Please see my review in the attached. 

Peter Köhler presents a series of box model experiments that explore the impact of changes in the isotopic signature of geological sources, related to carbonate weathering and/or volcanic emissions on the carbon cycle over orbital timescales. His model experiments reveal that the 400,000 periodicity observed in the 13C cycle (notably absent from any other climate variables) may have been related to changes in the isotopic composition of sources. Interestingly, the isotopic signatures for both sources required to best align the simulations with the existing benthic foraminifera 13C reconstructions, have most power in the obliquity band, suggesting that the ultimate forcing mechanism may relate to continental ice dynamics.

The manuscript is certainly a valuable contribution and should be published in Climate of the Past upon revision. It tackles an intriguing, yet outstanding, issue in paleoclimatology and offers a relatively simple solution. I am not a modeler myself, and I have to admit that the argumentation can be quite cumbersome to read in places. In particular, it would be great if §3 could be somewhat streamlined to make it easier to grasp for a wider audience.

Major comments -

I understand that you change the contribution of external (i.e. geological) sources of CO2 to the system (d13Crock and d13Cv) to understand how the global ocean mean d13C may have changed over time and whether these changes in the input term can reproduce the 405 kyr eccentricity cycles apparent in the Plio-Pleistocene marine d13C record.

However, as I understand, the output terms (i.e. CaCO3 vs. Corg export and burial) are left to evolve untuned in the model? In particular, it is unclear to me whether the 13Corg composition is allowed to vary in space and time. Given that recent reconstructions showed that the burial of Corg and CaCO3 in pelagic sediments varied quite substantially (Cartapanis et al., 2016 & 2017), it is not clear to me how these output terms are considered in the box model.

Also, both d13Crock and d13Cv are characterized by abrupt (millennial-scale?), large-amplitude (> 10 permil) oscillations across time (Fig. 3 bottom panels), which seem unrealistic to me. Can you maybe elaborate how these abrupt shifts can be explained? Are these related to high- vs. low-latitude erosion/weathering processes that vary on G-I timescales?

Minor comments -

p. 1, l. 13. Pre-Pliocene parts of the Cenozoic were largely ice free in the Northern Hemisphere only. Ice sheets have been present on Antarctica at least since the Oligocene.

p. 5, l. 101 – not sure what is meant by “wider tropics”?

p. 7, l. 141. Maybe a naive question, but does the sediment model include Corg cycling or only CaCO3? p. 11, l.263 – Can you briefly summarize what these internal processes are?

p. 16, l. 330-335. I would expect that changes in land ice load/sea level would induce a delayed response in subaerial/submarine volcanism due to the inertia of processes in the upper continual crust/mantle.

Panels d and e in figure 1 as well as all panels in figure 4 are somewhat difficult to read. Wouldn’t make sense to have these figures in colour?