Global Biogeochemical Cycles

Nanoscale lignin particles as sources of dissolved iron to the ocean

Authors


Corresponding author: R. Krachler, Institute of Inorganic Chemistry, University of Vienna, Waehringerstrasse 42, A-1090 Vienna, Austria. (regina.krachler@univie.ac.at)

Abstract

[1] Primary production in large areas of the open ocean is limited by low iron concentrations. Rivers are potential sources of iron to the ocean, however, riverine iron is prone to intense flocculation and sedimentation in the estuarine mixing zone. Here we report the detection of iron-rich nanoparticles in a typical peatland-draining creek which are resistant against salt-induced flocculation i.e., their behavior is in sharp contrast to the well-known behavior of Fe colloids in river waters. Sample fractionation by AsFlFFF (Asymmetric Flow Field Flow Fractionation) revealed that these powerful iron carriers are in the size range of only 0.5–3.0 nm hydrodynamic diameter. They were isolated from the water phase using solid phase extraction/gel permeation chromatography, and analyzed by a CuO oxidation/GC-MS method. Our results suggest that the particles consist mainly of lignin catabolites and that gymnosperm as well as angiosperm tissues are contributors to the seawater-resistant iron-bearing DOM. Lignin phenols, which have no autochthonous source in the ocean, have been nevertheless found in low concentrations throughout the entire Arctic, Atlantic, and Pacific oceans. It is therefore tempting to speculate that peatland-derived iron-bearing lignin particles may have a sufficiently long half-life in ocean waters to sustain iron concentration in extended regions of the ocean.

1. Introduction

[2] The extremely low concentrations of iron in surface seawater limit biological productivity in large areas of the world ocean [e.g., Hunter and Boyd, 2007; Gnanadesikan and Marinov, 2008; Breitbarth et al., 2010]. Likewise, low Fe availability can have a critical influence on the composition and structure of algal communities, because of differences in requirements among species [Hiemstra and van Riemsdijk, 2006]. Due to the very low solubility of Fe(OH)3 in seawater [Liu and Millero, 2002], dissolved Fe tends to precipitate and has a relatively short residence time in the surface ocean. Continental weathering of silicates, and, subsequently, river input of dissolved weathering products is one of the possible sources of iron to the ocean, with the constraint that the river-borne iron input is controlled by estuarine sedimentation processes. It is well accepted that the vast majority of “dissolved” Fe in river water exists as colloid particles (mainly hydrous ferric oxides and iron-binding humic substances) [Fox, 1988; Dai and Martin, 1995; Wen et al., 1999, Baalousha et al., 2006]. Aggregation of these colloids, due to the major change in ionic strength upon mixing of river water with seawater, causes a massive removal of the Fe in the estuarine mixing zone [Sholkovitz, 1976, 1978; Sholkovitz et al., 1978; Nowostawska et al., 2008].

[3] However, this removal is not necessarily complete. Iron may exhibit different behaviors depending on a number of variables [Powell and Wilson-Finelli, 2003]. Geographic position, size, and types of vegetation cover and land use of a given river basin are critical factors influencing both, riverine DOM (dissolved organic matter) quality [Vogt et al., 2004], and the riverine iron transport from land to the sea. For instance, iron transport by peat-bog-derived humic substances (HS) has been identified as an important carrier mechanism for riverine Fe [Krachler et al., 2005, 2010]. The influence of this Fe source may reach further out to sea than previously expected [Pan et al., 2011].

[4] The discharge of organic carbon in riverine DOM to the oceans (0.4 × 1015 g per year) [Opsahl and Benner, 1997] consists primarily of the degraded remains of terrestrial vegetation. The biodegradation products of lignins play a special role in riverine DOM since they are more refractory to further degradation than other biodegradation products of plant tissues. The most important aspects of lignin biodegradation are depolymerisation and solubilization, which result from oxidative reactions following β-ether linkage cleavage by extracellular enzymes. A water-soluble catabolite (APPL, acid-precipitable polymeric lignin) is observed during the process of lignin degradation [Borgmeyer and Crawford, 1985; Ko et al., 2009].

[5] Peat-bogs are generally dominated bySphagnum mosses that form the bulk of living and dead biomass, but vascular plants are also present (e.g., sedges, shrubs and trees). The decomposition of their lignins and formation of water soluble APPL has been reported to proceed in the acrotelm by microbial activity under aerobic conditions [Kracht and Gleixner, 2000; Djurdjević et al., 2003]. In the catotelm peat, further degradation of APPL occurs, particularly through demethylation of methoxyl groups under anaerobic conditions in an acidic environment [Kuder and Kruge, 1998]. Diagenetically modified lignin degradation products constitute a major component of the HS leaching out from peat soils to streams and rivers [Zaccone et al., 2008].

[6] HS from peat soils have functional groups which make them capable of reacting with Fe(II) and/or Fe(III) [Guillon et al., 2001; van Schaik et al., 2008] and, as a consequence, they show high efficiency in mobilizing iron from the soils. Peat-draining “black” rivers are thus rich in colloidal iron [Batchelli et al., 2010; Jones et al., 2011].

[7] The higher dissolution of Fe hydroxide in natural seawater than in inorganic synthetic seawater or ultraviolet (UV)-irradiated seawater, respectively, has been attributed to the presence of natural organic Fe(III) chelators in the ocean waters [van den Berg, 1995; Rue and Bruland, 1995; Kuma et al., 1996; Liu and Millero, 2002]. The extremely low concentrations of these ligands, however, make their direct detection a formidable task [Ibisanmi et al., 2011]. There are many mechanisms of DOM production in the upper ocean [Nianzhi et al., 2010] and autochthonous DOM e.g., bacterial siderophores have been suggested as ligands for Fe(III) in seawater [Hunter and Boyd, 2007; Hopkinson and Morel, 2009]. Investigations by Hiemstra and van Riemsdijk [2006] have shown that the behavior of dissolved iron in the ocean may be partly explained by the interaction of iron with DOM that has humic substances (HS) like properties. Recently, Laglera and van den Berg [2009]developed a new method, based on cathodic stripping voltammetry, to measure iron-binding humic substances (HS) in seawater and found a covariation of iron and HS in the coastal waters of the Irish Sea. They concluded that the iron is transported from the estuarine waters to the open sea by land-derived humic ligands. Their finding means that terrestrial components (terrigenous HS) should be considered as an important iron chelating ligand class in addition to the autochthonous marine ligands, but the identity of the terrigenous HS ligand remained undiscovered.

[8] Here we suggest that unique lignin degradation products, leaching out from acidic peatland soils, may be promising candidates as land-derived humic carriers for iron in the ocean waters.

2. Materials and Methods

2.1. Study Area and Sample Collection

[9] The present study was conducted in the River Halladale catchment area, a coastal wetland ecosystem in North Scotland (UK), which is rich in (partly restored) Sphagnum-dominated domed peatlands and blanket bogs. Freshwater samples were collected during a field trip April 6–10, 2009 from the creek Craggie Burn (a River Halladale tributary, River Halladale has a mouth into the sea), 85 m above sea level, geographic coordinates: N 58°25.821′ W 3°54.320′. Nearly pristine coastal seawater was collected from the Skerray Harbour, geographic coordinates: N 58°32.300′ W 4°18.100′. The geographic location is shown inFigure 1. Contamination of this seawater has not been measured, however, the area is only thinly populated and human caused factors that impact marine water quality seem to be negligible.

Figure 1.

Map of study area and location of sampling sites.

[10] All reagents used in this study were high purity chemicals by Sigma-Aldrich. All solutions were prepared with Millipore “Synergy UV” ultrapure water. Glassware and polyethylene-ware followed a rigorous cleaning protocol involving soaks in detergent and HNO3baths, and were subsequently rinsed with high-purity water. Ten L water samples were collected from the surface using a clean polyethylene beaker which had been rinsed with sample water before filling, and the samples were then passed through a sterile 0.2μm Sartobran 300 capsule cellulose acetate membrane filter using a Watson Marlow peristaltic pump. The filtered samples were stored in polyethylene bottles in the dark at 4°C. The pre-treatment immediately after sampling by filtration (0.2μm) of the natural freshwater or seawater, respectively, yields samples which are stable for several weeks at 4°C [von der Kammer et al., 2004; Baalousha et al., 2005; Baalousha, 2009].

[11] 100 mL portions of the filtered samples were used for the analysis of iron, silicon and sodium. After adjusting the pH to approximately 1.8 by adding concentrated nitric acid, the acidified samples were stored for several days. Subsequently, iron and silicon concentrations were determined using inductively coupled plasma-optical emission spectrometry (ICP-OES) (Optima 5300 DL by Perkin-Elmer). Na+concentrations were determined using flame atomic absorption spectrometry (flame-AAS) (Aanalyst 200 by Perkin-Elmer). Dissolved organic carbon (DOC) was measured in non-acidified filtered samples using a total organic carbon (TOC) analyzer by Shimadzu. Potassium hydrogen phthalate was used as an internal standard. Field instruments for portable water analysis by WTW (Munich, Germany) were used to measure temperature, pH and specific electrical conductance in the unfiltered water samples immediately after sampling. Results of the physical and chemical characterization of the water samples are shown inTable 1.

Table 1. Creek Water and Coastal Seawater (North Scotland), In Situ Measurements, and Experimental Results of Filtrates (0.2 μm Cut-Off)a
 T (°C)EC (μS cm−1)pH (25°C)NPDOC (mg L−1)Na+ (mg L−1)Fe2+/Fe3+ (mg L−1)Si4+ (mg L−1)
  • a

    NPDOC = non purgeable dissolved organic carbon. EC = specific electrical conductance. ND = not determined.

fresh creek water, April 200910.7131.106.115.0016.560.722.28
coastal seawater, April 200910.248,000.008.11.8510,486.50NDND

2.2. Colloidal Distribution and Mixing Experiments

[12] The fractionation of aquatic natural inorganic and organic nanoparticles by Asymmetric Flow Field Flow Fractionation has been optimized by Dubascoux et al. [2008]. One of the main advantages of Asymmetric Flow Field Flow Fractionation lies in its ability for coupling with various detectors permitting complementary information to be obtained [Hassellöv et al., 2006]. In the present investigation, the metal distributions in different size fractions of the river-borne humic colloids were measured via coupling Asymmetric Flow Field Flow Fractionation to Inductively Coupled Plasma Mass Spectroscopy (ICP-MS). The iron carrying capacity as a function of colloidal size was investigated by simulating estuarine mixing of the fresh water with seawater. Increasing percentage of filtered natural seawater was added to the creek water samples in order to observe the distribution behavior of iron as well as other prominent metals, i.e., lead and manganese.

[13] Field-flow-fractionation was performed using an Asymmetric Flow FFF – AF4, from the AF2000 MultiFlow FFF Series from Postnova Analytics (Germany). The AF2000 Series is based on a crossflow field as driving force for the separation. Conditions were optimized to increase the resolution of the lower molecular weight fractions. As accumulation wall, we used a membrane made from regenerated cellulose (RC) with pore size 1 kDa (Postnova Analytics). A PTFE spacer of 500μm in height was used. The carrier was 25 mM NaCl solution prepared with ultrapure water. The carrier must provide colloidal stability against aggregation (particle-particle interaction) and against adsorption of the colloids on the channel surfaces. Channel flow rate and timings were 5 mL min−1and 25 min. The eluent was directed from the AF4 through a photodiode array detector (PDA) SPD-M20 (Shimadzu Scientific Instruments) and UV absorption was measured atλ = 220 nm. The injection volume was 100 μL for each analysis. Macromolecular polystyrene sulfonates of 4 different molecular weights (Polysciences, Eppelheim, Germany) were used as standards for the molecular weight calibration (1.92, 16.8, 32.9 and 63.9 kDa). For the analysis of iron in the water samples, the AF4 system was coupled with an ICP-mass spectrometer. The instrument used was a Perkin Elmer Elan 6100 without collision cell. To avoid polyatomic interferences on56Fe the analysis was performed on the 57Fe isotope, which resulted in a low sensitivity for this element. This is visible in the strong noise proportions in the Fe signals and limited the analysis to samples with elevated iron content. However, the carrier of the FFF, which is introduced to the ICP-MS, is extremely dilute, iron is already a main element here, all other elements are equal or lower concentrated and therefore relatively good D.L.s were obtained.

[14] Hydrodynamic diameters were calculated from mean diffusion coefficients using the Stokes-Einstein equation.

[15] The procedures for laboratory mixing experiments were similar to those described previously by Stolpe and Hassellöv [2007]. Portions of the filtered creek water samples from the Craggie Burn (April 2009, Table 1) were mixed with filtered coastal seawater from Skerray Harbour (April 2009, Table 1) to final salinities of 0, 2.5 and 5.0. The mixtures were shaken for 2 min, and thereafter left to react for 20 min in a refrigerator at 4°C. 100 μL of the mixture was injected into the AF4.

2.3. Isolation and Identification of Iron-Binding Dissolved Lignin Phenols

[16] Methods for isolation and quantification of dissolved lignin phenols in seawater have been developed by Hedges and Ertel [1982], and Louchouarn et al. [2000, 2010].

[17] A mixture of solid sea salts [Kester et al., 1967] (Table 2) was dissolved in 8 L of the 0.2 μm cut-off filtered creek water sample from the Craggie Burn to a final salinity of 35, and the pH was adjusted to pH = 8 with small amounts of 1 M NaOH. This leads to an unstable pH which tended to fall to 7.8. Iron contamination through salting-out (addition of sea salts in order to induce colloid coagulation) was <0.1%. The oxygen-saturated solution was shaken and thereafter left to react for 3 days in a refrigerator at 4°C. After the designated reaction time, the sample was re-filtered through Sartobran 300 cellulose acetate membrane filter capsules (Sartorius, pore size 0.45μm + 0.2 μm) in order to isolate the seawater-soluble iron carriers.

Table 2. Salts Used for Experimentsa
 Mass (g L−1)
NaCl23.926
MgCl2.6H2O10.831
CaCl2.2H2O1.518
Na2SO44.008
KCl0.667
NaHCO30.196
KBr0.098
H3BO30.026
SrCl2.6H2O0.024
NaF0.003

[18] Solid-phase extraction (SPE) of DOM in the saline filtrate was performed on pre-packed columns (C-18 Mega Bond Elute/40 g Sorbent, Varian). Cartridges were pre-treated with methanol followed by acidified (pH = 2) Millipore “Synergy UV” ultrapure water. The saline filtrate was acidified to pH = 2 using HCl > 30% (Sigma-Aldrich, for trace analysis), and, immediately after acidification, pumped through the SPE cartridge with a peristaltic pump (Masterflex) at a flow rate of 50 mL min−1. While the low pH may lead to decomposition of the iron(III)-HS complexes, the decomposition reactions are slow which makes it possible to use SPE. After the sample was extracted, the cartridge was rinsed with 1 L of acidified (pH = 2) Millipore “Synergy UV” ultrapure water to remove residual salts, and then with 1 L ultrapure water to remove HCl. Thereafter, without delay, the organic material was eluted from the column in one fraction using 100 mL methanol. The eluent was collected into a clean glass flask and evaporated to a volume of 10 mL at 40°C (Rotavapor, Büchi). The 10 mL sample was diluted to 100 mL with Millipore “Synergy UV” ultrapure water. Five mL portions were pumped through a Sephadex LH-20 gel filtration column (GE Healthcare) (1 m, 25 mm i.d.) with a flow rate of 0.5 mL min−1. The mobile phase was 10% methanol, which was degassed prior to use. As detector, a Lambda 35 UV-Vis spectrometer (Perkin-Elmer) was used, and the UV absorption at 242 nm was measured. Ten mL samples of the eluent were collected and iron was measured therein with inductively coupled plasma-optical emission spectrometry (ICP-OES) (Optima 5300 DL, Perkin-Elmer).

[19] Figure 2 shows results of the gel permeation chromatography. The peak fraction with retention time 450–470 min was collected for further analysis. The sample was evaporated to dryness under Ar and lignin components were analyzed using the classical alkaline CuO oxidation and extraction scheme as described by Hedges and Ertel [1982], Kögel and Bochter [1985], and Goñi and Hedges [1992]. All analyses were performed at least in triplicate. Using an inert atmosphere glove box, a Teflon-minibomb was charged under an argon stream with 0.5 mg of the sample, 4.0 mL 8% (wt/wt) Ar–sparged aqueous NaOH solution, 1.00 g of CH2Cl2 extracted CuO (powdered), and 50 mg Fe(NH4)2(SO4)2 · 6 H2O. The Teflon-minibomb was loaded into a steel bomb and the CuO oxidation was carried out at 170°C. After 3 h of heating, the sample was allowed to cool to room temperature before the Teflon-minibomb was opened. The contents of the minibomb were washed with 30 mL 8% NaOH into a centrifuge tube and rotated at 4000 rpm for 10 min. Ethyl vanillin was used as surrogate standard and was added to the NaOH extract. The NaOH extract was then acidified to pH = 1 with 6 M HCl (p.a.) and extracted with distilled diethyl ether which had been previously treated with a saturated aqueous solution of Fe(NH4)2(SO4)2 · 6 H2O to remove peroxides. The ether extract was passed through an anhydrous Na2SO4 column and the ether was then removed under an Ar stream. The product was dissolved in 5 mL pyridine. In order to estimate loss/degradation of analytes during CuO oxidation, a mixture containing 0.1 mg of each of the 9 pure substances was subjected to the CuO oxidation procedure. The percentages of analytes recovered are shown in Table 3.

Figure 2.

Gel permeation (size exclusion) chromatography on a Sephadex LH-20 column of the seawater-soluble DOM fraction from creek Craggie Burn. UV absorbance (λ = 242 nm) versus retention time (min). The UV signal (red curve) represents the DOM and is superimposed with the iron signal (blue curve). The mobile phase is 10% methanol. Flow rate 0.5 mL min−1. Large particles elute first, smaller particles elute later.

Table 3. GC-MS Results of Trimethylsilyl Derivatives of CuO Oxidation Products
 Target Ions (m/z)Retention Time (min)Percentage of Lignin Phenols Recovered From a Mixture of the Pure Substances After CuO Oxidation (%)Concentration (mg L−1)aPositively Identified in Open Ocean Waters Around the World [Hernes and Benner, 2002]Concentration (ng L−1) in Water Samples From the Bermuda Station [Hernes and Benner, 2006]
  • a

    Concentrations measured in the fraction with retention time 450–470 min; see Figure 2. Losses during CuO oxidation are not taken into consideration.

ethyl vanillin167/195/23815.0068.4---
vanillin194/193/20913.3790.50.82 ± 0.08+8.24
acetovanillone193/208/22315.6274.40.44 ± 0.04+3.48
vanillic acid267/297/31219.52109.01.21 ± 0.12+14.0
syringaldehyde224/239/25417,7668.70.19 ± 0.02+4.07
acetosyringone238/253/26819,5262.60.31 ± 0.03+2.73
syringic acid327/312/34222,8829.70.11 ± 0.01+11.7
p-coumaric acid219/293/30823,8746.9---
ferulic acid338/249/32327,4617.0---

[20] For the GC-MS analysis, 200μL of the sample together with 200 μL bis-N, O-trimethylsilyltrifluoroacetamide were transferred into a vial and heated to 60°C for 10 min. Separation of the trimethylsilyl derivatives was performed immediately after silylation by gas chromatography on a Clarus 500 GC (Perkin-Elmer) fitted with a Perkin-Elmer Elite-Plot Q capillary column (30 m, 0.25 mm i.d.). The samples were carried through the capillary column by Helium (1.3 mL min−1) under a 1/13 split ratio. The temperature was raised from 100°C to 270°C at 4°C min−1and was held at 270°C for 10 min. The mass spectrometer was operated in the EI mode (70 eV) and three characteristic ion masses were scanned for each analyte. After a 12 min solvent-delay the scan time was 35.5 min. Positive identification of the analytes was performed using retention times and by comparing the relative abundance of the three characteristic ions in each sample to those produced by the 9 standard substances [Louchouarn et al., 2000]. Peak area ratios of the analyte to the standard were used to construct calibration curves. Seven concentration levels were used in calibration curves, which were linear for all of the target analytes with correlation coefficients (R2) of greater than 0.998. Limits of detection (LOD) were calculated using the equation LOD = 3σ/k where σ is the signal variability of the blank and k is the slope of the calibration curve. LOD ranged from 10 to 80 μg L−1.

3. Results and Discussion

3.1. Colloidal Distribution and Mixing Experiments

[21] The UV signal extracted from the AF4 channel pinpoints the DOM size range in the 0.2 μm cut-off filtered creek Craggie Burn water (Figures 3a and 3b), showing very small particles about 0.5–3.0 nm in hydrodynamic diameter which is typical for fulvic compounds [Buffle et al., 1998]. Larger DOM colloids with hydrodynamic diameter ≫3 nm are absent in this sample. The samples eluting from the AF4 channel further detected by ICP-MS show the metal distribution among the measured size ranges (Figures 3a and 3b). While Pb and Mn give peak signals mainly in the DOM domain, Fe further shows secondary adjacent peaks with significantly higher peak area. These secondary peaks point out the presence of an iron-rich nano-sized phase in the fresh water. Occurrence of inorganic Fe-rich colloids differing in size from the organically bound Fe could be related to the oxidative formation of FeOx in low pH peat bog drainage when mixed into higher pH creek waters since the Fe(II) oxidation rate is strongly pH dependent. As can be seen from Figures 3c–3f, seawater mixing affects only the larger iron-bearing colloid portion. The components of this size region (3–50 nm) are prone to fast coagulation when attractive van der Waals forces start exceeding the repulsive electrostatic forces, as a result of both contraction of the electrical double layer, and of decrease in surface net charge magnitude [Eckert and Sholkovitz, 1976; Boyle et al., 1977; Hunter and Leonard, 1988; Hunter et al., 1997; Mylon et al., 2004]. With increasing salinity, the secondary peaks of iron subside, and, concomitantly, a decreasing trend in the concentrations of Pb and Mn is observed. The fading secondary iron peaks in a salinity gradient can be explained by flocculation induced selective removal of iron oxides, whereas the surviving iron signal in the DOM size region points to an organically chelated iron fraction which is completely resistant against increasing concentrations of sea-salts.

Figure 3.

AF4-fractogramms of creek Craggie Burn water. The UV signal represents the DOM and is superimposed with the metal signals: (a and b) fresh creek water sample, (c and d) creek water and seawater mix at salinity 2.5, and (e and f) creek water and seawater mix at salinity 5.0. The focus time is 10 min.; the elution time of a polystyrene sulfonate standard of 40,000 g mol−1 equals 20 min. Smaller particles elute before the larger ones.

[22] Organically chelated iron has been detected recently in the River Thurso plume by a voltammetric titration method by Batchelli et al. [2010]. These authors concluded that the majority of dissolved iron in the coastal waters was present as iron-humic complexes supplied by the river. River Thurso is very similar (and next) to River Halladale and the creek Craggie Burn, respectively. Thus the DOM-associated iron peaks inFigures 3c and 3ecorroborate the iron-humic transport across the river-estuarine boundary as observed byBatchelli et al. [2010].

[23] In contrast to the DOM-bound iron, the likewise small-sized lead and manganese nanoparticles are vulnerable to salt-induced flocculation (Figures 3d and 3f). The Mn and Pb signals vanish both under elevated ionic strength, which could be explained by flocculation of very small MnOx particles in the size range of the DOM signal. Both synthetic and natural Mn oxides show higher sorption efficiency for Pb2+ than iron oxides [O'Reilly and Hochella, 2003]. The parallel elution patterns of Pb and Mn in the creek Craggie Burn can be speculated as binding of Pb2+ to small MnOx particles. The simultaneous disappearance could be explained by both, binding of Pb2+ to 1–3 nm small MnOx particles, and aggregation and settling under high ionic strength.

[24] Stolpe and Hassellöv [2007] performed mixing experiments of boreal black creek water (Delsjöbäcken creek) with synthetic seawater, followed by size fractionation with Asymmetric Flow Field Flow Fractionation. Differing from the elution profiles observed in the present study, Stolpe and Hassellöv [2007] found in the Delsjöbäcken creek DOM, Fe, Pb and Mn peak signals mainly in the 3–50 nm size range. After seawater addition, a removal of Fe, humic acids, Pb, Mn and other metals in the 3–50 nm size region was observed, whereas the 0.5–3.0 nm DOM peak and the associated iron peak were unchanged by the addition of seawater. The latter observation is in excellent accordance with the present results. The authors concluded that the removal of total dissolved Fe by seawater addition was ≥90% in the boreal Delsjöbäcken creek. In the present investigation, about 16.7% (0.12 mg L−1) of the dissolved iron in the temperate Craggie Burn was bound to small-sized DOM particles in the size-range 0.5–3.0 nm and was therefore not prone to sea-salt induced flocculation.

3.2. Isolation and Identification of Iron-Binding Dissolved Lignin Phenols

[25] Ten salts, typically found in the seawater environment, were dissolved in the filtered creek water sample in concentrations as proposed by Kester et al. [1967] (salinity 35). As a result, a large part of the organic colloids together with the majority of the dissolved iron was removed from the water phase by flocculation. After a reaction time of 3 days, the sample was filtered (0.2 μm cut-off) and the iron concentration as well as the DOC concentration was measured in the filtrate. The results showed that 16.7% of the iron and 52% of the DOC initially found in the fresh creek water sample was still present in the water phase at salinity 35. The iron concentration in the saline filtrate was 0.12 mg L−1. The Fe-binding DOM isolated from the saline filtrate by a combination of solid phase extraction and gel permeation chromatography on a Sephadex LH-20 column was subjected to alkaline CuO oxidation and GC-MS according to the method developed byHedges and Ertel [1982]in order to detect APPL-derived phenols.Table 3presents the results obtained from GC-MS. Eight lignin phenols, which are the most abundant lignin oxidation products, were determined in the seawater-resistant Fe-DOM fraction: vanillin, acetovanillone, vanillic acid, syringaldehyde, acetosyringone, syringic acid, p-coumaric acid, and ferulic acid. Of these, vanillin, acetovanillone, vanillic acid, syringaldehyde, acetosyringone, and syringic acid could be positively identified. Additionally, 11 unidentified oxidation products were detected at lower concentrations. The sum of lignin phenol concentrations was ∼3 mg L−1 and the corresponding Fe concentration 0.47 ± 0.02 mg L−1 (see Figure 2). These results show that a large fraction of the iron was complexed with lignin phenols.

[26] The oxidation state of the iron within these lignin complexes is not known. However, EXAFS studies on iron-fulvic acid complexes byvan Schaik et al. [2008](who were using fulvic acid from soil which very likely contained a significant fraction of lignin moieties) have shown that the iron was octahedrally configured with inner-sphere Fe-O interactions at 1.98 Ǻ which points to Fe(III) (at pH = 4).

[27] Syringyl phenols are known to be unique oxidation products of angiosperm lignins [Opsahl and Benner, 1997]. The simultaneous presence of syringyl and vanillyl phenols suggests that both angiosperm and gymnosperm vegetation contribute to the iron-carrying DOM.

[28] Lignin is a unique biogeochemical tracer for the quantification of terrigenous material in the ocean, and can also serve as marker of oceanic water masses since there is no autochthonous source for lignin in the ocean [Hedges and Mann, 1979]. Opsahl and Benner [1997] found the following 5 lignin phenols in DOM isolated from surface waters of the open Pacific and Atlantic oceans: vanillin, acetovanillone, vanillic acid, syringaldehyde, and syringic acid. In water samples from Station Aloha in the Pacific Ocean, acetosyringone was also present [Hernes and Benner, 2002]. The concentrations of lignin phenols in the open ocean were reported by Benner et al. [2005], Hernes and Benner [2002, 2006], and Louchouarn et al. [2010] to be in the range of 10−6 to 10−8 g L−1. These low concentrations point to a short residence time of lignins in the ocean [Hedges et al., 1997]. Lignin phenols may nevertheless play an important role as iron-binding ligands in the ocean waters.

[29] To have at least a ballpark figure for the significance of lignins as iron-chelating ligands in the surface waters of the oceans, one could multiply a typical oceanic lignin concentration by the iron:lignin ratio R found in our creek water sample (R ≈ 0.15). The average lignin phenol concentration in the North Atlantic surface waters has been reported to be ∼38 ng L−1 [Hernes and Benner, 2006]. The result of our crude initial estimate (0.15 × 38 = 5.7 ng L−1) points to an estimated concentration of ∼0.1 nmol L−1 iron in the North Atlantic Ocean from the peat source. In the North Atlantic Ocean, the ambient dissolved iron concentration in the euphotic zone has been reported to be 0.07–0.23 nmol L−1 [Martin et al., 1993]. Peat-bog derived iron chelating lignin phenols may therefore be responsible for ∼50% of the dissolved iron in the North Atlantic Ocean. As reported byBenner et al. [2005], the concentrations of dissolved lignin phenols in polar surface waters are sevenfold higher than those in the North Atlantic Ocean. The large contribution of terrigenous DOM contributed by Arctic rivers is responsible for the elevated concentrations of lignin phenols in the surface waters of the Arctic Ocean, and there is a significant physical transport of lignins to the North Atlantic Ocean e.g., by the East Greenland Current.

[30] The inverse linear relationship between dissolved iron and salinity in the central Arctic Ocean which has been found by Klunder et al. [2012]demonstrates once more the important role of Arctic rivers in the delivery of (lignin-bound) iron to the Arctic Ocean. But there are also other important sources of iron nanoparticles to the ocean, i.e., oxyhydroxide nanoparticles which are being introduced into the oceans from glacial meltwaters [Raiswell et al., 2006].

[31] Once in the oceans, the lignin-bound iron will be in competition with organic complexing agents of marine origin.Batchelli et al. [2010]reported that the iron in humic-rich coastal waters in North Scotland, though strongly bound to terrigenous DOM, may nevertheless be available for complexation by siderophore-type marine ligands.

4. Conclusions

[32] In the present paper, we report the detection of an iron-binding DOM fraction in a peat-bog-draining creek which is characterized both by extremely small colloid size (0.5–3.0 nm) and an unexpected solubility in seawater. This unusual resistance against salt-induced flocculation is in sharp contrast to the known behavior of Fe colloids in river waters. Our GC-MS results show that the Fe-carrying DOM contains moieties which are built up of lignin phenols identical to those which have been found in low concentrations throughout the entire Arctic, Atlantic, and Pacific oceans [Opsahl and Benner, 1997; Benner et al., 2005; Hernes and Benner, 2002, 2006; Louchouarn et al., 2010]. The investigated tributary Craggie Burn can be regarded as a typical peatland-draining creek, our findings are therefore consistent with the concept that seawater-resistant Fe-bearing lignin particles are integral components of DOM in peatland-influenced river waters. For example,Yoshimura et al. [2010]found evidence that organically bound iron supplied from the Amur River contributes significantly as an iron source for the highly productive Okhotsk Sea ecosystem. Given that a part of the drainage area of the Amur River is covered by boreal peatlands dominated by gymnosperm vegetation, it is very likely that the black waters of River Amur contribute large quantities of nanoscale iron-carrying lignin to the ocean.

[33] Terrigenous DOM has been shown to be a quantitatively important source of macronutrients to the ocean which accounts for about 25% of global ocean productivity [Opsahl and Benner, 1997]. Our present results indicate that peatland-derived terrigenous DOM might be a source of the micronutrient iron as well. Peatlands in Siberia, Canada, Alaska and Northern Europe cover more than 3.4 × 106 km2 [Limpens et al., 2008]. As phytoplankton growth in much of the world's ocean is highly dependent on the availability of iron [Galbraith et al., 2010; Moore et al., 2002], the effects of climate variability and land use change on production rates and properties of DOM in these large peatlands may in turn affect primary production in the ocean waters.

Acknowledgments

[34] We highly appreciate the financial support by the Austrian Science Fund (FWF) under grant P19629-N19. We thank John Leonard, Forsinard, for help with the field work.