|This is a completely rewritten paper that compares the climate response to 1.5C and 2.0C global mean warming in uncoupled, slab-ocean and fully coupled simulations with the NorESM model. Compared to the previous version, the manuscript has greatly improved featuring 1) a more focussed approach, 2) the inclusion of fully coupled simulations and 3) a clearer motivation for using the slab-ocean version of the model. I appreciate the efforts that went into this. I do have, however, major and minor concerns that the authors should address before I can recommend publication of this manuscript. |
1) The paper focusses on 'the role of ocean and sea-ice feedbacks', implying that the difference in response between the model versions is only due to the fact that feedbacks are allowed in the slab ocean and fully coupled version and not allowed in the AMIP version. An important aspect that is not elaborated on however is the fact that the prescribed SST and SI fields in the AMIP runs are taken from the CMIP5 multi-model mean (more specifically: the HAPPI-mip protocol), and not from the coupled NorESM model. If the AMIP boundary conditions were taken from the coupled NorESM model instead, the difference between the coupled and uncoupled response and hence the 'role of ocean and sea-ice feedbacks' would presumably be substantially smaller (this could be tested). My guess is that the larger polar amplification in the coupled model is not because ocean and sea-ice feedback amplify the polar amplification, but simply because of the fact that the coupled NorESM has a larger polar amplification than the multi-model mean of the CMIP5 models (the boundary conditions of the AMIP model). This puts into question the authors' interpretation of results, in particular the importance of ocean and sea ice feedbacks in explaining the response difference between the different model versions. I don't think this is a show-stopper. Documenting the difference in response between AMIP, SO and fully coupled model versions is a usefull excersize, but I do have concerns regarding the attribution of this differences to coupled ocean and sea ice feedbacks.
2) While I appreciate the addition of the fully coupled simulations, the authors have not included a description of how the scenarios for these simulations were constructed (section 2.2). How was it determined that the combination of RCP2.6 forcings and the adjusted CO2 evolution would result in global mean temperature stabilization? What was the physical reasoning behind these choices? Sanderson et al. (2017) constructed the scenario using an emulator, Sigmond et al. (2018) established stabilization by switching off all anthropogenic emissions, but how did you determine the scenario?
3) I find the structure of sections 2-4 non-intuitive. I would make this one section with section 2.1 describing the model and section 2.2 describing the model versions.
4) There are still quite a number of typos, e.g. P. 1 , p. 25: increase --> increases, P. 1, p. 29: it-->is, p. 2:l. 10: is to presented --> is to be presented
5) l. 22: The combination of 'Compared to the AMIP runs' and 'relative to the present day climate' is confusing. Perhaps remove 'relative to the present day climate' ?
6) p. 2, l. 29, p. 6, l. 9 and p. 19 l. 14: An other relevant paper that should be cited here is doi: 10.1038/s41558-018-0124-y who performed 1.5C and 2.0C stabilized warming simulations with a coupled model, by switching off all anthropogenic emissions in a 'free-CO2' mode
7) p. 4, line 24: 'specific to our set-up': a bit confusing, this suggests that the points listed below this statement are specific to the NorESM model and hence differ from the standard HAPPI specifications, but I don't think that is meant by the authors
8) p. 6, l. 3: please also include the warming relative to the preindustrial run.
9) p. 6, l. 28: Here is should be noted why the authors did not use the AMIP-15 and AMIP-20 boundary conditions to calibrate the SO model. If they had chosen that, the difference between the AMIP and SO responses (and hence the assessment of the importance of atmosphere-ocean feedbacks) would be much smaller (see my comment #1). Also, on p. 10, l. 29-30 the authors state that 'The experiments with NorESM1-HappiSO are designed to be comparable to the NorESM1-HapppiAMIP experiments'. Based on this statement it would more sense to me to use the AMIP-PD, AMIP-15 and AMIP-20 fields as an input to the calibration.
10) p. 9, l. 28: 'SST is (in this connection) the mixed-later temperature: I suggest changing the notation to something like T_mix. Also, SST_ext should be defined here, and not later in the paper.
11). section 4.1: I think find this a bit confusing, it would make more sense if in p. 10 l. 16 alpha is not set to 0 (is this a typo?).
12) p. 11, l. 9: It may be useful for the reader to include here an explanation for why the AMIP forcing agents are used in the SO runs, and not those used in the coupled runs (with the adjusted CO2 concentrations).
13) p. 12, l. 5: I'm not sure how the smaller 1.5K-PD and 2.0K-PD differences can be explained by a cold bias in PD? Shouldn't the cold bias in the PD cancel out in the response? The 1.5K, 2.0K and PD are all expected to suffer from a cold bias, correct? (the same applies to p. 14, l. 1-3)
14) Fig. 18: The observational estimates (presumably shown by 'solid black contours'), are unclear. They are hardly visible and it is not clear what the contour levels are. It's probably better to show the obseverations in separate panesl. Also, there seems to be something wrong with the colors in panels e, f, i, j, with positive (>+2.5%) responses all the way down to the UK.
15) Section 8: The last 2 column of table 4 should be discussed here (not in section 5). What are the observed mean values of Sea ice extent/area, and how do they compare to the CLP-PD and SO-PD values? Fig. 19 suggest to me that the interannual variability in the coupled model is biased low.
16) p. 17, l. 3 and l. 11: In line 3 it is noted that the SO has too little sea ice. This suggest that the ice-free frequencies under warming would be overestimated, correct? If yes this should be noted/discussed in lines 12-14.