Speaker
Description
Peatlands are globally important habitats, comprising a small fraction of the terrestrial surface yet storing large quantities of carbon. Predominantly found at high latitudes, northern hemisphere peatlands have received significantly more attention in scientific literature than their southern counterparts. The peatlands of Southern Patagonia are poorly studied, and accurate maps of their extent do not exist. These peatlands are subjected to anthropogenic disturbance in the form of peat mining and Sphagnum moss extraction. Using data from the European Space Agency’s (ESA) new BIOMASS satellite equipped with P-band Synthetic Aperture Radar (SAR), coupled with machine learning techniques, this study investigates the potential for this new technology to address the knowledge gap in our understanding of the extent of Southern Patagonian peatlands. Additionally, this study investigates if peatland disturbance can be detected using P-Band SAR. Results showed moderate success in classifying peatlands separately from other habitats (70-80% accuracy), with feasible methods identified for further refining classification accuracy and making an accurate peatland inventory. Assessing peatland type and condition returned moderately successful results for identifying devegetated peatland (60-70% accuracy) but not mined peat (13-19% accuracy). This study would benefit greatly from ground-truth information tailored to the study design. This study showcases the increasing utility of remote sensing as a viable method of answering landscape-scale questions and both the advantages and disadvantages which machine learning methods can provide in assessing large datasets.