The North Atlantic cooling reconstructed at 130 ka is typically interpreted as a fingerprint of a relatively weak AMOC circulation in response to meltwater input13,14. Deglacial meltwater stratified the upper North Atlantic, inhibiting deep-water formation, reducing northward oceanic heat transport, thus decreasing North Atlantic surface temperatures26. Including a freshwater perturbation in our model reproduces this behavior. Increasing the freshwater perturbation results in i) a more pronounced North Atlantic cooling relative to the unperturbed state (Figure 2a-d, left panel), and ii) a tendency towards more depleted values in both δ18Osw and δ18Oc. (Figure 2a-d, middle and right panels). Injecting freshwater into the North Atlantic IRD belt (Figure 2b-d) also leads to a weakening of North Atlantic Deep Water (NADW) formation, which increases with the magnitude of freshwater input (as demonstrated by the AMOC index, Figure 4a). Density anomalies are indeed transported from the central Atlantic to deep-water formation sites, which occurs in the Labrador and Nordic Seas in our model. Changing the location of freshwater injection from the IRD belt to the Labrador and Nordic Seas results in more pronounced North Atlantic cooling and δ18Osw and δ18Oc depletions (Figure 2e), because meltwater input now directly occurs over convection sites.
When compared to the reconstructions, the simulation without any freshwater influence yields ocean temperature anomalies which are too warm (LIG-PI Anomaly of -1.76, RMSE of 2.3 relative to the reconstructions), whereas including freshwater tends to over-estimate the cooling anomaly (LIG-PI anomaly (RMSE relative to reconstruction) of -3.56 (1.76), -4.03 (2.09), -4.17 (2.25) for the 0.05 Sv, 0.1 Sv, and 0.2 Sv injections). Changing the injection location leads to a further over-estimation of the anomaly (PI anomaly of -3.39, RMSE 2.19). For the isotopic signatures in seawater, increasing the injection strength leads to too depleted values (RMSE of 0.5, 0.8, and 1.14 for 0.05, 0.1, and 0.2 Sv); the simulation with an alternative injection again yields values which are too depleted (RMSE of 2.64). From this set of freshwater perturbation experiments, the 0.05 Sv injection into the IRD belt is best suited to reproduce the proxy reconstructions, yielding comparable temperature and isotopic anomalies.
We are able to disentangle and quantify three mechanisms influencing the δ18Oc signature in foraminiferal calcite resulting from a freshwater perturbation (described in detail in the Supplementary Information) using our stable-oxygen isotope enabled simulation. This is achieved by re-performing the 0.2Sv injection experiment, but without modifying the isotopic signature of the seawater. This allows us to separate changes induced by the physical effects of a freshwater perturbation from the changes caused by injection meltwater with a strongly depleted δ18O signature. With this technique, we are able to determine δ18Oc changes purely due to oceanic temperatures, yielding an enrichment of +1.0 to +2.5 ‰ in the Nordic seas (Figure 3-a). Secondly, we can detect changes due to modified circulation following the freshwater influence, which induces a slight enrichment of +0.5‰ in the west Atlantic (Figure 3-b), and lastly changes due to injection of strongly depleted meltwater which causes a dilution of the δ18O signature in the seawater (Figure 3-c). In this case, the entire region becomes increasingly depleted, with signature changes of -1.0‰ to -2.0‰. While this entanglement is demonstrated for the 0.2 Sv case, these effects are still present in the weaker injection scenarios. As noted earlier, the perturbation of 0.05 Sv yields a markedly smaller RMSE between the reconstructions and the simulation compared to the other injection experiments and might be a suitable representation of the climate dynamics at work at 130 ka BP. Yet, in order to attain these isotopic and temperature values in the simulation, the injection of freshwater needs to be sustained for at least 300 years. This perturbation strength and duration would already result in a sea level increase of 1.4 meters. The proxy reconstructions cover a time span of 131 to 129 ka BP, and maintaining the injection for the entire period would imply a sea level increase of 9.3 meters; which is at the upper limit of the reconstructed sea level changes for the entire LIG period spanning from 130 to 115 ka BP31,32. Furthermore, the dilution effect described above would be continuously active during these 2000 years, leading to increasingly depleted δ18O seawater values as the freshwater forcing is maintained. Therefore, although the 0.05 Sv injection scenario provides a possible explanation and yields the smallest model/data mismatch based upon the RMSE values, we also explore an alternative mechanism to explain the ocean state reconstructed in the proxy data.