Wetlands are the largest source of methane to the atmosphere, with tropical wetlands comprising the most significant global wetland source component. The stems of some wetland-adapted tree species are known to facilitate egress of methane from anoxic soil, but current ground-based flux chamber methods for determining methane inventories in forested wetlands neglect this emission pathway, and consequently, the contribution of tree-mediated emissions to total ecosystem methane flux remains unknown.
In this study, we quantify in situ methane emissions from tree stems, peatland surfaces (ponded hollows and hummocks) and root-aerating pneumatophores in a tropical forested peatland in Southeast Asia.
We show that tree stems emit substantially more methane than peat surfaces, accounting for 62–87% of total ecosystem methane flux. Tree stem flux strength was controlled by the stem diameter, wood specific density and the amount of methane dissolved in pore water.
Our findings highlight the need to integrate this emission pathway in both field studies and models if wetland methane fluxes are to be characterized accurately in global methane budgets, and the discrepancies that exist between field-based flux inventories and top-down estimates of methane emissions from tropical areas are to be reconciled.
Natural wetlands are the single largest source of atmospheric methane (CH4) and are known to contribute significantly to interannual variations in the growth rate of this potent greenhouse gas (Hodson et al., 2011). Gas transport through herbaceous plants adapted to wet soil is well documented (Brix et al., 1992; Whiting & Chanton, 1996) and enables ingress of oxygen (O2) to the root zone but coincidental venting of soil-produced CH4 to the atmosphere. Plant stems are a particularly efficient means for release of CH4 from wetland soil because the pathway bypasses highly active populations of methanotrophic bacteria situated at the oxic–anoxic interface in the subsurface.
Trees also have the capacity to cope with soil anoxia through development of morphological adaptations such as hypertrophied lenticels, adventitious roots and enlarged aerenchyma. These structures promote gas exchange between the atmosphere and the rhizosphere (Megonigal & Day, 1992; Kozlowski, 1997), in particular, entry of O2 to the root zone. Recent studies have demonstrated that temperate zone trees adapted to wet soil also facilitate egress of soil-produced CH4 (Rusch & Rennenberg, 1998; Vann & Megonigal, 2003; Terazawa et al., 2007; Gauci et al., 2010; Rice et al., 2010) via gas transport through aerenchyma tissue and emission to the atmosphere through stem lenticels (S. R. Pangala et al., unpublished). Tropical mangroves similarly possess specialized aerial roots (pneumatophores) to transport atmospheric O2 to submerged roots, which also release sedimentary CH4 to the atmosphere (Purvaja et al., 2004; Chauhan et al., 2008). However, mangroves occupy sulphate-rich intertidal zones, accounting for only c. 0.7% of tropical forested area (Giri et al., 2011), and consequently, CH4 flux from mangroves globally is small (1.95 Tg CH4 a−1; Chauhan et al., 2008) relative to other wetland sources.
Regardless, the capacity for wet soil-adapted trees to mediate CH4 emissions has been demonstrated unequivocally by studies of mangroves and temperate forested wetlands (Rusch & Rennenberg, 1998; Vann & Megonigal, 2003; Purvaja et al., 2004; Terazawa et al., 2007; Chauhan et al., 2008; Gauci et al., 2010; Rice et al., 2010). Notably the same morphological adaptations to wet soil conditions are common in trees that inhabit vast areas of highly productive freshwater swamp and peatland at low latitudes (Kozlowski, 1997; Parolin et al., 2006), yet measurements of CH4 emission from tropical forested wetlands typically focus on fluxes from the ground surface collected using closed chambers (Jauhiainen et al., 2005; Couwenberg et al., 2010). Upscaling of field measurements that exclude tree-mediated CH4 emissions may result in a significant underestimation of tropical CH4 fluxes. Moreover, the absence of the tree emission pathway in process-based models potentially limits their capacity to predict changes in trace gas exchange at the ecosystem level caused by internal or external perturbations.
Given that tropical wetlands account for the largest proportion of CH4 flux from global wetlands and that c. 53% of these ecosystems are forested (Matthews & Fung, 1987; Prigent et al., 2007), the aim of this study was to assess the extent to which trees may mediate CH4 export from anoxic soil in tropical wetlands and evaluate their contribution to ecosystem emissions relative to other CH4 emission pathways. We present in situ measurements of CH4 flux through trees and from the ground surface conducted in a tropical peat forest in Central Kalimantan (Indonesia, Borneo), Southeast Asia. Tropical peat forests of Southeast Asia are a significant reservoir of terrestrial organic carbon, storing c. 77% of tropical peatland carbon (Page et al., 2011). High rates of precipitation lead to elevated water-table levels, resulting in slow decomposition rates that favour both peat accumulation and CH4 production under anoxic conditions. Although significant quantities of CH4 are produced in the peat, CH4 typically is not released at high rates from the peat surface to the atmosphere because methanotrophic bacteria oxidize CH4 at the oxic–anoxic interface in soil and within the rhizosphere (Couwenberg et al., 2010). This study evaluated the extent to which trees in the wetland ecosystem function as conduits, enabling methane to bypass soil methanotrophs, thereby facilitating its release to the atmosphere. To our knowledge, this is the first study to measure tree-mediated CH4 emissions from tropical peat forests and also the first from any climatic zone to estimate the contribution of trees to total ecosystem methane flux.
Materials and Methods
Methane fluxes from tree stems, the peatland surface (ponded hollows and hummocks) and root-aerating pneumatophores were measured during a two week period in March 2011 in two 20 × 20 m (400 m2) plots during the wet season in a tropical forested peatland situated in the upper Sebangau River catchment in Borneo, Indonesia. The relatively undisturbed peatland is located c. 20 km southeast of Palanka Raya City in Central Kalimantan and has been described previously by Page et al. (1999). The study plots were established within mixed peat swamp forest, a forest type that extends up to 4 km from the margin of the peat dome into the interior, having a peat thickness ranging from 2 to 6 m. The climate is characterized by uniform temperature, high humidity and high rainfall intensity (c. 2800 mm a−1). The water table in these forests is above the soil surface during the wet season (c. 9 months), decreasing to 40 cm below the peat surface during the dry season (c. 3 months). During the study period, the water table depths were 4.7 ± 1.2 cm above the soil surface and 16 ± 3.5 cm below the soil surface in hollows and hummocks, respectively. The mean air and soil temperatures during the study period were 26.8 ± 2.2 and 24 ± 1.0°C, respectively. Temperatures in the region are usually relatively stable throughout the year, displaying negligible temporal variation (Jauhiainen et al., 2005).
The locations and distributions of trees, hollows and hummocks were mapped in each plot. The average area ratio of hollows to hummocks was 50 : 50 in the plots (56.4 : 43.6 in Plot 1 and 43.5 : 56.5 in Plot 2). Tree species in each plot were identified and every tree having a diameter ≥ 7 cm at 1.3 m height above soil surface (diameter at breast height; DBH) was measured for basal diameter and stand density. Approximately 87% of the trees measured had a DBH ≤ 20 cm (Fig. 1), similar to the DBH distribution reported for Southeast Asian tropical peat forests (Page et al., 1999 and references within) and some Amazonian forests (Macía, 2011). Stem diameter also was measured at 10 cm intervals between 20 and 130 cm above the soil surface for each of the eight tree species identified for CH4 measurements.
Static chambers used to measure CH4 fluxes from six locations per plot in ponded hollows and hummocks were based on the design described by Gauci et al. (2002). Approximately 30 fluxes were measured from each hollow and hummock per plot (i.e., 120 measurements in total). The static chambers were constructed from polyvinyl chloride (PVC) pipe and deployed on permanently installed soil collars (35 cm diameter × 30 cm height) inserted 5 cm into the peat surface 2 d before gas sampling. Each chamber had a removable lid equipped with a pressure regulator and sampling port. Care was taken while deploying the soil chamber to avoid disturbance of the peat surface. Gas samples that displayed evidence of induced ebullition at T = 0 min were rejected, representing c. 6% of samples collected. Static chambers used to measure CH4 fluxes from pneumatophores were constructed from PVC rectangular collars (40 × 30 × 40 cm) inserted 5 cm into the soil surface and enclosed one to three pneumatophores. During each measurement, a rectangular lid containing a gas sampling port and pressure regulator was placed on the collars.
Static chambers used to measure CH4 fluxes from tree stems were constructed from a design described by Rusch & Rennenberg (1998) and Gauci et al. (2010) with further modification (Fig. 2). The static chamber was constructed from flat Perspex® (Perspex Distribution Tamworth, Tamworth, UK) sheets (30 × 30 × 30 cm) assembled into a cube, which was then cut into two halves and held together using hinges and spring clips. Each cubic chamber had a 20-cm-diameter central opening to enclose the tree stem. A clear Perspex cylinder (20 cm diameter × 5 cm height) was attached to the central opening on either side of the chamber, which held a gas-impermeable foam strip (7 cm wide) against the tree stem, creating a gas-tight seal. A transparent sheet of gas-impermeable fluorinated ethylene propylene film (Adtech Ltd, Gloucestershire, UK) was attached to the outside of the Perspex cylinder, the foam strips and tree stem to further strengthen the gas-tight seal. Each chamber also contained a gas sampling port and pressure regulator. Pressure, temperature and humidity inside the stem chamber were continuously logged (TR-73U thermo recorder; T & D Corporations, Nagano, Japan) during measurements. Tree-stem CH4 emissions were measured twice from a minimum of four trees per species (stem diameter, 7.5–19.5 cm) for the eight dominant tree species chosen randomly within each plot. The eight dominant species within the two plots were: Mesua sp. 1, Xylopia fusca Maingay ex Hook. f. & Thomson, Shorea balangeran (Korth.) Burck, Diospyros bantamensis Koord. & Valeton ex Bakh., Tristaniopsis sp. 2, Litsea elliptica Blume, Elaeocarpus mastersii King and Cratoxylum arborescens (Vahl) Blume. These tree species accounted for c. 72% of all trees within the two plots.
Methane emissions from tree stems were measured at three intervals between 20 and 130 cm height above the peat surface. Gas samples were extracted from chambers at T = 5, 20, 40, 60, 80 min for trees and T = 5, 15, 30, 45 min for peat surface measurements. Soil porewater CH4 samples were collected at two locations within each plot from 50, 100 and 150 cm peat depth using the method employed by Gauci et al. (2002). Gas and water samples were stored and transported in 12 ml pre-evacuated Exetainer vials (Labco Ltd, High Wycombe, UK) and analysed within 4 wk of collection by cavity ring-down laser spectroscopy (Los Gatos Research RMA-200 Fast Methane Analyser; Los Gatos Research, Mountain View, CA, USA) using the method of Baird et al. (2010). The minimum flux that could be detected by this method based upon instrument sensitivity and chamber volume was 0.4–3.5 μg CH4 m−2 h−1. Methane fluxes were determined by least-squares linear regression analysis of the change in CH4 concentration with time in chambers after establishing the linearity of concentration change during the measurement period. Samples that displayed R2 < 0.90 with T = 0 concentrations being close to ambient concentrations (5% of data from hollows) were judged to represent natural ebullition events and so were included when characterizing ecosystem CH4 fluxes. All gas samples were analysed in duplicate. Dissolved concentrations of CH4 were determined by analysis of CH4 in the headspace of vials containing pore water and Henry's law as described by Blodau et al. (2007).
An increment borer (internal diameter, 5.1 mm; Haglöf Sweden AB, Långsele, Sweden) was used to extract wood samples at stem heights of 35, 75, 115 and 130 cm from the eight tree species. The wood samples were collected after the tree flux measurement campaign was concluded. Specific density of the wood was calculated based upon wood dry mass and volume (Williamson & Wiemann, 2010). Wood volume was measured using a water displacement method (Archimedes principle) and wood DW by oven-drying the samples at 103°C for 48 h. The stem diameter and wood densities at breast height (1.3 m) of the eight tree species studied are listed in Table 1.
Table 1. Tree diameter (DBH ≥ 7 cm) and wood specific density measured at 1.3 m stem height above soil surface for the eight tree species studied
Tree species studied
DBH range (cm)
Wood specific density range (g cm−3)
Tristaniopsis sp. 2
Mesua sp. 1
A relationship between measured tree-stem CH4 fluxes and corresponding stem sampling height for each species was established and was used to estimate the stem fluxes of CH4 along the length of the tree. Stem circumference was also measured at 10 cm intervals between 0 and 2 m height for each tree studied within the two plots and was used to establish a relationship between stem height and circumference. This relationship was later applied to the entire length of the tree, and stem surface area of the tree was estimated by considering the tree as a truncated cone. The total CH4 emission along the length of the tree for each tree species was estimated by multiplying the CH4 fluxes by the surface area (as estimated earlier) and the total number of trees per species. Tree-stem CH4 flux per plot was estimated by dividing total stem emissions from all tree species, including the tree species that did not emit CH4, and multiplying the resulting emissions per tree by the total number of trees. This approach assumes that a similar proportion of tree species and individual trees emitting and not emitting CH4 are present in other areas of the tropical forested peatland. The stem emission rates (2.5–10.6 mg CH4 per tree d−1) were used to estimate plot-level emissions.
All statistical analyses were conducted with SPSS software v.19 (SPSS, Chicago, IL, USA) using a significance level of P <0.05. Datasets were tested for normal distribution using the Shapiro–Wilk and Anderson–Darling tests. A general linear model (ANOVA repeated measures) along with Tukey's HSD test (P ≤0.05) was used for comparison of means. Relationships between stem-CH4 fluxes, stem diameter, stem sampling height and wood specific density were evaluated using regression models. The relative contributions of independent variables (stem diameter, wood specific density and concentrations of CH4 dissolved in pore water) to stem-CH4 fluxes at different stem heights were determined using multiple regression analysis. All independent variables were first tested for multicollinearity and homoscedasticity.
Results and Discussion
Seven of the eight tree species released significant quantities of CH4 from their stems (Fig. 3), with fluxes ranging from 17.0 ± 1.4 to 185 ± 7 μg m−2 h−1 on a stem surface area basis. Measurable stem emissions were not observed from Cratoxylum arborescens, the least prevalent tree species of the eight studied within the plots. The rate of CH4 flux significantly decreased (P <0.001) with stem height above the forest floor in all species (Fig. 3). Emissions from the peat surface averaged 32.9 ± 7.8 μg m−2 h−1 for hollows and 0.7 ± 0.5 μg m−2 h−1 for hummocks. In both study plots, tree-stem CH4 fluxes measured from 20 to 50 cm stem heights on each tree were larger than peat surface CH4 fluxes. The three dominant tree species in the plots (Shorea balangeran, Mesua sp. 1 and Xylopia fusca) had the highest rates and Elaeocarpus mastersii had the lowest average rate of CH4 egress. Stem CH4 flux rates from Diospyros bantamensis, Tristaniopsis sp. 2 and Litsea elliptica were similar in magnitude and not statistically different. Stem cores extracted across a range of stem heights in a subset of trees within each plot displayed no evidence of heartwood rot, which can result in CH4 production within trees (Zeikus & Ward, 1974; Covey et al., 2012). This observation, coupled with the finding that CH4 emissions decreased with height above the forest floor for all trees studied (Fig. 3) and the presence of significant concentrations of CH4 dissolved in soil water in the plots (113–1539 μmol CH4 l−1 at 50–150 cm soil depth), indicates that the anoxic peat soil was the main source of stem-emitted CH4, minimizing the likelihood of any substantial cryptic sources (e.g., tree holes; Martinson et al., 2010). The presence of an extensive root network reaching the CH4 production zone and a well-connected root–stem path for the transport of CH4 are prerequisites for this hypothesis.
Stem CH4 fluxes varied significantly between tree species (P <0.0001), exhibiting a significant relationship with stem diameter (R2 = 0.38; P <0.001; Fig. 4a) and wood specific density (R2 = 0.47; P <0.0001; Fig. 4b). Multiple regression analysis indicates that stem diameter, wood specific density and concentration of CH4 dissolved in pore water explain up to 80% (R2 = 0.808; P <0.0001) of stem CH4 flux variations (Table 2). These relationships were observed for fluxes measured at all stem heights (20–50, 60–90 and 100–130 cm above the soil surface). Stem diameter and wood specific density were inversely related to stem CH4 flux, whereas concentrations of CH4 dissolved in pore water were positively correlated with CH4 emission rates (Table 2). The latter relationship is consistent with findings from previous studies (Rusch & Rennenberg, 1998; Terazawa et al., 2007), but the observation of an inverse relationship between stem CH4 flux and diameter and wood specific density has not been reported to date. Notably, wood specific density is a well-known indicator of the functional traits and properties of wood, including porosity and anatomical composition, and varies within individual trees and between trees, commonly being influenced by ecophysiological factors such as flooding (Parolin & Worbes, 2000; Wittmann et al., 2006a,b). Therefore, the lack of any measurable CH4 emissions from Cratoxylum arborescens was probably a result of stem properties in the tree with larger stem diameter and higher wood specific density than other trees in this study, but may also have been a result of root distribution (i.e., roots failing to reach the CH4 production zone) and/or differences in transport pathways and CH4 egress points (e.g., CH4 transport through the transpiration stream and release via leaf surfaces that were not measured here).
Table 2. Results of multiple regression analysis of stem methane (CH4) fluxes measured at three stem heights (20–50, 60–90 and 100–130 cm above the soil surface), stem diameter and wood specific density measured at corresponding stem heights, and concentrations of CH4 dissolved in pore water at 50 cm below the soil surface measured within 2.5 m radius of the trees under investigation
0.808 (P <0.0001)
0.764 (P <0.0001)
0.693 (P <0.0001)
345 (P <0.0001)
239 (P <0.0001)
154 (P <0.0001)
Stem diameter (cm)
−11.2 (P =0.002)
−8.27 (P =0.002)
−4.57 (P =0.02)
Wood specific density (g cm−3)
−323 (P <0.001)
−190 (P =0.0008)
−151 (P =0.001)
Pore water concentration (μmol CH4 l−1)
0.646 (P <0.001)
0.253 (P =0.049)
0.263 (P =0.02)
Power function relationships between the rate of tree-stem CH4 emission and stem sampling height were determined for five of the seven tree species (Supporting Information, Table S1), suggesting that the entire tree may release CH4, albeit at much lower rates from higher portions. Methane emission rates along the length of trees were estimated using regression models based upon the power function relationships; however, CH4 fluxes from only the 0.1 to 3 m bottom section of tree stems were used to determine a conservative estimate of tree-mediated CH4 emissions in the ecosystem flux calculations, pending direct measurement and confirmation of CH4 emissions from higher portions of trees.
Peat surface CH4 fluxes per plot were estimated after deducting tree basal area and using a 50 : 50 proportion of hollow vs hummock coverage. The conservative estimate of total tree-mediated CH4 flux per plot (i.e., considering only the lowermost 3 m of tree emissions) is 6.7 ± 0.7 g ha−1 d−1, which is approximately twice the flux from hollows (3.9 ± 1.0 g ha−1 d−1; Fig. 5) and c. 62% of total ecosystem flux and the largest contributor of CH4 to the atmosphere from this ecosystem. Inclusion of tree emissions to an average height of 15 m based upon the power function relationships yields total tree-mediated CH4 emissions of 28.5 ± 3.4 g ha−1 d−1 or c. 87% of total ecosystem flux. We are aware of no other study to date that reports the significance of tree-mediated CH4 emissions and their contributions at an ecosystem scale. These findings suggest that exclusion of CH4 emissions from tree stems in field studies that use only ground chambers to measure CH4 flux in forested tropical wetlands may result in significant underestimation of total CH4 emissions from the ecosystem.
Our findings are likely also to be of relevance to other tropical forested wetlands beyond Southeast Asian tropical peat forests, which only account for c. 10% of forested tropical wetlands globally. Tropical peat forests in Southeast Asia are known to emit less CH4 than nutrient-rich tropical wetlands (Wassmann et al., 1992), because, in the latter soil, pH is higher (Bartlett et al., 1988; Koschorreck, 2000), CH4 production is greater and methanotrophy is generally less effective because of increased anoxic and stratified-water submerged sediments (Devol & Rickey, 1990; Bartlett et al., 1988; Koschorreck, 2000) as a result of higher water-table levels, in particular, within seasonally inundated wetlands where soils are submerged for prolonged periods and water column productivity contributes labile biomass to bottom sediment (Devol & Rickey, 1990 and references within). The relative proportions of CH4 flux via tree stems, the soil surface, herbaceous plants, and ebullition (i.e., release of CH4-rich gas bubbles) will almost certainly differ in other types of forested tropical wetland both spatially and seasonally, depending upon moisture regime. However, there are key similarities between all forested tropical wetlands that are likely to ensure a significant role for wetland-adapted trees in mediating CH4 flux.
First, the development of morphological adaptations to aerate root systems is a common feature in trees that inhabit seasonally or permanently wet soil (Kozlowski, 1997; Parolin et al., 2006). To date, the majority of tree species investigated that possess adaptive structures to facilitate O2 ingress also are capable of mediating CH4 egress (Rusch & Rennenberg, 1998; Vann & Megonigal, 2003; Purvaja et al., 2004; Terazawa et al., 2007; Gauci et al., 2010; Rice et al., 2010). Notably, six of the eight tree species investigated in this study in Borneo belong to families that are widely distributed amongst Amazonian wetlands (Elaeocarpaceae (Elaeocarpus mastersii), Ebenaceae (Diospyros bantamensis), Myrtaceae (Tristaniopsis sp. 2), Clusiaceae or Guttiferae (Mesua sp. 1), Lauraceae (Litsea elliptica), Annonaceae (Xylopia fusca); (Parolin et al., 2006; Wittmann et al., 2006a,b; Saatchi et al., 2007; Macía, 2011). Also, the wood specific densities of the related Amazonian wetland tree species correspond with the range reported in this study (0.22–0.87 g cm−3, Parolin & Worbes, 2000; Wittmann et al., 2006a,b). Moreover, it is well established that trees inhabiting Amazonian varzeas generally exhibit morphological adaptations that facilitate gas transport during periods of inundation (Parolin et al., 2006; Graffmann et al., 2008). Hence, there is considerable evidence to suggest that most wetland-adapted trees possess structures that enable CH4 egress from soil.
Second, wetland-adapted trees do not appear to be limited in their capacity to transport CH4 (Rusch & Rennenberg, 1998; S. R. Pangala et al., unpublished), but rather the amount of CH4 present in the subsurface is a more critical factor determining rates of CH4 flux from tree stems (Rusch & Rennenberg, 1998; Terazawa et al., 2007; Rice et al., 2010; S. R. Pangala et al., unpublished). Mesocosm experiments on common alder saplings by Rusch & Rennenberg (1998) and S. R. Pangala et al. (unpublished) demonstrate a strong positive linear relationship between CH4 concentrations in the rhizosphere and tree-stem CH4 fluxes. Notably rates of CH4 egress from tree stems in mesocosms greatly exceed in situ flux rates measured in this study, because rhizosphere concentrations of CH4 are artificially elevated in the mesocosm studies. In the Southeast Asian tropical peat forest, pore water from 0 to 50 cm depth contained a maximum concentration of 123 μmol CH4 l−1. The amount of CH4 in deeper peat in the Borneo peatland was greater (113–1539 μmol CH4 l−1 from 50 to 150 cm depth); however, c. 83% of root biomass occurs within 0–30 cm depth in the tropical peat forest and root abundance decreases exponentially with depth (Sulistiyanto et al., 2004; Jauhiainen et al., 2005; Verwer & van der Meer, 2010). By contrast, more nutrient-rich tropical wetlands typically contain higher concentrations of CH4 in shallow pore water. For example, shallow soil (0–30 cm depth) in Amazonian wetlands has been reported to contain dissolved CH4 concentrations of 175–1380 μmol CH4 l−1 (Bartlett et al., 1988; Koschorreck, 2000). High concentrations of CH4 in shallow soil are particularly common where standing water is present, because it impedes entry of O2 to support methanotrophy (Bartlett et al., 1988; Koschorreck, 2000). Ebullition may become an important pathway under such conditions (Wassmann et al., 1992; Bartlett et al., 1988; Koschorreck, 2000); however, high concentrations of CH4 at shallow depths, coupled with low O2 concentrations and the need for trees to aerate their root zone, present all the elements required for tree-mediated CH4 flux.
While our data demonstrate that there is significant potential for tree-mediated CH4 emission in other types of tropical forested wetlands, the actual contribution of CH4 export via trees to total ecosystem flux remains unknown. The majority of ground-based CH4 emission studies in tropical wetlands have been conducted using soil chambers and, as a result, tree-mediated CH4 fluxes are absent in scaled surface estimates of CH4 emissions. Notably, characterization of tree-mediated CH4 fluxes in other types of tropical forested wetland may help to reconcile discrepancies that currently exist between scaled ground-based CH4 fluxes and an unexplained excess of tropical atmospheric CH4 observed in atmospheric and space-borne measurements (Chen & Prinn, 2005; Miller et al., 2007; Frankenberg et al., 2008). Our findings may be particularly important given that other tropical CH4 sources suggested recently to account for the inconsistency between bottom-up and top-down inventories have been shown to be negligible globally (e.g., UV-driven aerobic fluxes from plants (Bloom et al., 2010) and tank bromeliads in tree canopies (Martinson et al., 2010)).
Process-based global emission models simulate CH4 production as a function of net primary productivity (NPP) and respiration (Walter & Heimann, 2000) and thus implicitly include emissions derived from productivity and decomposition processes in forests (Spahni et al., 2011). Such models typically generate CH4 emission estimates that are larger than scaled field measurements and which are more similar to estimates derived from inverse methods (Spahni et al., 2011). However, process-based models at present do not discriminate between herbaceous- and tree-mediated transport of CH4 (Walter et al., 2001) and some do not define pathways by which soil-produced CH4 is exported to the atmosphere (Spahni et al., 2011). Moreover, current models are parameterized based upon CH4 flux measurements from low herbaceous wetland canopies (Walter & Heimann, 2000) and consequently may not respond correctly when subjected to different environmental stimuli. For example, tropical forests possess dense multilayered canopies that are sensitive to variation in diffusive light; small increases in incident light intensities on normally shaded leaves stimulate NPP (Mercado et al., 2010), whereas no such interaction exists in northern wetlands dominated by short shrubs (Letts et al., 2005). If tree-mediated CH4 fluxes are a dominant contributor to ecosystem CH4 emissions from tropical forested wetlands, as suggested by our data, then there is a need for explicit inclusion of trees and relevant physiological responses in process-based emission models otherwise the capacity for such models to predict the effects of environmental change on trace gas fluxes may be limited. Accurate modelling of interannual variability in CH4 emissions and the long-term effects of climate change on CH4 fluxes from the tropics may rely upon parameterization of subtle responses of wetland-adapted trees to moisture and temperature.
Finally, current protocols for CH4 measurement in forested wetlands may require revision if we are to reduce uncertainties in global CH4 source estimates and provide accurate accounting of greenhouse gas exchange under different land-use scenarios (with potential economic consequences under the United Nation's Reducing Emissions from Deforestation and Forest Degradation (REDD) programme (www.un-redd.org)). The role of trees in the CH4 cycle should not, however, excuse deforestation, because our measured tree-mediated CH4 flux expressed in CO2 equivalents represents < 2% of total carbon emissions from deforested tropical peat forests (Hirano et al., 2007). Foremost, our study underscores the need for further study of tree-mediated CH4 emissions to determine whether wetland-adapted trees normally dominate ecosystem CH4 fluxes in all types of forested tropical wetland.
This study was supported by the Open University Centre for Earth, Planetary, Space and Astronomical Research, an Open University Charter Studentship to S.R.P., and Royal Society support to V.G. We thank Suwido Limin, Idrus Mohammed and Kitso Kusin for field assistance, and Susan Page, David Gowing and Sophie Green for valued contributions. We are indebted to Patrick Megonigal, Tim Moore and two anonymous reviewers for their helpful comments on earlier versions of this manuscript. Author contributions: V.G. conceived the study; S.R.P., E.R.C.H. and V.G. designed the study, which was carried out by S.R.P. and S.M.; S.R.P, E.R.C.H. and V.G. all contributed to the writing of the manuscript, on which S.M. also commented.