GENERAL COMMENTS

The authors have applied the BOREAS retrieval algorithm to obtain vertical profiles of aerosol extinction for a substantial data set of two different MAX-DOAS instruments in the Vienna area. They assessed the quality of these retrievals by comparing the profiles, the integrated AOD and the near-surface values with extinction profiles derived from a ceilometer, AERONET measurements and PM10 measurements respectively. The data set was large enough to be able to differentiate between the seasons. In addition they demonstrate that the two MAXDOAS instruments can be used to study the spatial variability of NO2 and aerosol over Vienna.

The paper is very well structured and clearly written. This is the first assessment of the BOREAS aerosol retrieval method with a large dataset (two instruments, and almost two years of data), and with multiple colocated comparison instruments, making it an important study. 

What is missing is a discussion of the issues of this particular retrieval method found in earlier studies (Boesch et al, 2018, Tirpitz et al, 2021, Friess et al, 2019): are these issues solved, can they be confirmed or disproven?

I am disappointed about the sole use of linear correlation coefficients to determine the quality of the retrievals, see specific comments below.

SPECIFIC COMMENTS

The use of a linear correlation coefficient to compare vertical profiles is not obvious to me. Correlation coefficients are typically used to assess a possible linear relationship between two datasets. Do we expect a linear relationship? And if so, between what and what? It is unclear how the corellation coefficient for vertical profiles is defined in this paper. Is it defined with respect to the average profile? And are relative or absolute differences used? Do all altitudes have equal weight in this definition? Please clarify how to interpret the correlation coefficient for vertical profiles.

A more common, and more informative way to compare profiles is to look at the absolute and relative difference profiles, and quantify the differences for certain altitude ranges. Please add this to your study.

 

Please clarify the differences in sensitivity and averaging kernels between the BOREAS retrievals on one hand and the ceilometer retrievals on the other hand, and whether these differences are expected to influence the comparison.

 

It is interesting to see that the observed average profiles often show an increase towards the lowest point, both for MAXDOAS and for Ceilometer (e.g. Fig. 2, fall, 10-16UTC, spring 8-14 UTC, and even summer 8-16 UTC). Please check if there is not an intrinsic different treatment for the lowest altitude (e.g. other vertical extent, or different interpolation). If not, can you elaborate on the possible reason why this is so often seen? 

Several average MAXDOAS profiles in Figure 3 (visible) show elevated aerosol between approximately 2 and 3 km altitude, while this is not seen in the ceilometer profiles. In Figure 2 (UV) this is much less pronounced. Please give an explanation, and also how this effects the observed differences.

The regression coefficients printed in Figures 2-6 have the wrong number of significant digits. The offsets have too many, and the slopes too few. Please use the number of significant digits that is justified by the uncertainty of the coefficients. Also do not print '+-', but rather '-'.

Figures 7 and 8 are very difficult to assess. I recommend adding a more map-like display. For example a figure with four subpanels for NO2, AOD, and the near-surface values, where each subpanel represents the spatial area considered and the values from fig 7 and 8 can be shown as colored bars (or parts of a circle) in the respective viewing directions, where the length of the stripe is somewhat representative for the area observed by the MAXDOAS.The use of 

p2, l19: "widely documented": please give references here

p6, l11: An elevation of 0 degree with a FOV of more than 0 will result in a vegetation signal. Is this measurement used in the retrieval? 

section 2.2.1: 

- second step to 'pre-select': what are the pre-selection criteria?

- third step: this is very complex; please add a figure illustrating the procedure.

- please explain what you mean by 'second-order difference'

p10, l8: How large is the effect of using measured p,T-profiles instead of US standard profiles? 

Section 2.2.3: the profiles are scaled by the AOD: do you mean that the profiles are scaled so that the AOD derived from the profile matches the AERONET AOD at a specific wavelength? It is not clear to me why you need the intermediate step at 910nm

Section 3.1 Only sucessful retrievals are evaluated here, which makes sense. However, it is also interesting to know how often the retrieval does not succeed (does not match the criteria). Please give numbers.

p13, l15: I don't fully agree with this assessment. It seems that the visible retrieval tends to result in higher aerosol levels at higher altitudes, which might be the reason for the worse correlation.

TECHNICAL CORRECTIONS

p2, l1: "Hile high correlation" should probably be "High correlation"?

p8, l11: "sza is taken from retrieved maxdoas data" is better changed to " sza is taken at the time of the maxdoas measurement"

p9, l29: is SCRIATRAN implemented in BOREAS or BOREAS in SCIATRAN?

p12,l27: intervalls -> intervals

 

p14, l12-17: this sentence is too long, please split in two or three sentences.

General Comments

The manuscript entitled “Evaluation of UV-visible MAX-DOAS aerosol profiling products by comparison with ceilometer, sun photometer, and in situ observations in Vienna, Austria” by Schreier et al. presents vertical profiles of aerosols retrieved from Multi-Axis DOAS. The MAX- DOAS observations are compared to co-located measurements of particulate matter, AOD from sun photometer, and backscatter profiles from a ceilometer.

Aerosols play a crucial role in the atmospheric system. They affect air quality, have an impact on radiative transfer, and provide surface areas for chemical reactions. Aerosol vertical profiles from MAX-DOAS provide a valuable contribution to the understanding of the role of aerosols in the boundary layer. Therefore, this manuscript fits well into the scope of AMT.

In general, the manuscript is well written. However, some aspects of the methodology are unclear and important information is missing. In particular, the discussion mainly focuses on regression coefficients between MAX-DOAS and co-located data, but it is not always clear which quantities are actually compared in the regression analysis (see Specific Comments). While regression coefficients provide information only on the precision of the measurements, there could be more emphasis on the discussion of slope and intercept of the linear regression analysis in order to assess the overall accuracy.

Smaller correlation coefficients between surface extinction/AOD from MAX-DOAS and co-located instruments are found during summer. The authors conclude that this is due to a poorer performance of MAX-DOAS during this season. I cannot really see any reason for this. Are the DSCD errors higher or is the information content lower in summer (this would require a discussion of the averaging kernels, see my comment below)? When looking at the right panels of Fig. 2 and 3, the smaller correlation coefficients instead seem to be caused by aerosols being less abundant during summer than in other seasons, leading to poor statistics. Furthermore, there is a single outlier in the Vis data (lower-right panel of Fig. 3) that is likely to have a strong impact on the slope of the regression line.

Apart from some general remarks on the vertical range of the retrieved profiles, a discussion on the vertical sensitivity based on averaging kernels is missing. At least some examples of averaging kernels for different atmospheric scenarios should be presented and discussed. Due to the limited vertical resolution of the MAX-DOAS extinction profiles, a quantitative comparison between MAX-DOAS and Ceilometer profiles requires the convolution of the high-resolution Ceilometer profiles with the MAX-DOAS averaging kernel according to Rodgers and Connor (2003) (see, e.g., Frieß et al., 2016 and Tirpitz et al., 2021). This is of particular importance because the vertical sensitivity of the aerosol profiles retrieved by the BOREAS algorithm appears to be limited to the lowermost 500 m (Fig. 9 in Bösch et al., 2018), which means that large fractions of the aerosol column are invisible for the MAX-DOAS instrument.

Specific Comments

Abstract L20-23: It is not correct that coincident measurements of temperature and pressure profiles were used here for the first time for profile inversion, see e.g. Friedrich et al. (2019), who used daily radiosondes.

P6, L15: Do you use the 0° elevation measurements for the profile retrieval? I could imagine that this leads to difficulties in the inversion, either due to blocking by surrounding buildings or trees, or due to the fact that the field of view includes both atmosphere and surface.

P8, L12: Mean vertical profiles of which quantities are used as input parameters for the RTM?

I suggest to structure the sub-sections of Section 2.2 in the order of their importance, starting with the MAX-DOAS profile retrieval, and to move Section 2.2.1 on cloud flagging to the end of section 2.2.

P8, L16: It is not clear to me what the term ‘Hence’ refers to. The term ‘Differential slant column’ is not defined yet and should either be explained here, or replaced by ‘measurements’.

P8, L19-24: The technical description of the pyranometer should be moved to section 2.1.

P9, L4: What do you mean with ‘daily total second-order difference’? Is this the mean of the second-order differences? Please clarify.

P9, L25: The term ‘oxygen dimer’ should be avoided; O₄ represents the O₂ collision complex.

P10, L7: Again, this is not the first time that measured atmospheric profiles of pressure and temperature from a co-located site are used for MAX-DOAS profile retrieval, see general comments.

Section 2.2.2: An error discussion regarding the retrieved aerosol profiles is completely missing. Retrieval errors and vertical resolution based on averaging kernels, as well as information content should be discussed here. A discussion of averaging kernels is particular importance because  the lack of sensitivity for aerosols at high altitude might explain parts of the discrepancies between MAX-DOAS and ceilometer, and because the ceilometer profiles should be convoluted with the MAX-DOAS averaging kernels in order to perform a quantitative intercomparison between both data sets (see general comments).

P11, L7: In which way has time been extracted from the backscatter profiles? Do you mean the time stamp of the profiles? This would not really be worth mentioning.

P11, L13,19 and 22: I think the term ‘assimilation’ is inappropriate here because it has a well-defined meaning in atmospheric science, namely to adapt a modelled atmospheric state to observational data in a statistically optimal way. Maybe gridding the data (in time and space) would be a more appropriate term.

P11, L15-19: Please add a sentence motivating why the temporal averaging has been done in such a quite complicated way, instead of just averaging over the duration of a MAX-DOAS scan.

Please describe how you are dealing with the missing aerosol information from the Ceilometer in the lowermost 50 m, where MAX-DOAS is most sensitive and variability in aerosol extinction is probably highest. What kind of extrapolation did you apply for the calculation of the extinction in the lowermost retrieval layer and for the determination of the AOD?

P12, L25: It is not clear what kind of data has been used for the calculation of the correlation coefficients. Did you correlate extinction at all heights, or just at the surface? Seasonally averaged or individual profiles?

P14, L11: Here it is stated that there is limited sensitivity of MAX-DOAS aerosol profiles above 4 km. When looking at Fig. 9 in Bösch et al. (2018), the sensitivity rather seems to be restricted to the lowermost 500 m only. Again, this means that a convolution of the Ceilometer profiles prior to the comparison is crucial – see general comments.

P16, L3: Here you distinguish between the availability of total columns and surface values. Shouldn’t the total column always be available if the surface value is available and vice versa, since both are derived from the according vertical profile?

Technical Corrections

Title of 2.1.4: ‘In situ’ -> ‘In situ aerosol measurements’

P9, L23: Remove ‘Briefly’ from the beginning of the sentence.

P9, L26: Remove comma after ‘thus’.

P12, L8: A threshold in difference between modelled and measured O4 DSCD of 1000 molec²/cm⁵ is extremely small, given that typical O4 DSCDs are in the order of 10⁴³ molec²/cm⁵. Is this a typo?

References

  Rodgers, C. D. and Connor, B. J.: Intercomparison of remote sounding instruments, J. Geophys. Res, 108(D3), 4116–4229, doi:10.1029/2002JD002299, 2003.

In this paper, aerosol extinction (AE) profiles, aerosol optical depth (AOD), and near-surface AE are retrieved from MAX-DOAS measurements acquired on cloud-free days during the September 201-August 2019 period at two stations in the vicinity of the Vienna (Austria) city centre. These retrievals are performed using the Bremen Optimal estimation REtrieval for Aerosols and trace gaseS (BOREAS) algorithm and are evaluated against co-located ceilometer, sun photometer, and in situ instrument observations covering all four seasons. The retrieved AE profiles are found to agree well with those from the co-located ceilometer in fall, winter, and spring, with correlation coefficients ranging between 0.85 and 0.99. During those seasons, a good agreement is also obtained with the ceilometer for the near-surface AE and between the MAX-DOAS and sun photometer AODs. The MAX-DOAS retrieval results appear to be less reliable in summer and the possible origins of the lower performance of BOREAS in those conditions are discussed. Finally, the spatial variability of AOD and near-surface AE over Vienna is assessed by analyzing the retrieved BOREAS aerosol profiling products in the different azimuthal pointing directions of the two MAX-DOAS instruments and for the different seasons.

This paper is well written and clearly structured and presents interesting results which fit well with the scope of AMT. I recommend the final publication of the manuscript after addressing the following major and specific comments:

Major comment: To my opinion, the present study suffers from two weaknesses: no uncertainty budget is presented for any of the MAX-DOAS retrieved quantities (AE profile, near-surface AE, and AOD) and there is no estimation and characterisation of the vertical sensitivity of the MAX-DOAS AE profile retrievals through the calculation and examination of the averaging kernels and corresponding DOFS. The uncertainties on the Ceilometer AE profiles and other ancillary data deserve also to be discussed. Both aspects (uncertainty and vertical sensitivity) of this major comment should be addressed in the revised manuscript.

Specific comments:

Page 3, lines 11-13: I think here it is worth explicitly mentioning that a large part of the sun photometer measurements effort is endorsed by the AERONET network. The AERONET http link could be also added. Then, no need to add the http link again on page 7, lines 7-8.

Page 10, line 20: the choice of the a priori scaling height (1.25km) should be justified here. Is it based on a sensitivity study using different a priori scaling height values?

Page 12, lines 7-9: The level of agreement between measured and simulated O4 DSCDs is used to select valid MAX-DOAS AE profile retrievals. I guess that the criteria used (absolute and relative difference smaller than 1000 molec2 cm-5 and less than 10%, respectively) is applied individually to all the elevation angles. I think this should be explicitly mentioned in the text.

Page 12, lines 13-14: It is said that temporal changes in pressure and temperature can affect the BOREAS retrieval. This general statement requires some explanation: how large can be this effect on a daily basis since daily atmospheric temperature and pressure profiles are used as input for the AE profile retrievals?

Page 13, lines 13-16: the higher correlation coefficient values between MAX-DOAS UV and ceilometer AE profiles is explained by the fact that in the UV, the MAX-DOAS instrument probes air masses closer to the ceilometer which is located at 2.25km from the MAX-DOAS. Is it a valid argument since the effective horizontal distance representative of the MAX-DOAS measurements of O4 in the UV can be as large as 10-15km or even more under clear-sky conditions? Was there any attempt to estimate the effective horizontal distances for both the UV and VIS channels?

Page 13, lines 25-29: According to the authors, BOREAS has difficulty in retrieving AE profiles and near-surface AE during summer. One possible reason is that during that season, AE profiles have a box-like shape and are therefore not well retrieved with the exponential a priori used. Did the authors make a sensitivity test using a priori profiles with box-like shape in order to see whether it improves the AE profile retrievals in summer? If it significantly improves their retrievals but also the agreement with the ceilometer and the sun photometer, I think the authors should consider to include such a sensitivity test in their manuscript.

Page 14, lines 9-17: According to the authors, lower BOREAS AODs are expected because of the limited sensitivity of MAX-DOAS profiling for higher altitudes, while AERONET AODs better represent elevated aerosol in the free troposphere (and stratosphere). The authors then argue that in Spring, Saharan dust events over Austria could potentially explained higher AERONET AOD, mentioning the detection of such events in the Austrian Central Alps. I think this explanation is very speculative and requires further investigation: for instance, have such events been detected during the period where the MAX-DOAS measurements presented in this study were performed? If yes, is it statistically significant, i.e. during how many days Saharan dusts stayed above the Vienna area?

Technical corrections:

Page 12, line 27: ‘intervalls’ -> ‘intervals’

Page 14, line 12: ‘tropsphere’ -> ‘troposphere’