The impact of changing surface ocean conditions on the dissolution of aerosol iron

The proportion of aerosol iron (Fe) that dissolves in seawater varies greatly and is dependent on aerosol composition and the physicochemical conditions of seawater, which may change depending on location or be altered by global environmental change. Aerosol and surface seawater samples were collected in the Sargasso Sea and used to investigate the impact of these changing conditions on aerosol Fe dissolution in seawater. Our data show that seawater temperature, pH, and oxygen concentration, within the range of current and projected future values, had no significant effect on the dissolution of aerosol Fe. However, the source and composition of aerosols had the most significant effect on the aerosol Fe solubility, with the most anthropogenically influenced samples having the highest fractional solubility (up to 3.2%). The impact of ocean warming and acidification on aerosol Fe dissolution is therefore unlikely to be as important as changes in land usage and fossil fuel combustion. Our experimental results also reveal important changes in the size distribution of soluble aerosol Fe in solution, depending on the chemical conditions of seawater. Under typical conditions, the majority (77–100%) of Fe released from aerosols into ambient seawater existed in the colloidal (0.02–0.4 µm) size fraction. However, in the presence of a sufficient concentration of strong Fe‐binding organic ligands (10 nM) most of the aerosol‐derived colloidal Fe was converted to soluble Fe (<0.02 µm). This finding highlights the potential importance of organic ligands in retaining aerosol Fe in a biologically available form in the surface ocean.


Introduction
During the past 20 years, extensive investigations of the marine biogeochemistry of iron (Fe) have demonstrated that this micronutrient limits primary production in vast areas of the world ocean [e.g., Martin et al., 1994;Coale et al., 1996;Boyd et al., 2000;Takeda and Tsuda, 2005]. In particular, the supply of Fe controls phytoplankton growth in high-nutrient, low-chlorophyll waters, where the macronutrients nitrate and phosphate are replete. In these regions, there is the potential for increased primary production, carbon export to the deep ocean, and associated atmospheric carbon dioxide (CO 2 ) drawdown to be realized by increased inputs of Fe [Martin, 1990;Kohfeld and Ridgwell, 2009].
Atmospheric deposition of aerosols is an important pathway by which Fe is introduced into the surface waters of the open ocean [Duce, 1986;Martin and Gordon, 1988;Jickells et al., 2005]. The percentage of the total Fe contained within aerosols that dissolves in seawater, known as the fractional solubility of aerosol Fe (%Fe s ), is known to vary substantially, depending on a number of factors. A review by Sholkovitz et al. [2012] notes that empirical estimates of the %Fe s range from <0.1% to >95% (for~1100 published values). Factors affecting the dissolution of aerosol Fe have been discussed by Baker and Croot [2010] and include the source and composition of the aerosols [Bonnet and Guieu, 2004], the particle size , the concentration of particles deposited at the ocean surface [Zhuang et al., 1990], and atmospheric conditioning [Zhu et al., 1997]. It is therefore important to consider differences in aerosol dissolution when constructing global biogeochemical models that include Fe input to the ocean. However, most global models have thus far assumed a constant value for the %Fe s [e.g., Aumont et al., 2003;Moore et al., 2004;Parekh et al., 2005;Tagliabue et al., 2009].
[e.g., Sedwick et al., 2007;Aguilar-Islas et al., 2010]. Therefore, the potential increase in anthropogenic aerosol emissions due to rising global population and the industrialization of developing nations [International Energy Agency, 2009] may be expected to increase the aeolian input of dissolved Fe (dFe) to the world ocean.
Following deposition, the physicochemical conditions of seawater, including temperature, pH, and dissolved oxygen (O 2 ) concentration, will impact on the dissolution of aerosol Fe and the subsequent solution speciation of aeolian Fe [Millero et al., 2009]. Inorganic Fe(III) species have an extremely low solubility limit in seawater (<80 pM) above which they precipitate as ferric oxyhydroxides [Liu and Millero, 2002]. The solubility of Fe(III) species is known to increase with decreasing pH, although this trend is not strong for seawater over the pH range of 7.5-9 [Kuma et al., 1996;Liu and Millero, 2002]. The effect of seawater temperature and dissolved O 2 concentration on Fe solubility in seawater is less clear. Liu and Millero [2002] report a higher solubility value for Fe(III) in 5°C seawater (0.5 nM) than in 25°C seawater (0.35 nM) and 50°C seawater (0.34 nM). However, Kuma et al. [1996] observed no change in the solubility of Fe(III) from 10°C to 20°C. In addition, the oxidation rate of the more soluble redox species Fe(II) is known to decrease with decreasing temperature, pH, and O 2 concentration [Millero et al., 1987;Croot et al., 2001;Hopkinson and Barbeau, 2007;Breitbarth et al., 2010].
Iron-binding ligands in seawater play a major role in Fe biogeochemistry. By keeping Fe in solution, ligands reduce the precipitation of ferric oxyhydroxides and thus the ultimate loss of dFe from surface waters via particle scavenging and export [Boyd and Ellwood, 2010]. Two classes of Fe-binding ligands have been characterized: stronger ligands ("L 1 "), the strongest of which occur mainly near the surface, and weaker ligands ("L 2 "), which are more abundant and occur throughout the water column [Rue and Bruland, 1995;Hunter and Boyd, 2007]. These L 1 and L 2 ligands have conditional stability constants (log K cond FeLi;Fe ′ , where i denotes the ligand class) of 11.1-13.9 and 9.7-11.95, respectively [Gledhill and Buck, 2012]. Numerous studies have shown that complexation by dissolved organic matter such as saccharides [Hassler et al., 2011], humic substances [Laglera and van den Berg, 2009], and other species that make up the L 1 ligands [Rue and Bruland, 1995] allows Fe to exist at concentrations well in excess of the solubility of inorganic Fe(III).
Anthropogenic greenhouse gas emissions are thought to be responsible for observed increases in sea surface temperature (SST) [Intergovernmental Panel on Climate Change, 2013] and reductions in surface seawater pH [Orr et al., 2005]. As a consequence of increased SST, the upper ocean is expected to become more stratified and less oxygenated [Whitney et al., 2007;Stramma et al., 2008]. Furthermore, the likely increase in anthropogenic aerosol emissions may result in an increase in the delivery of organic matter to surface waters, which may potentially contain Fe-binding ligands [Wozniak et al., 2013]. These changing environmental parameters clearly have potential consequences for aerosol Fe dissolution, but the nature and magnitude of these effects are uncertain.
The experimental study described here aims to elucidate the main controls on aerosol Fe dissolution by manipulating key parameters that are likely to be affected by future environmental change, namely, SST, pH, dissolved O 2 concentration, and Fe-binding ligand concentration. Filtration using different pore size filters (0.4 μm and 0.02 μm) provided insight into the size distribution of aerosol-derived dFe in seawater leachate, as well as possible dissolution mechanisms under the range of different physicochemical conditions. In addition, experiments using bulk aerosol samples collected over a full annual cycle allowed us to assess the impact of differing aerosol sources on aerosol Fe dissolution, in relation to differing physicochemical conditions in surface seawater.
Polyvinylchloride 10 L Teflon-lined, external-closure Niskin-type bottles (Ocean Test Equipment) were used to collect seawater samples at the deep chlorophyll maximum (~60-120 m depth), where Fe concentrations are typically lowest [Sedwick et al., 2005]. Upon recovery, samples were immediately transferred from the Niskin bottles into acid-washed 50 L low-density polyethylene (LDPE) carboys (Nalgene) inside a polyethylene-walled "clean bubble" under positive pressure. The seawater was subsequently filtered through 0.45 μm polytetrafluoroethylene (PTFE) Osmonics membrane capsules (GE Water Systems) via gravity filtration into 25 L LDPE carboys (Nalgene) and stored in the dark under ambient clean room laboratory conditions for no longer than 3 months before being used in leaching experiments. L 1 -type Fe-binding ligands in all bulk seawater samples were determined by competitive ligand exchange-adsorptive cathodic stripping voltammetry prior to leaching experiments. The ligand concentrations were found to be subnanomolar (~0.8 nM; n = 5) in all samples with a log K cond FeL1;Fe ′ of~12 (data not shown). To help structure the discussion, sample numbers AER 1-AER 4 were ordered to reflect the influence of North American air masses during sampling (from smallest to largest). Samples were taken from a high-volume aerosol sampler, which drew air through 8 × 10″ sheets of 2.2 μm pore size quartz microfiber Whatman QMA filters (GE Healthcare). The total volume of air sampled for AER 1, AER 2, AER 3, and AER 4 over these sampling periods was 8620 m 3 , 10,519 m 3 , 35,356 m 3 , and 24,243 m 3 , respectively. The aerosol filters were previously acid washed following a method described by Morton et al. [2013]. The aerosol sampler was located approximately 50 m above sea level on top of a 23 m high aluminum scaffolding tower. To avoid local aerosol sources and ensure that only marine air was sampled, pumps were switched off when the wind direction was not in the onshore sector (210-315°), or if the wind speed was less than 1 m s À1 . A procedural field blank was taken by loading an aerosol filter into the aerosol sampler, leaving it in the sampler (with pumps off) for 5 min and then removing it. Aerosol samples were removed and placed in zip-lock bags and stored frozen at À20°C for 6-24 months before use.

Seawater Preconditioning
Separate 1 L aliquots of the 0.45 μm filtered BATS seawater were preconditioned before use in the leaching experiments. Filtered seawater aliquots used in the variable temperature experiments were equilibrated to 4°C and 25°C, using a fridge and temperature-controlled laboratory, respectively. Filtered seawater aliquots used in the variable seawater pH experiments were preconditioned by sparging with filtered air/CO 2 mixtures using mass flow controllers until a stable pH was reached (between 48 and 72 h) [de Putron et al., 2011]. The sparging gas mixtures simulated equilibrium with atmospheric CO 2 concentrations of 400 ppm CO 2 (present-day conditions) and 1250 ppm CO 2 (possible future conditions), resulting in seawater pH values of 8.0 ("ambient seawater") and 7.6 ("acidified seawater"), respectively. "Total-scale" pH was calculated using the measured values of total dissolved inorganic carbon and alkalinity using "CO 2 sys." Similarly, anoxic conditions (<0.1% O 2 ) were created by purging dissolved O 2 from filtered seawater by sparging with filtered nitrogen (N 2 ) gas (for leaches 3 and 4 only). Filtered seawater aliquots used in organic ligand amendment experiments were preconditioned through the addition of a strong siderophore L 1 -type ligand (either 10 nM of aerobactin (EMC Microcollections) or desferrioxamine B (Sigma-Aldrich)) or weaker L 2 -type ligands (either 10 nM of protoporphyrin IX (Sigma-Aldrich) or 1 μM of glucuronic acid (Sigma-Aldrich)). These ligands were intended to simulate the range of Fe-binding ligands that exist in open ocean waters and were added in concentrations in excess (~1 order of magnitude) of typical open ocean surface water concentrations in the western North Atlantic [Cullen et al., 2006;Buck et al., 2014] in order to observe a clear response in aerosol Fe complexation.

Aerosol Leaching Experiments
Aerosol leaching experiments were designed to replicate the process of dry deposition in the open ocean as closely as possible within the laboratory. Leaches were performed using an acid-washed filtration tower (Savillex) comprising a 47 mm diameter perfluoroalkoxy filter assembly clamp, 250 mL evacuated Teflon collection vessel, and 0.25" diameter PTFE tubing. The PTFE tubing connected the collection vessel to a vacuum pump (GAST) via a 0.2 μm Acrovent air filter (Pall Corporation) and water trap. Acid-washed 0.4 μm pore size, 47 mm polycarbonate track-etched membrane filters (Nuclepore, GE Healthcare) were used to separate "dissolved Fe" in the aerosol leachate solutions. Triplicate subsamples from each aerosol filter sample were taken using a 20 mm diameter polished steel arch punch (Osborne) and allowed to thaw at room temperature before being used for each different treatment in the aerosol leaching experiments.
The leaching process ( Figure 1) involved four sequential leaches to test different equilibration times (leach 1 = 0-5 min, leach 2 = 5-10 min, leach 3 = 10 min-48 h, and leach 4 = 48 h-30 days). These leaching times were chosen to represent the typical range of residence times for aerosol particles in the upper surface ocean.
For leach 1, a 20 mm diameter aerosol filter subsample was placed on a polycarbonate filter mounted within the filtration tower (under vacuum). Then 250 mL of preconditioned filtered seawater was poured over the aerosol filter in a continuous leaching process, lasting 5 min, with the resulting leachate collected. This process was repeated for leach 2 using the same aerosol subsample used in leach 1. Leaches 3 and 4 were batch leaches where the same aerosol subsample and polycarbonate filters used in leaches 1 and 2 were placed in fluorinated polyethylene (FPE, Nalgene) bottles with 500 mL of preconditioned filtered seawater for 48 h and 30 days for leaches 3 and 4, respectively ( Figure 1). Previous studies employing an aerosol batch leach methodology have observed only negligible adsorption of dFe to bottle walls when fluorinated polymer bottles were used [Séguret et al., 2011]. To further minimize the effect of wall adsorption of dFe, the FPE bottles were first preconditioned for 24-72 h with filtered BATS seawater before batch leaches took place. To minimize the influence of ultraviolet light on Fe dissolution, leaches 1 and 2 were carried out under artificial light only, whereas leaches 3 and 4 were stored in darkness. The seawater leachate solutions that passed through the 0.4 μm filter membrane in each sequential leach were subsequently decanted into acid-washed 125 mL LDPE bottles (Nalgene). The entire leaching process was conducted in triplicate for every experimental treatment.
To determine "colloidal Fe" (cFe) and "soluble Fe" (sFe), 125 mL of seawater leachate samples were filtered through 0.02 μm pore size, 25 mm diameter aluminum oxide Whatman Anotop syringe filters (GE Healthcare) and collected in acid-washed 60 mL LDPE bottles (Nalgene) (Figure 1). In this study, following the definitions of Wu et al. [2001], dFe is defined as the <0.4 μm fraction, and sFe is defined as the <0.02 μm fraction. Colloidal Fe is thus inferred from the difference between dFe and sFe, representing the 0.02-0.4 μm size fraction (i.e., cFe = dFe-sFe). The Anotop filters used for ultrafiltration of aerosol leachate were set up and conditioned following the in-line filtration method detailed by Ussher et al. [2010]. Following aerosol leaches, all seawater leachate samples (including blanks) were acidified to pH~1.7 using HCl (Romil, UpA) and stored for >1 year before analysis. In addition to leachate samples, process blanks were prepared by passing ambient filtered seawater over a procedural field blank subsample (section 2.1) and a 47 mm polycarbonate filter. Following analysis, the Fe concentrations of the relevant process blanks were subtracted from Fe concentrations of all leachate samples. The dFe concentrations of the process blanks were typically <0.1 nM.

Determination of Iron in Seawater Leachates Using Flow Injection With Chemiluminescence
The concentrations of dFe and sFe in seawater leachate solutions were determined using flow injection with chemiluminescence detection (FI-CL) inside a Class 100 clean room (ISO 9001:2008), based on the method described by Obata et al. [1993]. Briefly, seawater leachate samples were buffered in-line to pH 3.5 with 0.35 M ammonium acetate (Romil, SpA) then loaded on a chelating column containing Toyopearl AF-Chelate-650 M iminodiacetate resin (IDA, Tosoh Bioscience). The seawater matrix major cations were removed using a 0.012 M HCl (Romil, SpA) rinse and then Fe was eluted from the column using 0.23 M HCl (Romil, SpA). The acid/analyte eluent was then mixed with 0.25 mM luminol, 0.5 M ammonium hydroxide, and 0.3 M hydrogen peroxide to initiate the oxidation of luminol [Rose and Waite, 2001], which produced a chemiluminescence signal detected by a photomultiplier tube (Hamamatsu).

Total Trace Metal Determination by Inductively Coupled Plasma Mass Spectrometry
Triplicate subsamples of AER 1-AER 4 and blanks were completely digested using concentrated hydrofluoric acid (28.9 M HF, Seastar, Baseline), concentrated nitric acid (15.8 M HNO 3 , Seastar, Baseline), and heat  at the University of Delaware. As acid digestion is a destructive process, it was necessary to use different subsamples of each aerosol sample in the aerosol leaching experiments and use the mean total Fe amount in fractional solubility calculations. Total element determinations were subsequently performed on the digests at Old Dominion University using an Element 2 (Thermo Fisher) inductively coupled plasma-mass spectrometer (ICP-MS) in medium resolution mode. Samples were spiked with indium ( 115 In) as an internal standard. Calibration standards were made up using a multielemental standard (QC Standard 4, PlasmaCal containing aluminum (Al), titanium (Ti), manganese (Mn), vanadium (V), chromium (Cr), Fe, cobalt (Co), nickel (Ni), copper (Cu), zinc (Zn), cadmium (Cd), antimony (Sb), and lead (Pb)) in the same 2% (vol/vol) HNO 3 matrix as the aerosol digest solutions.

Air Mass Back Trajectory Analysis
Air mass back trajectories were simulated using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model developed by the U.S. National Oceanic and Atmospheric Administration (available for download at http://ready.arl.noaa.gov/hyreg/HYSPLIT_pchysplit.php). Ten-day back trajectories at altitudes of 50 m, 500 m, and 2000 m were performed for each day within the period over which aerosol samples were collected. Daily back trajectories were used in the cluster analysis function of HYSPLIT to generate 10 mean back trajectories for each aerosol sampling period.

Aerosol Sample Characterization
The aerosol samples collected at Tudor Hill represent a seasonal-scale time series comprising a summer sample (AER 1), an autumn sample (AER 2), a spring sample (AER 3), and a winter sample (AER 4). Daily 500 m elevation 10 day air mass back trajectories for each day of the collection period of each aerosol sample are provided in Figure  These results are consistent with time series aerosol observations from Bermuda over a number of decades [Duce and Hoffman, 1976;Chen and Duce, 1983;Anderson et al., 1996;Sholkovitz et al., 2009].
To gain an appreciation of the relative contributions of anthropogenic versus crustal materials to the aerosols, total masses of Fe, V, Ni, Cu, Pb, and Sb in the aerosol samples (Table 1) were normalized to the mass of Al, and the resulting ratios were compared with those of average continental crust [Taylor and McLennan, 1995] to calculate enrichment factors (Table 2). Samples collected in the winter (i.e., AER 4) were more enriched (enrichment factors >10) in V, Ni, Cu, Pb, and Sb relative to aerosols collected in the summer (i.e., AER 1 and AER 2). These results are consistent with the back trajectory models ( Figure 2) and agree with the suggestion of Sedwick et al. [2007] that marine air over the Sargasso Sea contains a greater proportion of anthropogenic combustion products (relative to continental soil dust) in the winter. Summer Fe and Al atmospheric loadings were higher than those determined in the winter (Table 1), consistent with previous work that has shown high-Fe and -Al loadings in air masses dominated by lithogenic aerosols, relative to anthropogenic materials Buck et al., 2010a;Séguret et al., 2011].

Effect of Seawater Temperature, pH, and Deoxygenation on Aerosol Iron Dissolution
The total dFe leached from aerosols into preconditioned seawater samples of different temperature, pH, and O 2 concentration was calculated from the process-blank-corrected dFe concentrations and volumes of seawater leachate (Figure 3). An alternative representation of these results, as fractional solubilities of aerosol Fe (total dFe released into solution divided by total aerosol Fe), is shown in Figure 4.
The mean total dFe leached from each aerosol sample (AER 1-AER 4) into 4°C and 25°C seawater is shown in Figure 3a. The mean total dFe is the sum of all dFe released into seawater during the four sequential leaches and ranged from 0.7 ± 0.2 nmol to 4 ± 2 nmol for AER 1-AER 4 under these different seawater conditions. The uncertainty in the mean total dFe values of a given sample was dominated by differences between replicate leaches of subsamples of the same aerosol sample, with an RSD of 13-45%, rather than replicate Fe analyses of the same leachate solution, with an RSD of 5-6%, which is in accord with results of other studies [e.g., Morton et al., 2013]. Most of the aerosol Fe was released into solution during the first leach, with progressively less Fe released in the subsequent sequential leaches. It is clear that the dissolution of Fe from aerosols occurred rapidly; the majority (84-94%) of the aerosolderived dFe was released within the first 10 min of contact with seawater. Similar kinetics were observed for all leaches assessed in this study and in other studies of aerosol Fe solubility [e.g., Buck et al., 2006;Wu et al., 2007]. Importantly, the differences in the mean total dFe released into 4°C versus 25°C seawater were not statistically significant (two-tailed t test, p > 0.05) for any of the aerosol samples, suggesting that seawater temperature does not exert a major control on the dissolution of aerosol Fe.
The mean total sFe (<0.02 μm) leached from each aerosol sample (AER 1-AER 4) into 4°C and 25°C seawater is shown in Table 3. The mean total sFe from the four sequential leaches into these different seawater conditions ranged from 0.04 nmol to 0.41 nmol for AER 1-AER 4. The colloidal Fe size fraction (cFe, 0.02-0.4 μm) can be inferred by comparing the dFe and sFe in the seawater leachate. For all leaches, the majority (77-97%) of dFe released from the aerosols into the seawater leachate resided in the cFe size fraction (Table 3). It should be noted that these data represent the size distribution of dFe in solution following leaching from aerosols, not the amount of sFe and cFe that has been directly leached from the aerosols. The mean total sFe in solution leached from the aerosols at 4°C was greater than that leached at 25°C for aerosols AER 2 and AER 3. Conversely, for AER 1 and AER 4, the mean total sFe leached from the aerosols at 25°C seawater was greater than that leached at 4°C. However, the differences in the mean total sFe in the leachate using 4°C seawater versus 25°C seawater were not statistically significant (two-tailed t test, p > 0.05) when the differences in dFe for replicate leaches are taken into account, suggesting that seawater temperature does not exert a major control on the size distribution (sFe versus cFe) of dFe in the aerosol leachate solutions.
The mean total quantities of dFe leached from each aerosol sample (AER 1-AER 4) into ambient seawater (25°C, pH 8.0, oxygenated), acidified seawater (pH 7.6, sparged with air/CO 2 mixture, 1250 ppm CO 2 ), and "anoxic seawater" (sparged with N 2 for leaches 3 and 4) are shown in Figure 3b. The mean total dFe is the sum of all dFe released during the four leaches and ranged from 0.6 ± 0.2 nmol to 5 ± 2 nmol for AER 1-AER 4, for these different seawater conditions. Our results show that the total dFe released into acidified and Global Biogeochemical Cycles 10.1002/2014GB004921 anoxic seawater was not statistically different (two-tailed t test, p > 0.05) from the total dFe released into ambient seawater for all aerosol samples.
The mean total sFe in the seawater leached from each aerosol sample (AER 1-AER 4) into ambient, acidified, and anoxic seawater is shown in Table 3. The mean total sFe from the four sequential leaches into these different seawater conditions ranged from 0.02 nmol to 0.16 nmol for AER 1-AER 4. In all cases, the majority (92-98%) of dFe in the seawater leachate solutions resided in the cFe size fraction (Table 3). There were some differences observed between samples: for example, the mean total sFe leached into ambient seawater was greater than that leached into acidified seawater for two of the four aerosol samples (AER 2 and AER 3), whereas the opposite was observed for AER 1 and AER 4. However, similar to the dFe results, the differences in the mean total sFe in the various seawater leachate samples were not statistically significant (two-tailed t test, p > 0.05) when the differences in dFe for replicate leaching experiments are taken into account.
Thus, our experimental results suggest that seawater temperature, a Iron/aluminum (Fe/Al), vanadium/aluminum (V/Al), nickel/aluminum (Ni/Al), copper/aluminum (Cu/Al), lead/aluminum (Pb/Al), and antimony/aluminum (Sb/Al) mass ratios calculated by dividing the total mean mass of each trace metal with the total mean mass of aluminum in each aerosol subsample. Standard deviations from mean mass ratios (±1σ) represent the uncertainty of trace metal mean mass measurements only. b Iron (Fe), vanadium (V), nickel (Ni), copper (Cu), lead (Pb), and antimony (Sb) enrichment factors were calculated by dividing trace metal/aluminum mass ratios of aerosol samples by trace metal/aluminum ratios found in the upper continental crust. Abundances of trace metals in the upper continental crust taken from Taylor and McLennan [1995] and used to calculate mass ratios (Fe/Al = 0.4, V/Al = 0.0007, Ni/Al = 0.0002, Cu/Al = 0.0003, Pb/Al = 0.0002, and Sb/Al = 0.000002). Standard deviations from mean enrichment factors (±1σ) represent the uncertainty of trace metal mean mass measurements only.

B)
A) Figure 3. The mean total dissolved iron (dFe, nmol) leached from aerosol samples AER 1, AER 2, AER 3, and AER 4 during three replicate leaches with (a) 4°C and 25°C seawater and (b) ambient seawater (25°C, pH 8.0, oxygenated), acidified seawater (25°C, pH 7.6, sparged with air enriched in carbon dioxide), and anoxic seawater (25°C, sparged with nitrogen gas for leaches 3 and 4). The error bars represent the standard deviation (±1σ) on the mean total dFe leached from replicate subsamples of the same aerosol sample. Segments of each bar show the amount of dFe leached during each sequential aerosol leach (i.e., leaches 1-4). Sequential leaches 1, 2, 3, and 4 represent the following time periods for seawater that was exposed to aerosols: 0-5 min, 5-10 min, 10 min-48 h, and 48 h-30 days, respectively. pH (as controlled by pCO 2 ), and O 2 concentration did not exert a major control on the dissolution of aerosol Fe, nor on the size distribution of aerosolderived dFe in seawater, within the experimental uncertainties and ranges of the parameters examined. These observations are in accord with those of Kuma et al. [1996], who reported no difference in the solubility of inorganic Fe(III) in seawater at 10°C versus 20°C. In contrast, however, Liu and Millero [2002] reported higher solubility of Fe(III) seawater at 5°C versus 25°C (0.5 ± 0.07 nM for 5°C seawater and 0.35 ± 0.06 nM for 25°C seawater). Other studies have demonstrated an increase in the solubility of inorganic Fe with decreasing pH [Byrne and Kester, 1976;Kuma et al., 1996;Liu and Millero, 2002]. However, consistent with this study, Liu and Millero [2002] report that Fe solubility does not change in the pH range of 7.5-9, due to the dominance of the neutrally charged species Fe(OH) 3° [ Liu and Millero, 1999]. It should be noted that in the cases of Liu and Millero [2002] and Byrne and Kester [1976], equilibrium conditions were represented, whereas our leach experiments are not likely to represent equilibrium conditions. Data from our study also agree with empirical data that suggest little difference in the dissolution of aerosol Fe in seawater (pH~8) versus UHP (≥18.2 MΩ cm, pH 5.6) deionized water [Aguilar-Islas et al., 2010;Buck et al., 2013]. The 25°C seawater, 4°C seawater, ambient seawater from the pH and deoxygenation experiment (25°C, pH 8.0, oxygenated), acidified seawater (25°C, pH 7.6, sparged with air enriched in carbon dioxide), and anoxic seawater (25°C, sparged with nitrogen gas for leaches 3 and 4). Standard deviations (±1σ) from the mean total Fe leached from replicate subsamples of the same aerosol sample are shown for dFe. . Fractional solubility (%) of iron (Fe) from aerosol samples AER 1, AER 2, AER 3, and AER 4 leached with ambient seawater (combined mean of the temperature experiment and pH and deoxygenation experiment, 25°C, pH 8.0, oxygenated), 4°C seawater, acidified seawater (25°C, pH 7.6, sparged with air enriched in carbon dioxide), and anoxic seawater (25°C, sparged with nitrogen gas for leaches 3 and 4). The error bars represent the standard deviation (±1σ) on the mean fractional solubility of Fe leached from replicate subsamples of the same aerosol sample. Segments of each bar show the fractional solubility of each sequential aerosol leach (i.e., leaches 1-4). Sequential leaches 1, 2, 3, and 4 represent the following time periods for seawater that was exposed to aerosols: 0-5 min, 5-10 min, 10 min-48 h, and 48 h-30 days, respectively. Given the year 2100 projections for average SST to increase by 1°C [Meehl et al., 2007], seawater pH to decrease by 0.25 [Caldeira and Wickett, 2003], and oceanic oxygen minimum zones to expand [Shaffer et al., 2009], our data suggest that such changes will not significantly impact on aerosol Fe dissolution in surface ocean waters. However, it should be stressed that the combined effect of these changes, or the impact of such changes on other factors that impact aerosol Fe dissolution, cannot be ruled out. Our findings have important implications for modeling the impact of future changes in aerosol input to the global cycling of Fe and carbon, as well as phytoplankton growth and community structure. Specifically, our experimental results suggest that to a first approximation, models do not need to consider that aerosol Fe dissolution will vary as a function of projected variations in seawater temperature, pH, or oxygen saturation. While seawater temperature, pH, and O 2 concentration had no significant effect on aerosol Fe dissolution, we observed pronounced differences in aerosol Fe solubilities for aerosols from different source regions with different bulk compositions.

Effect of Aerosol Source and Composition on Aerosol Iron Dissolution
Seawater leaching of aerosol samples of different source and composition resulted in different fractional solubilities of aerosol Fe under all conditions ( Figure 4). Indeed, there is nearly an order of magnitude difference in the %Fe s for the more "anthropogenic"-type aerosol (AER 4) relative to the "mineral dust"-type aerosol (AER 1). The operationally defined %Fe s for aerosol AER 1 was consistently lower (0.3 ± 0.1% to 0.5 ± 0.2%) than that for aerosol AER 4 (2.5 ± 0.8% to 3.2 ± 0.8%) under all physicochemical conditions of the seawater leaching experiments. The dependence of aerosol Fe solubility on aerosol type and source has been noted in previous studies [Sedwick et al., 2007;Sholkovitz et al., 2009;Buck et al., 2010b;Séguret et al., 2011;Shelley et al., 2012;Baker et al., 2013;Wozniak et al., 2013]. It has been suggested that the elevated %Fe s in anthropogenically influenced aerosols reflects acidic processing at the particle surface [Meskhidze et al., 2005;Hsu et al., 2010], the presence of labile Fe sulfates and other soluble Fe phases [Oakes et al., 2012], and the tendency for anthropogenic aerosols to be smaller than mineral dust particles [Chen and Siefert, 2004;Jang et al., 2007]. Conversely, mineral aerosol particles tend to be larger, with much of the Fe contained in refractory minerals [Desboeufs et al., 2005]. Sholkovitz et al. [2009] show a linear relationship between the %Fe s and the V/Al mass ratios of bulk aerosol samples collected over the Sargasso Sea. That linear relationship and the V/Al mass ratios of bulk aerosol samples collected in this study were used to predict fractional solubilities in ambient seawater; these predictions were then compared with empirically determined fractional solubilities. While absolute values of the predicted and determined fractional solubilities differed (predicted values of 1.4%, 2.0%, 2.7%, and 3.8% for AER 1, AER 2, AER 3, and AER 4, respectively, versus observed values of 0.5%, 1.2%, 1.3%, and 2.8% for AER 1, AER 2, AER 3, and AER 4, respectively), the predicted trend of increasing solubility with increasing V/Al mass ratios matched the empirical values. Thus, our results indicate greater %Fe s in aerosols containing a larger proportion of anthropogenic material.
While this study is unable to provide insight into the mechanism of enhanced Fe solubility in non-soil-dust aerosols, the results add to the existing evidence that Fe in anthropogenic aerosols is more soluble than Fe in soil-derived mineral dust and that this variability dominates over potential changes in the fundamental parameters of temperature, pH, and O 2 concentration of the seawater in which aerosols are deposited. These results suggest that the potential for an increase in anthropogenic aerosol emissions accompanying the rising global population and the industrialization of developing nations [International Energy Agency, 2009] could conceivably increase the atmospheric flux of dFe to the ocean and may exert significant regional-scale impacts [Sholkovitz et al., 2009]. However, when considering a total flux including refractory material, soil dust is likely to continue to dominate the total global aerosol Fe deposition [Prospero et al., 2002], and therefore, the global importance of anthropogenic aerosol Fe input is likely to remain marginal. These results reinforce the idea that modeling efforts must consider the source of aerosols deposited in different ocean regions, as well as the potential for future changes in the composition of aerosols entering the ocean. following all leaches (including a control using seawater with no added ligands) fell in the narrow range of 0.8 ± 0.1 nmol to 1.2 ± 0.5 nmol. Interestingly, there was no statistical difference between the mean total dFe leached into each ligandamended seawater sample and that leached into unamended ambient seawater (two-tailed t test, p > 0.05), which contrasts with the results of previous work by Aguilar-Islas et al. [2010]. However, this earlier study did not include replicate leaches using subsamples of a single-aerosol sample. Our study has shown that the RSD associated with replicate leaches can exceed 50%, and while the mean total amounts of dFe leached into ligandamended seawater appeared to differ from that leached into ambient seawater, these differences were not statistically significant. The majority of the variation in dFe leached from subsamples of a single-aerosol filter reflects either the inherent variability of the leaching process (i.e., the leaching mechanism or variability in the Fe species or particle character) or the heterogeneity of Fe species with different fractional solubilities across the aerosol filter. Bulk analysis of an extended suite of aerosol digest solutions (data not shown) suggests that heterogeneity in the total Fe contained on the aerosol filter is unlikely to be responsible for the observed variability in dFe between replicate leaches (RSD on the mean total Fe in subsamples of a single aerosol sample is <3%).

Effect of Organic Ligands on Aerosol Iron Dissolution
In stark contrast, there were pronounced differences in the mean total sFe in leachate solutions after leaching with unamended ambient seawater versus seawater amended with the L 1 -type ligands aerobactin and desferrioxamine B (Figure 5b). The mean total sFe leached into ambient seawater leachate was 0.02 nmol, whereas leachate from seawater amended with aerobactin and desferrioxamine B contained 1.3 nmol and 1.1 nmol sFe, respectively. This large impact observed with L 1 -type ligand-amended seawater was not seen with the weaker organic ligands, glucuronic acid and protoporphyrin IX, for which leachate solutions contained 0.10 nmol and 0.16 nmol sFe, respectively.
In comparing the dFe and sFe in the seawater leachate solutions, it is apparent that the majority (˜100%) of the dFe in ambient seawater leachate existed as cFe. The same was true for all of the seawater leachate solutions from the temperature, pH, and dissolved O 2 experiments. This observation is consistent with the results of previous laboratory studies and field observations [Wu et al., 2001;Bergquist et al., 2007 A) Figure 5. The mean total amount (nmol) of (a) dissolved iron (dFe) and (b) soluble iron (sFe) leached from aerosol sample AER 1 into ambient seawater and seawater amended with various Fe-binding organic ligands (aerobactin, desferrioxamine B, glucuronic acid, and protoporphyrin IX). Error bars in Figure 5a represent the standard deviation (±1σ) on the mean total dFe leached from replicate subsamples of the same aerosol sample. No error bars are shown in Figure 5b, as replicate leaches were not performed. In both Figures 5a and 5b, the segments of each bar show the amount of dFe leached during each sequential aerosol leach (i.e., leaches 1-3). Sequential leaches 1, 2, and 3 represent the following time periods for seawater that was exposed to aerosols: 0-5 min, 5-10 min, and 10 min-48 h, respectively. Leach 4 was not performed in this experiment.

10.1002/2014GB004921
amended with the L 1 -type ligands aerobactin and desferrioxamine B, the majority of dFe in the leachate (77-98%) resided in the sFe fraction. However, for leaches with seawater amended with the weaker ligands, glucuronic acid and protoporphyrin IX, there was only a small increase in the proportion of sFe in the leachate compared with the control experiment using ambient conditions (sFe accounted for 11-21% of dFe). This observation suggests that the stronger L 1 -type ligands may play an important role in the dissolution of aerosol Fe in the surface ocean.
Based on the results of this study, we hypothesize a two-stage mechanism for the dissolution of aerosol Fe in seawater ( Figure 6). First, upon deposition of aerosols at the ocean surface, labile inorganic Fe [Fe(II) and Fe(III)] is released into seawater, whereupon it rapidly forms colloidal-sized ferric oxyhydroxides or is adsorbed or complexed by colloidal-sized organic matter with metal-binding functional groups (e.g., saccharides [Hassler et al., 2011], amino acids [Benner, 2011], and humic substances [Laglera and van den Berg, 2009]). After this first stage, the dissolved aerosol Fe exists predominantly in the cFe size fraction, as observed in the majority of our experimental leachate solutions. During the second stage of our hypothesized mechanism, strong L 1 -type ligands, such as siderophores, if present uncomplexed in sufficient concentration, form strong Fe-ligand complexes, whereby there is a transfer of dissolved aerosol Fe from the colloidal to the soluble size fraction, as low molecular weight (~300-1000 amu) Fe-ligand complexes [Macrellis et al., 2001]. The fact that not all Fe is "solubilized" by L 1 -type ligands, despite seawater being amended with large excess concentrations (10 nM) in our experiments, may reflect the physical or chemical association of some L 1 -type ligands with colloidal material present in the seawater.
An alternative hypothesis is that upon deposition at the surface ocean, aerosols may release colloidal-sized organic matter or inorganic phases with which Fe is already associated [Raiswell and Canfield, 2012; Paris and ©2014. The Authors. Desboeufs, 2013]. In seawater, L 1 -type ligands can then draw the aeolian cFe into the soluble fraction, as described above. However, we suggest that it is more likely that dFe released from aerosols associates with colloids already present in seawater. This conclusion is based on results for the leaching of aerosols using seawater preconditioned by filtration (<0.4 μm) and ultrafiltration (<0.02 μm), which show lower dFe concentrations in leachate derived using ultrafiltered seawater [Ussher, 2005]. In addition, Dammshäuser and Croot [2012] showed that aerosol Al and titanium (Ti) colloidal associations are very low following dissolution into seawater compared to soluble forms, potentially indicating low input of colloidal matter from aerosols.
More work is required to establish whether colloidal Fe is delivered by aerosols or forms in seawater following deposition and determine if aerosol characteristics such as particle size of deposited aerosols and aerosol composition (i.e., mineral dust versus combustion products) play a role in this dissolution mechanism. Certainly, a higher proportion of colloidal Fe has been observed in regions of fine-mode aerosol deposition in comparison to regions of coarse-mode aerosol deposition, and this fine aerosol size fraction results in higher fractional solubilities of Fe [Buck et al., 2010b;Ussher et al., 2013].
In summary, from our experimental results, it seems clear that stronger organic Fe-binding ligands play an important role in the dissolution of aerosol Fe. This observation has important implications for our understanding of the biogeochemical role of organic ligands with regard to the cycling of Fe. In the absence of such organic ligands, aerosol-derived dFe will remain largely in the colloidal size fraction in seawater; in which case, it is expected to be more rapidly lost from surface ocean waters over a period of weeks to months [Moran and Buesseler, 1992;. In the presence of sufficient concentrations of strong ligands, however, aerosol-derived dFe may be transferred from the colloidal to the soluble size fraction, where it may be expected to have a longer residence time in the euphotic zone [Boyd and Ellwood, 2010] and thus have greater biological availability to phytoplankton. It is important to note that the role that biological organisms play in aerosol Fe dissolution has not been considered in this study and should be included when attempting to understand the complete aerosol Fe dissolution mechanism.

Conclusions
The results of this experimental study suggest that plausible future variations in the temperature, pH, and O 2 concentration of seawater will have a minor impact on the dissolution of aerosol Fe. However, in accord with other studies, our data indicate that aerosol composition has the most significant effect on the dissolution of aerosol Fe, with North Atlantic aerosol samples with the greatest anthropogenic components having the highest fractional solubility. Hence, we surmise that future increases in the magnitude of global aerosol fluxes and the proportion of combustion aerosols, relative to mineral dust, are likely to be the most important drivers of changes in the atmospheric flux of dFe to the surface ocean. Therefore, it is reasonable for modelers not to prioritize the inclusion of seawater temperature, pH, and O 2 concentration effects on aerosol Fe dissolution, but instead, focus attention on the regional differences in the original source of deposited aerosols. Furthermore, our results highlight the importance of strong Fe-binding organic ligands in regulating the size distribution of aerosol-derived dFe in surface ocean waters, specifically in increasing the residence time and biological availability of aerosol-derived Fe in the surface ocean. A conceptual model of aerosol Fe dissolution is proposed in which strong Fe-binding organic ligands play a key role. The proposed two-stage mechanism highlights the significance of (i) the colloidal phase and (ii) the strong L 1 -type ligands in the soluble phase in retaining biologically available aerosol Fe in surface ocean waters following aerosol deposition.