Atmospheric Trace Metal Deposition near the Great Barrier Reef, Australia

: Aerosols deposited into the Great Barrier Reef (GBR) contain iron (Fe) and other trace metals, which may act as micronutrients or as toxins to this sensitive marine ecosystem. In this paper, we quantiﬁed the atmospheric deposition of Fe and investigated aerosol sources in Mission Beach (Queensland) next to the GBR. Leaching experiments were applied to distinguish pools of Fe with regard to its solubility. The labile Fe concentration in aerosols was 2.3–10.6 ng m − 3 , which is equivalent to 4.9–11.4% of total Fe and was linked to combustion and biomass burning processes, while total Fe was dominated by crustal sources. A one-day precipitation event provided more soluble iron than the average dry deposition ﬂux, 0.165 and 0.143 µ mol m − 2 day − 1 , respectively. Scanning Electron Microscopy indicated that alumina-silicates were the main carriers of total Fe and samples a ﬀ ected by combustion emissions were accompanied by regular round-shaped carbonaceous particulates. Collected aerosols contained signiﬁcant amounts of Cd, Co, Cu, Mo, Mn, Pb, V, and Zn, which were mostly (47.5–96.7%) in the labile form. In this study, we provide the ﬁrst ﬁeld data on the atmospheric delivery of Fe and other trace metals to the GBR and propose that this is an important delivery mechanism to this region. M.M.G.P), aerosol sample collection (M.S., R.G.R.), laboratory work (M.S., M.M.G.P., M.G.-R.), AMS data (Z.D.R., J.A.), MAAP data (R.S.H., M.D.K. and J.W.), wind data (R.S., R.G.R.), and fieldwork commanding (R.S., Z.D.R.).

sea during the day and those transported from the land at night, a short 12-h sampling period was applied from 6 a.m.-6 p.m. and 6 p.m.-6 a.m. (the first three samples were collected for 24 h). The sampling campaign was from 17 September to 7 October 2016, but samples chosen for analysis were collected between 19 September and 5 October 2016. Detailed information about sampling periods are given in Table S1. A high-volume air sampler, HiVol3000 (Ecotech), was placed on the roof of the AIRBOX container (Figure 1), and the sampling height was~25 m above sea level. Aerosol particulates were accumulated by pumping the air through an acid-cleaned full sheet (20 cm × 25 cm) cellulose filter (Whatman 41, W41) without a size cut-off (TSP). As a result, we directly measured concentrations of trace metals in air, and use these data to estimate trace metal deposition fluxes using the methods previously described [28,[66][67][68]. Before sampling, filters were cleaned for TM analysis according to GEOTRACES procedures [69,70]. After sampling, aerosol laden filters were folded in half (aerosol layer inside) and stored until analysis in two zip lock plastic bags in the freezer. There was no TM clean laboratory available in the field. Instead, all filter preparation and handling in the field was done inside an in-house made simple box consisting of plastic pipes covered by plastic sheeting without the high-efficiency particulate air (HEPA) filter air flow (Appendix A) that was regularly cleaned with UPW. Contamination issues were investigated by blank samples analysis (Appendix A) to assess the contamination provided by in-field sample handling. Before analysis, in the clean laboratory, two 47-mm diameter subsamples of the full filter sheet were taken using a pre-cleaned circular Ti punch and one was used for leaching and digestion experiments while the second subsample was used for major ion analysis.
Atmosphere 2020, 11, x 5 of 24 Figure 1. Location of the atmospheric deposition sampling station. The white cross on the insert map and Port Curtis, where concentration of trace metals in river water were determined by Angel et al. [71](white star on the main map) for assessment of atmospheric deposition fluxes in part 3.3. Black and red symbols on the insert indicate sugar cane mill (black) and iron smelter and steel manufacturing (red) locations (National Pollution Inventory). The green line indicates the Great Barrier Reef.

Back Trajectories Analysis
To distinguish remote aerosol sources and investigate air masses transport paths, samples were classified, according to their origin using back trajectories (BT) obtained from NOAA Air Resources Laboratory Hybrid Single-Particle Lagrangian Integrated Trajectory (HYPSPLIT) and the Global Data Assimilation System (GDAS) 05 model, with 3-h intervals and 72-h BT periods ( Figure S1) [72,73]. The shorter interval allowed us to assess the homogeneity of the air masses. The height of the sampler Rainwater samples were collected on the top of the AIRBOX container roof using an in-house made rain sampler. An HDPE funnel and collection bottle (LDPE) for rainwater were cleaned according to GEOTRACES protocols [70]. Sample bottles were capped and stored in double plastic bags in the freezer until analysis. Two rain samples were collected over the entire campaign and then analyzed for trace element concentrations. Both were collected on 22 September 2016 (local time). The rain water sampler was exposed at the beginning of the rain event and deployed until several minutes after the end of the rain event. Sampling times were 165 and 135 min for the first and second rain event, respectively. The first rain event was a heavy rainfall for which the amount of collected water exceeded the sampler capacity (>500 mL, exceeding 5 mm of rainfall in the event). The second rain event was a shower and provided about 1 mm of rainfall (~100 mL). Our results revealed three fractions of Fe contained in rain water: (1) soluble, defined as the fraction passing through a W41 filter, (2) total-dissolvable, defined as the difference between being suspended in the non-filtered and filtered sample before solution acidification, and (3) particulate, defined as the digest of the fraction remaining on the filter. The same type of filter was applied for rain water filtration, which was used for dry deposition collection leaching experiments (cleaned W41). The relative contribution of Fe fractions was similar for both samples. The amount of Fe deposited on the surface area unit, called Fe wet deposition flux (F wet ), was calculated based on the funnel inlet area size, volume of collected water, and Fe concentration (1).
where: [Fe] rain is a concentration of chosen fraction of Fe (soluble, suspended, and particulate) in collected rain water, V rain is a volume of collected rain water, and A inlet is a surface area of the funnel inlet.

Back Trajectories Analysis
To distinguish remote aerosol sources and investigate air masses transport paths, samples were classified, according to their origin using back trajectories (BT) obtained from NOAA Air Resources Laboratory Hybrid Single-Particle Lagrangian Integrated Trajectory (HYPSPLIT) and the Global Data Assimilation System (GDAS) 05 model, with 3-h intervals and 72-h BT periods ( Figure S1) [72,73]. The shorter interval allowed us to assess the homogeneity of the air masses. The height of the sampler location of 25 m above sea level was applied for generating backward trajectories. Based on these BTs, samples were divided into the following groups: marine (samples MB3, MB4, MB5, MB6, MB7, MB9, MB10) and terrestrial (samples MB11, MB12, MB13, MB18A, MB18, MB20). The remaining samples represent either mixed sources (MB19) or terrestrial aerosols, which were transported over the sea and returned to the coast (MB17, MB23, MB24, MB25) (Table S2).

Wind Direction Analysis
In addition to back trajectories, samples were classified according to the wind patterns to consider contribution of local sources. Wind data was recorded by a Thompson WS800 Meteorological Station located on the top of AIRBOX container, approximately two meters from the aerosol sampler. Samples were classified into three groups based on the dominating wind direction (Figure 2

Aerosol Sample Preparation
Samples collected on 47-mm filter punches were subjected to a three-step analytical protocol described by Perron et al. [3]and then analysed by Inductively Coupled Plasma Mass Spectrometry (ICP-MS) using a similar instrument setup as that reported in Bowie et al. [74]. Briefly, the protocol consists of three consecutive steps using the same 47-mm filter punch in each step.
Step 1, ultra-pure water (UPW) leach: The subsample was put into Savillex Perfluoroalkoxy (PFA) filter holder and 50 mL of UPW was passed through the filter using a vacuum pump and collected in a Teflon container. An aliquot of the UPW leach (9.8 mL) was collected and acidified to 1% (v/v) with distilled HNO3. This fraction was termed the "soluble fraction". Step 2, ammonium acetate (pH 4.7) leach: Following step 1, the wet filter was put inside a 15-mL centrifuge tube and 10 mL of 1.1 M ammonium acetate buffer was added. The filter was soaked for one hour in the buffer solution and agitated three times (at the beginning, after 20 min and after 40 min). In the final 5 min, samples were centrifuged to separate filter fibers from the solution. Lastly, the top 4.5 mL of solution was pipetted to 15 mL PFA (Savillex) vials and evaporated to dryness. The dry sample was re-suspended in 4.5 mL 1% HNO3 (v/v) and transferred into 10 mL auto sampler tubes for analysis. This fraction was termed the "leachable fraction." Step 3, total digestion using HNO3 and HF: The remaining buffer solution (5.5 mL) and filter were transferred to a 15-mL Teflon vial and evaporated to dryness. Samples were then digested for 12 h in 1.0 mL of concentrated distilled HNO3 and 0.25 mL of concentrated HF (SEASTAR, BASELINE ® ) at 120 °C. Samples were then evaporated to dryness, redissolved in 5 mL of 50% (v/v) distilled HNO3, and digested again for 12 h at 120 °C. Once the filter was digested, the samples were evaporated to dryness once again and re-suspended in 4.5 mL of 1% (v/v) HNO3. This fraction was termed the "refractory fraction." The sum of step 1 (soluble) and step 2 (leachable) fractions was termed the 'labile fraction,' while the sum of all three fractions was termed 'total.' Recovery of the digestion procedure was measured for the reference materials, which aim to mimic mineral dust. Arizona Test Dust and Köln Loess GeoPT13 were digested and analysed in the same way as the filter sub-samples. Digestion recovery for Fe was 102% and 104% and details are reported in Perron et al. [3].

Rainwater Sample Preparation
Rainwater samples were taken out from the freezer 24 h before processing, thawed, and intensively shaken before taking aliquot. The following fractions of rain water were prepared for trace element analysis by ICP-MS: (1) unfiltered rain water (9.8 mL) was pipetted into 10 mL of autosampler tubes and acidified to 1% (v/v) using distilled HNO3, (2) 50 mL of filtered rain was filtered through an acid cleaned W41 filter, and 9.8 mL of filtrate was pipetted into 10 mL auto-sampler tubes and acidified to 1% (v/v) using distilled HNO3. This fraction is more likely comparable to the soluble fraction of dry deposition. The filters after rainwater filtering were subjected to a total digestion procedure as used for dry deposition samples.

Aerosol Sample Preparation
Samples collected on 47-mm filter punches were subjected to a three-step analytical protocol described by Perron et al. [3] and then analysed by Inductively Coupled Plasma Mass Spectrometry (ICP-MS) using a similar instrument setup as that reported in Bowie et al. [74]. Briefly, the protocol consists of three consecutive steps using the same 47-mm filter punch in each step.
Step 1, ultra-pure water (UPW) leach: The subsample was put into Savillex Perfluoroalkoxy (PFA) filter holder and 50 mL of UPW was passed through the filter using a vacuum pump and collected in a Teflon container. An aliquot of the UPW leach (9.8 mL) was collected and acidified to 1% (v/v) with distilled HNO 3 . This fraction was termed the "soluble fraction". Step 2, ammonium acetate (pH 4.7) leach: Following step 1, the wet filter was put inside a 15-mL centrifuge tube and 10 mL of 1.1 M ammonium acetate buffer was added. The filter was soaked for one hour in the buffer solution and agitated three times (at the beginning, after 20 min and after 40 min). In the final 5 min, samples were centrifuged to separate filter fibers from the solution. Lastly, the top 4.5 mL of solution was pipetted to 15 mL PFA (Savillex) vials and evaporated to dryness. The dry sample was re-suspended in 4.5 mL 1% HNO 3 (v/v) and transferred into 10 mL auto sampler tubes for analysis. This fraction was termed the "leachable fraction." Step 3, total digestion using HNO 3 and HF: The remaining buffer solution (5.5 mL) and filter were transferred to a 15-mL Teflon vial and evaporated to dryness. Samples were then digested for 12 h in 1.0 mL of concentrated distilled HNO 3 and 0.25 mL of concentrated HF (SEASTAR, BASELINE ® ) at 120 • C. Samples were then evaporated to dryness, redissolved in 5 mL of 50% (v/v) distilled HNO 3 , and digested again for 12 h at 120 • C. Once the filter was digested, the samples were evaporated to dryness once again and re-suspended in 4.5 mL of 1% (v/v) HNO 3 . This fraction was termed the "refractory fraction." The sum of step 1 (soluble) and step 2 (leachable) fractions was termed the 'labile fraction,' while the sum of all three fractions was termed 'total.' Recovery of the digestion procedure was measured for the reference materials, which aim to mimic mineral dust. Arizona Test Dust and Köln Loess GeoPT13 were digested and analysed in the same way as the filter sub-samples. Digestion recovery for Fe was 102% and 104% and details are reported in Perron et al. [3].

Rainwater Sample Preparation
Rainwater samples were taken out from the freezer 24 h before processing, thawed, and intensively shaken before taking aliquot. The following fractions of rain water were prepared for trace element analysis by ICP-MS: (1) unfiltered rain water (9.8 mL) was pipetted into 10 mL of auto-sampler tubes and acidified to 1% (v/v) using distilled HNO 3 , (2) 50 mL of filtered rain was filtered through an acid cleaned W41 filter, and 9.8 mL of filtrate was pipetted into 10 mL auto-sampler tubes and acidified to 1% (v/v) using distilled HNO 3 . This fraction is more likely comparable to the soluble fraction of dry deposition. The filters after rainwater filtering were subjected to a total digestion procedure as used for dry deposition samples.

Trace Metal Analysis by ICP-MS
All samples from the three leaching steps and wet deposition samples were analysed for TMs content using ICP-MS (Element 2, Thermo Fisher Scientific) in the Central Science Laboratory, University of Tasmania. ICP-MS instrumental parameters and detailed analytical protocol description are reported in Perron et al. [3].

Aerosol Sample Preparation
A 47-mm diameter punch of the W41 cellulose filter was folded and transferred to a 10-mL glass container. A portion of 6 mL of UPW was added and the sample was left in an ultrasonic bath for 50 min. The extract was then filtered through 0.22 µm, 33-mm diameter polyethersulfone (PESdiameter, sterile, Millex ® syringe-driven filters, and were collected in 10 mL of auto-sampler glass vials. Thawed rain samples were intensively shaken and 5 mL was poured into 10 mL of auto-sampler glass vials. constant ratio in sea spray [75].
[nss-SO 4 where [SO 4 2− ] is the atmospheric concentration of SO 4 2− determined by IC and a constant value of 0.253 is the assumed ratio between sulfate and sodium in sea spray [75].

Black Carbon, Organic Carbon, and Levoglucosan Derivatives
Black carbon and organic carbon were measured as part of the Mission Beach campaign with instrumentation in the AIRBOX chemistry laboratory. Black carbon concentrations were measured by a Multi Angle Absorption Photometer (MAAP) (Thermo Scientific Model 5012) [76] with time resolution of 5 s. Organic carbon concentrations were measured by an Aerodyne compact time of flight Aerosol Mass Spectrometer (AMS) [77] and 10-min averaged measurement data were used for further calculation in this work. The instrument vaporises aerosols at 600 • C and uses a destructive electron ionisation process to detect non-refractory aerosol species. In this campaign, we measured the absolute concentration in air of compounds characterized by mass (m) to charge (z) ratios of 60 and 73 [78], which are levoglucosan (LG) decomposition compounds.
Levoglucosan is a proxy for cellulose degradation in biomass burning [47][48][49]. Due to technical issues, LG was only measured from September 25 to the end of the campaign and, therefore, we do not have results for marine samples and less samples available for other groups. Nevertheless, we checked for correlations between LG/Fe total and FFeS, in a similar approach to that used for BC.

Scanning Electron Microscopy
All measurements were performed using the Hitachi SU-70 analytical field emission Scanning Electron Microscope (SEM) at the Central Science Laboratory, University of Tasmania. An area of approximately 1 cm 2 of sample on Whatman 41 filter was covered by powdered carbon and analysed using an electron beam voltage of 15 kV. To find the best observation area, secondary electron (SE) images of the filter were collected [79]. Backscattered electrons (BSE) were used to determine average molecular masses of aerosols [79]. x-ray Energy Dispersive Spectrum (EDS) was applied to determine the elemental composition of the selected particles [79]. Two types of EDS measurements were applied: (1) in point measurements, where the electron beam hits a single point (within the particle surface) of a particle and (2) mapping of part of the filter or selected particle by the electron beam. Five samples were chosen for analysis (MB5, MB6, MB10, MB11, MB18A) based on predominant aerosol sources determined by leaching results, back trajectories, and wind patterns.

Enrichment Factor
The enrichment factor (EF) expresses the enrichment of a particular element in the particle relative to the global crust concentration. EF compares the ratio of the element (Z) with a mineral dust tracer (here Ti) in aerosols to the average content in the upper crustal layer [80] (3). Low values approaching 1 indicate that mineral dust is the main source of the element of interest in the atmosphere. A higher EF represents a higher contribution of non-crustal source. Elements were divided into three groups based on their median EF into (1) low, EF < 2, (2) moderate, 2 ≤ EF < 10, and (3) high, EF ≥ 10. This classification is similar to the one applied by Winton et al. [28] and Buck at al. [81] for an aerosol samples origin investigation.
where [Z] (atm.) indicates concentration of element Z in the atmosphere or in the average global crust [Z]( crust ).

Other Calculations
Fractional solubilities of Fe and other TMs are the relative contribution (in percent) of the atmospheric concentrations of (i) one fraction (soluble, leachable, or labile) to the atmospheric concentration to (ii) the total concentration of this element, which is the sum of soluble, leachable, and refractory fractions. FFeS was calculated by dividing the atmospheric concentration of soluble, leachable, and labile Fe by atmospheric concentration of total Fe (4).
For studying FFeS correlations with anthropogenic emissions and combustion/biomass burning particulates, a Pearson correlation was applied. The Pearson correlation coefficient 'r' was compared with critical Pearson r c (p = 0.05 two tailed) for the specific number of pairs (n) expressed as a degree of freedom (df) where df = n − 2, to test the correlation significance (if r ≥ r c then the correlation is significant). The following correlation coefficient (r) denomination was applied: 0.00-0.19 very weak, 0.20-0.39 weak, 0.40-0.59 moderate, 0.60-0.79 strong, and 0.80-1.0 very strong [82].
The Fe dry deposition flux (F dry ) was calculated by multiplying the Fe concentration in air [Fe] atm. by the velocity of dry deposition (v dry ) (5). Where [Fe] atm. is atmospheric Fe concentration and v dry is a dry deposition velocity.

Iron Provenance
Enrichment factor of iron ranged from 0.6 to 1.2 (one outlier of 6.8 for sample MB6) with a median (± SD) of 0.9 ± 0.2, which indicates that most Fe originates from crustal materials. In addition, total Fe was very strongly correlated with the mineral dust tracers Al (r = 0.990) and Ti (r = 0.943).
Backscattered electron imaging ( Figure 3) distinguished between two classes of particles, which included irregularly-shaped mineral dust and regularly-shaped carbonaceous particles. Among analyzed samples, almost all particulates containing Fe were alumino-silicates, which explains the high correlation between total Fe and Al. On the other hand, very few Fe oxide particulates were detected.
Atmosphere 2020, 11, x 9 of 24 Backscattered electron imaging ( Figure 3) distinguished between two classes of particles, which included irregularly-shaped mineral dust and regularly-shaped carbonaceous particles. Among analyzed samples, almost all particulates containing Fe were alumino-silicates, which explains the high correlation between total Fe and Al. On the other hand, very few Fe oxide particulates were detected.
SEM observations suggest that alumino-silicates are the main source of total Fe. Iron from alumino-silicate lattices is generally more bioaccesible than from oxides due to greater chemical weathering rates [83], which may partially explain the relatively high FFeS observed in this study as discussed below.

Iron Solubility
The contribution of contamination from filter handling in the field increased the amount of Fe in the blank sample by 14.3%, 5.7%, and 4.9% for soluble, leachable, and refractory fraction, respectively, while the average contribution of Fe from the blank in the sample was 1.54 ± 0.53%, 6.37 ± 3.78%, and 4.03 ± 2.10% for soluble, leachable, and refractory fractions, respectively. Precision of the applied leaching protocol was determined based on triplicate analysis and was reported as 1.9%, 8.5%, and 9.5% for soluble, leachable, refractory fraction, respectively [3]. More details about the initial method assessment is provided in Appendix A. Based on determined atmospheric concentrations of soluble, leachable, and refractory Fe fractions, we calculated FFeS and dry deposition fluxes (Figure 4 and Table S3). Labile Fe concentrations ranged from 2.3 to 10.6 (mean ± SD 6.9 ± 2.9) ng m −3 while total Fe concentrations ranged from 29.0 to 213.7 (mean ± SD 95.2 ± 53.9) ng m −3 . Atmospheric concentration of total Fe lies in the lower range of global model estimates forecast for this part of Australia,which are 40-1000 ng m −3 [84]. The labile Fe was 4.9%-11.4% (mean ± SD 8.0 ± 2.1%) of the total Fe content, and is at a similar level to the Fe solubility estimates from the global model, which forecasts an FFeS of 4%-10% for this part of Australia [85]. Between 52.6% and 83.7% (mean ± SD 69.4 ± 9.2%) of the labile Fe originated from the soluble (UPW) leach while the buffer leach provided an additional 16.3%-47.4% (mean ± SD 30.6 ± 9.2%). The fraction of labile Fe is inversely proportional to the total Fe content, which may be explained by the coexistence of two Fe pools characterized by (a) low and (b) high labile Fe content. The relative proportion of these factors drives the solubility of the mixture. SEM observations suggest that alumino-silicates are the main source of total Fe. Iron from alumino-silicate lattices is generally more bioaccesible than from oxides due to greater chemical weathering rates [83], which may partially explain the relatively high FFeS observed in this study as discussed below.

Iron Solubility
The contribution of contamination from filter handling in the field increased the amount of Fe in the blank sample by 14.3%, 5.7%, and 4.9% for soluble, leachable, and refractory fraction, respectively, while the average contribution of Fe from the blank in the sample was 1.54 ± 0.53%, 6.37 ± 3.78%, and 4.03 ± 2.10% for soluble, leachable, and refractory fractions, respectively. Precision of the applied leaching protocol was determined based on triplicate analysis and was reported as 1.9%, 8.5%, and 9.5% for soluble, leachable, refractory fraction, respectively [3]. More details about the initial method assessment is provided in Appendix A. Based on determined atmospheric concentrations of soluble, leachable, and refractory Fe fractions, we calculated FFeS and dry deposition fluxes ( Figure 4 and Table S3). Labile Fe concentrations ranged from 2.3 to 10.6 (mean ± SD 6.9 ± 2.9) ng m −3 while total Fe concentrations ranged from 29.0 to 213.7 (mean ± SD 95.2 ± 53.9) ng m −3 . Atmospheric concentration of total Fe lies in the lower range of global model estimates forecast for this part of Australia, which are 40-1000 ng m −3 [84]. The labile Fe was 4.9-11.4% (mean ± SD 8.0 ± 2.1%) of the total Fe content, and is at a similar level to the Fe solubility estimates from the global model, which forecasts an FFeS of 4-10% for this part of Australia [85]. Between 52.6% and 83.7% (mean ± SD 69.4 ± 9.2%) of the labile Fe originated from the soluble (UPW) leach while the buffer leach provided an additional 16.3-47.4% (mean ± SD 30.6 ± 9.2%). The fraction of labile Fe is inversely proportional to the total Fe content, which may be explained by the coexistence of two Fe pools characterized by (a) low and (b) high labile Fe content. The relative proportion of these factors drives the solubility of the mixture.  Reported values for Fe solubilities around Australia are sparse and variable between sampling sites. In this scenario, we are comparing data from the Misison Beach with results of FFeS of Australian aerosols and soils to get comparision in terms of ocean fetrilisation potential and sources. The FFeS results reported in this section are similar to data reported by Winton et al. [28] for aerosols collected at a coastal site in the Northern Territory (Australia) (2%-12%) and lie in the lower range of solubilities observed in Tasmania (0.5%-56%) [29]. The labile fraction of Fe in aerosols collected at Mission Beach is similar to those reported in aerosols collected at sea during the same period (4%-  Reported values for Fe solubilities around Australia are sparse and variable between sampling sites. In this scenario, we are comparing data from the Misison Beach with results of FFeS of Australian aerosols and soils to get comparision in terms of ocean fetrilisation potential and sources. The FFeS results reported in this section are similar to data reported by Winton et al. [28] for aerosols collected at a coastal site in the Northern Territory (Australia) (2-12%) and lie in the lower range of solubilities observed in Tasmania (0.5-56%) [29]. The labile fraction of Fe in aerosols collected at Mission Beach is similar to those reported in aerosols collected at sea during the same period (4-33%) [68]. Our results are higher than those reported by Mackie et al. [57] for readily soluble Fe (dissolving in the time range from minutes to hours) from soil samples collected in Thargomindah in Southwest Queensland (0.9 ± 0.3%). However, our findings are similar for dust samples collected in Burunga in the Mellee region between Victoria and New South Wales (5.6 ± 0.7%) [57]. Results presented in this study are also higher than previous findings for mineral dust from other regions, e.g., the fractrional Fe solubility was below 1% for Saharan dust [34,45] and 4.0 ± 0.5% for Chinese Loess [34]. FFeS also ranged from 0.02% to 0.4% for coarse dust from desert regions (Sahara, Arab Desert, and Thar Desert) collected in the Arabian Sea, while, for aerosols collected on the Bay of Bengal FFeS, it was much higher at 1.4-24% [86]. Srinivas et al. [86] also reported a correlation between FFeS and non-sea salt sulfate (nss-SO 4 2− ), which suggests emissions from biomass burning and/or fossil fuel combustion was also present in samples from the Bay of Bengal. The higher FFeS of particles may be explained by a higher content of combustion particles from biomass burning and anthropogenic emission, which are more soluble than mineral dust.

Drivers of Iron Solubility
Non-Crustal Emissions vs. Iron Solubility The correlations between soluble and leachable Fe fractions with total fractions of mineral dust tracers were much weaker than between total fractions of Fe and a mineral dust tracer. This indicated the existence of other sources of soluble Fe in aerosol samples. Thus, anthropogenic and crustal emissions were tested as a potential source of soluble and leachable Fe by analyzing correlation coefficients between them ( Table 1). The correlations between FFeS with total fractions of non-crustal elements were also calculated. To consider mineral dust loadings for each sample, total atmospheric concentrations of an anthropogenic element (AE) was divided by the total atmospheric concentration of Ti. In general, significant correlations, between fractional FFeS and Ti normalised anthropogenic elements, were only observed for the soluble Fe fraction, but not the leachable fraction. Therefore, we focused only on the observed moderate (0.40 < r < 0.59) correlations between soluble Fe and Cd, Co, Cu, and Mn, and strong (0.60 < r < 0.79) correlations between soluble Fe with total Mo, Pb, V, and Zn normalized to Ti (Table 1). To assign the non-crustal emissions to more specific sources, samples were divided into groups based on their marine and terrestrial origin, as identified by BT analysis ( Figure S1) and (see 0 for more information about BT analysis). This provided information about sources and areas of possible mixing along the transport path. Local wind direction patterns (Figure 2) (sea breeze and land breeze) were also noted to assess the contribution from local anthropogenic emissions (0 for details about wind direction classification). A significant and strong (0.80 < r < 1) correlation was found between soluble Fe with Cd (r = 0.964), V (r = 0.892), and Zn (r = 0.786) for sea breeze samples. This may be explained by anthropogenic emissions from ships in the GBR region and/or from emission along the northern Queensland coast (transport, power plants, and fuel combustion) due to air masses usually passing along the coastline just prior to sampling. Moderate (Cd, Co, Mn) and even strong (Cu, Mo, Pb, V, Zn) correlations were observed for marine samples, but these results were not significant due to a limited number of samples for which the critical value of the Pearson coefficient was high. Studies of land-derived samples indicated a significant correlation between soluble Fe and Mn (r = 0.878) and V (r = 0.833) within the terrestrial group, and between soluble Fe and Pb (r = 0.833) within the land breeze group. For V, we also observed strong but not significant correlations for the land breeze group. Similarly, strong but not significant (due to a limited number of samples) correlation for both terrestrial and land breeze samples were observed for Cd and Co. Non-crustal emissions were an important source of soluble Fe, particularly for sea breeze air masses. Terrestrial-based anthropogenic emissions indicated that soluble Fe coexists in the air with Mn and V, according to BT origin classification, and with Pb, according to wind direction origin classification. Sedwick et al. [45] reported higher FFeS for aerosols collected in Bermuda when air masses passed over North America compared to those that passed over the Sahara. North American air masses were characterized by low total Fe content accompanied by elevated values of V/Al, Fe/Al, and V/Mn indicating anthropogenic combustion products. On the other hand, Fu et al. [46] found high correlation coefficients between FFeS and K from biomass burning and the V/Fe ratio in spring in Shanghai, which suggests that both biomass burning and oil ash from ship emissions are responsible for an increase in FFeS. Correlations between FFeS with V and Pb were observed in our study, which confirms anthropogenic origin of labile Fe.

Combustion Products vs. Iron Solubility
Black carbon (BC) is a proxy for combustion processes including fossil fuel and biomass burning. A correlation between the ratio of atmospheric concentrations of BC to total Fe with fractions of soluble, leachable, and labile Fe was investigated. This was analogous to the procedure applied by Fu et al. [46] in which K + normalized to total Fe was used as biomass burning tracer and correlated with FFeS.
The soluble Fe was strongly positively correlated (r = 0.601, df = 15, p < 0.05) with BC/Fe total , which indicates that the FFeS increases with increased concentrations of BC. In addition, a very strong (r = 0.851, df = 4, p < 0.05) correlation was observed between soluble Fe with BC/Fe total within the group of sea breeze samples. Within this group, strong correlations between the percent of soluble Fe with Cd, V, and Zn were reported (Table 1). We, therefore, conclude that anthropogenic combustion processes, such as fossil fuel combustion, may provide soluble forms of Fe. No significant (p < 0.05) correlations were found for samples from other origins.
A moderate correlation (r = 0.50) was found between the percent of soluble Fe and both LG indicators (m/z 60 and 73). However, for this limited set of analyzed aerosol samples for which the LG data are available (df = 9), correlations were not significant. A significant (p < 0.05) correlation was observed only for sea breeze samples for which we calculated r = 0.95 and r = 0.92 (df = 3) for organics m/z = 60 and 73, respectively. This may indicate that FFeS originated from biomass burning emission from the land, and then transported above the sea before returning to the coast and/or from emissions from sources along the coast (e.g., sugar mills, which are also a source of LG). Within the sea breeze group, significant correlation was also found between soluble Fe and BC and Cd, V, and Zn. It has been reported that substantial proportion of anthropogenic elements, such as emissions from open mining, may be accumulated on the vegetation [87,88] and, consequently, then could be re-entrained to the atmosphere during the fire event.
The correlation between soluble Fe and both BC/Fe total and LG/Fe total in the atmosphere was observed in this study and agrees with previous reports [34,46,53] where combustion processes and biomass burning were found to provide a significant amount of soluble Fe. Therefore, we suggest that both sources enhance FFeS near the GBR. The Northern Queensland coast is both more densely populated compared to the inland and is a popular tourist destination with intensive ship movements across the GBRMoreover, this is an agricultural area with numerous sugar cane plantations and sugar mills, which produce energy by burning the bagasse, to generate electricity and steam for factory operations, while producing around 500 GWh [89]. Some of the sugar mills are in the proximity to the sampling site, in Tully (approx. 20 km S-E), South Johnstone (approximately 30 km N-E) and Mourilyan (approximately 35 km N-E) (National Pollution Inventory) (Figure 1). Similar to the enrichment of anthropogenic elements, we observed a higher correlation between the percentage of soluble Fe and total Fe normalized BC and LG during sea breeze conditions that may be explained by emissions from along the coast, from ship movements across the GBR, and emissions from the sugar mills located in close proximity to the sampling station. Bushfires occurring several hundred kilometers north-west and south-west of the sampling site around the time of sampling are another potential source of biomass burning emission [90]. However, due to limited LG data, this cannot be confirmed.
Observation of our samples by SEM confirm carbonaceous particulates were more abundant in some samples, particularly in these for which a high level of BC and LG was observed. Round shape particulates were more homogenously distributed across the filter area and their size range was more uniform compared to the minerals. These particles were 1-5 µm in size, which mainly consist of homogenously distributed C, P, and S. Some of the carbonaceous particles also contained homogenously distributed Zn, K, and Ca ( Figure 5). These particles may be classified as tar balls, which is a common particle from natural and industrial biomass burning [91,92].
Atmosphere 2020, 11, x 13 of 24 across the GBRMoreover, this is an agricultural area with numerous sugar cane plantations and sugar mills, which produce energy by burning the bagasse, to generate electricity and steam for factory operations, while producing around 500 GWh [89]. Some of the sugar mills are in the proximity to the sampling site, in Tully (approx. 20 km S-E), South Johnstone (approximately 30 km N-E) and Mourilyan (approximately 35 km N-E) (National Pollution Inventory) (Figure 1). Similar to the enrichment of anthropogenic elements, we observed a higher correlation between the percentage of soluble Fe and total Fe normalized BC and LG during sea breeze conditions that may be explained by emissions from along the coast, from ship movements across the GBR, and emissions from the sugar mills located in close proximity to the sampling station. Bushfires occurring several hundred kilometers north-west and south-west of the sampling site around the time of sampling are another potential source of biomass burning emission [90]. However, due to limited LG data, this cannot be confirmed. Observation of our samples by SEM confirm carbonaceous particulates were more abundant in some samples, particularly in these for which a high level of BC and LG was observed. Round shape particulates were more homogenously distributed across the filter area and their size range was more uniform compared to the minerals. These particles were 1-5 m in size, which mainly consist of homogenously distributed C, P, and S. Some of the carbonaceous particles also contained homogenously distributed Zn, K, and Ca ( Figure 5). These particles may be classified as tar balls, which is a common particle from natural and industrial biomass burning [91,92]. . Spatial elemental distribution in combustion particulates (tar balls). Grey image on the left is the BSE image showing the particle shape while the colour images observed by EDS x-ray microprobe show elemental distribution per single element (stated with measured X-ray emission line above the image) across the particle.

Aging Processes vs. Iron Solubility
Possible aging processes were studied by analyzing soluble major ion (MI) contents in our aerosol samples and correlating them with FFeS. Results indicate that relationships exist between the Figure 5. Spatial elemental distribution in combustion particulates (tar balls). Grey image on the left is the BSE image showing the particle shape while the colour images observed by EDS x-ray microprobe show elemental distribution per single element (stated with measured X-ray emission line above the image) across the particle.

Aging Processes vs. Iron Solubility
Possible aging processes were studied by analyzing soluble major ion (MI) contents in our aerosol samples and correlating them with FFeS. Results indicate that relationships exist between the presence of aging agents represented by anions (e.g., NO 3 − , nss-SO 4 2− ) and increased FFeS. Normalization of the MI atmospheric concentration to the total Fe atmospheric concentration was applied to minimize mineral dust contribution to MI content by following Fu et al. [46]. Strong and significant (p < 0.05) correlations between soluble Fe and total Fe normalized concentrations of NO 3 − (r = 0.669, df = 16) and nss-SO 4 2− (r = 0.724, df = 16) were found, which indicates that aging processes may increase Fe solubility. For marine samples, the correlation for these anions was even stronger (and significant, p < 0.05): r = 0.767 for both (df = 5). A strong and significant (p < 0.05) correlation (r = 0.925, df = 4) was also found between nss-SO 4 2− Fe total −1 and FFeS for sea breeze samples (collected 6 AM-6 PM). There were no significant correlations for either land breeze samples (collected 6 PM-6 AM) or terrestrial samples. Our results indicate that NO 3 − and nss-SO 4 2− anions coexist with soluble Fe. Both anions may exist in the atmosphere as acids or salts (commonly known as ammonium sulfate). Both nitric and sulphuric acids are derived from oxidation of precursor gases. Sulphur dioxide (SO 2 ) originates from combustion processes or oxidation of dimethyl sulphide (produced by surface ocean biota) [93]. Nitrogen oxides (NO x ) are produced from fossil fuel combustion and biomass burning [94], but also from lightening [95], soil emissions [96], and organic marine nitrate [97]. Nitric and sulphuric acid play two different roles: (1) as indicators of combustion processes, which provide highly soluble Fe and/or (2) as a source of protons, which increases FFeS for insoluble or low solubility Fe species such as minerals [98]. In this study, the correlation between FFeS and major ions indicates possible aging processes. However, correlations were also observed between FFeS and combustion and biomass burning emissions as well as elements of elevated EF. These observations suggest that major ions may simply be emitted together with the pyrogenic Fe.

Estimation of Fe Deposition Fluxes
Correlation coefficient between atmospheric concentrations of total Fe and total Ti was very high and their (atmospheric concentrations of total Fe and total Ti) ratio was similar to the value for the global average of the upper crust (see part 3.1.1. for details), which indicates a crustal origin that usually exists in the form of coarse particulates above 2.5 µm in size, and was confirmed by SEM images. A value of dry deposition velocity (V dry ) (4) is sensitive to the wind speed and humidity as well as particle size profile. For coarse crustal particulates, V dry has been estimated to range from 0.3 to 3.0 cm s −1 [66] and was assumed of 2 cm s −1 for coastal regions of Australia [28,67,68] with uncertainty of 50% [28,29]. The calculated dry deposition Fe flux ranged from 0.055 ± 0.027 to 0.258 ± 0.129 µmol m −2 day −1 (mean 0.143 ± 0.072) for the soluble, 0.017 ± 0.008 to 0.138 ± 0.069 µmol m −2 day −1 (mean 0.070 ± 0.035) for the leachable, and 0.802 ± 0.401 to 6.27 ± 3.14 µmol m −2 day −1 (mean 2.73 ± 1.37) for the refractory Fe fraction. Hence, the total Fe deposition ranged from 0.897 ± 0.448 to 6.61 ± 3.31 µmol m −2 day −1 (mean 2.95 ± 1.47). However, soluble and labile Fe appeared to be linked with the anthropogenic emissions/combustion processes (see parts 3.1. 3.1 and 3.1.3.2), which usually exist in fine mode (<2.5 m) and, consequently, their deposition flux may be lower (assumed as 0.2 cm s −1 in part 3.3 for anthropogenic elements). These finer particulates may reach more remote ocean areas.
The mean flux of labile Fe in this study was 0.213 ± 0.107 µmol m −2 day −1 , which is lower than results of marine aerosols collected at sea (Coral Sea Marine Region) during the same time period, 0.303 ± 0.590 µmol m −2 day −1 [68] calculated based on the same deposition velocity. Fluxes reported in this case are also similar to the lower results reported by Winton et al. [28] for Northern Territory, 0.2 ± 0.1-4 ± 2 µmol m −2 day −1 (also calculated for the same deposition velocity). The Northern Territory samples were collected during the dry season and had more land than marine-based origin compared to the samples collected in this study.

Iron in Rain Water
The fluxes of Fe provided by two rain events combined were 0.165, 0.195, and 0.659 µmol m −2 , for soluble, total-dissolvable, and particulate fractions, respectively. This corresponded to 16.2%, 19.2%, and 64.6% of total Fe in the rain, respectively. The second rain event provided more Fe (soluble as well as suspended and refractory) despite providing approximately five times less rainfall. Rain events tend to wash out the particulates from the atmophere at the beginning of the rain event. In case of the first rain events in our study, the amount of washed out iron was likely highly diluted by the cleaner rain water falling in the later parts of the rain event. The wet deposition flux of soluble Fe exceeded the dry deposition soluble Fe flux recorded on this day (0.073 µmol m −2 day −1 ) as well as the average dry deposition for the entire sample set (0.143 µmol m −2 day −1 ). Consequently, the amount of soluble Fe deposited in two rain events is equal to soluble Fe depostited by dry deposition for 28 h if the average for the sampling campaign soluble Fe flux is applied. These results highlight the importance of wet deposition as a source of soluble Fe during the dry season in Northern Queensland. It is worth noting the difference in Fe concentrations in two rain events occurring on the same day. Variability may be much greater when considering rain events occurring in different parts of the year and depending on BT of air masses. These wet deposition Fe results gives 0.248 µmol m −2 of soluble Fe flux in wet deposition in total during the sampling period. The average fraction of soluble Fe in collected rain water samples was 16.2%, which is twice as high as FFeS for dry deposition samples collected on the same day of the rain event (8.1% for soluble and 10.6% for labile). Similar FFeS for dry and wet depositions was also reported for most of the rain samples collected over the Sargasso Sea (typically below 4%) [45] and the Mediterranean Sea (0.5-27%) [99]. Theodosi et al. [99] also identified strong source and acidity influences on this parameter. Leaching experiments conducted on samples from the same region revealed FFeS of approximately 1% and 12% for Saharan dust and anthropogenic emission, respectively [100]. Our study confirmed the significant role of single wet deposition events as a source of soluble Fe, as previously reported [67,101]. Moreover, long-term records indicate that rainfall is typically higher than observed during our campaign, which suggests that the contribution of wet Fe deposition to the total Fe deposition may be even higher.

Atmospheric Deposition of Coral Toxins and Other Bioactive Metals
Origins of Cd, Co, Cu, Mn, Mo, Pb, V, and Zn were investigated based on the EF analysis, by following the method described in 2.10 using Equation (3). Due to concentrations spanning several orders of magnitude, log(EF) is presented on a linear scale ( Figure 6). Elements were divided into three groups based on their median EF into (1) low, EF < 2: Co, Fe, Mn, (2) moderate, 2 ≤ EF < 10: Mo, Pb, V, and (3) high, EF ≥ 10: Cd, Cu, Zn. Results indicate a significant contribution of non-mineral dust sources to the atmospheric TMs content at Mission Beach. Only Co and Mn (in addition to Fe) had a low EFs indicating mostly a crustal origin. Moderate contributions of anthropogenic emissions were observed for Mo, Pb, and V, which were classified as having a mixed origin overall. Lastly, Cd, Cu, and Zn had the highest EFs and, thus, we assumed they were mostly emitted by anthropogenic sources. However, Shotyk et al. [102] reported the relatively elevated EF for ancient peat samples originating from the mid-Holoscene and Boutron et al. [103] found severe variations and peaks of Cd in the Antarctic ice and snow in the last 155,000 years, which may not be simply accounted for by crustal emissions and volcano eruptions. Correlation of analysed TMs with the mineral dust tracers, Al and Ti, was weaker for elements of higher EF and the correlation increases with the drop of EF between elements, which confirms that Fe, Co, and Mn had a crustal origin while Cd, Cu, Zn, Mo, V, and Pb were rather independent on the mineral dust concentration (Figure 6). The main delivery of heavy metals to the GBR has been linked with riverine discharges associated with agriculture and stormwater runoff, and waste water from ships [104]. In this case, we studied atmospheric deposition as an additional source of heavy metals to GBR. Atmospheric concentrations of total and labile forms of elements, which were found to have toxic influences on corals, Cu, Zn, and Cd [21,[23][24][25], were determined. Deposition fluxes of labile fractions were estimated to assess the amount that may be available to marine biota (Table S4). The calculations were analogous to these for Fe (details in 3.1.2), but an order of magnitude lower deposition velocity (0.2 cm s −1 ) was applied for Cd, Cu, Mo, Pb, V, and Zn since they showed mixed or anthropogenic signatures, which typically exists in the finer form than mineral dust [66]. The mean (±SD) atmospheric concentration of total Cu was 1.54 ± 1.62 ng m −3 while the fraction of labile Cu was 52.6 ± 23.2%, which results in a flux of labile Cu of 1.74 ± 0.87 nmol m −2 day −1 . The mean ( ± SD) atmospheric concentration of total Cd was 0.069 ± 0.11 ng m −3 with 96.7 ± 4.0% in the labile form, which results in a flux of 0.11 ± 0.05 nmol m −2 day −1 of labile Cd. Analogous values for Zn were 3.50 ± 2.15 ng m −3 , 84.6 ± 19.5%, and 7.93 ± 3.95 nmol m −2 day −1 . All of these elements (Cd, Cu, and Zn) showed an anthropogenic signature and relatively high fraction of the labile form.
Deposition fluxes of potentially bioavailable forms of TMs obtained in this study were compared to the estimations for the riverine freshwater delivery based on data of discharged water volume to the GBR from Fabricius et al. [105] and TM concentrations measured in the Port Curtis estuary (Queensland) [71] (Table 2). We acknowledge the large uncertainty of this comparison as a result of (1) using short term data on atmospheric deposition (18 aerosol samples and 2 rain samples) not reflecting the seasonal variations, (2) using data from one sampling point for atmospheric deposition and two catchements for riverine input, which do not represent the diversity of the large area of GBR and its catchment, and (3) uncertainties of the calculation of deposition fluxes and riverine imputs. However, the aim of this comparison was to test the hypothesis that atmospheric deposition of heavy metals may be comparable to riverine discharge, which is currently considered to be the largest source of nutrients and contaminants. Table 2. Comparison of fluxes (in mol m −2 y −1 ) of Cd, Cu, and Zn to the GBR from the riverine system with dry and wet atmospheric deposition from this study. (*) Concentrations of TMs in river water Cd (7.7 ± 6.9 ng L −1 ), Cu (514 ± 115ng L −1 ), and Zn (153 ± 61 ng L −1 ) (based on data for Port Curtis and The Rivers [71]) volume of rivers freshwater to the GBR of 934,000,000 ML and the GBR area of 344,000 km 2 [105]. Wet deposition was estimated by assuming flux recorded during the campaign, which was The main delivery of heavy metals to the GBR has been linked with riverine discharges associated with agriculture and stormwater runoff, and waste water from ships [104]. In this case, we studied atmospheric deposition as an additional source of heavy metals to GBR. Atmospheric concentrations of total and labile forms of elements, which were found to have toxic influences on corals, Cu, Zn, and Cd [21,[23][24][25], were determined. Deposition fluxes of labile fractions were estimated to assess the amount that may be available to marine biota (Table S4). The calculations were analogous to these for Fe (details in 3.1.2), but an order of magnitude lower deposition velocity (0.2 cm s −1 ) was applied for Cd, Cu, Mo, Pb, V, and Zn since they showed mixed or anthropogenic signatures, which typically exists in the finer form than mineral dust [66]. The mean (±SD) atmospheric concentration of total Cu was 1.54 ± 1.62 ng m −3 while the fraction of labile Cu was 52.6 ± 23.2%, which results in a flux of labile Cu of 1.74 ± 0.87 nmol m −2 day −1 . The mean ( ± SD) atmospheric concentration of total Cd was 0.069 ± 0.11 ng m −3 with 96.7 ± 4.0% in the labile form, which results in a flux of 0.11 ± 0.05 nmol m −2 day −1 of labile Cd. Analogous values for Zn were 3.50 ± 2.15 ng m −3 , 84.6 ± 19.5%, and 7.93 ± 3.95 nmol m −2 day −1 . All of these elements (Cd, Cu, and Zn) showed an anthropogenic signature and relatively high fraction of the labile form.
Deposition fluxes of potentially bioavailable forms of TMs obtained in this study were compared to the estimations for the riverine freshwater delivery based on data of discharged water volume to the GBR from Fabricius et al. [105] and TM concentrations measured in the Port Curtis estuary (Queensland) [71] (Table 2). We acknowledge the large uncertainty of this comparison as a result of (1) using short term data on atmospheric deposition (18 aerosol samples and 2 rain samples) not reflecting the seasonal variations, (2) using data from one sampling point for atmospheric deposition and two catchements for riverine input, which do not represent the diversity of the large area of GBR and its catchment, and (3) uncertainties of the calculation of deposition fluxes and riverine imputs. However, the aim of this comparison was to test the hypothesis that atmospheric deposition of heavy metals may be comparable to riverine discharge, which is currently considered to be the largest source of nutrients and contaminants. Table 2. Comparison of fluxes (in µmol m −2 y −1 ) of Cd, Cu, and Zn to the GBR from the riverine system with dry and wet atmospheric deposition from this study. (*) Concentrations of TMs in river water Cd (7.7 ± 6.9 ng L −1 ), Cu (514 ± 115ng L −1 ), and Zn (153 ± 61 ng L −1 ) (based on data for Port Curtis and The Rivers [71]) volume of rivers freshwater to the GBR of 934,000,000 ML and the GBR area of 344,000 km 2 [105]. Wet deposition was estimated by assuming flux recorded during the campaign, which was an equivalent of proportion of recorded 6 mm of rainfall to the average annular rainfall to the GBR of 2010 mm.

Element
Riverine ( Atmospheric deposition provides more Zn in both dry (7.9 ± 1.4 µmol m −2 y −1 ) and wet (40.1 ± 5.9 µmol m −2 y −1 ) form when compared to riverine input (5.5 ± 2.5 µmol m −2 y −1 ). Wet deposition is also an important carrier of soluble Cd (1.26 ± 0.54 µmol m −2 y −1 ), ahead of riverine input (0.19 ± 0.17 µmol m −2 y −1 ) and dry deposition (0.04 ± 0.06 µmol m −2 y −1 ). On the other hand, rivers deliver the majority of dissolved Cu (22.0 ± 4.9 µmol m −2 y −1 ) more than wet (8.0 ± 0.7µmol m −2 y −1 ) and dry (0.6 ± 0.4 µmol m −2 y −1 ) deposition. Therefore, atmospheric deposition cannot be neglected as a source of toxins [21] as it may be delivered to remote sites of the GBR in a more effective way than freshwater discharges. Heavy metals have been previously shown to cause a reduction of spawning efficiency [23,24,106] and a strong accumulation of toxins by corals [107,108]. In addition, coral spawning occurs for a few nights in late spring or early summer before the wet season. This suggests that, at the time of spawning, riverine delivery of toxins is limited and, consequently, the atmospheric deposition contribution may be even greater. Iron fluxes could not be compared due to the lack of riverine Fe data. Further investigations are needed to understand the significance of atmospheric deposition of toxins to the GBR.

Conclusions
The application of leaching experiments followed by bulk analysis of trace metals in aerosol samples resulted in a unique data set of atmospheric TM deposition to a globally significant coral region known as the Great Barrier Reef. Our study showed that mineral dust, particularly alumina-silicates, was the main source of total Fe. However, Fe from mineral dust was not a dominant source of soluble and labile forms of Fe. We observed relatively high Fe solubility near the GBR, which was linked to Fe originating from anthropogenic emissions such as fossil fuel combustion and biomass burning. Furthermore, we revealed that a prevalent proportion of coral toxins such as Cu, Zn, or Pb delivered from the atmosphere originate from combustion processes including anthropogenic emissions. Due to their high solubility, they may enter the local food chains rapidly. Combustion processes are currently growing due to climate change (more frequent bushfires in Australia) and growing industrial and touristic development of Queensland (higher emission of anthropogenic aerosols). Therefore, ocean fertilization of the GBR by labile Fe may be expected to increase in the future. The previously mentioned processes are also a source of potential toxins, and our study indicates their contribution should not be neglected in the total toxic element delivery budget. Evaluation of effects of the atmospheric TM deposition on the GBR ecosystem was beyond the aim of this study. However, our study emphasizes the importance of atmospheric deposition of TMs in the GBR region including Fe and toxic elements. Our study is limited to land-based investigations in a rural part of Australia. Further work is required to understand the effect of natural and anthropogenic atmospheric TM emissions on coral reefs. Lastly, our results of Fe deposition fluxes reported here match the global Fe atmospheric deposition models in terms of atmospheric concentration of total Fe and FFeS [84], despite the fact that there is limited data for the models for the Southern Hemisphere. Our study also raises the importance of wet deposition, which may deliver a great quantity of TMs in a short time frame.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4433/11/4/390/s1: Figure S1. Air mass back trajectories of the aerosol samples and consequent origin classification, Table S1. Dry deposition aerosol samples log sheet. Measured volume was corrected to the temperature and pressure conditions, Table S2. Classification of aerosol samples based on their origin. Table S3. Fractional solubility, atmospheric concentration, and dry deposition flux of Fe, Table S4. Mean (±SD) values of total atmospheric concentration, labile fraction, and dry deposition flux of bioactive elements.

Iron Blanks
Two types of blank samples were analyzed to track sources of contamination during filter handling in the field when TM clean conditions were unavailable, and a simple house-made 'cleanbox' (Figure A1) was used to minimize contamination issues. The laboratory blank (LB) is a clean filter, which has not been used for aerosol collection (remaining double bagged until processing) while the procedural blank (PB) is a filter exposed on the sampler for 10 min with the vacuum pump being turned off. Results presented below are for a 47-mm diameter filter punches of both LB and PB filters. More details are available from Perron et al. [3]. The averaged values of four LB and three PB samples were 1.8 ng and 2.1 ng per filter, respectively, for the soluble Fe fraction, and 3.3 ng and 3.5 ng per filter for the leachable Fe fraction and 87.1 and 91.6 ng for the refractory Fe fraction. The difference between these two blanks (LB and PB) can be used as an indicator of contamination caused by handling of the filter. Consequently, the Fe content in PB was 14.3%, 5.7%, and 4.9% higher than in LB for soluble, leachable, and refractory fractions, respectively. Increase (between LB and PB) in the soluble fraction of Fe is considerate. The averaged values of four LB and three PB samples were 1.8 ng and 2.1 ng per filter, respectively, for the soluble Fe fraction, and 3.3 ng and 3.5 ng per filter for the leachable Fe fraction and 87.1 and 91.6 ng for the refractory Fe fraction. The difference between these two blanks (LB and PB) can be used as an indicator of contamination caused by handling of the filter. Consequently, the Fe content in PB was 14.3%, 5.7%, and 4.9% higher than in LB for soluble, leachable, and refractory fractions, respectively. Increase (between LB and PB) in the soluble fraction of Fe is considerate. However, the blank contribution in relation to the amount of soluble Fe contained in aerosol samples was, in most cases, below 2% (blue dots in Figure A2). This proves that applied in-field laboratory conditions and leaching protocol did not cause serious sample contamination and, consequently, they were suitable for TM clean handling of filters in the field.
The average PB has been used to correct sample concentration because this blank accounts for every stage of possible contamination. The single standard deviation from blank replicates has been used for uncertainty calculation. Generally, the average contribution of the PB to the Fe content in samples was 1.54 ± 0.53%, 6.37 ± 3.78%, and 4.03 ± 2.10 for soluble, leachable, and refractory fractions, respectively. Data for individual samples are given in Figure A2. However, the blank contribution in relation to the amount of soluble Fe contained in aerosol samples was, in most cases, below 2% (blue dots in Figure A2). This proves that applied in-field laboratory conditions and leaching protocol did not cause serious sample contamination and, consequently, they were suitable for TM clean handling of filters in the field. The average PB has been used to correct sample concentration because this blank accounts for every stage of possible contamination. The single standard deviation from blank replicates has been used for uncertainty calculation. Generally, the average contribution of the PB to the Fe content in samples was 1.54 ± 0.53%, 6.37 ± 3.78%, and 4.03 ± 2.10 for soluble, leachable, and refractory fractions, respectively. Data for individual samples are given in Figure A2.

Digestion Procedure Recovery
Recovery of the digestion procedure was measured for the reference materials, which aim to mimic the mineral dust. Arizona Test Dust and Köln Loess GeoPT13 were digested and analysed in the same way as the filter sub-samples. Recoveries using the digestion procedure presented in this case were a subject of analytical method development and was already reported by Perron et al. [3]. Reported recovery for Fe was 102% and 104% for Arizona Test Dust and Köln Loess, respectively [3]. The recoveries were 99% and 105% for Al, 97% and 103% for Co, 102% and 109% for Cu, 99% and 105% for Mn, 82% and 80% for Pb, 100% and 84% for Ti, and 94% and 100% for V [3].

Precision of the Leaching Protocol and Digestion
To test the sample's homogeneity (aerosol distribution across the filter) and precision of the applied leaching protocol, triplicate analysis of one sample has been conducted. Data from triplicate analysis was used to determine relative standard deviation and was 1.9%, 8.4%, and 9.5% for soluble, leachable, and refractory fraction of Fe, which gives 9.5% of uncertainty in total Fe fractions [3].
To test the sample's homogeneity (aerosol distribution across the filter) and precision of the applied leaching protocol, triplicate analysis of one sample has been conducted. Data from triplicate analysis was used to determine relative standard deviation and was 1.9%, 8.4%, and 9.5% for soluble, leachable, and refractory fraction of Fe, which gives 9.5% of uncertainty in total Fe fractions [3].