Assessment of the accuracy of Sentinel-3 OLCI L2 products retrieved by standard and regional algorithms for ecological monitoring of the Black Sea coastal and shelf waters

Abstract

The Black Sea is an inland sea strongly affected by climate change and anthropogenic activities. Regular monitoring is required to keep the sea "healthy" and productive. The Ocean Color (OC4), neural network of the Inverse Radiative Transfer Model (NN), generalized inherent optical properties (GIOP) algorithms and a combination of two algorithms of the band ratio and the color index (OCI) were validated using a database of co-located (day-to-day) measurements (145 matchup pairs) and the OLCI onboard Sentinel-3A/3B data. In situ data were collected around the Crimean Peninsula from August 2018 to February 2022. Comparison with in situ data showed, that these algorithms performed imperfectly retrieving chlorophyll a concentration (Chl-a), light absorption coefficient by colored detrital matter ${(a}_{\mathrm{CDM}})$ and phytoplankton, which limited the use of these standard satellite products. We validated regional three-band algorithm (Chl-CDM). Statistical analyses demonstrated that the Chl-CDM algorithm provided better performance for $a_{\mathrm{CDM}}\left(490\right)$ and total non-water absorption coefficient ($a_{\mathrm{tot}}\left(490\right)$) compared to standard algorithms. The Chl-CDM algorithm is suitable for all seasons due to seasonality in absorption parameterization used. Analysis showed limitation of applicability of the current version of Chl-CDM algorithm, namely underestimation of retrieval products in waters with relatively high content of the optically active components ($a_{\mathrm{tot}}\left(490\right)\ge 0.1$m⁻¹), which is connected with the particular spectral bands used. The reasons of observed underestimation of the products in summer and in early autumn might be caused by increasing water turbidity due to coccolithophore blooming. Additional satellite products (euphotic zone and spectral features of downwelling irradiance) were proposed. The Chl-CDM algorithm provides a major step forward for simultaneous retrieval of multiple inherent optical properties and two additionally proposed water quality indicators. This study highlights the need for further improvement of current version of the Chl-CDM algorithm to spread algorithm applicability to waters with high content of optically active components.