Changing the date of the first fall frost and the last spring frost is a very important phenomenon in agriculture that can be one of the consequences of global warming. One way to study future climate, is using general circulation models (GCMs). In this study, observations of temperature and precipitation were weighted by using Mean Observed Temperature-Precipitation (MOTP) method. This method considers the ability of each model in simulating the difference between the simulated average temperature and average precipitation in each month in the baseline period and the corresponding observed values. The model that had more weight, selected as the optimum model, because it is expected that the model will be valid for future. The outputs of GCM models based on the three emission scenarios A 2 and B 1 and A 1 B, downscaled by LARS-WG for six stations, Isfahan, Badroud, Khour va Biabank, Damaneh Fereidan, Kashan and Hamgin. The data were analyzed by software in 95% of confidence level (P 0.05). The results indicated that in all six stations in future period 2020-2049 based on the three scenarios, compared with baseline period 1971-2000, the first fall frost will occur later and the last spring frost will occur earlier. The first fall frost will occur later for 2 days (for Hamgin based on the B 1 emission scenario) to 8 days (for Khour va Biabanak based on the A 2 emission scenario) and the last spring frost will occur earlier for 2 days (for Hamgin and Damane Fereidan based on the A 1 B and B 1 emission scenarios) to 6 days (for Khour va Biabanak based on the A 2 emission scenario) . The best distribution functions obtained using the software EasyFit for first fall frost and the last spring frost for the baseline period and under climate change, were selected and compared. Key words: Climate change, Frosts, LARS-WG, GCMs
