General comments

The authors present a novel technique for measuring the 1064 extinction product from rotational Raman scattering in combination with 1064 depolarization ratios and compare their retrievals with sunphotometer measurements. This is an important study for the lidar community as these 1064nm products are still far from being a standard. It will reduce the uncertainties of the aerosol optical properties in the infrared spectrum and provide additional information on the spectral dependence that can be crucial for the aerosol typing and the microphysical inversions. The paper is well organized and written in a clear and easy to follow way. However, certain additions and clarifications are required before publication, mainly concerning the rotational Raman technique itself and the comparisons with the sunphotometer measurements.

Specific comments

-- Concerning the technique for measuring the 1064nm extinction coefficient (e.g. Lines 66-71)

Two interference filters (IFF) with a bandwidth of 9nm are placed one after the other before the infrared Raman channel in order to suppress the light from the elastic scattering. Even though the transmission information at 1064nm from the manufacturer seems sufficient for this, factors such as the temperature and/or the incidence angle of the collected light on the IFF could shift the transmission spectrum of the IFF allowing cross-talks from the elastic line. Have the authors checked that this is not the case experimentally? This can be checked by placing a third typical IFF centered at 1064nm in the rotational Raman channel and see if any significant amount of light is detected.

In addition, the Raman spectrum is affected by the air temperature. Is the bandwidth of the IFF sufficient to eliminate any temperature effects in the profiles? What are the uncertainties? Even if such issues are reported in a previous study, a reference must be added along with a few lines explaining the main findings

Temperature effects can be even more pronounced in the near range (below the full overlap region) where the angle of incidence of the collected light in the IFF is still expected to change with range, especially for biaxial systems. This translates to a shift of the central wavelength of the IFF which in turn could result to temperature dependencies, especially if the IFF central wavelength shifts to a more temperature sensitive region. In the near range this could create overlap-like effects to the signal. Values of the extinction profiles at 1064nm in figures 3 and 4 are not provided in the near range, probably due to overlap issues. Could temperature issues be the reason why the 1064nm extinction profile in figure 3 starts at 2km while the 355nm and 532nm profiles start at 1km? Please add a description explaining whether such issues are present. If they are, please provide some information on how they have been dealt with.

-- Concerning the comparisons with the AERONET inversions

The lidar and sunphotometer measurements are not simultaneous. They have a time difference that is larger than 11 hours (around 0 UCT for the lidar -middle time -, and 11:35 UTC for the first sunphotometer measurement). The atmosphere could have changed significantly in the meantime. The authors use only the inversion on 23rd of July in their analysis. The retrievals on 22 of February are also available. They should be used to indicate the degree of atmospheric change in Figure 4, Figure 5, and in the discussion. Looking at the AERONET size distributions on 22 and 23 it seems that there are substantial changes. Furthermore, the inversion on 3rd of March is also available. Having provided the degree of atmospheric change in a ~11hour interval from the 22/02 case, the authors could add the 2 inversions on 03/03 in the analysis since the time difference is not far greater (~14 hours). 

In addition, an AOD and Angstrom exponent comparison part or section should be included in the manuscript prior to the inversion comparisons in the Discussion as these AERONET products are far more accurate. Extinction and backscatter-related Angstrom profiles are also missing from figures 3 and 4. Please include them as well.

Moreover, it should be stressed in the Discussion section that differences with the AERONET retrievals, especially in the lidar ratio, are expected also due to the overlap function of the lidar system. For figure 1 and 3 it is obvious that the extinction profiles are available only above 1km and the 1064nm extinction profile above 2km! This is certainly not negligible, and could affect the comparison, even for the intensive properties since the aerosol type tends to be different inside the PBL. As mentioned above, a comparison with the AERONET AOD and Angstrom values will help quantify the degree of the uncertainties in the overlap region.

All the above points must be addressed in order to draw any firm conclusions from the comparisons with the columnar aerosol properties.  

-- Concerning the 1064 depolarization ratio profiles

The 1064nm PLDR profiles have different vertical structure than the 355nm and 532nm profiles which in general show similar patterns. This is especially visible in figure 1 where PLDR maxima and minima are reproduced in both 355nm and 532nm profiles but not in the 1064nm profile. Can the authors explain this behavior? Is this a matter of atmospheric variability since the 1064nm depol. measurements last only for the first 20 minutes. In this case, and serving mainly as a test, would the profiles be more consistent if the first 20 minutes are used so that all thee depolarization ratios can be measured at the same time?

Technical Comments

-- Line 111-112: Have the two instruments been intercompared? Please provided references.

-- Line 114, Lines 144-145: Here it is not clear what the reference value is and how the cirrus clouds facilitate the selection of a better value, especially for people outside of the lidar community. The authors should provide a clearer description and provide references.

-- Figure 1, 3, and the tables: How were the mean lidar ratio and Angstrom values calculated? From averaging the lidar ratio profile or by averaging the extensive products (backscatter, extinction) first and then dividing them? If not yet applied, the authors should switch to the later as this leads to less noise. 

-- Figure 4 and 5 on the right: Please add also the time of the lidar measurement in the figures

" First triple-wavelength lidar observations of depolarization and. extinction-to-backscatter ratios of Saharan dust" by Haarig et al.

This paper presents the first triple-wavelength lidar observations of depolarization together with extinction-to-backscatter ratios of Saharan dust and also compares the results with aerosol properties derived from ground-based photometers. The paper proposes an advanced lidar measurement techniques. It is well written and suitable for the journal. However, I found that some clarifications in the approach description are needed before publication.

Main comments:

1. The paper emphasizes the fact that the triple-wavelength lidar observations of depolarization together with extinction-to-backscatter ratios of Saharan dust were done for the first time. However, for scientists as myself, who is not specialist in lidar instrumentation, it is not fully clear the importance of this new advanced technique. I would encourage authors to emphasize the achievements. For example, as a reader I have following questions:

• What is the real advantage of the new technique? I could see triple-wavelength lidar observations of depolarization together with extinction-to-backscatter ratios of Saharan dust in other papers, for example, in Muller et al., (2012)?
• Is there any important differences (comparing with previous studies, e.g. Muller et al., 2012) were revealed? If yes, then what are those differences? May be, the new technique mainly confirms the previous results? If so, this should be stated.
• If I understood correctly, previously the measurements at 1064 were not simultaneous, but possible with few minutes of delay? If that is really a case, how critical is to have really simultaneous data, specifically thinking that more than 3 hours are required to get some meaningful data is necessary.
• Why in the figures 1 and 3 the different lidar characteristics are provided in different altitude ranges? These parameters can’t be retrieved simultaneously even with this advance technique? If yes, this shows some remaining serious limitation of the measurements. Also, the spectral dependencies of depolarization ratios, lidar ratios and other properties are different and often contradicting to each other especially at low and high altitudes.

1. In my opinion, there are several weaknesses in the methodology of comparisons of lidar data with AERONET in the present version of the manuscript. They need to be addressed and corrected:

• First of all, the AERONET information content is significantly different from that of lidar:
  • AERONET is measuring aerosol properties in total column while lidar in specific altitudes and doesn’t observe not negligible aerosol layer near ground;
  • Lidar measures signal in back scattering, while AERONET radiances are measured mostly in forward hemisphere with maximum scattering angle of ~ 140-150 degrees. The properties provided by the AERONET retrieval at 180 degrees are not directly constrained by the measurements.
• The above differences should be considered and discussed seriously in the AERONET and lidar comparisons:
• AOD and AE from AERONET are nearly directly measured by photometers and very reliable. It would be good if integrated extinction values and their spectral dependence compared AERONET AOD and AE, before comparing more complex retrieval products.
• The reported aerosol lidar ratios and depolarizations are not fully direct measurements and rely on some caveats and assumptions in processing. It would be nice if those assumptions and their accuracy are discussed. For example, lidar community tends to consider desert dust properties spectrally independent while, in reality, it is not true. Even AE of extinction for dust is rarely zero, and it can be notably positive or negative.
• In the considered observations, the observed aerosol seems to be not a pure dust in the entire vertical column. How the analysis assure that lidar characteristics derived at specific altitudes are for the same dust as seen in total column by AERONET?
• . The authors write: “The drop (I ASSUME “increase”) of the imaginary part of the refractive index at 440 nm compared to 675 nm is too strong in the AERONET inversion procedure, resulting in too high lidar ratios at 440 nm. In-situ studies could not confirm the spectral slope of the imaginary part used in AERONET inversions as discussed by Müller et al. (2012)." First, all AERONET “retrieves” imaginary part, not “uses” as the authors imply. Therefore, the spectral dependence is induced by the necessity of measurement fit. In these regards, I think the sensitivity to the imaginary part is likely higher in AERONET data than in lidar ones. Also, the obtained spectral slope tends to agree with the majority of other analyses of the dust (Shettle and Fenn 1979; WMO1983, Patterson et al. 1977; Sokolik et al. 1993; Koepke et al. 1997; Sokolik and Toon 1999, and there are many more recent data). In contrast, as I mentioned above the lidar community tends to assume no spectral dependence for desert dust properties. In these regards, I would suggest the authors to elaborate this aspect and if there is full confidence in the results to put a clear statement suggesting to use neutral or less pronounced spectral dependence for imaginary part of the dust. Some more extensive literature review of this subject is also desirable.

1. One of the weakest points of the paper is comparison with so-called GRASP retrievals. First of all, I would like to emphasize that GRASP and AERONET retrievals have the same origin (Dubovik and King, 2000 and Dubovik et al. 2006). Indeed, Giles et al. (2019) was reporting only updates in quality assurance of the AERONET results but not the change of the retrieval, Sinuyk et al. (2021) outlined few new elements in the retrieval but emphasized that overall the retrieval methodology remains the same as in Dubovik and King, 2000 and Dubovik et al. (2006). Therefore for the same data GRASP and AERONET retrievals should not show any significant differences. Certainly, GRASP is more flexible and can easily use the different (more complete observations than a 4 wavelengths) and this is why GRASP was used in the paper, I suppose… However, what was the RATIONAL for comparing lidar results with results of inversion of AOD data only (i.e. no angular sky radiances) as done by GRASP-AOD approach by Torres et al. (2017)? It is evident, that AOD HAS practically NO SENSITIVITY to aerosol properties at 180 degrees!!! Torres et al. (2017) clearly stated that GRASP-AOD strongly relies on the a priori assumptions about complex refractive index and particle sphericity. It is interesting that the authors see the better agreement of GRASP lidar ratios with the lidar data. This probably shows nice potential of GRASP-AOD retrieval once the constraints are correctly assumed by Torres et al. (2017). However, overall, I do not see much justifications of comparing AOD only retrievals with lidar results!
2. From my viewpoint, the comparisons of  GRASP results with the lidar ones would be much more convincing if GRASP would be used for inverting AEONET AOD and radiances at more wavelengths (even if at some new channels only AOD data would be added). Moreover, I would think that the best approach would be to use GRASP for simultaneous inversion of the lidar data and AERONET dada as illustrated by Lopatin et al. (2013) and especially Lopatin et al. (2021). Such approach would be the most appropriate to reveal the strength and shortcoming of GRASP aerosol model, since would show if the model can adequately reproduce all (lidar and photometer) data or not.
3. The authors emphasize serious limitations of Dubovik et al. (2006) spheroid model to reproduce the lidar measurements. Based on my personal experience I would probably agree that Dubovik et al. (2006) model would have difficulties to reproduce non-monotonic spectral dependence of depolarization ratio, as well as, the highest values of the depolarization ratio. However, I do not think that the comparisons of lidar data with AERONET and GRASP shown in the paper can be considered as a solid justification for such a statement. Once again, I would suggest trying the joint inversion of lidar and photometer data by GRASP as demonstrated by Lopatin et al. (2013, 2021). By the way, Lopatin et al. (2021) shown that using direct lidar measurements, e.g. volume depolarization instead of aerosol depolarization ratio results in better comparisons between photometric and lidar results.

Minor comment:

The following sentence seems to be awkward and incorrect: "Ln.33 The particle depolarization ratio sensitively depends on the form of the backscattering particles, whereas the dust lidar ratio is a function of particle shape”.



O. Dubovik