In the studied interval, glacials correspond to ice growth pulses with a ca. 40 ka pacing to be further examined, but each of unascertained duration. Glacials are intervals with a weak WBUC, noticed by finer grain-size and low SSmean values in the sedimentary record (Fig. 3). Adding to very low biogenic carbonate content, the glacials likely experienced low sediment accumulation rates over Eirik Drift. These features result in problematic linkages between sediment thickness and time when using linear interpolation between anchor dates for the setting of an age model. Nonetheless, the decreasing trend defined from the SSmean record throughout the mPWP (Fig. 3) suggests an evolution toward weaker WBUC during the mPWP, thus toward the near collapse of the Atlantic Meridional Overturning Circulation (AMOC) as reconstructed for the full glacial conditions of the Quaternary37.
Conversely, the predominance of coarser sediments and high SSmean values recorded during interglacials, and notably the late interglacials of the study sequence, unveil a higher flow rate of the WBUC (Fig. 3). Given the practically identical granulometric parameters from one interglacial to the next, one may thus infer the return to nearly similar WBUC strength each time. As the duration of glacials remains unascertained, that of interglacials are similarly difficult to estimate. During these warm intervals, surface conditions were variable38,39, as inferred here by the biogenic carbonate abundances (Fig. 3), thus adding some further variability to the sediment accumulation rate from one interglacial to the other.
The source of eroded materials during the mPWP glacials differs from that of interglacials. Glacials are illustrated by highly radiogenic Pb and unradiogenic Nd-bearing supplies (Fig. 3), mostly derived from the AB, NMB and KMB source areas (Fig. 1). On the contrary, sediments deposited during interglacials bear a stronger contribution from the PV source area, characterized by a highly radiogenic Nd and an unradiogenic Pb. Observing the long-term trends from ~ 3220 to ~ 980 ka (Fig. 3), both the glacial and interglacial Nd isotope compositions trend toward more radiogenic Nd endmembers, respectively from ~ -22 to -19 and from − 14 to ~ -11 eNd(0) values, whereas the glacial 207Pb/206Pb ratios trend toward less radiogenic values (from ~ 0.93 to ~ 0.88), suggesting a shift in sediment source relative contributions, from the older to the younger terranes. Combining the isotope trends and the grain-size data that indicated a weaker WBUC flow during glacials22,40, one may infer increasing PV terrane erosion from one glacial to the next, labelled by high radiogenic Nd signatures due to ice expansion in the eastern mountains (Fig. 1). Ice erosion thus spread from the mountainous regions of eastern (PV) and southern (KMB) Greenland, merging over the AB and NMB terranes of southeastern Greenland. The limited ice coverage during interglacials, which was restricted to mountains in eastern and southern Greenland (Fig. 1B), explains the spreading of radiogenic isotopic compositions between the PV and KMB endmembers illustrated in Fig. 2.
Assessing the effective orbital forcing of the glacial pulses over Greenland during the mPWP requires an age model with a time resolution of a few ky allowing to document potential linkages with the obliquity cycle, a resolution not necessarily reached using current approaches such as the oxygen isotope stratigraphy and paleomagnetic reversal ages. These approaches were indeed used for the setting of variable age models for core U130727,41, including the Blake-Mizen et al.' s model29 referred to in Fig. 3. The oxygen isotope data of planktic foraminifera were initially reported by Sarnthein et al. 27. They show d18O (vs VPDB) shifts from ~ 2.5‰, i.e., close to the mid-to-late Holocene isotopic composition of N. pachyderma34, in the same area42, to heavier values (~ 3.1‰) pointing to cooler conditions (Fig. 3). This + 0.6‰ shift is much smaller in amplitude than the ~ 2‰ glacial/interglacial offsets observed during the Quaternary34, thus pointing to lesser amplitude changes in the cryosphere/ocean budget during the mPWP. From a chronological viewpoint, several issues can be raised about the d18O record spanning the mPWP, some were discussed by Sarnthein et al.27 who proposed two possible age models for this interval. However, the correlation of site U1307-d18O record with the LR04 stack is not unequivocal as illustrated Fig. 3. Firstly, even in the upper part of the record, the age model of Blake-Mizen29 et al. or Sarnthein et al.27 involves large amplitude sedimentation rate changes, which show no relation with the glacial/interglacial oscillations documented from the grain-size parameters, i.e., with the effective variability of the WBUC. Secondly, there does not seem to be relationship between the heavier d18O values of N. atlantica and the glacial pulses inferred from all other parameters as illustrated Fig. 3. Nonetheless, a direct correlation of the glacial/interglacial cycles with the LR4 curve would make sense as both depict a similar frequency over the critical mPWP interval, but this would involve some fine-tuning of the age model illustrated Fig. 3.
The number and relative distribution of the glacial peaks over the studied interval as well as the LR04 stratigraphy oscillations (Fig. 3) suggest some linkage with the ~ 41 ky obliquity cycle, as observed in Antarctica for the same interval36,43. However, at site U1307, this frequency does not stand out clearly in a periodogram when using the Blake-Mizen et al. 29 age model (supplementary material) as misfits between glacial pulses and low obliquity values 44 suggests that the age model lacks accuracy at the required time resolution. Anchor ages from the paleomagnetic reversals (Gauss/Kaena/Gauss/Mammoth) are not necessarily more accurate, as their uncertainty for one standard deviation (± 1s) can be up to ± 15 ky45. Furthermore, the correlation between the N. atlantica from U130727 and the LR04 stack46 in equivocal, adding to the fact that the diachronous benthic vs planktic foraminifer d18O responses in particular during deglaciations47–49 may introduce additional error.
The planktic d18O record of site U1307 27 does not clearly match climate changes that can be reconstructed from other tracers (Fig. 3), notably those of the WBUC that are marked by high vs low accretion rates for interglacials vs glacials, respectively. Correcting for enhanced sedimentation rates linked to biogenic carbonate fluxes from carbonate-free sediment accumulation is ineffective, as the carbonate pulses during interglacials remain low and do not change the sedimentation rate significantly (see supplementary material). The specificity of the N. atlantica d18O record is further illustrated by the fact that it does not seem either to unequivocally correlate with the LR04 stack (Fig. 3) based on the age mode29l used here. This either raises the issue of the robustness of all age models proposed so far for core U1307, including that of Blake-Mizen et al.29 as discussed further below, or leads to infer a decoupling of the oxygen isotope composition of N. atlantica with that of the global ocean during the mPWP interval. Specific isotopic and physical properties of surface vs subsurface water masses in the Irminger and Labrador Sea basins, vs those of benthic environments would make sense, as observed during the last climatic cycle49. This decoupling has been particularly important during interstadial intervals and H-events, marked by lighter isotopic compositions of surface water, vs sub-surface to deep North Atlantic water layers.
Looking deeper into the potential linkage between the cold spells of the mPWP and the obliquity cycle, one should note that this interval was characterized by obliquity oscillations of reduced amplitude (Fig. 4), thus by generally lower overall insolation at high latitudes 35 allowing for some ice growth. Such low-obliquity intervals are marked by a ca. 1.2 Ma maximum frequency and have been associated with “third order eustatic cycles and Neogene glaciations” by Lourens and Hilgen (1997)50. A decrease in Earth’s axial tilt results in reduced annual insolation at high latitudes but in an increase at the equator35. Such changes in insolation gradients can generate atmospheric circulation shifts influencing the moisture transport and continental snowfall. In the Northern Hemisphere, for example, a reduction in obliquity enhances summer half-year snowfall by 78% and winter half-year snowfall by 22%35. Because low-amplitude obliquity oscillations will not be encountered within the next Ma (Fig. 4), one cannot consider the mPWP as a close analogue of the near future of the Earth climate, and more specifically here, of future ice conditions of the GIS.
Overall, a closer look at the potential relationship between the geochemical and sedimentological parameters with the obliquity cycles of the mPWP interval, shows important offsets, particularly before 3080 ka (Fig. 3). We have seen that neither the d18O record nor the paleomagnetic record can provide time-anchorages with a millennial time-resolution. We are thus left with an overall time frame for the mPWP glacial pulses, but their assignment to specific insolation and obliquity cycles within the interval would not go beyond the best guess.