Input and output fluxes of surface CO2 over the Late Quaternary

Ice core archives allow us to retrieve the atmospheric CO₂ concentration of the past 800,000 years characterized by periodically lower and higher CO₂ levels corresponding to ice ages and interglacials, respectively. However, there is no broadly accepted consensus regarding the leading drivers of such variability. To a large extent, what prevents us from identifying the mechanisms that underlie changes in atmospheric CO₂ concentrations is our inability to split the overall atmospheric CO₂ budget into its sources and sinks terms, thereby assessing the fluxes of carbon among different reservoirs. Here, we use a reversible-jump Markov chain Monte Carlo (rj-McMC) algorithm to invert the atmospheric CO₂ concentration dataset provided by the EPICA ice core based on a general forward formulation of the geological carbon cycle. We can quantify the most likely temporal variability of atmospheric carbon fluxes in ppm/yr throughout the last 800,000 years. Results suggest that temperature changes have been driving the variations of atmospheric CO₂ until the Mid-Brunhes Event (MBE), when the onset of a progressive cyclic increase of  the atmospheric carbon fluxes marks a distinct behavioral change of the climate system. We ascribe such change to mechanisms internal to the Earth system, possibly related to the deglacial triggering of global volcanism and associated feedbacks on climate or a combination of geological, biological, and physical processes. Regardless, our unprecedented quantification of past atmospheric input and output CO₂ fluxes provide (1) new constraints for climate carbon cycle and paleoclimate models to assess dominant climate-driving mechanisms, and (2) the benchmark for climate models intercomparison projects and better assessing the anthropogenic perturbation to the geological carbon cycle an associated climatic effect.