UC Santa Barbara

This paper presents an algorithm for the estimation of parameters in state‐space models that represent hydrologic processes in which the state variables are observed with error. The algorithm is based on the maximization of the conditional expectation of the likelihood function of the state equation. The estimation algorithm is numerically stable and guarantees local convergence under mild conditions. It is also shown that the estimation algorithm can be coupled with an optimal control method to yield a combined control estimation technique that can be easily implemented. An application of the theory and methods developed herein is given for flood routing via reservoir operation. Copyright 1985 by the American Geophysical Union.