Modeling Rainfall - Runoff Process in Lighvan Chai Basin Using Conditional Threshold Temperature Neuron

Necessity of river flow forecasting in constructional works, planning for optimal usage of water reservoirs, river training and flood warning has been well recognized. In this regard, the rainfall – runoff process has been widely studied using artificial neural networks modeling. In the current research, multi layer perceptron was applied to forecasting rainfall – runoff of Lighvan Chai snowy basin in East Azarbaijan province. The data of the basin includes daily rainfall, temperature, and runoff which their effects on the efficiency of network were studied at different steps. Getting along with the factors of rainfall and temperature at the current day, previous days and runoff in previous days in entrance matrix has led to the best results for neural networks. As the Lighvan Chai is a snowy basin, the effect of temperature and snowmelt on runoff is very important and a new neuron which is called conditional neuron of threshold temperature was introduced. Figure of this neuron is binary and the numbers are zero – one. The snowmelt temperature is the criterion of using these numbers. The results of neural networks model was compared to those from the dimensionless snowmelt hydrograph (DSH) including a greater efficiency of the neural networks.