This paper explores three important aspects of studies that assess the possible effects of climatic change on agricultural productivity at regional spatial scales. First, long-term historic and stochastically generated (WGEN) weather records are compared in terms of their statistical attributes using the climate conditions found in central France. Second, our results show that the use of CERES-wheat coupled with WGEN produced weather data provides an efficient method for assessing the impacts of changing climate on average agricultural production. Less confidence can be placed, however on the estimation of future agricultural risk and variability assessment. Finally, time series of climate variables with changed mean and variability are either constructed according to the methodology proposed by Mearns et al. (1992) or simulated with WGEN using the approach suggested by Riha et al. (1996). The climate change scenarios are compared in terms of their effects on wheat development and predicted yield with CERES-wheat using dally data from the sulphate integration of the HadCM2 General Circulation Model to drive the crop model. The comparison of the different approaches for the construction of climate change scenario demonstrates the relative importance of changes in the mean climate and short/long-term variability in the prediction of crop yield on a regional basis. The results also indicate that the strength of the yield response to such combined scenarios and sometimes even its sign, depends on the qualitative nature of the change. Therefore, assessments of future agricultural productivity based on this methodological approach must be regarded as speculative reserved.