Extended Data Fig. 1: Preindustrial Atlantic \(\left[ {{{{\mathrm{CO}}}}_3^{2 - }} \right]_{{{{\mathrm{as}}}}}\) vs DICas as shown in Fig. 1.
From: Millennial and centennial CO2 release from the Southern Ocean during the last deglaciation

Data are from ref. 15, based on the calculation method from ref. 16. Black curve represents the best fit of the data. Simply put, when a water mass sequesters more atmospheric CO2, it has lower \(\left[ {{{{\mathrm{CO}}}}_3^{2 - }} \right]_{{{{\mathrm{as}}}}}\) and higher DICas, and vice versa. For example, adding CO2 into a package of water would increase its DICas. At the same time, because the added CO2 would convert some \(\left[ {{{{\mathrm{CO}}}}_3^{2 - }} \right]_{{{{\mathrm{as}}}}}\) into bicarbonate, its \(\left[ {{{{\mathrm{CO}}}}_3^{2 - }} \right]_{{{{\mathrm{as}}}}}\) should decrease. Thus, the negative \(\left[ {{{{\mathrm{CO}}}}_3^{2 - }} \right]_{{{{\mathrm{as}}}}}\)–DICas correlation is exactly expected from the CO2 system theory30. See ref. 16 and Methods for detailed discussions.