Effect of permafrost thaw on CO₂ and CH₄ exchange in a western Alaska peatland chronosequence

We conducted our study within the Innoko Flats National Wildlife Refuge (NWR; 63.58 °N, 157.72 °W; figure 1), in the extensive lowlands developed on eolian and lacustrine silts south of the Innoko River, an area of degrading peatlands with a continental subarctic climate (USFWS 1987, Woodward and Beever 2011). Innoko Flats is the fifth largest NWR in the United States with an area of 15 582 km² and no road access. Over half the refuge is wetlands and much of the area is underlain by permafrost (USDOI 1987, Woodward and Beever 2011). Innoko Flats has a mean daily temperature of −17 °C in January and +14 °C in July (1993–2007; Woodward and Beever 2011), and mean annual precipitation in this region is 445 mm yr⁻¹ (1971–2000; Riordan et al 2006). The snow depth in April 2011 was 0.60 ± 0.09 m across 18 locations, consistent with previous observations for the refuge (Woodward and Beever 2011). The initiation age of the terrestrial peat deposit at Innoko Flats is estimated to be ∼2000–3000 ¹⁴C yr BP (Kanevskiy et al 2014).

Zoom In Zoom Out Reset image size

To capture thaw dynamics of lowland permafrost over millennial timescales, we evaluated trace gas fluxes, C stocks, temperature, inundation, and topographic effects across a thaw chronosequence of thermokarst bogs and an adjacent fen system (figure 1). A thaw chronosequence is a series of locations in which factors such as climate, potential vegetation, parent material, and landscape context are similar, but time since thaw differs. Four landscape units (site types) differing in thaw stage were identified based on landscape position and vegetation: forested plateaus where previous thaw has yet to cause a community shift, and ombrotrophic bogs at three stages of thaw, hypothesized to be young, intermediate, and old. An adjacent minerotrophic fen system was also evaluated, and is distinct from the bog chronosequence due to active water flow. Three field replicates of each landscape unit (site type) were identified based on vegetation and feature size, for a total of 15 measurement locations. All three old bog sites were within ∼10 m of collapse boundaries, where boundary elevations indicated 1–3 m of lowering of the peat surface, consistent with thawing of high ice content, and where forest-bog peat transitions were evident in cores (Jorgenson et al 2013).

In September 2009, core samples were collected at all sites for stratigraphy, carbon inventory, ice content and isotopic analysis (Kanevskiy et al 2014). Unfrozen organic soil was sampled using a variety of soil knives, scissors and corers, in order to obtain a sample of known volume. A SIPRE (Snow, Ice and Permafrost Research Establishment) corer was used to sample the frozen peat and underlying lacustrine and eolian silt. Peat core sections were analyzed for C concentration by combustion (Carlo Erba NA 1500) following oven drying (105 °C) and milling. In September 2010, boardwalks were constructed and placed at each site to minimize disturbance to the sample locations, and 37.5 cm diameter PVC collars were placed into the ground at the end of each boardwalk to a depth of ∼15 cm. The collars were placed in the plateau and bog locations in September 2010, eight months before initial flux measurements were taken. Collars were placed in the fen locations in May 2011, a few days before initial flux measurements at those locations.

Soil temperature at 5 and 25 cm depth was measured using thermistors and Hobo Pro V2 loggers (Onset Computer, Bourne, MA), in all feature types, from September 2009 to September 2011 but with gaps due to bear damage of loggers. Rainfall was measured from April to October 2011 using a tipping bucket and Hobo Pendant data logger (Onset Computer). Water table position was measured from April to October 2011 in both the fen and in two replicates of each bog type using absolute pressure transducers in Hobo U20 loggers (Onset Computer), installed in PVC wells that were anchored in the underlying permafrost. All continuous measurements were logged at one-hour intervals. Incident and outgoing shortwave and longwave radiation were measured near the PVC collars using a NR01 four-component net radiometer (Hukseflux Delft, The Netherlands) over eight days in September 2011.

Surveys were conducted to establish absolute elevations for water tables, well casings, and bog vegetation surfaces relative to a common benchmark near a frozen plateau site. Based on standard survey techniques and our repeated measurements, we estimate that the uncertainty of these elevations is 5–11 cm (table 1). Logged water table heights were corrected for atmospheric pressure (logged separately) and compared to manual measurements of water table height made during field campaigns. Logged water table records are incomplete after 28 July, 2011 due to bear damage of the atmospheric pressure logger.

Peat coring revealed transitions between forested plateau peat and bog peat based on macrofossil indicators, generally Chamaedaphne or Sphagnum riparium for bog peat. Forested peat was recognized by char, needles, and bark; bog peat was recognized by brown mosses and sedge peat. Permafrost thaw was inferred from proximity to surface collapse boundaries and the presence of this forest-bog peat transition in peat cores. Two radiogenic isotopes, ²¹⁰Pb and ¹⁴C, were used to estimate age of transitions. ²¹⁰Pb was measured on bulk peat of shallow strata where unsupported ²¹⁰Pb was detectable (∼60 cm); we used the constant rate of supply (CRS) method to calculate ages of basal horizon depths (Appleby and Oldfield 1978). We report ²¹⁰Pb ages only on profiles in which inventories of ²¹⁰Pb were confirmed by the absence of unsupported ²¹⁰Pb at the base of the profile; uncertainties in²¹⁰Pb ages from the CRS method are based on error analysis described in VanMetre and Fuller (2009). ¹⁴C was measured in macrofossils where identified or on bulk peat picked free of roots. We compared ²¹⁰Pb and ¹⁴C age results from a pair of sites at Koyukuk Flats, Alaska and found close agreement of age methods (see tables S2, S3); we therefore used CRS ages in younger materials (young and intermediate features) and ¹⁴C ages in older materials (old bogs and fen) at Innoko depending on available datable materials.

Net ecosystem exchange (NEE), ecosystem respiration (ER), and CH₄ exchange were measured during three ten-day field campaigns during the early (end of May to early June), middle (end of July to early August) and late (end of September) periods of the 2011 growing season using static chambers (Carroll and Crill 1997; see supplemental text). During flux measurements air temperatures inside and outside the chamber were manually measured using thermocouples (Fluke Corporation, Everett, WA). Clear chambers were used for NEE measurements and opaque chambers were used for ER and CH₄ measurements, with gas concentrations measured in-situ using an infrared gas analyzer (LI-COR or PP Systems; NEE and ER) or by manual collection for analysis of CH₄ by gas chromatography (GC). Photosynthetically active photon flux density (PPFD) was measured manually using a quantum meter (Apogee Instruments) within the clear chambers during NEE measurements.

Our gas flux sampling strategy sought to capture a range of light and temperature conditions over the course of each field campaign. Fluxes were measured in diurnal rounds for each replicate set of sites, with four to six measurements of both CO₂ and CH₄ flux during a 24 h period. The exception was at the fen locations, where CH₄ was consistently evaluated throughout the season, but CO₂ measurements were curtailed in September due to difficulty with access and site disturbance. NEE is defined as the sum of ER and gross primary production (GPP):

$$\begin{eqnarray}&&{\rm NEE}={\rm GPP}+{\rm ER},\end{eqnarray} \tag{ 2 }$$

where negative values of GPP and NEE (μmol CO₂ m⁻² s⁻¹) indicate CO₂ uptake. During the September trip, CH₄ ebullition fluxes were quantified based on the concentration of accumulated CH₄ in funnel traps (7.6 cm diameter) placed below the water table surface. Traps were backfilled with water then allowed to accumulate gases released by ebullition over a 24 h period, then sampled for assessment of volume and GC analysis.

Three separate trips were made and are distinguished by their month of campaign (May, July, September). Within each of these trips, which represent different seasonality, we evaluated the statistical effect of site type on the measured gas fluxes. To make this assessment, we used R (R Development Core Team 2011) and the lme function (nlme package) to fit a mixed effects model accounting for the repeated measures on three different locations for each site type and also fixed effects accounting for air temperature in the flux chamber and an interaction between site type and month of campaign. We performed three collections of family-wise controlled Tukey HSD tests to compare each site type to each other site type in order to ascertain which types are associated with a significantly different flux, on average, after accounting for temperature, within a given trip, at a significance level of p < 0.05.

This sequence of forested plateaus and collapse scar bogs captures dramatic changes in vegetation, feature size, and feature elevation with thaw progression (table 1, figure 2, table S1) (see also Camill et al 2001). The forested plateaus are characterized by an apparent shallow permafrost table (based on four years of observation) and a black spruce (Picea mariana) canopy with an understory of Rhododendron subarcticum, Rubus chamaemorus Sphagnum fuscum, and Cladonia spp. The three bog types differ in size of feature, relative elevation, thickness of bog peat, vegetation type, and forest/bog transition age. The young bogs are small collapse scars (2.0 ± 0.6 m² in area) within the permafrost plateaus that lack woody vegetation, exhibit thaw-induced lowering of the ground surface by 1.20 ± 0.11 m relative to plateaus, host shallow water tables with acidic pH (4.1–4.6), and are dominated by S. riparium, and S. lindbergii. Intermediate bogs form large scars (243.7 ± 82.1 m²) at elevations 2.6 ± 0.1 m below plateaus; contain shallow, acidic water tables (pH 3.9–4.3); and host vegetation dominated by Eriophorum scheuchzeri, S. riparium, and S. lindbergii. Old bogs are also acidic (pH 3.9–4.2), and larger (530.4 ± 182.4 m²), with 2.1 ± 0.6 m collapse, hummocky microtopography and vegetation dominated by Andromeda polifolia, Oxycoccus microcarpus, Carex rotundata, S. balticum, and S. flexuosum. The fens are slightly less acidic (pH 4.4–4.8), and are lower in elevation than plateaus by 2.84 ± 0.05 m; they form pronounced linear features with slow surface water movement, and support vegetation dominated by Carex limosa, C. rotundata, E. scheuchzeri, S. papillosum, and S. majus.

Zoom In Zoom Out Reset image size

Shallow water tables were maintained in all bog features during the 2011 growing season, as rainfall (20.5 cm) was balanced by water level decreases in bogs. In bogs, both surveys and logged water levels indicated decreases of ∼10 cm in water levels during 1 June–25 July. Comparable net gain was observed in survey results for 25 July–25 September, such that water levels in a given bog type on 25 September did not differ from those on 1 June, though the absolute elevation of water levels differed among bog types (figure 2).

Average total C stocks in terrestrial peats (bog plus forest peat) at Innoko Flats ranged from 28 ± 11 to 69 ± 12 kg C m⁻² along the chronosequence and were significantly reduced in young compared to forest and old site types (p = 0.05; figure 2; table 1). In addition, the mass fraction of bog peat increased with time since thaw (table 1).

Peat cores of young bogs revealed bog peat to within 20 cm of the surface underlain by forest peat. The CRS approach using basal depths in cores from two young bog sites indicates that bog peat close to the forest-bog transition in these features was about 15 and 18 yr old (range of 14–21 yr with uncertainties), with underlying forest peats indicating ages of 34 and 46 yr (table 1; table S2). Repeated collapse was indicated in cores of all three intermediate bogs, with bog peat above the most recent collapse transitions showed CRS ages of 45 and 37 yr (range of 22 to 68 yr with uncertainties; table 1; table S2), and an age in underlying forest peat of ∼100 yr (table S2). Bog peat in old cores ranged from ∼200 to 1400 yr in uppermost collapses based on ¹⁴C ages, and with maximum age constrained by macrofossils within underlying forest peat material dating to 1600 yr (table 1; table S2). Our age estimates for young, intermediate, and old bogs reflect most recent collapse timing, with ranges of 14–21 yr, 22–68 yr, and 200–1400 yr, respectively, prior to AD 2009, the time of sampling. Fen locations with active flow (IFIF in table S2) were comparable in age to old bogs (table 1, table S2).

CH₄ emissions at intermediate age sites were significantly higher than those of all other sites in July and were significantly higher than frozen plateau, young bog, and old bog sites in September (table 1; figure 3). The most pronounced seasonal peak in CH₄ emissions occurred at the intermediate age sites (figure 3). Mean CH₄ efflux rates from the intermediate age features were 123.3 ± 70.6 mg CH₄–C m⁻² d⁻¹ in July and 72.9 ± 37.8 mg CH₄–C m⁻² d⁻¹ in September (table 1, figure 3). In the permafrost plateau and young sites, CH₄ emissions were small, with low or negligible rates of uptake generally observed on the plateaus (−0.97 ± 1.24 mg CH₄–C m⁻² d⁻¹ in July), and low or negligible emissions observed at the young and old sites (1.37 ± 0.67 mg CH₄–C m⁻² d⁻¹ and 6.55 ± 2.23 mg CH₄–C m⁻² d⁻¹, respectively, in July). Fluxes of CH₄ at the fens in July were moderate and did not differ significantly from any other site type, producing 38 ± 11 mg CH₄–C m⁻² d⁻¹ (table 1, figure 3).

Zoom In Zoom Out Reset image size

In the NEE and CH₄ efflux data, we observed an interaction between site and date (trip) (figure 3), with a significant effect of air temperature on NEE and ER but not CH₄. In contrast with air temperature, logged 25 cm soil temperatures were warmer at the intermediate and old bogs compared to the young bogs and frozen plateaus (figure 4). At both intermediate and old bogs, temperatures remained above 0 °C at 25 cm depth for about two weeks longer at the end of the season, and soils at 25 cm depth were 2.5 °C warmer on average than frozen plateaus and young bogs (figure 4). In young bogs and frozen plateaus, cooler soil temperatures were associated with lower mean CH₄ emissions (table 1, figures 3 and 4). In intermediate bogs and fens, warmer 25 cm soil temperatures and shallow water tables were associated with enhanced CH₄ emissions. Frozen plateaus and old bogs were dryer at the surface, and had lower CH₄ emissions or uptake.

Zoom In Zoom Out Reset image size

Ebullition fluxes provide additional evidence of drivers. During September, one-time measures of shallow subsurface (20 cm depth) CH₄ ebullition fluxes were comparable to average surface diffusive fluxes at the intermediate bog and fen sites, but greater than surface diffusive fluxes at the old bogs (table 1). In addition, ebullition fluxes increased with depth in September 2011, particularly in intermediate and old sites, consistent with observations of dissolved CH₄ concentrations in porewaters that increased with depth in September 2009 (figure 5).

Zoom In Zoom Out Reset image size

Seasonal and chronosequence patterns were less evident in NEE, with little difference among site types (table 1, figure 3). NEE did not differ among site types except in July, when old bogs had reduced rates compared to intermediate bogs (table 1, figure 3). While variable, the NEE data follow the predicted response to light with scatter that may reflect variation in temperature (figure S1). The ER-temperature response and NEE-light and temperature response (equations (S3), (S4)) did not differ dramatically among landscape units (figure S1). Thus we observed an increase in the ratio of CH₄ efflux to NEE uptake—a metric for assessing net radiative forcing effects (Frolking et al 2006)—at the intermediate sites as a result of increased CH₄ efflux (table 1).

With increasing time since thaw, our surveyed elevations illustrate the progression of physical collapse and inundation, with a maximum collapse of over two meters in the intermediate bogs (figure 2, table S1). The depth of physical collapse seen in the intermediate bogs is consistent with the ice content in the upper 3–4 m of the frozen plateau sites (>2 m) (Kanevskiy et al 2014) suggesting that thaw drove all or most of the lowering (with limited influence from peat decomposition). Old bogs exhibit about 20 cm higher elevation relative to the intermediate bogs, apparently due to new bog peat accumulation (figure 2).

During the growing season, young and intermediate bogs maintained the shallowest water tables and supported a plant community dominated by S. riparium, and S. lindbergii, with a less consolidated surface peat mat compared to that of the old bogs, which were dominated by S. fuscum. These conditions at young and intermediate bogs suggest enhanced anoxia and atmospheric exchange that promotes CH₄ efflux and/or reduced opportunity for CH₄ consumption. Combined with the temperature data, this suggests that intermediate bogs support the most optimal conditions for CH₄ efflux—warmer subsurface temperatures and shallow, fluctuating water tables, with a floating mat that lacks the physical integrity of the older bog sites and thus is less likely to restrict CH₄ efflux (Kane et al 2013, Dise et al 1993, Heikkinen et al 2002, Coulthard et al 2009).

Comparing landscape units, CH₄ emissions were highest at intermediate bogs (figure 3) and were comparable to or exceeded previous observations in other thermokarst wetlands (Wickland et al 2006, Olefeldt et al 2012), consistent with the combination of both warmer and wetter conditions characteristic of those sites. Depth profiles of dissolved CH₄ (in September 2009) and CH₄ ebullition (in September 2011) support an interpretation of enhanced subsurface production at intermediate and old sites (figure 5). At intermediate sites, both dissolved CH₄ concentrations and CH₄ ebullition fluxes increased with depth, suggesting enhanced production relative to younger sites and enhanced surface emission relative to old bog sites, where subsurface fluxes and dissolved concentrations were also elevated in some cases (figure 5). Enhanced CH₄ efflux at intermediate sites could reflect the less indurated peat mat compared to old bogs, as well as plant transport due to the presence of e.g., E. scheuchzeri (Ström et al 2012, Joabsson 1999, Noyce et al 2014). At old bogs, methanotrophy and physical limitation by the solidified peat mat may limit CH₄ efflux at the surface (Zhuang et al 2007, Turetsky et al 2008, Parmentier et al 2011, Coulthard et al 2009), compared to ebullition flux at greater depth (figure 5).

Moderate CH₄ production was also observed at the fen locations (table 1, figure 2), consistent with previous work on fen systems (Olefeldt et al 2012). The fen system is distinct from the bogs due to groundwater flow and hence potential nutrient supply (e.g., Kokfelt et al 2010). Combining areas of active flow where most observations were made (IFIF sites) with relict fen areas now appearing to function more like bogs (including an older or relict fen area (IFOF) with a ¹⁴C date of ∼1650 yr; table S2), the fen system comprises a substantial proportion of the landscape (e.g., 26% of the area in figure 1). The contribution of the fen system to CH₄ fluxes may reflect the interaction of the regional flow system with peatland development following widespread thermokarst lake development 8–10 ka BP (Kanevskiy et al 2014). Our observations indicate that CH₄ emissions from the fen system merit further investigation (table 1) (Crawford et al 2013).

Peatlands in the boreal region typically store large amounts of C and therefore have acted as a long-term net sink for CO₂ (Smith et al 2005, Heyer et al 2002, Roulet et al 2007, Malmer et al 2005, Dunn et al 2007). At Innoko, we observed decreased total C stocks in young compared to frozen forested plateau and old bog sites, and the proportion of bog peat increased from 9% to 74% with time since thaw in young to old sites (table 1, table S1). Therefore, forest peat likely decomposed during bog peat accumulation, as seem in a similar thaw chronosequence at Koyukuk Flats (O'Donnell et al 2011a). At Koyukuk, O'Donnell et al observed a 57% reduction of forest peat after 300 yr of thaw, in an older deposit with larger initial C stocks (137 ± 37 kg C m⁻²). At Innoko, we observed a comparable decrease (55 ± 24%) between frozen plateau and intermediate+old sites over a time period of decades to centuries. Future analyses of peat in these permafrost lowland settings should investigate the tradeoff of forest peat decomposition for bog peat accumulation.

At Innoko, ratios of CH₄ emission to NEE (uptake) across the growing season ranged from negligible in the frozen plateau to 0.08 at intermediate bogs (table 1), indicating positive radiative forcing potential within a few decades of thaw as a function of enhanced methanogenesis at these warmest, wettest sites. In addition, enhanced CH₄ efflux during shoulder seasons might add to CH₄ flux (with minimal NEE) beyond what we observed during the growing season, enhancing net positive radiative forcing over the year (Mastepanov et al 2008, Jackowicz-Korczyński et al 2010). Although growing-season NEE rates remained similar to those observed elsewhere (Wickland et al 2006, Frolking et al 2011, Street et al 2012), enhanced CH₄ emissions during the growing season with intermediate stage thaw could yield a net positive radiative forcing from these sites for up to half a century or so using the approach of Frolking et al (2006) (figure S3), regionally at a level dependent on thaw area and duration of intermediate stage thaw.

The period of elevated CH₄ emission in intermediate stages of thaw at Innoko begins with a threshold transition from net negative to net positive radiative forcing and continues over timescales of decades to centuries. This threshold occurs at a landscape stage that is typified by reduced albedo, warmer temperatures, and shallower water tables. The timescales observed for this threshold at Innoko are generally consistent with previous experimental work on CH₄ production with peatland thaw in northern Canada (Turetsky et al 2007) and in Alaska (Klapstein et al in press); the work presented here suggests this threshold can occur after ∼32 to 68 yr of thaw but is not sustained for much more than ∼100 to 200 yr. Thus this thaw chronosequence approach implies a period of enhanced CH₄ emissions and subsequent reduced emissions with thaw progression over decades to centuries at Innoko Flats.

A conceptual model for thaw-gas flux response is suggested by our chronosequence approach at Innoko Flats, consistent with previous studies. Permafrost thaw and shifts in vegetation contribute to the altered hydrological and thermal conditions that predispose intermediate thaw features to enhanced CH₄ emissions. In black spruce forest, permafrost may be sustained by shading from the forest canopy and thick organic soils (Shur and Jorgenson 2007), as well as the presence of light colored Cladonia lichens that reflect radiation and limit ground heat flux from the forest floor (Stoy et al 2012). Once young bogs are initiated, inundation supports S. riparium communities at the expense of Cladonia spp., while continued shading by trees in the wet young bogs reduces incident radiation (table 1, S1). Young bogs remain cold as they create small depressions within the surrounding permafrost, but even their limited physical collapse promotes absorption of additional radiation and accumulation of runoff that further promotes thaw and maintains a shallow water table. Expanded collapse to form intermediate bogs eliminates shading through tree-kill (figure S2), decreases albedo (table 1) and continues to promote shallow water tables both from thawing ice rich permafrost and from rainfall accumulation.

The combination of warmer soil temperatures and shallower water tables leads to dramatically enhanced CH₄ production mid-summer and over the course of the growing season in the intermediate age bogs (figure 3). A positive C balance may persist, as indicated by bog peat accumulation over decades to millennia (table 1, figure 2) (O'Donnell et al 2011b) and CO₂ uptake outweighing CH₄ release during the growing season (and arguably over the annual cycle) in sites along this thaw gradient (table 1). Nevertheless, net positive radiative forcing effects may also occur for several decades (figure S3) (Frolking et al 2006). Eventually, the succession of vegetation and accumulation of peat over hundreds of years in old bogs physically elevates the land surface (figure 2). Along with the transformation to dwarf evergreen shrub and S. fuscum, soil and vegetation physically insulate the subsurface, which may in turn limit lateral thaw propagation, increase CH₄ consumption, and enhance absorption of solar radiation as indicated by reduced albedo (table 1) (Bonan et al 1992). Under more constant climate conditions, this vegetation succession could lead to re-aggradation of permafrost (Briggs et al 2014) and re-establishment of black spruce communities, but this outcome may be increasingly unlikely with a warming climate (Johnson et al 2013).

The role of fens in connection with subsurface waters is of key importance to the trajectory of permafrost thaw. Fens exhibit high water tables, reduced albedo, and tend toward higher CH₄ fluxes when compared to frozen plateaus (table 1). In some situations, fen locations can facilitate more rapid permafrost loss through thermal effects of flowing water (Jorgenson et al 2001). However, the spatial distribution of fens in permafrost landscapes is particularly variable and may not be easily generalized.