Carbon release from boreal peatland open water pools: Implication for the contemporary C exchange


  • Luc Pelletier,

    Corresponding author
    1. Department of Natural Resource Sciences, McGill University, Sainte-Anne-de-Bellevue, Quebec, Canada
    2. Center for Research in Isotopic Geochemistry and Geochronology, Université du Québec à Montréal, Montréal, Quebec, Canada
    Search for more papers by this author
  • Ian B. Strachan,

    1. Department of Natural Resource Sciences, McGill University, Sainte-Anne-de-Bellevue, Quebec, Canada
    Search for more papers by this author
  • Michelle Garneau,

    1. Center for Research in Isotopic Geochemistry and Geochronology, Université du Québec à Montréal, Montréal, Quebec, Canada
    2. Department of Geography, Université du Québec à Montréal, Montréal, Québec, Canada
    Search for more papers by this author
  • Nigel T. Roulet

    1. Department of Geography, McGill University, Montréal, Québec, Canada
    Search for more papers by this author


While peatland ecosystems overall are long-term net carbon (C) sinks, the open water pools that are characteristic of boreal peatlands have been found to be C sources to the atmosphere. However, the contribution of these pools to the ecosystem level C budget is often ignored even if they cover a significant area of the peatland surface. Here we examine the annual CO2 and CH4 ecosystem-atmosphere exchange, including the release following ice melt, from pools in a boreal maritime peatland, in order to estimate the annual loss of C from these water bodies. Over a 16 month period, dissolved CO2 and CH4 were measured periodically in five pools while continuous measurements of CO2 were made in one pool using a nondispersive infrared (NDIR) sensor. Fluxes were calculated using the thin boundary layer model and the eddy covariance technique (spring release only). We calculated an annual C release from pools of 103.3 g C m−2 yr−1 of which 15% was released during the spring ice melt. This release is the same order of magnitude, but with the opposite sign, as the average net ecosystem carbon balance for pool-free northern peatlands (−22 to −70 g C m−2 yr−1). We discuss the origin of the released C, as the magnitude of the release could have a significant impact on the contemporary C exchange of boreal peatlands.

1 Introduction

Peatlands are ecosystems characterized by a water table that is close to the surface and biomass productivity that exceeds decomposition, leading to large accumulation of partially decomposed organic matter define as peat. The long-term apparent carbon (C) accumulation rates (LORCA) in peatlands since the end of the last glaciation vary between 3.4 and 70.6 g C m−2 yr−1 [Turunen et al., 2002; Gorham et al., 2003; Yu et al., 2009]. Peatlands absorb carbon dioxide (CO2) through photosynthesis via the surface vegetation and release CO2 through autotrophic (growth and maintenance) and heterotrophic respiration (i.e., peat decomposition). Simultaneously, methane (CH4) is produced by anaerobic peat decomposition through acetate fermentation and/or CO2 reduction and is released to the atmosphere through a combination of diffusion, plant mediated transport, and ebullition. Although peatland ecosystems are generally a net sink for CO2 and a net source of CH4, the actual rates and direction of gas exchange can vary significantly among the different microforms (hummock, lawns, hollows, pools) as a function of the environmental conditions within a single peatland [e.g., Waddington and Roulet, 1996; Pelletier et al., 2007; Pelletier et al., 2011]. Studies looking at CO2 and CH4 exchange at the surface of open water pools have shown that these water bodies release C to the atmosphere [e.g., Hamilton et al., 1994; Repo et al., 2007]. In contrast, Macrae et al. [2004] found peatland open water pools in the subarctic regions to be net sinks for C and estimated LORCA between 7 and 26 g C yr−1. Macrae et al. [2004] attributed the C accumulation at the bottom to peat erosion resulting from wave action on the pool border. This situation has been highlighted in lake or reservoir studies where the systems are C sinks in sediments but also act as sources to the atmosphere [e.g., Prairie, 2008; Tranvik et al., 2009]

Peatlands' vegetated surfaces CO2 and CH4 fluxes have received more attention than their aquatic components. Measured fluxes from vegetated surfaces have been used to examine seasonal and interannual variability at the ecosystem level using eddy covariance (EC) towers and at the microform scale using manual and automatic chambers with a sporadic or continuous measurement approach [e.g., Lafleur et al., 2003; Forbrich et al., 2011; Pelletier et al., 2011; Lai et al., 2012]. On the other hand, pool fluxes have been measured using static chambers [e.g., McEnroe et al., 2009] or the thin boundary layer method [e.g., Hamilton et al., 1994]. However, recent studies have shown that floating chamber measurements tend to overestimate fluxes between 2 to 10 times especially on small water bodies [e.g., Lambert and Fréchette, 2005; Vachon et al., 2010]. The larger fluxes obtained from chamber measurements result from turbulence produced by the chamber that disturbs the air-water interface and therefore the gas exchange. In contrast, the thin boundary layer does not create disturbances at the air-water interface, as the flux is calculated using the gas concentration gradient between water and air, and the gas transfer coefficient. When combined with automated measurements of gas concentrations and other required environmental variables, this method allows high-frequency continuous flux estimation. However, the gas transfer coefficient for wind speed below 3 m s−1 remains difficult to establish [Cole and Caraco, 1998]

Most open water season average CO2 and CH4 releases from peatland open water pools are based on weekly measurements and therefore do not capture daily variation. Automated high-frequency dissolved CO2 measurements in peatland streams and beaver ponds have shown important diurnal and seasonal patterns in CO2 concentration [Dinsmore and Billett, 2008; Dinsmore et al., 2009]. Furthermore, a number of studies have shown a strong correlation between DOC concentration and dissolved CO2 in lakes; variations in DOC explaining the variability in dissolved CO2 between lakes [e.g., Hope et al., 1996; Jonsson et al., 2003; Roehm et al., 2009]. However, this type analysis has not been done for peatland open water pools.

GHG fluxes from peatland pools measured sporadically during the ice-free period range from 0.14 to 16.6 g CO2–C m−2 d−1 and −0.001 to 1.87 g CH4–C m−2 d−1 [Hamilton et al., 1994; Waddington and Roulet, 2000; Pelletier et al., 2007; Repo et al., 2007; McEnroe et al., 2009; Cliche-Trudeau et al., 2013]. Published total C release from pools during the ice-free season are limited and do not cover the entire open water season (generally late spring to early fall) with values ranging from 24 to 419 g C m−2 [Waddington and Roulet, 2000; Repo et al., 2007]. For example, the estimate from Repo et al. [2007] is based on 100 to 120 days while Waddington and Roulet [2000] covered a period from May to September. The open water season in boreal Canada can last more than 200 days (this study), and the resulting annual release of GHG's from peatland pools could therefore be significantly greater than has been reported to date. Furthermore, none of these studies account for the CO2 and CH4 that is released during the spring ice melt. It is hypothesized that the ice melt period plays an important role in the annual CO2 and CH4 emissions due to the buildup of gases linked to microbial activity in the sediments and pool water during winter. In a Finnish lake, Huttunen et al. [2001] estimated potential springtime emissions following ice melt to be between 31 and 51 g CO2–C m−2 and 2.3 to 17 g CH4–C m−2 for CH4. The contribution of gas released during the ice melt period in lakes has been estimated to represent 30% of the annual release for CO2 and between 40 and 60% for CH4 [Michmerhuizen et al., 1996; Karlsson et al., 2010].

In the present study, we use several techniques (headspace [CO2–CH4], submerged nondispersive infrared [NDIR] sensors and eddy covariance) to evaluate the annual dissolved CO2 and CH4 dynamics and to estimate the annual release of CO2 and CH4 from boreal peatland pools. The aims of this study are (1) to document the spatial and temporal variability in pool dissolved CO2 and CH4, (2) to estimate the release of GHG (CO2 and CH4) to the atmosphere during the spring ice melt, and (3) to establish an estimate of the annual release from these pools. We hypothesize that the net C exchange from peatland pools is on the same order of magnitude, but with the opposite sign, as the net exchanges reported in the literature for vegetated surface of peatlands. Furthermore, we hypothesize that the annual accumulated C exchange from pools is sufficiently large that if they are not accounted for, the net ecosystem exchanges for peatlands containing pools could be seriously in error.

2 Methods

2.1 Study Area, Peatland and Pool Descriptions, and Climate

The studied peatland is located near Baie Comeau on the North Shore region of Quebec, Canada. The peatland (49°08′N, 68°17′W; altitude: 19 m) sits in the center of the Manicouagan peninsula on deltaic sands left by the Outarde and Manicougan Rivers. The peatland covers ~600 ha and is classified as a raised bog [Glaser and Janssens,], dominated by Sphagnum fuscum, ericaceaous shrubs (mainly Chamaedaphnee calyculata, Rhododendron groenlandicum) and dwarf Picea mariana. Basal dates indicate that peat started to accumulate 4100 years B.P. (G. Magnan and Garneau M, A regional comparison of the long-term carbon dynamics within maritime ombrotrophic peatlands along the north shore of the Estuary and Gulf of St. Lawrence, northeastern Canada, submitted to The Holocene, 2013). The open water pools cover approximately 5% of the peatland surface and range in size from 2 to 9000 m2. The pool depths in the studied sector vary between 0.4 and 2 m which is similar to the range of pool depth observed in other regions of the world (Sweden: Foster and Fritz [1987] and Foster and Wright [1990]; Hudson Bay lowlands, Canada: Hamilton et al. [1994]; West Siberian Lowlands, Russia: Repo et al. [2007]; and Estonia: Karofeld and Tõnisson [2014]). The five pools selected are hydrologically independent from each other and are fed from precipitation; there are no streams connected to the pools. The studied sector is located more than 500 m from the closest peatland margin and approximately 5 m above the surrounding nonpeatland area. Peat cores (pools #1 and #5) showed no sediment accumulation at the bottom (A. Thibault, unpublished data, 2013). The 30 year climate normal (1971–2000) for mean daily temperature during the coldest and warmest months is −14.6°C in January and 15.6°C in July. The area receives on average 1014.4 mm in precipitation per year of which 361.5 mm is snow [Environment Canada, data available at].

2.2 Dissolved CO2 and CH4 Measurements

Five pools were chosen for dissolved gas measurements as representative of the range of pool sizes and depths found in the peatland (Table 1). Manual measurements of dissolved CO2 and CH4 in the five pools were made using the headspace technique [e.g., Kling et al., 1991; Dinsmore et al., 2009]. A sample of pool water (160 mL) collected in a Wheaton glass bottle was equilibrated with an ambient air sample (60 mL) by shaking under water for 1 min. A 20 mL sample of the headspace was then extracted and transferred to an evacuated Labco exetainer (Labco, High Wycombe, UK) for storage. Laboratory testing showed that the exetainers maintained an average vacuum level of 94.3 % (SD ± 1.3) after 18 days [Wittebol, 2009]. Analysis was done within 2 weeks after sampling on a Shimadzu GC-14B gas chromatograph equipped with a flame ionization detector (FID) for CH4 and a thermal conductivity detector (TCD) for CO2. Samples were hand-injected (5 mL) in a 1 mL injection. Column and detector temperature were 100°C and 315°C, respectively. Ultrahigh purity N2 was used as the carrier gas at a flow rate of 25 mL min−1. Pool water dissolved gas concentrations were calculated using Henry's law for the headspace and ambient concentrations. Measurements were made between June 2011 and October 2012 each time the site was visited during the open water and frozen seasons. These discrete measurements were performed to account for spatial and low-resolution temporal (1 week approximately every 50 days) variability in dissolved CO2 and CH4 concentration during the study period. The ice cover reached the bottom of the shallowest pools and technical problems such as ice forming in the peristaltic pump used for sampling resulted in only two to three pools sampled per site visit during the ice-covered season.

Table 1. Pool Physical and Chemical Characteristics, and Mean, Maximum, and Minimum Dissolved CO2 and CH4 During the Open Water Seasons as Obtained From the Headspace Measurement Techniquea
 CO2–C (mg L−1)CH4–C (µg L−1) 
Pool #Size (m2)Depth (m)MeanMinMaxMeanMinMaxDOC (mg L−1)
  1. a

    Standard deviation given in parentheses.

  2. b

    Mean CO2 and CH4 values are significantly different if they have no letters in common (CO2 and CH4—Kruskal-Wallis test followed by posthoc Steel-Dwass. DOC—One-Way ANOVA followed by Tukey-Kramer).

14992.00.55 (±0.18) (b)b0.280.7819.9 (±14.5) (b)3.2557.217.6 (±2.3) (b)
21280.70.99 (±0.50) (a)0.422.6915.2 (±16.9) (b)0.3058.617.5 (±3.3) (b)
37660.40.67 (±0.31) (b)0.261.6681.0 (±75.3) (a)0.2632721.4 (±2.8) (a)
425631.90.52 (±0.14) (b)0.270.7314.7(±31.7) (bc)1.5318316.7 (±2.9) (b)
518661.90.50 (±0.13) (b)0.250.796.3 (±3.4) (c)1.6815.817.8 (±4.0) (ab)

Automated continuous measurements of dissolved CO2 were made in pool #1 using NDIR CO2 sensors (CARBOCAP transmitter series Model GMT220, Vaisala, Finland) enclosed in a water tight, gas permeable polytetrafluoroethylene (PTFE) membrane [see Johnson et al., 2010] and connected to a datalogger (CR1000, Campbell Scientific, Edmonton, Canada). During summer 2011, two sensors were installed in pool #1: one at the surface and the other at 1 m depth. The sensor at the surface was removed before ice formation to prevent damage to it. The sensor was not reinstalled at the surface in 2012 as it was used to replace the sensor at 1 m that stopped functioning over the cold season. Automated measurements were made from August to mid-December 2011 and from May to August 2012. The measurements from the NDIR sensor were corrected for temperature and pressure [Tang et al., 2003]. The reference temperature and pressure for the Vaisala sensor were 22.5°C ±1% and 100.7kPa ±1%, respectively. Water and sediment (5 cm depth) temperature were measured every 30 min in pool #1 and pool #3 (2012 only) using thermocouples connected to a datalogger (CR1000, Campbell Scientific, Edmonton, Canada). In 2011, the water temperatures in pool #1 were measured at the surface, 100 cm and 200 cm. In 2012, measurements were made every 25 cm from the bottom of pools #1 and #3.

2.3 Flux Calculation

The diffusive fluxes of CO2 and CH4 from the pools were calculated using the thin boundary layer technique:

display math

where Cw is the concentration of gas in the water, Ca is the concentration of gas in the air Kx is a gas-specific exchange coefficient found from an expression of the gas-specific Schmid number (Sc), and

display math

where b = 0.66 for wind speed ≤ 3 m/s or b = 0.5 for wind speed > 3 m/s.

The Schmid number was found from Wanninkhof [1992] as a function of water temperature (T in °C) using

display math


display math

K600 is the gas exchange coefficient (cm h−1) normalized for CO2 at 20°C in fresh water with a Schmidt number of 600 and was approximated following Cole and Caraco [1998] as a function of wind speed at 10 m height (U10) as

display math

Wind speeds at 10 m height are not typically available and therefore these were approximated from measurements of wind speed on the meteorological tower at a height of 2 m above surface. The U10 was estimated using a logarithmic wind profile with a roughness length of 0.001 m and assuming a neutral boundary layer [Oke, 1987].

Fluxes were calculated from the continuous dissolved CO2 data from the NDIR sensor in (pool #1) and from the sporadic headspace measurements using continuous wind speed from the meteorological tower (see section 2.6) and water temperature measurements for other pools. Water temperature measurements for the headspace flux calculation were made at the time of sampling using a portable meter (model 3500, Kestrel, Birmingham, MI). Continuous open water season fluxes in pool #1 (2011 and 2012) and pool #3 (2012) were also estimated by linearly interpolating the instantaneous dissolved CO2 and CH4 from headspace measurements. Flux estimates were not made for pools #2, #4, and #5 as no continuous temperature measurements were made in these pools.

2.4 DOC Measurements

DOC measurements were performed once per field campaign in each of the five pools between May and October 2012. Water samples were filtered through 0.45 µm glass fiber filters, acidified, and kept in a dark refrigerator at 5°C. DOC analysis was performed on a Shimadzu TOC-V total organic carbon analyzer calibrated using standards of 10, 50, and 100 mg L−1, prepared from a 1000 mg L−1 stock solution made from Potassium Bipthalate. The composition of the DOC was evaluated by looking at UV absorption at 250, 254, and 365 nm using a spectrophotometer (Spectronic Genesys 10, Thermo Scientific, Waltham, MA) with distilled water as a blank. Disposable 10 mm path length cells were used for the analysis.

2.5 Eddy Covariance Measurements

Ecosystem level NEE-CO2 was measured using the eddy covariance (EC) technique in the spring of 2012 in order to measure the release of CO2 from pools during ice melt. The tower was located in the same sector as the pool dissolved CO2 and CH4 measurements were performed and its source area included a varying percentage of pool surfaces depending on the wind direction (Figure 1). The EC system consisted of a fast response three-dimensional sonic anemometer (CSAT-3, Campbell Scientific, Edmonton, Canada), a fine-wire thermocouple (FW05, Campbell Scientific, Edmonton, Canada), and an open path CO2/H2O analyzer (LI-7500, LI-COR, Lincoln, NE). The instruments were mounted on a tripod 2.5 m above the surface of the peatland. The variables used to calculate the flux using the eddy covariance method were recorded and stored on a 2 GB compact flash card using a data logger (CR3000, Campbell Scientific, Edmonton, Canada) at 5 Hz. The 30 min CO2 fluxes were computed from the 5 Hz data on a personal computer using an in-house script developed in Matlab (v.9.0, Mathworks, Natick, MA). CO2 fluxes were derived from the covariance between vertical wind speed and CO2 mixing ratio, corrected for the effect of temperature and water vapor fluctuation on air density [Webb et al., 1980]. A two-dimensional coordinate rotation was applied. No additional frequency response or energy budget corrections were applied to these fluxes. Details on the quality control procedure used for NEE-CO2 are presented in Bergeron and Strachan [2011]. Because of the sporadic nature of the spring release, no gap filling was done for this measurement period. The flux source area during ice melt was estimated using the Flux Source Area Model (FSAM) [Schmid, 1994].

Figure 1.

Location of the study area with pools and eddy covariance tower. The ellipse delineates the tower source area during the spring melt.

2.6 Supporting Measurements

Environmental measurements were made every 5 s throughout the study period (June 2011 to October 2012) and averaged over 30 min. The variables measured include components of the net radiation (CNR4, Kipp and Zonen, Delft, Netherlands), incoming and reflected photosynthetic active radiation (LI-190SA, LI-COR, Lincoln, NE), air temperature and relative humidity (HMP-45C, Vaisala, Helsinki, Finland), wind speed and direction (05103-10, RM-Young, Traverse City, MI), and precipitation (TE525M tipping bucket gauge, Texas Electronics, Dallas, TX). Daily time lapse photographs of pool #1 were taken using a game camera (Game Spy I-65 S, Moultrie, Alabaster, Al) to be able to determine ice cover extent and duration.

3 Results

3.1 Headspace Dissolved CO2 and CH4

Between June 2011 and October 2012, 219 discrete surface concentration measurements were made at the five selected pools in the peatland. The five pools sampled were constantly supersaturated in CO2 and CH4 relative to the atmosphere, and the measured concentrations varied significantly among some of the pools during the open water season (Table 1). The smallest (pool #2) pool had the highest average dissolved CO2, while the average dissolved CO2 from the other pools were not statistically different from each other (Table 1). Average dissolved CH4 was slightly more variable among the pools and the highest open water average dissolved CH4 value was observed in pool #3, which was the shallowest of the five pools sampled (Table 1).

Dissolved CO2 and CH4 in individual pools varied temporally over the study period (Figure 2). The CO2 and CH4 concentrations increased in all pools between June and August in 2011: increasing rates ranging from 3 and 9 µg L−1 d−1 for CO2 and from 0.05 to 1.3 µg L−1 d−1 for CH4 (Figure 2). The largest increases were observed in pool #2 for CO2 and in pool #3 for CH4 (Figure 2). This period was followed by a decrease in dissolved CO2 in all five pools in early November while CH4 concentrations continued to increase in three of the five pools (Figure 2). Although pool #3 saw its concentration in CH4 decrease in early November compared to values measured in August, values were still higher than the other four pools.

Figure 2.

(top) Dissolved CO2 and (bottom) CH4 in the five pools in the peatland. The frozen pool surface period is shown as the shaded area in each graph. Symbols represent the individual pool average headspace CO2 or CH4 for each field campaign.

Permanent ice cover formed on 22 November in 2011 (DOY 326) and lasted until 28 April 2012 (DOY 119) (156 days) as observed from daily time lapse photographs. Pool sampling was limited during the frozen water season because of the technically challenging conditions related to water sampling in subfreezing conditions. Pools #4 and #5 were sampled in January, and pools #1 and #4 were sampled in March 2012. Dissolved CO2 and CH4 increased considerably after the ice established at the surface of the pools (Figure 2). Maximum dissolved CO2 and CH4 concentrations under the ice were measured in March 2012 as 8.1 mg L−1 for CO2 in pool #4 and 472 µg L−1 for CH4 in pool #1.

The first 2012 open water measurements were made on DOY 138, approximately 3 weeks after ice melt. Dissolved CO2 values generally increase from spring throughout the open water season, but the pattern was not as clear as in the previous open water season and concentrations did not decrease in the fall as they did in 2011. However, the open water season increasing pattern in dissolved CH4 was more universal in 2012 as values increased in all pools from May to October with the average rates ranging between 0.15 and 1.94 µg L−1 d−1 (excluding pool #5 which was not sampled in October 2012).

3.2 NDIR Sensor Dissolved CO2

The dissolved CO2 30 min data at 1 m in pool #1 varied significantly over the study period with a range of CO2 concentration between 0.3 and 3.25 mg L−1 (including the frozen water season). The August 2011 dissolved CO2 values varied around 0.7 mg L−1 before increasing rapidly to systematically vary around 1.0 mg L−1 in early September. The dissolved CO2 decreased in mid-September before rising again to reach the maximum open water value of 1.7 mg L−1 on DOY 267. Following this peak value, the dissolved CO2 values decreased to 0.5 mg L−1 on DOY 279 after which it fluctuated between 0.6 and 0.8 mg L−1 until DOY 327 when ice started to form at the surface of the pool. From there, dissolved CO2 increased significantly at variable rates, with the largest rate observed between DOY 336 and 344 at 0.36 mg L−1 d−1. Maximum CO2 concentration reached 3.25 mg L−1 before the sensor failed on DOY 346. NDIR measurements resumed on DOY 141 in 2012.

Open water season 2012 NDIR dissolved CO2 increased from May to mid-August with values reaching 1.35 mg L−1 on DOY 228. However, the trend was not consistent as the longer-term increasing trend was interrupted by periodic drops in CO2 concentration. These drops correspond with declines in water temperature over the entire pool profile consistent with a decrease in air temperature and potentially strong winds that allowed pool mixing (Figure 3). NDIR dissolved CO2 values started to decrease in mid-August to reach a minimum of 0.68 mg L−1 on DOY 259. The decrease was similar to what was observed in 2011 but occurred approximately a month earlier. Dissolved CO2 remained around 0.9 mg L−1 until DOY 284, 2012 (Figure 2). While the overall trend in NDIR sensor dissolved CO2 was similar to that observed from the headspace measurements at the surface of the pools, the NDIR CO2 values measured at 1 m were greater in August 2012. The CO2 concentrations measured by the NDIR sensor at 1 m were strongly positively correlated with water temperature. The strongest relationships were found when the data were split into large periods corresponding to the open water season patterns such as the long-term CO2 concentration decrease from August to November 2011 and a period of increase from May to August 2012. Strong and significant relationships were found for both 30 min and daily averaged data (Table 2). The parameters of the relationships are statistically different between the two periods and from the overall relationship (ANCOVA, p < 0.01).

Figure 3.

Daily average dissolved CO2, sediment, and water temperature from pool #1, along with daily average wind speed, air temperature, and total daily precipitation during the 2012 open water season. Note that water temperature depths are relative to the bottom of the pool.

Table 2. Best Exponential Relationship Parameters Between Pool #1 Temperature at 1 m and Dissolved CO2 From NDIR Sensor
Daily Average Dissolved CO2
2011242–325LogCO2 = 0.013 (±0.001) × Temp. − 0.24 (±0.02)0.53<0.00183
2012141–221LogCO2 = 0.036 (±0.003) × Temp. − 0.94 (±0.05)0.72<0.00181
All data-LogCO2 = 0.002 (±0.001) × Temp. − 0.16 (±0.02)0.01<0.149249
30 min Dissolved CO2
2011242–325LogCO2 = 0.013 (±0.0001) × Temp. − 0.24 (±0.02)0.48<0.0013984
2012141–222LogCO2 = 0.034 (±0.0001) × Temp. − 0.89 (±0.01)0.61<0.0013921
All data-LogCO2 = 0.001 (±0.0001) × Temp. − 0.16 (±0.002)0.01<0.00112015

There was a strong agreement between the discrete surface concentration measurements and NDIR dissolved CO2 in 2011 (Surface NDIRCO2 = 1.00(±0.06) × headspaceCO2 + 0.04(±0.04), r2 = 0.99, p < 0.05). Surface NDIR dissolved CO2 in pool #1 measured from August to October 2011 followed a similar pattern as the CO2 measured at 1 m and also the discrete surface measurements made at the surface for the same period (Figures 2 and 4). The NDIR CO2 concentrations from the two depths (30 min data) were strongly correlated, with surface values being generally higher than those measured at 1 m (Surface CO2 = 0.78(±0.01)*1 m CO2 + 0.17(±0.01), r2 = 0.66, p < 0.05, n = 4254). For daily average CO2, the difference between the two depths was smaller (Surface CO2 = 0.80(±0.05)*1 m CO2 + 0.17(±0.04), r2 = 0.76, p < 0.05, n = 88). The relationship with daily averages attenuates the differences observed in the diurnal dissolved CO2 pattern (30 min data) between the two depths, the diurnal variation being more pronounced at the surface than at 1 m especially in August and September (Figure 4). We found a significant positive relationship between the cumulative daily photosynthetic active radiation (PAR) and the amplitude of the diurnal pattern in dissolved CO2 measured at the surface of pool #1 with the NDIR sensor (r2 = 0.41, p < 0.05).

Figure 4.

Diurnal pattern in dissolved CO2 in pool #1, measured by the NDIR sensors at the surface (solid symbols) and 1 m depth (open symbols) in August, September, October, and November (1 m only) 2011.

3.3 Dissolved Organic Carbon

There was no clear temporal pattern in DOC concentration in the pools during the 2012 open water season (Figure 5). Concentrations of DOC ranged between 8 and 26 mg L−1 over the study period. Average DOC concentration was higher in pool #3, although not different from pool #5, and there was no significant difference between open water pools #1, #2, #4, and #5 (Table 1). The relationships between DOC and CO2 concentration during the 2012 open water period were significant (p < 0.05) only in pool #2 where dissolved CO2 decreased with increasing DOC concentration (r2 = 0.36, p = 0.04) and in pool #5 where it increased with increasing DOC concentration (r2 = 0.48, p = 0.04). The relationships between SUVA254 and CO2 concentration for individual pools show significant positive relationships (r2 > 0.6, p < 0.05) in all the pools except pool #5 (Figure 5).

Figure 5.

(top) Average DOC concentration from the five pools in 2012, (middle) relationship between headspace measurements dissolved CO2 and SUVA254, and (bottom) relationship between dissolved CO2 and dissolved organic carbon concentration.

3.4 CO2 and CH4 Evasion

The pools were constantly supersaturated in CO2 and CH4 over the study period and were therefore constant sources of CO2 and CH4 to the atmosphere. The fluxes calculated from headspace measurements during the open water season ranged from 0.04 to 2.09 g CO2–C m−2 d−1 and 0.0001 to 0.16 mg CH4–C m−2 d−1 (Table 3 and Figure 6). As would be expected given how the fluxes are calculated, the patterns among pools were similar to the patterns observed for dissolved CO2 and CH4; pool #2 had the highest average CO2 flux, while pool #3 had the highest average CH4 flux (Table 3). Interpolated fluxes from pools #1 and #3 were highly variable for both CO2 and CH4 as a result of changes in dissolved gasses concentration over time and wind speed variability (Table 3 and Figure 6). The range of fluxes calculated from interpolated data was therefore greater than for the individual flux calculated with headspace data measurements. However, the average fluxes for the two methods were similar (Table 3). The interpolated fluxes from the two pools were different, and pool #3 had higher variability in evasion rates for both CO2 and CH4. The CH4 release in both pools increased from May to October 2012, but the increase was more pronounced in pool #3 (Figure 7).

Table 3. Mean, Minimum, and Maximum Fluxes From Headspace Measurements, Interpolation, and NDIR Sensora
 CO2–C (g m−2 d−1)CH4–C (g m−2 d−1)
Pool #MeanMinMaxMeanMinMaxMeanMinMaxMeanMinMaxMeanMinMax
  1. a

    Standard deviation given in parentheses.

10.36 (±0.1)0.160.570.33 (±0.2)0.091.900.45 (±0.3)0.062.480.021 (±0.013)0.0050.0510.020 (±0.013)0.0050.110
20.84 (±0.4)0.362.09-  ---0.014 (±0.012)0.0000.051---
30.47 (±0.2)0.160.920.40 (±0.2)0.081.76---0.078 (±0.042)0.0020.1570.090 (±0.085)0.0080.641
40.35 (±0.2)0.120.75------0.016 (±0.030)0.0020.120---
50.34 (±0.2)0.040.56------0.004 (±0.003)0.0030.016---
Figure 6.

(top) Daily average CO2 evasion from pool #1 calculated with the NDIR sensor CO2 measurements, (middle) manual headspace measurements (symbols) and headspace measurement interpolation (line), and (bottom) pool #3 manual headspace measurements (symbols) and headspace measurement interpolation (line). The period of frozen pool surface is shown as shaded areas in each graph.

Figure 7.

(top) Daily average CH4 evasion from pool #1 manual headspace measurements (symbols) and headspace measurement interpolation (line) and (bottom) pool #3 manual headspace measurements (symbols) and headspace measurement interpolation (line). The period of frozen pool surface is shown as the shaded area in each graph.

The 30 min CO2–C flux calculated from the NDIR sensor in pool #1 ranged from 0.06 to 2.48 g m−2 d−1 with a slightly higher average than the interpolated data of 0.45 (±0.3) g m−2 d−1 (Table 3 and Figure 6). Fluxes followed a similar pattern to the CO2 concentrations, with fluxes increasing from early August 2011 to reach a maximum release in September before decreasing until ice cover formed. The first fluxes calculated in May 2012 were low and similar to the last fluxes in November 2011. Similar to 2011, the 2012 open water season fluxes increased to reach their maximum in August before decreasing until the end of the experiment. When comparing with the interpolated fluxes, a similar pattern to the NDIR sensor for 2011 is seen as fluxes decreased in both case (Figure 6). The patterns were slightly different in 2012; the interpolated data fluxes did not increase as much as the NDIR fluxes (Figure 6). The maximum flux from the interpolated data was approximately 1.9 g CO2 m−2 d−1 compared to 2.48 g CO2 m−2 d−1 for NDIR sensor.

3.5 CO2 and CH4 Release Upon Ice Melt

3.5.1 Discrete Measurements

The release of CO2 and CH4 following ice melt was estimated using the headspace measurements made in March 2012 in pools #1 and #4 assuming that (a) concentrations measured manually in March represented the CO2 and CH4 concentrations just before ice melt, (b) CO2 and CH4 are well mixed across the pool profile, and (3) all excess CO2 and CH4 (relative to the atmosphere) would get released upon ice melt. Based on these assumptions, we estimated that pools #1 and #4 released, respectively, 15.1 and 15.9 g CO2–C m−2, and 0.9 and 0.8 g CH4–C m−2 following ice melt at the end of April 2012.

3.5.2 Eddy Covariance Tower Measurements

The NEE-CO2 during, and right after, ice melt shows large losses of CO2 (DOY 118 and 119 inclusively), with fluxes reaching 0.17 mg CO2–C m−2 s−1 (Figure 8). Wind direction during that period was 325° (SD ± 7) and the tower source area included pool #4. Fluxes were rejected for the first part of DOY 118 due to high Li-7500's automatic gain control (AGC) values linked to condensation or water droplets on the open path analyzer lens. The average release for the DOY 118-119 period was 0.1 mg CO2–C m−2 s−1 which corresponds to a release of 16.8 g CO2–C m−2.

3.6 Annual Release Estimation From Pool #1

The annual period, for the purposes of our study, was defined as 1 September 2011 to 31 August 2012. The annual CO2 release from pool #1 was calculated using the CO2 flux estimated from the NDIR sensor data (section 3.4) and the spring melt release from headspace measurement (section 3.5.1). Since the NDIR sensor was not functioning before 20 May 2012, the CO2 flux for that 3 week period was estimated using a conservative average value of 0.22 g CO2–C m−2 d−1 obtained by averaging fluxes for the periods including November 2011 and late May 2012. The total CO2 release from pool #1 was estimated at 98.7 g CO2–C m−2 yr−1 of which 79.2 g CO2–C m−2 came from the flux calculated using the NDIR sensor data and 15.1 g CO2–C from the ice melt release. The remaining 4.4 g CO2–C m−2 comes from the estimated missing flux data between 1 and 20 May 2012. If the annual release estimation had been made using the flux from the interpolation of headspace measurements, total CO2 loss from pool #1 would have been 77.3 g CO2–C m−2 yr−1. The annual CH4 release from pool #1 was estimated using the headspace interpolation combined with estimated spring evasion. The total annual CH4 release from pool #1 was 4.6 g CH4–C m−2 of which 0.9 g CH4–C m−2 was from the spring release. Combining CO2 and CH4 release from pool #1 resulted in an annual loss of 103.3 g C m−2.

4 Discussion

The results from this study show that the temporal variability of CO2 and CH4 concentrations in pools and the spatial variability among pools are important. The continuous supersaturation of CO2 and CH4 in the pools with respect to the atmosphere results in a net flux of gases towards the atmosphere. Previous studies have shown the importance of CO2 and CH4 release from pools by focusing mainly on the ice-free season primarily using sporadic measurement techniques such as the headspace or floating chamber techniques [e.g., Hamilton et al., 1994; Pelletier et al., 2007; McEnroe et al., 2009]. Thus far, annual C release from peatland pools had been estimated from a small number of discrete measurements, therefore missing the temporal variability and not accounting for spring C release. Furthermore, we examine the source of the released C and magnitude of the pools fluxes relative to the components of peatlands.

4.1 Pools Dissolved Gas Concentrations Controls and Origin of CO2

Pools sampled in this study were always supersaturated in CO2 and CH4 with respect to the atmosphere, with dissolved concentration ranges of 0.3 to 8.1 mg CO2–C L−1 and 0.4 to 472 µg CH4–C L−1. These concentrations are within the range of values found in the literature for similar water bodies of 0.3 to 16 mg CO2–C L−1 and 0.8 to 766 µg CH4–C L−1 [Hamilton et al., 1994; Riera et al., 1999; Repo et al., 2007]. Different processes drove the dissolved gas concentrations across the site and over time. The shallowest pool (#3) consistently had the highest dissolved CH4 of all five pools. This result may be explained by the higher measured sediment temperatures; although only measured in pools #1 and #3, sediment temperature was constantly higher in pool #3. In a laboratory peat column experiment, Moore and Dalva [1993] reported 6.6 times higher CH4 production when temperature increased from 10°C to 23°C. The average 2012 sediment temperature in pool #3 was 16.6°C compared to 14.6°C in pool #1. The higher temperatures in pool #3 are a direct result of pool depth (pool #3 was sampled at 25 cm and pool #1 at 200 cm deep). Temperature also seemed to play an important role in CO2 concentrations in pool #1, with specific relationships between CO2 and temperature corresponding to the fall water temperature cooling (2011) and summer warming (2012) periods (Table 2). The two relationships suggest that another factor influences the concentrations, as the rates of CO2 increase and decrease were different between the two periods. While higher temperature favors microbial degradation of organic matter and therefore higher CO2 concentrations, the slower decrease in concentration in the fall could potentially be explained by the pool mixing, which would bring CO2 from the bottom of the pool.

Pool #2 had the highest average CO2 values during the open water season. This pool was surrounded by relatively dense shrub vegetation compared to the other four pools sampled and the vegetation sheltered the pool surface from the wind, potentially reducing CO2 evasion and favoring CO2 accumulation in water [Kwan and Taylor, 1994]. The sheltering effect could also explain the high CO2 values measured in May, which could represent CO2 that accumulated under the ice layer and that was not released immediately following ice melt. Although no continuous temperature data is available for this pool, the shelter created by the vegetation could have also inhibited mixing and promoting warmer water. Pool size could also explain the higher CO2 measured in pool #2. In a study examining the relationships between peatland pool morphology and the CO2 fluxes, McEnroe et al. [2009] observed a negative relationship between pool surface area and the average CO2 fluxes. A similar relationship can be observed here (Tables 1 and 3). Roehm et al. [2009] observed a negative relationship between lake area and average CO2 concentration suggesting that the C loading from the surrounding watershed was an important driver for the dissolved CO2 dynamics in lakes. However, peatland open water pools are different from lakes as they are hydrologically isolated from the larger catchment they are in and receive their water from rainfall and lose water through evaporation. Although carbon loading has been shown to explain CO2 concentration variation in boreal lakes, the overall variability in headspace CO2 over time and space could not be explained by the DOC concentrations at our site. Within pools, our results showed negative and positive significant relationships between dissolved CO2 and DOC in pools #2 and #5, respectively. Note that no samples were collected in pool 5 during the last field campaign and the relationship is therefore based on a shorter period of time and smaller sample size. The relatively narrow range of DOC concentrations could explain the overall weak relationships with CO2 concentrations (Figure 5). Concurrently, the negative relationship observed in pool #2 seems to be driven by high CO2 concentrations measured early in the open water season that could be the result of a sheltering effect (see above). Because of the low pool water pH (<4.5), we can assume that CO2 represented the dominant inorganic carbon species.

As for the dissolved CO2-DOC relationship, the observed link between SUVA254 and CO2 concentration in the five pools does not provide clear evidence that the open water DOC is the main source for CO2. Specific UV absorbance (SUVA254) has been shown to correlate positively with DOC aromaticity and with lower microbial biodegradability of dissolved organic matter [Kalbitz et al., 2003; Weishaar et al., 2003]. Therefore, if DOC represented the main source of dissolved CO2 in our pools, we would expect CO2 concentrations to decrease with increasing SUVA254. Here however, the significant relationships observed in the four small pools are positive, with dissolved CO2 increasing with increasing SUVA254. However, in a lab study Olefeldt et al. [2013] found UV-mediated degradation of highly aromatic leachates to produce greater C loss than for lower aromacity compounds. This finding suggests that more recalcitrant dissolved forms of terrestrial organic matter reaching aquatic ecosystems could be rapidly degraded and released to the atmosphere. This could explain the positive relationships observed in four of the five pools in our study. Conversely, the significant positive relationship found between cumulative daily photosynthetic active radiation (PAR) and the amplitude of the diurnal pattern in dissolved CO2 at the surface in pool #1 in 2011 suggests that DOC still plays a role in CO2 variations, at least at a diurnal scale. The increase in CO2 concentration during the daytime at the surface of pool #1 could be the product of DOC photoxydation. This is also supported by the less pronounced diurnal variability at 1 m, which can be explained by the limited light penetration at depth. DOC photooxidation represents an important factor in lake CO2 supersaturation, and the reaction has been linked to energy absorption, which decreases with water depth [Granéli et al., 1996; Bertilsson and Tranvik, 2000].

The open water peatland pools studied here differ largely from lake ecosystems. Peatland pools are not connected to each other or to a stream, and their basins are composed almost uniquely of organic matter, providing an almost unlimited amount of substrate for decomposition. With pools being located on the highest portion of the ombrotrophic peatland, the size of individual pool catchments is relatively small and their water input comes exclusively from precipitation. A large DOC input into the pools would be required to sustain the CO2–C release measured in the present study. We estimated the size of the individual pool watershed using a topographic survey of the site done by Simard [1976] and estimated the May to October runoff to 149 mm based on precipitation (368 mm) and evaporation (219 mm) measured by the EC tower (unpublished data). Using a DOC concentration of 30 mg L−1, we found DOC import to the pools could range between 8 and 38 g C m−2 of pool area for the period between mid-May and mid-October 2012. It should be stressed that the pools are hydrological isolated for most of May to October, similarly to Quinton and Roulet [1998], and that using runoff to evaluate the potential water flow through the pool catchments is probably an over estimation. Nevertheless, we use this estimate in order to be conservative with our calculations. Based on the CO2 fluxes calculated from the daily average flux (NDIR in pool#1 and headspace measurement in other pools), we estimate that DOC could represent between 16 and 30% of the CO2 loss from our pools during the same period. This is consistent with the results presented by Hamilton et al. [1994] where limited groundwater movement was observed and the source of the C fluxes at the surface of the water was attributed to the decomposition of the peat at the bottom of the pools. Hamilton et al. [1994] also found the lowest CO2 and CH4 concentration in a pool that had a mineral bottom.

Figure 8.

NEE-CO2 eddy covariance measurements from 9 April to 9 May 2012. The period of frozen pool surface is shown as the shaded area.

Further analysis of our data showed that fluxes calculated using the NDIR-CO2 data from pool #1 (Figure 6) were also strongly correlated with the sediment temperature during the entire study period (Figure 9a). An Arrhenius plot (Figure 9b) of CO2 fluxes at the surface and sediment temperature of pool #1 suggests that the fluxes could be tightly related to CO2 production in the pool's sediment. It also indicates that the pool is close to steady state; i.e., the flux is approximately the same as production and is greater than the changes in CO2 storage in the peat at the bottom and pool water. This is circumstantial evidence that the origin of the emitted CO2 lies mainly in the peat at the bottom of the pool and not the open water DOC.

Figure 9.

(top) Relationship between sediment temperature and CO2 flux at the surface of pool#1 and (bottom) Arrhenius relationship for the same data set. The flux data were binned into temperature classes of a 1°C interval, and the mean and standard deviation of the fluxes were computed for each bin.

4.2 CO2 and CH4 Evasion From Pools

Fluxes presented in this study range between 0.04 and 2.5 g C m−2 d−1 for CO2 (headspace and NDIR) and 0.004 and 0.64 g C m−2 d−1 for CH4 which is well within the range of values presented in literature of 0.14 to 16.6 g C m−2 d−1 for CO2 and −0.001 to 1.87 g C m−2 d−1 for CH4 [Hamilton et al., 1994; Repo et al., 2007; Waddington and Roulet, 1996; Pelletier et al., 2007; Cliche-Trudeau et al., 2013]. The estimated spring melt evasion of 15 g C m−2 for pool #1 is, although smaller, consistent with values of 28 to 35 g C m−2 published for a small 11 m deep lake in Finland [Huttunen et al., 2004]. Our estimation also matches the flux measured by the eddy covariance tower during the spring melt (17 g C m−2). However, pool surfaces represented approximately 58% of the eddy covariance tower source area during the ice melt (Figure 1). Since the CO2 release from the terrestrial portion was limited due to the cold peat surface, the release measured by the eddy covariance tower is an underestimation of the actual contribution from the pool. At the time of spring melt, peatland vegetated surface soil temperature was below 2°C, limiting heterotrophic respiration, and the vegetation had not recovered from dormancy [e.g., Larcher, 2003]. To our knowledge, no other study has reported the spring melt release from peatland pools using eddy covariance. These results provide corroboration to the estimates from the discrete measurement techniques and have the potential to greatly improve annual estimates of release from aquatic ecosystems. Direct measurement of the CO2 release during spring melt using the eddy covariance has been made in a Finnish Lake with 30 min flux reaching maximums between 0.01 to 0.06 mg CO2–C m−2 s−1 depending on the year [Huotari et al., 2011], compared to a maximum 0.17 mg CO2–C m−2 s−1 presented here.

Estimates of the annual C release from peatland pools including spring release are nonexistent in the literature. The most relevant example where annual C release has been measured is Riera et al. [1999] who measured two bog lakes in Wisconsin, USA. In their study, Riera et al. [1999] estimated an annual C evasion for the two bog lakes of 84 and 124 g C m−2, including the spring release. Our annual estimate of 105 g m−2 yr−1 from pool #1 is well within this range. However, our annual release estimate remains conservative as no dissolved CO2 and CH4 value were measured just before the spring melt. This is also supported by the EC tower measurements during spring melt which included terrestrial area and, therefore, underestimated the flux from the pool. Riera et al. [1999] observed an increase in dissolved CO2 and CH4 up until ice melt. Furthermore, our results do not include the release of CH4 through ebullition. In two lakes adjacent to a subarctic peatland in Sweden, Wik et al. [2011] estimated the release of CH4 through ebullition as 0.07 ±0.3 mg m−2 d−1 during the frozen season and 0.02 ±0.07 mg m−2 d−1 during the open water season. Applied to our site, this would represent a release of 13.8 g CH4–C through ebullition, increasing the annual C loss to 117.1 g. This annual loss is larger than the range reported by Repo et al. [2007] of 22 and 66 g C (4 to 7 days of headspace and chamber measurements in July and August extrapolated to 100 to 120 days) and falls within that of Waddington and Roulet [2000] of 86 to 419 g C (weekly chamber measurements from May to September).

4.3 Effects on Contemporary C Budget Estimates

Peatland net ecosystem carbon balances (NECB) that include annual eddy covariance NEE-CO2, CH4 loss and DOC export are limited [Roulet et al., 2007; Nilsson et al., 2008; Dinsmore et al., 2010; Koehler et al., 2011; Olefeldt et al., 2012], Dinsmore et al. [2010] being the only one including CO2 and CH4 release from water bodies (streams). The published annual NECB values range between a net source of 14 g C m−2 to a net sink of 101 g C m−2 [Roulet et al., 2007; Nilsson et al., 2008; Dinsmore et al., 2010; Koehler et al., 2011]. However, the sites where these continuous measurements were made either have no pools, nonpermanent pools, or pools that are located outside of the eddy covariance tower source area that measured the NEE-CO2. A 6 year NEE-CO2 study has been published for a heterogeneous surface fen in Finland where the source area consisted of 60% flarks (shallow pools), 10% moist lawns, and 30% dry strings [Heikkinen et al., 2002; Aurela et al., 2004]. Although the flarks are inundated most of the time, emergent vegetation composed of sedges (Carex spp.) and mosses is present. Microform level measurements have shown the flarks to be sink for CO2 [Heikkinen et al., 2002] in contrast with the deep permanent pools. If a large part of the released C comes from the mineralization of the sediments, like our results suggest, peatlands with permanent pools could have a significantly different net ecosystem carbon balance considering the release of 103.3 g C m−2 yr−1 measured in the present study. Using an average NECB value of 30.9 g C m−2 yr−1, calculated from the published NECB, and assuming 50% of the release 103.3 g C m−2 yr−1 comes from the sediments, we estimate that peatlands that have pool covers greater than 37% could represent net sources of C to the atmosphere. In a Swedish peatland where pools covered less than 6% of the site, Waddington and Roulet [2000] found that pool release turned their site into a net C source, based on a open water season chamber-based peatland carbon exchange evaluation. The impacts of these findings are important as they imply that the current NECB of a significant portion of the boreal peatlands is negative, as they would be losing C to the atmosphere. However, peatland pool coverage data are currently lacking for most regions of the world. Peatland rich regions such as the Hudson Bay lowlands and Western Siberia lowlands exhibit large pool surface area that can represent more than 30% of the peatland surface in some area [Roulet et al., 1994]. Certainly this is an area of research that should be explored further.

5 Conclusion

This study highlighted the magnitude and spatial and temporal variability in CO2 and CH4 release from boreal peatland pools. Our measurements showed that pool GHG concentration varied on a diurnal and seasonal basis and with pool sizes and depths. All pools were constant sources of CO2 and CH4 to the atmosphere during the ice-free season, and the annual C release from pool #1 was 103.3 g m−2 of which spring melt release represented 15%. Overall, the different methods used in this study showed good agreement. The CO2 headspace measurements and derived fluxes for pool #1 displayed similar trends as the concentration and fluxes obtained with the NDIR sensor, although the later captured more variation over short time periods. Spring melt release estimated from headspace measurements was also similar to the flux measured using the eddy covariance tower, emphasizing the potential of eddy covariance flux measurements over aquatic ecosystems.

Although pools are a common feature in boreal peatlands, ecosystem level carbon exchange studies have yet to be made on this type of peatland. Our results suggest that 37% pool cover on a peatland could turn the ecosystem into a net C source. Therefore, we suggest that studies should evaluate the surface area of peatland pools in the Hudson Bay and James Bay Lowlands in Canada and in the Western Siberia Lowlands to get a true estimate of peatland net carbon balance.


The authors wish to acknowledge the financial support of the National Sciences and Engineering Research Council of Canada through an NSERC-CRD grant to MG and a Discovery grant to IBS. LP was funded through fellowships from NSERC (PGSD), the McConnell Foundation and the McGill Graduate Excellence program. The authors would like to thank Tim R. Moore (McGill University), Patrick Crill, Brent Thomson and Martin Wik (Stockholm University) and the two reviewers for useful discussion; Kerry Dinsmore and Mike Billett, (Center for Ecology and Hydrology, Penicuik, UK), and Fraser Leith, (University of Edinburgh) for advice on data treatment and analysis; Alain Tremblay and J-L. Fréchette, (Hydro Quebec); and the field and lab assistance of H. Asnong, A. Lamalice, V. Lefrancois, J. Minville (UQAM); M. Dalva, E. Christensen, S. Crombie, C. Lefrançois, C. Watt, R. Chen, F. Ferber (McGill).