General comments:

The authors investigate the ice accretion on a wind turbine. The article seems original because it allows to quantify the effect of the ice surface roughness on the performance losses of the wind turbine. For this, the authors use numerical simulations. Being well aware of the strong uncertainties on the roughness input data, they performed an interesting parameterization. Moreover, they have performed quite fine simulations of the ice accretion on each studied section by a time-efficient multi-step approach. 

Specific comments:

- Introduction : line 101, "The icing event was long enough for ice horns to form, to combine the effects". The author are supposed to address rime-ice conditions from an earlier comment. The term "horn" is more often used for glaze-ice shapes. 

- Methodology : 

  * line 129, what is the "wind shear exponent"?

  * line 131, what does the OpenFAST simulation imply for the ice accretion simulation? For instance, do the wind turbine operation data account for the retroaction of the ice shape growth? (rotational velocity, etc.)

  * line 146, it seems to me a good idea to define an average power value. But why use the Weibull distribution rather than another one? 

  * line 206, this is not clear to me how and why this extrapolation is performed. If it is common practices, is there any reference available?

  * line 209, "the average flow field is resolved down to the Kolmogorov length scale." This seems misleading to me. This looks more like the definition of DNS simulations. The low-Re approach is related to the description of the turbulent boundary layer structure and requires y+=1 to capture the region of the viscous sublayer.

 * line 236, is there any reference for uhMesh? what kind of mesh generation technique is used?

 * line 247, "the output had a 1P component". What does that mean?

- Validation:

 * line 273, since the description of the setup is diluted over several sections, it is not fully clear to me what experimental conditions are simulated in section 3.1.

 * line 298, since the residual seems to be of importance for the methodology, it would be worth describing exactly how it is computed.

- Results and discussion:

 * line 327, does the roughness always cover the whole ice surface (in the std and ext case)? On the contrary, can it cover the blade surface further than the ice?

 * line 338, "the ice shape was mainly responsible for the aerodynamic penalty". It would be interesting to know the polar for the smooth-wall simulation of the iced shape to support this assertion.

 * line 343, "due to the supposed early transition", I do not understand this early transition. Is the transition modeled for the rough-wall simulations? If not, wouldn't it be fairer to compare against the clean simulations without transition?

 * line 356, "This effect is peculiar since roughness should have little effect on the aerodynamic coefficients when ie horns are well developed." Is there an explanation?

 * line 399, "Once more, our results agree with those by Etemaddar et al.", in which sense do the results agree? They may be consistent with each other but they are not in agreement (except if the figures in the table are wrong).

 

 

Technical corrections:

- Lines 62 and 66: 20 microns, 25 microns

- Figure 10, page 14: why are there systematically 2 curves for "Clean" (and the slope is not recovered)?

- Line 370, define what TSR and Cp are

- Reference Lavoie et al (line 494): The journal article https://doi.org/10.2514/1.C036492 is probably more accessible to most readers

- Reference McClain et al (line 498): The journal article  https://doi.org/10.4271/2019-01-1993 may also be more accessible to most readers than the conference article referenced. However, this is not exactly the same topic (although the main information that roughness evolves both in space and time is also given).

This paper investigated the effects of roughness on airfoil performances under ice conditions. In this paper, NERL 5MW model was applied to perform numerical simulations. To analyse the power performance of the iced blade, DLC1.1 was considered with two different roughness cases, W_std and S_std. Detailed comments are addressed below.

Comments

• IEA task 19 published a report about available technologies for wind energy in cold climates where detailed IPS cases were reported. This report should be reviewed.
• Recently, there were studies to investigate surface roughness effects for simulating ice accretion. These papers should be reviewed. (https://arc.aiaa.org/doi/10.2514/1.J060641, https://arc.aiaa.org/doi/10.2514/1.J059222)
• In section 2.4 the authors introduced the extended roughness area: 25%, 18%, 15%, 13%, and 11% along the blade span shown in Fig. 3, respectively. I think it is one of the most important parts of this paper. But there is no detailed description of how these values were introduced. It must be clearly described.
• In Figs. 11 and 12 two different comparison studies were presented. It was shown that the current numerical results were not able to accurately predict the ice shape compared to the experimental test results. The author mentioned that “the ice impingement limit on the lower surface was underestimated”. It might be due to that impinged water does not freeze at the surface and exists as a water film. Therefore, it might be good to check the heat transfer rate.
• In Fig. 13, how do the authors ensure the predicted ice shapes are correct? Moreover, the ice shape at the blade tip areas (section A-C) seems very irregular horn shapes. Is it obtained under the rime ice condition? In general, more validation studies are required to prove the current numerical model's accuracy.
• On page 20, section 4.3, why is after the optimum TSR value interested?
• Figure 23 shows the power curve under turbulent wind conditions. Why clean airfoil case where no ice is accumulated could not obtain the rated power, 5MW, at the rated wind speed? Since there is no ice accumulated with the clean airfoil model, it should produce the rated power at the rated wind speed.

Overall, this paper needs more validation studies to prove that the current model is valid for 3D wind turbine rotor simulations. Furthermore, roughness model validations are required. Many studies have already investigated the power performance of a wind turbine with and without ice accumulations. Therefore, there is no novelty in evaluating the power performance with a CFD tool. One of the most exciting parts of this paper is considering the surface roughness effects. However, it is not clear how different surface roughness was considered and implemented into the simulations. Based on the aforementioned comments, this reviewer recommends rejecting this paper.