Short-term effects of thinning , clear-cutting and stump harvesting on methane exchange in a boreal forest

Forest management practices can alter soil conditions, affecting the consumption and production processes that control soil methane (CH4) exchange. We studied the short-term effects of thinning, clear-cutting and stump harvesting on the CH4 exchange between soil and atmosphere at a boreal forest site in central Sweden, using an undisturbed plot as the control. Chambers in combination with a highprecision laser gas analyser were used for continuous measurements. Both the undisturbed plot and the thinned plot were net sinks of CH4, whereas the clear-cut plot and the stump harvested plot were net CH4 sources. The CH4 uptake at the thinned plot was reduced in comparison to the undisturbed plot. The shift from sink to source at the clear-cut and stump harvested plots was probably due to a rise in the water table and an increase in soil moisture, leading to lower gas diffusivity and more reduced conditions, which favour CH4 production by archea. Reduced evapotranspiration after harvesting leads to wetter soils, decreased CH4 consumption and increased CH4 production, and should be accounted for in the CH4 budget of managed forests.


Introduction
Methane (CH 4 ) is the second most important carbon greenhouse gas, with a radiative forcing at least 25 times higher than carbon dioxide from a 100-year perspective (Shindell, et al., 2009).Consumption of CH 4 by methanotrophic bacteria in the aerobic part of the soil profile (Harriss et al., 1982) and production of CH 4 by archaeans in the anaerobic water-saturated part of the profile (Ehhalt, 1974) and at anaerobic micro-sites (von Fischer and Hedin, 2002;Kammann et al., 2009) often occur simultaneously (Le Mer and Roger, 2001;Megonigal and Guenther, 2008).Generally, well-aerated forest soils are net sinks of atmospheric CH4 (Van Amstel, 2012).Consumption in soils is the second largest sink of CH 4 after tropospheric oxidation by hydroxyl radicals with a global sink capacity estimated recently at 28-32 Tg CH 4 yr −1 (Kirschke et al., 2013).The soil sink capacity is higher in forest soils than in grasslands and arable land (Dutaur and Verchot, 2007), and therefore the global CH 4 budget is sensitive to disturbances in forests.Conversion of natural forests to arable land, increased N deposition from the atmosphere and N fertilization of agricultural lands are estimated to have reduced the global CH 4 soil sink by about 30 % between 1880 and 1980 (Ojima et al., 1993).
Disturbances, including forest management practices, can also have an impact on the soil CH 4 exchange by altering soil conditions, such as soil moisture (Zerva and Menucuccini, 2005;Castro et al., 2000), water table depth (Zerva and Menucuccini, 2005), bulk density (Mojeremane et al., 2012), soil temperature (Zerva and Menucuccini, 2005;Thibodeau et al., 2000), nutrient content (Smolander et al., 1998) and pH (Smolander et al., 1998).CH 4 oxidation in soil has been observed to be controlled by diffusivity (Koschorreck and Conrad, 1993;Whalen and Reeburgh, 1996;Gulledge and Schimel, 1998).A well-drained coarse soil facilitates the exchange of oxygen and CH 4 between the atmosphere and the deeper soil levels, where CH 4 is consumed (Verchot et al., 2000).By contrast, increased soil moisture and soil compaction reduce the diffusivity and promote anoxic environments in which CH 4 can be produced (Koschorreck and Published by Copernicus Publications on behalf of the European Geosciences Union.E. Sundqvist et al.: Short-term effects of thinning, clear-cutting and stump harvesting Conrad, 1993;Whalen and Reeburgh, 1996;Gulledge and Schimel, 1998).Changes in water table depth also influence the CH 4 exchange by altering the relative extent of anaerobic and aerobic zones in the soil (Whalen and Reeburgh, 1990).Temperature is also an important driver of CH 4 production, with higher temperatures leading to higher CH 4 production, while consumption by methanotrophs is not as strongly enhanced (Dunfield et al., 1993).Increased nitrogen content in the soil has been shown to inhibit CH 4 consumption in several studies (Steudler et al., 1989;Hutsch et al., 1993;Wang and Ineson, 2003).This is due to competition by certain nitrifiers, which might occupy the same niche in the soil.These nitrifiers have an enzyme similar to methanotrophs and are also able to oxidize CH 4 , though possibly at a lower rate (Hutsch et al., 1993).
Summarizing the effects of forest management practices on CH 4 exchange is difficult, since relatively few studies have been made on this topic, and they have covered a range of management practices, soil types and forests.However, several studies reported that clear-cutting led to reduced CH 4 uptake, possibly due to increased soil moisture (Wu et al., 2011), increased nitrogen availability (Steudler et al., 1991;Bradford et al., 2000), changes in pH (Bradford et al., 2000) and erosion (Kagotani et al., 2001).A shift from soil CH 4 sink to soil CH 4 source has been reported due to a rise in water table depth combined with increases in substrate availability (Zerva and Mencuccini, 2005) and due to increases in soil moisture (Castro et al., 2000).The same shift from sink towards emission has been seen following soil compaction by skid trails and machinery, as a part of clear-cutting (Teepe et al., 2004) and thinning (Keller et al., 2005).One study on a clear-cut drained peat soil showed no substantial changes in CH 4 exchange (Huttunen et al., 2003).
Site preparation by mounding at clear cuts can have a negative impact on CH 4 exchange from a climate perspective.In one study, compaction of the soil by excavators during mounding increased CH 4 emissions (Mojeremane et al., 2012).CH 4 emissions from stagnant water in hollows created during mounding can sometimes exceed the consumption in the mineral soil on top of the mounds (Mojeremane et al., 2010).However, bedding after clear-cutting has resulted in reduced CH 4 emissions (Castro et al., 2000).Drainage can also reduce CH 4 emissions following clear-cutting, but its positive effect on CH 4 emissions was outweighed by increases in CO 2 emissions when drainage was conducted on saturated peaty soils (Mojeremane et al., 2012).
Stump harvesting for bioenergy production has recently been proposed as a way of substituting fossil fuel CO 2 emissions in Sweden.To our knowledge, there are no publications on the effects of stump harvesting on CH 4 exchange, although it is likely to have a similar effect to other clearcutting and site preparation actions.There are a few studies on the effect of thinning on CH 4 exchange in a forest.Reduced CH 4 uptake due to increased nitrogen availability has been reported (Thibodeau et al., 2000).A study at three thinned plots in a temperate beech forest reported slightly reduced emissions at one plot, whereas the other two were not significantly different from the control plots (Dannenmann et al., 2007).Another study in a temperate forest actually showed an increased CH 4 uptake after thinning, as opposed to a decrease at two adjacent clear-cut areas (Bradford et al., 2000).Some studies found no significant changes in CH 4 exchange after thinning (Wu et al., 2011;Sullivan et al., 2008).
The objective of this study was to quantify the short-term CH 4 exchange at four sites: an undisturbed forest plot, a thinned forest plot, a clear-cut plot with stumps remaining and a clear-cut plot with stumps removed.The comparison between the different treatments is facilitated because all four sites are within a defined area and have a common soil type.We also wanted to investigate how soil moisture, soil temperature and water table depths influenced the soil CH 4 exchange.

Site description
The CH 4 exchange measurements took place in a forested area on the southern edge of the boreal zoon, at Norunda research station in central Sweden (60 • 05 N, 17 • 29 E).Hourly automated chamber measurements were made using a system that was moved between four differently managed plots (Fig. 1).One plot contained undisturbed 120-year-old mixed pine (Pinus sylvestris) and spruce (Picea abies) forest, which had not been thinned or fertilized in several decades.The other three plots were recently (2008)(2009)(2010) impacted by either thinning, clear-cutting or stump harvesting.Thinning was done in order to simulate continuous cover forestry, rather than to increase growth.
Measurements were made using four chambers at the thinned plot, and five chambers at each of the other plots.The chamber locations were named U 1 -U 5 at the undisturbed plot, T 1 -T 4 at the thinned plot, C 1 -C 5 at the clear-cut plot and S 1 -S 5 at the stump harvested plot.At the clear-cut and stump harvested plots, half of the chamber frames were positioned on bare soil, where organic and mineral soil layers were mixed.The disturbance was caused either by stump harvesting or by site preparation to facilitate the establishment and growth of new plants.The remaining frames were placed on soil surfaces with intact vegetation.The clear-cut and stump harvested plots had been fertilized in 1976, 1988 and 1998.Ground vegetation was sparse and dominated by bilberry (Vaccinium myrtillus) and feather mosses (Hylocomium splendens and Pleurozium schreberi).There were more shrubs and grass at the clear-cut site, following the soil's disturbance.The soil was a glacial till (Lundin et al., 1999) with an organic layer of 3-10 cm depth.For the period 1980-2010, the mean air temperature was 6.5 • C, and the mean annual precipitation was 576 mm (measured 30 km south of Norunda).

Timing of measurements
Thinning took place in November 2008, the clear-cutting in February 2009 and stump harvesting in May 2010.Both the clear-cut plot and the stump harvested plot were mounded and planted in May 2010.The chamber frames were installed in 2005 at the undisturbed and thinned plots, and in June 2010 at the clear-cut and stump harvested plots, to allow time for soil and vegetation to recover from the disturbance.
Due to equipment limitations, measurements were conducted at one plot at a time.Measurements at the thinned plot were made from 1 August 2009 to 31 May 2010, at the undisturbed plot from 7 July 2010 to 4 October 2010, at the stump harvested plot from 7 October 2010 to 20 October 2010 and at the clear-cut plot from 21 October to 9 November 2010.Winter data at the thinned plot from 1 December 2009 to 14 April 2010 were not used in the analyses due to uncertainties in the measurements caused by snow and frost.

Equipment
We used automated, transparent chambers of Polymethyl methacrylate in combination with a high-precision off-axis integrated cavity output spectroscopy (ICOS) laser gas analyser -DLT-100; Los Gatos Research (LGR) -for simultaneous concentration measurements of CH 4 , CO 2 and H 2 O.The chambers had a volume of 110 litres and covered a surface area of 0.2 m 2 .Gas concentrations in the chambers were measured after closure by recirculating the air through the gas analyser for 6 min.The flow rate between chambers and manifolds was 8-10 L min −1 .This air stream was subsampled and passed through the analyser at a flow rate of 1.2 L min −1 .A fan was installed in each chamber, designed to ensure sufficient mixing of chamber headspace air without disturbing the laminar boundary layer on the ground.Soil moisture was measured in the chambers at a 0-5 cm depth with a Ml-2× ThetaProbe from Delta-T Devices.The soil temperature was measured at a 5 cm depth inside the chambers using a type T thermocouple.Soil temperature measurements at the thinned plot did not work properly and so temperature data from the undisturbed plot, 125 m away, were used instead.

Water table
There were differences in height between the chamber frames relative to the groundwater table.One pipe with continuous measurements of the groundwater table was located 125 m from the thinned plot and 30 m from the undisturbed plot (Fig. 1).The groundwater table at these plots was treated as horizontal.At the clear-cut and stump harvested plots, the groundwater table was measured manually in seven pipes at each plot, on the 8 and 20 October 2010 and 2 November 2010.Some of these pipes are shown in Fig. 1.An inverse distance-weighting model was used to calculate the height of the groundwater table in relation to the ground surface for 40 m 2 areas surrounding the chamber frames.The groundwater table was also measured continuously at one position on the clear-cut plot.

Soil sampling
Soil samples were taken in order to determine organic carbon (C) and nitrogen (N) content and pH in the top 20 cm of the soil, including the humus layer, where the chambers had been positioned.The litter layer was not taken into account.Sampling was done in November 2010 at the clear-cut and stump harvested plots and in September 2012 at the undisturbed and thinned plots.
At the clear-cut and stump harvested plots, humus layer samples were taken, at the border between organic and mineral soil layers, using a 10 cm × 10 cm quadratic frame.The mineral soil was sampled with a 15.9 cm 2 steel corer to a depth of 20 cm, but it was subdivided in the field into 0-10 www.biogeosciences.net/11/6095/2014/Biogeosciences, 11, 6095-6105, 2014 and 10-20 cm layers.Humus samples were treated individually, while the mineral soil samples were pooled plot-wise for each soil layer.The samples, folded in plastic bags, were transported in cooling boxes to the laboratory, where they were kept fresh at 4-5 • C during the preparation process before the final analyses.Soil samples were passed through either a 5 mm (humus samples) or a 2 mm (mineral soil) mesh.Stones and gravel > 2 mm diameter not passing the mesh were always rejected, as were any roots.The sieved soil material from each sample was carefully mixed and divided into a number of subsamples for determination of soil pH (H 2 O) and total C and N content.Fresh weight : dry weight ratios were determined after drying the sub-samples at 105 • C for 24 h.Soil layer pH was determined with a glass electrode in the supernatant after shaking for 2 h on a rotary shaker, and sedimentation in an open flask for another 22 h.The proportion of fresh soil to distilled water was 1 : 1 by volume, compared to about 1 : 10 for dry matter : water for humus, and 1 : 2.5 for mineral soil.Total C and N content was determined, using vacuum-dried soil samples at 60 • C for 24 h in a Carlo Erba NA 1500 analyser.Because soil pH was always below 6, we assumed that there was no carbonate C, and all C analysed was assumed to be organic C.
At the undisturbed and thinned plots, a cylindrical metal corer with an 11 cm 2 opening was pressed horizontally into the humus layer, and also at a 5 cm and 10 cm depth in the mineral layer.At some of the measurement locations (T1, T4, U2, U6), large stones, rocks and roots occupied a large volume of the mineral soil, so that sampling at a 10 cm depth in the mineral soil was not possible.The soil samples were kept below 5 • C until they were analysed.
The total amount of C and N in the soil samples at the undisturbed and thinned plots was analysed with an element analyser (Elementar Analysensysteme GmbH, Germany).The pH value was measured after 2 h equilibration with a 0.1 M barium chloride solution (Orion Research model Microprocessor Ionanalyzer/901).The extractions were made on fresh material.Before determining the bulk density, the samples were oven dried for 48 h at 100 • C and then sieved through a 2 mm mesh.

Data analyses
The rate of change of CH 4 concentration (dC CH 4 /dt) within the chamber was calculated using a linear fit to the first 2 min of concentration data measured by the gas analyser, beginning immediately after chamber closure.We calculated the r 2 values for the fits of five different slopes, which were lagged at 10 s intervals after chamber closure.The fit with the highest r 2 value was then selected.The CH 4 flux (J CH 4 flux ) was calculated as J CH 4 flux = dC dt V A , where C is the molar density (µmol m −3 ), V (m 3 ) is the chamber volume and A (m 2 ) is ground surface area.Fluxes with an r 2 value higher than 0.3 were generally kept for further analyses.An r 2 of 0.3 was the limit when the fluxes were significantly different from 0. A few outliers that passed the r 2 limit were visually sorted out based on normalized root mean square error.Data kept for further analyses corresponded to 98 % of the data at the undisturbed plot, 97 % of the data at the thinned plot, 84 % of the data at the clear-cut plot and 77 % of the data at the stump harvested plot.
Minimum flux detection (MDF) limit was calculated as MDF = σ t , where t is the measurement time for one specific measurement and σ is the standard deviation for the concentration measurement.For a chamber the size as used in this study, the MDF for a single measurement was 2.8 µmol m −2 h −1 .For daily average values of hourly measurements, this value is reduced to < 1 µmol m −2 h −1 , since the MDF value should be divided by the square root of the number of measurements.It is important to note that, while fluxes below the MDF cannot be securely detected, they must still be considered.For example, consider time series where fluxes decrease smoothly from an emission peak to an uptake.In the transition phase from net emissions to net uptake, fluxes will be close to 0. Removing fluxes < MDF could possibly bias the result towards a stronger sink or source than those for which times series from the individual chambers give support.Therefore, the fluxes within the MDF interval will also be kept in the analyses.Removing fluxes within the MDF interval for a single measurement (±2.8 µmol m −2 h −1 ) would lead to a decrease in the number of flux measurements by 0, 16, 13 and 30 % for the undisturbed, thinned, clear-cut and stump harvested plots, respectively, and not change the mean exchange of CH 4 at any of the plots by more than 0.6 µmol m −2 h −1 .
Correction of the measured CH 4 concentrations for dilution by water vapour was only possible at the undisturbed, clear-cut and stump harvested plots after water vapour measurements started in June 2010.This means that daytime data (global radiation > 20 W m −2 ) from the thinned plot had to be excluded from the analyses.During the night, the dilution effect had very little impact.
The impact of the environmental variables (soil temperature, soil moisture, and water table depth) on CH 4 exchange was analysed separately by Spearman linear correlations using the "corr" function, and by multiple linear regressions on standardized data using the "stepwisefit" function (both of which used Matlab, version R2009b).The stepwise regression analyses were performed by bi-directional elimination.P values were used in the selection process.The analysis was made on standardized data to adjust for the disparity in variable sizes, which makes the outcome of the analyses, the coefficients, comparable.The coefficients would be useful in the modelling of CH 4 exchange.A variable with a larger coefficient has a higher impact on the CH 4 exchange.Standardization for a data point x i was made by x i = x i − x σ , where x is the average of all data points and σ is the standard deviation of all data.An r 2 value for the overall model was also calculated, showing how much of the variance in CH 4 exchange is not explained by the environmental variables included in the analyses.
The significance of mean values at the measurement locations was calculated with the t test function (also Matlab, version R2009b).

Environmental conditions
There were differences in soil moisture and soil temperatures among the plots.On average, the undisturbed forest plot, with measurements exclusively from the summer season (July through September) had the driest and warmest records, and the measurement locations were also further above the groundwater table than at other plots (Fig. 2, Table 1).The clear-cut plot, which was measured in October and November, showed the coldest and wettest conditions, including the highest water table.Four of the five measurement locations at this plot were, on average, less than 15 cm above the groundwater table (Fig. 2, Table 1).The thinned plot and the stump harvested plot had similar average moisture and temperature conditions, but the measurements at the thinned plot proceeded over a longer time period and thus the conditions varied more.The thinned plot also had a generally deeper groundwater table than the stump harvested plot (Fig. 2, Table 1).
Soil N and C content and pH were higher at the clearcut and stump harvested plots than at the undisturbed and thinned plots (Table 1).

CH 4 exchange
The mean CH 4 exchange of all measurement locations within the plots were as follows: the undisturbed plot and the thinned plot were net CH 4 sinks of −10 µmol m −2 h −1 and −5 µmol m −2 h −1 , respectively, while the clear-cut plot and the stump harvested plot were net sources of 13.6 µmol m −2 h −1 and 17 µmol m −2 h −1 , respectively (Fig. 2).However, the CH 4 exchange varied within the plots.At the clear-cut and stump harvested plots, both net sources and net sinks existed (Fig. 3).Plot T 3 and T 4 at the thinned plot shifted between net daily CH 4 sinks and net daily CH 4 sources on a few occasions (Fig. 3b).Fluxes ranged from −7.2 to −11.6 µmol m −2 h −1 at the undisturbed plot, from −0.3 to −8.6 µmol m −2 h −1 at the thinned plot, from −3.0 to 32.5 µmol m −2 h −1 at the clear-cut plot and from −2.9 to 74.0 µmol m −2 h −1 at the stump harvested plot (Fig. 3).

Drivers of CH 4 exchange at the undisturbed and thinned plots
Linear regression analyses between CH 4 exchange and climatic variables showed that, for most measurement locations at the undisturbed and thinned plots, consumption significantly (p < 0.001) increased with decreasing soil water content, decreasing water table depth and increasing temperatures.Exceptions to this were net CH 4 uptake at locations T 3 and T 4 , which decreased with increasing temperatures, and net CH 4 uptake at locations T 2 and T 4 , which decreased with decreasing soil moisture (Table 2).Figure 5 shows an example of the CH 4 exchange response to temperature and soil water conditions at plot U 4 .
Monthly multiple linear regression analyses (Table 3) added some temporal information to the CH 4 exchange at the undisturbed and thinned plots.At the undisturbed plot, the water table depth affected CH 4 consumption in August.In September 2010, temperature was the most influential variable at all measurement locations.In July 2010, the result was less distinct, showing some measurement locations with a higher dependency on water table depth, soil moisture and some measurement locations with a higher dependency on temperature.The clearest result at the thinned plot was a dependency on soil moisture at measurement locations T 1 and T 3 in August 2009 and at locations T 1 and T 2 in April 2010 (Table 3).Soils were wetter than average in August and April, due to heavy rains in June and July 2009, and snowmelt in spring 2010.However, according to the r 2 value of the overall model, there is a lot of unexplained variance in the CH 4 exchange at all measurement locations.

Drivers of CH 4 exchange at the clear-cut and stump harvested plots
Generally, at the clear-cut and stump harvested plots, the measurement locations with net emissions of CH 4 had either a relatively short distance to water .5-49.9 0-1 a C and N pool to a depth of 20 cm in the mineral soil (litter layer excluded).b pH (BaCl 2 ) for the undisturbed and thinned plots, and pH (H 2 O) for the clear-cut and stump harvested plots, were measured at a 0-10 cm depth in the mineral soil.
disturbed by site preparation, but plot S 4 was a CH 4 sink, while plot S 5 was a CH 4 source (Fig. 4).At the majority of the measurement locations on the clearcut and stump harvested plots, higher temperatures correlated significantly (p < 0.05) with lower CH 4 emissions, or, in one case, with a higher net uptake.Both negative and positive significant correlations between CH 4 exchange and soil moisture were found at a few measurement locations, but the soil moisture range at those measurement locations was very small.At two measurement locations with net emissions at the clear-cut plot, there was a significant (p < 0.05) negative correlation between CH 4 exchange and water table depth, so that a deeper water table depth gave higher CH 4 emissions (Table 2).The multiple linear regression confirmed the significantly negative correlation between CH 4 exchange and temperature at six measurement locations.

Discussion
All measurement locations at the undisturbed forest plot were sinks of CH 4 throughout the measurement period, which is consistent with the generally drained, drier and warmer soil conditions at the plot (Fig. 2).The measurement locations at the thinned plot were also net sinks of CH 4 , although reduced in comparison to the undisturbed plot.By contrast, the clear-cut and stump harvested plots were net sources of CH 4 .Since the measurements at the different plots were conducted at different times of the year, seasonality and annual variations can probably explain some of the differences in CH 4 exchange and soil conditions.However, it is not likely that differences in water table depth between the plots are due solely to seasonal variations.In the autumn of 2010, the water table was, on average, more than 1 m higher at the clearcut and stump harvested plots than at the undisturbed plot.In addition to this, the mean CH 4 exchange for the autumn period October to November at the thinned site did not differ much from the mean CH 4 exchange for the whole measurement period, indicating that average seasonal variations are small (Fig. 2).Precipitation was, on average, higher during the measurement period at the thinned site than during measurements at the other plots, which did not cause a switch from CH 4 sink to CH 4 source.The clear-cut and stump harvested plots are located on a plateau, which is uphill from the thinned and undisturbed plots, and hence topography should not be responsible for the higher water table at the clear-cut and stump harvested plots (Fig. 1).
Water table depth, soil moisture and soil temperature were all shown to be important drivers of CH 4 exchange, as demonstrated by the linear and multiple linear regression analyses.However, it appears that the rise of the water table and increased soil moisture caused some of the measurement locations to shift to CH 4 sources.This is consistent with results by Zerva and Menucuccini (2005) and Castro et al. (2000).Temporal shifts to CH 4 emissions after snowmelt and summer precipitation, as were seen at measurement locations T 3 and T 4 , were also reported by Wang and Bettany (1995).
A majority of net emitting measurement locations at the clear-cut and stump harvested plots (C 2 , C 3 , C 4 , C 5 , S 1 ) were positioned less than 21 cm above the water table, and had a volumetric soil moisture content above 40 % (Table 1).Measurement location T 3 , when it had temporarily shifted to a CH 4 source, also had volumetric soil moisture content above 40 %.Net emissions were also measured at measurement locations S 2 and S 5 with water table depths at 30-40 cm and volumetric soil moisture contents of 23-40 %.Fiedler and Sommer (2000) found a threshold value of water table depth at 15 cm, below which only minor annual emissions were measured.The three measurement locations at the clear-cut and stump harvested plots that showed net consumption of CH 4 were further than average above the water table for those plots (Fig. 4).
Temperature seemed to have a stronger impact on CH 4 exchange in drier conditions.Figure 5 illustrates a high correlation (r 2 = 0.74) between soil temperature and CH 4 exchange at measurement location U 4 when excluding data points with soil moisture above 22 % and a distance to the water table of less than 1.25 m.The threshold value of 22 % was selected after visual inspection of the data.If all of the data from wetter conditions were included (volumetric soil moisture content > 22 % and water table < 1.25 m away), the corresponding r 2 would equal 0.47.This is consistent with the results from the multiple linear regression analyses showing that water table depth had a significant impact on the CH 4 exchange at all measurement locations in August 2010, when the water table depth varied strongly.In contrast, throughout September, there were no major precipitation events, and soil temperature was the most influential variable.Soil moisture was rarely below 30 % at the thinned plot, thus the temperature dependence was less.In autumn (September to November 2009), all measurement locations at the thinned plot were stable sinks of CH 4 , even though the soil temperature was, at times, below 5 • C.
At the clear-cut and stump harvested plots, where most measurement locations were net sources of CH 4 , we would expect a positive correlation between soil temperature and  CH 4 exchange, so that higher temperatures led to higher net emissions of CH 4 .Methanogens generally respond better than methanotrophs to increased temperatures (Dunfield et al., 1993).However, this was not the case; a majority of the measurement locations showed a significantly negative correlation between temperature and CH 4 exchange.The result is difficult to explain since CH 4 production and oxidation are not measured separately.Soil temperature profiles at the clear-cut and stump harvested plots (data not shown) show that, during the measurement period, changes  in surface temperature, associated with periods of cloudy conditions and precipitation, at a 5 cm depth are larger than those at a 20 and 40 cm depth.Methanotrophs are expected to be located closer to the soil surface than methanogens, and the larger temperature increase at the surface might compensate their lower response to temperature, which could explain why net CH 4 exchange is negatively correlated to soil temperature during this period.The highest CH 4 emissions were found at four of the five disturbed measurement locations; that is, at sites of bare soil where organic and mineral soils were mixed.The soil at disturbed measurement locations seemed less compact than at measurement locations with intact vegetation, so the disturbance probably did not inhibit diffusion.The availability of fresh organic material was possibly higher at disturbed measurement locations.Fresh, labile organic matter would promote heterotrophic uptake of O 2 and increase the soil's water retention, thereby promoting the activity of methanogenic archaeans (Wachinger et al., 2000).The one disturbed measurement location, which showed net CH 4 consumption, S 3 , was positioned on top of a mound with a relatively large distance to the groundwater table (Fig. 4).
Since this is a study of the short-term effects of forest management practices on CH 4 exchange, there are no data on how long-lived these effects are.Sudden shifts from sinks to sources and back again, due to changes in soil water conditions, are evident, as we have seen at the thinned plot (Fig. 3b).It might take years (Tate et al., 2006) to several decades for a soil to regain its full sink capacity.The recovery time for the soil CH 4 sink strength of forests on abandoned agricultural land was more than 100 years (Prieme et al., 1997;Smith et al., 2000).Increasing CH 4 uptake with time after afforestation can be an effect of an increase in the population of CH 4 oxidizing bacteria with time (Bárcena et al., 2014) or better soil diffusivity and soil aeration with time  (Christiansen and Gundersen, 2011;Peichl et al., 2010).A better soil aeration with time could be due to an increase in root biomass, which means that the roots loosen the soil over time and absorb more water (Peichl et al., 2010).Hiltbrunner et al. (2012) found that the soil CH 4 sink capacity of abandoned agricultural land increased with stand age up to 120 years.This was due to the increased transpiration of older forests and their ability to shield the forest floor from precipitation, which resulted in more favourable conditions for methanotrophic activity.
Uptake rates by forest landscapes might be overestimated (Grunwald et al., 2012;Fiedler et al., 2005).A study by Grunwald et al. (2012) found that wet forests were as important as wetlands for the CH 4 budget of European forests.Fiedler et al. (2005) found that, if 2.3 % of a forest area consisted of wet soil, the forest could turn from a sink to a source of CH 4 .As mentioned, after clear-cutting, water table depth decreased and soil moisture increased.Wetter soils and a higher groundwater table are common consequences of clear-cutting, and it is therefore important to consider their impact on the CH 4 budget in managed forests, especially if the recovery time for the soil CH 4 sink is several decades.In this study, the effects of thinning on the CH 4 exchange were not as pronounced as those for clear-cutting, although the plot average consumption was reduced in comparison to the undisturbed plot.Any forest management practice that reduces disturbance and leaves a continuous forest cover might be a better alternative from a global-warming perspective.

Conclusions
Our study on the short-term effects of boreal forest management on CH 4 exchange shows that the undisturbed plot and the thinned plot remained net CH 4 sinks, while the clearcut and stump harvested plots were net CH 4 sources.Linear regression analyses between CH 4 exchange and climatic www.biogeosciences.net/11/6095/2014/Biogeosciences, 11, 6095-6105, 2014 variables showed that, for most measurement locations at the undisturbed and thinned plots, net CH 4 uptake increased significantly with decreasing soil moisture, decreasing water table depth and increasing temperatures.A higher water table and increased soil moisture were likely to be responsible for the shift to CH 4 emissions at the clear-cut and stump harvested plots.At most of the measurement locations, which showed net emissions, the soil was almost saturated and the water table was within a few decimetres of the soil surface.Clear-cutting of the forest resulted in a raised groundwater table and increased soil moisture.These effects should be accounted for in the CH 4 budget of managed forests.

Figure 1 .
Figure 1.Schematic picture of the different plots and some of the groundwater pipes.Three more pipes are located within the clearcut and at the stump harvested plots, but are covered by the plot symbol.The background consists of a digital elevation model showing the height above sea level for each square metre (Digital elevation model (DEM) data kindly provided by N. Kljun et al.).

Figure 2 .
Figure 2. Average CH 4 exchange rates, soil moisture and soil temperature at the four sampling plots.Data from the entire measurement period at each plot are included.The dashed line at the thinned plot represents average values for October and November, since measurements at the clear-cut and stump harvested plots were conducted during this part of the year.

Figure 3 .
Figure 3.Time series of daily mean CH 4 exchange, daily precipitation and daily mean soil temperature at the measurement locations.(a) Undisturbed plot; (b) thinned plot; (c) clear-cut plot; (d) stump harvested plot.

Figure 4 .
Figure 4. CH 4 exchange (µmol m −2 h −1) at all individual measurement locations with associated level of groundwater table.The water table depth at plot C 5 is close to zero and that is why the bar is not visible in the diagram.Measurement locations at which soil surface was disturbed during site preparation.

Figure 5 .
Figure 5. Correlation between CH 4 exchange (µmol m −2 h −1 ) and soil temperature ( • C) at measurement location U 4 .The different colours represent different soil moisture and water table depths.

Table 1 .
Information regarding vegetation, C and N pool, pH, soil moisture (5th and 95th percentiles)and depth to water table at the individual chamber locations.Chamber locations were named U 1 -U 5 at the undisturbed plot, T 1 -T 4 at the thinned plot, C 1 -C 5 at the clear-cut plot and S 1 -S 5 at the stump harvested plot.
table or were disturbed by site preparation, or both, although there were exceptions.Plots S 4 and S 5 had the same water table depth and were not www.biogeosciences.net/11/6095/2014/Biogeosciences, 11, 6095-6105, 2014

Table 2 .
Correlation coefficients C and corresponding P values for the linear regressions between CH 4 exchange and soil temperature, soil moisture and water table depth.The r 2 shows how well the combined variables explain the variance in the CH 4 exchange.The correlation analyses are based on data from the entire measurement period.* significant: p < 0.001; * significant: p < 0.05; no data: At the time for measurements on the stump harvested plot, the water table depth was only measured manually on a few occasions and therefore no linear regression could be made for this period. *

Table 3 .
Coefficients from multiple linear regression analyses.A value is given only if the variable significantly contributes to explain the variation in the CH 4 exchange.The r 2 shows how well the combined variables explain the variance in the CH 4 exchange.S.m represents soil moisture, S.t denotes soil temperature, W.t shows water table depth and n.d means no data is available.
table.The water table depth at plot C 5 is close to zero and that is why the bar is not visible in the diagram.Measurement locations at which soil surface was disturbed during site preparation.