|The authors have addressed most of my comments adequately, and I find that the manuscript has improved considerably. However, there are a few remaining issues that have to be taken into account before publication. |
1) The rationale for using a PI OHT is discussed in much more detail. In my opinion this discussion and the new Figure 2 makes the study more convincing. There are however two additional suggestions that I would like to make.
In Figure 2 a rather wide colour scale is used, ranging between -26°C and 0°C, while the MARGO anomalies do not exceed -12°C. The figure would be a lot more informative if a more narrow scaling range would be applied (i.e. between -12°C and 0°C).
I agree that using "BO2009 OHT" is much less confusing than the term "LGM OHT". However, I find the discussion of the simulated LGM SSTs still confusing. For instance, on line 148: "We find that the annual mean LGM SST response in CAM3 using PI OHT is in better agreement with the response in CSSM4". When discussing the SSTs simulated with CAM3, I would suggest making a distinction between LGM[PI subscript] SSTs and LGM[BO subscript] SSTs.
2) In my previous review, I proposed to include a flow diagram to explain the full, sequential experimental setup. Although the authors have considerably improved Section 2.1, I still think this could make the methodology easier to understand.
3) The authors have added a sentence to explain the applied modern mixed-layer depth. "Aside from areas covered by perennial sea ice, simulated changes of the LGM mixed-layer depth are small compared to PI (of order 10 m; Brandefelt and Otto-Bliesner, 2009); following Löfverström et al. (2014) we therefore use a modern annual mean mixed-layer depth in all simulations". As explained by the authors in their reply to the reviewers, the simulated mixed layer depths from the LGM experiments were not saved on the CCSM3 data server. I suggest to simply mention this in the manuscript as well, since I consider the small difference between LGM and PI mixed-layer depths not a good argument to take the PI mixed-layer depth if the LGM mixed-layer depth is also available.
4) The statistical significance of the simulated anomalies is now tested using a Student t-test. However, if I look at the figures, it is not entirely clear what results are not statistically significant. For instance, in Figures 3d, f and h, the simulated temperature anomalies are presented. The caption explains that all results shown in coloured shading are statistically significant at a 95% level. But what about the white areas on the map? Are they all statistically not significant, or do they also include grid cells with small anomalies between -0.1 and 0.1 that are statistically significant?
Line 134: typo, "the the"
Line 365: should be "of Beghin"