State estimation of conceptual hydrological models using unscented Kalman filter
Abstract Unscented Kalman filter (UKF) has its origin in transforming the Gaussian random variables for nonlinear estimation and has received little attention in the context of state estimation of conceptual hydrological models. This paper introduces UKF to estimate state variables of a conceptual hydrologic model. A symmetric point approach and a scaling framework are used for performing the sample generation process of UKF. This paper investigates the application of UKF for state estimation with a synthetic case study in which both the simulated state, the true state, and the corrected state are precisely known. The results show that the use of UKF can improve the performance of both the model outputs and the state variables as the difference between the corrected trajectories and the true trajectories decreases rapidly and tends to vanish after only a few iterations. Our results and comparisons also demonstrated the capability and usefulness of UKF for state estimation in two real basins.