Abstract: Nonpoint sources of P in agriculture soils have been identified as one of the main causes of freshwater eutrophication. Studies have showed that decades of P fertilization at rates exceeding the amount removed by crops have resulted in widespread accumulation of P in agriculture soils in Taihu Lake watershed, and the accumulation of P will increase the potential of sol P loss to surface waters. Therefore, studying the spatial variability of soil P at large scale will be helpful to develop management practices for controlling agriculture nonpoint P pollution. In this paper, a geostatistics method, with Geographic Information System (GIS), was applied to study the spatial variability of total soil P in the typical area of Taihu Lake watershed. The results showed that the spatial distribution features of the total soil P could be quantitatively described by semivariogram which could be fitted by exponential model. The ratios of nugget to sill at different lag intervals varied from 0.396 to 0.445, indicting that the total P in soil of the studied area had a relatively good spatial autocorrelation, and the range of autocorrelation extended to about 11km. The maps of the total soil P made by block kriging based on the fitted model depicted that the river shore land of northern part of Jiangyin county and low limnetic plan of the studied area generally had more P than the other places,and much more attention should be paid to such places for the sake of protection of surface water quality.