Comparison of Physically and Image Based Atmospheric Correction Methods for Sentinel-2 Satellite Imagery

Abstract

Atmospheric correction is the process to retrieve the surface reflectance from remotely sensed imagery by removing the atmospheric effects (Scattering and Absorption). The process determines the optical characteristics of the atmosphere and then applies it in order to correct the atmospheric effects on satellite images. Two main categories of atmospheric correction methods can be identified, the ones that rely on radiative transfer modeling and the image-based ones. In this study, four methods are compared, three physically-based (6S, FLAASH, Sen2Cor) and one image-based (DOS) for their effectiveness on atmospheric correction of Sentinel-2 high resolution optical imagery. A Sentinel-2 image, acquired on a clear day over Heraklion, Greece was used. Ancillary information on the aerosol optical thickness from the Moderate Resolution Imaging Spectroradiometer (MODIS) was used for the physically based methods. In line with similar studies using Landsat images, the physically based methods perform better than the image-based ones also for the Sentinel-2 imagery. Nevertheless, their high computational demand and the need for ancillary atmospheric information makes them difficult to apply. Different atmospheric correction methods showed different results for specific land cover types, suggesting that the selection of the suitable method is also application dependent.