Emissions of methane from northern peatlands: a review of management impacts and implications for future management options

Abstract Northern peatlands constitute a significant source of atmospheric methane (CH 4). However, management of undisturbed peatlands, as well as the restoration of disturbed peatlands, will alter the exchange of CH 4 with the atmosphere. The aim of this systematic review and meta‐analysis was to collate and analyze published studies to improve our understanding of the factors that control CH 4 emissions and the impacts of management on the gas flux from northern (latitude 40° to 70°N) peatlands. The analysis includes a total of 87 studies reporting measurements of CH 4 emissions taken at 186 sites covering different countries, peatland types, and management systems. Results show that CH 4 emissions from natural northern peatlands are highly variable with a 95% CI of 7.6–15.7 g C m−2 year−1 for the mean and 3.3–6.3 g C m−2 year−1 for the median. The overall annual average (mean ± SD) is 12 ± 21 g C m−2 year−1 with the highest emissions from fen ecosystems. Methane emissions from natural peatlands are mainly controlled by water table (WT) depth, plant community composition, and soil pH. Although mean annual air temperature is not a good predictor of CH 4 emissions by itself, the interaction between temperature, plant community cover, WT depth, and soil pH is important. According to short‐term forecasts of climate change, these complex interactions will be the main determinant of CH 4 emissions from northern peatlands. Drainage significantly (p < .05) reduces CH 4 emissions to the atmosphere, on average by 84%. Restoration of drained peatlands by rewetting or vegetation/rewetting increases CH 4 emissions on average by 46% compared to the original premanagement CH 4 fluxes. However, to fully evaluate the net effect of management practice on the greenhouse gas balance from high latitude peatlands, both net ecosystem exchange (NEE) and carbon exports need to be considered.


| INTRODUCTION
The concentration of methane (CH 4 ) in the atmosphere has increased from 722 ppb during the pre-industrial period to 1,819 ppb in 2012, due to increased anthropogenic emissions Whalen, 2005). Methane is the second most important greenhouse gas (GHG) after carbon dioxide (CO 2 ). Although it contributes less than 0.5% of the atmospheric carbon (C) gas concentration, it constitutes about 20% of the global radiative forcing (IPCC, 2013). This is because CH 4 has a much stronger radiative forcing (i.e., is 34 times stronger than CO 2 ) (IPCC, 2013). For an emission pulse of similar mass of C, CH 4 creates a powerful immediate radiative forcing at the start, but due to its shorter atmospheric lifetime, this declines faster than for CO 2 (IPCC, 2013).
In wetland soils, CH 4 is produced in the anaerobic zones of submerged soils by methanogens, is oxidized to CO 2 by methanotrophs in the aerobic zones, and is emitted to the atmosphere when the balance between the production and consumption is positive (Le Mer & Rodger, 2001). Peat soil accumulation derives from a positive water balance and a water table (WT) close to the soil surface, which results in anaerobic conditions preserving organic material (Belyea & Clymo, 2001;Lai, 2009), and that is also a prerequisite for methanogenesis, the terminal step of anaerobic organic matter mineralization (Hou, Wang, Chen, & Patrick, 2000;Yavitt & Williams, 2000). Both the CH 4 directly produced in the peat soil and atmospheric CH 4 can be oxidized as an energy source, or used for biosynthesis by methanotrophs (Conrad, 1996;Hanson & Hanson, 1996).
Methane can also be released to the atmosphere in bubbles (ebullition) which take place when there are gas pockets in the waterlogged soil, or the dispersal of the gas is prevented by a layer of dense peat or ice (Baird, Beckwith, Waldron, & Waddington, 2004;Tokida et al., 2007). Air pressure has an important role in establishing the timing and quantity of CH 4 ebullition (Tokida et al., 2007). In the Aapa mires (fens), CH 4 confined under ice layers can be released in the spring thaw, representing about 11% of the annual emissions (Tokida et al., 2007). In these situations, large quantities of CH 4 (>40 g CH 4 m −2 ) may be released to the atmosphere over periods of minutes to hours (Glaser et al., 2004;Rosenberry, Glaser, Siegel, & Weeks, 2003), where CH 4 bubbles are transported through the peat too fast to allow oxidation to occur. Methane can also be released to the atmosphere via vascular plants (Joabsson et al., 1999;King, Reeburgh, & Regli, 1998).
Under anoxic conditions, vascular plants in wetlands may transport O 2 through specialized, aerenchymatous tissues, by which CH 4 can also be released to the atmosphere (Joabsson et al., 1999). The exchange of O 2 and CH 4 through vascular plants between the anoxic zone and the atmosphere may have contrasting effects on CH 4 emissions in northern peatlands. Methane production by methanogenic archaea could be inhibited by the transport of O 2 into otherwise anaerobic layers, or oxidized due to release of O 2 into the rhizosphere. Due to this bypass release of CH 4 , the net emission to the atmosphere tends to increase when aerenchymatous vascular plants are present (Joabsson et al., 1999). Further, CH 4 has low solubility in water (23-40 mg/L at 0-20°C) and could escape through sediment into the atmosphere by either diffusion or ebullition. The gas could be transported through vascular plants (Joabsson et al., 1999) or diffused slowly upward through peat soils where the methanotrophic bacteria are able to oxidize it to CO 2 . Analyzing a large UK data set on CH 4 emissions from soils, Levy et al. (2012) found that where plant species composition data (percentage cover of aerenchymatous plant species) were available, this provided the highest explanatory power of CH 4 fluxes to the atmosphere.
Northern peatlands represent a crucial ecosystem for regional GHG budgets because they store large amounts of C (Loisel et al., 2014). However, the ratio between decomposition and conservation of the C depends on the vegetation types present, for example, Sphagnum mosses are more resistant to decomposition compared to sedges and other vascular plants and thereby retain more C over time (Rydin & Jeglum, 2006). Peatlands can be divided into two main categories, depending on their hydrology and nutrient status. These are (1) ombrotrophic peatlands (bogs) which receive water and nutrients from atmospheric deposition and thus are acidic and poor in nutrients and (2) minerotrophic peatlands (fens) which receive water and nutrients from the surrounding mineral soils in the catchment. Nutrient status in fens varies from close to ombrotrophic nutrient-poor conditions to mesotrophic/eutrophic conditions, mainly controlled by the ratio between the peatland and mineral soil area, and the mineral nutrient status in that catchment (Clymo, 1983). Differences between the peatland types are also reflected in vegetation composition, primary production, organic matter decomposition, and C gas emissions (Clymo, 1983;Nilsson et al., 2001). Peatlands are also classified into aquatic, forb, graminoid, lichen, moss, nonvegetated, shrub, and treed based on the general form of the vegetation cover, rather than on species (Adams et al., 1997).
Wetland restoration is one method with which northern countries could aim to meet their GHG targets under the Kyoto Protocol (Bain, Hornsey, Bongiorno, & Jeffries, 2012). In contrast to drainage, restoration raises the WT, increases water saturation, and thus may increase CH 4 emissions (Saarino, Winiwarter, & Leitao, 2009). The WT level controls the balance between CH 4 and CO 2 emissions and the rate of CH 4 emissions to the atmosphere is therefore very sensitive to WT depth (Price & Ketcheson, 2009;Sirin & Laine, 2008). The most prevalent restoration method is drain blocking, which could restore the WT to its initial state (Holden et al., 2007), or raising the WT by gully and ditch blocking (Evans, Monteith, & Cooper, 2005). Other restoration methods include planting and reseeding of bare surfaces, or re-establishment of natural peatland vegetation, which is important, as vegetation is a major factor in peat formation (Petrone, Price, Waddington, & von Waldow, 2004;Vitt, 2006).
Predicted changes in climate, including rising temperatures, changes in the amount, intensity, and seasonal distribution of precipitation and amount of snow fall and cover (IPCC, 2013), could affect the dynamics of hydrology in northern peatlands and could increase methane production (FAO, 2008). Additionally, the exploitation of peatlands for agriculture, energy, and horticulture under intensive management also greatly influences the rate of mineralization (CO 2 emissions) (Laine et al., 1995). Higher CH 4 emissions could lead to a positive feedback on climate change and thereby further disturbance of peatland C stocks (Friedlingstein et al., 2006). It is suggested that climate change reduces the capacity of northern peatlands to absorb atmospheric carbon dioxide (Wu & Roulet, 2014) and this depends on how management, and the interaction with climate change, will affect CH 4 emissions. The aim of this systematic review and meta-analysis was to collate and analyze published studies to improve our understanding of the factors that control CH 4 emissions and the impacts of management on the gas flux from northern peatlands. The specific hypotheses that we tested were as follows: (1) Methane emission is mainly controlled by WT, plant community, temperature, and pH; (2) management, especially drainage and restoration, significantly affects CH 4 emissions; and (3) climate change will significantly reduce the capacity of northern peatlands to absorb the atmospheric C.

| Data collection
To locate all papers that have reported CH 4 emissions from northern peatlands, we performed a comprehensive search on the Web of Science database (accessed between January 2013 and July 2016) using the keywords: pristine peatlands, methane emissions, drainage, restoration, fens, bogs, mire, and northern peatlands. In an attempt to gain a comprehensive coverage, we also checked all references in the papers found in the Web of Science search. Only studies which covered at least one growing season and measured at weekly or more frequent intervals were selected. These searches resulted in 87 studies reporting measurements of CH 4 emissions taken at 186 sites covering different countries, peatland types, and management systems ( Fig. 1). To indicate the direction of the methane flux, we used the atmospheric science sign convention, that is, a negative sign represents uptake of CH 4 gas by the ecosystem. In cases where a site has several years of flux data, the average flux of these years was used. If the flux values covered the growing season only, we estimated the annual flux values based on a previously used factor, generated from studies with full annual measurements coverage, whereby winter fluxes were estimated to constitute 15% of the annual CH 4 fluxes (Maljanen et al., 2010;Saarino et al., 2007). All CH 4 flux values were converted to g C m −2 year −1 . The overall CH 4 flux average ± SD (g C m −2 year −1 ) for "natural peatlands" was based on site averages reported in each publication and did not account for the variation between years at a single site. Some studies are repeated in more than one table because they include more than one site of different management systems.
For the studies included in this meta-analysis, CH 4 fluxes were measured using different methods which may differ in their ability to capture ebullition fluxes. These are manual chamber measurements, autochambers, and eddy covariance flux towers. Also, different methods were used to measure soil pH, for example, using pH probe/meter in deionized water or 0.01 M CaCl 2 in 1:1 and 1:2 or 1:5 (v:v) soils: solution ratios. We assumed the pH results to be equivalent and, where a range of values were reported, we took the mean. Where air temperature was reported, we used the mean annual temperature in degree Celsius (°C) as variations between years were minor. The WT was reported relative to the surface in centimeters (cm) and we used the convention of negative values representing distance below the surface. Where a range of water levels was reported over the study, we used the mean value in the meta-analysis. In this review, we have adopted the classification of fens/bogs, for consistency, as most of the sites included were classified into fen or bog. From the descriptions of the sites in each paper used in this study, we assigned a peatland type of either fen or bog or wooded fen and bog.

| Data analysis
We used Minitab 16 (Minitab Inc., State College, PA, USA) and R version 3.3.0 (R Development Core Team, 2016) for data exploration, conditioning, and analysis. We split the literature studies into three groups for analysis: natural, drained, and restored. We used different analytical procedures for each group appropriate for the available published data.

| Natural peatlands
The data collected from natural peatlands covered 56 studies and 108 sites. The predictive variables available to test the response variable of annual methane flux were as follows: latitude, longitude, duration of measurement, mean annual air temperature (T), mean pH (pH), and mean WT as covariates and bog, fen, and woodland as random factors. Data exploration using matrix plots determined that latitude was colinear with mean annual air temperature as were mean pH and WT depth with their maximum and minimum values and these were excluded from the analysis. Annual precipitation, evapotranspiration, and water flow through were not available for most studies so observed WT depth was used as the explanatory variable relating to both water supply and the oxidation status. Normality in flux and residual was tested and the flux was log-transformed. A one-way ANOVA test was performed to test whether there was a significant different in emissions between bog, fen, and wooded peatlands. Next a linear mixed-effects model (LMM) was applied to test annual methane flux relationships with environmental variables and type of peatland using the "lmer" method (version 1.1-12) (Bates, Mächler, Bolker, & Walker, 2014), while p-values were calculated in order to confirm the significance of the relationships using the lmerTest package version 2.0-30 (Kuznetsova, Brockhoff, & Bojesen Christensen, 2013) in R version 3.3.0 (R Development Core Team, 2016). The package "piece-wiseSEM" version 1.1.3 (Lefcheck and Jonathan, 2016) was used to calculate values for explained variation for obtained linear mixedeffects models. Not all variables were available at all sites with pH being available for the least, so the LMM was performed on samples that had pH values (n = 36) and then repeated on samples without the variable "pH" (n = 76). Then, package "missMDA" version 1.10 (Josse & Husson, 2016) was applied to impute missing data values, resulting in 108 samples to which the LMM was applied for all samples and all variables. Multiple linear regression analysis was applied to estimate the variation explained by two environmental variables.
The package "akima" version 0.5-12 was used to create interpolated contour plots (Akima & Gebhardt, 2015) of pairs of the environmental parameters as x and y with annual CH 4 emissions as the z variable. This was made for both the available study data and the imputed data to verify that the data trends were similar and the imputed values are valid. As WT and peatland type explained 42% of the variability, we performed linear regressions of these variables against the log-transformed annual CH 4 flux, with and without identified outliers. Then, we estimated the regression model of annual CH 4 flux and mean annual water level by nonlinear least squares, using the R function "nls." F I G U R E 1 World map showing the experimental sites included in this study, across the northern peatlands. Symbols: Green = pristine, brown = drained, blue = restored; the red areas indicate histosols

| Drained peatlands
The data were tested, using paired t-tests on all paired sites where both natural (n = 42) and artificially drained (n = 61) (i.e., lower water table for using in agriculture, forestry or mining) peatlands had CH 4 emission measurements.
In addition, a t-test was performed to see whether there was a significant difference between drained fens (n = 26) and drained bogs (n = 35). The effects of different land use systems/vegetation cover (cropland [n = 4], grassland [n = 7], and woods [n = 29]) on drained peatland methane emissions were also tested using one-way ANOVA.

| Restored peatlands
The impacts of restoration system on CH 4 emissions from peatlands were investigated. The management systems tested using paired ttest were as follows: rewetting (n = 16), and restoring by vegetation and rewetting (n = 16).

| Methane emissions from northern natural peatlands
Our results show that natural northern peatland (pristine) sites are important sources for CH 4 emissions to the atmosphere (Table 1) with an overall average annual flux (mean ± SD) covering all sites, vegetation, and locations being 12 ± 21 g C m −2 year −1 . The median is 4.3 C m −2 year −1 . However, emissions between the sites were highly variable with a 95% CI of 7.6-15.7 g C m −2 year −1 for the mean and 3.3-6.3 g C m −2 year −1 for the median. A t-test (t = −1.99) shows that CH 4 emissions from the fen sites (n = 59) mean 15.4 g C m −2 year −1 are significantly higher (p = .05) than those from the bog sites (n = 49) 7.1 g C m −2 year −1 (Fig. 2). A linear regression between log CH 4 flux and mean WT depth for different peatland types showed significant correlations for bog (n = 87, r 2 = 0.54, p < 0.01) and fen (n = 45, r 2 = 0.13, p < 0.01), but not for wooded fen and bog (n = 7, r 2 = 0.36, p = 0.09) (Fig. 3a). When four outliers were removed, the correlation was significant for fen (n = 43, r 2 = .22, p < .001) but not for bog (n = 33, r 2 = .36, p = .8) or wooded fen and bog (n = 7, r 2 = .36, p = .09) The contour plots in Fig. 5A show a trend toward higher CH 4 emissions with a high water table and high pH and with lower temperature, peaking at a mean annual air temperature around 2°C. The LMM results with samples that had a pH value (30 observations/samples) showed that pH is a statistically significant factor (p = .04). The proportion of variance explained by the fixed factor(s) alone is 34% of CH 4 flux variation. The proportion of variance explained by both the fixed and random factors is 53%. The LMM results with samples in cases when the variable "pH" was omitted, where the number of observations is 76, shows that "peatland type" and WT are statistically important factors (p < .05 & p < .01, respectively). The proportion of variance explained by the fixed factor(s) alone is 19% of CH 4 flux variation. The proportion of variance explained by both the fixed and random factors is 42%.
When missing data values are imputed using missMDA, the contour plots in Fig. 5B show a similar pattern to those of the raw data in Fig. 5A, validating the technique. When LMM analysis is made on the imputed data with 108 observations, it shows that peatland type (p < .05), pH (p < .001), WT (p < .001), and air temperature (p < .01) are statistically important factors in determining CH 4 flux. The proportion of variance in CH 4 flux explained by the fixed factor(s) alone is 31%.
The proportion of variance explained by both the fixed and random factors is 63% (Table 2).

| Methane emissions from drained peatlands
A t-test shows that the difference in CH 4 emissions between drained (n = 61) and natural peatlands (n = 42) is significant (t = 7.25, p < 0.001) (Fig. 6a). Drainage reduced the CH 4 flux by, on average, 84% compared to the original emission values with a mean of 8.3 g C m −2 year −1 . Drainage reduced CH 4 emissions from the fen ecosystems by more than that from bog ecosystems, and a t-test showed a significant difference (t = 2.46, p < .015) between fens and bogs ( Fig. 6B). This effect is similar for all types of drained peatland regardless of the land use and vegetation cover. A paired t-test to assess the effect of drainage for paired sites of bogs and fens showed that for the bogs (t = 4,443; p < 0.001; n = 25) and for the fens (t = 3,762; p < 0.01; n = 17). A one-way ANOVA shows that the difference in CH 4 emissions after drainage between the land use/land cover of crops (n = 4), grass (n = 7), natural (n = 21), or woodland (n = 29) is significant (F = 2.98, p < 0.05) (Fig. 6c).

| Methane emissions from restored peatlands
Only 16 sites explicitly measured the effect of rewetting peatlands that had previously been drained for many uses, including forestry, cropping grazing, and mining. There were insufficient data for each category of initial land use, but considering the entire dataset (n = 16) rewetting increased methane flux by an average of 1.3 ± 6.5 g C m −2 year −1 (46%). However, a paired t-test showed that the change in CH 4 flux due to rewetting was not statistically significant with mean flux before restoration being 3.0 ± 3.1 g C m −2 year −1 and after restoration being 4.2 ± 6.3 g C m −2 year −1 (p = .37) with a pooled standard deviation of 6.0 (Fig. 7).
This indicates a different response to rewetting between sites, which all have different previous anthropogenic management, land use, and initial peatland type. The published data are insufficient to identify why CH 4 emissions from the different sites respond differently after rewetting.

| Methane emissions from northern natural peatlands
This review and meta-analysis shows that natural northern peatlands are a significant source for CH 4 emissions to the atmosphere due to prevailing waterlogged conditions (Huttunen, Nykänen, Turunen, & Martikainen, 2003). This is in agreement with other previous studies carried out by Nilsson et al. (2001), Christensen et al. (2003), Zhuang et al. (2006, Lai (2009) and Turetsky et al. (2014). However, high variability was observed between the sites with 95% CI of 7.6-15.7 g C m −2 year −1 for the mean and 3.3-6.3 g C m −2 year −1 for the median, especially on flooded peatlands (Couwenberg & Fritz, 2012). The type and composition of dominant peatland vegetation (Bubier, 1995;Turetsky et al., 2014) influence CH 4 emission dynamics, both by adding labile C substrates for CH 4 production  and by maintaining gas conduits, which affect the production, oxidation, and transportation of CH 4 (Joabsson et al., 1999). Bogs and fens differ in biotic and abiotic factors. These biotic and abiotic differences lead to the fens having the highest methanogenic activity (Juottonen et al., 2005), highest litter degradation rate (Aerts, Verhoeven, & Whigham, 1999), and thereby highest CH 4 emissions , compared to the bogs. However, both fen and bog ecosystems (Granberg et al., 1997;Lund, Christensen, Mastepanov, Lindroth, & Strom, 2009;Nilson et al., 2001;Rinne et al., 2007) are sources for CH 4 emissions which may cause these peatland types to be a net GHG source to the atmosphere (Drewer et al., 2010). The microtopography of a peatland is not uniform, with many hummocks and hollows, which can result in highly variable CH 4 emissions from the same site (Lai, 2009 T A B L E 1 (Continued)

| 7089
AbdAllA et Al.  and the higher CH 4 productions in the hollows due to high WT and temperature (Bubier et al., 1993).
Natural northern peatlands have an important impact on climate change Lai, 2009;Nilsson et al., 2001;Turetsky et al., 2014), and climate change has an impact on northern peatlands. In the Nordic region, under climate change, temperature is predicted to increase and WT to decrease (Forster et al., 2007). . These two variables explain 8.5% of CH 4 flux overall variation (p < .05); (b) mean annual temperature and soil pH (n = 33). These two variables explain 16.3% of CH 4 flux overall variation (p < .05); (c) soil pH and mean water table below the surface (n = 32). These two variables explain 17.8% of CH 4 flux overall variation (p < .05). (B) when data were imputed (n = 108): (a) mean annual temperature and mean water table below the surface. These two variables explain 7.6% of CH 4 flux overall variation (p < .01); (b) mean annual temperature and soil pH. These two variables explain 19.7% of CH 4 flux overall variation (p < .001); (c) soil pH and mean water table below the surface. These two variables explain 16.0% of CH 4 flux overall variation (p < .05) Shannon, Weltzin, Pastor, & Bridgham, 2008). High temperature will also result in melting of the permafrost and release of CH 4 to the atmosphere, which may provide a positive feedback to climate change in the short term (Friedlingstein et al., 2006;Olefeldt, Turetsky, Crill, & McGuire, 2013). The larger unsaturated zone will lead to less CH 4 emissions, and some dry sites may become sinks for CH 4 (Worrall, Burt, & Adamson, 2006). McCalley et al. (2014) reported that microbial community response to permafrost thaw will regulate CH 4 dynamics.
However, the majority of models forecast a significant warming-related decrease of CH 4 emissions from northern peatlands (Bridgham et al., 2013;Frolking et al., 2011). In most temperate wetlands, over the long term (300 years), C sequestration is expected to compensate for the warming role of CH 4 , turning most wetlands to net C sinks with net negative radiative forcing (Mitsch et al., 2013).

| Methane emissions from drained peatlands
The drainage practices in northern peatlands clearly reduce CH 4 emissions under all types of land use and vegetation, on average by 84% (Table 3). Drainage practices improve aeration (Schrier-Uijl, Veenendaal, Leffelaar, van Huissteden, & Berendse, 2010) leading to lower CH 4 emissions. They decrease C input, from decomposing plants, to the methanogenic anaerobic layer (Basiliko, Yavitt, Dees, & Merkel, 2003;Bergman, Svensson, & Nilsson, 1998;Bergman et al., 2000). Drainage also increases CH 4 oxidation to CO 2 and thereby reduces CH 4 emissions (Holden, 2005;Moore & Dalva, 2006;Sundh, Nilsson, Mikkela, Granberg, & Svensson, 2000). Moreover, Yrjälä et al. T-test indicates a significant difference (t = 7.25; p < .001) between the two groups, (B) CH 4 flux of drained peatland by type, bog and fen, and t-test indicates a significant difference between bog and fen (t = 2.46; p < .05). (C) CH 4 flux of drained peatland by land use: cropland, grassland, natural, and woodland/shrubs; ANOVA shows significant differences between the four groups (F = 2.98; p < .05) F I G U R E 7 Effects of rewetting on annual CH 4 flux from peatlands. Box and whiskers plots showing median, 25 and 75% median quartiles, mean (⊕), 95% confidence interval (whiskers), and outlier values (*). The left box shows initial methane flux with anthropogenic drained land management, whereas the right one indicates CH 4 flux after restoring vegetation and/or rewetting Pullin (2012) and Turetsky et al. (2014). However, drainage influences CH 4 emissions from fens more than from bogs. This is because WT depth in the fen sites is more sensitive to drainage compared to the bog sites (Maljanen et al., 2010).

Drainage ditches themselves can become new anaerobic zones,
with similar characteristics to the undrained peat, with similar or even increased CH 4 emissions (Huttunen et al., 2003;Schrier-Uijl et al., 2010;Sundh et al., 2000). In fen meadows in the Netherlands, Schrier-Uijl et al. (2010) found that ditches and bordering edges contributed up to 60-70% of the total farms' CH 4 emissions. These higher emissions from drainage ditches could be large enough to compensate for the reduced CH 4 emissions by drainage on the remainder of the drained peatland area (Minkkinen, Byrne, & Trettin, 2008). In contrast, Minkkinen, Laine, Nykänen, and Martikainen (1997) reported that CH 4 emissions from ditches in a drained peatland plantation in Finland during the summer represent about only 4.5% of CH 4 emissions. Sundh et al. (2000) found that CH 4 emissions from harvested and drained peat can be kept lower than that from virgin peatland by keeping the ditches clear and free from vegetation.
The microbial production of CH 4 is anaerobic, while the production of CO 2 is aerobic. Therefore, the production and consumption of these two greenhouse gases in peat soils are highly dependent on the oxygen availability in the soil and, thus, the depth of the water table (Aerts & Ludwig, 1997). In fen and bog peatlands, drainage decreased CH 4 emission but increased CO 2 emission by more than one order of magnitude (Von Yamulki, Anderson, Peace, and Morison (2013). This reduction in CH 4 emissions, in association with the primary productivity of vegetation, could decrease the total climate forcing of peatlands over the coming century (Worrall et al., 2011). There is a probability of 69% that drainage will result in an overall improvement in the GHG budget due to less CH 4 emissions (Worrall et al., 2011). Nevertheless, the timescale over which this GHG budget is calculated has an influence, since loss of CO 2 after drainage can be very long-lasting (Maljanen et al., 2010). In contrast, Oleszczuk et al. (2008) noted that drainage could increase CO 2 emissions, with CO 2 having a longer atmospheric lifetime relative to CH 4 , so the loss of C and lower C sink capacity in drained peatlands could result in increased climate forcing over time.
This uncertainty in CH 4 changes over time is due to the limited longterm (>10 years) studies on drained northern peatlands.

| Methane emissions from restored peatlands
Restoration of drained northern peatlands by rewetting increased CH 4 emissions compared to the original prewetting emission. In this meta-analysis, restoration increased CH 4 flux by 46% (Table 4).
Here, the open water pools behind ditch blocks increase the gas emissions (Baird, Holden, & Chapman, 2009). Hahn-Schofl et al. (2011) reported significantly higher CH 4 emissions from flooded fen grasslands in Germany, because of high availability of fresh organic matter.
Methane emission could be reduced by creating different vegetation compositions (Komulainen, Nykänen, Martikainen, & Laine, 1998;Tuittila et al., 2000;Waddington & Day, 2007) that lead to changes in the methanogenic community and peat properties (Basiliko, Knowles, & Moore, 2004). Mahmood and Strack (2011) reported a significant correlation between CH 4 emissions and vegetation cover on an abandoned peatland. This is because vegetation stimulates CH 4 emissions by providing substrates for gas production and transportation to the atmosphere (Wilson, Farrell, Muller, Hepp, & Renou-Wilson, 2013).
In Canada and Ireland, CH 4 emissions from restored cutover peatlands increased in the first 3 years following restoration due to the fresh substrates provided by the new vegetation cover (Waddington & Day, 2007;Wilson et al., 2013). Fast decomposing litter following restoration of a bog peat could result in higher CH 4 flux, which could dominate GHG emissions up to 30 years following rewetting (Vanselow-Algan et al., 2015).
As discussed earlier, vascular plants can play an important role in transporting CH 4 from soils to the atmosphere through aerenchyma (Couwenberg & Fritz, 2012;Henneberg, Elsgaard, Sorrell, Brix, & Petersen, 2015). The establishment of vascular vegetation following extraction is generally more extensive on cutover fens than on cutover bogs (Graf, Rochefort, & Poulin, 2008). Although a combined transportation of O 2 with CH 4 by aerenchyma tissues could reduce CH 4 emissions, previous studies reported higher emissions from vascular plants, especially sedges (Waddington, Roulet, & Swanson, 1996). Roulet, Ash, and Quinton (1993) and Roulet and Moore (1995) reported approximately 23-57 times greater CH 4 emissions from Missing values were imputed using missMDA software. This produced 108 observations for LMM analysis. Peatland type, pH, WT, and air temperature are statistically important factors in this case. The proportion of variance explained by the fixed factor(s) alone is 31% of CH 4 flux variation. The proportion of variance explained by both the fixed and random factors is 63%. ¶ Significant codes: 0 = ***; 0.001 = **; 0.01 = *.
T A B L E 2 Relationships between annual CH 4 flux and environmental variables (WT and pH) and type of peatland using linear mixed-effects model (LMM)  (Baird, Belyea, & Morris, 2009;Beetz et al., 2013;Samaritani et al., 2011;Strack & Zuback, 2013). Published data on CH 4 emissions from long-term (>10 years) rewetted northern peatlands are limited and therefore, changes in gas emissions over time remain uncertain.
The increase in CH 4 emissions due to restoration must be considered when land use strategies to reduce emissions are developed.
Likewise, assessing the suitability of peatland restoration processes requires a better understanding of C processes and dynamics changed by the restoration. Previous studies have resulted in the development of a guide called the "North American Peatland Restoration Guide" (Quinty & Rochefort, 2003). Application of this strategy can result in the return of a plant community that is dominated by species characteristic of peatlands (Quinty & Rochefort, 2003). The new emerging plant community, and altered hydrology resulting from restoration, should help to return GHG dynamics to those more similar to natural peatlands. In addition to producing high CH 4 emissions, rewetting also increases the dissolved organic carbon (DOC) and thereby the amount of DOC lost to rivers (Dawson & Smith, 2007;Strack, 2008). However, it is still unknown how much is eventually lost to the atmosphere in the form of CO 2 . Thus, to have a clear picture of the advantages/disadvantages of restoration to preserve C stocks of northern peatlands, long-term investigations on the overall greenhouse gas balance are of great importance.
Future sustainable and climate-friendly management strategies are needed. These management practices should focus on preventing peatland degradation. Targets of climate protection on managed peatlands could be met by converting arable land to grassland, decreasing land-use intensity, and re-establishing the original ground WT (Byrne, Chojnicki, Christensen, Drösler, & Freibauer, 2004;Freibauer, Rounsevell, Smith, & Verhagen, 2004). Petrescu et al. (2015) reported that intensity of land management (e.g., conversion of natural peatlands to agricultural land) can strongly influence net climate footprint of wetlands and could eventually result in positive radiative forcing.
They suggested that estimates of future releases of GHG inventories based on IPCC guidelines for wetlands should consider the relationship between CH 4 and CO 2 fluxes, the intensity of management, and the land use/land cover change on both the net GHG balance and thereby radiative forcing.

| CONCLUSIONS
In this review, we investigated the factors that control CH 4 emissions and impacts of management in northern peatlands (latitude 40° to 70°N). The study covered a total of 87 studies taken at 186 sites Value from Beetz et al. (2013).
c Annual values were estimated from the original seasonal measured values. Methane gas flux during winter was considered as 15% from the annual flux (Maljanen et al., 2010;Saarnio et al., 2007).
T A B L E 4 (Continued) covering different countries, peatland types, and management systems. We found CH 4 emissions from natural northern peatlands to be highly variable with a 95% CI of 7.6-15.7 g C m −2 year −1 for the mean and 3.3-6.3 g C m −2 year −1 for the median and an overall annual average (mean ± SD) of 12 ± 21 g C m −2 year −1 . emissions from northern peatlands.

ACKNOWLEDGMENTS
This work was funded in part by the GHG-Europe Project (EU grant agreement number: 244122).