Evaluation of the ACOLITE atmospheric correction algorithm at a tropical coastal site

We assess the performance of the ACOLITE (version 20231023.0) atmospheric correction algorithm using <i>in situ</i> radiometric instruments at a tropical coastal site. Specifically, we evaluate the ACOLITE remote sensing reflectance (<i>R</i>_<i>rs</i>) derived from Landsat-8, -9, Sentinel-2A, and -2B observations at the Integrated Marine Observing System (IMOS) Lucinda Jetty Coastal Observatory (LJCO), Queensland, Australia. The ground-based data used in our evaluation include both multi-spectral (AERONET-OC) and hyperspectral (Sea-Bird Scientific HyperOCR) measurements. Generally, we find excellent agreement (<i>r</i>² > 0.8), low root-mean-squared-error (RMSE; ∼2.4 × 10 ^-3 sr^-1), low mean bias error (MBE; ∼1.1 × 10 ^-3 sr^-1) and low mean absolute percentage error (MAPE; ∼25 %) between the ground and satellite measurements; however, performance varies when broken down by sensor channel, satellite platform and validation dataset used. Considering individual channels, across all sensors, we find that ACOLITE performs best for the green band (560 nm; <i>r</i>² = 0.91; RMSE = 1.68 × 10 ^-3 sr^-1, MBE = 3.88 × 10 ^-4 sr^-1, MAPE = 11 %) whereas the poorest performance was found for the coastal aerosol band (443 nm; <i>r</i>² = 0.65; RMSE = 3.36 × 10 ^-3 sr^-1, MBE = 2.73 × 10 ^-3 sr^-1, MAPE = 40 %). We also find that better agreement is found when the satellite retrievals are compared against HyperOCR rather than AERONET-OC. Overall, the results show that ACOLITE provides reliable and accurate <i>R</i>_<i>rs</i> for the coastal waters encountered at LJCO.