Soil nutrient distribution has obvious spatial differences. The reasonable sampling number in field for soil nutrient surveying is based on the spatial variability of nutrient factor and the required data precision. The larger the spatial variability is, the more precise the data need to be, then the more sampling number needed. The calculated result indicates that most of the reasonable number of the soil factor can be reduced sharply, which means that a lot of cost of sampling can be saved. According to the result from self-correlation and semi-variance function of geo-statistics method, the self-correlation coefficient values have large difference in various directions, independence interval also changes with soil nutrient factors. As well, the results provide support to decide sampling point density, protracted isoline and fertilize accurately in precision farming.