Abstract
Visual shipboard and aerial surveys are the most widely used form of cetacean assessment but are limited by sighting probabilities, as cetaceans are often not readily detectable at the surface and are highly mobile, daylight, and weather. In comparison, acoustic surveys are constrained by vocal activity and the ability to identify species. The combination of both methods, however, is more powerful than either method alone; passive acoustic data can supplement visual data collection, particularly with the development of methodologies to identify species by their vocalizations with minimal constraints. An algorithm written by Julie Oswald at Bio-Waves, Inc. called ROCCA (Real-Time Odontocete Call Classification Algorithm) was evaluated using PAMGUARD to test for the program’s ability to identify and discriminate between five Atlantic delphinid species using passive acoustic data. The data used in this study were collected in the northern and southern regions of the western North Atlantic Ocean aboard the R/V Bigelow by NOAA/NEFSC and the R/V Gordon Gunter by NOAA/SEFSC in the summer of 2013 using a towed hydrophone array. Results showed regional differences in correct classification rates, with SEFSC data associated with a higher Kappa statistic (0.314) than NEFSC data (0.102), but an overall better recognition in both regions for Tursiops truncatus and Stenella frontalis than for Globicephala spp. and Stenella coeruleoalba. Contour variables were also analyzed for each species to determine unique species characteristics. The results and analyses provide developmental insight and strategies regarding how to further refine ROCCA for real-time use in both regions and to contribute to improved marine mammal management by enhancing visual abundance studies with acoustic data.