The chosen validation data were at a relatively high induction larger than 1/3. For induction factors lower than 0.3 the results when using uncorrected 2-D airfoil data normally give very good results compared to measurements. It would be nice if at least one case with a lower induction was included. The not perfect match for the loads shown in Figure 3 could be due to the 3-D correction of the airfoil data or for the BEM codes and the empirical Glauert correction. In Figure 11 some synthesized airfoil data are shown that indicate a quite different stall behavior than the prescribed ones used in the airfoil data dependent codes. Since the inflow angle and chordwise pressure distributions were measured at a few spanwise sections in the Danaero experiment it could be nice to see how this fit with the prescribed airfoil data.

What is exactly meant on page 13 with the sentence, “these compressible solvers reveal a different suction level causing the integral loads to improve for the fine mesh”

It is nice to see how the fully blade resolved codes give quite similar results 😊

In Figure 6 it would be nice to know what tip loss model was used for the BEM and perhaps also a discussion on how the decambering effect can affect the way that loads decrease when approaching the blade tip.

The paper use the Danaero data, but suddenly in the conclusion the New Mexico data are mentioned and the challenges to reproduce those. A more elaborate discussion of this is missing.

The paper is well written, but the quality of the Figures should be improved, since it is not always so easy to see the details. Also are the airfoil sections in Figure 4 a and b upside down, meaning that the suction side is the lower one ? If, yes, then it is inconsistent with the pressure plots.

Indeed, this is a paper, so to say “long-due”. It clarifies several issues regarding aerodynamic modeling of wind turbines and its level of confidence. To my opinion the paper provides valuable information in a concise way and I would strongly recommend its publication not exactly in its present form. There are a few points that to my opinion require some attention:

• In terms of modeling, it is mentioned that most CFD simulations were done in fully turbulent mode which leaves some doubt regarding the results. Please clarify this point
• To my opinion, the part that concerns the choice of data to be used in models that rely on look up tables (and these all except fully resolved CFD) is important. It was also the conclusion of the AVATAR project, that when non-CFD modelling is compared to CFD, the polars must be obtained from the CFD. In the present case the improvement was substantial, to a level that would suggest to specifically include it in the conclusions and why not in the abstract.
• With regard to Fig 2, I would recommend to add a figure comparing the BEM and FV as groups in the way done in Fig 6.
• In Fig 4 I noted that apparently there is good agreement regarding the placement of the stagnation point (this is not the case in the yawed cases). To my opinion this indicates that the flows produced by FV and CFD are similar which also seen in Fig 6 (it is said that FV and CFD agree better in terms of loads).
• The comment regarding the “poorer” grid independency trend compressible codes have in comparison to the incompressible ones, was also a conclusion drawn in AVATAR and was attributed to the pre-conditioning need in the compressible codes in low Ma conditions. The slope of convergence depended on the type of pre-conditioning
• Are all simulations in Fig 9 and 10 rigid or only the CFD ones? How do CFD and BEM results obtained with synthetized polars compare? Perhaps instead of adding the original BEM results to add the CFD ones in Fig 11.  
• Although without any doubt CFD reproduces reality well, perhaps a point could be made as regards the quality over the inner region of the blade (in Fig 7 and 8 at the two inner sections the prediction of stall is not as good as in the axial case).