|I reviewed the previous version of this manuscript, and recommended major revisions. In this version the authors have made numerous revisions, including additional analyses and various clarifications. These have strengthened the manuscript, and I appreciate the authors’ efforts. The reason for not including additional aircraft information makes sense.|
The paper is better but some aspects needs further clarification and possibly correction. Because of this, I recommend further revisions. I would be happy to review again if the Editor feels this would be useful. I expect the next revision is likely to be the final one before acceptance. It is an interesting study, just is still not quite up to ACP standards in my opinion. My comments on this version are as follows:
1. Page 2 line 8: I believe this should say top of atmosphere reflectance rather than albedo. This and the other applications (e.g. Wells, Seidel/Popp) used directional (radiance/reflectance) data rather than albedo. The authors should check the use of reflectance vs. albedo throughout as some instances refer to directional measurements (e.g. satellite measurements) while others are directionally integrated (e.g. CERES flux retrievals).
2. Looking at the revised paper and Supplement, I still have questions about the aerosol model used for SSA retrieval (section 3.2). That refers to tables S5, S6, and S7. Section 3.2 says “aerosol models from OPAC” (models being plural) but the Supplement seems to describe only one optical model. Is the same model used globally? Table S5 makes sense, but I do not think tables S6 and S7 are really phase function values (those would be a function of angle). I am not sure what ‘streams’ refers to here and am guessing these might be related to SBDART’s angular decomposition for the discrete ordinate method? In any case I would just give spectral extinction (as provided in Table S5) together with size distribution and refractive index (as well as perhaps further derived parameters like spectral SSA and asymmetry parameter) rather than providing spectral angular phase function. Or is perhaps the same spectral extinction used everywhere but SSA is manually varied in the calculation (in which case is e.g. phase function also varied)? These aspects remain unclear. So, as well as rewording section 3.2 to clarify if this is one model (manually varied SSAs) or multiple models (e.g. different size distributions and refractive indices forward propagated through Mie code), these tables should be corrected. The spectral dependence of Table S5 suggests to me that this is a fairly mixed fine/coarse aerosol model which implies potentially larger errors for fine-dominated or coarse-dominated aerosol plumes, but again it is not quite clear what is done.
3. Page 11 lines 12-15 and page 12, line 2: I don’t think this statement about POLDER is accurate. If the authors are using the GRASP POLDER product, the POLDER inversion is multispectral, multidirectional, and multi-polarisation so has good constraints on SSA across the visible. Plus, POLDER did not have any UV bands. It is not clear which POLDER product was used here (this should be stated). If it was not GRASP, then it ideally should have been, as this is to my knowledge the best one that is available at present. I also think more discussion should be given to the big offsets between the data sets over some land regions – the authors may be interested in the Chen et al (2020) paper which evaluates GRASP POLDER retrievals: https://doi.org/10.5194/essd-12-3573-2020
4. Section 5: a little more detail is needed here in order to understand what was done and how general these results are. Is the refractive index perturbation to the real part only or is something also done for the imaginary part? How many simulations were done for this (one case or thousands) and are the +/- numbers in the right column of the table the mean uncertainty, the range, or what? Finally, adding the terms on the right hand side in quadrature suggests a total uncertainty on SSA around 0.044, is that correct? I think this “total uncertainty” number should be quoted directly and given more prominence as it is important for the reader to understand expectations of this technique.
5. Page 16 lines 11-12 and 20-21: the wavelength difference between 440 and 550 nm is indeed significant and important. Dust, for example, can have sharp gradients of SSA across the visible. Smoke is flatter but still not zero. It would be better to estimate the AERONET SSA at 550 nm, by interpolating between values for 440 nm and 675 nm. I recommend the authors do this. It will likely improve the comparison in Figure 7a and perhaps 7b.
6. Section 8: I would prefer if this were written in paragraph form, which is more common, but that is more a decision for the Editor and journal style guide.
7. Page 18 line 14: the authors claim “improved” accuracy over previous techniques to estimate SSA, but I do not see evidence that this is the case? I agree that there are fewer data gaps. The wording should be checked and either backed up with facts or removed (please do not exaggerate findings).