Rainfall-runoff modeling using conceptual, data driven, and wavelet based computing approach

Abstract

The current study demonstrates the potential use of wavelet neural network (WNN) for river flow modeling by developing a rainfall-runoff model for Malaprabha basin in India. Daily data of rainfall, discharge, and evaporation for 21 years (from 1980 to 2000) have been used for modeling. In the modeling original model, inputs have been decomposed by wavelets and decomposed sub-series were taken as input to ANN. Model parameters are calibrated using 17 years of data and rest of the data are used for model validation. Statistical approach has been used to find out the model input. Optimum architectures of the WNN models are selected according to the obtained evaluation criteria in terms of Nash–Sutcliffe efficiency coefficient and root mean squared error. Result of this study has been compared by developing standard neural network model and NAM model. The results of this study indicate that the WNN model performs better compared to an ANN and NAM model in estimating the hydrograph characteristics such as flow duration curve effectively.