Modeling of hyperconcentrated sediment-laden floods in Lower Yellow River

This paper presents a rapid forecast model for simulating hyperconcentrated sediment-laden floods in the Lower Yellow River. The model is a hybrid of a conventional one-dimensional mathematical model for unsteady sediment-laden flow and an artificial neural networks model for encapsulation of numerical results. The former provides detailed river flood routing information under typical scenarios, whereas the latter extracts modeling outputs from the former and establishes a station-specific model for efficient flood forecasting. Three typical floods that occurred in the Lower Yellow River in 1977, 1982, and 1996 are simulated. Not only the hybrid model predictions are found to be in close agreement with measured data, but also the computational speed is significantly enhanced. It is found that sediment transport is of significance with regard to the flooding behavior of hyperconcentrated flows. Therefore, the model presented herein is of particular use for rivers with high sediment concentration.