Dual state-parameter optimal estimation of one-dimensional open channel model using ensemble Kalman filter

Abstract

In this paper, both state variables and parameters of one-dimensional open channel model are estimated using a framework of the Ensemble Kalman Filter (EnKF). Compared with observation, the predicted accuracy of water level and discharge are improved while the parameters of the model are identified simultaneously. With the principles of the EnKF, a state-space description of the Saint-Venant equation is constructed by perturbing the measurements with Gaussian error distribution. At the same time, the roughness, one of the key parameters in one-dimensional open channel, is also considered as a state variable to identify its value dynamically. The updated state variables and the parameters are then used as the initial values of the next time step to continue the assimilation process. The usefulness and the capability of the dual EnKF are demonstrated in the lower Yellow River during the water-sediment regulation in 2009. In the optimization process, the errors between the prediction and the observation are analyzed, and the rationale of inverse roughness is discussed. It is believed that (1) the flexible approach of the dual EnKF can improve the accuracy of predicting water level and discharge, (2) it provides a probabilistic way to identify the model error which is feasible to implement but hard to handle in other filter systems, and (3) it is practicable for river engineering and management.