Hyperspectral sensors acquire images in the reflective domain([0.4-2.5 m]), with a high spatial resolution (several meters), in numerous spectral bands simultaneously. These images enable the estimation of both atmospheric aerosols and gases. The present work aims at characterizing industrial plumes. The particles emitted by factories have widely unknown properties and the current databases do not include the information needed to understand and model their radiative impact. So, the goal of the first part of this study is to characterize the physical and optical properties of such aerosols, from samples in situ collected. It has led to the establishment of a model, enabling the description of the radiative impact of a plume emitted by a metallurgical plant. In a second part, a method to jointly estimate CO2 and aerosols in a plume is proposed. A sensitivity analysis and the application of this method to two AVIRIS images have shown that, for optically thick plumes, this joint estimation leads to an error, in the CO2 estimation, divided by a factor of two, compared with existing techniques (e.g. JRGE). Finally, a method to estimate the aerosol type, EARTH (Estimation of Aerosol Type using Hyperspectral data), has been developed, specifically designed for industrial plumes. Its application to simulated and real data (CASI image) has shown that it enables to identify, without any a priori information, the presence of metallic particles in an optically thin plume.
