Operational remote sensing of aerosols over land to account for directional effects

The assumption that the ground is a Lambertian reflector is commonly adopte d in operational atmospheric corrections of spaceborne sensors. Through a s imple modeling of directional effects in radiative transfer following the s econd simulation of the satellite signal in the solar spectrum (6S) approac h, we propose an operational method to account for the departure from Lambe rtian behavior of a reflector covered by a scattering medium. This method r elies on the computation of coupling terms between the reflecting and the s cattering media and is able to deal with a two-layer atmosphere. We focus o n the difficult problem of aerosol. remote sensing over land. One popular s ensing method relies on observations over dense dark vegetation, for which the surface reflectance is low and quite well defined in the blue and in th e red. Therefore a study was made for three cases: (1) dark vegetation cove red by atmospheric aerosols, (2) atmospheric aerosols covered by molecules, and finally (3) dark vegetation covered by atmospheric aerosols covered by molecules. Comparisons of top-of-the-atmosphere reflectances computed with our modeling and reference computations made with the successive-order-of- scattering code show the robustness of the modeling in the blue and in the red for aerosol optical thicknesses as great as 0.6 and solar zenith angles as large as 60 degrees. The model begins to fail only in the blue for larg e solar zenith angles. The benefits expected for aerosol remote sensing ove r land are evaluated with an aerosol retrieval scheme developed for the Med ium-Resolution Imaging Spectrometer. The main result is a better constraint on the aerosol model with inclusion of directional effects and a weaker ef fect on the optical thickness of the retrieved aerosol. The directional sch eme is then applied to the aerosol remote-sensing problem in actual Indian Remote Sensing Satellite P3/Modular Optoelectronic Scanner images over land and shows significant improvement compared with a Lambertian algorithm. Mo reover, it confirms our main theoretical conclusion. (C) 2001 Optical Socie ty of America.