Preliminary assessment of Sentinel-2 data for shallow waters’ bathymetry in Ria de Vigo, Spain

Detailed bathymetric data is essential to obtain precise and accurate shallow-water seabed maps. The Secure and Sustainable Supply of Raw Materials for EU Industry (S34I) project researched improved seabed mineral mapping while exploring the connection between seashore and coastal areas in the Ria de Vigo, part of a funnel-shaped estuarine inlet system, on the northwestern coast of Spain. This study represents a preliminary assessment of Copernicus Sentinel-2 data to achieve Satellite-Derived Bathymetry (SDB) using ensemble machine learning algorithms (Gradient Boosting, LightGBM, CatBoost and AdaBoost). For SDB model training, in-situ multi-beam echo sounder data obtained by the Spanish Navy covering a small part of the Ria de Vigo (Spain) were used. To maintain temporal consistency between satellite and in-situ data, Sentinel-2 images acquired in 2021 and covering the study area were filtered considering a maximum of 5% cloud coverage. Thirty-nine images were retrieved and pre-processed using the ACOLITE Processor. The pre-processing included atmospheric correction and removal of sun-glint, masking and the computation of water quality parameters such as turbidity, suspended matter concentration and Chlorophyll-a distribution. The best water-quality image was selected, i.e., the one with the lowest turbidity, sediments, and algae content. The obtained SDB map shows a correlation coefficient (R) of 0.62 and a Root Mean Square Error (RMSE) of 4.05m. In the future, multi-scene composites will be created, and the results will be validated against new in-situ measurements.