## Abstract

The uncertainties of integral aerosol properties calculated using aerosol size distributions retrieved from multiwavelength observations of aerosol optical depth have been determined for a variety of typical atmospheric aerosol size distributions and refractive indices. The results suggest that more information about the aerosol composition, as well as more information about the sizes that are less efficient, in the optical sense, is needed to improve the shape of the retrieved size distributions. All the calculations in this paper assume spherical homogeneous particles. The sensitivity results refer to these conditions. The moments of the retrieved size distributions are systematically underestimated and errors can be as large as −82%, −30%, and −35% for the total number of particles, the total surface, and the total volume, respectively. The errors in the mass scattering efficiency, the effective radius, and the total volume depend very much on whether the actual volume size distribution is monomodal or bimodal. For a known refractive index, the total scattering coefficient, the hemispherical backscattering coefficient, and the extinction coefficient, as well as the hemispheric backscattering to total scattering ratio and the asymmetry factor, are obtained with absolute values for the average errors less than 4%. Similar behavior was expected for cases with uncertainty in the refractive index, especially for parameters defined by the ratio of two integral properties. However, it turns out that the hemispheric backscattering coefficient and the hemispheric backscattering to total scattering ratio were poorly retrieved, reaching errors of 29% in several cases, while the asymmetry factor was very well recovered with absolute values of the average errors always under 7%. When the wavelength dependence of the refractive index is included, the retrieved size distribution is very unrealistic, with average errors in the hemispheric backscattering coefficients and the hemispheric backscattering to total scattering ratio around 30% at some wavelengths. However, even in this case the errors in the retrieved asymmetry factor stay under 8%. Thus, for spherical and homogeneous particles, the spectral optical depth data can be used to determine the asymmetry factor with little sensitivity to the assumptions in the calculations. Furthermore, the retrieved size distribution can be used as an intermediate step to extrapolate one set of optical properties from another set of optical properties.