A Hierarchical Approach to Land Use and Land Cover Mapping Using Multiple Image Types

The main objective of this project was to derive digital land use and land cover information from satellite remote sensing data for Connecticut, Long Island, and those watersheds contributing to Long Island Sound. In doing so, four methods of land use and land cover classification were evaluated on two test quadrangles to determine which method was best suited, in terms of time and costs balanced with spatial and thematic accuracy. These four methods are: a single date Landsat Thematic Mapper (TM) computer-assisted classification, a TM/SPOT panchromatic image fusion and automated classification, a TM classification combined with on-screen digitization of SPOT panchromatic data, and a TM classification combined with on-screen digitization of Digital Orthophoto Quarter-Quadrangles (DOQQs). For Connecticut statewide land cover mapping, a layered approach was used, with each stage concentrating on a related set of land cover types. Twenty-eight detailed land use and land cover classes were mapped with an overall accuracy of nearly 80%. An accuracy of better than 88% was achieved for the six Level I categories found in Connecticut.