Agathe Bouet1,2, Antoine Collin2, Dorothée James2, Eric Feunteun3,2
1Association Planète Mer, France - 2Coastal GeoEcological Lab (EPHE-PSL), France - 3MNHN Station Marine de Dinard, France
Rocky intertidal shores host a rich biodiversity, including engineer species such as canopy-forming brown macroalgae. Structuring their environment, they are essential to many animal communities and offer multiple ecosystemic services such as climate and erosion regulation. These ecosystems, already fragilized by local anthropic pressures (recreational fishing, over frequentation), are facing actual global changes (climate change, sea level rise, pollution…). Their monitoring and understanding are essential to ensure better conservation and preservation practices.
Remote sensing techniques using unmanned airborne vehicle (UAV) and multispectral sensors enable to monitor at a very high spatial resolution (decimetre scale) and higher spectral resolution than RGB sensors. They also are agile enough to cover the low tide’s short window. UAV-based imagery coupled with machine learning can be used to conduct habitat classifications.
This new study aims to classify temperate intertidal rocky shores habitats along the Emerald coast, in Brittany. UAV-based data is collected using a multispectral imager MicaSense Altum-PT, offering RGB, infra-reds and topographic data.
For the sake of transferability, the free available software SAGA-GIS is used to perform classifications of seven representative habitats of a temperate intertidal rocky shore, comparing three algorithms (Support Vector Machine, Random Forest, and Maximum Likelihood). Calibration and validation ground truths are used to train and test our classifications and its accuracy is quantified by the overall, producer’s and user’s accuracies (OA, PA and UA, respectively).
The Support Vector Machine appears to be the best classifier (OA : 98.20%) compared to the Random Forest (OA : 96.27%) and the Maximum Likelihood (OA : 94.35%). Ongoing classification of two other study sites, using the same methodology will extend these results.
Biography
Agathe works for Planète Mer at the marine station of the French National Museum of Natural History in Dinard (Brittany, France). She coordinates a national coastal biodiversity observation program (BioLit). She is also conducting a research project with the Coastal GeoEcological Lab about the spatial modelling of rocky foreshores on the emerald coast, by combining combining ecological, geographical and citizen sciences. She is developing remote sensing skills while conducting the later project.