- How can PAM be used to estimate marine mammal populations and individual animal locations before, during, and after industry operations?
- Can new software modules be developed for PAMGuard that improve the classifiers and detectors for large whales, dolphins, porpoises, beaked whales and other toothed whales, and for dolphins by their whistles?
Software that can detect animals, classify them to species, and locate them in three dimensions (called DCL software) is essential to PAMGuard. The Programme funded three projects to improve PAMGuard’s DCL capabilities. Project 1 (AKOOSTIX) focused on improving software algorithms to identify vocalising mysticete species. Project 2 (SMRU) focused on real-time species identification of dolphins, porpoises, beaked whales and other odontocete (toothed) whales. Project 3 (University of Hawaii) focused on classifying dolphins using whistle patterns (referred to as ROCCA).
Project 1 improved an existing algorithm for detecting large whales and designed, tested, and completed a module for PAMGuard
Project 2 developed an automatic classifier for 17 species of toothed whales (and one moan detector for a large whale) and incorporated these into PAMGuard. With an appropriate hydrophone array, this module can also display bearing and range to the calling animal. The project also developed a second module that classifies groups of whistles, and that can be trained to classify new species. Finally, it developed a beaked whale click detector.
Project 3 made a 47% improvement in the ability of ROCCA software to identify the whistles of 7 species of toothed whales. It also made variants for species from three regions of the Pacific Ocean, added four new species to the classifier, and converted the ROCCA software from MATLAB to JAVA to make it more user-friendly.
Objectives and methods
- Identify and build upon existing detection algorithms that have been previously developed and test for several species of marine mammals
- Create open-source detection modules for species or animal groups for integration into PAMGuard
- improve PAMGuard’s ability to identify many cetaceans to species based on the sounds they produce.
The ability to classify animals to species is less important to seismic operations than is detecting animal presence and location relative to the safety zone. However, species identification is of primary importance to the regulatory and scientific communities. The PAMGuard standard user interface software and modules have greatly improved detection and classification capabilities (such as for beaked whales). Further improvements to PAMGuard may well increase the number of surveys that use the system.
Links to other research
- Project 1: Akoostix Inc. (Joe Hood)
- Project 2: SMRU/University of Southampton (Doug Gillespie, Paul White)
- Project 3: University of Hawaii (Whitlow Au, Julie Oswald)