Lability of dissolved organic carbon from boreal peatlands: interactions between permafrost thaw, wildfire, and season

This study was conducted at a peatland complex consisting of peat plateaus and thermokarst bogs in the boreal discontinuous permafrost zone of the Northwest Territories, Canada (61.19N, 120.08W) (Fig. 1). The climate of the region is characterized by short summers and long winters with average annual temperatures of −3.2 °C and total annual precipitation of 369 mm, of which 46% falls as snow (Quinton et al. 2009). The peatland was partially affected by wildfire in 2013, which burned some peat plateaus but did not burn the bogs due to their non-forested, wetter conditions. We established two transects at the site in 2016; one in a section affected by wildfire and one outside the fire boundary (Fig. 1). Each transect (<100 m in length) starts on a peat plateau, and traverses a thawing edge of the peat plateau into a thermokarst bog. The unburned transect included a peat plateau site (“unburned plateau”), a recently thawed site at the edge of the thermokarst bog (“unburned edge”), and a thermokarst bog site away from the thawing edge (“mature bog”). The burned transect consisted of a burned peat plateau site (“burned plateau”), a recent fire-induced thaw site at the peat plateau burned edge (“burned edge”), and a mature thermokarst bog (“mature bog”). The mature bog sites in each transect had similar vegetation, DOC characteristics, and similar results from the incubation experiments, and as such we will henceforth combine these into a single mature bog site. The unburned plateau was elevated by 1–2 m above the surrounding thermokarst bogs and had an open canopy of ∼5 m tall black spruce (Picea mariana) and an understory dominated by labrador tea (Rhododendron groenlandicum) and lichens (Cladonia spp.). All black spruce were dead at the burned plateau, and the ground was extensively charred with sparse labrador tea. The active layer (i.e., the seasonally thawed peat layer above the permafrost) was ∼60 cm at the unburned plateaus and ∼100 cm in the burned plateau. The water table at both plateau sites was generally found at the base of the seasonally thawing front, although some internal depressions have a water table ∼10–20 cm below the peat surface. The unburned edge was dominated by Sphagnum riparium and rannoch-rush (Scheuchzeria palustris), with a water table between 5 cm above the peat and 10 cm below the peat surface. Permafrost thaw at the unburned edge is likely to have occurred in the last few decades, as indicated by chronologies of cores taken at a nearby area (Pelletier et al. 2017). The burned edge site is likely to have undergone permafrost thaw following the wildfire, i.e., within 3 yr prior to the study, since the ground cover was identical to the burned plateau but the peat surface had collapsed, and the water table was at or above the peat surface. At both mature bogs, vegetation was similar and dominated by Sphagnum fuscum and leather-leaf shrubs (Chamaedaphne calyculata), indicating that permafrost has been absent for >100 yr (Camill 1999; Pelletier et al. 2017; Estop-Aragonés et al. 2018a). There was no evidence of wildfire affecting either of the mature bog sites. Peat depth at the mature bogs was between 250 and 300 cm, which we assume is representative of the overall site.

Porewater was collected at each site in spring (21 May), summer (18 July), and fall (12 Sept.) to determine seasonal changes in porewater chemistry, DOC composition, and lability. The spring sampling occurred during the receding limb of snowmelt freshet, suggesting that surface water in the peatland complex was hydrologically connected to the catchment stream network at this time, considering the shallow water table at the sampling sites (Fig. 2). Heavy storms during summer and fall can contribute significantly to the annual runoff generation in peatlands (Clark et al. 2007). In 2016, there were no major storms, however, and peatland hydrological connectivity to the catchment stream network after the spring sampling period was likely limited. Electrical conductivity ranged from 38 to 62 μS cm⁻¹ and pH from 3.8 to 4.2 (Table 1) suggest that all sites are largely ombrotrophic with minor groundwater connectivity.

Peat porewater was collected using MacroRhizon samplers with a 0.15 μm pore size (Rhizosphere Research Products, Wageningen, the Netherlands), thus removing most microbial biomass. Samplers were inserted into the soil at or just below the water table position to collect DOC that is likely to be exported during runoff generation because hydrological connectivity decreases with peat depth (Quinton et al. 2008). In the drier peat plateau sites, samplers were inserted close to the base of the seasonally thawed peat layer in spring and summer (10–40 cm depth) and at 40 cm depth in the fall sampling. At each site, porewater was obtained from three locations ∼10 m apart to account for site spatial variability, and pooled into a single sample until sufficient volume (∼2 L) was collected for analysis and incubation preparations. Samples were kept dark overnight in 4 L polyethylene bottles in a coolbox with ice packs (∼4 °C).

An aliquot of 60 mL from the collected porewater was transferred to an amber glass bottle and acidified (0.6 mL 2 mol L⁻¹ HCl) for analysis of DOC and total dissolved nitrogen (TDN) concentrations. DOC and TDN concentrations were determined within 8 d on a TOC-L combustion analyzer (Shimadzu, Kyoto, Japan), and all measurements are based on four injections, where the average standard deviation for injections of the same sample was 0.07 mg C L⁻¹ and 0.008 mg N L⁻¹. The DOC/TDN analysis used an eight-point calibration, with calibration points between 0.5 and 50 mg C L⁻¹ and 0.05 and 5 mg N L⁻¹, respectively; ranges which encompassed all analyzed porewater samples.

Another aliquot of 60 mL was stored in amber glass bottles for analysis of total dissolved phosphorous (TDP), and major anion and cation (Ca²⁺, K⁺, Na⁺, Mg²⁺, Cl⁻, SO₄²⁻, and Fe^2+/3+) concentrations. Concentrations of TDP were determined by standard acid hydrolysis methods for flow-injection analysis, with concentrations determined based on the stannous chloride method. Major ion concentrations were determined by ion chromatography and ICP-OES at the Natural Resources Analytical Laboratory at University of Alberta within 15 d of sampling.

UV–Vis absorbance was determined the day after sample collection in the field by transferring aliquots of 3 mL from the porewater sample into a quartz cuvette and measuring absorbance between 230 and 600 nm using a UV–Vis portable spectrometer with deionized water as the blank (Flame-DA-CUV-UV-VIS, Ocean Optics, Dunedin, FL, USA).

An incubation experiment was initiated the day after porewater collection to determine the interactive effects of site and season on porewater DOC lability as assessed by dark (i.e., microbial DOC mineralization) and light (i.e., photochemical and microbial DOC mineralization) treatments. We used the previously measured absorbance at 254 nm (A₂₅₄) to standardize physicochemical conditions for the incubation experiment by diluting the collected porewater with a 0.001 mol L⁻¹ NaHCO₃ solution to a target A₂₅₄ of 0.35. This dilution was done to standardize pH and DOC concentrations among samples, to reduce the likelihood of anoxia during the incubation time, and to standardize the amount of sunlight absorbed among light-exposed samples. Diluted samples were further inoculated at a 1% volume ratio with a solution consisting of 1:1 mixed stream and peatland porewater filtered using 1.6 μm pore size glass microfibre filters (Grade GF/A, Whatman). The stream water was collected on the day of incubation initiation from the stream which drains the studied peatland, at a downstream culvert where it crosses the road (Notawohka Creek, 61.16N, 119.92W). The coarser filtration of the inoculation sources does not exclude the microbial communities, and as such this inoculation was done to standardize the microbial community among samples, while excluding microbial grazers. Once the dilution was prepared, UV-light transparent 1 L tedlar bags (Keika Ventures, Chapel Hill, NC, USA) were filled with 400 mL of the prepared sample. Five replicates each for both the dark (tedlar bags covered with opaque tape) and the light treatments were prepared for each site, except for the pooled sample of the mature thermokarst bog sites, where eight replicates were prepared for both the dark and light treatments.

Incubations were carried out in a 0.1 ha shallow peatland pond (59°29′N, 117°10′W) for 7 d to mimic the potential in situ conditions for DOC processing during water export from peatlands. Bags were randomly placed in submerged floating mesh baskets attached to a PVC tube secured with cord to limit movement in the pond. We assessed DOC mineralization by quantifying changes in concentration and UV–Vis spectral properties of DOC between days 0 and 7. Water samples were drawn from each bag prior to deployment and after 7 d in spring, summer, and fall campaigns for absorbance, DOC, and TDN concentration analysis. Samples for DOC and TDN concentration analysis were acidified to inhibit further microbial DOC mineralization, and stored in cool, dark amber bottles until being analyzed within 8 d of collection, as described above. Sample for absorbance spectra was filtered with 0.7 μm glass microfiber filters (Grade GF/F, Whatman) and analyzed immediately in the portable UV–Vis instrument. In spring, tedlar bags were submerged 10 cm below the pond water surface, which yielded very low rates of DOC loss attributed to photodegradation (see results below). Thus, to assess the effect of photodegradation, we instead placed the tedlar bags at 2 cm depth during the summer and fall incubations to enhance incoming radiation. Incoming photosynthetically active radiation (PAR), which we assume to be proportional to incoming UV light, was monitored in a floating station above the water surface in the pond (Photosynthetic Light Smart Sensor, Onset, MA, USA). Due to longer days in summer, incoming PAR was doubled during summer relative to the fall incubation, averaging 422 and 214 μmol m⁻² s⁻¹, respectively. Loggers attached to the floating baskets were used to monitor pond water temperature (HOBO pendant temperature logger, Onset, MA, USA) during the incubation time, which averaged 14, 21, and 11 °C for the spring, summer, and fall incubations, respectively.

We attribute DOC loss during the incubation to microbial activity (biodegradation) in our dark treatment and to a combined effect of both photodegradation and biodegradation in our light treatment. Although we did not exclusively assess photochemical mineralization (light treatment includes biodegradation), we consider our light treatment to be more representative of natural DOC processing in aquatic networks as both processes co-occur in situ. We express biodegradation as the percent DOC loss during the incubation relative to the initial concentration (BDOC). We express photodegradation as absolute DOC loss in mg C L⁻¹ and calculate it by subtracting the site averages of DOC loss in the dark treatment from site average of DOC loss in the light treatment (light–dark DOC losses). Specific UV absorbance at 254 nm (SUVA) is a measure of DOC aromaticity and was calculated by dividing the decadic absorption coefficient at 254 nm by the DOC concentration of each sample (Weishaar et al. 2003). Iron(III) concentrations may interfere with the absorbance in this spectral range, but this effect was ruled out due to the low total iron concentrations relative to DOC concentrations and low pH in our samples associated with reduced iron forms (Poulin et al. 2014). The porewater C/N ratio, a potential indicator for the degree of decomposition of DOC, was calculated using the molar ratio of DOC and TDN.

Statistical analyses were carried out using R Studio (version 1.0.44). Data were checked for normality using the Shapiro–Wilk’s test and for homogeneity of variances using a Levene’s test. A two-way analysis of variance (ANOVA) was used to test the effects of site and season, and their interaction, on BDOC and on total DOC loss. Post hoc analyses were conducted using Tukey’s honest significant difference test. Differences in DOC losses from photodegradation between summer and fall were compared using a paired t test. Linear regressions were used to assess relationships between BDOC and several water quality measures (C/N ratio, SUVA, and ion concentrations).

Seasonal differences in porewater composition between sites are shown in Fig. 3. DOC concentrations, SUVA values, and C/N ratios increased from spring to fall in most sites, with the greatest change generally occurring between spring and summer. The unburned edge had the highest DOC concentrations, particularly during spring when it was almost twice as high as any other site. The unburned edge also had the most distinct DOC composition across all sites and seasons during the spring sampling, with the lowest observed SUVA and C/N ratios, and the highest TDP concentration. When comparing the unburned and burned plateau sites, we found that the burned site had higher SUVA and C/N in spring and summer, and higher TDP during all three sampling occasions. The burned edge was also found to have high concentrations of TDP, particularly in spring. There was a positive linear relationship between C/N and SUVA during spring (R² = 0.72, p < 0.05), but this relationship was not observed in summer or fall when the range of SUVA values were much narrower across sites.

Ion concentrations (Ca²⁺, Mg²⁺, Na⁺, Fe^2/3+, K⁺, Cl⁻, SO₄²⁻, and Al³⁺) generally increased from spring to fall, with the exception of the unburned edge where the highest concentrations were found in spring. In general, the unburned edge also generally had the highest ion concentrations (Fig. 4). When comparing the unburned and burned plateau, we found that the burned plateau had elevated concentrations of Fe^2/3+, K⁺, SO₄²⁻, and Al³⁺.

Concentrations of DOC at the start of the incubations, after dilution to a standardized absorbance (0.35 cm⁻¹) and addition of the inoculate, varied between 9.2 and 18.5 mg C L⁻¹ in spring, 7.0 and 12.2 mg C L⁻¹ in the summer, and 7.6 and 13.3 mg C L⁻¹ in fall. All incubations had detectable losses of DOC, ranging between 0.05 and 4.0 mg C L⁻¹. As such, it is unlikely that rates of microbial activity were restricted by low oxygen levels (∼9.0 mg O L⁻¹ at the start of the incubation).

Biodegradability of the DOC was assessed for the dark incubation from the change in DOC concentration over the 7 d incubation relative to initial DOC concentration (expressed as a %). There were statistically significant differences in BDOC between sites and seasons, with the main and interactive effects of these factors interpreted through post hoc analyses (Fig. 5, Table 2). The BDOC was greatest in spring (12.5%), followed by fall (4.4%), and least during summer (2.5%). During summer and fall, there were no differences in BDOC between sites, whereas during spring, there were significant differences between sites. The greatest BDOC during spring occurred at the young bog (25.6%), followed by the burned edge (14.0%) and the unburned plateau (9.5%), which all were greater than the burned plateau (4.8%) and the mature bog (2.8%) (Fig. 5).

In spring, BDOC was significantly negatively related to C/N (R² = 0.80, p < 0.05) and SUVA (R² = 0.76, p < 0.05) (Fig. 6). In spring, BDOC was also significantly positively related to TDN (R² = 0.95, p < 0.01) and TDP (R² = 0.81, p < 0.05) (not shown). None of these relationships were apparent during summer or fall, when BDOC overall was much lower and had no differences between sites. BDOC was not significantly related to any ion concentration during any season (all p > 0.1, data not shown).

No results from the light incubation in spring are shown, as DOC losses were not different from DOC losses during the dark incubations — resulting from submerging the samples too deep in the pond, thus restricting the amount of light which was absorbed by the DOC. During the summer and fall incubations, we found consistently greater DOC loss in the light incubation than the dark incubations (Fig. 7). Mean DOC loss in the light treatments ranged between 0.59 and 0.91 mg C L⁻¹ across all sites and both seasons. There were, however, no differences between sites (p = 0.88, F = 0.45, df = 4) or seasons (p = 0.77, F  = 0.045, df = 1), and no interactive effect (p = 0.45, F  = 0.94, df = 4) with regards to DOC losses during the light incubations (two-way ANOVA).

The additional DOC losses enabled by the light treatment compared with the dark incubation, either from direct photomineralization or photo-enhanced biodegradation, were calculated from the difference between mean site DOC loss during the light and dark incubation. We found that this DOC loss due to photodegradation was greater in summer than in fall, with the range across sites being 0.35–0.78 mg C L⁻¹ in summer, whereas being 0.21–0.42 mg C L⁻¹ in fall (Fig. 7). This seasonal difference in photodegradation was not statistically significant (paired t test, p = 0.131, n = 5; comparing mean site DOC loss in summer and fall); however, it is likely related to the amount of light that was absorbed by the DOC samples during the 7 d incubations. Incoming average PAR during the summer incubation was nearly double that during the fall incubation (422 vs. 214 μmol m⁻² s⁻¹). As such, the photodegradation losses were proportional to incoming light. We found no relationships across sites or season between biodegradation DOC losses and DOC composition (C/N or SUVA) or with ion concentrations.

We found large differences in porewater characteristics between different peatland sites, both with regards to DOC characteristics and general water chemistry. Our study design required us to pool sampled porewater from several locations within each site to prepare for the incubation experiment, and thus it did not yield replicate analyses of porewater chemistry for each site/season, which precludes statistical assessment of differences between sites. However, many of the key differences between sites with regards to DOC characteristics were in agreement with previous findings at other sites. This includes the lower aromaticity and higher lability of DOC during spring freshet, along with the increased aromaticity and decreased lability of DOC following wildfire. The recently thawed unburned edge site emerged as an important hotspot for biogeochemical cycling during the spring period, which might be linked to both its hydrological position within the peatland complex (where they often receive and convey runoff from peat plateaus to adjacent fens), the release of DOC from previously frozen peat layers, and the shifts in vegetation composition.

The lower aromaticity (SUVA) observed in spring at most sites could be explained by a relative dominance of labile DOC after the winter due to the effects of low but prolonged microbial activity at low temperatures. Freezing of the soils during winter may result in damage of belowground plant tissues and microbial biomass, leading to a release of simple sugars and amino acids, a feature described for riparian soils and related to decreased SUVA values after frost (Haei et al. 2012). A build-up of microbial biomass has also been shown to occur in wet meadow soils during winter (Edwards and Jefferies 2013). This labile DOC pool may thus dominate at the beginning of the growing season, during and following freshet, and result in reduced porewater DOC aromaticity. Additionally, as the water table is closest to the surface in the spring, shallow flow paths bypassing deeper, more aromatic DOC in subsurface soils likely rendered lower SUVA values (Pinsonneault et al. 2016; Broder et al. 2017). The C/N ratios of the DOC were also lowest in spring for all sites and generally increased during the growing season. Microbial biomass has lower C/N than plants and peat (Kuhry and Vitt 1996; Preston et al. 2012), and a greater proportion of this component accumulating in soil during winter could explain the consistently lower spring C/N values.

Wildfire on permafrost peat plateaus was found to lead to higher DOC aromaticity, increased C/N ratio, and reduced DOC lability in porewater. This effect was apparent both in the comparison between the burned and unburned peat plateaus, as well as the comparisons between the burned and unburned recently thawed and collapsed edges of the thermokarst bogs. Similar findings with higher SUVA of DOC following fire have previously been found in Alaskan streams (Larouche et al. 2015) and boreal peatlands (Olefeldt et al. 2013a, 2013b). This increase in DOC aromaticity is likely related both to the combustion of fresh plant material at the peat surface, which otherwise is a source of aliphatic compounds when leached (Pinsonneault et al. 2016), and to the formation of char, produced due to the impartial combustion of organic matter, which leaches hydrophobic, and highly aromatic DOC (Preston and Schmidt 2006). The effect on downstream water chemistry is, however, not limited to DOC characteristics, as wildfire was also found to increase the concentrations of TDP as well as several ions (Fe^2/3+, Al³⁺, K⁺, and SO₄²⁻). These shifts may be due to both the release of these constituents from peat following fire (Smith et al. 2001), as well as a reduction in plant uptake.

The unburned recently thawed edge had the most distinct porewater chemistry, especially in spring. These thermokarst bog edges are locations where peat plateaus are likely to have thawed in the last few decades (Pelletier et al. 2017). The collapse of peat plateaus lead to inundation of old peat, and colonization of hydrophilic Sphagnum species and sedges. The flooded, anoxic, conditions inhibit rapid degradation of the deeper, older peat (Estop-Aragonés et al. 2018b), whereas rapid Sphagnum and sedge growth at the surface is likely to lead to high rates of DOC production. Leachates of live Sphagnum have been shown to have low aromaticity and high biodegradability in western Canada, south of the permafrost boundary (Olefeldt et al. 2013a). Rapid growth of Sphagnum species has also been shown to yield high rates of DOC and TDN production and support higher concentrations of microbial biomass (Lindholm and Vasander 1990). Highly microbially labile algal DOC may also have contributed to the DOC pool in both the unburned and burned edges, as these were the only partially inundated sites in the study (see Figs. 1c and 1e) (Wyatt et al. 2012). Overall, we consider it likely that the distinct DOC characteristics at the unburned edge were due to high rates of DOC production associated with Sphagnum, sedge species, and algae, rather than the production of DOC from recently thawed deeper peat.

Despite the limited areal extent of recently thawed thermokarst bogs within peatland complexes, they are likely to contribute disproportionally to overall peatland DOC export to downstream aquatic ecosystems. Recent thermokarst bogs have been estimated to cover approximately 10% of peatland complexes in the study region (Gibson et al. 2018). However, they are located at points within the peatlands which receive water from their surrounding peat plateaus, and then convey it further downstream to streams or channel fens. As thermokarst bogs expand, many previously isolated thermokarst bogs also become hydrologically connected to streams (Quinton et al. 2009). As such, accelerated thermokarst bog expansion due to permafrost thaw may increase the peatland export of both nutrients (Abbott et al. 2014) and DOC, particularly the delivery of highly labile DOC during freshet.

Biodegradability correlated with the C/N and SUVA of the DOC pool but also with overall TDN and TDP concentrations. The DOC biodegradability was greatest in spring, when the largest differences in DOC properties occurred between sites. Similar findings in DOC derived from ombrotrophic peat vegetation have reported that C/N, SUVA, and TDN are strong predictors for BDOC with fresh organic materials relative to that in peat or litter having a greater biodegradability (Pinsonneault et al. 2016). This is in agreement to our observation of highest BDOC in the unburned edge site. Although there was no relationship observed between BDOC and C/N in the summer or fall, BDOC was always <10% when C/N exceeded ∼45, suggesting this may be a threshold value, beyond which DOC has very limited biodegradability. Increasing C/N typically correlates to higher aromatic C content (Fellman et al. 2008), and thus SUVA is also a strong predictor of BDOC as previously found (Fellman et al. 2008; Abbott et al. 2014; Larouche et al. 2015; Mann et al. 2015; Pinsonneault et al. 2016). Higher values for either of these indices represent microbially recalcitrant DOC, and subsequently reduced BDOC is expected. We observed little BDOC (∼5% or lower) when SUVA values exceeded 3.2 L mg C⁻¹ m⁻¹. However, short incubations (11 d) conducted on soil leachates from burned boreal peatlands observed higher biodegradability (10%–20%) even when SUVA exceeded 3 L mg C⁻¹ m⁻¹ (Olefeldt et al. 2013b), whereas longer incubations (40 d) on stream outflows from thawed permafrost in uplands observed no real impact of high SUVA on BDOC (Abbott et al. 2014; Larouche et al. 2015). Hence, the relationships between BDOC and simple DOC indices such as SUVA are not straightforward and universal, and are likely influenced by sources of the DOC and to what degree it has already been microbially and photochemically processed.

We observed higher DOC loss in light treatments relative to the dark treatments indicating photochemical degradation may be an important pathway for DOC transformation and degradation, in agreement to a growing body of evidence in northern aquatic ecosystems (Laurion and Mladenov 2013; Cory et al. 2014; Ward and Cory 2016). Unlike other studies, we did not find a relationship between DOC losses attributed to photochemical processes and the chemical composition of the DOC. However, our study could not include a light incubation during the spring period, and we had limited variability in DOC composition during the summer and fall incubations. We did find indications of greater photochemical DOC degradation during the summer than fall incubation, likely related to the greater amount of absorbed UV light during the longer days of the summer incubations. As such, our findings suggest that longer ice-free periods of boreal lakes may increase the amount of degradation of terrestrial DOC. Our findings also suggest that photodegradation may play an important role for overall degradation of terrestrially derived DOC, particularly outside of the freshet period and in regions with shallow lakes with long residence times.