Synopsis:

The manuscript entitled "Self-nested large-eddy simulations in PALM Model System v21.10 for offshore wind prediction under different atmospheric stability conditions" submitted by the authors Maria Krutova, Mostafa Bakhoday-Paskyabi,  Joachim Reuder und Finn Gunnar Nielsen presents results from a study in which the self-nesting option available in the large-eddy simulation code PALM was evaluated. The authors aimed at reproducing stable, convective and neutral boundary layers, respectively, that were observed at the site of the offshore met mast FINO1. Therefore, they compared the results of their simulations with data from that met mast. The authors draw conclusions on the grid spacing required to reproduce certain observations in the stable and neutral case. They confirm the result from a previous study that showed that using the nesting mode in PALM does not affect the results in case of the simulation of the neutral boundary layer. However, in non-neutral situations there are clear differences between the results achieved with the one-way and the two-way nesting in PALM. Applying the two-way nesting results in the generation of a secondary circulation. The authors recommend therefore to use the one-way nesting under these conditions.  

Recommendation:

The topic presented in the manuscript by the authors is of high relevance and therefore worth being published. 

It can provide valuable guidelines to the users of the PALM code. However, I have a couple of major comments that should be addressed by the authors before I can recommend to  accept the manuscript for publication. Therefore, my recommendation to the editor is major revision.

Specific comments:

a) Major comments:

1. Line 40: "... at the hub height of 119 m" From the paper it does not get clear to me why 119 m are decribed as hub height. According to the information provided on https://www.alpha-ventus.de/english the hub heights in the wind farm alpha ventus that is situated next to the FINO1 met mast are 90 and 92 m, respectively. Moreover, I do not understand why not one of the real measurement heights is applied. Doesn't the introduction of the additional height to which measured data is extrapolated mean an additional source of uncertainty? Later in the paper it is written that the boundary layer height at the times of the measurements under stable stratification might be lower than 119 m. The equation in line 45 is however only applicable in the surface layer, i.e., in the lowest 10 percent of the boundary layer. From my point of view this 

arises doubts whether the derived 119 m data can actually be used for comparisons in the case of the stable boundary layer.

2. Figure 1: It seems as if the domain of the precursor run has exactly half the width of the domain of the main run in the neutral case. This setup is not ideal in order to break up long streaky structures. An excerpt from the PALM documentation reads: "Note that the initial flow field has a perfectly regular structure with a periodicity of the precursor run. This regularity can persist for a very long time. To break up this regularity, use a domain width that is not an integer multiple of the precursor run domain width. The occurring flow field discontinuity at the lateral domain boundary initiates a fast break up of the regularity." (https://palm.muk.uni-hannover.de/trac/wiki/doc/app/examples/turbinf). Figure 2 

shows indeed elongated structures in the u- and the w-component that seem to extend over the whole length of the model 

domain. Are these structures realistic or only a result of the chosen setup? What does it mean for the comparability between simulations and measurements?

3. Line 74: Does the nesting procedure also apply linear interpolation of the wind speed components in the vertical 

direction close to the ground? Wouldn't it make more sense to apply a logarithmic interpolation here? A linear 

interpolation should actually result in an underestimation of the resulting wind speed, shouldn't it? 

4. Line 101: How can there be a geostrophic wind without Coriolis force?

5. Line 101: The simulations are run without a capping inversion. This is a difference to the situation in the real atmosphere. What does it mean for the comparability between the simulations and the measurements?

6. Line 106: "We run main simulations for one hour" Is this sufficient to get a stationary solution? For that the inflow should not change with time and the flow should at least flow once through the model domain. I assume that the simulation time used is actually to short. What was the averaging period used? From my point of view it does not make sense to start with the averaging period directly at the beginning of the large-eddy simulation.

7. Line 173-176: It did not get clear to me why the Coriolis force had been omitted in the simulations. Moreover, I miss a discussion on how strong the impact of neglecting the Coriolis force actually is and what it means for the comparability of the measurements and the simulations. I recommend to add a thorough discussion of the topic to the manuscript.

b) Minor comments:

1. Language check: Please check carefully the use of articles again. From my point of view in the current version of  the manuscript the article "the" is used in many places where it would usually not be used.

2. According to the information provided in the manuscript a deviation between the mean wind speed in the measurements and in the simulations is obvious (see table 1 and table 4). How does this impact the meaningfulness of the results? It would be good to add a discussion on that topic. At least it should be stated that even the mean flow condition observed in the experimental data is not met by the mean flow in the simulations. 

3. Line 31: Please change "... are then compared between the model results and measurements ..." to "... are then compared 

with measurements"

4. Line 60: "Universität Hanover" --> Please change either to "Universität Hannover" (my suggestion) or "Hanover University" or "University of Hanover". 

5. Line 78: "In the two-way nesting case, each child domain interpolates its solution back to the respective parent domain" Wouldn't it make sense to include also an averaging process in bringing the fine-grid data to the coarse grid? It should be considered that PALM uses the volume-averaging approach.

6. Line 83: "development and" --> "development"

7. Line 84: "reaches steady state" --> "reaches a steady state"

8. Line 103: The SBL case uses surface cooling over time. Does it mean that the inflow changes with time? If so, wouldn't this create vertical movements? In case that indeed vertical movements are observed, are any measures taken to damp those movements?

9. Line 111: I suggest using the notation a(t,x1) a(t,x2) instead of a(t) and b(t). a and b are introduced once as time 

series and once as points. The points however do not change with time.

10. Line 117: Please change "synchronity" to "synchronicity".

11. Table 5: Please add an information on the grid spacing used in the CBL and in the SBL cases.

12. Figure 3: It seems to me as even in the case of the one-way nesting the flow field is developing between the inflow boundary and the outflow boundary. The values of the u-component close to the outflow boundary seem to be lower than the values close to the inflow boundary. Is this behavior expected?

13. Line 233: "We performed nested LES of three stability cases for the same wind mean wind speed of 12.5 ms-1". I do not 

see this statement be supported by what is reported in the paper. Neither the measurements nor the simulations show a 

mean wind speed of 12.5 ms-1. Please revise the statement.

14. Line 256 and 260: "Deutsches Windenergi Institut" --> "Deutsches Windenergie Institut"