A DECISION-TREE CLASSIFICATION APPROACH FOR MAPPING NATURAL HABITATS USING HIGH-RESOLUTION REMOTELY SENSED IMAGERY

High-resolution imagery may be useful for mapping and monitoring land cover and land use change, but requires the development of accurate and repeatable techniques that can be extended to a broad range of environments and conditions. The development of these techniques is necessary to establish baseline conditions and monitor change effects so that efficient environmental management practices can be designed. In this study, a method was developed for mapping natural habitats using a Boolean classification system based on a decision tree approach. Decision tree classifiers have a substantial advantage over traditional classification methods due to their nonparametric nature, simplicity, and computational efficiency. The decision tree classification scheme was used to map land cover and land use in the St. Tammany Parish, LA area, a region comprised of a diverse range of natural habitats. A combination of field data, NASA Airborne Visible and Infrared Imaging Spectrometer imagery (AVIRIS), Digital Globe Quickbird satellite imagery, and supporting geographic information system (GIS) coverages, including land use maps from the 1988 National Wetlands Inventory (NWI) and National Land Cover Database (NLCD) from the National Oceanic and Atmospheric Administration were used as input to the decision tree classifier. Classifications were performed using 1982 and 2004 imagery for the following land cover classes: marsh, wetland forest, upland forest, agricultural-grassland, shrub-scrub, urban, and water. This report discusses the development of these maps and methods, including the results in land cover and land use change.