Previous research showed that anaerobic respiration using iron (Fe) oxides as terminal electron acceptor contributed substantially to ecosystem respiration (ER) in a drained thaw lake basin (DTLB) on the Arctic coastal plain. As DTLBs age, the surface organic layer thickens, progressively burying the Fe-rich mineral layers. We therefore hypothesized that Fe(III) availability and Fe reduction would decline with basin age. We studied four DTLBs across an age gradient, comparing seasonal changes in the oxidation state of dissolved and extractable Fe pools and the estimated contribution of Fe reduction to ER. The organic layer thickness did not strictly increase with age for these four sites, though soil Fe levels decreased with increasing organic layer thickness. However, there were surprisingly high levels of Fe minerals in organic layers, especially in the ancient basin where cryoturbation may have transported Fe upward through the profile. Net reduction of Fe oxides occurred in the latter half of the summer and contributed an estimated 40–45% to ecosystem respiration in the sites with the thickest organic layers and 61–63% in the sites with the thinnest organic layers. All sites had high concentrations of soluble Fe(II) and Fe(III), explained by the presence of siderophores, and this pool became progressively more reduced during the first half of the summer. Redox titrations with humic acid (HA) extracts and chelated Fe support our view that this pattern indicates the reduction of HA during this interval. We conclude that Fe(III) and HA reductions contribute broadly to ER in the Arctic coastal plain.
 The contribution of anaerobic respiration to the global carbon (C) cycle is currently underappreciated. A large proportion of the planet's soil C lies in Arctic peatlands, where permafrost blocks drainage leading to saturated, anoxic conditions [Tarnocai et al., 2009]. The height of the water table controls diffusion of oxygen (O2) into the soil, and so it has often been assumed that besides temperature, water table height is the ultimate control over soil respiration in such systems [Billings et al., 1982; Oechel et al., 1998; Olivas et al., 2010; Sommerkorn, 2008]. However, significant rates of soil respiration can occur anaerobically, given the availability of alternative electron (e−) acceptors [Zona et al., 2012]. Few studies have focused on how the availability of alternative terminal e− acceptors controls the flux of carbon dioxide (CO2) from soils. The use of ferrous iron (Fe(III)) and humic substances (HS) by microbes as terminal e− acceptors in anaerobic respiration has been studied extensively in wetlands and sediments [Heitmann et al., 2007; Keller et al., 2009; Knorr and Blodau, 2009; Lovley and Phillips, 1987; Lovley et al., 1996; Roden and Wetzel, 1996], but their importance in Arctic peat soils has only recently been recognized [Lipson et al., 2010, 2012].
 Drained thaw lake basins (DTLBs) cover roughly half of the landscape in the Arctic Coastal Plain near Barrow, Alaska [Hinkel et al., 2003]. DTLBs form when lakes drain due to erosional processes, and subsequently become vegetated [Billings and Peterson, 1980]. Over a roughly 5000 year cycle, the organic layers increase in thickness, and microtopography develops due to ice-wedge polygons and similar processes [Bockheim et al., 2004; Hinkel et al., 2003]. Our previous work has focused on a drained thaw lake basin (DTLB) in the Arctic Coastal Plain near Barrow, Alaska. It was found that Fe(III) was abundant both in dissolved forms and as solid-phase minerals, that the microbial community contained close relatives of known Fe(III)-reducing species, and that Fe(III) reduction could account for most of soil CO2 flux during short-term experiments with added Fe(III) and contributed an estimated 30% of ecosystem respiration over the summer [Lipson et al., 2010]. Redox cycling of Fe shapes the fine scale landscape patterns of soil chemistry at this site, and the abundance of Fe(III) and HS appears to limit the emission of methane (CH4) from this site, by providing a more thermodynamically favorable alternative metabolic pathway under anoxic conditions [Lipson et al., 2012].
 This previous work was done in a single, medium-aged DTLB. However, the most common age classification for DTLB is “old,” covering 30% of the region [Hinkel et al., 2003]. Therefore, we aimed to extend our understanding of Fe(III) reduction across the full range of basin ages, to see whether the importance of this process is widespread across the dominant landforms of the Arctic Coastal Plain, and how it changes with soil development. As DTLB ages, the organic layer tends to thicken, and so the underlying mineral layers become more deeply buried. For ancient DTLB, the organic layer is often so thick that the underlying mineral layer would be completely buried in permafrost. Therefore, we hypothesized that Fe availability decreases with soil age, assuming that the mineral horizon is the primary source of Fe for redox reactions occurring at the surface. This question has implications for the C cycle in this ecosystem as the Arctic climate warms and active layers deepen. Moreover, we asked to what extent does reduction of HS contribute to anaerobic respiration, and how does it interact with Fe(III) reduction? Previous work has given some indirect evidence for the role of HS as e− acceptors in this ecosystem [Lipson et al., 2010; Lipson et al., 2012], but specific information about HS in these soils is lacking.
2.1 Study Area
 This study took place in four DTLBs near Barrow, Alaska. These were classified by Hinkel et al.  as young (0–50 years), medium (50–300 years), old (300–2000 years), and ancient (2000–5500 years). The locations of these four basins are young, 71°15.363′N, 156°37.538′W; medium, 71°17.263′N, 156°35.388′W; old, 71°17.172′N, 156°38.663′W; and ancient 71°14.901′N, 156°34.861′W. Soils are classified as typic histoturbels (surface organic layers 20–40 cm thick) and aquiturbels (surface organic layers less than 20 cm thick), with an organic layer overlying fine, generally silt-rich lacustrine sediments [Bockheim et al., 2001]. Transects approximately 70 m in length were established in each basin with plastic matting to protect the tundra. Ten soil water samplers (Rhizon, type-MOM for metal studies, Sunvalley Solutions, Florida) were installed along each transect at intervals of approximately 6 m at a depth of 0–10 cm. Water table wells (perforated polyvinyl chloride (PVC) tubes) were installed in six locations per transect. Six PVC collars, 10 cm in diameter, were installed for respiration chamber measurements.
2.2 Field Measurements
 Measurements of pH, oxidation-reduction potential (ORP), dissolved oxygen (DO), and electrical conductivity were made at 10 cm depth (after making incisions in the tundra with a serrated knife) with Thermo-Orion electrodes and a portable meter. ORP was measured relative to an internal Ag/AgCl electrode and converted to standard redox potential, Eh (relative to standard hydrogen electrode), using the correction factors supplied by the manufacturer (+212 mV at 5 °C, +200 mV at 20 °C). Proper electrode functioning was tested periodically using saturated solutions of quinhydrone in buffers of pH 4.0 and 7.0. Thaw depth was measured by probing with a stainless steel rod, and water table was measured by insertion of a wooden dowel into the water table wells. Soil respiration was measured on three dates in September in seven to nine replicate polyvinylchloride collars (10 cm diameter) within each basin, using a PP Systems soil chamber (SRC-1) and portable IRGA (EGM-4). Curves of CO2 versus time were fit quadratically to estimate initial respiration rates.
2.3 Soil Measurements
 In early June 2010, when soils were still frozen, soils were collected using a masonry hole bit (3 cm diameter) and portable electric drill. Four horizontal sections, approximately 6 cm deep, were retrieved sequentially from each sampling location and immediately placed in pre-weighed 100 mL bottles containing 2 M HCl. Bottles were weighed to calculate the wet weight of soil in each extraction. Six such replicates, chosen at random, were collected from each basin (except the medium basin, from which only four were collected on this date). After 24 h, the acid extracts were filtered and refrigerated (and in the dark) until analysis of Fe (less than 2 weeks). The acid-extracted soil was rinsed with distilled H2O and then dried to estimate dry mass of soil in each extraction. Soil organic matter (OM) content was determined by loss on combustion at 550 °C. Total Fe and Fe(II) in the acid extracts were measured using 1,10-phenanthroline, with and without ascorbic acid, respectively, and Fe(III) was calculated by difference [Analytical Methods Committee, 1978; Lipson et al., 2010]. In calculations of soil Fe content, the volume of the extract was adjusted to include the water content of each sample (calculated from the wet weight minus the dry soil residue). The bulk density (BD, g dry mass cm−3) of these samples was estimated from the core dimensions and the water content of each sample. A smaller number of deeper cores were also collected in early June using a SIPRE core and a gas-powered drill. These samples were sliced into discrete horizontal sections, the wet mass of each was weighed, and a subsample was used for water and OM content to determine BD. Additional freeze-dried subsamples subsamples of these cores were used for determination of cation exchange capacity (CEC), using the compulsive exchange method with BaCl2 [Hendershot and Duquette, 1986]. Cations were analyzed on acidified filtered extracts, diluted 1:8 with lab deionized water, and analyzed with a Thermo ICAP 6300 Duo View ICP spectrometer. Exchangeable acidity was estimated by binding exchangeable Al3+ with NaF and titrating to a phenolphthalein endpoint. We calculated the porosity-weighted CEC (Qv), a property that directly relates to the bulk conductivity of soils [Cosenza et al., 2008], assuming an average solid density of 2.65 g/cm3 for silicaceous minerals and 1.5 g/cm3 for organic matter [Skaven-Haug, 1972].
 For the later soil sampling dates in early August, early September, and late September, when soils were thawed, a different approach was used. Soils were extracted using a Russian peat corer (Rickly Hydrological Company, Columbus, Ohio). Soil samples averaging about 24 cm in depth were removed and divided into four horizontal sections, noting the transition from organic to mineral layers when possible. Each section was sliced longitudinally using a serrated knife, and one piece was immediately extracted in pre-weighed 100 mL bottles containing 2 M HCl and analyzed for Fe as described above. The other piece was sealed in a Bitran ziplock bag and later used to calculate water and OM content. Because the cores extracted using this method were not regularly shaped and were deformed during collection, no direct measure of BD was possible. However, a highly consistent (R2 = 0.899 for the multiple regression including basin as categorical variable) logarithmic relationship between BD and OM for each basin was established using the June samples, and so BD was calculated from OM using this relationship (Figure S1 in the supporting information). This relationship is similar to that reported by previous authors [Bockheim et al., 2001]. The relationship between BD and ln(OM) for all basins fell into two distinct groups:
 If a calculated value exceeded the maximum or was less than the minimum measured value, the maximum or minimum measured value was used instead, so that all calculated values remained within the measured range of BD.
 To estimate net rates of Fe(III) reduction, we calculated seasonal changes in the total amounts of Fe(III) in the upper 24 cm of the active layer (corresponding to the average sampling depth). To do this, we assumed that total Fe levels do not change throughout the year, as water is mainly lost through evapotranspiration from these basins and the annual rate of Fe deposition from dust should be negligible compared to the large amounts stored in these soils. Therefore, we assumed seasonal variations in total Fe arise solely from random variation and sampling artifacts, and estimated total extractable Fe in each layer of each basin by averaging data across all dates to obtain the most robust estimate possible. For each sampling date, Fe(III) in each layer was obtained by multiplying the ratio of Fe(III)/total Fe in the HCl extracts times the seasonal average of total Fe per layer, and the layers were then summed across the profile.
 Additional cores were collected in early August, immediately sealed in two layers of Bitran and frozen for later analysis of Fe minerals by sequential extraction [Poulton and Canfield, 2005]. This method employs sequential extractions of various solvents to operationally define various mineral fractions. MgCl2 (1 M) is first used to remove Fe associated with the cation exchange complex, sodium acetate (pH 4.5, 48 h, 50 °C) extracts Fe carbonates such as siderite and ankerite, hydroxylamine (1 M in 25% acetic acid, 48 h) extracts easily reducible oxyhydroxides such as ferrihydrite and lepidocrocite, sodium dithionite (50 g L−1, pH 4.8 with 0.35 M acetic acid/0.2 M sodium citrate, 2 h) extracts more crystalline reducible oxides such as goethite and hematite, and ammonium oxalate (0.37 M, pH 3.2, 6 h) extracts magnetite. The first two steps of this extraction (MgCl2 and Na acetate) were performed under an N2 headspace with N2-bubbled solutions to prevent oxidation of delicate Fe(II) phases. Soils (5–10 g wet weight) were sliced and weighed while still frozen, placed into 50 cc screw cap tubes, and allowed to thaw under N2 during extraction (25 mL of each solution was used). Soils were centrifuged and supernatants were removed for analysis of total Fe after each extraction step.
2.4 Soil Pore Water Measurements
 Soil pore water samples were collected every 1–2 weeks from each basin into vacutainers (Becton Dickinson) attached to the water samplers with hypodermic needles. Samples were preserved by acidification to pH <2 using HCl added though the septa with a syringe. Samples were wrapped in a Bitran bag to further avoid oxidation and stored refrigerated and in the dark until analysis. Fe(II) and Fe(III) were analyzed using 1,10-phenanthroline as described above. Siderophores were measured using chrome azurol S, as described by Alexander and Zuberer  and modified by Lipson et al.  (to avoid interference from soluble Fe in these samples, samples were diluted by 10 and the amount of Fe in the reagent was increased). Deferoxamine (Sigma Chemical, St. Louis, Missouri) was used as a standard. Dissolved organic C (DOC) was measured with a colorimetric assay based on Mn(III) [Bartlett and Ross, 1988] modified for microplate reader [Lipson et al., 2012].
2.5 Extraction and Characterization of Humic Acids
 Humic acids (HA) were extracted from soils with a method similar to that described by Agnelli et al. . Frozen soils (5–10 g wet weight) were placed in 50 cc polypropylene screw cap tubes under an N2 atmosphere and extracted overnight in N2-bubbled 0.1 M NaOH/0.1 M Na4P2O7. HA were precipitated by acidification to pH 1 with HCl and redissolved in N2-bubbled NaHCO3 (50 mM). Redox titrations were performed similarly as described by Struyk and Sposito . HA samples were precipitated with HCl and redissolved in K2HPO4 buffer (pH 7.0, 0.1 M). One set of samples was chemically reduced for 48 h under 10% H2/90% N2 atmosphere in the presence of Pd catalyst (10% on activated charcoal, 2.5 mg/10 mL) [Peretyazhko and Sposito, 2006]. Samples were then flushed with N2 to remove excess H2. Samples were injected into an apparatus constructed from a 50 cc screw cap polypropylene tube, with the cap modified with three holes to accept an ORP electrode (Thermo-Orion 9179), an inlet for N2 gas, and an outflow port (which was also used as an injection port for liquids, with the gas flow turned off). Chemically reduced samples were titrated in this apparatus with solutions of K3Fe(CN)6. The electrode was allowed to equilibrate for 10 min after addition of titrant before recording the value. At the end of the 10 min interval, the rate of change in ORP was less than 2 mV min−1, as deemed acceptable in other studies [Struyk and Sposito, 2001]. Another subset of HA samples was allowed to oxidize by exposure to the atmosphere for several days. ORP was monitored in one sample during this process and the oxidation was essentially complete after several hours. These samples were titrated using solutions of Fe(II) citrate prepared by adding FeSO4 powder to a vial of degassed sodium citrate buffer (pH 5.0), replacing the stopper, and immediately re-flushing the headspace with N2. The midpoint redox potential was estimated from each titration curve by locating the inflection point [Struyk and Sposito, 2001]. The mass of HA in each solution was measured by precipitating subsamples into pre-weighed microcentrifuge tubes, rinsing with 1 mM HCl, and re-weighing after drying at 60 °C. The relationship between Eh and pH for HA was investigated in a subset of samples by resuspending HA in K2HPO4 buffers (0.1 M) of pH 6, 7, and 8, recording ORP, and fitting with linear regression. In separate experiments, fresh HA extracts (in NaHCO3 buffer) were incubated 1:1 with Fe(II) citrate (20 mM) for 30 min. After incubation, HA were precipitated with HCl and the amount of Fe(III) produced was measured in these samples and in controls with either no HA (to control for spontaneous oxidation of Fe(II) citrate) or with no Fe(II) citrate (to control for Fe(III) liberated from HA extracts).
2.6 Statistical Analysis
 General linear models and simple regressions were used for most analyses. Analyses of covariance (ANCOVAs) were used to compare basins (as a categorical variable) while controlling for date or depth (as continuous variables), and to detect differences in the slopes for these relationships among basins. In some cases, seasonal patterns were nonlinear and so curves were fit quadratically by including (date)2 in the model. Analyses of variance (ANOVAs) were run separately on each soil layer to compare OM and BD between basins at each depth. Tukey's test was used to adjust alpha levels in post hoc contrasts. A two-way ANOVA (with month and basin as the two factors) was used to test for differences among basins in net Fe(III) reduction rates. Standard rules of error propagation were used to combine errors of the source terms in estimates of the contribution of Fe(III) reduction to respiration [Taylor, 1997]. This analysis included error terms for (1) the seasonal change in redox state of extractable Fe in each soil layer, (2) total Fe content in each layer, (3) error resulting from summing the four layers, and (4) soil respiration in each basin. Data were log transformed when appropriate to fit assumptions of normality. Probability (P) value less than 0.05 was considered to be significant.
3.1 Soil Properties and Field Conditions
 As expected, the young basin had the thinnest organic layer, while the ancient basin had the thickest, generally exceeding the depth of our soil cores (Table 1). The OM profiles corresponded with these measurements: OM content dropped off more quickly with depth in the young basin than in the ancient basin. Unexpectedly, the medium basin had a thicker organic layer than the old basin. This is less clear in the OM profile, as the deepest layer of the medium basin averages together samples with deep organic layers and mineral layers with lower OM content than in the other basins (hence the larger standard error). The BD profiles essentially show the inverse pattern relative to OM content. In the two-way ANOVA for OM, basin, layer, and their interaction were all highly significant (P < 0.001); for BD, basin was not significant, but layer and the basin × layer interaction were (P ≤ 0.001).
Table 1. Properties of Soils From the Four DTLBs Used in This Studya
Values are means and standard errors. Across each row, values with the same letter (or no letter) are not significantly different by Tukey's test. For cation exchange capacity (CEC) and pore volume specific-CEC (Qv), the overall tests for differences among basins were marginally significant (P = 0.092 and 0.086, respectively). BS = base saturation, EC = electrical conductivity.
Organic layer thickness (cm)
Organic matter content (%)
Bulk density (g/cm3)
mean of all layers
 On average, CEC and Qv (CEC expressed per unit pore volume) were highest in the old and lowest in the medium basin, though differences among basins were only marginally significant (Table 1). Base saturation decreased with basin age. The exchange complex of the four basins was dominated by Ca, Fe, and Mg, together accounting for 75% of CEC on average (data not shown). Electrical conductivity of the soil pore water, measured at 10 cm depth, was highest in the medium basin and lowest in the old and ancient basins (Table 1).
 All four basins showed similar seasonal patterns in hydrology and electrochemistry (Figure 1). As the water table dropped due to evapotranspiration, DO and Eh (at 10 cm depth) increased (date in ANCOVA: P < 0.001 for both). The seasonal pattern of pH was more complex, initially decreasing and then increasing in the second half of the summer (the quadratic term, date2, was significant for each basin).
 The young and medium basins were more reducing than the old and ancient basins (all contrasts among basins in the ANCOVA were significant at P ≤ 0.001, except ancient versus old (P = 0.616). Eh and log-transformed DO were correlated (r = 0.572, slope = 54.7 ± 3.9, P < 0.001), and so the trends for DO were similar (all contrasts among basins were significant in the ANCOVA except young versus Medium; Figure 1d). The young basin was generally the least acidic and the old basin was the most acidic (all contrasts among basins were significant except ancient versus Medium; Figure 1a). The medium site had higher water table heights than the other basins (all contrasts were significant except young versus ancient and ancient versus Old; Figure 1c).
3.2 Solid-Phase Fe in Soils
 The amount of HCl-extractable Fe closely followed the mineral content of the soil, with higher levels in the deeper, mineral-rich layers (Figure 2a). When integrated over the soil profile, the lowest Fe content was found in the medium and ancient basins, those with the thickest organic layers and hence the deepest mineral layers (Figure 2b). Based on the sequential extractions carried out on a single date in August (Figure 2c), all four basins contained minerals which can serve as substrates for Fe(III) reduction (the reducible and easily reducible Fe oxides, such as goethite and ferrihydrite, respectively), as well as minerals that can result from Fe(III) reduction (siderite and magnetite). The sum of these fractions agreed well with total Fe in HCl extracts of the same soils (r = 0.833, slope = 0.884 ± 0.101, P < 0.001, data not shown) and show a similar pattern overall, with soils of the young basin containing the highest amounts of all mineral fractions.
 The redox state of Fe in soil HCl extracts changed seasonally and with depth (Figure 3). As expected, the ratio of Fe(II)/total Fe increased with depth on all dates, reflecting more reducing conditions deeper in the profile. The seasonal pattern, however, was somewhat surprising: Fe became more oxidized at all depths between mid-June and early August and then became more reduced from early August to late September (the data for early September closely overlap those for June, and so were omitted for clarity). This pattern is contrary to that expected from the changes in water table, DO and Eh (Figure 1), which show more reducing conditions in the first half of the summer.
 The seasonal changes in the reduction state of soil Fe with depth shown in Figure 3 were used to calculate net changes in the Fe(III) pool across the entire profile for each basin, assuming that total soil Fe stays constant throughout the season (Figure 4). Net reduction occurred in all basins between early August and late September. These values shown in Figure 4 were used to calculate the contribution of Fe(III) reduction to ecosystem respiration over this interval (Table 2). Net Fe(III) reduction was higher in the young basin than the others (P = 0.002 for month × basin interaction in two-way ANOVA. This interaction was not significant when young basin was excluded). Expressed as a percentage of respiration measured in chambers during this interval, Fe(III) reduction contributed more in the young and old basins (61–63%) than in the medium and ancient basins (43–45%) (assuming 4 mol Fe per mol CO2). Soil respiration was only marginally different among the basins (P = 0.087).
Table 2. Estimates of Net Fe(III) Reduction in the Upper 24 cm of Soils From the Four Basins Between the Early August and Late September Sampling Datesa
Change (g Fe (III)/m2)
Fe Reduction (g CO2 m−2 h−1 equiv.)
(g CO2 m−2 h−1)
Net changes over this period are expressed as hourly rates for comparison to chamber-based measurements of ecosystem respiration (average of three sampling dates in September), assuming 4 mol Fe reduced per mol CO2 formed. FeR/ER is the percentage of ecosystem respiration that could be accounted for by this estimate of Fe(III) reduction. Values in parentheses are standard errors.
3.3 Soluble Fe in Soil Pore Water
 Concentrations of dissolved Fe in soil pore water followed a different pattern than that of solid phase Fe, being higher in the young and medium basins than in the old and ancient basins (Figure 5). Siderophores and DOC followed a similar pattern. High concentrations of Fe-chelating organics such as citrate can interfere with the siderophore assay [Schwyn and Neilands, 1987], but given the DOC levels in Figure 5a, such interference is unlikely. Siderophore concentrations were highly correlated with dissolved Fe (Figure 5b). These patterns are explained in part by Eh: higher dissolved Fe is seen at lower Eh (Figure 6), which is lowest in the young and medium basins (Figure 1b).
 The redox state of dissolved Fe became more reducing throughout the season in all four basins (Figure 7). In all except the young basin, this pattern was quadratic, with a more rapid increase in the first half of the season. This seasonal pattern was distinct from that of solid phase Fe, which became more reducing in the second half of the season, and from water table, DO, and Eh, which showed more oxidizing conditions as the season progressed.
3.4 Redox Properties of Humic Acids
 We interpret the ratio of Fe2/total Fe in the soil pore water (Figure 7) as a reflection of the redox state of the HA with which they are in contact (see discussion below). To provide support for this interpretation, we investigated the capacity for HA extracted from soils to undergo oxidation-reduction reactions with chelated Fe(II) and Fe(III). Figure 8 shows redox titrations of various HA extracts from the ancient basin. HA that had spontaneously oxidized in the presence of air (to an Eh of ~320 mV at pH 7) were reduced by Fe(II) citrate, and HA that had been chemically reduced with H2/Pd (to values in the range of 80–160 mV at pH 7) were oxidized by Fe(III) (in the form of K3Fe(CN)6). Based on changes in the slopes of these curves (Figure 8b), the capacity to exchange e− fell in the range of 0.2–0.5 meq e− g−1 HA for these samples. This is in agreement with another experiment, in which ancient basin HA were incubated with excess Fe(II) citrate (10 mM) and conversion to Fe(III) was measured, leading to an estimate of 0.38 ± 0.13 meq g−1 HA. Based on this value, yields of HA extractions from these soils (~20 mg g−1), BD measurements, and the respiration data in Table 2, this e− accepting capacity could account for about 20% of soil respiration in the ancient basin, assuming it was concentrated during the first half of the summer. From the data in Figure 8, we estimate the midpoint potential at pH 7 (E°ʹ) for these HA to be 245 ± 16 mV, similar to the value of 234 mV reported by Osterberg and Shirshova . Between pH 6 and 7, the Eh-pH slope was −16.9 ± 0.9 mV. This is similar to the relationship found by Struyk and Sposito  and corresponds approximately to a ratio of one proton consumed per three electrons reduced (q/n ~ 1/3).
3.5 Relationship of Field-Measured Eh and pH to Potential Reduction of Fe and HA
 Based on mean conditions at our site, reduction of amorphous Fe oxides is thermodynamically favorable at 10 cm depth for a majority of the data set (i.e., most points fall below the solid black line in Figure 9). The E° value of 1.06 V used for Fe(OH)3 [Reddy and DeLaune, 2008] also falls within the range reported for ferrihydrite [Straub et al., 2001]. This analysis does not include the effects of siderophores or other Fe-binding ligands, which can increase rates of Fe reduction via solubilization, but can alter the thermodynamics in either direction depending on the relative affinity for Fe2+ or Fe3+ [Thamdrup, 2000]. Similarly, using the E° for HA calculated from the titration and pH experiments, and using the seasonal changes the redox state of dissolved Fe as a proxy for the redox state of HA, reduction of HA can occur readily early in the season (Figure 9, solid gray line) but will halt as oxidized HA accumulate (Figure 9, dotted gray line).
 The quantity of Fe oxides potentially available to fuel Fe(III) reduction was inversely related to the thickness of the organic layer, and declined somewhat with basin age, though not in a linear fashion. Given that there is variability in organic layer thickness within each age category [Bockheim et al., 2004] and that we only studied a single representative of each category, it is not surprising that the organic layer of the medium DTLB exceeded that of the old DTLB. However, the substantial presence of Fe in the active layer of the ancient basin, despite the depth of the buried mineral layer, was contrary to our initial hypothesis. This hypothesis did not take into account the high solubility of Fe in these soils and the fact that even ancient soils were once young: Fe dissolved from the mineral layer at younger developmental stages remains in the profile as the organic layer grows, cycling continuously among oxidized and reduced forms, in solution and as minerals precipitated in the surface organic layer. Cryoturbation (mixing due to freeze-thaw cycles) presumably also contributed to the transport of minerals upward through the profile. It is estimated that over half of the C stored in active and upper permafrost layers of these soils has been redistributed through cryoturbation [Bockheim, 2007]. Frost boils, where mineral material from depth was pushed up to the surface, are fairly common within the ancient basin. Also, shallower mineral deposits were found in rims of ice-wedge polygons in this basin. It is possible that solubilized Fe could migrate across the landscape from high to low areas. Finally, there are atmospheric sources of Fe to this region, mostly in the form of mineral aerosols from Asian deserts [Stone et al., 2005; Zdanowicz et al., 2006], though industrial emissions also may have an impact [McConnell and Edwards, 2008], as might wildfires and volcanos [Tomasi et al., 2007].
 In addition to having higher quantities of Fe in the active layer, the young basin also had higher net Fe reduction rates than the older basins. It is likely that this pattern is generally true for young DTLB, given their shallow organic layers (and hence more accessible mineral layers) and more consistently anaerobic conditions given the flatter, less developed topography. The flatter topography of young and medium-aged basins in general likely leads to lower Eh and hence higher Fe solubility, especially given the production of siderophores under these conditions. Conversely, the enhanced micotopography and increased aeration of old and ancient DTLB should generally reduce soluble Fe concentrations. The lower Fe levels in the ancient basin corresponded with lower net Fe reduction rates. More complex, oxic landscapes of ancient DTLB could help compensate for reduced anaerobic respiration. Additionally, reduction of HA could contribute significantly to ecosystem respiration (ER), as calculated earlier. The trend toward reduced chamber measurements of ER with increased DTLB age in the current study is consistent with that found by a more comprehensive and replicated study of CO2 fluxes using an eddy covariance method [Zona et al., 2010].
 This study has potential climate change ramifications for ecosystem respiration in DTLB of the Arctic Coastal Plain. An increase in active layer thickness brought on by warmer conditions could increase the availability of Fe(III) for anaerobic respiration, especially in ancient DTLB where the organic-mineral boundary currently lies in the upper permafrost. Furthermore, extremely warm summers with high rates of evapotranspiration can lead to very low water tables, as occurred in 2007 [Olivas et al., 2010]. These extreme events could more strongly re-oxidize Fe minerals in the active layer, increasing the potential for anaerobic respiration when flooded conditions return.
 The amount of late-season respiration that could be accounted for by reduction of Fe(III) was impressive, especially considering that the values were conservative in only including the upper 24 cm of the active layer (the average depth sampled in the study), while the lower 10–15 cm of the active layer could have also contributed. Furthermore, the soil chamber measurements probably included some plant respiration (though plants are starting to senesce in September), and if plant respiration were excluded, Fe(III) reduction would account for an even higher fraction. On the other hand, this analysis assumed a molar ratio of four Fe reduced per CO2 respired, and fermentative processes could reduce Fe with a lower yield of CO2 [Lovley and Phillips, 1986]. Also, CH4 emissions were not included in our estimate of ER, though these are generally small in comparison to CO2 for this system [Lipson et al., 2012; Zona et al., 2010; Zona et al., 2009]. Similar arguments apply to the estimate of HA reduction's contribution to ER in the first part of the season in the ancient basin (~20%). This estimate required the assumption that the complete e− accepting capacity of HA was used during this interval, although this assumption is reasonable based on our thermodynamic analysis and the use of the soluble Fe pool as a proxy for HA redox state. In any case, both the HA and Fe pools represent sizeable electron sinks, and the current understanding of ER in the Arctic should be revised to include these pathways.
 It was surprising to observe net oxidation of Fe at depths below the water table, especially during the first half of the season when conditions were more reducing. This pattern was corroborated by seasonal trends in pH, which is lowered when Fe is oxidized and increases when Fe is reduced because of the production and consumption, respectively, of protons in these two processes [Lipson et al., 2012; Reddy and DeLaune, 2008]. One possible mechanism of Fe oxidation at depth is the transport of oxygen to deeper layers by plant roots [Colmer, 2003], though DO measured at 10 cm depth did not generally reflect this. Given the high degree of Fe solubility in these soils, Fe might also circulate throughout the profile, becoming oxidized at the surface and diffusing to lower layers. Fe can also be oxidized under anaerobic conditions by oxidants such as nitrate, perchlorate, or light [Weber et al., 2006]. An intriguing possibility raised by Nielsen et al.  is the direct flow of e− through conductive soil components, rapidly coupling anaerobic processes at depth to aerobic processes at the surface. The bulk conductivity of soils results from ions in pore water and the electric double layer of colloid surfaces [Cosenza et al., 2008]. Peat soils have been shown to have a more active electrical double layer than inorganic sediments [Comas and Slater, 2004], and the relatively high CEC we report here, consistent with other reported values for soils of the North Slope [Borden et al., 2010; Ping et al., 1999], implies a high surface conductivity component of these soils, despite the relatively low conductivity of the pore water. In addition, the soils studied here have many ingredients that could enhance such conductive networks: Fe-reducing microbes such as Geobacter spp. that produce conductive biofilms [Malvankar et al., 2011; Regberg et al., 2011], nanophase Fe oxides that can form conductive networks with Fe-reducing microbial cells [Kato et al., 2010], and extensive humic substances that can form conductive networks with Fe oxides [Roden et al., 2010].
 While all the previously mentioned mechanisms may play a role, there is evidence that HA serve as oxidants for Fe(II) oxidation in these soils. A common method of inferring the redox state of HA is to incubate extracts [Peretyazhko and Sposito, 2006] or soil water [Heitmann et al., 2007] with a solution of Fe(III) such as ferric citrate. The soils studied here are unique in having high concentrations of dissolved Fe(III), and so HA extracted from soils have essentially already been incubated with chelated Fe solution. This precludes the standard measurement of the reducing capacity of soil HA, but presents the opportunity for an in situ measurement of the redox state of the HA: the ratio of Fe(II)/total Fe in the soil solution. The titration experiments show that HA from these soils can exchange e− with chelates of Fe(II) and Fe(III), and so it is reasonable to interpret the redox state of the soil solution Fe pool as an indicator of HA redox state. The seasonal patterns of this variable provide further support for this interpretation: soil solution Fe became more reduced as both solid phase Fe and bulk Eh measurements became more oxidized, showing the soil solution Fe was not merely responding to changes in bulk redox conditions. Independent of the interpretation of the dissolved Fe pool, a thermodynamic analysis (Figure 9) shows that oxidized HA would be reduced early in the season when Eh is relatively low, and that this process would slow as the season progresses, Eh rises, and reduced HA accumulate.
 The relationship between dissolved Fe and soil HA suggests a model in which dissolved Fe(III), solubilized by siderophores or other Fe-binding ligands, is reduced biologically and then reoxidized to Fe(III) by oxidized HA. This model is supported by a previous study in which chelated Fe(III) was injected into the soil [Lipson et al., 2010]: after injection, respiration was stimulated, Fe(II) increased in the soil solution initially, but then the soluble Fe pool gradually became reoxidized. At the end of the experiment, plots injected with Fe(III) experienced a drop in ORP of 67 mV, suggesting that the addition of soluble Fe(III) facilitated the transfer of e− from soil OM (or other energy-yielding substrates) to some redox-active soil pool, presumably HA. Hence, we propose that in the current study, dissolved Fe cycled rapidly during the first half of the growing season until the supply of oxidized HA was exhausted. Beyond this point, net reduction of solid phase Fe minerals occurred. This model is consistent with available thermodynamic data: upon flooding of soils at snowmelt and exhaustion of available oxygen, HA would be reduced first (E°ʹ = 245 ± 16 mV, this study), along with chelated Fe (typical E°ʹ = 96–385 mV) [Straub et al., 2005]. Reduction of solid phase Fe oxides would then occur (E°ʹ < 100 mV) [Straub et al., 2005]. All our available data fit this model. However, we cannot distinguish whether HA are used directly as e− acceptors, as is widely reported [Cervantes et al., 2002; Coates et al., 2002; Lovley et al., 1996], or are reduced indirectly by chelated Fe, as observed in our laboratory incubations, and essentially as proposed by Struyk and Sposito . It is clear in either case that HA and Fe interact intimately. The Eh-pH relationship we observed for HA implies that organic moieties such as quinones or phenolics are not the only e− accepting groups, as these would consume two protons per two e− accepted, theoretically producing a slope of 0.059 V/pH unit. The lower slopes we and others have found could be explained by tightly complexed Fe(III) within HA acting as a pH-independent e− acceptor [Osterberg and Shirshova, 1997; Struyk and Sposito, 2001].
5 Summary and Conclusions
 Total Fe and the substrates and products of Fe(III) reduction were most plentiful in the active layer of the youngest basin, which had the thinnest organic horizon. Fe content was related to organic layer thickness, as we had hypothesized. However, development of the organic layer does not necessarily proceed linearly and uniformly across all DTLBs as soils age, and so the medium basin we studied had a thicker organic horizon and lower Fe levels than the old basin. Furthermore, Fe minerals are mixed into upper layers as soils age, through cryoturbation and the solubilization and re-deposition of Fe minerals during redox cycles, and so even soils of the ancient DTLB, with deeply buried mineral layers, were not as depleted in Fe as we expected. As a result, Fe(III) reduction was an important respiratory process along the full age range of DTLB. The estimated contribution of Fe(III) reduction to ecosystem respiration was higher in the two DTLBs with the thinnest organic layers. This supports the idea that if active layers deepen as the Arctic warms, increased Fe will be available to support anaerobic respiration.
 Because of increased microtopography as the DTLB develops, the old and ancient sites had higher average Eh and DO across the landscape than the young and medium sites. More reducing conditions led to increased Fe solubility, which was aided by the production of siderophores. As a result of these relationships, Fe concentrations in the soil solution decreased with basin age.
 The redox state of the soluble Fe pool in these soils appears to reflect that of HA in the soil, as this variable was decoupled from field measurements of water table, Eh, and DO, and lab experiments showed that HA can undergo oxidation-reduction reactions with both chelated Fe (II) and Fe(III). These data and a thermodynamic analysis of the system indicate that HA were reduced during the first half of the growing season, when solid-phase Fe minerals were generally becoming more oxidized. Taken together these results suggest a model in which chelated Fe and HA rapidly cycle until the e− accepting capacity of HA becomes exhausted, at which point net Fe reduction occurs.
 Fe(III) reduction contributed substantially to ER in all four sites in the second half of the growing season, and HA may have played a major role during the first half of the season, either directly as e− acceptors or in the re-oxidation of Fe(II). Together, these anaerobic pathways constitute a significant pathway for C loss from this ecosystem.
 We thank Faustine Bernadac, CPS, Umiaq, and BASC for logistical support. Francis Bozzolo, Craig Emerson, Irene Hale, Marguerite Mauritz, and Kim Miller provided invaluable field assistance. This work was supported by NSF 0808604 to DAL and TKR.