Comparison of AOD, AAOD and column single scattering albedo from AERONET retrievals and in-situ profiling measurements
E. Andrews, J.A. Ogren, S. Kinne, B. Samset
The authors have extensively revised the manuscript and have responded satisfactorily to many of my comments/suggestions. However, some important issues remain which need to be resolved before publication.
Especially regarding Figure 6: Please explain how you estimated the error bar in yellow shading in Figure 6. You had stated previously in this revised manuscript that the in-situ uncertainty in SSA was 0.06 (and gave no AOD dependence for this) while for AERONET it is 0.03 for AOD(440)>0.2 and 0.05-0.07 for lower AOD values (Table 2 of Dubovik et al., 2000). How did you get SSA uncertainty in Figure 6 of 0.05 for AOD>0.20 and 0.08 for lower AOD? In the first draft that I reviewed the in-situ SSA uncertainty was estimated to be 0.04, yet the yellow shaded uncertainty limits in Figure 6 were the same as in the revised manuscript. It appears that the uncertainty shading in Figure 6 needs to be updated.
Some attempt should be made to make an estimate of in-situ uncertainty for extremely low concentrations which occur in winter at mid latitudes (such as at SGP and Bondville), for total column AOD values ranging from ~0.02 to 0.08 at 440 nm (Figure 8). It is not realistic to think that the uncertainty of the in-situ measured/derived SSA will not increase when these very low aerosol loadings occur at all flight levels, especially since the flight segments at lower levels are presumably still 5 minutes in duration. Concentrations at all altitudes on very low AOD days (<0.08 at 440 nm) would likely be similar or lower than high altitude concentrations on moderate to high AOD days, therefore the overall uncertainty of the in-situ data would likely be significantly greater for these low AOD days. The paper still lacks convincing discussion/analysis regarding in-situ SSA uncertainty on the lowest AOD days (occurring typically in the winter season), as it must be higher uncertainty than on moderate to high AOD days when aerosol concentrations at flight altitudes in the boundary layer can be an order of magnitude (or more) higher. (see Lines 399-400 & Lines 528- 532 in the revised MS for this topic/issue).
Lines 347-348: Regarding the AERONET Version 3 data, maybe say at the 'time of writing this manuscript' here.
Lines 354-355: This should be “if there was a corresponding Level 2 almucantar retrieval available, but AOD (440)<0.4.”
Lines 427-429: Please include the reference of Eck et al. (2014; ACP) here, as this paper specifically discusses the phenomenon you are discussing here, of the measurement of higher AOD in the near vicinity of cumulus clouds. This paper also shows that both lidar and in-situ measurements found similar aerosol increases as AERONET measured AOD near to cumulus clouds during this field campaign in Maryland.
Line 638: Direct sun AOD are more accurately called measurements, not retrievals (the AERONET retrievals are made with the Dubovik and King algorithm from almucantar sky radiance and spectral AOD input).
Lines 674- 675: That is not possible, as there is always aerosol above 4.2 km asl, such as mid- to upper troposphere plus stratospheric aerosol. Perhaps the word 'significant' needs to be added here, since the exact amount in AOD above the 4.2 km level is often unknown.
Lines 761- 762: Perhaps putting both Hybrid and AERONET on the y-axis label would make this plot (Figure 5) easier to interpret.
Lines 806- 808: It should be mentioned here in the text that for mineral dust there is significant absorption at 440 nm in the coarse mode from iron oxide content.
Lines 817- 824: Please add here the underestimation of coarse mode absorption at 440 nm due to dust particles that are under-sampled in the in-situ measurements.
Lines 894-898: This should be stated as greater absorption than in-situ here (rather than over-estimation) since the in-situ data is not recognized (by the scientific community) as truth or gold standard for SSA.
Lines 898-900: In the interest of completeness and unbiased reporting, the number of cases where agreement is excellent (within 0.02), given the in-situ uncertainty of 0.06, should be mentioned here in the text.
Line 907: Please change 'often' to 'mostly' here since most data shown in Figure 6 for AOD >0.2 at 440 nm are within the yellow shaded uncertainty bounds.
Lines 970 - 971: Not really true, since the almucantar associated measurements (L1.5*) of AOD are biased towards days with lower cloud cover due to sky radiance screening for clouds away from the solar disc. Additionally almucantar retrievals are only valid in the relatively early morning and late afternoon-evening in summer at mid-latitude (SZA>50, which occurs all day in winter) thereby missing the most active fair weather cumulus development times of day in summer. AOD measurements associated with more cloudy conditions are often higher due to increased particle hygroscopic growth, cloud processing of particles in aqueous phase chemistry, gas-to particle conversion that occurs more rapidly in cloud droplets, and convergence of aerosols associated with larger cloud systems (see Jeong and Li (2010) and Eck et al. (2014)).
Lines 983 - 984: This statement is not rigorously true, as noted in the previous comment.
Lines 986 - 987: These are very large and significant differences (“by up to 50%”) between in-situ and AERONET measured AOD. Please note this in the paper and attempt an explanation for the significant low summer season bias of the in-situ AOD measurements.
Lines 1107 - 1113: It would be useful to note that the fine mode volume median radius (and also fine mode effective radius) decrease as AOD decreases at all of these AERONET sites, thereby suggesting one physical mechanism for the lower SSA at lower AOD levels. Sub-micron sized particles scatter less efficiently when they are smaller, therefore even if the particle composition remains constant the smaller sized particles at lower AOD would result in lower SSA. Additionally the coarse mode fraction of AOD increases at lower AOD likely due to a relatively constant background level of coarse mode particles while the fine mode often shows much larger seasonal increases due to stagnation, humidification, cumulus cloud interaction in the summer (opposite seasonality for Fresno due to fog in the winter). The lower Angstrom Exponents seen in low AOD months at most of these sites is indicative of greater coarse mode fraction of AOD when AOD is lower. Since the in-situ instruments under-sample the coarse mode particles this can potentially be a source of bias in the in-situ determination of SSA (440 nm) at lower concentration levels (lower AOD days), since iron oxide absorption in coarse mode dust may be missed. The SGP and Bondville sites are located in rural areas that would be expected to have varying amounts of mineral dust loading, depending on regional soil moisture conditions and transport from drier regions to the west (direction of prevailing winds in many seasons).
Lines 1180-1182: Suggest changing to "…typically well within the reported uncertainty bounds especially in light of the in-situ value of 0.06"
Lines 1188-1191: It should also be mentioned that a fraction of the 440 nm absorption from coarse mode dust may be underestimated from in-situ measurements due to the large particle under-sampling bias.