Review on

Optical Properties and Shape-Dependent Complex Refractive Index Retrievals of Freshly Emitted Saharan Dust

By

Yus-Díez, J. et al.

egusphere-2025-2571

General comments

The paper is well written and structured. It deals with the never-ending topic of properties of desert dust by analysing them in situ next to the source, and thus contributing to a rather scarce dataset of desert dust observations. It is decently illustrated and referenced, althhough I’d put some suggestions if authors would like to add them later. However, at least how it feels to me, these are two papers, one is generally describing a field campain and data colelcted, which is almmost flawless apart of several minor suggestions, and another part dealing with retrievals of desert dust imaginary part of refractive index. I can accept only the first one, since methodologically section 2.6 is so weak, fixing it would require writing another paper, below I’d try to structure my concerns. Generally, I’d leave suggest a majour review, either by getting rid of the sections 2.6 and 3.3 (and I’m afraid changing the name and scope of the paper), either significantly rewriting them.

Major Comments:

Majority of corrections used in situ easurements and descried in methodology section are affected by particle shape assumptions, and accuracy of such corrections and resulting impact on observed values and most importantly retrievals are not discussed in the paper.

Issues in sections 2.6 and 3.3, describing the retrivals of imaginary part of refractive index. First of all authors do their best to disentangle shape, size and complex refractive index of desert dust particles — an impossible trick using the observations provided. With all due respect having a wide range of values that depend on selected real part and the non-sphericity modelling doesn’t give any insite of what this parameter is actually are, all effects combined the unceranty of such retrieval in real world scenario will be huge, even if we limit the analysis to the specific shape model, and disregard completely the uncertainty of the observations used in the retrieval, as authors did.

Ideally it would be nice to see another group of retrievals on the edges of the SSA estimation uncertainty. And methodologically speaking to vary also the corrections used to provide this estimations, including the truncation, density packing etc. to see all possible impacts on the parameters that are lately used in the retrievals. As I mentioned above characterising this wil be a hell of astudy, wich will presumably yield very sad and unsatissfactory results increasing the uncertainty of k “retrievals” skyhigh. Nonetheless, I would be pleased to be proven wrong on this one, if such revision will be provided.

Another issue is the non-sphericity model selection, which seems to be arbitrary, or at least not well justified, why exactly these shapes with these parameters were chosen? Why the most commonly accepted (AERONET/MODIS operational, plus a ton of advanced satellite mission products) spheroid model is disregarded completely? Authors conclude that shape assumption provides a significant change in k retrievals, but without making retrieval with spheroid model (Dubovik et al., 2006) comparison with AERONET doesn’t make much sense methodologically, not even rising questions about correctness as a whole of  in-situ vs AERONET comparison. There are also impressive next-gen approaches by Bi et al., 2018 and Saito et al. 2021 a,b to tackle non-sphericity. Their existence at least should be acknowledged, and a resonable explanation why specific models are selected for the study should be provided.

Generally, it is not fully clear for me why a rather complex (and inevitebly biased) procedure to estimate SSA was used, and then these values were fitted. Why not use the both absorption and scatering instead (as in DB19 e.g.). This also allows to simulate scattering with truncation and fit exactly what was measured.

Also method is not tested on “clean” simulated data and immediately applied to measurements, taking them as ideal, and not accounting for any possible errors and biases. Whole section 3.3 would be a proper set up for the syntetic data input generated for method testing, free of any noise and biases, but not real observations. And performing this test would also allow to perturb data and estimate retrieval accuracy by adding controlled noises and biases to the “ideal” dataset.

Also paper lacks consistency: some shapes were retrived, some shapes were compared to but not retrieved, PM10 measurements are analysed provided, but not retrived. It would be nice to see a more cosistent approach.

Minor comments:

Line 124: “at eight angular positions (0 ◦ , 10◦ , 25◦ , 40◦ , 55◦ , 70◦ , 90◦ , and 170◦ ).” Were this positions used in this study? If not why mention it?

Line 125: “Forward (0 ◦ ) and backward (170◦ ) scattering measurements were then corrected for non-ideal illumination of the light source and truncation errors using the correction scheme from Müller et al. (2011) for coarse particles”. Contradits with line 260 (see below). Please, clarify. Also, truncation correction in Muller et al. provided only for spheres. Maybe for PM2.5 impact isn’t small, but for PM10 I belive it’ll be more noticeable. It is not clear how his affects the k retrievals.

Line 128: I find using PM2.5 inlet to study dust particles, rather unorthodox. Please provide some explanation why such set up was selected, it is not clear.

Line 140: “This conversion requires assumptions about 140 the particles’ CRI and shape”. How then retrival can be done, if we need CRI to estimate PSD, to retrieve CRI? Can the uncertainty of such PSD estimations be assesses, same as it’s impact on k retrievals? Technically if you don have a unique solution, I is not retrieval, at least I wouldn’t call it this way. Provide a flowchart of the retrieval to make it clearer.

Line 213: “minimizes the discrepancy between simulated and measured SSA”. I do believe “measured” is a very strong term for an estimation that relies on multiple assumptions and comes from separate instrumentation and in addition has no accuracy characterisation for the “measured” values.

Line 226: having a flow chart of the “retrieval” will be nice, it’ll bring more clarity how everything organised.

Line 246: ”biaxial prolate ellipsoids” is there any reason oblate ellipsoids were disregarded?

Table 1. SSA values are provided with ± (not indicated what), is it variation in the dataset, is it accuracy? For the latter assuming there are uncertainties on each part of the SSA estimation, b_abs has uncertainty, they are interpolated to another wl (with uncertain AE), b_scat in uncertain too, and these are combined and divided to get SSA, I found the provided uncertainty of 0.001 unrealistically low. DB19 claims uncertainty of SSA340 up to 12%, which is a whopping 0.1 (two magnitudes higher). (Once again, how this error will propagate to k, is the question)

Line 260: “These computations provided the truncated scattering coefficient and the absorption coefficient. To avoid introducing additional assumptions, we defined a modified SSA, using the truncated scattering coefficient rather than an assumed full-phase-function scattering coefficient.” This phrase is confusing, if truncation correction were not used, why pu so much descriptions about, and was the effect of having/not having truncation on SSA and importantly retrievals of k studied?

Lines 270 and 425: “and the three inlet cut-off diameter efficiencies (1.7 and 1.7 ± 10%)”, please justify that 10% is realistic innacuracy due to all affecting factors combined. Do you account the effective chage of cut-off due to the shape and consecutive PSD changes?

Line 275: “We repeated the optimisation for multiple fixed values of n within the 1.48–1.55 range, in 0.01 increments.” If understood correctly you get 7 results with different k, due to the intrinsic ill-posity of the problem, which value out of this solution group is then selected?

Figure 7. Where are the subplots for A-F shapes similar to subplots of sph, di dist and bi fix?

Figure 8 Is confusing, what are shapes A,B,C,D,E and F? They are mentioned only in the caption of the figure 7 and never discussed in the paper. It is also not clear were retrievals performed for these shapes?

Table 2 and Table 3. Focusing on PM2.5 retrievals, were any made for PM10? The rest of article analyses properties at these cut-offs. This part lack consistency. Seeing (no) differences in k retrievals for PM2.5 and PM10 would be nice, and demonstration that they are comparable will be a good sanity check for the retrieval.

Figure 9. Where are shapes A-F as in Fig 7 and 8? A bit confused what are theese are and what are the logic in putting them on that figures. Why the box plots include only best estimates of n and cutoff, but not showing the realistic box plot as in fg 7 (b-d)?  

Line 218: “minimizing a quadratic cost function based on SSA” please provide the thresholds on SSA differences and typical cost function values achieved during retrievals, how the method knows when to stop? Was the SSSA fitted to satisfactory levels? And what these levels are.

Line 227: “incorporating extensive information from the polar nephelometer”, it would be nice to read what options were considered, and generally speaking see a reference to a paper on retrieval design choice/optimisation, or at least some of the results shared in this one. In DB19 they used both absorption and scattering and managed to get both n and k. Please, discuss.

Technical comments:

Figure 1. Middle panel text is too small to read. Caption (a,b and c,) should be (a, b and c), also a/b/c/ notion is too hard to distinguish, concider making them bigger or higher contrast.

Line 40-41: For instance, the presence of iron oxides in such minerals as goethite and hematite, plays a dominant role in dust absorption (e.g., Di Biagio et al., 2019).

It is not clearly mentioned anywhere but I presume aethaelometer and nephelometer share the inlet, cyclone and pump?

The name of the article states that complex refractive index is retrieved, even though, technically speaking, it is not, since n is fixed.

I’m not sure that figure colour schemes are following the guidelines for colour vision deficiencies. There’s a nice set that can be used: https://www.fabiocrameri.ch/colourmaps/

This study presents high-resolution field measurements conducted in the Moroccan Sahara Desert to investigate the optical properties of dust and to retrieve the imaginary part of the refractive index. It explores the influence of particle size and shape on these properties. The study addresses an existing gap in measurements of fresh dust optical characteristics and helps improve understanding of how size and shape affect dust refractive index retrieval.

However, there are still areas where the clarity and methodological rigor of the manuscript could be improved.

Major Comments

While the retrieval of the imaginary part of the refractive index (k) is performed, the aim of disentangling uncertainties due to particle size and shape is ambitious, especially in the context of real measurements. Reducing uncertainties in climate models presents a further challenge. Directly investigating these uncertainties often favors theoretical studies, as real-world measurements and retrievals can introduce numerous errors and biases. A primary concern here is the reliability of the retrieval method.

The authors use a modified SSA for retrieval, which differs from other studies (e.g., Di Biagio et al., 2019) that directly use scattering and absorption coefficients. Can the authors explain their choice of using SSA? What are the advantages and disadvantages compared to using scattering and absorption coefficients directly?

In Section 2.6, the authors state that a robust simultaneous retrieval of both the real part (n) and imaginary part (k) is not feasible. Could the authors provide more detailed explanations on this point? For example, Kong et al. (2024) demonstrated that discrepancies in particle size distributions (PSDs) between inversion models and inhomogeneous irregular dust particles can lead to biases in scattering coefficients, complicating the retrieval of n. Could a similar mechanism explain the limitations encountered in this study?

Additionally, it’s recommended that the authors illustrate the optical properties (e.g., scattering and absorption coefficients) computed from the retrieved refractive indices. It would be insightful to discuss the differences between these model-derived optical properties and the measurements, especially given that the retrieval of the refractive indices relies on the ratio of scattering to extinction coefficients (i.e., SSA).

A robust validation of the algorithm is needed. Testing the method by adding controlled noise and biases to synthetic data, and then evaluating the retrieval accuracy, would be highly beneficial. Beyond these points, what are the authors’ suggestions for improving refractive index retrieval in real measurement scenarios?

Reference:

Di Biagio, C., Formenti, P., Balkanski, Y., Caponi, L., Cazaunau, M., Pangui, E., Journet, E., Nowak, S., Andreae, M. O., Kandler, K., Saeed, T., Piketh, S., Seibert, D., Williams, E., & Doussin, J.-F. (2019). Complex refractive indices and single-scattering albedo of global dust aerosols in the shortwave spectrum and relationship to size and iron content. Atmospheric Chemistry and Physics, 19(24), 15503–15531. https://doi.org/10.5194/acp-19-15503-2019

Kong, S., Wang, Z., & Bi, L. (2024). Uncertainties in laboratory-measured shortwave refractive indices of mineral dust aerosols and derived optical properties: A theoretical assessment. Atmospheric Chemistry and Physics, 24(11), 6911–6935. https://doi.org/10.5194/acp-24-6911-2024

Minor Comments

Line 49: The citation sequence is incorrect. It would be clearer as: (e.g., Gasteiger and Wiegner, 2018).

Line 263: Please clarify whether the modified SSA is used in subsequent analyses.

Figure 2: Could the authors comment on the origin of some of the peaks observed in k525nm in panel (b)?

Figure 3: The font size is too small, particularly for the labels indicating different AAE values. Increasing the font size would improve readability.

Line 414: The text refers to “Figs. 6c, d,” but it seems this should be “Figs. 6c, f.”

Line 506: Typo: “Moreovover” should be corrected to “Moreover.”

Line 546-547: The dependence of optical properties on composition is minimally addressed. Please either modify the expression to reflect this or add more details to elaborate on this aspect.