30 August 2023

Sergio Rodríguez (reviewer).

Overview

This manuscript presents an interesting data set on the aerosol composition and solubility of iron in Namibia. Data presented in this manuscript shows that iron solubility in the Namibian coast is rather high, compare to data collected in other deserts, close to dust sources. Authors propose that photo-reduction processes, involving methane sulfonic acid linked to marine emissions, are involved in the increase of iron solubility in the ambient air. This is a very interesting study suitable for publication in ACP. It contributes to increase our global knowledge (https://doi.org/10.3390/atmos9050201) on the variability of iron solubility in the aerosols, with specific observations at Namibia. I listed below some questions and some specific points (SP) that be useful to prepare the final – revised version of the manuscript for publication in ACP.

Specific issues:

SP-01. Line 78-82. In the introduction authors describe the role of dust deposition on the Benguela upwelling. Just to highlight that a new study may be of interest to the authors to highlight the importance of Angola and Namibian dust and the Benguela upwelling. This recent study has shown that in the North East Atlantic skipjack tuna performs northward (Jan to Aug) and southward (Sep to Dec) migrations under the Saharan Air Layer, tracking the seasonal shift of massive dust deposition; this study also indicates that the migration of skipjack tuna between Gabon (a mayor finishing area of skipjack tuna) and the area of southern Angola - Namibia (other mayor finishing area of skipjack tuna) may be modulated by dust inputs. Skipjack tuna migrates to southern Angola and Namibia in the dust season of high column dust load, this is the period when they are caught in abundance, see Fig.2A9, 2B3 and 2C3 of this is the study ( https://doi.org/10.1016/j.atmosenv.2023.120022, sorry for the self-citation, but I think it may be of interest for you). As in the case of NE Atlantic, this suggests that dust deposition (rich in Fe, P and bio essential trace elements) over upwelling waters (rich in Si and N) support zooplankton rich areas, optimal for fish larvae, molluscs, cephalopods, and large predators.

SP-02. A major issue in the is the extraction technique used for the determination of dissolved iron from dust aerosol samples. In this study, authors determined the concentrations of dissolved iron (by ICP-AES) by acidifying (1% nitric acid) the extraction of the sample in deionized water used for the determination of ions and cations (analysed by ion chromatography). The values of the dissolved iron concentrations, and thus iron solubility, strongly depend on the extraction technique and (specially) on the pH of the dissolution, in such a way that more acid dissolution, higher iron solubility.

Did authors measure pH of the dissolutions?

As far as I can remember now, there are, at least, three broad-group of extraction techniques:

1) the one used by the authors (I suggest to include references to other studies which have used the same technique), where the pH of the dissolution may be influenced by the acidification technique used and the presence of ionic balance between acidic species and basic species, including the buffering capacity of buffering capacity of CaCO₃.

2) the extraction of the sample in ammonium acetate leach at pH 4.7, e.g. as used by Baker and Jickells (2017) (http://dx.doi.org/10.1016/j.pocean.2016.10.002)

and

3) solid phase techniques extraction in real sea water (at the pH of the real sea water, pH ~ 8.1), e.g. as used by Rodríguez et al. (2021, https://doi.org/10.1016/j.atmosenv.2020.118092) and Ravelo et al. (2016, http://dx.doi.org/10.1016/j.atmosenv.2016.03.030); The use of real seawater allows considering the potential role of the organic ligands present in the ocean, which complex Fe to keep it in solution in excess of its solubility.

In principle, each extraction technique should be a proxy of an atmospheric process. Thus, iron solubility determined in deionized water would be a proxy of iron dissolution by in-cloud processes, often followed by wet dust deposition, whereas iron solubility determined in sea water would be a proxy of dry dust deposition in the ocean. In general, iron solubility in deionized water tends to be higher than in sea water (due to the lower pH of the former, 6 vs 8).

In the manuscript, authors did a too general (vague) comparison with other studies, both in the section 3.2 and in the abstract (lines 43-45) ;

According to section 3.2 and table 2, iron solubility during dust events is 7.9% (6.9% if removing 1 specific event, line 287), a value much higher than that observed in the dust aerosol samples from the Sahara, which is typically from 0.5 to 0.7 % in real sea water, see begging of section-3 of Rodríguez et al. (2021, https://doi.org/10.1016/j.atmosenv.2020.118092) (sorry for the self-citation again, but just to point that there is a summary and discussion on this topic with many references in this paper).

I suggest to authors to include a brief text (may be in the methodology and sections 3.2) describing that iron solubility will depend on the extraction technique, then authors may explicitly say that they compare their results with previous studies which have actually used the same technique (they may even compare with studies based on sea water extraction, for which lower solubilities are expected).

SP-03. This is just a suggestion. Equation 1 is used for estimate dust mass concentration. In this approach elements are assumed to be present as oxides, a hypothesis which is not actually true (since most of the mineral dust are Si and Al aluminosilicate minerals, clays), but nonetheless this is a approach widely used. Why is the factor 1.12 used?, is it to compensate the average contribution of any element (e.g. that aluminium accounts for 8% of the dust mass)?. If so, authros could verify it or simply determine the real one value the scatter plots of Al or Si vs gravimetric PM₁₀ (is available) in ochre dust samples.

This is important because in line 273-274 it is stated that total iron accounts for 5.8% of dust (total iron / EDM). This value is somewhat higher than that in Saharan dust, which is within 3.9 - 4.0% with an Fe/Al ratio = 0.5, which is lower than the 0.76 found by the authors in Namibia (so there is a data consistency).

If authors have gravimetric PM₁₀ concentrations and their samples are by far dominated by dust (ochre dust colour, as Fig.1B of the study https://doi.org/10.1016/j.atmosenv.2019.117186 ) they could do the scatter plot of Al versus PM₁₀ and determine the actual contribution of Al (similar to Fig. 1A of the study Rodriguez et al., 2020,  https://doi.org/10.1016/j.atmosenv.2019.117186 ), then, the EDM would be:

EDM (mg/m³) = (1 / slope (Al vs PM₁₀)) · Al (mg/m³)                          Eq-R1

Authors could also do it with silicon (which actually much better)

EDM (mg/m³) = (1 / slope (Si vs PM₁₀)) · Si (mg/m³)                           Eq-R2

In the Sahara slope (Al vs PM₁₀) = 0.079, i.e. Al accounts for 8% of dust

In the Sahara slope (Si vs PM₁₀) = 0.16, i.e. Si accounts for 16% of dust

According to my experience with Saharan dust samples, equation 1 may underestimate EDM by a 10%, compared to the Eq R1 and R2.

Just to remind that this (SP-02) is just a suggestion, in case authors find it interesting.

SP-04. Lines 148-155.

Fluorine is cited in the abstract, but not included here (methodology section).

How was fluorine analysed?

It should be described.

SP-05. Lines 195-206. Was the presence of fog verified with local in-situ measurements of local meteorological data of relative humidity (RH)? (table 1?), if these data are available, authors cloud just flag their data and compare them. Also, to validate their method.

SP-06, lines 202: < The FLC product does not specifically distinguish between fog and low clouds >.

To use local meteorological data of RH could help to distinguish fog from clouds.

SP-07. Lines 237-241. This is very interesting. I suggest to include into the body of the article (not in the supplement) a figure with some back trajectories over a satellite view (e.g. Google Earth) of the regions; it would be useful for reader that do not know the region. The number of events is rather low, so a composite could be done with this.

SP-08. Lines 242-246. The formation of fog is typical of the coast of subtropical deserts characterised by upwelling of deep cool waters. Trade winds plays a key role in prompting such upwelling (they are the actual prompters of the southern current and Benguela upwelling) so it should explicitly be cited.

SP-09. Section 3.2 presents the data of soluble iron, aluminium and silicone. Results are presented in Table 2 (comparing dust versus background conditions), then temporal evolution of SFe is shown in Fig. 2 (segregating dust from background conditions) and finally Fig 3 shows the plots of total Fe-vs-total Al, total Fe-vs-total Si, dissolved Fe-vs-total Al, dissolved Fe-vs-total Si. Iron solubility (%S) under dust and background conditions, values are very close, 7.9 and 6.8, respectively. According to line 287, if the dust event 11 is not considered, then iron solubility is similar under dust (6.9%) and background conditions (6.8%). Authors conclude that Fe and DFe have a (line 321).

In my modest opinion the data analysis is rather short and not conclusive.

It seems that under background condition there is also a significant amount of dust (probably up to exceeding 10 mg/m³, according to Fig.3) and that it is a source of iron.

How was background conditions defined?

A previous study in this site by the same group (Klopper et al., 2020; Atmos. Chem. Phys., 20, 15811–15833, 2020, https://doi.org/10.5194/acp-20-15811-2020 ) found very interesting results that may be use useful for the interpretation of the soluble iron data, they found that: 1) main sources of aerosols: sea salt, mineral dust, fugitive dust, industry and ammonium neutralized, 2) As, Zn, Cu, Ni and Sr attributed to combustion of heavy oils in ships, and 3) V, Cd, Pb and Nd of fugitive emissions from mining actives.

Is there a fraction of soluble iron linked to fugitive mining dust?, Is it contributing to the background?

Authors could use the results of Klopper et al. to identify if dissolved iron is linked to any of these sources and during background and even dust conditions. Even if most of total iron may be linked to dust, an important fraction of dissolved iron may be linked to other sources.

Have authors tried a source apportionment of soluble iron?

Authors could do a PMF as Klopper et al. (2020) or simply use the knowledge obtained in the study of Klopper et al. to apply the method used by Rodriguez et al. (2021, https://doi.org/10.1016/j.atmosenv.2020.118092) by scatter plots of soluble Fe to total Fe and soluble Fe to Ni/Al (Rodriguez used V, but authors may use Ni as tracer of heavy fuel oil combustion since V is linked to primary mining activities in their study site, according to Klopper et al., 2020).

How does the plot of iron solubility SFe (%) versus Ni/Al ratio looks?

It could serve to know if fuel oil combustion is contributing to soluble iron or not. Ni is a tracer of heavy fuel oil combustion, which is a source of soluble iron due to the emissions/formation of ferric sulphate and nanocrystals of magnetite aggregates (Fu et al., 2012; https://doi.org/10.1021/es302558m ), formed at temperatures >800 ◦C, followed by sulphuric acid condensation (Sippula et al., 2009 https://doi.org/10.1016/j.atmosenv.2009.07.022 ). According to Klopper et al. (2020) this region is affected by the emissions of ships in the Cape of Good Hope and these emissions may also be impacting during dust events since southern winds prevailed at the sampling site during the dust episodes (according to wind direction in table 1). This may help in the data analysis.

Other data analysis that may help to enrich data treatment (just suggestions):

Has the plot SFe (%) vs Fe a hyperbolic trend?

To include (as supplement or at least cite how they look) the plots of (1) dissolved iron to total iron, (2) dissolved iron to Ni/Al and (3) dissolved iron to sulphate/Al, nitrate/Al and fluoride/Al may help to understand the behaviour of the data and sources of soluble iron.

SP-10. The average iron solubility that authors found in Namibia, close to sources, is much higher than in other sites close to dust sources. I think that this is something that should be explicitly say in the section of results, but also in the abstract and in the conclusive remarks.

Average SFe during dust events found in several studies, = 6.9% in Namibia, 0.7% in Tenerife (close to Sahara), 1.3% in Barbados (distant to sources). Many studies have found SFe(%) < 1% during dust events and then increase up to 10% along several days of aerosol aging.

https://doi.org/10.5194/acp-10-9237-2010

https://doi.org/10.1071/EN09116

https://doi.org/10.1029/2004JD005082

https://doi.org/10.1016/j.atmosenv.2020.118092

SP-11. It would be interesting to label with A, B, C and D each plot of Fig 3 and make the proper reference to them in the text.

SP-12. According to line 287, if the dust event 11 is not considered, then iron solubility is similar under dust (6.9%) and background conditions (6.8%).

This is a very distinctive feature that should explicitly be compared with other studies. North African dust observations in the Atlantic have shown that iron solubility is lower during dust events (high dust concentrations) than in the background aerosol (dust at low concentrations), see as example: Baker and Jickell 2006 and Rodriguez et al. (2021) both included in the article.

SP-13. Section 4 Discussion. This section also includes a presentation of results with plots, which is something that usually goes into the section of results. Authors may consider to merge both sections as a single section Results and Discussion.

SP-14. Figure 5. It would be useful to put labels A (for FS(Fe)), B (Formate), C, D, E, F and G (nssSO4) and cited them in the text.

SP-15. Figure 5. Caption needs to cite that the first plot is dimensionless and not in mg/m³. Authors may also consider to put the first plot as %.

SP-16. Result describe in Lines 407-408 is very interesting. How does the plot %SFe vs MSA look?

SP-17. Section 4.2 is very clear, however section 4.3 is a little bit farragoes, it would be interesting to smooth and shorten the text.

SP-18. Conclusive remarks could be summarised, just focusing on the most relevant findings and the ideas proposed. I think that a sentence saying that average iron solubility in Namibia, close to sources, is (6.9%) much higher than in other sites close to dust sources (<1%). Then propose the photo-reduction processes, involving methane sulfonic acid linked to marine emissions, as potential process favouring such high iron solubility.

END OF REPORT========================================

Desboeufs et al. present a novel dataset of Fe solubilty from aerosols collected from Namibia, an understudied, but regionally important and poorly characterised, dust source. It was interesting to see the apparent seasonal variability and possible links with biogenic emissions from the Benguela EBUS. A small detail that is missing is the acknowledgement that differences in leaching methods also leads to some variations in aerosol Fe solubility, so the best we can hope for when comparing data from aerosols of broadly the same source using different techniques is that the values are broadly consistent (so, trends rather than absolute values is what we are looking at). It would be invaluable to future studies and comparisons, particularly those with a modelling component, if more details of the leaching method are included. I would also like to see more in depth analysis of the data, for instance, discussion on whether correlations are statistically significant. Other than that, I have made some minor suggestions for clarity, detailed below.

Line 31. Insert growth before limiting

Line 71. I think there is an extra ) after Etosha Pan which needs removing

Line 75. If you include the Southern Ocean in the list of where southern African dust can be transported to, the next sentence will be better linked.

Line 94. In the Methods you say that elemental determination was by WD-XRF not ICP(-MS?). I have re-read the Methods section and now see that it was total concentrations that were analysed by WD-XRF and soluble elements either by ICP-AES or ICP-MS. You could either clarify here that different methods were used for total and soluble elements, or just end the sentence at metals.

Line 101. As you are determining MSA from MQ water leaches, the MSA that you detected was water soluble rather than particulate, so it might be best to remove the word particulate from in front of MSA.

Line 103. Replace dust with Fe here

Section 2.2. What was the rationale for collecting different day/might samples and the 3 h gap between sample collection? Do you know if any large dust events or other unusual conditions were missed by sampling non-consecutive weeks?

Line 135. Does the 176 samples include the 13 blanks? Were 163 samples + 13 blanks or 176 samples + 13 blanks collected?

Line 149. Missing citation for the aerosol leach. Was the leachate filtered? If so, through what pore size? This is important information for future comparisons with the data set and needs to be included.

Line 260. enhances

Line 273. It might be worth noting that this is greater than the 3.5% in UCC (Taylor and McLennan, 1995) and closer to the value of 5.04% in Rudnick and Gao (2003), and that this could be due to natural differences in elemental abundance between UCC and Namibian source material and/or that UCC is slightly imperfect proxy for aerosol dust.

Line 283. %SFe is higher during dust events but not significantly so due to the large variability. And only by ~1% so it is hard to accept that that there is a real difference between dust and background in this data, especially as you state that it is the data from Dust 11 that skews the data. Perhaps some rewording needed to make it clearer that there is no difference between the %SFe in dusty/background samples. Perhaps swapping the last two paragraphs starting in line 282 and 291 around would help make this clearer. This similarity in dusty and background %SFe is noteworthy as it contrasts with other regions under the influence of episodic dust events, which should be mentioned.

Line 289. The fractional solubilities from the Baker and Gao studies both used ammonium acetate leaches, whereas, your study used MQ. In addition to mineralogy influencing fractional solubility, several recent studies (and now there is a SCOR Working Group is looking at this topic) have compared the use of different leaching schemes on trace element fractional solubility and concluded that caution should be used when comparing results from leaches using different leach media as the chemical composition of the leach media influences the amount of X that dissolves (from which fractional solubility is calculated, e.g., Perron et al. 2020). The point being your soluble concentrations and %SFe are likely lower than those of Baker and Gao (although we have no way of knowing for sure), supporting the argument that %SFe increases during atmospheric transport (e.g., Longo et al., 2016), while still producing results which are broadly consistent with the cited studies. While not suggesting this is the place for a discussion on leaching schemes, it is important to give as much detail as possible about the leach used in this study in the Methods to allow better comparisons between this dataset and others in the future. Therefore, you should mention the differences in leach media (and the impact on (fractional solubility) here.

Fig. 3. The equation in grey is very hard to read. Change to black.

Line 330. Differences in leaching methods also leads to some variations in aerosol Fe solubility, so the best we can hope for when comparing data from aerosols of broadly the same source using different techniques is that the values are consistent. The exciting finding is that although there was no difference in solubility between dusty and background samples in this study, except for Dust 11 and 13, there was a seasonality.

Fig. 4. The light yellow dots for Dust 4 are very difficult to see. Could a different colour be used?

Line 384. Soluble fraction

Line 397. As you were you determining MSA that has dissolved from particulates rather than particulate MSA, best to remove particulate, perhaps?

Line 401. This should be part of the previous paragraph

Fig. 5. Light pink is hard to see. Could you switch the order so that MSA is under the %SFe plot?

Line 420. ‘the closer they are in the circle’

Line 428. Does the PCA indicate which relationships are significant and which are not? It would be useful to have this information.

Line 466. Very long sentence.

Line 468. Include the range and median value.

Line 473. Concluding remarks

Line 484. ‘average water-soluble Fe fractional solubility is…’

Line 489. What do you mean by a benchmark lab experiment?

Line 490. I think some values are needed here for the soil Fe fractional solubility. This perhaps highlights the fact that soils aerosolised in the lab are not the exactly the same as the aerosols collected on filters.

Line 496. Replace conversely with however and continue from previous sentence (which should not be a stand-alone paragraph).

Line 501. Namibia

Line 507. Perhaps include same before photochemical

Line 509. ‘increased trace and major element solubility’

Line 511. Remove very. It is redundant if this is the first dataset of its kind. Perhaps it is also worth stating again that the conditions in the MBL result in deliquescent aerosols at the study site.