Applicability of IRS Satellite Images for Surveying Water, Soil and Vegetation Cover Condition of Najm Abad Region, Savojbolagh

Multi-spectral remotely sensed data is useful information source for the detection of surface changes and change detection is a major application of the remotely sensed data. This study is conducted to investigate capability of sensor Liss-III of IRS-P6Resource satellite data for providing land cover map, in Najm Abad of Savojbolagh region with 20000 ha area. The images of 26th June, 2006 were registered to digital map with scale of 1:25000. The RMSE of registered data was 0.58. Images were enhanced using contrast enhancement, making False Color Composite images (FCC), Principal Component Analysis (PCA), Vegetation Index and Digital Elevation Model (DEM). In order to determine the best band composition for using in classification and making FCC, Optimized Index Factor (OIF) and correlation techniques were used. For classification of images, unsupervised and supervised method (Box classification, Minimum Distance, Minimum Mahalanobis Distance and Maximum Likelihood classifier) were used. Ground truth map with sampling metod and field survey was provided. After classification, land cover map was provided with bareland, saline soil, rangeland and agricultural lands. The results of overall accuracy and kappa coefficient in different classification methods were as follows,: for Box classification classifier, 85.8% and 84%, for Minimum Distance classifier 81.3% and 79%, for Minimum Mahalanobis Distance 91% and 90%; and for Maximum Likelihood classifier estimate 93% and 92%. The results showed that Digital Elevation Model (DEM) cause to highest accuracy and Optimized Index Factor (OIF) cause to least accuracy for providing land cover maps. Also vegetation indices could not present acceptable results for land cover mapping.