Artificial Neural Network validation for rainfall-runoff relationship of Zayandehrud Dam Basin

Rainfall-runoff relationship is one of the most important and complex hydrological processes whose perception is very important in hydrology and water resources. A plenty of physical and statistical models have been developed for this that apply some parameters due to this relationship. The application of artificial neural network as a black box model is one of the methods to evaluate rainfall-runoff relationship. In this study، rainfall-runoff relationship of Plasjan Basin، upstream Zayandehrud River، is evaluated using multilayer perceptron network. Due to high variation of observed series، 3 daily rainfall series of regional stations and daily discharge of Plasjan station were first normalized and according to autocorrelation and cross correlation of rainfall-runoff data، 6 variables were selected for input of the network and a 4-hidden-layer network was found to be more valid comparing with other networks. The selected network was validated using comparison of the mean، standard deviation and probability density function of observed and simulated discharge.