Review for Atmospheric Measurement Techniques

Title: Assessment of NO2 uncertainty impact on aerosol optical depth retrievals at a global scale

Authors: Akriti Masoom, Stelios Kazadzis, Masimo Valeri, Ioannis-Panagiotis Raptis, Gabrielle Brizzi, Kyriakoula Papachristopoulou, Francesca Barnaba, Stefano Casadio, Axel Kreuter, Fabrizio Niro

General Comments:

This is an interesting and potentially useful paper on the biases in AERONET computed AOD due the application of climatological monthly averages of nitrogen dioxide (NO2) from OMI satellite data versus coincident in time accurate measurements of column integrated NO2 from ground-based Pandora instruments.  However, there are several significant issues (listed as 1-5 below) that the authors need to address before this manuscript is published.

(1)   First, the manuscript title suggests an assessment on a global scale when in fact there are no sites analyzed in either Africa or Australia and only one site in the entire continent of South America (as shown in Figure 1). Except for 8 sites out of the 33 investigated all are in three regions: western Europe, eastern half of North America, and northeastern Asia. Therefore the analysis cannot be considered global. Additionally, it is noted that more than two thirds of the station pairs analyzed in this study (Table 1) are in urbanized regions or in cities that would have significantly higher NO2 than rural sites (or small cities). It would be very useful to separately analyze the large urban and/or industrial region sites versus rural site data since the impact of accurate collocated NO2 data from Pandora on AERONET AOD will clearly be much more significant for the sites in urban/industrial regions versus the rural sites. It is unlikely that ~70% of all AERONET sites in the entire network (not just those collocated with Pandora) are located in urban/industrial regions therefore separate analysis of these two categories of regions would be important and valuable. For simplicity I suggest possibly including small cities that are adjacent to rural land or ocean as 'rural' therefore Boulder and Comodoro would both be rural in that that definition. In my opinion other sites in the rural category would be Dalanzadgad, Davos, Innsbruck, Izana, Lindenburg, Ny-Alesund, and Wallops. Even though Julich is not a high population density place it is still in an industrialized region therefore I would not categorize it as rural.

(2)   It is important to state in this paper that if PGN data were used to correct AERONET data then there would be discontinuities in the AERONET time series of AOD in both space and time since PGN data are not available for most years and most sites. Approximately 5-10% of AERONET sites currently have co-located PGN data and this decreases to 0% at the time before Pandora instruments existed and/or data are available.

(3)   Another aspect that needs to be emphasized in this manuscript is which AERONET measurement wavelengths are significantly affected and which are not affected by biases in column NO2 amount, since NO2 absorption does not impact all wavelengths equally. The AOD differences at AERONET measurement wavelengths other than 380 and 440 nm should also be given somewhere in this manuscript. If these are relatively small differences, then perhaps a table could provide the range of differences in AOD that occur when using the accurate PGN data instead of the OMIc values for NO2. The AOD difference values at 340, 500, 675, 870, 1020 and 1640 nm should be provided in this paper at least in summary form.

(4)   Regarding another important issue, you state on line 197: “…here we use 380–675 and 440–870 wavelength pairs for AE estimations”. Note that the 2 wavelength computations of AE (that you have suggested are utilized in this paper) differ from the multi-wavelength computations of AE provided by AERONET. For AE(440-870 nm) the AERONET computation uses the 440, 500, 675 and 870 nm AOD and computes it from the linear regression in logarithmic coordinates using all 4 wavelengths. Your two wavelength computation of 440-870 AE gives more weight to the 440 nm AOD which has large NO2 optical depth and therefore accentuates the AE change due to NO2 variability versus the AE changes that would occur in the actual AERONET product of AE(440-870) with 4 wavelengths input. The AE in this manuscript should be recomputed using all AOD within the wavelength range in order to provide an accurate estimate of the changes to the standard AERONET product of AE(440-870). Otherwise you would need to specify in the text that for the AERONET computations of AE the changes due to Pandora input would be smaller as compared to your computations of AE with fewer wavelengths. Also note that AERONET does not compute the 380-675 nm AE as you do so this is also not an AE computed product that users would download in the AOD files from the AERONET web page. If the 380-675 nm AE values remain in the paper then you need to make this clear to the reader. All AE computations available from the AERONET web page utilize 3 or more wavelengths: all AOD values within the wavelength range specified.

(5)   Finally, it is important to know if the large differences in NO2 between PGN and OMIc occur at high levels of AOD especially for stations such as Dhaka, Mexico City, Beijing, Seoul and Athens. Scatterplots of AOD(440) versus delta(AOD 440 nm) due to NO2 differences (PGN versus OMIc) for each station individually would provide important information about the relative changes in AOD and not just the absolute differences in AOD that are currently provided in the paper. For example it is important to know if the largest NO2 biases (when applying OMIc) occur at the highest AOD levels and also if AOD(440) nm is correlated with total column NO2 magnitude.

Specific comments:

Line 20: AOD data are more accurately described as measurements, not retrievals.

AOD is more of a direct measurement by sunphotometers as distinguished from the AERONET retrievals of size distribution and complex refractive indices from the combined inputs of spectral directional sky radiances and spectral AOD.

Line 25: Please specify here in the Abstract which wavelengths are significantly affected and which are not, since NO2 absorption does not impact all wavelengths equally.

Line 112: Again these are AOD measurements not retrievals such as from the sky radiance retrievals from the Dubovik algorithm. It is surprising that you utilized L1.5 data since final calibrations are not always applied yet and therefore the uncertainties are greater than for L2 data. Please explain in the text why L2 data were not utilized in this study, as it seems that most of the data were too recent (i.e. much 2023 data) to have post-deployment calibrations. The uncertainty of the L1.5 data that do not yet have final calibrations applied is ~2X greater (depending of length of field deployment) than that of L2 data (see Figure 20 in Giles et al., 2019). Please include this information in the text since many of the station data in Table 1 are for 2021-2023 only and therefore some may not include application of final calibrations to the data processing.

Line 138: Please give the range of distances between the AERONET and Pandora instruments for the 33 selected station pairs.

Line 149: What was the maximum time difference that was accepted for the time matching? Please specify in the text of the manuscript.

Line 168-170: Note that water vapor absorption is also subtracted from the 1020 nm total optical depth to get AOD at 1020 nm.

Line 286-289: This should be supported with some trend data on NO2 in Beijing from published literature (see Xu at al., 2023 in Atmospheric Environment) and with references included in the text. Similar references should be searched for Dhaka and provide the magnitudes of the observed changes in NO2 in the text of this paper.

Jing Xu, Ziyin Zhang, Xiujuan Zhao, Siyu Cheng, Downward trend of NO2 in the urban areas of Beijing-Tianjin-Hebei region from 2014 to 2020: Comparison of satellite retrievals, ground observations, and emission inventories, Atmospheric Environment, Volume 295,

2023, 119531, https://doi.org/10.1016/j.atmosenv.2022.119531.

Line 343-344: You had suggested earlier in the manuscript that the AE(440-870) and AE (380-675) were both computed from 2 wavelengths. However in the AERONET products the AE are computed from 3 or more wavelengths plus the 380-675 nm AE is not even provided as a product from the AERONET web page. In order to be more useful to the scientific community the AE in this manuscript should be computed in the same methodology as done by AERONET and with the same wavelength limits.

Line 351-352: This should be written a little more clearly. In fact there are no PGN NO2 corrections made at 675 and 870 due to the fact that there is no NO2 absorption at those wavelengths (not just that the corrections are not made). It is important to also include the effects of NO2 biases from OMIc at 340 and 500 nm in this paper.

Line 354-356: Yes indeed, therefore it would also be useful to show scatterplots of the AE differences as a function of AOD and as a function of NO2 amount for a few of the sites with the largest biases.

Line 371-373: This section is somewhat confusing since trends are only computed using the AERONET V3 AOD and AE values. The co-located PGN values of NO2 are not available for any sites for a long enough time period to actually compute the effect of correcting for NO2 biases on trends by including PGN data. I would suggest that this section could be removed from the paper since the effect of using PGN data on trends is not possible. Alternatively, the effect of using OMId versus OMIc (daily OMI and perhaps also daily TOMS versus OMI climatology) could actually provide something of a possible correction for NO2 effects on AOD to trends in AOD and AE.

Line 434-443: This seems particularly weak to include the discussion on trends in the conclusions since no corrections for NO2 biases could actually be applied to the data due to the short duration of the available PGN data sets (as shown in Table 1).

Line 447-448: It would be important to know if these high NO2 cases are associated with high AOD and therefore a smaller relative percentage of total AOD as opposed to absolute differences in AOD which you present.

Line 475, Figure A1, caption: “The numbers in the legend represent the ratio of mean optical depth difference…”  I do not see any numbers in the legend of Figure A1, please add them or clarify.

Review of: “Assessment of NO2 uncertainty impact on aerosol optical depth retrievals at a global scale”.

The paper is an extension of the already published “Evaluating the effects of columnar NO2 on the accuracy of aerosol optical properties retrievals” Drosoglou et al. 2023, who analyzed the effect for the site of Rome.

Extending the results to more sites worldwide is very interesting and can provide very useful information.

There are however some points that need a clarification.

• The title is misleading since it seems that the authors are evaluating how “the uncertainty in NO2 estimation” impacts over AOD measurements. I would suggest something like “Assessment of the impact of NO2 contribution on aerosol optical depth observations in several site worldwide locate”. AOD is not retrieved, because there isn’t any inversion analysis to perform. Moreover “global scale” is too much for the number and location of the sites studied in the work.
• In the abstract lines 27-28 it is not clear what a “deviation in NO2” is. It is understandable reading the text, but it should be clarified also   Why the authors preferred “deviation” instead of a simple “difference”?
• Could you explain the reason why you are looking for the NO2 effect only at 380 and 440 nm?
• Line 149: “the nearest matching PGN” to AERONET. Is there any threshold within searching the nearest measurement? The nearest could also be some with some days of difference.
• In the Sections 2.2.2 please describe (or cite a reference) for explaining Eq 3. Moreover in Eq 4: 1) explain the reason of the adding and subtracting each term, and 2) what is TNO2(l) in the third term after the first equivalence. Why this term disappears after the second equivalence?  The same explanations are necessary for the equivalences in Eq5. How delta_NO2 is defined. To facilitate the reading, please do a table that summarize the three lines 184-186 (adding also the variables in Eq 4 and 5 that are not defined) and the parameters in Eq. 3.
• In general, it is better doing an acronyms table.
• Lines 199-201: Angstrom exponents using AOD corrected for NO2 from PGN are calculated using two wls. But this method is different from the AERONET one, because the latter uses all the wls inside the intervals 380-675 / 440-870 and not only the range boundaries. Therefore, they can’t be compared.
• Section 3.2: “10% highest deviation cases” do you mean that you calculated the differences among NO2 estimations, then you took the highest and then you increased (or decreased) this difference of 10% for obtaining an extreme scenarios of NO2 differences ? Please the describe better the meaning of this sentence, also in the conclusion (line 420).

Some typos errors:

The numbers of sub sessions at  lines 118 and 128 are wrong ( 2.2.2=> 2.1.2 etc ).

Line 417: “Among these, 10 stations ..” => “Among these, 6 stations…”.