Carbon and carbon-13 in the preindustrial and glacial ocean

Abstract

Listen

Translate article

Despite their importance for Earth’s climate and paleoceanography, the cycles of carbon (C) and its isotope 13C in the ocean are not well understood. Models typically do not decompose C and 13C storage caused by different physical, biological, and chemical processes, which makes interpreting results difficult. Consequently, basic observed features, such as the decreased carbon isotopic signature (δ13CDIC) of the glacial ocean remain unexplained. Here, we review recent progress in decomposing Dissolved Inorganic Carbon (DIC) into preformed and regenerated components, extend a precise and complete decomposition to δ13CDIC, and apply it to data-constrained model simulations of the Preindustrial (PI) and Last Glacial Maximum (LGM) oceans. Regenerated components, from respired soft-tissue organic matter and dissolved biogenic calcium carbonate, are reduced in the LGM, indicating a decrease in the active part of the biological pump. Preformed components increase carbon storage and decrease δ13CDIC by 0.55 ‰ in the LGM. We separate preformed into saturation and disequilibrium components, each of which have biological and physical contributions. Whereas the physical disequilibrium in the PI is negative for both DIC and δ13CDIC, and changes little between climate states, the biological disequilibrium is positive for DIC but negative for δ13CDIC, a pattern that is magnified in the LGM. The biological disequilibrium is the dominant driver of the increase in glacial ocean C and the decrease in δ13CDIC, indicating a reduced sink of biological carbon. Overall, in the LGM, biological processes increase the ocean’s DIC inventory by 355 Pg more than in the PI, reduce its mean δ13CDIC by an additional 0.52 ‰, and contribute 60 ppm to the lowering of atmospheric CO2. Spatial distributions of the δ13CDIC components are presented. Commonly used approximations based on apparent oxygen utilization and phosphate are evaluated and shown to have large errors.