We have examined forty SeaWiFS images acquired during 1997 and 1998 off the west coast of Africa to develop theoretical models of Saharan dust radiative properties that could be used for atmospheric correction in this region, i.e., to predict the dust contribution to the measured reflectance in the visible from that measured in the near infrared (NIR). In contrast to nonabsorptive or weakly absorbing aerosols, the dust reflectance significantly decreases from the NIR to the blue because of the absorption of mineral particles in the visible. We chose simple but realistic vertical structures for the dust layer and examined the applicability of a set of aerosol size distributions and refractive indices with radiative properties computed from Mie theory. We found that 18 models (six aerosol size-refractive index distributions times three aerosol vertical distributions) were general enough to estimate the dust reflectance in the visible with an absolute RMS error of the order of 5%. We show that these dust models can be used within a spectral matching algorithm <Gordon et al., 1997> to effect atmospheric correction of ocean color imagery in dust-contaminated regions. We also found that our models all produce very similar top-of-atmosphere outgoing visible flux but that this flux may be significantly different from that predicted by other conventional dust models. ¿ 2001 American Geophysical Union