Mapping Drainage Canals in Southeast Asian Peatlands and their Implications for Peatland Degradation

Drainage canal networks associated with agricultural land use are a major contributor to peatland degradation in Southeast Asia. These canals are used to control water table depth and make the soil suitable for planting, but their presence has the negative impact of drying out peat soils near the ground surface. Drier soils in turn cause elevated fire risk, increased carbon release to the atmosphere, and subsidence. Although canals directly impact local peat hydrology, the effect of drainage intensity (i.e. canal density) in peatlands has not been quantitatively investigated, due to a lack of reliable canal maps in the region.

In this study, we trained a machine learning model to identify drainage canals and map their density throughout Southeast Asian peatlands using remote sensing imagery. Specifically, a fully convolutional neural network was applied to RGB 5m resolution Basemap imagery from Planet. Training data was generated by hand-labeling canals from satellite images, and validation of canal density was performed via comparison to independently labeled maps. A map of canal density was then produced across ISEA peatlands using images from 2017. We compared canal density with land use type and found that mean canal density is highest in industrial plantations. We also compared canal density with fire occurrence and subsidence data. This new dataset has potential applications for studies of peatland hydrology, land use change, and fire risk.