Carbon isotope evidence for a northern source of deep water in the glacial western North Atlantic

Significance

Understanding glacial ocean circulation is tied to understanding heat transport in the North Atlantic, growth and maintenance of ice sheets, and atmospheric CO₂ content. The proxy data are sparse, but, for decades, it has been thought that the North Atlantic did not produce deep water during the Last Glacial Maximum, unlike today. New C isotope results from a core at 5 km in the western North Atlantic indicate a young, glacial deep water may have been formed in the surface Labrador Sea by freezing and brine rejection. These data call for a revised circulation scheme where southern source water centered at 4.2 km would have been sandwiched between glacial North Atlantic intermediate water and the new dense bottom water.

Abstract

The prevailing view of western Atlantic hydrography during the Last Glacial Maximum (LGM) calls for transport and intermixing of deep southern and intermediate northern end members. However, δ¹³C and Δ¹⁴C results on foraminifera from a sediment core at 5.0 km in the northern subtropics show that there may have also been a northern source of relatively young, very dense, nutrient-depleted water during the LGM (18 ky to 21 ky ago). These results, when integrated with data from other western North Atlantic locations, indicate that the ocean was poorly ventilated at 4.2 km, with better ventilation above and below that depth. If this is a signal of water mass source and not nutrient storage, it would indicate that a previously unrecognized deep water end member originated along the western margin of the Labrador Sea, analogous to dense water formation today around Antarctica and in the Okhotsk Sea.

Over the past 30 y, paleoceanographers have learned that during the Last Glacial Maximum (LGM), North Atlantic Deep Water (NADW) shoaled and was replaced by Antarctic Bottom Water (AABW) that was transported from the south (1, 2). This reorganization affected climate through distribution of heat, salt, and CO₂. Carbon isotope results from a new sediment core at 5,010 m depth near Corner Rise (CR) Seamounts in the Sargasso Sea (Fig. 1) are the deepest paleo-data from the western North Atlantic, and they define a new mixing line between an LGM southern source of bottom water at 4.2 km and a young nutrient depleted northern source at 5.0 km.

Fig. 1.

Western North Atlantic locations discussed in this paper (Table 1), and details of the CR core location in the western North Atlantic. (A) General location of the CR seamounts in small box. For other core locations: BBOR and CS, core locations along Blake and Bahama Outer Ridges and Carolina Slope (32); BR, northeast corner of BR; HAP, Hatteras Abyssal Plain; LF, Laurentian Fan; MAR, Mid Atlantic Ridge. Red dots mark GEOSECS stations. Green line shows the historical position of the north wall of the Gulf Stream. (B) Bathymetry of the CR region, with a box showing the expanded view of core 17GGC location on (C) an abyssal hill. (D) A 3.5-kHz profile across the location of gravity core KNR197/10 17GGC reveals about 50 m of sediment draped on the north side of the hill, suggesting deposition from a southward-flowing benthic nepheloid layer.

Table 1.

Core locations in the western North Atlantic

Core name	Location abbreviation	Depth, m	Latitude	Longitude
KNR178 GGC-2	HAP	3,927	36°07.21'N	72°17.52'W
KNR197/10 CDH-10	MAR	2,955	37°11.998'N	34°15.903'W
KNR197/10 CDH-42	LF	3,870	43°27.871'N	54°45.344'W
KNR197/10 GGC17	CR	5,010	36°24.3'N	48°32.39'W
HU89003 PC8	BR	4,425	33°41.2'N	57°36.8'W
KNR140 JPC12	BBOR	4,260	29°04.446'N	72°53.933'W

Corner Rise Results

On R/V Knorr cruise 197/10, we surveyed the northern part of the CR region and cored the flank of an abyssal hill where 3.5-kHz sonar data show at least 50 m of sediment drape (Fig. 1D). Sediment collected from 5 km on CR by giant gravity core (GGC) 17 is very similar to the drift deposit at 4.4 km to 4.6 km on nearby Bermuda Rise (BR), but with more shells of the benthic foraminifer Cibicidoides (Fig. S1). Together with ice-rafted lithic particles, this genus reached peak abundance during the Heinrich Stadial 1 (HS1) interval on CR (Fig. 2B). Chronology of 17GGC is based on 11 accelerator mass spectrometer (AMS) ¹⁴C dates on the planktonic foraminifer Globorotalia inflata (Dataset S2), and linear interpolation assuming zero age at the core top. Ages are calibrated years before present, with only the standard 400-y reservoir correction [CALIB calibration program 6.1 (3)]. In general, sediment accumulated at 5 cm⋅ky⁻¹ to 10 cm⋅ky⁻¹, with higher sedimentation rates during a brief interval at the beginning of the Younger Dryas (YD; 40 cm⋅ky⁻¹) and during the 3,000 y preceding the LGM (23 cm⋅ky⁻¹) (Fig. 2C). The high sedimentation rate intervals may simply coincide with higher fluxes of glacial sediment to the deep ocean before events of cold climate, but we cannot exclude possible changes to the deep, southern recirculating gyre associated with the Gulf Stream (4). Absence of a sedimentation rate peak associated with the extreme cooling of HS1 may reflect remoteness of the site from the sources of Heinrich icebergs, but it is also consistent with minimal production of northern-source deep waters (2) and reduced sediment transport at that time.

Fig. S1.

Examples of the unknown species of Cibicidoides that are common in our CR samples. These are from an LGM sample (KNR197/10 17GGC, 176 cm to 178 cm). They are plano-convex, but without the calcified rim that forms a pseudokeel on Cibicidoides wuellerstorfi, without the pinched margin that is characteristic of Cibicidoides pachyderma, and without the rounder outline that is typical of Cibicidoides kullenbergi.

Fig. 2.

Analytical results from core KNR197/10 17GGC near CR. (A) Important features in the stable isotope data include δ¹³C of near 0‰ during the LGM (solid diamonds), the delayed increase of δ¹³C (at ∼13.5 ka) relative to the ∼14.5 ka onset of Bolling/Allerod (B/A) warming, and the postglacial decrease in δ¹⁸O (open diamonds) that began at ∼17 ka. Asterisks mark radiocarbon age control. (B) Abundance of Cibicidoides (open circles) and IRD grains (solid circles) (both >150 μm). The age difference between benthic and planktonic foraminifer (¹⁴C_B-P) is only 1,300 y at 15 ka and 1,200 y during the LGM. (C) Maximum rates of sedimentation occurred at the beginning of both the LGM and the YD. The sedimentation rate at >24 ka is constrained by an AMS ¹⁴C date at 28 ka. Data are compiled in Dataset S1 and are also shown vs. depth in Fig. S2.

Fig. S2.

(A) Stable isotope results on Cibicidoides. (B) Abundance of Cibicidoides (blue circles) and ice-rafted grains >150 microns (black diamonds). Data as in Fig. 2, but presented as a function of depth in core. Methods used in this study are identical to those in ref. 34.

The oxygen isotope history of 17GGC exhibits variability that is typical of the past 25 ky (Fig. 2A). The δ¹⁸O of Cibicidoides has a glacial−interglacial range of ∼2‰, about double the estimated ice volume effect (5), indicating local temperature and δ¹⁸O of water (or salinity) influences. Cibicidoides δ¹⁸O increased slightly into the LGM (∼18–21 ka), with mean LGM values of 4.34 ± 0.06‰ (n = 9). The δ¹⁸O decrease began between the sample dated to 17.0 ka and the next underlying sample interpolated to be 17.3 ka, and continued to the Holocene in steps centered on about 12 and 15 ka. These are the familiar Terminations 1A and 1B (6). The δ¹³C of Cibicidoides was close to zero during the LGM (0.06 ± 0.10‰), but, during HS1, it decreased by about 0.5‰ and stayed low until ∼13.5 ka, well into the Allerod. That half-permil decrease is comparable to the difference in δ¹³C of ΣCO₂ between AABW and NADW in the modern ocean (7). Following the minimum in Cibicidoides δ¹³C, values increased late in the Bolling/Allerod epoch, decreased into the YD, and increased again in the Holocene. AMS ¹⁴C dates on Cibicidoides at 15.0 and 20.4 ka are, respectively, 1,300 and 1,200 ¹⁴C years older than the G. inflata in the same samples.

Previous Work

Recently, as new core sites with higher-resolution sampling and more precise dating have become available, studies have identified temporal, spatial, and depth variability in stable isotope ratios of Cibicidoides that result from differences in bottom water properties during the LGM and deglaciation (8–11). These studies demonstrate it is possible to distinguish temperature and salinity effects on δ¹⁸O to show evolving bottom water temperature and salinity at different times and places. For example, it was reported that δ¹⁸O may have decreased earlier in the North Atlantic (17 ka on the Iberian Margin) than in the southwest Atlantic (15 ka) for sites at >3 km, and the decrease was larger (0.6‰) than can be accounted for by the secular change due to ice sheet melting (9). The decrease in δ¹⁸O at 17 ka is confirmed at CR, although the lead over South Atlantic δ¹⁸O is not significant because of uncertainty in the chronology there (9). However, the δ¹⁸O and Mg/Ca evidence for early freshening at 3.15 km along the Iberian Margin (12) suggests that some fraction of bottom waters at >3 km must have been generated by glacial melting and brine rejection during sea ice formation at a freshened high northern latitude surface, not the Southern Ocean (9). Before the present study, the early deglacial decrease in Cibicidoides δ¹⁸O had not been noted in the northwestern (NW) Atlantic because that genus is not consistently present in cores from its western margin.

Synthesis of Western North Atlantic C Isotopes

Although the δ¹⁸O on CR is typical of deep North Atlantic sites during the LGM, δ¹³C is higher than expected. This difference is seen by comparing the CR data to Cibicidoides data from other sites in the NW Atlantic in a depth profile for the LGM (Fig. 3 and Fig. S3). However, this compilation assumes little difference in properties between eastern and western locations in the NW Atlantic, which we can only support for paired analyses at about 2 and 3 km (Fig. S4). We interpret C isotopes in benthic foraminifera as quasi-conservative tracers for paleocirculation (5), although the data could reflect, in part, remineralization of biogenic C (13). The most recent analysis by Gebbie et al. (14) indicates that remineralization is less significant than previously thought, in essential agreement with Marchal and Curry (15), who estimated that organic matter remineralization makes little contribution today to the δ¹³C of ΣCO₂ in the deep (>1 km) Atlantic.

Fig. 3.

Compilation of LGM ventilation data from subtropical western North Atlantic core sites (16, 34) with the new data presented here (Dataset S3). (A) Stable isotope profiles of Cibicidoides δ¹³C (filled red circles) and δ¹⁸O (open black circles) on Carolina Slope, Bahama Outer Ridge (BOR), Blake Ridge, and BR. Cibicidoides are very rare during the LGM along the western margin, so, in most cases, the δ¹³C are single measurements where a few individuals were found. Otherwise, error bars are a nominal ±1σ. Multiple analyses were possible on BR, but the large scatter in Cibicidoides δ¹³C may represent greater influence of bioturbation and/or the presence of low δ¹³C in fluff layers where diatoms are preserved. Because the samples for stable isotope data were not dated, the curves represent average conditions for the LGM (18−21 ka). (B) Profile of the difference between benthic and planktonic (B-P) conventional ¹⁴C ages from NW Atlantic sediment cores during the LGM. Note similar trends in the two C isotope data sets. It is argued here that the reversal in the trends of δ¹³C and ¹⁴C_B-P below 4.2 km is evidence for a young source of North Atlantic bottom water during the LGM. Figs. S4−S7 put the new ¹⁴C_B-P data in a stratigraphic framework.

Fig. S3.

Comparison of δ¹³C results on Cibicidoides from CR with those from other deep sites in the subtropical North Atlantic. (A) Data are a work in progress and are discussed in Fig. S4. (B) Data from Skinner and Shackleton (12) display the typical increase in δ¹³C associated with the transition from HS1 to the Bolling/Allerod. (C) BR (GGC-1 in red) and (D) CR δ¹³C each show a late increase compared with the Iberian Margin.

Fig. S4.

Composite of Cibicidoides data from the west flank of the Mid-Atlantic Ridge (MAR) at 2,955 m water depth. Based on color changes, MC-9 was patched to CDH-10 from cruise KNR197/10 at 30 cm. This correlation indicates that the 29-m piston core (CDH, named for the late Dr. Charles D. Hollister, the pioneer of long coring) bypassed the upper 20 cm of seafloor. Cibicidoides data are based on C. kullenbergi, C. wuellerstorfi, or a combination of the two. The sedimentation rate is only about 4 cm⋅ky⁻¹, assuming the end of the LGM (shaded) occurred at 19 ka. For n = 5, average LGM δ¹⁸O = 4.62 ± 0.09‰, and average δ¹³C = 0.36 ± 0.13‰. These average data fall close to the data from Blake Ridge at 3 km (Fig. 3), indicating little δ¹³C contrast between the eastern and western sides of the western basin during the LGM. Likewise, at core KNR197/10 GGC5, from 2,127 m on the MAR, our three LGM measurements of δ¹⁸O (4.31± 0.05 ‰) and δ¹³C (1.25± 0.04 ‰) are very similar to results from Blake Ridge and Carolina Slope at similar depths (Fig. 3).

From Carolina Slope sites (1 km to 2 km water depth), δ¹³C steadily decreased down Blake Ridge to 4.2 km (Fig. 3A). Below that depth, the trend reversed, with higher δ¹³C at 4.5 km on BR, 4.7 km on Bahama Outer Ridge, and 5.0 km on CR. The large uncertainty in δ¹³C from BR is most likely related to the high accumulation rate of organic matter as revealed by preserved diatoms and bivalves (16) and the possible effects of organic matter oxidation and bioturbation on δ¹³C. The slight decreasing trend in δ¹⁸O below 3 km may not be significant, although it would be consistent with the Iberian Margin evidence for freshening (12). However, the δ¹³C reversal below 4.2 km is large and suggests the largest fraction of southern source (low δ¹³C) water was centered on 4.2 km, with increased mixing of a relatively high δ¹³C source below that depth.

In addition to δ¹³C, other geochemical proxies for deep ventilation and circulation during the LGM point to a blend of southern and northern source waters in the NW Atlantic. For example, Pa/Th results (2) and Nd results (17) from BR consistently show that the glacial mixture of bottom water must have contained a northern component. This finding is at variance with the prevailing view that there was little or no northern deep water present in the glacial North Atlantic (18, 19), but a recent synthesis of Nd isotopes in the western Atlantic has also been interpreted as evidence for a glacial source of NADW (20). Furthermore, Böhm et al. (21) found that some interstadials and the last interglacial on BR had lower εNd than in the modern ocean. This very low εNd could indicate there were both warm and cold climate processes that formed deep water in the North Atlantic that are inactive today.

Ventilation estimates based on the difference in conventional ¹⁴C age between pairs of planktonic and benthic foraminifera (or bivalves) (¹⁴C_B-P) during the LGM are consistent with the δ¹³C data (Fig. 3B). When the new CR data are added to previously published results from the NW Atlantic (16), and new results from the Laurentian Fan, Hatteras Abyssal Plain, and Blake Ridge (Figs. S5–S7, respectively), a similar pattern emerges that is interpreted here as decreasing ventilation down to 4.2 km, and increasing ventilation below that. McManus et al. (2) suggested that northern source export during the LGM was at intermediate depths over BR (glacial North Atlantic intermediate water), a result that was supported by later data (22). Our new CR results point to enhanced ventilation at least as deep as 5,010 m. Furthermore, lowest δ¹³C and highest ¹⁴C_B-P at 4.2 km is consistent with grain size results in three depth transects of Blake Ridge cores that indicate the fast-flowing deep western boundary current of modern circulation (3 km to 4 km) deepened to >4 km during the LGM (23).

Fig. S5.

Abundance of the benthic foraminifera Uvigerina peregrina (solid squares) and Cibicidoides (open diamonds) in KNR 197/10 CDH-42 from 3.87 km on Laurentian Fan. Radiocarbon-dated levels (conventional kiloyears) are indicated by arrows. At 852.5 cm, the planktonic date is on Globigerina bulloides, but all other PF dates are on Neogloboquadrina pachyderma (s.), and all BF dates are on U. peregrina. As on Hatteras Abyssal Plain at the same depth, Cibicidoides are very rare in the LGM. However, we took very large samples and tried twice to measure the stable isotopes on this genus, but both the δ¹⁸O and δ¹³C are widely scattered and seem to be irreproducible. In this respect the results are similar to those on BR.

Fig. S7.

Abundance of benthic foraminifera and ¹⁴C results of core KNR140 JPC-12 from 4,250 m on Blake Ridge. Nuttalides umbonifera abundance is represented by solid circles, and Cibicidoides are marked by diamonds. As with other sites on the western side of the western basin of the subtropical NW Atlantic, Cibicidoides are too rare to make stable isotope time series. Dates are on N. umbonifera for the three peaks (two are new data) and mixed planktonics at the 195- and 231-cm peaks, with the date at 259 cm based on fragments of G. inflata. The lower two peaks fall within the LGM as defined here.

Fig. S6.

Abundance of the benthic foraminifera U. peregrina (solid squares) and Cibicidoides (open diamonds) in KNR 178 GGC-2 from 3.93 km on Hatteras Abyssal Plain. Radiocarbon-dated levels are indicated by arrows. MPF denotes dates (conventional ¹⁴C ka) on mixed planktonic foraminifera, and BF denotes the same on U. peregrina. Dates that are not annotated are on the planktonic G. inflata (also noted as G.i.). The uppermost peak in Uvigerina abundance gave an age of 15.8 ka, 4,500 y older than the G.i. date from the same sample. As the planktonic dates indicate a rapid decrease in sedimentation rates between 10.85 and 16.85 ka, we suspect the anomalously large ventilation age may be associated with a hiatus. Note that this site is typical of cores from the western margin of the subtropical NW Atlantic in that Cibicidoides are only present in significant numbers in the Holocene. Thus, there are no Cibicidoides δ¹³C data from the LGM.

Unlike shallower locations where northern ventilation is thought to have resumed at about 14–15 ka (12), ventilation at 5 km on CR as represented by δ¹³C in Fig. 2 remained weak (between that of LGM and HS1) until ∼13.5 ka. However, results are ambiguous on BR. Whereas Pa/Th data indicate ventilation resumed 14–15 ka (2), δ¹³C indicates a 1-ky delay (Fig. S3). Either way, the available data suggest that postglacial resumption of ventilation in the North Atlantic evolved from the top down. That is, the tongue of glacial North Atlantic intermediate water thickened downward as suggested earlier (24), and more recently at sites >2.5 km off Brazil (10). Eventually, ventilation by convection in the Norwegian−Greenland Seas would have resumed, probably near the beginning of the Holocene.

Ocean and Climate Implications

What ventilation process could account for relatively young and nutrient-depleted water in the LGM NW Atlantic at 5 km? In the modern ocean, the densest waters are produced in the Nordic Seas by heat loss and mixing with cold Arctic waters, and around Antarctica by upwelling, heat loss, and brine rejection by freezing in polynyas and under ice shelves. During the LGM, much of the ocean heat transport to the Nordic Seas is thought to have been reduced, and new paleo-tracer data show deep Norwegian Sea waters were very old and nutrient rich (25), so we rule out the modern process in the Nordic Seas. Although oxygen isotope data indicate that brine rejection occurred in Nordic Seas (26), there is no evidence that enough dense water accumulated to spill over the sill at Denmark Strait (∼600 m).

Radiocarbon ventilation studies in the South Atlantic (27) and southwest Pacific (28) found a deep-water bulge in low Δ¹⁴C that is similar to what we observe in the western North Atlantic. The authors from both studies propose that deep waters during the LGM originated in the south, with relatively young northward flow underlying an aging southerly return flow. However, it should be noted that the bulge in the South Atlantic is defined by data at only three locations, and two have uncertain and different reservoir ages (29). Whereas a southern water mass origin is compatible with our results, the NW Atlantic data are far from the Southern Ocean data, with few between and none of >4 km depth. Instead, we seek a more local explanation, especially in light of the NW Atlantic Pa/Th, Nd, and the early decrease in δ¹⁸O on CR and the Iberian margin (12).

North Atlantic instrumental data suggest an alternative to an estuarine-type circulation driven from the south. The decadal observations that began in 1972 at Geochemical Ocean Sections Study (GEOSECS) Station 27 in Newfoundland Basin (Fig. 1) show that, using dissolved silicate as a tracer for AABW, bottom water between about 4 and 5 km had a variable AABW component (Fig. 4A). In 1972, there was a bulge of silicate at about 4.5 km that was undercut by low-silicate water that must have originated at Denmark Strait, according to the water properties mapped by Worthington and Wright (30). Not only is a northern source indicated by low silicate, but it is also supported by high Δ¹⁴C and tritium, each of which undercuts old water between about 4.3 and 4.5 km depth (Fig. 4C). This pattern of northern source water undercutting the southern source in Newfoundland Basin in 1972 was recorded by GEOSECS as far south as Station 28, but not as far south as Station 29 (Figs. 1 and 4B). Some of the pronounced silicate bulge in 1972 may be related to the negative phase of the North Atlantic Oscillation, which is hypothesized to cause enhanced convection in the Nordic Seas (31). The Newfoundland Basin ventilation changes are not exactly analogous to the LGM situation in the NW Atlantic because of the presumed absence of LGM convection in the Nordic Seas, but they show that the simplest way to produce a bulge in old water in the NW Atlantic is to introduce a local source of relatively young, dense water.

Fig. 4.

Repeat geochemical observations in Newfoundland Basin (Fig. 1) since 1972. (A) Four decades of repeat dissolved silicate measurements at the location of GEOSECS Station 27 in the Newfoundland Basin. (B) Silicate measurements at four 1972 GEOSECS stations that form a roughly north to south transect. In both A and B, silicate is used as a tracer for water of AABW origin. The GEOSECS Station 27 observations stand out in A because of the highest silicate concentrations at about 4,700 m, and the lowest at 5,000 m. The Si profile at Station 27 is similar to that of Station 28 to the south (Fig. 1), but, at southernmost Station 29, there is no evidence of the low Si water at the bottom. (C) The radionuclide tracers ¹⁴C and tritium at Station 27 show that waters at >4.7 km were younger than waters at ∼4.5 km in 1972. The data indicate that strong Denmark Strait Overflow Water (low Si, high Δ¹⁴C and tritium) must have undercut the older southern component in 1972, causing a steep reversal in gradient unlike any later time. Data are from the Global Ocean Data Analysis Project (GLODAP) (cdiac3.ornl.gov/waves/discrete/) and are listed in Dataset S4. CLIVAR, Climate Variability project; TTO, Transient Tracers in the Ocean project; WOCE, World Ocean Circulation Experiment.

Accepting the evidence for old and warm bottom waters in the Nordic Seas during the LGM (25), we propose the Labrador Sea could have been a site of deep LGM ventilation by brine rejection around its margins. There is no direct evidence for this proposition, but it would have been facilitated by katabatic winds blowing down a Laurentide Ice Sheet that was grounded on the broad Labrador continental shelf. The LGM glacial till along the Labrador shelf is thought to have been deposited ∼20 ka by a retreating ice margin that may have formed an ice shelf (32), and a basin such as the Labrador marginal trough could have accumulated brine. Offshore winds also could have promoted ice growth, dispersal, and brine formation similar to the several latent heat polynyas around the Okhotsk Sea that produce, today, the densest water in the North Pacific Ocean (33).

In conclusion, we have shown that there may have been a third component of deep North Atlantic circulation during the LGM, thereby adding a new dimension to our view of water mass distribution in the glacial ocean. The new C isotope results support the same conclusion based on Nd isotopes (20). If this new source of bottom water resulted from brine rejection and not heat loss to the atmosphere, there may not have been the same climate impact as modern NADW production. The new deep northern component is indicated by the reversal in C isotope trends to higher δ¹³C and lower ¹⁴C_B-P below 4.2 km in the western North Atlantic and more vigorous flow (23). Thus, in this view, water of southern source was sandwiched between glacial North Atlantic intermediate water and the new dense bottom water. The spatial extent of the deepest northern component can best be explored by new coring even deeper than 5 km near CR and up the west flank of the Mid-Atlantic Ridge.