The objective of this paper is to study the impact of critical source area (CSA) within an Annualized AGricultural Non-Point Source pollution models (AnnAGNPS) simulation at medium- large watershed scale. The impact of CSA on terrain attributes is examined by comparing six sets of CSA (0.5, 1, 2, 4, 6 and 8 km2). The accuracy of AnnAGNPS stimulation on runoff, sediment and nutrient loads on these sets of CSA is further suggested in this paper. The results are as followed: (1) CSA has little effect on watershed area, and terrain altitude. The number of cell and reach decreases with the increase of CSA in power function regression curve. (2) The variation of CSA will lead to the uncertainty of average slope which increase the generalization of land characteristics. At the CSA range of 0.5–1 km2, there is little impact of CSA on slope. (3) Runoff amount does not vary so much with the variation of CSA whereas soil erosion and total nitrogen (TN) load change prominently. An increase of sediment yield is observed firstly then a decrease following later. There is evident decrease of TN load, especially when CSA is bigger than 6 km2. Total phosphorus load has little variation with the change of CSA. Results for Dage watershed show that CSA of 1 km2 is desired to avoid large underestimates of loads. Increasing the CSA beyond this threshold will affect the computed runoff flux but generate prediction errors for nitrogen yields. So the appropriate CSA will control error and make simulation at acceptable level.
Keywords: AnnAGNPS simulation, critical source area (CSA), non-point source pollution loading, terrain attributes