Development of a machine learning detector for North Atlantic humpback whale song.

The study of humpback whale song using passive acoustic monitoring devices requires bioacousticians to manually review hours of audio recordings to annotate the signals. To vastly reduce the time of manual annotation through automation, a machine learning model was developed. Convolutional neural networks have made major advances in the previous decade, leading to a wide range of applications, including the detection of frequency modulated vocalizations by cetaceans. A large dataset of over 60 000 audio segments of 4 s length is collected from the North Atlantic and used to fine-tune an existing model for humpback whale song detection in the North Pacific (see Allen, Harvey, Harrell, Jansen, Merkens, Wall, Cattiau, and Oleson (2021). Front. Mar. Sci. 8, 607321). Furthermore, different data augmentation techniques (time-shift, noise augmentation, and masking) are used to artificially increase the variability within the training set. Retraining and augmentation yield F-score values of 0.88 on context window basis and 0.89 on hourly basis with false positive rates of 0.05 on context window basis and 0.01 on hourly basis. If necessary, usage and retraining of the existing model is made convenient by a framework (AcoDet, acoustic detector) built during this project. Combining the tools provided by this framework could save researchers hours of manual annotation time and, thus, accelerate their research.

An experimental sound exposure study at sea: No spatial deterrence of free-ranging pelagic fisha).

Acoustic deterrent devices are used to guide aquatic animals from danger or toward migration paths. At sea, moderate sounds can potentially be used to deter fish to prevent injury or death due to acoustic overexposure. In sound exposure studies, acoustic features can be compared to improve deterrence efficacy. In this study, we played 200-1600 Hz pulse trains from a drifting vessel and investigated changes in pelagic fish abundance and behavior by utilizing echosounders and hydrophones mounted to a transect of bottom-moored frames. We monitored fish presence and tracked individual fish. This revealed no changes in fish abundance or behavior, including swimming speed and direction of individuals, in response to the sound exposure. We did find significant changes in swimming depth of individually tracked fish, but this could not be linked to the sound exposures. Overall, the results clearly show that pelagic fish did not flee from the current sound exposures, and we found no clear changes in behavior due to the sound exposure. We cannot rule out that different sounds at higher levels elicit a deterrence response; however, it may be that pelagic fish are just more likely to respond to sound with (short-lasting) changes in school formation.

An echosounder view on the potential effects of impulsive noise pollution on pelagic fish around windfarms in the North Sea.

Anthropogenic noise in the oceans is disturbing marine life. Among other groups, pelagic fish are likely to be affected by sound from human activities, but so far have received relatively little attention. Offshore wind farms have become numerous and will become even more abundant in the next decades. Wind farms can be interesting to pelagic fish due to food abundance or fisheries restrictions. At the same time, construction of wind farms involves high levels of anthropogenic noise, likely disturbing and/or deterring pelagic fish. Here, we investigated whether bottom-moored echosounders are a suitable tool for studying the effects of impulsive - intermittent, high-intensity - anthropogenic noise on pelagic fish around wind farms and we explored the possible nature of their responses. Three different wind farms along the Dutch and Belgian coast were examined, one with exposure to the passing by of an experimental seismic survey with a full-scale airgun array, one with pile driving activity in an adjacent wind farm construction site and one control site without exposure. Two bottom-moored echosounders were placed in each wind farm and recorded fish presence and behaviour before, during and after the exposures. The echosounders were successful in detecting variation in the number of fish schools and their behaviour. During the seismic survey exposure there were significantly fewer, but more cohesive, schools than before, whereas during pile driving fish swam shallower with more cohesive schools. However, the types and magnitudes of response patterns were also observed at the control site with no impulsive sound exposure. We therefore stress the need for thorough replication beyond single case studies, before we can conclude that impulsive sounds, from either seismic surveys or pile driving, are a disturbing factor for pelagic fish in otherwise attractive habitat around wind farms.