Journal of Geophysical Research: Biogeosciences

Annual CO2 exchange between a nutrient-poor, minerotrophic, boreal mire and the atmosphere

Authors


Abstract

[1] Mires are key landscape elements at high latitudes and have certainly accumulated carbon during the Holocene, but their current carbon balance at the present time is very unclear. The major carbon flux is the land-atmosphere CO2 exchange and full-year data are still limited. Here we present data from 3 a (2001–2003) of continuous Eddy Covariance measurements at Degerö Stormyr (64°11′N, 19°33′E) an oligotrophic, minerotrophic mire in Sweden. The climate at the site is defined as cold temperate humid, with 30-a annual precipitation and temperature means of 523 mm and +1.2°C, respectively, while the mean temperatures in July and January are +14.7°C and −12.4°C, respectively. The length of the vegetation period was 153 ± 15 d during the measured years. The minerotrophic mire represented a net sink for the vertical exchange of atmospheric CO2-C during the 3 a, with an average net uptake of 55 ± 7 g (mean ± SD) CO2-C m−2 a−1. The growing season average uptake was 92 ± 10 g CO2-C m−2, of which approximately 40% (37 ± 5 g CO2-C m−2) was lost during the nongrowing season. The daily average uptake over the growing season was 0.65 ± 0.57, 0.73 ± 0.61, and 0.68 ± 0.62 g CO2-C m−2 d−1 in 2001, 2002, and 2003, respectively. The daily average net uptake for the month with highest uptake was 1.10 ± 0.33, 1.11 ± 0.63, and 1.22 ± 0.55 g CO2-C m−2 d−1 in July 2001, July 2002, and June 2003, respectively. The daily average efflux during the nongrowing season was 0.14 ± 0.28, 0.15 ± 0.20, and 0.20 ± 0.19 g CO2-C m−2 d−1 in the years 2001, 2002, and 2003, respectively.

1. Introduction

[2] Boreal and subarctic peatlands cover a small part of the global land area (only ∼3%) but store a considerable amount of carbon: 230–450 Gt, according to estimates by Gorham [1991], Immirzi et al. [1992], Turunen et al. [2002], and Lappalainen [1996], respectively. These values are equivalent to ∼20–30% of the global soil carbon pool [Schlesinger, 1997] and ∼40–60% of the roughly 730 Gt carbon currently held in the atmosphere as CO2 [Intergovernmental Panel on Climate Control, 2001]. The current fluxes of carbon between these ecosystems and the atmosphere are therefore considered to be important in global carbon budgets and may be even more important in the future if the exchange rates change in response to anticipated climatic changes.

[3] The carbon exchange at boreal mires is composed of several important fluxes. The major fluxes are uptake of CO2 by photosynthesis and direct loss to the atmosphere of CO2 through respiration, emission of methane (being the net effect of production and consumption, respectively) and loss through water runoff of organic carbon, inorganic carbon and as methane. At the ecosystem level the net ecosystem exchange (NEE, i.e., the net exchange of photosynthesis and ecosystem respiration respectively) rates are used to designate the land-atmosphere exchange of CO2 at the investigated site; the CO2 leaving the system via water runoff is not included in this flux term. The relative importance of the component fluxes varies between mire types and climate conditions. However, since NEE includes the most significant fluxes that add C to the mire ecosystem [e.g., Rivers et al., 1998; Roulet et al., 2006], understanding this component is essential for understanding net C exchange rates in mire.

[4] Currently, most estimates of the annual net carbon exchange in mires are based on data from peat cores. However, to be comparable to estimates of mire net C exchange from flux measurements the concurrent gaseous C losses must be accounted for by modeling. Even when modeling is applied the estimates of Long-term Apparent Rates of Carbon Accumulation (LARCA) may not provide valid estimates of current boreal mire C exchange, especially if the rates of peat accumulation have changed during the Holocene. Apparent rates of carbon accumulation seem to have fallen in recent times (the last thousand years) to 50–60% of LARCA, according to estimates based on peat cores from ecosystems in western Siberia [Turunen et al., 2001], Canadian low arctic and high subarctic regions [Vardy et al., 2000], Sphagnum fuscum Schimp. Klinggr. peat in western boreal Canada [Kuhry and Vitt, 1996] and “mature peat plateau phases” in the east European Russian Arctic [Oksanen et al., 2001]. Thus, on the basis of rates of apparent peat accumulation during the most recent part of the Holocene, the potential for C accumulation in boreal mires may have decreased [Gorham, 1991; Kuhry and Vitt, 1996; Turunen et al., 2001]. Compiling a comprehensive database on current annual NEE rates between mires and the atmosphere could greatly facilitate attempts to elucidate the current role of boreal mires in land-atmosphere exchange of carbon.

[5] Direct flux measurements of NEE have suggested that boreal mires may currently act as either sinks [e.g., Whiting, 1994; Oechel et al., 1995; Shurpali et al., 1995; Alm et al., 1999b; Lafleur et al., 2001a; Aurela et al., 2002] or sources of atmospheric CO2 [Whiting, 1994, Oechel et al., 1995; Shurpali et al., 1995; Carroll and Crill, 1997; Alm et al., 1999b; Lafleur et al., 2001a; Wickland et al., 2001; Aurela et al., 2002]. In comparison to LARCA values these results have led to conclusions by some authors that boreal and subarctic mires may be currently changing from sinks to sources of atmospheric carbon [e.g., Oechel et al., 1995]. Similar conclusions, suggesting that peatlands that are currently in equilibrium may change to being either sources or sinks of carbon following expected climatic changes, have also been drawn from simulations [Hilbert et al., 2000].

[6] Very few estimates of annual CO2 NEE based on direct flux measurements have been derived to date from full-year micrometeorological measurements. Micrometeorological-based estimates of annual exchange are available from: an ombrotrophic mire in southeastern Canada, Mer Bleue (45°40′N, 75°50′W) [Lafleur et al., 2001b, 2003]; a wet minerotrophic mire in northern Finland, Kaamanen (69°08′N, 27°17′E) [Aurela et al., 1998, 2001, 2002, 2004]; an ombrotrophic bog in Siberia (60°45′N, 89°23′E) [Arneth et al., 2002; Schulze et al., 2002]; an Irish blanket bog (51°55′N, 9°55′W) [Sottocornola and Kiely, 2005]; and a lowland temperate peatland (55°48.8′N, 3°14.40′W) in central Scotland [Billett et al., 2004].

[7] The estimates for the Canadian ombrotrophic mire represent net uptake rates over a 5-a period (1998–2002) varying between 9 g CO2-C m−2 a−1 (not significantly different from zero) to a significant uptake of 75 g CO2-C m−2 [Lafleur et al., 2003]. The estimates for the wet minerotrophic mire in northern sub-arctic Finland yield average uptake rates over 6 a (1997–2002) of 22 g CO2-C m−2 a−1, with net ecosystem exchange (NEE) rates ranging between −4 and −53 g CO2-C m−2 a−1 [Aurela et al., 2004]. Data on NEE from the ombrotrophic bog in Siberia in 1998–2000, which do not include the winter period and are therefore overestimates, vary from −43 to −62 g CO2-C m−2 a−1 [Arneth et al., 2002; Schulze et al., 2002]. The annual NEE values at the Irish blanket bog were −49 and −61 g CO2-C m−2 a−1 for the years 2003 and 2004, respectively [Sottocornola and Kiely, 2005] and average uptake over a 2-a period (1996–1998) was found to be 28 (±2.5) g C m−2 a−1 at the lowland temperate peatland in central Scotland [Billett et al., 2004].

2. Study Aims

[8] The major aims of this study were (1) to derive complete year budgets of the site-specific vertical land-atmosphere exchange of CO2 for an oligotrophic, minerotrophic boreal mire in northern Sweden, and (2) to characterize the variability in flux rates and cumulated C exchange during different periods of the year. To achieve these aims, we used data from an eddy covariance measurement system and a meteorological station measuring important air and soil climate variables. On the basis of 3 a of annual flux measurements the oligotrophic fen, Degerö Stormyr, constituted a substantial sink for the land-atmosphere exchange of CO2-C. Furthermore the annual NEE was fairly stable despite substantial variation in the annual weather conditions.

3. Materials and Methods

3.1. Site Description

[9] The study was conducted at Degerö Stormyr (64°11′N, 19°33′E) a mixed acid mire system covering 6.5 km2 situated in the Kulbäcksliden Experimental Forests near Vindeln in the county of Västerbotten, Sweden. The mire, which consists of a rather complex system of interconnected smaller mires, divided by islets and ridges of glacial till, is situated on highland (270 m above sea level) between two major rivers, Umeälven and Vindelälven, approximately 70 km from the Gulf of Bothnia. The depth of the peat is generally between 3 and 4 m, but depths up to 8 m have been measured. The deepest peat layers correspond to an age of ∼8000 a.

[10] The vascular plant community is dominated by Eriophorum vaginatum L., Vaccinium oxycoccos L., Andromeda polifolia L., Rubus chamaemorus L. with both Carex limosa L. and Schezeria palustris L. occurring more sparsely. Carex rostrata L. is characteristic in areas with direct minerogenic water inflow. The bottom layers of the carpets are dominated by Sphagnum majus Russ. C. Jens and Sphagnum lindbergii Schimp., and the lawns by S. blaticum Russ. C. Jens., while S. fuscum Schimp. Klinggr. and S. rubellum Wils. dominates the hummocks.

[11] The climate of the site is defined as cold temperate humid, the 30-a mean (1961–1990) annual precipitation and temperature are 523 mm and +1.2°C, respectively, while the mean temperatures in July and January are +14.7°C and −12.4°C, respectively [Alexandersson et al., 1991]. The length of the vegetation period (stable mean diurnal temperature over +5°C [Moren and Perttu, 1994]) over the measuring period (2001–2003) was 153 ± 15 d [Ottosson-Löfvenius, 2002, 2003, 2004]. The snow cover normally reaches a depth up to 0.6 m and lasts for 6 months on average.

3.2. Measurements

[12] The main instrument site was a 4 m tower located 200 m from the southeast border of the mire (Figure 1), accessed by a boardwalk. All data were collected and stored on a computer in a hut, 150 m south of the tower, on a small mire island with a sparse pine stand. Electrical power (240 V AC) was available at the site.

Figure 1.

Aerial photograph of Degerö Stormyr. The yellow area inside the periphery contributes 90% of the total 30 min measurements of the CO2 exchange, example from daytime, summer 2003 (0306–0309). The outer line delineates the nighttime footprint, and the middle line delineates the daytime footprint. The footprint inside the innermost circle corresponds to 5% of the measurements. The green star shows the location of the tower. Black flags represent other experiments at the site.

3.3. Eddy Flux Measurements

[13] The eddy covariance technique was used to measure the fluxes of CO2, heat and water vapor. The system used consisted of a Solent 1012R2 three-dimensional (3-D) sonic anemometer (Gill Instruments, England) and a LI-6262 closed-path infrared gas analyzer (IRGA; LI COR, Lincoln, Nebraska). The 3-D sonic anemometer was mounted on the tower 1.8 m above the surrounding mire on a 1.0-m-long boom oriented toward the north, and it was heated during the winter. The signals from the 3-D sonic anemometer were connected to a computer, which calculated the fluxes in real time, using EcoFlux software (In Situ Flux AB, Ockelbo, Sweden) and stored the data as both 10 Hz and 30-min averages. Schotanus corrections [Schotanus et al., 1983] of sensible heat flux were applied. A negative flux means an uptake of carbon by the mire, a positive flux an efflux.

[14] The IRGA was mounted in an instrumental box approximately 3 m south of the air intake, which was situated less than 5 cm from the measuring volume of the 3-D sonic anemometer. The air pump was placed behind the IRGA and pulled the air through the analyzer, which was connected to the intake through a 5 m long, 6 mm diameter tubing, with a particle filter (Acro®50 1 μm PTFE, Pall Gelman Laboratory, Ann Arbor, Michigan) between them. The gas analyzer was calibrated weekly against zero concentration (CO2-free N2) and ambient (technical air) CO2 concentration either automatically or manually. The technical air was sampled and its CO2 concentration was checked using a gas chromatograph each time the tube was replaced. The H2O concentrations were calibrated retrospectively against water vapor concentrations obtained from a ventilated MP100 sensor (Rotronic AG, Bassersdorf, Switzerland), which were also used for correcting the latent heat fluxes. Typical calibration intervals for H2O were 3–4 weeks. The EcoFlux software automatically calculates corrections for frequency losses in the tubing and a coordinate rotation is used to obtain zero mean vertical velocity. The whole eddy covariance system was delivered as a complete system with integrated hardware and software in an air-conditioned box with built-in lightning protection and power supply by In Situ Flux Systems AB, Ockelbo, Sweden.

3.4. Air and Soil Climate Measurements

[15] Air and soil climate variables were measured by sensors either mounted on the same tower as the 3-D anemometer, or in representative plant community areas within a hundred meters of the tower. The air temperature and humidity were measured by an MP100 temperature and moisture sensor (Rotronic AG, Bassersdorf, Switzerland) inside a self-ventilated radiation shield mounted 1.8 m above the mire surface. The snow depth was measured by an SR-50 ultrasonic sensor (Campbell Scientific, Logan, Utah) placed approximately 6 m from the flux tower. Water table depths and soil temperatures were measured in a lawn plant community 100 m northeast of the flux tower. The peat and water table surfaces were measured using a float and counterweight system attached to a potentiometer [Roulet et al., 1991]. Soil temperatures were measured by TO3R thermistors mounted in sealed, waterproof, stainless steel tubes (TOJO Skogsteknik, Djäkneboda, Sweden) at 2, 5, 8, 16, 24, and 32 cm depths. Precipitation was measured using an ARG 100 tipping bucket (Campbell Scientific, Logan, Utah) 4 m from the tower. Data from these sensors were scanned at 30-s intervals, and data from the soil temperature probes, the water table level recorder, and the precipitation gauge were scanned at 10-min intervals; all data were stored as either 30- or 60-min mean values by CR10X data loggers (Campbell Scientific, Logan, Utah) and exported daily to the main computer in the hut.

[16] The start and end of the vegetation season are defined as the dates when the daily mean temperature exceeds 5°C for a stable period of time in the spring and the daily mean temperature has dropped below 5°C for a stable period of time in the autumn [Moren and Perttu, 1994].

[17] The mire ecosystem was regarded as a sink when the derivative of the cumulated CO2 was negative (i.e., when there was a net uptake of CO2) and as a source when the derivative was positive (i.e., when there was a net loss of CO2). The growing and nongrowing seasons of each year were then regarded as the sustained periods of net uptake and losses, respectively.

3.5. Wind and Footprint

[18] Wind speed and direction were measured by the Solent 1012R2 three-dimensional sonic anemometer (Gill Instruments, England) used in the EC system. The ecosystem roughness length z0, was calculated by solving the expression

equation image

for z0, where u* and u (friction velocity and wind velocity, respectively) values were obtained from the EC system. The equation above is true for neutral and stable stratification within the boundary layer [Salby, 1996] and was solved for such conditions. The z0 value obtained was used to calculate “footprints,” i.e., the spatial distributions of the CO2 exchange measured and the flux footprints were calculated using the footprint model FSAM by Schmid [1994]. First, we verified the applicability of the parameterized version (MiniFSAM) by comparison with FSAM runs on a subset of data. Then, the MiniFSAM equations were used to calculate the dimensions of the elliptical flux source area for each half hour, separately for day and night, summer and winter. For each such scenario, corresponding polygon surfaces were then successively accumulated in an array in polar coordinates representing the surface around the flux tower. The resulting frequency distributions were then plotted as three-dimensional color maps and as line plots averaged over all wind directions. For winter estimates, the actual measuring height above the snow surface, updated in 30-min steps, was used in the model.

3.6. Data Quality

[19] The CO2 data and sensible heat flux data were visually scrutinized to detect (and remove) obvious outliers in the CO2 data, arising for instance from failure of the measurement system or spikes without any counteracting spikes. Values of u* > 1 from the sonic anemometer were also discarded. The meteorological data set was also scanned for anomalies and such values were removed.

[20] The possibility that the fugacity of the CO2 flux may have been affected by low u* was also investigated. To detect any possible systematic errors in the measurements, we determined the energy balance closure. Since soil heat fluxes were not measured, the energy balance was calculated on a daily basis to minimize the influence of these fluxes. The gas analyzer water vapor concentration measurements were verified against the measurements of ambient water vapor by the Rotronic sensor, which was calibrated regularly using an aspirated psychrometer (Assman, type 761, Wilhelm Lambrecht Gmbh, Göttingen, Germany). When the water vapor measurements from the gas analyzer deviated from the Rotronic data, the gas analyzer values were postprocessed and the fluxes corrected accordingly.

3.7. Gap Filling

[21] In order to integrate the annual NEE budget, missing data had to be replaced in some way. Several approaches have been used for this, but there is no generally accepted standard method [Falge et al., 2001]. Growing season gaps are often filled using some kind of model parameterization of daytime and nighttime fluxes [Aurela et al., 2001; Lafleur et al., 2003], while winter effluxes are estimated by averages, e.g., weekly averages, and <2 h gaps by linear interpolation [Aurela et al., 2001; Lafleur et al., 2001b]. Since the variations in half-hourly rates of CO2 fluxes at mire ecosystems seem to be less dynamic than those of various other ecosystems, e.g., forest ecosystems, because of the low-light saturation limits of Sphagnum species [see, e.g., Letts et al., 2005] other methods for replacing missing data could also be considered. Since species growing on the mire have low-light saturation limits, the fluxes are not as sensitive to daytime variations in incoming radiation, i.e., cloud cover, as the CO2 fluxes in forest systems. Most of the variation in CO2 fluxes within days is related to the diurnal variation of irradiance and variation of higher frequency is of less importance. Therefore we used the mean diurnal variation (MDV) method [Greco and Baldocchi, 1996; Jarvis et al., 1997] for gap filling, which has been previously used in attempts to estimate seasonal and annual sums in forest systems. In this method, ensemble averages are obtained from diurnal patterns of half-hourly data from days preceding and following the days with missing values. We applied this MDV approach in a “gliding” manner like a moving average. In order to replace incorrect or missing data the following protocol was used. The first choice was to use the MDV with a 15-d window spanning the period from 7 d prior to the gap to 7 d after it, e.g., if the 10.00–10.30 measurement for day x was missing, it was replaced by the average of the 10.00–10.30 values for the 7 d preceding and the 7 d following it.

[22] The precision and accuracy of this ensemble average method were evaluated as follows. During periods with at least 14 consecutive days for which complete half-hourly data were available, artificial 1-, 2-, 3-, 4-, and 5-d gaps were created, either sequential or separated by days for which measured data were retained. The single-day gap was repeated for all 14 of the days, while the other gaps were each repeated 5 times within the 14-d period (Figure 2). We defined a lower limit such that at least nine of the 14 values in the ensemble average had to be available in order to be used for filling a gap. If the half-hourly gaps were filled to such an extent that data (measured or filled) for more than 40 of the 48 half hours were available for a day, it was defined as a “good day” and a sum was calculated by multiplying the average of the half-hourly (measured and filled) data by 48. If less than 40 half-hourly data were available, the second choice method was applied, in which standard moving averages were calculated from the daily sums, and 7 + 7-d windows were applied as in the first choice method. This method based on moving averages of daily sums gives a correct daily average for the period, but probably does not accurately reflect the daily variation in controlling processes for the gap-filled days. Therefore the entire period with missing data was replaced with the same average to avoid misinterpretations of daily variations. The same procedures were used regardless of the season or year.

Figure 2.

Gap filling. The relative prediction error (SE) (from internal cross validation) in 30-min average CO2 flux (mg CO2-C m−2 s−1) data is a function of the proportion of missing data using moving averages (7-d) to replace missing data.

[23] The loss of data under normal conditions due to spurious measurements (e.g., measurements obtained in heavy rain or frost) was normally less than 5%. Power failure and computer hangups typically accounted for an additional 10%, which together with a few larger system failures resulted in the overall proportion of missing data being between 20 and 35% of the annual total (35%, 20%, and 31% for 2001, 2002, and 2003, respectively, Figure 3). Of the gap-filled data 90% were replaced according to the first choice, MDV, and the rest with the second choice method. This is comparable to, or lower than, reported proportions in other EC system-based mire studies, for instance, 33–44% of the data were missing in the study of the ombrotrophic mire Mer Bleu by Lafleur et al. [2003], 49% in the study of the wet minerogenic mire, Kaamanen in Finland by Aurela et al. [2002], and 43–51% in the study of a blanket bog in Ireland by Sottocornola and Kiely [2005].

Figure 3.

Proportions of missing values (30-min average flux rates) in 2-week periods during the 3 a of measurement.

[24] The artificial gaps we filled with the MDV method were compared with the measured data and we found a mean error of approximately 15% for large gaps (3–5 days during a 14-d period; Figure 2). The errors for most of the gaps were much smaller. The accuracy and precision of the MDV approach used (Figure 2) were as high as those of more complex approaches, e.g., various parameterizations [Lafleur et al., 2001a], nonlinear regression models [Aubinet et al., 2000; Sottocornola and Kiely, 2005] or neural networks [Papale and Valentini, 2003], and was therefore considered the best choice. The moving average window of 14 days applied was found to be an acceptable compromise, being large enough to minimize the effects of large flux values that occasionally occur, and short enough to avoid seasonal changes affecting the flux estimates. In our estimated budget for the 3 a considered, less than 16% of the estimated carbon fluxes emanate from gap-filled data.

3.8. Uncertainty in Eddy Covariance Measurements

[25] A random error of 20% (SD) was applied on 30 min fluxes [Morgenstern et al., 2004; Humphreys et al., 2006], both on measured and gap filled 30 min values and the same approach was also used on the gap filled data with daily means. The total annual error (SD) was then calculated as the square root of the sum of the respective variances. The standard deviation given for the 3-a mean of the annual CO2 uptake includes both the between year variation and the measurement error (square root of the sum of the respective variances).

4. Results

4.1. Climate

[26] The monthly average air temperatures were close to the 30-a averages (1961–1990) during 2001, while during 2002 the monthly average temperatures from midwinter until early autumn were above average, and in 2003 the mean monthly temperatures were above average during the winter but from early summer onward they were close to average (Figure 4a). The annual precipitation during 2001 was 60% higher than the 30-a (1961–1990) average (860 mm versus 523 mm), while the precipitation figures during 2002 (534 mm) and 2003 (559 mm) were close to the long-term average, although the temporal distributions deviated to some extent. The mean water level during the growing season was 7.4 ± 3.4, 17.3 ± 5.2, and 14.5 ± 6.0 cm below the vegetation surface in 2001, 2002, and 2003, respectively. The high precipitation and very close to normal temperatures during the year 2001 resulted in the water level remaining high during most of the “soil frost-free” season (Figures 4a and 4b). The year 2002 began with a warmer spring and less precipitation than normal and the water table dropped to −25 cm in lawn plant communities during June compared to −12 cm in June 2001 and did not recover to the 2001 level before intense precipitation during August 2003 (Figures 4a and 4b). The year with the largest variability in groundwater levels was 2002, when there were 8 months in which precipitation was lower than usual while temperatures were warmer than normal (Figure 4a). The length of the vegetation period was 169, 150, and 139 days during the years 2001, 2002, and 2003, respectively [Ottosson-Löfvenius, 2002, 2003, 2004].

Figure 4.

Three-annum data (2001–2003) on (a) air temperature (monthly average), and precipitation (monthly cumulated) compared with 30-a means (1961–1990), (b) snow depth, distance to water table level and soil temperature at 10 cm, (c) CO2 flux (daily average), and (d) cumulative CO2 exchange.

[27] No data for snow cover depth were available for the winter 2000–2001. During the winter of 2001–2002 most of the early winter snow (30 cm) melted in late December and the snow cover was maximal (55 cm) in March (Figure 4b). In the following winter, 2002–2003, there was a continuous accumulation of snow until February, when it reached a maximum depth of 65 cm, and it remained at nearly the same depth until it thawed in April. Immediately after the soil thaws, its temperature starts to increase and peaks in July.

4.2. System Performance and u* Dependency

[28] The energy balance closure was estimated by regressing the daily mean of the convective heat fluxes (H + LE) against net radiation (RNT) over the three growing seasons (Figure 5), giving the linear equation (H + LE) = 0.97 * RNT − 8.4, r2 = 0.82, n = 180, p < 0, 01. It is reasonable to assume that most of the deficit, i.e., 8.4 W m−2 + 3% ((1 − 0.97) × 100) of the net radiation is mainly due to soil warming during the growing season, and the energy balance closure was therefore considered satisfactory.

Figure 5.

Sensible and latent heat fluxes [W m−2] plotted against net radiation (Rn) [W m−2]. Data are from the growing seasons of 2001–2003. Sensible and latent heat flux data were calibrated against a Rotronic sensor, and Schotanus corrections were applied to the sensible heat fluxes.

[29] The nighttime CO2 fluxes (binned) for different u* values were not significantly different (Figure 6), and the variance in nighttime fluxes was similar for both high and low u* conditions. These patterns deviate from previously reported patterns [see, e.g., Gu et al., 2005] in that the variability is normally suppressed when conditions are stable, and the fluxes tend to decrease toward zero at very low u*. On the basis of the independency between both flux rate and u* and variance and u* the need for u* filtering was evaluated, and deemed to be unnecessary. Similar conclusions have earlier been drawn for other mire sites [Lafleur et al., 2003; Sottocornola and Kiely, 2005]. The storage term (calculated from one level) was very small and had negligible influence on the pattern. The storage term is probably small because of the low measurement height and the open landscape. Because of these findings we saw no justifiable reason for performing any kind of filtering for low u* values.

Figure 6.

Eddy covariance CO2 flux (Fe) and storage CO2 flux (Fs) plotted as functions of u*, Rg < 5 W m−2, during the 2001–2003 growing seasons. SD and SE are for Fe + Fs on half-hourly data.

4.3. Footprint

[30] The footprint area around the tower is slightly asymmetric and extends more toward the southeast during the summer and toward the south during the winter (Figures 1 and 7) . The dominating wind direction during the summer is biased toward the ENE while in the winter the prevailing wind is from the SSE (Figure 8). The footprint area (95% percentile of source distance, daytime) reached a radius of 22 m during the summer and 76 m during the winter, although the dominating area is much narrower according to the plot of the cumulated fetch distances (Figure 9). During wintertime the measuring height varied between 1.80 m and 1.15 m due to accumulation of snow. This means that periodically the measuring height (zm) is much lower in winter than during summer. The roughness, z0, is smaller during winter, only a few millimeters when snow is present, compared to 2 cm during the growing season when z0 is determined by the plant community structure. The footprint varies very little diurnally (night versus day) during the winter, but it varies substantially during the summer; a radius of about 22 m encompassing about 95% of the source distance during the daytime, increasing to 74 m during the nighttime. Nighttime footprints are generally larger than daytime footprints, due to differences in boundary layer stability during the night and day (Figures 7 and 9).

Figure 7.

Frequency distributions of half-hourly flux source area from the tower, for specific source directions in (a) summer nighttime (0306–0309)(Rg < 5 W m−2); (b) summer daytime, (c) winter nighttime (0211–0302) (Rg < 5 W m−2), and (d) winter daytime. X, Y scale is distance from tower [m]. Color scale shows absolute frequency distribution.

Figure 8.

Wind direction distribution for (a) summer days, (b) summer nights, (c) winter days, and (d) winter nights (Rg < 5 W m−2). Radial axis is % of measured half-hour averaged wind directions, angular axis is degrees (0°–360°) for summer (0306–0309) and winter (0211–0302).

Figure 9.

Cumulative distribution of half-hour fetch distance cumulated for all source directions, 0–360°. Summer nighttime (Rg < 5 W m−2) is indicated by the dotted black line; summer daytime is indicated by the solid black line; winter nighttime (Rg < 5 W m−2) is indicated by the dotted grey line; winter daytime is indicated by the solid grey line. Vertical lines illustrate intercepts with the 95% line (horizontal dotted black line). Summer nighttime and winter daytime vertical lines coincide. The measuring height during winter varied between 1.80 m and 1.15 m (due to the accumulating snowpack); thus it is periodically much lower than during the summer, which explains the relatively short footprint distance during winter time.

[31] The vegetation within the dominating source area during daytime in summer is very homogenous and is constituted by wet lawns and carpet plant communities (Figure 1). During the night, most of the flux emanates from the same microtopographical units and is dominated by the same plant communities. The more distant flux sources toward the east and south east are slightly drier and dominated by “dry” lawn and hummock plant communities, but due to their distant location they make a limited contribution to the total flux and the lack of influence from this variation in plant community composition is reflected in the half-hour average night values during the growing season (Figure 10). During winter time the source area extends somewhat toward the south and southwest, but it is still dominated by lawns and carpets (Figure 1). We consider the variation in plant community composition between source directions to be minor and assume that the entire source area represents the same mire type.

Figure 10.

Exchange of CO2 [mg CO2-C m−2 s−1], with specific wind directions for Rg < 5 W m−2 and u* > 0.1, May–August, 2001–2003. Angular axis is wind directions; radial axis is CO2 flux [mg CO2-C m−2 s−1].

4.4. CO2 Fluxes

[32] The annual NEE at the mire during the three study years, constituted an average uptake of 55 ± 7 g C m−2 a−1. The net carbon uptake during 2001 was 48 ± 1.1 g C m−2 a−1 (±SD); approximately 15–20% lower than the uptake during 2002 and 2003 when it was 61 ± 1.4 g C m−2 a−1 and 56 ± 2.1 g C m−2 a−1, respectively (Figure 4c and Table 1). On the basis of our definition of the growing season, as a period of net uptake, its length was 124 d in 2001 (7 May to 7 September), 132 d in 2002 (1 May to 9 September) and 146 d in 2003 (26 April to 18 September). The total uptake during the net uptake period was about the same in 2002 and 2003, while it was slightly lower in 2001 (Table 1). The nongrowing season efflux was on average about 37 ± 5 g C m−2 a−1 (Table 1), with between-year variations of similar magnitude to those in the growing season. The annual NEE, as a proportion of the growing season sink term, was relatively constant: 60%, 62%, and 57% for 2001, 2002, and 2003, respectively.

Table 1. Monthly and Seasonal Accumulated Net Ecosystem Exchangea
 MayJuneJulyAugustSeptemberGrowing SeasonNongrowing SeasonAnnual Total
  • a

    Average, in units of g CO2-C m−2.

2001−2.7−31.4−34.5−10.23.3−8133−48
2002−23.2−22.1−34.3−15.73.3−9736−61
2003−12.5−36.7−23.8−14.2−11.1−9943−56

[33] The growing season mean daily CO2 fluxes represented net uptake during each month from May to August in all 3 a and peaked during June or July. In 2003, net uptake also occurred during September (Table 1). The month with the highest uptake was July in 2001 and 2002 and June in 2003 (Table 2). The daily average uptake over the growing season was 0.65 ± 0.57, 0.73 ± 0.61, and 0.68 ± 0.62 g CO2-C m−2 d−1 in 2001, 2002, and 2003, respectively. The night efflux is relatively independent of wind direction (Figure 10), indicating that although the distribution of the plant species is not completely homogenous the variations do not significantly influence the CO2 flux. The daily average efflux during the nongrowing season was 0.14 ± 0.28, 0.15 ± 0.20, and 0.20 ± 0.19 g CO2-C m−2 d−1 in 2001, 2002, and 2003, respectively.

Table 2. Daily Mean CO2 Exchange in Each Month During the Net Uptake Perioda
 MayJuneJulyAugustSeptember
  • a

    Average ± SD, in units of g CO2-C m−2 d−1.

2001−0.09 ± 0.21−1.05 ± 0.60−1.10 ± 0.33−0.32 ± 0.370.11 ± 0.48
2002−0.75 ± 0.31−0.74 ± 0.24−1.11 ± 0.63−0.50 ± 0.390.11 ± 0.39
2003−0.40 ± 0.40−1.22 ± 0.55−0.77 ± 0.64−0.46 ± 0.53−0.37 ± 0.62

5. Discussion

5.1. System Performance and Methodological Evaluations

[34] We used the energy balance closure during the growing season as a quality measure of the measurement system. The 3% closure deficit and offset of 8.4 W m−2 in the radiation balance could most likely be explained by the soil heat flux storage term. The regression was based on data from the growing seasons of 2001–2003, during which the soil heat flux was directed into the soil. To evaluate the reliability of the assumption that the closure deficit and offset at Degerö Stormyr during this period represented the soil heat flux, we used data on net radiation to calculate the absolute values for the soil heat flux. The mean net radiation values in May, June, and July at Degerö Stormyr over the 3 a were 80, 87, and 92 W m−2, respectively. On the basis of a closure deficit of 3% and an offset of −8.4 W m−2 the average soil heat fluxes for these 3 months were estimated to be 10.8, 11.0, and 11.2 W m−2, respectively. These values compare quite well with corresponding soil heat flux values for hollows in a Sphagnum mire in central Sweden (11.0, 11.3, and 6.6 W m−2, respectively) published by Kellner [2001]. This comparison indicates that it is valid to attribute both the offset and closure deficit to the soil heat flux rather than unreliable measurements. The energy balance derived by Lafleur et al. [2001b], accounting for the soil heat flux, for an ombrotrophic mire in Canada had a closure of 93%, while there was a mean lack of closure of 20% for the energy balances at 22 Fluxnet sites analyzed by Wilson et al. [2002]. Thus we judge the flux measurements performed in our study to be of high quality.

5.2. Annual Fluxes

[35] The oligotrophic minerotrophic mire, Degerö Stormyr, appears to be a significant CO2-C sink, the annual flux being significantly different from zero. The average annual NEE at Degerö Stormyr was −55 ± 7 g CO2-C m−2 a−1; similar to the annual average of −56 ± 31 g CO2-C m−2 a−1 (1998–2002) recorded at the bog Mer Bleu, Canada [Lafleur et al., 2003], but higher than at the wet minerotrophic mire Kaamanen in northern Finland, −22 ± 20 g CO2-C m−2 a−1 during 1997–2002 [Aurela et al., 2004]. NEE has also been measured at a Siberian ombrotrophic bog by the EC technique during the snow-free seasons of 1998–2000 [Arneth et al., 2002; Schulze et al., 2002]. Using the acquired data and average winter time fluxes from other sites of 0.1 or 0.2 g C m−2 d−1 [Lafleur et al., 2001b; Aurela et al., 2002; this study] the average annual NEE for the Siberian ombrotrophic bog could be estimated to be 35 or 19 g CO2-C m−2 a−1, respectively. A blanket bog in Ireland reportedly had net uptake rates of 49 and 61 g CO2-C m−2, during 2003 and 2004, respectively [Sottocornola and Kiely, 2005]. In central Scotland, the net uptake across a lowland temperate peatland was found to be 28 (±2.5) g CO2-C m−2 over a 2-a period (1996–1998) [Billett et al., 2004].

[36] Estimates of the annual CO2 exchange at mires based on chamber measurements give similar or slightly higher values than available eddy covariance-based estimates. Estimates of annual NEE from a large palsa mire, Vaisejäggi, with similar vegetation to that at Degerö Stormyr of −39.6 and −138 g CO2-C m−2 have been published for 1998 and 1999 (a very wet year), respectively, by Nykänen et al. [2003], using winter estimates based on snow profile measurements. In contrast, annual NEE from the Finnish ombrotrophic bog Ahvensalo at Ilomantsi (65°N), represented a net loss of 82 g CO2-C m−2 during the year 1994, in which the summer was exceptionally dry, according to Alm et al. [1999b].

[37] The sites for which 3 a or more years of reported annual NEE data from EC measurements are available represent a wide range of climate conditions, spanning latitudes between 45°N and 69°N, annual average temperatures ranging from −3.6°C (Siberian site) to +5.6°C (Mer Bleu) and annual cumulative precipitation from 395 mm (Kaamanen) to 910 mm (Mer Bleu). If the sites for which EC measurements are available in Ireland and Scotland are also considered, the temperature and cumulated precipitation ranges are extended to +10°C and >2000 mm, respectively. This climatic range also covers most of the sites used for chamber measurements cited in the above paragraph.

[38] The climates at mire ecosystem sites where annual NEE (direct land atmosphere exchange of CO2-C) have been estimated cover most of the climatic range under which mires develop, and with few exceptions most estimates of average annual NEE rates at mires range between −20 and −60 g CO2-C m−2 a−1. Even if the estimates for annual NEE in mire ecosystems are limited, the data seem to indicate that the current annual NEE in mires is on average less than −60 g CO2-C m−2 a−1. The two mires with the highest recorded average annual net NEE rates are the oligotrophic minerogenic mire Degerö Stormyr, and the raised ombrotrophic mire Mer Bleu (−55 ± 7 and −56 ± 31 g CO2-C m−2 a−1, respectively). Although the averages were derived over quite limited time periods and with varying weather conditions, it is worth noting that the averages are quite similar, although they represent very different types of mire. Both are oligotrophic mires, but Degerö Stormyr is minerogenic while Mer Bleu is a classical ombrotrophic mire. Furthermore, although their reported average annual NEE rates are similar, the ecological controls are different. For example, the average water table at Degerö Stormyr is normally 10–20 cm below the mire surface while it is more than 50 cm below the surface at Mer Bleu.

5.3. Seasonal Fluxes

[39] The average cumulative growing season net uptake at Degerö Stormyr was slightly higher than both Mer Bleue in Canada (76 g CO2-C m−2 a−1 [Lafleur et al., 2003]) and Kaamanen in Finland (76 g CO2 C m−2, recalculated from Aurela et al. [2004]) and lower than at the Finnish mire Kaamanen (188 g CO2-C m−2 a−1 [Aurela et al., 2004]). The wintertime or nongrowing season efflux was low and more or less constant, but as stressed earlier by various authors, including Aurela et al. [2002] and Lafleur et al. [2001b], it is the length of the season that makes the low winter time efflux a significant term in the annual budget. At Degerö Stormyr the wintertime efflux is equivalent to 40% of the growing season uptake. The effluxes during the winters of 2001–2002 and 2002–2003 were 36 and 43 g CO2-C m−2, respectively; comparable to the 32.5–36.0 reported at Mer Bleue [Lafleur et al., 2003] and 41 g at an ombrotrophic bog at Ahvensalo [Alm et al., 1999a].

[40] The lengths of the growing and nongrowing seasons are very important determinants of the annual NEE. At high latitudes, the soil frost commonly lasts one or several months after the spring equinox, inhibiting water uptake by plants. During this period the incoming daily PAR is approaching annual maximum values. An earlier onset of the growing season will thus increase the length of the part of the growing season with high net primary production (NPP). Changing the date of the transition from the nongrowing to the growing season by a week at Degerö Stormyr would have changed the annual NEE by 17% on average over the three study years.

5.4. Daily Fluxes

[41] The average uptake rate at Mer Bleue during the net sink period was 0.81 CO2-C m−2 d−1 [Lafleur et al., 2001b], slightly higher than at Degerö Stormyr. During the summer (June to September 1998) mean daily NEE flux values varied considerably, from losses of 1.3 g CO2-C m−2 d−1 to a maximum uptake of 2.3 g CO2-C m−2 d−1. The mean daily fluxes in late fall and the snow covered periods (5 November to 6 April in 1998–1999), were fairly constant, with effluxes of 0.3 g CO2-C m−2 d−1 [Lafleur et al., 2001b]. The highest monthly daily average CO2 net uptake at the minerotrophic mire Kaamanen in northern Finland was about −1.6 g CO2-C m−2 d−1 and occurred in July. The highest daily effluxes, about 1.1 g CO2-C m−2 d−1, were observed in early June and August (1997). The daily NEE values in April were about 0.6 g CO2-C m−2 d−1 [Aurela et al., 2001] and maximum daily NEE values of about −2.4 g CO2-C m−2 d−1 were observed in July, while the highest daily respiration rates, of about 0.68 g CO2-C m−2 d−1, were observed just before and just after the sink period (1998) [Aurela et al., 2002]. Compared to Degerö Stormyr, the Kaamanen and Mer Bleue mires both appear to have higher daily uptake rates. The large variation in daily NEE also indicates that the mire has the potential to, within certain limits, respond to a changed climate.

5.5. Diurnal Maximum/Minimum Fluxes

[42] The maximum daytime uptake rates during the growing period at Degerö Stormyr were typically −0.10 mg CO2 m−2 s−1, and the nighttime maximum respiration rate was typically close to 0.07 mg CO2 m−2 s−1 (95% percentiles). Corresponding values at the bog Mer Bleue in southern Canada were −0.45 and 0.20 mg CO2 m−2 s−1, respectively [Lafleur et al., 2003]. At an open peatland in north central Minnesota the peak uptake varied from 0.15 to 0.24 mg CO2 m−2 s−1 [Shurpali et al., 1995], while the measured maximum net uptake rates from a minerotrophic wetland near Thompson, Manitoba, Canada, were 0.55 mg CO2 m−2 s−1 at midsummer in 1994 and 1995 [Joiner et al., 1999]. The highest individual downward flux densities at Kaamanen were about 0.25 mg CO2 m−2 s−1 and occurred at the end of July, while the highest respiration rates of 0.15 mg CO2 m−2 s−1 were observed later in August [Aurela et al., 2001]. Aurela et al. [2002] further report typical daytime peak values of about −0.20 mg CO2 m−2 s−1 in July and a typical nighttime respiration rate in summer of 0.10 mg CO2 m−2 s−1.

[43] Relative to the compared mire sites, both the maximal daytime uptake and the nighttime release at the minerogenic Degerö Stormyr were relatively low. The relatively high net CO2 effluxes during the autumn probably result from both autotrophic respiration in the large plant biomass and the combined effects of relatively high soil temperatures (which promote high rates of heterotrophic respiration) and low levels of PAR (which only allow low rates of photosynthesis).

6. Conclusions

[44] The boreal oligotrophic minerotrophic mire Degerö Stormyr represented a substantial net sink for CO2-C according to NEE values derived from EC measurements during 2001–2003, which are comparable to values in the higher range of previously reported annual net uptake rates. However, it should be stressed that to derive values for mire C exchange that are comparable to long-term averages from peat cores both methane emissions and the export of carbon through runoff also need to be included. The annual CO2 net uptake was fairly similar between years, despite large differences in weather conditions; the years 2001 and 2002 being relatively wet and relatively dry, respectively. Although our data only span 3 a, the results indicate that the uptake and release processes have similar responses to changes in weather conditions; that is, if photosynthesis decreases, the loss of C through respiration also decreases. The temporal partitioning into a nongrowing season and a growing season was of major importance for the annual budgets, and approximately 40% of the carbon uptake sequestered during the growing season was lost during the nongrowing season. However, the major effect of changes in the relative lengths of the growing and nongrowing seasons is the impact they have on uptake. The average net daily CO2 uptake during the growing season is substantially higher than the average net daily CO2 release during the nongrowing season, and thus the increase in numbers of days with a net uptake is more important than the decrease in days with net release. It is not only the number of days of the growing season that is important, starting a week earlier in spring affects the annual budget more than an additional week in autumn.

[45] When CO2 exchange rates from a number of mires are compared, it is evident that the nongrowing season efflux rates are quite similar and the differences in annual uptake are mainly determined by the growing season flux rates, but they are also affected by the relative length of each season.

Acknowledgments

[46] Financial support from the following sources is acknowledged: The Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (grant 21.4/2003-0876 to M. N.) and the Swedish Research Council (grant 621-2001-1911 to M. N. and grant 621-2003-2730 to L. K.). The Kempe Foundation is also gratefully acknowledged for grants for the micrometeorological instruments. This study is a contribution from the Nordic Centre for Studies of Ecosystem Carbon exchange and its Interactions with the Climate System funded by the Nordic Council of Ministers.

Ancillary