Background
Canadian peatlands store substantial amounts of organic C within their peat soils, making these ecosystems globally significant carbon sinks. Yet, conventional models for peatland C cycling have been constrained to site-specific applications, and to quantify peatlands potential to act as nature-based solutions to climate change there is a need to advance these models from the site-scale to national-scale. Thereby, to enable forecasting of C fluxes in peatlands under future scenarios, or to generate real-time estimates where no in-situ measurements exist, we applied machine learning algorithms trained on in-situ fluxes from a variety of peatland sites across Canada to estimate net ecosystem exchange (i.e., the balance between carbon dioxide taken up by plants and microorganisms versus released to the atmosphere). Additional model inputs include remotely sensed or gridded climate data products, as they are widespread both geographically and temporally enabling flux estimation to be expanded to regional analyses.
Activity Outline
- Assemble a dataset of peatland physical, hydrological, and biogeochemical properties
- Use a robust machine learning model to identify key environmental drivers and predict greenhouse gas emissions
- Investigate how well the model can predict peatland greenhouse gas emissions in different regions