Data Assimilation And Parameters Inversion Of Water-sediment Numerical Model

Improving the predicted accuration of water level, discharge, sediment concentration is import to the application on flood risk analyses. Water-sediment numerical model and field observation are the two main data sources to investigate the real river. On the one hand, water-sediment numerical models are based on the theory of dynamics and are the form of mathematical description from the real physical phenomenon. So the water-sediment numerical models have structural errors. On the other hand, the field observations have errors attributing to the instrument, observed thechnic s and data preprocess. With the technical development of field observation and information communication, it is necessary to research a data assimilation system for water-sediment numerical model. In the data assimilation system, the observations and numerical results are integrated and mixed together for more accurate predictions. The practice of the data assimilation system can develop the existing water-sediment numerical model to a real-time model in which the predited values of flow and sediment c oncentration are optimized dynamiclly.An ensemble Kalman filter system for water-sediment numerical model is built up. One dimensional numerical model is used to model the transport of water and sediment and the Preissmann four implicit scheme is adopted to solve the equations. The state-space equations of the variables(water level, discharge, sediment concentration, roughness) are constructed based on the theory of control system. A variational data assimilation system of suspended load is built up. A cost function which measures the difference between the observations and the numerical result is introduced first. The transport equation of suspended load is taken as a constraint of the cost function. The adjoint equations for the cost function are deduced and the gradients of the transport parameters including saturation recovery and sediment carrying capacity are constructed. A coupled data assimilation system of water-sediment numerical model is built up based on the advantage and disadvantage of the two assimilation method. In the coupled data assimilation system, the values of water level and discharge are calculated using ensemble Kalman filter and the value of suspended sediment concentration is calculated based on the variational method. Forthmore, the background error of the coupled assimilation system is estimated by ensemble Kalman filter and the parameters of water-sediment model are calculated in a way of inverse problem.The ensemble Kalman filter system is tested using historical flood data i n 2011 of the lower Yellow River. The variational data assimilation system of suspended load is applicated using historical flood in 2009 of the lower Yellow River. The performace of the assimilation system is tested and the effect of the assimilation system to the area without observations is analyzed. The coupled data assimilation system is applied in the lower Yellow River using the flood data in 2013.