Spectral Derivatives of Optical Depth for Partitioning Aerosol Type and Loading

Round 1

Reviewer 1 Report

Major comments

This study discusses about the partition of aerosol components (dust DD, biomass burning originated BB, and anthropogenic pollution AP) based on  spectral dependence of aerosol optical thickness (AOD). Though the paper is well-organized, I have several questions related to the methodology and data interpretation. I think it is nearly impossible to get information of chemical composition from only AOD and its spectral dependency (or Angstrom exponent ANG). My major concerns are outlined below

1. The fundamental idea of this paper is to see the behavior of first and second order derivates of AOD. They argue that the first order derivative gives information about aerosol size, and the second order derivate about refractive index (or single scattering albedo SSA) (Line 186) by referring Holden and LeDrew (1998). I could not find any information in the referred paper that could justify the statement of authors made in this study (the second derivative gives information about SSA). It needs to be justified both from theoretical and observation perspectives how the second order derivate is linked with SSA. Without such discussion, the proposed principle, and discussion based on this principle are not valid.
2. What is AP? Anthropogenic aerosols can be diverse based on their sources and mixing state? For example, black carbon and sulfate both are AP. The refractive index and size of AP can be completely different depending on how (in what proportion) are they mixed. Thus, the assumed data of refractive indices and particle radii (Table 1) are not universally constant to consider that AP and BB can have those fixed values. They can have strong spatial and temporal variations depending on the state of mixing, meteorological condition etc. If someone assumes a slightly different values of refractive index and aerosol size, Figure 4, which is the foundation of this paper, can be completely different. I think Figure 4 is capable to give information of only aerosol size, but not SSA (Note that AOD is largely determined by size distribution; the refractive index has a mere effect on AOD).
3. Similarly , the assumed size for dust seems to not represent coarse mode aerosols. Moreover, dust aerosols can exist in both fine and coarse mode. Dust aerosols often mix with BC and OC etc in the real atmosphere. How this paper can cope such scenarios?
4. Do authors assume a monomodal distribution of each aerosol type or bimodal distribution? If Bimodal distribution has been assumed, then which aerosol is in fine and coarse mode? 
5. The effect of relative humidity (RH) has been completely neglected. Note that RH can change both aerosol size and refractive index as well as Figure 4. 

 

Minor comments

1. 6S model is used for AOD calculation. AOD can be calculated very easily by using size distribution and refractive index information in mie calculation. I wonder why authors need to use 6S model.
2. AERONET data of different sites have been used. Aerosol type is determined based on location. I think it is not a good way to distinguish aerosol type, taking into account the effect of long-range transported air mass depending on the season.
3. AOD (440nm) > 0.8 is used as a criteria for SSA screening, which reduced data in a great amount. It has been verified that AOD (440nm) > 0.4 is enough. Why to choose only large AOD values ?

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The manuscript entitled “Spectral derivatives of optical depth for partitioning aerosol type and loading” presents a new approach for characterizing aerosol type and differentiating mixtures.

The method is well described and is highly reproducible with all the information provided by the authors. This kind of articles provides a new insight in the aerosol apportionment, highlighting the advantages but also the potential limitations if it is used with satellite data.

I highly recommend the publication of this manuscript and I have only some minor questions for the authors, whom I would like to congratulate for their work.

Minor comments

• Abstract: “normalized derivative aerosol index” is written with uppercase initials in page 5 but not here. Please, decide a common criterion
• 3: why is the second derivative of AOD equal to (Del)²AOD/AOD_ref
• 3c) In some figures you label the axis with “normalized (Del)AOD” but you could indicate “NDAI”. I am not asking for changing it, but I would only like to know the reason for using different terminology
• Lines 304-305: The references to Fig. 4c and 4b are incorrect. You should change c) and b)
• Figure S1: Could you use the same terminology that in the text for NDAI(A,mean) and the others?
• Figure S2 shows the normalized first-order derivative for different AOD_ref and you mention it in the text, but you do not discuss the selection of 0.87 um and the limitations of using other reference values.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The manuscript by T.-H. Lin et al. presents a new approach to using the normalized first- and second-order spectral derivatives of aerosol optical depth (AOD) with respect to wavelength to partition three major aerosol types (i.e. dust (DS), anthropogenic pollutant (AP) and biomass burning (BB)) in aerosol mixtures. This NDAI (Normalized Derivative Aerosol Index) approach was examined and validated using theoretical simulations from the 6S (Second Simulation of a Satellite Signal in the Solar Spectrum) model, and in-situ measurement datasets from AERONET sites near source regions of DS, AP and BB emissions.

The paper is comparatively well written. The manuscript material and results fall within the scope of Remote Sensing journal and will be of wide interest to the remote sensing community.

My recommendation is to publish the paper in the Journal after minor revision. The manuscript has several shortcomings and needs to be improved. Find please my specific and technical comments below.

 

Specific comments and recommendations

 

1. Introduction (page 3, lines 111–112). Citation: “... and CALIPSO (Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation) aerosol products”. I no longer found mention of CALIPSO in the entire article.
2. Introduction (page 3, lines 112–117). The last sentence of the Introduction sounds like the conclusions of the research done. Perhaps, it would be better to reword this sentence accordingly and move it to Section 5.
3. Methodology (page 3, line 139). Citation: “...standard deviation (σ), referring to WMO datasets, are listed in Table 1…” I did not find sigma (σ) in Table 1 (see the last row in Table 1).
4. Methodology (page 5, Equation (3)). I see the difference between the first-order derivative (Eq. (1)) and normalized first-order derivative (Eq. (2)). According to Eq. (3), this is the normalized second-order derivative. However, Figs. 3b and 3d show both second-order derivative and normalized second-order derivative. Why do you use only one equation (Eq. (3)) to describe both types of the second-order derivatives?
5. Results and Analysis (page 8, line 280). Why were the spectral pairs 0.44-0.47, 0.55-0.66, and 0.66-0.675 μm excluded from consideration?
6. Results and Analysis (page 8, line 282). Citation: “The values of ∇τ_{(λ1, λ2)} became almost flat…” I think it should be “The values of ∇τ_(λ1,λ2) for DS aerosols became almost flat…”
7. Results and Analysis (page 8, lines 287–288). Why were the other sequences excluded from consideration, for example, 0.44-0.47-0.55 μm? See please also the comment 5.
8. Results and Analysis (page 12, lines 382–385). Citation: “The relationship between first and second-order spectral derivatives basically follows a quadratic polynomial along with the mixing weight of components (R² ≅ 0), but it also can be well described by the linear relationship (R² ≅ 0.99) within 0.44–0.87 μm spectra, as Figure 7 shows”. Figure 7 does not show both linear and quadratic curves with the corresponding determination coefficients R².
9. Results and Analysis (page 13, lines 416–417). What do the expressions AODF_NDAI and AODF_INP mean? Perhaps, these are fAOD_NDAI and fAOD_INP, respectively?
10. Results and Analysis (page 13, lines 417–418). Citation: “The differences of the NDAI approach from initial inputs are less than 8% after detailed comparison in Figure 8b”. According to Fig. 8b, the differences are less than 0.08%.
11. Results and Analysis (page 13, line 420). Citation: “This may be closer to a quadratic polynomial relation, as Figure 7 indicates”. See please comment 8.
12. Results and Analysis (page 14, Figure 8 caption, line 430). Citation: “… while AOD fraction of DS decreased from 0.0 to 1.0 accordingly”. Are you sure that "decreased"?
13. Results and Analysis (page 14, Figure 8 caption, lines 431–432). Citation: “(b) The difference between fAOD_NDAI and fAOD_INP is shown as a percentage”. This contradicts with the statement in lines 417–418.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

General comment

 

The paper reports an analysis of a methodology to apportion aerosol types into three potential categories (dust, biomass burning, and anthropogenic pollution) based on spectral derivatives of remote-sensed aerosol optical depth. The use of AOD to apportion some aerosol categories is not particularly new and also the three typologies of aerosol studied are quite standard. However, the analysis is well done and it could be useful for readers. The topic is suitable for the Journal. I found that some aspects are not clear and that it is a little weak the discussion on real world applications in which three, or even more, categories of aerosol will be present. So that I suggest to consider the paper for publication only after a major revision addressing my specific comments.

 

Specific comments

 

One aspect that should be discussed in more detail is the application to real world cases. The paper focus mainly on mixing of two typologies of aerosol, however, real cases will have at least the three investigated categories for all the time. Some more details on the application with real-world mixtures is necessary especially for the applicability of the approach and its uncertainty.

 

Another aspect is that real cases will also have additional categories such as sea spray, secondary aerosol and so on. How these categories of aerosol will influence the partitioning with the method used here or the uncertainty of the results?

 

Figure 3. Are these several collapsed curves, an average or what?

 

Figure 6 starts from tau=0.8, instead, figure 5 starts from zero. So that it is not really true that figure 6 is the same as figure 5 after a normalisation as written in the paper. Please correct and explain. It is also necessary to discuss if the separation of the three typologies would be effective also without normalisation because looking at Figure 5 it seems this way. What normalisation really achieves?

 

Section 4.4. It should be clarified what is the percentages of cases with tau greater than 0.8 in experimental data and comment on the applicability of this approach. I mean if a relevant fraction of real cases is under 0.8, the method is not applicable and this limit its usefulness. A discussion on this aspect should be provided.

 

Lines 490-516. These sentences are actually a summary and a perspective discussion and should be transferred to a conclusions section that is missing in the paper.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 4 Report

The paper has been improved during revisions and authors answered to my questions. I believe that the paper is ready for publication.