SWIR-based atmospheric correction for Satellite Ocean Color using Principal Component Analysis decomposition over the la Plata River highly turbid waters

The quality of information provided by ocean color imagery relies on the availability of an accurate atmospheric correction algorithm, which turns to be more complicated in highly turbid coastal regions, such as in the Río de la Plata River (RdP) (Argentina-Uruguay). In these waters, the usual black pixel assumption in the Near Infra-Red (NIR, 700–1000 nm) bands is often invalid due to high backscattering from suspended particulate matter (SPM) present in the water. In this work, an atmospheric correction scheme is presented to estimate water reflectance in the 865 nm NIR band. This scheme is based on shifting the black pixel assumption to the Short-Wave-Infra-Red (SWIR, 1000–3000 nm) bands and a Principal Component Analysis (PCA) decomposition of simulated atmosphere-interface reflectances. To estimate the latter component in the top of atmosphere (TOA) signal, the weight of each of these PCA eigenvectors is determined from the SWIR bands in a per-pixel-basis. The algorithm was theoretically tested from a set of simulated TOA reflectances performed considering atmospheric conditions and in-situ water reflectance data from RdP. Four schemes were analyzed using different sets of SWIR bands present in MODIS and SABIA-Mar (future Argentinian-Brazilian ocean color mission) sensors. These sets differ from each other in their correlation to the NIR and the validity of the black pixel assumption. Without considering instrument noise, the scheme with better performance is the one in which water reflectance is negligible in the SWIR bands considered: 1640 nm (SABIA-Mar/MODIS) and 2130 nm (MODIS).