Carbon dioxide supersaturation in peatland waters and its contribution to atmospheric efflux from downstream boreal lakes



[1] Carbon fluxes at two boreal peatland-dominated catchments in northeastern Alberta were investigated through the analysis of fen and lake water chemistry and the measurement of partial pressure of carbon dioxide (pCO2) using headspace gas analysis. All waters had low pH (<5.3) and Gran alkalinity (<2.3 mg L−1), high DOC (>15 mg L−1), and were supersaturated with carbon dioxide (CO2) with respect to the atmosphere (CO2: 1.2–54 times atmospheric). Nonetheless, CO2 concentrations in the study lakes were significantly lower compared with surface water pools in associated fen systems. Average atmospheric flux of CO2 from the two lakes were 0.18 and 0.48 g C m−2 d−1, while potential fluxes from small surficial pools on the fen complexes were an order of magnitude higher. The higher average efflux estimated at one of the lakes (relative to the other) was attributed to shorter lake residence time and smaller relative lake area (i.e., higher relative carbon loading). Carbon mass balances for the lakes suggest that they act as conduits for dissolved CO2 from surrounding fen complexes to the atmosphere. In one of the two study catchments (with negligible groundwater sources), inputs of dissolved CO2 from fen surface waters supported a substantial component (∼30%) of the lake atmospheric CO2 efflux.

1. Introduction

[2] Globally, peatland ecosystems cover an area estimated at 420–450 million hectares and play an important role in the carbon (C) cycle [Gorham et al., 1984]. The vast majority of these ecosystems are found in the boreal regions of the northern hemisphere [Gorham, 1991] with approximately one third of all boreal peatlands in Canada. Boreal peatlands are recognized as being important net long-term sinks for atmospheric carbon dioxide (CO2) [Moore et al., 1998], but demonstrate interannual variation that can switch them from being sinks to sources [Roulet et al., 2007]. Eddy covariance techniques are often used to measure CO2 exchange over peatlands [e.g., Frolking et al., 1998] as these C-flux studies are considered to provide the best estimate of ecosystem balances. This approach is limited in that the underlying processes remain undefined; fluvial loss of C in the form of dissolved organic carbon (DOC) and dissolved CO2 are not often considered in these studies [Worrall et al., 2005]. Notably there are several investigations of multiyear peatland carbon balances that include exports of C in stream waters draining the peatland [Billett et al., 2004; Roulet et al., 2007; Nilsson et al., 2008; Dinsmore et al., 2010]. Peatlands exhibit direct atmospheric CO2 uptake through fixation; however their surface waters are highly supersaturated with CO2 relative to the atmosphere [Cole et al., 2007] thereby demonstrating potential for CO2 release. A number of studies have measured dissolved concentrations and atmospheric efflux of CO2 in streams draining C rich soils and identified them as hot spots for CO2 degassing to the atmosphere [Dawson et al., 2004; Billett and Moore, 2008; Dinsmore et al., 2009]. Considerably less is known about CO2 concentrations in open waters on peatland (fen) surfaces, and consequently their potential to emit CO2 to the atmosphere.

[3] As a result of hydrologic transport, the majority of C in lakes is of terrestrial origin [Molot and Dillon, 1996], and lakes, like peatland drainage waters, are frequently supersaturated with CO2 [Sobek et al., 2005]. Overall, lakes are an important exit route for terrigenous C, and the majority of lakes worldwide typically act as net sources of CO2 to the atmosphere [Cole et al., 1994]. Allochtonous C entering lakes is largely in the form of DOC [Hope et al., 1996; Striegl et al., 2001], and much of this can be converted to CO2 as in-lake respiration generally exceeds primary productivity [del Giorgio et al., 1997]. Nonetheless, inputs of CO2 rich groundwaters can also contribute directly to CO2 emission [Striegl and Michmerhuizen, 1998]. Further, C burial in oligotrophic and dystrophic lake sediments is an important component of C cycling [Hanson et al., 2004] and storage in sediments may be comparable to evasion losses [Molot and Dillon, 1996]. The controls on CO2 supersaturation in boreal lakes have received considerable attention in recent years [e.g., Algesten et al., 2003; Sobek et al., 2003; Kortelainen et al., 2006; Tranvik et al., 2009] and global CO2 emission from surface waters is expected to be impacted by both anthropogenic influences and climate change [Tranvik et al., 2009].

[4] Traditionally the partial pressure of CO2 (pCO2) has been estimated from pH and bicarbonate (HCO3) concentrations [e.g., Stumm and Morgan, 1996], but direct measurements [e.g., Cole et al., 1994; Kling et al., 1991] are favored, particularly in light of poor predictability in freshwaters with relatively high DOC concentrations and low pH [Herczeg and Hesslein, 1984]. Comprehensive studies of large data sets have generally relied on both approaches, as those obtained via direct measurement were rare [Sobek et al., 2005]. Direct measurements however, are becoming more widely used to record pCO2 in both streams and lakes [e.g., Dawson et al., 2004; Hope et al., 2004; Jonsson et al., 2007; Dinsmore et al., 2009]. Considerable advancement in the understanding of the sources and sinks of aquatic C has been made of late; however, in Canada studies have been limited to a select number of regions [e.g., Hamilton et al., 1994; Dillon and Molot, 1997; Dinsmore et al., 2009; Roehm et al., 2009], this despite the vastness of the boreal forest biome. Water chemistry and landscape predictors of pCO2 are inconsistent between regions [Whitfield et al., 2009] and there is a need to better understand the role of boreal freshwaters as active pipes [e.g., Cole et al., 2007]. Additional and more diverse measurements that better describe dissolved C in surface waters are required. This is particularly true of the surface waters that supply lakes in peatland dominated catchments, as these lakes are known to have high CO2 efflux rates.

[5] This study combined detailed water chemistry and hydrological data at two boreal catchments in western Canada in order to describe C fluxes. The objectives of this study were to document surface water pCO2 at two lakes and their associated fen complexes, to estimate the potential atmospheric flux of CO2 from the fen-lake system surface waters, and to identify the contribution of dissolved CO2 in surface waters to emissions from the lake surface. Surface water pCO2 and DOC (and other water chemistry parameters) were measured across the lakes and fen dominated peatlands.

2. Study Region and Sites

[6] The two study catchments NE07 and SM08 are situated approximately 60 km northeast (57.13°N, 110.89°W) and 50 km south (56.21°N, 111.19°W), respectively, of the city of Fort McMurray, Alberta (Figure 1). The catchment NE07 contains a small lake (NEL) and one large fen peatland (NEF). The SM08 catchment contains a lake (SML) that is bordered by two large fen complexes, SME and SMW, on the east and west ends of the catchment, respectively (Figure 2). Both lakes are characterized as acidic and dystrophic. The study sites were selected due to their contrasting hydrological behavior (reflecting the range in character typical for the region), SM08 being evaporative, and NE07 being a more throughflow system (more of the precipitation leaves as runoff from the lake rather than as evaporation to the atmosphere).

Figure 1.

River and stream networks in northern Alberta, Canada, and the location of the NE07 and SM08 study sites (open circles) in relation to the town of Fort McMurray (open square).

Figure 2.

Location of sampling points (solid circles) across the fen complexes and lakes at the study catchments, (left) NE07 and (right) SM08.

[7] NE07 is the smaller of the two catchments (Table 1), and NEL has a maximum depth of 2 m [Bennett et al., 2008]. The catchment is dominated by the NEF fen complex that extends southward from the lake and stretches for over 1 km from the shore (Figure 2). Small pools of water found in surficial depressions are scattered across the fen, with larger areas of standing water present along the middle of the fen. The peatland is mostly open, with short (1–2 m) black spruce (Picea mariana) and tamarack (Larix laricina) sparsely distributed. SM08 is located in the Stony Mountains; the lake (SML) has a larger relative area (Table 1) and a maximum depth of 2.5 m [Bennett et al., 2008]. SME is the larger fen complex, stretching for nearly 2 km southeast from the eastern shoreline, where it is open to the lake (Figure 2). Trees are largely absent across both fen complexes (SME and SMW). Large areas of these fen complexes were covered with surface water, particularly toward the distal (SME) and central (SMW) areas of the complexes, while smaller pools were found elsewhere. The depth of peat is approximately 2 m and movement of water in the peatlands occurs predominantly in the acrotelm (the shallow oxic zone spanning the range of water table heights) due to decreasing hydrologic conductivity with depth. Runoff through the acrotelm to the lakes has been observed following precipitation events at the fen complexes.

Table 1. Average Annual Runoff (Q), Catchment Area, Peatland Area, Lake Area, Lake Volume, Retention Time, and Wind Speed at 10 m (U10) for the Study Catchments
Qm yr−10.1130.082
Catchment areaha579960
Peatland area%77.566.5
Lake area%1.924.8
Lake volumem31.12 × 1051.69 × 106
Retention timeyr0.261.40
U10m s−12.943.37

3. Methods

3.1. Sample Collection and Analysis

[8] Measurements of pCO2 and surface water chemistry were made along transects in each catchment. Water samples were collected from surficial pools at 100 m intervals along two transects at each of the fen complexes; repeat measurements were made during mid June and late August, 2006. Lengthwise transects ran roughly through the middle of each of the fen complexes in an approximate linear fashion starting at the lake and ending at the fen apex. Widthwise transects ran perpendicular to the lengthwise transects and spanned the widest area of each fen complex. The transects ranged in length from 600 to 1900 m (Figure 2). Surface water samples were also collected along the length of each lake in June and August. In total, 124 samples were collected (both periods combined).

[9] The pCO2 was measured in situ using a headspace analysis technique [Cole et al., 1994; Jonsson et al., 2003; Sobek et al., 2003]. Water samples for pCO2 analysis were collected from a depth of 10 cm by submerging 125 mL clear glass bottles (both lake and fen); the bottles were completely filled with water and sealed with rubber septums to ensure that no air was present. A gas headspace was established by collecting 25 mL of ambient air from 2 m above ground and against the wind at each sampling location and injecting the air into the overfilled glass bottle while simultaneously extracting 25 mL of water using a double plastic syringe system. The sample bottle was then shaken for 1 min to equilibrate dissolved CO2 within the headspace [Cole and Caraco, 1998]. Once equilibrated, 20 mL of the headspace gas was then removed from the bottle using the emptied gas syringe by simultaneously injecting 20 mL of the sample water back into the bottle. Immediately following, the 20 mL headspace sample was slowly injected into an infrared gas analyzer (IRGA; PP Systems EGM-4 Environmental Gas Analyzer for CO2) using three-way stopcocks and rubber tubing for instantaneous analysis. Use of a larger sample bottle (∼1 L) traditionally used for headspace analysis was precluded by the physical limitations of sampling surficial pools (rather than deeper open waters). The headspace sample volume was sufficient to obtain a stable reading on the IRGA, and comparison with values obtained using a larger bottle and headspace (60 mL) demonstrated close agreement. For each sample, atmospheric pCO2 and barometric pressure (using the IRGA) and sample extraction temperature were measured.

[10] At each sampling location (in fen and lake), bulk water samples for all remaining chemical analyses were collected in HDPE bottles. Conductivity, pH, and water temperature were measured in the field using a portable pH/conductivity meter. Water samples were returned to the laboratory on ice and analyzed for a full suite of chemical parameters including HCO3, DOC, dissolved inorganic carbon (DIC) and Gran alkalinity within 1 week of collection. Gran alkalinity and HCO3 were measured using a PC-Titration Plus® system. Dissolved organic and inorganic carbon was determined using a Shimadzu® TOC-VCPH carbon analyzer.

3.2. Calculations and Analysis

3.2.1. Values of pCO2

[11] “True” values of pCO2 measured in situ were corrected according to Cole et al. [1994] using Henry's law and the CO2 fugacity-pressure relationship given by Weiss [1974].

3.2.2. CO2 Flux

[12] The potential CO2 flux from the surface waters of the study lakes to the atmosphere was estimated according to Raymond et al. [1997]:

equation image

where kCO2 is the gas transfer coefficient (cm h−1), CO2(water) is the concentration of CO2 in the surface water, and CO2(sat) is the concentration of CO2 at equilibrium with the atmosphere above the surface water [Cole and Caraco, 1998] (e.g., saturation concentration). The gas transfer coefficient for CO2 was estimated using temperature-dependent Schmidt numbers for CO2 (ScCO2) from Jähne et al. [1987].

equation image

The exponent n (equation 2) varies depending on the process dominating diffusion; however −0.67 was used here based on Jähne et al. [1987], Cole and Caraco [1998], and Rantakari and Kortelainen [2005]. The piston velocity for sulphur hexafluoride (k600) normalized to a Schmidt number of 600 was calculated using the function derived by Cole and Caraco [1998] using wind speed at 10 m (U10):

equation image

Daily average wind speed at a height of 2 m above lake and fen was measured at each catchment during the 2006 open-water season (7 months spanning April through October), and used to calculate U10 according to Crusius and Wanninkhof [2003].

[13] Potential CO2 fluxes from the surface waters of the fen complexes were also estimated under the assumption that they behaved similarly to the lakes, but using two wind speeds: 100% and 10% of observed speed, as air movement over the fens is expected to be slower owing to microtopographical features and the presence of vegetation. At wind speeds of less than 3 m s−1, the effect of wind on CO2 flux is not expected to be large [Cole and Caraco, 1998]. The calculations using the two wind speeds were used to illustrate a potential range in CO2 flux rates from pools on the fen surface.

[14] The data sets were tested for outliers (75th/25th percentile ± 1.5 times the interquartile range, respectively), and data analysis followed their removal (n = 7). A Kruskal-Wallis test was used to compare pH, DOC and pCO2 across the five study areas (NEL, SML, NEF, SME, SMW). The two sampling periods were similarly tested in order to identify potential changes in concentration at the lakes and fens. Correlations between measured parameters were tested for significance using Spearman's rank correlation coefficient (ρ). An alpha value of 0.05 was used as the minimum level of significance. Average pCO2 based on both sampling periods was used for atmospheric flux calculations.

3.2.3. Carbon Mass Balance

[15] A mass balance approach was used to describe the sources and sinks of various forms of C within the two study lakes. The purpose of these calculations was to investigate, in a general sense, the forms and amounts of C entering and exiting the lakes in order to identify the origin of the CO2 degassed from the lakes. The annual (2006) fluxes of C entering and leaving the lake were calculated as:

equation image

Ongoing investigation of the hydrology of the study catchments made it possible to calculate mass balances at NEL and SML. The key period of hydrologic relevance in the region is the open water period, as it is during this time that the majority of water movement (and therefore fluxes to and from the lake) occurs in the study catchments. As such, it was necessary to assume that observations of water chemistry available from the two sampling periods were reflective of the averages for the open water period. Dissolved inorganic carbon was estimated as the sum of dissolved CO2 and HCO3 (HCO3 was rarely detected in surface waters, and on average accounted for <10% of DIC when present). For groundwaters, HCO3 concentration was used to approximate DIC (pCO2 not measured). Inputs of DIC and DOC to the lake (DICIn and DOCIn, respectively) were calculated using average bulk deposition to the lake (S. Chang, unpublished data, 2010), and surface water (this study) and groundwater (K. Tattrie, unpublished data, 2010) fluxes according to contributions from the catchment as surface and groundwater [Schmidt et al., 2010]. Loss to the atmosphere (CO2Efflux) was estimated according to equation 1 using average CO2 concentrations (June and August measurements) and wind speed (K. Tattrie, unpublished data, 2010) for the open water period. The C sedimentation rate (300 mg C m−3 yr−1) presented by Flanagan et al. [2006] and lake volume were used to estimate C loss to the sediments (CSed). Average lake DIC and DOC concentrations (this study) and 2006 discharge (Table 1) were used to calculate outflow losses (DICOut, DOCOut).

[16] The influence of hydrologic CO2 on atmospheric emissions was explored by estimating the sources of CO2 in the catchment (e.g., in-lake mineralization of DOC and CO2 present in catchment waters in dissolved form and therefore available for degassing from the lake surface without further transformation (hydrologic CO2)). Hydrologic CO2 is made up of dissolved CO2 entering the lake both in surface water (surface water CO2) and groundwater (groundwater CO2) inflows. For each lake, surface water CO2 was calculated using DIC concentration (DIC ≈ CO2) and water yield (of surface water), and related to the total pool of CO2 available for atmospheric efflux according to:

equation image

The terms DICS, DICG and DICP are inputs of dissolved inorganic carbon to the lake from surface waters, groundwaters, and precipitation, respectively. An analogous calculation was used to estimate the proportion of CO2 efflux attributable to groundwater CO2. This approach assumes that CO2 generated from DOC, calculated as the difference between DOCIn and DOCOut and CSed, is lost to the atmosphere, as is the difference in DIC between inputs to (DICIn) and outputs from (DICOut) the lake; collectively this represents the total amount of CO2 available for efflux to the atmosphere. In reality a fraction of the DOC converted to CO2 would be lost from the lake via runoff, so these calculations may be a somewhat conservative estimate of the proportion of efflux attributable to hydrologic CO2.

4. Results

[17] The lakes had significantly higher pH (p < 0.05) and lower DOC (p < 0.01) than the fen waters, and the DOC concentration was approximately twofold higher at NEL than SML (p < 0.001; Table 2). Dissolved organic carbon concentrations were similar at the three fen complexes; nonetheless measurements within each fen complex were more variable than in the lakes. Bicarbonate concentrations in surface waters were very low at NE07, while no HCO3 was detected at SM08. At NE07 there was no significant difference in pCO2 between water samples with and without HCO3. The partial pressure of CO2 was not correlated with pH, alkalinity or HCO3, and was only weakly correlated with DOC (ρ = 0.28; p < 0.01). A subset of this data, lake pCO2 and DOC, was significantly correlated and demonstrated a stronger relationship (ρ = 0.73; p < 0.01; Figure 3). Surface water DIC across the study catchments was strongly dependent on dissolved CO2 concentrations.

Figure 3.

Dissolved organic carbon (DOC) and partial pressure of carbon dioxide (pCO2) in fen complex (open circles) and lake (open squares) surface waters at the two study catchments (ρ = 0.28; p < 0.01). The regression line demonstrates the relationship for the lakes (ρ = 0.73; p < 0.01).

Table 2. Average (Both Sampling Periods: June and August) pH, Conductivity (Cond), Gran Alkalinity (Alk), Dissolved Organic Carbon (DOC), Partial Pressure of Dissolved Carbon Dioxide (pCO2), Estimated Efflux, and Estimated Efflux at 10% of Measured Wind Speed (U10/10; Fens Only) for the Lakes (NEL, SML) and Fen Complexes (NEF, SME, SMW)a
ParameterUnitsNEL (5)SML (8)NEF (48)SME (34)SMW (29)
  • a

    Standard deviations are in parentheses, and the number of samples for each study area is included beside the name.

pH 5.27 (0.15)4.84 (0.33)4.18 (0.77)3.88 (0.18)3.91 (0.11)
Condμs cm−136.6 (1.2)21.1 (5.0)40.9 (17.0)56.7 (13.2)70.0 (10.5)
Alkmg L−12.3 (0.6)−0.6 (0.8)0.2 (7.4)−9.7 (5.9)−8.8 (2.2)
DOCmg L−134.3 (3.7)15.7 (2.1)40.3 (11.2)44.1 (8.5)45.8 (6.1)
pCO2μatm1760 (491)849 (310)10101 (3803)7865 (2072)9876 (2665)
Efflux (U10)g C m−2 d−10.46 (0.18)0.18 (0.12)3.4 (1.3)2.8 (0.8)3.6 (1.0)
Efflux (U10/10)g C m−2 d−12.1 (0.8)1.6 (0.4)2.1 (1.0)

[18] Measurement of dissolved CO2 using the headspace gas analysis revealed a wide range in pCO2 concentrations across the two catchments; lake concentrations were more consistent (Table 2 and Figure 4), likely due to the homogeneity afforded by their shallow depth and regular mixing by wind. The partial pressure of CO2 at NEL (mean of 1760 μatm) was significantly higher than at SML (mean of 849 μatm) (p < 0.01), however the lake concentrations were significantly lower than the peatland water concentrations (p < 0.001; Figure 4). Comparison of the June and August sampling periods at the five individual sites did not reveal significant differences in pCO2; however, pooling the data for the three fen sites suggested that pCO2 was lower during August (mean decrease = 2127 μatm; p < 0.01). There was no significant difference between June and August pCO2 when lake data were pooled. Atmospheric pCO2 was not different between sites or sampling periods.

Figure 4.

Box plots of the partial pressure of carbon dioxide (pCO2) in the lake and fen complex waters at the two study catchments. Each box encloses 50% of the data with the median value displayed as a line and the top and bottom of the box indicating the upper and lower quartiles, respectively. Mean values are shown (open square).

[19] Calculations of saturation levels for lake and fen surface waters revealed consistent supersaturation relative to overlying air. The level of supersaturation ranged widely, but was much higher in fen surface waters (range: 3.5–54 times atmospheric), compared with the lakes (range: 1.2–6.8 times atmospheric). The majority of peatland surface water samples were 20 to 30 times supersaturated with respect to the atmosphere. Average daily fluxes of CO2 from the supersaturated lake surfaces to the atmosphere were estimated to be 0.18 (SML) and 0.48 (NEL) g C m−2 d−1 for the open water period (Table 2). Potential atmospheric flux estimates for the fen waters exposed to the atmosphere averaged 3.1, 3.0 and 2.3 at 100% of observed wind speed and 2.1, 2.0, and 1.6 g C m−2 d−1 at 10% of observed wind speed for NEF, SMW, and SME, respectively (Table 2).

[20] Carbon mass balance calculations provided reasonable agreement between inputs and outputs. At NEL, loss of DOC in runoff and efflux of CO2 to the atmosphere represented the largest losses of C from the lake (Table 3). At SML, the dominant loss was estimated to be to the atmosphere (Table 3). At both catchments, DOC losses from the lakes were much lower than inputs, and the majority of the C emitted to the atmosphere is likely derived from DOC. At SML, a greater proportion of DOC inputs are lost to the atmosphere and sediments due to greater relative lake area and longer residence time. The two catchments contrast in their sources of hydrologic CO2 due to varying hydrological character. Nearly all of the hydrologic CO2 input to NEL occurs via surface water (96%) while at SML groundwater CO2 inputs predominate (85%). The contribution of surface water CO2 and groundwater CO2 to atmospheric efflux at NEL total approximately 30 and 1%, respectively. For SML, contributions of hydrologic CO2 to atmospheric losses were approximately 3 and 33% for surface water and groundwater, respectively.

Table 3. Components of the Carbon Mass Balances at NEL and SML Normalized to Lake Areaa
  • a

    Inputs of carbon to the lake include dissolved inorganic carbon from surface waters (DICS), groundwater (DICG), and precipitation (DICP) and dissolved organic carbon (DOCIn). Export of carbon from the lake is represented via dissolved inorganic and organic carbon in runoff (DICOut and DOCOut), atmospheric emission (CO2 efflux), and carbon sedimentation (C sediment).

DICSg C m−2 yr−1350.5
DICGg C m−2 yr−10.96.5
DICPg C m−2 yr−10.60.7
DOCIng C m−2 yr−125018
DICOutg C m−2 yr−15.30.2
DOCOutg C m−2 yr−11704.4
CO2 effluxg C m−2 yr−18830
C sedimentg C m−2 yr−10.30.3

5. Discussion

5.1. Surface Water Chemistry

[21] Measurement of pCO2 concentrations across the five study sites revealed that CO2 concentrations were very high in fen surface waters, but exhibited considerable variability on these landscapes. There are few observations of CO2 concentrations in peatland surface waters available for comparison, as measurement often takes place at considerable depth in the peat [e.g., Nilsson and Bohlin, 1993]. Hope et al. [1996] reported CO2 supersaturation for wetland waters ranging from 114 to 213 (times atmospheric), which are higher than found in the current study (mean: 26.9), but Hope et al. [1996] sampled interstitial rather than surficial waters. Hamilton et al. [1994] reported CO2 concentrations ranging from 2 to 90 times atmospheric for boreal wetland ponds, and their flux estimates are within the range in estimates reported for the lake and fen waters in the current study; the ponds may be structurally intermediate to the lake and fen surface waters compared herein. High variability in pCO2 across the fen complexes was likely due in part to morphology of the sample locations; the extent of surface water pools and the surrounding vegetation community were variable across the complexes, which may result in differential respiration, mineralization and decomposition intensities.

[22] At NEF, HCO3 potentially originates from local groundwater inputs [e.g., Whitfield et al., 2010]; however, regression of pCO2 with pH, alkalinity and HCO3 yielded no evidence of a relationship in surface waters at the study catchments. In the absence of HCO3 as a source of CO2, DOC originating in peatlands may play an important role [e.g., Nilsson et al., 2008]. Whitfield et al. [2009] reported a positive relationship between lake pCO2 and relative peatland area of catchments in the region, suggesting an organic carbon influence. Evaluation of the data set as a whole (fen and lake samples) indicated that CO2 and DOC are only weakly related. The fen data vary widely and the difference between lake and fen surface water character appeared to drive the correlation; as such no clear predictive relationship exists. Among the lake data, however, pCO2 and DOC were more strongly related (Figure 3). This is consistent with other studies that have shown organic C to be a good indicator of pCO2 concentration in lakes [Hope et al., 1996; Riera et al., 1999; Sobek et al., 2003; Algesten et al., 2005] where respiration of organic carbon and decomposition of sediments can be the primary sources of CO2 lost to the atmosphere [Striegl et al., 2001]. Given the high DOC concentrations in the lake and fen complex surface waters in the present study, mineralization of organic matter likely contributes to the elevated pCO2 levels. The pCO2 at the study lakes was much lower than in the surrounding fens, and NEL and SML exhibited differences in both DOC and dissolved CO2 concentrations (Table 2 and Figure 4), suggesting that contrasting catchment structure and hydrology are important influences.

[23] Both catchments have extensive peatland areas (Table 1) yielding water to the lakes that is high in both DOC and CO2. The NE07 catchment exhibits throughflow hydrology (e.g., lower lake retention time), which may result in higher lake pCO2 values than at SML owing to the more rapid runoff of fen surface water high in DOC and dissolved CO2. Worrall et al. [2005] estimated that 20–36% of soil respiration can exit from peat as hydrologic CO2 and Hope et al. [2004] linked stream CO2 to whole catchment respiration where connections between soil and stream are tight. Therefore lake pCO2 strongly depends on catchment processes. At SM08, where relative lake area is much higher, lower concentrations of DOC and pCO2 in SML likely occur due to lower relative loading of allochtonous C, and greater time for in-lake processes (e.g., mineralization, efflux) to occur. This is consistent with other studies that report lower pCO2 for lakes with larger surface area and lower flushing rates [Hope et al., 1996; Striegl et al., 2001; Sobek et al., 2003].

5.2. Atmospheric CO2 Efflux Estimates

[24] Lakes are well known for their role in transferring terrestrially fixed carbon to the atmosphere in boreal [Algesten et al., 2003], temperate [Hanson et al., 2004], and tropical [Richey et al., 2002] regions. Likewise the lakes in the present study were estimated to emit CO2 to the atmosphere. A number of explanations for the high concentrations of CO2 observed in lakes worldwide have been proposed, including in-lake respiration of allochtonous organic carbon [del Giorgio and Peters, 1993], inputs of CO2 from supersaturated groundwaters [Striegl and Michmerhuizen, 1998; Carignan et al., 2000] and photochemical mineralization [Graneli et al., 1996]. In boreal ecosystems, mineralization of imported organic matter is believed to play a central role in the supersaturation of CO2 in lake waters [Kling et al., 1991; Cole et al., 1994]. At the study catchments, the large difference in fen and lake DOC concentrations suggest that in-lake respiration and/or mineralization are important for CO2 production. Given that supersaturation levels were high during both sampling periods, and that the open-water season lasts only 7 months, these lakes are more than likely net sources of CO2 on an annual basis, which is consistent with most lakes worldwide [Cole et al., 1994; Sobek et al., 2005]. Fluxes from the study lakes (range: 0.18–0.48 g C m−2 d−1) are somewhat lower than estimates for reservoirs in Canada [see Hope et al., 2001], but are comparable to average annual flux estimates for boreal lakes in Scandinavia (−0.03–0.38 g C m−2 d−1 [Jonsson et al., 2003]; 0.049–0.075 g C m−2 d−1 [Rantakari and Kortelainen, 2005]) and lakes in Alaska (mean: 0.25 g C m−2 d−1 [Kling et al., 1991]).

[25] The loss of CO2 from pools of standing water at the fen surface was estimated following the approach for lake surface waters. Fen complex surface waters likely behave differently to lakes, thus these estimates are meant to explore potential atmospheric CO2 efflux from these surface water pools, which while numerous, are small and shallow. There is the potential for significant efflux of CO2 to the atmosphere from the surficial pools regardless of wind speed, with flux rates ranging from 3 to 20 times estimated lake efflux (due to the highly supersaturated state of fen waters). Drivers of CO2 emissions in fens might include precipitation [e.g., Rantakari and Kortelainen, 2005]; where turbulence and water flow stimulated by precipitation may be more important than wind speed for gas flux from sheltered fen surface waters. The challenges associated with determining surface water coverage due to spatial variability as well as variation in pCO2 preclude application of a uniform atmospheric CO2 flux estimate for the full complexes, but the estimates generated here can be compared to respiration rates for similar systems in the region. Glenn et al. [2006] estimated total ecosystem respiration using eddy covariance techniques (averaged for the 6 month period: May–October) as 1.39 and 1.82 g C m−2 d−1 for two boreal peatlands in Alberta. These respiration rates are similar to but lower than the efflux estimates calculated for surficial pools in the current study. The estimates for the current study reflect potential CO2 emission from pools, which are likely hot spots of emission and not representative of the wider landscape. If estimating CO2 flux from fen waters using the boundary condition approach for lakes is a reasonable approximation, fluxes from large peatland surficial pools exposed to the atmosphere can be (locally) significant and are likely higher than surrounding (drier) areas. In peatlands where surface water coverage is extensive, these losses may represent a quantitatively important flux that should be addressed where eddy covariance approaches that integrate over select areas are conducted.

[26] Lower pCO2 concentrations in the fen complex surface waters during August sampling (pooled data) suggest possible release of CO2 directly to the atmosphere from fen waters in this study. Fen waters had significantly warmer temperatures in August (average increase ∼2.5°C), which could stimulate enhanced outgassing of CO2 and lower pCO2 [e.g., Dawson et al., 1995]. Precipitation is highest during the summer months (∼65 mm month−1) and the rainfall could also be a potential driver of CO2 loss. Alternative mechanisms for the difference observed between the two sampling periods could be the horizontal movement of waters high in pCO2 toward the lake with replacement of waters with low pCO2 from surrounding areas, or a change in respiration rate. Despite this observation of lower pCO2 concentrations in August, it remains to be resolved to what extent CO2 is emitted to the atmosphere from fen surface waters, as Clymo and Pearce [1995] reported that vertical transport of CO2 through waterlogged peat is considerably slower than horizontal transport, and thus where surficial pools are small and shallow, CO2 efflux may be limited by the rate of horizontal movement to these exposed locations. In contrast, lake pCO2 concentrations were consistent between sampling periods. Continuous supply of waters with high DOC and CO2 concentrations from the fen complexes, and mixing of the epilimnion are likely responsible for more stable values observed in the lakes during the study period. Other studies have found seasonal patterns in pCO2 [e.g., Kling et al., 1991; Sobek et al., 2003; Kortelainen et al., 2006]; however, differences are typically observed between winter and the open water seasons, and while spring pulses following ice out may occur in the study area, lake sampling was not conducted early enough for this to be captured.

5.3. Sources of CO2

[27] Mass balances serve as a useful and widely used tool for describing catchment behavior. The calculations performed in the current study demonstrated the importance of CO2 efflux for the C balance of the study catchments, with estimated losses to the atmosphere comparable to runoff losses at both catchments. Lower DOC concentrations in the lakes relative to the fens indicated the key role of DOC as a source of dissolved CO2 (through mineralization), and as a source of C to the sediments. The mass balances however, do not demonstrate total agreement between inputs and outputs and each component of the calculations is subject to some uncertainty that may contribute to these gaps. Any measurement errors are anticipated to be small, and for DIC and DOC these would largely be offset by the mass balance calculation itself. A catchment-based mass balance approach is also sensitive to the water fluxes used, and calculating annual C balance at these lake catchments is complicated by interannual variability in hydrologic influences [e.g., Schmidt et al., 2010]. The hydrology of these catchments has been studied in detail over several years, however, and the modeled discharge (surface water, groundwater and lake runoff) estimates used are specific to the year of study. Thus uncertainty due to the water fluxes used is expected to have only a minor influence on the results. In lakes where residence times are longer than one year, hydrologic influences on water chemistry can be expected to carry over between years (as at SM08) and the use of observations from a single year may have contributed to the greater difference calculated for this catchment's balance. No measurements of C loss to the sediments were made, and it remains unknown whether the estimated sedimentation rate used adequately reflects the behavior of these catchments with contrasting throughflow. Similarly, while expected to be small, loss of carbon from the catchments through ebullition was not included in the efflux estimates and C accumulation in the sediments may be overestimated. Finally, the assumptions used to approximate the character of the lake and peatland surface water during the ice-free period (from two sampling periods) may contribute the greatest uncertainty to the gap in the balances. Any difference between the average chemistry of the sampling periods and the “true” open water season average could lead to errors in the representation of runoff inputs and exports, and atmospheric efflux of CO2 from the lake surface. Despite these limitations there was reasonable agreement between total inputs and losses, and while understanding of hydrogeochemistry at these catchments remains imperfect, this exercise proved useful for the characterization of catchment C contributions to CO2 efflux from these lakes.

[28] Boreal lakes in other regions show little change in DIC between lake inflows and outflows and attribute CO2 emissions to mineralization of DOC [Sobek et al., 2006]. Catchment structure (hydrology, morphology, drainage ratio, wetland coverage) affects the amount of DOC that is imported from the catchment to the lake [Rasmussen, 1989; Hope et al., 1996; this study]. The greater atmospheric efflux rate at NEL is sustained by a larger relative (to SML) C loading. Rapid flushing of NE07 generates a large supply of DOC to the lake, but with limited time to break down, the difference in DOC concentration between fen and lake is moderate, and hydrologic CO2 is also important for atmospheric CO2 fluxes from NEL. Most of the dissolved CO2 is received in NEL from surrounding peatlands (water yield from groundwater is less than 1% of total [Schmidt et al., 2010]), and surface water CO2 makes up approximately 30% of the total atmospheric CO2 emissions from the lake surface. At SML, water residence time is longer, leaving more time for DOC mineralization processes to occur, and DOC derived CO2 emissions are important. Owing to a greater contribution of groundwater to this lake, the hydrologic contribution to atmospheric CO2 efflux is dominated by groundwater CO2, estimated to supply approximately one third of total lake CO2 efflux. Clearly, input of CO2 to the lakes via net heterotrophy and CO2 injection (e.g., hydrologic CO2) [Cole and Caraco, 1998] are important mechanisms of supersaturation at the two study lakes and injection of hydrologic CO2 is important to atmospheric efflux at both catchments. At the throughflow catchment, where surface waters account for most of the water yield to the lake, surface water CO2 rather than groundwater contributions to efflux predominate.

6. Summary and Conclusions

[29] Peatlands can act as a C source in some years [Worrall et al., 2005] and it is important that C accumulation in these ecosystems is not overestimated, particularly when extrapolating site or regional measurements to larger areas or to the whole of Canada, where peatlands cover approximately 12% of the land surface [Gorham, 1991]. Detailed measurement of pCO2 across the two boreal catchments dominated by fen complexes revealed significantly higher concentrations in their surface waters relative to the lakes. Accordingly, fluxes of CO2 to the atmosphere from fen complexes may necessitate extension of the active pipe concept from boreal lakes to include peatland surface waters that interface between terrestrial and aquatic systems. The risk of overestimating carbon storage in well connected peatland dominated catchments with throughflow hydrology is of particular concern, as more rapid flushing of these catchments will likely yield increased transfer of DOC and hydrologic CO2 from terrestrial areas to lakes and promote downstream atmospheric efflux. In these boreal fen-lake complexes that are typical for the region, hydrologic CO2 from both ground and surface waters can be an important source of the CO2 emitted from the lake surface.


[30] This research was undertaken, in part, thanks to funding from the Canada Research Chairs Program and a Natural Sciences and Engineering Research Council (NSERC) Discovery grant. Financial support was also provided by the Cumulative Environmental Management Association (CEMA) and an NSERC Collaborative Research and Development Grant awarded to J. Aherne and S. A. Watmough, as well as NSERC scholarships awarded to T. A. Seabert and C. J. Whitfield. The authors gratefully acknowledge A. Jonsson and S. Sobek for their guidance with respect to methods development. J. J. Gibson and K. Tattrie's provision of wind speed, runoff, and water yield data, and contributions to the discussions around this work are greatly appreciated. We also thank H. M. Baulch and H. Broadbent for their assistance with method development and laboratory analyses at Trent. Comments from anonymous reviewers greatly enhanced the manuscript.