DONATE

IAHR Document Library


« Back to Library Homepage « Proceedings of the 25th IAHR World Congress (Tokyo, 1993)

River-Flow Forecasting Under Uncertainties by Simple System Models and Kalman Filtering

Author(s): Chao-Lin Chiu; Weizu Yu; Randy D. Crissman

Linked Author(s): Chao-Lin Chiu

Keywords: No Keywords

Abstract: In forecasting unsteady flows in natural rivers, a well-known method is to solve a pair of hyperbolic-type partial differential equations (representing the momentum and mass conservation principles), by a numerical technique such as the method of characteristics. It is complex and requires a great deal of computer time. The complexity further increases when uncertainties in flow resistance arise due to such factors as the wind and ice formation that vary with time. To deal with such uncertainties, an effective measure is to continuously observe or monitor the water levels along the reach of interest and use the data to modify and update the flow resistance represented by such measures as Manning's n. The updated flow resistance can then be used to modify the flow forecasts. In estimation theory, the procedure used to combine the predicted and observed values to estimate (update) the system parameters and variables is "filtering. " The Kalman filter is one of the possible filters that can be used for such a purpose (Chiu 1978, Gelb 1974). In Kalman filtering, the observed data is combined with a mathematical system model. The mathematical system model may consist of the pair of partial differential equations mentioned above, which include Manning's n to be updated at each time point. The mathematical model may also be other simpler models. In fact, simpler models when coupled with observed data in Kalman filtering often give acceptable flow forecasts. This is one of the advantages of using the filtering scheme (Chiu, etc.1976). The objective of the present study is to compare three simple system models for flow forecasting. These three models are: (1) Linear Time-Varying Reservoir Model (LTVRM); (2) Non-Linear Reservoir Model (NLRM); and (3) Muskingum routing model. These models will be compared relative to the partial differential equation model mentioned earlier, in forecasting the flow into the Grass Island Pool in the upper Niagara River (Crissman, etc.1992).

DOI:

Year: 1993

Copyright © 2025 International Association for Hydro-Environment Engineering and Research. All rights reserved. | Terms and Conditions