Sediment fluxes dominate glacial–interglacial changes in ocean carbon inventory: results from factorial simulations over the past 780 000 years

Abstract. Atmospheric CO2 concentrations varied over ice age cycles due to net exchange fluxes of carbon between land, ocean, marine sediments, lithosphere, and the atmosphere. Marine sediments and polar ice cores archived indirect biogeochemical evidence of these carbon transfers, which resulted from poorly understood responses of the various carbon reservoirs to climate forcing. Modelling studies demonstrated the potential of several physical and biogeochemical processes to impact atmospheric CO2 under steady-state glacial conditions. However, it remains unclear how much these processes affected carbon cycling during transient changes of repeated glacial cycles and what role the burial and release of sedimentary organic and inorganic carbon and nutrients played. Addressing this knowledge gap, we produced a simulation ensemble with various idealised physical and biogeochemical carbon cycle forcings over the repeated glacial inceptions and terminations of the last 780 kyr with the Bern3D Earth system model of intermediate complexity, which includes dynamic marine sediments. The long simulations demonstrate that initiating transient glacial simulations with an interglacial geologic carbon cycle balance causes isotopic drifts that require several hundreds of thousands of years to overcome. These model drifts need to be considered when designing spin-up strategies for model experiments. Beyond this, our simulation ensemble allows us to gain a process-based understanding of the transient carbon fluxes resulting from the forcings and the associated isotopic shifts that could serve as proxy data. We present results of the simulated Earth system dynamics in the non-equilibrium glacial cycles and a comparison with multiple proxy time series. From this we draw several conclusions. In our simulations, the forcings cause sedimentary perturbations that have large effects on marine and atmospheric carbon storage and carbon isotopes. Dissolved inorganic carbon (DIC) changes differ by a factor of up to 28 between simulations with and without interactive sediments, while CO2 changes in the atmosphere are up to 4 times larger when interactive sediments are simulated. The relationship between simulated DIC (−1800–1400 GtC) and atmospheric CO2 change (−170–190 GtC) over the last deglaciation is strongly setup-dependent, highlighting the need for considering multiple carbon reservoirs and multi-proxy analyses to more robustly quantify global carbon cycle changes during glacial cycles.