Visual and digital classification of landsat tm data for soil, physiography and land use mapping in axios alluvial plain, thessaloniki, Greece

Abstract

Abstract Visual interpretation of analogue images and digital analysis of Landsat‐5 TM CCT (August 1987) were applied (l) for land use/land cover classification, (2) to study the relationships between physiography, soils and land use classes of the survey area and (3) to construct an integrated land system map of Axios alluvial plain (Thessaloniki, Greece). The study of physiography and soils, as well as the final land system map were carried out by the application of visual interpretation of LANDSAT TM black/white and colour composites at 1: 125,000–1: 250,000 scale. Crops and land use classes were classified by (1) visual interpretation and (2) image processing techniques, using supervised and unsupervised classification algorithms, which allow the interpreter to select training sites of known features and to identify similar areas throughout the scene. The results show that at reconnaissance level and using visual interpretation, a sufficient volume of data (soil, land use, physiography) can be collected in a short term period at low cost. Computered‐aided image classification was checked in 100 sites, in order to define the classification accuracy. Maximum classification accuracy (about 85% overall) was obtained using supervised classification and includes rice and other irrigated and non‐irrigated crops as well as bare land. Red, red‐brown colored soils (Inceptisols, Alfisols), covered with cereal straw, were distinguished well, when unsupervised classification was applied on band 6 (thermal band). Land use types are strongly correlated with physiography and soils in the survey area and four well differentiated land systems were defined with similar soils, climate, physiography, geology and climate.