The manuscript compares systematically wind speed and turbulence quantities obtained from scanning Doppler wind lidar measurements in virtual mast (VM) mode with corresponding sonic anemometer measurements on co-located meteorological towers. The topic is interesting and highly relevant for a wide range of atmospheric boundary layer applications (e.g. wind energy meteorology) where our present measurement capabilities are limited by the availability and height of existing masts. Proving that lidars could extend our corresponding measurement capabilities will therefore open a wide range of new applications. The topic fits very well in the scope of AMT and I think that the manuscript can be considered for publication after some major revisions.

General comments:

My two main critics are related to a) the description, handling and interpretation of the vertical velocity component and b) the analysis with respect to atmospheric stability presented in in section 4.1.

• a) It has to be carefully explained how your data have been tilt corrected, because this will strongly influence your results (see also specific comments 7b, 9 and 13). If I understand correctly, you argue that the assumption of 0 average vertical wind speed is backed up by the sonic anemometer measurements on the masts. But if you apply tilt correction to the sonics, that is of course no surprise. Only a wind speed and wind direction dependent analysis of systematic deviations could reveal what portion of the tilt is caused by instrument mis-alignment and what by potential tilt of the streamlines due to the topography. This has to be elaborated in much more detail throughout the manuscript.
• b) Stability is for sure a parameter to be investigated here, and I see this part of the analysis as the most important and novel investigation of your study, Unfortunately, is your use of two stability classes in my opinion not appropriate for this purpose. I suggest, to re-perform the analysis with at least 3 stability classes including a near-neutral range. In this context it would be very helpful to see a histogram of the Richardson numbers occurring in your analysis (that is a plot I really miss in the study), that then could guide you to a proper selection of the near neutral range. In case you see also a decent number of very stable and very unstable conditions, you could even consider to extend your analysis to five stability classes.
• As a last general comment I suggest to rework/rephrase the introduction with respect to structure and non-precise scientific writing (I mentioned a few examples in my specific comments). 

Specific comments:

1. line 45: dual RHI scanning has recently also been used for the detection and characterization of thermal updrafts in the CBL (Duscha, C., Pálenik, J., Spengler, T., and Reuder, J.: Observing atmospheric convection with dual-scanning lidars, Atmos. Meas. Tech., 16, 5103–5123, https://doi.org/10.5194/amt-16-5103-2023, 2023. ); this work also documents the potential of retrieving valid data below a fixed user-defined CNR threshold (comment 9) 
2. line 73: "University of Porto, 2020"; is there a more proper reference, e.g. once again Fernando et al.?
3. line 73: "were configured with different scanning strategies"; please rephrase, you can't configure a strategy
4. line 73/74: "enabling the retrieval of multi-lidar measurements"; non-precise formulation, please rephrase; you use multiple lidar measurements to retrieve some other parameters
5. line 90: replace "on" by "in"
6. naming of the towers/virtual masts (table 1 and throughout the whole text): Do you really need the complicated double numbering/labeling; it would be much easier readable if you would go for one clear and understandable abbreviation. My suggestion WS2, WS3, ... for the WindScanners, and maybe T1, T2, T3 for the towers, that would then nicely coincide with the corresponding virtual masts VM1/2/3? As it is it is really complicated to read and requires continuous look up again.
7. line 115: can you elaborate a bit more on the pre-processing;
   
   a) which criteria was used for spike detection?
   
   b) what exactly do you mean with tilt correction (Planar Fit?, Double-rotation?, Triple-rotation?). This will have an important influence on the interpretation of the data afterwards.
8. line127-128: I feel that -22dB is a very conservative threshold, can you elaborate on the amount of data you are losing by applying this threshold;
9. line 145: "... assuming the vertical wind component is zero (w = 0)" ; how confident are you that this assumption holds in the complex environment of Perdigao? (see also my comments 7b and 13)
10. line 198/199; " is the radial velocity error, assuming that is identical in both lidars"; Do you also assume that the error is constant along the beam?; my experience with the scanning WindCube systems is that they have an individual "focus" area where they are performing better, which could cause both distance dependent variations in the errors, as well as differences between the different lidars. This could have considerable implications on your error estimates. Maybe you can elaborate a bit more on that, I assume that DTU has quite good control on their deployed lidars with respect to this behavior.
11. line 211: "assuming that u and v are not correlated"; aren't u and v closely correlated by the wind direction?
12. line 226: replace ")" by "]"
13. subchapter 4.2.2 Vertical velocity (lines 359-363): What kind of are you using for the vertical velocity (see also my comment on that before in section 2.2 describing the tower data)?  This could distinctly influence your results as the different tilt correction methods (that are basically designed to bring the vertical wind speed on average to zero) would cover potential systematic vertical velocities, e.g. caused by the terrain. For that it would be helpful to look into the non-corrected raw data and a potential systematic wind direction and wind speed dependent bias in the vertical velocities. 
14. line 365 "progressively lower sampling rates": How did you lower the sampling rate, by just picking e.g. every 10th value or averaging over the ten corresponding values and using the mean for further analysis?
15. figure 9 and corresponding text lines 369-374: wouldn't it be much more straightforward/"honest" to present this (at least for the velocity) for the horizontal velocity instead of only one component to avoid any potential wind direction influence?
16. the references Menke et al., (line 498) and Pitter et al. (line 516) seem to be incomplete