Review comments on “Inferring iron oxides species content in atmospheric mineral dust from DSCOVR EPIC observations” by Go et al. submitted to ACP

&&&GENERAL COMMENTS&&&

The iron oxides contents of goethite and hematite in mineral dust play a key role to quantify the dust light absorption, and then influence its radiative effect. Even though the different spectral behavior of refractive indices, the direct retrieval of goethite and hematite concentration from remote sensing measurements is difficult due to the limited information content. This paper, based on the existing EPIC MAIAC product of aerosol type and spectral imaginary refractive indices, proposed a method to infer columnar goethite and hematite mass/volume concentrations by fitting the EPIC MAIAC spectral (UV-Vis) imaginary refractive indices assuming the Maxwell-Garnett effective medium mixture of non-absorbing host and absorbing hematite and goethite. The results are evaluated with in situ measurements. Overall, this study is well-written and within the scope of ACP. I would recommend this paper to be published in ACP after some comments and concerns being addressed.

1. Since it’s a sequential approach relying on the product of EPIC MAIAC spectral (UV-Vis) imaginary refractive indices (k), the quantitative validation of spectral k product is critical while not included in this study. The SSA validation by Lyapustin et al. (2021, FRS) may imply the quality of k. However, I would suggest to perform the validation of EPIC MAIAC k product directly and to quantify the uncertainty of derived hematite and goethite concentrations due to input k uncertainty.
2. In my view, EPIC MAIAC dust detection is another key information used to select pixels and perform the hematite and goethite inversion. The authors claimed that some dust + mixed/smoke aerosol cases, which are classified into dust, may affect the retrievals for some specific regions. In this sense, I would encourage to include the maps of AE in the analysis which may provide some insights for dust detection.
3. How do you evaluate the derived columnar goethite and hematite mass concentration with in situ near-surface measurements? How do you assume the vertical distribution? Please clarify in the text.
4. The quality/clarity of the figures (e.g., Figures 1, 5, A1, S1) can be improved.

&&&SPECIFIC COMMENTS&&&

Figure 1: the quality/clarity of the figure should be improved.

Line 107: a recent study by Wang et al. (2021, 10.1016/j.atmosenv.2020.117959) developed an algorithm to derive aerosol components from effective density and spectral refractive indices.

Section 2.1: You mentioned the validation results of MAIAC EPIC SSA. Have you evaluated the MAIAC EPIC imaginary part of refractive index (k), since the fitting of MAIAC EPIC k is then used to derive hematite and goethite?

Line 180: the link does not work, please check.

Line 215: why don’t you fit real part together with imaginary part?

Line 227: what do you mean flexible retrievals?

Equations 11-13: So, you use the coarse mode volume concentration to approximate the total, right? Then the C_{v, hema} = C_vc x f_hema is named as the hematite total volume concentration. Is the coarse mode hematite volume concentration more precisely?

Line 283: Could you explain why the algorithm does not retrieve dust aerosol over South America and southern Africa?

Figure 5: There are more cloud-free pixels in the RGB images than that in the AOD, SSA, Hema and Goet plots. Could you explain how these pixels are selected?

Line 449: How do you convert hematite and goethite volume / mass concentrations to the iron-oxide mass fraction? sum of them? It seems missed in the methodology.

Line 469: with a size range of 2-6 days? Please check.

Line 486: The colocation method should include also in the main text. What’s the spatial resolution of the product? Why the spatial window is different over Australia (+/-3 degree)?

Line 586: Why do you use the threshold AOD>1.0? Is it for dust detection?

Line 589: Since it may contain some mixed / smoke aerosol cases, I would suggest to include maps of AE in Fig S2-S5 that may provide some insights for aerosol types.

Line 634: real refractive indices -> real part of refractive indices

Line 708: the reliable product of spectral imaginary part of refractive indices is a precondition

This is a very valuable study that describes a methodology of deriving hematite and goethite concentrations from the UV-VIS single-viewing space imager EPIC aboard the DSCOVER. Obviously, deriving such detailed aerosol property as hematite and goethite concentrations from this type of space sensor requires numerous hypothesis and a priori information, however, the authors do quite a careful work on their educated definition. The methodology is then applied for a series of carefully selected case studies that fulfill the conditions for an optimal application of the methodology. The study also presents a revision of a large diversity of the refractive indexes of hematite and goethite reported in literature, and an analysis justifying the choice made in the study. Elements of validation of the obtained hematite and goethite concentrations using in situ data are presented as well, which is not an evident task for satellite derived aerosol property. I would like also to acknowledge the work done on revision of the literature. The study is certainly in the scope of ACP, it is a solid work and indeed worth the publishing. I have just a few main questions which I believe addressing could strengthen the study, but I leave to the authors to decide if to include the elements of reply in the manuscript or not.

The evaluation of the uncertainty in the derived percentage of iron oxide due to different refractive indexes assumed is very interesting and informative (Fig. 11). I would think about at least two other assumptions that require evaluation of the uncertainty in the derived concentrations. First, the real part of the refractive index, which is fixed. However, generally, the values of real and imaginary parts are related and the imaginary part is varying here. Specially, the real part of iron oxides is quite high (Fig. 1, a), so the real part of mixture is expected to vary quite a bit depending on the iron oxide fraction. I would suggest calculating the effective refractive index and simulate satellite signal for internal mixture of nonadsorbing dust host and iron oxides for the corresponding real and imaginary refractive indexes and then deriving (under the fixed real part assumption) the hematite and the goethite concentrations for this synthetic signal. How will it compare to the initially used hematite and goethite fractions? Indeed, it is mentioned in the paper that Di Biagio et al. 2019 conclude that the real part is generally source- and wavelength-independent with a range of 1.48-1.55, but this range seems to be big enough to cause the derived iron oxides fractions variability. Second, 1 km aerosol height is assumed. The assumption is justified by generally good agreement of EPIC derived absorption with AERONET, but this is on average and fluctuations in specific cases are expected. The dust over hot desert surfaces, specially over Sahara, is lifted to rather higher than 1 km altitudes and sensitivity of UV to the dust altitude is known. What if to conduct a similar test as in the case of real refractive index? That is, to simulate the satellite signal for dust at different altitudes and invert for the iron oxides fractions using the fixed (1 km) dust altitude. These exercises can evaluate uncertainty in the derived fractions due to these two assumptions and provide a valuable error bar. The effect of presence of carbonaceous aerosols (mixture with smoke) can be evaluated in the same manner.

This work presents the application and discuss the results of a new algorithm that applies to MAIAC EPIC data to retrieve the mass concentration of hematite and goethite content in the global atmosphere over land. The aim of the study is of relevance for dust investigations, from the earth’s radiative budget analyses to the biogeochemical studies. The retrieval scheme and the Maxwell Garnett approxmation applied seem reasonable. Hypothesis and steps of the retrieval are clearly presented. Data from laboratory and field observations are used to compare and validate the results. Regional and seasonal variability of hematite and geothite mass concentration retrievals are analysed. The paper i well written and properly organised.

I am in favour of the publication of the paper after the authors have adressed some points below.

• The validation of k from MAIAC, which is the starting point of the analysis, is actually missing. Can you compare the k values against lab and field observations as you did for the retrieved hematite and goethite mass concentrations ?
• To my understanding the algorithms applies only on cases with AOD larger than 0.6, when MAIAC EPIC provides outputs of k, AOD and b. This is one point to discuss when analysing the regional/seasonal concentrations of hematite and goethite, perhaps. What is the impact of this assumption on the retrieval and its exploitation/application? is it a real "global" climatology that is obtained?
• This retrieval applies to the AOD for the coarse fraction, therefore hematite and goethite are referring to the coarse dust AOD only. Is this assumption biasing the retrieval in such a way ? considerations on the size-dependent composition of dust and distribution of Hematite and Goethite as a function of size should be added.
• Figure 3 can probably show goethite dataset for real and imaginary refractive indices as well, similar for Table 1, the corresponding information for goethite is missing

• Line 284 : what do you mean with « significant » events ? in AOD or other ?
• Line 291 : Di Biagio et al . (2019) analysed aerosols generated from natural soil samples and not soil properties, check elsewhere in the paper (in example sect 3.2 and following text) and clarify this aspect which is of relevance for the validation of the retrieved atmospheric aerosol hem and goet concentration from the present study
• What is the impact of fixing the real refractive index on the retrieval ?
• I would discuss possible perspective applications over ocean, an aspect which could be of great interest for biogeochemical studies
• Please, clearly state in the abstract the conditions to which this retrieval applies (land, AOD>0.6, ..)