Performances analysis of various atmospheric correction algorithms for monitoring Irish transitional and coastal (TrC) waterbodies using Sentinel-3 OLCI imagery

The availability of free and open data policies has led to the wide use of remote sensing (RS) products for assessing water quality (WQ) in the world. According to the literature, multispectral satellite products, including Landsat 8/9 OLI, Sentinel-2 MSI, Sentinel-3 OLCI, etc. RS products most widely utilize in the field of water quality monitoring. Among them, European Space Agency (ESA) recommended Sentinel-3 OLCI (Ocean and land color instrument) for conducting wide-scale observations of the optical and in-optical properties of ocean waters. Several recent studies have identified various challenges, including atmospheric conditions, surface reflection, adjacency effects, and sensor characteristics in RS products. For the purposes of the reliable and accurate assessment of WQ (various indicators like dissolved oxygen (DOX), pH, chlorophyll a (CHL) etc.), it is essential to eliminate the effect of the atmosphere from the RS products. However, the research aim was to compare the performance four widely used atmospheric correction (AC) algorithms such as C2RCC, ACOLITE, L2gen, POLYMER etc., using Sentinel-3 OLCI data in order to retrieve the dissolved oxygen (inactive), and chlorophyll a (active) in Irish TrC waterbodies. To assess the performance of AC, the study has developed 8188 models using the 25 ML/AI algorithms. The performance of the algorithms were evaluated using five identical metric R2, RMSE, MSE, MAE, PABE. Based on the initial assessment, the ACOLITE-AC could be effective to obtain more accurate WQ assessment, particularly DOX, and CHL using the extreme gradient boosting algorithm. The result of the research could be helpful for optimizing the atmospheric intervention in RS process in order to obtain accurate assessment of WQ in TrC waterbodies, Ireland.