|Review for Atmospheric Measurement Techniques - Revised Manuscript|
Title: Retrieval of UV-Visible aerosol absorption using AERONET and OMI-MODIS synergy: Spatial and temporal variability across major aerosol environments
Authors: Vinay Kayetha, Omar Torres, Hiren Jethva
The revised version of this manuscript is much improved from the original, with a more complete section on the estimated uncertainty in the SSA retrievals from the new satellite-AERONET algorithm and better and more complete references to the published literature. However, the uncertainty in AOD is underestimated in the calculations of SSA uncertainty since the AERONET point source uncertainty is assumed for the entire 50 km radius area of the input satellite data. Therefore the authors have assumed exactly homogeneous AOD over a 100 km diameter circle on earth, which is physically unrealistic. The authors should address this issue in a second revision. More details on this issue are given below in my specific comments. Also a related problem is the computation of the uncertainty in AAE in this revised manuscript. Again, the uncertainty in AOD is assumed to be zero and the authors have only accounted for the effects of uncertainty in SSA on the computation of AAE uncertainty. This is particularly important for the AAE(354-388 nm) since a small error in spectral AOD can cause a large error in AAE for such a narrow wavelength interval. More details are given below in specific comments. This aspect of uncertainty in AAE also needs to be addressed and discussed in a 2nd revision of the manuscript.
Other issues that the authors should address in a revised manuscript are given below in ‘Specific Comments”:
Line 34: Please add this after ‘almucantar plane’: (plus hybrid scans to lower solar zenith angles)
Line 201: Why not use OMI measurements of ozone or a realistic latitude dependent climatology of ozone? Did you show that ozone amount does not matter in the retrieval? Did you use NO2 measurements from OMI, or what NO2 amount did you assume in the RTM?
Line 229-231: This spatial and temporal averaging would certainly increase the difference in AOD between the AERONET point measurements and the AOD that exists in the 50 km radius plus 2-hour difference. Additionally, the delta in AERONET AOD versus the actual AOD in the satellite pixels will increase as a function of increasing AOD since AOD in general becomes less homogeneous in space and time as AOD increases.
Lines 242-243: These Angstrom Exponents are not computed from only two wavelengths as suggested by the authors. These are computed from 3 to 4 wavelengths of AOD with linear fit in logarithmic coordinates. The first three are 3-wavelength values (i.e. 380-500 uses 380, 440 and 500 nm AOD data) while the 440-870 AE uses the 440, 500, 675 and 870 nm AOD data to compute the Angstrom Exponent.
Lines 259-260: It should be noted that urban aerosols have a wide range of absorption, this is not exclusively a weakly absorbing category (see Dubovik et al. 2002 and Giles et al. 2014). Likewise, biomass burning (or carbonaceous) aerosols exhibit a very wide range of absorption (see Giles et al 2014 and Eck et al. (2003 GRL)), depending largely on the relative contributions of the two phases of combustion (flaming and smoldering). There is extensive overlap in absorption between the two categories of urban and biomass burning. This needs to be discussed in this manuscript as the labels of urban and biomass burning (sometimes called ‘carbonaceous’ in your paper) often do not make sense in the way your classification system works.
Lines 269-272: Therefore this is a climatology of aerosol layer height and should be clearly stated as such here. Climatological versus actual ALH can vary significantly for any given observation. Was this variation from climatology quantified in the uncertainty of ALH? It would also be expected to vary regionally, as some regions have greater variance in ALH both seasonally and day-to-day.
Lines 286-287: Six-hour surface pressure from NCEP/NCAR reanalysis at 2.5 degree lat-long spatial resolution is interpolated to each AERONET site location and altitude and is provided with the AERONET files of AOD. It would have been much more accurate to have used those values of surface pressure rather than compute it from station altitude.
Line 360: Why stop at AOD(440)=0.4? It would be useful to show estimates at higher AODs also. The uncertainty in SSA retrieval will decrease significantly at higher AOD levels.
Line 371-373: This is the uncertainty in measured AOD at the AERONET site. However you have used input satellite data over a 50 km radius and +-2 hour interval from the AERONET site location. Therefore the variability in AOD over space and time certainty exceeds the point measurement uncertainty at an AERONET site by about a factor of ~50% to 100%. I therefore believe that you have underestimated the uncertainty in your SSA retrievals due your assumption of AOD uncertainty that is not representative of a 100 km diameter satellite average AOD.
Lines 394-395: It is well known that fine mode particle size increases as AOD increases in many regions due to aging processes of coagulation and condensation (see Dubovik et al. 2002; Eck et al., 2010; Eck et al., 2012). Therefore your errors due to the use of climatological size distribution averages will be biased as a function of AOD.
Lines 414-415: Are all pixels assumed to be cloud contaminated or a fixed percentage of pixels assumed to be cloud contaminated in these calculations?
Lines 422-423: I cannot agree with your statement of minimal absorption in the UV from trace gases since NO2 absorption peaks at 380 and 440 nm. Also NO2 column abundance varies tremendously across the globe and also seasonally. In winter in East Asia (China and South Korea) the NO2 amounts are very high and result in significant absorption in the UV and 440 nm. It appears as though you are basically computing SSA due to aerosols plus NO2 in eastern China and Korea thus overestimating aerosol absorption in these regions. AERONET utilizes a global monthly climatology of NO2 at 0.25 degree resolution derived from OMI data in order to correct the AOD and sky radiances for NO2 absorption effects. This bias in your SSA retrievals which are maximum in China and South Korea need to be discussed in the text.
Lines 456-457: However it seems like you ignore this significant component of atmospheric variation in pressure due to meteorology in your calculation of uncertainty.
Lines 477-479: It seems you have neglected a significant source of uncertainty in your computations of AAE uncertainty. The uncertainty in AOD is also a significant factor especially when the wavelengths are close together such as for AAE(354-388 nm). Therefore your uncertainty estimates can be considered minimum values since it has been assumed that spectral AOD have zero error in your computations.
Lines 496-498: It can be expected that this is the more typical situation since the SSA retrievals are independent at each wavelength. Therefore I would expect the AAE uncertainty to be very large, much higher than your previous analysis that assumes the same bias in SSA in both wavelengths.
Line 508: What uncertainty in AAE do you assume here?
Lines 526-527: Any ideas on why the difference is so large for the Lake Argyle site? Small sample size or surface reflectance uncertainty?
Lines 529-531: However you should also note in the text of this manuscript that the surface reflectance is a relatively small source of error in the AERONET retrievals with upward viewing sky radiance measurements.
Line 540: The Figure 9 y-axis labels need to be clarified and/or changed, since it is not possible to know what is being plotted without reading the text first. The current y-axis labels just give a wavelength and a satellite name, so it is impossible to interpret by itself.
Lines 545-547: This could be partially explained by more sensitivity to the variability in particle size for fine mode aerosols coupled with significant departures on some days from the climatological values used in the retrievals. For dust there is much less sensitivity to particle size.
Line 565: However the surface reflectance is only a significant source of error in AERONET for low AOD magnitude, ~ <0.2 at 440 nm.
Line 567: Add this after almucantar plane: (or hybrid scan)
Line 569 & 571 : replace 'weak' with 'relatively strong' in both lines
Line 578: I suggest that you keep consistency in your labeling/categorizing of the aerosol type that you often call 'carbonaceous'. Immediately below in section 6.1 you call this type 'biomass burning'. It would be clearer to the reader if you consistently used the term 'biomass burning' throughout the manuscript.
Lines 596-597: September is also a month of significant biomass burning smoke in Missoula, while June typically has a very minor amount of smoke. Please correct this statement.
Lines 606-607: Should include country for each site name i.e. Brazil and Bolivia in this case.
Lines 611-612: Are these UV values of SSA for the JJA or SON months? Please clarify this sentence.
Line 619: Please include the country names: Zambia and South Africa.
Line 634: It is well known that there is always some dust present in the Sahel and Sudanian zones in the dry season. This is the reason for the relatively low AE of 1.3 since these are mixtures of fine and coarse mode particles. The presence of dust is the reason for the relatively flat spectral SSA at Ilorin. If these were all fine mode biomass burning particles with much black carbon then the SSA would decrease with increasing wavelength.
Lines 646-648: Cairo is a very large city, metropolitan area population of 21 million, with many emissions from industry and traffic. It is well known for very high levels of pollution from industry and vehicles. It is not possible to isolate the properties of the aerosol from agricultural burning alone. Mixture of biomass burning plus urban aerosols is inevitable. Please convey this in the text. It is quite odd to even include this site in the Biomass Burning section of this paper.
Line 653: Here is another example of the confusion that your 'carbonaceous' aerosol classification causes. These are dominantly urban/industrial aerosols at Beijing and XiangHe, not predominantly biomass burning aerosols. Your inclusion of this site under the section "6.1 Biomass Burning" is wrong and therefore very misleading.
Line 655: Do you mean these are AAE values here, if so then clearly state it.
Lines 662-663: Please also mention in the text the other significant aerosol sources in the Indo-Gangetic Plain region in northern India such as brick kilns that burn coal and therefore emit much black carbon, power plants that burn coal and also heavy vehicular traffic in the cities such as Kanpur. You give the impression in the manuscript that there is only biomass burning going on in the region which is both false and misleading.
Lines 754-756: Also it is likely that a much smaller sample size in MAM contributes to the difference with JJA since a few unusual cases in spring season may significantly affect the average.
Lines 764-768: Please note that NO2 absorption has not been adequately accounted for in your retrievals in the urban regions where NO2 column amounts are highest. Therefore it is important that you mention that your retrieved SSA are likely biased low especially at 380 and 440 nm where the NO2 absorption is highest. Maps of NO2 from OMI and TROPOMI clearly show high NO2 amounts over urban regions therefore affecting all of your retrievals for all urban sites, but especially so for China.
Lines 785-786: This probably due to biomass burning aerosols in SON mixing with urban aerosols, not due to a change in the urban aerosol absorption as you seem to be implying. Please rephrase this sentence.
Lines 793-794: The AE=1.3 strongly suggests a mixture of fine and coarse mode aerosols. Even 10-20% of the AOD from coarse mode particles (soil dust, etc) can result in substantial flattening of the SSA spectra. This seems a more likely explanation than your suggestion of a mixture of black carbon and organic carbon, which does not seem to make much sense since both are fine mode particle types.
Lines 796-797: Averaging 13 sites together in Europe over a vast geographic area is not a very rigorous approach. In fact you point out at the end of this paragraph that three of these sites apparently had significantly higher absorption in SON than the other sites. I would suggest some further discussion of the range of SSA values over these 13 sites.
Line 805: change 'observing' to absorbing' here
Line 806: There is always a mixture of organic and black carbon from fossil fuel combustion, so this sentence essentially tells us nothing.
Lines 812-813: Yes, aerosol humidification is summer results in a large shift in the fine mode particle size to larger particles relative to winter. Since you only apply a yearly mean aerosol size distribution it seems likely that you are underestimating the winter-summer difference in SSA since the larger particles in summer scatter light much more efficiently.
Lines 840-841: This is incorrect. The cerrado vegetation type dominates as a source of biomass burning aerosol only at the Cuiaba site, not 'at most sites considered here'.
Lines 853-854: It should also be mentioned here that the uncertainty of AAE for such a narrow wavelength interval of 354-388 nm is very high. It is higher than your estimated values in Section 3 since you assumed that AOD in both wavelengths was perfect (no error in AOD was assumed).
Line 861: This can be true for China where the fine mode particle size is very large due to aging and humidification processes (thereby reducing the AE value) but for the Sahel the reason for the AE is mixing with coarse mode dust.
Line 940: I cannot see a valid justification for such a large and polluted megacity as Cairo to be included in the Biomass Burning section. Even when biomass burning occurs near Cairo there is certainly a mix ox aerosol types since the urban aerosol sources remain strong producers of aerosol throughout the year.
Lines 948-950: Again I find that sites in NE China are well known to be dominated by urban/industrial pollution. Including this Chinese region in the Biomass Burning section is very problematic. Your classification system between biomass burning (that you call carbonaceous half the time) and urban is dubious at best. This ambiguity in these two classifications needs to be discussed in the text so that the reader can be aware of the large overlap between these two aerosol types in your analysis. Even if there is a month or season with biomass burning in the NE China region the overall aerosol could only be described as mixed since the urban aerosol loading is still very high.
Line 968: Replacing 'carbonaceous' with 'biomass burning' here would be clearer for the reader, especially since you sometimes classify urban aerosol as carbonaceous (see comment above).
Line 973: Please provide urban site names here.
Line 975: Also provide site names here.