The Columbia River basin is important to the economies of Canada and the United States, particularly in terms of hydroelectric power generation and industrial production. Variations in streamflows affect both hydropower and industrial production through changes in water supply and changes in risk of flood damage. The sources of the Columbia River in western Canada lie in mountain glaciers and high-precipitation areas. As in many other areas of the world, the available climate station network is inadequate in number, in distribution, and in quality for distributed hydrological modeling. An alternate source of data lies in atmospheric models, which are now able to generate climate data at scales suitable for hydrological modeling. This study evaluated the use of data generated by numerical weather prediction models at various scales. It was found that such data are suitable for modeling, although the differences in scale were found to be less important than was originally anticipated.¿ 1997 American Geophysical Union