Review on “Sensitivity of winter Arctic amplification in NorESM2” by Seland Graff et al. 2025, submitted to ESD

Main comments:

This manuscript explores the uncertainties in projections of Arctic amplification due to several key parameters that are known to be important contributors to uncertainties in climate projections. The structure and readability of the manuscript is very clear, and the description of the experiment set up is convincing. The only major concern I have is that the removal of a known bug influencing mixed-phase clouds is not really a sensitivity experiment. The bug should have been corrected, and another sensitivity experiment designed. In addition, the other sensitivity experiments should have been using the mixed-phase cloud bug fix.

While most of the parts are very clearly written and presented, the result section including some of the figures needs some clarification / correction as detailed in the minor comments.

Furthermore, there are several relevant and recent papers that have not been cited in the manuscript (for example Cai et al., 2025; Douville, 2023). There also is an older paper by Bracegirdle and Stephensen (2012) that is relevant. Methodologies and results should be compared and contrasted to the present study.

Minor comments:

Abstract, lines 14-16: Similar to results from Pithan et al.?              

Line 47: can the rectification of an error be declared as a model improvement (see also main comments)?

Lines 131/132: with the cloud physics changed, how can we know that we are still in a well-balanced state? Would there be a spin-up effect in the beginning of the historical cloud simulation?

Line 257: after the 60 years, is the model spun up?

Line 264: 3.7 or 7.0 W/m2?

Lines 284/285: Are these numbers for 30-year periods or for last year minus pre-industrial (after having applied the running mean?), or something else? Please specify.

Line 303: “more than exceeding”: duplication.

Lines 334-339: the cold bias is large compared to the warming signal, resulting in a situation where the difference future simulations minus PHC3.0 is smaller than the difference PHC3.0 minus historical simulations in many cases (as shown in Fig. S2). This merits a mention to allow the reader to judge the magnitude of the cold bias.

Line 380: “is know” -> “is known”

Lines 408/410: more Barents Sea than Kara Sea?           

Lines 418-420: The North Atlantic Warming Hole signature seems to be present in EOF2 of the NH ts while it is out of the area for the Arctic.

Line 486: “low-level cloud cover increases across all seasons”: can’t see this from Fig. 12b, and only partly from Fig. S7b, f, and j.

Line 531: staced -> stacked

Line 570: I’m surprised that it is only 14.4 K and not around 16 K plus minus some error bar. I would have expected around 16 if looking at the crossing of the red and the black dashed lines in Fig. 16c. Here, at least a brief description of the method to determine the emergent constraint is required.

Fig. 1: At first, I struggled to understand why there are coloured lines in panel a), and grey lines in panel c), given the different legends in panels a) and c). I think there would be space to have both legends in both panels. There are similar cases in other Figures as well.

Fig. 3h: not sure what to make out of a negative range (between 65 and 70 N at around 700 to 800 m depth). Maybe better to say “difference” instead of “range”, as it is SSP585 minus SSP126? Similar applies to Fig. 5h.

Figs. 5 and S4: in principle you are showing the sea ice thickness averaged over each model grid cell. Meter per unit area and calling this a volume is a bit confusing. Furthermore, it is difficult to judge the meaning of panel (a) without the simulated distribution of sea ice thickness in the historical period. Is practically all sea ice gone or is there still a substantial amount left during the last 30 years?

Fig. 6, caption, third last line: “… transition to and seasonal …” -> “… transition to a seasonal …”

Fig. 6, caption, last line: “Units are in mill km2”: only valid for panel a).

Fig. 7, caption: check references to the panels.

Fig. 9, caption, line 6: even the historical period is characterized by small trends. For computation of the variance, has a detrending been applied?

Fig. 14, caption, line 1: fluxe -> flux

Fig. S3: noteworthy AMOC strength peak in all simulations but hist/ssp585-piAerOxid. How about commenting on this?

Fig. S7, caption: August and May swapped

References:

Bracegirdle and Stephensen, 2012: On the Robustness of Emergent Constraints Used in Multimodel Climate Change Projections of Arctic Warming in: Journal of Climate Volume 26 Issue 2 (2012)

Cai et al., 2025: Lessened projections of Arctic warming and wetting after correcting for model errors in global warming and sea ice cover | Science Advances

Douville, 2023: Robust and perfectible constraints on human-induced Arctic amplification | Communications Earth & Environment

This manuscript clearly presents the experimental setup and effectively addresses key parameters contributing to uncertainties in climate projections. The methods are clearly described and outlined to the reader. In my opinion, the removal of a known bug affecting mixed-phase clouds is more of an update and not an independent experiment.

Enclosed below you can find some minor technical comments regarding the figures:



Fig. 5: Include units in the colorbar within the figure.

Fig. 7: Indicate lack of units in the colorbar with something like [-]

Fig. 13:  Include units in the colorbar within the figure.