Abstract: Soil texture is one of compositional data in the geosciences. Spatial interpolation for soil texture must meet four conditions，which including positive definiteness, a constant sum of interpolated values at a given position, error minimization and unbiased estimation. The study adopted compositional Kriging and ordinary Kriging based on data transformed by asymmetry logratio transformation (ALR) and symmetry logratio transformation (SLR) to predict spatial distribution of each soil particle composition. The precision and fitting effect were assessed by utilizing the root mean squared errors (RMSE) and mean squared deviation ratio (MSDR). The results showed that the interpolation results by compositional Kriging and ordinary Kriging based on data transformed by ALR and SLR could meet the four conditions in spatial interpolation. Values of RMSE for different soil particle composition by compositional Kriging were the least and the precision was the highest. For clay, the relative improvement of accuracy to the reference method ordinary Kriging based on ALR was close to 17%. On the whole, fitting effect by compositional Kriging was better than that by other methods. The ranges were wider for compositional Kriging, and its prediction results could better reflect the relations of different soil particle composition with elevation，soil parent material and water area distribution.