Increasing industrial activity and, consequently, greenhouse gas emissions have disturbed the global climate balance, especially in recent years, which is referred to as climate change. This phenomenon has affected various systems of human life, therefore, its investigation has become one of the most important scientific discussions in recent years. Since rainfall and temperature changes have a direct impact on water resources, one of the main goals of the experts is to predict the future of the climate more realistically. The purpose of this study is to investigate the uncertainty of the downscaling methods in Zayandeh-rud basin. For this purpose, the output of the five general atmospheric circulation models, which is the most credible tool for examining the effects of climate change, were obtained from the fifth Assessment Report of the Intergovernmental committee on Climate Change for Isfahan and Koohrang stations for the base period 1971-2000, the mid-term 2006-2010 and the upcoming 2020- 2049 for three scenarios of RCP2.6, RCP4.5 and RCP8.5. By verifying the accuracy of the climate parameters of each model in the mid-2006-2010 period, it was found that each GCM model alone has no ability to predict rainfall and temperature and it is better to use weighted models. The results showed that the RCP8.5 Scenario for Isfahan Station and RCP2.6 Senario for Koohrang Station stations are more consistent with real data.The weighted models were eventually introduced to the LARS-WG downscaling models and the neural network. Then, their uncertainty was investigated using Bootstrap method. The results showed that the rainfall parameter was more reliable than temperature, so that in most months, especially in spring, summer and autumn, the predicted precipitation was at 95% confidence range. In the event that the confidence range of the minimum and maximum temperature was lower and in most months, the predicted values were not considered at confidence ranges. Also, the results showed that the neural network did not have good ability to predict the minimum and maximum temperatures, and finally, it was found that in the LARS model, the number of months in which the climate parameters are in the confidence range is more than the model neural network model, and as a result, the LARS model is more suitable with less uncertainty for downscaling the climate parameters. The results of LARS downscaling showed that Isfahan station with a decrease of 15.83 mm in precipitation and an increase of 1.22 and 1.32 degrees Celsius in temperature for minimum and maximum temperatures Is in a more critical situation, compared to Koohrang Station, with a decrease of 10.41 mm in precipitation and an increase of 0.7 and 0.5 degrees Celsius in temperature for minimum and maximum temperatures Key Words Climate Changing, IPCC, Downscaling, LARS-WG, Neural Network, Bootstrap, Uncertainty