Prediction of Zarrinehrud River Run-Off in the Climate Change Condition using Artificial Neural Networks

In the present research, the climate change effect on variation of surface runoff of Zarrinehrud located in the Miandoab plain was investigated. In this direction, the scenarios including A1B, A2 and B1 via LARS-WG downscaling model and with applying the HadCM3 general circulation model and artificial neural network model in two different periods (2046-2065, 2080 -2099) were studied. For this purpose, the best combination of input parameters of the MLP artificial neural network model was selected to estimate the runoff among various meteorological parameters with time delay of zero and one day and runoff parameter with one-day delay. Then, the meteorological data predicted by the LARS-WG in the future were used as inputs for the selected neural network model and consequently the runoff was predicted. The comparison of results between observed and simulated data by LARS-WG model using different statistical and error measurement indices indicates that there is no significant difference between simulated and observed values. Performance analysis of the artificial neural network model indicates that the mentioned model has good and suitable accuracy to simulate the runoff variations in the studied area. The results showed that the average annual runoff in the period of 2046-2065 will increase about 4.62 CMS than base period and it will decrease about 14.7 CMS during the period 2080-2099 compared to the base period.