In the manuscript entitled “The impact of seismic noise produced by wind turbines on seismic borehole measurements” by Limberger et al., the authors present a parameter study on the impact local geology and experiment geometry have on recordings of seismic surface waves, which were emitted by wind turbines. These tests are performed using Salvus, a well-established spectral element wave propagation simulator. The goal of these tests is to give insights into and guidelines for required borehole depths for sufficient suppression of surface waves. The manuscript is a valuable contribution to the ongoing discussion on approaches to accommodate the increasing societal need for wind turbines in the operation of seismological observatories.

The manuscript is in a mature state and I don’t see any major issues. While not all source effects are accounted for in the authors’ description of the wind turbine source, I believe the fairly straightforward description they choose is likely sufficient to support the conclusions the authors draw from the simulations. I don’t think further simulations are necessary, but a few minor changes to figures, some wording, and the discussion would improve the accessibility of the manuscript. I believe addressing my comments below constitutes a minor revision.

A) Wind turbine noise. The fundamental benefit of borehole installations is implicit throughout the text but never explicitly stated. The desire for a reduction of surface wave noise through amplitude reduction of all surface waves implies that body waves carry the signal of interest. Surface wave “signals” are impacted to the same extent as “noise” in terms of relative amplitude reduction. To me this appears to be one of the major reasons other approaches to wind turbine noise reduction, which the authors introduced briefly in the introduction, have significant merit. At least some of them aim to differentiate surface waves from particular sources from other surface waves, e.g, in ML-based denoising. I’d appreciate a discussion of this aspect by the authors to gain insight into their thoughts on the position of the approach in this study relative to the overarching problem of wind turbine noise in seismic recordings and which direction they deem as most promising and why.

B) The wind turbine source. The authors choose a simple description of the wind turbine source, a vertically acting sinusoidal force. While the authors discuss some of the drawbacks and potential for future improvement of this choice, a few questions remain. A vertically acting force produces no Love waves, and they then may only emerge due to wave type conversion. To me, it appears more likely that the majority of movement of the wind turbine tower base is horizontal, in accordance with the eigenmodes of the tower (which correspond well to the spectral peaks a wind turbine produces). The authors implicitly acknowledge this, but I wonder what the impact on the results would be. Would this really matter for the intended message of this manuscript? Maybe not, I’m also not sure. Additionally, in the considerations here, all frequencies are treated as equally important (because only relative amplitude reduction at each frequency is investigated). Wind turbines are known to dominantly generate certain frequencies, as the authors also utilise for their relative amplitude reduction estimation on field data. This brings up two questions for me: What is the benefit of repeating the simulations for each frequency (with a tapered sinusoid) instead of formulating a source term more comprehensively, e.g., a sum of harmonics with high amplitudes only for those harmonics corresponding to wind turbine eigenmodes (and blade passing frequency)? Because certain frequencies dominate, the issue of wind turbine noise overpowering body wave signals is worse for those frequencies. What is the authors’ view on this?

C) Figure 3: Two aspects: 1) Maybe it could be helpful to quantify the mismatch between a) and b), at least for the first analytical prediction (colored background). Currently, the authors describe the agreement between them as “very good”. I tend to agree but it would give more confidence to quantify this. This could be done by computation some similarity measure, or maybe by showing a difference plot between the two. I’d expect them to be not exactly the same and it would be interesting to see (and discuss?) what differences emerge and why. 2) The dashed lines corresponding to lambda, lambda/2, lambda/3 were confusing to me at first. I think it would be helpful to expand in the text a bit more on the meaning of these lines. In some way, each of these lines represents an entirely different distribution (colored background) where amplitude is 1 above the line, and 0 below. At least, that’s how I understand what is colloquially meant by the term “penetration depth”. Of course, we know that there is no cut-off depth for surface waves in that sense, but it could be helpful for the reader to state that in your model a lambda/3 penetration depth means relative amplitudes are still 90% (or whatever the corresponding value exactly is, and the other wavelengths with lower values) to make the connection clearer.

D) Figure 8: The representation of the data points extracted from field data could be a bit more precise. Currently, the measured amplitude reductions (34%, 71%, 73%) are marked by what appear to be hand-drawn dashed markings. The exact mismatch (which is likely very low) does not become clear. Maybe it would be helpful to plot a colored circle (with the color corresponding to the field data measurement) on top of the simulation results in the background to give a better visual indication of how well-matched they are. I found the lines confusing at first, because they are measured for one frequency (on data) and then expanded to all frequencies (from modelling). The exact relation between the two does not become clear.

E) Colors. I recommend to use accessible colormaps instead of jet for all figures. For more details on why see Crameri, F., Shephard, G.E. & Heron, P.J., 2020. The misuse of colour in science communication. Nature Communications, 1–10. doi:10.1038/s41467-020-19160-7.

F) Author credits. While I don’t know whether EGUsphere gives specific rules for author attributions, the authors may find the CRediT system useful to cover all relevant aspects (https://credit.niso.org).