|Title||Automatic classification of whistles from coastal dolphins of the southern African subregion|
|Publication Type||Journal Article|
|Year of Publication||2017|
|Authors||Erbs, Florence, Elwen Simon H., and Gridley Tess|
|Journal||The Journal of the Acoustical Society of America|
|EndNote Rec Number||11584|
|Keywords||bioacoustics,biocommunications,sampling methods,underwater sound|
Passive acoustic monitoring (PAM) is commonly used to generate information on the distribution, abundance, and behavior of cetacean species. In African waters, the utilization of PAM lags behind most other continents. This study examines whether the whistles of three coastal delphinid species (Delphinus delphis, Tursiops truncatus, and Tursiops aduncus) commonly encountered in the southern African subregion can be readily distinguished using both statistical analysis of standard whistle parameters and the automated detection and classification software PAMGuard. A first account of whistles recorded from D. delphis from South Africa is included. Using PAMGuard, classification to species was high with an overall mean correct classification rate of 87.3%. Although lower, high rates of correct classification were also found (78.4%) when the two T. aduncus populations were included separately. Classification outcomes reflected patterns observed in standard whistle parameters. Such acoustic discrimination may be useful for confirmation of morphologically similar species in the field. Classification success was influenced by training and testing the classifier with data from different populations, highlighting the importance of locally collected acoustic data to inform classifiers. The small number of sampling populations may have inflated the classification success, therefore, classification trials using a greater number of species are recommended.