Abstract: The study on spatial variability of soil nutrients in farmland plays an important role in controlling the regional agricultural nonpoint source pollution and implementing the precision agriculture. Topsoil samples (0-20 cm) were collected from 223 sites in farmland at two scales of sampling grid (1 m×1 m and 10 m×10 m) in the Yellow River Irrigated Area in Ningxia Hui Autonomous Region, China. The objectives of this study were to investigate the spatial variability of total nitrogen content (TN), total phosphorus content (TP) and pH value at the two sampling scales using the conventional and geostatistical means. The conventional statistical analyses showed that the coefficient of variability ([WTBX]CV[WTBZ]) of pH value was 0.02 and in a slight spatial variability, which indicated that the scale effect was not obvious. With the decrease of the sampling scales, the [WTBX]CV[WTBZ] value of TN decreased but that of TP increased. The effects of scales were significant for TP and TN. Geostatistical analyses showed that there was a moderate or high spatial correlation between TP and TN at the two sampling scales, and their spatial distribution was jointly affected by structural and random factors. There was a weak spatial correlation of pH value at the 1 m×1 m scale due to the effects of random factors, such as irrigation and fertilization. There was a certain anisotropic effect of TN, TP and pH value at the 2 scales, and also a high spatial variability northeastward. The crossvalidation showed that the average error of Kriging interpolation was close to 0, and the error of mean square root was close to 1. The interpolation figures could be used to accurately reflect the spatial distribution of soil nutrients. The research results indicated that there was a certain scale effect and also an anisotropy effect for soil nutrients in the Yellow River Irrigated Area.