Modeling of tree species and structural attributes from high resolution multi-spectral imagery using decision tree analysis for east coast eucalypt forests of Australia

Tree clusters formed from 2m resolution multi-spectral imagery using the Un iversity of Melbourne TIDA algorithm have been classified using decision tr ee analysis to produce models which predict the distribution of tree specie s, stem diameter, tree height, and an index of tree physiological age. Data from field sites at Batemans Bay, on the south coast of New South Wales, A ustralia, has been applied to map a range of forest types from rainforest t o tall moist old growth forest and dry open forest. The method uses image s pectral and spatial attributes as well as terrain and climate attributes. T he reliability of these models is compared against the reliability of fores t type and structure maps produced by interpretation of 1:25,000 colour aer ial photographs.