Iron-rich colloids as carriers of phosphorus in streams: A field-flow fractionation study

Water Research 99 (2016) 83e90

Contents lists available at ScienceDirect

Water Research journal homepage: www.elsevier.com/locate/watres

Iron-rich colloids as carriers of phosphorus in streams: A field-flow fractionation study Stijn Baken a, *, Inge C. Regelink b, Rob N.J. Comans c, Erik Smolders a, Gerwin F. Koopmans c a b c

KU Leuven, Department of Earth and Environmental Sciences, Kasteelpark Arenberg 20 bus 2459, 3001 Leuven, Belgium Alterra, Wageningen University and Research Centre (WUR), P.O. Box 47, 6700 AA Wageningen, The Netherlands Wageningen University, WUR, Department of Soil Quality, P.O. Box 47, 6700 AA Wageningen, The Netherlands

a r t i c l e i n f o

a b s t r a c t

Article history: Received 20 October 2015 Received in revised form 13 February 2016 Accepted 24 April 2016 Available online 26 April 2016

Colloidal phosphorus (P) may represent an important fraction of the P in natural waters, but these colloids remain poorly characterized. In this work, we demonstrate the applicability of asymmetric flow field-flow fractionation (AF4) coupled to high resolution ICP-MS for the characterization of low concentrations of P-bearing colloids. Colloids from five streams draining catchments with contrasting properties were characterized by AF4-ICP-MS and by membrane filtration. All streams contain free humic substances (2e3 nm) and Fe-bearing colloids (3e1200 nm). Two soft water streams contain primary Fe oxyhydroxide-humic nanoparticles (3e6 nm) and aggregates thereof (up to 150 nm). In contrast, three harder water streams contain larger aggregates (40e1200 nm) which consist of diverse associations between Fe oxyhydroxides, humic substances, clay minerals, and possibly ferric phosphate minerals. Despite the diversity of colloids encountered in these contrasting streams, P is in most of the samples predominantly associated with Fe-bearing colloids (mostly Fe oxyhydroxides) at molar P:Fe ratios between 0.02 and 1.5. The molar P:Fe ratio of the waters explains the partitioning of P between colloids and truly dissolved species. Waters with a high P:Fe ratio predominantly contain truly dissolved species because the Fe-rich colloids are saturated with P, whereas waters with a low P:Fe ratio mostly contain colloidal P species. Overall, AF4-ICP-MS is a suitable technique to characterize the diverse P-binding colloids in natural waters. Such colloids may increase the mobility or decrease the bioavailability of P, and they therefore need to be considered when addressing the transport and environmental effects of P in catchments. © 2016 Elsevier Ltd. All rights reserved.

Keywords: Natural water Field-flow fractionation Iron Phosphorus Colloids Nanoparticles

1. Introduction The environmental impact of phosphorus (P) is mostly related to excessive emissions to natural waters, which contribute to cultural eutrophication of water bodies throughout the world (Schindler, 2012). Eutrophication may trigger summer algal blooms, which cause deep-water anoxia, odor problems, fish kills, toxin production, reduced biodiversity, and reduced recreational and amenity values of water bodies (Smith, 2003). The fate of P and its effects on aquatic ecosystems depend on its chemical speciation. In natural waters, P may occur as free ionic orthophosphate, polyphosphates,

* Corresponding author. Present address: European Copper Institute, Avenue de Tervueren 168 b-10, 1150 Brussels, Belgium. E-mail address: [email protected] (S. Baken). http://dx.doi.org/10.1016/j.watres.2016.04.060 0043-1354/© 2016 Elsevier Ltd. All rights reserved.

organic P compounds, and P associated with inorganic particles (Van Moorleghem et al., 2011). Small particles, i.e. colloids (1e1000 nm) or nanoparticles (1e100 nm), are of interest due to their mobility and high specific surface area (Buffle and Leppard, 1995a). The colloidal P may constitute a large part of the “dissolved” (<0.45 mm) P load of streams (Filella et al., 2006; Haygarth et al., 1997) and contributes to P transport in the soil-water continuum (Jarvie et al., 2012; Turner et al., 2004; van der Salm et al., 2012). The environmental effects of colloidal P differ from those of free orthophosphate: for example, colloidal P is less available to algae, and the eutrophication risk associated with colloidal P is therefore lower than that of free orthophosphate (Baken et al., 2014). Clearly, the role of colloidal P needs to be appreciated in order to understand the transport and effects of P in natural waters. Colloidal P in natural waters is associated with a variety of particles. Many studies highlight the presence of Fe in such

84

S. Baken et al. / Water Research 99 (2016) 83e90

particles, but P-bearing colloids may additionally contain various other constituents, including organic C, Al, Ca, and Si (Gunnars et al., 2002; Jones et al., 1993; Mayer and Jarrell, 1995; Shaw et al., 2000). Analysis of the speciation of Fe has shown that the Fe in these colloids can be present as Fe oxyhydroxides (Lofts et al., €stedt et al., 2013), as structural Fe in clay minerals 2008; Sjo (Regelink et al., 2013a), or as ferric phosphates (Lienemann et al., 1999). The Fe-bearing colloids may bind P very effectively: molar P:Fe ratios in ferric phosphate colloids of up to 0.5 have been reported (Gunnars et al., 2002; Lienemann et al., 1999), and one study tentatively identified CaeFeeP colloids with a molar P:Fe ratio close to unity (Vega and Weng, 2013). Characterization of colloids by asymmetric flow field-flow fractionation (AF4) coupled to UV-DAD and ICP-MS detection, allows for continuous size separation and online detection of element concentrations. Such studies on natural water samples have mostly emphasized the existence of two main types of colloids: humic-rich colloids (0.5e5 nm) and Fe-rich colloids (3e40 nm) which are usually assumed to consist of Fe oxyhydroxides (Andersson et al., 2006; Baalousha et al., 2006; Benedetti et al., 2003; Gottselig n et al., 2003; Stolpe et al., 2013, 2010). These et al., 2014; Lyve colloids can carry a wide range of other elements, which may be bound e.g. by surface complexation or coprecipitation. The association of P with such nanosized particles is poorly documented, likely due to the difficult detection of low P concentrations by ICPMS. The available AF4 studies have yielded contrasting insights: one study identified clay particles as the key vectors of colloidal P in drainage waters (Regelink et al., 2013a), whereas others showed that colloidal P is associated with Fe and Al rich nanoparticles (Gottselig et al., 2014; Stolpe et al., 2010). It is unclear which processes determine colloidal P speciation in streams. Such information would enable a more accurate evaluation of the impact of P on surface water quality and of the eutrophication risk. The first objective of this work was to demonstrate the potential of AF4-ICP-MS to characterize P-bearing colloids in natural waters. The main challenge was to quantify low P concentrations, since colloidal P in natural waters typically occurs at low concentrations in the mg L
1 range. The second objective was to elucidate how catchment geochemistry determines the nature, size, and properties of these colloids. Phosphorus containing nanoparticles (1 kDae150 nm) in streamwater samples from five contrasting catchments were characterized using AF4 coupled to UV and high resolution ICP-MS detection. This was complemented by isolation and analysis of “coarse colloids” (100e1200 nm) by membrane filtration from the same samples. 2. Materials and methods 2.1. Sampling Five Belgian streams draining catchments of contrasting size, geology, and hydrology were selected. The Schwarzbach and Roer streams drain the Hautes Fagnes, an upland peat area in eastern Belgium with no agricultural land use (Wastiaux et al., 2000). The Kleine Nete catchment is a lowland catchment predominantly fed by iron-rich groundwater (Baken et al., 2015). The upper part of the Meuse catchment is dominated by calcareous sedimentary rocks, whereas the Belgian part of the Meuse drains the Ardennes massif which mostly consists of slate rocks (de Mars et al., 2000). The Dijle catchment drains the undulating plateau of the Belgian loam belt, which is dominated by luvisols developed on calcareous loess (Notebaert et al., 2009). All three latter catchments have agricultural land, and surface runoff likely contributes significantly to the total discharge in the latter two. More details on the sampling locations and the catchments are available in Table S1 in the

Supplementary Material. Sampling was done on January 14, 2014. The pH, water temperature, dissolved oxygen (DO) concentration, and electrical conductivity (EC) were measured in situ (Table 1). The streamwater samples were filtered in the field using membrane filters with nominal pore sizes of 1.2, 0.45, and 0.1 mm (details in Section 2 of the Supplementary Material). The concentrations of major and trace elements (ICP-MS), Fe(II) (by colorimetry (Viollier et al., 2000)), Fe(III) (as the difference between total Fe and Fe(II)), organic and inorganic carbon (elemental analyzer), and major anions (anion chromatography) were determined in the filtrates as described in Section 2 of the Supplementary Material. The UVabsorbance at 254 nm of the filtrates was determined spectrophotometrically as an indicator of humic substances (Weishaar et al., 2003). The term ‘coarse colloids’ is used in this study to indicate the size fraction isolated by membrane filtration, i.e. between 100 and 1200 nm. The composition of the coarse colloids was calculated as the difference between the element concentrations in these filtrates. This size range should only be taken as an indication of the true size of these colloids, since the nominal filter pore sizes were not verified and since membrane filtration is prone to artifacts (Gimbert et al., 2005; Horowitz et al., 1996). 2.2. Asymmetric flow field-flow fractionation The nanoparticles in filtered streamwater samples (0.45 mm; Atlas Filtri AC-BX filter cartridge) were characterized by AF4 (AF2000, Postnova Analytics) on January 29e31, 2014. The streamwater samples were stored for 2 weeks at 4  C pending AF4 characterization; we were unable to reduce this storage time due to the extensive preparations involved in setting up and testing the coupled AF4-ICP-MS system. An in-depth discussion of the concepts and theory of AF4 can be found elsewhere (Giddings, 1984; Von der Kammer et al., 2011). The AF4 protocol was similar to that of a previous study (Regelink et al., 2013b). The polyethersulfone AF4 membrane had a nominal pore size of 1 kDa. A 3 mM NaHCO3 solution at pH 8.3 was used as carrier. The injected sample volume was 2 or 10 mL, depending on the expected concentrations of P in the nanoparticles. The elution protocol lasted for 2400 s. The cross flow rate was relatively high (3 mL min
1) during the first 900 s in order to separate the smallest particles with a high resolution. The cross flow rate then decreased linearly within 120 s to 0.2 mL min
1 and remained at that value during the final 1380 s in order to separate the large nanoparticles at a lower resolution. Size calibration showed that four proteins with hydrodynamic diameters between 3.3 and 17 nm were well separated within the first part of the elution protocol, whereas polystyrene spheres of 20, 46, and 102 nm were separated in the second part. For both parts of the elution protocol, the relationship between retention time and hydrodynamic diameter was linear. According to this size calibration, colloids with a size up to approximately 150 nm could be detected using this protocol, although this upper limit is uncertain because it is an extrapolation: calibration was performed with particles of at most 102 nm. Colloids >150 nm did not elute within the specified time and were, therefore, not detected. The elution protocol, the size calibration, and the rinsing procedure of the AF4 channel between different runs are described in Section 3 of the Supplementary Material. In this study, particles detected by AF4 (1 kDae150 nm) are referred to as “nanoparticles”. An overview of the terminology used for different operationally defined size fractions used in this study is shown in Table S4. The actual size of natural nanoparticles may differ somewhat from the value obtained by using the above calibration, e.g. due to nonspherical particles or nonideal behavior of the particles (Baalousha and Lead, 2012). In addition, some artifacts may be introduced due to the

S. Baken et al. / Water Research 99 (2016) 83e90

85

Table 1 Chemical composition of filtered (<0.45 mm) streamwater samples. pH

Schwarzbach Roer Kleine Nete Meuse Dijle

6.3 6.8 7.2 7.9 7.9

DO

Hardness

EC

Na

Mg

K

Ca

Si

DOC

SUVA

Fe

Fe(II)

Al

P

DIC

Cl

SO4

NO3

mM

mM

mS cm
1

mM

mM

mM

mM

mM

mM

L (mg$cm)
1

mM

mM

mM

mM

mM

mM

mM

mM

0.34 0.34 0.31 0.28 0.31

0.08 0.15 1.10 1.25 3.09

36 67 565 323 866

0.09 0.17 1.95 0.43 1.28

0.04 0.06 0.25 0.20 0.49

0.01 0.01 0.20 0.05 0.12

0.04 0.09 0.85 1.05 2.61

0.06 0.06 0.15 0.09 0.24

0.59 0.45 0.74 0.38 0.38

0.033 0.036 0.025 0.019 0.018

5.5 4.9 16.2 0.7 0.8

2.3 1.8 2.5 0.4 0.2

5.7 2.8 1.1 0.8 0.2

0.1 0.2 1.7 1.8 3.7

0.05 0.14 1.34 2.19 4.94

0.10 0.24 1.82 0.48 1.87

0.05 0.06 0.94 0.22 0.64

0.04 0.06 0.13 0.25 0.51

DO: dissolved oxygen; EC: electrical conductivity; DOC: dissolved organic carbon; DIC: dissolved inorganic carbon; SUVA: specific UV-absorption.

mismatch between the carrier composition (pH and ionic strength) and the background composition of the streamwater samples, since equilibration of the colloids with the carrier solution may occur (Neubauer et al., 2013b, 2011). The AF4 was coupled online to a UV-detector (UV-DAD, SPDM20A, Postnova Analytics) and to a high-resolution ICP-MS (Element 2, Thermo Scientific). The UV signal at 254 nm was used as a proxy for humic substances (Weishaar et al., 2003). It was corrected for the known contribution of Fe(III) to the UV-absorbance using a previously established correction equation (Poulin et al., 2014). The ICP-MS was operated in medium resolution mode and monitored the concentrations of Fe, P, Al, Ca, and Si at intervals of 7.4 s. It was calibrated by off-line aspiration of calibration standards at the beginning and end of each measurement day. The calibration slopes measured at the end of each measurement day were mostly within 5% (and always within 10%) of those at the beginning, showing that signal drift of the ICP-MS was minimal. For each sample, the signals recorded during elution were converted to net signals by subtraction of the baseline, for which we used the median signal measured during the focusing step of each AF4 run. The ICP-MS signals were subsequently converted to element concentrations using the calibration slopes. Integration of the obtained peaks across a range of retention times yielded the element concentrations in the corresponding size range. The limit of quantification (LOQ) of each element was based on the noise in the baseline and on the signal observed during elution of blank samples, i.e. the so-called “blank fractionations” which correspond to injection of carrier solution. The blank fractionations allow to take into account the signal picked up during elution, e.g. due to sample carry-over or pressure-induced baseline drift (Gottselig et al., 2014). The LOQs are reported and discussed in Section 4 of the Supplementary Material. 3. Results and discussion 3.1. General streamwater chemistry The sampled streams can broadly be subdivided into two categories based on water hardness and pH (Table 1). The Schwarzbach and Roer are soft waters (Ca þ Mg < 0.2 mM) and have pH below 7. Compared to typical European streams (Salminen, 2005), they contain low concentrations of most common ions and nutrients. The low nutrient concentrations can be explained by the absence of agricultural land use and wastewater inputs in these catchments. The relatively high concentrations of Fe and the high specific UV absorbance, which shows that the dissolved organic matter mostly consists of humic substances (Weishaar et al., 2003), are indicative of peat soils in these catchments. In contrast, the Kleine Nete, Meuse, and Dijle streams are harder waters (Ca þ Mg > 1 mM) with pH above 7 (Table 1). The high Fe concentrations in the Kleine Nete indicate that the phreatic aquifers in this catchment are hosted by glauconite-rich deposits, which supply Fe to the groundwater upon weathering (Baken et al., 2013). The contrast between soft and hard waters coincides with the

contrasting size of the colloids in these waters, which were analyzed by AF4 (1 kDae150 nm) and membrane filtration (100e1200 nm) (Fig. 1, Table 2, Fig. S3, and Table S5). The colloids in the soft waters are dominated by small nanoparticles in the size range between 1 kDa and 40 nm. In contrast, the colloids in the harder waters are mostly larger (40e1200 nm). Due to the limited number of samples, however, it is not certain to what extent these observations can be generalized. Because of these contrasting properties, colloids in soft waters and harder waters are discussed separately. 3.2. Colloidal vectors in soft waters In the soft waters of the Roer and Schwarzbach, membrane filtration showed that no more than 14% of the Fe, and no more than 7% of Al, Ca, or Si is present in the 100e1200 nm size range (Table 2 and Table S5). Thus, the majority of the Fe falls within the sizerange of nanoparticles which have been characterized by AF4. The AF4 fractograms show that iron and humic substances (1 kDae40 nm) are the main constituents of the colloids (Fig. 1), which is in agreement with previous work on peat draining streams (Jirsa et al., 2013). The size distributions of nanosized Fe and of the UV signal (which indicates humic substances) are characterized by a main peak around 2e6 nm with a long tail extending to larger particle sizes (Fig. 1). The fractograms exhibit a step around 20 nm, i.e. in the region where the cross flow changed. This is a consequence of the fractionation procedure (details in Section 5 of the Supplementary Material). The step should not be interpreted as a distinct peak, but rather as the continuation of the signal under a decreased cross flow. Interestingly, the main UV peak maxima are detected at lower sizes (2e3 nm) than the main Fe peak maxima (3e6 nm). This suggests that Fe oxyhydroxides, and not complexes of Fe with humic substances, are the dominant Fe forms in this size range (Benedetti et al., 2003). This is supported by speciation calculations with WHAM7.0, which predict that 77e90% of the Fe(III) in these samples is present as colloidal hydrous ferric oxide and that the remainder is present as complexes with humic substances. Ferrihydrite, a common Fe oxyhydroxide in colloidal material from €stedt et al., 2013), occurs as nanosized primary natural waters (Sjo particles in the 2e10 nm size range (Hiemstra, 2013; Hochella et al., 2008), but larger particles are formed upon aggregation (Angelico et al., 2014). Probably, the main Fe peak in these samples (3e6 nm) consists of small primary ferrihydrite-like particles, while the tail of the size distribution extending towards large particle sizes (up to 150 nm) reflects aggregates of such primary particles (Neubauer et al., 2013a). In the Roer, the Fe signal exhibits a small secondary peak at around 80 nm, which may reflect such aggregates. The surprisingly large fraction of Fe oxyhydroxides (and low prevalence of Fe-humic complexes) implies that these nanoparticles can act as vectors of P (see Section 3.5). Up to 40% of the Fe in the two soft water samples was present as Fe(II) (Table S5). Oxidation of Fe(II) may to some extent have occurred during sample storage, even though Fe(II) may also be

86

S. Baken et al. / Water Research 99 (2016) 83e90

Fig. 1. Fractograms of selected elements (in nM) and the UV absorbance (arbitrary units) after field-flow fractionation of natural nanoparticles from streamwater. The data are plotted versus retention time (top axis) and hydrodynamic diameter (bottom axis). The step around 20 nm (1000 s) is a consequence of the decreased cross flow rate. Left: soft water samples (hardness <0.2 mM); right: moderately hard to very hard water samples (hardness >1 mM).

stable in oxic waters: in a streamwater sample very similar to the ones studied here, the Fe(II) was well preserved over time (Gaffney et al., 2008). We could not discriminate between Fe(II) and Fe(III) in nanoparticles, and therefore the speciation of the colloidal Fe(II) remains unknown. The Fe(II) may be complexed by humic substances, it may be adsorbed by Fe oxyhydroxides, but it can also be present as free Fe2þ ions in solution (Appelo et al., 2002; Catrouillet et al., 2014; Gaffney et al., 2008). The main UV peak, which reflects free humic substances, is followed by a secondary peak which coincides with the main Fe peak (Fig. 1). Most likely, the humic substances in this secondary

peak are associated with Fe oxyhydroxides. Humic substances may be incorporated in or adsorbed on nanosized Fe oxyhydroxides and thereby prevent aggregation of these nanoparticles (Eusterhues et al., 2014; Philippe and Schaumann, 2014). The nanoparticles also contain Al and Ca, but their concentrations are much lower than those of Fe (Fig. S3). The nanosized Ca may be complexed by humic substances, or it may be coprecipitated with Fe oxyhydroxides (Gunnars et al., 2002; Voegelin et al., 2010). Given the great similarity between the size distributions of Fe and Al, the nanosized Al may be present in association with the Fe oxyhydroxide nanoparticles, e.g. as impurities. Silicon, which is

S. Baken et al. / Water Research 99 (2016) 83e90 Table 2 The concentrations of Fe, Al, and P in different size fractions of colloids in streamwater samples, as measured by AF4 (1 kDae40 nm and 40e150 nm) and membrane filtration (100e1200 nm). The total concentrations in filtered (<1200 nm) samples are also shown. Size range

Fe

Al

P

mM

mM

mM

Schwarzbach

<1200 nm 1 kDae40 nm 40e150 nm 100e1200 nm

5.7 1.13 0.26 0.47

5.8 0.17 0.07 0.16

0.07 0.021 0.018 <0.07

Roer

<1200 nm 1 kDae40 nm 40e150 nm 100e1200 nm

5.2 2.1 1.00 0.73

2.9 0.37 0.17 0.22

0.20 0.06 0.04 <0.07

Kleine Nete

<1200 nm 1 kDae40 nm 40e150 nm 100e1200 nm

24 0.40 3.8 18

1.40 0.03 0.19 0.87

2.1 0.07 0.29 1.45

Meuse

<1200 nm 1 kDae40 nm 40e150 nm 100e1200 nm

0.89 0.04 0.20 0.46

1.00 0.02 0.17 0.41

1.87 0.04 0.07 <0.07

Dijle

<1200 nm 1 kDae40 nm 40e150 nm 100e1200 nm

1.72 0.01 0.06 1.45

0.31 <0.04 <0.04 <0.02

4.0 0.05 0.10 0.51

indicative of clay minerals, was not detected in these nanoparticles. The organic matter, Al, Ca, and Si in these colloids is discussed in detail in Section 5 of the Supplementary Material. In summary, the colloids in the two soft water samples consist of 1) free humic substances, and 2) nanosized associations between Fe oxyhydroxides and humic substances, with small amounts of Al and Ca. 3.3. Colloidal vectors in hard waters Contrary to the soft water samples, the Fe in the Kleine Nete, Meuse, and Dijle streams is mostly present between 40 and 1200 nm (Fig. 1 and Table 2). The small nanoparticles of these streams are dominated by free humic substances as evidenced by the sharp UV peak (maximum around 2e3 nm). This UV peak is associated with small Ca and Fe peaks, indicating Ca and Fe complexes with humic substances. In the Kleine Nete, Fe is the key constituent of the colloids in the 40e1200 nm size range. The UV signal exhibits a broad peak between 40 and 150 nm, indicating humic substances associated with the Fe-rich colloids. This latter interpretation is somewhat uncertain: due to the high Fe concentrations in these colloids, the correction of the UV-absorption for Fe was around 65%. The concentrations of colloidal Al are more than an order of magnitude below those of Fe, and no colloidal Si or Ca was detected, meaning that the Fe is present in oxyhydroxides rather than in clay minerals. A previous study identified ferrihydrite-organic matter aggregates with sizes between 1 and 100 mm in the Kleine Nete (Baken et al., 2013). These aggregates originate from in-stream oxidation of Fe(II), which is supplied by the groundwater. Together with the data presented here, this suggests that Fe-rich colloids and particles in the Kleine Nete consist of ferrihydrite-organic matter aggregates with a very wide size distribution, ranging from 40 nm up to 100 mm. In the Meuse, the 40e150 nm size range is dominated by large peaks of Al, Fe, and Si with very similar concentration profiles. The molar ratio of Si:Al is 1.3 across the whole 40e150 nm size range,

87

and 1.5 near the peak maximum. The Al and Si peaks are indicative of aluminosilicate minerals, most likely clay minerals. Previous field-flow fractionation studies have identified clay minerals in this size range in natural water samples (Baalousha et al., 2006; Chanudet and Filella, 2006; Regelink et al., 2013a). Iron is present in this size range at concentrations similar to those of Al and Si. Most likely, the Fe is present as part of the mineral lattice by isomorphic substitution and as a Fe oxyhydroxide coating on the surface of the clay minerals (Arias et al., 1995; Fontes, 1992; White, 2006). The coarse colloids (100e1200 nm) contain Fe and Al in approximately equal molar concentrations. The Si in this size range could not be detected by membrane filtration, but this is likely due to the relatively high detection limit for Si (the ICP-MS used for these measurements was different from the one coupled to AF4). However, since the 40e150 nm size range contains clay mineral nanoparticles, and since the coarse colloids contain Al, it seems likely that clay minerals with associated Fe oxyhydroxides were present in the Meuse across the entire 40e1200 nm size range. The large nanoparticles of the Dijle (40e150 nm) consist of Fe, humic substances, and Al. The Al concentration in this size range is much lower than that of Fe, and no Si or Ca was detected. After membrane filtration, only Fe was detected in the 100e1200 nm size range. Thus, the speciation of Fe in the 40e1200 nm size range is likely dominated by Fe oxyhydroxide aggregates. Since the concentration of colloidal Al is much lower than that of Fe, clay mineral colloids likely play a minor role in the Dijle. Some shortcomings of the methods used in this study must be noted. First, membrane filtration is prone to serious artifacts, and the 100e1200 nm size range should be considered as a rough estimate of the true size of these colloids (Gimbert et al., 2005; Horowitz et al., 1996). Second, not all nanoparticles may have been detected by AF4-ICP-MS. The Fe concentrations measured after AF4 are between 25 and 72% of those measured after membrane filtration with a nominal pore size of 100 nm (Table S5). This difference may be caused by small Fe species, such as Fe-fulvic acid complexes (Beckett et al., 1987; Benedetti et al., 2003), which may pass the AF4 membrane during focusing, or by large Fe colloids which did not elute before the fractionation was terminated. Alternatively, nanoparticles may be lost by adsorption to the AF4 membrane, an adverse effect which has been observed in previous AF4 studies (Vega and Weng, 2013). This causes some particles or particle types not to be detected. For this reason, the nanosized concentrations reported in Table 2 may be an underestimation of the true concentrations in that size range. Unfortunately, we were unable to determine to what extent adsorptive losses occurred in our AF4 experiment. Third, this study only involves samples from a limited number of streams (5) which were only sampled once. Therefore, one must be cautious to generalize the findings of this study, since it is not certain that our findings can be extrapolated to other streams or to different hydrological regimes. Fourth, the samples were stored for 2 weeks at 4  C prior to AF4-ICP-MS characterization. The properties of colloids may change if samples are stored for more than a couple of days (Buffle and Leppard, 1995b). For example, aggregation of colloids has likely occurred to some extent during storage, or part of the Fe(II) may have been oxidized to Fe(III), both of which potentially resulted in a shift to larger particle sizes. The P may have been processed by microbial activity, which also affects the size distribution. Therefore, the reported size distribution may differ from that in the original samples. Due to time constraints and to the extensive preparations involved in setting up and testing the AF4-ICP-MS system, we were unable to sample the streams shortly before analysis or to apply this technique to more samples.

88

S. Baken et al. / Water Research 99 (2016) 83e90

3.4. Colloids reflect the catchment geochemistry The nature and size of the colloids appear to be in agreement with the geochemistry of the catchments they originate from. The Schwarzbach and Roer drain acidic upland peat and receive soil percolates rich in humic substances, resulting in small associations between Fe oxyhydroxides and humic substances. The iron-humic substances colloids may be formed as water percolates through the soil, or they may be formed by in-stream oxidation of Fe(II) (Baken et al., 2015; Gaffney et al., 2008; Pokrovsky et al., 2005). The soft water (low ionic strength, low concentrations of divalent cations) and the high concentrations of humic substances largely prevent aggregation of these primary nanoparticles. Conversely, in the harder, more alkaline waters of the Kleine Nete, Meuse, and Dijle, the colloids are larger (40e1200 nm), which is likely because the hard water has higher ionic strength and concentrations of divalent cations, which promote colloid aggregation (Philippe and Schaumann, 2014). The Kleine Nete is predominantly fed by Fe-rich groundwater (Baken et al., 2013), which results in colloidal Fe concentrations more than an order of magnitude above those in the Meuse and Dijle. The high colloid concentrations in the Kleine Nete may be an additional factor explaining the relatively large size of the aggregates. The Meuse contains colloidal clay minerals, which are likely weathering products of the Ardennes mountain range where loamy soils developed on slate rock are common. The clay particles are transported by surface runoff, which is an important water flow path in this catchment due to impervious soils, hilly topography, and steep slopes in the Belgian part of the Meuse catchment (de Wit et al., ~ oz and Klaassen, 2006). We expected to find 2007; Murillo-Mun clay minerals in the Dijle as well, because it drains an area with calcareous loess soils where clay minerals (e.g. illite) are common and where surface runoff may transport such minerals to the streams. These clay minerals were not detected in this study, but they may have been present as >1200 nm aggregates due to the very hard water. 3.5. The association of P with colloids The two soft water streams (Schwarzbach and Roer) contain very low P concentrations (<0.2 mM). Little or no P is present in the 100e1200 nm size range, but part of the P is present in the form of nanoparticles (1 kDae150 nm; Table 2). In the Roer, the concentrations of nanosized P could be well detected, whereas in the Schwarzbach, they were close to the LOQ. For both streams, the size distributions of nanosized P exhibit a maximum at around 5 nm and a long, slowly decreasing tail which spans the entire size range up to 150 nm. As discussed before, the step around 20 nm is due to the changing cross flow and must be interpreted as the continuation of this tail. The overall size distribution of P is similar to that of Fe. Therefore, the Fe oxyhydroxide-humic substances associations are likely the main vectors of colloidal P, which is in agreement with two previous field-flow fractionation studies (Gottselig et al., 2014; Stolpe et al., 2010). The molar P:Fe ratios of the colloids, as calculated from the fractograms of both elements, were between 0.02 and 0.1. Apart from orthophosphate, part of the P in these oligotrophic, low-P systems may be present as organic P. However, we could not discriminate between both P forms in this experimental set-up. In the Kleine Nete, more than 80% of the P is bound to colloids in the 40e1200 nm size range. The size distribution of P measured by AF4-ICP-MS almost exactly matches that of Fe. In the large nanoparticles (40e150 nm), the molar P:Fe ratio is constant at 0.07 (Fig. 2) which equals that in the 100e1200 nm fraction. This suggests that the chemical composition of colloids in the 40e1200 nm

Fig. 2. Molar P:Fe ratios in nanoparticles from streamwater, as determined by fieldflow fractionation.

size range is more or less uniform and consists of P bound to Fe oxyhydroxide aggregates. The size distribution of P in the Meuse and Dijle exhibits a large peak between 50 and 100 nm and matches that of Fe fairly well (Fig. 1). In the Meuse, where Fe oxyhydroxide-clay mineral associations are the main colloids, the molar P:Fe ratio is between 0.3 and 0.5. Natural Fe oxyhydroxide colloids containing such high P concentrations have previously been identified in this size range (Gunnars et al., 2002; Lienemann et al., 1999). Laboratory studies have shown that P-bearing Fe oxyhydroxides or Fe hydroxyphosphate minerals may form upon oxidation of Fe(II) in the presence of P (Griffioen, 2006; Senn et al., 2015; van der Grift et al., 2014). Therefore, in the Meuse, the colloidal P is associated with Fe oxyhydroxide or Fe hydroxyphosphate-clay mineral colloids. In the Dijle, the molar P:Fe ratio in the 50e150 nm size range is around 1.5, which is above that of Fe hydroxyphosphate. This P must be present in another form, but no explanation is readily available. It was not bound to Al oxyhydroxides, clay minerals, or Ca phosphates, given the low concentrations of Al, Si, and Ca in this size range (Fig. S3). Possibly, this P may be present as phytic acid, which has 6 P atoms and strongly binds to Fe and Al oxyhydroxides (Turner et al., 2002; Yan et al., 2014). However, this explanation remains largely speculative. In the coarse colloids (100e1200 nm) of the Dijle, the P:Fe ratio is 0.4 (Table 2), indicating that the P in this size range may be bound to Fe oxyhydroxides or may be present as Fe hydroxyphosphate colloids. The colloidal P in the Dijle may therefore be bound by Fe-rich colloids, but another unknown P carrier may also be present. Overall, the molar P:Fe ratios of the nanoparticles measured in this study vary widely between 0.02 and 1.5, indicating strong differences in levels of P binding by Fe-rich particles. The partitioning of phosphate between dissolved and colloidal species in natural waters is controlled by the formation of a solid solution of Fe oxyhydroxide and Fe phosphate (Fox, 1989). Our results are generally in line with this view: the filtered (<1200 nm) samples of the Meuse and Dijle have a molar P:Fe ratio above 2. In these streams, the Fe-rich colloids almost contain as much P as they can possibly host, as evidenced by the high P:Fe ratio of the colloids (i.e. the colloids are nearly saturated with P, Fig. 2) (Voegelin et al., 2013, 2010). These colloids likely consist of Fe hydroxyphosphate (van der Grift et al., 2014), and most of the P is likely present as truly dissolved species (<1 kDa, Table 2), e.g. free orthophosphate anions. In contrast, the three other streams have molar P:Fe ratios below 0.1, and more than 50% of the P is colloidal. These colloids have low P:Fe ratios (Fig. 2) and likely consist of P bound to Fe oxyhydroxides.

S. Baken et al. / Water Research 99 (2016) 83e90

Overall, these data on a limited number of samples suggest that the partitioning of P between colloids and truly dissolved species is related to the P:Fe ratio in filtered water samples. Waters with a low P:Fe ratio have mostly colloidal P, whereas waters with a high P:Fe ratio mostly contain truly dissolved P because the Fe-rich colloids are saturated with P. Further research is needed in order to determine to which extent these findings on only 5 samples can be generalized. 4. Conclusions This study shows that AF4 coupled to HR-ICP-MS is a powerful tool to characterize P-bearing colloids in natural waters. Application of this technique to five contrasting streamwater samples revealed the presence of highly diverse colloids. The limited number of samples suggests that soft water streams mostly contain small colloids (3e150 nm). In contrast, harder waters contain larger aggregates (40e1200 nm) of Fe oxyhydroxides, humic substances, clay minerals, and possibly ferric phosphate minerals. The colloidal P is mostly associated with Fe oxyhydroxides, showing that Fe plays a key role in P transport in catchments. The molar P:Fe ratios of colloids vary between 0.02 and 1.5, indicating strong differences in level of saturation of Fe oxyhydroxides with P. In summary, colloidal Fe is predominantly present as Fe oxyhydroxides or as Fe hydroxyphosphates, and the size of these aggregates increases with increasing water hardness. Colloidal P is mostly associated with Fe rich colloids, but the P-saturation of these colloids strongly varies between different streams. The Fe rich colloids may increase the mobility or decrease the bioavailability of P, and they therefore need to be considered when addressing the transport and environmental effects of P in catchments. Acknowledgements Thanks to Peter Nobels (Wageningen University) and Kristin Coorevits (KU Leuven) for assistance. S.B. thanks the FWO-Research Foundation Flanders for a PhD fellowship. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.watres.2016.04.060. References Andersson, K., Dahlqvist, R., Turner, D., Stolpe, B., Larsson, T., Ingri, J., Andersson, P., 2006. Colloidal rare earth elements in a boreal river: changing sources and distributions during the spring flood. Geochim. Cosmochim. Acta 70, 3261e3274. http://dx.doi.org/10.1016/j.gca.2006.04.021. Angelico, R., Ceglie, A., He, J.Z., Liu, Y.R., Palumbo, G., Colombo, C., 2014. Particle size, charge and colloidal stability of humic acids coprecipitated with ferrihydrite. Chemosphere 99, 239e247. http://dx.doi.org/10.1016/ j.chemosphere.2013.10.092. Appelo, C.A.J., Van Der Weiden, M.J.J., Tournassat, C., Charlet, L., 2002. Surface complexation of ferrous iron and carbonate on ferrihydrite and the mobilization of arsenic. Environ. Sci. Technol. 36, 3096e3103. http://dx.doi.org/10.1021/ es010130n. Arias, M., Teresa Barral, M., Diaz-Fierros, F., 1995. Effects of iron and aluminium oxides on the colloidal and surface properties of kaolin. Clays Clay Min. 43, 406e416. http://dx.doi.org/10.1346/CCMN.1995.0430403. Baalousha, M., Lead, J.R., 2012. Rationalizing nanomaterial sizes measured by atomic force microscopy, flow field-flow fractionation, and dynamic light scattering: sample preparation, polydispersity, and particle structure. Environ. Sci. Technol. 46, 6134e6142. http://dx.doi.org/10.1021/es301167x. Baalousha, M., von der Kammer, F., Motelica-Heino, M., Baborowski, M., Hofmeister, C., Le Coustumer, P., 2006. Size-based speciation of natural colloidal particles by flow field flow fractionation, inductively coupled plasma-mass spectroscopy, and transmission electron microscopy/X-ray energy dispersive spectroscopy: colloids-trace element interaction. Environ. Sci. Technol. 40, 2156e2162. Baken, S., Nawara, S., Van Moorleghem, C., Smolders, E., 2014. Iron colloids reduce

89

the bioavailability of phosphorus to the green alga Raphidocelis subcapitata. Water Res. 59, 198e206. http://dx.doi.org/10.1016/j.watres.2014.04.010. Baken, S., Salaets, P., Desmet, N., Seuntjens, P., Vanlierde, E., Smolders, E., 2015. Oxidation of iron causes removal of phosphorus and arsenic from streamwater in groundwater-fed lowland catchments. Environ. Sci. Technol. 49, 2886e2894. http://dx.doi.org/10.1021/es505834y. € stedt, C., Gustafsson, J.P., Seuntjens, P., Desmet, N., De Schutter, J., Baken, S., Sjo Smolders, E., 2013. Characterisation of hydrous ferric oxides derived from ironrich groundwaters and their contribution to the suspended sediment of streams. Appl. Geochem 39, 59e68. http://dx.doi.org/10.1016/ j.apgeochem.2013.09.013. Beckett, R., Jue, Z., Giddings, J.C., 1987. Determination of molecular weight distributions of fulvic and humic acids using flow field-flow fractionation. Environ. Sci. Technol. 21, 289e295. http://dx.doi.org/10.1021/es00157a010. Benedetti, M.F., Ranville, J.F., Allard, T., Bednar, A.J., Menguy, N., 2003. The iron status in colloidal matter from the Rio Negro, Brasil. Colloids Surfaces A Physicochem. Eng. Asp. 217, 1e9. http://dx.doi.org/10.1016/S0927-7757(02)00553-8. Buffle, J., Leppard, G.G., 1995a. Characterization of aquatic colloids and macromolecules. 1. Structure and behavior of colloidal material. Environ. Sci. Technol. 29, 2169e2175. http://dx.doi.org/10.1021/es00009a004. Buffle, J., Leppard, G.G., 1995b. Characterization of aquatic colloids and macromolecules. 2. Key role of physical structures on analytical results. Environ. Sci. Technol. 29, 2176e2184. http://dx.doi.org/10.1021/es00009a005. Catrouillet, C., Davranche, M., Dia, A., Bouhnik-Le Coz, M., Marsac, R., Pourret, O., Gruau, G., 2014. Geochemical modeling of Fe(II) binding to humic and fulvic acids. Chem. Geol. 372, 109e118. http://dx.doi.org/10.1016/ j.chemgeo.2014.02.019. Chanudet, V., Filella, M., 2006. A non-perturbing scheme for the mineralogical characterization and quantification of inorganic colloids in natural waters. Environ. Sci. Technol. 40, 5045e5051. http://dx.doi.org/10.1021/es060255y. de Mars, H., Ransijn, M., Verbraak, P., Schuttelaar, M., Vercoutere, B., Buskens, R., €ntatie en project2000. Internationale ecologische verkenning Maas - Orie aanpak (fase 1). Report. Rijkswaterstaat directie Limburg, Maastricht, the Netherlands. de Wit, M.J.M., van den Hurk, B., Warmerdam, P.M.M., Torfs, P.J.J.F., Roulin, E., van Deursen, W.P.A., 2007. Impact of climate change on low-flows in the river Meuse. Clim. Change 82, 351e372. http://dx.doi.org/10.1007/s10584-006-91952. Eusterhues, K., Neidhardt, J., H€ adrich, A., Küsel, K., Totsche, K.U., 2014. Biodegradation of ferrihydrite-associated organic matter. Biogeochemistry 119, 45e50. http://dx.doi.org/10.1007/s10533-013-9943-0. Filella, M., Deville, C., Chanudet, V., Vignati, D., 2006. Variability of the colloidal molybdate reactive phosphorous concentrations in freshwaters. Water Res. 40, 3185e3192. http://dx.doi.org/10.1016/j.watres.2006.07.010. Fontes, M.P.F., 1992. Iron oxide-clay mineral association in Brazilian Oxisols: a magnetic separation study. Clays Clay Min. 40, 175e179. http://dx.doi.org/ 10.1346/CCMN.1992.0400206. Fox, L.E., 1989. A model for inorganic control of phosphate concentrations in river waters. Geochim. Cosmochim. Acta 53, 417e428. http://dx.doi.org/10.1016/ 0016-7037(89)90393-1. Gaffney, J.W., White, K.N., Boult, S., 2008. Oxidation state and size of Fe controlled by organic matter in natural waters. Environ. Sci. Technol. 42, 3575e3581. Giddings, J.C., 1984. Field-flow fractionation. Sep. Sci. Technol. 19, 831e847. http:// dx.doi.org/10.1080/01496398408068596. Gimbert, L.J., Haygarth, P.M., Beckett, R., Worsfold, P.J., 2005. Comparison of centrifugation and filtration techniques for the size fractionation of colloidal material in soil suspensions using sedimentation field-flow fractionation. Environ. Sci. Technol. 39, 1731e1735. http://dx.doi.org/10.1021/es049230u. Gottselig, N., Bol, R., Nischwitz, V., Vereecken, H., Amelung, W., Klumpp, E., 2014. Distribution of phosphorus-containing fine colloids and nanoparticles in stream water of a forest catchment. Vadose Zone J. 13 http://dx.doi.org/10.2136/ vzj2014.01.0005. Griffioen, J., 2006. Extent of immobilisation of phosphate during aeration of nutrient-rich, anoxic groundwater. J. Hydrol. 320, 359e369. http://dx.doi.org/ 10.1016/j.jhydrol.2005.07.047. Gunnars, A., Blomqvist, S., Johansson, P., Andersson, C., 2002. Formation of Fe(III) oxyhydroxide colloids in freshwater and brackish seawater, with incorporation of phosphate and calcium. Geochim. Cosmochim. Acta 66, 745e758. http:// dx.doi.org/10.1016/S0016-7037(01)00818-3. Haygarth, P.M., Warwick, M.S., House, W.A., 1997. Size distribution of colloidal molybdate reactive phosphorus in river waters and soil solution. Water Res. 31, 439e448. http://dx.doi.org/10.1016/S0043-1354(96)00270-9. Hiemstra, T., 2013. Surface and mineral structure of ferrihydrite. Geochim. Cosmochim. Acta 105, 316e325. http://dx.doi.org/10.1016/j.gca.2012.12.002. Hochella, M.F., Lower, S.K., Maurice, P.A., Penn, R.L., Sahai, N., Sparks, D.L., Twining, B.S., 2008. Nanominerals, mineral nanoparticles, and Earth systems. Science 319, 1631e1635. http://dx.doi.org/10.1126/science.1141134. Horowitz, A.J., Lum, K.R., Garbarino, J.R., Hall, G.E.M., Lemieux, C., Demas, C.R., 1996. Problems associated with using filtration to define dissolved trace element concentrations in natural water samples. Environ. Sci. Technol. 30, 954e963. http://dx.doi.org/10.1021/es950407h. Jarvie, H.P., Neal, C., Rowland, A.P., Neal, M., Morris, P.N., Lead, J.R., Lawlor, A.J., Woods, C., Vincent, C., Guyatt, H., Hockenhull, K., 2012. Role of riverine colloids in macronutrient and metal partitioning and transport, along an uplandlowland land-use continuum, under low-flow conditions. Sci. Total Environ.

90

S. Baken et al. / Water Research 99 (2016) 83e90

434, 171e185. http://dx.doi.org/10.1016/j.scitotenv.2011.11.061. Jirsa, F., Neubauer, E., Kittinger, R., Hofmann, T., Krachler, R., von der Kammer, F., Keppler, B.K., 2013. Natural organic matter and iron export from the Tanner Moor, Austria. Limnologica 43, 239e244. http://dx.doi.org/10.1016/ j.limno.2012.09.006. Jones, R.I., Shaw, P.J., De Haan, H., 1993. Effects of dissolved humic substances on the speciation of iron and phosphate at different pH and ionic strength. Environ. Sci. Technol. 27, 1052e1059. http://dx.doi.org/10.1021/es00043a003. Lienemann, C.-P., Monnerat, M., Dominik, J., Perret, D., 1999. Identification of stoichiometric iron-phosphorus colloids produced in a eutrophic lake. Aquat. Sci. 61, 133. http://dx.doi.org/10.1007/s000270050058. Lofts, S., Tipping, E., Hamilton-Taylor, J., 2008. The chemical speciation of Fe(III) in freshwaters. Aquat. Geochem. 14, 337e358. http://dx.doi.org/10.1007/s10498008-9040-5. n, B., Hassello €v, M., Turner, D.R., Haraldsson, C., Andersson, K., 2003. CompeLyve tition between iron- and carbon-based colloidal carriers for trace metals in a freshwater assessed using flow field-flow fractionation coupled to ICPMS. Geochim. Cosmochim. Acta 67, 3791e3802. http://dx.doi.org/10.1016/S00167037(03)00087-5. Mayer, T.D., Jarrell, W.M., 1995. Assessing colloidal forms of phosphorus and iron in the tualatin river basin. J. Environ. Qual. 24, 1117e1124. http://dx.doi.org/ 10.2134/jeq1995.00472425002400060010x. ~ oz, R., Klaassen, G.J., 2006. Downstream fining of sediments in the Murillo-Mun Meuse river. In: Ferreira, R.M., Alves, E.C.T.L., Leal, J.G.A.B., Cardoso, A.H. (Eds.), River Flow 2006, 1. Taylor & Francis, pp. 895e905. http://dx.doi.org/10.1201/ 9781439833865.ch94. €hler, S., von der Kammer, F., Laudon, H., Hofmann, T., 2013a. Effect Neubauer, E., Ko of pH and stream order on iron and arsenic speciation in boreal catchments. Environ. Sci. Technol. 47, 7120e7128. http://dx.doi.org/10.1021/es401193j. Neubauer, E., von der Kammer, F., Hofmann, T., 2013b. Using FLOWFFF and HPSEC to determine trace metal-colloid associations in wetland runoff. Water Res. 47, 2757e2769. http://dx.doi.org/10.1016/j.watres.2013.02.030. Neubauer, E., von der Kammer, F., Hofmann, T., 2011. Influence of carrier solution ionic strength and injected sample load on retention and recovery of natural nanoparticles using flow field-flow fractionation. J. Chromatogr. A 1218, 6763e6773. http://dx.doi.org/10.1016/j.chroma.2011.07.010. Notebaert, B., Verstraeten, G., Rommens, T., Vanmontfort, B., Govers, G., Poesen, J., 2009. Establishing a Holocene sediment budget for the river Dijle. Catena 77, 150e163. http://dx.doi.org/10.1016/j.catena.2008.02.001. Philippe, A., Schaumann, G.E., 2014. Interactions of dissolved organic matter with natural and engineered inorganic colloids: a review. Environ. Sci. Technol. 48, 8946e8962. http://dx.doi.org/10.1021/es502342r. , B., Schott, J., 2005. Fe-Al-organic colloids control of trace Pokrovsky, O.S., Dupre elements in peat soil solutions: results of ultrafiltration and dialysis. Aquat. Geochem 11, 241e278. http://dx.doi.org/10.1007/s10498-004-4765-2. Poulin, B.A., Ryan, J.N., Aiken, G.R., 2014. Effects of iron on optical properties of dissolved organic matter. Environ. Sci. Technol. 48, 10098e10106. http:// dx.doi.org/10.1021/es502670r. Regelink, I.C., Koopmans, G.F., van der Salm, C., Weng, L., van Riemsdijk, W.H., 2013a. Characterization of colloidal phosphorus species in drainage waters from a clay soil using asymmetric flow field-flow fractionation. J. Environ. Qual. 42, 464e473. http://dx.doi.org/10.2134/jeq2012.0322. Regelink, I.C., Weng, L., Koopmans, G.F., van Riemsdijk, W.H., 2013b. Asymmetric flow field-flow fractionation as a new approach to analyse iron-(hydr)oxide nanoparticles in soil extracts. Geoderma 202e203, 134e141. http://dx.doi.org/ 10.1016/j.geoderma.2013.03.015. Salminen, R. (Ed.), 2005. Geochemical Atlas of Europe. Part 1: Background Information, Methodology and Maps. Geological Survey of Finland, Espoo, Finland. Schindler, D.W., 2012. The dilemma of controlling cultural eutrophication of lakes. Proc. Biol. Sci. 279, 4322e4333. http://dx.doi.org/10.1098/rspb.2012.1032. Senn, A.-C., Kaegi, R., Hug, S.J., Hering, J.G., Mangold, S., Voegelin, A., 2015. Composition and structure of Fe(III)-precipitates formed by Fe(II) oxidation in water at near-neutral pH: Interdependent effects of phosphate, silicate and Ca. Geochim. Cosmochim. Acta 162, 220e246. http://dx.doi.org/10.1016/ j.gca.2015.04.032. Shaw, P.J., Jones, R.I., De Haan, H., 2000. The influence of humic substances on the

molecular weight distributions of phosphate and iron in epilimnetic lake waters. Freshw. Biol. 45, 383e393. http://dx.doi.org/10.1111/j.13652427.2000.00634.x. € stedt, C., Persson, I., Hesterberg, D., Kleja, D.B., Borg, H., Gustafsson, J.P., 2013. Sjo Iron speciation in soft-water lakes and soils as determined by EXAFS spectroscopy and geochemical modelling. Geochim. Cosmochim. Acta 105, 172e186. http://dx.doi.org/10.1016/j.gca.2012.11.035. Smith, V.H., 2003. Eutrophication of freshwater and coastal marine ecosystems: a global problem. Environ. Sci. Pollut. Res. Int. 10, 126e139. Stolpe, B., Guo, L., Shiller, A.M., Aiken, G.R., 2013. Abundance, size distributions and trace-element binding of organic and iron-rich nanocolloids in Alaskan rivers, as revealed by field-flow fractionation and ICP-MS. Geochim. Cosmochim. Acta 105, 221e239. http://dx.doi.org/10.1016/j.gca.2012.11.018. € v, M., 2010. Size and composition of colloidal Stolpe, B., Guo, L., Shiller, A.M., Hassello organic matter and trace elements in the Mississippi River, Pearl River and the northern Gulf of Mexico, as characterized by flow field-flow fractionation. Mar. Chem. 118, 119e128. http://dx.doi.org/10.1016/j.marchem.2009.11.007. Turner, B.L., Kay, M.A., Westermann, D.T., 2004. Phosphorus in surface runoff from calcareous arable soils of the semiarid western United States. J. Environ. Qual. 33, 1464e1472. http://dx.doi.org/10.2134/jeq2004.1814. Turner, B.L., Paphazy, M.J., Haygarth, P.M., Mckelvie, I.D., 2002. Inositol phosphates in the environment. Philos. Trans. R. Soc. B Biol. Sci. 357, 449e469. http:// dx.doi.org/10.1098/rstb.2001.0837. van der Grift, B., Rozemeijer, J.C., Griffioen, J., van der Velde, Y., 2014. Iron oxidation kinetics and phosphate immobilization along the flow-path from groundwater into surface water. Hydrol. Earth Syst. Sci. Discuss. 18, 4687e4702. http:// dx.doi.org/10.5194/hessd-11-6637-2014. van der Salm, C., van den Toorn, A., Chardon, W.J., Koopmans, G.F., 2012. Water and nutrient transport on a heavy clay soil in a fluvial plain in The Netherlands. J. Environ. Qual. 41, 229e241. http://dx.doi.org/10.2134/jeq2011.0292. Van Moorleghem, C., Six, L., Degryse, F., Smolders, E., Merckx, R., 2011. Effect of organic P forms and P present in inorganic colloids on the determination of dissolved P in environmental samples by the diffusive gradient in thin films technique, ion chromatography, and colorimetry. Anal. Chem. 83, 5317e5323. http://dx.doi.org/10.1021/ac200748e. Vega, F.A., Weng, L., 2013. Speciation of heavy metals in river Rhine. Water Res. 47, 363e372. http://dx.doi.org/10.1016/j.watres.2012.10.012. Viollier, E., Inglett, P.W., Hunter, K., Roychoudhury, A.N., Van Cappellen, P., 2000. The ferrozine method revisited: Fe(II)/Fe(III) determination in natural waters. Appl. Geochem 15, 785e790. http://dx.doi.org/10.1016/S0883-2927(99)00097-9. Voegelin, A., Kaegi, R., Frommer, J., Vantelon, D., Hug, S.J., 2010. Effect of phosphate, silicate, and Ca on Fe(III)-precipitates formed in aerated Fe(II)- and As(III)containing water studied by X-ray absorption spectroscopy. Geochim. Cosmochim. Acta 74, 164e186. http://dx.doi.org/10.1016/j.gca.2009.09.020. Voegelin, A., Senn, A.-C., Kaegi, R., Hug, S.J., Mangold, S., 2013. Dynamic Feprecipitate formation induced by Fe(II) oxidation in aerated phosphatecontaining water. Geochim. Cosmochim. Acta 117, 216e231. http://dx.doi.org/ 10.1016/j.gca.2013.04.022. Von der Kammer, F., Legros, S., Hofmann, T., Larsen, E.H., Loeschner, K., 2011. Separation and characterization of nanoparticles in complex food and environmental samples by field-flow fractionation. TrAC Trends Anal. Chem. 30, 425e436. http://dx.doi.org/10.1016/j.trac.2010.11.012. Wastiaux, C., Hallneux, L., Schumacker, R., Streel, M., Jacqmotte, J.-M., 2000. Development of the Hautes-Fagnes peat bogs (Belgium): new perspectives using ground-penetrating radar. Suo 51, 115e120. Weishaar, J.L., Aiken, G.R., Bergamaschi, B.A., Fram, M.S., Fujii, R., Mopper, K., 2003. Evaluation of specific ultraviolet absorbance as an indicator of the chemical composition and reactivity of dissolved organic carbon. Environ. Sci. Technol. 37, 4702e4708. http://dx.doi.org/10.1021/es030360x. White, R.E., 2006. Principles and Practice of Soil Science, fourth ed. Blackwell Publishing, Carlton, Australia. Yan, Y., Li, W., Yang, J., Zheng, A., Liu, F., Feng, X., Sparks, D.L., 2014. Mechanism of myo-inositol hexakisphosphate sorption on amorphous aluminum hydroxide: spectroscopic evidence for rapid surface precipitation. Environ. Sci. Technol. 48, 6735e6742. http://dx.doi.org/10.1021/es500996p.