Abstract
This research compares whale and marine vessel detection methods through performance metrics adapted from machine-learning models. Monitoring whale habitat use and vessel infractions in exclusion zones can inform adaptive management for whale recovery efforts. Land-based cetacean observation (LBCO) surveys and dedicated vessel surveys (DVS) were conducted during the summer of 2023 and are considered the gold standard methods for this study. Data collected for comparison from alternative detection methods include a citizen science network, thermal imaging, acoustic, radar, and automatic identification systems (AIS). The citizen science network was the most reliable method for whale detection of all species observed. Vessel detection methods demonstrated similar overall detection reliability, as radar consistently had higher recall values while AIS consistently had higher precision values. Differing scenarios where human observation is unlikely to be the gold standard are discussed and are recommended as a topic for continued research.