Mapping underwater ecosystems through remote sensing: novel approaches to expand the scales of global biodiversity tracking.

Ocean environmental conditions are changing at unprecedented rates causing profound ecosystem transformations. These changes are especially important along the coast where intensifying human activity tends to concentrate. As a result, underwater vegetated ecosystems are severely degrading and substantial areas of  both seagrass meadows and macroalgal forests have already been lost. Preserving these crucial foundation species is an  imminent societal challenge, but to accomplish this, we first need to be able to robustly survey coastal waters at a broad spatial and temporal scales. In this context, satellite remote sensing has emerged as a magnificent approach to expand the scope of in situ observations in a cost-effective way. Here, we test whether Sentinel-2 will be useful to monitor underwater habitats in the “Freus de Ibiza y Formentera” Marine Protected Area. We use training data to calibrate machine learning algorithms that classify satellite imagery into four broad habitat types: sand, macroalgae, Posidonia oceanica and Cymodocea nodosa. The resulting distribution map reaches 20 m deep and shows an overall accuracy of 78% when compared with available ground truth data. In conclusion, this work shows that Sentinel-2 is a useful tool for coastal habitat mapping which can be relevant to monitor Essential Biodiversity Variables, such as the ecosystem distribution of seagrass meadows and macroalgal forests at large spatial and temporal scales.