Global wetlands are, at estimate ranging 115–237 Tg CH4/yr, the largest single atmospheric source of the greenhouse gas methane (CH4). We present a dataset on CH4 flux rates totaling 12 measurement years at sites from Greenland, Iceland, Scandinavia and Siberia. We find that temperature and microbial substrate availability (expressed as the organic acid concentration in peat water) combined explain almost 100% of the variations in mean annual CH4 emissions. The temperature sensitivity of the CH4 emissions shown suggests a feedback mechanism on climate change that could validate incorporation in further developments of global circulation models.
 Northern wetlands are significant sources to the atmosphere of the important greenhouse gas methane (CH4). Globally wetlands are the largest single source to the atmosphere with recent estimates ranging from 115 to 237 Tg CH4/yr [IPCC, 2001], and between one third and half of these emissions are from northern wetlands (here defined as all non-tidal peat-forming wetlands north of 50°N). These emissions constitute more than 75% of the total estimated natural emissions of CH4 to the atmosphere [IPCC, 2001]. Wetland distribution and activity are therefore thought to have been main determinants for variation in the atmospheric CH4 concentration in the past [Chappelaz et al., 1993]. Holocene records of variations in atmospheric CO2 and CH4 are closely coupled to past temperatures as obtained from Antarctic and Greenlandic ice cores [Petit et al., 1999; Monnin et al., 2001] suggesting a functional relationship between temperature and the sources (and sinks) of these trace gases. Interannual variations in the growth rate of the atmospheric methane concentration have been suggested partly attributed to the wetland source [Dlugokencky et al., 2001]. Important feedback mechanisms on future climate change arising from changing exchanges of CO2 between the terrestrial biosphere and the atmosphere have recently been identified [Cox et al., 2000]. A related question is how possible changes in the CH4 emissions from wetlands will affect the further development of the greenhouse effect, as these emissions are significant. Using the most conservative current estimate for CH4 emissions from wetlands of 115 Tg CH4/yr this amounts to 1.9 Gt C/yr in CO2 equivalents with a time-horizon of 20 years and a global warming potential (GWP) of 62 [IPCC, 2001]. In perspective this equals one third of the global anthropogenic fossil fuel related atmospheric emissions [IPCC, 2001]. Thus, changing wetland CH4 emissions could impose important implications for the understanding of past and prediction of future atmospheric greenhouse gas concentrations and the associated radiative forcing of Earth. Many independent studies have in recent years been devoted to understanding the controls on wetland CH4 emissions and the key factors influencing seasonal variations in fluxes. Controlling factors identified include soil temperature, water table position and net ecosystem exchange (NEE) of CO2 [Dunfield et al., 1993; Whalen and Reeburgh, 1992; Joabsson et al., 1999]. Attempts at synthesizing different independent studies to identify large scale controls on regional and inter-annual differences in CH4 fluxes have been made [Crill et al., 1992; Bartlett and Harriss, 1993; Bubier and Moore, 1994] but the individual field studies have differed in experimental design and subsequently their findings have prohibited the identification of single factors as indicator of large scale variations in CH4 fluxes. NEE has been suggested as an integrating proxy of CH4 flux strengths across large latitudinal gradients [Whiting and Chanton, 1993], and the present study was partly motivated by an objective to investigate this relationship further at a wide range of sites and with fully comparable methods.
2. Study Sites
 The sites span almost 20 latitudinal and more than 100 longitudinal degrees in the northwestern Eurasian and Greenlandic North (Zackenberg: 74°30′N, 21°00′W; Hestur: 64°35′N, 21°36′W; Abisko: 68°22′N, 19°03′E; Kevo: 69°45′N, 27°18′E; Plotnikovo: 57°01′N, 82°35′E). They represent typical high latitude wetlands for the individual regions with water tables at or close to the soil surface (i.e. excluding dry peatlands with minor CH4 emissions) and climate conditions ranging from arctic continental in Greenland to cool temperate in West Siberia. Although all sites represents typical peat forming wetlands with soils characterized as histosols the soil properties span a range of C/N values from 20 in Iceland to 55 in Siberia. The vegetation at all sites is dominated by typical wetland species such as Eriophorum spp. and Carex spp. with a varying subcanopy of mosses (Sphagnum spp.). The total aboveground biomass, however, differ significantly among sites from almost 1700 g dw/m2 in Siberia to less than 50 g dw/m2 in Greenland.
 At all sites a range of moisture conditions were included with average water tables ranging from the surface to 10 cm below and each site consisted of measurement plots replicated a minimum of 12 times. CO2 and CH4 flux measurements were carried out using a modified version of the standard closed chamber technique [e.g. Joabsson and Christensen, 2001; Livingston and Hutchinson, 1995; Vourlitis et al., 1993; Waddington et al., 1996]. Daytime measurements were carried out weekly or biweekly distributed evenly through the growing season during two or three years yielding a total of 12 measurement years. Here we compare the annual mean values in order to identify possible factors with a predictive capability for large-scale inter-annual variations in CH4 emissions.
 Each measurement plot was equipped with aluminium or steel bases with a water channel for sealing. Depth in the soil was between 5 and 20 cm depending on stability and seal. Plexiglas chambers (squared, 10–30 liters) that reduced PAR <10% were used. An IRGA system (Li-Cor 6200 or PP Systems EGM-2) was used for continuous measurement of CO2 over a 4-minute chamber installation period logged at <1 minute intervals. The IRGAs were equipped with desiccant (magnesium perchlorate) in the flow line to the sample cell. A probe for logging temp., PAR and RH was situated inside the chambers. Small internal fans were installed and operated in connection with all chamber measurements.
 The sampling routine involved one 4-minute light measurement (for net CO2 exchange) followed immediately by one more where the chamber was completely darkened (for total respiration). System photosynthesis was then estimated as the difference between those two measurements. The same chambers were used for the CH4 flux measurements although these were darkened to avoid temperature changes. CH4 samples were taken from the tubing (a t-piece) connected to the IRGA system in order to avoid touching the chamber during the installation and they were taken during the same (within <0.5 hours) time period (light, temp. conditions) as the CO2 flux measurements. Chamber installation periods for CH4 flux measurements were kept as short as possible (10–30 min) and depended on flux rates. Syringe samples were taken in duplicates three–four times during chamber installation. CH4 analysis was carried out on standard gas chromatographs equipped with flame ionization detectors. Methane flux was calculated based on a linear regression of the chamber concentration change with time and all measurements with an r2 < 0.8 were left out.
 For organic acid analyses pore water samples were collected from the peat profiles at each of the sites during the 1999 and 2000 field seasons. Water was sampled at 5 and, depending on site properties, 15, 20 or 25 cm depth of the profile and immediately sterile filtered, flushed with N2 for 1 min and frozen. The samples were thawed immediately prior to analysis of organic acids using an anion exchange HPLC system [Ström et al., 1994; 2002].
4. Results and Discussion
 A Pearson correlation analysis for linear relationships between the independently measured variables and seasonally averaged CH4 emissions showed that only soil temperature and organic acid concentrations had consistent significant correlations. Mean seasonal NEE alone showed no correlation with the CH4 fluxes. This is contrary to earlier findings based on flux measurements across a range of different wetland ecosystems [Whiting and Chanton, 1993]. From our study it appears that while NEE may correlate with seasonal CH4 flux variations at individual sites [Joabsson and Christensen, 2001; Waddington et al., 1996; Bubier et al., 1995; Christensen et al., 2000; Bellisario et al., 1999] it seems less suited for explaining large scale variations in CH4 fluxes. Rates of photosynthesis and respiration undoubtedly increase when moving from cold to warmer environments but the net exchange of CO2 does not necessarily change accordingly. This is indicated by long term accumulation rates of carbon in the relatively warm western Siberian lowlands that are not different to accumulation rates in peatland habitats with a much colder summer climate [Turunen et al., 2001]. The relationship between NEE and CH4 flux shown for individual sites in many studies [Joabsson and Christensen, 2001; Waddington et al., 1996; Bubier et al., 1995; Christensen et al., 2000; Bellisario et al., 1999], is therefore difficult to apply to large climatic gradients as the two will not change proportionally in response to increasing mean seasonal temperatures. This is seen clearly in the data shown in Figure 1 where CH4 emissions increase with increasing soil temperature, while NEE does not show any straightforward response (results not shown).
 The correlations show that mean seasonal soil temperature act as a strong predictor for methane flux rates and with an exponential correlation it explains up to 84% of the variance in methane emissions. (Figure 1). Interestingly the exponential relationship seen in Figure 1 is similar to standard temperature responses of CH4 production as seen over the same temperature range in laboratory studies of methanogenic monocultures [Panikov, 1995; Svensson and Roswall, 1984]. The water table was relatively high at the studied sites and the differences in water table did not show any consistent predictive power of large-scale variations in CH4 emissions. Variations in the water table have been shown repeatedly in field studies to be a strong controlling factor on seasonal variations in CH4 flux [e.g. Daulat and Clymo, 1998; Waddington et al., 1996; MacDonald et al., 1998]. However, when comparing large scale variations in CH4 flux the water table act more as an “on-off switch”, and once the water table is around or above 10 cm from the surface, i.e. in the wetland ecosystems where the substantial methane emissions occur, other processes take over the control on large scale variability. This indicate that the parts of northern wetlands with the most significant CH4 emissions may be rather robust with respect to their response to water table changes and that a limited degree of wetland drying, if kept within a certain threshold, does not necessarily lead to decreasing CH4 emissions as assumed in modelling studies [Christensen and Cox, 1995; Cao et al., 1998; Walter and Heimann, 2000]. On the contrary a substantial drying of northern wetland areas beyond the threshold could lead to a “switching off” reducing CH4 emissions dramatically.
Figure 2 includes the result of analyses of organic acids in the peat water at all the studied sites. Organic acids in general, and acetate in particular, are very important precursors and substrates for CH4 formation and therefore good indicators of methanogenic substrate availability. We were able to identify acetate, lactate, formate, proprionate, pyruvate, malate, oxalate and citrate in most samples with acetate overall constituting more than 60% of the total and in some cases >90%. Organic acids are typical fermentation products of anaerobic degradation of organic matter. However, most vascular plant species also actively exude organic acids from their roots [Jones, 1998]. One of the main factors controlling belowground C allocation and exudation is the intensity of photosynthesis [Kuzyakov, 2002]. This provides a direct species dependent linkage between vascular plant production and the substrate availability for CH4 formation, and this linkage has the potential to affect methane emission rates [Joabsson and Christensen, 2001]. This species-dependent provision of substrate for CH4 production in response to increased growth is a strong contributing factor to the variation around the otherwise straightforward temperature relationship identified in Figure 1. A strong overall correlation between organic acids and the mean methane emissions can be seen in Figure 2. If only acetate, instead of the sum of all measured organic acids, is correlated with CH4 emission rates the explained variance amounts to 92% (P = 0.028), which clearly demonstrates the particular importance of acetate as a precursor for CH4 formation.
Figure 1 illustrates the importance of temperature as control on large-scale variations in annual CH4 emission rates from northern wetlands with a water table at or close to the soil surface (i.e. the areas where CH4 emission matters for the atmosphere). Figure 2 indicate that an important indirect effect of temperature is through increasing the general carbon turnover and also the labile carbon pools providing substrates for CH4 emissions. Vascular plant species-specific root exudation patterns may modulate the substrate availability for methanogenic bacteria and hence account for the variance not explained by temperature alone.
 Although a range of interrelated factors undoubtedly will determine the interactions between plant and microbial responses to climate change, Figure 1 show simple and strong temperature sensitivity with respect to wetland CH4 emissions. This finding supports that CH4 emissions from wetlands may have changed rapidly in response to initial warming or cooling in the possibly adding to the explanation of the tight coupling between the atmospheric concentration [Dlugokencky et al., 2001] and temperature in the ice core records [e.g. Petit et al., 1999]. It also represents a potential feedback mechanism on future climate change from the northern wetlands, which are mostly located in the continental regions expected to experience the strongest warming in coming decades [IPCC, 2001]. A two degree average summer warming over the temperature range where most northern wetlands are located, would according to the relationship shown in Figure 1 result in a 45% increase in mean seasonal CH4 emissions. Using the conservative end of estimated total global wetland CH4 emissions this would result in the equivalent of an extra almost 1 Gt C/yr in CO2 equivalents using a GWP for CH4 of 62 [IPCC, 2001] emitted to the atmosphere. This is a considerable number in carbon budget terms, whether it may provide a significant feedback mechanism in the global climate models remain to be studied.
 We thank W. S. Reeburgh and N. T. Roulet for valuable comments on earlier versions of this manuscript. This work was funded by the European Union (EU) and the Swedish Environmental Protection Agency through the 4th Framework CONGAS project in the EU Environment and Climate Programme. P. Crill and one anonymous reviewer provided constructive and valuable reviews of this paper.