Taking mercury-contaminated farmland in Tongren city, Guizhou province as study area, we conducted five interpolation methods, including inverse distance weight, radial basis function, global polynomial, local polynomial and kriging, for the collected soil mercury content data. The interpolation results were evaluated by cross validation. The results indicated that the five methods can accurately reflect the spatial distribution characteristics of the study area and the interpolation effect and precision of kriging method are optimal. In kriging method, the accuracy of removing secondorder trends is highest, and the interpolation effect of exponential model is the best. The variation function model of kriging method shows that soil mercury content in this region is mainly affected by natural factors, but less affected by random factors.