Although climate simulation of the GCMs is an important basis to configure the future climate, there are numerous uncertainties in a simply application of the GCM outputs.Monte Carlo probability analysis is a powerful tool to determine the probability of projected climate, and understand better the uncertainties.We applied the Monte Carlo technique in the GCM climate simulations for the future 100 year, and analyzed the probabilities in the responses of lake water changes to the future climate for Qinghai Lake, Hulun Lake and Daihai Lake in the western China. The results show that the temperatures in year 2020s, 2050s and 2080s would have stable 2-5℃ increases at a 75% chance.The changes in temperature would exceed to the gauges of last 50 years and be equivalent to reconstructions during the Holocene.The 75% chance climate change would lead to lake water changes ranging-5% to+10% in Qinghai Lake Basin, -7% to+5% in Hulun Lake Basin and+2% to-+12% in Daihai Basin.Although the future changes in precipitation are equivalent to the last 50-yr gauges and lower than changes in the Holocene, the speed of lake water change was much high.Under the conditions of the projected climate at 75% chance, the 3 lakes would have cumulative 30%-45% changes with the amplitudes of-10% to+10%.The fast responses of the lake water would warn us that we should have a preparation enough for the future changes in water resources.