|I was asked to comment on the revised version of the manuscript by Hutchinson and colleagues. I have read the revision along with authors’ response to the comments provided by the two previous reviewers. In general, I feel that authors have done an excellent job addressing reviewers’ comments and improving the presentation of the manuscript.|
In this study, Hutchinson et al. conducted a thorough and up-to-date review on the EOT study. The review includes marine and terrestrial proxy data, model simulations, and model-data comparisons, as well as paleogeography and forcing mechanisms. Even though I have been working on this topic for a while, I could still get some new information from this review. Re-defining and clarifying the terminology used in the EOT study in Section 1 is also very helpful to eliminate the ambiguity of the terminology used in current studies. Thus my overall assessment on this review is very positive and I’d like to see this review officially published after technical corrections/minor revisions. Below I provide some minor comments/suggestions for authors’ consideration in their revision, which are mostly for clarification and correcting typos.
Comments and Suggestions:
1. Low-latitude TEX86 records: It is pretty clear that some of the low latitude TEX86 records showing large amplitude of EOT cooling is problematic. Liu et al. (2009) made it clear that the signal, especially from Site 803 and 998, may be related to non-thermal factors, and suggested that low-latitude cooling should be on the order of 2-3C. I hope these problematic TEX86 data were not used in authors’ data-model comparisons (Fig. 8). Otherwise, it would affect authors’ assessment on the model performance. The data points based on these TEX86 records could also be removed from Fig. 3 in order not to mislead readers.
2. Authors correctly discussed influencing factors for individual temperature indicators, but I feel the completeness of the individual records also matters in order to derive reliable temperature changes across the EOT. For instance, the available 336 UK’37 SST record only has a couple of data points from a snapshot time interval. Thus it may not be representative of the mean SSTs for the broadly defined Late Eocene and Early Oligocene periods. As currently data resolution in many SST records is not very high and some key intervals are missing due to sediment hiatus, this could largely contribute to the uncertainty in the estimates of SST changes for the two intervals. Authors could add one or two sentences somewhere to address this issue.
3. The time interval to represent post-EOT: Authors used a broad time interval between 33.9 Ma and 30 Ma to represent post-EOT. However, based on the relatively well resolved 1404 SST record, SST values between 29 Ma and 31 Ma could reach the late Eocene level, thus the mean SST over this broad time interval could potentially underestimate post-EOT temperature changes. I would suggest using a time interval broadly consistent with the defined interval for the EOGM, 33.9-33.16 Ma for post-EOT. On the other hand, I recognize that this approach could reduce the number of records to be qualified for this requirement. If the number is significantly reduced, I would suggest discarding my suggestion here, but perhaps it is better to mention in the text that the shorter time interval is more appropriate but due to practical reasons not adopted. Lastly, I note that 1404 SST record was not used in Fig. 3, which could fit the data gap around 40N.
4. CO2 changes across the EOT: Authors mentioned the intriguing rebound of CO2 in Pearson et al. (2009) record. Indeed, such a feature is quite common in alkenone-based CO2 records reconstructed from Southern Ocean. Pagani et al. (2011) suggested that oceanic changes in the Southern Ocean due to opening gateways could bias the long-term CO2 trend, and the reconstruction from tropical sites 925/929 is more representative of atmospheric CO2 changes across the EOT. Also recently Zhang et al. (2020, GCA, 281, 118-134, Fig. 10) have adjusted the assumed b values to allow the reconstructed CO2 values falling into a reasonable range. I’m not sure whether it is worth for authors to discuss this in the text, but it does provide a slightly better justification for the CO2 changes used in model experiments (~300 ppm from data and halving in model), as tropical sites provide larger CO2 changes across the EOT than southern Ocean sites.
Other minor ones:
1. Table 1: EOGM should be 33.65 Ma to ~33.16 Ma in order to be consistent with the ~490 kyr duration. A typo here? Fig. 1 also suggests 33.16 Ma.
2. Line 325, Meridional Overturning Circulation: Authors could briefly mention/summarize the various timing of the onset of AMOC suggested in previous studies.
3. Line 496: The UK’37 index saturates at ~29C for the Muller et al. calibration (1C makes slight difference).
4. Line 508-509: see my previous comment # 1.
5. Line 519: I believe they suggest an apparent decrease in CO2 over that period, not increase.
6. Line 576-578. I feel that the 2C surface cooling is OK for tropics, but for high latitude southern hemisphere, the cooling should be >5C. I recognize that some of the records could not resolve the detailed structure of cooling, but if only 2C cooling is assumed at the first step in the Southern Ocean, then additional cooling has to take place at the second step? Perhaps rephrase this sentence?
7. Line 721: I think this is related to the completeness of the SST record. The same might be also true for Site 277 and 1090 where data points representing the EOGM interval are generally missing.
8. Line 924-926: So the shallow Barents Sea gateway re-opened some time later? Is there any geological evidence for that?