Beach wrack mapping using unmanned aerial vehicles for coastal environmental management

Abstract As a common phenomenon along the global coastline, beach wrack, which is part of the blue carbon ecosystems (BCEs), has significant ecological values. However, the excessive accumulation of beach wrack can be a nuisance for local residents and tourism. Meanwhile, beach wrack can become a source of greenhouse gas due to the decomposition. Hence, effective monitoring of beach wrack has become a priority for coastal environmental management. As a cost- and labor-saving approach, unmanned aerial vehicles (UAVs) can perform customized flight tasks and achieve aerial images with sub-decimeter spatial resolution. This study investigated the feasibility of using UAVs to map wrack on three different types of beaches. The method of object-based image analysis (OBIA) was applied to classify the aerial images. Three typical machine learning methods, K-Nearest Neighbor (KNN), Support Vector Machine (SVM), and Random Forests (RF), were examined with different feature spaces at several segmentation levels. The results showed that the three machine learning methods performed well with the overall classification accuracy >75%. The tested algorithm, SVM with only RGB as feature space at the segmentation scale 50, was geographically transferable to beaches with different characteristics. This study demonstrated that UAVs can be developed as an applicable tool for beach wrack mapping and monitoring, which will help to better explore the role of beach wrack in BCEs and assist the local municipalities in environmental management of the coastal zone.