Caracterisation, modelisation et validation du transfert radiatif d'atmospheres non standard; impact sur les corrections atmospheriques d'images de teledetection

This study is based on data acquired with an airborne multi-altitude sensor on July 2004 during a nonstandard atmospheric event in the region of Saint-Jean-sur-Richelieu, Québec. By non-standard atmospheric event we mean an aerosol atmosphere that does not obey the typical monotonic, scale height variation employed in virtually all atmospheric correction codes. The surfaces imaged during this field campaign included a diverse variety of targets : agricultural land, water bodies, urban areas and forests. The multi-altitude approach employed in this campaign allowed us to better understand the altitude dependent influence of the atmosphere over the array of ground targets and thus to better characterize the perturbation induced by a non-standard (smoke) plume. The transformation of the apparent radiance at 3 different altitudes into apparent reflectance and the insertion of the plume optics into an atmospheric correction model permitted an atmospheric correction of the apparent reflectance at the two higher altitudes. The results showed consistency with the apparent validation reflectances derived from the lowest altitude radiances. This approach effectively confirmed the accuracy of our non-standard atmospheric correction approach. This test was particularly relevant at the highest altitude of 3.17 km : the apparent reflectances at this altitude were above most of the plume and therefore represented a good test of our ability to adequately correct for the influence of the perturbation.;Standard atmospheric disturbances are obviously taken into account in most atmospheric correction models, but these are based on monotonically decreasing aerosol variations with increasing altitude. When the atmospheric radiation is affected by a plume or a local, non-standard pollution event, one must adapt the existing models to the radiative transfer constraints of the local perturbation and to the reality of the measurable parameters available for ingestion into the model. The main inputs of this study were those normally used in an atmospheric correction : apparent at-sensor radiance and the aerosol optical depth (AOD) acquired using ground-based sunphotometry. The procedure we employed made use of a standard atmospheric correction code (CAM5S, for Canadian Modified 5S, which comes from the 5S radiative transfer model in the visible and near infrared) : however, we also used other parameters and data to adapt and correctly model the special atmospheric situation which affected the multi-altitude images acquired during the St. Jean field campaign.;We then developed a modeling protocol for these atmospheric perturbations where auxiliary data was employed to complement our main data-set. This allowed for the development of a robust and simple methodology adapted to this atmospheric situation. The auxiliary data, i.e. meteorological data, LIDAR profiles, various satellite images and sun photometer retrievals of the scattering phase function, were sufficient to accurately model the observed plume in terms of a unusual, vertical distribution. This distribution was transformed into an aerosol optical depth profile that replaced the standard aerosol optical depth profile employed in the CAM5S atmospheric correction model.;Based on this model, a comparison between the apparent ground reflectances obtained after atmospheric corrections and validation values of R*(0) obtained from the lowest altitude data showed that the error between the two was less than 0.01 rms. This correction was shown to be a significantly better estimation of the surface reflectance than that obtained using the atmospheric correction model. Significant differences were nevertheless observed in the non-standard solution : these were mainly caused by the difficulties brought about by the acquisition conditions, significant disparities attributable to inconsistencies in the co-sampling / co-registration of different targets from three different altitudes, and possibly modeling errors and / or calibration. There is accordingly room for improvement in our approach to dealing with such conditions. The modeling and forecasting of such a disturbance is explicitly described in this document: our goal in so doing is to permit the establishment of a better protocol for the acquisition of more suitable supporting data.;The originality of this study stems from a new approach for incorporating a plume structure into an operational atmospheric correction model and then demonstrating that the approach was a significant improvement over an approach that ignored the perturbations in the vertical profile while employing the correct overall AOD. The profile model we employed was simple and robust but captured sufficient plume detail to achieve significant improvements in atmospheric correction accuracy. The overall process of addressing all the problems encountered in the analysis of our aerosol perturbation helped us to build an appropriate methodology for characterizing such events based on data availability, distributed freely and accessible to the scientific community. This makes our study adaptable and exportable to other types of non-standard atmospheric events.;Keywords : non-standard atmospheric perturbation, multi-altitude apparent radiances, smoke plume, Gaussian plume modelization, radiance fit, AOD, CASI.