Control of non-point source pollution, particularly in terms of nutrients, has advanced to a new phase where the identification of pollution sources and processes is necessary for complex river systems. The Soil and Water Assessment Tool (SWAT) is a process-based model that has been applied to manage aquatic environments in multiple river basins worldwide. In this study, nitrogen (N) loads from point and non-point pollution sources in the Cau River Basin, one of the three most polluted river basins in Vietnam, were simulated. Three scenarios with different combinations of pollution sources were developed to evaluate the major contributions of natural sources, cultivation, industry, households, craft villages and livestock in the study area. A set of hydrological and water quality parameters was successfully calibrated and validated with data observed over multiple years. The model results showed good performance in stream flow simulations at all four hydrological stations. The best performance for the N load simulation was under a scenario representing the dominant contribution of the non-point sources, which caused high amounts of N runoff and was likely triggered by rainfall. The modelling approach presented in this study provides an example for establishing a modelling platform for complex pollution systems.