Abstract: The scours' maximum depth and position changes have a full influence on spur dikes' safety, and are important parameters for evaluating the dikes' safety. Taking the Zhangnan waterway on the Changjiang River as an example, this paper analyses the relationship between incoming flow, sediment and scours' changes. The influencing factors of scours are the riverbed boundary conditions (water depth, river width), spur body properties (spur length, picking angle), incoming flow and incoming sand factors (annual runoff, annual sediment transport, days of different runoff and sediment transport). Based on the BP neural network's theory, a prediction model of the erosion changes near the spur is built. The result shows that the predicted values are approximate to the true one, which proves that this prediction model is feasible and effective. The result of this study can provide reference for the improvement of spur dikes' safety assessment.