Fog samples were collected using two pre-cleaned Caltech Active Strand Cloud Water Collectors (CASCCs) with a droplet size collection efficiency of 50 % at 3.5 µm in diameter (Demoz et al., 1996). One of the collectors was a stainless-steel version, ss-CASCC, described by Herckes et al. (2002), while the other was a “classic” Teflon CASCC developed by Demoz et al. (1996). The collectors were run in parallel, at an airflow rate of about 25 m3min-1, and were thoroughly cleaned before sampling using ultrapure water (>18MΩcm, Millipore) for the CASCC and acetone (HPLC grade 99.9 %), followed by ultrapure water for the ss-CASCC. Blank samples were taken routinely between each fog event by nebulising ultrapure water on the CASCC and ss-CASCC. Samples collected with the ss-CASCC were analysed for organic acids, while samples collected with the CASCC were analysed for pH, inorganic ions, and elements. Samples for major ion analysis were prepared immediately after collection from the CASCC by pouring 1 mL into a plastic vial and were stored at -18°C. Samples for organic acid analysis were prepared immediately after collection from the ss-CASCC by pouring a 950 µL sample of non-filtered fog water and 50 µL chloroform into a 1.5 mL glass vial and storing it at 4 °C. Aliquots for dissolved trace metal analysis were prepared by filtering the samples through a 0.22 µm filter (PES, Millipore) and acidifying them with high-purity nitric acid (EMD Chemicals, 67 %–70 % Omni Trace Ultra) to a pH of less than 2.

Nine fog samples were collected (Table 1) through five fog episodes (see Sect. 4.1 for details of the fog occurrences). In addition to these fog samples, two samples of coastal seawater were collected on the seashore in pre-cleaned high-density polyethylene (HDPE) bottles on 10 September (at 16:30 UTC) and 11 September (at 13:17 UTC). They were aliquoted and analysed in the same way as the fog samples. To determine the amount of water collected with the CASCC and ss-CASCC, the sample bottles were weighed before and after sampling.

Cloud droplet measurements were conducted with a fog monitor 100 (FM-100, Droplet Measurement Technologies Inc., Boulder, CO, USA) located on a scaffolding about 2.5 m high. This forward-scattering spectrometer probe provides the droplet number concentration, size distribution, and liquid water content for particle sizes with an optical equivalent diameter of 2–50 µm. Droplet size is calculated on the basis of laser-scattered light, employs Mie theory for spherical water particles with a refractive index of 1.33, and is classified into 32 bin sizes of width 2 or 3 µm. The instrument was calibrated with glass beads of known diameters and refractive indices. Despite large uncertainties (Guyot et al., 2015; Spiegel et al., 2012), it is one of the most commonly used instruments for documenting the microphysical properties of fog (Gonser et al., 2012; Gultepe et al., 2006, 2009; Mazoyer et al., 2022; Niu et al., 2012; Wagh et al., 2023). Data were recorded every second and subsequently aggregated to 1 min averages.

The pH was measured by a pH meter (9110DJWP pH probe from Thermo Scientific) as soon as possible after sampling on non-filtered samples. The pH meter was calibrated daily using pH 4.01, 7.00, and 10.01 buffers.

Water-soluble ions were analysed by ion chromatography (IC, Dionex Corporation). A dual-channel Dionex ICS-3000 ion chromatograph equipped with conductivity detectors was used to measure the cations and anions. For the cation analysis, an eluent generator provided a concentration of 20 mM methanesulfonic acid at a flow rate of 0.5 mLmin-1 to perform the separation on a Dionex IonPac CS12A analytical column (3mm×150mm). The anion analysis was performed on a Dionex IonPac AS14A analytical column (4mm×250mm) using 1 mM sodium bicarbonate/8 mM sodium carbonate eluent at a flow rate of 1 mLmin-1. The complete runtime for both was 17 min. No fog samples showed concentrations below the detection limits, except for phosphate (two fog samples, all blanks, and the sea samples).

Organic acids including acetate, propionate, formate, methanesulfonate (MSA), glutarate, succinate, malonate, maleate, and oxalate were analysed using a Dionex DX-500 IC with a gradient pump, conductivity detector, and self-regenerating anion suppressor. A 20-fold dilution was used for formate quantification. Separation was performed using a Dionex AS-11HC analytical (4mm×250mm) column using a sodium hydroxide gradient at a flow rate of 1.5 mLmin-1. The complete run time was 65 min.

For dissolved trace metal determination, fog samples were aliquoted in the field into pre-cleaned (with nitric acid) HDPE bottles and stored at 4 °C until analysis. Prior to analysis, samples were filtered through a 0.22 µm filter (PES, Millipore) and acidified using high-purity nitric acid (EMD Chemicals, 67 %–70 % Omni Trace Ultra) to a pH of less than 2. The samples were analysed for 40 trace metals using a Thermo Electron X-Series 2 quadrupole inductively coupled plasma mass spectrometer (ICP-MS). All elements were analysed using the kinetic-energy-discrimination (KED) mode and, as the collision cell gas, a hydrogen (7 %) and helium mixture. Scandium, germanium, yttrium, indium, and bismuth were used as the internal standards. As for the TSP samples, dissolved metals are defined operationally as passing through a 0.2 µm filter, which may include nano-sized particles as reported in previous studies (Giorio et al., 2025; Shi et al., 2009). One difference between the concentrations of dissolved metals in fog and TSP samples may arise from the pH of the extraction solution, which was in both cases autogenic but may be more acidic for TSP samples (not measured) compared to fog samples (measured as being close to neutral).

Iron speciation, including the identification of the presence of iron–organic ligand complexes and the partitioning of iron species in the II and III oxidation states, was investigated by XANES at the FeK edge. The XANES analysis was performed at the SAMBA (Spectroscopies Applied to Materials based on Absorption) line at the SOLEIL synchrotron facility in Saint-Aubin, France (Briois et al., 2011). A Si(220) double-crystal monochromator was used to produce a monochromatic X-ray beam, which was 4000µm×1000µm in size at the focal point. The energy range was scanned from 7050 to 7350 eV at a step size of 0.2 eV.

Standards of Fe(II) and Fe(III) oxides, hydroxides, salts, and oxalate complexes (see Formenti et al., 2025, for details) were prepared as round pellets of 10 mm in diameter in a matrix of boron nitride (BN, molar mass 24.819 gmol-1) using an automatic hydraulic press (Atlas Power T8, Specac) available at the chemistry laboratory at SOLEIL and operated at a pressure of 5 t. The pellets were analysed in transmission mode. The TSP and fog samples were analysed in fluorescence mode in the same energy range and at the step resolution of the standard references. As in Formenti et al. (2014) and Caponi et al. (2017), samples were mounted in an external setup mode. The TSP samples were analysed as detailed in Formenti et al. (2025). For the fog samples, several aliquots of 100 µL solution were deposited on a Kapton sticky film and left to evaporate until a significant deposit was observed visually and the minimum Fe concentration of 0.4 µgcm-2 was obtained. The deposit was finally protected by sticking another portion of Kapton film on top of it, forming a pellet ready for analysis.

Data analysis was conducted with the FASTOSH software package developed at SOLEIL. The Fe speciation was obtained by the least-squares fit of the measured XANES spectra based on the linear combination of reference spectra of oxides, clays, oxalates of Fe(II) and Fe(III), and pyrite. Fits were conducted on the first derivative of the normalised spectral absorbance in the energy region between 7100 and 7180 eV, corresponding to -30 and +50eV of the K edge. The results of the fits were evaluated on the basis of the resulting chi square (χ2). Additional details of the reference standards and data analysis are reported in Formenti et al. (2025).

The coarse fraction of a fog sample was filtered out by aspirating a fog sample through a 0.2 µm filter. The supernatant was then allowed to sediment in order to remove any small but colloidally unstable particulate matter by resting an Eppendorf tube upright for 24 h. The resultant suspension was then sampled for analysis.

A Malvern Panalytical Zetasizer Nano was utilised to perform DLS on the prepared fog samples. The angle and intensity of light scattering caused by suspended particles can be correlated with particle size, thereby investigating the hydrodynamic particle size distribution of the suspension and confirming the presence of a colloidally stable phase.

Fog samples were prepared in the same manner for TEM. A 35 µL droplet of the suspension was deposited onto a TEM grid (Cu–C, 300 mesh) before blotting off the excess with filter paper. A Thermo Scientific (FEI Company) Talos F200X G2 TEM was used to capture images of the nanoparticles in the fog samples. Energy dispersive X-ray spectroscopy (EDS) was used to identify chemical elements occurring within the aforementioned nanoparticles and to perform elemental mapping within a nanoparticle to understand the distribution of elements. A Super-X EDS detector system with four windowless silicon-drift detectors was utilised for EDS analysis.

A weather detector (Vaisala PWD 22/52) provided 1 min average horizontal visibility using the forward-scatter measurement principle.

Standard meteorological parameters (wind speed and direction, air temperature, and relative humidity) were measured by a compact weather station (models CE155N, CE157N, and BA711; Cimel Electronique, Paris, France) installed 10 m above ground level on an extendable mast on the roof of the PEGASUS facility.

The average aerosol liquid water (ALW) content and pH for the aerosol fraction were calculated using the comprehensive version of the Extended Aerosol Inorganics Model E-AIM IV (Clegg et al., 1992, 1998; Wexler, 2002). The averaged relative humidity (RH) and temperature over the sampling period and aerosol concentrations of NH4+, Na+, Cl-, NO3-, and SO42- were used as input for the calculation. Ion balance was achieved by adding an appropriate amount of H+ or converting an appropriate amount of NH4+ into NH3. Acetic acid and oxalic acid were also included in the calculation and allowed to dissociate in solution (Clegg et al., 2003). Partitioning equilibria with gas components and solid precipitation were also considered. K+, Ca2+, Mg2+, and PO43- were not considered in the calculation as these species are not included in E-AIM. This can lead to an underestimation of the real pH of aerosols. The ISORROPIA II model could be used to consider these species, but a few studies have pointed out that ISORROPIA does not perform well at very high relative humidity (such as in this campaign) when run in reverse mode (i.e. when only particle-phase measurements are available) (Haskins et al., 2018; Hull et al., 2025; Peng et al., 2019). In addition, the absence of carbonate measurements can contribute to the uncertainty around the real aerosol pH. Including formic acid in the calculations did not show any difference, and therefore it was ignored.

Chemical speciation of metal ions in deliquescent aerosol was computed using the Visual MinteQ 3.1 model (available online at https://vminteq.com, last access: 2 September 2025). For the speciation, the following species were considered: Na+, NH4+, K+, Mg2+, Ca2+, Cl-, NO3-, SO42-, PO43-, oxalate, acetate, Al, Cu, Ni, Mn, Cr, Fe, Pb, and Zn. For metal ions, the dissolved concentration measured with ICP-MS was used. The oxidation state of the latter metal ions was assumed to be the following: Al(III), Cu(II), Ni(II), Mn(II), Cr(III), Pb(II), and Zn(II). In the case of Fe, we assumed that Fe(II) and Fe(III) were present in a 1:1 ratio as used in previous studies (Giorio et al., 2022b; Scheinhardt et al., 2013; Shahpoury et al., 2024a). This 1:1 ratio was used to obtain an average picture of Fe speciation, considering the dynamic nature of the Fe2+/Fe3+ couple in atmospheric aerosol with expected higher Fe2+ concentrations during the day and higher Fe3+ concentrations during the night (Mao et al., 2017). For the speciation, liquid-phase concentrations of the dissolved complex-forming compounds were calculated using the ALW obtained from E-AIM IV, assuming initially a complete dissolution. The pH was fixed at the value calculated by E-AIM IV, and the mean temperature over the sampling time was used. The speciation model was run according to the procedure described in previous studies (Giorio et al., 2022b; Scheinhardt et al., 2013). When specific chemical species were predicted to be present under oversaturation conditions, the calculation was repeated, allowing precipitation.

Chemical speciation of metal ions in fog was calculated at the measured pH value by means of the PITMAP software (Di Marco, 1998; Tapparo et al., 2020). Briefly, mass balance equations are solved. That is, the species concentration at equilibrium is obtained using the Newton–Raphson method (Press et al., 2007). Speciation was done considering metal ions and ligands quantified in the fog samples: formate, acetate, propionate, oxalate, malonate, succinate, glutarate, maleate, methanesulfonate, NO3-, Cl-, SO42-, PO43-, Mg(II), Ca(II), Sr(II), Ba(II), V (in the form VO2+), Cr(III), Mn(II), Fe(III), Co(II), Ni(II), Cu(II), Zn(II), and Al(III). For all metal ions, the thermodynamically stable oxidation state was considered. The input thermodynamic data (stoichiometry and stability constant of the complexes, including metal–aqua complexes) were obtained from the literature (The IUPAC Stability Constants Database, 2006).

Unlike Visual MinteQ, PITMAP does not explicitly calculate the amount of precipitated solids. Speciation of metals obtained by Visual MinteQ and PITMAP for three fog samples (H603, H901, and H1002) was qualitatively equal, with minor differences being observed only for species at very low concentrations (<10-7M) (see Sect. S1 in the Supplement).

Figure 1 illustrates the meteorological conditions during the campaign, which were characterised by remarkable stability in terms of temperature (around 12 °C) with a narrow range (<5°C) and humidity (RH 95±4%, minimum 78.5 %, and maximum 100 %), which is a clear indication of maritime influence. At the same time, a persistent stratocumulus cloud deck kept solar irradiance below 600 Wm-2 (Giorio et al., 2022a).

The visibility data show the presence of fog events (classified as visibility<1km) starting from 3 September 2017 and becoming persistent for extended periods from the night of 6–7 September 2017 (Fig. 1). Two fog events were particularly intense over 3 consecutive days, 9–11 September, with visibility reduced to below 100 m and with the liquid water content up to 0.3 gm-3.

Wind patterns were characterised by a gentle (<4ms-1) occasional land or sea breeze superimposed with moderate to strong (up to around 10 ms-1) breezes coming predominantly from the NW and NNW and from the SW (Fig. 1). Notable exceptions were 4 September (NE and NNW winds), 5 September (SWW winds), 6–9 September (SSW winds), and 11 September (SW winds). Air masses predominantly travelled along the coastline (NNW–SSE) before reaching the sampling site in Henties Bay. These air masses often travelled above the ocean or the Namib, transporting sea spray, marine biogenic emissions, and mineral dusts. When the winds blow from S and SE, sporadic anthropogenic influences from the towns of Henties Bay and Walvis Bay can be expected (Giorio et al., 2022a). During fog events, the wind speed is almost always lower than 4 ms-1 from the SSW to WNW. The observed fog was henceforth mostly due to the advection of marine stratocumulus or was locally formed, as confirmed by the fog microphysics (Weston et al., 2022) and chemical composition as shown in the sections below.

The major ions showed extremely high concentrations in fogs compared to other studies in coastal areas (Table 2), but this was in agreement with the findings of Gottlieb et al. (2019) for various sites in Namibia, while Fe, Zn, and Cu concentrations were of the same order or even reduced compared to other remote sites (Table 2) due to scarce vegetation sources in the hyperarid Namib region, where there is hardly any vegetation.

Table 2 shows that the fogs were more diluted than local seawater in terms of major ions (Na+, Cl, K+, NH4+, Mg2+, Ca2+, NO3-, and SO42-). This result is also corroborated by microphysics measurements that show rapid changes between fog, mist, and sea spray, the latter contributing significantly to reducing visibility (Formenti et al., 2019; Weston et al., 2022). Conversely, several elements and compounds, such as Al, Ti, Ga, Fe, P, nitrate, and all rare Earth elements (see Table S1 in the Supplement for elements not reported in Table 2), were more concentrated in coastal fog than in seawater. The elevated nitrate concentrations can be attributed to the atmospheric processes affecting the important marine phytoplanktonic emissions of reactive nitrogen species (organic nitrogen and/or NO), ending, by photooxidation, in particulate nitrate formation (Altieri et al., 2021) and henceforth explaining the presence of nitrate in the remote oceanic atmosphere. The correlation of the nitrate concentrations with markers of secondary chemistry, such as oxalate and MSA (R2=0.89 and 0.53, respectively), in our coastal fog samples supports this hypothesis. Carboxylic and dicarboxylic acids, i.e. notably highly oxidised species such as formate, acetate, succinate, and oxalate, also showed high concentrations in fogs. A remarkable correlation between formate and acetate (R2>0.99) suggests that they have similar sources or formation pathways. The high formate-to-acetate ratio (2.02) indicates the more important secondary formation of carboxylic acids in fog water as compared to other possible sources (Li et al., 2020). Surprisingly, U was more concentrated in seawater than in fog. A source attribution analysis is summarised in the Supplement in the form of enrichment factors (see Figs. S4 and S5 in the Supplement). Besides the local seawater, to represent the influence of the marine source, we present the enrichment factors calculated with respect to Al concentrations in the aerosol phase measured concurrently with the fog and representing the local dust content during the campaign (Formenti et al., 2025).

Generally, the chemical composition of the fog samples shows a large marine and mineral dust influence accompanied by highly processed organics which may have been formed in situ or in clouds and then advected on the coast. Indeed, more and more studies show atmospheric volatile organic compound (VOC) emissions from marine planktonic microorganisms. These compounds can then undergo atmospheric photochemistry inducing the formation of new ultrafine particles which may grow in size and contribute to cloud condensation nuclei (Giorio et al., 2022a; Halsey et al., 2017; Mungall et al., 2017; Peltola et al., 2022).

Solubility of metals in TSP samples is discussed for the whole campaign in Formenti et al. (2025). Here, we focus the discussion on the comparison between dissolved metals in pairs of TSP and fog samples. For the comparison, fog water-dissolved metal concentrations (µM) were transformed into air concentrations (ngm-3) using the liquid water content (gH2Om-3). This allows for comparison of fog concentrations with dissolved metal concentrations in TSP samples. The ratio of fog concentrations to TSP concentrations is shown in Table 3. When concentrations of dissolved metals in TSP samples are compared with concentrations in fog concurrently collected, higher concentrations of Al, Fe, Cu, and Ni are found in fog (Table 3). The Al, Fe, Cu, and Ni concentrations in fog are up to a factor of 4.7, 17, 5, and 6.5 higher than dissolved concentrations in TSP samples. Similarly, the Cr concentration in fog was up to 14.5 times higher than the dissolved concentration in TSP samples and Ti was detected in fog but not in the corresponding samples of the dissolved fractions of TSP samples.

This result suggests that the chemical processing of airborne particles in fog would enhance metal solubility, such as the formation of metal–ligand complexes and/or photochemical processes, if they are both locally produced and TSP samples act as condensation nuclei for fog formation. Previous studies have shown that Fe solubility can be driven by the formation of metal–ligand complexes with oxalates (Giorio et al., 2022b; Paris et al., 2011; Paris and Desboeufs, 2013). Similar processes can also impact Cu and Al solubility (Giorio et al., 2022b). Other organics, such as malonate, tartrate, or humic-like substances (HULISs) (Paris and Desboeufs, 2013), can also influence Fe solubility. This is also the case for formate and acetate, which, despite being weaker complexing agents, are better pH buffers (Upadhyay et al., 2011). Fe solubility is strongly dependent on pH (Desboeufs et al., 1999, 2003). However, this pH dependency does not explain the observations in this study, as Fe is supposed to be more soluble at acidic pH than at the neutral pH of the fog samples (Fig. 2).

A previous study of the same site showed that Fe solubility in PM10 was correlated with methanesulfonic acid concentrations, suggesting that photochemical reduction may be responsible for the enhanced Fe solubility (Desboeufs et al., 2024). Such a process would be compatible with the observations of a higher Fe solubility in fog samples, with the largest differences in dissolved Fe concentrations observed in samples collected during daytime (Table 3). It is worth keeping in mind that liquid water content (LWC) may be underestimated in nighttime fog due to the continuous running of the fog collectors until morning, which leads to an overestimation of the volume of air sampled. Formenti et al. (2025) show that fluoride could also play a role in Al, Ti, Si, and Fe solubilities. However, as all fog samples were collected in a period of low fluoride concentrations, the possibility can be excluded that this species could play a role in the enhanced dissolved metal concentrations in the fog samples.

The speciation of the most abundant metals (see Sects. 3.2 and 3.3) in the deliquescent TSP as well as in the fog samples is calculated to investigate which species are present in solution and could drive the solubility of each metal.

ALW content and pH, estimated using the E-AIM IV model, were in the ranges 0.122–412 mgm-3 and 0.1–3.2, respectively (Fig. 2). The high ALW indicates that the aerosol fraction, dominated by sea salt, was likely always in a deliquescent state due to the mild temperatures and elevated RH (observed from 6 September until the end of the campaign) (Fig. 1). Generally, the aerosol acidity was inversely proportional to the ALW content. It is worth noting that the calculated aerosol pH is very low (pH<3) compared to the fog samples (neutral pH) collected concurrently. The measured pH values for the fog samples ranged from 5.95 to 7.8, with a median value of 7.28, similar to that of seawater (Table 2). These pH values are toward the upper end of values reported in the literature, which, however, refer to urban and agricultural areas (Ge et al., 2022; Gundel et al., 1994; Paglione et al., 2021; Pye et al., 2020; Wen et al., 2005). The higher pH measured in this study in the fog samples is likely due to a high water content as well as the dominance of the sea spray component. The pH in fog might be at least partly driven by the presence of carbonates, as evidenced by the moderate correlation between -log⁡([Ca(II)]) and pH (r2=0.21), similar to what happens in seawater (Jiang et al., 2019). The E-AIM IV model does not consider K(I), Mg(II), Ca(II), and carbonates and therefore may underestimate aerosol pH compared to what is expected for sea spray aerosol (Pye et al., 2020).

Speciation of metal ions in deliquescent aerosol depended on ALW content, pH, and the number of inorganic and organic ligands in the samples. Due to the high ALW and low pH, a large fraction of the metal ions is in free form, with such a fraction being dominant during the second part of the campaign characterised by higher RH and fog events. For most of the metals, another significant fraction is bound to inorganic compounds such as SO42- for Al(III), Cr(III), Cu(II), and Ni(II) and Cl- for Ca(II), Mg(II), Mn(II), and Zn(II). Fe is the only exception, with Fe(II) being equally distributed between binding SO42- and Cl- and Fe(III) also being complexed with organics, in particular oxalate (Fig. S6 in the Supplement). In addition to Fe(III), a significant fraction of Al(III) and a minor fraction of Cu(II) are also predicted to bind oxalate. In addition, in the first part of the campaign when ALW was relatively lower, part of Ca(II) was calculated to precipitate as CaSO4. Given that pH was likely underestimated in our calculations, the real speciation picture may be shifted in favour of a larger proportion of hydroxide formation for some metal ions, which could limit their ability to dissolve in water.

Fog samples were characterised by much higher water content, ranging from 16–1117 mgm-3, and close to a neutral pH. Given the extremely different pH conditions, we would expect the speciation picture to be completely different compared to that of TSP.

The most striking and surprising difference is observed for Al(III) and Fe(III), which, according to our calculation, would form insoluble hydroxides in fog water (Fig. 3) and therefore should not be in solution unless they are present in a colloidal form (Shi et al., 2009). This result runs counter to observations from ICP-MS analysis, which show an even higher concentration of Al and Fe in solution in the fog samples compared to the dissolved concentration in TSP. In the case of Al(III), only a minor component is predicted to be bound with organics such as oxalate and succinate (Fig. 4b). Ni(II) speciation in fog shows similar results compared with TSP (Fig. 3), with the only difference being that Ni(II) is partly present as soluble hydroxide in fog in addition to being present in free form and bound to SO42- as in TSP. Only a minor component of Ni(II) is predicted to be bound to organics such as oxalate, malonate, and succinate (Fig. 4b). A notable difference between the two types of samples is instead observed for Cu(II), which becomes bound to oxalate and malonate for a fraction between 19 % and 43 % (Fig. 4a) in the fog samples, while in TSP only a minor fraction (<7.3%) was bound to the organics. In the case of Cr(III), speciation in TSP showed an almost equal distribution between free-form and bound SO42- for the only two samples in which Cr was detected in the dissolved fraction, while in fog it was mainly present as soluble hydroxide and a mixed PO43-–hydroxide complex (Fig. S7 in the Supplement).

Iron speciation was measured experimentally in order to elucidate the processes that may increase iron solubility in the fog samples compared to the TSP samples. Table 4 and Fig. 5 show the results of the linear combination fitting of the XANES spectra of the TSP and fog samples using reference spectra that are representative of various iron oxides, clays, pyrite, and complexes with oxalate formed by either Fe(II) or Fe(III). The results show that, for both TSP and fog samples, iron oxides and clays are by far the major components, as expected for samples affected by the transport of mineral dust. In the case of TSP, iron oxides and hydroxides contribute 39 %–52 % and iron clays contribute 38 %–52 % (Table 4), making the majority of Fe insoluble. In fog samples, the contributions of these two classes of minerals are more variable, with iron oxides and hydroxides contributing 27 %–56 % and iron clays contributing 32 %–64 % (Table 4). Iron-oxalate complexes, in particular those of Fe(III), contributed to some samples of both TSP and fog. Excluding samples that were close to the detection limit or affected by diffraction, iron oxalate contributed up to 6 % in TSP samples and up to 8 % in fog samples. Pyrite was found to contribute up to 7.8 % of the total iron concentration, which was attributed to the transport of fugitive dust from smelting areas (Formenti et al., 2025).

The presence of a significant amount of iron oxalate in more than one fog sample (Table 4) could explain the enhanced solubility of iron in fog compared to TSP samples. This amount is comparable to the dissolved fraction of iron (<2% in TSP in these samples). The results of the XANES measurements run counter to the results of the thermodynamic modelling that predicted iron to precipitate in fog samples in the form of hydroxides, but they are consistent with ICP-MS measurements that showed that dissolved Fe was indeed present in the fog samples. On the other hand, the thermodynamic modelling results are qualitatively consistent with XANES measurements of TSP samples, for which oxalates of iron complexes were expected and indeed experimentally detected. We hypothesise that iron-oxalate complexes are formed through aqueous-phase processing of iron during particle activation into droplets when the water content is lower and the pH is also lower than in the collected fog samples. These complexes may then remain in the fog droplets in a kinetically stable form before being converted into insoluble hydroxides. Additionally, Marcotte et al. (2020) showed that iron-oxide solubility in mineral dust decreases with increasing particle size. Therefore, as activation to droplets is a size-dependent process, there may be fractionation of soluble iron whereby, for example, smaller particles with a higher soluble iron content contribute more to fog droplets compared to larger particles with a lower soluble iron content. The XANES measurements also showed that iron was predominantly present in the Fe(III) oxidation state. This result would suggest that photochemical reduction to Fe(II) was not a significant process in the enhanced iron solubility in the fog samples.

In addition to organic iron formation during droplet activation, colloidal iron may be present in fog (Gledhiir and Buck, 2012), which would be detected as being dissolved in ICP-MS measurements. To test this hypothesis, TEM and DLS measurements were performed on fog samples before and after filtration through a 0.22 µm pore-size filter.

Particles identified by DLS in the sub-100 nm range, after filtration and sedimentation, reveal the presence of a colloidal phase within the fog samples (Fig. 6a). TEM confirmed this hypothesis, as distinct particles with diameters ranging from 50 to 150 nm were identified (Fig. 6b). These are likely to be the same particles observed using DLS. EDS showed that these particles had cores containing silicon, aluminium, iron, and sulfur, among other less prominent elements (Fig. 6c).