Improving Estimation of Suspended Sediment Concentration in Alluvial Rivers by Sediment Transport Model with Data Assimilation

Suspended sediment concentration (SSC) is the key factor in the interaction relationship between water, sediment and channel morphology in alluvial rivers. The sediment transport numerical model provides an important way to estimate the SSC to predict the river dynamics, such as aggradation and degradation, for the river and sediment management. Due to the uncertainties in the sediment transport modeling, it is still a great challenge to improve the accuracy of the SSC estimation. Data assimilation is a technical method which can combine the numerical model and the observations, improving the accuracy of model output by incorporating the observations into the model to update the model states and parameters. In this study, the Particle Filter (PF), a sequential Monte Carlo data assimilation algorithm, is employed to combine the sediment transport model with SSC observations to develop a data assimilation system to improve the SSC estimation in alluvial rivers. As time goes by, when the observed SSC data becomes available, the model state variable SSC and the three model parameters, the saturation recovery coefficient α, the coefficient k and the exponent m in the sediment carrying capacity formula, in the sediment transport model will be updated simultaneously according to the PF theory. The developed data assimilation system is applied to the lower alluvial reach of the Yellow River in China, which has the highest SSC in the world, to evaluate its performance. A series of synthetic experiments are designed to test the ability of the data assimilation system to update the model states and parameters from "initial values" to "true values". The results of synthetic experiments show that the PF data assimilation system has a strong ability to update the given "initial values" of model states and parameters to "true values" after a few times of assimilation. The accuracy of SSC estimation is improved significantly and the uncertainties of the model parameters are reduced effectively after the PF data assimilation. Finally, a real experiment is carried out with assimilating the historical observed SSC data in the Yellow River into the model. The results show the newly developed sediment transport model with PF data assimilation system is applicable to improve the accuracy of SSC estimation in real events in alluvial rivers.