Identification of Artificial Levees in the Contiguous United States

Abstract Artificial levees are anthropogenic structures designed to hydrologically disconnect rivers from floodplains. The extent of artificial in the contiguous United States (CONUS) is unknown. To better estimate distribution levees, we tested several different geomorphic, land cover, and spatial variables developed National Elevation Dataset, Land Cover database, Hydrology Dataset HR Plus. We used known levee locations Levee Database as training data. machine learning general logistic models’ ability detect a 100‐year hydrogeomorphic floodplain seven geographically diverse 8‐digit HUC basins. Random forest models outperformed other predicting location using representing geomorphic attributes, distance streams ranging size between stream order one through six. demonstrate our approach unknown conducted leave‐one‐out cross‐validation lower Mississippi Basin approximately 1,100 levees. This detected constituting 94% total length basin. Scaling up CONUS, applied high performing (overall accuracy 97%) random model cover variables. 182,213 km potential mostly along 2–6 Missouri River Basins, indicating that national database contains 20.4% length. Potential those documented modify 2% States.