Performance of glint correction algorithms for Sentinel-3 OLCI data

Rejane S. Paulino^1*Vitor Martins^1*Cassia B. Caballero¹Thainara M. A. Lima¹Daniel A. Maciel²Julio C. P. Santos²Bingqing Liu³

• ¹Mississippi State University, Starkville, United States
• ²Instituto Nacional de Pesquisas Espaciais, Sao Jose dos Campos, Brazil
• ³University of Louisiana at Lafayette School of Geosciences, Lafayette, United States

Sentinel-3 (A/B) Ocean and Land Colour Imager (OLCI) provides daily global coverage and spectral quality for monitoring optical water quality indicators across diverse aquatic systems. Accurate retrieval of remote sensing reflectance (Rrs) from OLCI imagery requires a series of radiometric correction procedures. Specifically, glint correction algorithms are essential in accounting for the impact of specular reflections from sunlight and skylight at the air-water interface, which can distort the measured radiance received by the satellite sensor. Despite its importance, the performance of glint correction algorithms remains underexplored for Sentinel-3 (A/B) OLCI images and represents a research gap for this sensor application. In this study, we analyzed the principles and performance of three image-based sunglint correction algorithms and one skyglint correction method across varying intensities of glint effects, using 570 Sentinel-3 images acquired between 2020 and 2024. Resulting Rrs retrievals were evaluated against the Aerosol Robotic Network for Ocean Color (AERONET-OC) observations at 11 coastal sites. All proposed sunglint correction methods improved Rrs retrievals compared to no glint correction over various optical water types. Among them, the combination of SCSh (i.e., a sunglint removal method designed for optically shallow waters) and SkyG (i.e., an analytical skyglint removal method) achieved the best overall performance, yielding the lowest absolute error (𝜀 < 58%) and number of spectra that were significantly overcorrected (n=99). However, challenges remain in the blue spectral range (400 – 490 nm), where the glint correction methods performed poorly compared to AERONET-OC observations, especially under medium and high-glint conditions. Moreover, glint-free images were overcorrected for all methods, highlighting the need for reliable glint detection and masking before applying corrections. Our findings demonstrated that while existing glint correction methods can significantly improve data quality under low and medium-glint conditions, the high-glint scenarios continue to pose difficulties. Addressing these limitations is essential to ensure the consistent and accurate use of the Sentinel-3 (A/B) OLCI data for aquatic monitoring.