Keywords: 
• slope scale /
• soil moisture /
• linear regression /
• state-space model /
• Loess Plateau

Abstract: Soil water content is one of the key factors affecting plant growth and eco-environment reconstruction on the Loess Plateau of China. To assess the spatial heterogeneity of soil water content and its potential influencing factors on a hillslope in the gully region of the Loess Plateau, the state-space approach and a classical linear regression approach were applied in order to identify and quantify the significant relationships between soil water content and elevation, contents of clay, silt, and sand, median soil grain size, and fractal dimension. The results showed that the soil water contents in different soil layers exhibited moderate variation, and were significantly influenced by the elevation, the contents of clay, silt, and sand, and by the fractal dimension. Autocorrelation for the six potential influencing factors were conducted, and cross-correlation functions indicated strong spatial dependences between the soil water content and the elevation, the contents of clay, silt, and sand, and the fractal dimension. The state-space approach simulated the soil water content much better than any equivalent linear regression method. The best state-space model included the elevation, the sand content, and the fractal dimension, which could explain 99％ of the variation in the soil water contents; the model accurately predicted the soil water contents along two transects. Consequently, the state-space analysis was verified to be an effective tool for estimating soil water contents in different soil layers on a hillslope on the Loess Plateau.