Pantnagar Journal of Research

Abstract

The main study was the estimation of runoff using present day and previous days rainfall and previous days runoff as a daily input variable using artificial neural networks (ANNs) and wavelet-based ANNs (WANNs). Rainfall-runoff data were collected, standardized, and selected as inputs using the Gamma test. The methodology for runoff estimation and modeling using ANNs and WANNs was applied to the regions of Narsimhpur and Mandla in Madhya Pradesh. As the number of neurons was increased, the correlation between rainfall and runoff was initially improved and then reduced. Therefore, an optimum number of neurons was identified at which the best correlation was achieved. Better correlation coefficients, least root mean square errors, higher Nash-Sutcliffe efficiency, and greater Willmott indices were obtained for WANNs models compared to ANNs models. These results can be utilized for runoff forecasting.