To quantify spatial variability of soil water is of important significance for ecological environment restoration in Karst regions.The geostatistics method is used to determine and quantify spatial variability of surface soil water(0～5 cm and 5～10 cm) for analysis of spatial patterns of soil water in a typical karst depression(150 m×50 m) in Guzhou.Using the grid sampling method(5 m×5 m space),341 sampling points are fixed in total.Based on autocorrelation analysis,spatial correlation structure of these soil moisture patterns is calculated with the aid of the Moran's I index.Results indicate that the soil water contents in the two layers have a simple pattern of periodic variations,and patches with the same spatial autocorrelation character correspond to soil continuum and rock distribution in area.Sample variograms are found to have an anisotropy character in the whole research area.And in different patches,the semivariograms show different structure characters.The soil continuum patch has a low sill and a long range,whereas the rock distributed patch has a high sill and a short range.Such soil water patterns can be explained by the effect of bare rock distribution and the topography of the research area.Based on the soil water content data,spatial scaling of soil moisture is further studied.The range of the soil water is increased with the sample spacing.This implies that when the spatial pattern is controlled by plural factors of different scales,sample spacing should be decided by the aim and precision of the research.The findings of this study may help improve the sampling design in the research on soil moisture in the Karst regions.