The importance of the treatment of elevation uncertainty and landscape position for barrier island habitat mapping and modeling

Authors: Nicholas M. Enwright*, USGS Wetland and Aquatic Research Center; Louisiana State University, Department of Geography and Anthropology, Lei Wang, Louisiana State University, Department of Geography and Anthropology, Hongqing Wang, USGS Wetland and Aquatic Research Center, Michael J. Osland, USGS Wetland and Aquatic Research Center, Laura C. Feher, USGS Wetland and Aquatic Research Center, Sinéad M. Borchert, Borchert Consulting at USGS Wetland and Aquatic Research Center, Richard H. Day, USGS Wetland and Aquatic Research Center
Topics: Coastal and Marine, Remote Sensing, Geographic Information Science and Systems
Keywords: habitat mapping, habitat modeling, uncertainty, lidar, dune, marsh, wetland
Session Type: Paper
Day: 4/6/2019
Start / End Time: 8:00 AM / 9:40 AM
Room: Stones Throw 1 - Granite, Marriott, Lobby Level
Presentation File: No File Uploaded


Landscape position, such as elevation and distance from shore, influences habitat coverage on barrier islands by regulating exposure to harsh abiotic factors, including waves, tides, and salt spray. Geographers often use topographic data to extract landscape position information for research on barrier islands and beach-dune environments. Researchers should consider lidar elevation uncertainty, especially when using automated processes for extracting elevation-dependent habitats from lidar data in low-relief coastal settings. Through three case studies on Dauphin Island, AL (USA), we show how landscape position and treatment of lidar elevation uncertainty can enhance habitat mapping and modeling efforts. First, we explored how Monte Carlo simulations affected automated extraction of intertidal wetlands. Next, we extended this approach to dune habitat and integrated these steps into a workflow to map habitats, such as beach, dune, and intertidal marsh. Lastly, we applied machine learning algorithms, including K-nearest neighbor, support vector machine, and random forest, to predict habitats using landscape position information from topobathymetric data. We validated the model results using a habitat map, and we assessed how well the model can hindcast habitats using historical data. We found the treatment of lidar elevation uncertainty led to an 80 percent increase in the areal coverage of intertidal wetlands when using automated extraction. For habitat modeling, we achieved the best results using random forest, which had a deterministic overall accuracy of about 67 percent and a fuzzy overall accuracy of 82 percent. Our research should interest scientists concerned with monitoring and forecasting habitats in dynamic coastal environments.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login