Whale Calls Research Articles

Song Notes and Patterns of the Mediterranean Fin Whale (Balaenoptera physalus) in the Ionian Sea

The Mediterranean fin whale emits two types of 20-Hz calls, known as “classic” and “backbeat”, that can be produced in irregular series or in patterned sequences called songs. The analysis of songs is recognized as a meaningful approach to study baleen whales and can be used to investigate populations’ identities. Mediterranean fin whale songs have been studied previously, but only in the western Mediterranean Sea. This work describes the structure of the songs recorded in the Ionian Sea. The inter-note intervals and the alternation of 20-Hz note types were considered to assess the occurrence of recurring patterns. Differences between patterned songs and irregular sequences were also investigated. Acoustic data were sampled continuously for about 10 months by the cabled observatory NEMO-SN1, deployed at 2100 m depth, 25 km offshore Catania; 28 call sequences were isolated and 10 of these were classified as either patterned songs or irregular sequences. Significant differences were observed in the spectral features of classic notes between songs and irregulars; four-note patterns were found repeatedly over different months, indicating a regular structure in detected songs. This work establishes a reference to interpret Mediterranean fin whale songs, and to assess the acoustic behavior of the population.

Distribution and seasonality of the Omura’s whale (Balaenoptera omurai) in Australia based on passive acoustic recordings

The Omura’s whale is one of the most recently described species of baleen whale. Initially known only from stranding and whaling specimens, it has now been identified in all ocean basins excluding the central and eastern Pacific. Unlike most baleen whales that migrate between the poles and the equator seasonally, the Omura’s whale is known to inhabit tropical to sub-tropical waters year-round. In Australian waters, there remain fewer than 30 confirmed visual sightings over the past decade. This study utilizes passive acoustic recordings from 41 locations around Australia from 2005 to 2023 to assess the distribution and seasonality of the Omura’s whale. The seasonal presence of Omura’s whale vocalizations varied by location, with higher presence at lower latitudes. Vocalizations were detected year-round in the Timor Sea and the Kimberley region. The most southerly occurrence of Omura’s whale vocalizations was recorded off the North West Cape, WA. In Australia, the vocalizations detected differ from those of other Indian Ocean Omura’s whale populations. Geographic variation of the vocalization was also observed within Australia, including vocalizations similar but not identical to those of the Omura’s whale detected in the Great Barrier Reef for the first time. The identified seasonal distribution and possible acoustic populations of the species in Australia provides valuable information to assess environmental and anthropogenic pressures on the Omura’s whale and to aid in creating management policies for the species.

A tale of whale and fiber: monitoring baleen whales with Distributed Acoustic Sensing (DAS)

Distributed Acoustic Sensing (DAS) converts fibers in existing underwater telecommunication fiber-optic cables into dense listening arrays of strain sensors. Recent advances in DAS interrogating technology enable increasing data quality, spatial coverage, and bandwidth, sparking interest in applied environmental sensing. Initially focused on seismic data collection, e.g., for earthquake monitoring, DAS has demonstrated abilities in detecting waterborne acoustic sources, including the low-frequency sounds of blue and fin whales (Balaenoptera musculus and B. physalus) over tens of km of fiber. DAS is instrumented from land, collects data in near-real time, and offers region-wide spatial coverage. Thus, this technology provides a unique opportunity for robust monitoring of endangered and threatened migratory baleen whales at ecologically meaningful spatial and temporal scales. This talk will introduce our research on using DAS to monitor low-frequency baleen whales, including detection and tracking, and the open-access DAS4Whales python package, developed to analyze terabytes of spatio-temporal data collected with DAS. We will also address the potential limitations of this new monitoring technology and provide a roadmap for further research and testing needed to exploit the technology's full potential.

Humpback whale: Yes/no? Challenges with passive acoustic monitoring of this well-described species

Humpback whales sing song and produce non-song sounds. Dozens of research articles have described their acoustic repertoire—even if only considering Australian waters. While song structure changes slowly over seasons, individual notes may exhibit multi-decadal stability. Irrespectively, automatically detecting humpback whale vocalizations remains an ongoing development activity while manual validation of automated detections continues to be problematic. While analysing archival recordings from Australian waters using the NOAA/Google convolutional neural network pretrained for humpback whale song [https://tfhub.dev/google/humpbackwhale/1], we found numerous instances of a particular type of sound with spectrographic features resembling those of humpback whale vocalizations. These sounds lacked the context of a song and occurred in chorus-like patterns, withthe same sound recorded at varying received levels and overlapping in time—indicative of many animals simultaneously producing nothing but this specific call. To investigate these phenomena further, we (1) built a custom detector for these specific calls, (2) ran the detector on archival recordings from North-western Australia, and (3) compared geographic and seasonal presences and absences to known patterns of humpback whale occurrence. The likelihood of these scenarios being humpback whales is discussed.

Does female choice for song complexity drive sexual selection in humpback whales?

Male humpback whales (Megaptera novaeangliae) sing a long, stereotyped, and culturally transmitted song display. While song likely functions in sexual selection, whether it is primarily directed at females and/or males is still debated. Most males within a single population sing the same, slowly evolving song type. However, multiple humpback whale song revolutions (where a song introduced from a neighboring population rapidly and completely replaces the existing song) have spread across the South Pacific region from the east coast of Australia to French Polynesia. How cultural evolution and sexual selection interact in this complex vocal display is not well understood. We investigated song-mediated female mate choice in humpback whales using 20 years of paternity data and song recordings from individually identified males in the small, South Pacific population of New Caledonia. As mating has never been observed in humpback whales, we assessed male reproductive success (calf paternity) and song complexity (unit repertoire and song length) of individual males. We found that males who successfully sired offspring had on average higher song complexity than males who did not sire offspring. This suggests females may select mates at least partially based on song complexity, and thus, song being primarily directed at females.

Building bridges: Doug Cato’s remarkable career spanning underwater acoustics and animal bioacoustics

Doug Cato’s research includes pioneering studies of the underwater acoustic environment surrounding Australia and numerous contributions to our understanding of marine mammal acoustic behavior. Moreover, his relaxed avuncular manner allowed him to assemble collaborations spanning disparate disciplines while mentoring the next generation of acoustics researchers in Australia. Throughout his career, Doug built bridges between underwater acoustics and animal bioacoustics, and between Australian and international researchers. This talk highlights two international collaborations applying signal processing techniques to humpback whale songs and leopard seal calling bouts. These projects estimated the information entropy and structure of the sequences of sounds in these marine mammal vocalizations. Comparing sliding window entropy estimates with Markov models supports Doug’s earlier observations that Australian humpback song follows the hierarchical structure proposed by Payne & McVay for North Atlantic humpbacks [Miksis-Olds et al., JASA, 2008]. The same comparison of entropy estimators for leopard seal calling bouts reveals that most of the sequential structure of in these animals’ calls is captured by Markov models.

Who’s knocking? Gray whale acoustic detection along California’s central coast

Passive acoustic monitoring (PAM) efforts are essential to marine mammal conservation. Bioacoustic techniques can be preferred as visual surveys are time intensive, costly, and limiting. California’s coastline is a biodiversity hotspot and a known migration corridor of multiple cetacean species, including gray whales, according to stock assessments provided by NOAA. Certain studies have indicated that gray whales vocalize during migration which can help confirm cetacean presence throughout the year. Off the coast of San Luis Obispo, California in an area that overlaps with the gray whale’s southern migration, our team detected an unidentified knocking noise. Although gray whale vocalizations can be difficult to confirm, we believe these knocks are produced by this transient species. In order to determine the source of these knocking sounds, the goals of this project are to (1) classify the repetitive noise as biotic or abiotic, (2) perform a comparative analysis of various species to narrow down possible biotic candidates, (3) utilize records from local whale watching companies for more accurate evidence of species presence at the time of our deployment, and (4) examine temporal variation in the occurrence of this knocking sound. Acoustic monitoring provides information about migration of gray whales necessary for conservation efforts.

Passive ocean acoustic waveguide remote sensing of vocalization behavior and spatial distribution of diverse marine mammal species in the Norwegian and Barents Sea

Leveraging data acquired using a 160-element coherent hydrophone array deployed in the Norwegian and Barents Seas during spring 2014, and the passive ocean acoustic waveguide remote sensing (POAWRS) technique is employed to enable instantaneous wide-area monitoring of marine mammal vocalizations over expanses exceeding 100 km in diameter. The vocalization behavior of diverse marine mammal species including Fin, Humpback, Minke, Sperm, and Beluga whales are analyzed, quantifying time-frequency characteristics and call patterns from their vocalization signals present in high-resolution beamformed power spectrograms. Previously developed automatic detection and machine learning algorithms are employed for clustering and classification of underwater acoustic events, facilitating the subsequent manual audio-visual inspection of whale calls. Bearing-time trajectories of repetitive species-specific vocalizations signals are utilized to estimate locations of the whale calls and hence their temporo-spatial distributions. Through comparative studies, we assess the presence and abundance of marine mammal species-specific vocalization signals in three distinct coastal regions off Norway, namely, Alesund, Lofoten, and Northern Finmark. In addition, we investigate correspondences of marine mammal distributions with concurrently imaged instantaneous wide-area fish distributions from distinct fish species to elucidate potential predator-prey interactions and habitat preferences.

Natural soundscape variation appears to influence the structure of stereotypic calls and repertoires in killer whales

Underwater soundscapes show dynamic variations in natural ambient sound levels at different locations and water depths and levels can vary over a course of a day. This soundscape dynamicity may pose a vocal challenge for highly mobile marine mammals whose habitat overlaps with many dynamic soundscapes. The most prominent communication tool of killer whales is a pulsated often multicomponent call which can be detected at more than 15 kilometers under quiet conditions. Social learning of call structure is a driver for stereotypic long lasting group- or population-specific call repertoires. Here we describe the potential vocal adaptations of killer whales to reduce the influence of sound propagation loss (PL) on their calls. Reliable propagation to identify groups and populations at considerable distances is important. In experimental field trials, the PL of killer whale calls and that of tones and sweeps was examined to determine if call component PL differs among soundscapes. PL is positively correlated with spatial and temporal ambient noise level variation. Furthermore, the use of burst pulses, a common feature of calls increased propagation distances at higher noise levels. We conclude that noise variation is one of the drivers of call structure and may also influence the call repertoire size.

Sub-unit classification of Pacific and Atlantic humpback high-complexity song units—Sub-unit diversity drives unit diversity

Humpback whale communication studies have primarily focused on holistic unit classification methods to identify thematic patterns of song structure. These various whole-unit, spectral-feature-extraction and classification procedures have resulted in low information entropy estimates of songs, akin to bird song. A recent study of high complexity, frequency-modulated song units shows that unit diversity and information entropy of humpback vocalizations is much higher and variable across regions of the Pacific and the Atlantic. Sub-unit-level feature-extraction and classification of these high-complexity units indicates that humpbacks have precision control of the generation of target frequencies, a characteristic of human vowel production. The number of distinct sub-units varies by geographic region and is comparable to English and Asian language phoneme sets.

Automatic detection and taxonomic identification of dolphin vocalisations using convolutional neural networks for passive acoustic monitoring

A novel framework for acoustic detection and species identification is proposed to aid passive acoustic monitoring studies on the endangered Indian Ocean humpback dolphin (Sousa plumbea) in South African waters. Convolutional Neural Networks (CNNs) were used for both detection and identification of dolphin vocalisations tasks, and performance was evaluated using custom and pre-trained architectures (transfer learning). In total, 723 min of acoustic data were annotated for the presence of whistles, burst pulses and echolocation clicks produced by Delphinus delphis (~45.6%), Tursiops aduncus (~39%), Sousa plumbea (~14.4%), Orcinus orca (~1%). The best performing models for detecting dolphin presence and species identification used segments (spectral windows) of two second lengths and were trained using images with 70 and 90 dpi, respectively. The best detection model was built using a customised architecture and achieved an accuracy of 84.4% for all dolphin vocalisations on the test set, and 89.5% for vocalisations with a high signal to noise ratio. The best identification model was also built using the customised architecture and correctly identified S. plumbea (96.9%), T. aduncus (100%), and D. delphis (78%) encounters in the testing dataset. The developed framework was designed based on the knowledge of complex dolphin sounds and it may assists in finding suitable CNN hyper-parameters for other species or populations. Our study contributes towards the development of an open-source tool to assist long-term studies of endangered species, living in highly diverse habitats, using passive acoustic monitoring.

The variable influence of anthropogenic noise on summer season coastal underwater soundscapes near a port and marine reserve

Monitoring soundscapes is essential for assessing environmental conditions for soniferous species, yet little is known about sound levels and contributors in Oregon coastal regions. From 2017 to 2021, during June–September, two hydrophones were deployed near Newport, Oregon to sample 10–13,000 Hz underwater sound. One hydrophone was deployed near the Port of Newport in a high vessel activity area, and another 17 km north within a protected Marine Reserve. Vessel noise and whale vocalizations were detected at both sites, but whales were recorded on more days at the Marine Reserve. Median sound levels in frequencies related to noise from various vessel types and sizes (50 – 4,000 Hz) were up to 6 dB higher at the Port of Newport, with greater diel variability compared to the Marine Reserve. In addition to documenting summer season conditions in Oregon waters, these results exemplify how underwater soundscapes can differ over short distances depending on anthropogenic activity.

New Marginal Spectrum Feature Information Views of Humpback Whale Vocalization Signals Using the EMD Analysis Methods.

Marginal spectrum (MS) feature information of humpback whale vocalization (HWV) signals is an interesting and significant research topic. Empirical mode decomposition (EMD) is a powerful time-frequency analysis tool for marine mammal vocalizations. In this paper, new MS feature innovation information of HWV signals was extracted using the EMD analysis method. Thirty-six HWV samples with a time duration of 17.2 ms were classified into Classes I, II, and III, which consisted of 15, 5, and 16 samples, respectively. The following ratios were evaluated: the average energy ratios of the 1 first intrinsic mode function (IMF1) and residual function (RF) to the referred total energy for the Class I samples; the average energy ratios of the IMF1, 2nd IMF (IMF2), and RF to the referred total energy for the Class II samples; the average energy ratios of the IMF1, 6th IMF (IMF6), and RF to the referred total energy for the Class III samples. These average energy ratios were all more than 10%. The average energy ratios of IMF1 to the referred total energy were 9.825%, 13.790%, 4.938%, 3.977%, and 3.32% in the 2980-3725, 3725-4470, 4470-5215, 10,430-11,175, and 11,175-11,920 Hz bands, respectively, in the Class I samples; 14.675% and 4.910% in the 745-1490 and 1490-2235 Hz bands, respectively, in the Class II samples; 12.0640%, 6.8850%, and 4.1040% in the 2980-3725, 3725-4470, and 11,175-11,920 Hz bands, respectively, in the Class III samples. The results of this study provide a better understanding, high resolution, and new innovative views on the information obtained from the MS features of the HWV signals.

Accommodating False Positives Within Acoustic Spatial Capture–Recapture, with Variable Source Levels, Noisy Bearings and an Inhomogeneous Spatial Density

Passive acoustic monitoring is a promising method for surveying wildlife populations that are easier to detect acoustically than visually. When animal vocalisations can be uniquely identified on an array of sensors, the potential exists to estimate population density through acoustic spatial capture–recapture (ASCR). However, sound classification is imperfect, and in some situations, a high proportion of sounds detected on just a single sensor (‘singletons’) are not from the target species. We present a case study of bowhead whale calls (Baleana mysticetus) collected in the Beaufort Sea in 2010 containing such false positives. We propose a novel extension of ASCR that is robust to false positives by truncating singletons and conditioning on calls being detected by at least two sensors. We allow for individual-level detection heterogeneity through modelling a variable sound source level, model inhomogeneous call spatial density, and include bearings with varying measurement error. We show via simulation that the method produces near-unbiased estimates when correctly specified. Ignoring source-level variation resulted in a strong negative bias, while ignoring inhomogeneous density resulted in severe positive bias. The case study analysis indicated a band of higher call density approximately 30 km from shore; 59.8% of singletons were estimated to have been false positives.

Variability in Temporal Characteristics of the South Eastern Indian Ocean Pygmy Blue Whale Song

The temporal characteristics of whale song are considered to be amongst the most stable elements of song production, though very few studies have looked at the timing of discrete sounds or the silences between them. This study indicates that the temporal characteristics of the song of the southeast Indian Ocean pygmy blue (SEIOPB) whale are highly variable. While the song repetition interval, or inter-song interval (ISI), appeared to remain consistent within a song event, defined as a consistent bout of singing presumed to come from an individual animal, it was found to be variable between song events. Song structure was a confounding factor in the analysis of variability in ISI with songs comprised of longer phrases naturally having a longer song repetition interval. The songs of the SEIOPB whale are characteristic of the species with only three primary song types. The three-unit song (units I, II and III) has the longest mean ISI of ~ 200 s, approximately twice that of songs composed of only two units (II and III) at a mean of ~ 100 s ISI which is approximately twice the ISI of a one-unit song (unit II only). The results of this study suggest that variability in ISI cannot be attributed to change in one element of a song but is driven by temporal changes in all elements of the song including phrase structure, unit duration and the length of the intervals between song units. Thus, the temporal characteristics of blue whale song appear to be under the control of the animal and may vary between individual whales, and within songs of an individual. It is concluded that the temporal characteristics of song potentially encode contextual information about the singer and may be influenced by audience effects, including audience composition and singer density.

DISCO: A deep learning ensemble for uncertainty-aware segmentation of acoustic signals.

Recordings of animal sounds enable a wide range of observational inquiries into animal communication, behavior, and diversity. Automated labeling of sound events in such recordings can improve both throughput and reproducibility of analysis. Here, we describe our software package for labeling elements in recordings of animal sounds, and demonstrate its utility on recordings of beetle courtships and whale songs. The software, DISCO, computes sensible confidence estimates and produces labels with high precision and accuracy. In addition to the core labeling software, it provides a simple tool for labeling training data, and a visual system for analysis of resulting labels. DISCO is open-source and easy to install, it works with standard file formats, and it presents a low barrier of entry to use.

ORCA-SPY enables killer whale sound source simulation, detection, classification and localization using an integrated deep learning-based segmentation

Acoustic identification of vocalizing individuals opens up new and deeper insights into animal communications, such as individual-/group-specific dialects, turn-taking events, and dialogs. However, establishing an association between an individual animal and its emitted signal is usually non-trivial, especially for animals underwater. Consequently, a collection of marine species-, array-, and position-specific ground truth localization data is extremely challenging, which strongly limits possibilities to evaluate localization methods beforehand or at all. This study presents ORCA-SPY, a fully-automated sound source simulation, classification and localization framework for passive killer whale (Orcinus orca) acoustic monitoring that is embedded into PAMGuard, a widely used bioacoustic software toolkit. ORCA-SPY enables array- and position-specific multichannel audio stream generation to simulate real-world ground truth killer whale localization data and provides a hybrid sound source identification approach integrating ANIMAL-SPOT, a state-of-the-art deep learning-based orca detection network, followed by downstream Time-Difference-Of-Arrival localization. ORCA-SPY was evaluated on simulated multichannel underwater audio streams including various killer whale vocalization events within a large-scale experimental setup benefiting from previous real-world fieldwork experience. Across all 58,320 embedded vocalizing killer whale events, subject to various hydrophone array geometries, call types, distances, and noise conditions responsible for a signal-to-noise ratio varying from -14.2 dB to 3 dB, a detection rate of 94.0 % was achieved with an average localization error of 7.01^circ. ORCA-SPY was field-tested on Lake Stechlin in Brandenburg Germany under laboratory conditions with a focus on localization. During the field test, 3889 localization events were observed with an average error of 29.19^circ and a median error of 17.54^circ. ORCA-SPY was deployed successfully during the DeepAL fieldwork 2022 expedition (DLFW22) in Northern British Columbia, with a mean average error of 20.01^circ and a median error of 11.01^circ across 503 localization events. ORCA-SPY is an open-source and publicly available software framework, which can be adapted to various recording conditions as well as animal species.

A theoretical account of whale song syntax: A new perspective for understanding human language structure

A theoretical account of whale song syntax: A new perspective for understanding human language structure

Adapting deep learning models to new acoustic environments - A case study on the North Atlantic right whale upcall

Passive acoustic monitoring is increasingly being used for studying marine mammals, leading to the accumulation of large acoustic datasets. Analyzing these datasets becomes impractical without automated detection and classification software. Detectors and classifiers based on deep neural networks have shown great potential, but their performance is often limited by the availability of sufficient quantities of annotated training samples, and their application restricted to the specific acoustic environment(s) from which their training data were collected. We address these limitations by employing transfer learning, a deep learning concept whereby knowledge from a source domain is transferred to a target domain. Specifically, we considered two different underwater acoustic environments as the source and target domains. The objective was to use a deep neural network that had been trained in one environment with abundant annotated training samples, and optimize its performance in the other environment where the annotated training samples were limited. Training and testing were conducted using three acoustic datasets containing North Atlantic right whale (Eubalaena glacialis) upcalls. Experiments show that adapting a trained model to the new environment led to a substantial improvement in recall from 70% to 85%, while maintaining a low false-positive rate of less than 5 per hour. The methodology is implemented as an open-source Python tool to facilitate the creation of more tailored deep learning-based acoustic detectors and classifiers for North Atlantic right whale vocalizations and other stereotyped marine mammal calls.

Exploring and verifying the acoustic presence of southern right whales (Eubalaena australis) off Elephant Island, Antarctica.

Passive acoustic monitoring (PAM) can be used to monitor acoustic presence and behaviour of cetaceans, providing continuous, long-term, and seasonally unbiased data. The efficiency of PAM methods, however, depends on the ability to detect and correctly interpret acoustic signals. The upcall is the most prevalent vocalization of the southern right whale (Eubalaena australis) and is commonly used as a basis for PAM studies on this species. However, previous studies report difficulties to distinguish between southern right whale upcalls and similar humpback whale (Megaptera novaeangliae) vocalizations with certainty. Recently, vocalizations comparable to southern right whale upcalls were detected off Elephant Island, Antarctica. In this study, these vocalizations were structurally analyzed, and call characteristics were compared to (a) confirmed southern right whale vocalizations recorded off Argentina and (b) confirmed humpback whale vocalizations recorded in the Atlantic Sector of the Southern Ocean. Based on call features, detected upcalls off Elephant Island could be successfully attributed to southern right whales. Measurements describing slope and bandwidth were identified as the main differences in call characteristics between species. With the newly gained knowledge from this study, additional data can be analyzed providing further insight into temporal occurrence and migratory behaviour of southern right whales in Antarctic waters.