| |
CRC PUBLICATIONS |
Adaptive Linear Models for Real-Time Flood Forecasting
Gnanathikkam Amirthanathan
Publication Type:
Technical Report
This is a publication of the initial CRC for Catchment Hydrology
CRC Program:
Flood Hydrology (Previous CRC)
|
Publication Keywords:
Modelling (Hydrological)
Flood Forecasting
Flood Warnings
Modelling (-Specific Names-II)
Rainfall/Runoff Relationship
Catchment Areas
|
Abstract / Summary:
Summary
Adaptive linear models have been used in the past for on-line forecasting of floods. However, there have been few investigations devoted to determining the suitability of these models for real-time flood warning systems for catchments in Australia. This report presents two linear models, namely, the Unit Hydrograph (UH) and ARMA-type models for real-time flood forecasting. As the parameters of the models can vary from storm to storm and also within a storm, a linear sequential optimal estimation technique (Kalman filter) is used to update the parameters of these models. A simple loss model consisting of initial and continuing loss components suitable for real-time operational purposes is used. The models are used to forecast runoff from rainfall for three Australian catchments. Linear river routing models are also developed and applied and a multiple input-single output (MISO) model is used to analyse a river reach with several tributary inflows. Proper a priori estimates of noise statistics are important for optimal performance of the filter. A procedure more suitable for short series of data is proposed and applied successfully. The real-time forecasting performance of these linear models is compared with that of non-adaptive methods and some of the reasons for the difference in performance are discussed.
This report is not available for downloading. Printed copies can be purchased from the Centre Office .
 |
1 Download(s) |
|
|
|
|
|
|
|
|
Centre Office:
CRC for Catchment Hydrology
Dept of Civil Engineering
Building 60
Monash University Vic 3800
Tel: +61 3 9905 2704
Fax: +61 3 9905 5033
|
|
|
