Abstract: In order to understand the spatial distribution of soil surface water content and its relations with environmental factors in grassland of the northern Loess Plateau, the autoregressive state-space models and classical linear regression models were used, to simulate the spatial distribution of soil surface water content in a grassland of the northern Loess Plateau, based on the saturated soil hydraulic conductivity (Ks), soil surface temperature (T), elevation (E) and litter mass (L). The results showed that state-space models could be applied to the wind and water erosion transitional area of the Loess Plateau where landscape factors varied greatly and the state-space models were consistently more effective than linear regression models. Among the mono-variable state-space models, Ks based models showed the best simulation result (R2 = 0.936). Among the multi-variable state-space models, Ks, E and L included model showed the best simulation result (R2 = 0.976), and the combination of such variables based models provided the best approach to explain the spatial variation of soil surface water content. State-space models are recommended for studying spatial relations between soil surface water content and other variables in the wind and water erosion transitional area of the Loess Plateau.