The spatial variability of soil nutrient is very important to the application of fertilizer, the sampling density of precision agriculture, and the sub-area of residence management of precision agriculture etc. With the examples of the agricultural soils of No.13 Village of Gongpeng Town and Xiguan Village of Enyu Town in Yushu City which are the typical black soil regions of Northeast China, applying semi-variance model and spatial autocorrelation model combined with GIS technology, a research on the differences of village-level spatial variability of agricultural soil available K is carried out in the thesis, which shows that the differences of averages and coefficients of variation in the same village are small, but large between different villages. The values of nugget and sill are close to the maximum ranges in No. 13 Village, but they are relatively different between different villages. The spatial autocorrelation of No. 3 and No. 7 lands in No. 13 Village are stronger than that of Xiguan Village.