Cryoturbation leads to iron-organic carbon associations along a permafrost soil chronosequence in northern Alaska

In permafrost soils, substantial amounts of organic carbon (OC) are potentially protected from microbial degradation and transformation into greenhouse gases by association with reactive iron (Fe) minerals. As permafrost environments respond to climate change, increased drainage of thaw lakes in permafrost regions is predicted. Soils will subsequently develop on these drained thaw lakes, but the role of Fe-OC associations in future OC stabilization during this predicted soil development is unknown. To fill this knowledge gap, we have examined Fe-OC associations in organic, cryoturbated and mineral horizons along a 5500-year chronosequence of drained thaw lake basins in Utqiaġvik, Alaska. By applying chemical extractions, we found that ~17 % of the total OC content in cryoturbated horizons is associated with reactive Fe minerals, compared to ~10 % in organic or mineral horizons. As soil development advances, the total stocks of Fe-associated OC more than double within the first 50 years after thaw lake drainage, because of increased storage of Fe-associated OC in cryoturbated horizons (from 8 to 75 % of the total Feassociated OC stock). Spatially-resolved nanoscale secondary ion mass spectrometry showed that OC is primarily associated with Fe(III) (oxyhydr)oxides which were identified by Fe Mössbauer spectroscopy as ferrihydrite. High OC:Fe mass ratios (>0.22) indicate that Fe-OC associations are formed via co-precipitation, chelation and aggregation. These results demonstrate that, given the proposed enhanced drainage of thaw lakes under climate change, OC is increasingly incorporated and stabilized by the association with reactive Fe minerals as a result of soil formation and increased cryoturbation.


Introduction
Increasing soil temperatures induced by climate change are accelerating permafrost thaw.
Since permafrost soils store approximately 1035±150 Pg organic carbon (OC) in the upper 3 m (Hugelius et al., 2014), rapid warming could potentially unlock vast amounts of OC and stimulate release of greenhouse gases to the atmosphere. Thermokarst landscapes are assumed to be most vulnerable to abrupt changes in a warming climate (Turetsky et al., 2020;Walter Anthony et al., 2018). They cover around 20 % of the northern permafrost region and are estimated to store around 30 % of the total OC (TOC) present in the upper 3 m of the circumpolar permafrost region (Olefeldt et al., 2016). As in other arctic lowland regions, thermokarst lakes are the dominating landscape feature of northern Alaska and follow a general thaw lake cycle (Grosse et al., 2013; This is a pre-print submitted to Earth ArXiv and has not been peer reviewed 2003). Lakes form by thermokarst processes and drain due to e.g. ice-wedge or coastal erosion.
After drainage, terrestrial soils develop in the drained thermokarst lake basins on the remaining sediments and ice-wedge growth continues which potentially reforms lakes. It is assumed that the reformation of thaw lakes will be absent in the future, due to insufficient ice wedge growth (Fuchs et al., 2019) which disrupts the thaw lake cycle and promotes the ongoing soil development on drained thaw lake basins. The soils developing on these thermokarst lake basins are increasingly influenced by cryoturbation which redistributes relatively young OC from the topsoil into deeper soil layers and contributes on average 55% to the soil OC density in the active layer and the nearpermafrost surface in permafrost soils from Alaska (Bockheim, 2007).
Previous studies on permafrost soils have intensively focused on the decomposition of the stored OC (Estop-Aragones et al., 2020;Hopple et al., 2020;Schädel et al., 2016), whereas only a few have focused on the stabilization mechanisms of OC which can mitigate the permafrost carbon feedback (Mu et al., 2016;Mu et al., 2020;Mueller et al., 2017;Osterkamp & Romanovsky, 1999;Patzner et al., 2020;Prater et al., 2020;Schuur et al., 2015;Wang et al., 2020). In soils, the fate of OC is determined by an interplay of various physical, chemical and biological components (Lehmann et al., 2020). The accessibility of OC for microorganisms is reduced by physical protection within soil aggregates and by interactions with minerals that stabilize OC via sorption, Despite the recognized role of reactive Fe minerals in the association with OC in permafrost soils, very little is known about the long-term development of Fe-OC associations especially in thermokarst regions. In order to address this knowledge gap, we quantified Fe-OC associations of soil cores from drained thaw lake basins spanning 5500 years of soil development by using selective chemical extractions. We further identified the reactive Fe minerals involved in the association with OC in different soil horizons and soil development stages along the chronosequence with 57 Fe-specific Mössbauer spectroscopy. This was complemented by visually highlighting spatial Fe-OC associations in soil horizons using correlative scanning electron microscopy and nanoscale secondary ion mass spectrometry. This approach allows us to evaluate biogeochemical drivers of Fe-OC associations along the soil profile and during progressive soil development after thaw lake drainage.

Sampling and soil properties
Permafrost soil cores along a chronosequence of drained thaw lake basins near Utqiaġvik in northern Alaska were collected in April 2010 during the sampling campaign described in . Based on vegetation succession and 14 C data, the soil chronosequence is classified into distinct age classes covering young (0-50 years), medium (50-300 years), old (300-2000 years) and ancient (2000-5500 years) drained basins (Hinkel et al., 2003). Briefly, four soil cores of each age class were collected using an 80-150 cm long corer from the Snow, Ice, Permafrost Research Establishment (SIPRE) with 7.5 cm diameter attached to a Big Beaver earth drill apparatus (Little Beaver, US). Samples were transported within 8 h to a cold room in Utqiaġvik where soil description followed. The frozen cores were cut with a chop saw into corresponding soil horizons and subsamples were oven-dried at 60°C for further analysis. Within our study, two soil cores of each age class were analyzed from locations spanning from 71°12´ to 71°16´ N in latitude and 156°25´ to 156°39´W longitude ( Figure SI 1). These soil cores were selected based on the core depth and on soil horizons to capture soil profiles representative for the individual age classes (informed by ). The selected soil cores along the chronosequence differ in maximum core length (sampled soil depth) ranging from 90 to 126 cm and reach the permafrost layer at around 30 to 50 cm depth ( Figure 1). Soil cores of the young age class show little to no cryoturbation and are mainly dominated by one organic horizon followed This is a pre-print submitted to Earth ArXiv and has not been peer reviewed by mineral horizons. The abundance of soil horizons with indications of cryoturbated soil material increases with the age of the drained thaw lake basins. Buried horizons are combined with, and referred to as, cryoturbated horizons in the following discussion as they both lead to a redistribution of organic material into deeper soil layers.  and description of soil horizons of replicate soil cores of the four different age classes. Abbreviations assigned to the distinct soil horizons are described below the figure. Colors indicate organic horizons (brown), cryoturbated/buried horizons (beige) and mineral horizons (grey).

Selective extractions
Soil horizons of two soil cores representative of each age class were selectively extracted to quantify different Fe mineral phases and the associated OC (Cornell & Giovanoli, 1988;Mehra & Jackson, 1960). To prepare for the chemical extractions, glass vials were washed with 1 M HCl, rinsed with deionized water and sterilized for 4.5 h at 180°C. Oven-dried soil samples (0.15±0.05 g) were weighed into glass vials in duplicates and degassed with N2 prior to extraction.
To target poorly crystalline and crystalline Fe(III) (oxyhydr)oxides (in the following considered as the total extractable Fe), samples were extracted anoxically in the dark over 24 h at 70°C with This is a pre-print submitted to Earth ArXiv and has not been peer reviewed 3.125 mL of a 6 M HCl solution (Foucher et al., 2000;Pehkonen, 1995;Peltier et al., 2005;Porsch & Kappler, 2011;Rutledge et al., 2010).
In parallel, a dithionite-citrate bicarbonate extraction was carried out to selectively extract reactive Fe minerals (defined here as reductively dissolvable by dithionite-citrate extraction) such as ferrihydrite, goethite, lepidocrocite, akaganeite and hematite nanoparticles (Cornell & Schwertmann, 2003;Coward et al., 2017;Mehra & Jackson, 1960;Raiswell et al., 1994). This extraction also allows for the quantification of OC that is mobilized during the reductive dissolution of reactive Fe minerals (in the following referred to as Fe-associated OC) (Lalonde et al., 2012;Mu et al., 2016;Mu et al., 2020;Patzner et al., 2020). We followed the dithionite-citrate extraction which is performed for 16 h on a rolling shaker under room temperature and anoxically in the dark (Coward et al., 2017;Wagai et al., 2013;Wagai & Mayer, 2007) and combined different approaches to account for potential difficulties when using this extraction method as previously discussed in detail by Patzner et al. (2020) and shortly discussed in the following. Due to the instability of dithionite in solution (Varadachari et al., 2006), powdered dithionite was added to the sample to reach a final concentration of 0.1 M by adding 3.125 mL of a 0.27 M trisodium citrate, 0.11 M sodium bicarbonate solution (pH 7, N2:CO2 (90:10, v:v) headspace). Sodium bicarbonate was used as a buffer (pH 7) to prevent hydrolysis and re-sorption of organic matter under acidic pH. Fisher et al. (2020) showed that under neutral pH, the dithionite-citrate extraction only partially dissolves reactive Fe minerals, thus leading to an underestimation of the amount of reactive Fe minerals in the analyzed soil cores. Trisodium citrate was used as a complexing agent to prevent re-precipitation of the mobilized Fe and sorption of OC (Mehra & Jackson, 1960). A pre-test carried out under the same conditions (pH, ionic strength) showed that without citrate, 43.5±17.6 % less reactive Fe and 37.2±4.4 % less reactive Fe-associated OC was obtained in an organic horizon of a medium aged soil core (Table SI 1). To account for effects of soil drying on reactive Fe minerals and Fe-associated OC, comparable permafrost soil material from a desiccating palsa in northern Sweden (Patzner et al., 2020) was dried under anoxic and oxic conditions at room temperature and 60°C. We obtained similar reactive Fe and Fe-associated OC contents by the dithionite-citrate extraction method for all drying techniques (Table SI 2).
To also account for the OC which is readily desorbed from the sample and not directly associated with reactive Fe minerals, a 1.85 M sodium chloride (NaCl), 0.11 M sodium bicarbonate solution (pH 7, N2:CO2 (90:10, v:v) headspace) was performed as control extraction under the This is a pre-print submitted to Earth ArXiv and has not been peer reviewed same conditions as the dithionite-citrate extraction (Table SI 3) (Lalonde et al., 2012;Patzner et al., 2020). It has to be noted that the NaCl control extraction was shown to also extract OC that was weakly bound to reactive Fe minerals (Fisher et al., 2020). As the OC quantified by the NaCl control extraction is subtracted from the OC mobilized by the dithionite-citrate extraction, the amount of Fe-associated OC is likely underestimated.

Geochemical analyses
The supernatant from the selective extractions was obtained by centrifugation (10 min, 13400 rpm) and subsequently analyzed for total Fe (6 M HCl extractable) and reactive Fe (dithionite-citrate extractable) using the ferrozine assay (Stookey, 1970) following the protocol from Hegler et al. (2008). Briefly, to quantify total Fe, 80 µL of hydroxylamine-HCl were incubated for 30 min with 20 µL of the supernatant which was diluted in 1 M HCl beforehand.
The dilution of samples ensured that the initial brownish color of the supernatant, resulting from OC, did not interfere with the spectrophotometric quantification. 100 µL of ferrozine was then added, the solution was mixed and incubated for 5 min to allow for the spectrophotometric quantification at 562 nm. Subsequently, reactive Fe contents were calculated by subtracting the Fe readily mobilized during the NaCl control extraction.
OC in the supernatant of the selective extractions and in the trisodium citrate-bicarbonate solution (citrate background) was analyzed by combustion at 750°C (Elemental analyzer, multi N/C 2100S, Analytik Jena GmbH, Germany). The Fe-associated OC was calculated by subtracting the OC background from the trisodium citrate-bicarbonate solution and the NaCl control extraction. As a consequence of the high OC background imposed by the use of citrate, Feassociated OC could not be quantified in four soil horizons (young 1: 56-82 and 82-117 cm, ancient 1: 87-116 cm and ancient 2: 34-50 cm), however the amount of Fe-associated OC was typically high and detectable in most samples.
The TOC content of horizons in soil cores from drained thaw lake basins was previously analyzed in the study of . Briefly, sample material was ground with a ball mill and measured in duplicates via dry combustion (Vario MAX CNS Analyzer, Elementar, Hanau, Germany). The measured carbon contents represent the TOC content due to absence of carbonates. This is a pre-print submitted to Earth ArXiv and has not been peer reviewed Stocks of reactive Fe and associated OC (Eq. 1) were calculated based on content, bulk density (BD) and layer thickness of the horizons in each soil core (Table SI 3). (1) To account for differences along the permafrost soil chronosequence, reactive Fe and associated OC stock of same horizons (organic, cryoturbated or mineral) within one soil core were summed up and averaged between replicate soil cores of one age class to represent the average total sum present in distinct soil horizons within one age class.

Fe mineralogy and spectroscopic analysis
To identify the mineralogy of the reactive Fe minerals potentially involved in the association with OC, 57 Fe-specific Mössbauer spectroscopy was applied. Prior to Mössbauer analysis, dried sample material from organic, cryoturbated and mineral horizons of two endmember soil cores (young, ancient) was mortared and loaded into plexiglass holders (1 cm 2 ), forming a thin film of sample material. Plexiglass holders were tightly closed with parafilm and transmission spectra were collected at 77 K and 5 K using a constant acceleration drive system (WissEL) in transmission mode with a 57 Co/Rh source. All spectra were calibrated against a 7 µm thick α-57 Fe foil that was measured at 295 K. Sample analysis was carried out using Recoil (University of Ottawa) and the Voigt Based Fitting (VBF) routine (Rancourt & Ping, 1991). The half width at half maximum was constrained to 0.124 mm s -1 during fitting.
For visualizing Fe-OC associations with correlative scanning electron microscopy (SEM) and nanoscale secondary ion mass spectrometry (NanoSIMS) analysis, the deepest horizon of an intact frozen soil core (medium 2, Figure 1) was chemically fixed, dried over an acetone row and subsequently impregnated with Araldite 502 (Araldite kit 502, electron microscope sciences, Hatfield, USA). The impregnated soil core was sectioned and polished (see details in Mueller et al. (2017)). Prior to SEM analysis (Jeol JSM 5900LV, Freising, Germany) in backscatter electron mode and NanoSIMS analysis, the intact cross section was gold-coated by physical vapor deposition under argon atmosphere to circumvent possible charging effects. The NanoSIMS measurements were recorded at the Cameca NanoSIMS 50 L (Gennevilliers, France) of the Lehrstuhl für Bodenkunde, TU München, Germany. Electron multiplier secondary ion collectors This is a pre-print submitted to Earth ArXiv and has not been peer reviewed were used for 12 C 14 N -, 27 Al 16 Oand 56 Fe 16 O -. Prior to analysis, impurities and the coating layer were sputtered away by using a high primary beam current and charging during the measurements was compensated by an electron beam generated by the electron flood gun of the NanoSIMS instrument (Mueller et al., 2013). The NanoSIMS data was analyzed using the Look@NanoSIMS plugin for MatLab (Polerecky et al., 2012), images were corrected for detector dead time and drift corrected.

Statistical analysis
An analysis of variance (ANOVA) was applied using RStudio (Version 1.3.959) to identify differences between soil age classes, combined with a post-hoc test to identify age classes that are different from another. Whenever normal distribution or homogeneity of variances was not given, a non-parametric Kruskal Wallis test was performed. For correlation analysis of non-normally distributed data, p-values (p) and correlation coefficients (r) are given for a Spearman (ps and rs) correlation. Additionally, linear regressions (R 2 ) were calculated to identify relationships between reactive Fe and associated OC across age classes.

Results and Discussion
3.1 Changes in reactive Fe and associated OC along the soil chronosequence 3.1.1 Greater differences in reactive Fe and associated OC between soil horizons than with depth Contents of reactive Fe and associated OC did not show a clear trend with depth, but were significantly different between soil horizons across the soil profiles (p<0.001, Kruskall Wallis test) ( Figure 2, Figure 3). Reactive Fe contents were highest in organic horizons (10 to 60 mg g -1 DW-soil -1 ) with a significantly higher content compared to cryoturbated (p<0.001, 5 to 20 mg g -1 DW-soil -1 ) and mineral horizons (p<0.001, 2 to 20 mg g -1 DW-soil -1 ) ( Figure 2, Figure 3).
Cryoturbated horizons incorporate, in comparison to directly over-or underlying mineral horizons, higher amounts of reactive Fe (Figure 2), even though overall differences in reactive Fe between cryoturbated and mineral horizons were not significant.
Horizons with elevated contents of reactive Fe in comparison to over-or underlying horizons were also identified to overlay the permafrost table (Figure 1), subject to annual This is a pre-print submitted to Earth ArXiv and has not been peer reviewed freeze/thaw cycles that experience fluctuations in redox conditions. At such redox interfaces, Fe(III) (oxyhydr)oxides are assumed to be effectively recycled forming poorly crystalline Fe minerals (Moormann & van Breemen, 1978) which counteracts the crystallization of reactive Fe minerals with time. Even though Chen et al. (2020) postulated the formation of more crystalline Fe minerals under redox fluctuations due to oxygen limitation, the association with high amounts of OC, as found in affected soil horizons, impede the crystallization of reactive Fe minerals (Schwertmann, 1966). Additionally, Fe(II) which is reduced in deeper, anoxic soil layers migrates upwards and subsequently is oxidized by oxygen and precipitates as Fe(III) (oxyhydr)oxides (Herndon et al., 2015;Liang et al., 1993). This likely explains highest contents and percentages (as % of total Fe) of reactive Fe found in organic horizons in our soil cores (~100% reactive Fe of the total Fe, Figure SI 2). In accordance to other studies from the same region, we also find highest amounts of poorly crystalline Fe minerals (represented here as reactive Fe) in organic horizons (Herndon et al., 2017) and overall higher amounts of reactive Fe compared to other permafrost regions (Mu et al., 2016;Patzner et al., 2020).
Fe-associated OC contents were also highest in organic horizons (20 to 80 mg g -1 DW-soil -1 ) and are significantly different from cryoturbated (p<0.001, 0 to 40 mg g -1 DW-soil -1 ) and mineral horizons (p<0.001, 0 to 25 mg g -1 DW-soil -1 ) ( Figure 2, Figure 3). It has to be noted that TOC contents were also greatest in organic horizons ( Figure  Cryoturbation redistributes mostly young, undecomposed OC into deeper soil layers (Ping et al., 1998), as was also shown by 14 C analysis for the same permafrost soil samples . The relocated OC in cryoturbated horizons was furthermore characterized to be more bioavailable compared to over-or underlying mineral horizons as indicated by O-N/alkyl C to alkyl C ratios obtained by NMR spectroscopy . Thus, it can be assumed that easily available OC is associated with reactive Fe minerals as the Fe-associated OC constitutes a large proportion of the TOC in cryoturbated horizons. Also, similar amounts of Fe-associated OC are present in other permafrost regions (Mu et al., 2016;Patzner et al., 2020), although Feassociated OC contents could be underestimated as a result of the NaCl control extraction (Fisher et al., 2020). Generally, reactive Fe and Fe-associated OC contents over all soil horizons and age classes correlate significantly (ps<0.001, rs=0.76) (Figure 3). By categorizing the percentage of Feassociated OC into four ranges (0-10, 10-25, 25-50, 50-100 %) (Figure 3), it can be shown that Feassociated OC constitutes a higher proportion of TOC in cryoturbated horizons compared to This is a pre-print submitted to Earth ArXiv and has not been peer reviewed organic and mineral horizons. Cryoturbated horizons show mainly 10-25 % Fe-associated OC, whereas organic horizons have between 0-25 % and mineral horizons have between 0-10 % Feassociated OC.
The depth distribution demonstrates that the content of reactive Fe and associated OC is controlled by soil horizons rather than by depth or age class and is supported by significant differences found between soil horizons.   (Mu et al., 2016) and a maximum of 7 % in soils overlying intact permafrost in Sweden when averaged over the soil profile (Patzner et al., 2020).
Across age classes of the permafrost soil chronosequence, reactive Fe contents are relatively constant (Figure 4 a), but significant differences between age classes were found for the content of Fe-associated OC (p=0.008) between the young and the medium (  Stocks of reactive Fe do not significantly change (6.22±0.33 to 6.40±0.34 kg m -2 from young to ancient, Figure 4 b, Table 1), but the distribution of reactive Fe in soil horizons differs along the soil chronosequence. Cryoturbated horizons store 10 % of the total reactive Fe stock in the young age class and 34-72 % in the older age classes (medium, old, ancient). Fe-associated OC stocks are significantly different between age classes (p=0.0285) (Figure 4 b, Table 1) and significantly increase from the young (3.95±1.59 kg m-2) to the medium age class (10.04±1.81 kg m-2) (p=0.0346). Changes in Fe-associated OC are driven by an increased incorporation of OC in cryoturbated horizons as they increase in abundance along the soil chronosequence.
% in the young age class to 42-75 % Fe-associated OC of the total Fe-associated OC stock in the older age classes (medium, old, ancient) (Figure 4 b). This underlines the shift from mineral horizons contributing most to the Fe-associated OC and reactive Fe stock in young (0-50 years) permafrost soils of drained thaw lake basins to cryoturbated horizons storing most reactive Fe and associated OC with progressive soil development. This is a pre-print submitted to Earth ArXiv and has not been peer reviewed

Differences in Fe mineral crystallinity between soil horizons
The crystallinity of Fe minerals can highly impact their capacity to interact with OC (Herndon et al., 2017). We therefore performed 57 Fe-specific Mössbauer spectroscopy analysis at 77 K and 5 K to identify and quantify the relative abundances of Fe minerals. Mössbauer transmission spectra of organic horizons of the young and ancient age class showed similar properties for spectra collected at 77 K ( Figure SI 5 a (Table SI 3 (Williams et al., 1980). Moreover, the preparation of the sample material under oxic conditions might lead to the abiotic oxidation of such oxygen sensitive ferrous Fe minerals. Taken this into consideration, and the lower abundance of reactive Fe minerals in the total extractable Fe pool compared to organic horizons (Table SI 3), suggests that the observed phase is probably an Fe(II)-bearing phyllosilicate (Poulton & Canfield, 2005;Raiswell et al., 1994). Overall, Mössbauer spectroscopy enabled the identification of ferrihydrite as the poorly crystalline mineral phase most abundant in organic horizons with lower abundance due to a more pronounced Fe(II) phase in cryoturbated and mineral horizons. Fe minerals of higher crystallinity such as magnetite are typically generated by dissimilatory Fe(III) reduction (Lovley et al., 1987) and present in permafrost soils from northern Alaska (Lipson et al., 2010), but are not involved in  (Wagai and Mayer, 2007). OC:Fe mass ratios largely This is a pre-print submitted to Earth ArXiv and has not been peer reviewed exceeding 0.22, indicate co-precipitation or chelation of organic compounds. Samples exceeding this threshold are illustrated as filled symbols, whereas framed symbols are below a ratio of 0.22.
Cryoturbated horizons show a high proportion of particulate organic matter where plant residues can for example promote soil aggregate formation by providing a priming structure for aggregating mineral particles (Mueller et al., 2017;Prater et al., 2020). The intricate connection of reactive Fe minerals, organic matter and the soil matrix in a cryoturbated horizon (medium 2) is illustrated by using combined SEM and NanoSIMS analysis ( Figure 5). The backscattered SEM image showed particulate organic matter (POM) and clay minerals (mi) enclosed by Fe minerals (FeO) in the soil cross section and delimited quartz grains (q) (Figure 5 a). Microscale distributions showed a close spatial association of Fe minerals visualized as 56 Fe 16 O-intertwined with POM as 12 C 14 N-and clay and silt minerals shown as 27 Al 16 O-as detected by NanoSIMS (Figure 5 b, c).
This highlights the role of reactive Fe minerals in retaining OC in soil horizons with emphasis on cryoturbated horizons. Thus, we propose that OC is primarily co-precipitated, chelated or aggregated with reactive Fe minerals.   Mueller et al. (2017)).
The type of Fe-OC interactions seems to be governed to a certain extent by the amount of TOC present in soil horizons, as supported by significant correlation between OC:Fe mass ratios and TOC content (ps=0.0003, rs=0.5230). Furthermore, the association of OC seems not to be limited by the amount of reactive Fe in our permafrost soil cores, in contrast to Fe-OC associations in forest soils . The Fe-associated OC in our study largely exceeds the theoretical maximum of sorbed OC onto Fe (oxyhydr)oxides, calculated as 0.22 times the reactive Fe content (Herndon et al., 2017) (Figure SI 4). High OC:Fe mass ratios in distinct soil horizons can possibly be explained by changes in solubility of redox active metals (Eusterhues et al., 2005), as it can also be observed at the redox interface in permafrost affected soils.
Besides the amount of OC, pH also influences the nature of the Fe-OC association (Wagai & Mayer, 2007). On average, higher OC:Fe mass ratios can be found in older age classes where pH is generally lower and TOC is higher compared to the young age class (young = 0.55 ± 0.54, medium = 1.85 ± 0.76, old = 1.03 ± 0.59, ancient = 1.68 ± 1.65) . In summary, our results show that Fe-OC associations are primarily formed by co-precipitation, chelation and aggregation in drained thaw lake basins of northern Alaska which is in accordance with findings from other permafrost regions (Mu et al., 2020;Patzner et al., 2020). Contrary findings in the same region, where Fe-OC associations are restricted to sorption (Herndon et al., 2017), highlights the heterogeneity of Fe-OC associations and the need for further investigations.

Potential fate of Fe-OC associations under climate change
The molecular diversity was long thought to be the main factor limiting the decomposition of organic matter. Lehmann et al. (2020) recently underlined that spatial and temporal variations in the soil shape the functional complexity and collectively determine OC persistence. As permafrost soils are highly variable ecosystems with annual freeze-thaw cycles, Fe-OC associations are considered to play an important role in retaining OC under a changing climate.
Increasing permafrost thaw can result in waterlogged and anoxic conditions under which Fe minerals are dissolved by dissimilatory Fe(III) reducing microorganisms that releases the Feassociated OC (Patzner et al., 2020). By this, greenhouse gas emissions could be exaggerated by microbial Fe(III) reduction which is directly coupled to OC mineralization leading to CO2 emissions (Lovley et al., 1987). As well as by the reductive dissolution of reactive Fe minerals mobilizing the previously bound OC which becomes available for microbial respiration.
Additionally, Fe(III) reduction can also be coupled to other microbial processes such as anaerobic This is a pre-print submitted to Earth ArXiv and has not been peer reviewed oxidation of methane which influences greenhouse gas emissions (van Huissteden et al., 2011) and thus the response of permafrost soils to climate change.
At the Arctic Coastal Plain of northern Alaska, microbial Fe(III) reduction contributes up to 63 % to the total respiration (Lipson et al., 2013) which emphasizes the important role of Fe minerals in the association with OC. Thermokarst regions of the Arctic Coastal Plain in Alaska are assumed to shift from a stable (Hinkel et al., 2007) to a susceptible environment under future climate (Nitze et al., 2020). Due to the lack of ground ice-wedge build-up under future climate scenarios, it is expected that thermokarst lakes will no longer follow the long postulated thermokarst lake cycle where reformation of thermokarst lakes was expected after around 5500 years of drainage (Bouchard et al., 2017;Fuchs et al., 2019;Jorgenson & Shur, 2007).
Consequently, OC introduced by cryoturbation can be associated and stabilized by reactive Fe minerals under increasing soil drainage which promotes oxic conditions (Herndon et al., 2020).
This association potentially preserves OC over thousands of years mitigating the rapid loss of respired C as greenhouse gases to the atmosphere.

Conclusions
We found that high amounts of OC are associated with reactive Fe minerals, identified as ferrihydrite, in soils forming on drained thaw lake basins in northern Alaska.

Conflicts of Interest
The authors declare no conflict of interest.   Wagai and Mayer (2007) following Herndon et al. (2017) This is a pre-print submitted to Earth ArXiv and has not been peer reviewed   Supplementary methods