EGU26-15816, updated on 14 Mar 2026
https://doi.org/10.5194/egusphere-egu26-15816
EGU General Assembly 2026
© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.
Poster | Monday, 04 May, 10:45–12:30 (CEST), Display time Monday, 04 May, 08:30–12:30
 
Hall X1, X1.120
Mapping Anthropogenic Disturbances in Canadian Boreal Peatlands using Satellite Imagery and a Machine Learning
Sanghyeon Song1, Oliver Sonnentag2, Mary Kang3, Matthew Fortier4, Mélisande Teng5, and Michelle Lin6
Sanghyeon Song et al.
  • 1Department of Civil Engineering, McGill University, Montréal, Canada (sanghyeon.song@mail.mcgill.ca)
  • 2Département de Géographie, Université de Montréal, Montréal, Canada (oliver.sonnentag@gmail.com)
  • 3Department of Civil Engineering, McGill University, Montréal, Canada (mary.kang@mcgill.ca)
  • 4Mila - Quebec Artificial Intelligence Institute, Montréal, Canada (fortier.matt@gmail.com)
  • 5Mila - Quebec Artificial Intelligence Institute, Montréal, Canada (tengmeli@mila.quebec)
  • 6Mila - Quebec Artificial Intelligence Institute, Montréal, Canada (michelle.lin@mila.quebec)

Peatlands cover 12% of Canada's territory, primarily in the boreal and Arctic regions, and have the capacity to absorb and store significant amounts of carbon. They are therefore gaining attention as a nature-based climate solution, which involves protecting, managing, and restoring natural ecosystems from further degradation, thereby reducing greenhouse gas emissions. Canadian peatlands have been disturbed by human activities, which can convert peatlands from net carbon sinks (absorbing more than they release) into net carbon sources (releasing more than they absorb), and these disturbances have intensified in recent years. In this context, mapping anthropogenic disturbances in peatlands is crucial for effective monitoring and management of peatlands and ensuring they continue to serve as carbon sinks. Available maps of anthropogenic disturbances in peatlands are typically confined to specific regions, disturbance types, or time periods. Therefore, our goal is to develop a comprehensive Canada-wide map of anthropogenic disturbances, covering both historical and recent periods. To do this, we develop an automated framework for mapping current and historical anthropogenic disturbances in Canadian boreal peatlands. The framework leverages machine-learning–based image segmentation models applied to Landsat satellite imagery and is designed to process the full 40-year (1984 to 2024) satellite archive to generate multi-class disturbance maps (i.e., agriculture, forestry, resource extraction, transportation, industry, residential, seismic lines) across multiple decades. By comparing disturbance maps through time, the spatiotemporal dynamics of anthropogenic disturbances in boreal peatlands can be examined. The resulting maps provide a foundation for improved understanding of peatland disturbance patterns and support researchers investigating peatland–climate interactions, government agencies developing policies for peatland protection and restoration, and Indigenous communities working to safeguard their traditional lands.

How to cite: Song, S., Sonnentag, O., Kang, M., Fortier, M., Teng, M., and Lin, M.: Mapping Anthropogenic Disturbances in Canadian Boreal Peatlands using Satellite Imagery and a Machine Learning, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15816, https://doi.org/10.5194/egusphere-egu26-15816, 2026.