Approaches for Identification of Delphinid Species in Passive Acoustic Data for Population Management Applications

Rebecca Cohen1, Kaitlin E Frasier2 and John Hildebrand2, (1)University of California, San Diego, Scripps Institution of Oceanography, La Jolla, CA, United States, (2)University of California San Diego, Scripps Institution of Oceanography, La Jolla, CA, United States
Abstract:
Utilization of marine passive acoustic data sets for cetacean population assessment and management is contingent upon our ability to acoustically discriminate between species. 16 delphinid species are known to inhabit the western North Atlantic based on historical sighting and stranding data. While their echolocation clicks are often recorded in large, high-quality data sets collected by moored autonomous passive acoustic sensors, the particular species present during these acoustic encounters is generally unknown. We investigated multiple approaches to matching distinct click types to the delphinid species generating them. In the first approach, we assigned species identity to acoustic encounters at five western North Atlantic mooring sites using colocated sighting data from shipboard and aerial visual surveys spanning eight years. In the second approach, we assigned species identity to acoustic encounters in towed acoustic array data collected concurrent with ship-based visual surveys in the western North Atlantic and the Gulf of Mexico. A total of 114 mooring encounters with seven species and 149 towed array encounters with nine species were investigated to characterize their spectral and temporal features and establish labeled click types consistently present for each species. In the third approach, we used an unsupervised clustering algorithm to identify recurring click types in four years of passive acoustic data from 11 mooring sites in the western North Atlantic, and compared time series of these unlabeled click types to historical sighting and seasonal movement data to determine likely species matches. Finally, labeled and unlabeled click types from all three approaches were compared to evaluate the consistency of putative species-specific identifying features across acoustic data types, and to determine the utility of these click types for application in delphinid species discrimination in moored passive acoustic data.