Most current and proposed satellite remote sensing of tropospheric aerosols relies upon radiance measurements that are interpreted using algorithms that determine best fits to precalculated scattered sunlight for one or more standard aerosol models. However, the number of different types of aerosol and the substantial space and time variations typically encountered can pose a severe uniqueness problem even for the multiple constraints provided by multispectral radiances of a scene at a number of observation zenith angles. Experience with polarimetry remote sensing on planetary missions has demonstrated that the measurement of polarization as well as the radiance can resolve such uniqueness problems. We use numerically accurate solutions of the vector radiative transfer equation for a realistic atmosphere-ocean model to theoretically simulate several types of satellite aerosol retrievals over the ocean utilizing radiance measurements alone, polarization measurements alone, and radiance and polarization measurements combined. We have restricted all simulations to a single near-infrared wavelength of 0.865 μm and assumed that aerosols are spherical, monomodal, and nonabsorbing. These simplifications permit a study of practical scope that tests the retrieval algorithms under exactly the same conditions, thus clearly demonstrating their relative capabilities. In agreement with previous analyses, we have found that radiance-only algorithms using multiple-viewing-angle observations perform far better than those based on single-viewing-angle measurements. However, even multiple-viewing-angle radiance measurements taken at a single wavelength are not always sufficient to determine the aerosol optical thickness, effective radius, and refractive index with high enough accuracy. In contrast, high-accuracy, single-wavelength, multiple-viewing-angle polarimetry alone is capable of uniquely retrieving all three aerosol characteristics with extremely high accuracy (¿0.015 in aerosol optical thickness, ¿0.03 μm in effective radius, and ¿0.01 in refractive index). Furthermore, the accuracy of the optical thickness retrieval can be slightly improved by simultaneously using radiance measurements. Our analysis demonstrates that algorithms utilizing high-accuracy polarization as well as radiance measurements are much less dependent on the availability and use of a priori information and can be expected to provide a physically based retrieval of aerosol characteristics (optical thickness, refractive index, and size) with accuracy needed for long-term monitoring of global climate forcings and feedbacks. ¿ 1997 American Geophysical Union