In view of the randomness and spatial structure of the variogram, the authors used the integrated variogram and weighted linear programming fitting methods to fit the model parameters in the main directions. This was conducted in accordance with the structure of the anisotropy, and the variogram calculation and automatic fitting of the spherical model were realized. In order to ensure the success rate of fitting spherical model, the algorithm also provided pairwise relative variogram methods which could effectively inhibit the impact of outliers on the variogram. This algorithm was realized in VC++6.0. Using the fitted model, the authors obtained grid files by interpolation of the sample region. Crossvalidation and contour map show a relatively high fitting degree.