Review of "An inter-comparison study on the impact of atmospheric boundary layer height on gigawatt-scale wind plant performance" by S. Ivanell et al.

This is a timely and ambitious paper that addresses the impact of atmospheric boundary layer (ABL) height and inversion thickness on the performance of large wind farms. The inter-comparison of five numerical solvers, spanning both LES and RANS approaches, is a valuable contribution to the field. The paper is well-structured, and the methodology is described in sufficient detail to allow reproduction. The authors are commended for their transparency in discussing model differences and limitations.

The paper has the potential to make a significant contribution to our understanding of ABL effects on wind farm performance. However, the authors should take care to distinguish between physical insight and model sensitivity, and to communicate the implications of their findings more clearly. With a major revision, the paper could be a valuable reference for both researchers and practitioners in wind energy.

The paper raises several concerns that should be addressed before publication. Most notably, the reported wind farm efficiencies (~60%) are significantly lower than typical assumptions (~85%) for similar conditions. While this may reflect a real and underappreciated physical effect, the lack of standardized inflow conditions and observational validation makes it difficult to separate physical insight from model artifacts. There is a risk that these results, if not carefully contextualized, could undermine confidence in high-fidelity modeling for wind resource assessment (WRA). Please include a discussion comparing your efficiency results with typical values used in WRA (e.g., 80–85%), and clarify whether the discrepancy is due to ABL effects or modeling setup. Consider adding a paragraph in the Discussion section explicitly addressing how these results should be interpreted by practitioners — i.e., not as a failure of LES/RANS, but as a call for better standardization and validation. Also, could this explain part of the overprediction bias in wind farm production? Similarly, the effect of blockage expressed as non-local efficiency appears to be much larger than even the most pessimistic estimates from the operational wind farm performance calculations. 

The absence of observational validation is a limitation. Even a qualitative comparison with met mast or LiDAR data would strengthen the conclusions. If such data are unavailable, please state this explicitly and suggest how future work could address this gap.

Several figures are placed far from their textual references, which makes the paper harder to follow. Consider reordering or duplicating key figures closer to where they are discussed.

The inflow profiles differ substantially between solvers, especially in the H150 case. While the authors acknowledge this, the implications for the efficiency metrics (η_w, η_nl, η_f) are not fully explored. Since these metrics assume identical inflow, the comparison across solvers is questionable. Could you quantify how much of the efficiency variation is attributable to inflow differences?

Another source of differences between the solvers seem to be how the wind turbines are represented. In Star-CCM+ for example, the forces are derived from the power and Ct curves, while they are calculated more explicitly in the EllipSys3D setup. Please comment. In addition, it would be important to know how the flow structure in terms of wind shear and wind veer may affect the power calculation in different solvers. For example, given a power curve, the wind turbine power is then usually calculated only using the hub-height wind speed, while the real power conversion also depends on the level of turbulence, wind profile, wind veer.

Specific Comments

L22 Stull 88 would be a better reference for the definition of a capping inversion.

L38 The GW discussion could profit from a calculation of the Froude number or an equivalent measure to determine the criticality of the flow over the WF. It would be for example interesting to see what happens in case when the inflow velocity is changed so that the flow at the WF is exactly critical.

L42 "adverse pressure gradient" needs some explanation

L64 It would be easier for the reader if this information was provided after Table 2.

L66 Please explain how the staggering was chosen. Is it to achieve minimal wake loss for the 270 degree wind?

L73, Figure 1. If possible, please simulate the 315 degree flow (or rotate the WF by 45 degrees) and analyze how that affects the non-local efficiency (blockage).

L195 "horizontal driving pressure gradient" is a sensitive subject in context of atmospheric CFD modelling attempts (in parallel with the Coriolis force), and requires more explanation in the paper. Has the flow achieved geostrophic balance during the precursor simulations (in all of the solvers) and the pressure gradient is then manipulated/adjusted during the production run to keep the winds aligned? Including a diagram of the model setup including the imposed pressure gradient would be best.

L232, Figure 2. The U and V velocity components look OK and indicate westerly wind (from 270 degrees) at the hub height, and the wind above the BL has a negative V component and is from WNW, which is correct (on the N hemisphere). But the wind direction plot, (d), shows 0 at hub height and negative direction change above the BL which would mean WSW. Please resolve this inconsistency and/or correct the plot accordingly. 

Figure 8. The horizontal wind velocity for the H500 case differs prominently from the H150 case (Figure 7). Most notably, the horizontal flow divergence is much stronger, and wavefronts are visible, indicating that the flow is supercritical. Please calculate or estimate the flow regime.

Figure 13. Claiming that the non-local (blockage) effects account for more than 20% power loss already in the first row of turbines, even in the more optimistic H500 case, is bold and significantly deviates from traditional assumptions, and will raise eyebrows. In the worse case this may lead to loss of confidence in the concerned modelling approach. Please see the general comment as well, and it would be recommended to present this aspect of the paper carefully.

L298 The transparency is appreciated, i.e. the purpose of the study may be to only illustrate the impact of the BL height, but the numeric results open important questions and call for revision of the traditional wind farm performance evaluations, and it may be questioned if just being open about the partial intent is sufficient for qualifying this paper for publication.

L309 Are we sure that the waves in Figure 18 are non-physical? Internal GW may propagate in a stratified fluid like this.

L319 Conclusions do not mention the huge contributions of internal and blockage effects to wind farm efficiency, which deserve some discussion.

Appendix A. U_d and gamma are not introduced. Please introduce them.

Comments on the manuscript entitled “An inter-comparison study on the impact of atmospheric boundary layer height on gigawatt-scale wind plant performance” by Ivanell et al. submitted Wind Energy Science journal.

In this work, the authors investigated the impacts of atmospheric boundary layer (ABL) height and capping inversion layer thickness on wind plant performance. It is an important topic considering the continuing growth of wind plant sizes. One strong point of the work lies on the inclusions of results from different models and different codes. Specific comments are as follows:

1. The work focus on the power output and mean streamwise velocity. As the momentum entrainment in the vertical direction plays an important role on the wake flow recovery, and therefore the wind farm performance. It is necessary to examine turbulence statistics around the top tip of the rotor for different ABL heights and capping inversion layer thicknesses.
2. There are several concerns regrading the conclusion, i.e., “the sensitivity of using different levels of modeling fidelity and numerical approaches overall is limited since the result generally show good agreement.”, drawn in section 5: (1) the overall good agreement among the several codes employed in this work may not be enough to conclude that the sensitivity is limited; (2) moreover, works in the literature have shown that recovery rates of wake flows depend on turbulence models and wind turbine models employed; and (3) it needs to be clear about for which quantities, the sensitivity is limited.
3. Differences among different codes are larger for the H150 case and the H500 case in comparison with the H500-dh500 case (figures 6, 13-15). It is suggested to discuss the underlying reasons.  
4. It is necessary to use the same naming convention for legends of different codes. For instance, “DNV-RANS, DTU-LES, ...”are employed in figure 12, while “NDV, DTU, ...” are employed in figure 13. It is suggested to use those in figure 12.
5. For the caption of each figure, make it self-explained. For instance, it is suggested to add the descriptions for each subfigure in figure 2.
6. A typo on line 258: “caes”