Marine versus Continental Sources of Iodine and Selenium in Rainfall at Two European High-Altitude Locations.

The essential elements selenium (Se) and iodine (I) are often present in low levels in terrestrial diets, leading to potential deficiencies. Marine I and Se emissions and subsequent atmospheric wet deposition has been suggested to be an important source of I and Se to soils and terrestrial food chains. However, the contribution of recycled moisture of continental origin to I and Se to precipitation has never been analyzed. Here we report concentrations and speciation of I and Se, as well as of bromine (Br), sulfur (S), and DOC-δ13C signatures for weekly collected precipitation samples (in the period of April 2015 to September 2016) at two high altitude sites, i.e., Jungfraujoch (JFJ; Switzerland) and Pic du Midi (PDM; France). Analysis of precipitation chemistry and moisture sources indicate combined marine and continental sources of precipitation and Se, I, Br, and S at both sites. At JFJ, concentrations of I and Se were highest when continental moisture sources were dominant, indicating important terrestrial sources for these elements. Furthermore, correlations between investigated elements and DOC-δ13C, particularly when continental moisture source contributions were high, indicate a link between these elements and the source of dissolved organic matter, especially for I (JFJ and PDM) and Se (JFJ).


■ INTRODUCTION
Importance of Iodine (I) and Selenium (Se) in Human Health. Many trace elements are essential for human health, however, low dietary intake can cause micronutrient deficiencies that have been estimated to affect over 3 billion people around the globe. 1 Two such essential trace elements, which are often deficient in human diets are iodine (I) and selenium (Se). 2,3 Iodine plays a crucial role in biochemical processes in the body as it is the main component of thyroid hormones. 4 Despite the fact that the body requires relatively small amounts of I, deficiency disorders caused by inadequate intakes are widespread. 2,5 Also, Se is an essential element for humans as it is needed in a small set of selenoproteins that serve redox functions. 6−9 Sources of both I and Se in the human diet are food crops (especially for Se), meat, dairy products, sea food, and food additives, especially for I in the form of iodized salt. 5,10 Marine-Derived Precipitation as a Source of I and Se to Terrestrial Environments. Precipitation in air masses with a marine moisture origin has been hypothesized to be an important source of I and Se to terrestrial environments and thus agricultural soils and food crops. 5,11 Particularly biogenic marine emissions have been suggested as a dominant source of I and Se to the marine atmosphere. 5,11−15 Marine biogenic I emissions are mainly in the form of methyl iodide (CH 3 I) and diiodomethane (CH 2 I 2 ), and for CH 3 I marine emission rates of up 610 Gg yr −1 have been reported. 16 Also abiotic production of volatile I species (e.g., I 2 ) constitutes an important pathway for marine I emissions. 17,18 These emitted I compounds photodissociate rapidly in the atmosphere to generate I atoms. Following gas-phase reactions (photolysis and oxidation), iodine oxides are formed (IO, I x O y (x = a − b, y = a − b) and hypoiodous acid: HOI), which are taken up in the aqueous phase. 17 In the aerosol, various inorganic I species may be present (e.g., I − , HOI, I 2 , ICl, IBr, IO 3 − ), however, the IX (X = Cl, Br, I) species are volatile and therefore partition more to the gas-phase than to the liquid-phase. 17 Hypoiodous acid reacts with various reductants (e.g., sulfite and DOM) to form I − , 19 which is together with IO 3 − the main I species found in rainfall. 12,20 For Se, fewer studies on volatile emissions are available but, similar to I, a dominant marine biogenic source for Se was suggested for the Southern Hemisphere. 21 A number of studies have measured volatile Se concentrations and fluxes in marine environments and estuaries. 25−29 These studies showed that dimethyl selenide (DMSe; [CH 3 ] 2 Se) was the main gaseous Se species, followed by dimethyl selenenyl sulfide (DMSeS; [CH 3 ] 2 SeS), and dimethyl diselenide (DMDSe; [CH 3 ] 2 Se 2 ). These volatile Se species undergo photolysis, hydrolysis, and gas-phase reactions with hydroxyl radical, nitrate radical, ozone, and halogen atoms 22 leading to relatively short estimated lifetimes, i.e., between 5 min and ∼6 h. 23 Wang and Tang 22 showed in a theoretical study that the reaction of DMSe and OH produces dimethylselenoxide (DMSeO), which can be further oxidized to elemental Se and Se oxide (SeO). Theoretical predictions indicate that SeO can be further oxidized to SeO 2 , via reaction with HO 2 . 24 SeO 2 is water-soluble and forms selenous acid (H 2 SeO 3 ), which can be further oxidized to selenic acid (H 2 SeO 4 ).
Marine versus Continental Sources. Although most studies indicate that marine environments are the main source areas of volatile organic I and Se emissions, evidence is increasing that terrestrial environments could also be important emitters of volatile organic I and Se compounds. For example terrestrial sources of CH 3 I have been reported for rice paddies, 30,31 forests, 32 salt marshes, and coastal wetlands. 33,34 A recent study in a pristine wetland in South Switzerland revealed substantial volatile fluxes of Se (max. 120 μg m −2 d −1 ). 35 Extrapolation of observed fluxes to global wetland areas equals 5−10% of the currently estimated marine Se volatilization. 35 Furthermore, in a more holistic approach using airplane measurements over the United States and the North Atlantic as well as data from a long-term groundobserving network 36 determined a terrestrial flux of CH 3 I around 2.7 μg m −2 d −1 , which was found to be largely derived from vegetation and soils. Interestingly, this flux is similar in magnitude to previous estimates of the marine fluxes. 36 However, it is likely that terrestrial emissions (e.g., from soils, vegetation, wetlands, and surface waters) of I and Se are still largely unknown due to the large heterogeneity of terrestrial environments and the lack of data from these different environments.
Atmospheric Deposition of I and Se. I and Se are deposited to the earth surface via both wet and dry deposition. 5,11 A number of studies have determined concentrations and speciation of I and Se in wet deposition. For I in rainfall, total concentrations from 7.88 nM for continental rainfall to 39.4 nM for marine rainfall were reported. 5 Furthermore, Gilfedder et al. 20,37 measured average I concentrations of 10 nM, 14 nM, and 11 ± 7 nM in rain samples from the Swiss Alps, the Black Forest, and Lake Constance, respectively, using ICP-MS (inductively coupled plasma mass spectrometry) as well as speciation by IC-ICP-MS (ion chromatography-inductively coupled plasma mass spectrometry). The majority of I was suggested to be organically bound I, followed by iodide (I − ), and iodate (IO 3 − ). 12,20,37 There are a number of studies that measured Se concentrations in rainfall at various locations. 11,38−52 These studies reported concentrations between 0.06 and 32 nM, with highest values measured close to industrial sites. Speciation of Se may give information on sources and atmospheric transformations; however, Se speciation in rainwater was only reported in a limited number of papers, e.g., refs 41, 42, 46, 47, 49, 50, 53, and 54. Most of these studies used hydride generation atomic absorption spectrometry (HG-AAS) or hydride generation atomic fluorescence spectrometry (HG-AFS) combined with oxidative or reductive pretreatments of the sample to quantify Se(IV), the sum of Se(IV) and Se(VI), and an operationally defined organic fraction of Se. 34,42 More recently, Wallschlager & London 54 analyzed a suburban rainwater sample using liquid chromatography coupled to HG-ICP-MS technology (IC-HG-ICP-DRC-MS) 54 and found selenate (SeO 4 2− ) and selenite (SeO 3 2− ). Apart from this single rainwater sample, ICP-MS based Se speciation data for rainwater are not available.
Although long-range atmospheric transport of marinederived I and Se has been suggested, 14,15,21,41,55−57 studies that directly link I and Se in rainfall to the origin of moisture sources are scarce. To study potential long-range transport and contributions of marine versus continental moisture sources, we investigated time series of I and Se concentrations and speciation in wet deposition at two high altitude sites, i.e., Jungfraujoch (Alps, Switzerland) and Pic du Midi (Pyrenees, France). Apart from I and Se, we determined bromine (Br) and sulfur (S) concentrations and speciation. Br and S share chemical properties and environmental pathways with I and Se; however, they generally have higher environmental concentrations than I and Se, e.g., about 10 3 times higher in oceans. 11,58,59 Furthermore, Br and S have well-known biogenic marine sources (e.g., bromoform (CHBr 3 ) and dimethyl sulfide (DMS)). 60 Shorter-lived bromo-organic compounds, such as CHBr 3 , decompose in the troposphere, whereas longer-lived bromo-organic compounds may slowly ascend across the tropopause. 61 Tropospheric inorganic Br species are produced from Br lost from sea salt aerosol (referred to as debromination), photolysis, and oxidation of CHBr 3 (by OH) and other short-lived bromo-organic compounds. Inorganic Br cycles between radical BrO x species (x = 1−3) and nonradical reservoirs (e.g., HOBr, BrO 3 − , and BrNO 3 ). 62 These species undergo heterogeneous reactions and can finally form Br − in aerosols, 62 which is the dominant Br species found in rainfall. 37 Atmospheric DMS can be oxidized both in the gas-and the aqueous-phase. In the gas-phase, two main reaction pathways occur: addition of an O atom and abstraction of an H atom. 63 The H-abstraction pathway in the gas-phase leads predominantly to SO 2 , whereas the addition pathway leads to MSA involving multiphase processes. 64 Gasphase oxidation is dominated by reactions with BrO, chlorine, and the hydroxyl (OH) radical. 64 Finally, SO 2 is either oxidized in the gas-phase (by reaction with the OH radical), or it can be taken up by cloud droplets and undergo aqueousphase oxidation. 65 Aqueous phase oxidation of DMS proceeds mainly via reaction with ozone, yielding DMSO, which is further oxidized to MSIA and MSA by HO radical and ozone. 64 Specifically, the objectives of this study were (i) to investigate the concentration and speciation of I, Se, Br, and S in relation to continental and marine origin of precipitation and (ii) to get first insights into potential biogenic versus anthropogenic sources of I, Se, Br, and S. Furthermore, this study presents the first quantitative and qualitative data on I and Se in precipitation at high-altitude sites and time series for speciation of I, Se, Br, and S in precipitation. These sites are particularly suitable to investigate long-range transport as they are often exposed to free tropospheric air. 66,67 At both sites, bulk precipitation was collected with an open sampler (JFJ bulk/PDM bulk) with a collection area of 0.04 m 2 . In addition to the bulk sampler at PDM, wet-only precipitation (referred to as "PDM wet-only", collection only during rain events) was collected with a sampler from Eigenbrodt (NSA181/E HD-PE), especially designed for high altitudes. The precipitation at JFJ (JFJ bulk) was collected weekly from July 2015 to September 2016 (28/07/15−27/09/ 16). The collected precipitation was filled in precleaned 5 L high density polyethylene containers (HD-PE) (Supporting Information, SI). In the case of ice formation, the sampler was brought inside and capped with a lid until the rainwater was thawed (approximately 6−12 h), which was then transferred into the 5 L containers and shipped to Eawag, where they were immediately processed. The precipitation samples from PDM were also collected weekly, starting at the end of April 2015  (Table S2).

Environmental Science & Technology
Sample Processing. The total amount of collected weekly precipitation [P w ] was determined by weight (Mettler Toledo XS6002S). The samples for quantification of elements (from JFJ, PDM, and NABEL) were filtered (0.22 μm syringe filters, Nylon-66). To test the optimum stabilization conditions the samples were stabilized once in acidic (1% HNO 3 (nitric acid 65%, Suprapur, Merck, and once in alkaline conditions (0.5% TMAH, tetramethylammonium hydroxide 25%, Sigma-Aldrich). As similar results were obtained for both matrices, for Se, S, and Br, only results for the HNO 3 matrix are presented and discussed. Total concentrations of Se, Br, and S in these two different matrices are given in the SI Spreadsheet. However, I was only determined in TMAH solution. Samples for anion and trace element speciation were filtered (0.22 μm) and refrigerated (4°C) in completely filled PP vials and glass tubes with Teflon lids (5 mL), respectively. For DOC determination, rainwaters were filtered with 0.45 μm syringe filters (Nylon-66) and stored in glass bottles precleaned with ultrapure water. For carbon isotope determination, filtered (0.2 μm cellulose acetate, vacuum filtration) samples were filled in 10 mL glass vials with Teflon lids that were precleaned with hydrochloric acid (HCl, 37%, ROTIPURAN, Roth).
Total  (Table S1, SI Spreadsheet). Further (trace) elements were determined (Na, Mg, K, Ca, As, Si, B, Li, V, Cr, Fe, Cu, Zn, Sr, Mo, Cd, Pb, and U). More details are shown in SI Spreadsheet. Internal instrument drift was corrected by simultaneous detection of internal standards (50 ppb Sc, 5 ppb Lu, Rh, Th). Instrument performance was checked with reference standards (NIST 1643f and a Fluka multi-anion standard solution for S and Br), yielding recoveries of 101 ± 9% for Se, 105 ± 7% for Br, and 128 ± 23% for S. Volume weighted mean (VWM) concentrations were calculated from elemental concentrations (c i ) and precipitation volumes (V i ) for the analyzed time period as follows: Speciation Analyses with LC-ICP-QQQ. Selected precipitation samples from both sampling sites in the period from April to September 2016 were analyzed for Se, I, S, and Br speciation by advancing a method, which was initially developed to determine organic and inorganic Se species only. 68,69 The speciation analyses were performed with liquid chromatography (Metrohm 940 Professional IC Vario or HPLC Agilent 1200 with multidraw kit) coupled to ICP-QQQ. Chromatographic separation was performed using a mixed-mode column (HyperCarb 100 × 4.6 mm, 5 μm from Thermo Fisher) and an isocratic formic acid mobile phase (240 mM formic acid, formic acid 98−100%, pro analysis, Merck) in 1% MeOH (methanol, Chromanorm, VWR) at pH 2.3 (adjusted with 25% ammonia solution, Suprapur, Merck). Injection volumes of 800 to 1200 μL were used to achieve detectable peaks for the low concentrated rainwater samples ( Figure S4). For ICP-QQQ detection, O 2 -mode was used which yielded comparable LODs for Se as reported for the former method using H 2 68,70 with 0.08 nM for 78 SeO 3 2− and 78 SeO 4 2− vs 0.08 and 0.07 nM, respectively (SI). However, the advantage of measuring with O 2 as reaction gas is the determination of Se, S, Br, and I speciation in a single run. The LODs for the S, Br, and I species were as follows: respectively, 0.02 and 0.4 μM for methanesulfonic acid (MSA) and 32  Trajectory-Based Moisture Source and Loss Analysis. Air parcel backward trajectories were calculated using the 3D wind fields from two different atmospheric model data sets. 75,76 On a global scale 6-hourly data from the ERA-Interim reanalyses 77 of the European Centre for Medium-Range Weather Forecasts (ECMWF) were used on a regular horizontal grid of 1°resolution with 60 vertical levels. On the regional European scale the higher spatial and temporal resolution analysis data of the limited-area model COSMO 78,79 was chosen to obtain a more realistic representation of the European orography and the coastal regions. Hourly COSMO data are provided operationally by MeteoSwiss with 7 km horizontal grid spacing and 60 vertical levels. The two trajectory data sets served as a basis to identify evaporative moisture sources of precipitation at the two sites JFJ and PDM following a Lagrangian procedure. 80 For more information on the two atmospheric analysis data sets used, the trajectory calculation and the moisture source diagnostics see SI and Figures S1, S2, and S3.
From the air parcel trajectories and the moisture source information several characteristic quantities of atmospheric moisture cycling were derived: 1. Moisture source contributions from different regions: in particular, the share of continental vs marine moisture sources were calculated. For this, the gridded moisture uptakes as obtained from the moisture source , representing the middle 50% of the data, which fall between the upper quartile (75% data below that score) and the lower quartile (less than 25% below that score). The median is indicated as a line and crosses show the averages. The whiskers are based on 1.5 × IQR. The points depicted left of the boxplots show the corresponding data. The symbol sizes in the distance to moisture source plots is equivalent to the share of marine moisture, i.e., the larger the symbol, the higher percentage of marine moisture sources.

Environmental Science & Technology
Article diagnostics 80 (see also SI Figures S1 and S2) were integrated over marine and continental areas respectively by using the land-sea mask of the ERA-Interim data set. The global ERA-Interim data set was used for this analysis because it covers the whole North Atlantic, which can represent a large share of the moisture sources at the two sites in particular in winter. 2. Weighted mean distance to moisture source: distance of the center of mass of ERA-Interim moisture sources from the sampling site. 3. Cumulative regional evaporation to precipitation ratio: relates the sum of all COSMO trajectory-based specific humidity changes associated with surface evaporation to the sum of all specific humidity changes associated with precipitation. This ratio provides an estimate of the history of moisture uptake vs moisture loss in the 5 days prior to arrival at the sampling site.  Tables S3 and S4), which may be explained by the volume of rainfall, i.e., the rainfall dilution effect, 81 which leads to generally higher concentrations when rainfall volume is low, and vice versa. But when correcting for precipitation volume (Table S3, in addition to Spearman rank correlations, r ranging from 0.3 to 0.6), partial correlations between the elements generally still remained, however at varying strength. Therefore, it is not the rainfall volume alone that explains observed correlations. Total concentrations of Se, I, and S were higher in spring, summer, and autumn than in winter (see Figure 1, SI Figures S6 and S7, and Tables 1 and  S5). For Br, concentrations in winter are about two times higher than those measured in the other seasons.
Origin of Precipitation. Contributions of marine and continental moisture sources for JFJ, established by trajectory-based calculations and moisture source diagnostics, are indicated in Figures S8 and S9. During winter, the percentage of marine moisture (on average 75%; winter 2015/2016) is significantly higher than in the other seasons with on average 53%, 42%, and 49% marine moisture sources for spring 2016, summer and autumn (averages for 2015/2016), respectively. Consequently, JFJ receives higher amounts of moisture derived from the continent in summer, spring and autumn than in winter (see Figure 2, Figure S9). This finding agrees well with previous studies, which showed that land moisture sources are dominant from late spring to summer, when air is transported from the planetary boundary layer to the high mountain stations. 67 In winter, however, JFJ is mainly influenced by free tropospheric air transported over larger distances. 67,82−89 To obtain more detailed information on marine moisture sources, these were further separated into Atlantic and Mediterranean sources (see SI). During winter, the Alps received markedly larger amounts of moisture from the Atlantic Ocean than in any of the other seasons (75% in winter vs 48% for spring to autumn). Contributions of Mediterranean moisture sources to JFJ were on average 11%, with 6 and 7% on average in summer and winter and slightly higher shares in spring and autumn (16 and 12%) ( Figure S10, S11). These results in combination with lower total Se, I, and S concentrations in winter at JFJ suggest that for these elements land sources and potentially Mediterranean sources are important.
Concentrations of Dissolved Organic Carbon (DOC). To get further insights into biogenic versus anthropogenic sources, we determined DOC content in rainfall as well DOC-δ 13 C. Concentrations of DOC over the entire sampling period ranged from 28 to 208 μM (average 70 μM, see Table 1, S5) for JFJ. Similar to concentrations of Se, I, and S, DOC contents were higher in spring (average 61 μM), summer (average 85 μM) and autumn (average 77 μM) compared to winter (average 44 μM) see SI Spreadsheet, differences tested with Wilcoxon Signed Ranks Test (p < 0.05). Concentrations of Se, I, S, and, to a lesser extent, Br correlate positively with DOC concentrations (Table S3, SI Spreadsheet). Also other studies reported higher DOC concentrations in summer than for winter as well as higher DOC levels in rain with continental sources compared to rain with marine sources. 90−94 Correlations between Se, I, S, Br, and Sources of DOC. For JFJ, DOC-δ 13  The volume weighted means were calculated from the sum of concentration × precipitation volume per event/sum of precipitation volumes within the sampling period. Note that for PDM no winter data are available.

Environmental Science & Technology
Article differences with greatest depletion (most negative values) in winter. Average values follow the order winter (−31.3‰) < spring (−30.3‰) < autumn (−28.7‰) and = summer (−28.4‰) (see Figure S12). For CO 2 data, more depleted values of δ 13 C in winter compared to summer (on the order of 4−6‰) have previously been described for JFJ 95,96 and other continental northern midlatitude sites 97 and Beijing. 98 These trends have been explained by a larger contribution of fossil fuel combustion in winter (isotopically depleted values) and a larger biogenic source contribution (e.g., vegetation) in spring and summer (enriched values). A larger contribution of fossil fuel derived OC in winter was also observed for aerosols at JFJ. 99 Furthermore, DOC-δ 13 C is weakly correlated to the moisture source of precipitation (ρ = −0.4 for marine, ρ = 0.4 for continental and ρ = 0.3) for Mediterranean moisture sources (see Table S3). These results agree with a higher terrestrial moisture source contribution in summer, and potentially indicate a contribution of OC derived from vegetation transported via anabatic winds. 84 However, further detailed analyses are required to determine the exact sources of dissolved organic matter (DOM) in our samples as, e.g., Ceburnis et al. (2011) 100 state that 13 C determination alone, cannot completely distinguish fossil fuel sources from biogenic sources.
For the whole analyzed time period, I and Se correlate positively with DOC-δ 13 C ( Figure 2 and Table S3: ρ = 0.6, r = 0.4, both for I and Se) and the strength of these correlations exceeds the strength of positive correlations between DOCδ 13 C and continental moisture sources (Table S3). Also, correlations between DOC-δ 13 C and S and Br are present, although generally at a lower strength as for I and Se ( Figure  S13 and Table S3). These results may indicate that particularly I and Se are related to the source of DOM, and that these elements have a more dominant biogenic terrestrial source in summer, in addition to other land sources including anthropogenic sources. However, this hypothesis requires further study. In contrast, during wintertime, no correlations of Se, I, S, and Br with DOC-δ 13 C-values are observed but instead, strong correlations with seasalt sulfate (ss-SO 4 ) (e.g., ρ = 0.9 and r = 1 for I; r = 0.9 for Br) ( Table S3, SI Spreadsheet). This finding indicates a more dominant marine source of I and Br in winter, agreeing with higher contributions of Atlantic moisture sources.
Correlations of Se, I, Br, and S to Chlorophyll a Exposure. To investigate a potential link between precipitation chemistry and emissions of volatile organic compounds of Se, I, Br, and S from bioproductive surface water bodies (marine environments and lakes), we tested if there was a correlation between chlorophyll a exposure at the moisture source and Se, I, Br, and S concentrations in precipitation. For JFJ, we found that chlorophyll a exposure at the moisture source correlates positively with continental moisture sources (ρ = 0.6, r = 0.5) and the fraction of moisture uptake over lakes (ρ = 0.4, r = 0.6, SI Spreadsheet). However, no significant correlations are present between chlorophyll a exposure at the moisture source and Se, I, S, and Br, indicating that biogenic emissions of these elements from surface waters do not represent a dominant source of these elements in precipitation at JFJ.
Precipitation Chemistry and Moisture Source Data for Pic du Midi (PDM). Concentrations of Se, I, Br, and S and Origin of Precipitation. As compared to JFJ, the Pic du Midi (PDM) site receives more precipitation with marine moisture. In general, PDM receives larger amounts of Atlantic moisture than JFJ as is expected because PDM is closer to the North Atlantic, i.e., the shortest direct distance is 150 km for PDM compared to 625 km for JFJ. At PDM two different samplers were used to analyze bulk deposition and wet deposition only. However, at JFJ rainfall data was only obtained using a bulk sampler (same type as used at PDM). Therefore, for PDM only precipitation data derived from the bulk sampler are discussed, to enable a comparison between the two sites. In contrast to JFJ, for PDM no data are available for winter and only limited data for autumn. Total concentrations of Se, I, S, and Br over the whole sampling period ranged from 0. 16 Figures S6 and S7), it can be seen that concentrations of these elements are generally higher in rainfall at PDM than at JFJ. These differences are likely not only explained by the difference in elevation as very low concentrations were also found at lower altitude stations in Switzerland (see Table 1, S5). During spring, precipitation with Atlantic moisture sources was always larger than 56% (56−71%), whereas during summer, this value ranged from 18 to 79% and was thus more variable (F-test, p < 0.05). Similar to JFJ, the contribution of precipitation with continental moisture sources were highest in summer, i.e., 21 to 69% (on average 44%) (see SI Spreadsheet, Figure S8).
Concentrations of DOC and Sources of DOC in Rainfall (DOC-δ 13 C Values). As for Se, I, S, and Br, also DOC concentrations at PDM are significantly higher than at JFJ (Median and Mann−Whitney test, p < 0.05). For samples with  (Table S3). The symbol size indicates the volume of collected precipitation. Corresponding figures for PDM are in Figure S14. higher contributions of rainfall from marine origin, lower average DOC contents were found (73 μM for samples with an oceanic moisture contribution >50% vs 129 μM for the other samples) (t-test and Mann−Whitney test for medians). Therefore, it is likely that the highest levels of DOC result from land sources. Compared to JFJ, DOC-δ 13 C-values from PDM span a smaller range, from −20.1 to −29.9‰ (average SD ± 0.7‰) for the complete sampling period in 2015 and 2016. There were no significant differences in average DOCδ 13 C-values for spring and summer and when DOC-δ 13 C data for the same weeks in summer for PDM and JFJ are compared, there is no statistically significant difference between both sites. Interestingly, at PDM, highest DOC-δ 13 C-values were measured when Atlantic contributions to rainfall were lowest (in summer) but also when Atlantic contributions were highest (in spring and early summer) ( Figure S14). Furthermore, for continental moisture source contributions larger than 40% (summer), DOC-δ 13 C values correlate positively to continental and Mediterranean moisture sources (ρ = 0.7, r = 0.7 for both) and negatively with Atlantic moisture sources (ρ = −0.9, r = −0.7) (see Table S4) as was also observed for JFJ, and which could be interpreted again by higher contributions of DOC derived from vegetation. In contrast, higher DOC-δ 13 C values in spring and early summer associated with higher contribu-tions of Atlantic moisture sources could point to contributions of marine biogenic DOC sources as described, e.g., in Avery et al. 90 and Gromov et al. 90,101 However, on average the values of DOC-δ 13 C-values for PDM, are lower than reported average values for marine biogenic DOC emissions i.e., −22‰ 101,102 or −18‰ to −20‰, 90 which is a further indication for mixed continental and marine DOC sources at PDM.

Environmental Science & Technology
Correlations between Se, I, S, Br, and Sources of DOC. In contrast to JFJ, there are no correlations between total concentrations of Se, S, Br, DOC-δ 13 C levels, and origin of precipitation for PDM (see Table S4). For I, there is a moderately strong correlation with the continental share of precipitation moisture sources (ρ = 0.4), which is more pronounced for periods when continental moisture sources exceeded 40% (mainly in summer) (ρ = 0.8, r = 0.7). For these events, I concentrations also correlate positively with DOCδ 13 C levels (ρ = 0.8, r = 0.8) which may indicate a biogenic land source in summer, however, again, this needs further investigation. Further information on precipitation chemistry can be found in the SI.
Correlations of Se, I, Br, and S to Chlorophyll a Exposure. For PDM, chlorophyll a exposure at the moisture source is positively correlated to marine (ρ = 0.5) and Atlantic moisture origin (ρ = 0.4) and negatively correlated to continentally

Environmental Science & Technology
Article derived moisture (ρ = −0.5) (see Table S4). This indicates that predominantly air masses derived from the marine environment have traveled over chlorophyll a enriched marine surface waters. Furthermore, positive correlations are found between chlorophyll a exposure and S (ρ = 0.5, r = 0.5) and Se (r = 0.6), which could indicate a contribution of a marine biogenic source for these elements at PDM. In the bioproductive period (April−September 2016), the correlation with chlorophyll a exposure holds for Se (r = 0.7) but not for S. This finding might be related to the so-called summerparadox, i.e., the time lag between maxima in marine phytoplankton biomass and DMS emissions, which has been reported for tropical to polar latitudes, 103,104 however, this needs further investigation. In addition, also relatively strong correlations for chlorophyll a and ss-SO 4 (r = 0.6, r = 0.8, respectively) are found in the bioproductive period, which may generally indicate marine inputs.
Speciation. S and Se Speciation. At both JFJ and PDM, only inorganic species of Se, I, S, and Br were detected, except for MSA, an oxidation product of DMS Figure S15. Figure 3 and Figure S15 indicate determined species as well as total concentrations. As expected, the majority of S (total S concentration) is present as sulfate (SO 4 2− ), ranging from 94.1 to 99.6% (8.4−854 μg S). MSA is present in amounts of 0.5 to 5.9% (0.1−12.6 μg S). For JFJ, we calculated an average contribution of ss-SO 4 and non sea-salt sulfate (nss-SO 4 ) of 41 and 59%, which is not much different than the values for PDM, i.e., with on average 50% ss-SO 4 and nss-SO 4 contributions (SI Spreadsheet). Seasonality of S speciation (no data for winter) was most pronounced for MSA, having contributions to total S at JFJ of 0.9−5.4% for summer >0.5−4.2% for spring >0.7− 1.4% for autumn reflecting the relation between MSA and biogenic DMS productivity. However, as also SO 4 2− is an important product of DMS oxidation, formed in a different reaction pathway than MSA, which is formed via both gas-and aqueous-phase reactions, 64 the concentrations of MSA cannot be directly linked to DMS fluxes. Previously, the ratio between 100 × MSA and nss-SO 4 was used as a proxy for marine air masses. This ratio was found to be higher in marine air (3.3 ± 2.4) than in continental air (1.8 ± 0.8), 105 largely due to the relatively lower nss-SO 4 levels in marine air masses. 105 This trend is also observed in our data as marine moisture sources correlate positively to the 100 × MSA/nss-SO 4 ratio both for JFJ (ρ = 0.5) and PDM (ρ = 0.7) (SI Spreadsheet).
As was also reported in previous studies e.g., 54 Se is mainly present in its most oxidized form, SeO 4 2− with on average contributions of 65 ± 12% at JFJ and 84 ± 8% at PDM. The concentrations range from < LOD (0.07 nM) to 0.42 nM (48−82%, 0.15 ± 0.1 nM) at JFJ and from 0.14 to 0.85 nM (66−94%; 0.41 ± 0.19 nM) for PDM (see Figure 3, SI Spreadsheet). The concentration of the more reduced species, SeO 3 2− is < LOD (0.08 nM) to 0.15 nM for JFJ and < LOD to 0.19 nM LOD for PDM. Generally, the fraction of SeO 3 2− relative to SeO 4 2− is higher for JFJ samples (18−52%) than for PDM (6−34%). However, the concentration of SeO 4 2− correlates positively to the continental share of moisture sources of precipitation (JFJ: ρ = 0.5, PDM: ρ = 0.6) (see Figure S16, SI Spreadsheet), which indicates that the more continental location of JFJ does likely not explain the larger share of SeO 3 2− at JFJ than at PDM. The predominance of inorganic anionic species indicates that any small organic Se compounds present have been oxidized in the atmosphere, which can be expected considering the short lifetime of DMSe. 23 However, for many samples, the total amount of quantified species is lower than the total Se concentrations and therefore missing fractions occur. For Se unidentified fractions make up to 59% (of total Se) for JFJ and 65% for PDM (for wet-only sampler even up to 80%). These unidentified fractions may be explained by larger organic molecules, which could not be detected using our method. Therefore, further research is required to elucidate the nature of these unidentified species.
I and Br Speciation. Iodide (I − ) is the main species of I that was found, with concentrations ranging from 0.6 to 6.9 nM (78−97%; 2.7 ± 1.8 nM) at JFJ and from 1.6 to 8.2 nM (56− 88%; 4.2 ± 2.1 nM) for PDM. Iodate (IO 3 − ) concentrations range from 0.1 to 0.7 nM (2−21%, 0.4 ± 0.2 nM) for JFJ and between 0.3 and 3.9 nM (11−44%, 1.8 ± 1.4 nM) for PDM (see Figure S15), indicating that the share of IO 3 − is higher at PDM than at JFJ in the analyzed time period (April− September). These results are in line with previous findings indicating generally higher fractions of IO 3 − in marine-derived precipitation compared to continental precipitation, 12,20,37,106 however, we did not find any significant trend between the moisture sources of precipitation and I speciation, neither for PDM nor for JFJ ( Figure S16). The higher share of I − may be explained by more reduced forms of I, e.g., CH 3 I, derived from local sources, which undergo photolysis and oxidation in the gas-phase, producing HOI and OI, among others. These species are expected to be taken up in the aqueous phase, where they can be transformed to I − via reaction with sulfite. 17 Another explanation for the lower share of IO 3 − at JFJ is reduction of IO 3 − in the presence of DOM, i.e., I − formation was demonstrated to occur via reduction of IO 3 − in the presence of DOM at low pH. 17,107 Therefore, even small changes in pH and DOM content and type may lead to differences in I − to IO 3 − ratios between different locations. Apart from I − and IO 3 − , there were also variable fractions of unidentified I. For PDM the share of missing species ranged from −4 to 63% (average 34 ± 22%) and for JFJ this fraction ranged from −47 to 57% (0 ± 37%). Negative values are explained by a higher sum of species than totals for 7 out of 13 events, which is likely explained by the very low I concentrations leading to higher uncertainties. The fraction of unidentified I likely indicates the occurrence of species which could not be resolved by our method, e.g., complexed organics or aerosols as has been suggested previously. 12,20,37,106,108,109 In contrast to I, Br is only present as bromide (Br − ) and bromate (BrO 3 − ) concentrations are below the LOD (1.1 nM) ( Figure S15, SI Spreadsheet). As total Br levels are generally higher for PDM as for JFJ, also Br − levels are higher, i.e., on average 41.6 ± 16.3 nM vs 12.7 ± 6.2 nM, respectively. Dissolved total Br in rainwaters from Plynlimon (Wales) was also found to be present as Br − , 110 however, concentrations were much higher (on average 163 nM) due to direct marine influence. The absence of BrO 3 − , in contrast of IO 3 − , is explained by the fact that BrO 3 − formation is a slow process and also not achievable by all main atmospheric oxidants. 21 Similar to Se, I, and S, also for Br, fractions of missing species are found, i.e., up to 53% for JFJ (mean 21%) and up to 63% for PDM (mean 34%). Again, these unidentified fractions may be explained by the presence of organic species, as was previously suggested. 37 Marine versus Terrestrial Sources. The relationships between elemental concentrations in combination with

Environmental Science & Technology
Article information on the origin of precipitation, indicates that in winter when marine moisture source contributions are highest, concentrations of I, Se, and S at JFJ are lowest. It is likely that the concentrations in winter are low due to a lack of continental sources but also due to a potential rainfall dilution effect during more long-range atmospheric transport. In contrast, Br concentrations are higher in winter than in other seasons. Furthermore, Br concentrations correlate strongly with ss-SO 4 (see Table S3). This may indicate transport of Br via submicrometer sea-salt aerosols, as has been reported previously. 111 Sources of S and Se in winter cannot be resolved but these elements may also have been derived from marine (submicron) aerosols (strong correlation to I and Br, which may indicate similar transport pathways) and/or continental emissions (correlation with DOC in winter time, e.g., DOC derived from fossil fuel emissions). In the bioproductive period, we find indications that Se, I, S, and Br have terrestrial sources (e.g., positive correlations with nss-SO 4 , see Table S3) that are related to a biogenic source of DOC-δ 13 C, in addition to marine contributions for I and Br (positive correlations with ss-SO 4 , see Table S3). Similar to I and Se, terrestrial methyl bromide (CH 3 Br) emissions, e.g., from coastal ecosystems, organic-rich soils, 112 vegetation, 33,113,114 and coastal wetlands 33 were found to be important. Butler (2000) suggested that terrestrial sources could balance atmospheric Br budgets, for which 40% of the sources are missing. 112 Also, the ratio of Br/I concentrations is generally higher in the JFJ samples than in the PDM samples (Table 1), which may imply that relatively more Br is derived from terrestrial sources.
At PDM, continental trace element contributions are less pronounced due to the generally higher contribution of marine sources. However, when continental moisture source con-tribution was high (>40%), we found indications of terrestrial sources of I, linked to a biogenic source of DOC-δ 13 C as well as nss-SO 4 correlations with I, Se, and S and DOC-δ 13 C (see Table S4).
In conclusion, our data suggest mixed marine and continental source contributions for I, Se, Br, and S at both sites, however, there are clear seasonal differences, illustrated in Figure 4. Higher shares of continental moisture sources in spring, summer, and autumn, and higher concentrations of I and Se at JFJ in these seasons suggests a more local cycling, i.e., terrestrial emissions, atmospheric transport, and deposition, of these elements. The link between concentrations of I, Se, Br, and S and the source of dissolved organic matter (i.e., correlations with DOC-δ 13 C values), especially in summertime and when continental moisture sources dominate, potentially indicates biogenic land emissions, although this needs further investigation. As emissions from terrestrial environments are still largely unknown due to the heterogeneous nature of terrestrial landscapes, more work needs to be done in measuring and understanding these trace element emissions from different types of environments. When this information becomes available, biogeochemical models, aimed at predicting the cycling and distribution of I and Se, can be refined, enabling more accurate predictions of environmental sources and sinks of I and Se, and especially their inputs into agricultural soils and terrestrial food chains.