A map of global peatland extent created using machine learning (Peat-ML)

• Published in: Geoscientific Model Development Vol. 15; no. 12; pp. 4709 - 4738
• Main Authors: Melton, Joe R., Chan, Ed, Millard, Koreen, Fortier, Matthew, Winton, R. Scott, Martín-López, Javier M., Cadillo-Quiroz, Hinsby, Kidd, Darren, Verchot, Louis V.
• Format: Journal Article
• Language: English
• Published: Katlenburg-Lindau Copernicus GmbH 20.06.2022; Copernicus Publications
• Subjects: Aerial photography; Algorithms; Analysis; Aquatic birds; Autocorrelation; Carbon; Carbon content; Carbon cycle; Climate change; Climate effects; Climate models; Climatic changes; Datasets; Errors; Freshwater; Geomorphology; Hydrologic cycle; Hydrological cycle; Hydrology; Inland water environment; Learning algorithms; Machine learning; Mapping; Mathematical models; Modelling; Peat; Peat-bogs; Peatlands; Regions; Remote sensing; Soil; Soils; Statistical models; Surface water; Vegetation; Waterfowl; Wetlands
• ISSN: 1991-9603; 1991-959X; 1991-962X; 1991-9603; 1991-962X
• DOI: 10.5194/gmd-15-4709-2022

Peatlands store large amounts of soil carbon and freshwater, constituting an important component of the global carbon and hydrologic cycles. Accurate information on the global extent and distribution of peatlands is presently lacking but is needed by Earth system models (ESMs) to simulate the effects of climate change on the global carbon and hydrologic balance. Here, we present Peat-ML, a spatially continuous global map of peatland fractional coverage generated using machine learning (ML) techniques suitable for use as a prescribed geophysical field in an ESM. Inputs to our statistical model follow drivers of peatland formation and include spatially distributed climate, geomorphological and soil data, and remotely sensed vegetation indices. Available maps of peatland fractional coverage for 14 relatively extensive regions were used along with mapped ecoregions of non-peatland areas to train the statistical model. In addition to qualitative comparisons to other maps in the literature, we estimated model error in two ways. The first estimate used the training data in a blocked leave-one-out cross-validation strategy designed to minimize the influence of spatial autocorrelation. That approach yielded an average r2 of 0.73 with a root-mean-square error and mean bias error of 9.11 % and −0.36 %, respectively. Our second error estimate was generated by comparing Peat-ML against a high-quality, extensively ground-truthed map generated by Ducks Unlimited Canada for the Canadian Boreal Plains region. This comparison suggests our map to be of comparable quality to mapping products generated through more traditional approaches, at least for boreal peatlands.