A method is presented for simultaneous determination of the aerosol optical thickness (&tgr;a), particle size (rm, geometric mean mass radius for a lognormal distribution) and the single scattering albedo (ω0, ratio between scattering and scattering + absorption) from satellite imagery. The method is based on satellite images of the surface (land and water) in the visible and near-IR bands and is applied here to the first two channels of the Advanced Very High Resolution Radiometer (AVHRR) sensor. The aerosol characteristics are obtained from the difference in the upward radiances, detected by the satellite, between a clear and a hazy day. Therefore the method is mainly useful for remote sensing of large-scale air pollution (e.g., smoke from a large fire or concentrated anthropogenic pollution), which introduces dense aerosol into the atmosphere (aerosol optical thickness ≥0.4) on top of an existing aerosol. The method is very sensitive to the stability of the surface reflectance between the clear day and the hazy day. It also requires accurate satellite calibration (preferably not more than 5% error) and stable calibration with good relative values between the two bands used in the analysis. With these requirements, the aerosol optical thickness can be derived with an error of Δ&tgr;a=0.08--0.15. For an assumed lognormal size distribution, the particle geometrical mean mass radius rm can be derived (if good calibration is available) with an error of Δrm=¿(0.10--0.20) μm, and ω0 with Δω0=¿0.03 for ω0 close to 1 and Δω0=¿(0.03--0.07) for ω0 about 0.8. Uncertainties in the type of the size distribution further increase the uncertainty in rm. The method was applied to AVHRR images of a forest fire smoke. The aerosol optical thickness and particle size derived from the AVHRR images are compared to the analysis of sun photometer measurements from the ground.