ABSTRACT
Radiance data measured by a hyperspectral sensor contain atmospheric effects, which include absorption by atmospheric water vapor and gases (e.g., oxygen and ozone), atmospheric molecular scattering (Rayleigh effect), and aerosol absorption and scattering. These atmospheric effects need to be corrected by converting the at-sensor radiance data to surface reflectance in order for hyperspectral data to be used for quantitative remote sensing. This chapter begins with an introduction of atmospheric effects on hyperspectral data. It describes statistics-based atmospheric correction approaches, including empirical line method, internal average relative reflectance, flat-field correction, cloud shadow method, and dense dark vegetation algorithm. The chapter describes the radiative transfer modeling based atmospheric correction techniques for land and water/ocean applications separately, because water/ocean surfaces are much darker than land surfaces and the air-water interface is not Lambertian, very accurate modeling of atmospheric absorption, scattering effects, and the specular water surface reflection effects is required. The chapter describes six popular radiative modeling based atmospheric correction techniques for land applications. It describes various radiative modeling based atmospheric correction techniques for water/ocean applications, including black-pixel NIR algorithm, NIR similarity spectrum algorithm, NIR-SWIR algorithm using turbid water index, self-contained atmospheric parameter estimation, modified black-pixel NIR algorithm, and direct inversion approach using neural network.
