Soil erosion due to by poor land use/cover, improper land management practices, and cultivation of steep slope land and abandoned grazing, is arguably the most critical problem in Koga watershed. It is resulting subsequent sedimentation, reducing water quality, decreasing agricultural productivity. Hence understanding and evaluating the spatial distribution of sediment yield with management scenario is paramount as the watershed has large reservoir volume in Koga watershed. The main objective of this study is predicting the runoff and sediment to provide a basis for future scenario analysis of water resource management of Koga watershed by using AnnAGNPS model. To achieve the overall objective of this study, all necessary data for Koga watershed were assembled. The AnnAGNPS model was used for simulating the stream flow and sediment. Calibration of the model was carried out from 1988 to 2001 and validation from 2002 to 2007. The result of sensitivity analysis indicated that the CN was the most sensitive parameter to runoff and peak runoff rate whereas LS and K-factor were for sediment yield following RF, and these parameters were subjected to calibration. For model calibration, R2 of 0.69, 0.35, 0.55; NSE of 0.69, -0.38, 0.55; RSR of 0.54, 1.14, 0.67; and PBIAS of 0.07%, -80.56% and 4.09% were obtained for surface runoff, peak runoff rate, and sediment load, respectively which was very good agreement with the observed values. Validation results indicated an R2 of 0.76, 0.54, 0.62; NSE of 0.76, 0.38, 0.62; RSR of 0.43, 0.71, 0.56, and PBIAS of 2.31%, -36.58% and 5.68% for surface runoff, peak runoff rate, and sediment load, respectively. This indicated that the model provided from fair to excellent performance for the three outputs. Generally, the results obtained from applying AnnAGNPS on Koga watershed demonstrate that the model has significant potential as a management tool for evaluation of the effectiveness of alternative BMPs scenarios and their impact on soil erosion and identification of hot spot area of erosion, and investigation of sediment delivery characteristics. The total annual soil loss in the watershed was 276370 tones. Out of the total watershed area only 21.5% was responsible for 78.8% of total soil erosion, and experienced above the recommended tolerable limit of soil loss. From the land area that was above the tolerable limit of soil loss, 4.99%, 39.71% and 34.10% belong to severe, very severe and extremely severe erosion class. Thus, conversion of 21.5% of highest eroding cropland cells either to forest or grassland would reduce soil erosion, sediment yield and sediment load by 78.8%, 85.4%, and 83.6% respectively, indicating that the maximum soil erosion in the watershed was reduced in to the tolerable limit of soil loss. Therefore, this scenario should be primarily practiced and more conservation practices should be adopted at the hot spot areas. This ultimately would help to reduce the sedimentation in Koga dam which could result in reduction of storage capacity.