Tracking the changes of iron solubility and air pollutants traces as African dust transits the Atlantic in the Saharan dust outbreaks

Abstract We studied the solubility, in real sea water, of iron present in the African dust outbreaks that traverse the Atlantic. Based on measurements of soluble iron (sFe) and aerosol chemistry, we found iron solubilities within the range of 0.4–1.8% in Tenerife, 0.4–3.1% in Barbados and 1.6–12% in Miami. We apportioned the concentrations of sFe between the three sources and processes that we identified: (1) dust, (2) heavy fuel oil combustion emissions, associated with an excess of vanadium and nickel, and (3) atmospheric processing, which is influenced by acidic pollutants. We tracked the propagation of the dust-front of the African dust outbreaks across the Atlantic, which are associated with dust peak events at the impacting sites. During the westward transport across the Atlantic, the contribution to sFe from dust decreased (63%, 42% and 11% in Tenerife, Barbados and Miami, respectively), whereas the contribution due to atmospheric processing increased (26%, 44% and 80% in Tenerife, Barbados and Miami, respectively). In these Saharan-dust outbreaks, the concentrations of sFe due to heavy fuel oil combustion were significantly lower (mostly


H I G H L I G H T S G R A P H I C A L A B S T R A C T
• Iron solubility increases from ~0.7% to 4.7% as Saharan dust fronts transit from Africa to the Americas. • Dust, heavy fuel oil combustion and atmospheric processing are identified as sources of soluble iron. • Fresh dust accounts for 63, 43 and 9% of soluble iron in Tenerife, Barbados and Miami. • Aged dust accounts for 26, 45 and 74% of soluble iron in Tenerife, Barbados and Miami. • Oil combustion accounts for 10, 12 and 16% of soluble iron in Tenerife, Barbados and Miami.

Introduction
Marine microorganisms account for ~ half of Earth's primary productivity, i.e. the conversion of carbon from carbon dioxide to organic molecules via photosynthesis and chemosynthesis, a process that also replenishes the oxygen budget (Chapman, 2013). This process requires nutrients to support the growth of primary producers. In many regions of the open ocean, atmospheric deposition is the main supplier of those nutrients. Nitrogen and, to a lesser extent, phosphorus have historically been identified as limiting macronutrients (Falkowski, 1998;Moore et al., 2013). Iron is a limiting micronutrient for phytoplankton, especially in the high-nitrate, low-chlorophyll (HNLC) regions, which account for 30% of the oceans, particularly in the Southern Ocean and the equatorial and subarctic Pacific (Mahowald et al., 2018). The atmospheric deposition of Fe is estimated to support 30-50% of global export productivity, whereas that of nitrogen, silicon and phosphorus is estimated to support 3-5%, ~0.2% and ~0.1% of global export productivity (Duce et al., 2008;Jickells et al., 2014;Krishnamurthy et al., 2010), respectively. It is thought that desert dust was the main source of iron to the ocean in pre-industrial times. Ice core records suggest that increased dust inputs to the ocean may have reduced atmospheric CO 2 in glacial versus interglacial periods, linked to iron modulated primary productivity (Martínez-Garcia et al., 2009;Ridgwell and Watson, 2002).
Since the advent of the Anthropocene, huge amounts of chemicals have been transferred from the atmosphere to the ocean, where they act as nutrients (N, Fe and P), toxins (e.g. Pb, Cu and Hg) and pH regulators (carbon dioxide), with implications on ocean ecosystems, carbon sequestration and climate (Duce et al., 2009;Mahowald, 2011). The composition of the marine biota community and its biological productivity may be altered due to the inputs of co-limiting trace metal (e.g. Zn, Co and Mn; Ahlgren et al., 2014;Browning et al., 2014;Saito et al., 2008), whereas other elements (e.g. Cu) may be toxic for plankton (Jordi et al., 2012;Paytan et al., 2009). Fertilization experiments have found that increased growth rates linked to iron-addition also shift the community composition and the biomass accumulation (de Baar, 2005;Boyd et al., 2007). The response of these marine communities to such deposition of chemicals has also implications in the Ocean Deoxygenation .
It is generally accepted that the solubility of Fe (i.e., the ratio of soluble to total Fe, %) deposited in the ocean will affect its bioavailability (Schulz et al., 2012). The solubility of Fe in dust samples collected from soil is typically low (as low as 0.1%), i.e. much lower than that observed in the Fe contained in the atmospheric aerosols (~0.45% for dust and up to 80% for other aerosols; Schroth et al., 2009;Shi et al., 2009Shi et al., , 2012. Mineralogy has been found to play an important role on the inherent solubility of dust (Gaston, 2020;Journet et al., 2008). High values of Fe solubility in the atmospheric aerosols are attributed to the emissions of pyrogenic particles (Ito et al., 2019) and atmospheric processing (Gaston, 2020;Kanakidou et al., 2018). The Fe solubility of pyrogenic particles may reach values of ~20% for coal and for biomass burning fly ashes and ~80% for oil combustion, specially low grade fuels, as heavy fuel oil (HFO) (Ito, 2013;Wang et al., 2015). The high Fe solubility in fly ashes linked to HFO combustion is attributed to ferric sulphate and to aggregated nanocrystals of magnetite (Fu et al., 2012;Schroth et al., 2009), formed at temperatures >800 • C, followed by sulphuric acid condensation (Sippula et al., 2009). In coal fly ash, rapid acid chemistry on aluminosilicates glasses formed at high temperature combustion followed by cooling process contributes to quick Fe leaching . During transport, atmospheric processing can increase the solubility of iron containing minerals (Kanakidou et al., 2018) by a set of processes still under research, including proton-promote dissolution due to the reaction of iron with inorganic and organic acids (Ito, 2015;Longo et al., 2016;Meskhidze, 2003;Shi et al., 2009), ligand-controlled dissolution due to the presence of organic ligands, as oxalate, that bind Fe in organic soluble complexes (Meskhidze et al., 2017;Wozniak et al., 2013) and photo-reduction involing the reduction of Fe(III) to Fe(II) (Chen and Grassian, 2013;Johnson and Meskhidze, 2013;Paris et al., 2011). Variability in pH and acidic conditions contributes to the formation of reactive iron nanoparticles in cloud (Shi et al., 2009). Several factors affect the efficiency of these iron dissolution processes including particle size (Baker and Jickells, 2006;McDaniel et al., 2019), dust mineralogy including the mineral form of iron present and the presence of semiconductor minerals, light, and the presence and types of acids, inorganic and organic ligads (Fu et al., 2010;Hettiarachchi and Rubasinghege, 2020;Rubasinghege et al., 2012;Wiederhold et al., 2006). In cloud processing of dust and pH variability contribute to the neo-formation of iron nanoparticles that increases iron reactivity (Shi et al., 2009).
Source-specific values of Fe solubility have been used in atmospheric transport models for quantifying the natural and anthropogenic inputs of soluble Fe to the ocean (e.g. ~ 0.45% for dust and 20%, 80% and 18% for coal, oil and biomass-burning fly ashes, respectively (Ito, 2013;Ito and Feng, 2010;Johnson and Meskhidze, 2013;Wang et al., 2015). Atmospheric chemical transport models have also incorporated acid dissolution schemes, based on laboratory experiments and kinetic process, in which the initial values of Fe solubility in combustion fly ashes and dust aerosol evolves depending on mineralogy, meteorology, pH, solar radiation and the mixing with inorganic (sulphuric, nitric and hyper-chloric) and organic (oxalate) acids (Ito and Feng, 2010;Ito and Xu, 2014;Johnson and Meskhidze, 2013).
The increasing volume of experimental data is providing a broad view on the variability range of iron solubility over the ocean (Baker et al., 2016;Baker and Jickells, 2017;Ingall et al., 2018), which shows that there are still major uncertainties regarding the sources and atmospheric processes contributing to the enhancements in Fe solubility (Aguilar-Islas et al., 2010;Ingall et al., 2018;Ito et al., 2019;Kanakidou et al., 2018;Mahowald et al., 2018;Meskhidze et al., 2016b;Raiswell et al., 2017). Models suggest that the amount of soluble Fe deposited in the ocean may have doubled since the industrial revolution (0.06-0.12 Tg/y − 1 ), not only because of the emissions of pyrogenic iron but also because of the acid dissolution of dust (Ito and Shi, 2016). There is a clear need to develop tools for apportioning measured soluble iron between the different sources and for assessing the relative importance of atmospheric processes in contributing to increasing iron solubility (Baker et al., 2016;Mead et al., 2013).
The majority of dissolved iron in the oceans is bound to iron-binding ligands or colloids (Buck et al., 2015). In the surface of the tropical and subtropical North Atlantic ocean, underlying the Saharan-dust plume, dissolved iron is predominantly found as colloids (Fitzsimmons et al., 2015). In contrast, no such preference is observed in regions where dust concentrations are low such as the Southern Ocean (Boye et al., 2010) and the Subarctic Pacific (Nishioka et al., 2001), where deposition of soluble iron is dominated by combustion sources (Ito, 2015).
We present the results of an experiment focused on studying the change of iron solubility after African dust plumes had crossed the Atlantic. We tracked the propagation of dust outbreaks across the North Atlantic and studied the evolution of aerosols composition, iron solubility and source-tracers. Our objective was to measure soluble iron concentrations before and after transit, to link changes to the impact of specific source types, and to identify the processes that might lead to changes in iron solubility.

Sampling and chemical determinations
From 30 June to September 1, 2015 we collected daily samples of aerosols in three sites of the North Atlantic: (i) at Izaña Observatory in Tenerife island (28.308 • N, 16.500 • W), located at ~2400 m above sea level (m.a.s.l.), directly exposed to the dusty air arriving from the inner Sahara, (ii) Ragged Point, on Barbados (13.165 • N,59.432 • W), ~45 m. a.s.l., exposed to the easterly trade winds blowing in the tropical central North Atlantic and (iii), Virginia Key (25.732 • N, 80.162 • W), an island 4 km east off the coast at Miami, ~30 m.a.s.l., directly exposed to the trade winds arriving from the tropical Atlantic and the eastern Caribbean. The three sites have produced multi-decadal dust records (Prospero, 1999(Prospero, , 2003Zuidema et al., 2019).
At the three sites, the samples were collected on Whatman™-41 cellulose filters at the flow rates of 30 m 3 /h in Tenerife and 40 m 3 /h in Barbados and Miami. A characterization of the blank of these filters and detection limits, within the frame of the GEOTRACES program, was presented by Morton et al. (2013). The analytical methods that we used were described previously  and in the Supplementary Material (SM). These analyses allowed the determinations of dust (Prospero, 1999), ions and cations (Na + where ' = ' indicate charge − 2 for sulphate; Prospero, 1999) and elemental composition (Na, Mg, Al, Si, S, Cl, K, Ca, Ti, V, Cr, Mn, Fe, Ni, Cu, Zn, Sr; Calzolai et al., 2015;Lucarelli et al., 2014Lucarelli et al., , 2018. We determined soluble iron in 55 samples collected before, during and after dust peak events. Soluble iron from the three sites was determined by extracting the samples in seawater (pH = 8.1) collected offshore in Northern Tenerife and filtered through a 0.45 μm filter (Ravelo-Pérez et al., 2016). The extraction of the soluble iron in seawater collected in the same site avoid uncertanites linked to heterogeneity in seawater composition, particularly organic ligands (Buck et al., 2006).
The uncertainity linked to soluble iron concentrations was estimated based on replicate analyses of the leaching that yielded a average precision of ±7% (n = 7 samples). The uncertainity linked to PIXE was determined (in ng/m 3 ) for each sample (considering X-ray counting statistics, areal densities of certified standards and aerosol sampling parameters); the average ratio of this uncertainity to the concentration value was ~10% for Fe and 14% for Al. Because of their low concentration, this ratio was higher for trace elements: ~47% for V > 5 ng/m 3 and ~30% for Ni > 0.9 ng/m 3 , and higher for lower V and Ni  and Miami (D). AOD at Tenerife (A), Cape Verde (B, Sao Vicente sitegrey triangles and Calhau -grey dots), Barbados (C) and Miami (D). Arrows O1 to O4 mark the location of the dust-front linked to the four identified dust outbreaks that transited the Atlantic. O4 didn't reach Miami.
concentrations. Iron solubility (Fe-S) was determined as: The uncertainity interval (Fe-S min , Fe-S max ) of each Fe-S data was determined considering the uncertainty of sFe (ΔsFe) and Fe (ΔFe): The width of the lower side (Fe-S -Fe-S min ) and upper side (Fe-S max -Fe-S) of the uncertainty interval, with respect to the Fe-S central position, was used to estimate Fe-S uncertainty. The average ratio of this uncertainity to the concentration value, in the whole data set, was 18% for Fe-S data. This uncertainity is higher than that of sFe and Fe due to the amplification effect of the ratios (equations (2a and 2b)). The same analysis yields values of 94% for the V/Al ratios, due to the low concentrations of V and the amplification effect of equations (2) and (3) ratios. The uncertainity intervals for sFe, Fe, Fe-S and the V/Al ratio are shown in some plots (Figs. 1B and 4A-4B).

Complementary data
Complementary data were used to track dust transport across the Atlantic: (i) Aerosol Optical Depth (AOD) measurments of the AERONET network at the three sampling sites and in two sites in Cape Verde (Calhau and Sao Vicente), and (ii) surface dust concentrations (μg/m 3 ) and column dust (kg/m 2 ) obtained with MERRA-2 model (Barkley et al., 2019;Gelaro et al., 2017). AOD is a measure of the attenuation (extintion) of the sunlight, caused by the atmospheric aerosols, when light propagates from the top of the atmosphere to the Earth surface, directly related to column dust load.

Source apportionment of soluble iron
We used two approaches, the minimum ratio and the minimum border line, to apportion soluble iron between the different sources. Theses two methods were previously used for segregating the contributions of primary and secondary sources of organic aerosols (Turpin and Huntzicker, 1995;Na et al., 2004) and of ultrafine particles (Kulmala et al., 2016;Reche et al., 2011;Rodríguez and Cuevas, 2007). As far as we know, this methodology is applied for the first time to soluble iron.
We used the excess of vanadium, with respect to its content soil dust content, as a proxy to trace HFO combustion emissions (Ault et al., 2010;Sedwick et al., 2007;Trapp et al., 2010a). Thus, the vanadium linked to dust (V d ) and to HFO combustion (V c ) was determined as: where, (V/Al) d is the maximum value of the vanadium to aluminium ratio in North African dust. Several V/Al ratios observed in soils have been published, ranging from (dimensionless mass ratios) 0.9⋅10 -3 (Taylor and Mclennan, 1985), to 1.17⋅10 -3 (Mason, 1966;Taylor, 1964) and to 1.3⋅10 -3 (Rudnick and Gao, 2003) for the global average in the Earth crust and 1.13⋅10 -3 in North African soil desert dust (Eglinton et al., 2002). When applying the minimum ratio method, we used a value (V/Al) d = 1.15⋅10 -3 . Thus, we assumed that the V/Al in Saharan dust may experience a certain variability, with a maximum equal to this threashold value. In aerosol dust samples with a V/Al ratio < (V/Al) d , the measured vanadium is attributed to dust (equation (4a)) and V c = 0 (equation (5a)). In samples with a V/Al ratio > (V/Al) d , V d is calculated using the (V/Al) d (equation (4b)), whereas V c is calculated as the excess of vanadium with respect to this (V/Al) d threashold (equation (5b)). When applying the lower border line method, the characteristic (V/Al) d above which the iron solubility increases, linked to HFO combustion emissions, was experimentally determined for each site, as detailed below.
Analytical data used in this study and the source apportionment calculations are available in Rodríguez et al. (2020a). Figure 1A shows the average dust distribution during our study and the main industrial areas of North Africa. Dust transport occurs primarily in the Saharan Air Layer, an elevated air stream located at altitudes 1-5 km off North Africa and < 3-4 km in the western North Atlantic (Prospero and Carlson, 1972;Tsamalis et al., 2013). Dust inputs to the ocean are estimated in 450 Tg/year and ~43% occur in the Atlantic (Huneeus et al., 2011;Jickells, 2005). Thus, this study was Other approaches focus on mimicking in-cloud processing (e.g. sample extraction at pH 2 to 5 in deionized water) (Baker et al., 2016;Baker and Jickells, 2017;Buck et al., 2010;Meskhidze et al., 2016a). We will refer to soluble iron and iron solubility as sFe and Fe-S, respectively.  Fig. 2 shows dust concentrations and AOD in Tenerife, in Barbados and in Miami. AOD at Cape Verde was also included (Fig. 2B). We tracked dust transits across the Atlantic with the MERRA-2 model; Fig. 3 shown an example on 23-31 July 2015. Measurements and model showed a good agreement ( Figure S1-S2), a fact also observed in previous studies (Kramer et al., 2020).

Transits of dust across the North Atlantic
Massive exports of dust occur in events of pulsating nature, at the rate of ~1 or 2 events/10 days, linked to east-to-west shifts of meteorological systems into North Africa (Cuevas et al., 2017;Rodríguez et al., 2020b). These dust outbreaks start with a dust-front expanding (South-North) along the African coast, across thousands of kilometres, e. g. from Cape Verde to Canary Islands (Fig. 3A). Then the outbreak propagates westward (Fig. 3B-E). The arrival of a dust-front to each measurement site resulted in increases in dust and AOD (Fig. 2). We detected four massive dust outbreaks; the peak-dust events associated with each dust-front is highlighted in Fig. 2 (outbreaks 1 to 4, O1-O4). Fig. 3 shows outbreak O3. Dust-fronts first impacted on Tenerife and/or Cape Verde, 3-6 days later in Barbados and 7-12 days later in Miami. Dust-fronts expand (South-North) along thousand of kilometres, consequently, a dusty air-parcel that impact Tenerife does not necessarily impacts Barbados and/or Miami. Dust impacting at a specific site could be linked to different dust source regions in North Africa. This is corroborated by the back-trajectories (Stein et al., 2015) associated with O1 to O4 at each site ( Figure S3).
The arrival of the dust-fronts is associated with an increase in dust concentrations at each site, up to reach an average value, during the outbreaks O1-O4, of ~100, 53 and 19 μg/m 3 in Tenerife, Barbados and Miami, respectively (Table 1A). After the dust-fronts passed, dust concentrations decreased, resulting in a dust background levels across the Atlantic (<20, <10 and < 8 μg/m 3 at Tenerife, Barbados and Miami, respectively).

Iron solubility
At the three sites, iron showed a high linear relationship with aluminium and dust (r 2 > 0.9), with a slope (mass ratio, g/g) of Fe-todust within the range 0.034-0.039 and of Fe-to-Al 0.42-0.44 ( Figure S4). These values are within the range of the typical Fe-to-Al ratio (0.4-0.5) and mean content of Fe in African dust (3-4%) (Rodríguez et al., 2020b;Pérez García-Pando et al., 2016;Scheuvens et al., 2013). This points that dust is the major source of iron at the three sites, a fact observed previously (Ravelo-Pérez et al., 2016;Trapp et al., 2010aTrapp et al., , 2010b. Fig. 1B shows the plot of Fe-S versus Fe in Tenerife, Barbados and Miami. The inverse relationship observed in previous global studies (Sholkovitz et al., 2012), with low Fe-S (~0.4%) under high Fe concentrations and high Fe-S (~13%) under low Fe concentrations, is observed. There is a westward increase of Fe-S, with most of the values within the range 0.4-1.8% in Tenerife, 0.4-3.1% in Barbados and 1.0-12% in Miami. Previous studies also observed increases in Fe-S in dusty air, from 0.4% to 10%, after a few days of atmospheric transport (Ito and Feng, 2010;Luo and Gao, 2010;Meskhidze, 2005). Thus, we focused on identifying the sources and processes that contribute to such increases in Fe-S along the path of African dust outbreaks. Fig. 4A1-4A3 shows the scatter plot of sFe versus Fe. At each site, the data set tends to be distributed between two linear boundaries, one with a minimum slope S min and the other a maximum slope S max . These represent the minimum and maximum iron solubilities, respectively. In Tenerife and Barbados, we found S min = 0.004, which corresponds to a minimum Fe-S = 0.4% (orange dashed line in Fig. 4A1-4A2). This result is consistent with previous studies which found that the solubility of iron contained in desert dust aerosols is ~0.45% (Ravelo-Pérez et al., 2016;Sedwick et al., 2007;Shi et al., 2012;Theodosi et al., 2010), which is also the input value in many models (Ito, 2013;Ito et al., 2019;Ito and Feng, 2010;Johnson and Meskhidze, 2013;Wang et al., 2015). The data above this line (green arrow, Fig. 4A) of slope S min indicates that sources other than dust provide sFe, increasing the Fe-S up to a maximum value represented by S max (red line in Fig. 4A1-4A3). These additional inputs result in a maximum Fe-S = 1.8% in Tenerife (S max ; Fig. 4A1) and = 3.1% in Barbados (S max ; Fig. 4A2). At Miami (Fig. 4A3), the slopes S min (=0.016) and S max (=0.119) are significantly higher than in Tenerife and Barbados, a result that indicates that in addition to dust, there are important additional contributions to sFe.

Traces of heavy fuel oil combustion
Fe-S tends to increase when the ratios V/Al and Ni/Al increase above the thresholds V/Al ~1.15⋅10 − 3 and Ni/Al ~2.5⋅10 − 4 (Fig. 4B1-4C1). These thresholds are very close to those representative of the V and Ni ratios to Al in Saharan dust (Eglinton et al., 2002;Sholkovitz et al., 2009). Data with a V/Al and Ni/Al ratios higher than those dust-thresholds indicate an excess of V and Ni in the aerosols, with respect to dust, likely linked to the presence of traces pollutants linked to HFO combustion (Trapp et al., 2010a) which usually containing soluble ferric sulphate and nanocrystals of magnetite (Fu et al., 2012;Schroth et al., 2009). By using equations 3, 4 and 5, we estimate that ~23, 15 and 28% of V measured at Tenerife, Barbados and Miami during this experiment is due to HFO combustion, respectively, i.e. dust is still the dominant source of V during dust outbreaks. The V/Al ratios we observed in dust outbreaks ( Fig. 4B1-4B3) are close to those reported previously (<2⋅10 − 3 ), but lower than those found in the polluted marine atmosphere (V/Al: 10-80 ⋅10 − 3 ; (Sedwick et al., 2007;Wozniak et al., 2013Wozniak et al., , 2015. Similar results are found for the Ni/Al ratios: 2-14⋅10 − 4 in Saharan dusty air ( Fig. 4B1-4B3) and 150-300⋅10 − 4 in the polluted marine atmosphere (Wozniak et al., 2013).

Traces of acid processing
Fe-S also increases when the ratios of nitrate (Fig. 4D1-4D3), nsssulphate ( Fig. 4E1-4E3) and ammonium to aluminium increased. The formation of these salts involves nitric and sulphuric acids and potential acidity during some stages after the emissions of their precursors (Chen and Grassian, 2013;Li et al., 2017;Longo et al., 2016). The reaction of dust with acids contributes to Fe dissolution, with an efficiency modulated by minerals with buffering capacity (Ito and Feng, 2010). Organic acids, linked to water-soluble organic matter mixed with dust (Zamora et al., 2011), may also contribute to sFe (Wozniak et al., 2013). Carboxylic acids may complex Fe and maintain it with an enhanced solubility (Hawkins et al., 2010;Okochi and Brimblecombe, 2002).
Although previous studies found that other sources (e.g. biomass burning or coal power-plants (Kanakidou et al., 2018;Mahowald et al., 2018) may also contribute to sFe), we did not identified their involvement during our study. This result is not surprising in the summer when biomass burning is at a minimum in North Africa.

Source apportionment of soluble iron
We used two approaches, the minimum ratio method (Na et al., 2004;Turpin and Huntzicker, 1995) and lower border line (Kulmala et al., 2016;Reche et al., 2011;Rodríguez and Cuevas, 2007), and the above-described relationships (Fig. 4) for apportioning sFe between the contributions of dust (sFe d ), HFO combustion (sFe c ) and atmospheric processing (sFe a ), by equations (6)-(9): The contribution of sFe d was determined using S min (= 0.004), representative of the solubility of Fe in dust (0.4%) observed in this ( Fig. 4A1 and Fig. 4A2) and in previous studies (Ravelo-Pérez et al., 2016;Schroth et al., 2009;Sedwick et al., 2007;Theodosi et al., 2010). Two approaches were used for determining sFe c (equation (8)), based on the minimum ratio and the lower border line. This sFe c component only contributes to sFe when the V/Al ratio is higher than a (V/Al) d threshold. Finally, sFe a was calculated with equation (6b).
In the minimum ratio (min-R) approach to determine sFe c (equations (8a)-(8b))), C min = 0 and Z min is the minimum value found in the 100⋅ (sFe c + sFe a /Fe)/(V/Al) ratios in the collection of samples at each site, where sFe c + sFe a is the difference between sFe and sFe d (equation (6a)). We found Z min values (dimensionless mass ratio) equal to 44.5 in Tenerife, 355.7 in Barbados and 988.0 in Miami. The low Z min values in Tenerife are due to the low Fe-S at this site. A threashold value (V/Al) d = 1.15⋅10 − 3 was used (mass ratio), as proxy of the maximum amount of vanadium in Saharan dust (Eglinton et al., 2002).
The lower border line (LBL) approach to determine sFe c (Rodríguez and Cuevas, 2007), Z min and C min are calculated by regression analysis, as the fitting equation of the lower border line of the plot of 100⋅(sFe c + sFe a )/Fe versus V/Al, where sFe c + sFe a is determined as the difference between sFe and sFe d (equation (6a)). This plot is shown in Fig. 5 (red dots), where, in green colour, are highlighted the two points (circles) representative of the lower border of the data set and the resulting regression line and equation. This (green) line represents the minimum increase of iron solubility associated with the increases of vanadium with respect to aluminium, and it is considered a proxy of the contribution of HFO combustion to soluble iron due to emissions of fly ashes containing soluble ferric sulphate and nanocrystals of magnetite (Fu et al., 2012;Schroth et al., 2009). In this method the (V/Al) d threshold is determined as: The regression equation and (V/Al) d thresholds obtained with this method are shown in green and blue, respectively, in Fig. 5. The (V/Al) d thresholds (ng/μg) above wich we observe an increase in iron solubility is 1.543 in Tenerife, 1.050 in Barbados and 1.556 in Miami. This experimental procedure to determine (V/Al) d allow to take into account the potential variability in the V/Al ratios in dust and in the relationship between sFe c and V c in HFO combustion emissions.
We applied these two (R-min and LBL) methods (equations (6)-(9)) to the collected data set. The correlation of sFe c with V c (Fig. 6A) and the increasing trend of sFe a with sulphate and nitrate ( Fig. 6B-C) support the consistency of the approaches.
We first focused on the peak dust events due to the impact of the dust-fronts at each site (Table 1), described previously. The overall data analysis indicates that during the Atlantic transit of the dust Saharan outbreaks, dust contributions (sFe d ) to sFe decrease from 63% (Tenerife) down to 9% (Miami), whereas sFe a + sFe c (joined effects of air pollutants) accounts for 35% (Tenerife) to 91% (Miami) of sFe (Tables 1E and  1F).
During the impact of the dust-outbreak fronts, Fe concentrations decreased across Tenerife, Barbados and Miami, 3.3, 1.7 and 0.8 μg/m 3 , whereas Fe-S increased: 0.69%, 1.00% and 4.71% (Table 1C), respectively. This westward increase of Fe-S is associated with (i) a decrease of sFe d , linked to (size depended) dust deposition across the Atlantic (13.3-3.4 ng/m 3 ; Table 1E-1F), and (ii) an increase in the contribution of atmospheric processing (sFe a 5-6 to 24-31 ng/m 3 ; Table 1E-1F), consistent with previous studies that found increasing amounts of soluble Fe(II)-sulphate on Saharan dust because of the effects of acids, even after 10 days of atmospheric transport (Longo et al., 2016). In Tenerife, sFe a accounts for ~25-27% of sFe (Tables 1E-1F), which suggests that atmospheric processing of dust may already have occurred over North Africa (Ravelo-Pérez et al., 2016). Regional emissions of acid precursors in the Caribbean  and in-cloud processing (Buck et al., 2010;Shi et al., 2009) may account for the observed increase of sFe a from Barbados to Miami. We observed that sFe c concentrations are Fig. 6. A) Soluble iron due to HFO combustion (sFe c ) versus vanadium due to HFO combustion (V c ). A1) data of Tenerife are multiplied by 10, scale factor. B-D) Soluble iron due to atmospheric processing sFe a versus nitrate, none sea salt sulphate and ammonium. Plots include all data (not only dust-front impact events) collected in summer 2015. Results obtained with the minimum ratio and lower border line methods are shown in columns (A1-C1) and (A2-C2), respectively.
mostly ≤5 ng/m 3 (Table 1D), which are lower than those that we estimate for the polluted marine atmosphere, sFe c = 10-200 ng/m 3 (based on equations (8b)-(8c) and previous data of Sedwick et al., 2007and Wozniak et al., 2013, 2015. This rather low contribution of sFe c is due to the fact that the V/Al ratios are not much higher than the (V/Al) d thresholds in dust. Fig. 7 shows all data of sFe concentration at each site, including the contributions of sFe d , sFe c and sFe a , generated in this study, classified from the highest to the lowest sFe value. This plot include all available data, i.e. peak dust events due to the impact of the dust-fronts at each and other days. In Tenerife, increases in sFe are linked to increases in the contributions of sFe d and sFe a . In Barbados, the contribution of sFe a is somewhat higher than in Tenerife, whereas in Miami the contribution of sFe a is dominant. For sFe concentrations within the range 20-80th percentiles, sFe d and sFe a + sFe c accounts for 57% and 43% (37 + 8% and 30 + 13%, based on min-R and LBL, respectively) in Tenerife, 33% and 67% (61 + 6% and 53 + 14%) in Barbados and 10% and 90% (61 + 29% and 71 + 18%) in Miami, respectively. These results support the idea that the mixing of dust with acid pollutants is an important source of soluble iron as Saharan-dust outbreaks traverse the Atlantic.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Fig. 7. Concentrations of soluble iron (sFe) classified from highest to lowest values. The contributions of dust (brown), HFO combustion (grey) and atmospheric processing (green) are highlighted in brown, grey and green colour, respectively. Two approaches were used: A1-C1 is based on the minimum ratio method, whereas A2-C2 is based on the lower border line method. Plots include all data (not only dust front events) generated in summer 2015. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)