Wavelet Based Neural Networks for Daily Stream Flow Forecasting

Abstract--Unlike other hydrological time series data, rainfall and runoff time series data are highly variable spatially and temporally. However, stream flow measuring point is fixed in a river, therefore mapping of temporal variability of runoff can be handled with suitable algorithms. Time series analysis requires mapping complex relationships between input(s) and output(s), because the forecasted values are mapped as a function of patterns observed in the past. In order to improve the precision of the forecasts, a Wavelet Neural Network (WNN) model, based on a combination of wavelet analysis and Artificial Neural Network (ANN), has been proposed. The WNN and ANN models have been applied to daily runoff in the West flowing Rivers of India. The calibration and validation performance of the models is evaluated with appropriate statistical indices. The results of daily runoff series modelling indicated that the performances of WNN models are more effective than the ANN models. Further, WNN model performance has been evaluated separately for Low, Medium and High runoff values and found that the selected model is suitable for various runoff patterns from a catchment. Keywords-- Time series, ANN, WNN, Runoff, India I