Abstract: The fundamentals of a stochastic weather generator (WGEN) run for synthetic daily climate scenarios are discussed in detail, and in accordance with the statistic attributes of each stochastic process embodied and commonly available information from general circulation models (GCMs), its parameters programme is presented and specified through statistic techniques (such as parametric analysis of dynamic data etc.). A stochastic experiment of WGEN is presented and carried out on the basis of the principles of Monte-Carlo numerical calculation. Proceed from the overall analysis of temporal and spatial statistic characteristics, evaluation of the present climate simulation in Northeast China from this experimenthas been made. It is concluded that WGEN has fairly good skills in generating maximum tempe rature, minimum temperature, precipitation and radiation. The probability property of data sequence observed and simulated is rather identical as well. Therefore, daily climatic change scenarios with different climatic variabilities can be generated with the combination of yearly and monthly output of GCMs at large spatial resolution, and adjustment of WGEN parameters. Thus, a promising approach is given to link climate models with various impacts models, and study the impacts of the change in climate variability.