Reflectance measurements in the visible and infrared wavelengths, from the Moderate Resolution Imaging Spectroradiometer (MODIS), are used to derive aerosol optical thicknesses (AOTs) and aerosol properties over ocean and land surfaces, separately. Both algorithms employ radiative transfer (RT) code to create lookup tables, simulating the top-of-atmosphere (TOA) reflectance measured by the satellite. Whereas the algorithm over ocean uses a vector RT code that includes the effects of atmospheric polarization, the algorithm over land assumes scalar RT, thus neglecting polarization effects. In the red (0.66 m) and infrared (2.12 m) MODIS channels, scattering by molecules (Rayleigh scattering) is minimal. In these bands, the use of a scalar RT code is of sufficient accuracy to model TOA reflectance. However, in the blue (0.47 m), the presence of larger Rayleigh scattering (optical thickness approaching 0.2) results in nonnegligible polarization. The absolute difference between vector- and scalar-calculated TOA reflectance, even in the presence of depolarizing aerosols, is large enough to lead to substantial errors in retrieved AOT. Using RT code that allows for both vector and scalar calculations, we examine the reflectance differences at the TOA, assuming discrete loadings of continental-type aerosol. We find that the differences in blue channel TOA reflectance (vector–scalar) may be greater than 0.01 such that errors in derived AOT may be greater than 0.1. Errors may be positive or negative, depending on the specific geometry, and tend to cancel out when averages over a large enough sample of satellite geometry. Thus, the neglect of polarization introduces little error into global and long-term averages, yet can produce very large errors on smaller scales and individual retrievals. As a result of this study, a future version of aerosol retrieval from MODIS over land will include polarization within the atmosphere.