Publications

Song, ZG; He, XQ; Bai, Y; Dong, XY; Wang, DF; Li, T; Zhu, QK; Gong, F (2023). Atmospheric correction of absorbing aerosols for satellite ocean color remote sensing over coastal waters. REMOTE SENSING OF ENVIRONMENT, 290, 113552.

Abstract
The atmospheric correction (AC) of satellite ocean color remote sensing, which is affected by the vertical distribution of absorbing aerosols (AA), is prone to large uncertainties in the waters with AA, especially for remote sensing reflectance (Rrs(& lambda;)) retrievals in the blue and ultraviolet bands. In this study, a new AC algorithm called the Ocean Color-XGBoost with radiative transfer (RT) simulation Retrieval Tool (OC-XGBRT), which is based on extensive radiative simulation and machine learning model, was developed to retrieve the Rrs(& lambda;) at the short wavelength bands of visible light in the presence of AA while considering their vertical distributions. The goal of the OC-XGBRT AC algorithm is to improve the quality of ocean color satellite products over inland and coastal waters with AA and heavy aerosol loadings. The images acquired by Moderate Resolution Imaging Spectroradiometer (MODIS) onboard the Aqua satellite (MODIS-Aqua) over six regions with the cases of AA were utilized to assess the performance of OC-XGBRT. The Rrs(& lambda;) retrieved by OC-XGBRT were validated with in situ measurements from the Aerosol Robotic Network-Ocean Color (AERONET-OC) and the SeaWiFS Bio-optical Archive and Storage (SeaBASS). The results showed that OC-XGBRT provided more accurate Rrs(412 nm) and Rrs(443 nm) products than the National Aeronautics and Space Administration (NASA) SeaWiFS Data Analysis System (SeaDAS) v7.5 standard near-infrared (NIR) AC algorithms, the POLYnomial-based approach established for the atmospheric correction of MERIS data (PLOYMER), and the Ocean Color-Simultaneous Marine and Aerosol Retrieval Tool (OC-SMART) over coastal and inland waters, with mean absolute percentage deviation (APD) and root mean square error (RMSE) values less than -36.9% and - 5.5 x 10-4 sr-1, respectively. Application results of OC-XGBRT in six typical regions further demonstrated that the quality of retrieved Rrs(412 nm) and Rrs(443 nm) had substantially improved compared with the original satellite products under the condition of AA. Moreover, the implementation of the OC-XGBRT contributes to a significant enhancement in the spatial coverage of utilizable Rrs(& lambda;) values at blue bands. Our proposed OC-XGBRT AC algorithm has the potential to process ocean color satellite data under the condition of AA.

DOI:
10.1016/j.rse.2023.113552

ISSN:
1879-0704