Alternate Title

Validation of Image-Based Species Identifications of Black Corals (Order Antipatharia) on Mesophotic Rocks


Biodiversity, an important measure of ecosystem health, is challenging to ascertain using sampled specimens in remote deep-sea environments. As image-based identifications become a predominant method for deep-sea species characterizations, there is a need to evaluate the accuracy of species- and genus-level identifications from video and still images to provide a reliable measure of biodiversity. This study presents a validation of the ability to make accurate image-based identifications of black coral species in the northwestern Gulf of Mexico from standard-definition video collected by a remotely operated vehicle. Results indicate that the greatest number of misidentifications occurred at species-level groupings (42.2% error), whereas genus-level groupings possessed 12.0% error, and identifications to kappa groupings had no error. We recommend genus-level groupings to maintain accurate identifications while maximizing estimates of biodiversity.