Passive acoustic monitoring on the seasonal species composition of cetaceans from the marine cable hosted observatory

Information on the species diversity and habitat use of cetaceans can help us to understand the community ecology of marine top predators. Passive acoustic monitoring has been widely applied in the cetacean research in recent years since cetacean vocalizations can be effectively detected through automatic detection methods. However, species identification based on tonal sounds remains challenging due to the high intra-specific variation. In order to examine the seasonal changing pattern of species diversity and species-specific occurrence, we applied an automatic detection and classification algorithm on the acoustic recordings collected from the marine cable hosted observatory (MACHO) off the northeastern Taiwan. The representative frequencies of cetacean tonal sounds within 4.5-48 kHz were detected by the local-max detector. Twelve statistical feature vectors were extracted based on the distribution of representative frequency and were used in the discriminant function analysis to classify four cetacean groups. The correct classification rate was 72.2% based on the field recording collected from onboard surveys. Analysis on one-year MACHO recordings revealed that the species diversity was highest in winter and spring. In addition, different patterns of seasonal occurrence were observed between different cetacean groups. Short finned pilot whales and Risso´s dolphins were the most common species, they mainly occurred in winter and summer. False killer whales were mostly detected in winter and spring. On the contrary, smaller delphinids, such as spinner dolphins, spotted dolphins, and Fraser´s dolphins were mainly detected in summer. Bottlenose dolphins represent the least common species. The current results show that the seasonal occurrence of multiple cetacean species can be effectively monitored by a marine observatory. In the future, the biodiversity, species-specific habitat use, and inter-specific interaction of cetaceans can be investigated through an unde- water acoustic monitoring network.