Assessing high-resolution remotely sensed data for mapping and monitoring wetlands in southern Ontario, Canada

Wetlands within the Great Lakes basin are among the most productive ecosystems in this region of North America. Despite the many functions and values of these ecosystems, human activities over the past 200 years have resulted in significant wetland degradation and loss within the Great Lakes basin. Wetland management agencies (e.g., Ducks Unlimited Canada and the Canadian Wildlife Service, Environment Canada) intend to develop procedures for detailed mapping and monitoring of wetlands; procedures that are uniform, consistent and can be readily applied to large geographic areas, such as the Great Lakes basin.; The overall goal of this study is to develop improved methods for wetland mapping and monitoring in the Great Lakes basin at levels of detail necessary for management purposes. Using test sites from an inland wetland complex in South Dumfries Township and a shoreline wetland site located along the north shore of Lake Erie, this study focused on assessing the capabilities of high-resolution remote sensing instruments (i.e., IKONOS and the Compact Airborne Spectrographic Imager (casi)) to provide detailed information about wetland vegetation and adjacent land-use and land-cover types at these locations.; IKONOS and casi data were acquired over the selected test sites during the 1999 and 2000 growing seasons. The IKONOS data included both multispectral (4 m x 4 m pixel) and panchromatic (1 m x 1 m) data. The casi data were acquired in spatial mode (2 m x 2 m and 1 m x 1 m pixels) and spectral mode (4 m x 4 m pixel) with a spectral resolution of approximately 10 nanometres (72 bands). In this study, both standard and non-standard image-analysis techniques were used to determine the information content of these data. Results to date indicate that high-resolution data can provide detailed and accurate information about wetland ecosystems including the size, shape, extent, and distribution of wetlands; wetland type; and information on adjacent land-use and land-cover types.; The most useful and accurate results, with Kappa values over 90%, were obtained by applying a maximum-likelihood classification algorithm to selected green, red, and near-infrared bands of the 1 m spatial-mode casi data. These data permitted the greatest differentiation between vegetation types, particularly within the inland wetland site. (Abstract shortened by UMI.)...