Quality assurance in the identification of deep-sea taxa from video and image analysis: response to Henry and Roberts

• Published in: ICES Journal of Marine Science
• Main Authors: Howell, Kerry L., Bullimore, Ross D., Foster, Nicola L.
• Format: Article in Journal/Newspaper
• Language: English
• Published: Oxford University Press (OUP) 2014
• Subjects: Anton Dohrn Seamount; George Bligh Bank; Hatton-Rockall Basin; Hebrides Terrace Seamount; Rockall Bank; Lophelia pertusa; Northeast Atlantic
• Online Access: http://dx.doi.org/10.1093/icesjms/fsu052; http://academic.oup.com/icesjms/article-pdf/71/4/899/29148731/fsu052.pdf; http://academic.oup.com/icesjms/article-pdf/71/4/899/23614665/fsu052.pdf

New high-resolution image data obtained from the Hebrides Terrace Seamount and analysed by ourselves and Henry and Roberts (Henry, L-A., and Roberts, J. M. Recommendations for best practice in deep-sea habitat classification: Bullimore et al. as a case study. ICES Journal of Marine Science, 71: 895–898.), suggested that we may have misidentified Solenosmilia variabilis as either Lophelia pertusa or Madrepora oculata in a previously analysed dataset from the Anton Dohrn Seamount (published in Bullimore et al., 2013). Therefore, we undertook a reanalysis of our entire image data holdings from multiple sample sites and identified possible records of S. variabilis from four sites previously sampled: Anton Dohrn Seamount, Rockall Bank, George Bligh Bank and the Hatton-Rockall Basin. The reanalysis of our image data holdings together with historic data from the wider literature suggests that, in the Northeast Atlantic region, S. variabilis is distributed from 888–2803 m (mean ∼1500 m) with reef habitat present only on Anton Dohrn Seamount. In this paper we discuss the use of video and imagery as a survey and monitoring too and make recommendations of best practice in data acquisition and analysis. We highlight the need for the development of training materials for deep-sea field identification in order to achieve reliable, replicable and comparable datasets among observers, and suggest possible quality assurance procedures.