An unsupervised approach to the classification of semi-natural vegetation from Landsat Thematic Mapper data. A pilot study on Islay

Publisher Website
Google Scholar
Abstract
Spectral classes resulting from an unsupervised maximum-likelihood classification of Landsat Thematic Mapper imagery are found to provide the basis for a thematic map of broad habitat types over an area of complex semi-natural vegetation. Contingency tables are used to assign spectral classes to cover types, in addition to calculating classification accuracy. Detailed cover categories identified on the basis of ecological divisions are poorly represented by the spectral classes, but broader cover categories chosen such that they have some spectral homogeneity, in addition to ecological significance, show good agreement. Photographic prints of the satellite imagery were found to be of value both for determining and for mapping cover categories in the field.