Quantifying the ability of imaging sonar to identify fish species at a subtropical artificial reef

Abstract

Abstract Imaging sonars (ISs) are high-frequency acoustic devices that are increasingly being used to study fish in marine and freshwater habitats. Acoustic devices are limited in quantifying species richness, and previous attempts to identify fish species using IS have mostly focused on assemblages of low species richness or high morphological diversity. This study aimed to determine the ability of IS for identifying fish species at a subtropical artificial reef off Perth, Western Australia. Several fish traits that could be defined using IS were identified and described for all fish species observed with simultaneous optical footage. These traits were used to create a clustering algorithm to infer the species identity of IS detections of the five most abundant species at the reef. The identities of all fish from two species (Chromis westaustralis and Neatypus obliquus) were inferred with 100% success, though no individuals from the remaining three species (Seriola dumerili, Coris auricularis, and Pempheris klunzingeri) were correctly identified. An alternative clustering-based approach to categorising fish detected by IS independent of taxonomic inference was also implemented. Overall, this study demonstrates that IS can identify reef fish with variable success, and proposes an alternative method for describing fish assemblages irrespective of species identity.