Using low‐cost drones to monitor heterogeneous submerged seaweed habitats: A case study in the Azores. Issue 11 (25th July 2019)

• Record Type: Journal Article
• Authors: Kellaris, Alexandros; Gil, Artur; Faria, João; Amaral, Ruben; Moreu‐Badia, Ignacio; Neto, Ana; Yesson, Chris
• Is Part Of: Aquatic conservation. Volume 29:Issue 11(2019)
• Page Start: 1909
• Page End: 1922
• Publication Date: 2019-07-25
• Subjects: algae -- alien species -- aquaculture -- archipelago -- coastal -- monitoring -- remote sensing; Aquatic ecology -- Periodicals; Conservation of natural resources -- Periodicals; Aquatic resources -- Periodicals; 333.95216
• DOI: 10.1002/aqc.3189
• Languages: English
• ISSNs: 1052-7613

Abstract

Abstract: Remote sensing is a powerful monitoring tool for seaweeds, providing large‐scale insights into their ecosystem benefits and invasive impacts. Satellites and manned aircraft have been widely used for this purpose, but their spatial resolution is generally insufficient to map heterogeneous seaweed habitats. In this study, the potential of low‐cost and high‐resolution drone imagery to map heterogeneous seaweed habitats was assessed on Azorean coasts, where an invasive and commercial species, Asparagopsis armata, is present. A Phantom Pro 3 drone equipped with a visible‐light sensor was used to create photomosaics in three sites on São Miguel island, and ground‐truth data for various seaweed groups were collected with exploratory kayak sampling. Support‐vector machine, random forest, and artificial neural network algorithms were used to construct predictive models of seaweed coverage. Wind, clouds, and sun glint were the most significant factors affecting drone surveys and images. Exploratory sampling helped locate relatively homogeneous seaweed patches; however, the data were limited and spatially autocorrelated, contributing to overoptimistic model evaluation metrics. Moreover, the models struggled to distinguish seaweeds deeper than 3–4 m. In conclusion, using drones to monitor heterogeneous seaweed habitats is challenging, especially in oceanic islands where waters are deep and weather is unpredictable. However, this study highlights the potential use of photo‐interpretation to collect modelling data from drone imagery, instead of time‐consuming exploratory ground‐truth sampling. Future studies could assess drones to map seaweeds in less challenging conditions and use photo‐interpretation to improve collection of modelling data.