Risk assessment for accidental pollution plays an important role in river water quality management. Bayesian networks can be applied to represent the relationship between pollution sources and river water quality intuitively. A time sequential Monte Carlo algorithm, integrated with pollution sources model, water quality model and Bayesian reasoning, is developed to quantify river water quality risk under the collective influence of multiple pollution sources. A case study shows that multiple pollution sources have obvious effect on water quality risk of the receiving water body, which means that integrated management should be developed for multiple risk sources. The model could also support the decision-making process of river basin management through identification of critical pollution sources.