Assessment of pollution of water bodies from non point sources (NPS) is a complex, requisite long series data and time consuming task. The accuracy of NPS pollution models depends to a great extent on how well a model selects input parameters describing the relevant characteristics of the watershed. It is certain that the promoting precision of input parameters affects simulation results of runoff, sediment and nutrients yield for the entire watershed. In this study, a basic database, which includes DEM, soil sort and landuse map, climate data, and land management data, is established using GIS. The generation and formation of non point source pollution involves a great uncertainty which makes pollution monitoring and controlling very difficult. Understanding the main parameters that affect NPS pollution uncertainty are necessary for planning and design of control measures. On the basis of the results of parameter sensitivity analysis, the sensitive parameters of soil and water assessment tool (SWAT) model are identified, and then model parameters related to stream flow and nutrient loadings are calibrated and validated by observed values. The results show that simulated values are reasonably compared with observed data. Spatial temporal distribution features of NPS pollution in the Qingjiang River basin are revealed. NPS pollution mainly takes place in flood season. The critical risk areas of soil erosion are identified. Stream flow and nutrient loadings (including total nitrogen (TN) and total phosphorus (TP)) in Qingjiang River Basin are simulated. The surface runoff and nutrient yield results indicate that average annual runoff and the output of TN and TP provide better understanding on stream flow and nutrient loadings corresponding to various variation conditions of land use mode, agricultural tillage operation and natural rainfall etc.