ACCURATE ANALYSIS OF MULTI-SPECTRAL BANDS OF LANDSAT TM IMAGES IN MAPPING LAND SURFACE WATER INFORMATION IN EBINUR LAKE, XINJIANG, CHINA

Water is an important resource for the social-economic development and growing population of the earth. However, the appearance coverages of land surface water (LSW) such as vegetation cover, river-beds, and water bodies are complex and have no clear boundaries in remote sensing images. Thus the accurate analysis of water information is difficult to map. In this paper, we tested and analyzed for accuracy of eleven water spectrum index models of the multi-spectral bands in Landsat TM (thematic mapper) images from the Ebinur Lake area in Xinjiang Autonomous Province, China. The eleven spectrum relationship models include normalized difference water index (NDWI), modified normalized difference water index (MNDWI), automatic water extraction index with no shadow (AWEI_nsh), automatic water extraction with shadow (AWEI_sh), vegetation index 1 (VI-1), vegetation index 2 (VI-2), vegetation index 3 (VI-3), water index (LBV_B where L is naked land radiance level, B is water body visible-infrared radiation, V is vegetation radiance variation and B is balance), national wetland inventory (NWI), enhanced water index (EWI), and revised normalized different water index (RNDWI). All have advantages and disadvantages in extracting water information in Landsat TM Images that were acquired in 2011. We found two relationship models, VI-2 the extraction water body from the pixels at 92.95% and VI-3 from the non-exploited ground at 85.04% to be the most optimize and accurate methods of water extraction information and they enhance the advantages over other models. Using the optimal mask water model proposed by Wu, et al (2008) calculation of these two relationship models have achieved accuracy up to 93.80%, and the disturbance from vegetation and non-exploited land is minimized. Through comparison with other commonly used methods, the results of our study show the performance of our proposed method is superior to the others. Therefore VI-2 and VI-3 are the best indicators for LWS mapping of the Landsat TM images. These methods would provide potential method for quantitative evaluating of temporal changes of the water bodies in global Landsat TM aerial images.