Mapping Subtidal Marine Forests in the Mediterranean Sea Using Copernicus Contributing Mission
Mediterranean subtidal reefs host ecologically significant habitats, including forests of <i>Cystoseira</i> spp., which form complex benthic communities within the photic zone. These habitats are increasingly degraded due to climate change, invasive species, and anthropogenic pressures,...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-07-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/17/14/2398 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Mediterranean subtidal reefs host ecologically significant habitats, including forests of <i>Cystoseira</i> spp., which form complex benthic communities within the photic zone. These habitats are increasingly degraded due to climate change, invasive species, and anthropogenic pressures, particularly in the eastern Mediterranean. In support of habitat monitoring under the EU Natura 2000 directive and the Nature Restoration Regulation, this study investigates the utility of high-resolution satellite remote sensing for mapping subtidal brown algae and associated benthic classes. Using imagery from the SuperDove sensor (Planet Labs, San Francisco, CA, USA), we developed an integrated mapping workflow at the Natura 2000 site GR2420009. Aquatic reflectance was derived using ACOLITE v.20250114.0, and both supervised classification and spectral unmixing were implemented in the EnMAP Toolbox v.3.16.3 within QGIS. A Random Forest classifier (100 fully grown trees) achieved high thematic accuracy across all habitat types (F1 scores: 0.87–1.00), with perfect classification of shallow soft bottoms and strong performance for <i>Cystoseira</i> s.l. (F1 = 0.94) and <i>Seagrass</i> (F1 = 0.93). Spectral unmixing further enabled quantitative estimation of fractional cover, with high predictive accuracy for deep soft bottoms (R<sup>2</sup> = 0.99; RPD = 18.66), shallow soft bottoms (R<sup>2</sup> = 0.98; RPD = 8.72), <i>Seagrass</i> (R<sup>2</sup> = 0.88; RPD = 3.01) and <i>Cystoseira</i> s.l. (R<sup>2</sup> = 0.82; RPD = 2.37). The lower performance for rocky reefs with other cover (R<sup>2</sup> = 0.71) reflects spectral heterogeneity and shadowing effects. The results highlight the effectiveness of combining classification and unmixing approaches for benthic habitat mapping using CubeSat constellations, offering scalable tools for large-area monitoring and ecosystem assessment. Despite challenges in field data acquisition, the presented framework provides a robust foundation for remote sensing-based conservation planning in optically shallow marine environments. |
|---|---|
| ISSN: | 2072-4292 |