2.1. Study Site
 Gas fluxes were studied intensively during the summer of 1998 and 1999, and occasionally during the late autumn and winter to permit calculations of gaseous carbon budgets for the whole year.
 The study site (69°49′N, 27°10′E, 295 m. a.s.l.) (Figure 1a) is in the western part of a large palsa mire, Vaisjeäggi. The climate of this northernmost part of Finland is subarctic. Weather data (from 1962) are available from Kevo weather station located 10 km southwest of the study site. The mean air temperature for 1990–2000 was −1.2°C, and annual mean precipitation 456 mm (40% as snow) (Table 1). The length of the average snow-free period is 155 days, and on average there are 146 days when the maximum temperature is below zero and 230 days when the minimum temperature is below zero [Finnish Meteorological Institute, 1991]. Vaisjeäggi, with its exposed landscape, is located 188 m higher than the Kevo weather station; therefore the air temperature is somewhat lower during summer compared to the Kevo weather station (Table 1).
Figure 1. Location of (a) research area and (b) the collars at experimental area at Vaisjeäggi; T1–T4, transects; WS, weather station; PS, peat surface level measurement pole. Only the collars without permanent shading are included to this study.
Download figure to PowerPoint
Table 1. Climatological Statistics From Kevo Weather Station
|Year||Precipitation,a mm (Min.–Max.)||Temperature,b °C (Min.–Max.)||Global radiation, MJ m−2 (Min.–Max.)|
|January||30 (12–51)||32||18||−13.1 (−20.5–−8.7)||−12.8 [n.d.]||−20.5 [−16.4]||2 (1–3)||2||2|
|May||25 (6–36)||25||9||2.9 (0.7–5.6)||3.0 [n.d.]||1.3 [1.2]||504 (438–580)||457||508|
|June||65 (7–115)||83||72||9.4 (6.2–12)||7.8 [5.6]||12.0 [11.3]||499 (399–589)||514||522|
|July||59 (27–129)||55||82||12.8 (10.7–14.2)||14.2 [12.3]||13.3 [11.3]||458 (376–531)||472||448|
|August||62 (28–161)||31||161||10.8 (8.5–13.2)||10.1 [8.8]||8.5 [7.3]||321 (244–405)||347||268|
|September||33 (5–66)||48||27||5.7 (2.5–7.9)||5.1 [3.5]||7.8 [7.0]||164 (110–200)||142||154|
|October||48 (9–97)||58||71||−1.0 (−8.4–3.8)||−0.2 [−1.5]||1.7 [0.6]||56 (43–69)||50||51|
|Average||456 (296–597)||489||597||−1.2 (−3.3–0.2)||−3.3 [n.d.]||−1.8 [n.d.]||2561 (2144–2798)||2577||2531|
 The site is on bedrock with a peat layer of 4.5 m at the center of the experimental site and 3.5 m at the palsa margin. The study site had different surfaces based on morphological, hydrological and vegetation characteristics. Four study transects were set up on the mire to encompass the various functional surfaces (Table 2). Transects T1 and T2 were located on the wet surfaces, transect T3 was at a wet palsa margin south of a steep degrading palsa wall, and transect T4 was on the palsa top (Figure 1b). The studied palsa was elevated over 1.5 m from the mire surface. The palsa margin was partially flooded during early summer. On the wet transects, Sphagnumlindbergii dominated in some plots, with a mixture of Sphagnumlindbergii and Sphagnumriparium in others. The most common vascular plants were Eriophorum angustifolium, Eriophorum russeolum, Vaccinium microcarpum and Carex limosa. At the palsa margin, Sphagnum riparium occupied the ground layer, and vascular plants included E. angustifolium and E. russeolum. On the palsa, the vegetation consisted of Vaccinium vitis-idaea, Betula nana, Empetrum nigrum, Rubus chamaemorus, Ledum palustre, Dicranum polysetum, Andromeda polifolia and lichens like Cladina rangiferina and Cladonia species. On the top of the palsa, there were spots without vegetation coverage and some cracks, as some of the palsa edges were breaking down.
Table 2. Water Content, Bulk Density, pH, and Contents of C, P and N (mg/g) in the Upper Soil Profiles of Various Transects at Vaisjeäggi mire in 1998
| ||Transect 1||Transect 2||Transect 3||Transect 4|
|0–10 cm||10–20 cm||20–30 cm||0–10 cm||10–20 cm||20–30 cm||0–10 cm||10–20 cm||20–30 cm||0–10 cm||10–20 cm|
|Peat water content, %||96.2||95.6||94.9||95.9||95.5||95.6||96.9||87.5||87.9||65.2||74.1|
|Bulk density (dry), g/L||27.8||34.2||39.4||29.8||35.1||39.2||17.4||101.5||95.9||100.9||150.7|
|Bulk density (wet), g/L||717.1||781.0||776.1||736.8||783.5||902.3||552.6||810.5||795.8||290.3||582.1|
|pH, in collars||3.9||n.d.||n.d.||4.0||n.d||n.d.||4.0||n.d.||n.d||4.0a||n.d.|
|C (dry weight), mg/g||419.0||429.0||432.0||427.2||438.7||437.9||465.1||453.3||469.6||n.d.||n.d.|
|P (dry weight), mg/g||0.38||0.43||0.54||0.47||0.33||0.26||0.93||0.67||0.75||n.d.||n.d.|
|N (dry weight), mg/g||6.83||9.37||12.01||7.20||8.09||7.24||8.15||6.65||11.45||n.d.||n.d.|
 There were some differences in the contents of total carbon, total nitrogen and total phosphorus between transects. Carbon and phosphorus contents were highest at the palsa margin (Table 2). The nitrogen content, peat pH (Table 2), and vegetation composition (see above) suggested that the palsa margin was minerotrophic mesotrophic and the other wet surfaces were ombrotrophic or minerotrophic oligotrophic [Eurola et al., 1995].
2.2. Weather and Soil Physical-Chemical Characteristics
 Local weather data were obtained from a Vaisala MAWS weather station in 1998 from July 9 to October 7 and in 1999 from July 2 to September 29. Photosynthetically active radiation (PAR) was measured with a Li-190SA Quantum sensor, air temperature at a height of 1.5 m with a QMH101 sensor and precipitation with a tip bucket rain gauge (QMR101). The radiation and temperature data were measured every 10 min, from which the mean hourly values with standard deviations were calculated. There were additional HOBO loggers storing air and peat temperatures during the winter. These data, in addition to the data from a nearby (2 km distance) weather station set up for the LAPP project [Lloyd et al., 2001], were used to supplement some data gaps in our weather station data.
 The changes in the vertical position of the peat surface were examined using wooden poles fastened to the mire bottom at transects T1, T2 and T3 [see Roulet, 1991]. Peat pH was measured in situ from the collars of transects T1, T2 and T3 with a portable pH meter WTW 302 equipped with a Hamilton Flustrode probe and a temperature probe. The results shown here were measured in 1998 on August 10 from a depth of 8 cm (Table 2). Samples to determine peat bulk density and nutrient concentrations were taken with a volumetric peat sampler (8 cm × 8 cm). These samples were dried for 48 hours at +60°C. Total contents of N and P were measured using a FIA technique after digestion with sulphuric acid, and the total C content was determined using a Leco carbon analyzer at the University of Lund. Peat properties and nutrient contents are shown in Table 2.
2.4. Measurement of CO2 and CH4 Fluxes
 Boardwalks were constructed to prevent any disturbance to peat gas storage during measurements or damage to vegetation when regularly visiting the site. The collars (56 × 56 cm) were permanently inserted into the peat to a depth of 15–30 cm in the second week of June, 1998. The upper part of the collars had a groove for the water seal needed for the chamber measurements. The upper parts of the collars were kept flush with the peat surface. Transects T1 and T2 both had 18 collars, with six collars at the palsa margin (T3) and on the palsa (T4) (Figure 1b). Adjacent to each of the collars were perforated tubes inserted into the peat to measure water table depth. Frost depth and peat temperatures were also measured regularly near the collars. Every second collar was covered permanently with tent reducing radiation, these manipulated collars are not included to this study.
 During intensive study periods, CH4 fluxes were measured weekly and CO2 fluxes once or twice each week. In 1998, the intensive measurements started in the third week of June and continued until October 10. In 1999, the first flux measurements were made in the second week of June and the intensive study period lasted until September 29. The winter fluxes were measured from April 12–16, 1999 and May 9–10, 2000.
 The transparent closed chamber was made of 1.6 mm polycarbonate (60 cm × 60 cm × 25 cm) and was used in the CO2 exchange measurements. The chamber had an automatically driven cooler to prevent temperature increase and a muffin fan ensured mixing of the air during the measurement [Alm et al., 1997]. The CO2 concentration in the chamber was measured with a portable infrared gas analyzer (Li-6200 Portable Photosynthesis System, LI-COR, inc., Lincoln, Nebraska, USA) during an interval of 90 to 120 s and was equipped with a pump circulating all sample air (1.3 L min−1) through a desiccant to the CO2 analyzer. The net CO2 exchange (NEE) was measured at ambient illumination and at an illumination that was 70% of the ambient (the chamber was shaded with a portable dome shaped tent, area 2.3 m−2 and height 1.0 m, made of hessian sackcloth). These measurements at reduced PAR completed the radiation data needed to establish the light response curves between photosynthesis and PAR [Bubier et al., 1998; Tuittila et al., 1999]. Immediately after the CO2 measurement, the transparent chamber was removed from the collar for 1 min, reinserted and darkened with an opaque cloth. The total release of CO2 (Rtot) was then measured with the same IR analyzer. The PAR-sensor (LI-190SA) and radiation shielded temperature sensor were inside the chamber and these data were logged along with time and CO2 concentration data. During the CO2 flux measurements, soil temperatures were measured near the collars with a multichannel temperature probe “Hessu Kevo,” or with a Fluke 52 thermometer at the surface and at depths of 5, 10, 15 and 20 cm. The depth of the frost layer was measured with a metal rod (diameter 5 mm). The water table was measured from the perforated pipes with a ruler. The snow depth was measured with a carbon fiber rod. All depth measurements were fixed to the peat surface.
 A dark, closed aluminum chamber (56 cm × 56 cm × 20 cm) was used for methane flux measurements. There was a battery-operated fan in the chamber and the chamber was equipped with a capillary tube to retain atmospheric pressure inside the chamber when sampling. During the 12-min measuring period, four samples were taken into gas tight plastic syringes [Nykänen et al., 1998] from the chambers at the wet transects. For the palsa, a measuring period of 20 min was used. Separate syringes were used for the palsa and wet transects because of the great difference in the methane fluxes. The syringes were left open for at least 24 hours before the next sampling. The CH4 concentrations were analyzed within 24 hours with an HP 5890 gas chromatograph with an FI-detector, HayeSep Q 1/8″ × 1.8 m column and a manual sampling valve with 0.5 mL loop. The peak areas were analyzed with an HP integrator 3369. The gas standard applied had 9.89 ppm ± 3% CH4 and 401 ppm ± 3% CO2 in synthetic air (Aga, Sweden). The CV of analyses of repeated runs of standards was less than 0.5%. Methane flux was calculated from the linear slope of the change in concentration over time, thus CH4 from ebullition is not included in the flux rates. The same environmental background variables for the CO2 fluxes were measured for CH4, except that the temperatures in the peat were measured at depths of 0, 2, 5, 10, 15, 20, 25, 30, 35, 40 and 50 cm. Winter fluxes of CH4 and CO2 were measured on April 16, 1999, and on May 10, 2000, using a snow profile method [Sommerfeld et al., 1993; Alm et al., 1999b]. In 1999, measurements were made from 48 snow profiles. On the palsa, chambers were also used during late winter 2000.
2.5. Modeling Gaseous Carbon Fluxes
 Statistical analyses were performed using SPSS for Windows [SPSS, Inc.]. Analyses of variance, correlation and regression analyses were used to study the relationships between the CH4 fluxes and the environmental parameters.
 The annual CO2 flows were reconstructed in a similar way as previously applied for carbon balance studies on peatlands using the chamber technique [Alm et al., 1997; Tuittila et al., 1999]. The net ecosystem CO2 exchange (NEE) in situ at prevailing radiation and at reduced radiation (shading) were calculated from the linear change in the CO2 concentration logged at 5-s intervals during the measuring period of 90–120 s. When plotting the changes in the CO2 concentration against time, the best fit period (the highest r2) of six measurements made during 30 s was used. For the respiration measurements, similar measuring periods were used as above.
 Negative NEE values were used when the CO2 fixation by the vegetation exceeded the total respiration of vegetation and soil. With this approach, respiration had positive values. Similarly, the CH4 emissions were positive, and uptake from atmosphere to soil had negative values.
 The gross photosynthesis (PG) was calculated using data obtained with the full (NEEf) and reduced radiation (NEEr). PG is the result of subtracting respiration (Rtot) from NEE, i.e. PG1 = NEEf − Rtot, and PG2 = NEEr − Rtot. Using this approach, two PG:s were obtained at every measuring event. In these two measurements, other environmental conditions except radiation were constant, which strengthened the modeling. The Rtot is the sum of CO2 produced by plant dark respiration, and by respiration of microbes and soil fauna.
 PG was assumed to be dependent on solar irradiation (I) in a functional form of a rectangular hyperbola with parameters Qmax (asymptotic maximum value of photosynthesis in optimal light) and half saturation parameter k (amount of radiation when photosynthesis is of maximum). The linear part of equation (1) includes air temperature (Tair) multiplied by a constant (t). CO2 release (Rtot) was related to air temperature with a logarithmic model. For CO2, modeling air temperature was selected because it had the best correlation with the CO2 fluxes (Rtot and PG). The response function for PG with solar irradiation (I) and air temperature (Tair) as independent variables is shown in equation 1, and Rtot with Tair as the independent variable is shown by equation (2).
 To account for the high-frequency variation in irradiation, equation (1) was calculated twice with two irradiation values (hourly averaged PAR ± 1 S.D). The mean of these calculations was the hourly PG [Smolander, 1984; Alm et al., 1999a].
 During the summer, the net CO2 exchange (NEE) was calculated for every hour using the formula
To avoid negative PG values in modeling, PG in equation (1) was marked as 0 when PAR was below 4 μmol m−2 s−1.
2.6. Reconstruction of the Annual Carbon Balance
 The annual CO2 balances were calculated using the sum of three periods. First, the spring fluxes of CO2 (Julian days 152–182) were modeled using combined CO2 data from years 1998 and 1999 at the transect levels. In the calculation of actual annual spring fluxes, weather data for that particular year were used. Since PG and Rtot measurements began in the second half of June, the fluxes for early June were approximated from the models made for late June. During Julian days 152–182, the increase in respiration and photosynthesis rates were assumed to increase linearly from the rates during the winter. During winter, there was no photosynthesis. The CO2 balance during summer was modeled separately for both years using data from Julian days 183–280 from individual collars. The CO2 emissions for the late autumn and winter (Julian days 281–151) were calculated from the flux measurements carried out with the snow gas gradient method on the transects (see above). The CH4 fluxes for Julian days 162–258 were calculated for every collar by summing the weekly mean fluxes multiplied by the hours of the week. Late autumn fluxes (Julian days 259–270) were extrapolated from measurements made at the transect level during both years, similarly the fluxes for Julian days 271–167 were extrapolated from the measurements made with the snow gas gradient technique at the transect level during winter.
 Standard error and 95% confidence limits were calculated according to Heikkinen et al. [2002b] for variation in the gaseous carbon balance (GCB) at various transects: GCB ± t × S.E (n = 3, t = 4.3; n = 9, t = 2.3). The standard error (S.E.) for the growing season flux estimate was calculated from SD = (var (GCB))1/2, where
Winter and spring fluxes were excluded from uncertainty analysis.