A semiempirical approach for downscaling general circulation model (GCM) based daily atmospheric circulation patterns (CP) and predicting local climatological variables under climate change is developed. Specifically, the daily 500-hPa surface outputs of the Canadian Climate Center (CCC) and Max Planck Institute (MPI) (Germany) GCMs are linked stochastically, using a split sampling approach, to local temperature and precipitation in Nebraska. Three series of data are analyzed: historical data, 1¿CO2 GCM results and 2¿CO2 GCM results. Between these three data sets, no significant difference can be detected in either CP typology (constructed by principal component analysis and k means method) or stochastic properties of daily time series (Markov matrix). On the other hand, the average geopotential height of the 500-hPa pressure field exhibits significant change, labeled the ΔCO2 effect, between the 1¿CO2 and 2¿CO2 cases. Accordingly, climate change is assumed to be represented by the historical average geopotential height augmented by the ΔCO2 increment. It is found that both the CCC and MPI GCMs lead to predicting a winter temperature increase of 3¿--6 ¿C, a smaller but significant increase in spring and fall temperatures, and no increase in summer. The probability of precipitation occurrence is found to remain almost unchanged, as well as the dry period duration. The estimates of local response to climate change depend upon the location and the GCM used for downscaling the CP. The MPI GCM, which includes an ocean-atmosphere coupling, appears to yield smaller downscaled changes than the purely atmosphere-based CCC GCM.