The traditional Kriging interpolation method contains only limited types of variogram models to express the spatial variation of variable, which is difficult to depict the distribution of actual data exactly (especially for the spatial structure of complicated data).A Support Vector Machine-Kriging(SVM-Kriging) method is proposed in this paper by introducing Least Square Support Vector Machine (LS-SVM) to fit the experimental variogram of actual data.The SVM-Kriging method is compared with the traditional Kringing method, and the results show that the variogram of SVM-Kriging method from actual data fields can avoid the subjectivity and arbitrariness of selecting types of variogram models.The SVM-Kriging method has good objectivity and adaptability, effectively improving the Kriging interpolation results.