Empirical line model for the atmospheric correction of sentinel-2A MSI images in the Caribbean Islands. Issue 1 (1st January 2018)

• Record Type: Journal Article
• Authors: Ariza, Alexander; Robredo Irizar, Marina; Bayer, Steven
• Is Part Of: European journal of remote sensing. Volume 51:Issue 1(2018)
• Page Start: 765
• Page End: 776
• Publication Date: 2018-01-01
• Subjects: ELM -- Sentinel-2A -- SEN2COR -- ATCOR -- Caribbean Islands -- atmospheric correction; Remote sensing -- Periodicals; Remote sensing; Electronic journals; Periodicals; 621.3678
• DOI: 10.1080/22797254.2018.1482732
• Languages: English
• ISSNs: 2279-7254

Abstract

ABSTRACT: An Empirical Line Model (ELM) was tested to correct Sentinel-2A (MSI) images acquired in the tropical archipelago of San Andrés, Colombia. This approach uses a linear regression to model the relationship between the average ground reflectance and radiance on bands 2, 3, 4, and 8, for 32 spectrally homogeneous targets. The model was validated from eight targets measured on different land-covers trough the estimated coefficient of determination R 2 . The result of the prediction equations observed was high, with a value of R 2 :0.91 performed at the 0.01 level of significance for the four wavebands, against the R 2 :0.77 of SEN2COR and R 2 :0.81 of ATCOR Correction Models. Complementary, a quantitative approach was proposed to determine the suitability of the ELM, based on the spectral response on six land-covers types for every band after correction. A separability index (M) was used from a set of independent targets to estimate the effectiveness of spectral classification of land-covers. The more evident results of the correction are on the vegetation cover in the NIR band (785–900 nm), where the ELM has 55% and 58% more separability than the SEN2COR and ATCOR models, respectively. Additionally, the absolute difference between the Top-of-Atmosphere (TOA) and Bottom-of-Atmosphere (BOA) images was calculated, finding the highest differences in the NIR band with 0.094 in the L1C-TOA reflectance image, and 0.013 in the ELM-BOA image. Finally, a sensitivity analysis on the Normalized Difference Vegetation Index (NDVI) to estimate the performance of the spectral response of ELM on vegetation cover was employed.