Application of spectral angle mapping model to rapid assessment of soil salinization in arid area

Salinization is a major cause of land degradation in arid area, where it is difficult to assess the extent of soil salinization rapidly. Remote sensing is regard as one of the methods which contribute to detecting salt-related surface features. The objective of this research is to analyze spectral characters of some land surfaces salinized to several extents, and extract useful information with spectral angle mapping (SAM) method, in the area of Ebinur lake, Xinjiang, Northwest China. In this study, multispectral image of ASTER is used, including 9 bands, visible and infrared. Minimum noise fraction (MMF) transformation which determine the inherent dimensionality of image data is used to segregate noise in the data, and to reduce the computational requirements for subsequent processing. Then we use a method of pixel purity index (PPI) to choose pure pixels standing for some land surfaces, contrasting with ground-base data. These land surfaces include wet salty crust, dry puffy salty crust, salinized meadow, sparse vegetation, luxuriant vegetation, sand dune, and diluvium, which are salinized strictly, moderately, slightly and none, respectively. A quantified result of salinized soil grades is achieved, ultimately. This method is resultful for rapid assessment of soil salinization, especially in remote arid area where conventional methods are restricted.