Abstract: As a global environmental problem,soil salinization has induced different types and degrees of land desertification and degradation,and affected the ecological environment stability and regional ecological security together with soil pollution,soil degradation,soil erosion and soil desertification. China was affected by soil salinization seriously in wide distribution,especially in the arid and semi-arid areas in the northwest China. Besides, the trend of the area of soil salinization is continuous to expand and becomes more and more severe. Therefore it is very important to monitor the soil salinization in the arid and semi-arid areas scientifically,accurately and rapidly. Multi-spectral remote sensing technology has become a new method to detect and monitor the soil salinization due to the advantage that it can obtain the feature information of large area repeatedly. However,the remote sensing images have a large number of data,bands and redundant information,which brought difficulties for the interpretation of remote sensing images. It is difficult to accurately distinguish the soil salinization for mild and moderate salinization soil. However,hyperspectral technology has broader spectral response range and high spectral resolution at nanometer level,which can help show the difference of structure and composition of the material, and reflect the subtle spectral characteristics of surface features more accurately. In this study,the delta oasis between the Weigan River and the Kuqa River,Xinjiang,China was selected as the study area,which was influenced by soil salinization seriously. Based on the different soil hyperspectral data and its soil salt content,and using fifteen spectral transformation methods including logarithm,RMS,continuum removal,first-order,second-order RMS and so on,this paper processed the hyperspectral data to find the most sensitive wavelengths,and established a soil salinization monitoring model by combining with the different spectral index models. Then apply the model to the Landsat-TM remote sensing images to achieve the high-precision quantitative inversion of soil salinity at large scale. The results show as follows:(1)The first order differential spectral transformation is the best method for extracting the soil spectral information,with a higher correlation with soil salinity;(2)The surface spectral model based on parabola model and optimal soil salinity index SI 3 has the best effect,R² = 0.871 2;(3) Compared with soil salinity monitoring model(R² = 0.592 5)built by traditional multi-spectral remote sensing technology,the model(R² = 0.700 2)converted by scale transformation showed better effect. The study provided a scientific and effective way to achieve high-precision,large scale remote sensing inversion of soil salinity in arid area.

Key words: the surface spectral modeling, scale transformation, soil salinization