In this work we present a super-resolution approach for deriving high-spatial-resolution and high-temporal-resolution ocean colour satellite datasets. The technique is based on DINEOF (data-interpolating empirical orthogonal functions), a data-driven method that uses the spatio-temporal coherence of analysed datasets to infer missing information. DINEOF is used here to effectively increase the spatial resolution of satellite data and is applied to a combination of Sentinel-2 and Sentinel-3 datasets. The results show that DINEOF is able to infer the spatial variability observed in the Sentinel-2 data to the Sentinel-3 data while reconstructing missing information due to clouds and reducing the amount of noise in the initial dataset. In order to achieve this, the Sentinel-2 and Sentinel-3 datasets have undergone the same pre-processing, including a comprehensive, region-independent, and pixel-based automatic switching scheme for choosing the most appropriate atmospheric correction and ocean colour algorithm to derive in-water products. The super-resolution DINEOF has been applied to two different variables (turbidity and chlorophyll) and two different domains (Belgian coastal zone and the whole of the North Sea), and the sub-mesoscale variability of the turbidity along the Belgian coastal zone has been studied.
Located in
Library
/
RBINS Staff Publications 2025
