River-derived humic substances as iron chelators in seawater.

The speciation of iron(III) in oxic seawater is dominated by its hydrolysis and sedimentation of insoluble iron(III)-oxyhydroxide. As a consequence, many oceanic areas have very low iron levels in surface seawater which leads to iron deficiency since phytoplankton require iron as a micronutrient in order to grow. Fortunately, iron solubility is not truly as low as Fe(III) solubility measurements in inorganic seawater would suggest, since oceanic waters contain organic molecules which tend to bind the iron and keep it in solution. Various iron-binding organic ligands which combine to stabilize dissolved iron have been detected and thoroughly investigated in recent years. However, the role of iron-binding ligands from terrestrial sources remains poorly constrained. Blackwater rivers supply large amounts of natural organic material (NOM) to the ocean. This NOM (which consists mainly of vascular plant-derived humic substances) is able to greatly enhance iron bioavailability in estuaries and coastal regions, however, breakdown processes lead to a rapid decrease of river-derived NOM concentrations with increasing distance from land. It has therefore been argued that the influence of river-derived NOM on iron biogeochemistry in offshore seawater does not seem to be significant. Here we used a standard method based on 59Fe as a radiotracer to study the solubility of Fe(III)-oxyhydroxide in seawater in the presence of riverine NOM. We aimed to address the question how effective is freshwater NOM as an iron chelator under open ocean conditions where the concentration of land-derived organic material is about 3 orders of magnitude smaller than in coastal regions, and does this iron chelating ability vary between NOM from different sources and between different size fractions of the river-borne NOM. Our results show that the investigated NOM fractions were able to substantially enhance Fe(III)-oxyhydroxide solubility in seawater at concentrations of the NOM ≥ 5 μg L- 1. Terrigenous NOM concentrations ≥ 5 μg L- 1 are in no way unusual in open ocean surface waters especially of the Arctic and the North Atlantic Oceans. River-derived humic substances could therefore play a greater role as iron carriers in the ocean than previously thought.

Introduction

Iron limitation restricts photosynthesis and nitrogen fixation in extended regions of the world ocean (Hunter and Boyd, 2007; Moore et al., 2009; Kendall et al., 2012). Complexation reactions with Fe(III)-binding organic ligands are critical in regulating iron solubility and potentially bioavailability (Rose and Waite, 2003). Consequently, these natural organic molecules exert a major influence on the global carbon cycle and play an important role in regulating climate (Sarmiento et al., 2010). Many different Fe-binding ligands are produced by marine biota (Gledhill and Buck, 2012). Recent work by Laglera and van den Berg (2009) has shown that some sorts of humic substances may be of great importance as iron(III) carriers in coastal waters (land-derived humic substances) and in the deep ocean, respectively (autochthonously produced marine humic substances).

Land-derived natural organic matter (NOM), which consists to a large proportion of relatively recalcitrant humic substances, is continuously discharged by Arctic rivers and other rivers into the ocean (~ 0.4 Pg of carbon per year) (Opsahl and Benner, 1997; Hernes and Benner, 2006). The river-borne NOM contains humic substances with functional groups that make them capable of binding iron and keeping the iron dissolved (Guillon et al., 2001; van Schaik et al., 2008; Karlsson and Persson, 2010; Jones et al, 2011; Krachler et al., 2012; Stolpe et al., 2013; Lesher et al., 2013). Iron–humic complexes might be unavailable for uptake by phytoplankton, however, several mechanisms have been proposed whereby the humic-bound iron can be converted into more labile forms, for example, photochemical reactions or ligand exchange with marine siderophores (Nieto-Cid et al., 2006; Batchelli et al., 2010; Kuhn et al., 2014).

The transport of NOM-bound iron across the estuarine mixing zone is severely impeded by flocculation and sedimentation (Sholkovitz et al., 1978; Dai and Martin, 1995; Nowostawska et al., 2008). The estuarine removal processes prompt the conclusion that terrigenous humic ligands might be of little importance for iron complexation in the ocean (Laglera and van den Berg, 2009). On the other hand, several studies have highlighted the role of iron-binding NOM, as well as copper-binding NOM, in the coastal plumes of humic-rich rivers (Buck et al., 2007; Batchelli et al., 2010; Muller and Batchelli, 2013). Laglera and van den Berg (2009) and Laglera et al. (2011) could show that land-derived humic substances play an important role as an iron chelating ligand class in the Irish Sea. Yoshimura et al. (2010) reported that the Amur River plume which is transported by the east Sakhalin current is a major source of bioavailable iron to the Okhotsk Sea. Recently, Klunder et al. (2012) measured dissolved iron in the Arctic shelf seas and surface waters of the central Arctic Ocean and found evidence for impact of Arctic river water. Thus a significant flux of iron-loaded NOM from land to the sea appears possible.

However, breakdown processes of riverine NOM in the marine environment through microbial oxidation and bleaching by solar radiation (Powell and Wilson-Finelli, 2003; Fichot and Benner, 2014; Fichot et al., 2014) lead to a rapid decrease of NOM concentrations with increasing distance from land (Opsahl and Benner, 1997; Obernosterer and Herndl, 2000; Laglera and van den Berg, 2009). The concentration level of land-derived NOM in the open ocean has been estimated by Benner and co-workers (Benner et al., 2005) who used lignin phenols as biomarkers. Their results revealed that terrigenous DOC (dissolved organic carbon) is present throughout the Atlantic and Pacific Oceans in the range of 0.7–2.4% of bulk DOC. However, because there is no sufficiently precise knowledge on the iron-binding properties of highly diluted terrigenous NOM in seawater, it is not clear whether or not this land-derived NOM may play an important role for the iron nutritional status of the Atlantic and Pacific Oceans.

In order to answer this question, we studied the influence of low concentrations of terrigenous NOM on iron solubility in oxygenated seawater. We used Field-Flow Fractionation for precise determination of molar masses of NOM fractions, combined with iron solubility measurements using 59Fe as a radiotracer. As natural seawater contains a variety of siderophores, organic exudates, and autochthonous humic materials which combine to stabilize dissolved iron (Hopkinson and Morel, 2009), we used artificial seawater made of ultrapure chemicals to see the influence of land-derived NOM alone.

Materials and methods

Samples

Peatland-influenced water samples were collected from the Craggie Burn, a tributary of River Halladale in the Flow Country of Caithness and Sutherland in North Scotland which is containing the largest expanse of blanket bogs in Europe (Muller and Tankéré-Muller, 2012) (geographic coordinates: N 58°26′ W 3°54′), and from the Tannermoor brook in Upper Austria (geographic coordinates: N 48°30′ E 14°52′). Sampling for experiments was performed during July and August 2011–2014. The sampling locations have been described in detail previously (Krachler et al., 2005, 2010, 2012). Anthropogenic contamination of the sampling sites is regarded as minimal, as the creeks drain unspoiled peatlands and are themselves free-flowing and unpolluted. Water samples were drawn from the surface using a HNO3 cleaned polyethylene bottle that had been rinsed with ultrapure water. Immediately following collection, we pumped the water samples through 0.2 μm filters (Sartobran 300 Capsules) to remove particulate material, eukaryotes and bacteria, and the filtrates were rapidly filled into acid-cleaned sterile polyethylene bottles to prevent microbial degradation of natural organic Fe chelators in the filtered samples. The samples were stored in the dark at 4 °C until further treatment.

Chemical analyses

We measured concentrations of DOC by a high-temperature oxidation method using a TOC-V total organic carbon analyzer, Shimadzu, with potassium hydrogen phthalate as internal standard. For pH measurements in seawater, we used an iodine/iodate electrode from SCHOTT Instruments (IL Micro pHT-A). The electrode was calibrated using standard buffer solutions to DIN 19 266 (SI Analytics GmbH, Mainz, Germany). pH was thus determined on the NBS scale. The Fe concentrations of the samples were analyzed by ICP-MS (Agilent Technologies 7700 ×, Waldbronn, Germany). We monitored 56Fe, using He as a collision cell gas to account for polyatomic interferences. In addition to Fe, the following trace metals were monitored: 27Al, 47Ti, 55Mn, 63Cu, 64Zn, 75As, 139La, 140Ce and 208Pb. The results revealed that the concentrations of these competing metals in the NOM samples were significantly lower than the concentration of Fe. Iron was in parallel detected in the samples by means of graphite furnace atomic absorption spectrometry (GF-AAS) using a PinAAcle 900Z (PerkinElmer) after microwave digestion. Anion concentrations (sulfate and chloride) were analyzed by Ion Chromatography (DIONEX ICS-1000).

Synthetic seawater

We preferred synthetic seawater over UV-digested natural seawater. During UV digestion for the breakdown of metal complexing organic ligands in natural seawater, acidification is necessary and sample temperatures are relatively high (80–90 °C) (Achterberg et al., 2001). CO2 would escape from the seawater and it would be difficult to re-establish the natural carbonate system and Ca2 + concentration. Therefore, artificial seawater was prepared which contained the major components of sea salt as proposed by Kester et al. (1967). Due to their high concentrations in seawater, Mg2 + and Ca2 + are important competitors reducing the Fe(III) complexation of organic ligands (Hiemstra and van Riemsdijk, 2006). Takahashi et al. (2014) calculated the degree of CaCO3 saturation for surface ocean waters and found that tropical and subtropical waters are supersaturated by a factor of 4.2 with respect to aragonite and 6.3 for calcite, whereas the subpolar and polar waters are supersaturated by 1.2 for aragonite and 2.0 for calcite. In order to create an environment similar to that of the surface ocean, we prepared CaCO3 supersaturated (by a factor of ~ 2), oxygenated inorganic seawater with pH = 8.2, using ultrapure water and high-purity chemicals. The artificial seawater (salinity S = 35) was made under clean room conditions. Ultrapure mineral acids were prepared by double sub-boiling distillation of 65% nitric acid and sub-boiling distillation of 37% hydrochloric acid of p.a. grade (Merck KGaA, Darmstadt) respectively, using a duo PUR quartz sub-boiling unit (MLS Lab Systems GmbH, Leutkirch). Ultrapure water was obtained by sub-boiling distillation of purified water (18.2 MΩ·cm) using an ultra-clear system (SG water GmbH, Barsbüttel). Sodium hydroxide monohydrate (Suprapur, Merck) and purified water were used to prepare the 0.5 M NaOH solution. The following salts were used: NaCl Sigma Aldrich 38979, trace select; Na2SO4 Sigma Aldrich 204447, ≥ 99.99%; KCl Sigma Aldrich 05257, trace select; NaHCO3 Sigma Aldrich 31437, p.a. > 99.7%; KBr Sigma Aldrich 90737, trace select; H3BO3 Sigma Aldrich 202878, 99.999%; NaF Sigma Aldrich 450022, 99.99%; MgCl2·6 H2O Alfa Aesar 10797, 99.999%; SrCl2·6 H2O Sigma Aldrich 204463, 99.995%; Na2CO3 Fluka 71347, trace select; and CaCl2 anhydrous Sigma Aldrich 499609, 99.99%. The synthetic seawater was stored in a refrigerator at 4 °C. Prior to the experiments, it was filtered using acid-washed 0.2 μm Sartobran 300 Capsules in order to remove possible iron contaminations. By measuring Fe(III)-oxyhydroxide solubility in our synthetic seawater (see Fig. 1) and comparing our results with literature data we could show that Fe complexing organic ligands were absent. DOC concentrations in our synthetic seawater and in the ultrapure water, respectively, were < 8 μmol L− 1.

Fig. 1

Comparison of the solubility of Fe(III) (mol L− 1) as a function of pH in synthetic seawater (present study, blue squares) and in 0.7 M NaCl (data points from Liu and Millero, 1999) at 25 °C. Data points of the present study represent the average of three measurements that agreed to ± 15%.

Gel chromatography

Gel chromatography on Sephadex LH-20 can be used for isolating representative fractions of riverine NOM with high iron and lignin phenol contents. The general procedure for isolating these fractions has been described elsewhere (Krachler et al., 2012). We used an ÄKTA purifier with conductivity monitor. A UV detector set to 242 nm was used to detect the NOM. Eluent was 4.5% methanol (HPLC grade from Fisher Scientific UK) dissolved in ultrapure water. Flow rate was 0.1 mL min− 1. We used two series-connected SR 25/100 columns (tube height 1000 mm) from GE Healthcare. For sample preparation, the filtered creek water was concentrated under vacuum in a rotary evaporator (Rotavapor R-210, Büchi) and dried. As a reference material, we used Suwannee River NOM (RO isolation, 1R101N, from the International Humic Substances Society). 10 mg portions of the solid NOM were shaken with 10 mL synthetic seawater for 30 min, and then stored in a refrigerator at 4 °C for at least one week. After storage, the suspension was filtered through a syringe filter (Acrodisc 25 mm, 00.2 μm GHP membrane, PALL). Iron and DOC were measured in the filtrate. Subsequently, the filtrate was separated on a Sephadex LH-20 column. Sample volume was 2 mL. After separation, 10 mL fractions were collected using a Frac-900 fraction collector. In order to determine what fractions of the initial DOC in the original samples were recovered in these NOM fractions, the 10 mL fractions were dried at room temperature and the dried samples were re-dissolved in ultrapure water. Iron and DOC concentrations were measured in the solutions.

Field-Flow Fractionation

Field-Flow Fractionation (FFF) provides a continuous and high-resolution separation of nanoparticles as a function of their diffusion coefficients. When coupled to various detectors, FFF yields a wealth of information on particle properties including size and chemical composition (Baalousha et al., 2011). For Asymmetric Flow Field-Flow Fractionation (FlowFFF) coupled to UV–Vis detection and Inductively Coupled Plasma Mass Spectrometry (ICP-MS) we used a method that has been described previously (Neubauer et al., 2013a). In brief, the separation took place in a thin (0.75 mm in this instance) channel with a flow field applied perpendicular to the main parabolic flow of a mobile phase. An ultraviolet/visible diode array detector (UV-DAD, Agilent Technologies 1200 series, Waldbronn, Germany) with the absorption wavelength set to 254 nm was used to detect NOM. An ICP-MS (Agilent Technologies 7700 ×, Waldbronn, Germany) was coupled to the outlet of the UV-DAD detector and the flow split before entering the ICP-MS, reducing the flow rate from 1 mL min− 1 to 0.35 mL min− 1. The ICP-MS was calibrated using standards between 5 and 100 μg L− 1 Fe. We used 15 mM ammonium carbonate/ammonium carbamate, adjusted to pH 7 with HNO3 (ultrapure), as the mobile phase for the samples. All injections (100 μL for the samples) were performed with an autosampler (Agilent Technologies 1200 series, Waldbronn, Germany). Recovery runs were performed by injecting 10/20 μL without applying any cross flow (injection with elution). The fractograms were integrated using the OriginPro 7.5 software (OriginLab Corporation, Northampton, USA) to obtain the peak areas, and recoveries were calculated. Measurements were done in triplicate. The standard deviation of the UV–Vis signal and the Fe signal was < 5% for the triplicate measurements. Calibration with polystyrene sulfonate molar mass standards (1100 g mol− 1; 3610 g mol− 1; 10,600 g mol− 1) was performed at higher ionic strength of the mobile phase (50 mM) (Neubauer et al., 2011).

Radiotracer technique

Radiotracer techniques based on 59Fe and in-situ preparation of a Fe(III)-oxyhydroxide solid phase have been employed by Kuma et al. (1996), Liu and Millero (1999; 2002), and Heller et al. (2013) to determine the solubility of iron(III)-oxyhydroxide in NaCl solutions and in seawater. Following a related methodology, Krachler et al. (2010) have investigated the iron carrying capacity of river water upon mixing with seawater.

Here, we use the 59Fe technique originally developed by Kuma et al. (1996) and Liu and Millero (1999, 2002) in a slightly modified version for measuring the solubility of iron(III)-oxyhydroxide in synthetic seawater in the presence of NOM.

After separation on Sephadex LH-20, the NOM fractions were dried at room temperature using a Savant ISS110 SpeedVac Concentrator from Thermo Scientific. For preparation of the stock solutions, weighed quantities of the solid NOM were dissolved in synthetic seawater. DOC and Fe were measured in the stock solutions. The stock solutions were diluted with synthetic seawater in a serial dilution process in order to generate NOM samples for the radiotracer experiments. For these experiments, clean 50 mL PE tubes were used. Each tube was filled with the amount of artificial seawater needed to add up to 40 mL together with the other solutions. 10 μL of the NOM sample and 10 μL of a FeCl3 solution in 0.5 M HCl were added. FeCl3 solutions in appropriate concentrations were prepared by dilution of a stock solution with 0.5 M HCl. The total amount of FeCl3 added to the sample was in all cases sufficiently high to allow precipitation of Fe(III)-oxyhydroxide. Then 10 μL of a 59Fe radiotracer (59FeCl3 in 0.5 M HCl NEZ037 from PerkinElmer) with a half-life of 44.6 days was added. Finally, NaOH (0.5 M) was added in small increments in order to adjust the pH to 8.2 or to another desired value. The total amount of iron (WFe) in 40 mL (6–10 ng) was consisting of the contributions from the synthetic seawater (< 0.1 ng), the FeCl3 solution, the tracer solution (< 0.1 ng), and the NOM sample. A reference sample was prepared with 39.99 mL 0.5 M HCl (ultrapure) plus 10 μL tracer solution. The tubes with the air-saturated samples were sealed gastight and vigorously shaken in a dark room at controlled temperature (± 2 °C) for a time span of 1–2 weeks to allow chemical and isotopic equilibrium to be established. Due to the low amounts of iron used in these experiments (6–10 ng), and the low NOM concentrations, co-precipitation of the NOM with Fe(III)-oxyhydroxide or adsorption of the NOM by Fe(III)-oxyhydroxide particles can be assumed negligible (Hiemstra and van Riemsdijk, 2006).

The separation of the solid phase (Fe(III)-oxyhydroxide) from the liquid phase was performed by centrifugation using a Hermle Table Top Centrifuge Z 326 K (4 h at 4000 rpm, RCF value 2665 ×g). Control experiments using filtration (0.025 μm cut-off) instead of centrifugation have been performed in our earlier work. The results from the experiments using filtration were in good accordance with the results from centrifugation experiments (Krachler et al., 2010). Using centrifugation made it possible to treat a number of samples simultaneously. A soft stop of the rotor after centrifugation and rapid removing of the aliquots (5 mL) was necessary to avoid resuspension of the solid phase. The 5 mL aliquots were pipetted into clean polyethylene tubes of identical geometry, acidified with 10 μL 0.5 M HCl, and measured by γ-spectrometry (1099 keV and 1292 keV) using a Ge detector GX8021/S (Canberra™). Counting times were between 10 and 300 min in order to achieve low counting uncertainties of about 3%. All radiotracer experiments were performed in triplicate. Total dissolved Fe concentrations, [FeT], were calculated using the following equationwhere cpm(sample) are the net counts per minute of the 59Fe activity of a 5 mL aliquot of the final solution, cpm(HCl) are the net counts per minute of the 59Fe activity of a 5 mL aliquot of the reference sample (where no precipitation takes place), and WFe [g] is the total amount of iron (solid and dissolved) present in the 40 mL sample. The accuracy of the measurements of [FeT] is ± 15%. We assumed reduced precision in [FeT] applying Gaussian quadratic error propagation for the numerous steps of the laboratory procedure that contributed to the uncertainty, e.g., serial dilution steps.

All sample manipulations were performed in a clean room laboratory (EN ISO 14644; laminar flow ISO 8, metal-free class 100 workbenches). As a reference material, Suwannee River NOM (RO isolation, 1R101N, from the International Humic Substances Society) was investigated, using the same technique.

Data analysis

The stability constants determined here are “conditional stability constants” based on molar concentrations. Conditional stability constants as defined e.g., by Rue and Bruland (1995) have been widely used for determining dissolved iron speciation in seawater. In contrast to true thermodynamic stability constants, which are based on activities, conditional stability constants are valid only for the pH and salt composition of seawater and only for a limited range of ligand concentrations.

For determination of conditional stability constants, we assumed that the iron was present as Fe(III). For the Fe(III)–L complex, 1:1 stoichiometry (1 Fe atom per 1 ligand molecule) was assumed according to Rue and Bruland (1995). Due to the very efficient separation method applied here, we were able to assume that iron-binding ligands comprised nearly the total NOM in our samples (see Fig. 2). X-ray absorption fine structure studies by van Schaik et al. (2008) have shown that mononuclear Fe(III)–fulvic acid complexes dominate in solution, whereas bi- and poly-nuclear Fe(III)–fulvic acid complexes tend to be water insoluble. The conditional stability constants of the complexes have therefore been calculated by using the molar concentrations of total NOM assuming mononuclear complexes with 1:1 stoichiometry, whereas the electrochemical methods in wide use calculate log K values (and corresponding ligand concentrations) for individual operationally defined “binding sites”.

Fig. 2

Asymmetric Flow Field-Flow Fractionation experiments. The FFF chromatograms show the molar mass distribution of Fe and NOM (measured as UV–Vis absorption at 254 nm) in seawater-soluble NOM fractions which have been collected after separation on a Sephadex LH-20 column. Results are shown for the three Craggie Burn fractions (CB_0, CB_1, and CB_2) and for the three Suwannee River fractions (SR_0, SR_1, and SR_2): (a) CB_0; (b) CB_1; (c) CB_2; (d) SR_0; (e) SR_1; (f) SR_2. The UV signal (solid red curved line) represents the NOM and is superimposed with the iron signal (solid blue curved line). The molar mass of the Fe-NOM can be obtained from the peak maximum of the Fe signal. The resulting molar masses are listed in Table 2.

We have the following mass balance equations:

[FeT] represents the total dissolved iron concentration in the sample which can be determined experimentally by the radiotracer method. [FeL] is the concentration of the undissociated Fe(III)–L complex. [L] denotes the free ligand concentration. [LT] is the total ligand concentration as derived from the mass of NOM added to 40 mL inorganic seawater and the molar mass of this NOM determined experimentally by FlowFFF. [Fe′] is the sum of the concentrations of all dissolved inorganic iron(III) species. We assume thermodynamic equilibrium of the dissolved inorganic iron(III) species with the solid iron(III)oxide-hydroxide phase present in the sample. The solubility product expression for Fe(OH)3 is the following:

According to Eq. (4), [Fe3 +] is constant for given values of Ksp and [OH−], which implicates that [Fe′] is constant, too. The reaction between Fe′ and the free ligand L can be described by the following simplified equation:

The mass action law representing this equilibrium is:

Eq. (6) describes the conditional stability constant Kcond as defined by Rue and Bruland (1995). [Fe′] can be assumed to have the same value in all samples of a given run, regardless of the NOM concentration. We have:where [FeT]0 represents the total dissolved iron concentration in equilibrium with the solid iron oxide phase in inorganic seawater (NOM concentration [LT]0 = 0 mol L− 1).

Eqs. (2) through (6) yield

If the data can be interpreted as 1:1 complex between Fe(III) and ligands with closely similar Kcond, a plot of [FeT] against [LT] will yield a straight line according to Eq. (8).

The conditional complex stability constant Kcond can then be derived from the y-intercept d and the slope k of this straight line:

Obtaining rigorous estimates of parameters for two ligand classes requires the application of non-linear regression (Pižeta et al., 2015). However, due to the very efficient separation method applied here, we found only one homogenous ligand class in each of our NOM samples. Simple linear regression was therefore used in order to estimate ligand parameters and error bars.

Results and discussion

Separation and characterization of NOM fractions

We collected NOM from a tributary of River Halladale in North Scotland (Craggie Burn, CB), and from a creek draining a raised peat-bog in Upper Austria (Tannermoor brook, TM). Suwannee River NOM from the International Humic Substances Society (Suwannee River, SR) was used as a reference material. In a first step, the seawater-soluble NOM fractions were isolated from the samples. The enhanced ionic strength reduced initial DOC in the freshwater samples by 53–54% and iron by 80–85%, in accordance with our previous results (Krachler et al., 2005; Jirsa et al., 2013). Further separation of the seawater-soluble fractions was carried out on a Sephadex-LH-20 column followed by UV detection (λ = 242 nm). The NOM samples of different origin (CB, TM and SR) showed nearly identical patterns. The UV-absorbing components were split into 3 baseline-separated peaks, as shown for CB in Fig. 3, and for SR and TM in Fig. 4. The employed size exclusion column contained hydroxypropylated dextran as stationary phase. Accordingly, retention of the investigated NOM-fractions, which are polar and/or moderately polar compounds, is not only based on their size and shape, but also controlled by weak electrostatic forces leading to a separation of three fractions with similar size but different polarity. Fractions of DOC recovered in the different components separated by gel chromatography are shown in Table 1.

Fig. 3

UV absorbance (λ = 242 nm) versus volume of eluent (mL) for the seawater-soluble NOM fraction of the Craggie Burn. The UV signal (solid red curved line) represents the NOM and is superimposed with the conductivity signal (solid blue curved line), the chloride concentration (green-colored bar graph), and the sulfate concentration (brown-colored bar graph). The mobile phase is 4.5% methanol in ultrapure water. Flow rate is 0.1 mL min− 1. The UV-absorbing NOM is clearly separated into three peaks (CB_0, CB_1 and CB_2). Larger molecules elute first, smaller molecules elute later.

Fig. 4

Gel chromatography on Sephadex-LH-20 of the seawater-soluble NOM fractions (a) of Suwannee River NOM (standard material purchased from the International Humic Substance Society), and (b) of a peat-bog draining creek (Tannermoor in Upper Austria). UV absorbance (λ = 242 nm) versus volume of eluent (mL). The mobile phase is 4.5% methanol in ultrapure water. Flow rate is 0.1 mL min− 1. In both samples, the UV-absorbing NOM is clearly separated into three peaks (SR_0, SR_1, SR_2; TM_0, TM_1, TM_2). Larger molecules elute first, smaller molecules elute later.

Table 1

DOC concentrations in peatbog-influenced creek waters and fractions of DOC recovered in different components separated by gel chromatography.

Sample	DOC [μmol L− 1]± 5%	DOC recovered from creek water (%)
Initial creek water sample from the Craggie Burn (CB)	4030
Initial creek water sample from the Tannermoor (TM)	6780
CB salta	1870	46.4
TM salta	3120	46.0
CB_0	150	3.7
CB_1	130	3.2
CB_2	60	1.5
TM_0	270	4.0
TM_1	250	3.7
TM_2	130	1.9

Seawater soluble fraction of initial creek water DOC.

After separation, the Craggie Burn and Suwannee River fractions (CB_0, CB_1, CB_2; SR_0, SR_1, SR_2) were pre-concentrated (1:100) and analyzed using Asymmetric Flow Field-Flow Fractionation (FlowFFF). The aim was to characterize the NOM fractions and the associated Fe that resisted aggregation and flocculation induced by the addition of seawater. The FFF chromatograms in Fig. 2 show both the elution patterns of the UV absorption signal and the iron signal. The iron peak signal is in all samples in the domain of the UV-absorbing NOM. The molar mass distribution was derived from the retention volume of Fe and NOM by calibration with polystyrene sulfonate molar mass standards (1100 g mol− 1; 3610 g mol− 1; 10,600 g mol− 1) according to Neubauer et al. (2011). The molar mass of the iron-binding NOM was obtained from the peak maximum of the Fe signal. The resulting molar masses are listed in Table 2.

Table 2

Results of FlowFFF analysis: Fe concentration and pH of the samples, Fe and UV–Vis recovery from the FlowFFF analysis, and molar mass (MM) at the mode (peak maximum) of the Fe signala.

	Feb[μmol L− 1]	DOC[μmol L− 1]	pH	Recovery UV–Vis%	Recovery Fe%	MMc[g mol− 1]
Peat-bog draining creek
CB_0	87.9 ± 4.5	12,060 ± 600	5.9	84	93	3320 ± 100
CB_1	22.1 ± 1.1	2450 ± 120	5.3	46	102	2420 ± 90
CB_2	29.0 ± 1.4	3730 ± 260	6.5	38	101	2810 ± 100
Suwannee River NOM
SR_0	17.5 ± 0.9	2500 ± 130	5.5	82	105	3450 ± 110
SR_1	19.5 ± 0.9	2090 ± 100	5.2	60	100	2120 ± 100
SR_2	17.3 ± 0.9	1900 ± 100	4.5	51	108	2580 ± 90

Shown in Fig. 3.

From ICP-MS bulk measurement.

± denotes the confidence interval (P = 0.99).

The molar mass for the fractionated Craggie Burn NOM ranged between 2400 and 3300 g mol− 1, and between 2100 and 3400 g mol− 1 for the fractionated Suwannee River NOM (Table 2). Samples CB_0 and SR_0 exhibited the widest molar mass distribution of the NOM (shown in Fig. 2) which is in accordance with the results of gel chromatography (Figs. 3 and 4). Sample CB_0 contained a higher Fe concentration than the other fractions from the gel chromatography separation (Table 2). This is also reflected in the FlowFFF analysis that revealed that more Fe was associated with the NOM in this fraction compared to CB_1 and CB_2. The latter fractions exhibit a similar Fe and NOM distribution and have a lower molar mass compared to CB_0. This is in accordance with the elution chronology of the fractions from the gel chromatography. The results show that the higher molar mass fraction (CB_0) is more relevant for Fe binding than the lower molar mass fractions (CB_1, CB_2) similar to what was observed in our previous FlowFFF studies (Neubauer et al., 2013a; Neubauer et al., 2013b).

Determination of Fe(III)-oxyhydroxide solubility in seawater

Verification of the 59Fe method used in the present work was carried out by measuring Fe(III)-oxyhydroxide solubilities in our synthetic seawater at 25 ± 2 °C as a function of pH. Equilibration time was 1 week. The results of these experiments are in perfect agreement with the measurements of Liu and Millero (1999, 2002) who determined solubilities of Fe(III)-oxyhydroxide in a 0.7 M NaCl solution at 25 °C as a function of pH using a 59Fe method. Our results are shown in Fig. 1, together with the data by Liu and Millero (1999).

After separation on the Sephadex LH-20 column, the seawater-soluble NOM-fractions (TM_0, TM_1, TM_2, SR_0, SR_1, SR_2, CB_0, CB_1, CB_2) were collected and dried at room temperature. Total ligand concentrations [LT] (nmol L− 1) were derived from the mass of dry NOM added to a known volume of inorganic seawater (40 mL) and the molar mass of this NOM determined by FlowFFF (Table 2). The iron(III) complexing ability in seawater of the NOM fractions was studied using 59Fe as a radioactive tracer. Results are given in Fig. 5. The diagrams in Fig. 5 show the total dissolved iron concentration [FeT] (nmol L− 1) in synthetic seawater at pH = 8.2 in equilibrium with in-situ precipitated Fe(III)-oxyhydroxide (aged 1–2 weeks) as a function of the NOM concentration (nmol L− 1). Conditional complex stability constants log (Kcond) have been derived from the y-intercept and the slope of the linear regression lines, using Eq. (9). Conditional complex stability constants (present study and literature data) are given in Table 3.

Fig. 5

Results of gamma-ray spectrometry. Total concentration of dissolved iron [FeT] (nmol L− 1) in synthetic seawater (pH = 8.2) in equilibrium with in situ precipitated iron(III)-oxyhydroxide versus NOM concentration (nmol L− 1). Temperature (°C), y-intercept and slope of the linear regression line, and the correlation coefficient from the regression analysis are also shown. NOM concentration = [LT] is the total ligand concentration as derived from the mass of NOM added to 40 mL inorganic seawater and the molar mass of this NOM determined experimentally by FlowFFF.

Table 3

Conditional complex stability constants for Fe(III) complexes with seawater-soluble riverine NOM in seawater.

NOM fraction	Method	NOM concentration during measurements (mg L− 1)	Log Kcond	Reference
Suwannee River fulvic acid	Competing ligand/UV–Vis	10	10.4	Rose and Waite (2003)
Suwannee River fulvic acid	Competing ligand/cathodic stripping voltammetry	1–4	10.6 ± 0.2	Laglera and van den Berg (2009)
Suwannee River NOM fraction SR_0	Radiotracer technique	0.0005–0.004	10.6 ± 0.4	Present work
Suwannee River NOM fraction SR_1	Radiotracer technique	0.0005–0.004	9.7 ± 0.6	Present work
Suwannee River NOM fraction SR_2	Radiotracer technique	0.0005–0.004	9.5 ± 0.6	Present work
Peat-bog draining creek NOM fraction TM_0	Radiotracer technique	0.0005–0.004	10.4 ± 0.4	Present work
Peat-bog draining creek NOM fraction TM_1	Radiotracer technique	0.0005–0.004	10.5 ± 0.6	Present work
Peat-bog draining creek NOM fraction TM_2	Radiotracer technique	0.0005–0.004	9.5 ± 0.6	Present work
Peat-bog draining creek NOM fraction CB_0	Radiotracer technique	0.0005–0.004	9.8 ± 0.6	Present work
Peat-bog draining creek NOM fraction CB_1	Radiotracer technique	0.0005–0.004	8.8 ± 0.6	Present work
Peat-bog draining creek NOM fraction CB_2	Radiotracer technique	0.0005–0.004	8.5 ± 0.6	Present work

Over the observed temperature range 10 °C–40 °C, variations of log Kcond with temperature were negligible. This is in accordance with Hiemstra and van Riemsdijk (2006) who found that Fe complexation by terrigenous NOM is almost temperature independent. Iron complexation at different temperatures has also been investigated by Hassler et al. (2013). The levels of affinity for iron(III) differ considerably between NOM from different sources, and also between different size fractions of the NOM (see Table 3). However, all investigated NOM fractions exhibit significant affinity for iron(III) in seawater even if highly diluted. Our result for the conditional complex stability constant for fraction SR_0 of Suwannee River NOM is in excellent accordance with the result by Laglera and van den Berg (2009) for the conditional complex stability constant for Suwannee River fulvic acid (FA) in UV-digested seawater (see Table 3).

Fig. 6 visualizes the fraction of total dissolved iron bound to NOM (%) as a function of NOM concentration (nmol L− 1), as resulting from our experimental data (samples TM_0 and TM_1). As it can be seen from the diagram, in synthetic seawater with NOM concentration 1.7 nmol L− 1 (5 μg L− 1), 97% of the total dissolved iron is bound to the organic ligands, i.e., Fe(III) solubility is enhanced by a factor of F = 33. In seawater with NOM concentration 8.5 nmol L− 1, (25 μg L− 1), 99% of the dissolved iron is associated with the NOM, and the solubility of Fe(III) is enhanced by a factor of F = 100 (Fig. 6).

Fig. 6

Fraction of total dissolved iron bound to peat-bog derived NOM (%) versus NOM concentration (nmol L− 1), as calculated from the data points shown in Fig. 5 d), e), and j) (samples TM_0 and TM_1). NOM concentration = [LT] is the total ligand concentration as derived from the mass of NOM added to 40 mL inorganic seawater and the molar mass of this NOM determined experimentally by FlowFFF.

Conclusions

The present study shows that the affinity for Fe(III) differs considerably between aquatic humic materials from different sources, and also between different size fractions (Table 3). However, most of the river-derived humic fractions studied here were able to substantially enhance Fe(III)-oxyhydroxide solubility in seawater at concentrations of the humic material ≥ 5 μg L− 1. Opsahl and Benner (1997) and Benner et al. (2005) studied the distribution of terrigenous dissolved organic material in the ocean and found that terrigenous DOC (dissolved organic carbon) is present throughout the Atlantic and Pacific Oceans in the range of 0.7–2.4% of bulk DOC and that terrigenous DOC concentrations are 7–16 fold higher in Arctic surface waters than in Atlantic and Pacific surface waters. The near-surface bulk DOC concentration in the North Atlantic Ocean is about 60 μmol L− 1 (Hansell and Carlson, 2001). Combining these data we see that land-derived NOM is present in the surface waters of the open North Atlantic Ocean at concentrations of about 10–35 μg L− 1. River-derived humic substances may therefore play a greater role as iron carriers in seawater than previously thought, and could be of particular importance in surface waters of the Arctic and North Atlantic Oceans.

Spatial–temporal variations in ligand concentrations in marine surface waters have the potential to impact primary production via changes in iron limitation (Völker and Tagliabue, 2015). Over the last decades, an increase in both NOM and iron concentrations has been observed in boreal blackwater rivers which points to a significant impact of climate change and permafrost thaw on these rivers. The increase in riverine iron and NOM may enhance the flux of iron-binding NOM to the Ocean (Kritzberg and Ekstrom, 2012; Kritzberg et al., 2014; Muller and Tankéré-Muller, 2012; Ilina et al., 2014). Benner et al. (2005) have shown that 25–33% of the NOM discharged to the Arctic Ocean by rivers is transported by sea currents to the North Atlantic. The observed change in river biogeochemistry could therefore give rise to enhanced primary production in extended marine regions and carbon export from the atmosphere into the deep sea, inducing a negative climate feedback.

Author contributions

RK was involved in study design and wrote the manuscript, RFK collected the samples and contributed to the experimental work, GW performed the radiochemical part of the work, ML, MFCR, and FJ performed experiments, EN, FVDK, and TH performed the FFF and ICP-MS analyses, SH contributed to the work in the clean room laboratory, BK was involved in study design, and all authors contributed material to the final version.