POSTER Session 2

DINEOF SUPER-RESOLUTION OCEAN COLOR DATA USING SENTINEL-2 AND SENTINEL-3 IN SYNERGY

Coastal regions are very dynamic, with high temporal variability occurring at small spatial scales. Analysis of ocean color satellite data in such regions should be done at adequate spatial and temporal scales, but no single sensor provides them. We exploit the synergy between Sentinel-2/MSI and Sentinel-3/OLCI to derive a unique estimate, at high spatio-temporal resolutions, using DINEOF (Data Interpolating Empirical Orthogonal Functions). Due to differences in the spectral band sets and viewing geometries between Sentinel-2/MSI and Sentinel-3/OLCI sensors, CHL algorithms developed for the OLCI sensor often cannot be directly applied to the MSI sensor. This results in CHL products from the two sensors frequently exhibiting diverging values. We increase the coherency between the Sentinel-2/MSI and Sentinel-3/OLCI CHL products using a machine learning technique (LightGBM). More complex CHL algorithms from the Sentinel-3/OLCI sensor are transferred to the Sentinel-2/MSI sensor. Subsequently, DINEOF extracts the small scales of variability from Sentinel-2 and infers them into Sentinel-3 data, through projection of an EOF basis. As a result, we obtain a cloud-free, high spatio-temporal resolution dataset. The approach is tested in 4 coastal sites around European Seas, representing different water masses (Belgian coast in the North Sea, Venice Lagoon, Gironde estuary and southwest Baltic Sea). DINEOF also improves the quality of the data by decreasing the banding effects caused by the staggered position on the detectors which are affected differently by sun glint, an effect that is most visible at low chlorophyll concentrations.

Aida Alvera-Azcarate, University of Liege, Belgium

Dimitry Van der Zande, Royal Belgian Institute of Natural Sciences (RBINS), Belgium

Alexander Barth, University of Liège, Belgium

Joppe Massant, Royal Belgian Institute of Natural Sciences (RBINS), Belgium

Code of Conduct
Privacy Policy