Please see the attachment.

General Comments

The manuscript entitled “Noise filtering options for conically scanning Doppler Lidar measurements with low pulse accumulation” presents the study of a series of post processing methods that can be used to reduce erroneous velocity estimations that are generated by a Doppler lidar. The objective is to increase the measuring accuracy of scanning Doppler lidar in those cases where the scanning speed of the lidar’s line-of-sight results in a low laser pulse accumulation. The authors investigate the impact of the increased accuracy of the radial speeds acquired by wind lidar profiler on the estimation of atmospheric turbulence kinetic energy.

The authors have a done a good work to present, given the length of the article, a well written and structured manuscript. However, I think that the article is too long. This makes difficult focussing on the differences between the various approaches that are presented in the manuscript and more importantly to identify the message regarding the approach that the authors recommend. For example, the ideas for the new filtering techniques are presented after 22 pages of the manuscript. For this reason, I think that the authors should consider reducing significantly the length of manuscript, by focusing on the new approaches that they implement and test, to improve its readability.  

Furthermore, the authors state they apply different filtering methods to reduce the noise in Doppler lidars, however I think that rather the noise they reduce the biases of the estimated velocity. A velocity bias in a wind lidar measurement is usually a result of a combination of low atmospheric backscattering and of both the post processing of the Doppler spectra (e.g. noise normalization and thresholding), as well as the frequency estimator of the mean Doppler shift frequency. The latter two parameters are instrument specific. Therefore, I think that it is important to include in the article this information related to the wind lidar used, since they could potential explain the “type B” errors presented in Fig. 1. If it is possible it would be also interesting to show typical Doppler spectra from the “type A” and “type B” errors.

Specific Comments

1. ”Section 2. Measurements”: for the understanding of the velocity measurements presented in the article it is going to be useful to include information regarding the terrain features around the area where the lidar was located, as well as the local wind climatology.
2. Line 139. The authors state that according to their experience random velocity biases are not uniformly distributed in Doppler lidars. I think that this comment should be specific to the instrument used, otherwise please add references here.
3. Lines 162 - 163. Is there a noise threshold applied to the Doppler spectra prior the estimation of the Doppler shift in the lidar used in this study? And if yes, is this threshold always the same or does it vary in time? And how are the SNR values presented in Fig. 1 estimated?  
4. Lines 166 - 170. The authors write “we can also rule out that the occurrence of type B noise is a systematic DL78 problem. However, couldn’t the “type B noise” presented in Fig. 1[i] be a result of RIN noise in the specific Doppler lidar system? If it is not a systematic problem of the DL78 instrument, then what is the origin of this bias?
5. Line 197. I think that there is an error in Eq. (4), according to the way the parameter “C” is defined it should multiply the ACF not divide.
6. Page 13. Figure4, why all the range gates are used in this figure? Shouldn’t we expect to see a difference in the features of these plots with height?
7. Page 19. Figure 9, what does it mean that there is a 43.19% data availability in the unfiltered data? Why isn’t it 100%? And how it is possible that the data availability of the data after the application of the CNS filter is higher than the unfiltered?
8. Line 399. How is it possible to get negative TKE values?
9. Line 515. How is the Δr= 0.5 m/s selected? Is it related to the velocity resolution of the Doppler lidar?
10. Line 658. The use of median absolute deviation has been used in previous studies of filtering random errors from Doppler wind lidars, as in the work of:

              Karagali, I., Mann, J., Dellwik, E., & Vasiljević, N. (2018, June). New European wind atlas: The Østerild balconies      

              experiment. In Journal of Physics: Conference Series (Vol. 1037, No. 5, p. 052029). IOP Publishing.

              Please note that I am neither the author nor one of the co-authors of the above publication.

1. Line 661. Eq. (9) where is the number 0.6745 coming from?
2. Line 685. Fig. 17, what is the parameter “g” in the caption of the figure?
3. Line 709. The authors write “First, by calculating the previously accepted MIN/MAX radial velocity values for each azimuthal direction.”. What happens the first time that the filter is applied where there are no previously accepted min and max values?
4. Section 5.1 It is very difficult to visually detect the differences between the full approach I and II in Figs 20 and 21. Given the length of the article the authors could consider removing these figures and keep in the text the key findings of the testing of the two methods.
5. Figure 24. The authors present here a comparison between the estimated TKE using the suggested combined filter (CF) approach versus the estimated TKE from a sonic anemometer. The results in respect to RMSD and data availability are comparable to the ones got using the CNS filter. It is claimed that the advantage of the CF is that it gives more systematic results when comparing it to the TKE estimation based on a SNR filter which are presented in Fig. 24 c. The results presented in Fig. 24 c show indeed that there is an improvement in the agreement in the estimated TKE values using the two filtering approaches, however they do not say something whether they are accurate or not.
6. Lines 887 – 889. My hypothesis is that the systematic appearance of close to zero values is probably an issue with either the way that the Doppler spectra are being processed in the cases of low SNR by the StreamLine lidar or an instrument specific technical issue. I think that this should be stated, because they way that it is written right now one could understand as a generic issue of wind lidar profilers. 

Technical Corrections

I found only few parts that need a technical correction. Please find below a short list of points that either need a clarification or correction.

1. Line 23. What is meant with “all-sky-scanner technique”
2. Line 253. I suggest replacing the sentence “is a useful tool to get clear results” with “reduces the variance of the background noise”
3. Line 437. Replace “horizonatl” with “horizontal”
4. Line 534. Delete one “the” after the “Compared to”
5. Line 689. Replace “explained with” with “explained by”

Finally, please replace the italic fonts with roman fonts for writing units.