This is a useful modelling study with a finding that differences in ocean stratification due to TKE MLP parameters have significant implications for the Arctic freshwater content. For global NEMO modellers would be nice to know how these various simulations perform in other sea-ice covered seas, not just in the Arctic. Also, is there a specific configuration that is recommended and differs from default settings?

The text could be improved significantly. Its expressions are often confusing and the writing style could be clearer, in particular scientific terms and simulation names could be used in a more uniform manner, which would help reading a lot. I give some detailed comments to assist on this:

• line 58: Instead of writing 'several regions', be specific and write '3 regions'
• line 58: 'by varying the mixing scheme' is more clear, if that is meant here. 'varying the mixing just below the mixed layer TKE
  
   parameters' is hard to understand.
• line 61: Change to 'The vertical turbulent kinetic energy (TKE) closure scheme'
• line 63: Change to 'integrated by Blanke and Delecluse (1993) into the OPA model'
• line 75: 'ocean's surface boundary layer'
• line 84: Can f_r be > 0.1? If it can, would it make sense to try higher values?
• line 84: Do you mean that 5% of TKE is redistributed below MLD or in the MLD and below it?
• line 94: Explain what ORCA1 is.
• line 97: 'Rathore et al. (2024)' [year is missing]
• Table 1 and others: Table captions are usually above tables.
• line 114: Is pss Practical Salinity Scale? Does it equal psu which is more commonly used?
• line 120: 'de Boyer Montegut (2024)' [typos in the reference]
• line 125: 0.03 kg/m$^3$ [there should be space between number and unit, change also line 127]
• line 127: '(e.g. de Boyer Montegut et al. 2004)' [remove double parenthesis, change also line 130]
• Table 2: It is difficult to follow in text what simulation is what. Perhaps number simulations in this table from 1 to 9 and use
  
   those numbers consistently in text when refererring to particular simulations.
• line 137: 'We computed'
• line 140: 'The ITP data are from'
• Figure 2 caption: 'Basins following Peralta-Ferriz and Woodgate (2015).'
• lines 155-157: Seems that latitude-longitude boxes can not be used to define the regions drawn in Figure 2.
• lines 157-162: This text does not belong here but fits better in Introduction.
• line 172: 'deeper than modelled ML'
• lines 176-178: It is hard to see these salinity differences from Figure 4 which shows absolute salinities. A solution is to add 
  
  Control-LOPS salinity difference panels to Figure 4. 
• line 181: 'data are averaged spatially and temporally'
• line 189: Here is a place where indicating the parameter would be a reader: 'Increasing the fraction of surface TKE (rn_efr) that'
• line 190: Do you mean 0.75 rather than 0.08?
• lines 191-192: The effect of nn_tau is not shown in Figure 5, so this sentence in this context is a bit misleading.
• line 194: 'could arguably be caused'
• line 195: 'However, we recomputed'
• line 198: Is the MLD std modelled and/or regional, what are the summer months used? This information should be added.
• line 201: 'and is only 8 m'
• line 204: 'with MDL at least 20 m'
• line 207: '20 m' [add space between number and unit]
• line 207: 'in the Canada Basin and the Eurasian Basin in March'
• line 209: The spatial distribution looks similar in September too, not only in March.
• lines 215-216: The effect of deeper ML on surface salinity is mentioned. What about temperature?
• line 226: Could continue that the TKE MLP parameters do not improve the temperature maximum below 200 m because the temperature maximum is due to the heat advection at that depth.
• line 266: Note that there is also negative sea-ice thickenss bias in March.
• line 266: Is the East coast the Siberian coast?
• line 267: 'negative biases between'
• line 268: Discrepancies between OSI-SAF and model are not relatively minor in September but look quite substantial. What is the OSI-SAF uncertainty? Would make sense to use other satellite sea-ice concentration data for comparison. It is known that different data sets provide somewhat different sea-ice concentration and the ice edge.
• lines 295-296: The link between stratification and sea-ice melt is interesting. What would be the role of ice-albedo feedback strengthening this process?
• line 299: Would be reader friendly to remind what these three sensitivity experiments are, not just list their parameter values. Here Tables 1 and 2 could be used/ cited.
• Figure 11. Differences seem often not very clear and there can be biases both ways. Perhaps calculating RMS or mean absolute differences per basin would give clearer and more quantitative measure.
• Figure 12: Caption: 'MLD, sea-ice concentration and sea-ice thickness'. How much these timeseries correlate? Seems that year-to-ear variability is captured rather well due to the same atmospheric forcing and free drift in summer.
• line 317: 'is close to'
• line 325: What is short-term?
• line 332: 'exhibit closer resemblances'
• line 346: Can you link the MLD increase to the erosion of the Arctic halocline?
• line 370: 'in these regions than the 1-f_i option'
• line 371: 'the LOPS climatology'
• line 372: 'upper ocean vertical properties'
• line 398: 'their coverage'
• line 401: 'to entirely remove TKE MLP'
• line 405: 'this parameterization lacks a'
• lines 406-407: Do you have a reference for the claim of time dependence of the NEMO TKE MLP scheme?

The present manuscript focuses on improving the choices that go into a mixing scheme used in the NEMO ocean model. The authors have evaluated a vertical mixing parameterization based on turbulent kinetic energy (TKE) in the Arctic region. They have checked the mixed layer penetration effects modeled in the TKE based scheme. The MLD penetration effects account for additional mixing caused by near inertial effects and ocean swells. They find that it impacts the evolution of mixed layers below sea-ice and impacts the formation of sea-ice itself. They also checked the influence of adding information about the sea-ice concentration in the mixing scheme and it leads to trends in the mixed layer depths that agree with observations.



The authors convincingly argue about the choices of tunable parameters used in mixing parameterizations. Unconstrained turbulence mixing parameterizations lead to major sources of uncertainty in climate projections. It is vital to improve mixing parameterizations, especially those of upper ocean processes which regulate the upper ocean response to surface forcing. Because of the important nature of the work done in this manuscript, I recommend this article to be published in the journal ‘Geoscientific Model Development’.



I have a few minor comments / suggestions to improve the clarity of the manuscript:

1. Line 75: “The parameterization is activated in NEMO when the parameter nn_etau is set to 1 (and deactivated when nn_etau=0).”

This belongs in the NEMO documentation rather than in this article. If the authors want to document this, they can consider adding these kind of sentences to a supplementary material for additional documentation. 

2. Line 75: It would be helpful to write the evolution equation of TKE. It would help in understanding how the various terms such as N^2, velocity, and energy (e at time t) go into the evolution of e at time t+dt.

3. The equation written in line 75 is a bit confusing. On the left hand side, the authors have written e(t+dt,z), and on the right hand side is there a missing term:  e(t,z) ?

Should the equation be:

e(t+dt, z) = (( d e(t,z) / dt ) \times \delta t) + e(t,z) + e_inertial(t,z),

Where all the d are partial derivatives?

Please check the equation in line 75. 

4.  Table 1: This table is not necessary. You have already given a short summary around line 100.

5.  Line 125: You have already defined MLD (0.03) at line 115.

6.  Line 195:  Do you mean standard deviation of MLD?

7. Line 80: Using the symbol f_r should be enough for the manuscript. Describing the namelist parameter is not needed.  

8. Line 95: Please provide a reference for ORCA1. Is this a specific configuration or some different model is not clear.

9. Line 130: For table 2, you could provide an experiment number as an additional column before the ‘parameter’ column. In general, I find that writing down both: variables and namelist variables used in NEMO code to be confusing. The authors could consider using only symbols rather than namelist variables. All the symbols and their corresponding namelist variables could be added to a table in an appendix. 

10. Line 140: “The WOA23 dataset is available at the NOAA website.”

No need for this sentence. You can reference WOA in the earlier sentence.

11. Around lines 195-205: These sentences seem more suitable in the discussion.

12. Line 240: When you mention the Brunt-Vaisala frequency, please state to refer to results in Figure 8.

13. Line 370: ’Surprising’ seems like a strong word. Stratification is sensitive to OSBL mixing parameterization, so it is not that surprising. Just remove the word “surprising”.