A simple approach to upscale methane emissions from peatlands using Planetscope satellite data and machine learning algorithm

Peatland plays a significant role in methane (CH4) emissions, and methane dynamics are governed by ecohydrological variables and site heterogeneity. Emission quantification from different stages of peatland is vital to understanding the impacts of peatland on climatic feedbacks for effective rehabilitation of these sensitive ecosystems. Chamber measurement and eddy covariance techniques are widely used to understand methane dynamics. These measurements are either at a point or footprint scale, making it challenging to upscale these emissions to the site scale considering the heterogeneity of peatlands. Here, we present a simple approach to upscale methane emissions from closed chambers using PlanetScope high-resolution satellite data along with the random forest algorithm and weighted-area approach. This methodology was tested at three peatlands covering near-natural, under-rehabilitation, and degraded sites in Ireland for a span of two years. The annual vegetation maps were mapped with an accuracy of 83% at the near-natural site and around 98-99% at the under-rehabilitation and degraded sites. The highest site-scale fluxes were observed at the near-natural site (2.25 and 3.80 gC m−2 y−1), and the site-scale fluxes were close to net zero for the under-rehabilitation (0.17 and 0.31 gC m−2 y−1) and the degraded site (0.15 and 0.27 gC m−2 y−1). As a step forward, this approach will be applied to upscale eddy covariance fluxes from three fen sites in the Netherlands. Overall, the easy-to-implement methodology proposed in this study shows potential to apply it across various heterogeneous land-use types to assess the impact of peatland rehabilitation on methane emissions.