Intra-specific variability in delphinid whistle structure: Implications for acoustic species identification

About

Passive acoustic methods are commonly used to study and monitor marine mammals. Identifying sounds recorded from marine mammals to species can be challenging and often necessitates using statistically based classification analysis. Training and testing classifiers requires visually validated, single-species recordings and large amounts of data, which can be difficult to obtain. Combining data recorded from different locations to train classifiers would allow for larger sample sizes, however, due to geographic variation in signal structure, it has been suggested that classifiers should be trained using recordings made in the locations where they will be used. To examine the effect of geographic variation on classifier performance, whistles were measured from three delphinid species that occur in both the eastern tropical Pacific (ETP) and northwest Atlantic (NWAt) oceans: striped dolphins (Stenella coeruleoalba), bottlenose dolphins (Tursiops truncatus), and short-finned pilot whales (Globicephala macrorhynchus).