General:

The submitted article “Modulating surface heat flux through sea ice leads improves Arctic sea ice simulations in the coupled EC-Earth3” by T. Tian, R. Davy, L. Ponsoni, S. Yang, provides interesting results on the effect of surface heat flux through leads on Arctic climate representation in one global climate model. It highlights the importance of representing the effect of small-scale processes.

The study is in most parts well written. The figures are nice and easy to understand. However, some of the conclusions from this study should be better supported by additional analysis or better explanation. 

I thus recommend to accept this submission after revision that considers a few major and a somewhat larger number of minor comments.

Main comments

1. The authors highlight that the modulation is not only affecting Arctic sea ice but also improving the sea ice representation in EC-Earth3. They set up three improvement goals in the introduction (reduced sea ice, better trend, and Arctic ice minimum in September instead of August). These goals are only partly met. Sea ice is slightly less extended and slightly thinner in the improved model version when averaging over 1980-2014. However, August is still the month with lowest sea ice extent and the trend is not improved. The heat flux modulations seem further to have little impact in a present day climate but is more pronounced in a colder climate, and would likely have rather little impact in a warming climate. 

Further, the authors do not show that the modulation of the heat flux is improving the heat flux in high-ice covered areas or atmospheric stratification in EC-Earth. This should be done to the extent possible. Otherwise, any potential improvement of the sea ice could be due to a compensation of errors. We should have in mind that many other processes including large scale ocean and atmospheric circulation are strongly affecting sea ice. 

As long as it is not well shown that the modulation of heat fluxes is really improving the related local processes, this study, which provides interesting and relevant results, should be seen as a sensitivity study to understand the impact of modulating the heat flux over leads, and not try to sell it as an improvement that solves long-standing biases in global models.

2. The cold and warm climate simulations are very short (section 2.2). Although Figures S2 / S3 indicate that the 30-year period chosen for analysis seem to be rather stable for sea ice area, volume and global air temperature, a 20-year spin-up is very short. I would expect, and earlier studies showed this, that climate warms (depending on the model by maybe around 0.2 - 0.6 degree C) after initializing from a transient run and repeating the forcing from that year before stabilizing. EC-Earth3 seems not to show such a warming after initialization from the transient run, or the warming is very small and short in time, or it is not visible due to internal variability and the shortness of the simulations. It should be checked which of these alternatives is true.

A 30-year period is also short for comparison between the cold and warm climate, particularly given that EC-Earth3 shows huge internal variability on centennial scales. Maybe, the somewhat surprisingly much larger effect of the modulating heat flux in ECE3L in the cold climate compared to the transient run can partly be explained by internal variability. I suggest to make the cold and warm simulations in total at least 100-year long to get more robust results.

3. While it is very nice that the ensemble of historical simulations is large with 20 members, the method to initialize the ensemble of historical simulations from only two historical members of EC-Earth3 in 1960 might lead to an underestimation of the spread in the ECE3L ensembles compared to the original ECE3 ensemble. The large spread across CMIP6 ECE3 members due to long-term large internal variability (AMOC) might not sufficiently be captured by this method based on 2 members only. It would be good to see where these 2 members that are used to create the ECE3L ensemble are placed in the cloud of the original 25 ECE3-members (e.g. AMOC, global mean temperature, Arctic ice volume). 

Minor comments:

1. Title: Since the improvement (reduction of sea ice) seems to be limited to the earlier part of the historical period that is analysed here, and there is little evidence that modulating the heat flux improves present day sea ice, I suggest to change the title. Maybe something like: “Impact of modulating surface heat flux through sea ice leads on Arctic sea ice in EC-Earth3.” 

Please delete the “coupled” before EC-Earth3. Either write the “global coupled climate model EC-Earth3” or only “EC-Earth3”. To my understanding EC-Earth3 is as default coupled.

 

2. Line 4: without reading the entire article it is impossible to know, what “cold” and “warm” climate refers to. “cold” sounds like PI or even colder and “warm” like some time in the future.  I recommend to say something like: …one pair using 1985-forcing (cold climate) and the other 2015-forcing (warm climate). 

3. L7: “two CMIP6 historical ensembles”. Sounds like an ensemble of historical simulations from different CMIP6-models. Make clear that you performed an historical ensemble with ECE3L and compare it to an ECE3 ensemble. 

4. L8: It is unclear why ECE3L is in () . I guess what you mean is that both the ECE3 and ECE3L ensemble means closely resemble the mean states in the respective cold-ensemble. 

5. L10/11: It does not seem very logical that the mean climate states are similar in the cold climate and the transient run, but the effect of the modulated heat flux so much smaller in the transient run. If the mean sea ice state of the transient run is similar to the cold climate, the argumentation that diminishing ice leads to a smaller impact of modulating the fluxes in the transient run does not make sense. In fact, your figures seem to show that sea ice is even a bit thicker in the transient run than in the “cold” climate.  

6. L11/12: I disagree to the statement that the local amplification rate in ECE3L is significantly improved compared to ECE3. Maybe in parts of the Greenland and Labrador Seas but in contrast ECE3 is clearly better in the Barents Sea. Please explain clearer (in section 5.1) why you would judge that this is a significant improvement or delete this sentence.    

7. L27: “most climate CMIP6 models struggle to reproduce the rapid decline since the mid-2000s” – this suggests that most CMIP6 models underestimate the observed trend but is this really true? The CMIP6-model mean (e.g. Notz and Community) slightly underestimates the trend between 2000 and 2014 but if we would consider the observed ice area until 2023 and compare versus e.g. hist+ssp2-45, the CMIP6 model ensemble mean is well representing the trend. Differences among models are large (Keen et al. 2021) but is Keen et al. 2021 really stating that CMIP6 models generally are underestimating the trend since mid-2000s?

Further, observations show a rapid decline until 2012 but no further decline thereafter, thus it “rapid decline since the mid-2000s” is not entirely correct from todays (year 2024) view.

Please modify this statement.

8. L 52/53: This is an important question that is phrased here but I do not find any answer to it in conclusion or abstract of this article. To me it seems that your conclusion would be “no”. The cold climate shows less ice/ is warmer with heat flux modulation but the warm climate shows only little change, thus the delta sea ice between warm and cold states is actually smaller in the ECE3L than the ECE3 runs. Surprisingly, this is not really reflected in the amplification rate or the trends. 

Please take up the question again later, e.g. in the conclusions.

9. L87: “does not supply additional heat to warm the atmosphere”. Please explain:

Do you mean the amplification of heat flux in winter is compensated by below-1-values in summer, and in the annual mean there is no additional heat flux to the atmosphere? And why should not the heat flux into the atmosphere be increased in a coupled model if it would be more realistic?

10. L 110-115: “These simulations were a subset of a 25-member ensemble.” How many EC-Earth3 members are you using: 25 or 20 or what do you mean with “subset”? 

11. L118: How do you perturbed the atmospheric initial conditions? 

12. L177:  “..in EXPWarm during the 50-year cycle (as revealed in the full time series in Fig S2)” . Figure S2 does not show the full 50-year time series but only year 21-50, please correct. As stated before, 100-year runs would provide more robust results, and please show really the entire time series as it would be very interesting to see how quickly the sea ice changes evolve in ECE3L.

13. L194/195: Do you want to indicate that ocean heat transport from the south into the Arctic increases in ECE3L, and that the increased heat is decreasing more sea ice? In case this is the hypothesis, please show ocean heat transports into the Arctic. However, I believe that it is more likely that sea ice is generally getting thinner. At the ice margins, ice is getting so thin, that it drastically drops while in the Central Arctic sea ice concentration will still be close to 100% in winter even if ice thickness is reduced by 1 m or more. 

14. L 230: Model spread is smaller in ECE3L than ECE3 -  this might be an artifact of the initialization methodology; see main comment 3.

15. L235: “can refine estimates of the declining sea ice trend”. Yes, but it would reduce the trend since the effect of heat flux modulation is larger in a colder climate. Is not that in contrast to one of your goals to make the trend in CMIP6-models or specifically in EC-Earth3 more realistic (= larger) ? (Although as stated before I disagree that CMIP6 models generally underestimate the observed sea ice trends).

16. L239-245: Linked to minor comment 5: Do you have any explanation why the impact of modulating heat flux in the transient period is so much smaller than in the cold climate, although the cold climate has even slightly thinner ice than the transient run?

17. L270: “has significantly improved its accuracy …”. But the improvement seems to be time-dependent and mainly for a (past) colder climate. If I follow your argumentation, then you would not expect large impact in a warmer future climate as well or would you? Any improvement of models is good but how relevant is this improvement to “improve predictions of future Arctic conditions” (Line 275) then?

18. L291: “The temperature trend maps demonstrate that ECE3L outperforms ECE3”: Linked to comment 6, I disagree: both ECE3L and ECE3 show trend patterns that are quite different from the reference data sets. There are maybe a few small areas where ECE3L fits a little bit better but there are other areas where it is worse. And if we look at the local amplification maps, it is even more difficult to see a clear improvement. We need also have in mind that we compare ensemble means to one realization of the reality.   

19. L365/ 366: “In a warmer climate, the modulating factor can either increase or decrease sea ice states …” Would it not be most probable that the effect would be small in a future warmer Arctic if we would follow your argumentation from the section before? We know from future climate simulations in the Arctic that ice thickness will further decrease and atmospheric stability decrease. What would you expect, given this knowledge, how the modulating factor would affect sea ice states in a warmer climate, increase, decrease or little effect?

20. L370: Please explain (or modify or delete this statement) how a parameterization in a model can “inform effective strategies for mitigating the impacts of climate change”? 

 

21. L375: in code and data availability: what about code and data of ECE3L?

Technical corrections:

L71: delete “particularly”

L74/75: write “sea ice in the Central Arctic” or just “it in the Central Arctic” instead of only “Central Arctic”.

Caption Figure 1: “Seasonal air temperature”: I think you show “monthly air temperature” or maybe the “annual cycle”.

Caption Fig S4 and S5: “under a constant”: “climate” missing? ; “as in Fig. ??”: replace “??”

L395: It seems to be strange to write into the reference that “The Blue Action project receives …”. Consider deleting. 

General: The manuscript aims at improving the Arctic sea ice state by introducing a modulated factor for sensible heat flux (based on Davy and Gao, 2019) depending on ice lead characteristics and the atmospheric boundary stability (enhancing in winter stable conditions and reducing it in summer unstable conditions). The scheme is shown to be effective during cold climate, although its impact seems to be less in normal or warm climate. The paper is well written. Unfortunately, some important details are missing and prevent a full acceptance of the manuscript at this stage. The heat exchange is not clearly established in view of the modulating factor and the cold and warm experiments are not clearly defined.

Major comments:

L87: “We emphasise that in ECE3L, the ocean does not supply additional heat to warm the atmosphere” (z for emphasize?) I am still scratching my head on this. However, no substantial addition is made to clarify the statement. This manuscript clearly states that the factor is applied to surface sensible heat flux: (L77-78) “we introduced a factor Alead to the surface sensible heat flux (SSHF) within the coupled ECE3 framework, to better represent the heat exchange through leads in sea ice ». Does it mean that the heat exchange between air and ocean-ice is non conservative? According to Davy and Gao (2019), only the heat flux over open water should be amplified but I was not able to ascertain the exact method used in the manuscript. Also, why stopping at sensible heat? Latent heat should be also pretty high over ice leads. Some results from Davy and Gao (2019) could be added in appendix since theirs is a project report to their funding agency (i.e., not clear whether it was peer-reviewed). Moreover, the details could be moved to the main text rather than the appendix since it is the core of the paper.

L99 states that “The [cold and warm climate] simulations used constant forcing with a repeating seasonal cycle corresponding to the respective climate states ». However, I am still scratching my head on how you do this. In a coupled model, you have only variations is solar radiation at the top of the atmosphere and aerosol forcing. Nudging perhaps of one of the components? Please give more detail.

Methodology: Some figures show the ensemble envelop (5-95% percentile), which is interesting since it gives us the statistical significance of the results. However, except for ice volume, they seem to be hardly significant (i.e., an overlap is visible), which needs to be stated in the text.

L194-195 states that “The findings suggest that during Arctic winters, the decrease in sea ice concentration is mainly due to heat advected by the ocean from the south, rather than air-sea heat exchange through sea ice leads ». I think I see the same ice reduction in S6 for concentration below 70% but I am not sure I understand the statement any better. Since the modulation factor is one over these regions, it must a non-local effect as mentioned by the authors. However, since we are dealing with a coupled system, it could be the ocean or the atmosphere. I am less inclined to think it is the ocean, as the authors do, since you would need to explain a change in the Arctic ability to pump more Atlantic waters northward, than a simple advection/diffusion of the warmer atmospheric boundary layer southward. So, please elaborate.

Minor comments:

L56: Essential Climate Variables : why the capitalization here?

L248: what was the concentration threshold value used to define the ice edge? Please add.

L1 of Fig1 caption: please remove “compared”

L2 of Fig1 caption: “based on” I think you meant “relative to”, i.e., the bars and maps are the difference between ECE3L and ECE3 base runs (?) for the two climate experiments.

Fig2: the volume does show statistical differences but not the area (at least not clearly). To be mentioned in the text.

Fig.7: can the authors add the spread? I am worried that the significance is less than visually shown.

Fig.9: The plots do not show a statistically different mean (the two envelops overlap).

L1 of Fig.10: ECEL should be ECE3

Fig.11: the solid circle appearing in the key of the map is misleading. Its interior should be lightly colored as the area covered.

Fig.13 and L291-295: Despite the authors’ vigorous statement that ECE3L Arctic temperature trend is better than ECE3, both modelled trends are still quite far from the observed one. So, we are still missing out on the amplification factor in climate simulations. I realize that it is touched upon in the previous paragraph L287-290 discussing S10 but I must say that the S10a was not showing clearly this underestimation (but it is clearer in Fig.13), unless we are talking about something subtle hidden behind the spread of the ensemble? Why would there be a compensation between a global overestimation and the Arctic underestimation? Can you elaborate on this please, and possible in view of what the authors intent with the supplement (see below)? 

Appendix:

L480: units are missing from a1 and a2.

L481: is it not too large for winter Arctic Atmospheric Boundary Layer? Maybe add the expected and reasonable values for comparison

Supplement figures: What is the goal of them: do the authors intent to keep them there (please revise the captions then with attention), or ultimately discard them?