The paper presents a three-year WRF mesoscale simulation study of a Belgian–Dutch offshore wind-farm cluster, with and without the Fitch wind-farm parameterization. The simulations are validated using three offshore lidars, and model performance is evaluated for different stability conditions. The main question the paper addresses is the impact of the wind-farm parameterization on the results. The authors find that including the wind-farm parameterization in the mesoscale simulations improves accuracy in both intra-farm and downstream wakes.

Overall, this is a well-designed study that uses quantitative metrics to assess performance across different boundary-layer stabilities. The addition of SAR data alongside the lidars is a clear strength. I have a few general comments and questions I would like the authors to address before I can recommend this work for publication.

General Comments

1. Is there a specific reason for focusing on only one height when the lidars provide velocities at multiple heights? Showing at least one representative vertical profile (e.g., ) would help illustrate the vertical structure of wakes and blockage.
2. Please comment on the criteria used to select the specific dates/times for the SAR analysis. Were they chosen based on visible wakes, wind direction, stability, or data availability? A short description of the selection process and potential selection bias would be helpful.
3. Is the surface roughness used in the SAR-based wind retrieval spatially variable in your analysis? How is this variability calculated or prescribed? A brief explanation of the roughness treatment should be added. In addition, how many SAR snapshots were analyzed in total, and what fraction of them showed clearly visible wakes?
4. Is it possible to quantify the mean wake length for different stability classes (e.g., using a threshold deficit to define “wake end”)? Even a simple diagnostic would strengthen the physical interpretation.
5. Please discuss the limitations of using the Obukhov length diagnosed by the PBL scheme, particularly given that Monin–Obukhov similarity is known to break down in strongly stable conditions. How might this affect your stability classification and related conclusions?

Specific Comments

1. Instead of referring only to the lidars by their site names, it may improve readability to also refer to them by their functional roles (e.g., “upwind/free-stream,” “intra-farm,” and “far-wake” or “downstream”) throughout the text.
2. Providing the WRF namelist file (at least for the innermost domain) as supplementary material would significantly help reproducibility.
3. Figure 7 uses “r” for the correlation coefficient, whereas Figures 6 and 8 use “ρ.” Please standardize the notation across figures for consistency.