the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 18 May 2026
Mineralogical composition of transported desert dust over Cabo Verde and comparison with model predictions
Abstract. The mineralogical and chemical composition of desert dust particles strongly influences their cloud-forming ability and radiative effects. This study provides quantitative estimates of the main mineralogical and elemental components of desert dust during atmospheric transport above Cabo Verde based on in-situ measurements from the ASKOS campaign in summer 2022, obtained using impactors mounted on unmanned aerial vehicles. Simulations from the METAL-WRF model were used for comparison with sun-photometer observations of total dust load and with in-situ measurements of relative elemental mass fractions of key elements. Across all cases, particle chemical signatures were dominated by illite/muscovite (62%), followed by smectite (9%), kaolinite (9%), quartz (7%), feldspar (5%), calcite (4%), gypsum (3%), and Fe-oxide/Fe-hydroxide (1%). Trajectory and source–receptor analyses combined with satellite observations revealed enhanced calcite fractions for air-masses originating from northern Mali, whereas air masses from southern Mali exhibited increased proportions of Fe-oxide/hydroxide. Good agreement was found between METAL-WRF-derived total dust mass concentrations and independent AERONET observations (slope = 0.62, r = 0.87). Based on in-situ measurements, Si was the dominant elemental component (~25%), followed by Al (~12%), Fe (~6%), Ca (~2.7%), and S (~0.4%). While METAL-WRF reproduced the mean relative abundances of Fe and Ca over the 20-day period, it did not capture the case-to-case variability. Nevertheless, Fe exhibited good agreement, within overlapping uncertainty, between modelled and measured values for most cases, which is particularly relevant for studies of ocean biogeochemistry and dust-related radiative processes.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Atmospheric Measurement Techniques.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(6144 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 23 Jun 2026)
- RC1: 'Comment on egusphere-2026-921', Anonymous Referee #1, 21 May 2026 reply
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 245 | 73 | 17 | 335 | 16 | 19 |
- HTML: 245
- PDF: 73
- XML: 17
- Total: 335
- BibTeX: 16
- EndNote: 19
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
General Assessment
This cross-institutional study presents, according to the authors, the first combination of UAV-mounted free-stream impactor measurements and a regional mineralogical dust model (METAL-WRF) to characterize the elemental and mineralogical composition of Saharan dust during atmospheric transport at multiple altitudes. The observational dataset: 14 analyzed cases, approximately 26 000 individual particles characterized by SEM-EDX, supported by AERONET retrievals, HYSPLIT and FLEXPART-WRF trajectory analyses, and satellite imagery represents a significant field effort. The scientific questions are well-motivated, and the multi-platform observational strategy is commendable. The topic is of clear relevance to dust radiative forcing, ocean biogeochemistry, and cloud microphysics. However, the manuscript requires substantial revision before it can be accepted. Several methodological limitations that directly affect the discussion results are not adequately addressed or quantified. The narrative is at times unfocused, attempting to advance too many sub-arguments simultaneously without a single clear thread connecting them.
Some concerns are discussed and the specific comments are listed below with clear reference to lines in the manuscript.
Comments by section
Abstract
The abstract reads as somewhat disconnected, listing results without a clear narrative arc. The authors should ensure the abstract communicates: (1) the specific gap being addressed; (2) what was done; (3) the key quantitative findings; and (4) the principal conclusion and its implications. As currently written, the transition between the mineralogical results and the model comparison feels abrupt. Introduction
Lines 33–34: The term 'distinct functions' is imprecise. Replace with 'distinct roles in atmospheric processes’ and be specific about which properties are relevant in the context of this paper.
Lines 36–37: The use of 'milling' as a dust production process is ambiguous and potentially misleading in the context of aeolian dust. 'Milling' might refer to an industrial or laboratory process. If the cited studies used laboratory-milled mineral samples, this must be stated explicitly. Particles analyzed in the laboratory can have different surface properties from naturally produced aeolian dust, and generalizing those findings to atmospheric dust without acknowledging this distinction can be misleading.
Lines 46–48: This is an opportunity to expand on the regional and local characteristics that drive differences in dust properties, since source-region variability is precisely what this manuscript addresses. Other regional and global modelling efforts, and other soil mineralogy atlases beyond those currently cited, should be acknowledged here. Other soil map atlases should be mentioned in this context.
Line 72-75: The citation list supporting the statement that model validation 'has been primarily done using ground-level in-situ measurements of mineralogical composition' could be expanded. The authors should cite Nickovic et al. (2013), Ilic et al. (2022), Mifka et al. (2022), and Gonçalves Ageitos et al. (2023), which represent directly relevant regional and global modelling efforts addressing dust mineralogy.
Lines 74–78: The claim of being the 'first time worldwide' that such a combination of measurements and models has been attempted is likely correct but should be reframed to emphasize the scientific value of the findings. What insights does this first-of-kind combination provide that were not previously possible? The novelty claim is more compelling when tied to its scientific consequences.
Missing references: The introduction would benefit from citing the Barreto et al. African dust campaigns and the Journet et al. work on dust mineralogy, as well as the EMIT imaging spectrometer programme (which the authors mention in the conclusions as future work but do not introduce in the introduction where it would provide context for why satellite-based mineralogy constraints are needed).
Free-stream impactor analysis
Line 123: 'Particle areas' is ambiguous. This likely refers to the projected cross-sectional areas of particles on the substrate, not to particle surface area. Clearer phrasing is needed to avoid confusion.
Line 127: 'Lower Z than oxygen' is informal lab shorthand and should not appear in a journal manuscript. Write, for example, 'elements with atomic number lower than that of oxygen' and define Z on first use.
Line 131: The stoichiometric assumption that iron is present exclusively as Fe³⁺ (Fe₂O₃) is introduced without justification or reference. In freshly emitted Saharan dust, a non-negligible fraction of iron occurs as Fe²⁺ in silicate mineral matrices (illite, biotite, amphiboles). The sensitivity of the derived elemental mass fractions to this oxidation state assumption should be noted, or a justification provided for why Fe³⁺ is the appropriate assumption for transported dust.
Line 134: What are the implications of using a common average density of a mineral group in this method? The goal is to understand the composition of a particle and some minerals, namely hematite and goethite have larger density than for example quartz. This should be clarified or justified in the manuscript.
Lines 140: What are the classification uncertainties of the method and how do they affect the discussion?
Model description
The model description does not specify an upper particle size threshold corresponding to the 25 µm threshold applied in the observational analysis. For a meaningful model–measurement comparison, this should be clarified.
FLEXPART-WRF is introduced as a separate modelling framework from METAL-WRF, running at a slightly different horizontal resolution (20 km vs. 22.5 km). The rationale for this separate configuration should be explained. Was it technically possible to run FLEXPART as a module within METAL-WRF?
Line 168: The appendix table referenced in the text as Table A1 is labeled Table A2 in the Appendix. This must be corrected throughout.
The manuscript does not discuss the uncertainty introduced by the choice of soil mineralogy atlas used in METAL-WRF. Given that the implementation of EMIT atlas is planned as future work, can discussion of the currently available soil mineralogy data sets (Claquin/Nickovic, Journet and EMIT) and how they relate to possible active source area during the campaign can be included in the discussion of the results. Furthermore, it could contribute to the Fe\Al ratio discussion.
Backward transport analysis The FLEXPART-WRF emission sensitivity fields indicate regions where surface emissions would most likely have contributed to the observed aerosol population, but they do not confirm that emissions actually occurred there at the relevant time. The satellite provides qualitative support but is itself subject to retrieval uncertainties and detection thresholds. The convergence between trajectory sensitivity and satellite dust signal is largely qualitative and the paper presents it as more confirmatory than the methodology warrants. The authors should temper the language around source attribution.
Lines 177–178: The sentence introducing the satellite constraint is ambiguous. The authors should restructure: one sentence stating the motivation and type of observational constraint being introduced, followed by a second sentence identifying the specific satellite products used. The current version buries the key information inside a long clause.
Atmospheric conditions
The subsection title 'Atmospheric Conditions over the analyzed period' implies a discussion of meteorological conditions, but the section begins immediately with the number of selected data points. There is no characterization of the synoptic conditions during the campaign, no assessment of whether June 2022 was a typical or anomalous dust month relative to regional climatology, and no classification of the sampled cases by synoptic regime, trajectory cluster, or aerosol load category. Without this context the reader cannot judge whether the 14 cases are representative of the broader transport regime or constitute a potentially unrepresentative sample.
Lidar measurements
The campaign included continuous lidar observations providing vertically resolved aerosol information, which is used in the present manuscript solely for cloud screening purposes (Table 2). No comparison between the modelled vertical structure of the dust layer and the lidar-derived aerosol profiles is presented. Given that METAL-WRF provides dust concentrations at multiple vertical levels and that the in-situ measurements are point samples at specific altitudes within the SAL, the absence of a vertical model evaluation represents a significant gap.
Line 188: 'Consistent dataset' is undefined. What criteria determine consistency here? This should be stated explicitly.
Table 2: The relevance of the lidar-based cloud screening column is not explained in the text. Why does cloud presence matter for the interpretation of the impactor samples? This should be clarified, either in the table caption or the surrounding text.
Lines 200–213: The discussion of synoptic and local weather and aerosol conditions appears suddenly and without connection to Table 2, which is not referred to in this passage. The subsection needs restructuring so that the meteorological context, the case selection criteria, and the aerosol characterization flow logically.
Line 205: 'Dust-related objectives' is referenced before they have been formally stated. Either state them here or refer to where they are defined.
Line 211: 'Diverse range of aerosol scenarios' is vague. These scenarios should be classified as suggested above so that the reader understands the variability captured by the dataset.
In-situ mineralogical composition
Line 227 paragraph: It should be stated explicitly that the reported average mineralogical fractions are average across all 14 cases shown in Figure 4. This is currently assumed but not stated.
Lines 235–236: The connection between air masses remaining above 2 km altitude and the dominance of fine clay minerals is not demonstrated. The statement needs a more explicit mechanistic link.
Lines 237–240: The selection of cases for detailed analysis is framed as 'slightly more distinct' without quantitative justification. The authors should state an explicit thresholds that motivated the case selection. The order in which cases are then analyzed should also be explained upfront. The section should be reorganized into clearly labeled subsections either by the case characteristics dependent on the thresholds or chronologically to improve readability.
Figures 5–8: Individual figures span 2–3 pages and the layout has significant design problems: panels are not matched in width or height, some panels overlap bounding boxes, and there is excessive white margin around individual panels. The authors should substantially reduce the number of panels in each figure and move supporting panels to supplementary material, retaining in the main text only those panels that are directly discussed. The figure design should be standardized across all panels as much as possible.
Total columnar mass comparison
Figure 9: A scatter plot is not the most informative way to demonstrate that the model leads or lags the observations in time. A timeseries comparison with multiple model lead/lag lines in the style commonly used in numerical weather prediction verification would more directly illustrate this behavior and allow the reader to assess the temporal offset. AERONET observational uncertainties should be included in whichever plot type is chosen.
The averaging window of ±3 hours is later referred to as a 6 hour window which is correct but slightly confusing. Perhaps choose the preferred way of referring to it.
Elemental mass fractions
Title and terminology: 'Elemental mass percentages' could be changed to 'elemental mass fractions' throughout, as this is the more common term in the literature. The authors may retain percentages as units but the suggested quantity name should be fractions.
Lines 346–347 and Figure 10: The method used to evaluate model uncertainties is not stated in the text. It can be implied that it is the ±3 hour window but this must be clearly stated.
Line 363: 'Comparable' to what, specifically? The reference point is not clear. State the comparison explicitly.
Lines 365–371: The context paragraph is weak. What the current and previous studies may have in common is dust from the SAL, with trajectory analyses suggesting emission from similar West African source regions. This should be made explicit. The acknowledged differences in sampling altitude and location are correct but the discussion does not help the reader understand whether the marginal differences in reported elemental fractions are scientifically meaningful and in what way.
Lines 372–373: The selection of Fe and Ca for case-by-case comparison cannot be justified solely by good model–measurement agreement. The text should explain the scientific motivation for each element and then note that the model agreement for these elements makes the comparison meaningful.
Lines 381–383: It is not clear why Figure 11 and Figure A3 are presented as separate figures apparently showing the same data (scatter plots of in-situ vs. model elemental fractions), one without and one with regression lines. Either consolidate into a single figure or provide a clear rationale for the separation. The paragraph should also begin by explaining what analysis was performed and why, before presenting the results.
Conclusions
Lines 416–417: 'No statistically correlation is found' is a grammatical error.
Lines 417-418: The inference that METAL-WRF is 'better suited for climatological-scale applications' while the WRF model suite was developed and used extensively for NWP is an interesting statement. Do the results clearly indicate this? Alternative explanations, including soil atlas uncertainty, meteorology are mentioned later in the section after this conclusion is made.
Lines 423–427: The attribution of elemental underestimations to ‘limited set of mineral species’ or ‘missing minerals’. Can this be expanded in accordance to previously suggested expansion of the discussion to other soil mineralogy atlases not used in this study?
Lines 431-435: The observational uncertainty sources such as the collection efficiency is listed as one factor among many, whereas it potentially affects the results significantly.
The conclusions should reflect, if possible to evaluate this, the significance of uncertainty sources rather than listing them uniformly.
Recommendation
Major Revisions. The authors are encouraged to respond point-by-point to each concern above. The reviewer is willing to re-evaluate the manuscript following revision and notes that the underlying dataset and scientific concept are strong enough to support a solid publication after revision.