Real-time stage and discharge estimation by a stochastic-dynamic flood routing model

A dynamic-stochastic model for flood forecasting based on the Kalman filtering theory is described. To reduce the computational time that the Kalman filter involves, this model is performed using the reduced rank square root algorithm, which approximates the covariance matrix by a matrix of lower rank. By using the special banded structure of the Jacobian matrix of the discretization of the Saint Venant equations by the four points Preissman scheme, this algorithm is made computationally efficient for a flood routing model. The noise terms are treated as coloured noise processes, with one term for every grid element. An application to estimate the state of a one-channel reach is described. Copyright (C) 2000 John Wiley & Sons, Ltd.