Historical Land use/Land cover classification and its change detection mapping using Different Remotely Sensed Data from LANDSAT (MSS, TM and ETM+) and Terra (ASTER) sensors : a case study of the Euphrates River Basin in Syria with focus on agricultural irrigation projects
This thesis deals spatially and regionally with the natural boundaries of the Euphrates River Basin (ERB) in Syria. Scientifically, the research covers the application of remote sensing science (optical remote sensing: LANDSAT-MSS, TM, and ETM+; and TERRA: ASTER); and methodologically, in Land Use/Land Cover (LULC) classification and mapping, automatically and/or semi-automatically; in LULC-change detection; and finally in the mapping of historical irrigation and agricultural projects for the extraction of differing crop types and the estimation of their areas. With regard to time, the work is based on the years 1975, 1987, 2005 and 2007. Initially, preprocessing of the satellite data (geometric- and radiometric- processing, image enhancement, best bands composite selection, transformation, mosaicing and finally subsetting) was carried out. Then, the Land Use/Land Cover Classification System (LCCS) of the Food and Agriculture Organization (FAO) was chosen. The following steps were followed in LULC- classification and change detection mapping: visual interpretation in addition to digital image processing techniques; pixel-based classification methods; unsupervised classification: ISODATA-method; and supervised classification and multistage supervised approaches using the algorithms: Maximum Likelihood Classifier (MLC), Neural Network classifier (NN) and Support Vector Machines (SVM). These were trialed on a test area to determine the optimized classification approach/algorithm for application on the whole study area (ERB) based on the available imagery. Pre- and post- classification change detection methods (comparison approaches) were used to detect changes in land use/land cover-classes (for the years 1975, 1987 and 2007) in the study area. The remote sensing methods show a high potential in mapping historical and present land use/land cover classes and its changes over time. Significant results are also possible for agricultural crop classification in relatively large regional areas (the ERB in Syria is almost 50,335 km²). Change trends in the study area and period was characterized by land-intensive agricultural expansion. The rapid, more labor- and capital- intensive growth in the agricultural sector was enabled by the introduction of fertilizer, improved access to rural roads and markets, and the expansion of the government irrigation projects. Irrigated areas increased 148 % in the past 32 years from 249,681 ha in 1975 to 596,612 ha in 2007.