The authors assess the particle backscatter coefficient profiles from Aeolus/ ALADIN using co-located ground-based lidars at three locations in Greece, together with auxiliary model and satellite datasets. They attempt to attribute discrepancies between space borne and ground based lidars to (i) the natural variability of aerosols, (ii) instrument or retrieval limitations and/or (iii) spatial temporal co-location issues. This paper needs substantial improvements before it is worthy of publication in AMT.

Major comments

In a very general sense, numerous sentences throughout the paper are too wordy and their structure just too complicated. For example, “are utilized towards an optimum characterization of the probed atmospheric conditions under the absence of a classification scheme in Aeolus profiles” could be replaced by something along the lines of “are used to characterize the atmosphere as Aeolus/ALADIN does not provide an atmospheric classification product”. This makes for a lengthy paper. Other examples are “obtained results” instead of simply writing “results” or “probed Atmosphere” instead of “atmosphere” throughout the paper.

We recommend that the authors:

. Use ALADIN in all caps or Aeolus/ALADIN consistently (instead of small caps and/ or interchangeably using both names). Also, they should spell it out once when first introduced (i.e., “Atmospheric LAser Doppler Instrument”).

. Add a table describing the lidar(s) at each of the three locations. This table could contain, for example, the lidar’s name, a small description, its limitations and uncertainties, its products. It could also contain information on how cloud screening is performed and possibly a dominant type of aerosols at each location (together with references related to these stations).

. Add a brief description of the two models used in this study and, especially, their limitations. The way the analysis is written might sometimes give the impression that model results are considered as accurate as observations.

. Add more text comparing their results to those in studies such as Baars et al. (2021) and Abril-Gago et al. (2022). In a more general sense, it would be helpful to open the paper by describing in more detail what their study brings to the table/ how it complements other studies.

. Specify the way they average the MODIS AOD and add more analysis to their spatial characterization of aerosols at the three different stations. We believe that there is more to the characterization of aerosol spatial variation than simply averaging the AOD in different boxes. The authors should refer to previous studies such as Anderson et al., (2003), Sayer and Knobelpiesse (2019) or Shinozuka and Redemann (2011)

. Discuss the limitations of their cloud screening of the Aeolus/ALADIN aerosol backscatter profiles using SEVIRI.

. Discuss whether the Aeolus/ALADIN, with its 87 km horizontal resolution, is in fact able to characterize the aerosol natural variability at the three locations.

. Avoid strong statements such as “very good performance” when it pertains to a specific altitude, one case study and no quantification of the differences between space borne and ground based lidar in that case.

. Use additional satellite derived aerosol information (e.g., CALIOP, TropOMI) to further characterize the aerosol during their case studies.

. Specify why they use AERONET level 1.5 instead of more accurate level 2 data (quality assured). Also, they should show AERONET-derived FMF and SSA at different wavelengths throughout the day (in addition to what they already do -- spectral AOD and angstrom exponent). This would be like Figure 8 in Abril-Gago et al. (2022). Let us remind the authors that, in addition to a size difference, a difference in SSA at two wavelengths from AERONET could point to the presence of dust versus smoke (e.g., Russel et al., 2014; Kacenelenbogen et al., 2022).

. Specify why they show Aeolus/ALADIN profiles that are cloud contaminated (or “unfiltered”) in their analysis. It is not obvious why there would not be more disagreement between cloud-unfiltered space-borne and cloud-filtered ground-based profiles compared to cloud-filtered space and ground profiles. Or is this a way to test their SEVIRI-based cloud filtering method?

. Figures should be called in the order they appear.

. Shorten the conclusion

Detailed comments:

Title: The authors might want to add “ALADIN” and “aerosol” in the title for increased searchability

Line 29: Why not give examples after “a variety of aerosol species.”

Line 32: Why is PANACEA spelled out but not AERONET, CAMS, MERRA-2 etc…? We recommend either spelling none or all of them.

Line 33: we recommend writing “sunphotometry observations”… “model reanalysis” …” modeled air mass back trajectories”

Line 36: Again, multiple sentences throughout the paper are too wordy. For example, “are utilized towards an optimum characterization of the probed atmospheric conditions under the absence of a classification scheme in Aeolus profiles” could be replaced by something along the lines of “are used to characterize the atmosphere as Aeolus/ALADIN does not provide an atmospheric classification product”

Line 40: “very good” is too strong a statement here.

Line 44: We recommend writing “46 identified cases when using [this time frame] at all three stations…”

Line 47: “positive tendency” could be replaced by “improvement” and “both Aeolus vertical scales” by “multiple Aeolus vertical scales”

Line 48: we recommend to replace “justified” by “explained” + “in the vertical the Aeolus performance”

Line 49: “performance decreases” followed by the explanation for that decrease is not clear

Line 83: We recommend “Such observations are provided by networks… or by dedicated experimental airborne (Ansmann et al., 2011; Weinzierl et al., 2016) or shipborne campaigns (Bohlmann et al., 2018)”

Line 87: We recommend “characterization of aerosol vertical structure at global (e.g., Liu et al., 2008) … was performed using CALIOP … and CATS… respectively on the CALIPSO and the ISS…”. CATS could use other references such as Lee et al., (2019)

Line 108: “good quality” needs a reference

Line 116: ALADIN needs to be in all caps throughout

Line 117: We find the description in Flament et al., (2021), a little clearer (e.g., “The UV laser beam is linearly polarized at the laser output. It goes through a quarter-wave plate before being routed towards the telescope and is thus transmitted towards the atmosphere with a circular polarization…”)

Line 126-128: This is important and should be explained in more detail. This paper is about validating Aeolus/ ALADIN. The limitations of the lidar should be clearly explained and other papers should be referenced.

Line 130: We recommend “ALADIN” and why are “continuous” calibration and validation needed? Please explain.

Line 136: We recommend “L2A aerosol optical properties”

Line 138: Regarding the “excellent agreement” here, we recommend adding some nuance. These results for a case study with a strong non-depolarizing aerosol, were ~satisfying only between ~4 and 8km.

Line 142 – 156: If the type of aerosols over the three regions is discussed here, then you might consider not repeating it elsewhere (e.g., section 4). In general, we recommend adding a table describing the lidar(s) at each of the three locations. This table could contain, for example, the lidar’s name, a small description, its limitations and uncertainties, its products. It could also contain information on how cloud screening is performed from these ground-based lidars and possibly the dominant type of aerosols at each location (together with references related to these stations).

Line 184: HSRL was already introduced on line 116.

Line 187: We recommend “are backscattered”

Line 206-215: What is the purpose of describing algorithms that are not used in the study (e.g., ICA and MCA)?

Line 216: We recommend “the primary and most reliable”

Line 217: We recommend “measured signals in the Mie channel”

Line 224: We recommend “signals in each channel”; also, the sentence is not clear

Line 260 and section 4: Again, the three stations, type of lidar(s), products, uncertainties, limitations (e.g., overlap), etc. could really use a table. That table could also show a predominant aerosol type over the region and a median and standard deviation AOD from satellite(s).

Line 263: We recommend using “different” instead of “adverse”

Figure 1-i: It would be helpful to write “all three stations are within Xkm of each other”.

Line 272 to 275: The authors must mean “to ensure the consistency of all lidar-derived observations”?

Line 279: We recommend deleting “measurements” here.

Line 280-281: Why is this assumption plausible? Does it remain to be tested?

Line 286-288: Doesn’t this apply to all three stations? Also why not add biomass burning aerosols here?

Line 338: We recommend “Aerosol spatial variability in the vicinity of the PANACEA sites”. A description of the dominant aerosol type at each station would fit well here but then the authors would have to delete it from the introduction to avoid repetition.

Section 4.4: The purpose behind studying the spatial variability could be explained more clearly. Our understanding is that the authors are attempting to characterize spatial variability to explain a potential disagreement between Aeolus/ ALADIN and ground-based lidars. A disagreement could be due to imperfect spatial co-location and/or simply ALADIN’s 87 km horizontal resolution. The authors are studying horizontal variability by using total column integrated AOD and that should be mentioned as well. There could be minimal horizontal variability but a strong vertical variability. It is also not clear how the authors have computed the mean AOD from MODIS. Is it a arithmetic or geometric mean? It does make a difference -- see e.g., Sayer and Knobelspiesse, (2019). Spatial characterization analysis usually uses mean and standard deviations within each satellite grid cell and/or the variation between consecutive satellite pixels or airborne measurements within a region (e.g., Anderson et al., 2003 or Shinozuka and Redemann et al, 2011).

Fig. 2: The orange arrow is hard to see; It is also not clear if the analysis involving the 46 cases considers the closeness of the actual track to the station (e.g., better spatial colocation on July 1^st)

Line 381: The authors should discuss this “temporal window extension” in more detail and attempt to explain its consequences.

Line 385-397: We recommend “derived from radiances measured by SEVIRI”, “indication of cloud presence”; the limitations of using this SEVIRI cloud mask should be discussed. For example, could SEVIRI be missing small broken water clouds? What about cirrus clouds? and what would be the consequences on the Aeolus/ ALADIN aerosol profiles?

Line 403: Regarding (ii), how will the authors differentiate the effects of natural variability, the imperfect co-location and the errors in the Aeolus/ ALADIN instrument? See e.g., section “nature versus noise” in Anderson et al., 2003. Regarding (iii), this was already demonstrated in numerous studies.

Line 406: We recommend “(…) Cal/Val study to facilitate the interpretation of our findings and to identify possible upgrades in the Aeolus/ALADIN retrievals.”

Line 409: We recommend “the results are depicted in Figure 3”

Line 411: We recommend “… Aeolus retrievals are provided at a coarse horizontal and vertical resolution …”

Line 420: We recommend “To depict the spatial patterns (…)”

Line 423: The fact that MERRA-2 and CAMS provide “aerosol products of high quality” is a strong statement and should be explained. The explanation should include model evaluation results from previous studies. Model aerosol optical properties and model aerosol speciation have serious limitations, which should absolutely be mentioned in the text.

Line 425: Why not use AERONET Level 2 (quality assured) instead of Level 1.5?

Line 427: We recommend “characterization of the aerosol load and size over the station”

Line 432: Figures should be called in the order they appear. Figure S4 is introduced before S1, S2 or S3.

Line 435: We recommend adding “at 550 nm”

Line 437: The truth should be in the sunphotometer direct measurements. This sentence, as written, could be interpreted as things being the over way around.

Line 438: The Angstrom exponent should be briefly explained here (i.e., difference of AOD at two (or more?) wavelengths that informs on the particle size) and references for typical dust angstrom exponent should be added to the text (e.g., Dubovik et al., 2002)

Figure S2: This10 min-worth of high VLDR content looks suspiciously high compared to the consecutive profiles in the curtain plot. How do the authors explain that dust was present for only 10min and then suddenly disappears?

Line 442: Why would this case be ideal for evaluating Aeolus/ ALADIN as we know Aeolus cannot measure non-spherical dust properly?

Line 449: What is meant by “statistical uncertainty margin”?

Line 473: The authors mention “fine particles” but an explanation is missing here; we recommend “until their arrival over…”

 Line 480: We recommend “AOS are mainly attributed …”; the models seem to be taken, once again at face value here (i.e., they seem to be treated the same as observations, but model species (and their spatial variation) are sometimes not reliable)

Line 484: We recommend “ALADIN reproduces the layer’s structure well”

Line 493-496: This sentence is not clear. The fact that MERRA-2 and CAMS aerosol optical properties and speciation disagree cannot be directly connected to a good or bad performance between models and AERONET. One would need to directly compare aerosol optical properties from MERRA-2 or CAMS to aerosol optical properties from AERONET. Also, the way this is written could make it sound like AERONET provides aerosol species, which it does not. Instead, it measures aerosol optical properties, which can be used to define aerosol types and that can be indirectly translated into aerosol chemical species in certain cases (Kacenelenbogen et al. 2022).

Line 502: We recommend “Aeolus performance depends on altitude according to Polly…”.

Line 554: The reader needs to be reminded which case studies are included here -- are those the 46 case studies of line 383?

Line 563: The statement referring to “the contribution of depolarizing particles is quite low based on the ancillary dataset” needs more explanation and needs to be supported by some results.

Figure 4: The metrics need to be described, like in Abril-Gago et al. (2022). Authors should specify why they show Aeolus/ALADIN profiles that are cloud contaminated (or “unfiltered”) in their analysis. Is is not obvious why there would not be more disagreement between cloud-unfiltered space-borne and cloud-filtered ground-based profiles compared to cloud-filtered space and ground profiles. Or is this a way to test their SEVIRI-based cloud filtering method?

Line 586: We recommend “Fig. 4” instead of “Fig. 5”.

Line 591-592: The authors should mention that “SCA mid-bin” is expected to perform better than SCA

Line 597-599: This is a repeat from line 137

Line 624: Do the authors mean low SNR instead of high SNR?

Figure 6-7: Again, why show cloud contaminated Aeolus/ALADIN profiles? Also, why show SCA instead of SCA_bin as the latter is expected to lead to better results (already shown in Fig. 5)?

Line 641: Again, can Aeolus/ALADIN profiles still be cloud contaminated after applying the SEVIRI cloud mask?

Line 641-643: This statement about not using QA flags appears too late in the text. It should be in the method or the Aeolus/ALADIN section.

Line 654: “many similarities” needs to be described in more detail.

Line 682-688: The authors should explain why they expect a difference in performance between the ascending and descending orbital data. Grouping the data per orbit direction seems inconclusive and we question the usefulness of mentioning the results.

Line 701-705: This appears too late in the text. It should be part of the comparison method between Aeolus/ALADIN and ground-based lidars.

Line 702: It is not clear what the authors mean by “the theoretical assumptions”.

Line 719-722: Again, the authors should add some nuance to the discussion on model performance.

References:

Abril-Gago, Jesús, et al. "Statistical validation of Aeolus L2A particle backscatter coefficient retrievals over ACTRIS/EARLINET stations on the Iberian Peninsula." Atmospheric Chemistry and Physics 22.2 (2022): 1425-1451.

Anderson, Theodore L., et al. "Mesoscale variations of tropospheric aerosols." Journal of the Atmospheric Sciences 60.1 (2003): 119-136.

Baars, H., Radenz, M., Floutsi, A. A., Engelmann, R., Althausen, D., Heese, B., et al. (2021). Californian wildfire smoke over Europe: A first example of the aerosol observing capabilities of Aeolus compared to ground-based lidar. Geophysical Research Letters, 48, e2020GL092194. https://doi.org/10.1029/2020GL092194

Dubovik, Oleg, et al. "Variability of absorption and optical properties of key aerosol types observed in worldwide locations." Journal of the atmospheric sciences 59.3 (2002): 590-608.

Flament, T., Trapon, D., Lacour, A., Dabas, A., Ehlers, F., and Huber, D.: Aeolus L2A aerosol optical properties product: standard correct algorithm and Mie correct algorithm, Atmos. Meas. Tech., 14, 7851–7871, https://doi.org/10.5194/amt-14-7851-2021, 2021.

Kacenelenbogen, Meloë SF, et al. "Identifying chemical aerosol signatures using optical suborbital observations: how much can optical properties tell us about aerosol composition?." Atmospheric Chemistry and Physics 22.6 (2022): 3713-3742.

Lee, Logan, et al. "Investigation of CATS aerosol products and application toward global diurnal variation of aerosols." Atmospheric Chemistry and Physics 19.19 (2019): 12687-12707.

Russell, Philip B., et al. "A multiparameter aerosol classification method and its application to retrievals from spaceborne polarimetry." Journal of Geophysical Research: Atmospheres 119.16 (2014): 9838-9863.

Sayer, Andrew M., and Kirk D. Knobelspiesse. "How should we aggregate data? Methods accounting for the numerical distributions, with an assessment of aerosol optical depth." Atmospheric Chemistry and Physics 19.23 (2019): 15023-15048

Shinozuka, Y. and Redemann, J.: Horizontal variability of aerosol optical depth observed during the ARCTAS airborne experiment, Atmos. Chem. Phys., 11, 8489–8495, https://doi.org/10.5194/acp-11-8489-2011, 2011.

The authors perform an assessment of the SCA backscatter coefficient product from the Aeolus satellite by comparison to ground based lidar observations. They have split their work in two parts. First, an analysis of four dedicated, illustrative test cases is presented, including a creditable multitude of ancillary data that provides information on the aerosol origin and type. In a second part, all available collocation cases over the chosen lidar stations contribute to a statistical analysis of bias and RMSE, spanning the current mission lifetime. In lack of a cloud mask within Aeolus’ data products, the authors efficiently filter the data themselves and can thereby show moderate to good performance of Aeolus backscatter coefficients. However, the findings suggest that particularly the retrieved backscatter coefficients closest to the ground are not reliable since they suffer from low SNR. The retrieved backscatter coefficient above the ground is biased due to surface reflectance.

However, there are some substantial changes and clarifications necessary before publication of the work.

General (Major) Comments

- I agree with referee 1 that the wording and sentence structure throughout the manuscript makes it often more difficult to grasp. That is particularly because of numerous insertions into the sentences, separated by commas or parentheses, and maybe a general trend for nouns over verbs. To provide only one example from L.468 “Under the prevalence of the Etesian winds (Tyrlis and Lelieveld, 2013), a typical pattern dominating over the broader Greek area during summer months, when winds blow mainly from NNE directions, anthropogenic aerosols from megacities (Kanakidou et al., 2011) and particles originating from biomass burning in the eastern Europe and in the surrounding area of the Black Sea (van der Werf et al., 2017) are transported southwards.”

The main clause “Under the prevalence of winds [...] aerosols [...] are transported southwards.” is stretched out too much.

- Some parts of the manuscript seem not to contribute to or distract from the scope of the paper. Some sections or paragraphs could potentially be shortened or omitted, by asking who the audience of this work is. E.g. the second section with the Aeolus instrument description contains very general information that is mostly not used throughout the rest of the manuscript and can therefore be referenced (see suggestions in specific comments). Also, the conclusion can be made more compact by separating it into a conclusion and an outlook section, or can be condensed in other ways (also see suggestions in specific comments).

- It is good that the authors assess the aerosol climatology via the MODIS-Aqua AODs. However, the performed analysis of concentric circles seems not well suited for the assessment of the horizontal heterogeneity, see specific comments regarding L.351-366

- throughout most of the text, the authors do not differentiate between the performance of the Aeolus satellite itself and the performance of the retrieved SCA co-polar backscatter coefficient within the L2A product. This needs to be clarified, particularly since two significantly improved optical properties products are available as of March this year (see specific comments)

- the currently implemented collocation method appears to me to have an offset of about 45 km in flight direction, since only the start of a BRC but not its center location is used for the distance calculation to the lidar ground stations. If I did not miss something, this will need to be adjusted, making necessary to reanalyse the data and update the corresponding plots

- Subsections 6.1.3 and 6.1.4: In my opinion, the descriptions and conclusions of the individual Aeolus lidar profiles in Fig. 3 may be much to detailed and flawed. I explain in my specific comment on L.502-503, that there is reason to believe that the discussed discrepancies are just noise induced and therefore the reached conclusions are not valuable or generalizable. I recommend the following procedure: As a first validation step, I encourage the authors to provide Figure 3 with all negative SCA backscatter values shown. This will provide an impression of the actual noise level encountered in the SCA backscatter in the different test cases. I expect to see values up to minus 0.5-1 Mm-1sr-1 in some cases in accordance with e.g. Fig. 8 in Ehlers et al. (2022, doi.org/10.5194/amt-15-185-2022). If that is indeed the case, then the discrepancies along the profiles may be mostly noise induced and the current, detailed conclusions must be reconsidered, i.e. the authors should test for the hypothesis and make accordingly changes to the text. In this case, especially the statement in the abstract L.41-43 “The level of agreement between spaceborne and ground-based retrievals varies with altitude when aerosol layers, composed of particles of different origin, are stratified (8th July 2020, 5th August 2020).” is contestable.

Specific comments

L.38-43 This could be more compact, considering it is in the abstract. Particularly the discussion of the 4 test cases seems very specific and could be condensed into a shorter sentence.

L.41-43 The “level of agreement” is not strictly defined and, hence, seems subjective. In my opinion, this statement is presented too confident. In fact, this is concluded from two single BRC of Aeolus. The authors themselves stress the issues with collocation, so I am not convinced at this point that the remaining variations in the profiles are caused only by the stratification, but can originate from horizontal inhomogeneity of the atmosphere’s aerosol load. The supplementary material helps only little, since the models provide AOD only.

L.74-80 This sentence is too long and the last part seems not to fit in gramatically.

L.82 “as well as the geometric features of the particle's layers” What are the “geometric features”, if not the already-mentioned vertical structure? Please omit or specify.

L.94-115 This paragraph provides an overview of the L2B wind product development and application. Considering the paper's scope of aerosol backscatter assessment, I recommend to omit/condense it.

L.116-128 This paragraph might be better placed in / merged with the second section about the ALADIN instrument.

L.169, Section 2; While reading this section I was reminded of other Aeolus related works. In fact, I was wondering whether the degree of detail is relevant for the audience of your manuscript, or if you could get away with a more high-level description of the Aeolus typical vocabulary only (as in the L2a user guide). Essentially, the information here can be looked up in the Aeolus Science Report or many other papers. But since your work focuses on validating the data rather than e.g. modifying the L2A processing chain or including so-far unknown instrumental effects, it may be a consideration to omit most parts for brevity.

L.192-194 It is unclear with which property the angle increases. Please make the formulation unambiguous by changing the statement to something like e.g. “The 35 degree off-nadir pointing corresponds to an angle of about 37.6 degree with the Earth surface, due to its curvature”.

L.205, Section 3; This description of the L2A data product is outdated at least with the start of the new baseline 2A14 from 29th March 2022. It has been decided to remove the ICA product completely and two new optical property products have been added, namely the SCA-MLE Optical Properties (Ehlers et al., 2021) and the AEL-PRO Optical Properties (from adjusted EARTHCARE algorithms). Both products are expected to bring considerable improvement over the SCA, since the inverse retrieval problem is solved not algebraically but via state-of-the-art methods (Maximum Likelihood Estimation, Optimal Estimation, respectively), see Ehlers et al. (2021) for the SCA-MLE product. These changes are tracked e.g. in the Aeolus Level 2a Processor Input/Output Data Definition available here: https://earth.esa.int/eogateway/documents/20142/37627/Aeolus-L2A-Input-Output-Data-Definitions-Interface-Control-Document Please give an adequate description of the data product, in order to put your analysis in the correct context. To my knowledge the Aeolus mission data has not yet been reprocessed with the new processors, which then offers potential for future studies.

L.238 NITWT is not the name of the method, but the name of the variable that allows for simpler notation.

L.243-250 This paragraph is a perfect introduction to then mention the SCA-MLE and AEL-PRO optical properties data products, which aim to mitigate such problems to a big part. A brief description could be added hereafter to update the section. It must be stressed that also the backscatter profits from the processing update!

L.250-252 This is not a primary reference for the zero-flooring. The primary reference is Flament et al. (2021) or the L2A Algorithm Theoretical Baseline Document, section 6.2.2.1, see here: https://earth.esa.int/eogateway/documents/20142/37627/Aeolus-L2A-Algorithm-Theoretical-Baseline-Document

L. 351-366, Criterion for spatial homogeneity; The authors want a measure for spatial homogeneity of the atmosphere’s aerosol load on an instantaneous base. However, the presented, concentric, climatological analysis is not suited for these needs for at least two reasons. Also, the description lacks some detail. The two main points below:

i) Figure 1 provides one AOD value per concentric circle and location. So the reader has to assume that in addition to the spatial average a temporal average over the 10 year period has been performed. This average is not mentioned in the text and the word “climatological” appears only in the figure caption. Therefore, the word “Annual” in Fig. 1i and 1ii is misleading.

Now, averaging the AOD pattern over time will potentially smoothen out most of the horizontal heterogeneity of AOD that is present on a daily basis. However, the latter is the desired property in order to assess the quality of collocation. An (oversimplified) counter example goes as follows: Assume AOD pixels follow a chess board pattern (with changing locations over time due to wind). This would show a lot of heterogeneity, hampering collocation. But due to the two averages, one over the rings and one over time, the developed criterion would indicate perfect homogeneity.

ii) Another shortcoming is the possibly very location specific outcome of this analysis: In my opinion, there is no reason to either favour Aeolus’ frequent observation location or the ground-based lidar location as a center for the concentric circles. However, if the center was chosen e.g. 80 km away from pollution sources such as Athens or Thessaloniki, then their increased AOD pixels would be averaged with all pixels from a whole ring of unrelated locations, including ocean, meadows and villages 160 km away. This way, pollution sources will be hidden by averaging, if the circles’ center is not coincident with them, making the presented analysis little robust.

Standard tools such as 2D autocorrelation functions of the AOD “images” would not suffer from such shortcomings.

L.374 You take the beginning of the scan as the location of an Aeolus BRC, however, its middle is more representative as centerpoint of the measurement but lies about 45 km further away. When considering Figure 2(i), imagine now that by random chance, Aeolus had started scanning each BRC 5 km earlier. Then the BRC that is now red would not be considered in the analysis at all, though still closest.

This means, your collocation criterion is currently offset by about 45 km in flight direction, which is quite a lot! This must be fixed and can be done, e.g. approximately, by applying the running average filter ([0.5 0.5]) over the current latitude and longitude arrays, or by assuming the satellites’ speed and direction. Otherwise, the location of the center measurement within the BRC can be extracted from elsewhere in the L2A product, to the best of my knowledge.

Fig.2 The tips of the orange arrows are barely visible, please enhance.

L.380-383 The collocation criteria should be objective, so please quantify by up to how much time they have been relaxed, another hour? 2 hours?

L.395-397 When reading the manuscript, my burning question was, how many of the above mentioned cases/BRCs remained after filtering. I only found this information much later in the text. Could the authors please consider moving this information up here?

L.408-410 Can the authors motivate here why explicitly these cases where chosen? Also, I wonder which criterion was applied to choose a single BRC out of each case, presented in Fig. 3. The spatially closest? The visually most representative? I presume that at least for one of the cases there was more than one BRC to consider.

L.442-444 The word “ideal” is exaggerated.

L. 484-486 As the authors report themselves earlier, the backscatter coefficient in SCA and SCA midbin is essentially identical, just averaged onto two different scales. Hence, I do not support this argument of overestimation/underestimation and find it misleading. Also, with a quick look it seems that both are overestimating. Do the authors mean that the layer reaches too far up in SCA midbin? Please specify.

L.502-503 This sentence is stated with a suggestion, while in fact the information should just be that Aeolus and PollyXT do not agree over the entire profile, and where. As you are well aware this mismatch can have various reasons but a lack of performance of Aeolus.

At this point I want to also mention, that the error bars on the L2A products are not found to be accurate, and hence suggest a wrong sense of precision, see the recent work of Adrien Lacour from Meteo France and e.g. Fig. 8 in Ehlers et al. (2022): In this test case, the MLE retrieval brings the optical properties much closer to the ground truth than SCA and SCA midbin. So the gaping disagreement between the SCA or SCA midbin and the ground truth is apparently due to noise in the cross-talk corrected particulate (Y) and molecular (X) signals (beta_p=Y/X*beta_m). The true magnitude of this noise can also be illustrated with the magnitude of the negative backscatter values in almost clear atmosphere, which unfortunately are not shown in Figure 3. However, Figure 6 gives a good idea of the spread of negative values in SCA, indicating a ballpark value of 0.5 up to 1 Mm-1sr-1 around the GROUND beta = 0.

Now, the discussion following in L.502 to 507 focuses entirely on the value of two Aeolus bins between 2 and 4 km altitude, the errors of which are most likely underestimated. However, the discrepancy with the ground truth is not much bigger than the approximate noise amplitude estimated from Figure 6 above. Therefore, it is very much possible that these discrepancies in just these 2 bins are indeed caused by noise! Therefore, it is not reasonable to generalize from these results an altitude dependent performance and to conclude a contradiction between the observations on such a weak fundament. This can only be done statistically.

L.530-535 I see no to little reason to underline that SCA midbin is "better” in this particualar case. As you point out, SCA midbin has just worse resolution, which helps to reduce the mismatch here, because averaging consecutive bins also reduces the noise. This is no characteristic of this particular lidar profile but follows from the math: In general, SCA midbin is worse at high SNR because of the resolution loss, but appears better in low SNR due to the additional smoothing that the average implies.

L.576-577 Can you specify how the ground profiles have been rescaled to match vertically the Aeolus bins? Were the ground-based observations averaged onto both different scales or simply sub-sampled? The latter is not preferred. Please provide an explanation or formula.

L.579-585 The range bin index is a tricky reference to perform the analysis on, but I see the need for this implementation. However, it should be stressed in a seperate sentence that, this way, one may mix up bins of e.g. 250 m size with bins of 1 km size, which have different noise properties. This is important for interpreting the reliability of RMSE and bias.

L.586-596 Reading the text while looking at the figure, I cannot follow the choice of the authors to discuss the groups of bins 1-3, 4-12 and 12-23 separately. Can the motivation be explained? The bias and RMSE within the group 4-12 is anything but homogeneous.

L.593 “the most important finding is that Aeolus is not capable to reproduce satisfactorily the backscatter profiles” I find this a bolt statement to make here. It is not Aeolus but specifically the current Aeulos SCA product in absence of cloud flagging. The cloud-flagged observations, presented some lines thereafter, let you draw a very different conclusion!

L.600-603, point ii); It is not motivated how increased noise causes bias in backscatter coefficients (I assume that “overestimation” is used synonymously to bias, if so, please use “bias” throughout the manuscript whenever appropriate). This is explained in Sec. 4.1 & 4.2 in Ehlers et al. (2022), so maybe reference here as well?

L.615 It should be already explained here, that this low positive bias is due to omitted negative backscatter values (this can be seen in the scatterplots), as you do later in L.662-665. To my knowledge, the corresponding L2A processor parameter has been adjusted so that negative backscatter values in SCA midbin are not just omitted in the newer baselines!

L.622-625 It should be made clear that this statement regards only the bin closest to the surface. Also, “level” should be replaced by “bias”.

L.644-646 I would not use the word contradiction. I can simply not be said based on metric 1,2 and 3 which product is “better”. However, it should also be mentioned that the SCA midbin scatterplot contains less data due to the inherent flagging of negative values, see scatterplots, and hence the analyses are not strictly comparable.

Also, the discussion whether SCA backscatter or SCA mb backscatter is “better” depends simply on SNR, as has been addressed in my comment on L.530.

L.672 please use “bias” as in Table 1, instead of “overestimation”, in order not to confuse the reader whether or not these are two different statistical properties.

Table 1&2; provide units!

L.668-678 This paragraph describes the statistics of the unfiltered data in detail. At this point, it has already been made clear to the reader that the unfiltered data is not suited for statistical analysis. Hence, Table 1 may as well be moved into the Appendix and may be kept for the sake of completeness, including a hint in the text. Instead, a bit more detail about the metrics for the filtered backscatter profiles would be appropriate in L.678-681.

L.722 What does “they” refer to?

L.748 It's rather “SCA backscatter coefficients” to be specific.

L.763 Please specify and write “Aeolus’ SCA backscatter product” instead of “Aeolus”.

L.781-790 Partly repeats the cross-polar misdetection mentioned above in L.757. Also, this paragraph does not contain a conclusion from your analysis seems detached. It resembles more of an Aeolus-2 future mission outlook? Maybe move into a separate section “outlook”, if the information is crucial in your opinion?

L. 793-797 The content of this text should be moved into Section 3, since these products have been released already at the end of March 2022. Specifically, it needs to be clarified that not only the extinction but also the quality of the backscatter coefficients (especially precision) is significantly increased with the Maximum-Likelihood Estimation (MLE), making new Cal-Val studies worthwhile once there processed data is available.

L.804-810 This also reads as a mission outlook rather than as a part of your conclusion and may be dropped or moved into a separate section “outlook”.

Technical corrections

L.230 Refer to the “C coefficients” as cross-talk coefficients as above. In general, using words as “so-called” and setting words in quotation marks should be avoided. It suggests little reliability.

L.241 “downwards”, same comment as in L.230

L.343 The formulation seems odd. Just write “in Section 5” and omit the part in parentheses.

L.369-370 This sentence is wordy/bulky. Better: “The Aeolus L2A backscatter profiles are compared to the measurements of three PANACEA lidar stations.”

L.384 replace “rest” by “remaining”

L.481-483 The information in the parentheses is different from the information in the text (SCA vs. Ground and Ground vs. Aeolus-like Ground observations).

L.586 This should be Fig. 4 instead of 5.

L.653 replace “not any” with “no”

Ref list: Ehlers et al. (2022) is not included though cited in the text?

Gkikas et al. compared the Aeolus L2A particle backscatter coefficient retrievals with ground-based lidar measurement in Greece. The authors showed Aeolus SCA and GRD backscatter profiles for 4 cases and statistic assessments for 46 collocated cases. It is not clear if the 4 cases are representative for the SCA backscatter coefficient product. For the statistic assessment, the authors showed that the SCA (SCA mid bin) cloud filtered backscatter profiles have better agreement with the GRD backscatter profile than the unfiltered profiles. The authors used AERONET, CAMS, MERRA-2 aerosol data to describe the aerosol situations for the 4 cases but not compared AOT from the auxiliary data with L2A. It may give readers a feeling that more auxiliary information than the Aeolus L2A data is used in the paper. The paper is well-written, good structure and lots of references. Some long sentences can be rewritten to make the paper easy to read.

Specific comments

Abstract

Line 27 Change ’hydrometeors’ to clouds. I think hydrometeor is too broad here.

Please provide the L2A data version (Baseline) in the abstract, because there are different L2A versions available.

It would be nice to provide some numbers in the abstract.

Introduction

It is impressive that the authors have cited so many papers throughout the manuscript.

Line 285 ‘lat = 35.86 N, lon-23.31 E’.

The degree symbol is missing. Please check the texts with ‘lat=, lon = ‘ throughout the manuscript.

Line 307 ‘...at 354 and 532 nm…’

Is it 354 or 355 nm?

Sect. 5 collocation between Aeolus and ground-based lidars.

It is not clear how the Aeolus and ground-based lidar are matched in altitude bins. Could you explain it in the texts?

Sect. 6.1 results

Please explain why these 4 cases are selected. Are they the best cases?

Sect. 6.2

Lines 554 – 555. Please move this sentence to the earlier section. It is important to know the L2A data version.

Lines 576-577 ‘… the GRD profiles have been rescaled to match Aeolus vertical product resolution’

How is the rescaling performed?

How many Aeolus profiles are used in the statistic assessment? Later I saw it is in the figures.

Lines 672 -673.

Units are missing after the values.

Line 796

‘… and the EarthCARE derived AEOL-FF and …’

Change AEOL-FF to AEL-FM.