Machine learning with satellite imagery to document the historical transition from topographic to dense sub-surface agricultural drainage networks (tile drains)

Image library of (1) tile-drained landscapes and (2) tile-drain types used for training a machine-learning model that identifies (1) tile-drained landscapes and (2) differentiates two types of tile-drained areas visible in satellite imagery. These images were sourced from WorldView, Quickbird, and GeoEye satellite imagery (copyright DigitalGlobe) and cropped to features of interest. Imagery has a ground resolution of 0.34 - 0.65 m.