A comparison of supervised, unsupervised and synthetic land use classification methods in the north of Iran

Land use classification is often the first step in land use studies and thus forms the basis for many earth science studies. In this paper, we focus on low-cost techniques for combining Landsat images with geographic information system approaches to create a land use map. In the Golestan region of Iran, we show that traditional supervised and unsupervised methods do not result in sufficiently accurate land use maps. Therefore, we evaluated a synthetic approach combining supervised and unsupervised methods with decision rules based on easily accessible ancillary data. For accuracy assessment, confusion matrices and kappa coefficients were calculated for the maps created with the supervised, unsupervised and synthetic approaches. Overall accuracy of the synthetic approach was 98.2 %, which is over the 85 % level that is considered satisfactory for planning and management purposes. This shows that integration of remote sensing data, ancillary data and decision rules provides better classification accuracy than traditional methods, without significant additional use of resources