Sediment Depth Scale

The depth scale previously used for Cariaco Basin paleoclimate and ¹⁴C records used in IntCal was based on ODP Site 1002 Hole D, from which the majority of samples for ¹⁴C dating were obtained. A core-break occurred in Hole D near 14 mbsf, which was bridged using continuous sediments from adjacent Hole E (Hughen et al. Reference Hughen, Lehman, Southon, Overpeck, Marchal, Herring and Turnbull2004a). For consistency with sample identifications at ODP sediment core repositories, the Hole D depth scale was maintained, despite creating an anomaly in the age-depth curve around 14 mbsf (30–32 cal kyr BP) (Figure 3). Here, we have revised the depth scale by restoring 56 cm of missing sediment to the Hole D depth scale between 14.2 and 14.76 mbsf. Both original and adjusted depth scales are reported in Table 1.

Figure 3 Close-up view of Cariaco-Hulu calendar age plotted versus Cariaco Basin sediment depth. 56 cm of sediment missing from a core-break below 14 mbsf have been restored to the original depth scale (gray), and the revised age-depth model (blue) showed an anomalous tie point around 32 cal kyr BP (red). This tie-point has been removed from the current age model.

Table 1 ¹⁴C and calendar age data for the Cariaco Basin Site 1002D/E marine-derived ¹⁴C calibration record. Lab identification and species information are provided for each ¹⁴C date. All errors are 1 sigma. Dates obtained by averaging two or three replicate measurements are indicated on right. Data points removed due to anomalous values (compared to means and uncertainties taken from replicates or immediately adjacent bracketing samples) are also indicated on right with an “X”; (a) indicates samples from core breaks, (b) small samples deviating by >4 sigma, (c) samples from poorly performing wheel deviating by >4 sigma, (d) non-suspect samples deviating by >4 sigma. CAMS-LLNL refers to the Center for AMS at Lawrence Livermore National Laboratory. UCI refers to the Keck AMS Facility at the University of California at Irvine. Results labeled “NSRL/NOSAMS” were prepared at the INSTAAR Laboratory for AMS Radiocarbon Preparation and Research at University of Colorado, Boulder (along with associated standards and process blanks) and provided as pressed graphite targets for measurement at the National Ocean Sciences AMS Facility at the Woods Hole Oceanographic Institution. NOSAMS-reported results were δ¹³C- and blank-corrected at NSRL. Results labeled “NSRL/UCI” were prepared at INSTAAR (along with associated standards and process blanks) and provided as pressed graphite for measurement and δ¹³C- and blank-correction at UCI.

Tie-Points

In this study, the tie points used to align the Cariaco Basin and Hulu Cave paleoclimate records were selected primarily at the rapid transitions between the tropical analogs of “stadial” and “interstadial” millennial climate events (i.e, “warmings” at high latitudes and increased precipitation in the tropics) (Figure 2). These are generally the same tie points used in Hughen et al. (Reference Hughen, Southon, Lehman, Bertrand and Turnbull2006), except that most of the less-rapid transitions between “interstadial” and “stadial” events (i.e, high-latitude “coolings” or tropical “dryings”) have been eliminated due to the lower precision of the matches. We note that during the later Glacial period, from ~15–25 cal kyr BP, the tie points are less distinct, with greater variability in the Cariaco record than in Hulu. The remaining individual tie points were evaluated to identify anomalous changes in sedimentation rate resulting from the revised depth scale. All identified tie-points were assumed to represent approximately synchronous events in the Hulu Cave and Cariaco Basin records, with an uncertainty in this contemporaneity of ±100 cal yrs (1σ) for each tie incorporated into chronology construction.

One tie point, related to a relatively minor climate event correlated with Greenland Interstadial (GIS) 5.1 immediately preceding Heinrich Event (HE) 3 (Rasmussen et al. Reference Rasmussen, Bigler, Blockley, Blunier, Buchardt, Clausen, Cvijanovic, Dahl-Jensen, Johnsen, Fischer, Gkinis, Guillevic, Hoek, Lowe, Pedro, Popp, Seierstad, Steffensen, Svensson and Winstrup2014) (Figure 2), created a sharp change in sedimentation rate at 30–32 cal kyr BP (Figure 3) and likely contributed to anomalously old calendar ages within this section in the Cariaco record and IntCal13 (Reimer et al. Reference Reimer, Bard, Bayliss, Beck, Blackwell, Bronk Ramsey, Buck, Cheng, Edwards, Friedrich, Grootes, Guilderson, Haflidason, Hajdas, Hatté, Heaton, Hoffmann, Hogg, Hughen, Kaiser, Kromer, Manning, Niu, Reimer, Richards, Scott, Southon, Staff, Turney and van der Plicht2013). This spurious link has therefore been removed, eliminating the anomalous sedimentation rates and resulting in a smoother age-depth curve overall (Figure 4). The tie points were used as input into a Gaussian process model (Heaton et al. Reference Heaton, Bard and Hughen2013) to create a continuous calendar age-depth model for the Cariaco Basin with associated age uncertainty estimates that propagate both our uncertainty in the Hulu timescale and the precise contemporaneity of the chosen ties.

Figure 4 Cariaco-Hulu age-depth models shown over the entire time scale. The revised age-depth curve (blue) is smoother overall and shows fewer abrupt changes than previously (gray).

Marine Reservoir Age

Comparing the newly updated Cariaco Basin ¹⁴C dataset to Hulu Cave ¹⁴C in Figures 5 and 6, both plotted under the assumption of a constant level of MRA/DCF ¹⁴C depletion, does however reveal several periods where apparent offsets exist, mostly consisting of younger ¹⁴C ages (higher Δ¹⁴C values) for the Cariaco Basin data. These are most evident around 18, 23, and 30 cal kyr BP. Potential explanatory factors for these periodic offsets include periodic misalignment of the Cariaco tuning; uncertainties in the calendar age chronologies which are not indicated in Figure 5 but are incorporated in construction of the IntCal20 curve; and/or that an assumption of constant MRA/DCF depletion over time is not appropriate.

Both systems sample atmospheric ¹⁴C indirectly. The world’s ocean sequesters ¹⁴C away from replenishment at the surface long enough for significant decay to occur, and thus surface waters appear older than the contemporary atmosphere with a global average, present day, MRA of ~400 years, although with substantial spatial variability (Reimer and Reimer Reference Reimer and Reimer2001). The Cariaco Basin MRA has been measured in recent, pre-bomb sediments of known calendar age (Hughen et al. Reference Hughen, Overpeck, Peterson and Trumbore1996), and by comparison with overlapping tree rings (Kromer and Spurk Reference Kromer and Spurk1998; Kromer et al. Reference Kromer, Friedrich, Hughen, Kaiser, Remmele, Schaub and Talamo2004) further back in time during Deglaciation, to yield a value of 420 ± 50 ¹⁴C yrs (Hughen et al. Reference Hughen, Southon, Lehman and Overpeck2000). Speleothems also exhibit ¹⁴C depletion as groundwater feeding Hulu Cave filters through overlying soil and bedrock, absorbing “¹⁴C-dead” CO₂ and giving the growing speleothems a DCF equivalent to ~450 ± 70 ¹⁴C yrs (Southon et al. Reference Southon, Noronha, Cheng, Edwards and Wang2012).

Over the period from 11–14 cal kyr BP, it is possible to obtain an estimate of MRA within the Cariaco Basin that is independent of tuning using the varved section of the record (Hughen et al. Reference Hughen, Southon, Lehman and Overpeck2000). This section of the Cariaco record, with its varve-counting chronology, can be compared with radiocarbon measurements from Northern Hemisphere (NH) tree-rings taken from the IntCal20 database (Reimer et al. Reference Reimer, Austin, Bard, Bayliss, Blackwell, Bronk Ramsey, Butzin, Edwards, Friedrich and Grootes2020 in this issue) that provide a direct record of atmospheric radiocarbon. Consequently, this estimate is also independent of the Hulu Cave’s DCF. In Figure 7 we plot the resultant MRA estimate, obtained via a Bayesian spline. As we see, the MRA is relatively constant despite past abrupt climatic changes, such as the end of the Younger Dryas (YD) cold event around 11.5 cal kyr BP, but shows a very clear reduction in MRA during the onset of the YD around 12.8 cal kyr BP where it drops by several hundred years to near zero. For the Hulu cave speleothem, an analogous comparison to NH atmospheric trees indicates that the DCF with Hulu has remained highly constant (see Figure 6 in Heaton et al. Reference Heaton, Köhler, Butzin, Bard, Reimer, Austin, Bronk Ramsey, Grootes, Hughen, Kromer, Reimer, Adkins, Burke, Cook, Olsen and Skinner2020b in this issue) over the entire 11–14 cal kyr BP period.

Figure 7 Cariaco MRA estimate obtained via a Bayesian spline when comparing the varved section of the record from 11–14 cal kyr BP to atmospheric tree-ring data from the IntCal20 database. The MRA estimate we obtain over this period is independent from both tuning and Hulu Cave’s DCF. Each plotted point corresponds to the observed offset between an individual ¹⁴C determination from the Cariaco varved record and the IntCal20 curve, which in this period is based only on tree-ring samples. The observed offsets are shown with 1σ uncertainties representing the calendar age uncertainty of the varved Cariaco ¹⁴C determinations, and 1σ uncertainties in ¹⁴C which incorporate both the uncertainty in the Cariaco ¹⁴C sample and the IntCal20 curve at that calendar age. The MRA is relatively constant across the abrupt climate change at the YD termination (~11.5 cal kyr BP) but shows a clear reduction to near zero during the YD onset (~12.8 cal kyr BP).

The above independent comparisons of the varved Cariaco and Hulu records against NH tree-rings that provide direct sampling of atmospheric ¹⁴C suggest that the periodic (and short lived) offsets between Hulu and the tuned Cariaco are likely due to MRA changes as opposed to mistuning.

Inferred MRA Changes within Cariaco Basin over Time

As a consequence of this adaptive approach to modeling Cariaco Basin’s MRA, alongside construction of the IntCal20 curve, we simultaneously obtain a posterior estimate of the MRA within the Cariaco Basin over time. This is shown in Figure 8a and is of potential independent interest. We discuss this posterior estimate below, illustrating the changing offset in the Cariaco Basin compared to Hulu cave. In building the IntCal20 (Reimer et al. Reference Reimer, Austin, Bard, Bayliss, Blackwell, Bronk Ramsey, Butzin, Edwards, Friedrich and Grootes2020 in this issue) and Marine20 (Heaton et al. 2020b in this issue) curves, marine datasets were corrected for changes in MRA through time. Global MRAs were calculated using an ocean GCM (OGCM) (Butzin et al. Reference Butzin, Köhler, Heaton and Lohmann2020 in this issue) and applied to individual datasets according to region. However, the modeled MRA variability for the Cariaco Basin appears much too high (Figure 8b), possibly due the OGCM’s inability to capture the unique bathymetry and hydrographic setting of the basin. We therefore use the posterior estimate of the MRA to correct the Cariaco ¹⁴C calibration curve back through time (Heaton et al. 2020 in this issue). Due to the density of Hulu Cave measurements in the combined IntCal20 database, and the assumption of an approximately constant DCF over time within each Hulu speleothem in IntCal20 construction, this ¹⁴C offset between the Cariaco and Hulu data is the primary driver for the obtained MRA estimate.

Figure 8 (a) Posterior mean estimate (shown in black with the 95% credible interval) of Cariaco Basin’s MRA through time, obtained simultaneously to IntCal20 curve construction. This MRA estimate is obtained adaptively as an additive Bayesian spline at the same time as the IntCal20 curve is constructed based upon the observed offset between Cariaco and other individual ¹⁴C calibration datasets used in IntCal20. The procedure captures the same timing, direction and relative magnitude of Cariaco MRA changes as seen in direct comparisons with Hulu Cave data (following figures), likely due to the density of Hulu Cave measurements in the combined IntCal20 datasets and the assumption of constant DCF depletion. (b) Cariaco posterior mean MRA estimate shown together with the estimates for the nearest open-ocean site to the Cariaco Basin obtained by the LSG OGCM under three different climate scenarios when driven by a preliminary estimate of atmospheric Δ¹⁴C obtained from the Hulu cave speleothems alone (Butzin et al. Reference Butzin, Köhler, Heaton and Lohmann2020 in this issue). Regional first-order Hulu-based LSG OGCM estimates were used to incorporate the other marine datasets into IntCal20 (Reimer et al. Reference Reimer, Austin, Bard, Bayliss, Blackwell, Bronk Ramsey, Butzin, Edwards, Friedrich and Grootes2020 in this issue) but do not appear appropriate for the Cariaco Basin, as they overestimate the variability and magnitude of MRA values through time.

Looking closely at both our posterior MRA estimate and the apparent deviations between Cariaco Basin and Hulu Cave ¹⁴C through time, we can evaluate the timing of ¹⁴C change relative to rapid climate shifts in the paleoclimate record. As discussed above in Figure 7 comparing against trees that directly sample the atmosphere, the YD onset is clearly seen to coincide with an apparent reduction of Cariaco MRA by several hundred years. This is also seen when the Cariaco data is compared against the Hulu data alone (Figure 9). Similarly, Cariaco MRA appears to drop by several hundred years during the time of Heinrich Event 1 (Figure 10). The MRA decrease follows event H1b and persists throughout the transition into event H1a. However, there is no apparent change in MRA during the climatically more severe H1a event (Figure 10). The calendar age of this MRA decrease is bracketed by tie points at the end of H1b and the onset of H1a (17,148 and 16,150 cal yr BP, respectively), and falls within the older half of an anomalous “grey layer” that bears evidence of increased freshwater input and fluvial runoff (Yurco Reference Yurco2010). However, in order to explain the Cariaco ¹⁴C age plateau as resulting from rapidly increased sedimentation (from terrestrial runoff), ~63 cm of sediment would have had to be deposited nearly instantaneously soon after event H1b. In the absence of evidence for such an extreme sedimentation event, this MRA decrease around the time of H1 appears real. Although the precise calendar age of this feature has relatively high uncertainties from the tuning process, the relative timing of the MRA decrease well before the climatic H1a stadial suggests that increased atmospheric Δ¹⁴C associated with reduced AMOC (e.g., Hughen et al. Reference Hughen, Overpeck, Lehman, Kashgarian, Southon, Peterson, Alley and Sigman1998) may not have played a role here. Rather, the apparent reduction in Cariaco MRA is likely associated with decreased mixing of deep waters due to increased stratification in surface waters of the tropical North Atlantic or Cariaco Basin region.

Figure 9 Detailed comparison of timing between Cariaco Basin paleoclimate and ¹⁴C excursions relative to Hulu Cave (considered here as changes in MRA), focusing on the period of the Younger Dryas cold event. (upper) 550 nm reflectance from Cariaco sediments, showing periods of greater windiness (down) and greater rainfall (up). (middle) Cariaco Basin (blue) versus Hulu Cave (red) age-age plots. (lower) Cariaco Basin (blue) versus Hulu Cave (red) Δ¹⁴C plots. Reduced Cariaco MRA appears as lower ¹⁴C ages but higher Δ¹⁴C. Cariaco data are corrected with a constant 420 ± 50 ¹⁴C yrs MRA (Hughen et al. Reference Hughen, Southon, Lehman and Overpeck2000), and Hulu data have a constant DCF correction of 450 ± 70 ¹⁴C yrs (Southon et al. Reference Southon, Noronha, Cheng, Edwards and Wang2012). These uncertainties are propagated into the plotted error bars.

Figure 10 Same as for Figure 9 but focusing on the period surrounding Heinrich Event 1.

The general timing of periods of reduced Cariaco MRAs reveals that an episode occurred approximately every 6–7 cal kyrs, suggestive of the intervals between Heinrich Events (Rasmussen et al. Reference Rasmussen, Bigler, Blockley, Blunier, Buchardt, Clausen, Cvijanovic, Dahl-Jensen, Johnsen, Fischer, Gkinis, Guillevic, Hoek, Lowe, Pedro, Popp, Seierstad, Steffensen, Svensson and Winstrup2014). To test for a potential link between climate and MRA changes, we examine the detailed evidence for reduced Cariaco MRA during earlier Heinrich events (Figures 11–14). Although we do in fact observe reduced Cariaco MRA near the occurrence of Heinrich Events, the timing in general is off, with reduced MRA beginning up to 2000 cal yrs after the onset of the climatic events (Figures 11–14). In addition to reduced MRA, we also see evidence for a period of increased MRA between 28 and 25 cal kyr BP (Figure 15). Although there was likely an overall increase in MRA during the Glacial period, due to reduced carbon reservoirs in the atmosphere, biosphere and shallow ocean sediments (e.g., Hughen et al. Reference Hughen, Lehman, Southon, Overpeck, Marchal, Herring and Turnbull2004a), this would not explain a short interval of increased MRA from 25–28 cal kyr BP. This interval instead appears to be part of the cyclical pattern of Cariaco MRA changes (Figure 8a) and may represent a period of relatively weak stratification and strong deep mixing (upwelling).

Figure 11 Same as for Figures 9 and 10 but focusing on the period surrounding Heinrich Event 2.

Figure 12 Same as for Figures 9–11 but focusing on the period surrounding Heinrich Event 3.

Figure 13 Same as for Figures 9–12 but focusing on the period following Heinrich Event 4.

Figure 14 Same as for Figures 9–13 but focusing on the period following Heinrich Event 5.

Figure 15 Same as for Figures 9–14, but focusing on a period between Heinrich Events, but characterized by increased Cariaco MRA (reduced Δ¹⁴C).

We note that, due to potential variability in both Hulu DCF and Cariaco MRA through time, the magnitude and even exact timing of changes in the Cariaco Basin MRA cannot be determined with precision. In particular, our current estimates are heavily based upon the assumption that the DCF within Hulu Cave remained approximately constant from 0–55 cal kyr BP. If global MRA were adjusted upward due to Glacial boundary conditions, it would increase the apparent magnitude of periodic Cariaco MRA reductions, and even slightly broaden the durations of the events. Regardless, the general pattern of quasi-periodic cycles of MRA variability appears to be robust (Figures 8, 16). Overall, the disagreement in relative timing between reduced MRA in the Cariaco Basin and Heinrich Events (Figure 16) precludes reduced AMOC during H events as a potential cause. The data suggest cyclic decreases in Cariaco MRA to values below 100 ¹⁴C years, despite Glacial boundary conditions that should have somewhat increased surface ocean MRA around the globe (Hughen et al. Reference Hughen, Lehman, Southon, Overpeck, Marchal, Herring and Turnbull2004a; Butzin et al. Reference Butzin, Köhler, Heaton and Lohmann2020 in this issue). These low reservoir ages are an extraordinary finding without analog in the modern ocean and are difficult to reconcile with physical or geochemical models of ocean circulation. At a minimum, these results suggest periodic swings in surface stratification in the Cariaco Basin region with times of greatly reduced mixing of ¹⁴C-depleted “older” water from below. However, a clearer vision of true atmospheric ¹⁴C throughout the Glacial period may be necessary in order to draw a firmer conclusion. The Lake Suigetsu record provides such an atmospheric record, but the magnitude of scatter in the current dataset precludes the identification of MRA shifts, and extension of dendrochronologies such as from floating Kauri sequences (Turney et al. Reference Turney, Fifield, Hogg, Palmer, Hughen, Baillie, Galbraith, Ogden, Lorrey, Tims and Jones2010) may be the best solution.

Figure 16 Comparison of timing between Cariaco Basin paleoclimate and ¹⁴C excursions relative to Hulu Cave, shown over entire timescale of the data sets: (upper) 550 nm reflectance from Cariaco sediments, showing periods of greater windiness (down) and greater rainfall (up). The general timing of Heinrich Events in the Cariaco record is shown by light blue bars. (lower) Cariaco Basin (blue) versus Hulu Cave (red) Δ¹⁴C plots. Periods of reduced Cariaco MRA occur every 6–7 cal kyrs (indicated by light blue bars) but appear distinct from Heinrich Events.