Résumé
Coastal bathymetry is critical to inform various aspects of coastal zones management. If high-resolution bathymetric data collected using airborne LiDAR and multibeam acoustic sonars are now available in many coastal areas, the high cost of those surveys remains a limiting factor that prevents mapping most of the worlds coastal zones. Satellite-derived bathymetry could be used to derive coastal bathymetry in those regions, but the inversion models used by those methods require in-situ depth measurements that are rarely available. Here, we report on the combination of Pleiades-1A very-high resolution imagery (0.5m) collected in May 2020 by the French space agency CNES and ICESat-2 laser altimetry data acquired in November 2018 by NASA. ICESat-2 data were first filtered using the DBSCAN algorithm to extract photons characterizing the seafloor and were then corrected for the refraction effect. Bathymetry was derived from the Pleiades images using ICESat-2 depth measurements to train the algorithm. The method was tested in the waters of the French archipelago of Mayotte in the Indian Ocean, an area which is part of a larger marine protected area (MPA), allowing to create a 0.5m resolution bathymetry down to 23 meters depth. Comparisons of the resulting bathymetry with Litto3D airborne LiDAR data collected by the French hydrographic agency SHOM helped assess a root mean square error (RMSE) of 0.93 m. Findings suggest that high-resolution bathymetric grids (0.5m) can be derived with a sub-meter vertical accuracy in optically clear coastal waters. Such a method could help produce useful bathymetry for the coast of many developing countries lacking such baseline data, helping better inform the management of their coasts.