Land-use and land-cover classification in semi-arid regions using independent component analysis (ICA) and expert classification

• Published in: International journal of remote sensing Vol. 35; no. 24; pp. 8057 - 8073
• Main Authors: Namdar, Mohammad, Adamowski, Jan, Saadat, Hossein, Sharifi, Forood, Khiri, Afsaneh
• Format: Journal Article
• Language: English
• Published: Abingdon Taylor & Francis 17.12.2014
• Subjects: Animal, plant and microbial ecology; Applied geophysics; Biological and medical sciences; Classification; Density; Earth sciences; Earth, ocean, space; Exact sciences and technology; Fundamental and applied biological sciences. Psychology; General aspects. Techniques; Independent component analysis; Internal geophysics; Land cover; Land use; Slopes; Teledetection and vegetation maps; Vegetation; Watersheds; Middle East; Iran; Asia; image processing; processes; Thematic Mapper; maps; Vegetation index; Independent component analysis; sampling; remote sensing; decomposition; classification; extraction; drainage basins; Geographical division; land use; Semi arid zone; imagery; segmentation; Third World; Plant cover; statistical analysis; slope angle; methodology; land cover
• ISSN: 0143-1161; 1366-5901
• DOI: 10.1080/01431161.2014.978035

This study was focused on addressing the need for accurate land-use/land-cover classification (LULC) maps in Iran and in other similarly developing countries. To generate and validate a new LULC map for northeastern Iran's 2037.5 km 2 Hable-roud watershed, a step-by-step process was developed and implemented, consisting of image preprocessing, extraction of training and reference sampling locations, decomposition of multi-spectral thematic mapper bands into features by independent component analysis methods, classification using these features and slope maps, enhancement of land-use classes through image segmentation and zonal statistics, then through consideration of normalized difference vegetation index and climatic zones, followed by ground truthing. This newly developed approach provided maps that distinguished dryland farming, irrigated farmland, forest plantations, and low-, medium-, and high-vegetation density rangelands, while currently available maps for the watershed lef t 39% of lands unclassified or in combined classes. The new maps' ground-truthing-based overall accuracy and kappa coefficient were 88.3% and 0.83, respectively. In order to develop such an improved LULC map, it was necessary to go beyond the mere analysis of reflectance information, to incorporating climatic and topographic data through this newly proposed step-by-step approach.