Agricultural Best Management Practices (BMPs) are effective ways to reduce agricultural nonpoint source pollution from their source area to receiving water bodies. Characterization of BMPs in a watershed model is a critical prerequisite for evaluating their impacts on water quantity and water quality in a complex system. However, limited research has reported about the representation of BMPs in fully distributed models. This paper presents a stepwise procedure for representation of several BMPs and assessment of their hydrologic impacts with a fully distributed model, SEIM (Spatially Explicit Integrated Modeling). A case study is conducted in the 73 km² Luoyugou watershed located in the Loess Plateau of China, where rainstorm erosion accounts for more than 60% of annual sediment load in average. Three BMPs are selected in this study including (i) conversion from farmland to forest, (ii) terrace, and (iii) no-till farming. These management practices are represented in the model through the alteration of model parameters characterizing their physical processes in the field. The results of scenario assessment for a historical storm event showed that the maximum sediment reduction after terrace is about 97.3%, the average sediment reduction after no-till farming is about 9.5%, and the average sediment reduction after conversion from farmland to forest is 75.6%.