The retrieval of aerosol properties over land is more complicated due to the relatively strong contribution of the land surface reflectance to the radiation measured at the top-of-atmosphere (TOA). Another problem caused by the anisotropic earth surface which is a second order effect, can create systematic biases in the aerosol retrieval. But the simple Lambertian surface assumption is still widely used in most aerosol retrieval algorithms for the single satellite view. In this paper, radiative transfer simulations with coupling surface-atmosphere are employed to assess how much uncertainties are introduced from the Lambertian surface assumption in satellite aerosol optical depth (AOD) retrieval. The result shows that it has great impacts on aerosol retrieval especially at lower aerosol loading. The uncertainties mainly depend on the anisotropy of the target. The more difference lies between the reflectance along observing direction and the average ' reflectance ρ̅_t, ρ̅_t, ρ_t of the whole BRDF surface, the larger error happens in aerosol retrieval.