ESR prepress abstract   -  DOI: https://doi.org/10.3354/esr01384

ABSTRACT: With fewer than 360 individuals remaining, critically endangered North Atlantic right whales Eubalaena glacialis (henceforth, ‘NARW’) embody New England’s foremost marine conservation challenge. Every year, a large portion of the NARW population visits Cape Cod Bay (CCB), MA, USA, a critical foraging area. Traditionally, aerial surveys have documented the abundance and distribution of NARWs in CCB. In this work, we demonstrate abundance estimation through passive acoustic monitoring, utilizing recordings from an array of 5 marine autonomous recording units (MARUs) deployed from February to June 2019. We first trained, validated, and applied a deep-learning-based detector for NARW upcall vocalizations, achieving a precision of 0.857 and recall of 0.896. Next, we matched duplicate detections across the MARU array through time-difference-of-arrival association. Lastly, after calibrating acoustic cue counts to concurrent aerial surveys conducted by the Center for Coastal Studies, we estimated daily NARW abundance in CCB across the foraging season. We demonstrated diel and seasonal patterns in acoustic phenology consistent with prior studies. Upcall rates were higher at night, particularly just after sunset, than during daylight hours. We observed low presence of NARWs in late February, with presence rising in early March, peaking in late March and early April, and ultimately decreasing through mid-May. While many sources of uncertainty limit the precision of abundance estimates, PAM offers a cost-effective, generalizable, and largely automated approach for detecting NARW abundance trends applicable to both conservation management and ecological studies.