Method of watershed scaled erosion prediction is different from plot scale erosion prediction. The objective of this study was to predict watershed scaled erosion by measuring sediment yield and using Agricultural Non-Point Source Pollution (AGNPS) model. The plot scale experiment was set in randomized block design with 7 treatments i.e combination of ridges and raised beds. In the experiment, surface erosion was measured with soil pan method, watershed scale erosion (sediment yield) by sampling at the outlet. The method of watershed scale prediction were using AGNPS model. The result shows that AGNPS model was applicable to predict Tugu Utara sub watershed erosion (sediment yield) and simulate its land use management. The simulation of AGNPS model, with crop factor of 0.18 and slope length 5 m significantly reduced watershed erosion. This implies that AGNPS model is reliable to predict Tugu Utara sub watershed erosion and the model proves that watershed erosion is different from plot scale erosion.