Spatial Discrimination of Salt–affected Soil Surfaces Using Landsat TM Data – A Preliminary Study

Salinity has plagued irrigated agriculture throughout history (Tanji, 1990), and soil salinization is an old–yet– new problem in arid regions. For example, there are approximately 100 million ha of salt–affected soils in China, accounting for one tenth of the total land area (Li, 2009). The development of effective salinity control practices requires an understanding of the causal relationship between soil and water salinization and salt formation and movement in the soil. Thus, the first step in establishing an innovation in the salinity control practices is to develop a means for assessing the degree of soil salinization and its spatial distribution easily and efficiently. Remote sensing is a possible candidate for an effective tool in detecting temporal and spatial changes in soil salinity. However, in general, irrigated lands are studded with saline plots of various sizes (Wang et al., 2008), so that fields with the crops in vigor are intermingled with ones in which the crops lack vigor, or plots almost bare of vegetation. As a result, how to cope with mixed pixels is one of main problems in utilizing remote sensing for this purpose effectively. Soil salinity is indicated by (a) presence of white crust and efflorescence and (b) soil parameters such as EC and SAR, and also by (c) invasion of salt–tolerant weeds and (d) lack of plant vigor (Metternicht and Zinck, 2009; Araki et al., 2010). The former two, (a) and (b), reflect “soil” conditions, and the latter two, (c) and (d), reflect “vegetation” conditions at each site. In this study we regard a saline area as being composed of salt– affected patches which differ in degree, and try to make discrimination of salt–affected soil surfaces by composing canonical discriminant functions based on the discriminating variables which indicate the “vegetation” and “soil” conditions, and using these functions as salinity indicators. This paper describes the methodology and results of a preliminary study made in the middle reaches of the Heihe River, China, using the Landsat TM data.