Determination of Land Cover using Landsat TM Imagery

In this study, land cover maps of Bozcaada district were developed using Landsat satellite images obtained in 2006, 2007 and 2008. In addition to original images (6 band Landsat TM), the new images constituted with image processing techniques were also used. A total of ten images were formed by supervised classification method using principal component analysis (PCA), normalized difference vegetation index (NDVI), and tasseled cap (TC) transformation methods. Accuracy analysis were conducted for land cover maps using the data obtained in land and high resolution Formasat (2 m spatial resolution) satellite images. While the highest average classification accuracy was for 3 band image obtained by PCA, the lowest average classification accuracy was for the image obtained by combining the NDVI images of three years. It was found that the highest average classification accuracies were calculated for the image that was formed by the combination of 18 band Landsat images acquired in three years, and 9 band images formed by the first three bands of TC images. It was calculated that, the digital maps formed by using PCA and TC analysis have higher accuracies than that of multi-year NDVI images in the determination of land cover for Bozcaada and similar locations.