• cetacean
• conservation
• offshore
• passive acoustic monitoring
• towed hydrophone

Cetacean distribution and abundance within Irish waters is poorly known, especially offshore. The present thesis focuses on the use of passive acoustic monitoring (PAM) during offshore surveys in Irish waters, in order to acquire a better understanding on habitat choice and abundance of odontocetes in offshore Irish waters. By using a towed hydrophone array, diagnostic whistles, clicks and burst pulsed sounds from odontocetes were recorded during 79 days spent at sea. A total of 6 species were identified using the manual method of analysis of the audio files, consisting in the identification of vocalizations from spectrograms by eye while simultaneously listening to acoustic files using spectrograms.
The manual method proved to be very powerful in detecting vocalizations from Sperm whale, Long finned pilot whale and Harbor porpoise. Common dolphin, Bottlenose dolphin and Northern bottlenose whale could not be identified directly using only vocalizations, but the simultaneous associated sighting of the animals was needed to confirm the species identification. With the exception of the Harbor porpoise, which seems to be confined to shallow waters, and the Common dolphin, which was encountered almost everywhere, the other species identified were encountered mainly on steep depth gradients along the shelf edge, suggesting this area to be an important habitat for most of the odontocetes likely to be encountered in Irish waters.
In the last years attempts have been made in order to develop automated software able to identify cetacean vocalizations automatically. The efficacy of click and whistles identification using the automated methods of analyses was evaluated by comparing the performances of the automated software PAMGUARD to those of the manual method of analysis. The results show that, in order to correctly identify clicks, PAMGUARD should be always used in conjunction with the manual method of analysis. The manual method, although requiring much more time than the automated method, seems to be the only system able to identify whistles with fair accuracy.