Abstract

The accurate estimation of methane, a greenhouse gas, from natural sources such as peatlands is vital for overall assessment of global climate change. Investigations that rely on process-based models that derive CH4 emissions from natural wetlands as a function of variables such as soil temperature, water table depth and aerenchymateous vegetation measured at plot scales may produce biased results when extrapolated to landscape, regional and global levels. This is because assumptions such as average water table depth, vegetation patchiness, and microclimate that must be ‘scaled up’ from the plot to larger scales. Heterogeneity is an inherent characteristic of wetlands, and the relationships between CH4 and variables deriving methane emissions are non-linear. These limitations can be overcome by independent remote sensing data with multi-spatio-temporal resolutions acquired from different sensors and platforms. This paper discusses novel techniques from information theory applied to preserve the information content of edaphic and vegetative characteristics of a peatland ecosystem in Scotland. Our results demonstrate shifting patterns in the information retrieved as the pixel resolution decreases. We report the changes in processes and patterns across different scales which would help improve the overall performance of the process based models at global scales.