Understanding Arctic terrestrial ecosystem carbon dynamics is important as these ecosystems cover large areas, store large amounts of soil organic carbon and are projected to be most affected by global warming. Mapping and modeling carbon sink-source relationships at landscape scale in these highly variable environments are a key challenge for improved representation of these ecosystems in earth system models.
This project aims to resolve local scale landscape heterogeneity of carbon dynamics in Arctic terrestrial ecosystems by combining field observations with high spatial and spectral resolution datasets derived from drones and satellite imagery. This way we want to bridge the tremendous spatial mismatch from plot scale ecological measurements and experiments, over landscape scale understanding of the ecological processes to global scale monitoring programs and climate projections using earth system models. Machine-learning will be used as a technical mean to connect variables across large spatial datasets and to investigate high information density in hyperspectral datacubes. Our work is focused on the Abisko region.
Siewert, M. B.,(in press) 2018: High-resolution digital mapping of soil organic carbon in permafrost terrain using machine-learning: A case study in a sub-Arctic peatland environment, Biogeosciences, https://doi.org/10.5194/bg-2017-323.