References

Our research investigates the spatial and temporal variability of methane (CH4) emissions in two drained eutrophic peat areas (one intensively managed and the other less intensively managed) and the correlation between CH4 emissions and soil temperature, air temperature, soil moisture content and water table. We stratified the landscape into landscape elements that represent different conditions in terms of topography and therefore differ in moisture conditions. There was great spatial variability in the fluxes in both areas; the ditches and ditch edges (together 27% of the landscape) were methane hotspots whereas the dry fields had the smallest fluxes. In the intensively managed site the fluxes were significantly higher by comparison with the less intensively managed site. In all the landscape element elements the best explanatory variable for CH4 emission was temperature. Neither soil moisture content nor water table correlated significantly with CH4 emissions, except in April, where soil moisture was the best explanatory variable.


Introduction
It is of great importance to assess the contribution of the trace gas methane (CH 4 ) to the greenhouse gas effect and the related global warming. Northern peatlands are believed to be significant sources of CH 4 , estimated to emit between 20 and 50 Tg yr −1 (Mikaloff Fletcher et al., 2004a, 2004b. In northern oligotrophic and eutrophic managed peatland systems, net uptake and emission rates have been found to depend In the Netherlands, eutrophic peatlands have been drained for centuries and in the last 50 years peatlands have been drained more deeply to make agriculture possible, resulting in peat oxidation. These peatlands are therefore major carbon sources of CO 2 (Langeveld et al., 1997;Schothorst, 1977;Veenendaal et al., 2007). Burgerhart (2001) and Van den Bos (2003) have suggested that peat oxidation can be reduced if agri-5 cultural peatlands are transformed into wetland nature by raising the water table and by reducing agricultural intensity, thus altering the carbon cycle and probably turning sources into sinks. There is great uncertainty, however, about the impact of such measures on the CH 4 balance. Hendriks et al. (2007) found that in an area in the centre of the Netherlands where intensive farming had ceased 14 years previously and the 10 water table had risen, a very small sink of 71 g CO 2 -equiv m −2 yr −1 had developed. They attributed this to a decrease in CO 2 emissions and an increase in CH 4 emissions from ditches and waterlogged soil however, they had no data from the pre restoration situation.
We investigated spatial and temporal variability of CH 4 emissions in two drained peat 15 areas -one intensively managed and the other less intensively managed -and examined the correlation between CH 4 emissions and soil temperature, air temperature, soil moisture content, water table and management over almost two years. We monitored CH 4 flux measurements at discrete points within landscape elements representing different micro topographical conditions. We aimed to determine the mean flux associated 20 with the landscape element, and to provide a spatially integrated flux measurement for the study areas. We compared the annual CH 4 emission balances of both areas.

Site description
The experimental sites (Oukoop, intensively managed dairy farm and Stein, less in-Introduction  (Fig. 1). The climate is temperate and humid, with mean annual precipitation of about 800 mm and an annual long-term mean temperature of 9.8 • C. Nol et al. (2008) calculated that 21% of the polder is open water (ditches and small permanent pools) 6% is ditch edges (waterlogged land bordering the ditches), <2% is drainage trenches (located in the middle of the field, con-5 taining water in winter) and >71% is drier land with a fluctuating water table. The soils consist of a clayey peat or peaty clay top layer of 25 cm overlying 12 m eutrophic peat deposits. The polder is below sea level: its mean elevation is between 1.6 and 1.8 m below the Amsterdam Ordnance Datum (NAP). The depth to the groundwater varies from 70 to 15 cm; perched water tables occur after heavy rain, when the soil impedes 10 water infiltration. Both sites have been described in detail by Veenendaal et al. (2007). As Stein has become a bird reserve, its management is less intensive than previously. Intensive dairy farming on the two experimental parcels of land here was stopped more than 20 years ago. During the measurement period, both parcels were used as hayfields; they were mown three times after 15  From February 2006until September 2007 additional sampling points, distributed over two different land parcels per location, were sampled at both research locations to study spatial variability. We stratified both locations into four landscape elements with differing soil/water temperature and soil moisture conditions: permanently water-filled ditches, ditch edges, narrow drainage lines or trenches in the middle of the fields, and the field area with fluctuating water table (henceforth referred to as " field"). In each of the two fields, there were four sample points in the ditches, four points in the ditch edges, two or three sample points in the drainage trenches and eight or nine sample points in the fields. 15 Fluxes of CH 4 were determined using a modified closed chamber method (Hutchinson and Mosier, 1981). Gas flux was measured using a Photo Acoustic Field Gas Monitor (INNOVA 1412 sn, 710-113, ENMO services, Belgium) connected by Teflon tubes to a PVC chamber (Van Huissteden et al., 2005). Samples were taken from the headspace of the closed, dark chamber (30 cm diameter, 25 cm height) that was placed on a collar. 20 A small fan was installed in the chamber to homogenize the inside air and a water lock was placed to control inside pressure. On land we used water between the chamber and the collar to seal the chamber from the ambient air during the measurement. At the ditches we used floaters and a lever system to gently lower the chamber onto the water surface, carefully avoiding the effect of pressure differences. We used external 25 silica gel and soda lime filters to minimize cross-interference of CO 2 and water vapour with methane at high concentrations. Our gas analyser was calibrated and tested for Introduction drift every year at the NMI (Nederlands Meet Instituut: Delft, The Institution of Standards). Occasional cross-checks with a standard calibration gas established that the instrument did not drift. All measurements were taken during the day, between 8 a.m. and 4 p.m. Each flux measurement consisted of five point-measurements taken at oneminute intervals.

5
In addition to each flux measurement, soil or water temperature was measured at 10 cm depth and soil moisture content was determined in the top 5 cm of soil at the sample points, using a HH2 Delta-T device (Delta T Devices, Llandindrod Wells) calibrated for the soil type. The water table was recorded with pressure sensors installed in a steel frame to a depth of 70 cm into the soil at one or two places in the field (e+ 10 sensor L-50, Eijkelkamp Agrisearch Equipment BV, Giesbeek, Netherlands). Water levels were logged hourly. Any gaps in the data were filled with average values from other sensors in the surrounding area.

Calculations and statistical analyses
CH 4 fluxes were calculated using linear regression of the changes in concentration 15 over time, because the closure times of the chambers were short. First, the data quality was assessed: outliers resulting from disturbances, chamber leakage or instrument failures were removed from the data set. Annual mean net CH 4 emissions were estimated by trapezoidal integration of mean net CH 4 emission over time (design-based approach, Van den Pol-Van Dasselaar and Oenema, 1997;Velthof et al., 1996) and by 20 linear regression of natural logarithm-transformed CH 4 data (model-based approach, Hendriks et al., 2007).
The statistical significance of differences between landscape elements within sites was calculated with one-way ANOVA; analysis of covariance, with temperature as covariate, was used to ascertain the statistical significance of differences in the fluxes 25 from the landscape elements of the two sites. Correlations between natural logarithmtransformed emissions and independent variables were calculated using step-wise multiple linear regression analysis (case-wise elimination of variables). Statistical 5,2008 Methane emissions in drained peat agro-ecosystems For the calculations of the contribution of animals and manure on the intensive dairy farm, we used the method with simple emission factors as described by Hensen et al. (2005). Though the most recent IPCC report (IPCC, 2007) mentions a global warming potential (GWP) of 25 to convert CH 4 emissions to CO 2 equiva-5 lents, we used a (GWP) of 23 for CH 4 , at a 100-year time horizon (IPCC 2001, UN-FCCC/CP/1997/7/Add.1/Decision2/CP.3, e.g. Lashof et al., 2000) to allow comparisons to be made with other research.

Seasonal and spatial variation of methane fluxes
We compared the landscape elements for the period that measurements of emissions from all landscape elements ran parallel (January 2006 to September 2007). The average soil moisture content and air temperature during this period are shown in Fig. 2: highest temperatures were in July and August and lowest temperatures in December and January. 15 Emission rates varied greatly, depending on the time of the year. For instance, at both sites in 2006 and 2007, ditch emission rates were highest in June, July, August and September; field emission rates were highest in March, April, May and June (Fig. 3). We also observed that in February 2006 emission rates were slightly higher from the intensively managed site after thaw and manure application. There was a 20 seasonal effect: over 85% of the total annual CH 4 emissions were observed in summer and maximum emission rates in ditches were ten times those of the fields. Ditches showed episodic, exceptionally high emission values: for example in 2006 on 27 and 28 September: 366.05 (n=6) and 123.8 (n=4) mg m −2 hr −1 for the intensively managed and less intensively managed areas, respectively. During these measurements we ob- Spatially, the fluxes from edges and ditches differed significantly between the two sites: in the intensively managed site they were up to 1.5 and 3 times higher, respectively ( Fig. 3). At both sites, the CH 4 fluxes from the ditches (21% of the area) 5 were significantly greater than those from the fields, ditch edges and drainage trenches (P <0.01). At both sites, the fluxes from the ditch edges (6 % of the landscape) were significantly greater than those from the fields (P <0.01). The lowest emissions were from the fields (>71% of the landscape surface area). The emission rates of drainage trenches (<2% of the landscape) did not differ significantly from the emission rates of 10 fields and ditch edges.
Overall, the correlations of CH 4 emissions with soil moisture were weak in both the intensively and the less intensively managed sites. Adding moisture in a stepwise re- 2007, when soil moisture appeared to be a stronger predictor for the emission rates at both sites (Fig. 5). The water table in both sites fluctuated greatly during the measurement period: it was high (−15 cm) in winter and low (−65 cm) in summer. The water table did not show any significant correlation with CH 4 emissions from the fields.

5
The CH 4 emissions from ditch edges correlated significantly with soil temperature, but the best explanatory variable was air temperature (Fig. 4). The correlation between emission rates and soil moisture in the edges of our fields was not significant, except for April in the less intensively managed site where again a significant positive correlation occurred (r 2 =0.862; P <0.05; n=7). 10

Annual methane balances
For the estimation of annual terrestrial CH 4 balances we used 2 methods: (1) trapezoidal integration over time (for 2005, 2006 and 2007) and (2) linear regression with temperature as explanatory variable (for 2006) ( Table 1). The regression-based estimates are based on hourly air temperature data. Fluxes were estimated per landscape 15 element and multiplied by the area occupied by the landscape element, providing a spatially integrated flux measurement. Emissions from drainage trenches were not used in the calculations of the annual CH 4 balance because their contribution is negligible. Comparison of the fluxes estimated by these two methods sometimes revealed large differences. 20 In the fields, integrated annual fluxes based on the daytime measurements were higher but of comparable magnitude to the estimates from regression. However in the ditches and ditch edges, large episodic venting events, in the ditches mainly caused by turbulent water, caused averages to be up to 20 times higher. Even when these are excluded, the mean values are much higher than the values estimated from regres-25 sion. Our exclusive use of daytime measurements might have resulted in overestimation of methane fluxes. We estimated regression-based methane fluxes for two days and found that when using data between 12 p.m. and 4 p.m. only, the estimated daily 1245 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version Interactive Discussion methane fluxes were higher: by up to 45% on a winter day and up to 50% on a summer day. Regression-based estimated methane fluxes at different times on 21 December and 4 July are given in Fig. 6.

Field methane emissions compared to farm-based emissions
It is interesting to compare measured field methane emissions with farm-based 5 emissions. We estimated farm fluxes using emission factors for dairy cows (E d ), heifers (E y ), calves (E c ), manure (E s ) and farmyard manure (FYM) (274, 170, 48, 53 and 40 g CH 4 day −1 animal −1 or m −3 , respectively) as described by Hensen et al. (2005). Using these factors, emission (Q) from the farm was calculated as:  landscape) turned out to be CH 4 hotspots. As Table 1 shows, linear regression revealed that in the intensively managed site they emitted 6.3 and 4.4 times more CH 4 , respectively than the dry field; this compares with 1.7 and 2.5 times more in the less intensively managed site.
BGD 5,2008 Methane emissions in drained peat agro-ecosystems The CH 4 emission rates from the permanently water-filled, 30-60 cm deep ditches could be predicted by water temperature or air temperature. The exceptionally high emission peaks (up to 800 mg m −2 hr −1 in the summer) may have resulted from ebullition events, during which CH 4 quickly passes through the top layer in the water column where oxidation can take place, but they probably mainly occurred when both wa-5 ter temperature and wind velocity were high. In a peatland study in Central Finland, Minkkinen and Laine (2006) found that the methane fluxes from ditches with flowing water were generally higher than from ditches with standing water. They argued that the diffusion rates were higher in flowing water because when water is turbulent the boundary layer is thinner. The contribution of very high fluxes to the annual balance may lead to overestimation in the integration-based approach. Even when using a regressionbased approach, annual averages of methane emission must be presented with some caution. Our comparison of trapezoid integrated calculated means and the regressionbased temperature-dependent calculations highlights the uncertainty in the means: for example, all the flux measurements were performed during the day, so when trapezoid 15 calculations were used, the fluxes may have been overestimated due to diurnal temperature changes when integrating over time (cf. Mikkelä et al., 1993;Chanton et al., 1993). It will be recalled that when using data from 12 p.m. to 4 p.m. only, we found that the daily methane fluxes were 45% higher on 21 December (winter) and 50% higher on 4 July (summer). 20 Our results demonstrate that field emissions are an exponential function of soil and air temperature: the regression-based fluxes in the intensively managed site were statistically significantly higher in 2006 and 2007 despite varying soil moisture contents. The seasonal distribution of emission rates varied between sites. In the less intensively managed site, high CH 4 fluxes were concentrated in the summer period, while in the 25 intensively managed site they were concentrated in early spring and summer, partly associated with field applications of slurry. Van den Pol-Van Dasselaar et al. (1999) also reported higher CH 4 emissions after manure application (but the increase they found was not statistically significant). We would argue that the reason for the enhanced CH 4 BGD 5,2008 Methane emissions in drained peat agro-ecosystems

Printer-friendly Version
Interactive Discussion production is the combination of wet soil, the application of easily decomposable organic material and the anaerobic conditions in the slurry. The only month for which we found significant correlation between soil moisture and CH 4 emission rates was Aprilthe period when the field begins to dry out after being waterlogged in winter and when air temperature may rise rapidly from 10 to 25 • C. The correlation between depth to 5 water table (65 to 15 cm) and CH 4 fluxes at both sites was not significant: the highest fluxes occurred at intermediate and sometimes high water tables. In both sites the water table fluctuated not only seasonally, but also because of hydraulic regulation by the Dutch water board. In both sites, the impermeable soil prevented water from infiltrating after heavy rain, resulting in perched water tables in winter. The large variation in the 10 water table could explain the weak correlation between water table and CH 4 emissions. Ditch-edge CH 4 fluxes correlated significantly with soil and air temperature, but not with soil moisture content except, as in the fields, for April (positive correlation). At both sites the edge fluxes were significantly higher than the field fluxes. The edges border the ditches and so were damp for most of the year, with soil moisture contents >60%. 15 These damp conditions give rise to a different vegetation than in the dry fields: in some places Iris pseudacorus and Typha angustifolia are present. These aerenchymatic plants might cause additional fluxes because CH 4 diffuses rapidly through their stems.

Methane fluxes in other peatland ecosystems
Our measurements showed that both our drained sites are a net source of CH 4 the 20 annual regression-based means were 258 and 114 kg ha −1 for the intensively managed site and less intensively managed site, respectively. These values are in the same order of magnitude as fluxes found in other managed and unmanaged peatland ecosystems (Table 3). Van den Pol-Van Dasselaar et al. (1998a), who studied CH 4 emissions in grassland on peat soils in a nature reserve elsewhere in the Netherlands, 25 reported great spatial variability; though the emission values in fields were similar to ours, the emission rates they found in saturated land were higher.
It is particularly interesting to compare our results with those reported by Hendriks  (2007) for a site some 35 km from our research area. They studied the full greenhouse gas balance of a peatland on comparable soils that had not been intensively farmed for >10 years but left unmanaged, with a high water table. Their estimated total weighted (per landscape element), regression-based annual methane flux of 417 kg ha −1 , was higher than both our estimated, weighted regression-based annual 5 methane fluxes: 258 and 114 kg ha −1 for the intensively and less intensively managed sites, respectively. The estimated annual emission rate of 18.72 mg m −2 hr −1 in ditches reported by Minkkinen and Laine (2006) is also higher than the regression-based fluxes from our ditches. The extremely high emission rates from ditches found by Minkkinen and Laine (2006) and by Bubier et al. (1993) were similar to the extreme values we found at turbulent water conditions. However, when we took account of the farm-based emissions (which are ∼64% of total emissions), our estimated annual methane flux was 725 kg ha −1 in the intensively managed site. Compared with the values found in formerly intensively farmed peatland (Hendriks et al., 2007), this estimated annual methane flux is ∼42% lower, whereas the fluxes from our less intensively managed site 15 are ∼84% lower. For the total balance (i.e. CO 2 plus CH 4 ) we used data from Veenendaal et al. (2007) who performed CO 2 eddy correlation measurements in our intensively and less intensively managed sites. The intensively managed site had a net annual CO 2 emission of 122 g C m −2 , whereas the less intensively managed site had a net annual CO 2 up-20 take of 57.6 g C m −2 . The resulting annual sources obtained when the terrestrial annual methane emissions were included (using global warming potentials of 23 for CH 4 , IPCC 2001) were of 567 and 138 g C m −2 CO 2 eq, respectively. The restored site studied by Hendriks et al. (2007) was found to be a very small sink of 269 g C m −2 . These estimates of the total balance in the intensively and less intensively managed sites do not

Hour of day
Ditch emissions (mg m -2 d -1 ) Ditch emissions (mg m -2 d -1 ) Fig. 6. Diurnal variation in regression-based estimated ditch emissions (mg m −2 d −1 ) based on temperature during a day in winter (red squares, y-axis right), and a day in summer (blue circles, y-axis left).