Automated stream planform mapping: A test study

The nature of streams and their fish habitat is now a critical parameter to forest management. Mapping of the main features of stream courses is valua ble baseline information for channel assessment and fisheries surveys, dete rmining stream classes for prescribing logging setbacks and monitoring chan ges over time due to natural events or logging activity. Eighty centimeter casi imagery was acquired over Tofino Creek on the west c oast of Vancouver Island, British Columbia. Using a spectral angle mapping algorithm, seven surface types were classified (deep water, shallow water, sand, gravel and cobble, woody debris (e.g., logs), conifer and deciduous). Comparison with ground truth taken within days of the imagery indicates ge neral agreement of surface types and accuracy estimations greater than 80 p ercent for most classes. Individual logs and piles of woody debris were con sistently detected. Problems do arise within shadowed areas, boundaries of stream features and forest and with some zones of sand, gravel and cobble.