Whale Signals Research Articles

Machine learning with taxonomic family delimitation aids in the classification of ephemeral beaked whale events in passive acoustic monitoring.

Passive acoustic monitoring is an essential tool for studying beaked whale populations. This approach can monitor elusive and pelagic species, but the volume of data it generates has overwhelmed researchers' ability to quantify species occurrence for effective conservation and management efforts. Automation of data processing is crucial, and machine learning algorithms can rapidly identify species using their sounds. Beaked whale acoustic events, often infrequent and ephemeral, can be missed when co-occurring with signals of more abundant, and acoustically active species that dominate acoustic recordings. Prior efforts on large-scale classification of beaked whale signals with deep neural networks (DNNs) have approached the class as one of many classes, including other odontocete species and anthropogenic signals. That approach tends to miss ephemeral events in favor of more common and dominant classes. Here, we describe a DNN method for improved classification of beaked whale species using an extensive dataset from the western North Atlantic. We demonstrate that by training a DNN to focus on the taxonomic family of beaked whales, ephemeral events were correctly and efficiently identified to species, even with few echolocation clicks. By retrieving ephemeral events, this method can support improved estimation of beaked whale occurrence in regions of high odontocete acoustic activity.

Studies of the acoustic activity of white whales (in captive conditions and the possibility of their signals using to control behavior of fish in fishing process

The purpose of the work: identification of acoustic activity of belugas in conditions of aviary keeping and justification of the possibility of using their signals in fishing.The material of research was digital audio recordings of hydroacoustic signals of belugas collected from 2016 to 2018.Methods used: analysis of hydroacoustic data, visualization of the daily acoustic activity of belugas in different seasons, identification of stereotypes of acoustic behavior and characteristic signals in captive conditions and during fish hunting. Novelty: for the first time a substantiation for the possibility of using signals of one of the species of toothed whales — belugas registered in the captive conditions, to influence on the behavior of hydrobionts and solve practical problems of fishing and fish protection is given.Results: the highest acoustic activity of white whales is observed during the day and noticeably decreases at night with increase and decrease in the morning and evening hours. Maximum acoustic activity precedes the feeding of dolphins.The most common tones are those containing the fundamental frequency and harmonics. Characteristic features of the sounds are high variability of duration from 0.25 to 2.5 s, pronounced frequency components in the 500–2500 Hz spectrum zones, amplitude and frequency modulation, sound pressure levels of the signal up to 500 Pa /1 m.Acoustic activity of belugas during fish hunting and in the period preceding feeding in captive conditions, stereotypical. Typical are low-frequency frequency-modulated calls and whistles in the hearing range of fish, as well as a curtain of air bubbles and body and tail slaps. Possible ways of using beluga whale signals to increase the efficiency of fishing are proposed.Practical significance: the results of the study open up the possibility of using the signals of belugas to remotely control the movement of fish, create artificial concentrations and deterrent of hydrobionts in the intended areas.

Overlapping Whistle-and-Click Separation for Whale Signals

Overlapping Whistle-and-Click Separation for Whale Signals

Unsupervised beluga whistle end-point detection under burst pulse interferences

Passive acoustic monitoring is an increasingly important tool for whale research. Accurately detecting the end-point of whale whistle is an essential task in the study of marine mammal calls. However, the detection accuracy of the beluga whistle is reduced due to the non-stationary burst pulse interference in the marine environment. In this paper, an unsupervised two-stage beluga whistle end-point detection method is proposed to solve the above problem. Based on the high Q-factor wavelet decomposition, a rough whistle detection method is proposed to remove the silent time and the low intensity pulse. On this basis, local density adaptive spectral clustering is designed to further distinguish whistle and strong pulse interference based on the sparsity difference in time-frequency domain. The performance of the detector is tested with real signals of beluga whales, and its F 1-score is calculated. The result shows that the detector is obviously better than the traditional whistle detectors under the background of burst pulse interferences, and has higher robustness. In the future, the presented method is supposed to be applied for detecting whistles of some other whale species.

Adaptive echolocation behavior of bats and toothed whales in dynamic soundscapes.

Journal of Experimental Biology has a long history of reporting research discoveries on animal echolocation, the subject of this Centenary Review. Echolocating animals emit intense sound pulses and process echoes to localize objects in dynamic soundscapes. More than 1100 species of bats and 70 species of toothed whales rely on echolocation to operate in aerial and aquatic environments, respectively. The need to mitigate acoustic clutter and ambient noise is common to both aerial and aquatic echolocating animals, resulting in convergence of many echolocation features, such as directional sound emission and hearing, and decreased pulse intervals and sound intensity during target approach. The physics of sound transmission in air and underwater constrains the production, detection and localization of sonar signals, resulting in differences in response times to initiate prey interception by aerial and aquatic echolocating animals. Anti-predator behavioral responses of prey pursued by echolocating animals affect behavioral foraging strategies in air and underwater. For example, many insect prey can detect and react to bat echolocation sounds, whereas most fish and squid are unresponsive to toothed whale signals, but can instead sense water movements generated by an approaching predator. These differences have implications for how bats and toothed whales hunt using echolocation. Here, we consider the behaviors used by echolocating mammals to (1) track and intercept moving prey equipped with predator detectors, (2) interrogate dynamic sonar scenes and (3) exploit visual and passive acoustic stimuli. Similarities and differences in animal sonar behaviors underwater and in air point to open research questions that are ripe for exploration.

Source separation and arrival angle estimation using a single acoustic vector sensor

Extracting individual time signatures from two angularly-separated sources overlapping in time and bandwidth typically requires coherent array processing on multiple hydrophones, a process whose performance is degraded by limited aperture and spatial aliasing. Vector sensors, which measure both acoustic pressure and particle velocity, can measure an acoustic field’s directionality in a single, compact sensor package. Standard beamforming techniques can be applied to the outputs of a vector sensor to suppress one source’s signal, but its efficacy is limited by the angular insensitivity of the resulting cardioid beam pattern. Here we use fundamental physical principles to demonstrate that measurements on a 2-D vector sensor can completely determine the azimuths, amplitudes, and relative phases of two horizontal plane waves of the same frequency, achieving source separation and direction-of-arrival estimates for both. Interference between the waves creates a reactive intensity whose orientation is perpendicular to the bisector of the two wavenumber vectors. This observation leads to closed-form inversion formulas. Data examples of the technique are illustrated using overlapping humpback whale songs off Hawaii, and on weak airgun and bowhead whale signals embedded within intense directional noise from a nearby drilling site in the Beaufort Sea. [Work sponsored by ONR TFO.]

Tracking killer whale movements in the Alaskan Arctic relative to a loss of sea ice

Killer whales (Orcinus orca) are highly intelligent and versatile predators. Though they have a consistent presence in southern Alaskan waters, they are not known to frequently or consistently venture into the typically ice-covered U.S. Arctic, where they risk ice entrapment. However, with the trend of decreasing sea ice in the Arctic Ocean, these top predators may be venturing into waters that were previously inaccessible to them. To expand our knowledge of killer whale movements in the Arctic, eight years of data (2010–2018) from four AFSC passive acoustic moorings (funded by NOAA S&T, the U.S. Navy, and BOEM) located in the Chukchi, Beaufort, and northern Bering Seas, were manually analyzed for the presence/absence of killer whale signals to examine interannual trends in the acoustic presence of killer whales. Results from our data indicate that killer whales spend more time than previously recorded in the Arctic, as far as the edge of the Chukchi Borderland, directly following the decrease in sea ice in these areas. These results speak to a changing Arctic, both in terms of the presence of killer whales themselves and in terms of the impact that increased killer whale predation could have on Arctic food webs.

Vertical sonar beam width and scanning behavior of wild belugas (Delphinapterus leucas) in West Greenland.

Echolocation signals of wild beluga whales (Delphinapterus leucas) were recorded in 2013 using a vertical, linear 16-hydrophone array at two locations in the pack ice of Baffin Bay, West Greenland. Individual whales were localized for 4:42 minutes of 1:04 hours of recordings. Clicks centered on the recording equipment (i.e. on-axis clicks) were isolated to calculate sonar parameters. We report the first sonar beam estimate of in situ recordings of wild belugas with an average -3 dB asymmetrical vertical beam width of 5.4°, showing a wider ventral beam. This narrow beam width is consistent with estimates from captive belugas; however, our results indicate that beluga sonar beams may not be symmetrical and may differ in wild and captive contexts. The mean apparent source level for on-axis clicks was 212 dB pp re 1 μPa and whales were shown to vertically scan the array from 120 meters distance. Our findings support the hypothesis that highly directional sonar beams and high source levels are an evolutionary adaptation for Arctic odontocetes to reduce unwanted surface echoes from sea ice (i.e., acoustic clutter) and effectively navigate through leads in the pack ice (e.g., find breathing holes). These results provide the first baseline beluga sonar metrics from free-ranging animals using a hydrophone array and are important for acoustic programs throughout the Arctic, particularly for acoustic classification between belugas and narwhals (Monodon monoceros).

Bowhead and beluga whale acoustic detections in the western Beaufort Sea 2008-2018.

The Distributed Biological Observatory (DBO) was established to detect environmental changes in the Pacific Arctic by regular monitoring of biophysical responses in each of 8 DBO regions. Here we examine the occurrence of bowhead and beluga whale vocalizations in the western Beaufort Sea acquired by acoustic instruments deployed from September 2008-July 2014 and September 2016-October 2018 to examine inter-annual variability of these Arctic endemic species in DBO Region 6. Acoustic data were collected on an oceanographic mooring deployed in the Beaufort shelfbreak jet at ~71.4°N, 152.0°W. Spectrograms of acoustic data files were visually examined for the presence or absence of known signals of bowhead and beluga whales. Weekly averages of whale occurrence were compared with outputs of zooplankton, temperature and sea ice from the BIOMAS model to determine if any of these variables influenced whale occurrence. In addition, the dates of acoustic whale passage in the spring and fall were compared to annual sea ice melt-out and freeze-up dates to examine changes in phenology. Neither bowhead nor beluga whale migration times changed significantly in spring, but bowhead whales migrated significantly later in fall from 2008–2018. There were no clear relationships between bowhead whales and the environmental variables, suggesting that the DBO 6 region is a migratory corridor, but not a feeding hotspot, for this species. Surprisingly, beluga whale acoustic presence was related to zooplankton biomass near the mooring, but this is unlikely to be a direct relationship: there are likely interactions of environmental drivers that result in higher occurrence of both modeled zooplankton and belugas in the DBO 6 region. The environmental triggers that drive the migratory phenology of the two Arctic endemic cetacean species likely extend from Bering Sea transport of heat, nutrients and plankton through the Chukchi and into the Beaufort Sea.

Deep embedded clustering of coral reef bioacoustics.

Deep clustering was applied to unlabeled, automatically detected signals in a coral reef soundscape to distinguish fish pulse calls from segments of whale song. Deep embedded clustering (DEC) learned latent features and formed classification clusters using fixed-length power spectrograms of the signals. Handpicked spectral and temporal features were also extracted and clustered with Gaussian mixture models (GMM) and conventional clustering. DEC, GMM, and conventional clustering were tested on simulated datasets of fish pulse calls (fish) and whale song units (whale) with randomized bandwidth, duration, and SNR. Both GMM and DEC achieved high accuracy and identified clusters with fish, whale, and overlapping fish and whale signals. Conventional clustering methods had low accuracy in scenarios with unequal-sized clusters or overlapping signals. Fish and whale signals recorded near Hawaii in February-March 2020 were clustered with DEC, GMM, and conventional clustering. DEC features demonstrated the highest accuracy of 77.5% on a small, manually labeled dataset for classifying signals into fish and whale clusters.

Vertical sonar beam width of wild belugas (Delphinapterus leucas) in West Greenland

Echolocation signals of wild beluga whales (Delphinapterus leucas) were recorded in 2013 using a vertical, linear 16-hydrophone array at two locations in the pack ice of Baffin Bay, West Greenland. Individual whales were localized for 12 min of 1:04 h of recordings where on-axis clicks were isolated to calculate sonar parameters. We report the first sonar beam estimate of in situ recordings of belugas with an average 3-dB asymmetrical vertical beam width of 5.7 deg, showing a wider ventral axis. This is narrower than the commonly used estimates in the literature obtained from captive whales, suggesting beluga sonar beam width may be different in captive and wild contexts and is not necessarily symmetrical. Apparent source levels ranged from 210 to 220 dB pp re 1 μPa and whales were shown to vertically scan the array from 120 m distance. Our findings support the hypothesis that highly directional sonar beams and high source levels are an evolutionary adaptation for Arctic odontocetes to reduce noise and surface echoes from ice. These results provide the first baseline beluga sonar metrics from free-ranging animals using a hydrophone array and are important for acoustic programs throughout the Arctic, particularly for acoustic classification between belugas and narwhals (Monodon monoceros).

Solution of the Problem of Classification of Hydroacoustic Signals Based on Harmonious Wavelets and Machine Learning

Two types of real hydroacoustic signals of whales are classified based on the harmonic wavelet transform (HWT) coefficients (fast implementation), windowed Fourier transform (FT) (spectrogram), and conventional FT using the k-NN algorithm. The accuracy of the classification is estimated for various signal-to-noise ratios (SNRs). In order to reduce the dimension of the feature space during classification using the k-NN algorithm, the use of the modulo N reduction method is proposed. The efficiency of the use of harmonic wavelets in the classification of complex nonstationary signals is experimentally proved. The applicability of speech processing methods for the classification of underwater bioacoustic signals is confirmed. The discussed methods are initially developed taking into account the characteristics of human speech, but, nevertheless, showed good results even without being tuned to the characteristics of the classified signals. The problem of classifying two types of whales by the sounds they make using a neural network is solved.

Assessing the Possibilities of Identifying Bowhead Whales on the Arctic Shelf

Real recordings of bowhead whale signals and numerical simulation were used to investigate the propagation of such signals in an Arctic-type shallow-water waveguide. The authors analyzed attenuation and distortion of whale signals with increasing distance in the following conditions: (1) a soft and hard bottom, (2) different whale depths, (3) surface waves, (4) additive noise, and (5) reception by a single hydrophone and a vertical array. Signal attenuation was estimated by the change in sound pressure level (SPL). Signal distortion was estimated by the change in the correlation coefficient for the spectrograms of received signals and the reference spectrograms. It is shown that, compared to signals from a single receiver, the attenuation and distortion characteristics of the array output signals are more stable and resistant to changes in whale depth, to possible, under Arctic conditions, significant variations in sound speed in the bottom along the acoustic path, and to the effects of additive noise. The results can be used for passive monitoring of bowhead whales on the Arctic shelf.

ORCA-SPOT: An Automatic Killer Whale Sound Detection Toolkit Using Deep Learning

Large bioacoustic archives of wild animals are an important source to identify reappearing communication patterns, which can then be related to recurring behavioral patterns to advance the current understanding of intra-specific communication of non-human animals. A main challenge remains that most large-scale bioacoustic archives contain only a small percentage of animal vocalizations and a large amount of environmental noise, which makes it extremely difficult to manually retrieve sufficient vocalizations for further analysis – particularly important for species with advanced social systems and complex vocalizations. In this study deep neural networks were trained on 11,509 killer whale (Orcinus orca) signals and 34,848 noise segments. The resulting toolkit ORCA-SPOT was tested on a large-scale bioacoustic repository – the Orchive – comprising roughly 19,000 hours of killer whale underwater recordings. An automated segmentation of the entire Orchive recordings (about 2.2 years) took approximately 8 days. It achieved a time-based precision or positive-predictive-value (PPV) of 93.2% and an area-under-the-curve (AUC) of 0.9523. This approach enables an automated annotation procedure of large bioacoustics databases to extract killer whale sounds, which are essential for subsequent identification of significant communication patterns. The code will be publicly available in October 2019 to support the application of deep learning to bioaoucstic research. ORCA-SPOT can be adapted to other animal species.

Predator-prey interactions in the Southern California Bight

The Southern California Bight (SCB) is an important foraging area for a variety of cetacean species, including blue whales and beaked whales. However, the interactions of these species with their prey (krill and deep-water squid, respectively) can be very difficult to observe. Moorings combining passive and active acoustics were deployed at two locations in the SCB in two consecutive fall and winter seasons. A High-frequency Acoustic Recording Package sampling at 200 kHz recorded the full range of baleen and beaked whale signals. The active acoustic system on each mooring included two Simrad Wide Band Acoustic Transceivers (WBAT). A bottom-mounted WBAT had an upward looking 70 kHz transducer and the one at 300 m depth had upward looking 70 kHz and 200 kHz transducers. Over the course of four months, blue whale songs and social calls were recorded on both moorings. Beaked whale echolocation clicks were commonly detected on the mooring deployed offshore, with only rare beaked whale encounters on the coastal mooring. Backscatter data indicated overlap in the presence of cetacean predators and their prey. Scattering layer dynamics were dominated by regular diel migration, with differences in scatterer abundance between mooring sites, as well as on 2–5 day time scales.The Southern California Bight (SCB) is an important foraging area for a variety of cetacean species, including blue whales and beaked whales. However, the interactions of these species with their prey (krill and deep-water squid, respectively) can be very difficult to observe. Moorings combining passive and active acoustics were deployed at two locations in the SCB in two consecutive fall and winter seasons. A High-frequency Acoustic Recording Package sampling at 200 kHz recorded the full range of baleen and beaked whale signals. The active acoustic system on each mooring included two Simrad Wide Band Acoustic Transceivers (WBAT). A bottom-mounted WBAT had an upward looking 70 kHz transducer and the one at 300 m depth had upward looking 70 kHz and 200 kHz transducers. Over the course of four months, blue whale songs and social calls were recorded on both moorings. Beaked whale echolocation clicks were commonly detected on the mooring deployed offshore, with only rare beaked whale encounters on the coastal...

Littoral Acoustic Demonstration Center passive acoustic localization and tracking of marine mammals in the Northern Gulf of Mexico

The Littoral Acoustic Demonstration Center (LADC) consortium, founded in 2000, has collected underwater acoustic data in the northern Gulf of Mexico since 2001 using Environmental Acoustic Recording Systems (EARS) buoys. These hydrophone systems were developed by the U.S. Navy, and originally LADC could record underwater signals only up to 11.7 kHz. Since then the EARS have been upgraded several times and currently record up to 192 kHz, thus allowing recordings of not only sperm whale clicks and codas, but also signals of beaked whales and dolphins. One goal of LADC is to develop an acoustic method for identification of individual animals, with the goal of an “acoustic catalog” similar to the existing photographic catalog of individual whale flukes. Localization and tracking is an important component for acoustic identification. Some localization and tracking techniques and procedures developed and used by LADC will be presented and discussed.

Seasonal acoustic environments of beluga and bowhead whale core-use regions in the Pacific Arctic

The acoustic environment of two focal Arctic species, bowhead (Balaena mysticetus) and beluga (Delphinapterus leucas) whales, varied among the three core-use regions of the Pacific Arctic examined during the months in which both species occur: (1) January-March in the St. Lawrence Island/Anadyr Strait region, (2) November-January in the Bering Strait region, and (3) August-October in the Barrow Canyon region. Biological noise (consisting of the signals of bowhead whales, walrus and bearded seals) dominated the acoustic environment for the focal species in the St. Lawrence Island/Anadyr Strait region, which was covered with ice throughout the months studied. In the Bering Strait region whales were exposed primarily to environmental noise (in the form of wind noise) during November, before the region was ice-covered in December, and biological noise (from bowhead and walrus) again was prevalent. Anthropogenic noise dominated the Barrow Canyon region for the focal species in late summer and fall (August through October); this was also the only region in which the two species did not overlap with sea ice. Under open water conditions both near Barrow Canyon and in Bering Strait, noise levels were tightly correlated with wind. However, with climate-change driven increases in open water leading to rising noise levels across multiple fronts (atmospheric, biological, anthropogenic), the relatively pristine acoustic environment of Arctic cetaceans is changing rapidly. Characterizing the acoustic habitat of these regions before they are further altered should be considered a management and conservation priority in the Arctic.

Seasonal acoustic ecology of beluga and bowhead whale core-use areas in the Pacific Arctic

The acoustic ecology of Arctic marine mammals is driven by anthropogenic, biotic, and abiotic factors each of which may influence the behavioral ecology of each species. The acoustic environment of bowhead (Balaena mysticetus) and beluga (Delphinapterus leucas) whales in three core-use regions of the Pacific Arctic was examined during the months in which both species occur in these regions. The Anadyr Strait region in winter was dominated by the signals of bowhead whales, walrus and bearded seals. In Bering Strait in late fall and winter, wind noise predominated in November but once the region was ice-covered, bowhead and walrus were the main sources of noise. Barrow Canyon in late summer and fall was the only region in which anthropogenic sources overlapped with both whale species. Overall, ambient noise levels were low in the Pacific Arctic when compared to other ocean basins in which anthropogenic noise dominates low frequencies. However, climate change-driven increases in open water are leading to rising noise levels from increased human use of the Arctic, increased storminess, and increased presence of vocal subarctic whales. These “new” sources of sound may be altering the underwater soundscape and possibly influencing the acoustic ecology of Pacific Arctic cetaceans.

A comparison of whale tracking algorithms in the Indian Ocean using the Comprehensive Test Ban Treaty Hydrophone Data

This paper will present results comparing several geo-location and tracking algorithms applied to whale signals as monitored by the Comprehensive Test Ban Treaty (CTBT) sensor network. The performance of three tracking algorithms are compared using several hours of acoustic data collected off Cape Leeuwin, Australia, containing the apparent broadcast call of a blue whale (Balaenoptera musculus). The first approach used cross-correlation in time on the entire signal, while the second used a smaller time sample that covers the leading or trailing edge of the signal and creates a motion vector from the pair. The third approach used the Cross-Ambiguity-Function Mapping technique to generate 2-D energy maps on a geographic plot. Accuracy and energy detection algorithms based on blind-processing of the data are compared.

Waveguides and whales: A review about using acoustic waveguide theory for localizing whales and inverting for their propagation environment

Low-frequency acoustic propagation in shallow-water waveguides is conveniently modeled using sets of normal modes that propagate with different group velocities. Analytical solutions of simple environmental models, such as the Pekeris waveguide, were almost immediately applied to estimate ranges of impulsive sources by measuring the relative arrival times of the modal components, but to the author’s knowledge, it was not until the early 1990s that such techniques were first applied to low-frequency baleen whale calls, in a technical report by Gerald D’Spain. Since then waveguide-based methods have been used to track blue whales off Southern California, right whales in the Bering Sea, sei whales in the Atlantic, and bowhead whales in the Chukchi and Beaufort Seas. Some of these applications have exploited time-separated modal arrivals on a single hydrophone, while others have used vertical arrays to measure modal interference patterns, and subsequently estimate an animal’s range, depth, and local propagation environment. A recent promising development has been the introduction of nonlinear time sampling methods that permit mode isolation without the need for vertical arrays, enabling range estimation of certain whale signals from existing single-hydrophone data, even when the modal arrivals are not cleanly separated in time.