Hyperspectral Imaging of Seagrass and Macroalgae Using Uncrewed Aerial and Surface Vehicles for Coastal Habitat Mapping

Sammendrag

Seagrass and macroalgae are critical components of shallow coastal habitats undergoing fragmentation due to environmental and anthropogenic stressors. Mapping and monitoring their extent are essential for understanding and managing ecosystem changes. Hyperspectral imaging (HI) is an emerging tool for ocean mapping, providing spectral reflectance per image pixel for benthic habitat mapping. This study evaluated the use of multiscale hyperspectral mapping of shallow seagrass and macroalgae habitats using two platforms: an Uncrewed Aerial Vehicle (UAV-HI, 10 × 10 cm spatial resolution, 20,200 m2 coverage) and an Uncrewed Surface Vehicle (USV-UHI, 1 × 1 cm resolution, 230 m2 coverage). A spectral angle mapper (SAM) classification algorithm was applied with varying thresholds to assess classification performance. As much as 2316 m2 (11.5% of total area) was estimated to be seagrass by spectral angle mapper algorithms; of this, 150 m2 was mapped at the centimeter scale by the USV. The overall accuracies of the SAM classifications reached 89% and 86% for the UAV and USV, respectively, for a subsampled overlapping area. With the training data available, the UAV performs better in the accurate classification of seagrass and brown algae; however, this is collected at a coarser resolution (10 × 10 cm). The lower accuracy across both platforms is explained by a higher rate of false positives in the sediment class, which was supported by corresponding Intersect over Union (IoU) and recall metrics. The classification and accuracy assessment presented in this study support the proposed uses of both platforms. These results contribute to the development of a methodological approach to hyperspectral imaging applications for mapping seagrass and macroalgal habitats, with clear relevance for environmental monitoring and management.