Synthesizing greenhouse gas fluxes across nine European peatlands and shrublands – responses to climatic and environmental changes

In this study, we compare annual ﬂuxes of methane (CH 4 ), nitrous oxide (N 2 O) and soil respiratory carbon dioxide (CO 2 ) measured at nine European peatlands ( n = 4) and shrublands ( n = 5). The sites range from northern Sweden to Spain, covering a span in mean annual air temperature from 0 to 16 ◦ C, and in annual precipitation from 300 5 to 1300 mm yr − 1 . The e ﬀ ects of climate change, including temperature increase and prolonged drought, were tested at ﬁve shrubland sites. At one peatland site, the long-term ( > 30 yr) e ﬀ ect of drainage was assessed, while increased nitrogen deposition was investigated at three peatland sites. The shrublands were generally sinks for atmospheric CH 4 whereas the peatlands 10 were CH 4 sources, with ﬂuxes ranging from − 519 to + 6890 mg CH 4 -C m − 2 yr − 1 across the studied ecosystems. At the peatland sites, annual CH 4 emission increased with mean annual air temperature, while a negative relationship was found between net CH 4 uptake and the soil carbon stock at the shrubland sites. Annual N 2 O ﬂuxes were generally small ranging from –14 to 42 mg N 2 O-N m − 2 yr − 1 . Highest N 2 O emission oc- 15 curred at the sites that had highest concentration of nitrate (NO − 3 ) in soil water. Further-more, experimentally increased NO − 3 deposition led to increased N 2 O e ﬄ ux, whereas prolonged drought and long-term drainage reduced the N 2 O e ﬄ ux.


Introduction
Shrublands are natural or semi-natural nutrient poor ecosystems, characterized by patches of low stature vascular vegetation alternating with bryophytes or bare soil. Commonly, the shrublands on poorly drained, deep organic soils (> 30 cm) are called peatlands (Lai, 2009) while shrublands on mineral soils, which are typically exces-5 sively drained, are called shrublands sensu stricto (Beier et al., 2009). These ecosystem types make up approximately 7 % of the European land area (EEA, 2006;Montanarella et al., 2006) and contribute valued ecosystem services, such as habitat provision, recreation, water purification and carbon (C) sequestration (Kimmel and Mander, 2010). For instance, northern peatlands were estimated to sequester 0.07 Gt C yr −1 , which over time has accumulated, resulting in deep organic soil profiles that constitute a major store of terrestrial C (Gorham, 1991;Clymo et al., 1998). Responses of these ecosystems to climate change and other anthropogenic disturbances are therefore important for the overall European greenhouse gas (GHG) budget. In Europe, peatlands are most abundant under the subarctic and temperate climates 15 of the Nordic countries, where Finland and Sweden together account for 65 % of the European area of peat soils (Montanarella et al., 2006 Lai, 2009). Presence and composition of vascular vegetation can also affect the CH 4 flux dynamics, both through root exudation of labile C substrates for CH 4 production , and by providing gas conduits, which may influence CH 4 production, oxidation and transport processes (Joabsson 5 et al., 1999). In contrast, dry Calluna heaths and Mediterranean shrublands are usually CH 4 sinks, where CH 4 uptake most often correlates positively with temperature and negatively with soil water content (Castaldi and Fierro, 2005;Carter et al., 2011). The latter is due to the slower gas diffusion in water than in air, which causes soil water to limit CH 4 diffusion towards the zone of methanotrophic activity (Dunfield, 2007). This 10 zone is mainly located in the upper 20 cm soil layer (Roslev et al., 1997;Stiehl-Braun et al., 2011a, b), but can go deeper as aridity increases (Striegl et al., 1992). Studies have also reported differences between peatlands and shrublands in the key drivers of nitrous oxide (N 2 O) fluxes. At least two microbial processes may contribute to N 2 O emissions from soils (Baggs, 2011). These are nitrification and denitrification, 15 which occur under aerobic and anaerobic soil conditions, respectively, and generally increase with nitrogen (N) availability. In a shallow peatland in Scotland, a spatial analysis revealed a negative correlation between N 2 O emissions and soil pH, possibly an indirect result of limited soil N availability for the microbes at locations with higher pH values (Dinsmore et al., 2009), where the vegetation constitutes a stronger N sink. The 20 temporal N 2 O dynamics were primarily controlled by soil temperature with an apparent switch from consumption to production at about 8 • C (Dinsmore et al., 2009). Fluxes of N 2 O in shrublands have only been investigated in a few studies, and knowledge of key drivers specific for this ecosystem type is sparse (e.g., Curtis et al., 2006;Carter et al., 2011). In four British moorlands, Curtis et al. (2006)  Decomposition of soil organic matter and plant root respiration together make up soil respiratory CO 2 emission, which in peatlands typically correlates positively with both temperature and water table depth (Smith et al., 2003;Danevčič et al., 2010). Similarly for shrublands, soil CO 2 emissions were shown to be strongly controlled by temperature in an analysis based on two years of data from four European heathlands 5 (Emmett et al., 2004).
The relative contribution to global warming of the three greenhouse gases CH 4 , N 2 O and soil respiratory CO 2 may vary between peatlands and shrublands. For instance, annual GHG budgets for a minerotrophic fen in Finland and a shallow peatland in Scotland showed that N 2 O fluxes were generally of minor importance, whereas 10 CH 4 emissions and net ecosystem CO 2 exchange exerted the greatest impact on the ecosystem's feedback to global warming (Drewer et al., 2010). For comparison, in an old Kunzea ericoides shrubland in New Zealand, GHG emissions were dominated by CO 2 release from the soil while pronounced CH 4 uptake more than counteracted N 2 O emissions on a CO 2 -equivivalent basis (Price et al., 2010). Similar studies for European 15 shrublands are currently lacking.
In future, the exchange of GHGs between ecosystems and the atmosphere may be affected by predicted changes in climate, including increasing temperatures and increased duration and frequency of summer droughts; the latter expected in western and southern Europe (IPCC, 2007a). For instance, Emmett et al. (2004) reported 0-20 19 % increases in soil CO 2 effluxes in response to 0.5-2 • C warming across four of the five shrublands described in this study. Furthermore, Carter et al. (2011) found a 20 % increase in CH 4 uptake rates under elevated temperature in a dry temperate heathland (DK-Bra in this study). Emissions of N 2 O from this site were reduced and stimulated by drought and elevated temperature, respectively, but only when these treatments were 25 combined with elevated atmospheric CO 2 concentrations (Carter et al., 2011).
In addition to the anthropogenic contribution to climate change, anthropogenic activities continue to affect the hydrology of peatlands through extensive drainage operations to enable forestry, agriculture or peat harvesting for horticultural purposes and fuel. Drainage of peatland was particularly widespread in the northern temperate zone before the 1980s (Kimmel and Mander, 2010). Such drainage has led to elevated CO 2 effluxes (von Arnold et al., 2005a,b;Salm et al., 2009) and to decreased CH 4 emissions (Roulet and Moore, 1995;Minkkinen et al., 2002;Lai, 2009), although emissions of CH 4 can be significant from drainage ditches in nutrient-rich fens (Minkkinen and 5 Laine, 2006). By contrast, N 2 O emissions from ombrotrophic bogs are relatively insensitive to changes in the water table, as the efflux is probably limited mainly by slow N transformation rates caused by low pH and nutrient availability (Martikainen et al., 1993). A further anthropogenic influence on GHG fluxes in peatlands and shrublands is the 10 effect of enhanced atmospheric N deposition, which is less well understood. Most oxidized N (NO x ) pollution is derived from fossil fuel combustion, to satisfy energy and human transportation needs (Dignon and Hameed, 1989). Reduced N (NH x ) pollution is predominantly an agricultural bi-product, increasing in direct proportion to fertilizer use, animal numbers and body mass (Asman et al., 1998). Thus, there is little likelihood 15 that the global N pollution will decline in the immediate future (Galloway et al., 2004;Peñuelas et al., 2012). Bragazza et al. (2006) investigated peatlands along an atmospheric N deposition gradient ranging from 2 to 20 kg N ha −1 yr −1 and found increased soil respiratory CO 2 emissions under higher N deposition rates. This relationship was explained by the removal of N constraints on microbial metabolism and by improved 20 litter quality. The effects of increased N availability on CH 4 exchange are dependent on site specific properties. Some soils show inhibitory effects on CH 4 oxidation (Crill et al., 1994;Christensen et al., 1999), while others show minor or no effects (Saarnio and Silvola, 1999). If the abundance of vascular plants increases in peatlands as a result of higher N availability, this may increase CH 4 emissions by providing gas conduits 25 and improving litter quality (Joabsson et al., 1999;Nykänen et al., 2002). Increasing the amount of inorganic N in a soil will also enhance the potential for N 2 O emissions both through nitrification and denitrification processes (Firestone and Davidson, 1989 Experimental manipulation studies at the field scale, as well as natural gradient studies (space for time substitution), are valuable tools for predicting how ecosystems will respond to climate change. While manipulation studies are well-suited to study effects of changes in specific experimentally controlled climate drivers, they are intrinsically constrained in time (years to two decades). Natural gradient studies are better suited to study effects on longer time scales (decades to centuries), but suffer from possible biases caused by changes in factors other than the gradient component under investigation, for example, climate. Studies based on a combination of experimental manipulations performed across multiple sites covering larger natural gradients should combine the strengths of both strategies and increase the reliability of the results.
Within the framework of the NitroEurope Integrated Project, we investigated how the soil-atmosphere exchange of the three greenhouse gases, CH 4 , N 2 O and CO 2 , responded to simulated changes in drought, warming, drainage or nitrogen addition at nine experimental peatland and shrubland sites situated across wide natural gradients in precipitation and temperature. The aims of the study were i) to identify environmental 15 variables that correlated with GHG fluxes across the nine sites or within an ecosystem type, ii) based on treatment responses at the site level, to evaluate whether specific variables had a direct effect on GHG fluxes and thereby could be identified as key drivers, and iii) to quantify the relative contribution of each GHG to the total global warming response for each experimental treatment.

Study sites
The nine experimental peatland and shrubland sites included in the study were all part of the NitroEurope project. The climatic conditions at the field sites ranged from subarctic conditions in northern Sweden to a Mediterranean climate in Spain, and from Introduction  (Fig. 1). The ecosystems were selected to represent open landscapes in these regions and included four peatlands and five shrublands ( Table 1). The peatlands were all ombrotrophic bogs, while the shrublands covered heather-moorland (UK-Clo), dry Calluna heath (DK-Mol, DK-Bra, NL-Old) and Mediterranean garrigue shrubland (ES-Gar). Soil type and plant species composition differed between the sites, but all sites 5 had dwarf shrubs in common, mainly Calluna vulgaris, Empetrum nigrum or Erica spp. (Table 1).

Experimental manipulations
At the Estonian peatland EE-Män, the long-term effect of drainage was studied in a part of the bog where a drainage ditch system established in the 1950s (Veber, 1974) was reconstructed in 1975. Substantial lowering of the water table has resulted in significantly improved tree growth (Niinemets et al., 2001;Portsmuth et al., 2005). Thus, when this study took place, Pinus sylvestris was 10-12 m tall in the drained part of the bog compared to 1-3 m in the non-drained part. The effect of increased nitrogen deposition was investigated at three peatland sites. At the two Swedish sites SE-Sto and SE-Fäj, 15 fertilized plots received NH 4 NO 3 three times per year, corresponding to a total input of 4 g N m −2 yr −1 (Lund et al., 2009). At the Scottish peatland UK-Whi, treatment plots were given an extra 5.6 g N m −2 yr −1 in wet deposition using an automated sprinkler system, which provided ca. 120 artificial rain events per year coupled to rainfall (Sheppard et al., 2011). Nitrogen was applied either in the reduced form as NH 4 Cl or in the Introduction

Soil greenhouse gas flux measurements
Fluxes of CH 4 , N 2 O and soil respiratory CO 2 were measured monthly or twice a month using static chamber methods described by e.g. Carter et al. (2011) and Sowerby et al. (2008). The chamber design and the measurement procedure varied slightly across the nine sites. Overall, well before simultaneous measurement of CH 4 and N 2 O fluxes 5 started, a permanent soil collar was installed in each plot on which a chamber was placed during measurements. Three to four times during the enclosure period of 20-180 min, a sample of headspace air was collected using a syringe through a septum in the chamber. At the Spanish shrubland ES-Gar, however, only two gas samples were collected during the 15 min enclosure period to avoid adverse effects on 10 headspace air pressure in the small chambers (volume 0.8 l) used at this site. Gas samples were stored in plastic syringes (SE-Sto), pre-evacuated vials (UK-Clo, NL-Old) or non-evacuated vials using a double-needle approach (EE-Män, UK-Whi, DK-Mol, DK-Bra, ES-Gar) before analysis for CH 4 and N 2 O concentrations by gas chromatography. One exception was the Swedish site SE-Fäj, where headspace concentrations of CH 4 15 and N 2 O were determined in the field using a photoacoustic gas analyzer (Lund et al., 2009). Previous studies showed that gas flux rates measured using a photoacoustic gas analyzer were statistically identical to flux rates based gas sampling and analysis by gas chromatography (Ambus and Robertson, 1998;Lund et al., 2009). Soil CO 2 emissions were measured within permanent collars without vegetation, i.e. where any 20 aboveground plant growth was removed on a regular basis. Generally, a chamber was placed on the collar and the build-up of CO 2 during the enclosure period was monitored by a portable infrared gas analyzer. At three out of nine sites (EE-Män, NL-Old and ES-Gar), vascular plant free patches enabled soil CO 2 emissions to be determined concurrently with CH 4 and N 2 O fluxes using the same chamber and gas chromatogra- 25 phy approach. For all three gases, the flux rates were calculated using linear regression of headspace concentration versus time, except for long enclosure periods (≥ 120 min), where CH 4 uptake was determined by fitting a first-order function. Introduction Annual cumulative fluxes were obtained by linear interpolation between measurement days. At the Swedish peatlands SE-Fäj and SE-Sto, gas flux measurements were not conducted during the five and eight coldest months of the year, respectively. When calculating the annual CO 2 fluxes for these two sites, we assumed that the contribution of the winter period to annual fluxes is 9 % at SE-Fäj (Lindroth et al., 2007) and 22 % 5 at SE-Sto (Larsen et al., 2007). These ratios were also used to estimate annual CH 4 and N 2 O fluxes, assuming that CO 2 fluxes represented the general level of microbial activity.

Precipitation, temperature and soil moisture
The meteorological measurements included precipitation and air temperature. Soil 10 temperature probes were installed at either 5 cm depth (EE-Män, SE-Fäj, DK-Bra, ES-Gar) or 10 cm depth (UK-Whi, UK-Clo, DK-Mol, NL-Old) and data were logged at least twice a day. Volumetric soil water content was measured at least twice a week by Time Domain Reflectometry probes installed in the top 6 cm (DK-Mol), 10 cm (SE-Sto, SE-Fäj, UK-Clo, NL-Old), 15 cm (ES-Gar) or 20 cm soil layer (DK-Bra). For UK-Whi and 15 EE-Män, volumetric soil moisture in 0-10 cm depth was calculated from monthly or seasonal measurements of gravimetric soil water content. At the peatland sites, water table depth was measured manually in dipwells or automatically using a pressure transducer. 20 Bulk N deposition was measured with bulk deposition collectors with a minimum diameter of 10 cm. Concentrations of NH + 4 , NO − 3 and total N in the sampled precipitation were determined by ion chromatography or colorimetric assays. If bulk N deposition was not assessed at the site, then total N deposition reported in the literature for this specific area was used instead. Nitrate leaching was determined from modelled water 25 percolation combined with NO the rooting zone using porous suction cups (Schmidt et al., 2004;Larsen et al., 2011). Soil water was also sampled in the upper part of the soil profile, generally just below the O-horizon, and analyzed for NH + 4 and NO − 3 concentrations (e.g., Schmidt et al., 2004).

Plant and soil analyses
Aboveground biomass C and N at the shrubland sites were determined using plant 5 tissue C and N analyses combined with biomass estimation based on pinpoint data (Peñuelas et al., 2004;Beier et al., 2009). Briefly, pinpoint measurements were conducted by lowering a sharpened pin through the vegetation, recording species and height for each plant hit with the pin. Total plant biomass was subsequently estimated using a site specific correlation between pinpoint data and actual biomass obtained by ion chromatography or colorimetric assays (e.g., Andresen et al., 2009). For upper and lower soil layers, soil pH was determined in a suspension of soil in distilled water, 0.01 M CaCl 2 or 1 M KCl. Furthermore, bulk density, total C and N contents, and microbial biomass C were measured as described by Beier et al. (2009), and C stock in the 0-10 cm soil layer was estimated.

Responsiveness
Relative responsiveness (RR, %) of CH 4 , N 2 O and CO 2 fluxes to the different treatments was calculated as where Treat is mean treatment flux across sites and Cont is mean control flux across sites (given in mg C or N m −2 yr −1 ). Similar to Liu and Greaver (2009), we refrained from calculating relative responsiveness in cases where the flux changed direction in response to a treatment. This occurred in the N addition experiments, where N 2 O fluxes changed from net uptake to emission. For each gas species and treatment, we 5 also calculated the absolute responsiveness (AR) in CO 2 -equivalents as where molar ratio (MR) is 12/16, 28/44 and 12/44, and global warming potential (GWP) is 25, 298 and 1 for CH 4 , N 2 O and CO 2 , respectively (IPCC, 2007b). To compare the relative contribution of each gas species to the total global warming response 10 for a specific treatment, we calculated the fractional importance (FI, %) of each gas species as where |AR 1 | to |AR 3 | is the numerical value of the absolute responsiveness for each of the three gases. Numerical values were used in order for the total global warming 15 response of all three gases to sum up to 100. To clarify, when assessing the overall response patterns across sites and treatments, we calculated RR, AR and FI for all treatments, also including treatments that did not show significant effects on flux rates in the cross-site statistical analysis described below.

20
To assess treatment effects on GHG fluxes across the sites, two-factor analyses of variance (ANOVA) with Treatment, Site and their interaction as fixed effects were conducted using the PROC MIXED procedure of SAS (SAS Institute, 2003). The random effects included Site and Treatment × Site. The GHG flux datasets were unbalanced as the number of replicates was unequal between sites, however the MIXED proce-  . More specifically, equal weight is given to each site regardless of the number of replicates when determining treatment effects. Data were log-transformed as required to obtain normality and homogeneity of variance. Main effects and interactions with P ≤ 0.05 were considered to be significant, and were interpreted using differences of least squares means.

Precipitation, temperature and soil moisture
Annual precipitation during the study period ranged from 311 mm at the subarctic peatland Storflaket in Sweden (SE-Sto) to 1351 mm at the Welsh shrubland Clocaenog (UK-Clo) ( Table 2). Mean annual air temperature in control plots ranged from −0.3 • C 10 at the subarctic site SE-Sto to 15.6 • C at the Spanish shrubland Garraf (ES-Gar). The passive night-time warming generally raised annual soil temperature at 5 or 10 cm depth by 0.5 • C, and decreased the soil water content by 2 % vol. (Table 2). Experimental drought caused a decline in mean annual soil moisture ranging from a reduction of 1.5 % vol. at the Danish shrubland Brandbjerg (DK-Bra) to 10.2 % vol. at the wet 15 shrubland UK-Clo. Drainage of the Estonian peatland Männikjärve (EE-Män) lowered the water table from an annual depth of 15 cm to 38 cm below soil surface. Characteristics of the upper and lower soil layers at the sites, as well as data on aboveground biomass, N deposition and leaching are found in the Supplement. 20 Annual CH 4 fluxes in control plots ranged from uptake of 519 mg C m −2 yr −1 at the Danish shrubland Mols (DK-Mol) to emission of 6890 mg C m −2 yr −1 from the Scottish peatland Whim (UK-Whi) ( Table 3). For the peatland sites, a clear relationship was found between annual CH 4 emission and mean annual air temperature (Fig. 2) contrast, CH 4 fluxes at the temperate shrublands correlated with the carbon stock in the top 10 cm soil layer (Fig. 3). Permanent drainage of the Estonian peatland EE-Män reduced the CH 4 efflux substantially (P = 0.003), whereas no consistent effect on the CH 4 flux was observed for prolonged drought or warming at the shrubland sites (P ≥ 0.49) (Fig. 4, Table 3). Effects of increased N input were tested at three peatland 5 sites and overall no uniform effect was found on the CH 4 efflux (P = 0.65) (Table 3). At UK-Whi, though, additional deposition of oxidized N, corresponding to 5.6 g NO 3 -N m −2 yr −1 , resulted in an almost 300 % increase of the annual CH 4 emission (P = 0.031). Concurrently, soil pH rose from 3.7 to 4.0 (see Supplement).

10
Annual fluxes of N 2 O generally consisted of low efflux rates, although net N 2 O uptake was observed in control plots at the Swedish peatland Fäjemyr (SE-Fäj) and the Scottish peatland UK-Whi (Table 3). Within each of the two ecosystem types, peatland and shrubland, the highest N 2 O efflux occurred at the site, which had the highest soil water NO − 3 concentration (Fig. 5). The stimulatory effect of NO − 3 on the N 2 O efflux was 15 supported by the N deposition experiments. More specifically, the two peatlands SE-Fäj and UK-Whi changed from N 2 O sinks into N 2 O sources as a result of increased NH 4 NO 3 and NaNO 3 input, respectively (P = 0.011) (Fig. 6a). Overall, differences in volumetric soil water content across sites did not explain the variation in annual N 2 O fluxes for control plots (Fig. 6b). At the site level, however, reduced soil moisture caused 20 by drainage or drought consistently reduced the N 2 O efflux (P = 0.0005). According to differences of least squares means, this was especially pronounced for drainage at the Estonian peatland EE-Män and for drought at the Dutch shrubland Oldebroek (NL-Old) (Site × Treatment; P = 0.034). The warming treatment had no effect on N 2 O fluxes (P = 0.47) (

Soil respiratory CO 2 emission
Soil CO 2 emissions ranged between 518 to 732 g C m −2 yr −1 at six of the nine sites investigated (Fig. 7a), but was considerable lower at the remaining three sites. As expected, the lowest rate was observed at the northernmost and coldest site SE-Sto (231 g C m −2 yr −1 ), where fluxes were only made on vegetated plots and there-5 fore are ecosystem-level respiratory CO 2 losses. Consequently, soil respiratory CO 2 losses should be even lower than the rate reported here. Soil respiration was also considerably lower at the temperate shrubland NL-Old (310 g C m −2 yr −1 ), and at the southernmost and warmest shrubland ES-Gar (390 g C m −2 yr −1 ), compared to the six sites with highest emissions. Across the five sites where warming was applied as treat-10 ment, the temperature rise of 0.5 • C in the soil had no effect on soil CO 2 emissions (P = 0.83) ( Table 3). At the site level, the CO 2 efflux seemed to be more controlled by soil moisture. More specifically, prolonged drought and long-term (> 30 yr) drainage overall caused a reduction of soil CO 2 emissions (P = 0.003) (Fig. 7b). This was especially pronounced for drainage at EE-Män and drought at the dry shrubland DK-Mol, 15 whereas drought tended to have the opposite effect at the wet shrubland UK-Clo (Site × Treatment; P = 0.0086). Increased N input at three peatland sites had no effect on annual CO 2 emissions (P = 0.69) ( Table 3).

Responsiveness of greenhouse gas fluxes
Relative to control plots, the observed treatment response was generally higher for 20 CH 4 and N 2 O fluxes than for CO 2 fluxes (Table 4). However, when comparing the fractional importance of each GHG to the total numerical global warming response (GWR), the changes in CO 2 emissions dominated the response in all treatments (ranging 71-96 %), except for NO − 3 additions where 89 % of the total GWR was due to a change in CH 4 emissions. Relative to the other GHGs, the impact on GWR from changes in N 2 O 25 fluxes was generally low; only 4 % of total GWR across the shrubland sites and absent across the peatland sites. Across all treatments and gas species, there was a negative

Discussion
In Europe, open landscapes represented by peatlands and shrublands cover about 7 % of the land area (EEA, 2006;Montanarella et al., 2006). In the current study, we 5 investigated greenhouse gases fluxes between these ecosystems and the atmosphere in order to identify factors that seemed to control cross-site variation in flux rates, either directly or indirectly. Furthermore, we assessed the effect on the GHG fluxes of anthropogenic disturbances, such as temperature rise and prolonged droughts caused by climate change, drainage of peatlands and increased atmospheric nitrogen deposition.  Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | dependent than those associated with methane oxidation (Dunfield et al., 1993). However, the simple temperature relationship of CH 4 emissions may also include other interrelated factors, most importantly plant productivity affecting substrate quality and quantity, but also plant aerenchyma development allowing CH 4 produced in the soil to be released into the atmosphere (Joabsson et al., 1999). Based on the current dataset 5 we are not able to distinguee whether temperature affected the CH 4 efflux directly or indirectly, and most likely several of the mentioned mechanisms acted simultaneously. Despite this, we suggest that temperature could be used as a common indicator when estimating the magnitude of CH 4 emissions from peatlands across wide regional scales. Furthermore, using space for time substitution our results indicate that CH 4 10 emissions from peatlands will rise in future as a result of the predicted temperature increases caused by climate change.
In a global analysis of CH 4 uptake by aerobic soils, Dutaur and Verchot (2007) found that soil texture strongly controlled CH 4 uptake with coarse textured soils consuming more CH 4 than other texture classes. All temperate shrublands in our study had 15 a sandy mineral soil layer, while the soils differed in the properties of the organic layer, which consisted of peat at the heather-moorland (UK-Clo) and of mor humus at the dry Calluna heaths (DK-Mol, DK-Bra, NL-Old). Net CH 4 uptake at the four temperate shrublands correlated negatively with the carbon stock in the upper 10 cm soil layer, suggesting a switch from net CH 4 uptake to net CH 4 emission as soils become en-20 riched in organic matter. To our knowledge, such a relationship has only been reported in very few studies. For instance, Singh et al. (1997) showed that CH 4 uptake in dry tropical forest soils correlated with both carbon and nitrogen content of the soils; two variables which increased in parallel. Within soil types, the content of organic matter and soil water often correlate, either because organic matter increases the water 25 holding capacity (Hudson, 1994) or because high soil moisture hinders aerobic decomposition (Jungkunst and Fiedler, 2007). In line with this, the two sites with the highest carbon stock of the four temperate shrublands (i.e., UK-Clo and NL-Old) also had the highest soil moisture ( Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | exchange could be indirect via an effect on soil moisture, which limits CH 4 diffusion towards the zone of methanotrophic activity (King, 1997;Dunfield, 2007). Singh et al. (1997), however, suggested a more direct control mechanism, where reduced net uptake of CH 4 in carbon-rich soils was due to decreased O 2 content in the soil atmosphere, resulting from higher decomposition rates. In our study, though, we found no 5 relationship between carbon stock and soil respiratory CO 2 emission at the four temperate shrubland sites. Potentially, organic matter stimulated methanogenesis by acting as a substrate for the process, leading to lower net CH 4 uptake in the carbon-rich soils in our study. Most likely, the correlation between soil organic C and CH 4 uptake was caused by a combination of these indirect and direct mechanisms, and we suggest that 10 future studies aim at investigating this relationship in more details.
Previous studies of N additions to wetlands reported both positive and negative responses in net CH 4 emissions, and the mechanisms involved are still under debate (Bodelier, 2011). In the present study, NO − 3 addition at the Scottish bog UK-Whi increased the CH 4 efflux by almost 300 %, whereas NH + 4 or NH 4 NO 3 additions had no 15 overall effect. At the Scottish bog, the addition of NO − 3 raised pH in peat water from 3.7 to 4.0 (see Supplement) and increased the availability of dissolved organic carbon (DOC), possibly because the slightly higher pH enhanced the solubility of organic compounds in the soil (Evans et al., 2008). In contrast, NH + 4 addition had no effect on pH or DOC. For comparison, Murakami et al. (2005) raised the pH of four tropical 20 acid peat soils by ca. 2 pH units and observed a vast increase in the CH 4 production potential. The authors suggested that this was due to increased substrate supply for methanogenic microorganisms derived from decomposed organic matter such as organic acids from peat humus. A similar mechanism could explain the enhanced CH 4 efflux in response to NO − 3 addition at the Scottish bog UK-Whi. 25 At the Swedish peatland sites SE-Fäj and SE-Sto, addition of NH 4 NO 3 had no significant effects on the CH 4 flux. Although insignificant, the opposing effects of N addition on CH 4 fluxes at SE-Fäj (increased efflux) and SE-Sto (decreased efflux), illustrates the importance of peatland type and site specific properties, such as plant composition and

N 2 O fluxes
Latest advances in knowledge suggest that three separate processes contribute to N 2 O emissions from soils. These are nitrification, denitrification and dissimilatory nitrate re-5 duction to ammonium (DNRA), which are thought to be important sources of N 2 O in oxic, sub-oxic and anoxic microsites of the rhizosphere, respectively (Baggs, 2011). At individual field sites, daily N 2 O emission rates are typically reported to increase with soil moisture in the range from 50 to 90 % water filled pore space (WFPS) (Smith et al., 2003). This relationship is probably caused by the associated decrease in O 2 diffusion within the soil, leading to larger anaerobic microsites and thereby increased N 2 O production via denitrification (Smith et al., 2003) and possibly also DNRA. However, when soil moisture exceeds 90 % WFPS, N 2 O emissions may decline as N 2 is the final product of denitrification in most waterlogged soils (Davidson, 1991), although pH may also influences the N 2 O/N 2 loss ratio of denitrification (Šimek and Cooper, 2002). 15 In the present study, the lack of relationship between N 2 O emission and volumetric soil moisture across sites was possibly because N 2 O production is controlled by gas diffusivity, which is poorly reflected by volumetric soil moisture when comparing soils that differ in total porosity. At the site level, however, net N 2 O emissions declined consistently in response to reduced soil moisture caused by drought treatment, supporting 20 previous results from drought experiments in a spruce forest (Goldberg and Gebauer, 2009). Furthermore, N 2 O emissions from the Estonia peatland EE-Män declined due to long-term drainage, which is in contrast to results from previous drainage experiments in ombrotrophic bogs (Martikainen et al., 1993;Regina et al., 1996). Apparently, increased tree growth at the drained area reduced the availability of soil nitrate (see Supplement), which in combination with lower soil moisture caused the decline in N 2 O emissions. Analysed across sites, the shrubland and peatland with highest NO − 3 concentration in soil water also had the highest annual N 2 O emission within each ecosystem type, which illustrates that N 2 O production in these nutrient poor ecosystems was limited by N availability. This was supported by data at the site level, where N additions (< 60 kg N ha −1 yr −1 ) that included nitrate led to increased N 2 O efflux from two peatlands, suggest-5 ing that denitrification or DNRA were involved in N 2 O production in these ecosystems, and that the processes were stimulated by increased substrate availability.

Soil respiratory CO 2 emission
Annual soil respiratory CO 2 emissions (R s ) ranged from 231 to 732 g C m −2 yr −1 in control plots across the nine sites. The highest fluxes were observed at the temper-  NL-Old could therefore result from a lower quality of the soil C pool compared to the other sites. Thus, our results indicate a pattern across the sites where soil respiration increased with temperature, but that the increase was offset in the warmest ecosystem due to low soil moisture, resulting from low precipitation combined with high temperatures. At the site level, the was no significant effect of the warming treatment across 10 sites, which may be due to the relatively mild temperature increase achieved with the passive warming technique of 0.5 • C in the soil.
The prolonged drought or drainage manipulations showed stronger effects on soil respiration rates than both warming and N additions. Reduced soil moisture due to drought or drainage generally decreased soil respiration rates at the dry shrublands 15 and the wet, long-term drained peatland EE-Män, but increased soil respiration rates at the wet shrubland UK-Clo. As drainage most often leads to increased CO 2 emissions in hydromorphic soils, at least in the short-term (Jungkunst and Fiedler, 2007;Danevčič et al., 2010;Maljanen et al., 2010), our results highlight how results may differ depending on the time since the experimental treatments were initiated. The drainage 20 treatment at EE-Män was started more than 30 yr ago. Currently, the easy degradable fractions of the peat at EE-Män have already been decomposed and carbon cycling in the ecosystem has reached a new balance. In contrast, the drought treatment at UK-Clo has only been running for eight years, thus it is likely that the long-term effect may differ from the current short-term effect. Furthermore, the different response directions to prolonged drought between the sites show how important the initial conditions are for the strength and directions of the response to the treatment.

Experimental manipulations
The quantitative effect on GHG fluxes of an experimental treatment is to a certain extent controlled by the magnitude of the treatment effects on environmental drivers. The question is whether our experimental manipulations have caused realistic changes in these drivers, which enable us to predict changes in GHG fluxes in future? On an 5 annual basis, the prolonged drought treatment at the shrubland sites reduced moisture in the top soil by 10-20 %, which is a consistent decline when taking into account that the drought treatment was only carried out for about two months per year. In contrast, the warming treatment generally raised annual soil temperature at the sites by 0.5 • C.
For comparison, annual mean soil temperatures in control plots at the Welsh shrubland Thus, the overall warming effect of 0.5 • C was less than the interannual variability at the site level. Furthermore, surface temperatures in Europe are projected to increase 2-5 • C during this century, depending on which scenario is used (IPCC, 2007a). In conclusion, the lack of experimental warming effect on any of the three GHGs in our 15 study could relate to an insufficient temperature increase obtained by the passive nighttime warming treatment. By contrast, drainage about 30 yr ago of the Estonian peatland EE-Män has more than doubled the size of the oxic zone by lowering the water table from an annual depth of 15 cm to 38 cm below soil surface. As a result, the emission of all three GHGs 20 had dropped dramatically after 30 yr of treatment. Furthermore, annual N deposition was raised from 0.2, 1.5 and 0.8 g N m −2 to 4.2, 5.5 and 6.4 g N m −2 at the peatlands in northern (SE-Sto) and southern Sweden (SE-Fäj) as well as Scotland (UK-Whi), respectively. It is unlikely that total inorganic N deposition in these regions of Europe will reach such high levels in future (Galloway et al., 2004). However, global warming 25 will lead to increased soil mineralization rates, releasing more inorganic N and other nutrients for plant uptake (Rustad et al., 2001;Mack et al., 2004). Also, large regions of South Asia are projected to receive more than 5 g N m −2 yr −1 by 2050. Apart from the CH 4 response to nitrate addition at the Scottish peatland UK-Whi, responses in GHGs fluxes to these relatively high N application rates were minor, illustrating resilience in the peatland ecosystems towards increased N deposition in the short term. However, in the longer term, changes in peatland plant composition and structure may occur, affecting the exchange of greenhouse gases. For instance, a warmer and drier climate 5 with increased nutrient turnover will inflict competitive disadvantages for Sphagnum mosses, as they are adapted to cold, wet and nutrient-poor conditions (Clymo, 1984), and possibly stimulate growth of vascular plants.

Responsiveness
We observed some clear and interesting patterns in the responses of the three GHG 10 fluxes across the nine investigated ecosystems. First of all, the observed responses to the manipulations of climate and atmospheric N deposition were stronger in relative terms for CH 4 and N 2 O fluxes than for soil CO 2 emissions. However, when comparing the fractional importance of each GHG to the total global warming response, then the observed changes in soil CO 2 efflux dominated the response for the drainage, 15 drought and warming treatments, where at least 91 % of the global warming response was caused by changes in soil CO 2 fluxes. The N addition treatments in peatlands that included NO − 3 differed from this overall pattern by the dominance of a change in CH 4 emissions (89 % of global warming response) compared to changes in the other gas species. This profound effect therefore also partly explains why responses in CH 4 20 fluxes played a slightly more dominant role across the four peatlands (56 % of the global warming response) compared to soil CO 2 emissions (44 % of global warming response) and N 2 O fluxes, which did not contribute significantly. In contrast, across the five shrublands including both drought and warming treatments, the response in soil CO 2 emissions dominated (94 % of global warming response) by far over responses in Based on 109 different studies on effects of N addition on ecosystem GHG budgets, Liu and Greaver (2009) reported an overall negative global warming feedback of net ecosystem CO 2 exchange, but also that this was largely offset (53-76 %) by increased CH 4 and N 2 O emissions. When comparing the different treatments in our study, the N addition experiments led to a substantial positive climate forcing response. However, 5 this was caused by a strong increase in CH 4 emissions, whereas Liu and Greaver (2009) found that N 2 O was the primary contributor to global warming across the studies in their review. Warming effects were very small and not significant in the overall cross-site ANOVA. By contrast, the drought and drainage treatments led to a uniformly negative climate forcing response. Again, it must be emphasized that the response for the drained peatland in this study is the long-term response and it is likely that a substantial amount of carbon was lost from the ecosystem in the years immediately after drainage. But at least across the five shrubland ecosystems there was a clear pattern of overall negative global warming feedback in response to the drought treatment.
In general, we recognize that the nine sites included in this study span differences in 15 soil types, plant species, pre-treatment conditions and present climate, and that the experimental treatments differed between sites. For this reason, generalized conclusions based on the mean responsiveness presented here should be drawn with caution. Especially, we lack information on responses in net primary production, and it is likely that the absolute responsiveness of net ecosystem CO 2 fluxes would differ from that of soil 20 CO 2 emissions alone. Finally, none of the nine experiments included interactive effects with other climate drivers, which have previously been shown to be important for the combined response to multiple changes (Larsen et al., 2011).

Conclusions
With respect to environmental variables that seemed to influence the magnitude of 25 GHG fluxes across the nine sites, temperature was identified as a driver of CH 4 emissions at the peatland sites. In contrast, net CH 4 uptake at the temperate shrublands sites correlated negatively with the soil carbon stock, either due to a indirect effect via soil moisture or because soil carbon acted as a source for methanogenesis. Nitrate availability seemed to be a driver of N 2 O emissions at both peatland and shrubland sites, as the highest N 2 O efflux within each ecosystem type occurred at the site, which had the highest NO − 3 concentrations in soil water. Soil respiratory CO 2 emission 5 showed maximum annual values at the temperate sites with mean annual temperature of 8-10 • C. Outside this range, in each end of our temperature gradient, soil respiration was either limited by low temperatures at the subarctic site SE-Sto or by low summer soil moisture at the Mediterranean site ES-Gar.
In terms of total climate forcing response, prolonged drought and long-term (> 30 yr) 10 drainage resulted in consistent negative feedback to global warming. However, substantial amounts of CO 2 were presumably lost from the drained peatland immediately after drainage. Furthermore, N addition to peatlands gave rise to considerable positive feedback to global warming, primarily due to the response of CH 4 emissions to NO − 3 addition at one site. Finally, experimental warming had no effect on any of the 15 three GHGs, which could be explained by a modest temperature increase of 0.5 • C imposed by the passive night-time warming treatment. When comparing the fractional importance of each GHG to the total numerical global warming response, then the change in CO 2 efflux dominated the response in all treatments, except for NO − 3 addition where change in CH 4 emissions accounted for the major part. Therefore, within 20 the applied range of water, temperature and nutrient manipulations, soil CO 2 emission plays a dominant role with respect to ecosystem feedback on global warming.

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version