Modelling carbon isotopes to examine ocean circulation and the marine carbon cycle

Understanding how the Meridional Overturning Circulation (MOC) might change in response to anthropogenic climate change has important repercussions for predictions of surface climate and the carbon cycle, but our ability to interpret future changes in the context of natural variability is limited by the short observational record (approximately 15 years). Key insights into the temporal variability of the MOC can be gained from geochemical tracers (e.g. carbon isotope ratios, δ13C and ∆14C) in geological archives. However, interpreting proxy records is complex because the isotopes respond to both physical and biogeochemical processes. Isotope-enabled models enable us to directly compare simulated and observed values, and investigate plausible mechanisms for the measured signals. Nonetheless, carbon isotopes are not routinely included in numerical climate models because of the computational cost of fully spinning up the deep ocean circulation and the marine carbon cycle (which requires run lengths between 5000 and 15,000 years).

This thesis uses carbon isotopes to investigate ocean circulation and the marine carbon cycle in FAMOUS, a fully coupled atmosphere-ocean General Circulation Model that offers a unique balance between speed and complexity. Using the most recent published generation of the model, this study demonstrates that small regional salinity drifts (that occur because of inaccuracies in the formulation of the hydrological budget) can lead to significant changes in the representation of the MOC in multi-millennial climate simulations, even under constant pre-industrial boundary conditions and when the global hydrological budget has been forcibly closed. An earlier generation of the model, which provides a more accurate and stable representation of the pre-industrial MOC, is therefore used for the implementation and validation of a new carbon isotope scheme (13C and 14C). The simulated δ13C distributions capture the physical and biogeochemical behaviour of the model well, but are offset from observed values because of inaccuracies in the biological pump and the large-scale ocean circulation. The ∆14C tracer is less sensitive to these biases and shows good agreement to observations, both spatially and temporally, which demonstrates the skill of the model in representing carbon uptake and transport. Radiocarbon ages are typically interpreted in terms of ventilation, but comparing the simulated 14C ages to idealised water ages suggests that 14C is only a good ventilation tracer in well-mixed regions, where the physical component of the solubility pump is a more dominant control on dissolved inorganic carbon distributions than the chemical component. The local balance between physical and biogeochemical processes should therefore be considered when interpreting ∆14C in proxy records to avoid drawing erroneous conclusions about palaeocean circulation.

Overall, this study demonstrates the utility of including carbon isotopes in numerical climate models: it aids our interpretation of geochemical tracers in geological archives, and provides a sensitive and holistic tuning diagnostic for evaluating and improving model performance.