Soil is a necessary nutrition library of crop growth, its nutrient spatial variability exists the whole crop growth period with the characteristics of long and complex, so the spatial variability of soil nutrient prediction is one of the hot research problems to be urgently solved in precision agriculture. This paper first introduces GIS and statistical analysis organic combination of research method and model, on the basis of this, based on GIS spatial data level of analysis and point-line-side buffer, we will accurate analysis of point-line-sides mutation interval through the variation Euclidean distance and distance weighting interpolation method. Finally through the comparison of mean value and T test, we can predict the approximate interval of soil nutrients spatial variability, to a certain extent, it can provide theory and technology support for the scientific prediction of soil nutrients spatial variability.