Individual tree CH4 flux observations were made as follows in each of the following sites. Chambers used at all sites were constructed from gas-impermeable polyethylene terephthalate or polycarbonate plastic sheet. The chamber characteristics, accuracy and precision of the method when used either in manual syringe mode for later analysis on Los Gatos analysers or in continuous mode (with in situ analysis by means of field portable microportable and ultraportable greenhouse gas Los Gatos analysers (MGGA and UGGA, respectively)) are detailed in ref. 22. To summarize, and unless specified below, two ‘sleeve’ chamber sizes were used, small (25 × 16 × 1.5 cm3) and large (30 × 24 × 1.5 cm3), depending on the dimensions of the woody surface being measured. All fluxes were measured in the middle part of the day (10:00–15:00) or at times between 09:00 and 18:00. Fluxes measured in continuous flow mode had variable deployment and measurement periods, depending on the rate of concentration change, which was observed in real time. A photograph of a typical chamber deployment is presented in Extended Data Fig. 1. The minimum flux that could be detected using the modified fast methane analyser (FMA) analysis method32 based on instrument sensitivity and chamber volume was 0.4–3.5 µg of CH4 m2 h−1. The minimum flux that could be detected using the LGR UGGA and the LGR MGGA in real-time measurements based, respectively, on instrument sensitivities of 4 ppb with 1 s precision and 2 ppb with 1 s precision, as well as on chamber volume, is less than 1 µg of CH4 m2 h−1.
Cuniã Nature Reserve, Amazonia, Brazil (63° 5′ W, 8° 1′ S)
Stem CH4 emissions from mature trees (equal to or more than 10 cm diameter) were measured from two free-draining forested plots. The two plots (20 × 30 m2) were located in the Madeira River catchment, a white-water river system and one of the largest tributaries of the Amazon. The Madeira drains the Andean area upstream, which results in high suspended and dissolved solids concentrations in water, with neutral to alkaline pH33,34. The MAT is 26 °C and mean annual rainfall is 2,500 mm (ref. 35).
Methane flux measurements from mature trees stems (n = 50 per plot) at five stem heights (20, 60, 100, 140 and 180 cm above the soil surface) were performed during a period of transition towards the dry season when the water tables in both the plots were more than 10 m below the soil surface. Measurements were carried out between 15 and 25 March 2013. Methane fluxes were measured using static chambers as described in refs. 3,22, with air samples from the static flux chambers drawn using 30 ml syringes and immediately transferred to a 12 ml exetainer (Exetainer) for later analysis of CH4 using the modified LGR CH4 laser-based analyser3 (LGR UGGA).
Gigante Peninsula, Barro Colorado Nature Monument, Panama (9° 6′ N, 79° 54′ W)
Measurements in semi-evergreen tropical forest were carried out between 18 and 27 November 2015 in the five control plots of the Gigante Litter Manipulation Project, approximately 5 km south of Barro Colorado Island, Panama, Central America. A full description of the litter manipulation experiment is given in refs. 36,37. The MAT at the weather station on Barro Colorado Island is 26 °C, mean annual rainfall is 2,600 mm and there is a strong dry season from mid-December to mid-April38.
We measured tree stem CH4 fluxes at 30, 75, 130 and 200 cm height from two common tree species: the fast-growing canopy tree Simarouba amara (Aubl.) and the shade-tolerant subcanopy tree Heisteria concinna (Standl.) (12 trees per species). Tree stem gas fluxes were measured using a flexible chamber (45 cm × 30 cm × 19 mm polycarbonate) as described in refs. 4,20. Gas samples were taken by syringe from a septum in the middle of the chamber at 0, 5, 10 and 15 min and injected into pre-evacuated 12 ml borosilicate vials. All samples were analysed in the United Kingdom using off-axis integrated cavity output spectroscopy (FMA-200 fast methane analyser).
Wytham Woods, Oxfordshire, United Kingdom (51° 46′ 42″ N, 1° 19′ 42″ W)
Measurements at the temperate site were conducted in four control plots of a litter manipulation experiment39,40 at Wytham Woods, an old growth (about 120 yr) mixed deciduous woodland in Oxfordshire, United Kingdom. The canopy at the study site is dominated by ash (Fraxinus excelsior L.), beech (Fagus sylvatica L.), sycamore (Acer pseudoplatanus L.) and oak (Quercus robur L.)41. MAT was 10 °C (ref. 40). In each 25 × 25 m2 plot, three individuals each of ash and sycamore were randomly selected, making a total of 24 trees. Tree stem CH4 fluxes were sampled using the same chamber design, sampling heights and procedure as described for Gigante above, except that gas samples were collected at 0, 3, 6 and 10 min. All samples were analysed in the United Kingdom using off-axis integrated cavity output spectroscopy (FMA-200 fast methane analyser).
Skogaryd, Sweden
The stem and soil flux measurements were conducted in the Skogaryd Research Catchment, near Vänersborg in southwestern Sweden, on three occasions in spring 2014 (2–3 April, 14–16 April and 28–30 April). Tree stem CH4 fluxes were measured from Norway spruce (Picea abies (L.) Karst.) and Scots pine (Pinus sylvestris (L.)), based on the dominance of these tree species in the European boreal forest42 and at the study site. Methane fluxes from tree stems were measured in real time using static chambers connected to a laser-based CH4 analyser (LGR UGGA) as described in ref. 3. Stem CH4 emissions were measured from mature tree stems (equal to or more than 10 cm, n = 9 per species) from three stem heights (20, 60 and 100 cm above the soil surface). The water table was more than 5 m below the surface in the plot and soil was characterized as a histosol. The mean air temperatures during the three sampling occasions were 13.1, 14.2 and 18.1 °C, respectively, with all measurements carried out during daytime (09:00–18:00). MAT is 6.9 °C.
Following stem flux measurements at the Skogaryd and Cuniã study sites, we assessed the CH4 oxidation potentials (both high affinity and low affinity) in the tree stems by extracting wood cores at 30 and 130 cm stem height from a subset of the trees (Cuniã, 80 trees; Skogaryd, 9 trees per species).
Wood cores were extracted at four cross-sections radially from bark to pith at the same location on the stem where CH4 flux measurement were performed using a 5.1 mm increment corer. These cores were immediately transferred into a 50 ml vial and incubated on the same day of sampling to quantify the potential rates of CH4 production and oxidation. Incubations were carried out onsite in the dark, at field temperature of 25–27 °C and 11–15 °C, respectively. An initial headspace CH4 concentration of 6.5 and 750 ppm was maintained for high- and low-affinity CH4 oxidation potentials, respectively, and the cores were incubated aerobically for 48 h. Headspace samples from all incubations were extracted at 4, 24 and 48 h and CH4 concentrations analysed using methods described in refs. 3,12.
Further measurements made in the Amazon floodplain
Sampling design and measurement protocols for fluxes presented in Fig. 2 are detailed in refs. 19,20. In summary, we established three temporary plots (60 × 60 m2) in the floodplains of three principal rivers of the Amazon, the Negro (black water), Solimões (white water) and Tapajós (clear water). The CH4 fluxes were measured from a total of 108 trees (36 across each plot) at vertical intervals above the forest floor during low water in January 2018. We returned in the exceptional dry season of 2021 (October) to make measurements from a subset of trees at 5 m above the forest floor and also to sample for methane isotopes.
Chamber CH4 isotopes in the Amazon
For δ13C-CH4 analysis, 30 ml gas samples were collected from tree woody surfaces and soil surfaces at the Negro and Solimões floodplain forests during the dry season of 2021. Samples were taken from air and from flux chambers on the soils surface and tree stem surface at 5 m above the forest floor using gas-tight syringes and then transferred to pre-evacuated 12 ml borosilicate vials fitted with double wadded caps (Exetainer). Vials were over-pressurized to prevent ingress of air from pressure or temperature changes during transport to the laboratory. The δ13C values of CH4 were analysed using a cavity ring-down spectrometer (model G2201-i, Picarro) coupled with a custom-built auto-sampler and are reported relative to the Vienna Pee Dee Belemnite standard. The instrument was calibrated for δ13C-CH4 using isotopic reference gases with isotope ratios of −23.9‰, −54.5‰ and −66.5‰ (Isometric Instruments). The overall analytical precision based on replicate measurements of reference gases was ±0.4‰.
CH4 uptake global estimate methods
Our study leveraged TLS technology to develop a new surface area allometry43. TLS provides high-resolution, three-dimensional representations of tree structures, enabling precise surface area estimations. We scanned a total of 2,161 trees across 22 plots in tropical forests, temperate conifer forests, temperate broadleaf forests, temperate dry eucalypt forests and tropical savannahs to capture a wide range of tree morphologies. We built two woody area index (WAI) allometries (one for tropical forests, one for the rest of the world) to predict woody surface area from tree stem diameter (diameter at breast height, DBH) based on three-dimensional models of trees43. Each tree was scanned from several angles to ensure comprehensive coverage and cylindrical models were fit with TreeQSM44. These models were analysed using the treestruct R package45, which resulted in woody surface area and DBH for each tree. We then applied hierarchical generalized additive models46 to build the surface area allometric equations. These models were designed to capture the complex, nonlinear relationships between tree surface areas and their respective DBH (Extended Data Fig. 3).
To scale our TLS-derived allometry to forest plots, we integrated tree census data from 70 global ecosystem monitoring plots that were chosen to have close to 100% canopy cover and be generally greater than 5 m tall. Six Centre for Tropical Forest Science plots across the tropics were also used, which, in addition to their greater than 10 cm census, included trees down to 1 cm diameter. In most plots, trees of more than 10 cm DBH were measured, and we used plots for which trees were measured down to more than 1 cm DBH (six ForestGeo plots47) to estimate the percentage contribution of trees with 1–10 cm DBH to total woody surface area. We applied our allometry to each tree in these plots, thereby estimating the total woody surface area for each plot. We then performed a weighted average, based on plot size, across all plots to determine mean surface area for tropical forests.
The integration of TLS data with forest plot census data not only validated our allometric model but also allowed the extrapolation of surface area estimates across different biomes and forest structures.
Global scaling using satellite remote sensing data
For global-scale extrapolation, we used a combination of satellite remote sensing datasets. This approach allowed us to estimate woody surface areas across the world’s ecosystems, integrating our allometric model with global forest cover data.
We used The Nature Conservancy’s Ecoregions map to determine biome extents and both the 1 km global consensus land cover project48 and the MODIS MOD44B Version 6 Vegetation Continuous Fields (VCF49) 2019 product, as well as the 1 km consensus land cover map, to determine per-pixel forest cover. This dataset provided a global view of vegetation cover, quantified on a continuous scale for each pixel.
Combining these datasets, we were able to scale our allometric model from individual trees to a global scale. For each pixel, we determine the per-hectare woody surface area based on the undisturbed forest plots in that ecoregion and applying our allometry-derived WAI. We then scaled (multiplied) that potential value by the proportion of that pixel covered by forest, according to the 1 km consensus map and MODIS VCF.
To account for climatic variations in our model, we used the ERA5 monthly climate dataset50. This dataset provides mean monthly temperatures, which we converted to MAT. With the woody surface area in each pixel known (Extended Data Table 4), we then applied the CH4 uptake versus MAT regression across the globe (Extended Data Table 4).
Finally, we aggregate the CH4 uptake across biomes to determine per-biome figures (Extended Data Table 3). The entire approach is summarized in Extended Data Fig. 2.
Global upscaling uncertainty analysis
We examined how woody surface area was influenced by various parameters, in particular the branch size fraction. Small branches and twigs carry greater uncertainty in our scaling because of challenges in estimation of their area from TLS measurements. Exclusion of twigs smaller than 2 cm in diameter from our analysis resulted in a 30% reduction in calculated uptake at the biome level and exclusion of branches smaller than 5 cm in diameter resulted in a 61% reduction in uptake (Extended Data Table 6). This represents uncertainty in our biome and global-scaled estimates, which may be reduced in future with more measurements in these smaller branch size fractions and better estimation of their surface area.
A second source of uncertainty is the relatively low uptake flux estimates derived from tropical trees at 2 m above the forest floor for our upscaling estimates, as opposed to fluxes that were twice as large when measured at 5 m (Fig. 2b), which is likely to counterbalance any such reductions from excluding small branch area size fractions. Because of the limited size of our dataset at 5 m height, we chose to use only the smaller 2 m flux values for our scaling. If the 5 m values are more typical of the whole tree, which seems plausible as most of the tree surface area is above 5 m and further away from any soil-generated methane carried through and lost from lower portions of the tree trunk, then our biome-scaled fluxes would increase by up to 100%. Hence the possible biases in small branch fluxes and flux sampling height probably work in opposite directions and cancel each other out to some extent. These uncertainties can only be reduced by a greatly expanded series of measurements of woody surface methane measurements in tropical trees at a range of heights and branch sizes, coupled with fine-scale assessment of small branch surface area.
We further considered forest structure uncertainty introduced through the WAI allometry. We considered variability in TLS-derived WAI across the 60 census plots spanning the tropics, which were used to inform the metric. The weighted mean of the surface area per hectare is 41,176 m2 (the WAI of 4.12 we applied to our estimates). Using Cochrane’s51 formula for variance of a weighted mean, we identify an s.e. of 221 m2 with the 95% confidence interval (1.96 × s.e.) of 433 m2 on either side of the mean, which, when propagated across our upscaling approach, falls well within the broader uncertainties already detailed.
A further uncertainty concerns local hydrological or humidity control of woody surface methanotrophy functioning. We have therefore eliminated water-limited biomes from our low estimates (Extended Data Table 3) and so provide a representative estimate spread that takes into account this uncertainty. Finally, there is some variability in the CH4 exchange behaviour of floodplain trees with respect to hydrology. They act as large point sources of CH4 when inundated, contributing to the comparatively well-known global wetland CH4 source, but our data show they also take up CH4 during the dry season, albeit with orders of magnitude smaller fluxes. Given the small area of tropical floodplain forests versus all tropical and subtropical moist broad-leaved forest (less than 1.5%), their contribution to global CH4 uptake is negligible.
Estimating the CO2 equivalence of methane uptake and comparisons with ecosystem C dynamics
To examine the relative importance of CH4 uptake, we compared it to the C fluxes and stocks of forests. Although metrics such as global warming potential (GWP) have long been used, a recent consensus has developed that expressing CH4 emissions as CO2 equivalent emissions using GWP-100 overstates the effect of constant CH4 emissions on global surface temperature by a factor of 3–4 (ref. 52) while understating the effect of any new CH4 source or sink by a factor of 4–5 over the 20 years following the introduction of the new source (IPCC AR6). A more accurate indication of CO2-we emissions is to equate a constant 1 t yr−1 of CH4 source that is more than 20 years old with 8 t of CO2-we yr−1 but to account for the warming (or cooling) impact of a 1 t of CH4 yr−1 step-change in CH4 emission rate by adding (or subtracting) 120 t of CO2-we yr−1 over the 20 years following the change. Hence, a new constant source (or sink) of CH4 introduced in year 1 is equated with 128 t of CO2-we for years 1–20 and 8 t of CO2-we thereafter53. For direct comparisons with ecosystem C stocks and fluxes, we converted all CO2-we values to C only.
Effects of changing forest area analysis
Tropical deforestation entails the sudden loss of this woody surface CH4 sink, corresponding to a net loss of 128 t of CO2-we per t of CH4 for the first 20 years following deforestation (Methods) or a total ‘stock loss’ of 2,560(20 × 128) t of CO2-we per t of CH4 yr−1 of the CH4 sink. Tropical forest was lost at a rate of 10.3 million ha yr−1 over the period 2002–2018 (ref. 54). This results in 0.59 Mg of CO2-we-C ha−1 yr−1 sink reduction (2.15 Mg of CO2-we ha−1 yr−1) or a total of 6.04 Tg of CO2-we-C from the act of deforestation, a small extra climate impact, dwarfed by the release of biomass C stocks (1 Pg of C yr−1). It remains unknown how quickly a methanotrophic community equivalent to that of a mature forest takes to develop, but, if it develops quickly, the CH4 sink in young secondary forests is likely to be similar to that in mature forests. Hence, the CH4 sink benefits of new forest could manifest much more quickly than the C storage benefits. Assuming a similar woody surface area and CH4 sink per hectare as for mature forests and a CO2-we of 128 for the first 20 year timeframe of interest, the tree CH4 sink would add an extra greenhouse gas mitigation value of 0.131 and 0.586 Mg of CO2-we-C ha−1 yr−1 in temperate and tropical forests, respectively, corresponding to a 7% and 12% extra climate benefit of new trees in these respective biomes.