Applying a Recursive Update Algorithm to a Distributed Hydrologic Model

Abstract

A real-time flood forecast algorithm was developed using a distributed hydrologic model together with recursive state variables updating to utilize online hydrologic information. The Kalman filter was applied to the distributed model using Monte Carlo simulation methods and several other techniques. The Monte Carlo simulation methods permit the algorithm to successfully project forward the state variables and their error covariance through the nonlinear hydrologic system. A huge computation burden would arise if the filter were applied to update the very large number of distributed state variables at the same time; this was avoided by assuming spatial similarity for the distributed state variables. Applying a factor, calculated from the total amount of the model internal variables, to the distributed state variables provided computational efficiency in the update algorithm. The distributed hydrologic model coupled with the Kalman filter algorithm offers improved simulation results with encouraging forecast accuracy.