Surface reflectance estimation from satellite imagery with inhomogeneous atmospheric conditions

Surface reflectance, in principle, can be estimated from at-sensor radiance if we construct a physical model that incorporates several atmospheric an topographic parameters. Among them, spatial inhomogeneity of aerosol optical depth is the most important but difficult factor. In this paper, we propose a method simultaneosly estimating surface reflectance and optical depth by establishing their functional relation for every possible value of at-sensor radiance. It is noted that atmospheric scattering tends to affect shorter wavelength and thin haze has negligible effect on infrared bands. For effiicient calculation, correction parameters such as path radiance and diffuse solar irradinance are approximated by polynomial functions of optical depth and surface elevation. Simple derivations of background radiance and enviornmental irradiance are also given as functions of averaged reflectance of neighbor pixels and spherical albedo of the atmosphere.