Subtropical North Atlantic Temperatures 60,000 to 30,000 Years Ago

See allHide authors and affiliations

Science  22 Oct 1999:
Vol. 286, Issue 5440, pp. 756-759
DOI: 10.1126/science.286.5440.756


A reconstruction of sea surface temperature based on alkenone unsaturation ratios in sediments of the Bermuda Rise provides a detailed record of subtropical climate from 60,000 to 30,000 years ago. Northern Sargasso Sea temperatures changed repeatedly by 2° to 5°C, covarying with high-latitude temperatures that were previously inferred from Greenland ice cores. The largest temperature increases were comparable in magnitude to the full glacial-Holocene warming at the site. Abrupt cold reversals of 3° to 5°C, lasting less than 250 years, occurred during the onset of two such events (Greenland interstadials 8 and 12), suggesting that the largest, most rapid warmings were especially unstable.

Annually dated records of isotope paleotemperature from Greenland ice cores depict a highly volatile climate during the last glacial period [80,000 to 10,000 years ago (ka)] (1). Many of the largest temperature excursions occurred from 60 to 30 ka during marine isotope stage (MIS) 3, an interval characterized by intermediate ice sheet size, high-latitude radiation receipts, and atmospheric CO2concentrations. Similar excursions are seen in faunal records of high-latitude sea surface temperature (SST) (2) and geochemical records of deep ocean ventilation (3,4), consistent with numerical modeling results showing a large dependence of high-latitude sea and air temperatures on the rate and mode of ocean thermohaline circulation (5). There are also indications of related SST change at lower latitudes (6–8), but these are primarily based on planktonic foraminiferal isotope records that may be influenced by factors other than temperature. The SST of the warm ocean is nonetheless expected to play a crucial role in amplifying and propagating climate change because the partial pressure of water vapor, an abundant and effective greenhouse gas, depends exponentially on temperature (9). Here, we present alkenone-derived SST records from Bermuda Rise sediments in the northwest Sargasso Sea and from high-deposition rate sites in the southwest Sargasso Sea in order to evaluate the temperature history of the subtropical Atlantic Ocean during MIS 3.

The Bermuda Rise is a sediment drift deposit northeast of the islands of Bermuda. Lateral sediment focusing within the North American Basin augments deposition at the site (10), so that late Quaternary sedimentation rates range from 10 to 200 cm/1000 years (1 ky) (11), some 5 to 100 times the open ocean average. As a result, Bermuda Rise sediments provide exceptional resolution in time. Core MD95-2036 (from 33°41.444′N, 57°34.548′W, at a water depth of 4462 m) is 52.7 m in length and contains sediments of Holocene through penultimate glacial (MIS 6) age (11). We determined SSTs by alkenone paleothermometry (12) in contiguous 1- or 2-cm intervals throughout the 12-m section of the core corresponding to MIS 3. Sedimentation rates averaged 30 cm/ky during the interval, so that single samples represent 33 to 67 years of deposition on average.

Lipids were extracted from 1 to 4 g of freeze-dried sediment with a pressurized fluid extractor, and alkenone abundances were quantified by gas chromatography with flame-ionization detection (13). Down-core results are presented in Fig. 1 in units of the alkenone unsaturation ratio (Uk 37) and as estimated SSTs from the regression relation of Prahl and others (12,14). Our semiautomated analytical procedure allows the routine analysis of 50 samples per week with an external precision of 0.0058 Uk 37 units (0.17°C) (15). Dispersion of duplicate measurements in the interval from 2750 to 3000 cm exceeds the external precision because of a contamination problem that was encountered and eliminated early in our study (16).

Figure 1

Alkenone-derived SSTs (red circles) and sediment lightness values (a proxy for the CaCO3 content of the sediment) (black lines) shown to depth in Bermuda Rise core MD95-2036, with an expanded view of results for MIS 3 (60 to 30 ka).Uk 37 values were converted to SST with the equationUk 37 = 0.034T + 0.039 (12, 14). The precision of the analysis (1σ error bar) is 0.00585Uk 37 units or 0.17°C. Scatter between 2750 and 3000 cm is the result of contamination by partially coeluting compounds (16). Warm interstadial events (numbered in blue) were identified on the basis of visual correlation to the GISP2 isotope temperature record (1). Two ice-rafted debris (IRD) maxima in Bermuda Rise sediment cores, corresponding to Heinrich events 4 and 5 (H4 and H5) (8), are also shown (vertical dashed lines).

Reconstructed SSTs ranged from ∼15.5° to 21°C during MIS 3 and are associated with changes in sediment lightness (Fig. 1), a proxy for the CaCO3 content of the sediment at this location that has been previously correlated to the Greenland record of paleotemperature (11). Millennium- to century-scale SST minima (stadials) were typically between 15.5° and 17°C, and maxima (interstadials) were between 19° and 21°C. The mean warming from stadial minima to interstadial maxima for 12 millennium-scale SST oscillations was 3.1°C (a range of 1.7° to 5.3°C). In addition, stadial minimum temperatures reached successively colder levels during MIS 3, from 17.3°C before interstadial 15 (IS-15) to 15.3°C before IS-5. The large SST changes documented in core MD95-2036 are consistent with ∼1–per mil millennial variations in planktonic foraminiferal δ18O measured in nearby core KNR31-GPC5 from the Bermuda Rise (8), although there are notable differences in the two proxy records that suggest either an important salinity influence on the planktonic δ18O (17) or changes in the season or depth of foraminiferal calcification.

Our temperature estimates are largely insensitive to the regression relation used to convertUk 37 values to SST. For example, a calibration from a recent global compilation of core-top and contemporary mean annual SSTs (18) yields results that are nearly identical to those we obtained using a relation (12) based on culture experiments (14). In addition, measurements of box-core samples spanning the past 2500 years of sedimentation at the Bermuda Rise yield an average SST of 21.8° ± 0.5°C (n = 49) with the culture relation (19). Considering that this interval includes the Little Ice Age cold excursion (20), the 2500-year average is close to the modern annual mean and production-weighted SSTs at 0 m of 22.8° and 22.5°C, respectively (21). Changes in the contribution of fine-grained sediment to the site have not influenced measured Uk 37 values (22).

Reconstructed millennium- and century-scale SST oscillations at the Bermuda Rise are unexpectedly large in light of climate proxy data (23) and numerical models (24) indicating 2° to 5°C of warming between full glacial and present-day or Holocene–average temperatures in the region. In order to determine whether such variations were a local response to the movement of an oceanographic front or were associated with more widespread temperature change, we also measured Uk 37ratios across selected events in cores from the Blake (KNR140-JPC27; from 30°01′N, 73°36′W, water depth of 3975 m) and Bahama (KNR31-GPC9; from 28°15′N, 74°26′W, water depth of 4758 m) outer ridges, in the southwestern Sargasso Sea (Fig. 2). SSTs rose abruptly by 2° to 2.5°C during the transitions into IS-8, IS-12, and IS-14 at both locations (25), equivalent to one-half the warming observed at the Bermuda Rise (Fig. 1). Absolute SSTs in the southwestern Sargasso Sea were higher than those at the Bermuda Rise throughout the studied intervals. At the Blake Outer Ridge, stadial cool episodes were warmer by 1° to 1.5°C, and at the Bahama Outer Ridge, the stadials were warmer by 2° to 3.5°C. In contrast, maximum interstadial temperatures at both southwestern locations were within 0.5° to 1.0°C of those at the Bermuda Rise for coeval events. The larger divergence of lower latitude SSTs from those at the Bermuda Rise during cold periods suggests that diminished poleward heat transport contributed to stadial cooling, an expected consequence of weakened thermohaline circulation (26). SST relations among these study sites thus support the interpretation of Bermuda Rise SSTs in terms of a regional climatic signal rather than a localized response to the movement of an oceanographic front.

Figure 2

Alkenone-derived SSTs (red circles) and CaCO3 content of the sediment (black line) for two sites in the southwestern Sargasso Sea. The CaCO3 content of the sediment was used to identify the position of interstadial events (blue numbers) (25). SSTs rose abruptly at the onset of IS-7 and IS-8 at the Blake Outer Ridge (core KNR31-GPC9; 28°15′N, 74°26′W) and at the onset of IS-11, IS-12, and IS-14 at the Bahama Outer Ridge (core KNR140-JPC27; 30°01′N, 73°36′W). The amplitude of warming was about half that at the Bermuda Rise in the northern Sargasso Sea.

To place SST results from core MD95-2036 on an estimated age scale, we correlated variations in sediment lightness with weight-percent CaCO3 variations in core KNR31-GPC5 (correlation coefficient r = 0.90), previously dated with radiocarbon and oxygen isotopic stage boundaries (3). However, this method produced ages for the SST events that were 2000 to 5000 years older than their apparent counterparts in Greenland paleotemperature (δ18O of ice, δ18Oice) records, probably due to radiocarbon dating errors associated with carbonate dissolution and poor calibration and to uncertainties in the SPECMAP age scale (27). We therefore developed an alternative age model by maximizing the correlation between Bermuda Rise SSTs and layer-counted variations in Greenland δ18Oice. Using 146 tie points, we obtained a correlation coefficient of 0.83 between the two series (28). This correlation results in a linear relation of amplitude between the two paleotemperature records (Fig. 3A); all major variations in the ice core are observed at the Bermuda Rise.

Figure 3

(A) Bermuda Rise SSTs (red circles) and central Greenland δ18Oice (blue diamonds) for MIS 3 on the GISP2 ice core time scale (28). The position of IRD peaks associated with Heinrich events 4 and 5 in the sediment core is shown for stratigraphic reference (dashed vertical lines). Cold reversals of 3° to 5°C occur during the onset of (B) IS-8 and (C) IS-12 in less than 250 years. (D) High instantaneous rates of sedimentation (black line) of 100 to 1000 cm/ky characterize cold stadial periods and the transitions into interstadials, and interstadial periods are characterized by lower sedimentation rates of 3 to 10 cm/ky. The sampling interval was 2 cm before 47.2 ky B.P. (dashed vertical line) and 1 cm afterward. The interval of time represented by each sample (green crosses) varies from ∼1 to 20 years during stadial periods to 100 to 300 years during interstadials.

Although this age model lacks absolute chronologic control, on the basis of the linear covariation of millennium- and century-scale features, we reason that it provides the best possible estimate of relative age. We know of no mechanism by which to delay the climate signal while preserving the linear relation of amplitude for both brief and long-lasting events (Fig. 3A). Furthermore, because the heat capacity and mixing time of the atmosphere are small with respect to those of the ocean, the atmospheric adjustment to such large oceanic changes is expected to be nearly instantaneous, on the order of days to years. Absolute uncertainty of the time scale is approximated by the error associated with counting annual layers in the Greenland Ice Sheet Project 2 (GISP2) ice core, which is 5% (29), or 1500 to 3000 years in MIS 3. Relative uncertainty of the SST chronology is 10 to 100 years during stadial cool episodes, when inferred resolution and sedimentation rates (Fig. 3D) are high, and 100 to 1000 years during interstadial warm periods, when inferred resolution and sedimentation rates are low (30).

The inferred amplitude lock of the SST and δ18Oice series (Fig. 3A) suggests that Bermuda Rise SST variations were one-third to one-half of Greenland air temperature variations, depending on the δ18Oice-temperature relation used to estimate isotopic paleotemperature (31). The SST record also contains abrupt features not yet identified in the ice core. The most important of these are large (3° to 5°C) oscillations that interrupt rapid warmings at the onset of IS-8 and IS-12 (Fig. 3, B and C). These events probably occurred entirely within the stadial-interstadial transitions marked by δ18Oice in GISP2 and Greenland Ice Core Project (GRIP) ice cores, a period of not more than 250 years (28). Estimated rates of sedimentation, as determined by correlation to GISP2, rose during these transitions (Fig. 3D), whereas associated SSTs sank to minima that were substantially colder than preceding stadial temperatures (Fig. 3B). We thus speculate that transitional cooling episodes were associated with enhanced meltwater and iceberg delivery of sediment to the North Atlantic Basin during Heinrich events 5 and 4 (at the onsets of IS-12 and IS-8, respectively) (2) and consequent meltwater suppression of North Atlantic thermohaline circulation (4, 32). A limited number of benthic foraminiferal Cd/Ca measurements from IS-8 in core MD95-2036 and additional Cd/Ca and δ13C measurements from correlative levels in core KNR31-GPC5 support diminished rates of North Atlantic Deep Water formation during Heinrich event 4 (8). Similarly abrupt climatic shifts and deep ocean changes have been previously inferred in association with this event in both the subpolar North Atlantic (2, 32) and the western equatorial Atlantic (7).

Our results demonstrate that large millennium- to century-scale changes in SST were not restricted to the polar and subpolar North Atlantic, but extended well into the subtropics. Indeed, rapid SST changes in the northern and southwestern Sargasso Sea (2° to 5°C and 1° to 2.5°C, respectively) were as large or larger than full glacial-Holocene mean annual SST differences reconstructed by the CLIMAP (Climate: Long-Range Investigation, Mapping, and Prediction) project (23) with foraminiferal transfer functions. They are also comparable to alkenone-derived SSTs from Bermuda Rise core KNR31-GPC5, indicating a deglacial temperature increase of 5.4°C (from 16.5° to 21.9°C) (33). Amplitudes of the largest MIS 3 SST events in the subtropical North Atlantic have a distribution similar to that observed in a number of coupled atmosphere-ocean models simulating SSTs for Last Glacial Maximum versus modern boundary conditions (24). The presence of such large SST variations over the warm ocean may help to explain observations of abrupt climate events at locations distant from the subpolar North Atlantic and Greenland [for example, (6,34)] through direct thermal forcing and the temperature–water vapor feedback.

  • * Present address: Department of Environmental Science, Barnard College, Columbia University, 3009 Broadway, New York, NY 10027, USA.


View Abstract

Navigate This Article