IAHR Document Library

« Back to Library Homepage « Proceedings of the 20th IAHR APD Congress (Colombo, 2016)

Hybrid Neuro-Wavelet Models for Forecasting Streamflow

Author(s): Shreenivas Londhe; Shweta Narkhede

Linked Author(s):

Keywords: Streamflow forecasting; Wavelet transform; ANN

Abstract: The accurate streamflow forecasts are an important component of watershed planning and sustainable water resource management. Accurate and timely predictions of high and low flow events at any watershed location can provide stakeholders the information required to make strategic, informed decisions. Over the past two decades, Neural Networks are well established as recognized tools that offer effective solutions for modeling and analyzing the behaviour of complex dynamical systems along with simulating and forecasting hydrological applications. Many researchers have used ANN successfully to predict streamflows for a few days in advance. Timing errors nevertheless appear to be a common problem in most of the Neural Network univariate forecasting models (for that matter for all univariate data driven models). To develop operational solutions on measures that possess temporal insensitivities is therefore problematic and could deliver forecasts that are incorrect with respect to the timing of highermagnitude events. This technical issue must be resolved if such tools are to be transferred into an operational settings. This study explores the potential benefits of applying a timing error correction procedure to Neural Network streamflow forecasting models using data processing tool of wave transform coupled with ANN model for two stations from Krishna basin in Maharashtra, India to explore the possibility of correcting the time lag errors in forecasting 1-day, 3-days, 5-days and 7-days ahead streamflow.


Year: 2016

Copyright © 2022 International Association for Hydro-Environment Engineering and Research. All rights reserved. | Terms and Conditions