Towards a Method for Detecting Macroplastics Floating in Coastal Waters using Sentinel-2 Earth Observation Data.
We present four case studies where Sentinel-2 was used to detect aggregations likely to include floating macroplastics, namely Scotland, Ghana, South Africa, and Canada. Level 1C Sentinel-2 data were corrected using the ACOLITE dark spectrum fitting atmospheric correction, and floating aggregations were detected using a Floating Debris Index (FDI) developed for the Sentinel-2 on-board Multi-Spectral Instrument (MSI). In all case studies, patches of debris were detectable on sub-pixel scales, and appeared to be largely composed of a mix of plant materials and plastics. This finding on composition was supported by the literature, local environmental reports, and photos posted to the social media platform, Twitter. By leveraging spectral shape to identify pixels dominated by plastics, and a Naive Bayes algorithm for classification, it appears possible to distinguish floating macroplastics from natural materials in mixed aggregations.
Our planned next steps are to automate detection and classification of floating aggregations through machine learning and artificial intelligence approaches. In order to further develop links between debris accumulation and macroplastic abundance, however, access to standardised in situ data is essential.