In this study, the effects of the Lambertian assumption on the top-of-atmosphere reflectance are evaluated through comparison with calculations derived using a more detailed bidirectional reflectance distribution function under different atmospheric, lighting, and viewing conditions. The numerical experiments are performed for background, dusty, and cloudy models of the atmosphere in spectral channels of 0.44 and 0.87 μm. In the case of a background or dusty medium over the terrestrial surface, the overestimation of the top-of-atmosphere reflectance in the forward-scatter viewing direction and underestimation in the backscatter one are observed. The angular range as well as magnitude of the discrepancy is noticeably narrower and lower, respectively, when the atmosphere is more turbid and the wavelength is shorter. The use of the Lambertian assumption instead of "true" ocean reflectance leads to a significant underestimation of the top-of-atmosphere reflectance in the forward-scatter direction and a moderate overestimation of reflectance in the backscatter one. The ocean reflectance generally exhibits a high dependence on wind, which affects the reflected solar radiation distribution around the forward-scatter direction. Analysis of simulation results for an overcast sky showed that, in general, the multiple scattered radiation smoothes the anisotropy effects. However, there are conditions at which the choice of a bidirectional reflectance distribution function model is significant: in the case of thin cirrus cloudiness over the ocean at large solar zenith angles and stratus cloudiness with an optical thickness of at least 5 over a vegetation cover or ocean in the near-infrared region of the spectrum.