DARK OBJECT SUBTRACTION MODEL OF ATMOSPHERIC CORRECTION FOR LANDSAT-8 AND SENTINEL-2 SURFACE REFLECTANCE PRODUCT: A CASE OF EVALUATION (WATER, VEGETATION, DESERT, URBAN)

Abstract

As the fields that need the remote sensing imagery are getting increased, the need of an effective and cheap method for atmospheric correction is getting more important. There are some methods that depend on in-site measurements and some depend only on the image itself the latter is less money and time consuming. The Dark Object Subtraction (DOS) is one of the most popular methods for atmospheric correction. In this paper the DOS and cosine of the solar zenith correction COST algorithm is applied on Landsat-8 and DOS on Sentinel-2 data in different atmospheric conditions (July and December). The data are classified into four different land cover classes: water, vegetation, desert and urban. Both qualitative and quantitative analysis are implemented to compare the corrected DOS scenes to the Level 2A of each satellite and also compared to the results of QUick Atmospheric Correction (QUAC) in case of Landsat-8, as for the Sentinel 2 the results are compared to Sentinel-2 Correction (Sen2Cor) results. The results show the effectiveness of the DOS in the case of Landsat-8 for the vegetation, water, and urban scenes, as the Sentinel-2, DOS shows good results in the vegetation and water scene. However as for the desert scene the result was not satisfying especially in clear weather in both satellites due to the lack of dark object.