Background & Summary

In order to decipher the mechanisms at play in observed past climate changes, it is necessary to establish a common temporal framework for paleoclimate records from different archives and from different locations. Determining the lead/lag relationships between different climatic and circulation changes can help to identify the underlying causes and foster development of conceptual hypotheses to be tested with climate model simulations. Also, paleoclimate data-model integration studies, such as groundtruthing of transient modeling analyses, timeslice comparisons of proxy data, or data assimilation, necessitate consistent paleoclimate records chronologies in calendar years.

Here we focus on the last 40 ky because it is the time span covered by radiocarbon dating and the sole period for which it is possible to establish calendar age timescales for marine cores with a precision approaching that of ice core or speleothem records.

Radiocarbon dating of marine records is complicated, however, by a difference between the surface water 14C/12C ratio (expressed as ∆14C, in ‰) and that of the contemporaneous atmosphere, due to the balance between the input of atmospheric 14C and its removal by radioactive decay in the water column, advection, and mixing with older waters. This difference in ∆14C is termed the “reservoir age” of the surface waters. Previous studies have revealed that surface reservoir ages have not remained constant over time at high latitudes of the North Atlantic and Southern Ocean (i.e. poleward of ~38°N and of ~40°S) due to changes in the location and vigour of deep-water formation1,2,3,4.

In those high-latitude regions, it is thus necessary to use an alternative dating strategy in lieu of 14C dating of marine organisms. Here we adopt a strategy that has been widely applied (e.g. refs4,5,6,7) and has been adopted by the INTIMATE (Integration of Ice core, Marine and Terrestrial records of the North Atlantic) group when surface reservoir ages can not be assessed8. This strategy consists of synchronizing the sea surface temperature (SST) signal recorded in marine cores with the air temperature signal recorded in polar ice cores. This dating approach is based on the observed thermal equilibrium between the ocean’s surface water and overlying air. Previous studies have demonstrated that changes in air and sea surface temperature were synchronous across the last deglaciation9 and some of the last glacial rapid climate changes10 over the North Atlantic region. Moreover, modeling studies of the last deglaciation11 or last glacial millennial climate changes12,13 show that both increases and decreases in North Atlantic (Southern Ocean) SST and in air temperature above Greenland (Antarctica) are synchronous.

Currently, the Greenland NorthGRIP (NGRIP) ice core can be considered the best-dated continuous continental paleoclimatic archive over the last 50 to 75 ky. The NGRIP Greenland Ice Core Chronology 2005 (GICC05) calendar age scale has been established by annual layer counting with estimated uncertainties of 50 y at 11 calendar ky BP (i.e. calendar ky before 1950, noted ka hereafter), 100 to 450 y for the 11–30 ka interval, and 450 to 800 y for 30–40 ka14 (y or ky referring to durations and ka to dates). Moreover, a common chronology for Greenland and Antarctica ice cores has been developed based on their records of 10Be and atmospheric CH4 concentration15,16. This dating effort yielded the Antarctic AICC2012 age scale for four Antarctic ice cores, which is fully consistent with the GICC05 age scale over the last 60 ky16. Using the GICC05 and AICC2012 age scales as alignment targets for high latitude SST records of the north and south hemispheres respectively, it is thus possible to directly compare marine records from both hemispheres on a common time frame.

Here, we present the first set of consistently dated Atlantic sediment cores from 92 locations distributed between 68°N and 53°S, and between 400 and 5000 m water depth (Fig. 1, Online-only Table 1, ref.17), together with consistently derived dating uncertainties. This new data set enables paleoclimate scientists to (i) examine relative phases between Atlantic records (e.g. planktonic and benthic oxygen and carbon isotopes, Pa/Th); and (ii) use the spatial and temporal changes recorded in Atlantic sediments to constrain paleoclimate model simulations.

Fig. 1
figure 1

Location of the 92 dated Atlantic sediment cores (see Online-only Table 1 for precise coordinates and water depths of the cores). The figures were generated using the Ocean Data View software53, the ETOPO bathymetry54 (left panel), and the WOA13 mean annual salinity55 along a mid-Atlantic north-south section (right panel). The salinity section illustrates the distribution of the cores with respect to the main modern water masses.

Methods

We compiled existing paleoceanographic data from Atlantic sediment cores covering part of or the entire 0–40 ka interval, with sedimentation rates of at least 5 cm/ky, for which there exists the following dating means: radiocarbon dates for mid and low latitudes sediment cores, and SST or magnetic records for sediment cores located poleward of ~38°N and ~40°S. New cores were added to fill gaps with respect to the available geographical and water depth coverage, and additional radiocarbon dates were produced to improve the existing age models of some cores (Online-only Table 1).

In mid and low latitudes (i.e. between ~40°S and ~38°N), reservoir ages can be assumed not to have strongly varied in response to ocean circulation changes of the last glacial and deglaciation. The same is true at all latitudes during the Holocene. Thus, in mid and low latitudes, and during the Holocene at higher latitudes, the sediment cores were dated by means of calibrated radiocarbon ages. For this, 1427 published and 104 new radiocarbon dates have been calibrated using the Bayesian calibration program “MatCal”18, and the IntCal13 and SHCal13 calibration curves19,20 for North and South Atlantic cores, respectively.

We accounted for both spatial and temporal variability in 14C reservoir ages. To estimate spatial variations in reservoir ages we extracted bomb-corrected reservoir ages from the GLobal Ocean Data Analysis Project for Carbon (GLODAP) data set21. Prior to extracting these surface reservoir ages, GLODAP data were re-gridded to a 4° × 4° grid, whereby the mean and standard deviation for the GLODAP data points from the upper 250 m for each 4° × 4° grid cell were calculated. The modern surface water reservoir age at a given site is then obtained from the nearest grid node to the core site (Fig. 2). In the case of certain sites that are out of range of the GLODAP grid, such as those in the Gulf of Mexico, we have extrapolated the GLODAP 4°x4° grid to these areas. This spatially varying component of the reservoir age is subtracted from the laboratory 14C age before calibration (with error propagation). The error used for this spatial reservoir age component is either the computed GLODAP standard deviation, or 100 14C yr, whichever is greater. For pre-Holocene dates, a minimum of 200 14C yr is used instead of 100 14C yr.

Fig. 2
figure 2

Average reservoir age extracted from the GLODAP data re-gridded to a 4° × 4° grid and averaged over the upper 250 m of the water column. These values can be downloaded from Figshare56.

To also consider temporal changes in reservoir age, we further applied a correction to account for the impact of atmospheric CO2 concentration changes upon surface water 14C activity. At the Last Glacial Maximum (LGM), the lower atmospheric CO2 concentration induced an increase in atmospheric ∆14C of ~30‰ due to the speciation change, everything else being equal22. This ~30‰ increase in atmospheric ∆14C in turn caused a ~250 y increase in surface water reservoir ages22. To account for this temporal change in surface reservoir age, we linearly scaled a reservoir age correction to atmospheric pCO2, whereby a correction of 0 14C y corresponds to present day pCO2, and 250 14C y to LGM pCO2. For pCO2 values, we consulted the composite atmospheric CO2 record of Antarctic ice cores23. This age-dependent component of the reservoir age is added to the IntCal13 (or SHCal13) 14C age record before calibration.

Even in regions where surface reservoir ages can be predicted based on the evolution of atmospheric CO2, as described above, increased uncertainties in radiocarbon-dated chronologies can still arise from bioturbation biases (e.g. ref.24). Thus, in the best cases, when bioturbation biases and local changes in past surface reservoir ages remain limited, sediment core dating uncertainties mainly arise from the conversion of radiocarbon ages into calendar ages. In these cases, uncertainties are less than 150 y for the time interval 0–11 ka, of about 400 y for the 11–30 ka interval, and of 600 to 1100 y for the 30–40 ka interval19. In all other cases, dating uncertainties are larger.

Almost all our age-depth models of low- and mid-latitude cores (51 out of the 92 cores, see Online-only Table 1) are entirely based on calibrated 14C ages. In three cores (GeoB3910, MD09-3246 and MD09-3256Q), located on the Brazilian margin in a region under the influence of the Intertropical Convergence Zone, it is possible to take advantage of the simultaneous recording of rainfall increases during Greenland stadial periods in the marine cores and in U-Th dated speleothems from the adjacent continent to improve the marine age models. Rainfall increases are recorded both by XRF-Ti/Ca peaks in the marine cores, and by δ18O decreases in the speleothems25. By aligning the XRF-Ti/Ca in the marine cores to the speleothem δ18O, it is possible to improve the precision of the marine age models around 40 ka and to extend them beyond the limit of 14C dating. Importantly, the speleothem record from El Condor cave26, to which we have aligned the three marine cores, has been shown to be in phase, within dating uncertainties, with the NGRIP air temperature record in the GICC05 age scale25,27. Our alignment of GeoB3910, MD09-3246 and MD09-3256Q marine cores to El Condor speleothem is thus consistent with the NGRIP GICC05 age scale.

For cores located north of ~38°N (26 cores) and south of ~40°S (2 cores), and ODP Site 1060 for which there exist no 14C dates but where planktonic foraminifer census counts exhibit a clear NGRIP signal28, we have used calibrated radiocarbon ages only over the Holocene portion (i.e. after the end of the Younger Dryas, dated at 11.65 ka in the GICC05 age scale29), and aligned their glacial and deglacial portions to NGRIP or EPICA Dronning Maud Land (EDML) air temperature signal. We used different types of chronological markers to derive these 29 age-depth models:

  1. (1)

    Tie points defined by aligning high latitude SST records to NGRIP air temperature proxy record on the GICC05 age scale for North Atlantic cores, and to EDML air temperature on the AICC2012 age scale for South Atlantic cores;

  2. (2)

    Tie points defined by aligning magnetic properties of northern North Atlantic and Nordic Seas cores to the NGRIP air temperature signal on the GICC05 age scale;

  3. (3)

    Dated tephra layers.

The dating procedures (1)-(3) are described in detail below. The alignment procedures (1) and (2) by essence impede the assessment of leads and lags between the aligned records. For instance, leads/lags between SST and polar air temperatures, or among SST records from high latitude marine cores, are by construction not significantly different from zero. In contrast, this dating approach gives access to the relative timing of circulation changes recorded at different water depths in cores located on depth transects.

(1) We aligned SST records to polar ice core air temperature proxy records using the AnalySeries program30. NGRIP alignment targets correspond to the rapid transitions out of and into Greenland stadials, as dated and listed in refs29,31 (Online-only Table 2). Tie points were generally defined by aligning rapid warmings recognized in both the ice core and marine core, as recommended in ref.8. In rare cases, rapid and well-defined coolings have been aligned. In a few cases, when the SST record resolution was too low or the signal shape ambiguous, maxima or minima have been aligned. Remaining ambiguities in the identification of alignment tie points were solved in most cases by fulfilling the condition that the tie point age is younger or equal to the calibrated 14C ages obtained by assuming no other change in surface reservoir age than the temporal evolution due to changing atmospheric pCO2. Not fulfilling this condition would result in negative surface reservoir ages, which is not physically possible (see Supplementary Fig. 1 for an example).

SST alignment to Antarctic temperature variations was made at marked transitions in the temperature record, such as Antarctic Isotopic Maxima32, the onset of the early and late deglacial warming, or the beginning of the Antarctic Cold Reversal.

In addition, we used the following three alignment targets in the North Atlantic:

  1. (i)

    A first alignment target is based on the observation that the cooling marking the beginning of Heinrich Stadial 1 in three independently dated North Atlantic cores is synchronous with the sharp increase in dust flux recorded in the Greenland ice cores and dated at 17.48 ka ± 0.21 ky on the GICC05 age scale33. This observation is consistent with this cooling being coeval with an increase in dust transport from Asia to Greenland, as observed during other Greenland stadials34.

  2. (ii)

    Two other alignment targets correspond to the beginning and the end of the warm event identified in ref.35 within Greenland stadial 3 (GS-3) in several North Atlantic cores between 24 and 25 ka. This warm event within GS-3 is not clearly recorded in Greenland ice (δ18O) or gas (δ15N) isotopic records, but corresponds to a marked decrease in dust flux. Here again, we aligned the beginning and end of the warm event to the corresponding changes in the NGRIP dust flux dated on the GICC05 age scale at 25.05 ka ± 0.35 ky and 24.1 ka ± 0.33 ky, respectively.

For consistency, the alignment tie points in high latitude cores were all defined by the same person. Similarly, one single person defined all the alignment tie points in the three Brazilian cores. Also, the SST records used in the present study are all based on planktonic foraminifer census count data. When SST reconstructions based on full census count data were not available, we used the percentage of the polar species Neogloboquadrina pachyderma (left coiling) as a proxy for SST. This approach has been described and validated in a number of studies (e.g. refs36,37,38,39). In two North Atlantic cores (ODP Site 1060 and core MD08-3180Q) we used the percentage of warm species instead, because the percentage of N. pachyderma was too low. In the particular case of the Iberian margin, it has been shown that Globigerina bulloides δ18O co-varies with SST40,41 and we have used G. bulloides δ18O as a proxy for SST when no SST estimates were available.

Both age and depth uncertainties are defined for each tie point. The depth uncertainty directly depends on the sampling resolution of the SST curve: it is taken as half of the depth interval corresponding to the rapid warming (or more rarely cooling), or as half of the width of the SST maximum or minimum, when maxima or minima have been aligned. In instances of ambiguities that could not be tested by the constraints provided by 14C dates, we attributed an uncertainty to the depth of the tie point, large enough to encompass the two events (warmings, or more rarely, coolings or SST maxima or minima) which could both be aligned to the same target. The uncertainty on the tie point ages is the GICC05 dating precision of the transitions between Greenland stadials and interstadials, with one sigma uncertainties defined as half the cumulative ‘maximum counting error’ in the GICC05 age scale29,31. Similarly, the dating uncertainty of the alignment tie points defined with respect to AICC2012 is the dating error given in ref.16.

(2) In high northern latitudes, when SST records are not available, for some cores it is possible to instead use high-frequency variations in magnetic susceptibility (MS) recorded during the last glacial period. The rapid oscillations in magnetic properties in sediment cores on the flow path of North Atlantic Deep Water (NADW) in the Nordic Seas and North Atlantic have indeed been shown to be in phase with the Greenland ice δ18O or air temperature signal42. Support for this synchronicity comes from tephra and geomagnetic field (Laschamp inclination excursion) marine records. These marine records become aligned with tephra and cosmogenic nuclide Greenland records when the MS tuning to Greenland is applied (e.g. refs43,44,45).

We dated five cores located north of 62°N by aligning their MS records to the NGRIP ice δ18O signal (Online-only Table 1). MS tie points and their associated uncertainties were defined using the same method as described for the alignment of SST signals to ice core records. The MS records of four of these five cores have been previously shown to be in phase with the Greenland air temperature signal42. More recently, the identification of tephra layers in core MD99-2284 demonstrated that this core’s MS record is also in phase with the NGRIP δ18O record43. This can be explained by the fact that changes in MS arise from changes in the efficiency of the transport of fine grained magnetic particles by deep currents from the source to the site of deposition42. The fact that the MS signal is in phase in cores located north and south of the sills separating the Nordic Seas from the North Atlantic basin, suggests that the source of magnetic minerals could be at the sills, with the strength of the overflow from the Nordic Seas directly proportional to the strength of the inflow into the Nordic Seas.

(3) We used dated tephra layers as additional chronological markers over the last 55 ky in 10 of the northernmost cores (Online-only Table 1). The following four tephra layers have been recognized both in Greenland ice cores and in certain North Atlantic and Nordic Seas marine cores: the Saksunarvatn Ash46, the Vedde Ash46, the Faroe Marine Ash Zone (FMAZ) II46,47, and the widespread rhyolitic component of North Atlantic Ash Zone (NAAZ) II (II-RHY-1)46,48 (Online-only Table 2).

Age-depth relationships were built for each core accounting for both the age and depth uncertainties of the 14C dates and chronological markers, using the age-depth modeling routine “Undatable49 (Fig. 3). This new rapid age-depth modeling routine was ideal for this project as it allowed us to run and re-run age models for the many sediment cores that we have analyzed. Moreover, this age-depth modeling routine computes a conservative age-depth uncertainty, through the provision of bootstrapping and sediment accumulation rate uncertainty49 (Fig. 3). Default values for bootstrapping percentage and sedimentation rate uncertainty were set to 10% and 0.1 respectively. In the presence of age reversals, we progressively increased the bootstrapping percentage in order to make sure that the dating uncertainty computed by Undatable was large enough to encompass most calibrated 14C ages, leaving out only outliers beyond 2 sigma dating uncertainty. This way, we take into account increased dating uncertainty associated with the existence of age-depth scatter, which may be related to sedimentation hiatuses, abundance changes, or bioturbation. Also, we considered tephra layers as the most reliable age-depth constraints and, therefore, a priori excluded them from the bootstrapping process (e.g. ref.50).

Fig. 3
figure 3

Example of age-depth plot produced by Undatable. Age-depth model produced for North Atlantic core RAPID-10-1P with bootstrapping set to 10% and sedimentation rate uncertainty set to 0.1 (see ref.49 for details). Blue, yellow and red probability density functions indicate the radiocarbon and alignment tie points, and tephra age-depth constraints, respectively. The grey cloud indicates the probability density cloud of the age-depth model, whereby darker colors indicate higher age-depth probability. The blue and black broken lines represent 68.27% and 95.45% confidence intervals, respectively. The red line indicates the age-depth model median.

In some North Atlantic cores (7 out of 92, cf. Online-only Table 1), we used 14C dates together with SST alignment tie points to NGRIP. These cores are located at the northern edge of the region where surface reservoir ages may be assumed not to have strongly varied in response to ocean circulation changes, and are characterized by large changes in SST which parallel the NGRIP ice δ18O signal. In those cores, we used alignment tie points to complement calibrated 14C dates when the latter were too sparse.

Finally, although the focus of this work is the time interval 0–40 ka, we used dating information available beyond 40 ka to ensure the robustness of the computed sedimentation rate and age-depth relationship around 40 ka.

Data Records

The present set of age-depth models contains three text files per marine sediment core17. The first text file (“age depth input”) contains an overview of the 14C ages and other age constraints used in the age-depth model. More specifically, a first section provides all the available 14C raw data, the reservoir age and calibration curve used, as well as the calibrated ages together with the 68.3% highest posterior density interval(s), and specifies which 14C dates have been used to generate the age-depth model. A second section provides the definition of the alignment tie points: the tie points depth and its uncertainty, the tie points age and its uncertainty, the nature of the tie points and the nature of the uncertainty of the tie points age. The second text file (“udinput”) contains the input for the age-depth model in the Undatable format. The third text file (“_admodel_ka”) contains the computed age-depth relationship and associated dating uncertainties. In addition to the complete set of data records archived on Seanoe17, the 92 “_admodel_ka” text files can be found on Pangaea51.

Notably, the fact that the 14C raw data are provided makes the present data set easy to update using a future 14C calibration curve. Also, tie point depths are provided, allowing updates of the age-depth models if higher resolution SST records are produced.

In addition to these three text files, the age-depth model plot produced by the Undatable routine (see Fig. 3 for an example) is provided for each core, as well as a plot of the aligned SST or MS record, ice core record, and chosen alignment tie points (see Fig. 4 for an example) for the cores which have been partially or completely dated by alignment to an ice core record17.

Fig. 4
figure 4

Example of North Atlantic and Nordic Seas cores dated by alignment of their SST records to the NGRIP ice δ18O signal. Top panel: planktic foraminifer-based warm season surface temperature of core MD99-228157,58; middle panel: % N. pachyderma of core MD99-2281 and MD04-284559,60 (both panels: diamonds and squares above the x-axis indicate calibrated 14C ages and alignment tie points, respectively). Bottom panel: NGRIP ice δ18O record on the GICC05 age scale61. Grey bands highlight the Younger Dryas and Heinrich stadials 1–4 chronozones as defined in Online-only Table 2.

Technical Validation

The information relative to the validation of the age models entirely based on 14C dates can be found in the publications describing the Undatable age-depth modeling routine49 and the “MatCal” Bayesian calibration program18. The age-depth model plot provided for each core (e.g. Fig. 3) shows the calibrated 14C dates together with the computed age-depth relationship and dating uncertainty, as well as the bootstrapping percentage and sedimentation rate uncertainty values used in the computation.

Concerning the age models based on the alignment of SST to NGRIP air temperature, a first validation step involved comparing the resulting dated SST signals of different marine cores among themselves and with the NGRIP air temperature signal. An illustration of such a comparison is given in Fig. 4 for North Atlantic core MD04-2845 and Norwegian Sea core MD99-2881. Moreover, available 14C data over glacial and deglacial portions of cores dated by alignment to NGRIP provide a verification of the tie points selection since surface water reservoir ages should not be negative (e.g. Supplementary Fig. 1). Interestingly, around Heinrich stadial 4 (38 to 40 ka), our age-depth models yield ages which are systematically older than the calibrated ages obtained using IntCal13 and modern surface water reservoir age values, in agreement with the recent findings of ref.52 showing that IntCal13 is too young with respect to GICC05 during that time interval.

The age models based on the alignment of MS to NGRIP ice δ18O have been validated by comparing the resulting dated MS signals with the NGRIP ice δ18O signal (Supplementary Fig. 2). Moreover, these age models have been validated by climate-independent tie points, such as tephra layers in core MD99-228443, MD95-201044 and ENAM93-2145, or changes in the Earth’s magnetic field intensity in core MD99-228144.

The age models making use of the alignment of Ti/Ca to speleothem isotopic records have been validated by comparing the radiocarbon-dated upper portion of the cores with the U-Th dated speleothem signal (Supplementary Fig. 3). This validation was the initial step that led to the use of speleothem isotopic records to complement the dating of the three cores from the Brazilian margin since it demonstrates that terrigenous input at these sites is coeval with the precipitation events recorded in the speleothems25.