Floodplains perform several important ecosystem services, including storing water during precipitation events and reducing peak flows, thus reducing flooding of downstream communities. Understanding the relationship between flood inundation and floodplains is critical for ecosystem and community health and well-being, as well as targeting floodplain and riparian restoration. Many communities in the United States, particularly those in rural areas, lack inundation maps due to the high cost of flood modeling. Only 60% of the conterminous United States has Flood Insurance Rate Maps (FIRMs) through the U.S. Federal Emergency Management Agency (FEMA). We developed a 30-meter resolution flood inundation map of the conterminous United States (CONUS) using random forest classification to fill the gaps in the FIRM. Input datasets included digital elevation model (DEM)-derived variables, flood-related soil characteristics, and land cover. The existing FIRM 100-year floodplains, called the Special Flood Hazard Area (SHFA), were used to train and test the random forests for fluvial and coastal flooding. Models were developed for each hydrologic unit code level four (HUC-4) watershed and each 30-meter pixel in the CONUS was classified as floodplain or non-floodplain. The most important variables were DEM-derivatives and flood-based soil characteristics. Models captured 79% of the SFHA in the CONUS. The overall F1 score, which balances precision and recall, was 0.78. Performance varied geographically, exceeding the CONUS scores in temperate and coastal watersheds but were less robust in the arid southwest. The models also consistently identified headwater floodplains not present in the SFHA, lowering performance measures but providing critical information missing in many low-order stream systems. The performance of the random forest models demonstrates the method's ability to successfully fill in the remaining unmapped floodplains in the CONUS, while using only publicly available data and open source software.
Keywords: CONUS; Ecosystem services; EnviroAtlas; Flood; Geographic information systems; Machine learning.
Published by Elsevier B.V.