Sen2TimeFusion: a remote sensing image processing chain to improve deeper seafloor feature extraction using Sentinel-2 time-series imagery - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

Sen2TimeFusion: a remote sensing image processing chain to improve deeper seafloor feature extraction using Sentinel-2 time-series imagery

Résumé

Coastal ecosystems are facing increasing environmental pressures and anthropogenic disturbances, leading to a worldwide degradation of their health. This situation generates an urgent need to improve methods used to study and monitor those ecosystems and their vulnerability to different stressors. Optical remote sensing (RS) is commonly used to map and monitor coastal benthic habitats. Because of limitations brought by water characteristics on ind ividual satellite images (e.g., surface reflection, water column absorption, turbidity), RS images are often only useful to map clear shallow waters when light conditions are the most favorable. With rapid improvements in RS sensors and the acquisition of images over longer time series at high spatial and temporal resolutions, large collections of images acquired at a same site can be processed together to improve the ability to extract useful information on marine environments. To deal with such large time-series of satellite data, we have developed processing chains that allow different levels of processing (data download, pre-processing for atmospheric and cloud effects, and production of indices and temporal statistics). In addition, the possibility of automating the various steps reduces human intervention and processing time, allowing computation over large regions and long periods of time. Seafloor information was derived from a time series of selected cloud-free images in the blue spectral band, less sensitive to disturbances related to water characteristics (e.g., surface states and water quality and absorption). The signal to noise ratio was improved by averaging pixel values, revealing information on the seabed at greater depths than using traditional analyses of individual images. The number of images used (up to several hundred) determines the maximum depth up to which the seafloor can be perceived (beyond 70m in places). This novel method has been applied to remote areas of several islands of the Southwest Indian Ocean region (i.e., Reunion Island, Madagascar and Seychelles deep banks) that remained poorly mapped until now, providing important baseline data at these locations. With the guarantee offered by the European Space Agency (ESA) that the Sentinel satellites will be maintained and the data will be available for any years to come, long time series will make it possible to carry out temporal analyses to highlight and quantify spatial and temporal changes in benthic habitats over large areas. Such analysis can greatly help inform management strategies for the conservation of many types of shallow benthic habitats, such as coral reefs and seagrass beds.
Fichier non déposé

Dates et versions

hal-04440404 , version 1 (06-02-2024)

Identifiants

  • HAL Id : hal-04440404 , version 1

Citer

Pascal Mouquet, Christophe Révillion, Rodolphe Devillers. Sen2TimeFusion: a remote sensing image processing chain to improve deeper seafloor feature extraction using Sentinel-2 time-series imagery. Geohab 2023, Geohab, May 2023, Saint-Gilles (Reunion Island), Réunion. ⟨hal-04440404⟩
18 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More