Skinner et al., present a compilation of ocean-atmosphere radiocarbon age offset (B-Atm) across the last deglaciation. A Bayesian interpolation method is further applied to the global compilation to provide interpolated fields as well as globally averaged B-Atm values. By performing a suite of experiments with a model of intermediate complexity, the authors further assess whether the reconstructed changes in deglacial B-Atm are due to ventilation changes or atmospheric radiocarbon dynamics.

The manuscript presents a unique and useful data compilation and tackles an important problem, i.e. deglacial changes in the carbon cycle.  It is of high scientific quality and well written.

1) My main comment is related to the presentation of the modelling work and of the some results. I am puzzled by Figure 9. I find it very surprising to see such a different CO2 response across the different background conditions. For example, at BA assuming the symbol gets smaller as the forcing gets stronger, there is a linear decrease in CO2 as the SO winds decrease. However, no such relationship exists at 41ka and the LGM. Can the authors explain this? Maybe the colors have been mixed up? In addtion, please clearly state which symbol is which (size).

I don’t think it is necessary to display Table 2 in the main manuscript. There is a lot of information there, that should be simply conveyed in (an improved) Fig. 9. In addition, INT, FIX and CONST need to be defined in the caption of Table 2.

Fig. 10a: I find it a bit surprising to only use the PI sensitivity experiments in that figure, when the previous figure apparently showed a different sensitivity of the DpCO2/D14C as a function of background state. I suppose using the BA would be more appropriate. I am also confused as to the necessity of the PI simulations.

2) Conclusions

I find the conclusions a bit unclear. From the abstract L. 26-28, and L. 555-557, it is mentioned that “evolving ocean-atm exchange can account for 1/3 of the total GIG CO2 rise”. Most studies suggest that the terrestrial carbon reservoir was smaller at the LGM than during the Holocene. This implies that the deglacial CO2 rise was due to a decrease in oceanic carbon, which in simple terms implies regional increase in CO2 outgassing and/or reduced CO2 ocean uptake…. therefore ocean-atm exchange… what am I missing here? What do the authors mean?

It is further suggested that “half of the CO2 rise appears to have been associated with the BA”. It is a confusing statement, particularly in the abstract as lacking context. Can the authors more precisely state what they mean?

3) Minor edits:

• L. 86-88: I do not understand the logic in that sentence.
• L. 241: “and” did not consider?
• L. 253-256: This sentence is confusing.
• L. 487: Figure 9 does not seem to suggest that the biases are attenuated for large changes in ventilation (i.e. no bias change as SO winds are decreased, even if colors/symbols have been mixed up).
• Fig. 8: Please provide the latitudinal boundaries used to define North Atlantic, North Pacific and Southern Ocean.

In this manuscript, authors Skinner et al. perform several tasks that are of importance for the understanding of deglacial changes in atmosphere and ocean Δ14C: 1) By using an interpolation method, they produce gridded three dimensional fields of radiocarbon ages in the global ocean for different time slices associated with the last deglaciation. They then use models to attribute those changes and their relationship with air-sea CO2 transport to different processes in the ocean interior, and discuss the implications fro atmospheric Δ14C. The paper is clearly written and the figures are adequate. Follow some comments: 

Major comment:

The only major comment I have is regarding the concept of transport rates governing ocean Δ14C. Throughout the manuscript, transport rate is discussed as a factor governing atmospheric-benthic Δ14C offsets. In the paper Muglia and Schmittner (2021), an analysis is performed with an ensemble of LGM model simulations. They find that mean ocean radiocarbon ages are much more closely related with deep ocean water mass structure than with overturning transport (please see Figs. 5 and 6 in that paper). I believe the authors should consider changing the attribution of Δ14C changes to deep water mass transport to deep water mass structure, and reflect that in the final version of their paper. 

Minor comments: 

Lines 60-70: Please include the values and uncertainties (if available) of ocean-atmosphere radiocarbon age offsets calculated by the cited literature. 

Line 320: "A few data points". Imprecise. Say the number of points. 

Line 324: "This comparison highlights the Indian basin as an important target for future work". What type of future work? Please specify. 

Line 327: I don't understand the correlation coefficients expressed here. Are you calculating a correlation coefficient between data and a gridded interpolation calculated from the same data? If that is the case, what is the purpose of such calculation?

Line 337: I can't use Figure 5 to compare with LGM with the modern state because the modern state is not plotted. 

Figure 8: d14O? You probably mean d18O. 

Line 1174: Typo "indicate are for constant"

Figure 8: Please specify how the splines where calculated. What data did you use? Did you calculate them from time slices or using the x-axis of the age models?

Figures 9 and 10: Please us the same color scheme for the experiments in these two figures. 

Data availability comment:

 

The production of 3-dimensional past global fields of Δ14C based on data interpolation is very useful for the paleoceanography community. It will be good if the authors make those fields available on a repository.