Many applications of Earth-observing satellite radiometer measurements in the visible and near infra-red require surface reflectances, necessitating the use of atmospheric correction algorithms. Performing these corrections usually requires the assumption of Lambertian surfaces, a limitation that becomes serious with the new generation of multi-look satellite radiometers. In this letter, we present and test theoretically a multiple-channel approximation that does not assume a Lambertian surface, but instead assumes that the wavelength dependence of the shape of the surface bidirectional reflectance factor distribution function is negligible in comparison with the wavelength dependence of the surface and atmospheric scattering properties. The approximation is characterized by a mean ratio, k, that may be retrieved by a suitable atmospheric correction process along with surface reflectances and the important atmospheric parameters, without assuming a canopy or soil model or using ''dark'' pixels to obtain the atmospheric optical thickness. We test the approximation with a canopy model that includes the hot-spot effect, and, when using ATSR-2 configurations, it yields at best an order-of-magnitude reduction in the error of the surface reflectance (at the surface) over the Lambertian assumption, and in more stringent conditions, a factor of three improvement. We indicate how the approximation might be implemented in an atmospheric correction algorithm. ¿ American Geophysical Union 1995