General Comments

The manuscript describes the sampling and analysis of mineral dust samples collected in the Moroccan Sahara in September 2019. For inter-comparison the dust particles were collected by three sampling methods, including flat-plate sampler (FPS), free-wing impactor (FWI), and a multi-stage deposition impactor (MOUDI). Real time particle size distributions were also measured by optical particle counter (OPC). The samples were analyzed by automated scanning electron microscopy (SEM), generating analyses on particle morphology and chemical composition, of more than 300,000 individual dust particles. Normalized elemental results were assembled into eight particle classes by a set of rules, each class approximating a likely dust mineral. These were labelled as hematite-like, quartz-like, feldspar-like, calcite-like, gypsum-like, halite-like ammonium sulphate-like, kaolinite-like, etc. The results were classified by particle diameter into eight size bins from, < 1 to 128 μm, each size bin containing the particle counts in that size bin. Inter-comparative plots of number abundances of the mineral-like and other chemical parameters abundances are provided.  Discussion of the iron comparative number distribution plots for the FPS, FWI, two stages of the MOUDI, as well as for the OPC are given, with discussion of the mixing state thereof. Aspect ratios are compared with those from other studies.

The title clearly reflects the content of the paper and the abstract provides a concise summary.

It adequately addresses the in situ sampling and SEM analysis of mineral dusts, of importance to a better understanding of airborne minerals and their potential impact on Global climate.   

Although automated SEM analysis has been known for more than two decades (Engelbrecht et al., 2009), the application thereof to dust analysis remains novel, especially regarding the interpretation of data. Research presented in this paper further explores the automated SEM procedures and technology. A Global library of airborne dust mineralogical and morphological data will facilitate the interpretation of satellite signals.  To readily compare with other similar studies, the chemical results should be calculated to reflect mass fractions or mass percentages, not number abundances.

Specific Comments

1. Particle number abundances (fractions) per “volume equivalent diameter” are presented in figures and text in the manuscript. In a few instances (lines 16, 447, 532, 574) the authors do refer to volume fractions. As shown in the below figures (Engelbrecht et al., 2017), there is a substantial bias towards the number of smaller particles. To better compare with other bulk analytical techniques, individual particle volume or mass (Ott and Peters, 2008a; Prakash et al., 2016; Engelbrecht et al., 2009; Engelbrecht et al., 2017; Marsden et al., 2018; Marsden et al., 2019; Ott and Peters, 2008b) can be assessed from SEM based particle area measurements (Engelbrecht et al., 2009; Prakash et al., 2016; Ott and Peters, 2008b).  Similarly, for intercomparison of chemical parameters between particle size bins, it is preferable to express abundances as equivalent volumes, and if particle densities are known, particle mass.

  It is recommended that the number concentrations be converted to at least volume concentrations throughout this manuscript, and that the chemical parameters be assessed in terms of volume fractions or volume percentages. The figures illustrating the chemical parameters will need to be re-compiled to reflect volume (or mass) concentrations and not number abundances. In this way the results can be readily compared to other SEM based studies, as well as to chemical results generated by other analytical techniques such as SPMS, ICP-MS, XRF, etc.

2. Results from this study should be discussed in conjunction with individual particle dust studies from other desert regions, including North Africa and the Middle East (Engelbrecht et al., 2009; Engelbrecht et al., 2017; Prakash et al., 2016).

Some Technical Corrections to consider

Line 43.properties and mineral composition

Line 47.  deposit faster due to gravity, ….

Line 53.  ..dust is more optically absorbing when…

Line 57.  ..not only on the source composition but also on the feldspar…

Line 80.  …are the interdependencies among particle size, composition,…

Line 82.  …mixing state upon both particle composition and size…

Line  83. …mineralogical composition of windblown dust are relatively new …

Line 89.  Unclear, rephrase sentence. “One reason for such measurement scarcity is that source areas, particularly the host prolific ones, are typically located in remote areas and subject to harsh conditions.”

Line 91. Dust storms result in high particulate concentrations and filter overload, providing a challenge to analyze by automated SEM.

Line 96. Three different instruments were used to sample mineral dusts, for the measurement of chemical and physical properties of individual dust particles by SEM coupled with an energy-dispersive X-ray analyzer (EDX).

Line 99.  …individual dust minerals to ultimately help understand…

Line 101. Such information is required to better constrain climate models that consider mineralogical variations in their representation of the dust cycle (Perlwitz et al., 2015a; Scanza et al., 2015; Li et al., 2021). This is timely, given the prospect of global soil mineralogy retrievals using high-quality satellite hyperspectral measurements (Green et al., 2020).

Line 117. Under favorable weather conditions, dust is frequently emitted from this source area. (González-Flórez et al., in prep.).

Line 129.  …FWI were collected twice daily with a typical….

Line 138. …particle size cutoff is determined by the impaction …

Line 145.  The main controls for particle …

Line 153. …and fourth stages were selected for detailed analysis.

Line 154. ..when particles impact on the collection substrate…..

Line 156. ….depending on the mineralogical properties of particles….

Line 162. Previous studies (Kandler et al., 2008………

Line 162. Add references (Engelbrecht et al., 2009; Engelbrecht et al., 2017; Prakash et al., 2016)

Line 166. ….without any pretreatment. Particles were automatically detected and analyzed …….

Line 167.  …(BSE) imagery was used for particle detection, as dust particles ….

Line 170. .. number of input counts by the EDX detector.

Line 174. Accounting for matrix dependent efficiencies of  the….

Line 176. The SEM-EDX results are normalized to 100 %, including those of the elements C, N, O, Na, Mg, Al, Si, P, S, Cl, K, Ca, V, Cr, Mn, Fe, Zn, and Pb. …….

Commnent: What is the minimum particle size measured by automated SEM-EDX?

Line 228. SEM-EDX measures the elemental composition

Line 229.  …can therefore be readily identified (e.g. gypsum, quartz, calcite).

Line 238. X is defined as the ratio of the mass concentration of the element considered to the sum of the mass concentrations of all the measured elements.

Line 243.  …method was not applied to quantify…

Line 246.  Their most prevalent chemical components, …

Line 247.  The mineral labels were assigned from the most prevalent measured elemental concentrations. There was no actual phase identification ……

Line 252.  Aerosol PSDs of suspended mineral dusts were also obtained by Optical Particle Counter (OPC)…….

Line 264. …..only error considered is the Poisson counting error. Explain or delete sentence

Line 268   Based on set classification rules, particles….

Line 275 …particle types were found mainly in particles with…

Line 284. ….Fe content of quartz-like particles is generally low and variable, suggesting that Fe oxides are not an integral part of the quartz-like particles.

Line 289.  Ca-feldspar-like particles are quite rare in the samples. Explain why, plagioclases are seldom pure albite, often oligoclase or andesine, depending on their provenance. This anomaly perhaps to do with the SEM analytical procedure?

Line 301. Figure 2 y-axis. Are relative abundances by number or volume (or mass)?

Line 326. Ammonium sulfate is the prevalent sulfate species in atmospheric aerosols and generally of anthropogenic origin.

Line 373. On average across all particle sizes, by number, 26% of particles are illite-like, ……

Line 416. …. along the air stream onto the lower stages.

Line 422. While the dependence of particle size distribution on sample mineralogy is quite strong, the temporal …

Line 431.  There are fertilizers that do not contain K, such as di-ammonium phosphate (DAP), mono-ammonium phosphate (MAP). Hydroxy apatite is not commonly found in the Morocco phosphate deposits, the mined francolite generally contains F.

Line 439. Figure 8. Phase compositions of size bins for samples collected under specific meteorological conditions at specific times and on certain days.

Line 439. The total number of analyzed particles is given for…..

Line 441. feldspar not a typical Fe bearing mineral, only K-feldspar in very small amounts.

Line 474, Figure 9. Relative number abundances?

Engelbrecht, J. P., McDonald, E. V., Gillies, J. A., Jayanty, R. K. M., Casuccio, G., and Gertler, A. W.: Characterizing mineral dusts and other aerosols from the Middle East – Part 1: Ambient sampling, Inhalation Toxicology, 21, 4, 297-326, doi: 10.1080/08958370802464273, 2009.

Engelbrecht, J. P., Stenchikov, G., Prakash, P. J., Lersch, T., Anisimov, A., and Shevchenko, I.: Physical and chemical properties of deposited airborne particulates over the Arabian Red Sea coastal plain, Atmospheric Chemistry and Physics, 17, 11467-11490, https://doi.org/10.5194/acp-17-11467-2017, 2017.

Marsden, N. A., Flynn, M. J., Allan, J. D., and Coe, H.: Online differentiation of mineral phase in aerosol particles by ion formation mechanism using a LAAP-TOF single-particle mass spectrometer, Atmospheric Measurement Techniques, 11, 1, 195-213, 10.5194/amt-11-195-2018, 2018.

Marsden, N. A., Ullrich, R., Möhler, O., Hammer, S. E., Kandler, K., Cui, Z., Williams, P. I., Flynn, M. J., Liu, D., Allan, J. D., and Coe, H.: Mineralogy and mixing state of north African mineral dust by online single-particle mass spectrometry, Atmospheric Chemistry and Physics, 19, 2259-2281, https://doi.org/10.5194/acp-19-2259-2019, 2019.

Ott, D. K., and Peters, T. M.: A Shelter to Protect a Passive Sampler for Coarse Particulate Matter, PM_{10 - 2.5}, Aerosol Science and Technology, 42, 4, 299-309, 10.1080/02786820802054236, 2008.

Prakash, P. J., Stenchikov, G., Tao, W., Yapici, T., Warsama, B., and Engelbrecht, J. P.: Arabian Red Sea coastal soils as potential mineral dust sources, Atmospheric Chemistry and Physics, 16, 18, 11991-12004, doi:10.5194/acp-16-11991-2016, 2016.

Overview

This work describes SEM-EDX and particle size analysis of freshly collected mineral dust from the Moroccan desert. The authors combine multiple collection methods and describe how mineralogical composition changes over the particle size distribution. Measurements of mineralogical composition in mineral dust aerosol are still sparse, so the dataset presented by the authors is valuable and relevant to ACP. The work suffers from issues in presentation and does not always transition from data presentation to the expected interpretation. I recommend that this work be accepted only after the authors revise the work.

Overall Comments

1. The authors often spend a significant amount of time in the results citing previous work but not directly connecting the discussion to the current analysis in a clear way. This work will be significantly more impactful if the authors work on conciseness.
2. The mineral identification is very roughly explained, and no validation is presented. At a minimum, citations need to be included showing that the development of user ID’d mineralogical classes is an appropriate method for these data. If the authors were only presenting a few major classes, what’s presented may be acceptable, but 22 mineral classes require some validation. If the authors could present some analysis of known standards, it would greatly strengthen the defensibility of this dataset.
3. The change in mineral composition/size distribution of particles from soil to suspended mineral dust is an important point. It would improve the paper to include a direct comparison to the soil data here (not just reference to a forthcoming paper).
4. Aggregate data presentation. Throughout the work, the authors present compositional size distributions that seem to have aggregated multiple collection methods with different collection efficiencies, target size ranges, and sampling intervals (for some the sampling intervals are very short.) The authors should more explicitly describe this aggregation process so it can be accessed for validity.

Specific Comments

Supplement: For the particle ID code (the key component of this text), the text should be included, not just a screenshot. The authors should also list the software used here.

39: replace “catalyzer” with catalyst

84: On a global scale, the mineralogical maps presented by Journet et al. 2014 are relatively high resolution and based on a large number of mineralogical measurements, albeit with notable gaps. The authors should instead emphasize the gap in knowledge bridging soil mineralogical and size distributions to those found in dust aerosol, as this is a major challenge.

2.7: The authors should note that Optical particle counters usually rely on assumed refractive indexes and particle shapes in order to determine distributions. Thus, in a dust-dominated system with irregular particle shapes, the OPC is likely to be highly uncertain. This should be, at minimum, discussed and perhaps adjusted for in the discussion of the data.

2.8 Is the confidence interval simply the confidence of the fit of the binomial distribution to the observed particle size distribution, or are other uncertainties included?

307: What is the typical percentage or range of Ca-rich particles in the larger size classes? For other types, percentages or ranges were included.

3.1: It would be useful to the reader to see the percentages of particles in a table (average, range, median, etc) for each compositional class and size type. The size classes referenced in the text are inconsistent between compositional classes making comparisons difficult. This may be suitable for the supplement.

373: I would caution against presenting an average of percentages over the different size classes. This average loses information such as particle number concentration and masks compositional differences between size classes.

375: Do you believe the ammonia sulfate to be soil/dust sourced? It seems likely that this is secondary aerosol, which would support its presence in the small-size bins.

380:  The change in mineral composition/size distribution of particles from soil to suspended mineral dust is an important point. It would improve the paper to include a comparison to the soil data here (not just reference to a forthcoming paper).

380-393: This presentation reads like a collection of unrelated thoughts. I recommend rewriting to emphasize how this puts your data into context.

Figure 4: Please explain what “normalized” means in the left axis. What are the data normalized to? How is this different that dn/dlog(d), and why is it used?

3.3: It seems an important point to parse whether larger particles are mostly aggregates of internal mixed smaller particles or have uniform composition. There is a good discussion in this section, but I wonder if additional analysis could be performed. Could you compare the elemental ratios Fe allocated to the “complex” categories oversize classes to gain some insight into this? For example, does the Fe:Al ratio go up with size class for these complex categories? This would imply the internal mixing of high Fe particles. A “simple” two-source mixing model maybe could be used to estimate the Fe-oxi/hydroxides vs. structural iron. I’m not confident using the Si:Al ratio as a filter as you do in this section gives a good reference point. As your major iron hydroxides (hematite, Goethite) don’t contain Si, this just allows you to compare siliceous, and clay minerals, which I’m not convinced is a useful comparison for Fe.