Constraints on the Northwestern Atlantic Deep Water Circulation From 231Pa/230Th During the Last 30,000 Years

Global climatic changes during the last glacial and deglacial have been related to variations of the Atlantic Meridional Overturning Circulation (AMOC). Here, we present new and refined Pa/Th down‐core profiles extending back to 30 ka BP from the northwestern Atlantic along the Atlantic Deep Western Boundary Current, which is the main component of the southward deep backflow of the AMOC. Besides the well‐known Bermuda Rise records, available high‐resolution Pa/Th data in the northwestern Atlantic are still sparse. Our new records along with reconstructions of deep water provenance from Nd isotopes constrain the timing and magnitude of past changes in AMOC from an additional northwestern Atlantic region forming a depth transect between 3,000‐ and 4,760‐m water depth. Our extended and improved data set confirms the weakening of the AMOC during deglacial cold spells such as Heinrich Event 1 and the Younger Dryas interrupted by a reinvigoration during the Bølling‐Allerød interstadial as seen in the prominent Pa/Th records from the Bermuda Rise. However, in contrast to the Bermuda Rise records, we find a clearly reduced circulation strength during the Last Glacial Maximum in the deep Atlantic.


Introduction
Deep water formation in and around the North Atlantic represents an important feature of Earth's climate system, since it redistributes water masses between the surface and the deep ocean as well as between the Northern Hemisphere and the Southern Hemisphere. Accordingly, reconstructions of the Atlantic Meridional Overturning Circulation (AMOC) by means of various proxies and methods have been carried out extensively in the last decades in order to determine its behavior under very different past climatic boundary conditions (e.g., Curry & Oppo, 2005;Lynch-Stieglitz, 2017;McManus et al., 2004). Today, the majority of the deep western Atlantic basin is occupied by North Atlantic Deep Water (NADW), which is transported southward primarily by the Atlantic Deep Western Boundary Current (DWBC) (Figure 1; Johnson, 2008;Rhein et al., 2015). NADW is underlain by southern-sourced water (SSW), which reaches into the deep North Atlantic up to 40°N (Johnson, 2008). The modern distribution of these water masses is clearly reflected by nutrient concentrations such as phosphate (Garcia et al., 2014) or physical properties such as salinity and potential temperature (Broecker et al., 1985;Locarnini et al., 2013;Zweng et al., 2013). The distribution of these water masses changed on millennial scales and has been reconstructed with nutrient-based proxies such as δ 13 C and Cd/Ca (e.g., Curry & Oppo, 2005;Keigwin, 2004;Oppo et al., 2018). An alternative approach for the identification of different water masses is the radiogenic neodymium isotope proxy (denoted as εNd). Neodymium isotopes are extracted from the authigenic phases of bulk sediments, foraminifera, or fish teeth on which ideally the water mass Nd isotopic signature has been imprinted (e.g., Blaser et al., 2016;Gutjahr et al., 2008;Howe et al., 2016;Piotrowski et al., 2005). Different reconstructions showed that during the Last Glacial Maximum (LGM) as well as during short northern hemispheric cold spells such as Heinrich Stadials 1 and 2 (HS1 and HS2) and the deglacial Younger Dryas (YD), the balance of these water masses shifted toward a predominance of SSW filling most of the deep western Atlantic reaching north as far as 50-60°N (Curry & Oppo, 2005;Gutjahr & Lippold, 2011;Marchitto & Broecker, 2006;Roberts et al., 2010). However, recent studies suggested the presence of a deep northern-sourced water mass in the northwestern Atlantic also during the LGM as derived from stable carbon isotopes (Keigwin & Swift, 2017) and εNd data (Howe et al., 2016;Pöppelmeier et al., 2018). ©2019. The Authors. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
Alongside reconstructions of past water masses distribution, the knowledge of the past overturning strength from sensitive locations and different water depths is of equal importance. Reconstructions of bottom water circulation at the Blake-Bahama Outer Ridge (BBOR; Figure 1) using sortable silt suggested a reduced circulation during the LGM below 3,000-m water depth but with a strong shallow circulation cell above (Evans & Hall, 2008). These authors further suggest stronger circulation in the abyssal (>4,000 m) northwestern Atlantic, which they interpret as intensified SSW inflow. Sortable silt is a valuable proxy for reconstructing local bottom currents (McCave et al., 1995;McCave et al., 2017) but provides less information about the large-scale AMOC. For such large-scale AMOC reconstructions, the kinematic circulation strength proxy 231 Pa/ 230 Th excess is widely applied in the Atlantic McManus et al., 2004;Ng et al., 2018;Yu et al., 1996). Both, 231 Pa and 230 Th, are daughter isotopes from the radioactive decay of the uranium isotopes 235 U and 234 U, respectively. Uranium (and its isotopic composition) is homogeneously dissolved in the world ocean due to the long residence time in the order of 400 ka (Henderson & Anderson, 2003). Accordingly, the production of both daughter isotopes is of a constant activity ratio (0.093, hereafter "production ratio"). In contrast to uranium, protactinium and thorium are highly particle reactive, leading to short oceanic residence times of 100-200 and 10-40 years, respectively (Henderson & Anderson, 2003). Due to the slightly longer residence time of protactinium (roughly the time it takes for NADW to reach the Southern Ocean; Yu et al., 1996), 231 Pa and 230 Th are fractionated in dependence of the circulation strength (i.e., changes in AMOC intensity). Hence, low sedimentary 231 Pa/ 230 Th values today in the deep Atlantic reflect strong NADW advection (meridional 231 Pa export to the Southern Ocean), while higher values are consistent with a weaker AMOC state (Luo et al., 2010;McManus et al., 2004).
In the Southern Ocean, 231 Pa is preferentially scavenged from the water column into sediments due to its high affinity to particles consisting of biogenic opal (Chase et al., 2003). From this observation, a critical view on the usage of 231 Pa/ 230 Th as a circulation proxy arose, since 231 Pa/ 230 Th might be strongly linked to particle fluxes and their compositions (Chase et al., 2003). For regions with high particle fluxes and characteristic particle compositions (e.g., the Southern Ocean) or weak ocean circulation (e.g., the Pacific), this effect seems to control sedimentary 231 Pa/ 230 Th ratios (Anderson et al., 1983;Chase et al., 2003;Costa et al., 2017;Hayes et al., 2013). In contrast, in the western Atlantic, with its relatively moderate particle fluxes and a pronounced AMOC, 231 Pa is effectively exported from the North to South Atlantic by advection as derived from a 231 Pa deficit of at least 26% (Deng et al., 2018) (a former study by Yu et al. (1996) mentioned up to 50%) supporting the applicability of 231 Pa/ 230 Th as a circulation proxy in the Atlantic.
Previous studies using the 231 Pa/ 230 Th proxy have concluded that a strong circulation of the North Atlantic prevailed during the Holocene (Hoffmann et al., 2018) and pronounced weakenings of AMOC strengths occurred during past cold phases (Bradtmiller et al., 2014;Gherardi et al., 2009;Lippold et al., 2016;Figure 1. Overview map of the northwest Atlantic. Core locations presented/discussed in this study are indicated in red. The main study area is the northwestern Atlantic with the Blake-Bahama Outer Ridge and the Bermuda Rise. The blue arrow depicts the Deep Western Boundary Current path from the northern North Atlantic along the North American margin (Thornalley et al., 2013). The table gives core parameters and references. McManus et al., 2004;Ng et al., 2018). Nevertheless, a clear picture of the vertical structure of the deglacial circulation and the evolution of AMOC is still lacking, in particular due to insufficient spatial and temporal proxy coverage.
Here we present four new high-resolution 231 Pa/ 230 Th down-core profiles forming a depth transect at the BBOR (3,000 to 4,760 m; Figure 1) spanning the last 30 ka. The BBOR is a well-studied location (e.g., Evans & Hall, 2008;Keigwin, 2004) in the direct flow path of the DWBC that allows for the investigation of advection rates depending on the water depth. Additionally, we provide three updated and improved records from a prior study . With these new records from the deep BBOR, we are able to reconstruct the evolution and changes in AMOC strength in greater detail.

Setting and Age Model
We analyzed 231 Pa/ 230 Th from ODP Leg 172 Sites 1059, 1060, 1061, and 1062 back to 30 ka. Sites 1059, 1060, and 1061 are located on the crest of the Blake Outer Ridge, which is formed by drift sediments (Keigwin & Jones, 1994), while Site 1062 is located further south on the Bahama Outer Ridge (Figure 1). The four sites form a depth transect from 3,000-to 4,760-m water depth. In addition, we improved the 231 Pa/ 230 Th records (partly reanalyzed and recalibrated, see Table S5 in the supporting information) of KNR140 12JPC (hereafter 12JPC) on the BBOR and cores GeoB1515-1 and GeoB1523-1 from the Ceara Rise in the equatorial western Atlantic . Age models for Sites 1059 to 1062 are provided by Pöppelmeier et al. (2019) and are based on the correlation of carbonate concentration for Sites 1059 and 1062 (Grützner et al., 2002) to neighboring cores KNR140 GGC39 (Keigwin & Schlegel, 2002) and KNR31-GPC9 (Keigwin & Jones, 1994), respectively. Additionally, the age model consists of 12 14 C dates on Site 1060, two dates on Site 1062, and four recalibrated 14 C dates on Site 1059 (Hagen & Keigwin, 2002;Pöppelmeier et al., 2019). We further present new age models for GeoB1515-1 and GeoB1523-1 from the Ceara Rise, which have been improved by recalibrated 14 C dates and one new 14 C date for GeoB1515-1. This new 14 C date was measured at the LARA laboratory at the University of Bern, Switzerland (Gottschalk et al., 2018). We used the CALIB 7.1 online tool tied to the Marine13 curve (Reimer et al., 2013) for calibration of all 14 C dates with the standard 400-year reservoir age correction.

Analytical Procedure for Pa, Th, and U Isotope Measurements
Sediment samples were analyzed for the radioisotopes 230 Th, 231 Pa, 232 Th, 234 U, and 238 U. Purification and separation of the elements followed the protocol described in Süfke et al. (2018). Before chemical treatment samples were spiked with 233 Pa, 229 Th, and 236 U. The short-lived 233 Pa isotope (t 1/2 = 27 d) was milked from a 237 Np solution using the procedure described by Regelous et al. (2004). The 233 Pa spike was calibrated against the reference material UREM-11 (Süfke et al., 2018) and an internal pitchblende standard (Fietzke et al., 1999). Protactinium isotopes were measured using an Element 2 HR-ICP-MS in the Institute of Earth Sciences at Heidelberg University and a Neptune Plus MC-ICP-MS in the GeoZentrum Nordbayern at the Friedrich-Alexander University in Erlangen equipped with a retarding potential quadrupole filter. Uranium and thorium isotopes were measured with two Neptune Plus MC-ICP-MS at GEOMAR Helmholtz Centre for Ocean Research in Kiel and at the GeoZentrum Nordbayern in Erlangen.
A detrital correction ( 238 U/ 232 Th) of 0.55, in agreement to overall minima of bulk 238 U/ 232 Th, was applied to the measured activities of 231 Pa and 230 Th in accordance with the typical lithogenic activity ratio for 238 U/ 232 Th of 0.5 to 0.6 (Henderson & Anderson, 2003) in the western Atlantic. The lithogenic 238 U/ 232 Th, and hence the detrital correction on the calculation of 231 Pa/ 230 Th, may vary with time (Missiaen et al., 2018) in particular during times of high detrital and/or authigenic contributions. One control parameter for a potentially changing detrital factor is given by the 230 Th xs0 / 232 Th ratios (Missiaen et al., 2018). Throughout the glacial to deglacial parts of the records, the 230 Th xs0 / 232 Th activity ratios are relatively constant, pointing to a stable detrital phase, but show strong increases with the onset of the Holocene ( Figure S1). However, the Holocene sections of our 231 Pa/ 230 Th records are insensitive to changes in the detrital correction ( Figure S2). Furthermore, X-ray fluorescence (XRF) data (cf. sections 2.3. and 3.2.) support a uniform detrital sediment composition throughout the complete records ( Figure S4). Hence, we used a constant detrital correction of 0.55 that has also previously been used for Site 12JPC . The ingrowth of 231 Pa and 230 Th from authigenic uranium was calculated and corrected as described by Henderson and Anderson (2003). Finally, 231 Pa and 230 Th excess concentrations were decay corrected to the time of deposition. All individual isotope concentrations are provided in the supplement (Tables S1-S6).

Biogenic Opal and Major Elements
As well as the measurements of 231 Pa/ 230 Th, we analyzed the content of biogenic opal in the sediments. High fluxes of biogenic opal may increase 231 Pa/ 230 Th ratios independently of the circulation strength, due to the high affinity of 231 Pa to it (e.g., Chase et al., 2002Chase et al., , 2003Rutgers van der Loeff et al., 2016). Opal concentrations were analyzed by automated leaching following the procedure described by Müller and Schneider (1993). Furthermore, we measured the bulk sediment content of Al, Si, Ti, Fe, and K of discrete samples (Table S7) with a fourth-generation Avaatech XRF core scanner at the Institute of Earth Sciences at Heidelberg University, using a 10-kV Rh anode X-ray tube without a filter, a 1,000-mA current (500 mA for Site 1060), and a counting time of 30 s.

Selection Criteria for Compilation of Existing 231 Pa/ 230 Th Profiles From the West Atlantic for Comparison
In order to provide a comprehensive 231 Pa-/ 230 Th-based picture of the deep AMOC evolution in the western Atlantic sector, we compiled new and existing records from this basin. We only used sites from the western basin of the Atlantic since records from the eastern basin show noticeably different features and deglacial evolution due to different circulation regimes in both basins (Bradtmiller et al., 2007;Gherardi et al., 2009;Howe et al., 2017;Ng et al., 2018). For this compilation we excluded sites that are located on or near the Mid-Atlantic Ridge (MAR) in order to avoid potential effects of hydrothermal activity on regional 231 Pa/ 230 Th Lund et al., 2019;Pavia et al., 2018), which are not yet satisfyingly resolved (Bradtmiller et al., 2007;Gherardi et al., 2009;Lippold et al., 2016). We further excluded sites where records are not continuous from the LGM to the Holocene or from which the original authors consider specific time intervals as questionable (Gherardi et al., 2009;Lippold et al., 2011). Overall, we thus compiled two cores from the Bermuda Rise (Lippold et al., 2009;McManus et al., 2004), five from the BBOR (this study, Lippold et al., 2016), and five from the central/equatorial western basin (Bradtmiller et al., 2007;Lippold et al., 2016;Ng et al., 2018), which fulfill our criteria (Figure 1). In the further interpretation, we combined the two cores GeoB1523-1  and EW9209-3JPC (Ng et al., 2018) to a single record since both sites are situated at the same water depth and nearly the same location.

231 Pa/ 230 Th
All new records show a similar millennial-scale variability closely following the prominent Bermuda Rise record (McManus et al., 2004;Lippold et al., 2009; Figure 2). Youngest 231 Pa/ 230 Th values are consistently low (between 0.053 and 0.060) indicative of strong 231 Pa export that was established during the early Holocene (<10 ka) (Figures 2 and 3). Pronounced variability during the deglaciation including the prominent climatic episodes YD, Bølling-Allerød (B/A), and HS1 is present in all new records, except for the reevaluated Site 12JPC, which shows less variability during the YD and B/A events.   (Gutjahr & Lippold, 2011;Lippold et al., 2009;McManus et al., 2004;Roberts et al., 2010). The blue vertical bars indicate the cold periods YD, HS1, and HS2, respectively. The yellow vertical bar marks the B/A warm period. While 231 Pa/ 230 Th ratios of Site 1059 exhibit no distinct peak during HS2, the deeper BBOR cores display generally higher variability during this period and values as high as during HS1, including few values exceeding the production ratio of 0.093. In particular, Site 1060 shows a double peak feature similar to the record at ODP Site 1063 (Lippold et al., 2009).

Opal and Major Elemental Abundances
Low preserved opal concentrations below 4% were measured in all presented BBOR cores (Figure 2). Further, a higher opal content should be detectable by increased Si/Al ratios (McManus et al., 2004) but is not present in the new XRF records from the BBOR ( Figure S4). In addition to the Si/Al ratio, we used other elemental ratios obtained by XRF for analyzing the provenance and major composition of the sedimentary phase. Ratios of Ti/Al, K/Al, K/Ti, and Ti/Fe show little variation throughout the entire records ( Figure S4), pointing to an invariable detrital phase (Rothwell & Croudace, 2015), which in turn supports the selection of a constant detrital correction as outlined in section 2.2.

Potential Primary Particle Influences on 231 Pa/ 230 Th
As seen from its high 231 Pa/ 230 Th, 231 Pa exported southward via NADW ends in the sediments of the Southern Ocean (Rutgers van der Loeff et al., 2016) due to the opal-dominated particle flux in this region (Anderson et al., 2009;Chase et al., 2003;Walter et al., 1997). Therefore, the role of particle flux and particle composition on controlling sedimentary 231 Pa/ 230 Th besides ocean circulation is emphasized (Chase et al., 2002;Geibert & Usbeck, 2004). While biogenic opal has not been found to be a significant scavenging phase in the modern North Atlantic Ocean , we also exclude opal as a major driver of 231 Pa/ 230 Th at the BBOR for the past. Preserved opal concentrations are constantly low at all BBOR sites during the last 30 ka (even in the presence of high 231 Pa/ 230 Th variability). Comparing 231 Pa/ 230 Th and opal concentrations from all sites used for this study (Bradtmiller et al., 2007;Lippold et al., 2009Lippold et al., , 2016McManus et al., 2004;Ng et al., 2018), no persuasive correlation is present, and opal concentrations are always below 5% ( Figure S3).
Besides opal, authigenic Fe and/or Mn hydroxides are potential strong scavengers of 231 Pa and 230 Th (Hayes, Anderson, Fleisher, Vivancos, et al., 2015). Indeed, along the GA03 GEOTRACES North Atlantic Transect (Hayes, Anderson, Fleisher, Huang, et al., 2015), K d values of MnO 2 and Fe (OH) 3 have been estimated at magnitudes higher than for lithogenic particles or CaCO 3 (Hayes, Anderson, Fleisher, Vivancos, et al., 2015). This effect was observed emanating from hydrothermal plumes of the MAR (Hayes, Anderson, Fleisher, Vivancos, et al., 2015). However, due to the large distance between the MAR and the BBOR, scavenging by Mn/Fe phases originating from the MAR can be considered as negligible. Furthermore, increased scavenging of 231 Pa and 230 Th by MnO 2 has been reported in bottom water particles off the Mauritanian margin (Hayes, Anderson, Fleisher, Vivancos, et al., 2015). In contrast to the BBOR, the Mauritanian margin is a region of high upwelling intensity. There, 231 Pa can be scavenged by MnO 2 coatings formed from redox cycling from the respiration of high contents of organic matter (Hayes, Anderson, Fleisher, Vivancos, et al., 2015), but such conditions have not been observed at the BBOR. Although the mobility of Fe and Mn may have been quite different between glacial and interglacial conditions due to lower oxygenation, effects of increased 231 Pa and/or 230 Th scavenging are rather expected under very sluggish circulation regimes such as the Pacific (Korff et al., 2016).
Alongside particle composition the sheer amount of particles may also increase 231 Pa/ 230 Th, in particular at ocean margins (so-called boundary scavenging; Anderson et al., 1983;Hayes, Anderson, Fleisher, Huang, et al., 2015). Increased 231 Pa/ 230 Th have been found from the high accumulating and organic-rich sediments off the coast of West Africa (Christl et al., 2010;Hayes, Anderson, Fleisher, Huang, et al., 2015;Lippold, Mulitza, et al., 2012;Scholten et al., 2008), however still overprinted by an AMOC signal in their temporal evolution (Lippold, Mulitza, et al., 2012). The effects of boundary scavenging and increased particle fluxes at the high-productivity upwelling regions off West Africa are not comparable with the BBOR as discussed before.
On the other hand, the effect of bottom scavenging due to the occurrences of nepheloid layers (Hayes, Anderson, Fleisher, Huang, et al., 2015) may represent an additional sink for 231 Pa and 230 Th at the seafloor.
A first simple representation of bottom scavenging in a 231 Pa/ 230 Th-enabled model (Rempfer et al., 2017) did not yield a disturbed relationship between overturning strength and 231 Pa/ 230 Th on the larger spatial and temporal scales in the Atlantic. The implementation of bottom scavenging in the model, however, was global. Nepheloid layers were assumed to appear in all bottom grid cells with the thickest layers in the deepest water depths. Accordingly, the model is able of capturing a basin-wide relation rather than a local influence. Temporally variable occurrences of local and regional nepheloid layers may be able to influence a 231 Pa/ 230 Th profile to a certain extent possibly increasing sedimentary 231 Pa/ 230 Th ratios (Deng et al., 2014).
Today, parts of the northwestern Atlantic have been found to be covered by nepheloid layers (Gardner et al., 2018;Stahr & Sanford, 1999). Effects of nepheloid layers on Nd isotopes have been reported from the BBOR, for example, by offsets found between core tops and seawater Nd isotopic compositions and by the manifestation of unradiogenic anomalies in Nd isotopic signatures attenuating as a function of distance to the presence of strong benthic nepheloid layers (Pöppelmeier et al., 2019).
As of yet, there is little handle on reconstructing the extent and intensity of past nepheloid layers and, thus, how to assess the influence of past nepheloid layers on radionuclide scavenging. Based on analyzing changes in sediment focusing, Gutjahr et al. (2008) suggested reduced focusing at BBOR intermediate depth sites indicating reduced shelf-derived sediment redistribution before the Holocene. Nepheloid layers are mostly produced by upper ocean dynamics of surface eddy kinetic energy propagating downward stirring up bottom sediments (Gardner et al., 2018). During glacial periods the sea level was lower and the shelf contact area reduced. As a consequence, energy from shallow tidal mixing on the shelves was nearly absent, and the tidal energy increasingly dissipated in the deep ocean (Egbert et al., 2004;Wilmes et al., 2019). Bringing this tidal energy into the deep Atlantic would increase the mixing and turbulence there. As a consequence nepheloid layers during the LGM could have been greater and denser in the northwestern Atlantic. The greater suspended particle concentration could have increased the bottom scavenging of 231 Pa and 230 Th during the LGM (c.f. Deng et al., 2014;Hayes, Anderson, Fleisher, Huang, et al., 2015).
Based on these observations and as anticipation of the discussion on differences in 231 Pa/ 230 Th between the BBOR and the Bermuda Rise, we can deduce that nepheloid layers may not have had first-order control on the down-core evolution of the new 231 Pa/ 230 Th records during the Holocene. Bottom scavenging by nepheloid layers was presumably more intense before the Holocene. Paleoceanography and Paleoclimatology dominant based on above observations. Moreover, increased bottom scavenging by nepheloid layers during the LGM is also pending to be confirmed.

Stable Holocene Circulation
Today, the northwestern Atlantic basin is dominated by the southward export of NADW as main part of the AMOC. In contrast to this, in the northeastern Atlantic, NADW flows partly northward as it enters the basin as far south as the equator (Rhein et al., 2015). Concentrations of dissolved 230 Th and 231 Pa in the eastern part of the North Atlantic show a depth-dependent increase as a result of weaker export (Hayes, Anderson, Fleisher, Huang, et al., 2015). In contrast, the western basin exhibits a moderate increase in dissolved 231 Pa concentrations with water depth due to the southward advection of 231 Pa by the strong overturning circulation (Deng et al., 2018). We therefore aim for considering 231 Pa/ 230 Th-based AMOC reconstructions from regionally constrained data of the northwestern Atlantic Lippold, Mulitza, et al., 2012;Ng et al., 2018) instead of integrating 231 Pa/ 230 Th data from across the Atlantic (Bradtmiller et al., 2014;Lippold et al., 2016;Lippold, Luo, et al., 2012;Yu et al., 1996).
The notion of an active and strong Holocene AMOC (Keigwin & Boyle, 2000;Oppo et al., 2003) as indicated by low and relatively constant 231 Pa/ 230 Th at the Bermuda Rise is supported by low Holocene 231 Pa/ 230 Th values at the BBOR (average = 0.056 ± 3.2%; Figure 3). The strong overturning circulation is in accordance with strong NADW production as indicated by εNd indicating the prevalence of NSW in the northwestern Atlantic (Pöppelmeier et al., 2019). The absolute values of around 0.056 are well in agreement with previous 231 Pa/ 230 Th data compilations (Bradtmiller et al., 2014;Lippold, Luo, et al., 2012) for deep sites. An early Holocene AMOC strength overshoot as suggested by combined 231 Pa/ 230 Th and εNd records of lower time resolution  cannot be confirmed from the new data set.

Weaker Circulation During the LGM
All new down-core profiles display 231 Pa/ 230 Th values clearly higher during the LGM than during the Holocene (Figures 4 and 5) but on different absolute levels depending on the water depth. The shallowest Site 1059 (2,984 m) features the overall lowest LGM values of all BBOR records. Following the finding that sedimentary 231 Pa/ 230 Th is largely a signal integrated from the 1000-m water column above the seafloor (Luo et al., 2010;Thomas et al., 2006), it seems likely that Site 1059 at least partially recorded the influence from a shallow Glacial North Atlantic Intermediate Water (GNAIW) overturning cell above the core location.
The existence of such a shallow overturning cell has been suggested by several nutrient-proxy-based studies (e.g., Boyle & Keigwin, 1987;Curry & Oppo, 2005;Duplessy et al., 1988;Keigwin, 2004;Oppo & Lehman, 1993) as well as observations from sortable silt data from the BBOR (Evans & Hall, 2008). Highest LGM values are found at the depth interval from 3,500 to 4,000 m (Site 1060/1061; Figure 4a) suggesting either the influence of bottom scavenging or weaker water transport at these depths. The interpretation of a weaker water transport seems more plausible since the effect of bottom scavenging at the BBOR is subordinate to the large-scale Pa export from the North to the South Atlantic (cf. section 4.1). A slight decrease in 231 Pa/ 230 Th during the LGM is observed in the cores below 4,000 m (12JPC and Site 1062), indicative of a more active circulation in the abyssal northwestern Atlantic (Figures 2 and 3). Such a circulation depth structure has

10.1029/2019PA003737
Paleoceanography and Paleoclimatology also been reported from BBOR sortable silt reconstructions (Evans & Hall, 2008). Stable carbon isotope data (δ 13 C) from a depth transect along the BBOR show lighter δ 13 C values with increasing water depth interpreted as propagating SSW during the LGM (Evans & Hall, 2008;Keigwin, 2004). However, in the light of recent findings of past water mass distributions in the western Atlantic basin from a δ 13 C and Nd isotope perspective, an extensive advance of SSW into the North Atlantic may need to be seen more critically (Gebbie, 2014;Howe et al., 2016;Oppo et al., 2018;Spooner et al., 2018).
The study by Spooner et al. (2018) gives an estimation on the depth where the boundary between GNAIW and SSW was located at 30°S in the South Atlantic. Their data support a strong southward directed circulation at least in depths shallower than 2,600 m. Furthermore, they found faster flow speeds below~4,000-m water depth interpreted as northward propagated SSW. Spooner et al. (2018) therefore suggested that the boundary between GNAIW and SSW at 30°S was situated somewhere between 2,600-and~4,000-m water depth possibly in deeper than in shallow depths. This boundary is rather expected to descend toward 30°N instead of being stable on such a long distance due to the increasing dominance of NSW with shorter distance to the NADW formation areas (Gebbie, 2014;Howe et al., 2016;Oppo et al., 2018). Accordingly, a strong inflow of SSW in the deep northwestern Atlantic basin during the LGM appears unlikely. A more recent study based on Nd isotopes argues that the northwestern Atlantic was filled mainly with NSW rather than SSW during the LGM (Howe et al., 2016), which is supported by proxy-model comparisons concluding that the volume seized by NSW was not much different in the LGM than in the Holocene (Gebbie, 2014). The clearly more radiogenic Nd isotopic signatures before the deglacial are not necessarily an unequivocal evidence for the presence of SSW as interpreted in earlier studies (Gutjahr et al., 2008;Lippold et al., 2016;Roberts et al., 2010), since the εNd end-members for both SSW and NSW potentially have changed toward more radiogenic values (Gutjahr et al., 2008;Howe et al., 2016;Skinner et al., 2013). From these findings it is difficult to maintain the notion of SSW predominantly bathing the BBOR and the Bermuda Rise during the LGM. Analyses from all these sites provide very similar Nd isotope signatures indicating that the complete depth transect at the BBOR was bathed in the same water mass during the LGM (Gutjahr et al., 2008;Gutjahr & Lippold, 2011;Pöppelmeier et al., 2019;Roberts et al., 2010). In contrast to the uniform picture in Nd isotope signatures, 231 Pa/ 230 Th ratios at the Bermuda Rise (~4,500-m water depth; McManus et al., 2004;Lippold et al., 2009) were significantly lower than at the BBOR during the LGM nearly reaching Holocene levels (Figures 3 and 4a; note the different color at~4,500 m between~20 and~24 ka compared to the predominant color of the water column above and below in Figure 3).

Paleoceanography and Paleoclimatology
New glacial water mass reconstructions from the northwestern Atlantic do argue not only for a shallow GNAIW overturning cell but also for an abyssal northern water mass that potentially was formed by brine rejection (Howe et al., 2016;Keigwin & Swift, 2017;Pöppelmeier et al., 2018). Such an abyssal southward directed water mass advection could explain the low 231 Pa/ 230 Th observed at the Bermuda Rise, which might be corroborated by slightly lower peak LGM 231 Pa/ 230 Th values at the deepest BBOR Site 1062 compared to Sites 1060/1061 (Figure 4a).
Accordingly, when considering the depth structure of 231 Pa/ 230 Th (Figure 3), our new records are in line with the hypothesis of the northwestern Atlantic basin predominantly bathed by NSW rather than SSW during the LGM as inferred from Nd isotope records (Howe et al., 2016;Pöppelmeier et al., 2019).

Confirming Deglacial AMOC Variability
The 231 Pa/ 230 Th record of Site 12JPC (4,250 m) differs from the most proximal cores in the depth transect, namely, Site 1061 (4,036 m) and Site 1062 (4,760 m), during the deglaciation (11 to 19 ka). While the latter records display deglacial millennial-scale features, these are missing in the 12JPC record exhibiting a small plateau during the B/A only. Similarly, the Nd isotope record of 12JPC also shows only a gradual glacialinterglacial shift (Pöppelmeier et al., 2019). Sedimentary depositional processes (e.g., winnowing or sediment redistribution; Gutjahr et al., 2008;Pöppelmeier et al., 2019) may have played a role in smoothing out the original oceanographic signal at this site. Sediment focusing during the YD is smaller at 12JPC compared with Sites 1061 and 1062 ( Figure S5). On this basis, we exclude the deglacial part of 12JPC from the following discussion as was done by Pöppelmeier et al. (2019).
Substantial freshwater input into the North Atlantic has been associated with a reduced (Bradtmiller et al., 2014) Figures 2 and 3; below 0.093), pointing to a widely weakened but still active overturning (Bradtmiller et al., 2014). Increases in 231 Pa/ 230 Th from LGM to HS1 are less pronounced for the water depth interval from 3,500 to 4,200 m (Sites 1060, 1061, and 12JPC; Figures 3 and 5) featuring the highest LGM values of the northwestern Atlantic depth transect and thus calling for a weaker change in circulation strength between LGM and HS1. In contrast, the deepest Site 1062 features a more pronounced increase in 231 Pa/ 230 Th from the LGM to HS1, potentially due to an abyssal component of northern-sourced glacial water mass during the LGM (Howe et al., 2016;Keigwin & Swift, 2017;Pöppelmeier et al., 2018). Accordingly, increases in 231 Pa/ 230 Th from the LGM to HS1 below 4,200 m indicate a weakening of glacial abyssal NSW water mass admixture. Thus, the new data set further strengthens the notion of a sluggish AMOC during HS1 in the northwestern Atlantic for a wide range of water depths (Lund et al., 2015;Robinson et al., 2005;Bradtmiller et al., 2014; Figure 3).
Following HS1, the onset of the B/A warm period is marked by an abrupt decrease in 231 Pa/ 230 Th in the BBOR cores reflecting the invigoration of the AMOC very similar to the Bermuda Rise record (McManus et al., 2004; Figure 3). It has been proposed that the B/A was potentially marked not only by an AMOC reinvigoration but by an AMOC overshoot transiently producing more NADW than during the Holocene (e.g., Barker et al., 2010;Cheng et al., 2014). Such a B/A overshoot is not apparent from the new records, even if taken into account that there is a certain response time of sedimentary 231 Pa/ 230 Th to AMOC changes as well as a potential smoothing of the signal by bioturbation. If such a B/A overshoot was shorter than 200-500 years, it may not have been fully recorded in its whole character in the sediment due to an interplay of sedimentation rate and processes like bioturbation smearing sedimentary signals (Marchal et al., 2000;Rempfer et al., 2017;Yu et al., 1996). However, the high sedimentation rates at the BBOR (10-40 cm/ka) and in particular the duration of the B/A (~2 ka) are expected to allow any 231 Pa/ 230 Th minima to be fully resolved. Thus, our new data from the BBOR (Figures 2 and 3) confirm the relatively abrupt onset of deep circulation but do not favor an extraordinarily strong long-lived AMOC strength overshoot during the B/A.
For all cores, low 231 Pa/ 230 Th during the B/A are terminated by a sharp increase toward almost LGM-like values during the YD for the whole depth transect (Figure 3). During this last cold spell preceding the Holocene, the water mass distribution and circulation regime in the northwestern Atlantic basin below 3,000-m water depth was again similar to these during the LGM and HS1 (Pöppelmeier et al., 2019 (Lippold et al., 2009). However, even during peaks of high diatom abundances, the absolute opal bulk concentrations of the sediments do not exceed 6% (Böhm et al., 2015). Since Bermuda Rise and BBOR cores feature very high sedimentation rates, a fairly good preservation of opal can be expected leading to the assumption that the buried opal is representative of the past opal flux. On a basin-wide scale, there is no significant correlation of opal concentration with 231 Pa/ 230 Th (Bradtmiller et al., 2014;Lippold, Luo, et al., 2012)    in line with the timing of HS1 (Table S6). Most of these records exhibit clearly lower time resolution than the new BBOR records. For this reason, we only compare the time slices of the Holocene, LGM, and HS1 with our findings (Figure 5). The equatorial records show the common feature of low Holocene 231 Pa/ 230 Th and higher deglacial and glacial 231 Pa/ 230 Th levels and a more or less well resolved peak around HS1. These characteristics are in good agreement with the new higher-resolution BBOR records calling for a large-scale oceanographic feature on this timescale for this range of water depths. While the absolute 231 Pa/ 230 Th values of the shallower cores (GeoB1515-1; GeoB1523-1; EW9209 3JPC; 3,100 to 3,300 m) are in good agreement to the new BBOR records, 231 Pa/ 230 Th values from the deeper equatorial sites (EW9209 1JPC; RC 16-66; 4,000 to 4,500 m) are considerably lower ( Figure 5). One would expect the 231 Pa/ 230 Th ratio to increase with a longer traveling time of NADW (greater distance to the deep water formation zone). While 230 Th is quickly scavenged from the water column advected, 231 Pa can be supplied to sites further south from upstream (north), and the concentration is further increased by the continuous ingrowth from the decay of dissolved 235 U. However, the relationship between water mass aging and increasing 231 Pa/ 230 Th is still unclear (Deng et al., 2018). Lower 231 Pa/ 230 Th at greater depth further downstream of NADW (equatorial Atlantic) can rather be explained by the effect of building up a 231 Pa deficit relative to 230 Th within one individual overturning cell (Burckel et al., 2016;Luo et al., 2010). Alternatively, the lower equatorial 231 Pa/ 230 Th values observed may also reflect the effect of bathymetry and the narrowing of the DWBC flow path in the region, leading to increased flow speeds at greater depths. Taken together, all sites from below~3,000-m water depth and within the area influenced by the DWBC display very similar patterns in the 231 Pa/ 230 Th down-core profiles confirming the general notion of the relative AMOC strengths being most vigorous during the Holocene and clearly weaker during the LGM with the weakest overturning during HS1.

Conclusions
We present four new high-resolution 231 Pa/ 230 Th records from the deep northwestern Atlantic covering the time period from Heinrich Stadial 2 until today and resolve AMOC variability during climatic key intervals like the HS2, LGM, HS1, B/A, YD, and the Holocene. These new time series confirm the timing and magnitude of the millennial-scale climate variability previously established by the Bermuda Rise 231 Pa/ 230 Th records (such as gradual AMOC increase from the YD into the Holocene, no B/A AMOC overshoot, similar levels of HS1 and HS2 values but with higher variability of the latter) but with one exception. Whereas a strong LGM circulation was suggested by the Bermuda Rise record, the depth transect presented here shows a more complex circulation pattern. The BBOR records from 3,500 to 4,700 m suggest significantly reduced circulation strength during most of the LGM. In the water depth from 3,500 to 4,000 m, LGM 231 Pa/ 230 Th values are nearly indistinguishable from Heinrich Stadial 1 when the AMOC was weakened across the entire Atlantic. The shallowest location of our depth transect, Site 1059 at 3,000-m water depth, again features lower 231 Pa/ 230 Th values during the LGM on a level similar to the Bermuda Rise in accordance with a shallow GNAIW overturning cell above. Further, comparison to existing equatorial West Atlantic 231 Pa/ 230 Th records yields a uniform basin-wide picture confirming a strong Holocene circulation regime, a weaker LGM overturning configuration, and a mostly reduced circulation during Heinrich Stadials 1 and 2. Stefan Rheinberger is thanked for technical support during ICP-MS measurements. Further, we acknowledge Sönke Szidat for the new 14 C date for GeoB1515-1 and Benny Antz for preliminary work on the ODP cores. We thank the IODP core repository in Bremen for providing the samples. The comments of two anonymous reviewers as well as David J.R. Thornalley considerably improved the manuscript. This study has been funded by the Emmy Noether Programme of the German Research Foundation (DFG) Grant Li1815/4. Data can be found in the supporting information and on Pangaea (https:// doi.pangaea.de/10.1594/ PANGAEA.908156).