Monitoring haddock in the Stellwagen Bank National Marine Sanctuary using passive acoustics

Haddock (Melanogrammus aeglefinus) are important commercial resources in the western North Atlantic. Their distribution ranges from Newfoundland to North Carolina and are most abundant on Georges Bank and in the Gulf of Maine. Over the course of a decade, the Gulf of Maine warmed faster than 99% of the global ocean and these changes are likely to impact the distribution and reproduction of haddock. It is therefore increasingly important to closely monitor this species and ensure a sustainable fishery. The objective of this study is to monitor the spatial and temporal occurrence of haddock using the sound they produce. We manually annotated over 20 000 haddock calls at five different sites and four different years and used them to train a convolutional neural network that can automatically detect haddock calls in acoustic recordings. The detector was then used to analyze several months of passive acoustic data collected in the Stellwagen Bank National Marine Sanctuary. The analysis revealed that haddock calls were mostly detected during the spawning season (January to March) but were also detected as late as August. We discuss how passive acoustics can complement existing monitoring methods and support the management of this commercially and ecologically important species.