Phosphorus Interactions with Iron in Undisturbed and Disturbed Arctic Tundra Ecosystems

Abstract

Phosphorus (P) limitation often constrains biological processes in Arctic tundra ecosystems. Although adsorption to soil minerals may limit P bioavailability and export from soils into aquatic systems, the contribution of mineral phases to P retention in Arctic tundra is poorly understood. Our objective was to use X-ray absorption spectroscopy to characterize P speciation and associations with soil minerals along hillslope toposequences and in undisturbed and disturbed low-lying wet sedge tundra on the North Slope, AK. Biogenic mats comprised of short-range ordered iron (Fe) oxyhydroxides were prevalent in undisturbed wet sedge meadows. Upland soils and pond sediments impacted by gravel mining or thermokarst lacked biogenic Fe mats and were comparatively iron poor. Phosphorus was primarily contained in organic compounds in hillslope soils but associated with Fe(III) oxyhydroxides in undisturbed wet sedge meadows and calcium (Ca) in disturbed pond sediments. We infer that phosphate mobilized through organic decomposition binds to Fe(III) oxyhydroxides in wet sedge, but these associations are disrupted by physical disturbance that removes Fe mats. Increasing disturbances of the Arctic tundra may continue to alter the mineralogical composition of soils at terrestrial-aquatic interfaces and binding mechanisms that could inhibit or promote transport of bioavailable P from soils to aquatic ecosystems.

Short abstract

This study evaluates how the nutrient phosphorus interacts with soil minerals in Arctic tundra ecosystems to infer potential controls on its bioavailability to plants and microorganisms.

Introduction

Climate change is rapidly altering biogeochemical processes in northern high-latitude ecosystems. Permafrost thaw is releasing vast stores of previously frozen organic matter that can be decomposed and emitted as climate-warming greenhouse gases (i.e., CO₂ and CH₄).¹ Increases in plant productivity caused by warmer temperatures could partially offset these C losses,¹⁻³ but plant growth in Arctic ecosystems may be inhibited by low nutrient availability.⁴ In particular, Arctic soils are often low in bioavailable phosphorus (P), namely inorganic orthophosphate anions, from occlusion of P in soil organic matter and the corresponding slow rates of P mineralization due to colder temperatures, i.e., enzymatic dephosphorylation of organic matter.⁵⁻⁸ Yet, a full understanding of the factors influencing P bioavailability in Arctic tundra ecosystems is limited due to knowledge gaps in the occurrence, distribution, and speciation of reactive soil minerals that could bind and sequester phosphate.

Ferric iron (Fe) minerals adsorb large quantities of P (primarily as inorganic phosphate)⁹⁻¹¹ and may potentially limit its availability to plants and microorganisms. However, Fe(III) minerals are susceptible to reductive dissolution and solubilization during hydrologically driven redox cycles that can release P from soils into aquatic systems.^12,13 The susceptibility of Fe(III) minerals to dissolution is dependent on the physical size, composition, and crystallinity of the underlying mineral structure.¹⁴ Phosphorus can also be immobilized by coprecipitation with calcium (Ca) carbonates and coprecipitation or sorption to aluminum (Al) oxides and clay minerals, potentially to a similar or greater extent than Fe oxides.¹⁵⁻¹⁷ Conversely, the weathering and dissolution of Ca and Al phosphate minerals could release significant amounts of P previously bound to those minerals.¹⁸⁻²¹ However, unlike Fe(III) (oxyhydr)oxides, the mechanism of dissolution of Ca carbonate, Al oxide, and clay minerals is not directly redox driven but more dependent on soil pH.^22,23 Thus, knowledge of the quantity and speciation of soil minerals that interact with P at terrestrial–aquatic interfaces is critical for evaluating the ability of soil at terrestrial–aquatic interfaces to be sources or sinks of bioavailable P, with implications for control of the ecosystem carbon balance in Arctic tundra.

Several studies have reported the widespread occurrence of Fe microbial mats on the surface of saturated, organic-rich tundra soils.²⁴⁻²⁸ These biogenic Fe mats form through the activity of Fe(II)-oxidizing bacteria, which colonize soil environments where opposing iron and oxygen gradients supply dissolved Fe²⁺ and limit dissolved oxygen. Fe(II)-oxidizing bacteria form thin (∼0.5–2 cm) coatings of short-range ordered Fe(III) oxyhydroxides (e.g., ferrihydrite) that comprise biogenic Fe mats.^26,29 Due to the high phosphate sorption capacity of biogenic Fe(III) oxyhydroxides,⁹ biogenic Fe mats could form an unrecognized sink and barrier to P transfer across terrestrial-aquatic interfaces, which may contribute to P limitation either in soils or in adjacent aquatic systems. Although Fe cycling in Arctic systems has gained increasing attention,^27,30,31 there remains limited knowledge regarding the geochemical composition of biogenic Fe mats, their distribution on the landscape, and their ability to bind inorganic phosphate and potentially limit its bioavailability. To address this knowledge gap, we compared the geochemical compositions of organic-rich surface soils in upland, midslope, and lowland positions of two hillslope toposequences to biogenic Fe mats formed at terrestrial–aquatic interfaces of low-lying wet sedge tundra in the Alaska North Slope region. We also evaluated pond sediments in areas impacted by gravel mining and retrogressive thaw slumps to explore whether anthropogenic disturbance could influence the formation of biogenic Fe mats and consequently, the major pools of P present in Arctic surface environments.

Methods

All work for this study was conducted in the field (sample collection) or laboratory (bulk geochemical and X-ray characterization) and no unexpected or significant safety hazards were encountered.

Site Description

The Toolik Field Station (68.66109, −149.37047, mean elevation 820 m) is located on the North Slope of Alaska. Maps of the Toolik region and individual study areas are shown in Figure 1. The region is underlain by continuous permafrost (∼200 m depth) with a shallow active layer (∼10–75 cm) that develops annually during the summer season.³² Soils in this region are either acidic (pH < 6) and depleted in base cations (Ca, Mg, Na, K) or nonacidic (pH > 6), depending on the time since glaciation.³³ Land surfaces in the Toolik region are characterized by topographic-moisture gradients that span upland moist tussock tundra to relatively dry evergreen heath hilltops and lowland wet sedge meadows. The upland tussock tundra consists of thick organic soils with a maximum thaw depth of ∼50 cm. The wet sedge lowlands are dominated by meadows of nontussock forming sedges (e.g., Carex spp., Eriophorum spp.) and series of interconnected ponds (<2 m depth) approximately 10–15 m apart that span from a few meters to tens of meters across.^24,34,35 The wet sedge meadows generally have thick organic horizons (∼50 cm) that do not thaw to the mineral layer. The average gradient of vertical relief from tussock uplands to wet sedge lowlands is approximately 5–7 m over 50–250 horizontal meters.³² To evaluate the variation in soil geochemical properties along typical hillslope toposequences in the Toolik region, we collected organic soils from upland, midslope, and lowland positions along an acidic and a nonacidic hillslope located on the east and west side of Toolik Lake, respectively (Figure 1a).

Figure 1.

(a) Overview of all sampling locations with inset maps for (b) Toolik Lake Inlet (TLI) and (c) Oksrukuyik Creek (OKS) watersheds. (a) Hillslope transects and retrogressive permafrost thaw slumps are individually marked. (b, c) Mining-disturbed regions are indicated by the dotted line. A detailed list of all sampling locations is in Tables S1 and S2.

Biogenic Fe mats (n = 12) were collected from saturated wet sedge meadows within two small watersheds in the Toolik region (Figure 1a): the Toolik Lake Inlet (TLI) watershed and the Oksrukuyik Creek (OKS) watershed. Wet sedge meadows in the TLI watershed drain through a series of randomly distributed ponds directly into Toolik Lake. The OKS watershed (Figure 1c) is approximately 15 km ENE of the Toolik Field Station. Wet sedge meadows in the OKS watershed comprise a series of interconnected ponds that are likely precursors to a beaded stream that drains into Oksrukuyik Creek and ultimately into the Sagavanirktok River.³⁶ We collected “submerged” biogenic Fe mats from submerged plant materials in saturated soils and “pond” mats from sediment–water interfaces of small ponds within the wet sedge meadows. Pond sediments were collected from an additional four ponds in which biogenic Fe mats were absent, as confirmed by optical microscopy and DNA sequencing in Michaud et al.²⁴ Three of these ponds were formed on land surfaces excavated in the 1970s to extract mineral-rich gravel for construction of the Dalton Highway (notated as “mining-disturbed ponds” in Figure 1a). The fourth location (pond 4) was a pond with surface sediments dominated by a benthic diatom species. To our knowledge, there have been no disturbance events associated with pond 4. To evaluate natural disturbance events, surface sediments were collected from mud slurries forming adjacent to the thawing front of two retrogressive thaw slumps: Lake 89 and Wolverine (Figure 1a, Table S1).

Sample Collection

All biogenic Fe mats were collected in June and July 2019, except for one sample collected in August 2021 (Table S1). Five submerged and six pond biogenic Fe mats were collected in addition to surface sediments from three mining-disturbed ponds. Biogenic Fe mats were collected by aspirating the mat material with a sterile 25 mL serological pipet and transferring it into a sterile 15 mL conical centrifuge tube. Pond sediments in areas disturbed by gravel mining were collected by directly scooping out the top 1 cm of surface sediments with a sterile 50 mL conical tube. Retrogressive thaw slump samples were collected 5 m from the thaw slump headwall using a sterile 50 mL conical tube. All samples described above were preserved by freezing at −20 °C. The conical tubes containing the samples were vacuum-sealed in a plastic bag containing an oxygen-scrubbing packet (AnaeroPack, Mitsubishi Chemical Corp.) to prevent oxidation during shipment to Bigelow Laboratory for Ocean Sciences. Hillslope soil cores (<20 cm depth) were collected in July 2021 from upland, slope, and lowland positions along hillslopes located in acidic (pH < 6) and nonacidic (pH > 6) tundra on either side of Toolik Lake. Intact cores were separated by the horizon (surface and subsurface organic) based on visual differences in texture and then frozen for shipment to the University of Tennessee. Sample information for all samples, including coordinates and collection dates, is in Tables S1 and S2.

Sample Preparation

The conical tubes containing biogenic Fe mat, disturbed pond sediment, and retrogressive thaw slump samples were transferred into a vinyl anaerobic chamber (5% H₂, 95% N₂, <1 ppm of O₂; Coy Laboratory Products) and allowed to thaw at room temperature. The thawed samples were centrifuged at 5000g for 15 min to concentrate the solid material, and the pellet was stored inside the anaerobic chamber in a desiccator containing freshly rejuvenated desiccant. The pellets were allowed to dry for 3 weeks (desiccant refreshed weekly) then sent to the Canadian Light Source for X-ray absorption spectroscopy (XAS) analysis (see below). Intact cores of surface and subsurface organic horizons were freeze-dried then homogenized by horizon prior to Fe XAS and bulk geochemical analysis. Phosphorus speciation was determined on subsamples of freeze-dried soils collected before homogenization; therefore, P speciation for hillslope soils represents P species present in the soil but may not be representative of the bulk soil composition.

Bulk Geochemical Analysis

The soil organic matter content was approximated from the loss-on-ignition (LOI, %) of oven-dried soils and sediments during combustion at 1000 °C for 4 h. Total element concentrations were determined by inductively coupled plasma atomic emission spectroscopy (ICP-AES) following lithium borate fusion of the combusted soils. LOI and ICP-AES analyses were conducted by ALS Geochemistry (Reno, NV). Although combustion at 1000 °C can include both soil organic matter and carbonates, carbonate contributions are expected to be low in these acidic, organic-dominated materials. Combined bulk geochemical results are provided in Table S3. The soil bulk density and gravimetric water content were determined for each soil horizon in hillslope soil cores. Soil bulk density was determined as the mass of dry soil after freeze-drying per soil volume calculated using the horizon depth and core diameter. Gravimetric water content was calculated as the mass of soil water, i.e., mass difference between the wet and freeze-dried soil, per mass of dry soil for each horizon. Soil bulk density ranged from 0.03 to 0.20 g soil·cm^–3 (mean 0.08 g soil·cm^–3) and 0.06–0.23 g soil·cm^–3 (mean 0.12 g soil·cm^–3) in the organic soils from the acidic and nonacidic hillslope, respectively. The gravimetric soil moisture content (mean ± SEM) was 4.4 ± 0.7 and 4.4 ± 0.6 g H₂O·g soil^–1 for acidic and nonacidic hillslopes, respectively.

Additional soil geochemistry from around the Toolik Field Station (TOOL) has been reported by the National Ecological Observatory Network (NEON).³⁷ We evaluated concentrations of oxalate-extractable Fe and Mehlich-III total P in all organic horizons (n = 31) collected during initial site characterization in August 2018 from 18 distributed plots spanning dwarf shrub and herbaceous sedge land covers. Oxalate-extractable Fe includes organically complexed and non- and semicrystalline Fe in addition to magnetite (Fe₃O₄) and minor amounts of crystalline Fe and trioctahedral Fe phyllosilicates and was used to represent SRO Fe(III)-oxyhydroxides measured in our study.³⁸⁻⁴⁰ Mehlich-III total P is a measure of orthophosphate extractable by acetic acid and fluoride compounds and is used to represent total bioavailable P.⁴¹

X-ray Absorption Spectroscopy

Iron, P, and Ca K-edge X-ray absorption spectra were acquired for all biogenic Fe mat samples on beamline 06B1-1 Soft X-ray Microcharacterization Beamline (SXRMB) at the Canadian Light Source (CLS, Saskatoon, SK, Canada). Phosphorus spectra for the hillslope soils were also collected at SXRMB. Iron spectra for hillslope soils and Fe and P spectra for the thaw slump samples were performed at beamline 9-BM at the Advanced Photon Source (APS, Argonne National Laboratory, Argonne, IL). Samples for XAS analysis were dried and lightly ground with a mortar and pestle that were cleaned with methanol between each use. The powdered samples were transferred with a methanol-cleaned spatula onto C tape adhered to a Cu metal sample holder. Spectra were collected under vacuum in fluorescence and total electron yield (TEY) modes with a 7-element silicon drift detector. The distance from the sample to the detector was adjusted to optimize fluorescence counts for each element in each sample. Iron K-edge spectra were collected with energy steps of 5 eV from −88 to −8 eV, 0.35 eV from −8 to 40 eV, and 0.05 Å^–1 from 40 eV to 13 Å^–1 around the Fe K-edge. Fluorescence spectra were collected for environmental samples while total electron yield (TEY) spectra were collected for standards to avoid self-absorption effects. Phosphorus and Ca K-edge spectra were collected with energy steps of 2 eV from −20 to −6 eV, 0.15 eV from −6 to 20 eV, and 0.75 eV from 20 to 65 eV (P) or 100 eV (Ca). Phosphorus standards that were plotted in fluorescence were compared to TEY spectra, where available, to ensure consistent output between detectors. Spectra were deglitched, merged across replicate scans, and energy calibrated using Athena software.⁴² The E₀ of each spectrum was set to the zero-crossing of the second derivative. All Fe spectra were shifted to align E₀ for the Fe reference foil collected during each run with that reported for a reference database (E₀ = 7110.75 eV).⁴³ No energy shift was applied to P spectra collected at CLS given that the P spectrum acquired for a CaHPO₄ standard was consistent with that reported in Hephaestus (v0.9.26).⁴² Phosphorus spectra for thaw slump samples collected at APS were shifted by +3 eV to align with those collected at CLS. No energy shift was applied to Ca spectra since fitting was performed with standard spectra collected at the same beamline with the same energy calibration. Spectra were rebinned to 10 eV in the pre-edge region and 0.5 eV in the near edge X-ray absorption fine structure (NEXAFS) region to reduce noise, then baseline corrected using a linear regression in the pre-edge region and edge-step normalized using a second- or third-order polynomial to fit the postedge region. The normalization range for each P spectrum ranged from +27 ± 3 to +58 eV and was adjusted on the low end to ensure consistency and proper fitting, following Werner and Prietzel (2015). Files containing XANES and extended X-ray absorption fine structure (EXAFS) spectra and results of the linear combination fits are provided in the Supporting Information. The Fourier transformed Fe K-edge EXAFS spectra plotted in R space and the fitting results for P and Ca XANES spectra are shown in Figures S1–S5.

Average Fe oxidation states were determined using linear combination fits to μ(E) from −20 to +30 eV around the Fe K-edge with single-valence Fe standard compounds, i.e., biotite, vivianite, chlorite, Fe(II)-oxalate, Fe(III)-oxalate, Fe(III)-citrate, ferrihydrite, goethite, and hematite. Relative proportions of Fe(II) and Fe(III) were calculated as the sum of contributions from Fe(II) or Fe(III) standard compounds, respectively. Iron speciation was investigated using linear combination fits to k² χ(κ) from 3 to 10 Å^–1. Each spectrum was fit with all combinations of select standards using up to five standards. The best fit for each spectrum was selected as the fit with the minimum number of components that were within 10% of the best fit, using reduced chi-square (χ²) as a fitting parameter. If multiple fits met those criteria, the contributions of individual components were averaged across the best fits, and uncertainty was calculated as the standard error of the mean. In this approach, components were excluded if their addition did not significantly improve the fit. Best fit spectra were obtained with a single linear combination fit using the components reported by combinatorial fitting (Table S5). The final Fe components that contributed to the best fits were Fe(II)-bearing clays (chlorite (ripidolite)), crystalline Fe(III) oxides (goethite, magnetite), short-range ordered Fe(III) oxyhydroxides (lepidocrocite, 6L-ferrihydrite), and organic-bound Fe(III) (Fe(III)-citrate). Vivianite (Fe(II)₃(PO₄)₂·8H₂O) and Fe(III) phosphate were included as components in linear combination fits for samples with P/Fe molar ratios >5% of total Fe (n = 16 total samples).

Phosphorus speciation was investigated using linear combination fits to μ(E) from −20 to +30 eV around the P K-edge. Standards used in the fits included organic phosphorus (ATP, phytic acid), Fe-bound P (phosphate adsorbed to ferrihydrite, phosphate adsorbed to goethite), Al-bound P (phosphate adsorbed to Al(OH)₃, AlPO₄), and Ca-bound P (hydroxyapatite, CaHPO₄). Standard spectra reported by Werner and Prietzel⁴⁴ and Gustafsson et al.⁴⁵ were shifted by −2.0 eV to align with comparable standard spectra collected at CLS (i.e., phosphate adsorbed to goethite and CaHPO₄). Best fits were determined using the same criteria reported for Fe speciation. Standard spectra that generated negative contributions were removed from fits for individual samples by rerunning the linear combination fitting without that standard included as a contributor. Calcium speciation was investigated using linear combination fits to μ(E) from −20 to +30 eV around the Ca K-edge. Standards used in the fits included spectra from CaHPO₄ and hydroxyapatite powders collected during our beamtime as well as spectra for apatite, calcite, Ca-pectate, Ca-acetylacetonate, kaolin, and smectite previously collected at SXRMB by the beamline scientists. Summaries of the linear combination fit for all Fe, P, and Ca spectra are provided in Tables S3–S6. The mass proportions of different compounds were calculated as the fraction of that compound determined by linear combination fits multiplied by the total concentration of that element in the sample.

Statistical Analysis

All statistical analyses were performed using R Statistical Software (v4.3.0).⁴⁶ Statistical comparisons between sample groups reported in the text were calculated using two-sample t-tests with a confidence interval of 95%. When p-values were lower than 0.05 (i.e., p < 0.05), differences in the numerical results were considered statistically significant. All calculated results with uncertainties presented in the text are reported as mean ± SEM. All data figures were plotted using the ggplot2 R package (v3.4.3).⁴⁷

Results and Discussion

Bulk Element Concentrations

The bulk Fe and organic content of hillslope soils were similar across topographic positions (i.e., upland, slope, lowland) and vertically between surface and subsurface O horizons (Figure 2a,b). The exception was the subsurface O horizon in the lowland acidic tundra, which was substantially enriched in Fe (∼115 g Fe·kg soil^–1) relative to upland and nonacidic soils (Fe < 35 g Fe·kg soil^–1). Aluminum (Al) and Si concentrations were higher in subsurface O horizons than surface O horizons (∼155% enrichment of Al; ∼74% enrichment of Si; see Table S3), which presumably reflected higher mineral content in the subsurface O horizons due to mixing with the underlying mineral soil. In addition, the total amount of base cations was significantly higher in the nonacidic hillslope (22.6 ± 0.6 g·kg soil^–1) than in the acidic hillslope (13.0 ± 1.2 g·kg soil^–1; Table S3).

Figure 2.

(a) Loss-on-ignition (LOI, %), (b) total Fe concentrations (g Fe·kg^–1), and (c) average Fe oxidation state of materials collected from hillslope soils, biogenic Fe mats, and pond sediments in areas disturbed by mining. (c) Average iron oxidation states were calculated from linear combination fits to Fe K-edge XANES spectra (Table S4). The stars in (c) denote samples from permafrost thaw slump mud slurries. LOI and total Fe concentrations were not evaluated for the thaw slump samples.

Biogenic Fe mats exhibited geochemical properties that were distinct from tundra hillslope soils and from pond sediments in mining-disturbed areas (Figure 2). Biogenic Fe mats contained higher concentrations of Fe (219 ± 18 g Fe·kg^–1) and organic material (53.6 ± 1.3%) relative to the mining-disturbed ponds (35 ± 1 g Fe·kg^–1, 6.7 ± 0.9% organic material), but had lower amounts of base cations (i.e., Ca, Mg, Na, K), silicon (Si), and other presumably bedrock-derived elements (Table S3). Notably, Fe was enriched by at least 280% in pond sediments supporting biogenic Fe mats relative to pond sediments in mining-disturbed areas. There was a wider range of Fe concentrations in biogenic Fe mats formed at pond sediment–water interfaces (134–292 g Fe·kg^–1) compared to those associated with submerged plant material (229–255 g Fe·kg^–1; Figure 2b), but all other geochemical properties between the two groups of biogenic Fe mats were similar.

Iron Oxidation State and Speciation

Iron K-edge XAS was used to identify and compare the average oxidation state and speciation of Fe in biogenic Fe mats to the mining-disturbed ponds, tundra hillslope soils, and retrogressive thaw slumps (RTS). All samples contained mostly Fe(III) (50–95%), except for the Lake 89 RTS which contained 70% Fe(II) (Figure 2c, Table S4). The average oxidation state of Fe in hillslope soils was similar across hillslope positions (2.73 ± 0.02). There was a relatively low Fe oxidation state in the subsurface O horizon at the lowland position of the nonacidic hillslope that could indicate either more reducing conditions or the presence of Fe(II)-bearing silicates. Iron oxidation state was significantly higher (p < 0.05) in biogenic Fe mats from undisturbed wet sedge habitats (2.87 ± 0.04) than in hillslope soils (2.73 ± 0.02) or mining-disturbed pond sediments (2.69 ± 0.01).

In hillslope soils, concentrations and proportions of SRO Fe(III) oxyhydroxides were highest in lowlands, largely driven by Fe accumulations in subsurface horizons of the acidic tundra (Figure 3a). Organic-bound Fe(III) was proportionally highest in upland soils, consistent with previous studies showing that organic-bound Fe(III) may be more abundant than Fe(III) oxyhydroxides in acidic soils.^9,32,34,48 Fe-bearing clays increased slightly moving down the hillslope, but the differences between each hillslope position were not significant (i.e., p > 0.05). This may indicate that chemical weathering processes that cause the transport and/or formation of Fe-bearing clay minerals had a lesser effect on Fe speciation than redox-driven processes that result in the formation of SRO Fe phases and Fe(III)-organic matter complexes.

Figure 3.

(left panels) Mass concentrations (g·kg^–1 soil) and (right panels) relative proportions of Fe, P, and Ca soil minerals in hillslope soils, biogenic Fe mats, and pond sediments in areas disturbed by mining. Error bars represent standard error of the mean. Element speciation was calculated from linear combination fits to XAS spectra. The legend in the left panel also applies to the right panel. (a) Iron speciation was determined from K-edge EXAFS analysis. The “other Fe” category includes vivianite, phosphate adsorbed to Fe(III), and oxalate-bound Fe(II). (b, c) P and Ca speciation were determined from K-edge XANES analysis. Data shown for each hillslope soil includes combined results for each hillslope transect (acidic and nonacidic) and O horizons (surface and subsurface). No Ca speciation data were collected for hillslope soils.

Biogenic Fe mats were dominated by short-range ordered (SRO) Fe(III) oxyhydroxides, whereas mining-disturbed ponds contained mostly organic-bound Fe(III) and Fe(II)-bearing clays (Figure 3a, Table S5). Correspondingly, the average mass concentration of SRO Fe was significantly higher (p < 0.05) in biogenic Fe mats (135 ± 16 g·kg^–1) than in the mining-disturbed ponds (2.5 ± 1.3 g·kg^–1). These results align with the reported dominance of SRO Fe(III) oxyhydroxides in biogenic Fe mats.^24,49 In addition, SRO Fe phases were more prevalent in biogenic mats associated with submerged vegetation (73 ± 8%) than Fe mats at pond sediment-water interfaces (47 ± 11%), whereas the pond biogenic Fe mats had higher amounts of organic-bound Fe(III). The Lake 89 thaw slump sample contained mostly Fe(II)-bearing clays (45%) with 31% SRO Fe(III) oxyhydroxides and 9% organic-bound Fe(III), whereas the Wolverine thaw slump contained 52% SRO Fe(III) oxyhydroxides, 17% Fe(II)-bearing clays, and 17% organic-bound Fe(III) (Table S5).

Crystalline Fe(III) oxides, organic-bound Fe(II), and Fe-phosphate minerals comprised a relatively small proportion (<15%) of total Fe (Table S5). This suggests that the microbial dissolution and formation of SRO Fe phases may outcompete the ripening of SRO into crystalline Fe (oxyhydr)oxides in these dynamic redox systems. In addition, the greater abundance of organic-bound Fe(III) and Fe(III) (oxyhydr)oxides relative to crystalline Fe(III) oxides and organic-bound Fe(II) may indicate that all sampling locations experience consistent or transiently oxic conditions. Nutrient limitation may be another factor influencing Fe speciation. For example, although Fe (oxyhydr)oxides can bind large quantities of phosphate, tundra soils often contain low amounts of available phosphate which could limit the formation of Fe-phosphate precipitates (e.g., vivianite).⁵⁻⁸

From these results, we infer that Fe is mobilized and transformed across the transition from tussock uplands to wet sedge lowlands and into ponds and streams. The progressive increase in Fe, and in particular SRO Fe(III) oxyhydroxides, down the hillslope toward the lowlands is a key signature of increasing activity of Fe-oxidizing microorganisms, resulting in the formation of biogenic Fe mats which are primarily comprised of SRO Fe phases (Figure 2).^25,28

Calcium Speciation

Biogenic Fe mats associated with submerged vegetation contained primarily organic-bound Ca (43 ± 11%) followed by Ca²⁺ associated with clay minerals (31 ± 6%) and Ca phosphates (23 ± 4%). Biogenic Fe mats formed at pond sediment–water interfaces had a higher proportion of organic-bound Ca (57 ± 10%) and slightly more Ca phosphates (16 ± 1%) than Ca²⁺ associated with clays (14 ± 3%). The disturbed pond sediments exhibited the lowest proportion of organic-bound Ca (38 ± 6%) and similar amounts of Ca phosphates (32 ± 3%) and Ca²⁺ associated with clays (27 ± 4%). Calcium carbonate was a minor component in all collected samples (<10%). Overall, there was a significant enrichment of Ca phosphates in disturbed pond sediments relative to pond sediments supporting biogenic Fe mats (75% enrichment, p < 0.05), indicating a shift toward more Ca–P associations in mining-disturbed habitats.

Phosphorus Speciation

Hillslope soils contained high proportions of organic P species (66 ± 3%) with relatively small amounts of Ca- and Fe-bound P. There were no significant trends in P species along the hillslope toposequence. Biogenic Fe mats contained mostly phosphate bound to Fe (oxyhydr)oxides and had significantly higher (p < 0.05) proportions of Fe-bound P than the mining-disturbed ponds and hillslope soils (Figures 3b and S3 and S4, Table S6). Aluminum-bound P was higher in biogenic Fe mats formed at pond sediment–water interfaces (28 ± 4%) than those associated with submerged plant material (12 ± 0.3%). All biogenic Fe mats had low amounts of organic P (mean 6 ± 2%). The sediments in mining-disturbed ponds were dominated by Ca–P phases (e.g., CaHPO₄, (hydroxy)apatite) and contained similar proportions of Al- and Fe-bound P (∼25–30%). No organic P species were detected in any of the disturbed sites.

These results indicate that landscape position may determine the dominant processes driving phosphate bioavailability (Figure 4). In upland soils, dephosphorylation of organic matter by plant and microbial phosphatase enzymes may release phosphate into solution where it can be removed by microbial and plant uptake.⁷ In saturated lowland soils, Fe mats may immobilize phosphate cleaved from organic matter and/or sequester phosphate as it is transported downslope.⁵⁰ In addition, organic P, which comprised 35–86% of total P in hillslope soils, may form strong outer-sphere complexes with mineral surfaces.⁵¹⁻⁵³ At the pond sediment–water interface (i.e., pond mats), Fe mats may bind P leaching from soil water, dissolved in the water column, or released from the reductive dissolution of other Fe(III) phases in the sediments during periods of anoxia.^54,55 Our data suggest that SRO Fe(III) oxyhydroxides dominate phosphate sorption at redox interfaces in the wet sedge tundra (Figure 3a). Indeed, the Fe mats in our study contained the highest overall concentrations of Fe-bound P (0.37 ± 0.04 g·kg^–1; Figure 3, Table S5) despite having the lowest average concentration of total P (0.70 ± 0.07 g·kg^–1; Table S3). These results are consistent with previous observations that demonstrate a strong correlation between the concentration of low crystallinity Fe oxide phases with both the phosphate sorption capacity and concentrations of labile P in boreal and tundra systems.⁹ Furthermore, soils throughout the Toolik region exhibit a negative relationship between oxalate-extractable Fe and Mehlich-III P (Figure 5),⁵⁶ indicating that P bioavailability decreases where Fe(III) oxyhydroxides accumulate and supporting our assertion that Fe oxyhydroxides that form at redox interfaces are critical controls on P solubility in arctic tundra.

Figure 4.

Conceptual representation of the differences in P interactions between hillslope soils, biogenic Fe mats, and mining-disturbed pond sediments. The proportion of Fe(III) increased down the hillslope toposequence and into the undisturbed wet sedge meadows which supported biogenic Fe mats. Mining-disturbed ponds were associated with a loss of biogenic Fe mats. The dominant P species shifted from organic P in hillslopes to Fe-bound P in biogenic Fe mats and Ca-bound P in disturbed pond sediments.

Figure 5.

Mehlich-III extractable soil P and oxalate-extractable soil Fe were measured during the distributed initial characterization of 18 sites from around Toolik Field Station (TOOL). The data were reported by the National Ecological Observatory Network (NEON).³⁷ The represented soil samples were collected from organic horizons in the top 1 m of the soil profile in August 2018. Oxalate-extractable Fe includes organically complexed and non- and semicrystalline Fe in addition to magnetite (Fe₃O₄) and minor amounts of crystalline Fe and trioctahedral Fe phyllosilicates.³⁸⁻⁴⁰ Mehlich-III P includes orthophosphate extractable by acetic acid and fluoride compounds and is an estimate of total plant available P.

Environmental Implications

In this study, we identified differences in the speciation and associations of P with soil minerals across undisturbed and disturbed Arctic tundra ecosystems. Most notably, we observed a shift from P associations with Fe (oxyhydr)oxides that comprise biogenic Fe mats in undisturbed wet sedge meadows to interactions with Ca minerals in areas impacted by physical disturbances (Figure 3). Considering that frequent redox cycling supports the formation and dissolution of Fe(III) (oxyhydr)oxides,^25,28,57 we infer that biogenic Fe mats represent important redox interfaces in wet sedge tundra that may influence P bioavailability primarily via adsorption of phosphate to SRO Fe(III) oxyhydroxides (Figure 4). As such, processes that alter the geochemical and hydrologic characteristics of tundra systems (e.g., climate warming,⁵⁸⁻⁶⁰ mineral extraction,^24,61 land cover change⁶²) may affect the spatial distribution of these important redox interfaces and directly impact the accumulation of Fe (oxyhydr)oxides and regulate P bioavailability to tundra plants and microorganisms in a warming Arctic climate.

Landscape disturbances were associated with an absence of biogenic Fe mats and a shift toward more P interactions with Ca minerals and Fe(II)-bearing clays (Figure 3; Tables S5–S7).⁶¹ As both chronic and episodic disturbances are becoming increasingly common across the Alaskan Arctic, we contend that this could limit potential P retention at wet sedge redox interfaces by inhibiting microbial Fe cycling and inducing a shift from primarily redox-controlled P interactions to those that are more susceptible to changes in pH. For example, Fe(II) clays and Ca minerals are effective sorbents for phosphate and organic P^17,63−65 but are susceptible to dissolution and P release under acidic conditions (pH < ∼5).^22,23 Moreover, gravel mining may increase soil alkalinity and either enhance or inhibit P retention by promoting the coprecipitation of Ca phosphate species or raising the pH above the point of zero charge of soil minerals, respectively.⁶⁶⁻⁶⁸ A further compounding factor is that gravel mining in our study area occurred approximately 50 years ago, yet biogenic Fe mats have never reestablished. This observation provides a striking example of the long-term impacts of mining disturbance on microbial Fe cycling, and in turn, the ability of wet sedge redox interfaces to recover and mediate P bioavailability in postmining landscapes.

The retrogressive thaw slumps develop as a consequence of accelerated permafrost thaw that is becoming increasingly common under a warming Arctic climate.⁶⁹⁻⁷³ In contrast to the physical removal or organic soil horizons during mining, the high proportion of Fe(II) in RTS slurries (up to 70%) relative to ponds in mining-disturbed areas (28–32%) and the biogenic Fe mats (6–21%; Figure 2d) reflects the mixing of organic and mineral horizons during soil collapse that can solubilize soil minerals previously trapped in permafrost and export considerable amounts of C and bioavailable P into aquatic systems.^74,75 However, the proportion of Fe(III) and Fe(oxyhydr)oxides in RTS slurries appears to increase with RTS age, which may indicate that some microbial Fe cycling has been reestablished following initial soil collapse. It is possible that although RTS can significantly alter soil P interactions, the redox interfaces within the affected landscape may reestablish their function as mediators of P mobility over much shorter time scales than areas impacted by gravel mining.

The time scales of P retention by SRO Fe oxyhydroxides at redox interfaces remain unclear. Transient anoxia driven by fluctuating water tables promotes reductive dissolution of SRO Fe phases to release phosphate, thus providing a source of bioavailable P that may be resorbed as Fe(III) oxyhydroxides reprecipitate under oxic conditions. Under a transition to persistently anoxic conditions (e.g., soil flooding from increased Arctic precipitation⁷⁶ and thermokarst development⁷⁷), Fe reduction may release phosphate into P-limited aquatic systems and fuel primary production, thus partially offsetting C losses linked to permafrost thaw.^1,2,5,7 Repeated redox cycling may facilitate the transformation of SRO Fe oxyhydroxides into crystalline Fe phases⁷⁸ which confer a lower phosphate sorption capacity, but are less thermodynamically favorable to Fe-reducing bacteria⁷⁹ and could immobilize P over longer time scales.^9,78,79 Continued activity of Fe-oxidizing bacteria in wet sedge meadow habitats could counteract this mineral aging process.^14,80 and perpetuate the prevalence of SRO Fe(III) oxyhydroxides that may be long-term P sinks. These processes may be severely disrupted by future climate- and human-induced disturbances that inhibit microbial Fe cycling at redox interfaces and shift the quantity and composition of P associations with soil minerals. In addition, enhanced evapotranspiration and thawing permafrost may create deeper flow paths that limit potential interactions of soil P with Fe mats.⁸¹ Thus, our study indicates that redox interfaces in wet sedge meadows form important barriers regulating P transport into nutrient-limited aquatic systems but are sensitive to disturbance.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.est.3c09072

• Additional details on sampling locations, bulk element concentrations, and figures of XAS spectra; complete processed EXAFS spectra and normalized XANES spectra of all samples included in the manuscript and reference standards (PDF)

• Normalized P and Ca XANES spectra with linear combination fits (XLSX)

The authors declare no competing financial interest.

Notes

Notice This manuscript has been authored by UT-Battelle, LLC, under contract DE-AC05-00OR22725 with the US Department of Energy (DOE). The US government retains and the publisher, by accepting the article for publication, acknowledges that the US government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this manuscript, or allow others to do so, for US government purposes. DOE will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (https://www.energy.gov/doe-public-access-plan).