To achieve a better understanding of environmental issues related to soil erosion,¯ood-ing, and solute transport, it is necessary to understand the variability of soil moisture under different climatic and geological conditions and at different spatio-temporal scales. Moisture patterns based on ®ve depths (0±5, 15±20, 25±30, 45±50, and 70±75 cm) were determined in the Da Nangou catchment of the loess plateau of China from May to September 1999. Spatial structure of pro®le-averaged and layer-averaged soil moisture was analysed using semivariograms. The results indicated that soil moisture exhibits changing spatial dependence with time and depth. For pro®le-averaged soil moisture, high sills [4.6±14 %(v/v) 2 ] and high ranges (135±160 m) were observed during dry conditions. The sills [2.6±6.3 (%v/v) 2 ] and ranges (140 m) were small during wet conditions. However, soil moisture was described by linear semivariograms without a sill following a signi®cant rain event. For layeraveraged soil moisture, the sills increased with depth, but not for the ranges. The nuggets tended to increase with higher sills. This analysis provides insights into the variability of soil moisture on a large scale and interpolation strategies for extrapolating point measurements across the catchment.