability in subtropical North Atlantic thermocline ventilation. Nat. Commun. 3:803 doi: 10.1038/ncomms1811 (2012).