Sequential changes in ocean circulation and biological export productivity during the last glacial–interglacial cycle: a model–data study

We conduct a model–data analysis of the marine carbon cycle to understand and quantify the drivers of atmospheric CO2 concentration during the last glacial–interglacial cycle. We use a carbon cycle box model, “SCP-M”, combined with multiple proxy data for the atmosphere and ocean, to test for variations in ocean circulation and Southern Ocean biological export productivity across marine isotope stages spanning 130 000 years ago to the present. The model is constrained by proxy data associated with a range of environmental conditions including sea surface temperature, salinity, ocean volume, sea-ice cover and shallow-water carbonate production. Model parameters for global ocean circulation, Atlantic meridional overturning circulation and Southern Ocean biological export productivity are optimized in each marine isotope stage against proxy data for atmospheric CO2, δ13C and114C and deep-ocean δ13C,114C and CO2− 3 . Our model–data results suggest that global overturning circulation weakened during Marine Isotope Stage 5d, coincident with a ∼ 25 ppm fall in atmospheric CO2 from the last interglacial period. There was a transient slowdown in Atlantic meridional overturning circulation during Marine Isotope Stage 5b, followed by a more pronounced slowdown and enhanced Southern Ocean biological export productivity during Marine Isotope Stage 4 (∼−30 ppm). In this model, the Last Glacial Maximum was characterized by relatively weak global ocean and Atlantic meridional overturning circulation and increased Southern Ocean biological export productivity (∼−20 ppm during MIS 3 and MIS 2). Ocean circulation and Southern Ocean biological export productivity returned to modern values by the Holocene period. The terrestrial biosphere decreased by 385 Pg C in the lead-up to the Last Glacial Maximum, followed by a period of intense regrowth during the last glacial termination and the Holocene (∼ 600 Pg C). Slowing ocean circulation, a colder ocean and to a lesser extent shallow carbonate dissolution contributed ∼−70 ppm to atmospheric CO2 in the ∼ 100 000-year leadup to the Last Glacial Maximum, with a further ∼−15 ppm contributed during the glacial maximum. Our model results also suggest that an increase in Southern Ocean biological export productivity was one of the ingredients required to achieve the Last Glacial Maximum atmospheric CO2 level. We find that the incorporation of glacial–interglacial proxy data into a simple quantitative ocean transport model provides useful insights into the timing of past changes in ocean processes, enhancing our understanding of the carbon cycle during the last glacial–interglacial period.

5 Table S3. Model forcings for MIS across the last glacial-interglacial cycle. Proxy for Antarctic sea ice extent using ssNa fluxes from the EPICA Dome C ice core (Wolff et al., 2010), used to temporally contour MIS model forcings for salinity (Adkins et al., 2002) and polar Southern Ocean piston velocity. Global ocean salinity is forced to a glacial maximum of +1 psu and the polar Southern Ocean is forced to +2 psu, as modified from Adkins et al. (2002). Ocean volume forced using global relative sea level reconstruction of Rohling et al. (2009).

MIS CO2
(ppm) Tables S5-S7 shows ocean δ 13 C, CO 2− 3 and ∆ 14 C proxy data mapped and averaged into SCP-M boxes, and averaged in MIS slices.          Table S8. Welch's paired T-tests for the independence of deep-abyssal offsets in mean δ 13 C with respect of the penultimate interglacial period (MIS 5e), for the periods MIS 1-5e. The null hypothesis is that the deep-abyssal offset in mean δ 13 C in each MIS is not statistically independent of MIS 5e (i.e. statistically the same and not supportive of a change in deep-abyssal δ 13 C distribution that may be delivered by a changed ocean process). p-values>0.05 lead to the null hypothesis being accepted, whereas p-values <0.05 lead to the null hypothesis being rejected and confirm statistical independence of the deep-abyssal offsets relative to MIS 5e (perhaps supportive of a changed ocean distributive process in the glacial period). Deep-abyssal offsets for the Pacific-Indian during MIS 2-MIS 5d are statistically independent of MIS 5e, supportive of a changed oceanic distribution of δ 13 C throughout the glacial period. The MIS 1 Pacific-Indian deep-abyssal δ 13 C offset is not statistically independent of MIS 5e, indicating a similar deep-abyssal δ 13 C distribution between the last and penultimate interglacial periods. For the Atlantic Ocean, deep-abyssal mean δ 13 C offsets are not statistically independent with respect to MIS 5e (p-value >0.05, accept null hypothesis), until the period MIS 2-4. Atlantic deep-abyssal mean δ 13 C offset in MIS 1 is not statistically different from MIS 5e. Table S9 shows the parameter values for global overturning circulation (GOC, P si 1 ), Atlantic meridional overturning circulation (AMOC, P si 2 ) and Southern Ocean biological export productivity (Z) from the model-data experiments described in the main text.

Model-data experiment results
5 Table S10 shows the optimised model-data experiment results for atmospheric CO 2 , δ 13 C and ∆ 14 C, and the terrestrial biosphere.
Table S11 shows the optimised model-data experiment results for oceanic δ 13 C.
10 Table S12 shows the optimised model-data experiment results for oceanic CO 2− 3 .
7  Figure S2 shows the full range of experiments undertaken and the resulting model-data error mesh charts. These charts plot the difference between the model outputs and the proxy data, as a function of each of the three parameters varied in the model-data experiments (P si 1 , P si 2 and Z).

Model code and data
The model code, processed data files, model-data experiment results, and any (published) raw proxy data gathered in the course 5 of this work, are located at https://doi.org/10.5281/zenodo.3559339. No original data was created, or unpublished data used, in this work. Figure