[en] Two statistical approaches for linking large-scale atmospheric circulation patterns and daily local rainfall are described and applied to several GCM (general circulation model) climate simulations. The ultimate objective is to simulate local precipitation associated with alternative climates. The index stations are located near the West and East North American coasts. The first method is based on CART analysis (Classification and Regression trees). It finds the classification of observed daily SLR (sea level pressure) fields in weather types that are most strongly associated with the presence/absence of rainfall in a set of index stations. The best results were obtained for winter rainfall for the West Coast, where a set of physically reasonable weather types could be identified, whereas for the East Coast the rainfall process seemed to be spatially less coherent. The GCM simulations were validated against observations in terms of probability of occurrence and survival time of these weather states. Some discrepancies werefound but there was no systematic bias, indicating that this behavior depends on the particular dynamics of each model. This classification method was then used for the generation of daily rainfall time series from the daily SLP fields from historical observation and from the GCM simulations. Whereas the mean rainfall and probability distributions were rather well replicated, the simulated dry periods were in all cases shorter than in the rainfall observations. The second rainfall generator is based on the analog method and uses information on the evolution of the SLP field in several previous days. It was found to perform reasonably well, although some downward bias in the simulated rainfall persistence was still present. Rainfall changes in a 2xCO₂ climate were investigated by applying both methods to the output of a greenhouse-gas experiment. The simulated precipitation changes were small. (orig.)