Yi Li
College of Water Resources and Architectural Engineering, Northwest A&F University

Studied on spatial variability of soil hydraulic and saline properties had great significance on guiding the development and application of soil environment, soil tillage, and soil water movement. Saline-alkaline sail samples were taken in Xinjiang region at different scales of grids. Spatial variability of soil hydraulic and saline properties were analyzed by means of principles of classical statistics, geostatistics, spatial autocorrelation theory, wavelet variance map, the BP artificial neural network, multifractal theory, and joint multifractal theory. From statistical analysis, most soil hydraulic and saline properties varied to a moderate degree. From the analysis of Geostatistics, most semi-variance theoretical models of soil saline properties could be fit with spherical models. From the analysis of spatial autocorrelation theory for scheme one, Moran’s I coefficients of all the soil hydraulic and saline properties were quite similar in variation, which ranged from -0.8 to 0.6 at all the three scales. For scheme three, Moran’s I coefficients of all the soil hydraulic and saline properties ranged from -0.5 to 0.4. From the analysis of multifractal theory, in scale-free scale from -4 to 4, soil hydraulic and saline properties had multifractal feature. For scheme one, the multifractal features of θs, θ and ρ were weak, and the multifractal features of Ks, α and n were obvious. According to the calculations of generalized dimensions of soil hydraulic and saline properties, spatial variability of θs, θ and ρ were weak, the spatial variability of Ks was stronger than α and n, suggested that spatial structure of Ks was more complex. For scheme three, soil salt content, water droplet penetration time (WDPT), pH, Na+, Mg2+, and Ca2+ had multifractal structure, but multifractal features of pH was not obvious, suggested that spatial variability of pH was weak. From joint multifractal analysis for spatial correlation of soil hydraulic and saline properties, spatial correlation between salt content and Na+, between Na+ and Mg2+, between Mg2+ and Ca2+ were strong. Using rigrsure value to de-noise not only cleared up contaminated elements, but also reflected the origin characteristics of soil hydraulic and saline properties. Error ranges were less than 16%, so that with BP artificial neural network model to simulate the soil hydraulic and saline properties was feasible. In conclusions, different methods for analyzing spatial variability revealed thoroughly characteristics of soil water repellency and physical-chemical properties in a field in Xinjiang area.