Flood forecasting using Hybrid wavelet neural network model

Abstract

The dynamic and accurate flood forecasting of daily stream flow processes of a river are important in the management of extreme events such as flash floods, floods and optimal design of water storage structures and drainage network. This paper aims to recommend a best hydrologic models are linear stochastic models autoregressive moving average (ARMA), autoregressive integrated moving average (ARIMA) and nonlinear models like Artificial neural network (ANN) and Wavelet neural network (WNN) for flood forecasting of Vamsadhara river in the India (sub zone 4A). Wavelet neural network (WNN) is an hybrid modelling approach for forecasting of river flow using daily time series data. A discrete wavelet multi resolution method was employed to decompose the time series data of river flow into subseries (approximation) and high (details) frequency, and these sub series were then used as input data for the artificial neural network (ANN).