A new method for ecological surveying of the abyss using autonomous underwater vehicle photography. (18th November 2014)

• Record Type: Journal Article
• Authors: Morris, Kirsty J.; Bett, Brian J.; Durden, Jennifer M.; Huvenne, Veerle A. I.; Milligan, Rosanna; Jones, Daniel O. B.; McPhail, Stephen; Robert, Katleen; Bailey, David M.; Ruhl, Henry A.
• Is Part Of: Limnology and oceanography, methods. Volume 12:Number 11(2014:Nov.)
• Page Start: 795
• Page End: 809
• Publication Date: 2014-11-18
• Subjects: Limnology -- Methodology -- Periodicals; Oceanography -- Methodology -- Periodicals; 551.48
• DOI: 10.4319/lom.2014.12.795
• Languages: English
• ISSNs: 1541-5856

Abstract

<abstract abstract-type="main"> <title> <x xml:space="preserve">Abstract</x> </title> <p>The extent and speed of marine environmental mapping is increasing quickly with technological advances, particularly with optical imaging from autonomous underwater vehicles (AUVs). This contribution describes a new deep‐sea digital still camera system that takes high‐frequency (>1 Hz) color photographs of the seafloor, suitable for detailed biological and habitat assessment, and the means of efficient processing of this mass imagery, to allow assessment across a wide range of spatial scales from that of individual megabenthic organisms to landscape scales (>100 km<sup>2</sup>). As part of the Autonomous Ecological Surveying of the Abyss (AESA) project, the AUV Autosub6000 obtained > 150, 000 seafloor images (~160 km total transect length) to investigate the distribution of megafauna on the Porcupine Abyssal Plain (4850 m; NE Atlantic). An automated workflow for image processing was developed that corrected nonuniform illumination and color, geo‐referenced the photographs, and produced 10‐image mosaics ('tiles, ' each representing a continuous strip of 15‐20 m<sup>2</sup> of seafloor), with overlap between consecutive images removed. These tiles were then manually annotated to generate biological data. This method was highly advantageous compared with alternative techniques, greatly increasing the rate of image acquisition and providing a 10‐50 fold increase in accuracy in comparison to trawling. The method also offers more precise density and biodiversity estimates [Coefficient of variation (CV) < 10%] than alternative techniques, with a 2‐fold improvement in density estimate precision compared with the WASP towed camera system. Ultimately, this novel system is expected to make valuable contributions to understanding human impact in the deep ocean.</p> </abstract>