Relationship between weather variables in observed and WXGEN generated data series

In this study, we assess for five selected Canadian climatological stations the validity of selected simplifying assumptions used in the Erosion/Productivity Impact Calculator (EPIC) weather data generator, WXGEN. A useful feature of WXGEN is the small amount of input data required to generate long periods of record. Recent studies have shown some of the deficiencies in WXGEN generated records at selected locations. It is felt that some of the limitations are the result of oversimplification. A primary focus of this study is the failure of WXGEN to account for regional and seasonal differences in the relationship between weather parameters. It is known that there are pronounced regional and seasonal differences in the relationships between precipitation and daily average temperature across Canada. We show that data generated with WXGEN fails to mimic these regional and seasonal relationships. The question of whether alternatives to stochastic weather data generation may be more appropriate for estimating missing solar radiation data is addressed. An empirical model based on temperature and precipitation was found to provide better estimates than data from WXGEN. (C) 1998 Elsevier Science B.V.