An Approach for Mapping Peatland Distribution in Boreal, Temperate, Alpine and Tropical Zones using Multi-date, Multi-sensor SAR and Optical Data
Laura Louise Bourgeau-Chavez, Michigan Technological University, lchavez@mtu.edu (Presenter)
Michael Battaglia, Michigan Technological University, mjbattag@mtu.edu
Sarah Endres, Michigan Tech Research Institute (MTRI), slendres@mtu.edu
Jeremy A Graham, Michigan Tech Research Institute, jeremyg@mtu.edu
Rod Chimner, Michigan Technological University, rchimner@mtu.edu
Erik Lilleskov, Michigan Technoligical University, elillesk@mtu.edu
Dorthea J Leisman, Michigan Technological University, djleisma@mtu.edu
Peatlands occur in ecozones from the tropics to the arctic, and are estimated to globally cover just under 450 million ha, roughly 3–5% of the Earth’s land surface. Although they cover a small amount of land globally, peatlands are estimated to store more than 30% of the Earth’s soil carbon (C) and are at risk from land use and climate change. Knowledge of peatland type and distribution is crucial to balancing the global C cycle, understanding their vulnerability to change and for planning conservation and restoration activities. Peatlands are difficult to detect and map, and many regions of the world lack geospatial maps or are only recently discovering peatlands. An approach of multi-source radar and optical imagery from multiple dates was first demonstrated to be of high value in accurately mapping peatlands in regions of boreal North America. Since then it has been demonstrated as useful in temperate and tropical regions as well. The approach integrates field data collected from plot sizes that are commensurate with the resolution of the imagery and interpretation of high resolution aerial or satellite imagery to create suitable training and testing data. Next these training data are used in a machine learning algorithm with multiple dates of L- and/or C-band satellite SAR data combined with optical and DEM derivatives. The multi-date SAR data allows for inclusion of hydrological conditions (wet soil, low to high inundation) from different seasons while the multi-date optical imagery captures the vegetation phenology. Maps and accuracy statistics will be shown for vast peatland areas of Alberta and Northwest Territories Canada, as well as lowland areas of Peru, alpine regions of Ecuador, Peru and Colombia, and peatlands of the Great Lakes Region. Overall, accuracies have been greater than 93% for each of these regions.
Presentation: iPoster
Associated Project(s):
Presentation Type: Poster
Science Theme: Vegetation Dynamics and Distribution