An algorithm for aerosol optical depth τ retrieval from the Geostationary Observational Environmental Satellite (GOES) series is described, where the darkest pixels are used to create a spatial composite of surface reflectance. The data are calibrated and corrected for atmospheric extinction to retrieve the surface reflectance which is then used to retrieved τ. Analysis suggests that τ retrieval uncertainty is ¿18--34% depending on the certainty of the assumed radiative transfer model parameters. Retrieval uncertainty is less over low surface reflectances and at large scattering angles. The retrieval algorithm is validated against Sun-sky radiometer τ measurements for aerosols emitted by biomass burning in South America during 1995 and 1998. The relative differences between observed and retrieved τ are within the estimated uncertainty, having correlations ranging from 0.78 to 0.97. Further, the GOES retrievals are compared to τ retrieved using the Moderate-Resolution Imaging Spectroradiometer (MODIS) airborne simulator (MAS). The average relative difference in this comparison is 11%, thus retrieval validations are again within the estimated algorithm uncertainty. These results suggest that the GOES satellite can be used to monitor aerosols over land, while the agreement between MAS and GOES retrievals suggests the ability to combine the spectral abilities of MODIS with the temporal observations of GOES.