Mapping of peatlands in the forested landscape of Sweden using lidar-based terrain indices

Abstract

Globally, northern peatlands are major carbon deposits with important implications for the climate system. It is therefore crucial to understand their spatial occurrence, especially in the context of peatland degradation by land cover change and climate change. This study was aimed at mapping peatlands in the forested landscape of Sweden by modelling soil data against lidar-based terrain indices. Machine learning methods were used to produce nationwide raster maps at 10ĝ€¯m spatial resolution indicating the presence or not of peatlands. Four different definitions of peatlands were examined: 30, 40, 50 and 100ĝ€¯cm thickness of the organic horizon. Depending on peatland definition, testing with a hold-out dataset indicated an accuracy of 0.89-0.91 and Matthew's correlation coefficient of 0.79-0.81. The final maps showed a national forest peatland extent of 60ĝ€¯292-71ĝ€¯996ĝ€¯km2, estimates which are in the range of previous studies employing traditional soil maps. In conclusion, these results emphasize the possibilities of mapping boreal peatlands with lidar-based terrain indices. The final peatland maps are publicly available at 10.17043/rimondini-2023-peatlands-2 (Rimondini et al., 2023) and may be employed for spatial planning, estimating carbon stocks and evaluating climate change mitigation strategies.