Spatial variability of soil properties was studied, and comparison in estimation accuracy was conducted between three different sampling intensities. The structure of soil spatial variation change with sampling intensities and soil properties. Variation of soil pH, NH₄⁺-N, P and S was best fitted with their corresponding semi-variation model, respectively, whereas that of soil OM, Ca and Mg could be simulated by the same one model. Sampling intensities did not have anything certain with the fitting degree of models, effective range and spatial dependence of variation. Soil properties with significant model-fitting degree were estimated, with results showing that the accuracy of estimation with soil OM and S increased with sampling intensity, however, it was not the case with soil NH₄⁺-N and Ca.