Reasonable sampling sizes or spatial interpolation method is an assurance of the accuracy of soil spatial variability analysis which is important for sustainable use of soil resources. This paper discussed the influence of sampling sizes and spatial interpolation method on spatial prediction accuracy of soil fertility quality using Ordinary Kriging, Simple Kriging, Inverse Distance Weighted and Spline Function method by comparing random samples of 340, 170, 90, 50 and 30 from a total of 359 samples. The results indicated that the differences of prediction accuracy between different interpolation methods were not significant, whereas the differences of prediction accuracy between different sampling sizes were significant. This study suggested that a minimum sampling size of 90 should be used for soil fertility quality evaluate in the North China Plain. Moreover, this study compared this minimum sampling size with the appropriate sampling quantity acquiring by Cochran method, and found that the sampling quantity acquire by Cochran was smaller. So this study suggested using Cochran method without considering the spatial variability would produce inaccurate spatial prediction.