Introduction

Arctic and subarctic permafrost soils contain approximately half of global soil carbon but are vulnerable to thaw and changing hydrological conditions as air and ground temperatures increase1,2,3,4. Permafrost thaw and soil warming accelerate the decomposition of soil organic matter, increasing the emission of greenhouse gases such as CO2 and CH4 to the atmosphere5,6,7,8,9. Increased plant growth under a warmer climate might offset carbon losses10; however, gross primary production is constrained by phosphorus (P), an important limiting nutrient in many regions including the Arctic and Subarctic11,12. Iron (Fe) (III) (oxyhydr)oxide minerals, hereafter referred to as Fe oxides, strongly bind orthophosphate (i.e., phosphate), the major ion of P used by plants and microorganisms13,14, and have been proposed to control P bioavailability in temperate and tropical soils15,16,17, lakes18, and seawater14. However, the role of Fe oxides in regulating phosphate bioavailability in (Sub)Arctic ecosystems remains unclear.

A few recent studies have investigated Fe speciation and associated interactions with C7,19,20,21,22 and P23,24,25 across redox gradients in Arctic tundra and permafrost peatlands, but lack insight into how naturally occurring Fe oxides and their interactions with phosphate transform over time in response to soil conditions induced by thaw. The Fe oxides that occur in unsaturated soils underlain by permafrost or accumulate at redox interfaces in shallow, saturated soils in topographic lows can trap phosphate24,25,26,27 released through enzymatic dephosphorylation of organic matter28,29, potentially limiting phosphate uptake by plants and microorganisms30. Ground collapse and flooding associated with permafrost thaw can generate reducing conditions that dissolve Fe oxides and release bound organic matter, depleting the “rusty carbon sink”19,31. Impacts of thaw on Fe-P associations have not been examined but are critical to investigate in (Sub)Arctic permafrost-affected ecosystems that are warming faster than the global average32. Changing soil water saturation due to permafrost thaw affects the formation of Fe oxides and their ability to bind phosphate. Although ground collapse and flooding can promote reducing conditions, some soils may drain and become oxidizing as permafrost thaw lowers the water table33, favoring the formation of either organic-bound Fe(III) or short-range ordered (SRO) Fe oxides that possess high phosphate sorption capacities34,35. Incomplete understanding of the complex trajectory of Fe oxide formation and dissolution under shifting environmental conditions limits our ability to predict the impact of permafrost thaw on nutrient storage and bioavailability in (Sub-)Arctic ecosystems.

This study investigates how Fe in naturally occurring sediments transforms under contrasting redox regimes and influences phosphate retention along a permafrost thaw gradient in a Subarctic permafrost peatland. First, we acquired high-resolution soil redox (Eh) data to evaluate how permafrost thaw impacts soil redox regimes. Most previous studies have relied on sporadic Eh measurements to characterize redox regimes8,36, emphasizing a need for continuous Eh measurements to resolve how soil redox conditions vary over space and time. Second, we collected Fe-rich sediments from a local drainage channel and incubated them in surface soils across the permafrost thaw gradient to investigate how Fe speciation and Fe-bound phosphate respond to contrasting redox conditions. Third, we measured P concentrations in microbial biomass associated with the Fe-rich sediments to evaluate microbial acquisition of phosphorus from the Fe-rich sediments. We hypothesized that soil redox potential would shift from predominantly oxidizing to predominantly reducing conditions as saturation increased along the permafrost thaw gradient. We also predicted that more Fe would be lost from sediments incubated in saturated soils. We expected that the composition of Fe remaining in the sediments would shift towards SRO Fe oxides under fluctuating redox conditions as dissolved Fe2+ reprecipitated during oxidizing periods and that these newly formed SRO Fe oxides would retain phosphate. Finally, we hypothesized that microorganisms associated with the Fe-rich sediments would contain more biomass P when the sediments were amended with excess phosphate.

Here, we show that permafrost thaw generates persistently reducing conditions in soils that are flooded following thaw and ground collapse. Short-range ordered Fe oxides dissolve and release phosphate under reducing conditions in saturated soils, while organic-bound Fe(III) that does not bind phosphate is retained. Iron oxide dissolution and microbial biomass P increase when Fe-rich sediments are amended with phosphate, indicating that microorganisms may enhance Fe oxide dissolution and/or acquire associated phosphate. Overall, these findings indicate that permafrost thaw reduces the capacity for Fe oxides to immobilize and retain phosphate in permafrost peatlands.

Results and discussion

Permafrost thaw induces flooded, reducing conditions

Soil redox potential (Eh) transitioned from persistently oxic in the unsaturated palsa to temporarily Fe(III) reducing in the saturated bog to persistently Fe(III) reducing in the fen (Fig. 1 and Supplementary Fig. 1; Supplementary Data 1). Redox potential was stable and consistently high in palsa surface soil (average = +796 ± 14 mV for <6 cm depth), indicating oxidizing conditions that reflect the unsaturated soil regime. Redox potential was lower and more variable in the bog (average = +349 ± 107 mV; p < 0.001), reaching a minimum of 80 ± 60 mV in mid-July before steadily increasing to 661 ± 83 mV by early August. Thus, the bog remained in Fe-reducing conditions for the first 17 days of the incubation and then gradually transitioned to Fe(II) oxidizing conditions. The slow increase in Eh in the bog can be explained by the progressive drying of surface soils as surface water evaporated and/or drained into the fen. Average Eh was lowest in the fen (+172 ± 14 mV; p < 0.001) and peaked at +441 ± 107 mV in early July before declining to as low as +16 ± 24 mV, indicating conditions favorable for Fe(III) reduction throughout the incubation. The more persistent reducing condition in the fen likely reflects more saturated soils due to discharge from bog soils and/or connectivity with the regional water table37. Redox potential remained Fe(III) reducing at soil depths >8 cm in the bog and fen, with the fen maintaining lower Eh than the bog (Supplementary Fig. 1), consistent with previously reported single point Eh values of −50 to +100 mV8. The shift from persistent oxidizing conditions in the palsa to reducing conditions in the fen has been previously demonstrated to drastically alter biogeochemical reactions and greenhouse gas emissions in this system, for example, by increasing both methane production25,38,39 and methane oxidation8.

Fig. 1: Redox conditions in surface soils differ across a permafrost thaw gradient.
figure 1

Redox potential in surface soils in the palsa (yellow), bog (blue), and fen (green) were significantly different (one-way non-parametric analysis of variance, p < 0.001; n = 5,179). Black lines represent the average of measurements at 2, 4, and 6 cm depths below the ground surface (1, 2, 3, 4, and 5 cm for palsa) while colored areas represent the standard error of the mean. The transparent box represents the range of expected ferrihydrite stability between pH 4 (top) and 5 (bottom). The striped, black box on the right side of the graph represents range of expected goethite and lepidocrocite stability between pH 4 and 5. Measurements below the box are considered Fe(III)-reducing while measurements above the box are considered Fe(II)-oxidizing. The vertical dotted line represents the date of the first bag removal (7 days).

Full size image

Redox potential was oxidizing in surface water (+528 ± 7 mV) and decreased with depth in pore waters, reaching a minimum of +381 ± 19 mV in the deep fen in early July (Fig. 2; Supplementary Data 2). This redox potential was poised in the zone corresponding to ferrihydrite reduction potential, indicating that Fe redox cycling was likely occurring in the top 30 cm of the bog and fen. Higher dissolved Fe and P concentrations were strongly correlated with low Eh and higher pH (Supplementary Fig. 2), and the highest concentrations occurred at 5–10 cm depth in the bog and at 28 cm depth in the fen (Fig. 2). These dissolved Fe concentrations (<500 µmol L−1) were consistent with values reported in Perryman et al8. but lower than Fe2+ concentrations in rhizon-extracted pore water reported by Patzner et al.31, indicating that the sippers likely sampled more dilute mobile water while rhizons extracted more tightly held water in small pores.

Fig. 2: Soil water chemistry in bog and fen soils.
figure 2

Soil water pH, specific conductance (µS cm−1), redox potential (Eh; mV) and 0.45 µm filterable Fe and P in surface and pore water collected from the palsa (n = 1), bog (n = 33), and fen (n = 33) in early July 2019. Redox potential relative to the standard hydrogen electrode (Eh) was converted from oxidation-reduction potential measured in the field. The shaded gray box indicates the range of the approximate reduction potential for ferrihydrite between pH 4 and 5. Shaded blue and purple areas emphasize the range of values for bog and fen water and are not shown for specific conductance and Eh were they largely overlap. Values for individual samples are provided in Supplementary Data 2. Note that P was not measured on all samples.

Full size image

Soil water in the palsa (3.94; n = 1) was more acidic than in the bog and fen (4.49 ± 0.04 and 4.70 ± 0.06; respectively) (Fig. 2). The highest pore water pH occurred ~30 cm deep in the fen (5.49 ± 0.23; n = 3). Surface soil pH was acidic and increased from the palsa (3.88 ± 0.11) to the bog (4.27 ± 0.02) to the fen (4.40 ± 0.04) (Supplementary Table 1), consistent with previous studies8. Surface water taken from a nearby bog to the fen transition (east site; Supplementary Fig. 3) increased from 3.88 ± 0.05 (n = 3) in the bog to 5.67 ± 0.04 (n = 3) in the fen, indicating that the fen position in the experimental transect receives surface water primarily from bog drainage rather than lake outflow37. Similarly, the highest soil pH occurred in subsurface mineral soil at the east bog site (pH = 5.37).

Soil and sediment contain organic-bound Fe and SRO Fe oxides

The majority (90.0 ± 0.5%) of Fe in the initial Fe-rich sediments was present as hydroxylamine-soluble Fe (3.27 ± 0.14 mmol g−1), which is widely reported to extract SRO Fe oxides such as ferrihydrite and lepidocrocite40 (Table 1; Supplementary Table 2). We observed substantial extraction of organic material as indicated by the dark color of the extract solution and consider hydroxylamine-soluble Fe to represent both SRO Fe oxides and organic-bound Fe in these sediments. Relatively low concentrations of Fe were present as water-soluble (<0.1%; 0.19 ± 0.06 µmol g−1) and dithionite-soluble (9.8 ± 0.5%; 0.36 ± 0.03 mmol g−1) Fe. The results of these extractions were consistent with high concentrations of short-range ordered (SRO) Fe oxides (69% ferrihydrite + 7% lepidocrocite) and organic-bound Fe(III) (24%) identified via linear combination fits to XAFS spectra (Table 2; Supplementary Data 3 and 4; Supplementary Fig. 4). The composition of the Fe-rich sediments was similar to Fe in palsa surface soil reported here (21% SRO Fe oxides and 62% organic-bound Fe(III); Table 2) and by Patzner et al.31, although with higher amounts of SRO Fe oxides than organic-bound Fe(III). Palsa soil also contained 15% “other Fe” fit as chlorite (6 ± 6%) and vivianite (10 ± 10%), although the identity of those phases was not definitive. Iron in both the Fe-rich sediments and palsa surface soil was highly oxidized (98% and 93% Fe(III), respectively). Thus, the Fe-rich sediments contain Fe phases that are comparable to phases found in the palsa soils prior to thaw and ground collapse.

Table 1 Concentrations of Fe and soluble reactive phosphorus (SRP) sequentially extracted from incubated sediments by water, hydroxylamine hydrochloride, and citrate buffered dithionite solutions
Full size table
Table 2 Mean (±s.e.m.) Fe oxidation states and speciation of incubated Fe-rich sediments and palsa soils determined by linear combination fits to Fe K-edge XAFS spectra
Full size table

Phosphate added to the Fe-rich sediments was retained largely in the hydroxylamine-soluble fraction (62%; 25.8 ± 0.2 µmol P g−1; Table 1) and is attributed to sorption to SRO Fe oxides. A smaller but substantial proportion of added P partitioned into the dithionite-soluble fraction (38%; 15.9 ± 0.7 µmol P g−1), despite it comprising a smaller portion of extractable Fe (9.8 ± 0.2%), signifying sorption to more crystalline Fe oxides such as goethite. Very low concentrations of water-soluble phosphate (<1%; 0.04 ± 0.01 µmol P g−1) signify the strong sorption capacity of the Fe-rich sediments and the strength of phosphate binding to Fe oxides. The strong Fe-P associations that formed in Fe-P treatments reflect the strong correlation between acid-digestible P and acid-digestible Al + Fe (representing non-silicate phases) observed in permafrost and fen soils at Stordalen Mire25.

Soil redox and phosphate sorption influence Fe dissolution

Landscape position along the thaw gradient (p = 0.018), P treatment (p = 0.006), and time (p < 0.001) were all significant factors influencing Fe gain and loss from the sediments during incubation (Table 3). Total extractable Fe in all Fe-rich sediments decreased during their incubation in the bog (–0.21 ± 0.13 mmol g−1) and fen (–0.15 ± 0.16 mmol g−1) but increased slightly in the palsa (+0.14 ± 0.08 mmol g−1). Total extractable Fe decreased in Fe-P treatments (–0.33 ± 0.09 mmol g−1) but increased in Fe-only treatments (+0.17 ± 0.09 mmol g−1). Iron also increased slightly after 7 days (+0.08 ± 0.07 mmol g−1) but decreased by 70 days (–0.24 ± 0.12 mmol g−1), as averaged across all treatments (Fig. 3).

Table 3 Statistical probabilities (p-values) for differences in Fe and soluble reactive phosphorus (SRP) values across treatments
Full size table
Fig. 3: Iron loss and gain during incubation of Fe-rich sediments across a permafrost thaw gradient.
figure 3

Changes in total extractable Fe (mmol g−1) (a–c), hydroxylamine-soluble Fe (d–f), and dithionite-soluble Fe (g–i) extracted from Fe-rich sediments with no added phosphate (Fe-only; orange symbols) or with adsorbed phosphate (Fe-P; purple symbols) incubated in the palsa, bog, or fen for 7 or 70 days. Mean (±s.e.m.) values are shown as solid symbols connected by lines while individual replicates are shown as semi-transparent symbols. Extractable Fe is equal to the sum of water-, hydroxylamine-, and dithionite-soluble Fe. The gray boxes indicate the mean (±s.e.m.) of concentrations in the initial materials. Letters indicate significant differences between treatments. Significant differences across sites are not shown. Note the x-axis shown as categorial instead of linear to allow for better visualization of trends over time.

Full size image

Decreases in total extractable Fe derive from decreases in hydroxylamine-soluble Fe (Fig. 3). Losses in total extractable and hydroxylamine-soluble Fe were strongly correlated with losses in SRO Fe oxides for the Fe-P treatments (r = 0.94; p = <0.001), indicating that Fe losses were driven by dissolution of SRO Fe oxides (Figs. 4 and S7). However, Fe-only sediments lost small quantities of SRO Fe oxides (mostly <0.5 mmol g−1) but gained hydroxylamine-soluble Fe (Supplementary Fig. 5). This discrepancy may be explained by increases in organic-bound Fe(III). Although changes over time were not considered significant, Fe-only treatments ended up with more organic-bound Fe(III) than Fe-P treatments (p = 0.009) (Table 3), with an average gain of 0.11 ± 0.02 mmol g−1. From this result, we infer that Fe loss from the Fe-only treatments was both less than Fe loss from Fe-P sediments and partially offset by Fe gain as organic-bound Fe(III). Dissolved Fe2+ either released from reductive dissolution of the SRO Fe oxides or present in the surrounding soil water may have complexed with organic matter to form organic-bound Fe(II) and Fe(III), which is favored over SRO Fe oxide precipitation in these acidic conditions (pH < 5)41.

Fig. 4: Iron transformation during incubation of Fe-rich sediments across a permafrost thaw gradient.
figure 4

Iron speciation in sediments incubated in the palsa (a), bog (b), or fen (c), as determined by linear combination fits to EXAFS spectra. Orange symbols represent Fe-only sediments and purple symbols represent Fe-P sediments. Mean (±s.e.m.) values are shown as solid symbols connected by lines while individual replicates are shown as semi-transparent symbols. Iron species include short-range ordered Fe(III) oxides (triangles), Fe(III) bound to organic matter (circles), and other minor Fe species (≤15%) (squares). Orange and purple horizontal bars indicate the mean (±s.e.m.) of concentrations in the initial Fe-only or Fe-P materials, respectively. Letters indicate significant differences between P addition and time within each Fe phase; lowercase letters are used for the SRO Fe(III)-oxides and uppercase letters are used for the Organic-Fe(III). No significant differences were present in the Other Fe group. Note the x-axis shown as categorial instead of linear to allow for better visualization of trends over time.

Full size image

Iron loss from sediments incubated in the bog and fen is consistent with observed Fe(III) reducing conditions (Fig. 1) that favor microbial dissolution of Fe(III) oxides to release Fe2+ into solution. Patzner et al.19 similarly observed that synthetic ferrihydrite gained Fe when incubated in the palsa but was reduced and lost Fe when incubated in the fen, and reported increases in Fe(III)-reducing bacteria and aqueous Fe2+ from the palsa to the fen along the same thaw gradient. Similar microbial Fe(III) reduction is commonly observed in Arctic peat soils42,43,44,45. Although the bog transitioned to Fe(II) oxidizing conditions at the end of the redox measurements, precipitation patterns would indicate a return to Fe(III) reducing conditions for at least a short period toward the end of the incubation. These redox fluctuations likely limit cumulative Fe(III) reduction in the bog relative to the fen. No changes in extractable Fe were observed in the palsa, as we would expect minimal Fe(III) reduction and leaching in the oxidizing environment. Solid-phase Fe(II) also increased in all incubated sediments and was higher in the bog and fen than in the palsa (p = 0.008) (Supplementary Fig. 6), indicating that Fe(II) may have resorbed to Fe oxide surfaces following dissolution or formed minor quantities of secondary phases. Higher percentages of Fe(II) than non-Fe(III), “other Fe” phases may also indicate that SRO Fe and organic-Fe(III) pools contained appreciable Fe(II) that was too low to be resolved as a separate phase (Table 2). Overall, SRO Fe oxide dissolution in the fen after 10 weeks (19 ± 9%) was moderate compared with Fe losses from synthetic ferrihydrite in the same area over two weeks (~50%)19, reflecting lower reactivity of the natural material due to either different mineral properties or influence of the sediment matrix, e.g., mineral coatings or aggregates that limit microbial access to Fe.

Phosphate-amended (Fe-P) sediments lost significantly more Fe than the Fe-only sediments (p = 0.006), primarily as SRO Fe oxides, and this effect was most pronounced in the fen after the 70-day incubation (Figs. 3 and 4). Although phosphate sorption is known to inhibit ferrihydrite dissolution and formation of secondary Fe phases46, the Fe-P sediments were likely undersaturated with phosphate, leaving phosphate-free mineral surfaces that were susceptible to dissolution. Indeed, the P:Fe ratio of Fe-P sediments (10.5 µmol P mmol-Fe−1) was lower than the range reported for maximum phosphate sorption to synthetic Fe oxides (53–76 µmol P mmol-Fe−1)47. The presence of Fe-bound phosphate may have stimulated microbial Fe(III) reduction. Microorganisms experiencing phosphorus limitation can produce redox-active antibiotics and other metabolites that dissolve Fe oxides to release adsorbed phosphate48. Alternatively, phosphate sorption may have promoted the desorption of organic matter from the initial Fe-rich sediments, exposing more mineral surfaces for reductive dissolution49; however, sorbed phosphate would occupy those sites. Furthermore, P addition had no significant impact on the C content of the incubated Fe-rich sediments (C = 8.27 ± 0.10 wt.% in Fe-only vs. 8.59 ± 0.11 wt.% in Fe-P sediments; p = 0.1), suggesting no substantial impact of phosphate sorption on sediment C (Supplementary Table 3).

Dithionite-soluble Fe comprised only 9.8 ± 0.2% of extractable Fe and remained mostly unchanged during incubations (Fig. 3). However, small but significant differences were observed in the palsa and bog where dithionite-soluble Fe increased in Fe-P sediments but decreased in Fe-only sediments. Loss of dithionite-soluble Fe in the Fe-only sediments may result from reductive dissolution or dispersion of colloidal, crystalline Fe oxides47. Phosphate sorption in the Fe-P sediments likely passivated a substantial proportion of crystalline Fe oxide surfaces (dithionite-soluble P:Fe = 45 µmol P mmol-Fe−1) and inhibited dissolution. Increases in dithionite-soluble Fe in the Fe-P sediments are consistent with goethite formation through either aging of SRO Fe minerals50 or precipitation from Fe-rich pore water51,52. Periods of Fe(II) oxidizing conditions in the palsa and bog would allow for SRO Fe oxides to age into more crystalline phases51. Goethite also has a lower reduction potential than ferrihydrite, leading to higher expected stability in the observed Eh range (Fig. 1). Although no specific phase changes were observed, it is likely that the change (<5% of total Fe) was too small to be resolved with XAS. We do not attribute increases in dithionite-soluble Fe to residual enrichment of crystalline Fe phases given that no enrichment was observed in the fen where sediments experienced the highest total Fe loss.

Iron oxide dissolution and precipitation drive phosphate retention

Phosphate in the Fe-P sediments decreased by 13 ± 4% over time, although losses were only marginally significant (p < 0.1) and did not differ between sites (Fig. 5; Table 3). Changes in hydroxylamine-soluble phosphate were strongly correlated with hydroxylamine-soluble Fe (r = 0.74, p < 0.001; Supplementary Fig. 7) and total extractable Fe (r = 0.75, p < 0.001), indicating that SRO Fe oxide dissolution resulted in phosphate solubilization and mobilization into the surrounding environment. Dithionite-soluble phosphate was moderately correlated with dithionite-soluble Fe (r = 0.53, p = 0.03) but decreased over time despite increases in dithionite-soluble Fe (Supplementary Fig. 7). Consequently, there was proportionally less phosphate than Fe over time in these sediments, and increases in dithionite-soluble Fe may have mitigated but did not prevent phosphate loss. We propose that phosphate was solubilized during the dissolution of SRO Fe oxides but not immediately recaptured during the subsequent formation of organic-bound Fe(III) and crystalline Fe oxides during transient oxidizing conditions. Conversely, solubilized phosphate may have been acquired by microorganisms, as evident from increases in microbial biomass P (discussed below), or released into the surrounding soil solution. Additionally, the dilute, acidic soil water would have favored the desorption of surface-adsorbed phosphate from crystalline Fe oxides53. Secondary phases such as vivianite are unlikely to have formed given consistent phosphate loss during incubation of Fe-P sediments.

Fig. 5: Phosphate loss and gain during incubation of Fe-rich sediments.
figure 5

Changes in total soluble reactive phosphorus (SRP, i.e., phosphate) extracted from Fe-rich sediments with (diamonds) or without (circles) added phosphate incubated in the palsa (green), bog (blue), or fen (purple) for 7 or 70 days. Mean (±s.e.m.) values are shown as solid symbols connected by lines while individual replicates are shown as semi-transparent symbols. The gray box indicates the mean (±s.e.m.) of concentrations in the initial Fe-P material. All Fe-only treatments gained SRP, and the material incubated in the fen gained more SRP than in the palsa (p = 0.04). Iron-P sediments lost more SRP in the fen than in the palsa (p = 0.05).

Full size image

Iron-only sediments gained similar amounts of phosphate over time (p < 0.001) at all sites, but total P gain was less than initial P concentrations in Fe-P sediments (Fig. 5, Table 3). On average, 66 ± 1% of the acquired phosphate was present in the hydroxylamine fraction, consistent with the partitioning between the hydroxylamine and dithionite fractions observed for the initial Fe-P sediments. Like in the Fe-P sediments, hydroxylamine-soluble phosphate was correlated to hydroxylamine-soluble Fe (r = 0.57, p = 0.02) and total extractable Fe (r = 0.58, p = 0.02), while dithionite-soluble phosphate was correlated to dithionite-soluble Fe (r = 0.52, p = 0.04). However, differences in P sorption amongst sites were small despite large differences in Fe gain or loss. From this result, we infer that the incubated sediments had more capacity to adsorb phosphate than there was available phosphate in the solution.

Microbial uptake of phosphate sorbed to Fe oxides

Microbial biomass P was significantly higher in Fe-P sediments than Fe-only sediments after seven days of incubation (p = 0.005). The interaction between site and phosphorus addition was marginally significant (p = 0.118), as microbial biomass P concentrations in the Fe-P sediments were roughly two times greater in the fen compared to the palsa (Fig. 6). Specifically, microbial biomass P increased across the thaw gradient from the palsa (0.03 ± 0.003 µmol g−1) to the bog (0.05 ± 0.01 µmol g−1) to the fen (0.07 ± 0.03 µmol g−1) (Supplementary Table 4). In contrast, microbial biomass P concentrations in Fe-only sediments were low and similar across the thaw gradient (0.02 ± 0.001 µmol g−1). Phosphate addition (p = 0.001) was also a significant factor influencing salt (K2SO4) extractable P in sediments after the 7-day incubation period. Salt-extractable P concentrations from unfumigated sediments were higher in Fe-P treatments (0.03–0.04 ± 0.002–0.004 µmol g−1) compared to Fe-only treatments (0.01–0.02 ± 0.004–0.007 µmol g−1) (Supplementary Fig. 8). However, salt-extractable P constituted only a minor portion of phosphate added to the treatments (<1%), consistent with the water-soluble P and indicating that the majority of phosphate was associated with Fe oxides through inner sphere complexes rather than sorbed to the mineral surface through electrostatic interactions.

Fig. 6: Microbial P acquisition from Fe-P sediments.
figure 6

Microbial biomass phosphorus concentrations (µmol g−1) from Fe-only (open symbols) and Fe-P (closed symbols) sediments incubated across the thaw gradient for 7 days. Symbols and error bars represent mean concentrations ± standard error. Smaller points represent individual samples.

Full size image

Collectively, we observed that Fe-P sediments experienced more Fe oxide dissolution and gained more microbial biomass P than Fe-only sediments under Fe-reducing conditions in the bog and the fen. From these results, we infer that phosphate addition to Fe-rich sediments may have stimulated microbial activity that increased the dissolution of SRO Fe oxides and microbial P uptake. One explanation is that microorganisms actively dissolved the Fe oxides to acquire phosphate, as has been documented previously for microorganisms experiencing P limitation in laboratory studies48. Low concentrations of resin-extractable phosphate and high N:P ratios signify phosphorus limitation in Stordalen Mire, which may be exacerbated by phosphate sorption to oxide minerals25. Alternatively, it is possible that microorganisms passively acquired phosphate released through Fe oxide dissolution, although this explanation does not account for the higher Fe oxide dissolution observed for Fe-P sediments. It is also possible that P may have been incorporated into microbial biomass during phosphate addition (pre-incubation) rather than during incubation; however, Fe-P sediments incubated in the palsa had only slightly more microbial biomass P than Fe-only sediments, indicating any potential pre-incubation enrichment was minimal.

Implications of dynamic iron-phosphorus associations

Permafrost ecosystems store a substantial amount of soil organic carbon that is vulnerable to release into the atmosphere with soil warming and increasing permafrost thaw. Increased plant growth can sequester carbon from the atmosphere and offset losses due to decomposition but may be limited by phosphate bioavailability11. Iron oxides are increasingly recognized as an important control on phosphate solubility in northern tundra and boreal systems24,25,27, particularly in low-lying, saturated areas where Fe(II) oxidizes at redox interfaces54,55. Here, we observed that soils shifted from oxidizing to reducing redox regimes in response to ground collapse and flooding associated with permafrost thaw. High-resolution measurements revealed temporal variability in redox conditions, particularly in shallow bog soils that experience more frequent changes in soil saturation as the water table drops below the ground surface during late thaw37.

Decreasing redox potential associated with permafrost thaw influenced Fe-P associations in Fe-rich sediments (Fig. 7). Short-range ordered Fe oxides were most sensitive to reducing conditions and experienced the greatest Fe loss, reflecting the higher reduction potential associated with ferrihydrite compared to other Fe(III) oxides56. Organic-bound Fe(III) and crystalline Fe oxides remained relatively stable through the incubation, with Fe-only sediments often gaining organic-bound Fe(III), particularly in the palsa. This result contrasted with our initial hypothesis that SRO Fe oxides would reprecipitate and bind phosphate under transient oxidizing conditions. We attribute retention and slight accumulation of organic-bound Fe to the formation of stable Fe(III)-OM or Fe(II)-OM complexes that limit Fe(III) hydrolysis and polymerization41,57, particularly at low pH where Fe(II) is favored at higher reduction potentials. Consequently, multiple processes contribute to reduced phosphate sorption potential with permafrost thaw. First, Fe-bound phosphate is released into the solution during reductive dissolution of SRO Fe oxides. Second, although ferrihydrite can reprecipitate under transient oxidizing conditions and bind phosphate, acidic conditions and high organic content promote Fe complexation with organic matter rather than ferrihydrite precipitation. Thus, Fe that is retained in the soil is gradually converted into a form that does not sorb phosphate58, potentially explaining increases in P solubility (e.g., Fig. 2) and bioavailability25 observed in soils overlying partially thawed permafrost. Increases in P bioavailability may be only transient as soluble phosphate is subsequently taken up by plants and microorganisms, flushed out of the soil, or converted into another form as permafrost thaw progresses.

Fig. 7: Conceptual diagram of how Fe-associations mediate P retention and bioavailability across a permafrost thaw gradient.
figure 7

Palsa soils that are underlain by permafrost are relatively dry and contain short-range ordered Fe oxides that bind phosphate and organic-bound Fe(III). Permafrost thaw leads to ground collapse and flooding, generating transient to persistent anoxic conditions that drive reductive dissolution of SRO Fe oxides and phosphate release into solution, increasing bioavailability in bogs. Microorganisms can potentially enhance Fe oxide dissolution to acquire bound phosphate when P is limiting. Complete dissolution and losses of SRO Fe oxides in acidic (pH <5), persistently saturated soils potentially decrease P bioavailability by shrinking the reservoir of Fe-bound phosphate.

Full size image

Ferrihydrite accumulation might still be expected in less acidic (pH > 6), minerotrophic areas of the fen where dissolved Fe2+, sourced from deep mineral soils or transported laterally from palsa and bog areas, can oxidize and precipitate as SRO Fe oxides at redox interfaces24. Indeed, Chauhan et al.57 observed that particulate matter suspended in thaw ponds within fens contained more Fe and more ferrihydrite than particulate matter from acidic bogs within Stordalen Mire57. Accumulation of Fe oxides in the fen could drive P limitation to plants and microorganisms by sorbing bioavailable phosphate; indeed, available phosphate is low in the tall graminoid fens within Stordalen Mire where Fe/Al-associated P is high25. These Fe-P associations could retain phosphate and prevent its outflow from the mire. Although collapsed and flooded soils experience net Fe loss and possess a smaller “rusty sink”31, it is likely that phosphate sorption to Fe oxides is limited by low phosphate concentrations in solution rather than Fe oxide surface area. Indeed, Fe-only sediments incubated across the thaw gradient uniformly acquired phosphate regardless of Fe gain or loss.

Although phosphate sorption to Fe oxides is typically understood to limit bioavailability, we provide evidence that Fe oxides may serve as a phosphate reservoir that is accessible to microorganisms. Dissolution of SRO Fe oxides increased with added phosphate and was coincident with increases in microbial biomass P, which may indicate that microorganisms actively dissolve Fe oxides and acquire phosphate under P limitation48. Although microorganisms could also incidentally acquire phosphate released into solution during the reductive dissolution of Fe oxides, that scenario does not explain the enhanced dissolution of Fe-P sediments relative to Fe-only sediments. Additional studies are needed to confirm microbial enhancement of Fe oxide dissolution and P acquisition under field conditions.

Phosphate bioavailability to plants and/or microorganisms will potentially increase as ongoing warming thaws permafrost, adding new sources of mineral or organic P to the soil59,60. A decreasing Fe-phosphate sink due to Fe oxide dissolution may also temporarily increase P bioavailability but reduce long-term P storage in soils where Fe-OM complexes form. In other areas where Fe oxides accumulate24, the need for plants to compete with microbes and Fe oxide minerals for remaining phosphate might limit or even negate the effect of carbon sequestration through plant growth. Future studies should explore how Fe-mediated phosphate solubility determines P bioavailability to plants and microorganisms at the ecosystem scale, particularly focusing on impacts to carbon storage and greenhouse gas emissions.

Methods

Site description

Field incubations were conducted along a permafrost thaw gradient from July 1st, 2019, to September 10th, 2019 in Stordalen Mire (Supplementary Fig. 3) (Abisko, Sweden; 68.3495°N, 18.8312°E). Stordalen Mire is in the northern Swedish Subarctic, where peatland soils are underlain by discontinuous permafrost and experience seasonal freeze and thaw. The thaw gradient in our study was previously defined based on shifts in vegetation and hydrology19,61: a palsa environment underlain by permafrost, a bog underlain by partially thawed permafrost that is semiwet and wet ombrotrophic, and a minerotrophic fen with no underlying permafrost and complete water saturation61,62,63,64. The fen receives inflows from palsa and bog drainage and from a nearby lake37. Stordalen Mire is currently experiencing permafrost thaw due to regional climatic warming65 that is resulting in the expansion of the active permafrost thaw zone and conversion of palsa into bog and fen66. Air temperature measured at the Abisko Scientific Research Station, 5 miles northwest of Stordalen, averaged at 11.8 ± 3.0 °C while soil temperature at 5 cm averaged at 9.6 ± 1.4 °C across the study period (Supplementary Fig. 9). Precipitation (measured daily) totaled 73.0 mm during the incubation period with the majority falling in August (56.6 mm) (Supplementary Fig. 9). The 2019 year was wetter than average with an annual precipitation of 403.3 mm (2000–2020 average = 350.3 mm).

Site characterization

Redox potential (Eh) was continuously monitored with five redox potential probes (PaleoTerra) deployed in either the palsa soil (n = 1), the bog (n = 2), or the fen (n = 2). Surface soil Eh is reported as the mean (±standard error) of values recorded in the top 6 cm of soil. Soil conditions were considered Fe(II) oxidizing above a redox potential of +524 mV and Fe(III) reducing below +347 mV, with 347–524 mV representing the range of the Fe(II)-Fe(III) redox couple for ferrihydrite between pH 4–556,67. Soil moisture was continuously recorded in the top 5 cm of palsa soil (ECH2O EC-5, METER) (Supplementary Fig. 9). Additional details regarding the sensors and soil and water collection and characterization are reported in Supplementary Methods.

Surface water and pore water were collected from the bog and fen in early July 2019. Unfiltered surface and pore water were measured for pH, specific conductivity (µS cm−1), and oxidation-reduction potential (mV) that was converted to Eh. One pore water sample from the palsa was acquired by combining water from eight soil moisture samplers (0.15 µm pore size, Simpler) deployed at 5 cm depth. Soil water was filtered through a 0.45 μm cellulose acetate syringe filters (Sterilitech) and acidified with two drops of trace-metal grade nitric acid in the field for dissolved Fe and P analysis by inductively coupled plasma optical emission spectroscopy (ICP-OES).

Shallow soil cores were collected to 5 cm depth in the palsa, bog, and fen on July 11th, 2019. Additional soil cores were collected along a permafrost thaw gradient located northeast of the experimental thaw gradient (Supplementary Fig. 3). Soil cores were composited by horizon (organic or mineral) prior to analysis for pH (1:10 water), salt-extractable nutrients (1:5 soil to 0.5 M K2SO4 solution ratio), and C and N content. Additional small quantities of surface soil (<5 cm) and a homogenized soil core (0–60 cm) were collected from the palsa, frozen, and then freeze-dried within a week of collection.

Iron-rich sediment incubation and characterization

Iron-rich sediments were collected under flowing water in a drainage channel located near the Abisko Research Station (Supplementary Fig. 10). These Fe-rich sediments were selected to represent naturally occurring Fe phases and their associated microbial communities. Sediments were either enriched with phosphate or not and then incubated at 5 cm depth in the soil of either the palsa, bog, or fen for 7 or 70 days. As such, the sediment incubation experiment consisted of three factors. The first factor was the location and included incubations in either the palsa, bog, or fen. The second factor was incubation time (i.e., 0, 7, or 70 days). The third factor was phosphate addition, i.e., phosphate was either experimentally sorbed onto the Fe-rich sediment (Fe-P) or not (Fe-only). Each treatment was replicated three times. Sediments were quickly frozen in the field with liquid N2 (within 2 min of retrieval) and then freeze-dried prior to transport and processing in an anoxic chamber. More details on these methods are provided in Supplementary Methods.

Sequential extractions were performed on freeze-dried sediments to evaluate the loss or gain of water-soluble (dissolved or loosely sorbed Fe), hydroxylamine-soluble (e.g., ferrihydrite and lepidocrocite), and dithionite-soluble (targeting goethite and other crystalline oxides) Fe during sediment incubation40. We infer that soluble reactive P extracted in each step was labile and bioavailable (water-soluble) phosphate or phosphate associated with the extracted Fe phase68. X-ray absorption fine-structure (XAFS) spectroscopy was conducted at beamline 12-BM at the Advanced Photon Source (Argonne National Laboratory) to evaluate changes in sediment Fe composition during incubation. Linear combination fits (LCF) to the XANES region of the spectra were used to analyze the Fe oxidation state and relative proportions of Fe(II) vs. Fe(III) while LCFs to the EXAFS regions were used to determine Fe speciation (detailed methods in Supplementary Information). Major Fe phases determined via EXAFS and reported here are SRO Fe oxides (ferrihydrite + lepidocrocite), organic-bound Fe(III), and “other Fe”, which includes Fe(II) or Fe(II/III) phases representing <15% of total Fe (Supplementary Fig. 11).

Separate mineral bags were prepared and incubated in the same manner as described above but removed after 7 days and analyzed for P concentrations in microbial cells associated with the Fe-rich sediments (microbial biomass P). Briefly, microbial biomass P concentrations were determined by subtracting P concentrations of fresh (non-frozen) sediments extracted with K2SO4 from P concentrations of fresh sediments fumigated with chloroform prior to extraction with K2SO469,70.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.