Wetlands are the largest global source of atmospheric methane (CH), a potent greenhouse gas. However, methane emission inventories from the Amazon floodplain, the largest natural geographic source of CH in the tropics, consistently underestimate the atmospheric burden of CH determined via remote sensing and inversion modelling, pointing to a major gap in our understanding of the contribution of these ecosystems to CH emissions. Here we report CH fluxes from the stems of 2,357 individual Amazonian floodplain trees from 13 locations across the central Amazon basin. We find that escape of soil gas through wetland trees is the dominant source of regional CH emissions. Methane fluxes from Amazon tree stems were up to 200 times larger than emissions reported for temperate wet forests and tropical peat swamp forests, representing the largest non-ebullitive wetland fluxes observed. Emissions from trees had an average stable carbon isotope value (δC) of -66.2 ± 6.4 per mil, consistent with a soil biogenic origin. We estimate that floodplain trees emit 15.1 ± 1.8 to 21.2 ± 2.5 teragrams of CH a year, in addition to the 20.5 ± 5.3 teragrams a year emitted regionally from other sources. Furthermore, we provide a 'top-down' regional estimate of CH emissions of 42.7 ± 5.6 teragrams of CH a year for the Amazon basin, based on regular vertical lower-troposphere CH profiles covering the period 2010-2013. We find close agreement between our 'top-down' and combined 'bottom-up' estimates, indicating that large CH emissions from trees adapted to permanent or seasonal inundation can account for the emission source that is required to close the Amazon CH budget. Our findings demonstrate the importance of tree stem surfaces in mediating approximately half of all wetland CH emissions in the Amazon floodplain, a region that represents up to one-third of the global wetland CH source when trees are combined with other emission sources.

No abstract

Inland waters (lakes, rivers and reservoirs) are now understood to contribute large amounts of methane (CH4) to the atmosphere. However, fluxes are poorly constrained and there is a need for improved knowledge on spatiotemporal variability and on ways of optimizing sampling efforts to yield representative emission estimates for different types of aquatic ecosystems. Low-latitude floodplain lakes and wetlands are among the most high-emitting environments, and here we provide a detailed investigation of spatial and day-to-day variability in a shallow floodplain lake in the Pantanal in Brazil over a five-day period. CH4 flux was dominated by frequent and ubiquitous ebullition. A strong but predictable spatial variability (decreasing flux with increasing distance to the shore or to littoral vegetation) was found, and this pattern can be addressed by sampling along transects from the shore to the center. Although no distinct day-to-day variability were found, a significant increase in flux was identified from measurement day 1 to measurement day 5, which was likely attributable to a simultaneous increase in temperature. Our study demonstrates that representative emission assessments requires consideration of spatial variability, but also that spatial variability patterns are predictable for lakes of this type and may therefore be addressed through limited sampling efforts if designed properly (e.g., fewer chambers may be used if organized along transects). Such optimized assessments of spatial variability are beneficial by allowing more of the available sampling resources to focus on assessing temporal variability, thereby improving overall flux assessments.

No abstract