Summary

This is a well written paper on an important topic. The paper examines the spatial and temporal variation of so-called mesoscale winds over an area with high off-shore wind energy potential. The authors show that although considering the mesoscale part of the wind speed variability doesn’t change the annual power output significantly, it introduces variability into the power output. This variability is shown to peak in summer, and to have a more pronounced diurnal cycle with a daily peak at around 5pm. It is also shown to peak around coastal areas, although the authors do discuss the fact that model spin-up away from the boundaries may influence this result.

General comments

1. I really like the attempt to separate the ‘mesoscale’ and ‘synoptic-scale’ winds. However, I’m not really sure what physical processes are being picked up by this approach. The ’mesoscale wind’, is designed to align with the effective resolution of the model, rather than any physical arguments. The ‘synoptic-scale wind’, calculated over a 45kmX45km grid box, is still well within the stated length-scale of 100km for mesoscale weather systems. For example, a sea-breeze disturbance could cover the whole larger box, as could a large thunderstorm complex. Can the authors try to quantify/justify this choice of length-scales? Some kind of spectral analysis orspatial filtering with known spectral properties could be helpful here. Either the length-scales should be clearly justified by physical arguments about the scale of mesoscale weather systems, or by practical arguments about the scale of variability that is of relevance to large offshore wind farms.
2. When discussing the variability that the authors call long time-scales (6-12 hours), the focus is on diurnal processes over the land influencing the wind over the sea. However, there can also be a diurnal cycle over the water, especially in shallow areas where the water can warm or cool more rapidly. The authors did not state how the SSTs were specified in the model, nor whether they were being updated on daily or sub-daily time-scales. 
3. The authors looked at both temporal and spatial variability, but it would have been nice to see a greater attempt to relate these results. In particular, given that the temporal and spatial analysis should be capturing the same thing, why was it necessary to work with the spatial methods that place limitations on coastal areas? Why could the integrated periodograms over ‘mesoscale’ and ‘synoptic-scale’ periods have been compared, in a similar way to the MSVI? This would have given well-resolved maps of where the mesoscale variability was playing an important role?

Specific comments

1. Page 3, line 41: “The effects of turbulence can be taken into account in Large Eddy Simulations” -> I think it should say “partly taken into account”, since LES models only capture the larger part of the turbulence.
2. Page 3, line 53: “the majority of research is based on the onshore extent of the systems” -> There are a few references that look at the offshore part of land-sea breeze circulations that might be missing here. For example, Short et al. (2019) and Gille et al. (2005). In this context, the authors should also mention the land-breeze, which may be more relevant for offshore winds.
3. Page 2, line 38: There are more recent versions of the wind speed spectrum that you could refrence - eg. Kang et al (2016).
4. Page 3, line 45: Change first sentence to “Less is known about the impact of mesoscale weather systems on wind variability, for example in organised convection”
5. Page 3, line 55: Add reference to Trombe et al. (2014).
6. Page 3, lines 45-60: I think this section is missing discussion of a major source of mesoscale variability, which is from organised thunderstorms or MCS.
7. Page 4, line 100: What height is the ERA5 wind speed accuracy quoted for?
8. Page 5, line 109: ‘aggregated’ - clarify - is this interpolated, or averaged?
9. Page 5, line 130-133: The authors state that the time series is cut in overlapping sections. Later, it says that a ‘Hann window is used to cut the signal into sections’. Is this duplication?
10. Page 5, line 134: ‘fast natural variability’ - is it ‘fast’, or just removing the noise?
11. Page 6, line 135: ‘good estimate’ -> how do you know it’s ‘good’?
12. Page 6, line 137: ‘integrated over a 3-month interval’ - what does that mean?
13. Page 6, line 144: ‘period’ -> can the authors choose a different word? This could mean ‘a period of time’ or ‘periods from the spectrum’.
14. Page 7, line 169: suggest changing to “As the small window is contained within the large window’
15. Page 7, line 170: ‘everywhere’ -> ‘everywhere else’
16. Page 10, line 220: The comparison with lidar data at high frequencies is mentioned, but as far as I can see, this is not shown in the figures.
17. Page 10, line 225-227: Why would the diurnal effects only show up on the 12h time-scale, and not the 24-hour time-scale?
18. Page 11, line 230: It would be useful to contrast/compare the results to Vincent (2011).
19. Page 11, line 240: The authors mention the issues of spin-up around the edges of the domain. This is an interesting problem, but raises the question of whether the boundary removed from the edges was sufficient. Can we really trust the results, give these effects?
20. Page 11, line 246: Can the authors show the periodogram for power, as well as the integrated maps? The power time-series presumably has some constant sections where the wind speed is greater than 15 m/s or less than 3 m/s, and possibly some sudden jumps due to the cut-out speed being reached. What impact do these shocks have on the periodogram?
21. Page 14, figure 9: I don’t find this graph very useful. What is it supposed to be showing? Could the authors show it as an average annual cycle, averaged over the 10-year period? Or overlay a smoothed version so that the curve is more obvious?
22. Page 15, line 269: Why are the ‘mesoscale winds’ always more than the ‘synoptic wind speed’? This is attributed to the land-sea breeze circulation and other mesoscale phenomena. In some cases, this might not be the case, since if the sea-breeze opposes the background flow, then it will weaken the wind overall. I agree that usually, mesoscale phenomena would lead to more windy conditions, but it should not be assumed that this is the case.
23. Page 15, lines 270-272: This section lacks references

Technical Corrections

1. Page 2, section 30: ‘condense wind farm’ -> ‘condensed wind farm’
2. Page 7, line 179: ‘bellow’ -> ‘below’
3. Page 8, line 196: ‘extend’ -> ‘extent’

References

Gille, S. T., Llewellyn Smith, S. G., & Statom, N. M. (2005). Global observations of the land breeze. Geophysical Research Letters 32(5), 1–4. https://doi.org/10.1029/2004GL022139

 Kang, S.-L., and H. Won (2016), Spectral structure of 5 year time series of horizontal wind speed at the Boulder Atmospheric Observatory, J. Geophys.Res. Atmos.,121, 11,946 11,967,doi:10.1002/2016JD025289.

Richard Rotunno, Joseph Klemp, & Morris Weisman. (1988). On the A theory for strong, long-lived squall lines. Journal of Atmospheric Sciences 5(3), 463–485.

Short, E., Vincent, C. L., & Lane, T. P. (2019). Diurnal cycle of surface winds in the maritime continent observed through satellite scatterometry. Monthly Weather Review 147(6) 2023–2044. https://doi.org/10.1175/MWR-D-18-0433.1

Trombe, P.-J., Pinson, P., Vincent, C., Bøvith, T., Cutululis, N. A., Draxl, C., Giebel, G., Hahmann, A. N., Jensen, N. E., Jensen, B. P., Le, N. F., Madsen, H., Pedersen, L. B., & Sommer, A. (2014). Weather radars - The new eyes for offshore wind farms? Wind Energy 17(11). https://doi.org/10.1002/we.1659

Xia, G., Draxl, C., Optis, M., & Redfern, S. (2022). Detecting and characterizing simulated sea breezes over the US northeastern coast with implications for offshore wind energy. Wind Energy Science 7(2), 815–829. https://doi.org/10.5194/wes-7-815-2022

This paper conducts a 10-year mesoscale atmospheric simulation in Kattegat using the COSMO model. The goal of this paper, as stated by the authors is "to investigate what factors influence mesoscale wind speed variability, on what timescales this variability occurs, and how it affects wind power output in offshore wind farms". While I have come across many papers that conduct mesoscale wind resource assessments, I believe that a paper like this that digs deeper into the atmospheric mechanics is valuable. I really enjoyed the spectral analysis as well as the spatial analysis. I also appreciate that the authors have stored code and data on Zenodo. This is also the first mesoscale wind energy paper that I have read that doesn't use WRF, which is refreshing. That being said, I have major concerns regarding connections to the broader literature, novelty, methodology, and analysis. To the editor, I recommend a status of Major Revisions.

Major Concerns

* My first major concern has to do with the framing of this paper (L45-72).

    * To put it succinctly, I'm uncertain if the authors are claiming that either (a) few mesoscale wind simulations studies have been published or (b) if they are very specifically claiming that the mesoscale wind simulation studies that have been published have not sufficiently analyzed "mesoscale wind speed variability".

    * If they are claiming (a), then I strongly disagree.

        * The authors categorize the following as mesoscale phenomena: "thunderstorms, but also sea breeze systems, low-level jets and gravity waves". Many many papers have been published on how wind energy is affected by these, e.g. (Tomaszewski and Lundquist 2020 for thunderstorms, sea breezes in the North Sea by Steele et al. (2014), many papers for LLJs, many papers by Allaerts on gravity waves).

        * Additionally, there has been extensive work on the mesoscale through large-scale wind resource assessments (like the New European Wind Atlas and the WIND Toolkit) as well as smaller but important studies (e.g., Hahmann et al. 2015).

        * Relatedly, I believe that NEWA, the Dutch Offshore Wind Atlas (Kalverla et. al 2020), and possibly NORA3 (Cheynet et al 2022) all include Kattegat in their simulations, and these products should be mentioned somewhere if so

    * If (b),

        * the citations in L62-63 are a small set of mesoscale wind resource papers, and I don't believe these papers do anything unique regarding wind speed variability.

* My second major concern is one of novelty

    * I normally don't comment on novelty, but many mesoscale simulations have been conducted specifically in the North Sea, with domains that include Kattegat. A non-exhaustive list includes Hahmann et al. (2015) and the New European Wind Atlas (see the two papers in GMD). I believe that the Dutch Offshore Wind Atlas (see papers by Kalverla) and possibly NORA3 (Cheynet et al. 2022) may also cover this region. I don't see any references to these papers. The authors should clearly state why their work is novel to help the readers out.

    * I feel like I've seen many papers also do similar-but-different seasonal analysis before (e.g., Wang et al 2019 in California), and these should also at least be referenced

* My third set of concerns has to do with the methodology of this paper

    * Simulation setup: What is the timestep of the simulation? How frequently is data saved out? Did you run a single, 10-year long simulation or did you chunk up the job into smaller periods? Did you use spectral nudging to encourage the long simulation to not drift too far away from the expected ERA5 values? I'm unfamiliar with COSMO, but in WRF, it is known that the PBL scheme choice is very important. If multiple options are allowed in COSMO, which PBL scheme is used here? While not required by this journal, I highly encourage the authors to upload an example configuration file somewhere so that others may more easily reproduce this study in the future. There is possibly one on Zenodo, but the link is private, so I cannot view its contents

    * Validation efforts: Researchers often validate modeled winds against measured winds in order to built trust, but the validation study here instead erodes my confidence. I also don't see how validation furthers the authors stated goals in the intro. I believe that all the lidar analysis should be struck entirely. Please write a sentence or two that explicitly ties the motivation behind the validation back to the goal of the paper.

        * Lidar: The simulated winds do not include wakes, but the lidar is probably being waked. That waking can be significant (>1 m/s modifications). We also don't know how close the lidar is to the nearest turbine, and also the number of operational turbines changes throughout the comparison period, changing the waking strength. The authors attempt to minimize the effects of waking by looking at periods when wind farm availability stayed below 50%, but that is insufficient in my opinion. 

        * ASCAT: I am less familiar with this instrument, but I suggest that the authors point out that others in wind energy have used this data source before (e.g., Hasager et al. 2020). I mention this because I know there is some controversy regarding validating WRF against SAR (a different instrument), but if others in the wind energy sector have looked at ASCAT previously, then there is a stronger case for its use here

        * ASCAT: I haven't seen others make maps of winds at different percentiles. Why did you do this instead of simply comparing mean wind speed maps? You give a justification on L110, but I don't quite follow. Maybe if you provide a summary of what the "double penalty" is, that would help clarify things

        * ASCAT: L106: It's hard to judge if 229,503 WVC is a lot of data or a small amount of data. Could you give more context? Maybe an easy number to calculate is the number of aggregated WRF grid cells over this period.

        * ASCAT: If a model validates well at 10 m, that doesn't necessarily mean that the model is accurate at hub-height (see the Bodini extrapolation papers if you're interested). Please add that caveat in somewhere, as many today would consider validating against near-surface winds to be of minimal utility (for what it's worth, I am not one of those people).

     

* My fourth set of concerns has to do with the analysis

    * Fig 5: I really like the spectral analysis. One of the goals of this paper is to "to investigate what factors influence mesoscale wind speed variability", and as such, I would like to see a stronger physical justification as to why the wintertime shows stronger short-timescale TKE and the summertime shows stronger long-timescale TKE. Why would relatively warm SSTs impact the 20 min - 1 hour range as opposed to a different range? Why would sea breezes and nocturnal jets contribute to TKE specifically on the 6-12 hour range? Consider citing papers in this section to justify your analysis, I don't think you necessarily need to write any code to address this point. Also, consider mentioning here that you also further investigate this line of questioning later in the paper.

    * Fig 6: Nice figure! I think your analysis in L234-235 makes sense. I like that you have the 11 day case study to justify your hypothesis, but please include the figure somewhere, perhaps in an Appendix. Consider talking about Vincent et al (2011) citation in the preceding paragraph.

    * Fig 7: You use a power curve from a 120 m turbine, but I assume you are calculating power using winds measured at 100 m, correct? You should use a turbine for which you have reliable simulated winds. If you didn't save out winds at 120 m, I believe the NREL 5 MW turbine has a 90 m hub height, and you could interpolate between your modeled winds at 80 m and 100 m. In theory you could extrapolate your modeled winds up to 120 m, but I have a feeling that would introduce a lot of noise and also uncertainty, so I recommend against that approach.

    * L270-279: I don't think that the reader was warned that this type of analysis was going to be conducted. This paragraph felt like it came out of nowhere.

    * L281-286: I don't think that the reader was warned that this type of analysis was going to be conducted. This paragraph felt like it came out of nowhere.

Minor Concerns

* Figs 2 and 3: In accordance with WES "colour vision deficiency" publication guidelines, please use a different colormap than the rainbow ones. 

* L26-34: This is a suggestion and not a requirement. I found the discussion on farm density a bit hard to follow, and I wasn't certain why the authors were talking about density. Consider reorganizing the paragraph to move the thunderstorm example higher up.

* L37-38: Is there a latitude dependency for this peak? I would imagine that the timescale isn't also 4 days near the equator, but I may be wrong

* L41/42: Consider citing the review papers of Stevens and Meneveau (2017) as well as the Porté-Agel et al. (2020) review paper

* L96-97: You should cut this statement. If you wish to retain it, please consult the ERA5 wind energy validation that was done as part of NEWA and the Olauson (2018) paper

* L109: Why use RSMD instead of bias? I feel like every wind validation paper I have seen has used bias, not RSMD

* L128: You don't need to change this in the paper, and this is more for my education: do COSMO researchers talk about "periodograms"? In WRF, we call them spectra, though I suppose periodogram is more correct

* L131: Why use a window of 7 days? Could you put that into context of the mesoscale timescales you're interested in? As an aside, thank you for giving all these details on your FFTs, because people often neglect to mention these important details.

* L147-149: If you integrate the periodogram over all bins, that's just the TKE, right? If so, maybe mention here that you take a spectral approach because you can then focus on specific scales (which would be harder to do in the time-domain)

* L151: power curve

* L159 and 165: I recommend the authors state that "We define the MSVI..." and "We define the size of the small window...". When I read these sentences, I got the impression that some other paper specified these definitions, but I believe the MSVI is invented here.

* L191-192: This statement about the double-penalty seems very hand-wavey

* L212-214: I strongly disagree with this statement. COSMO may underpredict winds in simulations without turbines, and the wakes on the lidar would conveniently also lower the observed wind speed.

* L224-225: I appreciate that you conduct statistical testing. Is this test done to 95% confidence?

* L289: Is an RMSD of 1.35 m/s "good agreement"? Relative to what? Either compare to other papers or reword

* L298: I thought the short-timescale variability came from SST/air temperature differences, not convective systems?