Restoration of drained peatlands in southern Finland: initial effects on vegetation change and CO2 balance


Dr Jukka Laine, Department of Forest Ecology, P.O.Box 24, FIN-00014 University of Helsinki, Finland. Fax: + 358–9-1917605; E-mail: Jukka.K.Laine@Helsinki.Fi


1. Components of the CO2 balance for a drained minerotrophic fen and a drained ombrotrophic bog were measured for permanent plots using static chamber techniques for 1 year before and 2 years after a rewetting treatment in part of both sites. During the same period, changes in ground and bottom layer vegetation composition were monitored.

2. After the treatment, the water table rose, the average increase being 25 cm for the fen site and 20 cm for the bog site. In the untreated areas the average water table remained at the pretreatment level.

3. There was a clear change in vegetation composition in the rewetted area of the fen site where the cover of cottongrass Eriophorum vaginatum L. increased drastically. The change in vegetation composition seemed to be dependent on the nutrient status of the sites, being faster at the more nutrient-rich fen site.

4. The rates of CO2 efflux from the soil surface decreased on those rewetted plots where all vegetation had been removed. In 1996, the CO2 efflux rates from the soil surface of the untreated plots were about twice as high as from the rewetted plots with a high water table.

5. The change in water table levels and vegetation composition affected the seasonal (mid-May to end of September) CO2-C balances. For the rewetted area of the fen site, the CO2-C balance varied from 162 to 283 g m–2, being greatest in a plot with dense E. vaginatum cover and a high water table. For the rewetted area of the bog site, the CO2-C balance varied from 54 to 101 g m–2, being greatest in a hollow-level plot with a high water table and mire vegetation. For the untreated areas, the CO2-C balance was close to zero (3 g m–2 at the bog site) or negative (–183 g m–2 at the fen site), when carbon fixation by the tree stand was omitted.

6. The results are encouraging from the practical point of view as restoration of both fen and bog sites initiated vegetation succession and CO2-C balance development towards those of pristine mires.


In Finland, about half of the original peatland area (10·4 million ha) has been drained for forestry (Paavilainen & Päivänen 1995). Environmental issues and the need for the conservation of key biotopes have created pressure to restore drained peatland ecosystems. The present area of drained peatlands in national parks is 6000 ha and, if the planned extensions to the parks are realized, the area will increase tenfold (Heikkilä & Lindholm 1995). After experimental work (Vasander, Leivo & Tanninen 1992), drained peatlands in protected areas are being restored on a large scale. Restoration may also be considered outside protected areas, where profitable drainage for forestry has failed and the original mire belongs to key biotopes or threatened site types.

The aim of peatland restoration is to bring back a self-sustaining, naturally functioning mire ecosystem (Wheeler & Shaw 1995). To achieve this, the drainage ditches are blocked to raise the level of the water table and the predrainage inflow to minerotrophic sites is re-established by directing extra groundwater to the area. If the restoration is successful, the original peat-forming mire species, mainly Sphagnum spp. and sedges, will recolonize and the natural carbon accumulation processes are resumed (e.g. Heikkilä & Lindholm 1995; Wheeler & Shaw 1995).

Natural, pristine peatlands act as sinks for atmospheric carbon because of an unbalanced ratio of primary production to decomposition (e.g. Clymo 1984; Gorham 1991). The annual primary production of Finnish mire ecosystems has been reported to vary from about 100 g m–2 year–1 (Silvola & Hanski 1979; Vasander 1982) to 1400 g m–2 year–1 (Saarinen 1996). These values correspond to carbon accumulation rates of 15 g m–2 year–1 in fens to 24 g m–2 year–1 in bogs (Tolonen & Turunen 1996), with an average value of 18 g m–2 year–1 (Clymo, Turunen & Tolonen 1998). Drainage causes changes in mire vegetation, such that forest species become dominant (Laine, Vasander & Laiho 1995; Laine, Vasander & Sallantaus 1995) and shifts the focus of primary production from the field and bottom layers to the tree layer (Laiho & Laine 1997). Increased aeration of the surface peat significantly enhances organic matter oxidation, as found in studies concerning CO2 emissions from drained peatlands (Silvola, Välijoki & Aaltonen 1985; Glenn, Heyes & Moore 1993; Silvola et al. 1996a). In Finnish conditions, the increased above- and below-ground litter flow of organic matter from the tree stand into the drained soil system appears, however, to compensate for the increased oxidation carbon losses (Finér & Laine 1996b; Minkkinen & Laine 1998).

Thus, it can be expected that the successional postdrainage vegetation will start to develop back towards mire vegetation and that the net ecosystem CO2 balance will return to predrainage levels after the rewetting treatment. How fast and to what extent these processes are restored is still unknown due to the lack of studies on carbon cycling on restored peatlands. In this study, we present results from the restoration of two peatlands (minerotrophic and ombrotrophic) in southern Finland that had been drained for forestry but were partially restored in 1995. The aim of the study was to determine the initial effect of rewetting on (i) water level and vegetation composition, and (ii) the consequent change in the CO2-C balance.

Materials and methods

Research areas and rewetting treatments

The two study sites, Konilamminsuo mire (61°48′N, 24°17′E) and Viheriäisenneva mire (61°51′N, 24°14′E), are in southern Finland. The area belongs to the zone of southern boreal coniferous forests (Ahti, Hämet-Ahti & Jalas 1968) and the zone of eccentric raised bogs (Ruuhijärvi 1982). The long-term mean annual temperature of the region is 3 °C; for February and July the mean is –8·5 °C and 16·5 °C, respectively. The mean annual precipitation is 700 mm, of which 250 mm falls as snow. The long-term average cumulative temperature sum (daily mean temperatures > 5 °C) is 1150 degree days.

In its pristine state, Konilamminsuo mire (the fen site) was a minerotrophic tall-sedge pine fen (sensuLaine & Vasander 1996) and Viheriäisenneva mire (the bog site) was an ombrotrophic low-sedge bog. The fen site was drained 43 years and the bog site 30 years prior to commencement of our study. The ditch spacing used at the fen site was 50 m and that at the bog site, 30 m. The ditch depths ranged between 80 and 90 cm, which is the standard in Finland (Paavilainen & Päivänen 1995). The fen site was typical of the most common mire site type drained for forestry in Finland (Keltikangas et al. 1986), and it had a thriving pine Pinus sylvestris L. stand before the area was clear-cut and rewetted (Komulainen et al. 1998). The bog site is an example of a mire site type where drainage for forestry proved uneconomic.

Restoration, which involved clear-cutting the trees and blocking the drainage ditches, was started in February 1995. Both sites have a rewetted area and an untreated area where the original drainage treatment was not altered. In the rewetted area of the fen site, one of the drainage ditches was blocked and one was totally filled in. One additional ditch (feeder ditch) was excavated above the study area, which connected two drainage ditches and directed groundwater from the surrounding mineral soil to the restored area (restoration of the original minerotrophic geohydrology). Another additional ditch was excavated in the lower part of this fen study area to collect the water flowing through the area. The area restored was 1·1 ha, 0·6 ha of which were clear-cut. The removal of the P. sylvestris stand reduced evapotranspiration and increased light reaching the ground vegetation. In the rewetted area of the bog site, the drainage ditches were blocked by peat dams and the collector ditches were completely filled in. The area restored is 10·5 ha, and most of the stunted pines and seedlings were removed from the area of rewetted plots (0·5 ha).

Vegetation monitoring

In 1994 nine plots were established at the fen site and eight at the bog site. The locations for the plots were selected to be representative of different kinds of microtopography and vegetation cover. Plots FR1-FR6 at the fen site and BR1-BR5 at the bog site were located in the rewetted area (rewetted plots) and the rest of the plots (F7-F9 and B6-B8) in the untreated area. Plots were delimited with 0·36 m2 (60 × 60 cm) aluminium collars, inserted into the peat to a depth of 20–30 cm and surrounded with boardwalks to minimize disturbance during the measurements. The percentage cover class (scale 0·1, 1, 2, … 10, 15, 20, … 90, 91, … 100%) of each species was visually estimated from the plot by the same person at the end of each growing season each year. The nomenclature follows Moore (1982) for vascular plants, Ahti (1993) for lichens, and Koponen, Isoviita & Lammes (1977) for bryophytes.

In 1995, four additional plots were established: two (FR0 and BR0) in the rewetted area and two (F0 and B0) in the untreated area. All vegetation was removed from these plots and the peat surface was kept bare throughout the growing season. In May 1996, all the vascular plants were removed from plots FR1 and FR6 on the fen site and kept from regrowing by regular cuttings.

Carbon dioxide flux measurements

Carbon dioxide fluxes in the plots were measured using static chamber (aluminium or plastic chamber placed onto collar) techniques. Measurements were carried out during the growing season every second week in 1994, every week in 1995, and every third week in 1996. The CO2 concentration in the ventilated and thermostated transparent plastic chamber (112 dm3) was measured with a portable infrared gas analyser (ADC, LCA 2, Analytical Development Company, Ltd, UK) at intervals of 30 s after closing the chamber. The net CO2 exchange (PN) measurement was carried out in stable light during a period of 120–180 s. On days with high irradiance, artificial shades were used to extend the irradiance range to smaller values. After PN measurement, the chamber was removed and aerated before measuring total CO2 efflux (RTOT) from the vegetated plots and CO2 efflux from soil surface (RS) at the non-vegetated plots. The RTOT and RS measurements were made with the chamber covered with an opaque aluminium blanket. Photosynthetically active radiation (PAR; LI-COR quantum sensor LI-190SB, units µmol m–2 s–1), water table level and soil temperatures were monitored simultaneously with the gas flux measurements. Gas fluxes were calculated from the linear change of CO2 concentrations inside the chambers as a function of time. An estimate of gross CO2-assimilation (PG) in photosynthesis was calculated as a sum of net CO2 exchange (PN) and total respiration (RTOT) values (Alm et al. 1997).

RTOT was also measured in plots FR1-FR4, F7-F8 and BR1-BR4, B6-B7 with an opaque static aluminium chamber technique (Martikainen et al. 1992). A series of air samples (5, 15, 25 and 35 min after closing) were taken in situ from the headspace of the chamber using 60 mL plastic syringes and CO2 concentrations in the samples were determined in a laboratory with a gas chromatograph [HP 5890 Series II, for analysis method see Nykänen et al. (1995)] within 24 h after sampling.

Data analysis

The effect of the rewetting treatment on the rate of total CO2 efflux (RTOT) from the vegetated plots and on the rate of CO2 efflux from soil surface (RS) at the non-vegetated plots was analysed by comparing the seasonal plot means in an analysis of covariance with the –5 cm soil temperature as a covariate. Log-transformation was used to obtain a linear relationship between the dependent variables and soil temperature, and to minimize the heterogeneity of group variances. Following the results of the analysis of covariance, the similarity of plot means of RTOT and RS were tested using Tukey's HSD pairwise comparison.

The effect of treatment on potential rates of gross CO2 assimilation (PG) and net CO2 exchange (PN) was studied by analysing the values measured at the light-saturated level (PAR > 500 µmol m–2 s–1) using a one-way anova. This irradiance was used to minimize the heterogeneity of group variances and was based on the common observation that photosynthesis responds hyperbolically to irradiance and that the individual leaves of many C3 plants are unable to use additional light above 500 µmol m–2 s–1 (Long & Hällgren 1987). Following the anova the similarity of plot means was tested using a Tukey's HSD pairwise comparison.

The seasonal estimates for the CO2-C balance (from mid-May to the end of September) were calculated using the 1995 data with the most frequent sampling. At the fen site, the balance was calculated for plot FR1, which was used to represent an efficient rewetting treatment. Plot FR3 was used to represent an intermediate treatment, and plot F7 as the untreated. At the bog site, plot BR1 represented the hollow level and plot BR4 the hummock level and plot B6 was used as the untreated. The estimates were calculated using environmental response functions for PG and RTOT, with solar irradiance (I), peat temperature (T), water table level (WT) and effective temperature sum index (ETI) as the driving variables. Two variants of ETI (ETI1 and ETI2) were calculated to model the seasonal trend in the photosynthetic activity of different vegetation covers. ETI1 was calculated as the quotient of the cumulative temperature sum and of day number (17 May = 1), and results in a curve rising to a maximum in early July and descending towards autumn (Alm et al. 1997) (performed best for plots FR3 and F7 and BR1, BR4 and B6). ETI2 was calculated as the quotient of the cumulative temperature sum and of Julian day (17 May = 137), and results in a curve which rises to a maximum early in September (performed best for plot FR1). The CO2-C balance was calculated as a difference between PG (f(I, ETI)) and RTOT (f(T, WT)) for every hour and summed over the growing season. The response functions were of the form:

PG = b1I/(k+I) ETI(eqn 1)
logRTOT = b0+b1T +b2WT +b3WT2(eqn 2)

The coefficients (b0, b1, b2 and b3) were estimated using non-linear (eqn 1) or linear (eqn 2) parameter estimation techniques. Equation 1 is formally the same as the Michaelis–Menten function for enzyme kinetics. Here we use it to describe the saturating response (rectangular hyperbola) of photosynthesis to irradiance (I). Hourly averaged irradiance and peat temperature values were measured at an automatic weather station on Lakkasuo mire, 1 km east of the fen site and 7 km south-east of the bog site. A time series of solar irradiance (measured with LI-COR pyranometer sensor LI-200SZ, unit W m–2 s–1) was converted to quantum sensor units (LI-190SB, µmol m–2 s–1) using a quadratic relationship (r2 = 0·98, n = 48) between the pyranometer and quantum sensor readings. Weekly average water table levels used in the calculations were monitored in each plot. All the calculations were made with the SYSTAT software package (SYSTAT 1996).

The coefficients of determination for PG (eqn 1) ranged between 0·79 and 0·87 at the fen site and between 0·66 and 0·80 at the bog site, and those for RTOT (eqn 2) between 0·82 and 0·85 and between 0·81 and 0·88, for the fen and bog sites, respectively (Table 1).

Table 1.  The parameter values for the regression equations (see eqns 1 & 2) for estimating hourly gross CO2-assimilation, PG (mg m–2 h–1) and total CO2 efflux, RTOT (mg m–2 h–1). Coefficients of determination (R2) and degrees of freedom for regression (d.f.reg) and for residuals (d.f.res) are also shown. For a description of the plots, see Table 2
  Fen site
FR3F7Bog site
Parametersb0 1·62 1·89 1·95 1·44 1·79 1·45
 b1 0·041 0·047 0·042 0·038 0·046 0·058
 b2 0·062 0·019 0·0083 0·066 0·0101 0·0091


Effect of restoration on water table level

The cumulative precipitation during June–September in 1994, 1995 and 1996 was 267, 267 and 282 mm, respectively, values very close (98–104%) to the 30-year average for the region. The distribution of monthly precipitation during these 3 years was also close to the long-term average, and seasonal fluctuation in precipitation is clearly reflected in the water table dynamics of the sites (Fig. 1). In both rewetted areas, the water table was higher than before the treatment; the increase averaged 25 cm at the fen site and 20 cm at the bog site (Table 2). In the rewetted area of the fen site, the water table in plots FR1, FR2, FR6 and FR0, which were near the feeder ditch (˜30 m) was quite stable and was closer to the mire surface during the whole growing season than in plots FR3, FR4 and FR5, which were about 60 m from the feeder ditch (Table 2, Fig. 1). Some flooding was observed in the rewetted area of the bog site, especially in hollows and lawns during rainy periods. The average water table depth in the untreated areas of both sites remained at the pretreatment level during the study period (Table 2, Fig. 1).

Figure 1.

Weekly precipitation and water table dynamics in three selected plots at the study sites before (1994) and after the rewetting treatment (1995, 1996). For a description of the plots, see also Table 2.

Table 2.  Mean water table depth (cm) by site, plot and year. In the rewetted area of the fen site, plots FR1, FR2 and FR6 are about 30 m from feeder ditch and plots FR3, FR4 and FR5 are about 60 m from feeder ditch. In the rewetted area of the bog site, plots BR1 and BR5 are at hollow level, BR2 and BR3 are at lawn level and BR4 is on a low hummock. Plots F7-F9 and B6-B8 were in untreated areas of the sites. Plots FR0, BR0, and F0, B0 are plots where vegetation was removed, in the rewetted and untreated areas, respectively
  1.  –, not measured.

FenFR0– – –  7·0 3·515 7·0 6·412
 F0– – – 43·0 9·71434·011·69
 FR138·013·41510·0 4·334 9·5 4·818
 FR233·511·815 6·0 4·335 7·0 4·419
 FR631·512·112 2·5 3·320 8·5 5·611
 FR340·513·91715·0 8·83523·514·514
 FR442·513·61714·5 8·33220·513·114
 FR551·012·51120·5 7·62026·014·59
 F945·0 7·5839·5 8·02039·012·610
BogBR0– – – 23·011·81713·010·28
 B0– – – 35·011·51230·013·59
 BR127·57·115 9·5 9·93610·5 8·615
 BR535·55·6912·010·52412·5 9·810
 BR239·55·21315·510·63514·0 9·915

Effect of restoration on vegetation

At the time of plot establishment in 1994, the vegetation of both mires was in a successional stage, initiated by drainage, changing towards upland forest vegetation. At the fen site, the field layer was dominated by forest dwarf shrubs, such as Vaccinium myrtillus and V. vitis-idaea, with an admixture of mire shrubs, Ledum palustre and Vaccinium uliginosum. Although Eriophorum vaginatum was common, its cover was negligible. The bottom layer was characterized by forest mosses (i.e. Pleurozium schreberi) and Sphagnum spp. that grow in shade and moderately dry sites (i.e. S. angustifolium, S. russowii, S. magellanicum). At the bog site, the bottom layer of the lawn level was dominated by Cladonia species, which were also common on hummocks. The hummocks were dominated by Sphagnum fuscum while in moister lawn and hollow surfaces, S. rubellum and S. balticum were abundant. A typical species in open mires, Andromeda polifolia, was common in the field layer and Trichophorum cespitosum and Eriophorum vaginatum were sporadically present. Dwarf shrubs, such as Calluna vulgaris and Empetrum nigrum were found at both hummock and hollow surfaces.

Two years is a short period to monitor vegetation development but a clear change in vegetation composition was observed after the rewetting treatment. This change was especially clear at the fen site where Eriophorum vaginatum became a dominant species of the field layer in the first year after treatment and its cover continued to increase in the second year (Fig. 2). The increase was greatest near the feeder ditch. There was a moderate rise in the abundance of dwarf shrubs, such as Ledum palustre, Vaccinium myrtillus and V. vitis-idaea in the rewetted area but also in the untreated area. The cover of E. vaginatum also rose in one plot in the untreated area (Fig. 2), obviously as a result of an increase in light flux after clear-cutting the adjacent rewetted strip. In the bottom layer, the cover of Sphagnum angustifolium moderately increased and the increase was highest in plot FR1, where the field layer vegetation had been removed in 1996. At the ombrotrophic bog site, changes in the vegetation following the rewetting treatment were small in comparison to the minerotrophic fen site. The cover of shrubs such as Andromeda polifolia, Vaccinium oxycoccos and V. microcarpum showed a moderate increase in the hollow level. The cover of Empetrum nigrum increased on hummocks, while Calluna vulgaris started to die in the hollows. The cover of Cladonia species decreased while that of Sphagnum balticum, S. fuscum and Polytrichum strictum increased.

Figure 2.

Projection cover of Eriophorum vaginatum in the plots of the fen site at the end of the growing season before rewetting in 1994 and after treatment. The vascular plants were removed from plots FR1 and FR6 in May 1996. Plots in the drained, untreated area are shown with a dashed line, and those in the rewetted area, nearer the feeder ditch with a thicker solid line.

Effect of restoration on co2 efflux from soil surface at non-vegetated plots

The rate of CO2 efflux from the soil surface (RS) of the rewetted plot (FR0) of the fen site varied from 42 to 560 mg m2 h–1 and that of the untreated plot (F0) from 131 to 600 mg m2 h–1. The RS rates at the bog site were lower, varying from 36 to 290 mg m2 h–1 for the rewetted plot (BR0) and from 51 to 310 mg m2 h–1 for the untreated plot (B0). The mean rates of RS were significantly higher in the untreated plot than in the rewetted plot at the fen site in both years and at the bog site in 1996 (Table 3). The 1996 values of the rewetted plots were clearly lower than those in 1995.

Table 3.  Mean rate of CO2 efflux from the soil surface, RS (mg m–2 h–1) for plots where the vegetation was removed, adjusted to a mean soil temperature (at –5 cm) in the analysis of covariance. Those rates on the rewetted plots, which differ significantly (P < 0·05, Tukey's HSD pairwise comparison) from those on untreated plots, are shown in bold

Effect of restoration on total co2 efflux from vegetated plots

The rates of total CO2 efflux (RTOT) from vegetated plots were generally higher at the fen site than at the bog site (Table 4), and were related to the variation in peat temperature. The effect of rewetting on RTOT was more complicated than on RS. During the first year after rewetting, the mean rate of RTOT remained at the same value as in 1994 in all plots at the fen site except for plot FR6, which had the highest water table, where it decreased significantly (Table 4). During the second year of rewetting, the mean rate of RTOT significantly increased in the plots further from the feeder ditch (FR3, FR4 and FR5, Tables 2 & 4), which had deeper water table levels and where cover of Eriophorum vaginatum was still clearly increasing. In the plots where vascular plants were removed in 1996 (FR1 and FR6), the mean rates of RTOT were significantly lower than in all other plots.

Table 4.  Mean rate of total CO2 efflux, RTOT (mg m–2 h–1) from vegetated plots, adjusted to mean soil temperature (at – 5 cm) in the analysis of covariance. Those rates of RTOT after rewetting (years 1995 and 1996) which differ significantly (P < 0·005, Tukey's HSD pairwise comparison) from those before rewetting (year 1994) are shown in bold. For a description of the plots, see Table 2

In the case of the rewetted plots on the bog site, the trend was for the mean rates of RTOT to fall from 1994 to 1996 (Table 4). The mean rates of RTOT significantly decreased in the hollow-level plots with the highest water tables and they were, in general, significantly lower than in the lawn and hummock level plots during 1995–96 (Table 4).

Effect of restoration on potential gross co2 assimilation

There were no significant differences among the plots ( d_f. = 6, 27, F = 1·62, P = 0·20) in the mean rates of potential gross CO2 assimilation (PG) at the fen site in 1994. The rise of the water table and the following dramatic increase in Eriophorum vaginatum cover, however, resulted in a significant increase in the mean rates of potential PG of the rewetted plots but also for untreated plot F8. The increase for plot F9 may be attributed to an increase in the cover of Ledum palustre in 1996 (Table 5). After removal of the vascular plants from plots FR1 and FR6 in 1996, the mean rates of potential PG decreased to the values measured in 1994 or even lower (Fig. 3 and Table 5).

Table 5.  The mean potential (PAR > 500 µmol m–2 s–1, from June to August) rate of gross CO2 assimilation, PG (mg m–2 h–1) for each plot and year. Those rates of PG after rewetting (years 1995 and 1996) which differ significantly (P < 0·005, Tukey's HSD pairwise comparison) from those before rewetting (year 1994) are shown in bold. For a description of the plots, see Table 2
  1. –, missing data.

 FR690099715707213 490879
 F7– – – 1240641211107110
 F8– – – 83054121470668
 BR5850447 6502424 460399
 BR2480466 63026185203510
Figure 3.

Monthly averages of potential (PAR > 500 µmol m–2 s–1) gross CO2 assimilation (PG) during 1994–96 in plots FR1 and FR6 where all the vascular plants including Eriophorum vaginatum were removed in May 1996. The projection cover of E. vaginatum is given in parentheses, after the year, at the end of each line.

At the bog site, the response of the mean rates of potential PG to the raised water table levels varied with vegetation composition (Table 5). In plot BR5 at the hollow level, where Calluna vulgaris was suffering from a high water table, the mean rate of potential PG significantly decreased in the first year after rewetting (Table 5). In plot BR4, the increase in Empetrum nigrum cover caused a significant increase in the mean rate of potential gross photosynthesis (Table 5).

Effect of restoration on potential net co2 exchange

As the net CO2 exchange is the result of photosynthesis and respiration, the dynamics in Eriophorum vaginatum cover were clearly reflected in the rates of potential PN in the rewetted area of the fen site (Fig. 4, Table 6). In 1994, the mean rates of potential PN were quite close to zero and no significant differences were observed in plotwise comparisons (d.f. = 6, 27, F = 1·80, P = 0·16). In 1995, the mean rates of potential PN were clearly positive and rose together with increased E. vaginatum cover. In 1996, after removal of vascular plants, the mean rates of potential PN in plots FR1 and FR6 declined but on average remained positive (Fig. 4, Table 6). The changes in vegetation cover presumably also account for the observed increased rates of potential PN on the untreated plots (Table 6). At the bog site, the potential PN rates in 1994 were slightly positive for all plots except plot B7 but the effect of rewetting on rates of potential PN was clearly smaller than at the fen site (Fig. 5, Table 6).

Figure 4.

Rates of net CO2 exchange (PN) in relation to photosynthetically active radiation (PAR, µmol m–2 s–1) in three selected plots at the fen site. The curve for the calibration year (1994) is marked with dashed line. For a description of the plots, see Table 2.

Table 6.  The mean potential (PAR > 500 µmol m–2 s–1, from June to August) rate of net CO2 exchange, PN (mg m–2 h–1) for each plot and year. Those rates of PN after rewetting (years 1995 and 1996) which differ significantly (P < 0·005, Tukey's HSD pairwise comparison) from those before rewetting (year 1994) are shown in bold. For a description of the plots, see Table 2
  1. –, missing data.

 F7– – – 59058126006310
 F8– – – 3405212900648
Figure 5.

Rates of net CO2 exchange (PN) in relation to photosynthetically active radiation (PAR, µmol m–2 s–1) in three selected plots at the bog site. The curve for the calibration year (1994) is marked with a dashed line. For a description of the plots, see Table 2.

Seasonal estimates for co2-c balance

The seasonal CO2-C balances for the three plots at the fen site were positive, being 283 g m–2 in plot FR1, 162 g m–2 in plot FR3, and 60 g m–2 in untreated plot F7. The results for the untreated plot F7 were calculated for non-shadowed conditions because we did not make continuous measurements of irradiance under the canopy of the pine stand. Assuming that the quotient of irradiance under the canopy of the pine stand to that in the open mire is 0·5 (based on the observations in 1994, see also Grace & Marks 1978) and a similar decrease in the rate of PG, the seasonal CO2-C balance in plot F7 would be strongly negative (–183 g m–2). The expansion of Eriophorum vaginatum together with a high water table clearly explains the relatively high positive CO2-C balance in plot FR1 (Figs 1 & 6). The seasonal CO2-C balance at the bog site was clearly positive in the rewetted area, being 101 g m–2 in plot BR1, 54 g m–2 in plot BR4 and close to zero in the untreated plot B6 (3 g m–2). The water table level had the strongest effect on the seasonal CO2-C balance at the bog site through RTOT change; the deep water table increased RTOT in the untreated plot (B6) and the hummock plot (BR4) while the high water table decreased it in the rewetted plot (BR1) (Figs 1 & 6).

Figure 6.

Monthly CO2-C balance of selected plots at both study sites in 1995. Dashed lines in the columns indicate the monthly balance; positive values show gross CO2 assimilation (PG) and negative values show total respiration (RTOT). For a description of the plots, see Table 2.


The measurement of CO2 fluxes in field conditions is susceptible to many possible sources of error. For instance, the equipment we used was not very reliable in detecting small fluxes due to a large head space volume in the chamber. This was a particular problem when measuring the CO2 fluxes of the non-vegetated plots. A systematic error of 1·5 p.p.m. in the measurement of the CO2 concentration of the head space air at 30 s intervals would result in an error of 100 ± 23 mg CO2 m–2 h–1 in the CO2 flux.

The continuous presence of the measuring collars around the plots during the 3-year study period may have brought about changes in the thermal and hydrological properties of the peat within. However, collars have been widely used in field gas flux measurements (e.g. Crill et al. 1988; Alm et al. 1997), and there are few alternatives.

Summing up the couplets of CO2 flux measurements, PN (in the light) and RTOT (in the dark), gives biased estimates of gross CO2 assimilation (PG) in daylight due to photorespiration and post-illumination CO2 bursts (Coombs 1987). However, this procedure was needed in order to obtain separate statistical functions for the PG and RTOT (see also Alm et al. 1997).

The above factors may not cause a serious bias in our results, however, since we deal with differences between plots and years affected by similar experimental design. In addition, the observed CO2 fluxes accord rather well with those presented in other studies concerned with carbon cycling in natural and drained peatlands (e.g. Silvola & Hanski 1979; Moore & Knowles 1989; Freeman, Lock & Reynolds 1993; Moore & Dalva 1993; Alm et al. 1997).

Vegetation changes

The peatlands in our study differ from those restored in Central Europe and Canada. They still have an acrotelm and catotelm, and mire plants and their propagules are found either on the site or nearby. That is why we can rely on the natural regeneration of mire plants without the need for seeding or planting (cf. Rochefort, Quinty & Campeau 1997; Sliva, Maas & Pfadenhauer 1997).

The rewetting treatment had a clear effect on vegetation at both sites but the change was much greater at the more nutrient-rich fen site, indicating that the speed of vegetation succession depends on the nutrient supply at the site. This phenomenon seems to be similar to the vegetation changes after drainage, i.e. post-drainage vegetation change is faster at nutrient-rich sites than at nutrient-poor sites (Laine, Vasander & Laiho 1995). The clearest change in the composition of the ground vegetation after the removal of the tree stand and rewetting at the fen site was the large increase in the cover of Eriophorum vaginatum (Fig. 2). This species has a wide ecological amplitude, ranging from arctic tundra (Chapin, Van Cleve & Chapin 1979) to temperate blanket bogs (Wein 1973), and from pristine mires to harvested cut-away peatlands (Salonen 1990; Grosvernier, Matthey & Buttler 1995; Tuittila & Komulainen 1995). Owing to efficient vegetative reproduction, E. vaginatum may rapidly utilize additional resources, such as added nutrients (Tamm 1954; Vasander 1982; Shaver & Chapin 1986; Shaver, Chapin & Gardner 1986; Anderson, Pyatt & White 1995), light (Fetcher 1985), elevated CO2 (Tissue & Oechel 1987; Grulke et al. 1990) and altered competition (Wein & Bliss 1973).

In the moss layer, Sphagnum angustifolium cover increased, especially in plots where all vascular plants had been removed. Sphagnum angustifolium occurs as a dominant moss layer species in a variety of pristine pine mires and it has a relatively large amplitude in relation to water level, as found in studies concerning vegetation succession after water table drawdown (Laine, Vasander & Laiho 1995). The rapid increase may be explained by the fact that it was already found in all plots before rewetting.

At the bog site after 30 years of drainage succession, microtopographical variation in vegetation typical to natural bogs (e.g. Backéus 1972; Vasander 1984; Foster & Glaser 1986) could still be seen. However, hummock species were found to be growing on the lawn and hollow levels. The rewetting treatment increased the cover of hollow and lawn species in the wetter microhabitats and hummock species started to degenerate.

Co2 efflux

In the context of predicted global climate warming, a great deal of research has recently concentrated on the effect of water table drawdown on CO2 exchange from peatlands, mainly on CO2 emissions. Laboratory and field experiments have shown that the depth of the water table and peat nutrient concentrations are the key factors controlling aerobic microbial decomposition and consequent CO2 emissions (Glenn, Heyes & Moore 1993; Moore & Dalva 1993; Silvola et al. 1996a). After water table drawdown, emissions have been reported to increase 2·5-fold (Silvola, Välijoki & Aaltonen 1985) and even 3·4-fold (Moore & Knowles 1989), depending on the effectiveness of the drainage and nutrient status. Our results show that this process can be reversed (CO2 fluxes from peat decrease) when drained peatlands are rewetted. At the fen site, CO2 efflux from the soil surface still decreased after the first year of rewetting. This was probably caused by a decrease in soil respiration resulting from death and gradual decay of roots isolated from the surrounding root system by the collars (McClaugherty, Aber & Melillo 1984; Silvola et al. 1996b). Similar results of gradually decomposing roots have been reported by Finér & Laine (1996a).

The difference in CO2 efflux from the soil surface (RS) of non-vegetated plots and total CO2 efflux (RTOT) from vegetated plots in our study suggest that most of the CO2 released (50–70%) is derived from recent photosynthates in the respiration of leaves and roots and microbial respiration using root exudates. The greater use of recently fixed carbon in respiration was particularly clear in the case of the rewetted plots with high water levels and an increased cover of vascular plants. Our results indicate a higher proportion of recent photosynthates in CO2 cycling than the results of Silvola, Alm & Ahlholm (1992) and Silvola et al. (1996b), who suggested that 55–65% of CO2 fluxes originate from so-called ‘old carbon’. However, other studies have shown the importance of ‘younger carbon’ (Charman, Aravena & Warner 1994; Minkkinen & Laine 1998; Domisch et al. 1998), and particularly the importance of fine roots and root exudates in the carbon cycling of pristine and drained peatlands (Finér & Laine 1996b; Saarinen 1996).

Gross co2 assimilation and net co2 exchange

The increase in cottongrass cover at the fen site was clearly reflected in the gross CO2 assimilation (PG) as well as in the net CO2 exchange (PN). The efficiency of Eriophorum vaginatum in productivity and sequestering of CO2 is due to its growth form (Chapin, Van Cleve & Chapin 1979) and special characteristics of its leaves; there is an intensive lacunar system throughout the mesophyll for efficient CO2 diffusion and uniform distribution of chlorophyll (Kummerow et al. 1988). It can quickly grow new assimilating leaf biomass in early summer and leaf biomass continues to grow until late autumn (Robertson & Woolhouse 1984). Our results showed a similar seasonal trend, i.e. the rate of PG started to rise quite early in summer 1995, being almost twice as high in June as measured in 1994, and the rates still continued to rise until the late summer (Fig. 3).

Following removal of Eriophorum vaginatum and other vascular plants the rates of PG and PN decreased, reflecting the importance of these species in carbon binding. In these plots (FR1,FR6) where Sphagnum angustifolium became dominant after removal of the field layer, the seasonal variation in the rate of PG was small despite the water table fluctuating between 0 and 20 cm. These results are in accordance with those of Silvola & Aaltonen (1984), who reported that the relative photosynthetic rate of S. angustifolium markedly decreased only after the water content had dropped below 600% or rose above 2000% of dry mass. The rates of S. angustifolium photosynthesis measured in our plots were near the optimum reported by Silvola & Aaltonen (1984). In such relatively dry conditions, before rewetting and clear-cutting, the PG of field and moss layer were lower and PN negative in plot FR1 (Table 5 and 6), and the tree stand dominated carbon binding (Laiho & Laine 1997).

In a pristine bog, biomass and production have been found to be highest on hummocks and lowest in hollows (Vasander 1982, 1984; Backéus 1985). Our potential PG results for the bog site appeared to follow the same pattern; that is, the rate of PG clearly decreased with the gradual dying-off of hummock species in hollows due to excess moisture after rewetting. The increase in the cover of hollow species could not compensate for this effect. Simultaneously, on slightly drier hummocks, both the cover and PG of dwarf shrubs increased. This difference between hummocks and hollows is mainly due to the relatively high photosynthetic efficiency of vascular plants (e.g. Grace & Marks 1978; Silvola & Hanski 1979), particularly shrubs, which in pristine bogs are mainly concentrated on hummocks because of higher oxygen availability. On lawns, where the vegetation mosaic is formed by lichens and Sphagnum mosses, the rates of PG did not change after rewetting, because of the different photosynthetic responses of these species to changing water content (Kallio & Kärenlampi 1975; Silvola & Aaltonen 1984). Presumably, the decrease in the PG of lichens was compensated for by an increase in PG for Sphagna.

Effect of restoration on the seasonal co2 balance

The changes in the seasonal CO2-C balance of the plots with dense Eriophorum vaginatum cover at the fen site were already clearly seen 1 year after treatment, and E. vaginatum seemed to compensate for the lost carbon fixation capacity of the former tree stand. The seasonal CO2-C balances at the rewetted fen site (162–283 g m–2) were slightly higher than the values reported from a pristine pine fen (Alm et al. 1997). According to Alm et al. (1997), the CO2 balance integrated over the growing season varied from 108 to 160 g m–2, depending on the vegetation composition and microtopography of the measurement site.

At the bog site, the seasonal CO2-C balance (54–101 g m–2) was of similar magnitude and showed similar seasonal trends (Fig. 6) to those reported by Silvola & Hanski (1979) in a laboratory experiment. Despite higher rates of gross primary production for the hummock of the bog site, the seasonal CO2 balance was greatest in a rewetted hollow level plot with high water table. This suggests that the carbon balance at the bog site was highly dependent on decomposition, which was affected by water table dynamics. This explanation is consistent with the results of Alm et al. (1998), who reported relatively high (10–117 g C m) net losses of carbon for a pristine bog with a deep water table and high rates of decomposition during a very dry summer.

Both the vegetation and carbon balance at the sites have already approached those of pristine mires with corresponding nutrient levels during the first years after restoration. This is encouraging from the practical point of view as drained peatlands are now being restored in Finland, both in areas of nature conservation and as buffer zones between waterbodies and drained peatlands to diminish detrimental impacts of forestry operations. However, further research is needed to ascertain the differences between restored fens and bogs in terms of carbon accumulation efficiency.


This study was part of the Finnish Research Programme on Climate Change and was financed by the Academy of Finland, the University of Helsinki and The Finnish Society of Forest Science. We thank those who provided facilities at the Hyytiälä Forestry Field Station of the University of Helsinki; Jouni Meronen and Hannu Nykänen for help with the gas exchange measurements; and Dr Raija Laiho and Dr Michael Starr for advice about the manuscript.