Loss of permafrost can modify the export and composition of dissolved organic carbon (DOC) from subarctic peatlands by changing the hydrological regime and altering ecosystem structure and function. In Stordalen peatland complex (68.20°N, 19.03°E) recent permafrost thaw has caused a conversion of the palsa parts (an ombrotrophic, permafrost affected peatland type) into both bog and flow-through fen peatland types. Within the Stordalen peatland complex we estimated the DOC mass balance and assessed DOC composition for one palsa catchment, one bog catchment and two fens in order to assess the possible impacts of permafrost thaw on peatland complex DOC export. The fens were found to have higher net DOC export rates at 8.1 and 7.0 g C m−2 yr−1 than either the palsa or bog catchments, at 3.2 and 3.5 g C m−2 yr−1, respectively. The snowmelt period was more important for the annual DOC export from the palsa and bog catchments than for the fens, representing 65–100% of the palsa and bog catchment exports while 35–60% of the net fen exports. DOC exported from the palsa and bog catchments were characterized by high aromaticity, molecular weight, C/N ratios, and contained DOC of primarily terrestrial origin. The fens exhibited a shift in DOC composition between inflows and outflows that suggested that fens act as catchment locations for degradation and transformation of DOC. Permafrost thaw can thus alter the magnitude, timing, and composition of subarctic peatland DOC exports due to interactions among peatland type, permafrost conditions, and hydrological setting.
 Peatlands in the circumpolar permafrost region are estimated to store 280 Pg carbon (C): with half of this store in the discontinuous, sporadic or isolated permafrost zones [Tarnocai et al., 2009]. Over the second half of the twentieth century it has been found that peatlands containing surficial permafrost in both North America [Camill, 2005; Payette et al., 2004] and Eurasia [Åkerman and Johansson, 2008] have been thawing, and this is attributed to higher temperatures and increased snowfall. Permafrost thawing is projected to continue over the next century [Lawrence et al., 2008]. Loss of permafrost from northern peatlands causes a portion of the peatland C storage to become microbially and hydrologically available, but also leads to altered vegetation composition and ecosystem productivity [Camill et al., 2001; Malmer et al., 2005]. Changes due to permafrost thaw likely alter the quantity and chemical composition of waterborne organic C export, i.e., the sum of particulate organic carbon (POC) and dissolved organic carbon (DOC) export, with implications for both peatland and catchment carbon fluxes and balances [Frey and McClelland, 2009; Tranvik et al., 2009].
 DOC in soil solutions and aquatic environments is commonly dominated by high molecular weight compounds, fulvic and humic acids in particular, but also includes low molecular weight compounds such as aliphatic and aromatic organic acids, peptides, amino acids, monosaccharides and disaccharides, amino sugars, phenolics and siderophores [van Hees et al., 2005]. Because it is practically impossible to identify and quantify all the individual compounds that constitute the DOC pool, characterization of DOC composition is often done based on bulk characteristics such as polarity (hydrophilic/hydrophobic), acidity (acids/neutrals), functional group content, stoichiometry (e.g., C/N ratio), aromaticity and origin of DOC (terrestrial/aquatic) [Guggenberger et al., 1994; Kalbitz et al., 2000]. Some of these bulk characteristics (aromaticity, average molecular weight and relative terrestrial/aquatic origin) have been found to correlate well with simple spectrophotometric measurements [McKnight et al., 2001; Peuravuori and Pihlaja, 1997; Weishaar et al., 2003], which, in turn, have been linked to the biodegradability of DOC samples [Berggren et al., 2009b; Marschner and Kalbitz, 2003]. However, the link between DOC composition indices and biodegradability is not straightforward and caution is needed when interpreting DOC composition indices [Spencer et al., 2008; Wickland et al., 2007]. Terrestrial ecosystems vary in the chemical composition of exported DOC, where peatland exported DOC in general has higher aromaticity than that from nonpeatland ecosystems [Ågren et al., 2008]. Variation in DOC composition is also found among peatland types, where, e.g., soil pore DOC extracted from sedge dominated fens have lower aromaticity and higher biodegradability than samples from Sphagnum moss dominated bogs [Chanton et al., 2008; Roehm et al., 2009].
 Permafrost thaw in peatlands can alter peatland waterborne C export in several ways, including a differentiated response for DOC and POC exports. Radiocarbon dating of POC in the Yukon River, where permafrost thaw is ongoing, shows that POC is predominately composed of older C, i.e., likely partly derived from riverside erosion of old peat deposits [Guo and Macdonald, 2006]. Lake sediments have also revealed that sedimentation rates in lakes close to subarctic peatlands increases during periods of permafrost thaw [Kokfelt et al., 2010]. However, POC export to the Arctic Ocean represents only ∼15% of the DOC export [McGuire et al., 2010].
 Peatland DOC export has been proposed to increase after thaw because of increased hydrological access to previously frozen peat deposits, based on measurements of DOC concentrations in Siberian rivers along a latitudinal transect [Frey and McClelland, 2009]. However, permafrost thaw will also enable increased interaction of runoff with mineral soils and this has been associated with recent decreases in discharge-normalized DOC export in the Yukon River [Striegl et al., 2005]. In contrast to radiocarbon dating of POC, DOC samples in rivers draining into the Arctic Ocean are predominately young, i.e., largely derived from recent plant photosynthetic fixation [Benner et al., 2004; Raymond et al., 2007]. Radiocarbon dating of DOC from wetland dominated catchments undergoing permafrost thaw also indicate that the DOC export is predominately younger [Striegl et al., 2007]. These latter studies indicate that the main effect on peatland DOC export due to permafrost thaw is more likely associated with alterations in plant ecology and hydrology rather than exposure of previously frozen peat [Frey and McClelland, 2009; Guo et al., 2007].
 Peatlands in the subarctic region are often found in a complex of several different peatland types that differ in permafrost conditions, hydrological setting and vegetation composition [Masing et al., 2010; Quinton et al., 2009]. Permafrost thaw can alter the relative abundance of each peatland type [Camill, 2005; Malmer et al., 2005]. The studied Stordalen peatland complex, located in northern Sweden, contains three broad peatland types; palsas that have permafrost and bogs and fens that do not have surface permafrost. The bogs and fens also represent possible end states following ongoing permafrost thaw in the palsas. In our research we wish to answer the following question: What are the associations among peatland type, permafrost conditions, and degree of hydrological connectivity to the surrounding catchment, that explain the quantity and composition of waterborne C export? Our objectives are to measure the annual DOC/POC exports and to characterize the DOC composition from each peatland type of the Stordalen peatland complex. We hypothesize that the DOC export from the different peatland types is positively related to the degree of hydrological connectivity of this peatland type to the surrounding catchment. We further hypothesize that palsas export DOC of poorer composition for microbial degradation than DOC exported from the bogs and fens, due to both low productivity that limits the supply of fresh DOC and to the recalcitrant properties generally found in ombrogenic Sphagnum peat. Our work helps in the assessment of the possible implications of continued permafrost thaw on waterborne C export from the entire peatland complex.
2. Site Description
 Stordalen (68.20°N, 19.03°E: 351 m above sea level) is a 15 ha peatland complex located 10 km east of Abisko in northern Sweden. It is situated in the lower elevations of a 15 km2 catchment that drains into Lake Torneträsk. The catchment contains several peatlands similar to Stordalen, but is dominated by open canopy mountain birch forest (Betula pubescence subsp. czerepanovii) and has alpine heath above the tree line.
 The average annual temperature measured at the nearby Abisko Scientific Research station has increased by 2.5°C between 1913 and 2006, leading to annual mean temperatures above 0°C during the last few decades [Callaghan et al., 2010]. Annual total precipitation has also increased, from 306 mm for the period 1913–2009 to 336 mm during the period 1980–2009 (NORDKLIM, data available at http://www.smhi.se/hfa_coord/nordklim). The permanent snow cover period has decreased from 5.8 months to 4.9 months per year over the last 50 years, along with an increase in the maximum birch forest snow depth from 59 to 70 cm [Malmer et al., 2005].
 The Stordalen peatland complex, for the purpose of our study, is classified into three peatland types: palsa, bog and fen (Figure 1). Classification is based on hydrology and permafrost conditions but is clearly identified by vegetation composition, and is compatible with earlier ecological classifications of Stordalen by Sonesson and Kvillner  and Malmer et al. . Over the last 35 years there have been pronounced shifts in relative extent of the present peatland types, with 10% of the palsa (∼1 ha) being converted into bog and fen because of permafrost thaw [Malmer et al., 2005]. The palsa parts of Stordalen are ombrotrophic and raised 0.5–2.0 m above their surroundings because of ice lenses in the underlying silt. The palsa parts of Stordalen, with extensive coverage and a relatively flat interior, have been classified as palsa plateau by Seppälä . The permafrost underlying the palsa and the bog is between 10 and 20 m thick [Åkerman and Johansson, 2008]. The palsa is dominated by dwarf shrubs, e.g., Empetrum hermaphroditum, along with a moderate abundance of sedges, Eriophorum vaginatum, and a surface cover of mosses, Sphagnum fuscum and Dicranum elongatum, or lichens, Cetraria spp. and Cladonia spp. The peat is relatively thin on the palsa, between 0.4–0.7 m. Only the upper ∼0.25 m is ombrogenic Sphagnum peat and carbon dating suggests that the transition from fen to palsa conditions occurred between 800 and 300 years BP [Malmer and Wallén, 1996]. The active layer thickness follows the surface topography: thinner in hummocks (<0.7 m in late summer) and thicker in the wetter depressions (∼1 m) [Rydén and Kostov, 1980]. Drainage follows the frost table, filling first the internal depressions and then draining out of the palsa into bog, fen or lake via narrow (1–2 m) channels, similar to that described for Canadian treed palsa plateaus [Wright et al., 2009]. Longer and warmer thaw seasons has increased the maximum active layer depth from 0.6 to 0.8 m over the last 35 years [Christensen et al., 2004], and has been associated with a relative shift in dominance from bryophytes to lichens [Malmer et al., 2005].The bog parts of Stordalen are also underlain by permafrost, but wetter conditions cause a thicker active layer (>1 m) that reaches into the silt, hence there is no permafrost within the peat and frost heave is minimized, placing the bog surface at lower elevation than that of the palsa. Also the bog water table is generally closer to the surface than on the palsa. Vegetation is dominated by Sphagnum balticum and E. vaginatum. The peat layer is generally thicker than on the palsa, ranging from 0.5 to >1 m [Rydén et al., 1980]. The bog parts in Stordalen receive runoff from surrounding palsa or are entirely ombrotrophic, and in turn drain into the fens that surround the palsa and bog parts (Figure 1).
 The fens have no permafrost and receive a large amount of their water input from a lake to the east of the peatland complex (Figure 1). The catchment upstream of this lake covers 1.7 km2 (a catchment size approximately 40 times greater than the area of the fens), has a maximum elevation of ∼650 m above sea level, is dominated by birch forest and contains no other significant peatlands. The water table in the fens is at or above the peat surface throughout the year, with higher nutrient availability supporting Carex rostrata and Eriophorum augustifolium close to the fen inflows, mixed with increasing abundance of E. vaginatum and Sphagnum riparium closer to the outflows or away from the main water flow paths.
 This study monitored the hydrology and DOC fluxes for one palsa catchment (0.39 ha), one composite bog/palsa catchment (hence forth referred to as the bog catchment) (3.34 ha) and two separate fens, referred to as the north (1.34 ha) and south fen (3.72 ha) (Figure 1). The palsa and bog catchment limits were determined by a survey of the surface topography. The maximum elevation difference within each catchment is ∼1.5 m, but the hydrological gradient is lower in the bog parts. The palsa catchment outflow was gauged by installing a thin metal plated 20°V notch weir at a clearly defined channel outflow (Figure 1, site 1). The studied bog catchment contained ∼45% palsa peatland type in addition to the bog peatland type and was sampled at a flume box located at a narrowing of the surface flow paths close to the transition into the south fen (Figure 1, site 2). It was, however, not possible to directly quantify the mass flow of water at the bog catchment outflow because of it was too diffuse. The specific runoff measured from the palsa catchment was used as an approximation for the bog catchment, justified by the fact that the bog catchment nests the palsa catchment.
 The inflows and outflows of the two fens were sampled for water chemistry (Figure 1, sites 3–7), but it was only possible to directly measure discharge at the outflows. Discharge at the north fen outflow (Figure 1, site 4) was measured in a flume box, while the south fen had two outflow sites monitored using a small 45°V notch weir at the eastern outflow site (Figure 1, site 6) and a flume box at the western outflow site (Figure 1, site 7). The sum of discharge from the two south fen outflow sites is used to represent total south fen outflow discharge, and reported water chemistry for individual sampling occasions is either flow weighted (DOC concentration) or weighted for DOC export (all DOC composition indices) using data from these two outflow sites. We define “fen flow-through” as the discharge at the fen outflow per fen area (mm), a measure that represents fen downstream transport potential for waterborne C.
 Daily discharge at the fen inflows was estimated as the residual after solving the water budget of the fens, taking outflow discharge, precipitation, evapotranspiration and change in storage into account. Groundwater input to the fens was assumed negligible, supported by no differences in electrical conductivity measurements between fen inflows and outflows (measurements done at every sampling occasion by handheld Oakton pH/Con 10). Precipitation was measured at Stordalen using a tipping bucket. Hourly fen evapotranspiration was estimated by the Penman-Monteith equation with parameters (aerodynamic and canopy resistance) appropriate to wet northern fens [Wessel and Rouse, 1994] and using hourly data (net radiation, air temperature, wind speed and relative humidity) from a micrometeorological tower centrally located on the palsa/bog part of Stordalen peatland. Change in fen water storage was calculated from the recorded water level changes in the fens. The south fen water budget also took into account the inflow of water from the palsa/bog area that drains into the south fen (5.71 ha, Figure 1), with the specific runoff from this area assumed to equal the palsa catchment runoff.
 Water levels at the weirs, flume boxes, in the fens, and of the eastern lake were recorded hourly with Odyssey capacitance sensors (resolution <1 mm). The sensors were calibrated using manual stage measurements taken each occasion the sites were sampled for water chemistry. Flume box discharge was calculated by the velocity area method and velocity was measured with an Aqua Data Sensa Electromagnetic Current Meter RC2 (±0.01 m s−1 resolution). Discharge over the weirs was calculated using information from LMNO Engineering, Research and Software, Ltd (http://lmnoeng.com), and verified by manual calibration. Rating curves were constructed using second-order polynomial functions in the Curve Fitting Toolbox of MatLab R2009b for each site: north fen outflow: n = 11, r2 = 0.99; south fen west outflow: n = 30, r2 = 0.95; south fen east outflow: n = 16, r2 = 0.87; palsa outflow: n = 10, r2 = 0.99, all significant at p <0.001. We used the confidence bounds (95%) of the rating curves at the estimated average discharge rate for the week of peak snowmelt in 2008 to give an estimate of the uncertainty of our seasonal (April–October) cumulative discharge measurements that we use in subsequent error analyses. This yielded seasonal discharge uncertainties of 7% for the north fen outflow, 7% for the south fen west outflow, 15% for the south fen east outflow and 10% for the palsa outflow.
 Snow surveys were conducted within the palsa and bog catchments prior to the commencement of melt in early to mid-April 2008 and 2009 to estimate the snow water equivalent (SWE). Snow depth was measured on a grid of 100 by 200 m, n = 100, and a subset (n = 15) of those points were analyzed for snow water equivalent. The thickness of the thawed active layer was measured every 3 to 7 d in nine sites in the palsa catchment and a further twelve sites in the bog catchment between late April and mid-October of 2008 and 2009.
 Sampling frequency for water chemistry was dependent on hydrological conditions: every 1 or 2 d during snowmelt and other high flow events and every 3 to 6 d during lower flows. All sites were sampled in 2008, while in 2007 only the north fen was sampled and in 2009 only the palsa catchment was sampled. Sampling started before snowmelt and ended in late September in 2007 and mid-October in 2008 and 2009. Each site was sampled between 40 and 63 times each year. Water samples were filtered through 0.45μm glass fiber filters, acidified to a pH of 2 with 0.5 M HCl and kept in 25 ml vials in a dark refrigerator at 5°C. Absorbance at 250, 254 and 365 nm was measured on a Shimadzu UV-160A with a 1 cm quartz cuvette on same day as the water samples were collected. A Shimadzu TOC-V CPH analyzer was used for DOC and total nitrogen (TN) analysis. Calibration of the TOC-V CHP analyzer was performed with four standard solutions ranging between 1 and 100 mg C l−1 and 0.05 and 2.5 mg N l−1, using a linear regression (r2 >0.99). Standardized samples (10 mg C l−1, 0.25 mg N l−1) and Milli-Q water samples were included in each sample run to ensure no analytical drift within and between sample runs. Triplicate injections of each sample showed that the analytical standard deviation (SD) of each sample was 0.15 mg C l−1 and 0.05 mg N l−1. Fluorescence intensity at 450 and 500 nm with a 370 nm excitation was measured on a Perkin Elmer LS55. DOC and fluorescence analysis were performed within 3 weeks of sampling. Particulate organic carbon (POC) was analyzed by loss on ignition at 375°C for 16 h, after filtering 1 l of sample water through 0.45 μm glass fiber filter [Ball, 1964]. Each sampling site was analyzed for POC seven times, evenly distributed over the season, in 2008. Water temperature was measured hourly with HOBO temperature loggers installed at all water chemistry sampling sites.
 Daily DOC fluxes were estimated by multiplying average daily discharge with DOC concentrations, using linearly interpolated DOC concentrations for days between sampling occasions. From daily DOC fluxes we estimated monthly and seasonal (April–October) DOC fluxes, along with flow-weighted average concentrations. Because the peatland freezes in late October and there are no observable flows, the April to October fluxes can be assumed to represent the annual waterborne C fluxes of the peatland parts. Because of low sampling frequency and no apparent relationship between POC concentration and hydrology or season at any sampling site, annual POC export was estimated using the arithmetic mean POC concentration and total discharge. Fen net DOC and POC export was estimated as the difference between fen inflows and outflows and taking into account the additional inputs from precipitation and from the palsa/bog area that drains into the south fen, while net palsa and bog catchment exports only included catchment outflows and precipitation inputs.
 Monte Carlo simulations (using 10,000 realizations) were used to estimate the uncertainty (±2 SD reported from each simulation) of annual catchment runoff, area specific fen inflows and outflows and DOC/POC export estimates. The uncertainties for area specific fen outflows and palsa and bog catchment runoff was estimated by combining the uncertainties of the rating curves (see above) and an assumed uncertainty of the measured catchment or fen area of 10% (representing ±2 SD). Fen inflow uncertainties combined the uncertainties of all other fluxes of the fen water budgets (assuming 2 SD to be represented by 25% of the measured precipitation, 50% of the estimated evapotranspiration and a fixed 10 mm for the uncertainty of the change in fen storage over the season). DOC and POC flux estimates combined the uncertainties of area specific hydrological fluxes and their respective annual flow-weighted concentrations. We assume a 10% uncertainty (i.e., ±2 SD) of the flow-weighted seasonal DOC concentrations (similar to that assumed byNilsson et al. ), while uncertainty of the annual average POC concentrations were estimated using ±2 SD of measurements at each sampling site (varying between 50 and 100% of the average POC concentration). Uncertainty of precipitation DOC inputs were estimated analogously (±2 SD of DOC concentration measurements, representing 100% of the average concentration). Uncertainties of palsa and bog catchment net DOC and POC export rates combined uncertainties of catchment exports and precipitation inputs. Net DOC and POC export for the fens used a linked Monte Carlo simulation of both the hydrological fluxes and their respective DOC concentrations.
 Composition of DOC was assessed using four indices. Specific UV absorbance (SUVA254) is defined as the UV absorbance at 254 nm normalized for the DOC concentration (l mg−1 C m−1) and increases linearly with measured DOC aromaticity [Weishaar et al., 2003]. Reported values of SUVA254 in natural waters usually range between 0.5 and 6 l mg−1 C m−1, equivalent to a range of percent aromaticity between 5 and 45%. The UV absorbance ratio between 250 and 365 nm (a250/a365) is inversely related to the DOC weight-averaged molecular weight [Lou and Xie, 2006; Peuravuori and Pihlaja, 1997], with reported values in the range 3 to 8. Fluorescence index (FI) is a measure of relative terrestrial or microbial origin of DOC [McKnight et al., 2001], and is calculated as the ratio between emission intensities at 450 and 500 nm at 370 nm excitation. Reported values of FI are in the range 1.2–1.9, with higher values indicating a microbial source and lower values a terrestrial. The reported C/N ratio was assumed to be equal to the measured DOC/TN ratio, an assumption that was tested by analyzing a subset of samples for dissolved inorganic carbon (DIN). Analysis of DIN on five samples from each sampling site with a Foss FIAstar 5000 detected only minor concentrations of NO3- (all samples <0.01 mg N l−1) and NH4 (all samples <0.005 mg N l−1), supporting our assumption that only dissolved organic nitrogen significantly contributed to the TN. Monthly and seasonal estimates of average values for the composition indices were calculated by weighing for the magnitude of the daily DOC fluxes.
 All years of the study (using a hydrological year of 1 November to 31October) were warmer than the 1913–2009 annual average, with 2007 and 2008 >1 SD warmer (Table 1). The differences between annual temperatures measured at Abisko Research Station and Stordalen peatland complex were small, while summer precipitation at Stordalen peatland complex was greater during all three years (Table 1). Precipitation measured at Abisko Research Station for the 2007 hydrological year was above the 1913–2009 average because the winter (November to May) precipitation that was >1 SD above the average. This led to a large spring runoff in 2007 compared to 2008 and 2009. Precipitation of both the 2008 and 2009 hydrological years was below the 1913–2009 average, with a pronounced dry period between June and October in 2009 that had >50 mm less precipitation than in either 2007 or 2008 for the same period (Table 1).
Table 1. Annual and Seasonal Precipitation and Annual Mean Temperature for the Years of the Study Years and the Longer Term (96 Years) Means Measured at the Abisko Scientific Research Stationa
Nov 2006 to Oct 2007
Nov 2007 to Oct 2008
Nov 2008 to Oct 2009
Years run 1 November to 31 October. Means are given with ±1 SD. Observations are made at the Abisko Scientific Research station 10 km west of the Stordalen peatland complex. Precipitation from November to May represents snow storage and precipitation from June to October is primarily rainfall.
Annual temperature and summer precipitation was also measured directly at Stordalen peatland complex and is reported in parentheses.
4.2. Palsa and Bog Catchments Runoff and DOC Patterns
 Annual runoff (April–October) from the palsa and bog catchments was 123 ± 10 mm (±2 SD) and 101 ± 9 mm in 2008 and 2009, respectively (Table 2). In 2009 there was no runoff recorded after 18 June, while 24% of the annual total runoff in 2008 occurred after mid-June (Figure 2). Runoff during the week of the snowmelt peak was 54 mm in 2008 and 63 mm in 2009, i.e., >50% of the annual total in both years. The runoff during the melt period was greater than the estimated SWE measured prior to snowmelt: SWE was 47 mm in 2008 and 33 mm in 2009.
Table 2. Hydrological and Waterborne C Fluxes for the Studied Peatland Components of Stordalen Peatland Complex Between April and October of Study Yearsa
Specific Water Flux (mm)
Average DOC Concentrations (mg l−1)
Specific DOC Flux (g m−2)
Average POC Concentrations (mg l−1)
Specific POC Flux (g m−2)
All hydrological and waterborne C fluxes are normalized to their catchment (palsa and bog catchments) or areal extent (fens). Uncertainties in italics are based on Monte Carlo simulations (±2 SD), while nonitalics are based on direct measurements (±2 SD) or assumptions of uncertainty (see text for further detail). DOC, dissolved organic carbon; POC, particulate organic carbon.
Fen inflow from lake calculated as the residual of the fen water balances, including fen outflow, precipitation, evapotranspiration, change in storage ,and for the south fen also the inflow from the palsa/bog part that drains into the south fen.
Palsa/bog part outflow to the south fen estimated from the specific runoff from the palsa catchment and then rescaled to the area of the south fen. Palsa/bog part DOC concentration assumed to be equal to the sampled bog catchment.
Bog catchment runoff assumed to be equal to the palsa catchment runoff since they are nested.
 The annual net export of DOC from the palsa catchment was 3.2 ± 0.6 and 2.5 ± 0.5 g C m−2 yr−1 in 2008 and 2009, and 3.5 ± 0.7 g C m−2 yr−1 in 2008 from the bog catchment (Table 2). These estimates include a precipitation input of DOC of 0.5 ± 0.5 g C m−2 yr−1. The palsa and bog catchment outflow DOC concentrations varied by between a factor of 5 and 9 over a full season (Figure 2 and Table 3), with measured DOC concentrations negatively correlated with runoff (p <0.01 and r2 >0.4, see Figure 3). Paired t tests show that the palsa catchment had significantly lower concentrations than the bog catchment during April and May (n = 16, p <0.05) but higher concentrations during September and October (n = 11, p <0.01). Snowmelt period (April–June) export of DOC in 2008 was responsible for 65% of the annual DOC export from the palsa catchment and 74% for the bog catchment. DOC concentrations continued to rise in stagnant water behind the weir/flume box at the palsa and bog outflows after runoff had ceased during the 2008 summer, reaching 95 and 140 mg C l−1 for the palsa and bog, respectively (Figure 2). Concentrations of POC were much lower than DOC over the entire 2008 season, with all samples <2 mg C l−1 (Table 3) resulting in annual POC export <0.1 g C m−2 (Table 2).
Table 3. Means and Ranges of Measured DOC and POC Concentrations and DOC Composition Indices at Sampling Locations Within the Stordalen Peatland Complex
 Seasonal averages of DOC composition indices show that the palsa catchment had higher SUVA254 and lower FI and a250/a365 than the bog catchment (Table 3). Paired t tests indicate a significant difference in SUVA254 for all months in 2008 (p <0.05), with the exception of April (Figure 4), and in most months for FI and a250/a365. A progressive increase in SUVA254 occurred during snowmelt for both the palsa and bog catchments (Figure 4). This transition in DOC composition as indicated by SUVA254 occurred both in 2008 and 2009 once the active layer depth had reached 5 to10 cm (Figure 5). A transition in DOC composition was also evident in a250/a365 and FI data during early snowmelt (not shown), indicating decreased importance of microbially derived, nonaromatic, low molecular weight DOC once the active layer started to develop.
4.3. Fen Hydrology and DOC Patterns
 The south and north fen convey water from a 1.7 km2 upstream catchment to a stream (south fen) or lake (north fen) at their outflow. Of the total catchment runoff arriving at lake to the east of the fen, ∼80% passes through the south fen and ∼20% through the north fen. This led to annual (April–October) flow through, expressed as a depth equivalent (i.e., discharge divided by fen area) of 2790 ± 240 and 2280 ± 200 mm for the north fen in 2007 and 2008, respectively, and 5880 ± 780 mm for the south fen in 2008. With an estimated average level of stored water of ∼400 mm in the fens during spring and ∼200 mm during summer low flow conditions we can calculate a rough estimate of the average water residence time of between 3 and 30 d for the south fen depending on season (average water residence time ∼10 d for the full April–October season). The north fen had similar estimated water residence time during snowmelt but longer during the summer since flow through nearly ceased (average water residence time ∼25 d for the full April–October season).
 Precipitation, inputs of runoff from the bog/palsa portion to the south fen, evapotranspiration, and change in fen storage between April and October were all minor components of the hydrological budget relative to the flow through (Table 2). In 2008, none of these components exceeded 4 and 10% of the south and north fen flow through, respectively. Daily fen inflows, as these could not be measured directly, were estimated as a residual of the water budget. The resulting annual fen inflows were both estimated to be equivalent to ∼97% of their respective outflow discharges. Hence, the fens were hydrologically dominated by surface water flowing through them.
 The net DOC export from the north fen was 5.2 ± 3.7 and 7.0 ± 3.1 g C m−2 yr−1 in 2007 and 2008 and 8.1 ± 7.8 g C m−2 yr−1 from the south fen in 2008 (Table 3). The DOC fluxes at the fen outflows were ∼3 to 6 times greater than the resulting net fen exports taking into account the inflow, precipitation and palsa/bog DOC inputs (Table 3). The seasonal patterns of DOC concentration at the fen outflows were consistent both between the north and south fen and between years, with higher DOC concentrations prior to snowmelt, minimum concentrations during snowmelt and progressively increasing concentrations over the summer (Figure 6). Paired ttest show that north fen outflow DOC concentrations were significantly higher than inflow concentrations in all months of 2008 and 2009, (n size varied from 4 to 16 depending on the month, p <0.01). The flow-weighted average difference in DOC concentration between north fen inflows and outflows was 1.88 and 2.99 mg C l−1 in 2008 and 2009, while it was 2.53 mg C l−1 for the south fen in 2008 (Table 3). DOC concentrations in the south fen outflow were significantly higher than in its inflow during April–May and July–August 2008 (n = 5–13, p <0.05). These four months accounted for 86% of the estimated annual net DOC export from the south fen. There was a significant correlation between the difference in fen inflows and outflow DOC concentrations and flow through for the north fen (2007: r2 = 0.47, p <0.01. 2008: r2 = 0.26, p <0.01) (Figure 3), but not for the south fen. Average POC concentrations in fen inflows and outflows ranged between 0.47 and 0.65 mg C l−1, but the sample size was insufficient to determine if the fens were net sinks or sources of POC (Table 2).
 Fen inflows had lower seasonal average SUVA254 and C/N ratios and higher FI and a250/a365 ratios than their respective outflows (Table 3). This difference in DOC composition between fen inflows and outflows was significant for most months in 2008 (Figure 4). The difference between SUVA254 in inflows and outflows increased with warmer water temperatures and less flow through (Figure 7).
5.1. Peatland DOC Export
 We found that the estimates of annual (April–October) net DOC exports (specific DOC flux in Table 3) from the studied fens in Stordalen were two to three times higher than that from either the palsa or bog catchments (range 5.2 to 8.1 g C m−2 yr−1 versus 2.5 to 3.5). The palsa and bog DOC exports are in the lower range of reported peatland estimates [Billett et al., 2004; Dinsmore et al., 2010; Fraser et al., 2001; Koehler et al., 2009; Nilsson et al., 2008]. Since DOC concentrations at the palsa and bog catchment outflows (seasonal averages between 27 and 31 mg C l−1) were within the range reported for peatlands (25 to 40 mg C l−1 (see references above)), therefore we conclude it was their relatively small annual runoff (101 and 123 mm yr−1) that led to the low DOC exports. The runoff for the larger catchment in which the Stordalen peatland complex is situated for 2008 and 2009 was 180 and 270 mm [Olefeldt et al., 2012], indicating that the palsa and bog catchments yield less specific runoff than other landscape components. Less snow accumulation on the elevated palsa and greater potential water storage during summers due to deepening of the active layer would both act to reduce specific runoff. The assumption we made that the specific runoff from the palsa was applicable to the bog catchments likely overestimates the bog catchment runoff somewhat because the hydrological gradient in the bog is less than the palsa catchment. This would, in turn, lead to an overestimate of the bog catchment DOC export.
 In contrast to the palsa and bog catchments, the Stordalen fen DOC exports appear to be primarily substrate rather than transport limited. We arrive at this conclusion because increased daily fen flow through led to a decrease in the difference between fen inflow and outflow DOC concentrations indicating that an increase in water residence time results in a greater buildup of DOC in fen water. An increase in daily flow through thus leads to a proportionally smaller increase in fen net DOC export. This effect was also evident in 2008 when the south fen had 158% greater water flow through than the north fen but only a 16% greater net DOC export. Further, the north fen had higher net DOC export in 2008 than in 2007 despite having lower annual flow through of water in 2008. Hence, in comparison to the dependence of the palsa and bog DOC export on runoff magnitude, the net DOC export from the fens in Stordalen is less dependent on DOC transport potential, i.e., the annual fen flow through.
 A brief snowmelt period (∼4–8 weeks) is often responsible for more than half of the annual DOC export from northern catchments [Carey, 2003; Finlay et al., 2006; Laudon et al., 2004; Striegl et al., 2007]. In our study the snowmelt period (April–June) dominated the palsa and bog catchments DOC export (range 65–100% of annual export), but was less important for the fens (range 35–60%). While runoff ceased from the palsa and bog during the summers (no runoff at all was recorded from the palsa catchment in 2009 after mid-June), fen flow through and, as a consequence, fen DOC export was sustained throughout the summer driven by runoff from the larger upstream catchment flowing through the fens. A late season storm in October 2008 (41 mm of rain) showed that such events can contribute significantly to the annual runoff and DOC export for the palsa as it was responsible for 24% of the palsa annual runoff and 35% of the annual DOC export. In comparison, this event was less important for the annual DOC export from the fens. The frequency and maximum magnitude of large precipitation events (>20 mm d−1) have increased in the Stordalen area over the last three decades [Callaghan et al., 2010], a trend that thus could have a greater influence on the magnitude and timing of DOC export from the palsa and bog than for the fens.
 Permafrost thaw in peatlands has been hypothesized to increase the export of POC derived from old peat [Guo et al., 2007]. Although there is evidence of increased POC sedimentation during periods of permafrost thaw in the sediment records from downstream lakes in study the area [Kokfelt et al., 2010], we were unable to detect any significant POC fluxes from the Stordalen peatland complex during our study. For the palsa and bog catchments the annual POC export was <0.1 g C m−2 yr−1, but the sample size was insufficient to determine whether the fens were net sources or sinks for POC (confidence interval for the fen net POC fluxes between −2.2 and 3.4 g C m−2 yr−1).
 DOC export has been shown to be equivalent to between 25 and 50% of the net ecosystem productivity in northern peatlands [Dinsmore et al., 2010; Nilsson et al., 2008; Roulet et al., 2007], stressing the importance of waterborne C export for peatland net ecosystem C balance (NECB, sum of all ecosystem C fluxes [Chapin et al., 2006]). Bäckstrand et al.  report palsa plots to be a net source of C to the atmosphere (−30 g C m−2 yr−1), bog plots a net sink (29 g C m−2 yr−1) while fen plots are in near balance due to high total hydrocarbon (mainly methane) emissions (4 g C m−2 yr−1). If we combine our waterborne C fluxes with the atmospheric exchanges of Bäckstrand et al. , then DOC export has a larger relative influence on the NECB of the fens than for the palsa and bog areas. In turn, restricted DOC export and low methane emissions explain how the palsa and bog areas of Stordalen peatland have a NECB in the same range as more productive boreal bogs that have greater DOC export and methane emissions [Olefeldt et al., 2012].
 Observed changes in peatland type coverage in the Stordalen peatland between 1970 and 2000 indicate the palsa cover decreased by 1.05 ha (equivalent to 10% of the 1970 coverage), along with concurrent increases in bog (+0.35 ha, +8%) and fen (+0.66 ha, +15%) coverage [Malmer et al., 2005]. Combining these changes with our DOC export estimates for each peatland type (using data from 2008) suggests that the Stordalen peatland complex total DOC export has increased by ∼5% from 5.0 to 5.3 g C m−2 yr−1. However, this estimate assumes that DOC export from the fens is not affected by changes of the flow through magnitude, which is dependent on fen area. Increased fen dominance in Stordalen peatland is also expected to have shifted the timing of DOC export from snowmelt to the summer/autumn period. Altered timing of peatland DOC export could be of significance for its downstream fate in aquatic environments, where longer residence times in lakes during the summer/autumn period compared to during snowmelt implies a higher likelihood for retention through sedimentation, biodegradation and photodegradation. In addition to altered peatland DOC export magnitude and timing thereof, the fate of exported DOC in downstream environments can also be influenced by changes in DOC composition following peatland permafrost thaw.
5.2. DOC Composition
 In our study we assessed the DOC composition over full hydrological seasons at the palsa and bog catchment outflows, along with fen in and outflow sites using several indices. While the palsa and bog catchments represent headwater contributions to the greater catchment, the fens are flow through peatlands with the mass flux of DOC received by the fens from upstream sources being several times greater than the fen net DOC export. A straight comparison of DOC indices does thus not reflect compositional differences in DOC derived from different sources, e.g., fen and palsa peat derived DOC, but the results can highlight the different roles of minerotrophic and ombrotrophic peatlands at the catchment scale, which is pertinent because of the conversion of palsas into fens following permafrost thaw.
 Exported DOC from the palsa and bog catchments had higher SUVA254, C/N ratios and lower FI and a250/a365 than the birch dominated upstream catchment (represented by measurements at the fen inflow sites in the eastern lake), suggesting that the DOC exported from the palsa and bog is a poor substrate for microbial degradation. While higher SUVA254 and lower FI of DOC exported from peatland catchments than from nonpeatland catchments likely is a general pattern [Ågren et al., 2008], the low SUVA254 and high FI measured in the eastern lake in this study could also be due to the influence of DOC derived from groundwater sources [McKnight et al., 2001]. Incubations of peatland derived DOC with high aromaticity is often linked to low bioavailability [Kalbitz et al., 2003a; Peuravuori and Pihlaja, 1997], suggesting that DOC exported from the palsa and bog catchments would be a poor substrate for downstream microbial degradation [Berggren et al., 2009a; Berggren et al., 2010].
 In Stordalen, extracted DOC soil samples from within and below the active layer of the palsa have been found to have higher SUVA254 (between 4.8 and 6.6 mg C l−1 m−1) [Roehm et al., 2009] than observed in this study at the palsa catchment outflow in this study (individual samples measured between 1.59 and 4.4 mg C l−1 m−1). An additional DOC source other than that derived from the peat matrix is suggested in order to explain the composition at the palsa and bog catchment outflows. Particularly low SUVA254 was observed during early snowmelt before the active layer had started to develop, indicating that DOC derived from decomposing surface litter might provide DOC with relatively low aromaticity. While low SUVA254 values were only observed for a few days during early snowmelt, they were important to the seasonal average SUVA254 because of the high DOC export during this period, stressing the importance of snowmelt not only for the quantity of export but also composition of peatland DOC export.
 We also observed consistently lower SUVA254 in DOC samples from the bog catchment relative to that of the palsa. The difference in SUVA254 between the palsa and bog catchments could be a result of increased importance of DOC derived from recent photosynthetic activity in the more productive bog [Bäckstrand et al., 2010]. Sphagnum moss, which dominates the bog, has been found to contribute recently fixed atmospheric carbon to the DOC pool [Fenner et al., 2004], primarily as nonaromatic DOC that is largely available for rapid microbial degradation [Wickland et al., 2007].
 All indices of the fens' DOC composition (SUVA254, FI, a250/a365 and C/N ratio) changed significantly between inflows and the outflows toward the composition exhibited at the palsa and bog catchment outflows. The fen inflow to outflow difference in SUVA254, a250/a365 and C/N ratios became larger with lower fen flow through and higher fen outflow water temperature. Since there was no significant relationship between fen flow through and water temperature, the flow through and temperature relations with composition indices appear independent of each other. The change in DOC composition from fen inflow to outflow and its dependence on water residence time and temperature are likely the result of several biotic and abiotic processes including DOC production, sedimentation, transformation and degradation [Berggren et al., 2009b; Bertilsson and Tranvik, 2000; Kalbitz et al., 2003b; von Wachenfeldt and Tranvik, 2008]. A mixing model can be used to estimate the hypothetical DOC composition of DOC produced within the fen assuming that all DOC from upstream sources pass through the fen unaltered, i.e., that within-fen DOC production is the only process affecting fen DOC composition and export. SUVA254 was the most suitable index for this analysis as it is linearly correlated with DOC aromaticity. The mixing model SUVAFen × DOCFen = (SUVAOut × DOCOut) − (SUVAIn × DOCIn), where SUVA is the seasonal average SUVA254, DOC is the annual DOC flux and Out, In and Fendenotes DOC composition and flux at the fen outflow, inflow and for the fen net DOC export that is assumed to represent within-fen produced DOC. The mixing model was solved for the only unknown, SUVAFen, for the north fen in 2007 and 2008, and for the south fen in 2008. The residual value representing the SUVA254 of DOC produced within the fen was 3.21 ± 0.5 l mg−1 C m−1 (mean ± 2 SD). This estimate is similar to the composition at the palsa catchment outflow (2.94 l mg−1 C m−1, 2 year mean), but it is unlikely to be representative of DOC produced in the fens as DOC extracted from the fen peat has a SUVA254 of 2.24 l mg−1 C m−1 [Roehm et al., 2009]. In addition, Carex rostrata, a common sedge in the fens, is a significant source of nonaromatic, low molecular weight organic acids [Koelbener et al., 2010]. This suggests that the assumptions of the pure mixing model are invalid and infers that processes other than within-fen production of DOC, i.e., microbial DOC degradation, are required to explain the change in DOC composition from fen inflow to outflow. Microbial degradation of DOC results in the release of DOC with higher SUVA254 than the source material [Berggren et al., 2009b]. The fens in Stordalen receive a substantial supply (fen DOC inflow between 16 and 47 g C m−2 yr−1) of low aromatic DOC (SUVA254 <2.0 l mg−1 C) from upstream sources. Hence, while the fens are net sources of DOC to the catchment, they could simultaneously be important catchment locations for microbial degradation of DOC from upstream sources. Northern lakes act as locations of degradation of terrestrial DOC, with their sink strength related to water residence time and water temperature [Algesten et al., 2004; Mattsson et al., 2005] and it appears that an analogous situation can exist with the study fens with water residence time (indicated by water flow through rate) and water temperature affecting the extent of transformation in DOC composition from inflow to outflow. The fens in Stordalen are large sources of CH4 (32 g C m−2 yr−1) [Bäckstrand et al., 2010], and it possible that degradation of DOC from upstream sources plays some role in this, but this conjecture need to be tested. Microbial degradation in incubated DOC samples primarily occurs during the first 1–3 d [Kalbitz et al., 2003a; Roehm et al., 2009], which means the water resides in the Stordalen fens long enough to remove most of the bioavailable DOC. This stresses the importance of fens for the composition of DOC reaching downstream aquatic ecosystems, a function that might be enhanced as the fens expand within the peatland complex.
 The results in this study show that interactions between hydrology and permafrost are important in determining DOC export from a northern peatland complex. In the Stordalen peatland complex permafrost thaw leads to conversion from palsa into bog or fen. The palsa has low rates of DOC export compared to other portions of the complex because of a low DOC transport capacity as runoff generation is restricted. The palsas act as headwater catchments and supply downstream ecosystems with DOC that is characterized by having on average higher aromaticity, average molecular weight, C/N ratio and being primarily derived from terrestrial rather than microbial sources, characteristics that have been linked to low bioavailability. A brief snowmelt period is important for the palsa DOC export providing most of the annual DOC export, and early snowmelt DOC was of significantly lower aromaticity than when the active layer had started to develop.
 Transition from palsa to bog following thaw is not likely to alter the total DOC export significantly; the palsa and bog catchments in this study had similar flow-weighted DOC concentrations at their outflows. The composition of DOC exported from the bog catchment differed from that of the palsa: it had consistently lower SUVA254 and higher FI, indicating lower aromaticity and increased importance of microbially derived DOC, possibly linked to the different vegetation composition. In Stordalen it is possible that the bog type peatland is a transient peatland type and not a stable end state following thaw, since it is the surrounding palsas that hydrologically disconnect the bog from the upstream catchment. The bog areas could become hydrologically connected to the eastern lake once the surface permafrost in palsa parts thaws, which would turn the whole peatland complex into a single fen.
 The fens in Stordalen act as greater net DOC sources than either the palsa or the bog. Fen DOC export is also less limited by hydrological transport potential than palsa and bog DOC export – a large proportion of the annual fen DOC export occurs during the summer/autumn period when daily flow through is low while the annual palsa and bog DOC export is dominated by the snowmelt period. Shifts in DOC composition from fen inflows to outflows could partly be due to sources of highly aromatic DOC within the fens but also indicate substantial degradation of DOC derived from upstream sources. While terrestrial ecosystems are generally characterized as DOC sources and aquatic ecosystems as leaky DOC conduits [Cole et al., 2007], the fens are thus found to exhibit characteristics of both a terrestrial and aquatic ecosystem. Expansion of fens because of continued palsa thaw in a peatland complex such as Stordalen is thus likely to increase the peatland complex DOC export, particularly outside the snowmelt period, but also to increase the role of fens as catchment sites for DOC degradation because of increased water residence time and water temperatures. Altered quantity, timing and composition of DOC export from subarctic peatland complexes can in turn have downstream effects on aquatic secondary metabolism.
 The research was funded by a Natural Sciences and Engineering Council of Canada Discovery grant to N.T.R. Financial support for D.O. was provided by Sixten Gemzéus Stiftelse, Sweden-America Foundation and the Global Environmental and Climate Change Centre (GEC3). The analysis of DOC quantity and composition were done in collaboration with Abisko Scientific Research Station and the Climate Impacts Research Centre (CIRC) of Umeå University. We would like to especially thank Reiner Giesler. We also thank K. Bishop (SLU) and two anonymous reviewers for their very constructive comments and suggestions on our manuscript.