In this paper, the spatial variability of soil total P and Olsen-P at the depths of 0 approximately 10 and 10 approximately 20 cm in a field-scale was examined by using traditional statistics combined with geostatistics methods. A sampling grid of 30 m x 42 m including 49 pairs of soil sampling points was established in the Shenyang Experimental Station of Ecology, Chinese Academy of Sciences. The results showed that the variance coefficients were much higher for Olsen-P (46.56% approximately 56.42%) than for total P (11.68% approximately 13.33%). Both total P and Olsen-P in the two soil depths had strongly spatial structures and similar spatial correlation ranges. The parameters derived from best-fitted models showed that the spatial variability of both total P and Olsen-P was mainly affected by structural factors, with C/(C0 + C) being higher than 66% for all the variables. The kriging contour maps showed that both total P and Olsen-P in the two soil depths had similar spatial distribution patterns. A more precise sampling scheme could be made based on the spatial distribution pattern of test soil properties combined with traditional variance coefficients. To recognize the strong variability of soil Olsen-P in field-scale is of significance for better understanding the P cycling in farm ecosystems and for precise agriculture.