Sampling designing based on spatial distribut ion and variability of soil properties is an important issue in recent studies on soil sampling. In the present study, the variance quad tree method (VQT), combined with the semivariogram function, was used to design an optimal sampling scheme in a field of coastal saline soils. The ordinary kriging method was adopted to interpolate the sampling points fixed with the regular grid method and the VQT method, respectively, and their prediction errors were subsequently calculated for accuracy comparison. Results reveal that with the same size of samples, the VQT sampling scheme was much smaller than the grid one in prediction error, and hence its sampling efficiency was improved by 16% to 25%. The advantage of the VQT method is that this scheme samples sparsely in fields relatively even in soil properties and more inten sively in fields varying sharply. Thus it not only characterizes variability within the sampling area, but also improves sampling ef ficiency, and hence saves sampling cost.