Biogeochemical Cycling of Colloidal Trace Metals in the Arctic Cryosphere

Abstract

The surface waters of the Arctic Ocean include an important inventory of freshwater from rivers, sea ice melt, and glacial meltwaters. While some freshwaters are mixed directly into the surface ocean, cryospheric reservoirs, such as snow, sea ice, and melt ponds act as incubators for trace metals, as well as potential sources to the surface ocean upon melting. The availability and reactivity of these metals depends on their speciation, which may vary across each pool or undergo transformation upon mixing. We present here baseline measurements of colloidal (∼0.003–0.200 μm) iron (Fe), zinc (Zn), nickel (Ni), copper (Cu), cadmium (Cd), and manganese (Mn) in snow, sea ice, melt ponds, and the underlying seawater. We consider both the total concentration of colloidal metals ([cMe]) in each cryospheric reservoir and the contribution of cMe to the overall dissolved metal phase (%cMe). Notably, snow contained higher (cMe) as well as higher %cMe relative to seawater for metals such as Fe and Zn across most stations. Stations close to the North Pole had relatively high aerosol deposition, imparting high (cFe) and (cZn), as well as high %cFe, %cZn, %cMn, and %cCd (>80%). In contrast, surface seawater concentrations of Cd, Cu, Mn, and Ni were dominated by the soluble phase (<0.003 μm), suggesting little impact of cMe from the melting cryosphere, or rapid aggregation/disaggregation dynamics within surface waters leading to the loss of cMe. This has important implications for how trace metal biogeochemistry speciation and thus fluxes may change in a future ice-free Arctic Ocean.

Key Points

• Melt ponds act as a incubator for biotic and abiotic aggregation and disaggregation of colloidal trace metals

• There are higher proportions of colloids for Fe, Zn, Ni, Cd, and Mn in Arctic sea ice, snow, and melt ponds, compared to surface seawater

• There is colloidal Fe and Zn loss moving from snow to melt ponds and sea ice to underlying seawater during “incubation” in the cryosphere

Plain Language Summary

The Arctic Ocean is particularly vulnerable to climate change that results in increased freshwater inputs, such as river discharge and sea ice melt. This may also result in an increase or decrease in nutrient supply to the microorganisms living in the surface ocean who have metabolic requirements for survival. In fact, microorganisms need “trace” metals, found in very low abundances, that can be supplied by rivers and sea ice. However, the physical size of these metals may change how available they are to microorganisms. Here, we examine a size spectrum (∼0.003–0.200 μm) of important trace metals within snow, sea ice, melt ponds that form on sea ice, and the underlying seawater. These frozen reservoirs may serve as temporary “holding areas” for small trace metal particles to aggregate or dissolve, thus changing their availability or reactivity to microorganisms. Our goal was to assess what percentage of the overall sample is “small” (soluble <0.003 µm) versus “intermediate” (colloidal ∼0.003–0.200 µm) for each metal and how that changes between snow, sea ice, melt ponds, and seawater. As these frozen reservoirs decline in the Arctic it is important to understand their metal and nutrient supply to the surface ocean to predict how this may change in the future.

1 Introduction

The Arctic Ocean region is highly susceptible to a warming climate, particularly with respect to sea ice melt and riverine fluxes. Climate-induced changes to the hydrologic cycle are pronounced in the Arctic evident in trends of increasing water temperature, permafrost thaw, riverine discharge, and sea ice melt (Frey & McClelland, 2009; Macdonald et al., 2015; Peterson et al., 2002; Spencer et al., 2015). The Arctic Ocean, which already receives 11% of total global river water flux to just 1% of the world's ocean volume (Opsahl et al., 1999), will thus be acutely impacted by this freshening due to the isolated nature of the basin. Furthermore, freshening of the Fram Strait, the Arctic’s connection to the far North Atlantic, is expected to impart an influence of Arctic climate change onto the global ocean through increased freshwater discharge into the North Atlantic (Peterson et al., 2002). For example, negative salinity anomalies in the North Atlantic, explained by Arctic freshening, may slow the formation of deep water and thus the rate of global overturning circulation (Dickson et al., 2002; Karcher et al., 2012).

Importantly, these climate-driven changes to the Arctic are not just physical but also biogeochemical (Macdonald et al., 2015). Freshwater sources, such as rivers and sea ice are responsible for supplying essential nutrients to the central Arctic, where ice coverage is more extensive than in the productive, open coastal regions. Among these essential nutrients are trace metals, which can act as micronutrients by serving as metal cofactors in important metabolic enzymes in photosynthesis, nitrogen fixation, and carbon uptake/usage (Bruland et al., 2014; Sunda, 2012). Arctic rivers have high trace metal and nutrient loadings (Dai & Martin, 1995; Hölemann et al., 2005; Holmes et al., 2019) that can be transported offshore into the central Arctic via strong surface currents (Charette et al., 2020). Likewise, sea ice, which forms on the Arctic’s wide continental shelves and thus can be nutrient-rich and laden with shelf sediments, can supply a large flux of nutrients and trace metals (Aguilar-Islas et al., 2008; Hölemann et al., 1999; Tovar-Sánchez et al., 2010) to the surface Arctic upon melting (Measures, 1999).

The cryosphere, defined here as sea ice, snow, and melt ponds (Figure 1), is an important component of the Arctic freshwater inventory. As climate change accelerates both riverine discharge and sea ice melting, we can expect inventories of nutrients and metals in the cryosphere to change. These elemental reservoirs will be discharged directly into surface waters of the Arctic Ocean, which is expected to become fresher, more stratified, and less limited by light (Nummelin et al., 2016) as sea ice coverage is reduced. Additionally, less ice coverage may increase exposure to wind-driven mixing of intermediate waters (Lincoln et al., 2016; Rainville & Woodgate, 2009) and upwelling of subsurface nutrients. As a result of these chemical and physical changes, surface water phytoplankton community composition is also likely to respond, and the relative potential for primary production limitation by metal micronutrients could meet or exceed current limitations by light and/or nitrate (Arrigo et al., 2012; Rijkenberg et al., 2018; Taylor et al., 2013). Macronutrient and organic matter fluxes have been well reported in prior studies of the Arctic Ocean, but trace metal distributions are not as well studied, despite their importance in influencing productivity and phytoplankton community composition. For example, recent studies suggest that Fe may be presently limiting in the surface waters of the Nansen Basin (Rijkenberg et al., 2018) but not the Canada Basin (Jensen, Morton, et al., 2020).

Figure 1

Schematic representation of reservoirs in the Arctic cryosphere. Side view of snow, sea ice, melt pond, and the underlying seawater with major points of exchange and likely processes affecting colloidal trace metals. Modified from Marsay, Aguilar-Islas, et al. (2018). Note that the vertical scale is exaggerated here for illustration purposes.

Typically, trace metals are supplied to the surface ocean by aerosol deposition, riverine/freshwater discharge, or shelf sediment fluxes along continental margins. The fate and transport of trace metals is subject to changes in their physicochemical speciation, which includes both their size and their chemical composition. For example, a metal's speciation can affect its scavenging and aggregation fate with respect to settling (Cullen et al., 2006; Fitzsimmons et al., 2017; Jensen, Morton, et al., 2020; Wu et al., 2001) and/or its bioavailability to phytoplankton (Chen et al., 2003; Chen & Wang, 2001; Hassler & Schoemann, 2009; Hutchins et al., 1999; Rue & Bruland, 1995). Certain trace metal source fluxes carry unique metal physicochemical speciation that is different from the seawater into which they are deposited, such as riverine (Dammshäuser & Croot, 2012; Gledhill & Buck, 2012; Laglera et al., 2019; Oldham et al., 2017; Slagter et al., 2019) and aerosol inputs (Bergquist et al., 2007; Fitzsimmons & Boyle, 2014b; Fitzsimmons et al., 2015; Kunde et al., 2019). These external fluxes thus change the speciation of the receiving seawater and influence subsequent biogeochemistry (Laglera et al., 2019).

In today's Arctic, atmospheric deposition including snowfall often collects upon sea ice rather than being deposited directly into the surface ocean. In the summer, snow melt contributes to melt pond formation on the surface of sea ice before mixing with sea ice or seawater (Figure 1). Thus, snow, sea ice, and melt pond reservoirs act as “incubators” for elements, such as trace metals, over varying timeframes. During this incubation period, trace metal physicochemical speciation may change through biotic and abiotic particle interactions before the trace metals are released into the surface seawater. Sea ice, in particular, is host to many physical and biological processes that can affect aggregation and disaggregation of particles and colloids (Dieckmann et al., 2010; Janssens et al., 2016; Weeks & Ackley, 1986). The potential speciation transformations within these cryospheric reservoirs are the focus of this study.

Size speciation of dissolved trace metals is operationally defined here through ultrafiltration into a soluble or “truly dissolved” phase (<10 kDa, equivalent to ∼0.003 μm for globular proteins Erickson, 2009) and a colloidal phase (0.003–0.200 μm), which together add up to the total dissolved phase (<0.2 μm). Marine colloids are thought to be composed of macromolecules and are thus largely organic in nature (Guo & Santschi, 1997) but can also contain small amorphous or crystalline inorganic solids, which we call nanoparticles. The colloidal phase is thought to be an intermediary between the soluble (most bioavailable) and particulate (less bioavailable) phases (Honeyman & Santschi, 1989). Colloidal metals may be available to phytoplankton (Chen et al., 2003; Chen & Wang, 2001; Hassler et al., 2011); however, there are also studies showing that colloidal Fe, for example, is unavailable to phytoplankton (Wells, Mayer, Donard, et al., 1991) because of its nanoparticulate nature and/or its size resulting in slow diffusion to the cell surface (Chen & Wang, 2001; Rich & Morel, 1990; Wells, Mayer, & Guillard, 1991). Importantly, metals present in the colloidal size fraction need not be complexed with organic ligands, and metals present in the soluble size fraction can also include inorganic nanoparticulate species; thus, size partitioning by filtration does not define chemical complexation, and vice versa (Wilkinson & Lead, 2007). Complexation by organic ligands plays a vital role in maintaining sufficient concentrations of trace metals for optimal phytoplankton growth and also in preventing toxicity (Vraspir & Butler, 2009). While external sources of colloids to the ocean range from estuarine to sediment resuspension fluxes, the in situ oceanic cycling and aggregation-disaggregation dynamics of colloids dominate water column gradients in the marine colloidal metal distributions (Guo & Santschi, 1997; Santschi et al., 1997).

Across the global ocean, colloidal species of Fe have been studied more than those of other metals. Rivers and sediments are sources of colloidal Fe (cFe; Fitzsimmons & Boyle, 2014b; Wells, 2002), and in the riverine- and sediment-rich Arctic, we expected cFe species to dominate. However, results from the U.S. GEOTRACES GN01 study of the Western Arctic showed that cFe is a smaller fraction of dissolved Fe (dFe; 25 ± 16%) in the Arctic water column (Jensen, Morton, et al., 2020) than in the North Atlantic (50%–90%; Fitzsimmons et al., 2015). So, why are colloidal Fe concentrations so low in the Arctic, and does this trend extend to other metals? Overall, there is little data on colloidal metal distributions in the Arctic Ocean except in river estuaries, where colloids appear to be attenuated due to flocculation (Pokrovsky et al., 2014). In Western Arctic Ocean seawater, cFe appears to be enriched only on the shelf (65 ± 22%) and is likewise rapidly removed away from that source (Jensen, Morton, et al., 2020). Additionally, one study from the Eastern Arctic found %cFe ranges from 17% to 23% throughout the water column (Thuróczy et al., 2011). This suggests that either external sources to the Arctic have lower fractions of Fe colloids or alternatively generate fewer colloids post-input, or that perhaps longer incubation times, such as within the cryosphere, result in scavenging removal of colloidal metals.

Here, we use a unique and opportunistic sample set from the U.S GEOTRACES GN01 Western Arctic expedition. We measured the size partitioning of a suite of trace metals (e.g., Mn, Fe, Ni, Cu, Zn, and Cd) at six locations across three different cryospheric pools: snow, sea ice, and melt ponds, in addition to the underlying seawater (1–20 m). Previous work on this sample set has identified several potential processes occurring over time in the cryosphere (Marsay, Aguilar-Islas, et al., 2018), such as biological activity, particle scavenging, mixing of melt waters, and brine rejection (Figure 1; Weeks & Ackley, 1986). We apply chemical tracers of these various processes to compare their impact in different sample types to investigate the mechanisms driving the variability in dissolved metal size partitioning observed in the Arctic cryosphere. We aim to answer the following questions. (a) Do cryospheric reservoirs contain elevated concentrations of colloidal trace metals compared to seawater, and if so, for which metals? (b) What are the internal processes governing cMe transformations within the cryosphere prior to mixing into the underlying seawater?

2 Methods

2.1 Sample Collection

The U.S. Arctic GEOTRACES GN01 cruise left Dutch Harbor, AK on August 9, 2015 aboard the USCGC Healy, moving northward through the Bering Strait and Chukchi Shelf and across the Canada and Makarov basins along ∼170°W to the North Pole, before returning southward along 150°W to terminate sample collection on the Chukchi Shelf and return to Dutch Harbor on 12 October 2015. At six designated ice stations (Stations 31, 33, 39, 42, 43, and 46) occupied north of 80°N (Figure 2), trace metal clean samples were taken from snow (∼0.05 m³ volume), sea ice (first-year ice, upper 1 m), and underlying seawater (1, 5, and 20 m depth) through ice holes. At five of these six ice stations (Stations 33, 39, 42, 43, and 46), samples were also collected from melt ponds that had formed on the sea ice earlier in summer and had since frozen over. All ice stations were selected at locations where sampling was deemed safe for collection and in close proximity to a full water column sampling station for comparison.

Figure 2

Cryospheric sampling sites map, U.S. Arctic GEOTRACES GN01 cruise (2015). Relevant ice stations are shown as red dots, and full-water column stations are shown as smaller black dots (dissolved seawater data found in Jensen et al., 2019 for Zn, Zhang et al., 2020 for Cd, and Jensen, Morton, et al., 2020 for Fe and Mn). Snow, sea ice, and under-ice seawater samples were collected at Stations 31, 33, 39, 42, 43, and 46, and melt pond samples were taken at these same stations except Station 31. The gray arrow shows the trajectory of the transpolar drift where anything north of 85°N in this map is affected (Stations 31–43).

All plasticware was acid-cleaned prior to use, using established methods (Fitzsimmons & Boyle, 2012). Bulk snow samples at each of the six stations were collected for contamination-prone elements using an acid-cleaned high-density polyethylene shovel and were placed into an acid-washed low-density polyethylene (LDPE) bag. On board the ship, snow samples were melted, homogenized, and subsequently filtered through a 0.2 μm filter (Supor) and subsampled into 250 mL (for dissolved concentration analysis) and 500 mL (for ultrafiltration) LDPE bottles (Nalgene). Likewise, bulk sea ice was collected at each station using a trace metal clean ice corer and from each station four 1 m cores were homogenized in acid-cleaned LDPE melting chambers (Bolt et al., 2020; Marsay, Aguilar-Islas, et al., 2018). The meltwater was filtered through a 0.2 μm filter (Supor) into a 25 L carboy and subsampled into 250 and 500 mL LDPE bottles. Melt pond water was pumped into a 20 L carboy using a peristaltic pump and then subsampled on the ship (Marsay, Aguilar-Islas, et al., 2018). Finally, seawater was collected at 1, 5, and 20 m below the ice using a battery powered pump and Teflon-lined PVC tubing and was directly filtered (0.2 μm, Supor) into a 25 L carboy and subsampled into LDPE bottles. Information on depth, temperature, salinity, and chlorophyll a data from the snow and sea ice can be found in Table S1. It should be noted that the lack of sediment-laden sea ice in the samples, as well as only collecting the upper 1 m of sea ice, precludes consideration of the bottommost ice layer where sympagic algae is present (Bolt et al., 2020) and may bias our data.

2.2 Ultrafiltration

The subsamples from snow, sea ice, melt ponds, and seawater were ultrafiltered on board the ship within 2 h of the 0.2 μm filtration, in a trace metal clean environment. The full methods for the ultrafiltration are described in Jensen, Wyatt, et al. (2020) and Fitzsimmons and Boyle (2014a). Briefly, cross-flow filtration (CFF) employed a membrane with a 10 kDa pore size and a Masterflex pump with FEP tubing. The pump, filter, and tubing where all acid-cleaned and conditioned following the methods of Fitzsimmons and Boyle (2014a) and calibrated and conditioned with 100–500 mL of sample seawater, as volume allowed. Importantly, we required the permeate and retentate flows to be equal at 12.5 mL/min, with an overall pump flow rate of 25 mL/min. This means that the retentate solution was not returned to the feed water, effectively rendering this CFF method a “single-pass” ultrafiltration, utilizing back-pressure on the retentate flow to force permeate solution through the membrane. The permeate and retentate fractions were each collected in 60 mL LDPE (Nalgene) bottles following three 10% volume rinses of the bottles, caps, and threads. All samples were promptly acidified to pH < 2 (0.012 M HCl, Optima, Fisher Scientific).

Colloidal concentrations of each metal ([cMe]) were determined by subtracting the soluble metal concentration ([sMe]) derived from the 10 kDa permeate fraction from the overall dissolved metal concentration ([dMe]): [cMe] = [dMe] − [sMe] where Me = Mn, Fe, Ni, Cu, Zn, or Cd. In the text, we refer to %cMe to denote the contribution of colloids to the overall dissolved metal phase, which is calculated as:

(1)

(2)

The retentate solutions were also analyzed for their concentrations to assess total metal recovery through the CFF system. Since soluble-sized compounds are present in the retentate solution, and colloids are concentrated into this solution, we used the following equation to calculate recovery (Fitzsimmons & Boyle, 2014a):

(3)

where the CF = concentration factor was ∼2 for all samples but was determined accurately for each sample. Fitzsimmons and Boyle (2014a) previously showed that low Fe recovery in the CFF system was attributed to colloidal trapping within the mesh CFF membrane. Here, recoveries for all samples were assessed, and only samples where %Recovery > %sMe were used in this analysis, as otherwise poor ultrafiltration recovery could be confused with high colloidal loadings. For example, copper colloidal results for all the snow samples were eliminated by this criterion, and all metal data from the snow sample at Station 39 were also eliminated by low recovery. This agrees with a previous finding that %Recovery was often low and inconsistent across multiple metal colloidal fractions within samples frozen and then ultrafiltered (Jensen, Wyatt, et al., 2020), suggesting that the freezing/thawing process may bias our methods.

2.3 Analyses

Acidified samples sat for a minimum of 9 months at a room temperature before analysis, allowing for optimal desorption of metals from bottle walls (Jensen, Wyatt, et al., 2020). After this time, samples were initially pre-concentrated for dissolved and soluble trace metals using a SeaFAST-pico system (Elemental Scientific, Omaha, NE, USA) at Texas A&M University following the isotope dilution ICP-MS methods of Lagerström et al. (2013) adapted for offline analysis by Jensen, Wyatt, et al. (2020). Briefly, a 10 mL aliquot of seawater was loaded into the SeaFAST system, and buffered in-line to pH ∼6.3 with an ammonium acetate buffer (Optima, Fisher Scientific) while being loaded onto a column containing Nobias-chelate PA1 resin. After rinsing to release the major ion matrix, metals were back-eluted using 10% (v/v) nitric acid (Optima) into a 400 μL eluent (25x pre-concentration factor) for analysis. Eluents were subsequently analyzed in medium (Mn, Fe, Ni, Cu, and Zn) and low (Cd) resolution on a Thermo Element XR high-resolution inductively coupled plasma mass spectrometer at the R. Ken Williams Radiogenic laboratory at Texas A&M University. The accuracy, precision, and limits of detection for these elements are summarized in Table S2.

3 Results

Colloidal metals are known to aggregate or transform across salinity gradients (Gunnars et al., 2002), and thus it is useful to assess whether there is a pattern in %cMe across the cryosphere reservoirs at each sampling site to elucidate which of these reservoirs may supply colloids to the marine environment. While we do not have snow salinity data, Tables S1 and S3 show salinity data for sea ice (0–5.2) and for the melt ponds (1.4–25.76). For the purposes of this analysis, we have consistently ordered the reservoirs in each figure based on their approximate age, from snow (a “young” atmospheric source), to melt ponds (transient/heterogenous), and finally to sea ice and the underlying seawater. In so ordering the %cMe and (cMe) for each element, we noticed that %cMe decreased moving through the reservoirs from snow to melt pond to sea ice to seawater for Fe, Zn, Ni, and sometimes Cd and Mn. This was particularly evident at Stations 31, 33, and 46 (Figures 3-5, respectively).

Figure 3

Station 31 %colloidal = [colloid]/[dissolved] for each metal (left) and the size partitioned concentrations (right) across each cryospheric pool. Colloids are represented by black bars and the soluble phase by white bars. This station is an “aerosol-dominated” station where snow inputs of lithogenic metals were high, which was also true at Station 33 (Figure 4). No melt pond samples were collected at this station.

Figure 4

Station 33 %colloidal = [colloid]/[dissolved] for each metal (left) and the size partitioned concentrations (right) across each cryospheric pool. Colloids are represented by black bars and the soluble phase by white bars. Seawater [cMe] and %cMe is an average of samples from 1, 5, and 20 m. This station exemplifies the pattern observed in “aerosol-dominated” stations where snow inputs of lithogenic metals were high, which was also true at Station 31 (Figure 3).

Figure 5

Station 46 %colloidal = [colloid]/(dissolved) for each metal (left) and the size partitioned concentrations (right) across each cryospheric pool. Colloids are represented by black bars and the soluble phase by white bars. Seawater [cMe] and %cMe is an average of samples from 1, 5, and 20 m. This station exemplifies the background patterns observed at stations where neither aerosols nor seawater mixing appeared to exert a dominant influence.

Our first goal was to determine whether colloids were more abundant in cryospheric sources (snow, melt ponds, and sea ice) compared to the underlying seawater into which they are ultimately transferred, which would create the potential for the cryosphere to be a source of colloidal metals to seawater. Thus, we first examined the colloid contributions and spatial variability within the surface seawater itself (averaged across 1, 5, and 20 m depth). We found that %cMe were remarkably consistent within the upper 20 m of seawater at all six stations (Table 1), despite variable [dMe], indicating a characteristic pattern for surface Arctic seawater %cMe. Colloidal contributions in seawater ([cMe]_sw or %cMe_sw) were negligible for Mn and Cd (%cMn_sw = 0 ± 1% and %cCd_sw = 1 ± 1%), while colloids were significant fractions of the dissolved phase for Fe, Cu, Zn, and Ni (%cFe_sw = 38 ± 3%, %cCu_sw = 28 ± 2%, %cZn_sw = 26 ± 13%, %cNi_sw = 13 ± 2%), though colloids never dominated the dissolved Me concentration for any metal in this area. These results compare well with observed %cFe_sw across the surface ocean of the Western Arctic (40 ± 15%; Jensen, Morton, et al., 2020). Few published values exist for the %cMe_sw of Ni, Cu, Cd, and Mn in the open ocean, but our observations are consistent with previous coastal and open ocean studies (Jensen, Wyatt, et al., 2020; Roshan & Wu, 2015; Wen et al., 1999; Yang & Sañudo-Wilhelmy, 1998), while %cFe and %cZn appear to vary more with space and depth within the global ocean.

Table 1. Summary of %Colloids Within Each Reservoir (Snow, Melt Pond, Sea Ice, and Underlying Seawater From 0 to 20 m) and the Number of Measurements Incorporated From Each Pool (n)

Pool	n	Fe	Zn	Ni	Cu	Cd	Mn
Snow	6	74 ± 20%^a	70 ± 37%	48 ± 30%	–	79 ± 47%^a	56 ± 52%
Melt pond	5	67 ± 24%	90 ± 25%^a	34 ± 17%^a	41 ± 13%	25 ± 18%	4 ± 3%
Sea ice	6	70 ± 8%^a	52 ± 26%	43 ± 10%^a	21 ± 14%	38 ± 39%^a	13 ± 19%
Average cryosphere	17	71 ± 17% ^a	65 ± 27% ^a	42 ± 20% ^a	31 ± 17%	45 ± 40% ^a	25 ± 38% ^a
Seawater(0–20 m)	18	38 ± 3%	26 ± 13%	13 ± 2%	28 ± 2%	1 ± 1%	0 ± 1%

• Note. Percentage values are reported as mean ± 1SD. Average cryosphere values represent a combined average of snow, melt pond, and sea ice and associated ±1SD. The bolded “Average cryosphere” is the mean of the snow, melt pond, and sea ice values above.
• ^a %colloids is significantly greater than in the underlying seawater, using a two-tailed t-test (p < 0.05).

A second important conclusion is that %cMe was statistically greater in the cryospheric samples (snow, melt pond, and sea ice) than in the underlying seawater, as determined by a two-tailed t-test (Table 1, p < 0.05) for all metals except Cu, which had similar %cCu in seawater and cryosphere reservoirs. The %cMe were also, generally, much more variable in the cryosphere than in the surface seawater; note that the cryospheric %cMe standard deviations (SDs) were often 50% or more of the mean values (Table 1), indicative of the heterogeneity of our sample set. Analogous data with which we might compare our size-fractionated metal concentration results are very limited. However, studies in Antarctic pack ice found %cFe_ice and %cMn_ice averages of 75 ± 30% and 76 ± 19%, respectively (Lannuzel et al., 2014), which agrees well with our %cFe_ice value of 70 ± 8% for the Arctic but is significantly different from our Arctic %cMn_ice value of 13 ± 19% (Table 1), perhaps due to biogeochemical differences between pack ice (Antarctic study) and fast ice (this study). Overall, these Antarctic data reinforce our observation that colloidal contribution to the dissolved phase is larger and more variable in cryospheric reservoirs than typically found in seawater.

Stations 31 (Figure 3) and 33 (Figure 4) closest to the North Pole had received recent aerosol deposition that influenced their particulate metal concentrations (Bolt et al., 2020). Dust deposition to the area was estimated to be ∼500 μg/m²/day (Marsay, Kadko, et al., 2018). These two stations also had high concentrations of dissolved Fe and Zn in snow (>7 nmol/kg for dFe and >15 nmol/kg for dZn at both stations), with large colloidal contributions of >91% for %cFe_snow and >96% for %cZn_snow. This is in alignment with observations of higher %cFe_sw in surface seawater that has received recent aerosol deposition (Bergquist et al., 2007; Fitzsimmons & Boyle, 2014b; Fitzsimmons et al., 2015; Kunde et al., 2019). However, at these same stations, the colloidal concentrations and %cMe_sw were much lower in surface seawater (upper 1 m only), for both %cFe_sw (43% and 37% for 31 and 33, respectively), and %cZn_sw (0% and 36%, respectively), perhaps because the cryospheric samples had not yet mixed into the surface seawater below or due to differing biological processes between the relatively freshly deposited snow and the underlying surface seawater. Together, these increased aerosol fluxes to the snow create a clear decreasing trend of %cFe and %cZn, as well as a decrease in [cFe] and [cZn], moving from the snow, to the melt ponds and sea ice, and to surface seawater (Figures 3 and 4). Station 46 (Figure 5) shared this pattern of decreasing %cMe for Fe and Zn from snow to seawater, but it had much lower [cFe]_snow and [cZn]_snow (1.04 and 7.67 nmol/kg, respectively), and seawater (0.04 and 0.16 nmol/kg, respectively), likely due to lower recent aerosol input, although the %cFe_snow and %cZn_snow were still high (87% and 94%, respectively).

At these two aerosol-influenced stations (31 and 33), Ni, Cu, Mn, and Cd had high %cMe in the cryospheric reservoirs, relative to seawater, that like Fe and Zn decreased from snow to meltponds and sea ice, but unlike Fe and Zn these metals had lower [cMe] in the cryosphere compared to the underlying seawater, since their overall dMe concentrations were much lower in the cryosphere than in seawater (Figures 3 and 4). For instance, at Station 31 %cNi_snow was 93%, %cMn_snow was 96%, and %cCd_snow was 100%. The %cMe_sw values in the underlying seawater, in contrast, were a much lower 10%, 0%, and 0%, respectively. While this greater %cMe in the cryospheric sources could thus act as a source of colloidal metals upon melting, the trend in [cMe] told a different (and sometimes opposite) story because of the much larger dMe inventories in seawater than in the cryosphere for these metals. For example, [cNi] at Station 31 increased from 0.05 nmol/kg in the snow to 1.08 nmol/kg in the underlying seawater (Figure 3). At both Stations 31 and 33, [cMn]_snow and [cCd]_snow were higher than in seawater, but their total dissolved concentrations actually increased toward the seawater below, driven by an increase in smaller soluble-sized species (Figures 3 and 4). Copper appeared to show very little variation in the %colloidal space (∼20%–40% colloids across sea ice, melt ponds, and seawater), although this may be due in part to the lack of Cu size partitioning data for snow in this study.

Melt ponds at three other stations sampled—Stations 39, 42, and 43—were more “seawater-like” in that they had higher salinity (Figure 9; Table S3), suggesting greater seawater infiltrations (Marsay, Aguilar-Islas, et al., 2018). All stations (except Station 46) were within the trajectory of the Transpolar Drift (TPD) current that bisects the Arctic Ocean, primarily north of 84°N (Figure 2), and transports trace metals across the Arctic (Charette et al., 2020). This influence was reflected in the %cMe and [cMe] across all three cryospheric pools at Stations 39, 42, and 43 (Figures 6-8), which matched the underlying seawater much more than at the “aerosol-dominated” stations (31 and 33), especially for Fe and Zn. Rather than a decrease in %cFe or %cZn between snow, sea ice, and melt ponds, %cMe_snow were lower overall at these stations, which may be explained by a lower or less recent aerosol supply to these stations, as particulate metal concentrations were also lower in these cryospheric samples (Bolt et al., 2020). At these stations, the maximum %cMe was more likely to be found in the melt pond (Stations 39 and 42) or sea ice (Station 43) pools. For Fe, [cFe] appeared to increase from snow to melt pond to sea ice to seawater at these three stations, while [cZn] was more variable, with no clear pattern across the reservoirs (Figures 6-8). Likewise, for the other elements, broad patterns were less clear, although the dissolved fraction in the underlying seawater continued to be dominated by soluble-sized compounds, and seawater dissolved concentrations for Ni, Cu, Mn, and Cd were higher than for the cryospheric reservoirs (Figures 6-8).

Figure 6

Station 39 %colloidal = [colloid]/(dissolved) for each metal (left) and the size partitioned concentrations (right) across each cryospheric pool. Colloids are represented by black bars and the soluble phase by white bars. This station is a “Transpolar Drift-dominated” station, which was also true at Stations 42 (Figure 5) and 43 (Figure 8). Note there is no snow data for all elements except Fe at this station.

Figure 7

Station 42 %colloidal = [colloid]/(dissolved) for each metal (left) and the size partitioned concentrations (right) across each cryospheric pool. Colloids are represented by black bars and the soluble phase by white bars. Seawater [cMe] and %cMe is an average of samples from 1, 5, and 20 m. This station illustrates the pattern observed in “seawater-influenced” stations, which was also true at Stations 39 and 43 (Figures 6 and 8, respectively).

Figure 8

Station 43 %colloidal = [colloid]/(dissolved) for each metal (left) and the size partitioned concentrations (right) across each cryospheric pool. Colloids are represented by black bars and the soluble phase by white bars. This station is a “TPD-dominated” station, which was also true at Stations 39 (Figure 6) and 42 (Figure 7). Note that there is no Cu sea ice data due to poor %recovery.

Station 46 shared characteristics with other stations (Figure 5); however, its relatively low metal concentrations in snow suggest that it did not receive high aerosol inputs and that it is located outside of the bounds of the TPD. Both %cMe and [cMe] for elements such as Fe and Zn appear to decrease moving from snow to sea ice to melt pond to the underlying seawater. Even %cNi and %cCu appear to decrease but [cNi] and [cCu] do not. This station, absent of clear atmospheric or TPD influence, can be considered a “background” example of the Arctic cryosphere system in the Arctic with respect to colloidal metals.

4 Discussion

This study was guided by two major questions: (a) Do cryospheric pools have high %cMe and if so, for which metals, and why? (b) What are the internal processes governing cMe transformations within the cryosphere prior to mixing into the underlying seawater? To answer these questions, we began by grouping the stations by governing oceanographic processes and then evaluating the role of internal cycling within each reservoir and exchanges between reservoirs. Stations 31 and 33 (Figures 3 and 4) are more influenced by high/recent aerosol fluxes, while surface seawater at Stations 39, 42, and 43 (Figures 6-8) is more influenced by the TPD, and Station 46 (Figure 5) can be considered a “background” station for the Canada Basin. Note that this is a simplification for the purpose of identifying major trends; all stations receive some aerosol input and Stations 31, 33, 39, 42, and 43 were all within the bounds of the TPD. Additionally, we observed that metal groupings that shared similar trends in both %cMe and [cMe] across all stations were Fe and Zn, Ni and Cu, and Mn and Cd, which suggests that metal complexation chemistry and reactivity also plays a role in these patterns. We address each of our two driving questions in the Sections 4.1 and 4.2.

4.1 Do Cryospheric Pools Have Higher %cMe Than Seawater?

All metals except for Cu had significantly (p < 0.05) higher %cMe in the cryosphere compared to seawater (Table 1). Previous studies suggest concentrations of colloidal Cu and other trace metals are higher within low salinity environments such as estuaries than in seawater and are removed during flocculation (Dai & Martin, 1995; Gunnars et al., 2002; Sholkovitz et al., 1978). This study, by contrast, found colloidal Cu contributions were statistically indistinguishable in sea ice, melt ponds, and seawater (Table 1). Iron, which is thought to be >99% complexed by natural ligands in the ocean based on reactivity to added known ligands (Gledhill & Buck, 2012; Rue & Bruland, 1995), is also often found in the colloidal phase in both seawater and freshwater environments. Accordingly, we might expect Cu to follow a similar trend to that of Fe, as it is likewise significantly organically complexed (Moffett & Dupont, 2007; Vraspir & Butler, 2009). Instead, the relatively low and consistent %cCu here and in other similar studies (Jensen, Wyatt, et al., 2020; Roshan & Wu, 2018) suggests some equilibrium partitioning of Cu between soluble and colloidal phases, which would have to be facilitated by ligand exchange if our assumption of ligand complexation of dissolved Cu for seawater is also true for cryospheric waters. These ligands may be terrestrial in origin, transported in surface waters via the TPD (Slagter et al., 2017, 2019) and incorporated into melt ponds and sea ice, or biotic in origin as exopolymer saccharides and other bacterial byproducts are known to be abundant in Arctic sea ice (Krembs et al., 2002) and are also effective trace metal ligands (Hassler et al., 2011; Hassler & Schoemann, 2009).

Previous work using similar methods found that seawater dissolved Cd and Mn do not typically include a significant colloidal component, across multiple different environments, assessed using similar methods (Jensen, Wyatt, et al., 2020). Thus, it was surprising to find %cMe above even 5% for both Cd and Mn in snow, melt ponds, and sea ice. As described in Lannuzel et al. (2014), this is not without precedent for Mn in sea ice. Those authors attributed their observation of high %cMn_ice but low %cMn_sw to the aggregation and removal of colloids to the particulate phase as salinity increased, which is also observed in estuaries for Fe, Zn, Mn, Cu, and Ni (Dai & Martin, 1995; Hölemann et al., 2005; Sholkovitz et al., 1978). Alternatively, aggregation/disaggregation dynamics or even biological activity may serve to shuttle cMe back to the soluble phase. If this were the case we would expect that the [cMe] trend would follow the %cMe. However, for elements such as Cd and Mn, while cryospheric %cMe was high (particularly in snow), total [dMe] was relatively low, so [cMn] and [cCd] were relatively similar to the values in the underlying seawater, despite its low %cMe (<5%; Figures 4 and 5). Any aggregation that might result in altered concentrations of cCd and cMn in the cryosphere is thus unlikely to influence the underlying seawater.

A question remains, why is %cMe in snow high for many metals, including Cd and Mn, at Stations 31, 33, and 46? Particulate data from snow samples suggest high Fe and Al particle concentrations at Stations 31 and 33 (Bolt et al., 2020), indicative of high relative dust inputs. Air mass back trajectories support a European aerosol origin that may also contain higher levels of contaminants, such as Zn (Marsay, Kadko, et al., 2018). Previous studies in the North Atlantic (Bergquist et al., 2007; Fitzsimmons & Boyle, 2014b; Fitzsimmons et al., 2015; Kunde et al., 2019) and North Pacific (Aguilar-Islas et al., 2010) observed high %cFe_sw associated with elevated dust fluxes into the surface ocean. In the Arctic, dust inputs are lower than in the North Atlantic, and we did not observe an Arctic surface seawater peak in cFe at these three (ice-covered) stations (Table 1; Jensen, Morton, et al., 2020). However, snow that is, influenced by significant aerosol inputs appears to carry this higher colloidal load for Fe, Zn, Cd, and Mn. Arctic aerosols are known to range in size from 20 to 500 nm (0.02–0.5 μm) and tend to fall into the smaller range of 20–100 nm in the later summer and early fall (Freud et al., 2017). Thus, the higher %cMe_snow that we see across all elements in snow probably results from deposition of aerosols (>0.02 μm) into or with snow. Likewise, reversible scavenging of sorbed species by equilibrated particles may augment the dissolved phase in the form of colloids (Fitzsimmons et al., 2017).

In contrast to Stations 31 and 33, Station 46 snow did not have high particle loads that might indicate recent aerosol inputs but still had high %cMe. In addition, at this station %cMe_snow was higher for all metals than in the other cryospheric and seawater pools. One possible explanation is high organic loading. Arctic snow is rich in organic compounds that have both terrestrial and anthropogenic sources (Grannas et al., 2007) as a result of wet and dry deposition (Lei & Wania, 2004; Roth et al., 2004; Sempere & Kawamura, 1994). Additionally, labile dissolved organic carbon of terrestrial and marine origin is a known component of Arctic snow (Zhang et al., 2020). This could serve to promote soluble metal sorption and subsequent aggregation to the colloidal phase (Honeyman & Santschi, 1989; Wells, 2002; Wilkinson et al., 1997) or provide ligands in the colloidal phase that stabilize metal colloids, provide an adsorption surface for metals, and influence photochemical reactions occurring in the snow (Grannas et al., 2007) across all stations. Notably, a companion study of particulate metals in these GN01 cryosphere samples showed a progressive increase in particulate metal lability from snow to sea ice to seawater (Bolt et al., 2020). The authors suggest that labile particulate metals may have disaggregated to the dissolved phase during sample processing of snow, resulting in low measured lability of the remaining snow particles. Similar processes may be at play in our samples and will be discussed in the following section. Alternatively, the high %cMe for metals, such as Mn in snow and even sea ice (this study and Lannuzel et al., 2014), could be an artifact of the freezing process during sample deposition and subsequent thawing during collection (Bolt et al., 2020). This processing has been suggested to increase %cMe in laboratory experiments for metals such as Zn and Cd (Jensen, Wyatt, et al., 2020); unfortunately, we cannot currently sample snow's metal speciation in situ without a thawing treatment.

4.2 What are the Internal Processes Governing cMe Transformations Within the Cryosphere Prior to Mixing Into the Underlying Seawater?

We have shown above that sea ice and snow have elevated colloidal metal concentrations that could serve as a colloidal source to melt ponds and seawater upon mixing. However, does mixing between these reservoirs of different salinity, through a mechanism akin to flocculation in an estuary, actually drive aggregative/scavenging metal removal? To evaluate this, we could not rely on the individual station data above, where the extent of mixing between different pools could not be constrained. Instead, we must examine samples where we know how much seawater has mixed with cryospheric waters. Melt ponds are ideal case studies, as they constitute a unique reservoir where snow, sea ice, and seawater actively mix and incubate over various timescales, all the while influenced by scavenging onto particles and both phytoplankton and microbial processes (Marsay, Aguilar-Islas, et al., 2018). Previously, oxygen isotopes of water and salinity measurements in our five melt ponds were used to assign the volume contribution of sea ice, snow, and seawater in each melt pond (Marsay, Aguilar-Islas, et al., 2018). Here, we use those tracer-derived source partitioning estimates in a mass balance of our size-fractionated metal data for each cryospheric reservoir (end member) to investigate potential speciation transformations occurring within melt ponds.

First, we characterize the melt ponds using their salinity and tracer-based water volume apportionment (Table S3). The melt pond sampled at Station 33 had the lowest salinity (1.40) and highest percent contribution to volume from snow (94%; Marsay, Aguilar-Islas, et al., 2018) as well as %cMe_mp for Fe, Zn, and Cd (Figure 4). In contrast, Station 43 had the highest salinity (25.76) with a high contribution from seawater (79%) and accordingly presented more “seawater-like” %cMe_mp and [cMe]_mp (Figure 8). Together, these patterns support our hypothesis that mixing and incubation time spent within a melt pond may remove colloids via scavenging, biological uptake, and/or other processes.

We tested this hypothesis further by modeling the colloidal and soluble fractions within melt ponds based on the fractional contribution from each cryospheric and seawater reservoir (end member):

(4)

where f_SW is the fractional volume of melt pond (MP) water from seawater (SW), f_snow is the fractional volume of MP water from snow, and f_ice is the fractional volume of MP water from sea ice. This modeled colloidal concentration in the melt pond (hereafter referred to as [cMe]_model) can be compared to the measured melt pond concentrations ([cMe]_meas) in order to identify colloidal disaggregation and colloidal scavenging losses. If [cMe]_meas is lower than expected based on the [cMe] supplied by each pool ([cMe]_meas < [cMe]_model), then aggregation, biological uptake, or scavenging of colloids to the particulate phase or colloidal disaggregation to the soluble phase may have occurred. A similar analysis of the soluble metal mass balance would help confirm colloidal scavenging versus disaggregation in these cases. In contrast, if the [cMe]_model < [cMe]_meas, this would indicate that colloids are being generated within the melt pond from particle disaggregation or sorption of soluble metals onto colloids leading to aggregation. We assume in this exercise that the metal concentrations measured in each reservoir (snow, sea ice, and seawater) are representative of the waters that melted during the mixing that formed the melt pond. This maybe not be true for all melt ponds for reasons of either spatial (e.g., ice core depth in relation to the melt pond) or temporal variability. For example, it is worth noting that the snow that contributed to the melt ponds was from a previous melt season compared to the fresh snow sampled during this expedition, and thus the snow end members are likely an estimate. However, the pervasive metal-specific patterns that emerge below attest to the validity of this exercise.

The measured and modeled cMe and sMe data are tabulated in Table S4, and the melt pond salinity and difference between the measured and modeled values are summarized in Figures 9 and 10 in graphical form. From Figure 9, it is immediately evident that different elements have different characteristic behaviors during mixing of melt pond components: dissolved Fe and Zn (i.e., both sMe and cMe) are lost during incubation in the melt ponds, while dissolved Mn concentrations increase within the melt ponds, in both soluble and colloidal size fractions. Dissolved Ni and Cd have more intermediate behavior that suggests that they do not undergo significant aggregation or disaggregation during mixing in the melt ponds. We discuss these patterns in more detail below.

Figure 9

Table showing [cMe]_measured − [cMe]_model (first line) and [sMe]_measured − [sMe]_model (second line) for each station, with color to demonstrate the %difference between the modeled results and the measured value. Red indicates that the measured value is greater than the modeled, to varying degrees according to the color key. Blue indicates that the modeled value is greater than the measured, yielding a negative %difference according to the color key. Red indicates gain of cMe or sMe whereas blue indicates a loss in cMe or sMe over what we expect due to mixing of cryospheric reservoirs into melt ponds. Station 39 is excluded due to poor %recovery of the colloidal phase in the snow fraction for all metals.

Figure 10

Results of (colloid) and (soluble) concentrations in modeled (black) and measured (white) melt ponds for Stations 33, 42, 43, and 46. Melt pond 39 is excluded due to poor snow recovery, and Cu is omitted throughout for the same reason. Propagated error is shown as red bars in the positive direction only for space consideration. Error was assessed based on duplicate samples, long-term variability, and a baseline 10% error in fractions calculated by Marsay, Aguilar-Islas, et al. (2018).

First, we see that [cFe,cZn]_meas << [cFe,cZn]_model at all stations except Station 43 (for both cFe and cZn) and Station 46 (cFe only; Figure 9). The fact that this colloidal loss is accompanied by a soluble metal loss ([sFe, sZn]_meas < [sFe, sZn]_model) at Stations 33, 42, and 43 suggests strong biological uptake or particle scavenging from both soluble and colloidal phases to the particulate phase in the melt pond. Notably, Station 33 experienced a significant contribution from snowmelt (Table S3), which presumably also could have accumulated trace metals through dry aerosol deposition onto the snow, and thus uniquely at this station the “missing” dissolved Fe and Zn could have arisen from aggregation of aerosol-derived Fe and Zn due to high particle concentrations. The melt pond particle data reported in (Marsay, Aguilar-Islas, et al., 2018) indicate that Station 33 stood out as having the highest particulate Mn/P and Fe/P ratios, after lithogenic corrections, suggesting the presence of Mn and Fe oxides from shelf-derived particulate material, rather than algal material, that could promote scavenging. Lower particulate Al and Ti in the Station 42 melt pond indicated less contribution from crustal material and the particulate Fe/P and Mn/P ratios were more representative of cellular material (Marsay, Aguilar-Islas, et al., 2018), pointing toward biological uptake (instead of abiotic scavenging) as the mechanism of dissolved metal loss. Station 42 also exhibited loss of both soluble Cd and Ni, perhaps pointing to stronger biological uptake at that station. Arctic melt ponds are known to be biologically active and diverse, and productivity is often enhanced by exposure to light and nutrient supply from brine channels leading to high particulate organic carbon (POC) or chlorophyll a (chla) compared to surface seawater (Lee et al., 2012; Mundy et al., 2011; Sørensen et al., 2017). Without POC or chla concentrations at these melt ponds, we cannot trace biological uptake at the different stations with confidence.

Using the size-partitioned dissolved metal data, we had hoped to identify aggregation or disaggregation between soluble and colloidal compounds specifically, though this is challenging because the particulate metal inventories in some of these reservoirs were sufficiently large to affect the dissolved metal inventory (Bolt et al., 2020). An example of a soluble-colloidal metal exchange was observed at seawater-influenced Station 43, where ∼0.3 nmol/kg of [sFe]_meas (Schlitzer et al., 2018) was lost from the melt pond, which was approximately equal to the ∼0.4 nmol/kg “extra” [cFe]_meas observed (Figure 9). This, coupled with low particulate Fe (Marsay, Aguilar-Islas, et al., 2018), suggests possible sFe aggregation into cFe during incubation in that melt pond.

Station 33 showed the largest differences between [cMe, sMe]_model and [cMe, sMe]_meas in melt ponds, particularly for Fe and Zn, where we observed less cFe, cZn, and sFe, sZn than we expected based on mixing alone, and particularly for the colloidal size fraction. Station 33 had high aerosol deposition rates via snow (Kadko et al., 2018), and both %cMe and [cMe] were the highest at this station for metals such as Fe, Ni, Zn, Cd, and Mn, suggesting that colloid partitioning and abundance of these metals is linked to deposition rate or source. This melt pond also had the lowest salinity, so aggregation of colloids due to ionic strength effects on colloid double-layers would have been at a minimum. Particulate Fe and Zn in the melt pond at Station 33 both had high contribution from the crustal phase (Marsay, Aguilar-Islas, et al., 2018), which perhaps served to scavenge or aggregate these dissolved metals more efficiently at this station. Additionally, photochemical reduction could serve to solubilize colloidal or labile particulate Fe and Mn after deposition (Sunda et al., 1983; Sunda & Huntsman, 1994; Kuma et al., 1996).

Melt pond concentrations of soluble and colloidal Ni and Cd more closely followed conservative mixing of the end member concentrations, suggesting that little to no additional colloidal aggregation/disaggregation processes occur for these metals, though there were exceptions including Station 42 and Station 43 (Ni only). This major trend of conservative behavior aligns well with results from the combined dissolved phase that showed a correlation between salinity and dNi and dCd, indicating mixing was a dominant control (Marsay, Aguilar-Islas, et al., 2018). The soluble comparison points to production of sNi and sCd at Stations 43 and 46. This pattern contrasts with Fe and Zn, which largely show aggregation or scavenging of both phases across all melt ponds. As noted above, the unique observation at Station 42 of [sNi, sCd]_meas << [sNi, sCd]_model is likely due to biological uptake of sNi and sCd at this station.

Manganese provides a sharp contrast to the other elements in the melt pond mixing exercise (Figures 9 and 10). Both soluble and colloidal Mn, in contrast to Fe and Zn, was higher in the melt ponds compared to what was expected from conservative mixing of the cryospheric reservoirs ([sMn, cMn]_model < [sMn,cMn]_meas; Figure 9). Across all stations, this “extra dissolved Mn” was much more abundant in the soluble size fraction (3–18 nmol/kg more than expected) than in the colloidal size fraction (Figure 10), in line with low observed %cMn in these melt ponds. This scenario could only occur if particulate Mn was disaggregating or becoming solubilized into these smaller dissolved phases within melt ponds, as suggested in Marsay, Aguilar-Islas, et al. (2018). This is plausible as photoreduction can solubilize particulate Mn oxides in surface waters, such as melt ponds on the timescale of hours. This process would explain the higher measured concentrations of soluble Mn in melt ponds at all stations, relative to the mixing model (Figure 9; Marsay, Aguilar-Islas, et al., 2018; Sunda et al., 1983).

Differences between Fe and Mn in the melt ponds can be largely explained by differential behavior in scavenging and ligand stabilization. Cold temperatures, such as in an Arctic melt pond, may enhance the photoreduction of both particulate Mn and Fe species (Sunda & Huntsman, 1994, 2003). Any dissolved Fe formed as a result may be efficiently scavenged by remaining Mn oxides (Cheize et al., 2019), leaving behind a soluble Mn pool and a depleted dissolved Fe pool. As noted above, Station 33 had high particulate Mn and Fe concentrations suggesting increased Mn oxide formation or scavenging within the dissolved phase (Marsay, Aguilar-Islas, et al., 2018). Additionally, if ligands are present such that their concentration exceeds dFe, then when photoreduced Fe(II) is reoxidized, it can remain in solution as Fe-ligand complexes (Gledhill & Buck, 2012 and references therein). Instead, if the concentration of dFe exceeds the concentration of ligands available, then upon re-oxidation the dFe likely precipitates back into the particulate phase (Kuma et al., 1996). Without ligand data from any of the cryospheric reservoirs it is hard to move beyond speculation, but low ligand concentrations in melt ponds would also account for the observed difference between Fe and Mn in the melt ponds.

5 Conclusions

We used a unique dataset of snow, sea ice, melt pond, and seawater samples at six sites from the GN01 U.S. Arctic GEOTRACES section to address the following two questions: (a) Do cryospheric pools have higher %cMe than seawater? (b) What are the internal processes governing cMe transformations within the cryosphere prior to mixing into the underlying seawater? The answer to the first question is undoubtedly yes, as we found significantly higher %cMe in the cryospheric samples compared to underlying seawater for all metals except Cu (which was similar to %cCu in seawater). Aerosol deposition at Stations 31 and 33 was linked to higher %cMe in cryospheric reservoirs than at other stations. We also recorded metal-specific size partitioning trends, such as that Fe and Zn had highest %cMe in the cryospheric reservoirs (>50%). Even Mn and Cd, which do not have a significant colloidal fraction in seawater, had elevated %cMe, especially in snow. Arctic Ocean surface seawater, interestingly, had lower %cFe than in many other global ocean basins, supporting our hypothesis that cryospheric colloids may be removed prior to being delivered to seawater.

To answer the second question of what processes control colloidal partitioning during incubation in the cryosphere, we compared the modeled effect of mixing cryospheric reservoirs over an incubation time in melt ponds to our measured concentrations. This exercise confirmed that incubation of trace metals in melt ponds results in active transformations between the soluble, colloidal, and particulate metal phases, driven by processes such as soluble metal scavenging, abiotic colloid aggregation/disaggregation, photoreduction-driven dissolution, and/or biological cycling. These transformations were driven by particles present in the melt ponds: either crystalline particles (dust-derived or hydrogenous) that promoted abiotic aggregation, or cells that promoted biotic dissolved metal uptake. However, the dominant pattern that emerged from all five Arctic melt ponds was a characteristic (and unique) behavior of each metal. For example, soluble-sized Mn was released from particles in all five melt ponds, presumably from a particulate metal photoreduction processes, while soluble and colloidal Fe and Zn were both removed, perhaps by metal adsorption onto particles and/or a flocculation-like process during mixing of seawater and ice/snow in an organic-rich environment. Neither soluble nor colloidal Ni and Cd showed significant transformational trends in the melt ponds, showing either relative mass balance after melt pond mixing or, at most, some biological uptake. These conclusions are based solely on six stations within an ice-covered region of the Western Arctic Ocean sampled during late summer/early fall, and our observed spatiotemporal variability do suggest that geographic variability across the Arctic (with respect to major currents, dust inputs, etc.) and incubation time could result in more variability in metal colloid behavior and distribution across the wider Arctic cryosphere.

As climate change accelerates melting of the cryosphere, mixing between the reservoirs and with the underlying seawater will increase. Importantly, our findings indicate that high %cMe in snow, sea ice, and melt ponds at present does not translate into high %cMe in the surface seawater in the Arctic. Only dissolved Fe and Zn concentrations were high enough to potentially influence the surface seawater assuming conservative mixing, although their colloidal partitioning indicated aggregation during their incubation time in these cryospheric reservoirs. In a warmer future Arctic, we might therefore expect the direct delivery of more colloidal metals into surface waters, either by direct aerosol deposition into an ice-free Arctic seawater or by shorter incubation time in the cryosphere reservoirs where colloidal metals are removed prior to melting. This could have implications for both the bioavailability and residence time of micronutrient trace metals in future Arctic Ocean surface waters. This study provides a “snapshot” of colloidal dynamics in a late season cryospheric system, and future sampling efforts will ideally provide much-needed time-series measurements with increased spatial resolution. More data on dissolved and colloidal trace metals across the Arctic will help constrain residence times as well as the influence of important factors such as sea ice melt and snow deposition on metal speciation within the cryosphere and underlying seawater.

Acknowledgments

The authors would like to thank and acknowledge the Captain and crew of the USCGC Healy, Dave Kadko and Greg Cutter for proposing and enabling cruise leadership, and GN01 Supertechnicians Gabi Weiss and Simone Moos for sample collection at sea. The authors thank Luz Romero for assistance with ICP-MS analyses and maintenance and Angelica Pasqualini, Bob Newton, Peter Schlosser, and Tobias Koffman for contribution of their oxygen isotope measurements and freshwater model estimates used in the melt pond models. Additionally, this work would not have been possible without the SIO ODF team for salinity analyses. This work was supported by NSF Division of Ocean Sciences (OCE) award 1434493 and 1713677 to J. N. Fitzsimmons and R. M. Sherrell, 1438047 to C. S. Buck and W. M. Landing, and 1433717 to A. Aguilar-Islas and R. Rember. N. T. Lanning was funded through an NSF REU experience (NSF OCE 1455851) as well as through the NSF Graduate Research Fellowship award 1746932.

Open Research

Data Availability Statement

All dissolved metal data described above are available on the Biological and Chemical Oceanography Data Management Office website. These Research Data associated with this article can be accessed at: http://doi.org/10.1575/1912/bco-dmo.647259.4 (for salinity data), http://doi.org/10.26008/1912/bco-dmo.817259.1 (for dissolved metals). Stable oxygen isotope data used in this work can be found through EarthChem Library: https://doi.org/10.1594/IEDA/100633.