Review of Arctic Ocean Simulations in the CMIP6 Ocean Model Intercomparison Project (OMIP)

This paper analyses 19 ocean-sea ice models participating in the Ocean Model Intercomparison Project (OMIP) as part of CMIP6 and evaluates their performance in simulating Arctic Ocean properties. The simulations are evaluated mainly in terms of their mean climatological state, and to a lesser extent with respect to temporal variability.

The authors report that no significant improvements were made since the previous CORE II model simulations when it comes to simulating mean Arctic Ocean water mass and circulation properties. This is the main finding of the paper, with most models showing large biases in simulated mean hydrography.

The paper is nicely structured, well written and presents a valuable contribution to the modelling community. With some modifications it has potential to help guide future model development needed to improve ocean-sea ice models. Overall, I am pleased with the paper as is, but only have minor comments and suggestions. These are listed below along with some specific “in-text” comments. 

General Comments

Limited interpretation of the results  

The results are presented in a very “straightforward” manner which I presume is typical for GMD, but there is little interpretation and discussion of these results and various biases. I wish the authors would go beyond merely stating what the models are showing but interpret those results and put them into a larger context in the final discussion/conclusion section. For example, can the authors comment or speculate why there has been no improvements since CMIP5? 

Furthermore, a paper like this offers an opportunity to reflect on the direction for the ocean modelling community going forward. Which model biases deserve the most attention? How do we fix them? Will increasing model resolution fix all these issues or do we need a different strategy? I would appreciate it if the authors could give their expert thoughts on this as I think it would be extremely useful to the community at large. 

No evaluation of horizontal circulation 

The paper focusses primarily on the model’s capability of simulating Arctic Ocean hydrography and heat, volume and freshwater exchanges. Is there a particular reason you do not evaluate the horizontal circulation? I suspect this could help in understanding some of the biases in water mass properties and exchanges across the different sections.

MMM versus individual model performance 

The paper focuses mostly on the MMM state, which I think is fine. But it would be nice if you could pull out specific models more. Why are some models performing well in certain cases? Why are some doing poorly? This could be an opportunity to learn what works and what does not. Specifically, I would like to see a more detailed discussion on the effects of model resolution (both horizontal and vertical) and choice of vertical coordinate. 

Specific Comments 

L64: “some systematic biases … have been identified”. Could you list some examples?

L82: Consider making a new Methods section here.

L121: It would be helpful to the reader if you mention the typical resolution in the other models for comparison. 

L123-124: Can you please clarify why 400 m depth is chosen specifically? I guess, this is the depth of the AW layer, but this should be said explicitly. But I am wondering why you do not calculate and show the Ocean Heat Content for the AW layer? I think this could be a particularly useful diagnostic (for example for people interested in sea ice) and is easier to connect to the changes in ocean heat transport later. It would capture both biases related to the larger vertical extent of the AW layer and the temperature.

L164: It would be nice if you could motivate why the liquid freshwater content is an important metric to look at. Just one sentence at the start of the section. 

L171-172: Can you comment on why OMIP-2 has more freshwater content in the Beaufort Sea compared to OMIP-1? Is there an improvement from CMIP5? Also, how important is sea ice for the freshwater biases in the models?

L198: AW warming over the whole Arctic basin?

L207: I found the part about the re-initialization a bit difficult to follow. Could you describe this in a bit more detail? Perhaps in the Methods section. And can you quantify the impact of the re-initialization compared to the natural variability in the models?

L229-230: Consider reiterating why the halocline layer is important. For example, by insulating the AW from the sea ice. 

L224: I would be interested in seeing the temporal changes in mixed layer depth. Do some of the models simulate episodic deep convection in the Arctic basin and what impacts could this have? 

L245: It would be nice if you could discuss the broader implications of the simulated biases in halocline depth. Why is it important?

L265: Are the trends negative since 1990 for all the OMIP model simulations? Also, can you please comment of the importance of horizontal resolution on simulating transport through the BS. Do models with higher resolution better capture the gateway?

L268: “historical observations” - please clarify over which time period specifically. 

L360: Here I really miss a more detailed discussion on why there have been no major improvements in hydrography since CMIP5. It is a quite powerful statement, so it deserves more reflection. See also my general comment. 

L392-397: I am glad to see that you discuss the effect of resolution. However, this should be expanded upon in more detail. For example, how does resolution affect the model’s capability in simulating the volume, heat and freshwater transport through narrow straits? And what about the effects of vertical resolution?

L403-405: This last paragraph was not so clear. Consider reformulating. 

Figure comments:

For all figures: I would suggest putting the name of the plotted variable in the colorbar legend (not only the caption) and with units. This makes it easier to quickly see what the figure is showing without having to read the caption first.

Fig 3: Consider also showing the temperature and salinity anomalies with respect to PHC3.0. Also, can you please comment on the absence of very cold and fresh water south of the Fram Strait seen in PHC3.0. 

Fig 4: Label name and unit on colorbar 

Fig 7: Label name on colorbar

Fig 9: Add unit and name. It is difficult to see the lower values (blues) in the upper ocean. It would also be nice to show the time series of AW ocean heat content here. 

Fig 10: It is difficult to see the values >100 m. Can you extend the upper limit of the color scale so to better see MLD in the Barents Sea. Also, please clarify if it is the mean or max over the cold season? 

Table 1: It would be useful if you could list the type of vertical coordinate here too. The grid number for AWI-CM-1-1-LR is 126859 x 46 (x, y, z). Is this a typo? 

Table 2-7: Would it be possible to somehow highlight which models perform better relative to observations? Also, MMM in bold font, or double line before final row.