Signals from the deep: Spatial and temporal acoustic occurrence of beaked whales off western Ireland

Discriminating and classifying odontocete echolocation clicks in the Hawaiian Islands using machine learning methods

Passive acoustic monitoring (PAM) has proven a powerful tool for the study of marine mammals, allowing for documentation of biologically relevant factors such as movement patterns or animal behaviors while remaining largely non-invasive and cost effective. From 2008-2019, a set of PAM recordings covering the frequency band of most toothed whale (odontocete) echolocation clicks were collected at sites off the islands of Hawai'i, Kaua'i, and Pearl and Hermes Reef. However, due to the size of this dataset and the complexity of species-level acoustic classification, multi-year, multi-species analyses had not yet been completed. This study shows how a machine learning toolkit can effectively mitigate this problem by detecting and classifying echolocation clicks using a combination of unsupervised clustering methods and human-mediated analyses. Using these methods, it was possible to distill ten unique echolocation click 'types' attributable to regional odontocetes at the genus or species level. In one case, auxiliary sightings and recordings were used to attribute a new click type to the rough-toothed dolphin, Steno bredanensis. Types defined by clustering were then used as input classes in a neural-network based classifier, which was trained, tested, and evaluated on 5-minute binned data segments. Network precision was variable, with lower precision occurring most notably for false killer whales, Pseudorca crassidens, across all sites (35-76%). However, accuracy and recall were high (>96% and >75%, respectively) in all cases except for one type of short-finned pilot whale, Globicephala macrorhynchus, call class at Kaua'i and Pearl and Hermes Reef (recall >66%). These results emphasize the utility of machine learning in analysis of large PAM datasets. The classifier and timeseries developed here will facilitate further analyses of spatiotemporal patterns of included toothed whales. Broader application of these methods may improve the efficiency of global multi-species PAM data processing for echolocation clicks, which is needed as these datasets continue to grow.

Automatic data selection for validation: A method to determine cetacean occurrence in large acoustic data sets

Passive acoustic monitoring (PAM) can inform wildlife management by providing information on the distribution of cetaceans. This paper presents an automatic data selection for validation (ADSV) method to effectively identify all species acoustically present in large PAM data sets. The ADSV method involves the application of automated detectors, the automated selection of a portion of data for manual review, and the evaluation/optimization of automated detectors. Using an exemplar data set, results from the ADSV method were compared to a more intensive systematic manual review method. The two methods were found to have similar species occurrence results (hourly occurrence matching 73%-100%).

Techniques for distinguishing between impulsive and non-impulsive sound in the context of regulating sound exposure for marine mammals

Regulations designed to mitigate the effects of man-made sounds on marine mammal hearing specify maximum daily sound exposure levels. The limits are lower for impulsive than non-impulsive sounds. The regulations do not indicate how to quantify impulsiveness; instead sounds are grouped by properties at the source. To address this gap, three metrics of impulsiveness (kurtosis, crest factor, and the Harris impulse factor) were compared using values from random noise and real-world ocean sounds. Kurtosis is recommended for quantifying impulsiveness. Kurtosis greater than 40 indicates a sound is fully impulsive. Only sounds above the effective quiet threshold (EQT) are considered intense enough to accumulate over time and cause hearing injury. A functional definition for EQT is proposed: the auditory frequency-weighted sound pressure level (SPL) that could accumulate to cause temporary threshold shift from non-impulsive sound as described in Southall, Finneran, Reichmuth, Nachtigall, Ketten, Bowles, Ellison, Nowacek, and Tyack [(2019). Aquat. Mamm. 45, 125-232]. It is known that impulsive sounds change to non-impulsive as these sounds propagate. This paper shows that this is not relevant for assessing hearing injury because sounds retain impulsive character when SPLs are above EQT. Sounds from vessels are normally considered non-impulsive; however, 66% of vessels analyzed were impulsive when weighted for very-high frequency mammal hearing.

Foraging activity of sperm whales (Physeter macrocephalus) off the east coast of New Zealand

The occurrence and distribution of sperm whales in New Zealand waters is mainly known from whaling records or opportunistic sightings by the public and a systematic estimation of the abundance and distribution has never been conducted. In this study, we investigated the foraging activity and occurrence of sperm whales off the Eastern coast of New Zealand using passive acoustic monitoring techniques. Three acoustic recorders were moored to the ocean floor at different locations on the east side of the North and South Island to collect passive acoustic data from June 2016 until August 2017. A total of 53,823 echolocation click trains were recorded and analyzed to understand the spatial and temporal variation of sperm whale foraging activity. No difference in the foraging activity was found between night-time and day-time periods at any of the locations. Click train detections increased toward the south, suggesting increased foraging activity near Kaikoura. At each station, sperm whale foraging activity varied by month.

Click characteristics of northern bottlenose whales (Hyperoodon ampullatus) and Sowerby's beaked whales (Mesoplodon bidens) off eastern Canada

Passive acoustic monitoring (PAM) is crucial to expanding the knowledge of beaked whales, including the northern bottlenose whale (Hyperoodon ampullatus) and Sowerby's beaked whale (Mesoplodon bidens). Existing descriptions of clicks produced by these species are limited by sample size, number of individuals recorded, and geographic scope. Data from multiple encounters in the western North Atlantic are used to provide a quantitative description of clicks produced by these species. Recordings from nine encounters with northern bottlenose whales in Nova Scotia and Newfoundland were analyzed (N = 2239 clicks). The click type described had a median peak frequency of 25.9 kHz (10th-90th percentile range: 22.9-29.3 kHz), and a median inter-click interval (ICI) of 402 ms (N = 1917, 10th-90th percentile range: 290-524 ms). Recordings from 18 Sowerby's beaked whale encounters from Nova Scotia were analyzed (N = 762 clicks). The click type described had a median peak frequency of 65.8 kHz (10th-90th percentile range: 61.5-76.5 kHz), and a median ICI of 237 ms (N = 677, 10th-90th percentile range: 130-315 ms). These results will contribute to the development of methods to detect and classify beaked whale clicks to the species level, improving the effectiveness of PAM and enhancing scientific understanding and conservation efforts for cryptic and at-risk cetaceans.

Northern bottlenose whales in a pristine environment respond strongly to close and distant navy sonar signals

Impact assessments for sonar operations typically use received sound levels to predict behavioural disturbance in marine mammals. However, there are indications that cetaceans may learn to associate exposures from distant sound sources with lower perceived risk. To investigate the roles of source distance and received level in an area without frequent sonar activity, we conducted multi-scale controlled exposure experiments ( n = 3) with 12 northern bottlenose whales near Jan Mayen, Norway. Animals were tagged with high-resolution archival tags ( n = 1 per experiment) or medium-resolution satellite tags ( n = 9 in total) and subsequently exposed to sonar. We also deployed bottom-moored recorders to acoustically monitor for whales in the exposed area. Tagged whales initiated avoidance of the sound source over a wide range of distances (0.8-28 km), with responses characteristic of beaked whales. Both onset and intensity of response were better predicted by received sound pressure level (SPL) than by source distance. Avoidance threshold SPLs estimated for each whale ranged from 117-126 dB re 1 µPa, comparable to those of other tagged beaked whales. In this pristine underwater acoustic environment, we found no indication that the source distances tested in our experiments modulated the behavioural effects of sonar, as has been suggested for locations where whales are frequently exposed to sonar.