I am disappointed by the response of the authors and it reflects a missed opportunity to improve the manuscript based on input provided by the reviewers and E. Galbraith.
A marked-up version with the changes identified would have been helpful for the reviewer(s).
The authors should provide the simulations requested by E. Galbraith and reviewer 2. The authors should summarize the finding of earlier freshwater model experiments in a paragraph in the introduction. The authors should summarize the literature for changes in SH westerlies in another paragraph in the introduction. The authors should remove the misinterpretation of Tschumi et al, 2011 in the next manuscript.
1) Figure 10 reveals that the simulated d13C anomalies are in the range from 0.5 to – 1.6 permil whereas the observation based range is about a factor of two smaller. This is an important finding and should be noted in the result section. Moreover, the authors should try to find a model solution where they can reduce this large mismatch. The author should clearly spell out in the abstract and the discussion that this most likely leads (or at least “may lead”) to an overestimation of the response in CO2 to an AMOC shut-down for the LGM.
2) The authors apply a cost-efficient Earth System Model of Intermediate Complexity with an energy balance atmosphere. The advantage of such an approach, e.g. in contrast to a comprehensive ESM, is that many simulations can be carried out and the sensitivity of simulated responses to initial states and parameters can be explored.
The authors should extend the simulation FW015 (or preferentially a simulation that starts from LGM conditions) to a state where the AMOC is turned on again. They may restart their simulation and add a negative freshwater anomaly at 14.7 ka to mimic the onset of NADW at the beginning of the B/A.
I appreciate that it might be more difficult to get an LGM state, but this should also not be out of reach, given that the model has been initialized for the LGM before. Further, the authors state: “Even though it would be possible to use a simulation with LGM boundary conditions its value would be low if the deep ocean circulation and carbon cycle simulation is inconsistent with observations.” and “Whereas simulations with glacial boundary conditions (lower GHG, orbital parameters, ice sheets) are easy and have been done before (e.g. in PMIP)
It is possible (or even “easy”) and the authors should run the model for an initial state with a shallower than modern AMOC to see whether they can reduce the d13C data-model mismatch and what the implications are for atm. CO2 and d13C
It is a weak argument that the LGM ocean is poorly constraint and that their model might be unrealistic for the LGM. This argument applies equally for the simulations now described. We know that the modern ocean is different from the LGM state and the early termination, e.g. from d13C and 14C data etc. Still the authors use the modern condition as a starting point for the simulation. They do not hesitate to use the results for the modern conditions to come up with their strong and, as pointed out by E. Galbraith, likely unrealistic conclusion on the early CO2 rise.
3) A paragraph is needed in the introduction that summarizes the results from earlier freshwater hosing experiments. This would allow the reader to put this work in the context of the literature. It would provide a starting point for the work done here. In particular the findings that the CO2 responses, and related carbon stock changes on land and in the ocean, are found to be very sensitive to the initial state must in my opinion be discussed upfront. Similarly, responses were found to be smaller for LGM boundary conditions than for preindustrial conditions. A point that also need to be mentioned upfront. I have already provided various references in my initial review. Another study that is of interest in this context is the work by Menviel et al., QSR, 2012 where the results of transient glacial-interglacial simulations is discussed and model outcomes are compared with proxy data. This reveals that changes in AMOC for a reasonably realistic LGM state have a much smaller impact on atm. CO2 than postulated by Schmittner and Lund.
4) A paragraph is needed in the introduction that summarizes the results from earlier wind stress experiments on deglacial CO2. The available results suggest a weak influence on changes in SH westerlies on atmospheric CO2 for the deglacation. Thus, the results presented by the authors fall well within the range of earlier quantitative studies and do not provide much new insight. There is even one study available with the UVIC model (d’Orgeville et al., GRL, 2010) which is not mentioned either.
See for example Menviel et al., 2008, Tschumi et al, PO, 2008, d’Orgeville etal, 2010, Lee et al., PO, 2011, Voelker and Köhler, PO, 2014) for wind and CO2, or also work by England and Sijp and by Böning et al., NatGeo, 2008 etc. considering physics only.
The authors should condense the text and figures on SH westerlies as it is a confirmation of earlier work.
5) Abstract, line 3, and line 22-25: The authors focus here on SO wind changes as the only mechanisms for SO circulation changes. The text gives a wrong impression. While it is reasonably clear from the literature that the sensitivity of atm. to realistic wind stress and wind position changes is fairly weak, there are other mechanisms that may have been responsible for deep ocean circulation changes in relation to the Southern Ocean atmospheric window. It is clear from the literature that SO changes did take place, e.g. as indicated by opal records, the d13C and D14C records. The sentence “We propose that the observed early deglacial rise in atmospheric CO2 and the decrease in _13CCO2 may have been caused by an AMOC induced decline of the ocean’s biologically sequestered carbon storage without the need to invoke changes in Southern Hemisphere winds.” gives the impression that SO changes played no role at all. As mentioned above, it is not a new finding that SO wind changes played a small role. The more general statement, that there is no need to invoke SO changes to explain atm. CO2 is can hardly be reconciled with the available proxy evidence. In particular, as this MS makes no statement on the role of SO D14C and on SO opal fluxes. If the authors wish to talk about SO wind in the abstract, they should do this in a clean statement, e.g .” Our results confirm the findings of earlier model studies that changes in SO winds likely played a small role for the deglacial CO2 rise.”
6) Line 243 ff and line 283 The misinterpretation of the work by Tschumi et al., 2011 is to be removed.
Tschumi et al., 2011 quantify the influence of SO ventilation changes on CO2 and d13C, but they do not postulate that SO wind changes are responsible for ventilation changes. They clearly state this in their paper at several places. For example, at the beginning of the discussion:
“Second, changes in deep ocean ventilation are generated by varying the magnitude of the wind stress in the Southern Ocean by more than a factor of two. This is a convenient technical tuning mechanism to adjust the deep ocean ventilation age in the model. However, we stress the fact that these variations in SO wind strength exceed the range of realistic changes. The atmospheric CO2 response to realistic variations has been found to be rather modest (Tschumi, 2008; Menviel et al., 2008).
7) Line 267: The authors state that a deep ocean state estimate of the global LGM circulation and carbon cycle that is consistent with sedimentary _13CDIC reconstructions does not exist. While of course no model is perfect, there exist simulations of LGM conditions including isotopes. For example, Menviel et al., QSR, 2012 provide a set of simulations with different LGM states and their simulated d13C anomalies for the LGM as shown in their Figure 9 compare reasonably well with reconstructions. As mentioned above, simulated changes in atm. CO2 in response to changes in AMOC are relatively small and do not explain the initial deglacial CO2 rise.
8) The authors state in the discussion section when dealing with the d13C model-data mismatch: “
„the model would overestimate
changes in volume fluxes and perhaps carbon isotopes even if a complete AMOC collapse
did occur during HS1.“ They should spell out that this applies not only to the volume flux and carbon isotopes but also to atmospheric CO2 as indicated by previous studies.
Further comments:
line 133: Sentence is confusing. “Because biologically sequestered, organic carbon is isotopically light (_13Corg = -20‰) its loss increases deep ocean _13CDIC by 0:06 permil (Fig. 3f) and its gain decreases _13CDIC (by _ 0:3 permil) in the surface ocean and _13 135 CCO2 by _ 0:25‰in the atmosphere (Fig. 1d).”
My understanding is:
- The decrease in global DOC inventory as shown in Figure 3a tends to decrease d13C of DIC, atm. CO2 and land C as isotopically-light carbon is moved from DOC to the other carbon reservoir.
-The increase in land carbon storage tends to increase d13C in the atmosphere and ocean
-The decrease in the efficiency of the biological pump tends to decrease d13C in the atm. and the surface ocean and in DOC and POC and to increase d13C in the deep ocean.
The result of these three processes is a slight increase in the mean ocean signature of DIC.
Figure 1: Reference to Schmitt et al., 2012 seems missing here. |