Introduction

Organic-rich soils in high-latitude boreal and arctic regions contain large reservoirs of soluble organic carbon (Laudon et al. 2004; Pokrovsky et al. 2015). The “colloidal world” of organic-rich waters has received thorough attention in non-permafrost (Andersson et al. 2001, 2006; Gundersen and Steinnes 2003; Vasyukova et al. 2010) and permafrost regions (Guo et al. 2003; Stolpe et al. 2013; Pokrovsky et al. 2011, 2013, 2016a). The bulk of DOM present in organic-rich surface waters, which comprises colloidal humic and fulvic substances, is dramatically different from the original soil organic matter, likely because of a combination of photochemical (Shiller et al. 2006; Porcal et al. 2015; Doane 2017) and microbial degradation (Bosch et al. 2010; Fritzsche et al. 2012; Oleinikova et al. 2017a). Both processes led to the removal of organic carbon (Corg) and trace elements (TE) via coagulation of amorphous iron (Fe) hydroxide, loss of carbon via CO2 emissions, and production of low-molecular-weight (LMW) fractions of Corg and TE (Wetzel 1995; Moran and Zepp 1997; Bertilsson et al. 1999; Brinkmann et al. 2003; Oleinikova et al. 2017a, b). Another important factor potentially transforming DOM is the freezing of shallow boreal and Arctic ponds and lakes (Arp et al. 2011; Manasypov et al. 2015). Even more intense organic colloid transformation may take place during the transfer of DOM from the active layer to the underlying permafrost with successive freeze-thaw cycles due to full freezing of the soil column (Shur et al. 2005).

The cryo-transformation of organic and organo-mineral colloids due to freezing and thawing of aqueous solutions may coagulate colloids, decrease the concentration of organic carbon (Spencer et al. 2007), change the molecular structure of DOM (Belzile et al. 2002; Chen et al. 2016; Xue et al. 2015, 2016), and precipitate insoluble Fe and aluminum (Al) hydroxides as well as silica- and aluminium-rich minerals (Wada and Nagasto 1983; Keung et al. 1984; Dietzel 2005; Kokelj and Burn 2005; Ireson et al. 2013). Previous studies on freezing of natural water to optimize DOM (Giesy and Briese 1978; Spencer et al. 2007) and nutrient (Ron Vaz et al. 1994; Fellman et al. 2008) preservation have demonstrated partial removal of DOC, phosphorus (P), and trace elements (TE) from the dissolved fraction (< 0.22 or 0.45 µm). Cryogenic concentrations of major ions during freezing at the soil surface (Levy et al. 2012) and within the active layer (Lundin and Johnsson 1994; Kokelj and Burn 2005; Jessen et al. 2014) have also been reported. However, to our knowledge, there has been no experimental evaluation of the transformation of organic and organo-mineral colloids during freeze and thaw cycles.

The overall aim of our study is to evaluate the possible transformation of dissolved and colloidal carbon and TE due to seasonal freezing and thawing of organic-rich surface waters in the high latitude boreal and Arctic inland waters. We quantified colloidal transformation by measuring concentrations of DOC and TE in different size fractions during subsequent freeze and thaw cycles in surface waters from two contrasting thaw ponds located in the discontinuous permafrost zone in northern Sweden. We hypothesized that (1) dissolved + colloidal (< 0.45 µm) organic carbon and TEs would be removed from solution via coagulation, (2) the precipitation of secondary mineral phases would depend on the initial saturation state of the fluid with respect to these minerals, and (3) the breakdown of large organic and organo-mineral colloids would lead to production of small-size organic molecules.

Materials and methods

Study sites

The two thaw ponds are situated in two nearby peat palsa complexes (Storflaket and Stordalen) located about 3 km apart within the Stordalen valley (68°20′N, 18°58′E) in northern Sweden. The palsa mires are characterized by a peat plateau with a homogeneous peat thickness of about 0.5 m overlaying silty lacustrine sediments of glacial origin (Klaminder et al. 2008). The mean annual air temperature in the region between 2000 and 2009 was 0.6 ± 0.4 °C (mean ± standard deviation). Annual precipitation was 340 ± 56 mm with average snow coverage lasting from October to May. All climatological data were recorded at the Abisko Scientific Research Station (www.polar.se/abisko). The maximum thickness of the active layer (i.e., the seasonally thawing zone in late September) was about 0.5 m in the hummocks and between 1 and 3 m in the hollows in both of the palsa mires (Åkerman and Johansson 2008, Johansson et al. 2013). Detailed descriptions of the two palsa mires are published elsewhere (Malmer et al. 2005; Alewell et al. 2011; Johansson et al. 2013 Thompson et al. 2015). The Stordalen thaw pond is acidic (pH 3.6) with low concentrations of calcium (Ca), dissolved inorganic carbon (DIC), magnesium (Mg), and strontium (Sr), as shown in Table 1. The Storflaket thaw pond is circumneutral (pH 6.9) and demonstrates a higher concentration of total dissolved solids (TDS). The two shallow thaw ponds (< 0.5 m deep; Fig. S1 of Online Resource 1) were rich in DOC (55 and 112 mg/l in Storflaket and Stordalen, respectively) and fully oxygenated at the time of sampling (Table 1).

Table 1 General physical and chemical dissolved (< 0.45 µm) parameters of the studied thaw ponds
Full size table

The Stordalen pond is representative for the large zone of frozen peat bogs, where active thermokarst processes cause peat thaw and small pond formation. The low pH, high DOC, and low TDS are very typical for hundreds of thousands of small thaw ponds and lakes in Western Siberia (Pokrovsky et al. 2011, 2013, 2016a; Shirokova et al. 2013; Manasypov et al. 2014, 2015 and Polishchuk et al. 2017) and elsewhere. The typical feature of all these ponds is that they are located within ombrotrophic bogs and receive no groundwater because of their geomorphologic position in the landscape and the frozen nature of the surrounding organic soil. The Storflaket pond is more similar to minerotrophic bogs in that it has a circumneutral pH and higher concentrations of major ions such as Ca and DIC. Such hydrochemical features are also typical of thaw slump-affected ponds and lakes in northern Canada (Kokelj et al. 2009) as well as small rivers and streams draining boreal and subarctic peatland regions like those in Western Siberia (Pokrovsky et al. 2015, 2016b). Together, these two studied ponds are generally representative of common, yet distinctly different types of surface waters in boreal and subarctic wetlands.

Sampling, size fractionation, and analyses

On 21 October 2016, after the first ca. 2 cm of ice formation (Fig. S1 of Online Resource 1), the thaw ponds were sampled and brought to the laboratory for filtration and colloidal separation within 1 h after sampling. All the analytical approaches used in this study followed methods developed for boreal DOM-rich waters (Pokrovsky et al. 2011, 2016a, b; Shirokova et al. 2013). The analytes are listed and defined by type in Table 1. Approximately 1 L of unfiltered surface water was collected in sterile pre-cleaned light-protected polypropylene bottles for 0.45- and 0.22-µm filtration followed by a centrifugal ultrafiltration (UF) through 50- and 3-kDa single-use Amicon Ultracell 15-ml cartridges [see the description of the preparation and cleaning procedure and discussion of possible artifacts of ultrafiltration in Oleinikova et al. (2017a)]. The centrifugal ultrafiltration was run at 4 °C using an Eppendorf 5920 R refrigerated centrifuge (4000 rpm, 20 min). Vacuum filtration (0.45 µm) was performed using a Mityvac MV8255 PVC hand pump. The filtration and ultrafiltration allowed us to separate two group of colloids: the high molecular weight (HMW50 kDa–0.45 µm) and low molecular weight (LMW3–50 kDa) as well as the low-molecular-weight fraction (LMW< 3 kDa).

The pH was measured at the in situ temperature of the ponds (4 °C). Chloride and sulfate concentrations were measured by ion chromatography (Dionex 2000i) with uncertainties of 2% and detection limits of 0.05 mg l−1. The concentrations of DOC and DIC were determined using a Shimadzu TOC-VSCN Analyzer with an uncertainty of 3% and a detection limit of 0.1 mg/l as derived for the specific instrument in the Géosciences Environnement Toulouse (GET) laboratory (Prokushkin et al. 2011). The UV absorbance of the filtered samples was measured at 254 nm using a quartz 10-mm cuvette on a Cary-50 spectrophotometer. The specific UV absorbency (SUVA254, l mg−1 m−1) was used as a proxy for the aromatic C, molecular weight, and source of DOM (Weishaar et al. 2003; Ilina et al. 2014 and references therein). All major and trace elements were measured with an ICP-MS Agilent 7500 ce using both argon (Ar) and helium (He) modes to diminish the interferences. Indium and rhenium were used as internal standards of the interference types at concentrations of ~ 3 µg/l. The typical uncertainty for elemental concentration measurement ranged from 5 to 10% at 1–1000 μg/l to 10–20% at 0.001–0.1 μg/l. The MilliQ field blanks were collected and processed to monitor any potential sample contamination introduced by our sampling and handling procedures. The DOC blanks of filtrates and ultrafiltrates never exceeded 0.1 mg/l. For all major and most trace elements, the concentrations in the blanks were below analytical detection limits (≤ 0.1–1 ng/l for Cd, Ba, Y, Zr, Nb, REE, Hf, Pb, Th, U; 1 ng/l for Ga, Ge, Rb, Sr, Sb; ~ 10 ng/l for Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, As). The international geostandard SLRS-5 (Riverine Water Reference Material for Trace Metals) was used to check the validity and reproducibility of analysis. All certified major (Ca, Mg, K, Na, Si) and trace element (Al, As, B, Ba, Co, Cr, Cu, Fe, Ga, Li, Mn, Mo, Ni, Pb, all REEs, Sb, Sr, Th, Ti, U, V, Zn) concentrations of the SLRS-5 standard and the measured concentrations agreed with an uncertainty of 10–20%. The agreement with SLRS-5 for Cd, Cs, and Hf was between 30 and 50%, although the uncertainty in the analyses of these elements was between 5 and 10%.

The tests of Fe(II) presence at the beginning and end of freezing cycles were performed using the conventional ferrozine method (Viollier et al. 2000), employing a standard addition technique, to account for the presence of high DOM concentrations. The uncertainty of this technique was ± 10% and the quantification limit 50 µg l−1 Fe(II).

Experimental freezing and thawing cycles

Sterile-filtered (< 0.22 µm) waters were subjected to full freezing at − 18 °C in the freezer (12–24 h) and thawing at 4 °C (typically 24 h) in the refrigerator. After each cycle, a subsample of water was collected in a sterile laminar hood box under continuous stirring with a magnet stir bar and processed for filtration (< 0.45 µm) or ultrafiltration (3 and 50 kDa) as described below. After complete melting of ice in the refrigerator, we measured pH at 4 °C in the laboratory. Each sample was subjected to ten freeze-thaw cycles run in duplicate. The ten cycles chosen in this study corresponded to the typical range of annual soil freeze-thaw cycles across the high-latitude boreal and Arctic zone, which is between 5 and 15 (Henry 2008). Limited volumes of initial samples precluded a larger number of cycles and did not allow detailed colloidal characterization after each freeze-thaw cycle. As such, only at the 0th, 3rd, and 10th cycles was the 0.45-µm-filtered solution separated into colloidal fractions (3 and 50 kDa) as described above.

Thermodynamic modeling of metal speciation in the course of freeze-thaw cycles in the colloids

Trace element (metal cations) complexation with organic matter in filtrates and ultrafiltrates was calculated using the Visual MINTEQ (vMinteq) computer code (Gustafsson 2011, version 3.0, for Windows) in conjunction with the NICA-Donnan humic ion-binding model (Benedetti et al. 1995; Kinniburgh et al. 1999; Milne et al. 2003) and the vMinteq (CRITICAL) database. The vMinteq implements a database for the binding of metal cations to discrete carboxylic sites of the humic and fulvic acids (Allison and Perdue 1994). The calculation is performed assuming a constant content of 10 µeq carboxylate per mg DOC (Oliver et al. 1983). We considered a ratio of DOM to DOC of 2 and that 100% DOM is fulvic acid. This approach works efficiently for boreal organic-rich waters (Vasyukova et al. 2010; Ilina et al. 2016; Pokrovsky et al. 2016a). The input parameters of the model were the in situ temperature, pH, DIC, DOC, anions, cations, and filtered (0.45 µm) or ultrafiltered (3 and 50 kDa) TE concentrations at the 0th, 3rd, and 10th freezing-thawing cycles. The model was run at 4 °C, which corresponded to the temperature of the solution before filtration and ultrafiltration and after complete melting of ice in the refrigerator. The model calculation yielded a theoretical degree of complexation for divalent metals [barium (Ba), Ca, cadmium (Cd), cobalt (Co), copper (Cu), Mg, manganese (Mn), nickel (Ni), lead (Pb), Sr, zinc (Zn), Al3+, FeIII, lanthanum (La3+), thorium (ThIV), and uranyl (UVIO22+)] with DOM and the degree of solution supersaturation with respect to possible authigenic minerals such as amorphous Al hydroxide, allophane, imogolite, ferrihydrite, and lepidocrocite, which most likely form under the soil and groundwater conditions in the high-latitude boreal and Arctic Sweden (Gustafsson et al. 1995; Lundström et al. 2000).

Results

Impact of freezing on the total dissolved concentration, colloidal composition, and low-molecular-weight fraction

We identified three groups of elements according to the change in element concentrations in filtrates of progressively decreasing pore size from 0.45 µm to 3 kDa: (1) strongly affected by decreasing pore size (> 5× concentration change): Fe, DOC, P, all trivalent and tetravalent hydrolysates, Pb2+, and UVIO22+; (2) moderately affected by decreasing pore size (between 2 and 3× concentration change): alkaline earth elements (Ca, Mg, Sr, and Ba), V, Cr, Mo, and As and divalent metals (Cu, Ni, Co, Zn, Mn, and Cd); (3) not affected (within 20–30% variation) by the size separation procedure: Li, B, Si, Na, K, Rb, and Sb. The evolution of the DOC, Fe, Al, and TE compositions of surface waters during filtration and ultrafiltration and their comparison with other boreal and subarctic settings are described in Supplement 2 (Online Resource 2).

The freezing yielded generally transparent ice on the walls of containers with concentrated brown ice in the center (Fig. S3-1 of Online Resource 3). After the first thawing, the water from Storflaket produced a brown/yellow precipitate at the bottom of the container. In contrast, no precipitate was visible in the Stordalen palsa mire water even after the tenth cycle of freezing and thawing. The pH and proportion of Fe(II) of both samples remained very stable during the freeze-thaw cycles (within ± 0.05 pH units and ± 5%, respectively). The variations of pH are illustrated in Fig. S3-2 of Online Resource 3. After ten freeze-thaw cycles, the concentrations of DOC (< 0.45 µm) decreased by ca. 25 ± 5% and 6 ± 1% in the Storflaket and Stordalen waters, respectively. The decreases in concentrations of Fe, Al, P, Ti, V, Ge, Y, Nb, Cd, REEs, Pb, Th, and U were between 40 and 90% of their initial concentrations in the Storflaket water but only 5–15% in the Stordalen water (Table 2). Chromium (Cr), Mn, Co, Ni, Cu, and Ba decreased by 15–30% in the Storflaket and < 10% in the Stordalen water. Note the sizeable removal of Si (75 ± 10%) in both waters. Examples of element concentration changes during freezing/thawing cycles of Stordalen and Storflaket waters are shown in Online Resource 3 (Fig. S3-3 and S3-4, respectively). Depending on their affinity to Fe or DOC within size fractions after each freezing and thawing, the elements are divided into two groups. In both Stordalen and Storflaket waters, Al, Ga, As, Ti, V, Cr, Y, Nb, Zr, Mo, REEs, Cd, Hf, Pb, Th, and U followed the behavior of Fe during freeze-thaw cycles, whereas alkaline-earth elements and divalent transition metals decreased their concentration concomitant to that of DOC. The separation into two groups was based on the degree of correlation (R2 of the linear relationship) between the elements and DOC and Fe (see details in Online Resource 2).

Table 2 Percentages of element removals from the water (< 0.45 µm fraction) after ten freeze-thaw cycles
Full size table

We observed a strong contrast in the organic carbon and trace metal concentrations in the course of freezing/thawing of two ponds. The Corg/Fe and Corg/Al remained fairly constant or only slightly (within 10%) decreased in Stordalen water (Fig. 1a) but increased by a factor of 6 ± 1 in the course of the freezing experiment of Storflaket water (Fig. 1b, c). The OC demonstrated net and clear enrichment in the low-molecular-weight fraction (LMW< 3 kDa) (Fig. 2a, b). The proportion of OC in the LMW< 3 kDa fraction increased from 23% at the beginning to 57% at the tenth cycle and from 19 to 47% for Stordalen and Storflaket mires, respectively (Fig. 3a, b). Further, in the LMW< 3 kDa fraction of DOM, the SUVA254 nm, which reflects the content of aromatic DOM, decreased from 4.0 to 2.4 in Stordalen and from 4.3 to 1.8 in Storflaket in the course of ten freeze-thaw cycles. At the same time, the decrease of SUVA254 nm in the 0.45-µm fraction was absent in Stordalen and constituted only 15% of the initial amount in Storflaket (Table 2).

Fig. 1
figure 1

The ratio of Corg to Fe (a, b) and Al (a, c) in the course of freeze-thaw cycles of Stordalen (a) and Storflaket (b, c) waters

Full size image
Fig. 2
figure 2

Concentrations of OC (a, b), Fe (c, d), Al (e, f), and Ti (g, h) in three fractions of thaw pond water at the beginning and after the third and tenth cycles of freezing. Left panels, Stordalen; right panels, Storflaket

Full size image
Fig. 3
figure 3

Percentages of elements in low- and high-molecular-weight fractions during freeze-thaw experiments. Element proportions in the dissolved low-molecular-weight fraction (LMW< 3 kDa) in Stordalen (a) and Storflaket (b) waters and the fractions of high-molecular-weight colloids (HMW50 kDa–0.45 µm) in Stordalen (c) and Storflaket (d) waters. (E) Cr, Mo, and V in LMW< 3 kDa fractions of Storflaket waters. (f) Si proportions in LMW< 3 kDa and high-molecular-weight colloid (HMW50 kDa–0.45 µm) fractions in Stordalen waters

Full size image

The distribution of major and trace elements between colloids of different sizes was heavily impacted by freezing and thawing cycles in neutral Storflaket water compared with acidic Stordalen water. Thus, the percentage of Fe and Al found in the LMW< 3 kDa fraction increased by a factor of 3 over ten cycles of freezing/thawing in the Storflaket waters, whereas this change in the Stordalen water was within  ± 10% of the total dissolved fraction (see Fig. 2c–f for concentrations and Fig. 3a, b for relative fractions).

The concentration of HMW50 kDa–0.45 µm colloidal forms of organic matter remained constant (Stordalen) or decreased (Storflaket) in the course of freeze-thaw cycles (Fig. 2a, b). The relative proportion of Fe, Al, and Ti in HMW50 kDa – 0.45 µm colloids systematically decreased in the course of freeze-thaw cycles in the Storflaket waters (Fig. 2c–h). At the end of the tenth cycle, this decrease relative to the initial proportion was a factor of 15, 13, and 40 for Fe, Al, and Ti, respectively (Fig. 3 D). At the same time, the HMW50 kDa–0.45 µm fraction of these metals only slightly decreased in concentrations during freeze-thaw cycles in the Stordalen waters (Fig. 3 C). Significant increases in the LMW< 3 kDa fractions of the oxyanions Mo, Cr, and V were observed in circumneutral Storflaket waters (Fig. 3e). Finally, there was remarkable formation of colloidal Si in Stordalen waters as the proportion of Si in the HMW50 kDa–0.45 µm fraction increased from 1.3% at the beginning to 43% after ten cycles and the proportion of Si in the LMW< 3 kDa fraction decreased from 95 to 25% (Fig. 3f).

Thermodynamic modeling: metal-DOM complexation and mineral saturation indices

The vMinteq model yielded metal cation complexation with DOM in both LMW< 3 kDa and dissolved (< 0.45 µm) fractions during the freezing cycles (Fig. S4 of Online Resource 4). The fraction of metal complexed to DOM is defined as the ratio of metal-DOM complexes to total dissolved metal (organic + inorganic forms). With this approach, we calculated the degrees of TE associations with DOM and possible changes in metal speciation during the experiment. In the Stordalen thaw pond (pH 3.6), organic complexes of Na and K accounted for 45% of total dissolved metals, whereas all other elements were fully bound to DOM (98.5 to 100% of organic complexes). Element speciation in Storflaket thaw pond (pH 6.9) showed that organic complexes of Na and K constitute < 10% of dissolved metals. Between 20 and 40% of alkaline earth elements (Ca, Mg, Sr, Ba) and 60–80% of Co, Zn, Cd, and Ni were complexed to DOM. Finally, 95–100% of UO22+, Pb, Cu, Fe(III), Al, La, Ce, and Th was complexed to DOM.

We used the calculated saturation indices of aqueous solutions with respect to possible solid phases to evaluate the potential of the fluids to precipitate insoluble minerals of Fe, Al, and Si during freezing (Table S4 of Online Resource 4). The Stordalen water was strongly undersaturated (SI < − 7) with respect to all possible secondary minerals including amorphous Al(OH)3, ferrihydrite, lepidocrocite, imogolite, and kaolinite regardless of the organic matter size class (< 3 kDa and < 0.45 µm) and the number of freezing cycles. The Storflaket water exhibited a different saturation state as the initial solution (prior to freezing) was oversaturated with respect to Al(OH)3, ferrihydrite, lepidocrocite, imogolite, and kaolinite (SI = 0.43, 1.3, 4.4, 1.0, and 3.6, respectively). The initial LMW< 3 kDa fraction of this sample was undersaturated (SI < − 1) with respect to all possible minerals except kaolinite (SI = 0.86). After freezing, the 0.45-µm fraction of Storflaket water equilibrated with ferrihydrite and kaolinite and was undersaturated with respect to Al(OH)3 and imogolite (SI = − 0.78 and − 1.92, respectively). Further removal of Fe and Al in the course of the experiment led to undersaturation with respect to ferrihydrite and kaolinite (SI = − 0.50 and − 0.52, respectively) after the tenth cycle of freezing/thawing.

Discussion

Chemical nature of colloids

The colloidal phase (3 kDa–0.45 μm) of the acidic thaw pond water at Stordalen and circumneutral thaw pond at Storflaket exhibited consistent organic carbon and TE concentrations (ca.  ± 10%) with the values reported for surface waters of the discontinuous permafrost zone (Table S2-1). The elemental composition of colloids, the decrease of organic C, Fe, Al, and TE concentrations during ultrafiltration (Figs. S2-1, S2-2, S2-3 of Online Resource 2), and Fe-normalized TE distribution coefficients between colloids and the LMW fraction (Fig. S2-4) show the similarities between our results for two thaw ponds at Abisko to previously investigated bog waters and adjacent lakes and rivers in other permafrost and non-permafrost regions. The trivalent and tetravalent hydrolysates and Pb2+ are incorporated into Fe(Al) hydroxides of high molecular weight, which are stabilized by OM. In contrast, LMW and colloidal organic complexes are responsible for speciation of alkaline earth and divalent transition metals. As such, we believe that the results of the freezing experiments of thaw pond waters in this study are representative of similar landscape features, which are widespread in high-latitude boreal and Arctic regions.

Production of the low-molecular-weight fraction

We show a sizeable production of low-molecular-weight OC (< 3 kDa) for both acidic and neutral organic-rich waters. This phenomenon has not been reported by earlier studies of frozen and thawed DOM-rich pond water (i.e., Giesy and Briese 1978). In Storflaket, the Corg increase in the LMW< 3 kDa fraction was due to generation of non-aromatic OM in the course of freeze-thaw cycles. It can be further hypothesized that the decrease in SUVA and the increase in LMW OC occurred because of removal of aromatic carbon and the appearance of aliphatic organic material during the freeze-thaw process. The molecular nature of the initial and newly produced OC fraction in permafrost-affected waters is unknown and requires special molecular-level DOM study. Analogous to peat waters in northern Scotland (Batchelli et al. 2009; Krachler et al. 2010, 2012) and organic-rich rivers of the temperate zone (Remucal et al. 2012), the OC fraction may include small humic molecules and oligomeric lignin phenols, aliphatic compounds, as well as small-size fulvic molecules from degradation of plant residues. In the course of freeze-thaw experiments, the generation of the LMW Corg fraction may occur via disintegration of HMW organic colloids. The breakup of large DOM molecules was accompanied by the removal of aromatic carbon from the LMW fraction as SUVA254 nm in LMW< 3 kDa decreased from the beginning to the tenth freeze-thaw cycle by a factor of 1.6 and 2.3 in Stordalen and Storflaket, respectively. Therefore, we conclude that the aromatic carbon was preferentially removed from the LMW fraction and aliphatic compounds were generated in the course of freeze-thaw cycles.

The appearance of LMW organic ligands within the hydrological continuum “depressions → rivers → ponds → lakes” has been reported in the discontinuous permafrost zone (Roehm et al. 2009; Pokrovsky et al. 2011, 2016a; Shirokova et al. 2013). The first process responsible for LMW OC generation is the photodegradation of soil-derived allochthonous DOM in lakes (Lindell et al. 1995; Roiha et al. 2012) and bogs (Oleinikova et al. 2017b) of the high-latitude boreal zone. The second process of LMW ligand generation in the surface waters of permafrost zones is the biodegradation of large organic colloids and HMW aromatic carbon of allochthonous origin by heterotrophic bacterioplankton (Sleighter et al. 2014; Shirokova et al. 2017; Oleinikova et al. 2017a). The present study adds the possibility of a third and previously unknown mechanism of LMW organic carbon and metal complex generation in surface waters—full freezing of shallow organic-rich water bodies. We hypothesize that this process may occur in all bog waters, both neutral and acidic, of the high-latitude boreal and Arctic regions and requires several thaw and freeze phenomena over the course of the year.

Mechanism of Fe, Al, and TE removal during freezing via coagulation of high-molecular-weight colloids

Thermodynamic modeling of the solution saturation state with respect to common secondary minerals of the boreal zone demonstrates the possibility of precipitation of Fe, Al hydroxide, and semi-amorphous aluminosilicates from the Storflaket circumneutral water. The decrease of dissolved (< 0.45 µm) Fe, Al, and Si during freezing-thawing cycles strongly suggests the on-going precipitation of secondary phases. In the acidic Stordalen water, we hypothesize that the increase in HMW colloidal Si (Fig. 3f) is due to formation of amorphous aluminosilicates. The aggregation of Fe hydroxide HMW colloids into particles and removal of dissolved Fe in neutral waters of Storflaket (pH 6.9) are consistent with the fact that the surface charge of Fe(III) hydroxide at these conditions is low (the pH of the zero charge point is close to 7.0, Haq et al. 1991), which can inhibit repulsive forces and favor aggregation (Carlson and Schwertmann 1981). In contrast, because of repulsion between positively charged Fe oxyhydroxide particles and colloids at pH 3.6 (Stumm 1992), the freezing of acidic Stordalen bog waters did not coagulate dissolved Fe.

The main cause of organic and organo-mineral colloid coagulation in thaw lakes and ponds is annual freezing of the whole water column, which leads to the concentration of Fe, organic carbon, and related divalent metals and trivalent and tetravalent hydrolysates in the particulate phase (Manasypov et al. 2015). A mechanism of TE removal during such coagulation is the incorporation of TEs into Fe oxy(hydr)oxides via coprecipitation during full freezing. It is known that two colloidal pools are responsible for TE speciation in boreal aquatic environments, namely, (1) small organic-rich colloids and LMW< 1 kDa DOM and (2) high-molecular-weight Fe-rich colloids (Baalousha et al. 2006; Stolpe et al. 2013; Krachler et al. 2012). The LMW colloids dominate metal speciation in acidic thaw pond water and are highly stable during freeze-thaw. The HMW colloids occur mostly in neutral mire/fen waters of the permafrost zone and are subjected to strong removal during water freezing. If small streams are fed by mire waters that are subjected to full freezing, the observed impact of freezing on the DOC and element concentration may partially explain the high variability of the chemical composition of small streams that drain peat mires. Therefore, we hypothesize that highly variable interannual patterns of DOC, Fe, Al, and TE concentrations in small streams draining mires and peatlands having similar lithologies and environmental settings (Pokrovsky et al. 2015, 2016b; Huser et al. 2012; Tiwari et al. 2016) may be partially due to different regimes of freezing and the numbers of freeze-thaw cycles that affect adjacent bog waters.

Possible transformation of surface water colloidal chemistry and organic carbon under changing climate and increasing freeze-thaw frequency

Across high-latitude boreal and Arctic zones, the frequency of annual soil freeze-thaw cycles is projected to increase significantly in both warm and dry winters by 2050 (Henry, 2008). Soil freezing has long been recognized to exert a strong impact on DOC and nutrients (i.e., Wang and Bettany 1993; Ron Vaz et al. 1994; Haei et al. 2010; Fitzhugh et al. 2001). Results of the present study suggest that the increase in the frequency of freeze/thaw cycles in surface waters will (1) impoverish the neutral (fen-like) waters of DOC, Fe, Al, and a number of low-soluble trace elements and (2) produce low-molecular-weight organic ligands and complexes with metals. LMW ligand production is expected at the beginning and end of winter because of repetitive freezing. In addition to shallow bog waters of the high-latitude boreal zone, a larger territory where the soil temperature fluctuates around zero is the discontinuous permafrost zone of western Siberian peatlands (e.g., Romanovsky et al. 2010). Therefore, interstitial soil solutions of Siberian peatlands may become enriched in LMW organic carbon during annual freezing/thawing cycles.

The production of LMW organic ligands has important consequences for Arctic ecosystems. The LMW fraction of DOM is expected to be more bioavailable and susceptible to microbial transformation than the HMW colloids, consistent with the results of microbial processing of natural colloids from different terrestrial and aquatic sources (cf., Roehm et al. 2009; Berggren et al. 2010; Yang et al. 2016). As a result, the climate warming in high-latitude boreal and Arctic regions may increase the delivery of the potentially bioavailable LMW fraction of carbon and micronutrients such as V, Mo, and Cr in the coastal zone because of simultaneous enhancement of the photo-, bio-, and cryo-transformation of organic and organo-ferric colloids in the inland waters. At the same time, freezing and thawing of acidic peat waters may increase the concentration of HMW colloidal Si and decrease the concentration of bioavailable LMW< 3 kDa Si.

Low-molecular-weight organic ligands increase as water moves from hollows to depressions and from thaw ponds to thermokarst lakes (Pokrovsky et al. 2016a). Indeed, large (terminal) lakes accumulate effluents of small temporary water channels, suprapermafrost waters, hollows, and depressions. These shallow (< 0.5 m) water bodies are subjected to intermittent freezing and thawing at the beginning and end of winter. We argue that the full freezing of water in hollows and depressions of the permafrost zone may generate the LMW fraction of DOC and affect delivery of metal micronutrients to the rivers and larger lakes. In particular, small Fe- and DOC-rich rivers with neutral pH may represent key locations for the preferential removal of Fe, Al, and some TEs by freezing.

Conclusions

This study demonstrated a systematic change in DOC and metal concentrations and colloidal composition in the course of freeze-thaw cycles of pond water from discontinuous permafrost peat mires. The high-molecular-weight colloids (50 kDa–0.45 µm) of organic carbon, Fe, Al, and other metals were preferentially removed from solution during freeze-thaw cycles. The impact of freezing on Fe hydroxide coagulation and TE coprecipitation was much smaller in acidic bog waters than in the circumneutral fen waters. This is consistent with the supersaturation of the latter with respect to Fe and Al oxy(hydr)oxides. The freezing of acidic bog water yielded an increase in HMW colloidal Si at the expense of the LMW< 3 kDa fraction. We show an increase in (1) the concentration of the low-molecular-weight (< 3 kDa) fraction of organic matter in water of thaw ponds and (2) the relative proportion of the LMW< 3 kDa fraction of Fe, Al, and micronutrients (Mo, V, Cr) in fen waters after several cycles of freezing and thawing. As climate changes in high latitudes, we expect an increase in LMW forms of organic matter and metal micronutrients due to simultaneous enhancement of photo-, bio-, and cryo-transformation of colloids in surface waters. In particular, the increase in the frequency of freeze-thaw cycles may increase the delivery of bioavailable carbon and micronutrients from high-latitude boreal inland waters to the Arctic Ocean.