Volume 36, Issue 1 e2020PA003877
Research Article
Free Access

Distinguishing Glacial AMOC and Interglacial Non-AMOC Nd Isotopic Signals in the Deep Western Atlantic Over the Last 1 Myr

Maria Jaume-Seguí

Corresponding Author

Maria Jaume-Seguí

Departament de Dinàmica de la Terra i l'Oceà, Universitat de Barcelona, Barcelona, Spain

Lamont-Doherty Earth Observatory, Columbia University, Palisades, NY, USA

Correspondence to:

M. Jaume-Seguí,

[email protected]

Search for more papers by this author
Joohee Kim

Joohee Kim

Lamont-Doherty Earth Observatory, Columbia University, Palisades, NY, USA

Department of Earth and Environmental Sciences, Columbia University, Palisades, NY, USA

Search for more papers by this author
Leopoldo D. Pena

Leopoldo D. Pena

Departament de Dinàmica de la Terra i l'Oceà, Universitat de Barcelona, Barcelona, Spain

Search for more papers by this author
Steven L. Goldstein

Steven L. Goldstein

Lamont-Doherty Earth Observatory, Columbia University, Palisades, NY, USA

Department of Earth and Environmental Sciences, Columbia University, Palisades, NY, USA

Search for more papers by this author
Karla P. Knudson

Karla P. Knudson

Lamont-Doherty Earth Observatory, Columbia University, Palisades, NY, USA

Search for more papers by this author
Maayan Yehudai

Maayan Yehudai

Lamont-Doherty Earth Observatory, Columbia University, Palisades, NY, USA

Department of Earth and Environmental Sciences, Columbia University, Palisades, NY, USA

Search for more papers by this author
Alison E. Hartman

Alison E. Hartman

Lamont-Doherty Earth Observatory, Columbia University, Palisades, NY, USA

Department of Earth and Environmental Sciences, Columbia University, Palisades, NY, USA

Department of Analytical Services, Agricultural Experiment Station Chemical Laboratories, University of Missouri, Columbia, MO, USA

Search for more papers by this author
Louise Bolge

Louise Bolge

Lamont-Doherty Earth Observatory, Columbia University, Palisades, NY, USA

Search for more papers by this author
Patrizia Ferretti

Patrizia Ferretti

Dipartimento di Scienze Ambientali, Informatica e Statistica, Università Ca' Foscari Venezia, Venice, Italy

Search for more papers by this author
First published: 03 December 2020
Citations: 4


The global thermohaline circulation plays a major role in regulating global climate, driven by the formation of North Atlantic Deep Water (NADW). ODP Site 1063 on the Bermuda Rise, at the interface of NADW and Southern Ocean-sourced water, appears an ideal location to study the relationships between ocean circulation and climate. This study reports Nd isotope ratios at Site 1063 that extend to ~1 Ma. The data show Nd isotope values during portions of interglacials that are much lower than modern NADW. However, interglacial Nd isotope values at Site 607, located within the core of NADW, off the abyssal seafloor in the North Atlantic, are consistently similar to modern NADW. In contrast to glacial values, we infer that interglacial Nd isotopes at Site 1063 are not representative of NADW and do not solely record water mass mixing. We conclude that the low Ndisotope ratios reflect regional particle-seawater exchange as a consequence of input of freshly ground bedrock from the Canadian shield, which is eroded into the North Atlantic during major ice sheet retreats. The result is a deep, thin, and regionally constrained layer of seawater tagged with this anomalous low Nd isotope signature that is unrepresentative of the Atlantic Meridional Overturning Circulation (AMOC). We suggest that a benthic nepheloid layer, whose development is linked to the behavior of a deep-recirculating gyre system, regulated by the interaction between the Gulf Stream and the deep western boundary current, facilitates the periodic masking of the Nd isotope signature of the North Atlantic AMOC end-member in this region at these depths.

Key Points

  • Glacial-interglacial changes at the deep Bermuda Rise during the Middle and early Lower Pleistocene were recorded with εNd
  • Glacial εNd values in the deep Bermuda Rise record the AMOC, interglacial εNd values do not
  • Dynamic benthic nepheloid layers temporally mask εNd at the deep Bermuda Rise during parts of interglacials

1 Introduction

Ocean circulation plays a crucial role in regulating heat transport across the globe and carbon storage in the deep ocean, and various lines of evidence suggest the existence of different ocean circulation modes over glacial-interglacial cycles (e.g., Broecker, 1991; Rahmstorf, 2002). In the modern ocean, the formation of deep waters in the North Atlantic (North Atlantic Deep Water, NADW) and deep and intermediate water in the Southern Ocean (Antarctic Bottom Water and Antarctic Intermediate Water, AABW and AAIW, respectively) drives the global thermohaline circulation (Broecker, 1991; Rahmstorf, 2002). Reconstructing the distribution of these water masses is essential for understanding how ocean circulation has changed throughout time and its relationship to climate variability. Studies of the Atlantic arm of the global system, termed the Atlantic Meridional Overturning Circulation (AMOC), have shown that during interglacials, NADW flows deep into the South Atlantic where it is overlain by AAIW and underlain by AABW. During glacials, however, the export of NADW is reduced, and the deep Atlantic is characterized by more extensive AABW underlying a shallower, less extensive NADW (Boyle & Keigwin, 1982; Curry & Oppo, 2005; Lynch-Stieglitz et al., 2007).

The western North Atlantic represents an important mixing zone of deep and intermediate depth circulation. The existence of fast-flowing deep western boundary currents (DWBC) in the region is indicated by extensive sediment remobilization and the presence of large sediment drift deposits. Ocean Drilling Program (ODP) Site 1063 was recovered from one such drift deposit on the Bermuda Rise (33.41°N, 57.36°W, 4,584 m) (Figures 1a and 1b). With sedimentation rates of ~30–60 cm/kyr during glacials (Channell et al., 2012), it represents one of the highest temporal resolution archives known from the deep North Atlantic. Site 1063 is located in the present-day mixing zone of NADW and AABW (Figures 1a and 1b). Thus, from its location it appears to represent an excellent site to study changes in deep waters affecting North Atlantic circulation on glacial-interglacial and millennial timescales.

Details are in the caption following the image
(a) Map with the core locations and main deep water masses and currents: North Atlantic Deep Water (NADW; light green arrows) and Antarctic Bottom Water (AABW; dark blue arrows). Green stars point to regions of deep water formation: Labrador Sea Water (LSW), Denmark Strait and Iceland-Scotland Overflow Waters (DSOW, ISOW). Thin green arrows represent the two deep recirculating gyres (detailed in Figure S1). The Gulf Stream (dashed red arrow) flows on the surface between the two gyres. Small black arrows point to the variable sediment sources that affect the deep Bermuda Rise. Brown shaded areas in the western North Atlantic indicate the highest particulate matter concentrations in the benthic nepheloid layer described by Gardner et al. (2018). Colors in the American continent point to source regions with different εNd values (i.e., Appalachians (orange), Grenville Province (light purple), Archean Canadian shield (purple); after Casse et al., 2019 and Jeandel et al., 2007). CR: Corner Rise; BBOR: Blake-Bahama Outer Ridge. Bathymetric map composed with GeoMapApp (http://www.geomapapp.org) (Ryan et al., 2009); grey scale displays meters water depth (mwd). (b) Salinity profile of the Atlantic Ocean; PSU: Practical Salinity Units. Studied core locations and main deep water masses are shown (after Lambelet et al., 2016). Upper Labrador Sea Water (ULSW) and Labrador Sea Water (LSW) mainly contribute to Upper NADW while Denmark Strait Overflow and Iceland-Scotland Overflow Waters (DSOW, ISOW) contribute to lower NADW. In the modern ocean, DSDP Site 607 is bathed by Lower NADW (LNADW) while ODP Site 1063 is in the mixing zone between LNADW and AABW. This was generated with Ocean Data View software (Schlitzer, 2018) using the GLODAPv2 2019 database (Olsen et al., 2019).

Several studies using benthic foraminiferal δ13C to trace bottom water ventilation associated with AMOC variability over time at Site 1063 have revealed changes in the relative contributions of northern-sourced versus southern-sourced waters (NSW and SSW, respectively, here used to describe past analogs to NADW and AABW) over glacial-interglacial timescales during the late Pleistocene, with enhanced production of NSW during interglacials and enhanced SSW during glacials (e.g., Ferretti et al., 2005; Poirier & Billups, 2014). Neodymium isotope ratios have been used in abyssal Bermuda Rise sediments, including Site 1063, with the purpose of tracing glacial-interglacial AMOC variability, and they have shown large variations, including extreme low values during interglacials, far outside present-day range observed between the end-members used for tracing the global overturning ocean circulation, NADW and North Pacific Deep Water (NPDW) (Böhm et al., 2015; Gutjahr & Lippold, 2011; Roberts et al., 2010). The meaning of these results has been a matter of debate, with some studies concluding that the Nd isotope ratio of the NSW end-member has shown large fluctuations (Böhm et al., 2015; Gutjahr & Lippold, 2011) and others questioning whether the extreme values faithfully reflect the AMOC (e.g., Howe et al., 2016; Pöppelmeier et al., 2018).

Here we present a new Nd isotope record from Fe-Mn encrusted foraminifera and fish debris at ODP Site 1063 from 0.25 to ~1.0 Ma. Our study extends the time period of previous studies of this site and allows us to evaluate the Nd isotope variability over the last 13 interglacial-glacial cycles. We also compare the Nd isotope ratios of Site 1063 to a record from Deep Sea Drilling Project (DSDP) Site 607 (41.00°N, 32.95°W, 3,427 m), located to the northeast, on the western flank of the Mid-Atlantic Ridge within the core of present-day NADW (Figures 1a and 1b). This comparison enables us to potentially isolate the Nd isotope variability at Site 1063 that is attributable to changes in NSW prominence versus other non-circulation related processes that might impact the Nd isotope ratios.

2 Background

2.1 Nd Isotopes in the Oceans

The residence time of Nd in seawater is estimated to be ~500–1,000 years, shorter than or similar to the ocean mixing time (e.g., Goldstein & O'Nions, 1981; Tachikawa et al., 1999), and, as a result, Nd isotope ratios vary spatially and with depth, making them useful water mass tracers (e.g., Frank et al., 2002; Goldstein & Hemming, 2003). 143Nd is produced by radioactive decay of 147Sm (t1/2 = 105 Ga), which is responsible for the variability of 143Nd/144Nd ratios in the Earth. Nd in seawater is dominantly derived from the continents, where 143Nd/144Nd ratios mainly reflect how long the Nd has been in the continental crust (e.g., Goldstein et al., 1984). In this study Nd isotope ratios are expressed as εNd, the fractional deviation from an estimate for the average value of the Earth in parts per 10,000 (defined further in Materials and Methods). The North Atlantic, encircled by old cratonic continental crust (Figure 1a), has the lowest εNd values in the oceans, with NADW εNd ~ −13.5, while the Pacific, surrounded by young orogenic belts and volcanoes, has the highest εNd values, with NPDW εNd ~ −2 to −4 (e.g., Piepgras & Wasserburg, 1980). In between these end-members, the present-day distribution of Nd isotope ratios in intermediate and deep oceans is “quasi-conservative”, that is, mainly determined by the global thermohaline circulation and water mass mixing (e.g., Goldstein & Hemming, 2003; Jones et al., 2008; Piepgras & Wasserburg, 1980; van de Flierdt et al., 2016). If these water mass values are transferred to and preserved in sedimentary archives, and the values of the end-member water masses are known, then changes in the proportions of the North Atlantic and North Pacific end-members can be traced through time (e.g., Frank et al., 2002; Goldstein & Hemming, 2003; Piotrowski et al., 2005; Roberts et al., 2010).

The application of Nd isotopes to trace past ocean circulation is limited by a general lack of understanding of the processes that control the Nd chemical cycling in the oceans. For example, while Nd isotopes appear to behave as quasi-conservative tracers of water mass mixing in the deep ocean, Nd concentrations ([Nd]) appear non-conservative, showing systematic increases with depth (e.g., the “Nd paradox”; Goldstein & Hemming, 2003). In the modern ocean, sediments from rivers and atmospheric dust deposited in the ocean contribute Nd to surface waters. Decreasing Nd concentrations observed in surface and shallow waters and increasing Nd concentrations with depth have been interpreted as reflecting scavenging of Nd by sinking marine particles, which release or exchange Nd with increasing depth (e.g., Siddall et al., 2008). Tachikawa et al. (2003) concluded that neither Nd concentrations nor Nd isotopes in the modern ocean can be explained by riverine and/or dust inputs and estimated that up to 90% of the oceanic Nd may be added by boundary exchange at continental margins (Jeandel et al., 2007; Lacan & Jeandel, 2005). Recent studies suggest a benthic source to explain the excess of Nd in the oceans (Abbott, Haley, & McManus, 2015; Abbott, Haley, McManus, & Reimers, 2015; Abbott et al., 2019). Their “bottom-up” model suggests that deep sea sediments play an important role in the Nd budget, through pore water flow in the sediment column during early diagenesis, which impacts the Nd isotopic composition of the bottom waters and may add complexity to using εNd as a paleoceanographical tracer.

2.2 North Atlantic Circulation

The deep subtropical North Atlantic encompasses the convergence zone between southward flowing NADW and northward flowing AABW. The DWBC represents the westernmost branch and the fastest core of NADW flow. Formed in the Irminger Sea between Greenland and the Reykjanes Ridge south of Iceland by sinking Nordic Seas waters, the DWBC flows equatorward, incorporating the deep waters of the Labrador Basin, and skirting the continental slope (Figure 1a). At subtropical latitudes (~23–35°N), NADW can be generally characterized by two main water masses (Figure 1b). (1) Upper NADW (UNADW, εNd = −13.2 ± 1.0) results from mixing of Upper Labrador Seawater (ULSW; 1,000–1,800 m, εNd = −14.3 ± 0.3) and Labrador Sea Water (LSW; 1,400–2,800 m, εNd = −13.7 ± 0.9), which represent the least dense and best ventilated proto-NADW water masses (Lacan & Jeandel, 2004; Lambelet et al., 2016; Piepgras & Wasserburg, 1987; Stahr & Sanford, 1999). (2) Lower NADW (LNADW; 2,500–4,300 m, εNd = −12.4 ± 0.4) contains higher contributions from the Nordic Seas water masses Iceland-Scotland Overflow Water and Denmark Strait Overflow Water (ISOW and DSOW, respectively, εNd = −8.4 ± 1.4) (Lacan & Jeandel, 2004; Lambelet et al., 2016). The northward flowing AABW can be recognized below 4,500 m in the NW Atlantic, characterized by a significantly more negative εNd-signature (εNd = −12.2 ± 1.0) (Lambelet et al., 2016) than that in the South Atlantic (εNd = −8.5 ± 0.3; Jeandel, 1993) (Figures 1a and 1b).

At the surface, the warm Gulf Stream flows northeastward, driving the anticyclonic subtropical gyre. At depth, the NW Atlantic is characterized by the presence of two gyres, whose motion is directly linked to the northward flowing Gulf Stream and the southward flowing DWBC (Figure 1 and supporting information Figure S1) (Calvin Campbell & Mosher, 2016; Schmitz & McCartney, 1993). These gyres and currents respond to changing climatic conditions (Keffer et al., 1988; Keigwin et al., 1998; Lynch-Stieglitz et al., 1999) and exert a strong influence on the redistribution of sediments within the basin, even eroding the bottom seafloor if they are intense and deep enough. Today, the fast core of the DWBC flows at 3,500–4,000 m depth and the Gulf Stream separates from the continent and flows eastward along ~40°N latitude (Figures 1 and S1). Recent studies have shown a tight link between surface Gulf Stream meandering and the thickness and depth reached by the associated deep anticyclonic gyre. When the Gulf Stream meanders greatly, the deep anticyclonic gyre can expand down to ~5,000 m. On the other hand, during weak meandering periods, the gyre is constrained above 4,000 m (Gardner et al., 2017). As a result of this linkage, intense benthic nepheloid layers (BNLs) with high suspended particle loads develop in the near-bottom region below the gyres (Gardner et al., 2018) (Figures 1 and S1), with the highest particulate matter concentrations occurring below the Gulf Stream path. These processes bring sediment into the water column and focus it on the bottom, generating drift layers like the one on the Bermuda Rise where Site 1063 was recovered.

3 Materials and Methods

This study reports εNd data from Fe-Mn encrusted foraminifera and fish debris from Site 1063 (Figure 1). Mixed species of planktonic foraminifera from 97 samples were picked from the >125 μm size fraction and processed according to established procedures at Lamont-Doherty Earth Observatory (LDEO) (Pena & Goldstein, 2014). Foraminifera tests were crushed and cleaned of clay particles by ultrasonic agitation with MilliQ water and methanol and then dissolved in weak acetic acid. Fish debris from 95 samples (33 of which are replicates of foraminifera samples; Table S1) were picked from the >63 μm fraction, cleaned with methanol and water, and dissolved in 1:1 concentrated nitric acid-concentrated hydrochloric acid solution. Dissolved samples were run through Eichrom® TRU-Spec and Ln-Spec resin columns to isolate Nd (Pin & Santos Zalduegui, 1997).

The amount of extracted Nd in each sample was determined on a VG PQ ExCell quadrupole inductively coupled plasma mass spectrometer (ICP-MS) prior to measuring Nd isotope ratios on a ThermoScientific Neptune-Plus multi-collector ICP-MS at LDEO. We carried out eight different measuring sessions, with 10 to 28 samples, grouped by the Nd concentrations (10–200 ppb in the measurement solutions). Samples were bracketed with JNdi standards (Tanaka et al., 2000) with concentrations similar to those of the samples, to monitor instrumental drift and to calibrate each running session and estimate uncertainties. External reproducibility from routine runs of the JNdi standard ranged from ±0.000008 to ±0.000023. For foraminifera samples, the median concentration batch measured was 20 ppb, which have errors of ±0.000017 (n = 44). For fish debris samples, the median concentration batch measured was 50 ppb with an error of ±0.000009 (n = 31). 143Nd/144Nd ratios were normalized to 146Nd/144Nd = 0.7219 to correct for mass fractionation using the exponential fractionation law and then corrected for instrumental bias to a 143Nd/144Nd ratio of 0.512115 for JNdi-1 (Tanaka et al., 2000). All sample uncertainties reported are the 2σ external reproducibility of the averaged 143Nd/144Nd of bracketing standards during the same measurement session, and n is the number of bracketing standards. In most cases the internal precision is better than the external error. In cases where the internal error was greater than external error, the propagated error (square root of the sum of the squared errors) is reported. We also carried out analyses of the La Jolla Nd standard during every session. Across all the sessions, these gave an average value of 0.511859 ± 0.000006 (2σ standard deviation, n = 17). Nd isotope ratios are represented as εNd, the deviation from the Chondritic Uniform Reservoir (CHUR) value, representing the average estimated value for chondritic meteorites and the Earth, in parts per 10,000: εNd = [(143Nd/144Nd)sample/ (143Nd/144Nd)CHUR]-1]*104, where in this study 143Nd/144NdCHUR = 0.512638 (Jacobsen & Wasserburg, 1980).

4 Results

Results for Site 1063 are shown in Figures 2 and 3. We update the age model for the ~1,030 to 730 ka interval (Ferretti et al., 2005) by re-tuning the Site 1063 benthic δ18O record to the LR04 benthic stack. From ~730 to ~250 ka, the age model used is from Poirier and Billups (2014). Over the entire 1 Ma time interval studied (Figure 2), Site 1063 has high εNd variability extending far outside the range of the global NADW (~−13.5) and NPDW (~−4 to −2) end-members, consistent with previous studies (Böhm et al., 2015; Gutjahr & Lippold, 2011; Roberts et al., 2010). This variability is mainly due to the presence of εNd values much more negative than NADW during interglacials and strong interstadials, generally around −16 but lower than −19 in several cases (e.g., Marine Isotope Stages [MIS] 29, 27c, 25, 17c, and 9a, reaching −26 at MIS 27c). In contrast, glacial periods are characterized by more positive εNd values and much smaller variability (constrained between −10 and −14 through the entire time interval).

Details are in the caption following the image
Site 1063 εNd data analyzed in this study plotted versus age. Replicated data are highlighted in red. For those samples that have been replicated, the average value between the carriers has been used to construct the continuous record (black line). Site 1063 benthic foraminifer δ18O record is shown by a gray line (Billups et al., 2011; Ferretti et al., 2005; Poirier & Billups, 2014; Poli et al., 2000), plotted as a 3-point running mean of the record. The horizontal blue bar represents εNd of modern NADW. Numbers represent the different marine isotope stages (MIS).
Details are in the caption following the image
Crossplot of εNd extracted from foraminifera versus εNd from fish debris. The black line follows the 1:1 correlation. The gray area indicates the average 2 standard deviation for the data set. The brackets on the x and y axes represent gaps in values. Statistics can be found in supporting information (Table S1).

In order to assess whether the foraminifera samples reflect the original deep seawater signal or have been impacted by later diagenetic or pore water alteration, we crosschecked the foraminifera against fish debris from the same 33 samples covering nearly the entire time interval, from MIS 29 to 8 (1.03 to 0.24 Ma). The fish debris samples have very high Nd concentrations (typically 100–1,000 ppm) compared to foraminifera (typically with ~1–10 ppm); as a consequence, Nd isotope ratios in fish debris are much more resistant to change by later alteration (Huck et al., 2016; Martin & Haley, 2000; Martin & Scher, 2004). The results show an excellent agreement (R2 = 0.983) between foraminifera and fish debris (Figure 3 and Table S1), without systematic bias (Figure S2) and with only 4 out of 33 samples displaying values that are outside of analytical error. We conclude that both the foraminiferal coatings and fish debris represent reliable archives of the Nd-isotopic signal of past bottom waters at Site 1063 and interpret the data extracted from both carriers as a record of the deep seawater at this site (Figure 2).

5 Discussion

5.1 Nd Isotopes in the Deep Bermuda Rise and North Atlantic Water Masses

In the modern ocean, the deep Bermuda Rise is characterized by εNd ~ −12.5 (e.g., seawater values from the region between 31.7 and 37.5°N and 3,999–5,001 m depth have εNd = −12.3 to −12.7; in Lambelet et al., 2016), consistent with its location in the confluence between NADW (εNd ~ −13.5) and AABW (εNd ~ −8.5). However, our ~1 Myr record for Site 1063 displays a large range of εNd values, with many highly negative values as extreme as ~−26 (Figure 2). NADW is a multi-sourced water mass derived from the mixing of LSW (εNd ~ −13.5) and Nordic Seas (DSOW/ISOW, εNd ~ −8.5) end-members (Lacan & Jeandel, 2004). Temporal fluctuations in the εNd values at a deep sea core site downstream from NSW sources might be attributed to (1) changes in the fraction of NSW versus SSW, or (2) to changes in the relative contributions of the end-members that form NSW (LSW and DSOW/ISOW), or (3) changes to the εNd values of the NSW or SSW sources. In addition, (4) at individual sites the seawater εNd-signature may be changed by diagenetic and/or pore water alteration and other sediment-seawater exchange processes (e.g., Abbott et al., 2019; Blaser et al., 2019; Böhm et al., 2015; Goldstein & Hemming, 2003; Lacan & Jeandel, 2005; Pena & Goldstein, 2014; Piotrowski et al., 2004).

Studies reconstructing Nd isotope records in the abyssal Bermuda Rise from MIS 5 to the late Holocene have discussed the anomalously negative values observed, as low as εNd ~ −18.3, as being the result of changes in the εNd of the source waters of NADW and/or their export rates (Böhm et al., 2015; Gutjahr & Lippold, 2011; Roberts et al., 2010). Some studies have concluded that these occurrences reflect enhanced LSW production during peak interglacials and Dansgaard-Öschger interstadials (Böhm et al., 2015; Roberts et al., 2010). However, Howe et al. (2016) compared Holocene εNd data from different latitudes in the intermediate to deep North Atlantic and showed that highly negative εNd values, such as those at Site 1063, are not found at any North Atlantic sites outside of the deep Bermuda Rise. This observation, along with the fact that, in the present-day Atlantic, the NADW precursors do not have highly negative values (LSW has εNd values of ~−13.5, similar to NADW and DSOW/ISOW has ~−8.5 (Lacan & Jeandel, 2004; Lambelet et al., 2016; Piepgras & Wasserburg, 1987), indicates that changes in the εNd of the NADW end-members are unlikely to explain the negative εNd at Site 1063. Additionally, foraminifera stable isotope records and micropaleontological data suggest that LSW did not form during the early Holocene (Hillaire-Marcel et al., 2001), when extreme negative εNd values (~−16.2) are observed in core OCE326-GGC6, which was recovered at the location of Site 1063 (Roberts et al., 2010) (Figure 4). LSW formation, instead, resumed at ~7 ka, at the same time as NADW-like values characterize the εNd-record (Figure 4). Howe et al. (2016) suggest that the very negative εNd values recorded at the deep Bermuda Rise are the result of the influx of poorly chemically weathered material from the Canadian Shield into the Labrador basin during ice sheets retreat, combined with an active deep circulation that allows the southward advection of the modified bottom waters to Site 1063.

Details are in the caption following the image
εNddata from OCE326-GGC6—equivalent to ODP Site 1063 (orange)—for the last glacial-interglacial cycle. εNd data from Fe-Mn-coated foraminifera (green line), reductively cleaned fish debris (green circles) and detrital fraction (brown dashed line) (after Howe et al., 2016; Roberts et al., 2010, Roberts et al., 2012). The δ18O record from the NGRIP ice core is used as a reference. Vertical gray bars represent cold events that characterize the last termination: Heinrich Stadial 1 (HS-1), Bølling-Allerød (B/A) and Younger Dryas (YD). The light blue bar represents present-day NADW. The darker blue shadow on the εNd axis shows the range of εNd values recorded by the present-day seawater in the deep NW Atlantic close to ODP Site 1063 (Lambelet et al., 2016), which are slightly less negative than average NADW. Top schematic diagram shows depth and intensity variability in DWBC (black and thicker bars represent more intensity) (after Evans & Hall, 2008). Light pink bar represents the suggested depth at which an intense main core of the DWBC enhances the deep-gyre system. SSW enters the region during cold periods.

5.2 Pleistocene Deep North Atlantic: Site 1063 Versus 607 εNd Records

In order to inform our understanding of the evolution of the deep North Atlantic throughout the Early to Late Pleistocene and assess the causes of the negative εNd values observed at the Bermuda Rise, we compare Site 1063 with εNd records from DSDP Site 607 (Kim et al., 2020). Site 607 is located in the deep North Atlantic, northeast of Site 1063 in the present-day core of NADW (Figures 1a and 1b). It thus represents an alternative site to 1063 for tracing the variability of the deep North Atlantic. Site 607 is ~1 km shallower, on the western flank of the Mid-Atlantic Ridge, and farther away from terrigenous sediment sources than Site 1063, which is on the abyssal plain (Figure 1a).

The Site 607 εNd data show a strong coherence with benthic δ13C (R2607: 0.67) on glacial-interglacial timescales (Figure 5). Benthic foraminiferal δ13C is commonly used to reconstruct the past distribution of deep water masses, by attributing changes in bottom water ventilation to AMOC variability (e.g., Boyle & Keigwin, 1982; Curry & Oppo, 2005; Lynch-Stieglitz et al., 2007). Although the 13C/12C ratio is impacted globally by changes in the terrestrial/marine biosphere and regionally by primary biological productivity and physical effects such as air-sea exchange (e.g., Ravelo & Hillaire-Marcel, 2007), εNd values are not affected significantly by these same processes (Goldstein & Hemming, 2003). Because there is strong coherence between δ13C and εNd at Site 607 and δ13C and εNd interglacial values fall within the range of modern NADW (Figure 5), we are confident that both proxies track changes in the deep Atlantic water through time.

Details are in the caption following the image
Crossplots of εNd versus benthic foraminifer δ13C for (a) Site 607 and (b) Site 1063; εNd from Site 607 are from Kim et al. (2020), and δ13C data are from Raymo et al. (1990) and Ruddiman et al. (1989). For Site 1063, εNd data are from this study and δ13C data are from Billups et al. (2011), Ferretti et al. (2005), Poirier and Billups (2014) and Poli et al. (2000). Open symbols represent glacial data, and closed symbols represent interglacial data. Gray open symbols correspond with glacial stages prior to MIS 22, which are characterized by more NADW-like εNd values than glacial values from MIS 22-to-2 (Pena & Goldstein, 2014; Figure S3). Gray lines display the correlation between the two variables (R2607 = 0.67; R21063 = 0.31). (c) Cross-plot from combining both 607 (purple) and 1063 (orange) data. Horizontal bars represent the characteristic εNd and δ13C for the main water masses in the modern ocean: NADW (light blue) and AABW (yellow). The brown shadow in the background represents the gradient of contamination from an external source of Nd, from lighter (less affected) to darker (more affected) tones.

εNd values at Site 607 are constrained within a much narrower range than those at Site 1063 during interglacials (Figures 6 and 7). Site 607 εNd values are around the present-day NADW εNd value of ~−14 during most interglacials, while Site 1063 is characterized by εNd values that are consistently more negative during interglacials, generally fluctuating within εNd = −16 ± 2 but showing several negative peaks around εNd ~ −20 and reaching an extreme value of −26. This relationship between εNd values at Sites 607 and 1063 during interglacials, where the data from shallower Site 607 remain similar to present-day NADW through most of the interglacial periods, while the data from deeper Site 1063 show excursions to more negative values (Figure 7), is the opposite of what would be expected if the reason for the extremely negative εNd values was increased LSW contributions. Considering the water mass structure of the western North Atlantic, LSW (1,000–2,000 m) is a major contributor to UNADW; as such, UNADW has more negative εNd values than LNADW, whose major contributors are ISOW/DSOW (~−13.5 vs. −12.5, respectively, in Stations 15, 19, and 21 of Lambelet et al., 2016, located between 32°N and 37°N). In addition, the deeper western North Atlantic waters showing significant amounts of AABW have higher εNd values than LNADW (~−11.4 to −12.7, in Stations 15, 18, 21, 25, and 26 of Lambelet et al., 2016, located between 22°N and 37°N). Thus, in the modern ocean, εNd becomes more positive with increasing depth as the water masses transition from UNADW to LNADW and to AABW. It is reasonable to conclude that an inversion of the water mass structure of the western North Atlantic, such that LSW containing a larger proportion of the Nd with highly negative εNd values that today are only observed in shallow Baffin Bay waters (<−20), north of the Labrador Sea, is highly unlikely (Piepgras & Wasserburg, 1987; Stordal & Wasserburg, 1986).

Details are in the caption following the image
εNd records for Site 1063 (orange) and Site 607 (purple). Site 1063 data are from this study (thick orange line) from 250–1,100 ka and from Böhm et al. (2015) and Roberts et al. (2010) (thin orange line) for 0 to 150 ka. Thicker Site 607 data are from Kim et al. (2020) (thick purple line) and D. C. Lang et al. (2016) (thin purple line). Site's 607 age model has been modified from Ruddiman et al., (1989) by tuning the 607 benthic foraminifer δ18O record to the LR04 stack benthic foraminifer δ18O record (Kim et al., 2020). Glacial (open circles) and interglacial (closed circles) εNd maxima were chosen following the glacial-interglacial maxima defined by the LR04 stack (gray line) (Lisiecki & Raymo, 2005). Horizontal blue bar represents εNd of the present-day NADW. Numbers represent the different Marine Isotope Stages (MIS).
Details are in the caption following the image
(a) Interglacial and (b) glacial εNd records for Site 1063 (orange) and Site 607 (purple). In Site 1063 record, MIS 3 to 6 data (Böhm et al., 2015) were extracted from leachates instead of foraminifera and/or fish debris. Only the data that correspond with benthic δ18O < 4‰ in the LR04 stack benthic foraminifer record (gray line) (Lisiecki & Raymo, 2005) are included to plot the interglacial record. For glacial periods, every datapoint analyzed that fall within the range of each MIS δ18O maxima has been plotted. Color scheme in bars used as in Figure 5. Diamonds represent εNd from Fe-Mn crusts BM1969.05 (open) and TR079 D-14 (filled), recovered from intermediate depths in the NW Atlantic (Foster et al., 2007).

If Nd-isotopes at Site 1063 are simply taken to solely represent variability in NSW, this would suggest that NSW εNd signature is highly variable through time, often much more negative than NADW during interglacials. In contrast, if Site 607 faithfully represents NSW, it implies that the εNd signature of NSW remained similar to the present-day through time. Considerations of benthic δ13C-εNd relationships place additional constraints on the sources of Nd in Sites 1063 and 607. The sources of NADW (or NSW), such as LSW and DSOW, are well-ventilated water masses with high δ13C (Lambelet et al., 2016; Stahr & Sanford, 1999). In the Bermuda Rise record, the highest δ13C values do not correspond with the extreme negative εNd values, but rather to intermediate values (Figure 5), indicating that the extreme negative εNd signatures in Site 1063 are not derived from LSW or DSOW. Moreover, the interglacial stability of the NSW end-member is also supported by benthic δ13C records from intermediate depths in the subpolar North Atlantic (Raymo et al., 2004). These data from higher latitudes indicate that the relative contribution of Labrador Sea and Nordic Seas waters did not change significantly during the Pleistocene, consistent with the εNd data at Site 607 indicating that the εNd values of the NSW end-member during interglacials remained similar to NADW through the Pleistocene. Further evidence that very negative εNd values at Site 1063 during interglacials do not correspond with changes in NSW comes from a recent study in the Labrador Sea that concluded that deep waters were more negative during MIS 2 and 3 (~−16 ± 1) than during the Holocene (~−13 ± 1) (Blaser et al., 2020). This would generate LNADW with more negative εNd during glacials than during interglacials, opposite to what we observe at Site 1063. Therefore, it appears that the negative εNd values recorded at the deep Bermuda Rise in Site 1063 are not caused by the arrival of extremely negative εNd in LSW or DSOW but instead reflect regional processes in the midlatitudes, where Site 1063 is located.

Comparing Sites 1063 and 607, an important observation is that the pronounced differences between them occur almost exclusively during interglacials (Figure 7). During glacial maxima, both sites have similar εNd values, indicating that both sample a single water mass that filled the deep North Atlantic during cold periods. The sites show notable glacial differences (2–4 εNd units) only during MIS 8, 14, and 28. Interestingly, these are weak glacials (Lang & Wolff, 2011), indicated by benthic δ18O values (≤4.5‰), that are lower than average Late Pleistocene glacial values in the LR04 global benthic stack (Lisiecki & Raymo, 2005).

The Site 607 data offer an important perspective for the interpretation of the Site 1063 data through time. We suggest that Site 607 εNd provides a better reflection of the changing εNd of the deep North Atlantic related to AMOC than Site 1063, for the following reasons. (1) Site 607's location on the flanks of the Mid-Atlantic Ridge (Figures 1a and 1b), isolated from abyssal sedimentation processes and in the core of NADW, makes it a better site to monitor NSW. (2) During interglacials, Site 607 εNd values remain similar to present-day NADW (Figures 6 and 7); this is unlikely to be a coincidence. (3) The glacial-interglacial variability of Site 607 back through time, while different from Site 1063, is similar to the variability seen over the last glacial cycle at other North Atlantic sites, ranging from ~−13 to −14 during interglacials and −10 to −12 during glacials (Howe et al., 2016). The more positive glacial values are consistent with intrusion of SSW into the deep Atlantic during glacials. (4) The covariation of εNd and benthic δ13C in Site 607 (Figure 5) adds strong evidence that both proxies reflect the global overturning circulation in the Site 607 record. (5) Site 607 and several other sites in the equatorial (Sites 926 and 929) and South Atlantic (Sites 1267, 1088, and 1090) (Figure S3) all systematically show glacial-interglacial variability, with stronger SSW signals during glacials, and stronger SSW signals downstream from the North Atlantic at each point in time, with all εNd values bounded by those of the global end-members as reflected by Site 607 and North Pacific seawater (Figure S3). Such a combination of relationships cannot be maintained through such a long progression of glacial cycles if the εNd values represent regional processes. Rather, the only reasonable explanation for these observed north-south relationships through time is that the data in these deep sea cores reflect the AMOC.

If Site 607 εNd reflects AMOC variability, then the question arises as to the origin of the Site 1063 εNd variability. The agreement between εNd in fish debris and encrusted foraminifera in Site 1063 (Figure 3) strongly indicates that these data reflect the ambient seawater bathing the site throughout the study interval. As such, when Site 607 and 1063 εNd values converge during glacials (Figures 6 and 7), it suggests that Site 1063 is directly recording the influence of the AMOC. Instead, when Site 1063 εNd values are considerably more negative than Site 607 during interglacials, it suggests that AMOC shifts cannot be the cause. However, the origin of the substantial interglacial divergence remains equivocal.

5.3 Abyssal Source of Nd to the Bermuda Rise

Our εNd record at Site 1063 reveals that a source of highly negative εNd has been intermittently active during interglacials at least over the last 1 Ma (Figures 2, 6, and 7). However, as pointed out by Howe et al. (2016), similar negative εNd values have not been found in any other site in the intermediate and deep North Atlantic. We agree with their conclusions that the lack of widespread highly negative εNd signatures north of the deep Bermuda Rise indicates that the highly negative interglacial εNd values at Site 1063 do not reflect NSW. Rather, these values reflect an abyssal source of highly negative εNd at the Bermuda Rise that has negligible impacts on the reconstruction of AMOC processes using Nd isotopes. An abyssal source for the highly negative εNd is further supported by εNd data from the Sohm abyssal plain (Pöppelmeier et al., 2018). That study, at a deeper site in the Corner Rise (KNR197/10 GGC17), north of the Bermuda Rise, found εNd values much more negative than Site 1063's early Holocene values (~−20 vs. ~−16) (Figure 8). Further support that these highly negative εNd values do not represent the AMOC comes from the Site 607 εNd data indicating that during interglacial maxima, the εNd of the deep North Atlantic has remained stable through the Pleistocene, with values similar to present-day NADW (εNd ~ −14) (Figures 6 and 7). This stability is also supported by Fe-Mn crust records from intermediate depths in the western North Atlantic basin, which do not display the highly negative εNd values (Foster et al., 2007) (Figure 7). Although Fe-Mn crusts do not record rapid changes in ocean circulation due to their low (~30 kyr) temporal resolution, the cumulative εNd values that they display are consistently close to NADW through the last ~2 Ma. These Fe-Mn crusts, well situated at ~10° to the north and south of the Bermuda Rise at 1,800–2,000 m depth, are currently bathed by UNADW. Seawater sampled close to the Fe-Mn crusts show that modern water masses contain noticeable amounts of LSW, detectable through chlorofluorocarbon and dissolved O2 concentrations. However, their εNd signatures are characteristic of NADW (Lambelet et al., 2016), providing further evidence that increased LSW fluxes with highly negative εNd are not the source of the highly negative εNd values at the deep Bermuda Rise. All of these observations support our conclusions that the processes impacting Site 1063 are occurring in its midlatitude vicinity.

Details are in the caption following the image
(a) εNd data for the last glacial-interglacial cycle, comparing the different isotopic signatures recorded at deep sites in the Mid-Atlantic Ridge (MAR), the Bahama-Blake Outer Ridge (BBOR), the Bermuda Rise (BR), and the Corner Rise. (b) Location map with the sites discussed in the text and the flows of the main deep water masses. The deep MAR, represented in purple (Site U1313 at 3,427 m, equivalent to Site 607*) (Lang et al., 2016; Lippold et al., 2016), is used as a reference for εNd changes in the North Atlantic. The deep BBOR is represented in various tones of green (ODP Site 1060 at 3,481 m (light green), Site 1061 at 4,027 m (turquoise), and ODP Site 1062 at 4,761 m (dark green) (Pöppelmeier et al., 2019). The deep Bermuda Rise (BR) is represented in orange (ODP Site 1063* or its equivalent OCE326-GGC6 at 4,584 m) (Böhm et al., 2015; Roberts et al., 2010), and the abyssal corner rise (CR) is represented in blue (KNR197/10-GGC17 at 5,010 m) (Pöppelmeier et al., 2018). The NGRIP δ18O ice core record is used for time and climate reference. Gray bars represent cold periods.

5.4 Bermuda Rise Sediments and Nd Mobilization by Pore Water Flow

The drift deposits of the deep Bermuda Rise where Site 1063 is located are characterized by high sedimentation rates throughout the Pleistocene, although they have a considerable glacial-interglacial variability. Variability in glacial-interglacial sedimentation rates increases significantly during the Late Pleistocene, due to increased (up to 40–60 cm/kyr) detritus deposition during full-glacial periods, while they are lower during interglacials (~10 cm/kyr) (Channell et al., 2012; Keigwin & Jones, 1994; Keigwin et al., 1998).

The Bermuda Rise is in the open ocean, separated from the continent by a 5,500 m deep abyssal plain that is irregularly affected by deep currents (McCave, 2002). As a consequence, the drift sediments have variable sources (Figure 1a). Most of the sediment input to the Bermuda Rise comes from (1) transport, lateral advection, and focusing of fine-grained terrigenous detritus from North American margins by deep sea currents (such as the DWBC) and recirculating gyres (McCave, 2002); (2) resuspension of fine-grained sediments that have been transported downslope from the Eastern Grand Banks to the Sohm Abyssal Plain through the Laurentian Channel (Figures 1a and S1), possibly by turbidity currents (Keigwin & Jones, 1994; Laine & Hollister, 1981); and (3) pelagic sedimentation, which represents a relatively minor contribution (McCave, 2002). Such detritus from the margin of the northeastern United States and eastern Canada is expected to have εNd values of ~ −14 to −16 today (Figure 1), but detritus supplied by a melting ice sheet from the continental interior can contribute values as low as −20 to −28 (Casse et al., 2019). Therefore, the deep Bermuda Rise collects terrigenous detritus from multiple sources that can be expected to change through time, especially between glacial and interglacial periods. Such a change in detrital provenance since the last glacial maximum (LGM) is illustrated by terrigenous detritus εNd data from Bermuda Rise core OCE326-GGC6, with values ~−13 during LGM and ~−15 during the Holocene (Figure 4) (Howe et al., 2016; Roberts et al., 2012, 2010). A nearly identical glacial-interglacial change in εNd is observed in the detrital fraction at similar depths in the Blake Ridge (LGM −12.2; Holocene −15, Figure 4) (Gutjahr et al., 2008). The εNd shift of the detritus to more negative values during the deglaciation documents increased derivation from the North American cratonic shield regions during the Holocene (Figure 1a).

Diagenetic processes in oxygen-poor environments affecting pore waters can mobilize the manganese oxyhydroxides that hold the bulk of the rare earth elements, including Nd. In Site 1063, as already noted, only 4 of 33 analyzed foraminifera/fish debris sample pairs have εNd values that are outside of analytical error (Figure 3). The agreement between the foraminifera and fish debris suggests that diagenetic and/or secondary alteration by pore waters long after primary deposition is unlikely to be the cause of its unusually negative εNd values at Site 1063 and provides strong evidence that both archives record a deep seawater signal. However, the agreement by itself does not preclude the possibility that these archives obtain their εNd signatures from pore water flow at or near the time of deposition (e.g., Du et al., 2016; Haley et al., 2017). The occurrence of suboxic pore waters is often induced by the combination of high sedimentation rates and non-ventilated deep water masses. In Site 1063 these conditions were markedly enhanced during the LGM (Keigwin & Jones, 1994; Keigwin et al., 1998; Roberts et al., 2010, 2012). Thus, if pore water flux were the main cause of the negative εNd excursions at Site 1063, then the excursions should be observed during glacial periods, rather than only during interglacials (Figures 2, 6, and 7). However, the εNd data from Bermuda Rise core OCE326-GGC6 (Figure 4) indicate an offset between terrigeneous detritus and fish debris foraminifera during the LGM (~−13 vs. ~−11, respectively), while there is no offset during the Holocene (averaging at εNd ~ −15), suggesting that regional pore water flow under suboxic conditions did not significantly impact εNd values during the LGM (Roberts et al., 2010, 2012). Therefore, since the less oxic conditions occurred during glacials, when the negative εNd excursions were not observed and the Site 1063 εNd values agree with Site 607, pore water flow under such conditions cannot explain the negative εNd excursions at Site 1063. The integrity of the seawater signal at Site 1063 is further evidenced in the literature by agreement between core top εNd measurements from foraminiferal Fe-Mn coatings and present-day bottom seawater at the same location within measurement error (Roberts et al., 2010).

Recent studies in the NE Pacific and Tasman Sea (Abbott, 2019; Abbott, Haley, & McManus, 2015; Abbott et al., 2019) suggest that additional variables need to be taken into consideration when arguing modification of bottom waters εNd by pore waters, as authigenic Fe-Mn oxides seem to act as both a source and sink of pore water and bottom water Nd. The benthic flux model from Abbott, Haley, and McManus (2015) concludes that the potential for alteration of bottom waters εNd is determined by the Nd concentrations of the water mass ([Nd]wm), the difference between the isotopic composition of the flux (i.e., pore waters) and the water mass (εNd flux-wm) and the magnitude of the benthic flux itself (FNd). Therefore, a high FNd or/and a large ∆εNd flux-wm could alter the εNd of the bottom water, even with short time exposure of the water mass to the sediment. On the other hand, long time exposures would have a smaller impact on the εNd of the water mass if [Nd]wm were large or FNd or/and a ∆εNd flux-wm were small. Pore waters with altered εNd signatures can potentially affect the bottom waters εNd signature at the Bermuda Rise throughout the studied time interval (it is likely when the differences are observed between some fish debris-foraminifera samples). However, we need to consider that the regional contexts of the mentioned studies are very different to that of Site 1063. It is known that certain components of the sediments (such as volcanic or basaltic material) may be more reactive during diagenesis, becoming more influential sedimentary Nd sources (Abbott et al., 2019; Du et al., 2016). Therefore, the sediments from the deep Pacific are, in most cases, far more reactive than those of other basins because of the abundance of a widespread volcanic component. In fact, Abbott et al. (2019) already observed much higher [Nd] in pore waters in the deep Pacific than in the Tasman Sea. During glacial periods at Site 1063, sediments have less negative εNd values (showing Appalachian provenance); thus, ∆εNd flux-wm is smaller. We believe that this may contribute to the preservation of the bottom water εNd signature observed during glacials. However, it is difficult to fully explain the high variability observed in our εNd record at Site 1063 only as a result of Nd mobilization by pore water flow, especially when there are evidences pointing to an abyssal origin of the very negative εNd signature in the northwestern Atlantic, which dissipates toward shallower and southern sites, such as Site 1063 (Howe et al., 2016; Pöppelmeier et al., 2019, Figure 8).

5.5 BNL and Highly Negative εNd at Site 1063

Previous studies have suggested that the peak in the early Holocene at the Bermuda Rise (Figures 4 and 8) is the result of a BNL of particles with very negative εNd signature reaching the site (Howe et al., 2016; Pöppelmeier et al., 2019). We have presented evidence that agrees with the idea of an abyssal source of Nd to the bottom water during interglacial/interstadial periods. The transport of particulate and dissolved Nd through a BNL supports the observation that the extreme εNd signatures do not necessarily correspond with the highest sedimentation rates. In the modern ocean, intense BNLs, often with a thickness of >600 m, develop in the deep Bermuda Rise region as a consequence of high bottom kinetic energy and eddy flow (Gardner et al., 2018) (Figures 1 and S1). They are associated with high seawater Nd concentrations in excess of >30% greater than expected from conservative mixing of the regional deep water masses (Figure 9) (Hartman, 2015; van de Flierdt et al., 2016), indicating an additional external Nd source (Hartman, 2015). Modern seawater samples showing excess [Nd] near Site 1063 from close to the core of the BNL are characterized by lower εNd values (~1 εNd unit) than waters unaffected by the BNL (Hartman, 2015) (Figure 9). The effect of particulate Nd on the seawater εNd signatures is also supported by the discrepancy of ~1 ε unit observed between filtered and unfiltered seawater samples from the region (Lambelet et al., 2016; Piepgras & Wasserburg, 1987). Core top data from the Corner Rise, a deeper site (5,010 m) slightly north of Site 1063 in the Sohm Abyssal Plain, also within the limits of the current BNL (Gardner et al., 2018), show even more negative values (εNd ~ −14.7) (Pöppelmeier et al., 2018).

Details are in the caption following the image
Plots representing the results of a comparison between measured and calculated (a) Nd concentrations and (b) Nd isotopic compositions (middle) in a W-E transect in the North Atlantic (after Hartman, 2015). An offset over 6 in ∆[Nd] (dark red in the top plot) shows how the benthic nepheloid layer (BNL) in today's deep NW Atlantic works as an external source for Nd. This is coincident with a ∆εNd up to −1.5 ε units, suggesting that the particulate Nd may be affecting the bottom seawater signature. ODP Site 1063 is located within the region where bottom seawater may be affected by particle-seawater exchange of Nd. (c) Location map shows the W-E transect from GEOTRACES cruise GA03 where seawater samples were obtained. On top of it, main regions and deep water masses in the NW Atlantic. Brown shadow represents the benthic nepheloid layer (Gardner et al., 2018).

The recirculating gyres that generate the BNL are linked to the intensity and location of the Gulf Stream and the DWBC. Gardner et al. (2017) conclude that the highest particulate concentrations within the BNL are directly below the Gulf Stream path and that Gulf Stream meanders enhance the area of the BNLs and the capacity of the deep gyres to mobilize the surface sediments. While there are no paleo-records of BNLs, in this study we infer the behavior of past BNLs based on what we know from its relation to the DWBC and the Gulf Stream today and what we know from these currents in the past. For the most recent glacial-interglacial cycles there are εNd records with enough resolution to compare with established DWBC variability.

Studies of stable isotopes and physical properties (grain sizes, color, and magnetic properties) of sediments from the Bahama-Blake Outer Ridge (Figure 1), off the continental slope in the subtropical North Atlantic, have concluded that the depth and intensity of the main core of the DWBC have changed through the late Pleistocene (Figure 10a), generally becoming stronger and deeper during interglacials (Bianchi et al., 2001; Evans & Hall, 2008; Franz & Tiedemann, 2002; Hall & Becker, 2007; Yokokawa & Franz, 2002). During the most recent glacial-interglacial cycle, the continental detritus and foraminiferal/fish debris εNd values show significant differences during cold periods (i.e., the LGM and Younger Dryas), when the core of the DWBC was shallower than 2,500 m (Evans & Hall, 2008), and during the late Holocene, with a DWBC >3,000 m (Evans & Hall, 2008). Instead, the εNd values are similar during the warm periods (i.e., during the Bølling-Allerød and early Holocene), when the DWBC flows with high intensity at intermediate depths (<3,000 m) during the transition between full-glacial and full-interglacial modes (Figure 4). This indicates that the depth of the DWBC is linked to the incorporation of the terrigenous εNd signature to Site 1063.During full-interglacial stages (Figures 4, 10c, and 10d) MIS 9e, 5e, and Holocene interglacial maxima, the DWBC's main core was at ~4,000 m and showed high intensity (Bianchi et al., 2001; Evans & Hall, 2008; Yokokawa & Franz, 2002). Coinciding with the terminations (~10, ~125, and ~330 ka, Figures 10b and S4), Site 1063 records eNd peaks of ~−16 that increase gradually toward values similar to the modern seawater (~−12.5; blue star, Figures 10b and S4), suggesting that the BNL location and composition during full-interglacials may be similar to today, not (or marginally) affecting Site 1063. The peaks of ~−16 have been suggested to be the result of advected waters with more negative εNd (~ −20) from the Sohm Abyssal Plain toward the site (Pöppelmeier et al., 2018). Surface waters with εNd ~ −16 and high Nd concentrations in the modern ocean around the Grand Banks (Lambelet et al., 2016) suggest that the region may also be a source for those εNd signatures. The Grand Banks is an extensive platform that undergoes a lot of changes during glacial-interglacial cycles. For example, the area is submerged today but was exposed during glacial maxima as a consequence of the glacial sea level drop, and its surface partly eroded by the extended ice sheet (Piper et al., 1994; Shaw et al., 2006). Therefore, Grand Banks may also act as a source of poorly weathered, highly reactive terrigenous detritus that contributes Nd to the DWBC, reaching Site 1063 as it deepens and intensifies.

Details are in the caption following the image
(a) Schematic diagram shows depth and intensity variability in the fast core of the deep western boundary current (DWBC) for different time intervals (black and thicker bars represent more intensity) (after Bianchi et al., 2001; Evans & Hall, 2008; Franz & Tiedemann, 2002; Hall & Becker, 2007; Yokokawa & Franz, 2002). The light pink bar represents the suggested depth at which an intense main core of the DWBC enhances the deep-gyre system; (b) Site 1063 εNd for intervals MIS 10 to 8 (green, this study), MIS 6 to 4 (orange, Böhm et al., 2015), and LGM-Holocene (blue, Roberts et al., 2010). Notice that εNd-signature from modern seawater near Site 1063 (blue star, Lambelet et al., 2016) matches the observed values at previous MIS right after interglacial maxima, being the location of the core of the DWBC similar to today (3,500–4,100 m depth; Bianchi et al., 2001). Vertical gray bars represent cold periods.

Full-glacial stages (e.g., MIS 10, 6, and 2) are characterized by a less intense DWBC core, as shallow as ~2,200 m (Bianchi et al., 2001; Evans & Hall, 2008; Yokokawa & Franz, 2002). During the LGM, the terrigenous εNd data indicate that the sediment supply to the Bermuda Rise is limited to that of the Appalachians rather than the Canadian Shield (εNd ~ −13, Roberts et al., 2010, 2012), due to the extended ice sheet over the Shield. Foraminifera-fish debris εNd values are even more positive than during full interglacials (between ~−10 and −12 vs. ~−13; Figure 10b), suggesting that there is no alteration of the seawater signature and that there may be an increasing contribution of AABW to the site compared to the interglacials. Only weak glacials, such as MIS 8 or 4, show εNd variability between ~−16 and −12; this could be explained either by a source of Nd with negative εNd from the Grand Banks—there is evidence that the region was not covered by ice during MIS 8 (Piper et al., 1994)—or by advection of altered abyssal waters due to an active deep-recirculating gyre system over the site.

Following interglacial maxima (e.g., MIS 9e), the DWBC flow remains intense but gradually shoals to 2,500–2,800 m (Figure 10a), before deepening again to ~3,000 m during interstadials (i.e., MIS 9a) (Yokokawa & Franz, 2002). In Site 1063 (Figure 10b), MIS 10 to 8 is characterized by highly variable εNd values between −11.3 and −20.6. As already mentioned, during MIS 10 to 8 the highly negative εNd excursions are missing in the deep Bermuda Rise when the depth of main core of the DWBC was at its shallowest (MIS 8 and 10) and deepest (MIS 9a and 9e). Rather, these peaks are present when the DWBC flow is intense and at intermediate depths (<3,000 m). During MIS 6 to 4, the behavior of the DWBC is similar to MIS 10 to 8 (Bianchi et al., 2001), and εNd also shows a similar pattern. Thus, the occurrence of highly negative εNd values at Site 1063 appears to be strongly linked to DWBC flow speed changes (Figure 10).

We suggest that the highly negative εNd values reflect a third (hybrid) circulation mode (Figure 11b) during deglaciations and some stadial-to-interstadial warmings that combines the presence of an intense and shallow DWBC and a displacement of the Gulf Stream over Site 1063. The melting of the continental ice sheet brings sediments from the Canadian Shield with highly negative εNd values to the western North Atlantic basin and around Grand Banks (Figure 1) (Casse et al., 2019). On the one hand, the intense and shallow DWBC results in increased interaction with poorly weathered sediments from the continental slope around Grand Banks and further north, favoring exchange of Nd. On the other hand, increased interaction with an intensified Gulf Stream promotes increased meandering and deepening of the deep gyre system at ~5,000 m depth. The deepening and displacement of the deep gyre system mobilizes sediments by generating BNLs in the Bermuda Rise and the Sohm Abyssal Plain. The freshly ground bedrock by the Laurentide ice sheet with a low degree of chemical weathering contains Nd and other REE that are relatively mobile and able to exchange with the bottom water when mobilized by the BNL, imparting a highly negative εNd-signature on the deep water. This effect did not occur during glacials, when the continental detritus deposited on the Bermuda Rise had a different provenance with higher εNd (Figures 1 and 4) which we speculate was more chemically weathered and less reactive. More importantly, a weaker glacial AMOC, associated with the incursion of SSW and a shoaling of the NSW-SSW mixing zone, resulted in a weak deep recirculating gyre system that led to the reduction or even the dissipation of the BNL. Therefore, during full glacials or cold periods, the deep Bermuda Rise would not be affected by the particle-loaded seawater that forms the BNL, with the result that the εNd of the deep water reflected the water mass signature imparted from the AMOC-derived mix of NSW-SSW (Figure 11c).

Details are in the caption following the image
Schematic diagrams depicting the hypothetical behavior of the main water masses, currents and benthic nepheloid layer (BNL) during (a) full-glacial and (b) deglacial circulation modes. (c) The full-interglacial mode diagram is based on modern NW Atlantic circulation. The right side represents a map view of the region. The rippling in the Gulf stream (red arrow) represents the degree of meandering. Font and symbol sizes and arrow thickness relate to relative intensities of the streams and currents or continental weathering (black and white arrows on the continent boundaries). Big white arrows indicate the most probable continental sources of more negative εNd. The black line represents the section used in the profiles (left). Vertical profiles show the distribution of the main deep water masses in depth. Approximate water mass boundaries are marked as dotted white lines. Black circular arrows represent the deep-gyre system. Circles represent the southward flow of the Deep Western Boundary Current (blue) and the northward flow of the Gulf stream (red).

6 Conclusions

Nd-isotope records from ODP Site 1063 extending to ~1 Ma, in the deep Bermuda Rise, show extremely negative εNd values during parts of interglacials that are much lower than that of NADW. The same values are observed in fish debris and Fe-Mn encrusted foraminifera, indicating this signal represents the bottom water near the site. Several studies have suggested the cause of the extreme values to be changes in the εNd and/or production of the precursor water masses that form NADW. However, such signatures are not seen in other intermediate North Atlantic sites, including Site 607, located off the abyssal seafloor in the core of NADW, which shows εNd values similar to NADW during interglacials over the same time interval, fluctuating with higher values during glacials. εNd values at Site 607 covary with lower benthic δ13C values. The combined εNd-benthic δ13C data are consistent with NSW dominating the site during interglacials similar to today and incursions of SSW during glacials. The extreme negative εNd at Site 1063 during interglacial maxima are corroborated only by a deep site at 5,010 m in the Corner Rise in the Sohm Abyssal Plain, north of the Bermuda Rise (KNR197/10 GGC17), which shows extreme negative εNd values during the early Holocene, in this case even more negative than at the Bermuda Rise. We conclude that Site 607 better represents the deep North Atlantic and the AMOC over the past ~1 Ma. Further evidence is given by a north-to-south transect of the Atlantic basin showing data at interglacial and glacial maxima over this period, where at each point in time, the Site 607 data that fit well with more southerly locations show higher fractions of SSW through time. Such relationships are likely to occur only if the data along the transect reflect the AMOC. In contrast to the interglacials, at Sites 1063 and 607, the εNd values agree during most glacials. Thus, the data indicate that the Site 1063 data reflect the AMOC during glacials, while there is external Nd added to the deep Bermuda Rise during interglacials that is not propagated into the global thermohaline circulation.

We conclude that the extreme negative εNd values at Site 1063 reflect particle-seawater exchange as a consequence of input of freshly ground bedrock, with a low chemical weathering index resulting in enhanced rare earth element mobility, during major ice sheet retreat such as deglaciations and stadial-to-interstadial transitions. The result is a deep, regionally restricted thin layer of seawater tagged with this extreme εNd signature that is not representative of the AMOC. On the other hand, during glacials the Site 1063 εNd signature reflects the effects of the AMOC.

A mechanism that would facilitate this exchange is a BNL in the deep Bermuda Rise, generated by the dynamic interaction between the northward flowing Gulf Stream and the southward flowing DWBC. The intensity of the deep gyre system, the changing location of the NSW/SSW mixing zone, and the growth and retreat of the North American ice sheet determine the location, extensiveness, strength, and composition of the BNL. We propose three scenarios to explain the periodical masking of the deep Atlantic Nd isotope signature at Site 1063. (1) In “full-glacial mode,” as a result of a weak Gulf Stream and incursion of SSW, the BNL is weak or absent, and in εNd signature the deep Bermuda Rise reflects the incursion of SSW. (2) In the “deglacial mode,” also operating during transitions to interstadials, the combination of freshly ground bedrock from the interior Canadian Shield by the North American ice sheet, contributed by erosion associated with its retreat, along with an intense but southwardly displaced Gulf Stream and an intense DWBC, results in deepening of the deep gyres; this results in the development of the BNL and generates a deep water layer with an extreme negative εNd signature. (3) In the “full-interglacial mode,” operating in the present day, the BNL is mainly located north of the deep Bermuda Rise and its impact on Site 1063 is marginal. Further investigations to test this hypothesis would include targeting appropriate climate transitions and regions likely impacted by the BNLs.


  • AABW
  • Antarctic Bottom Water
  • AAIW
  • Antarctic Intermediate Water
  • AMOC
  • Atlantic Meridional Overturning Circulation
  • BNL
  • Benthic Nepheloid Layer
  • BBW
  • Baffin Bay Water
  • BR
  • Bermuda Rise
  • BW
  • Bottom Water
  • CR
  • Corner Rise
  • DBWC
  • Deep Western Boundary Current
  • DSDP
  • Deep Sea Drilling Project
  • DSOW
  • Denmark Strait Overflow Water
  • ISOW
  • Iceland-Scotland Strait Overflow Water
  • LGM
  • Last Glacial Maximum
  • Lower North Atlantic Deep Water
  • LSW
  • Labrador Sea Water
  • MIS
  • Marine Isotope Stage
  • NADW
  • North Atlantic Deep Water
  • NPDW
  • North Pacific Deep Water
  • NSW
  • Northern-Sourced Waters
  • ODP
  • Ocean Drilling Program
  • SSW
  • Southern-Sourced Waters
  • ULSW
  • Upper Labrador Sea Water
  • Upper North Atlantic Deep Water
  • εNd: Neodymium isotope ratio, 143Nd/144Nd, expressed as εNd = ((εNdsample/εNdCHUR)-1) × 104, the fractional deviation from the “CHondritic Uniform Reservoir,” an estimate for the average value of the Earth in parts per 10,000, here using 0.512639 (Jacobsen & Wasserburg, 1980).

    Conflict of Interest

    The authors declare no competing interests.


    We thank Maureen Raymo, Bärbel Hönisch, Jesse Farmer, Heather Ford, and Laura Haynes for discussions. This research was mainly supported by NSF grant OCE 14-36079. Additional support came from the CTM2016-75411-R project (MINECO, Spain) and the Storke Endowment of the Department of Earth and Environmental Sciences at Columbia University. M. J-S. and L. D. P. acknowledge support from predoctoral grant BES-2017-079622 and from the Ramón y Cajal program, respectively (MINECO, Spain). Samples were provided by the International Ocean Discovery Program (IODP), sponsored by the NSF and participating countries under management of Joint Oceanographic Institutions. S. L. G. and L. D. P. recruited funding and guided this project. M. J-S., L. D. P., and S. L. G designed the study and sampling strategy. M. J-S produced the εNd-record for ODP Site 1063. L. B. gave assistance with the isotopic analyses. P. F. provided part of the samples and developed the age model for the interval MIS 29-18. All authors provided substantial scientific and prose contributions to the final manuscript. This is Lamont contribution number 8459.

      Data Availability Statement

      Original data from this study will be stored on the EarthChem Library at the Interdisciplinary Earth Data Alliance (https://doi.org/10.26022/IEDA/111585).