Climate change is a natural phenomenon that occurs in the long time scale. The most important factors that intensify climate change are changes in reflected radiation from the sun, changes in Earth's orbit, greenhouse gases and continents drift. In this research the Barandoezchay watershed in West Azerbaijan Province with an area of 1088 km² was investigated. Due to the lack of Synoptic station in the area, neighbor stations and Thiessen Polygons method were used and the weights of each station was incorporated in the climatic parameters include minimum temperature, maximum temperature, rainfall and solar radiation that are the requirement of downscaling LARS-WG model. Then, using the weighting method, among nine general circulation models, two models were selected which had the highest weight and were performed for production of climate data until 2040 using the scenarios studied in these models. In the next step using Multilayer Perceptron Network, daily runoff was predicted until 2040 under five scenarios of two models. The results showed that under five scenarios of two GCM models, there are changes in the average of runoff from watershed in base period compared to future period. Such that the average of river discharge in the future in spring and summer compared to the base period will be changed and will be reduced, however in autumn and winter, the runoff average will increase compared to the base period, overall total runoff volume will be more in future period.