Estimating cetacean population density using fixed passive acoustic sensors: an example with Blainville's beaked whales

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.

Click detection rate variability of central North Pacific sperm whales from passive acoustic towed arrays

Passive acoustic monitoring (PAM) is an optimal method for detecting and monitoring cetaceans as they frequently produce sound while underwater. Cue counting, counting acoustic cues of deep-diving cetaceans instead of animals, is an alternative method for density estimation, but requires an average cue production rate to convert cue density to animal density. Limited information about click rates exists for sperm whales in the central North Pacific Ocean. In the absence of acoustic tag data, we used towed hydrophone array data to calculate the first sperm whale click rates from this region and examined their variability based on click type, location, distance of whales from the array, and group size estimated by visual observers. Our findings show click type to be the most important variable, with groups that include codas yielding the highest click rates. We also found a positive relationship between group size and click detection rates that may be useful for acoustic predictions of group size in future studies. Echolocation clicks detected using PAM methods are often the only indicator of deep-diving cetacean presence. Understanding the factors affecting their click rates provides important information for acoustic density estimation.

Narwhal (Monodon monoceros) echolocation click rates to support cue counting passive acoustic density estimation

Estimating animal abundance is fundamental for effective management and conservation. It is increasingly done by combining passive acoustics with knowledge about rates at which animals produce cues (cue rates). Narwhals (Monodon monoceros) are elusive marine mammals for which passive acoustic density estimation might be plausible, but for which cue rates are lacking. Clicking rates in narwhals were investigated using a dataset from sound and movement tag records collected in August 2013-2016 and 2019 in East Greenland. Clicking rates were quantified for ∼1200 one-second-long systematic random samples from 8 different whales. Generalized additive models were used to model (1) the probability of being in a clicking state versus depth and (2) the clicking rate while in a clicking state, versus time and depth. The probability of being in a clicking state increased with depth, reaching ∼1.0 at ∼500 m, while the number of clicks per second (while in a clicking state) increased with depth. The mean cue production rate, weighted by tag duration, was 1.28 clicks per second (se = 0.13, CV = 0.10). This first cue rate for narwhals may be used for cue counting density estimation, but care should be taken if applying it to other geographical areas or seasons, given sample size, geographical, and temporal limitations.

Implications of porpoise echolocation and dive behaviour on passive acoustic monitoring

Harbour porpoises are visually inconspicuous but highly soniferous echolocating marine predators that are regularly studied using passive acoustic monitoring (PAM). PAM can provide quality data on animal abundance, human impact, habitat use, and behaviour. The probability of detecting porpoise clicks within a given area (P̂) is a key metric when interpreting PAM data. Estimates of P̂ can be used to determine the number of clicks per porpoise encounter that may have been missed on a PAM device, which, in turn, allows for the calculation of abundance and ideally non-biased comparison of acoustic data between habitats and time periods. However, P̂ is influenced by several factors, including the behaviour of the vocalising animal. Here, the common implicit assumption that changes in animal behaviour have a negligible effect on P̂ between different monitoring stations or across time is tested. Using a simulation-based approach informed by acoustic biologging data from 22 tagged harbour porpoises, it is demonstrated that porpoise behavioural states can have significant (up to 3× difference) effects on P̂. Consequently, the behavioural state of the animals must be considered in analysis of animal abundance to avoid substantial over- or underestimation of the true abundance, habitat use, or effects of human disturbance.

A sperm whale cautionary tale about estimating acoustic cue rates for deep divers

Passive acoustic density estimation has been gaining traction in recent years. Cue counting uses detected acoustic cues to estimate animal abundance. A cue rate, the number of acoustic cues produced per animal per unit time, is required to convert cue density into animal density. Cue rate information can be obtained from animal borne acoustic tags. For deep divers, like beaked whales, data have been analyzed considering deep dive cycles as a natural sampling unit, based on either weighted averages or generalized estimating equations. Using a sperm whale DTAG (sound-and-orientation recording tag) example we compare different approaches of estimating cue rate from acoustic tags illustrating that both approaches used before might introduce biases and suggest that the natural unit of analysis should be the whole duration of the tag itself.

Using individual-based bioenergetic models to predict the aggregate effects of disturbance on populations: A case study with beaked whales and Navy sonar

Anthropogenic activities can lead to changes in animal behavior. Predicting population consequences of these behavioral changes requires integrating short-term individual responses into models that forecast population dynamics across multiple generations. This is especially challenging for long-lived animals, because of the different time scales involved. Beaked whales are a group of deep-diving odontocete whales that respond behaviorally when exposed to military mid-frequency active sonar (MFAS), but the effect of these nonlethal responses on beaked whale populations is unknown. Population consequences of aggregate exposure to MFAS was assessed for two beaked whale populations that are regularly present on U.S. Navy training ranges where MFAS is frequently used. Our approach integrates a wide range of data sources, including telemetry data, information on spatial variation in habitat quality, passive acoustic data on the temporal pattern of sonar use and its relationship to beaked whale foraging activity, into an individual-based model with a dynamic bioenergetic module that governs individual life history. The predicted effect of disturbance from MFAS on population abundance ranged between population extinction to a slight increase in population abundance. These effects were driven by the interaction between the temporal pattern of MFAS use, baseline movement patterns, the spatial distribution of prey, the nature of beaked whale behavioral response to MFAS and the top-down impact of whale foraging on prey abundance. Based on these findings, we provide recommendations for monitoring of marine mammal populations and highlight key uncertainties to help guide future directions for assessing population impacts of nonlethal disturbance for these and other long-lived animals.

Wildlife Population Assessment: Changing Priorities Driven by Technological Advances

Advances in technology are having a large effect on the priorities for innovation in statistical ecology. Collaborations between statisticians and ecologists have always been important in driving methodological development, but increasingly, expertise from computer scientists and engineers is also needed. We discuss changes that are occurring and that may occur in the future in surveys for estimating animal abundance. As technology advances, we expect classical distance sampling and capture-recapture to decrease in importance, as camera (still and video) survey, acoustic survey, spatial capture-recapture and genetic methods continue to develop and find new applications. We explore how these changes are impacting the work of the statistical ecologist.

Comparison of visual and passive acoustic estimates of beaked whale density off El Hierro, Canary Islands

Passive acoustic monitoring (PAM) offers considerable potential for density estimation of cryptic cetaceans, such as beaked whales. However, comparative studies on the accuracy of PAM density estimates from these species are lacking. Concurrent, low-cost drifting PAM, with SoundTraps suspended at 200 m depth, and land-based sightings, were conducted off the Canary Islands. Beaked whale density was estimated using a cue-count method, with click production rate and the probability of click detection derived from digital acoustic recording tags (DTags), and distance sampling techniques, adapted to fixed-point visual surveys. Of 32 870 detections obtained throughout 206 h of PAM recordings, 68% were classified as "certain" beaked whale clicks. Acoustic detection probability was 0.15 [coefficient variation (CV) 0.24] and click production rate was 0.46 clicks s - 1 (CV 0.05). PAM density estimates were in the range of 21.5 or 48.6 whales per 1000 km2 [CV 0.50 or 0.44, 95% confidence interval (CI) 20.7-22.4 or 47-50.9), depending on whether "uncertain" clicks were considered. Density estimates from concurrent sightings resulted in 33.7 whales per 1000 km2 (CV 0.77, 95% CI 8.9-50.5). Cue-count PAM methods under application provide reliable estimates of beaked whale density, over relatively long time periods and in realistic scenarios, as these match the concurrent density estimates obtained from visual observations.

Automated extraction of baleen whale calls based on the pseudo-Wigner-Ville distribution

Baleen whales produce a wide variety of frequency-modulated calls. Extraction of the time-frequency (TF) structures of these calls forms the basis for many applications, including abundance estimation and species recognition. Typical methods to extract the contours of whale calls from a spectrogram are based on the short-time Fourier transform and are, thus, restricted by a fixed TF resolution. Considering the low-frequency nature of baleen whale calls, this work represents the contours using a pseudo-Wigner-Ville distribution for a higher TF resolution at the cost of introducing cross terms. An adaptive threshold is proposed followed by a modified Gaussian mixture probability hypothesis density filter to extract the contours. Finally, the artificial contours, which are caused by the cross terms, can be removed in post-processing. Simulations were conducted to explore how the signal-to-noise ratio influences the performance of the proposed method. Then, in experiments based on real data, the contours of the calls of three kinds of baleen whales were extracted in a highly accurate manner (with mean deviations of 5.4 and 0.051 Hz from the ground-truth contours at sampling rates of 4000 and 100 Hz, respectively) with a recall of 75% and a precision of 78.5%.

Acoustic identification and classification of four dolphin species in the Brazilian marine area affected by the largest tailings dam failure disaster

Passive acoustic monitoring (PAM) is an increasingly used technique to access the occurrence, distribution, and abundance of cetaceans that may be visually unavailable most of the time. The largest tailings dam failure disaster occurred on 5 November 2015, when the Fundão dam collapsed, releasing over 50 million cubic meters of tailings into the Doce River basin; 14 days later, the tailings plume reached the Atlantic Ocean. PAM was implemented in the concerned area and cetacean species were acoustically identified. Whistles and clicks of visual and acoustic matches were used to predict and classify exclusive acoustic records through random forest models. The identified species were Guiana, rough-toothed, and bottlenose dolphins. Additionally, the franciscana, the most threatened cetacean in the western South Atlantic Ocean, was also acoustically identified. The whistle classifier had 86.9% accuracy with final frequency, duration, and maximum frequency ranked as the most important parameters. The clicks classifier had 86.7% accuracy with peak frequency and 3 dB bandwidth as the most important parameters for classifying species. Considering the potential effect of the increase in turbidity on sound transmission, such as attenuation, the presented classifier should be continuously improved with novel data collected from long-term acoustic monitoring.

Discovering marine biodiversity in the 21st century

We review the current knowledge of the biodiversity of the ocean as well as the levels of decline and threat for species and habitats. The lack of understanding of the distribution of life in the ocean is identified as a significant barrier to restoring its biodiversity and health. We explore why the science of taxonomy has failed to deliver knowledge of what species are present in the ocean, how they are distributed and how they are responding to global and regional to local anthropogenic pressures. This failure prevents nations from meeting their international commitments to conserve marine biodiversity with the results that investment in taxonomy has declined in many countries. We explore a range of new technologies and approaches for discovery of marine species and their detection and monitoring. These include: imaging methods, molecular approaches, active and passive acoustics, the use of interconnected databases and citizen science. Whilst no one method is suitable for discovering or detecting all groups of organisms many are complementary and have been combined to give a more complete picture of biodiversity in marine ecosystems. We conclude that integrated approaches represent the best way forwards for accelerating species discovery, description and biodiversity assessment. Examples of integrated taxonomic approaches are identified from terrestrial ecosystems. Such integrated taxonomic approaches require the adoption of cybertaxonomy approaches and will be boosted by new autonomous sampling platforms and development of machine-speed exchange of digital information between databases.

Detection probability and density estimation of fin whales by a Seaglider

A single-hydrophone ocean glider was deployed within a cabled hydrophone array to demonstrate a framework for estimating population density of fin whales (Balaenoptera physalus) from a passive acoustic glider. The array was used to estimate tracks of acoustically active whales. These tracks became detection trials to model the detection function for glider-recorded 360-s windows containing fin whale 20-Hz pulses using a generalized additive model. Detection probability was dependent on both horizontal distance and low-frequency glider flow noise. At the median 40-Hz spectral level of 97 dB re 1 μPa²/Hz, detection probability was near one at horizontal distance zero with an effective detection radius of 17.1 km [coefficient of variation (CV) = 0.13]. Using estimates of acoustic availability and acoustically active group size from tagged and tracked fin whales, respectively, density of fin whales was estimated as 1.8 whales per 1000 km² (CV = 0.55). A plot sampling density estimate for the same area and time, estimated from array data alone, was 1.3 whales per 1000 km² (CV = 0.51). While the presented density estimates are from a small demonstration experiment and should be used with caution, the framework presented here advances our understanding of the potential use of gliders for cetacean density estimation.

Convergence of emerging technologies: Development of a risk-based paradigm for marine mammal monitoring for offshore wind energy operations

The ability to gather real-time and near real-time data on marine mammal distribution, movement, and habitat use has advanced significantly over the past two decades. These advances have outpaced their adoption into a meaningful, risk-based assessment framework so critically needed to support society's growing demands for a transition to increased reliance on renewable energy. Marine acoustics have the capacity to detect, identify, and locate vocalizations over broad areas. Photogrammetric and image processing increases the ability to visually detect animals from surface or aerial platforms. Ecological models based on long-term observational data coupled with static and remotely sensed oceanographic data are able to predict daily and seasonal habitat suitability. Extensive monitoring around anthropogenic activities, combined with controlled experiments of exposure parameters (i.e., sound), supports better informed decisions on reducing effects. Population models and potential consequence modeling provide the ability to estimate the significance of individual and population exposure. The collective capacities of these emerging technical approaches support a risk ranking and risk management approach to monitoring and mitigating effects on marine mammals related to development activities. The monitoring paradigm related to many offshore energy-related activities, however, has long been spatially limited, situationally myopic, and operationally uncertain. A case evaluation process is used to define and demonstrate the changing paradigm of effective monitoring aimed at protecting living resources and concurrently providing increased certainty that essential activities can proceed efficiently. Recent advances in both technologies and operational approaches are examined to delineate a risk-based paradigm, driven by a diversity of regional data inputs, that is capable of meeting the imperative for timely development of offshore wind energy. Integr Environ Assess Manag 2022;18:939-949. © 2021 The Authors. Integrated Environmental Assessment and Management published by Wiley Periodicals LLC on behalf of Society of Environmental Toxicology & Chemistry (SETAC).

The influence of sea ice on the detection of bowhead whale calls

Bowhead whales (Balaena mysticetus) face threats from diminishing sea ice and increasing anthropogenic activities in the Arctic. Passive acoustic monitoring is the most effective means for monitoring their distribution and population trends, based on the detection of their calls. Passive acoustic monitoring, however, is influenced by the sound propagation environment and ambient noise levels, which impact call detection probability. Modeling and simulations were used to estimate detection probability for bowhead whale frequency-modulated calls in the 80–180 Hz frequency band with and without sea ice cover and under various noise conditions. Sound transmission loss for bowhead calls is substantially greater during ice-covered conditions than during open-water conditions, making call detection ~ 3 times more likely in open-water. Estimates of daily acoustic detection probability were used to compensate acoustic detections for sound propagation and noise effects in two recording datasets in the northeast Chukchi Sea, on the outer shelf and continental slope, collected between 2012 and 2013. The compensated acoustic density suggests a decrease in whale presence with the retreat of sea ice at these recording sites. These results highlight the importance of accounting for effects of the environment on ambient noise and acoustic propagation when interpreting results of passive acoustic monitoring.

Grouper source levels and aggregation dynamics inferred from passive acoustic localization at a multispecies spawning site

Four species of grouper (family Epinephlidae), Red Hind (Epinephelus guttatus), Nassau (Epinephelus striatus), Black (Mycteroperca bonaci), and Yellowfin Grouper (Mycteroperca venenosa) share an aggregation site in Little Cayman, Cayman Islands and produce sounds while aggregating. Continuous observation of these aggregations is challenging because traditional diver or ship-based methods are limited in time and space. Passive acoustic localization can overcome this challenge for sound-producing species, allowing observations over long durations and at fine spatial scales. A hydrophone array was deployed in February 2017 over a 9-day period that included Nassau Grouper spawning. Passive acoustic localization was used to find positions of the grouper-produced calls recorded during this time, which enabled the measurement of call source levels and evaluation of spatiotemporal aspects of calling. Yellowfin Grouper had the lowest mean peak-to-peak (PP) call source level, and Nassau Grouper had the highest mean PP call source level (143.7 and 155.2 dB re: 1 μPa at 1 m for 70-170 Hz, respectively). During the days that Nassau Grouper spawned, calling peaked after sunset. Similarly, when Red Hind calls were abundant, calls were highest in the afternoon and evening. The measured source levels can be used to estimate communication and detection ranges and implement passive acoustic density estimation for these fishes.

Estimating Florida manatee (Trichechus manatus latirostris) abundance using passive acoustic methods

Manatees are difficult to detect, particularly cryptic populations that inhabit areas with limited water clarity. The effectiveness of using vocal detections to estimate manatee abundance was evaluated in a clear water spring where manatees congregate seasonally. Vocalizations were extracted by a detection classifier that clustered sounds with similar spectral properties. Vocalization counts from recordings in Blue Spring, FL, USA were strong predictors of manatee abundance. The link between independent visual counts and abundance estimates from passive acoustic monitoring was used to provide an estimate of 1.059 (95% confidence interval 0.963-1.127) vocalizations/manatee/5-min, which might be used elsewhere for cue counting of manatees.

Analytical approaches for evaluating passive acoustic monitoring data: A case study of avian vocalizations

The interface between field biology and technology is energizing the collection of vast quantities of environmental data. Passive acoustic monitoring, the use of unattended recording devices to capture environmental sound, is an example where technological advances have facilitated an influx of data that routinely exceeds the capacity for analysis. Computational advances, particularly the integration of machine learning approaches, will support data extraction efforts. However, the analysis and interpretation of these data will require parallel growth in conceptual and technical approaches for data analysis. Here, we use a large hand‐annotated dataset to showcase analysis approaches that will become increasingly useful as datasets grow and data extraction can be partially automated.

We propose and demonstrate seven technical approaches for analyzing bioacoustic data. These include the following: (1) generating species lists and descriptions of vocal variation, (2) assessing how abiotic factors (e.g., rain and wind) impact vocalization rates, (3) testing for differences in community vocalization activity across sites and habitat types, (4) quantifying the phenology of vocal activity, (5) testing for spatiotemporal correlations in vocalizations within species, (6) among species, and (7) using rarefaction analysis to quantify diversity and optimize bioacoustic sampling.

To demonstrate these approaches, we sampled in 2016 and 2018 and used hand annotations of 129,866 bird vocalizations from two forests in New Hampshire, USA, including sites in the Hubbard Brook Experiment Forest where bioacoustic data could be integrated with more than 50 years of observer‐based avian studies. Acoustic monitoring revealed differences in community patterns in vocalization activity between forests of different ages, as well as between nearby similar watersheds. Of numerous environmental variables that were evaluated, background noise was most clearly related to vocalization rates. The songbird community included one cluster of species where vocalization rates declined as ambient noise increased and another cluster where vocalization rates declined over the nesting season. In some common species, the number of vocalizations produced per day was correlated at scales of up to 15 km. Rarefaction analyses showed that adding sampling sites increased species detections more than adding sampling days.

Although our analyses used hand‐annotated data, the methods will extend readily to large‐scale automated detection of vocalization events. Such data are likely to become increasingly available as autonomous recording units become more advanced, affordable, and power efficient. Passive acoustic monitoring with human or automated identification at the species level offers growing potential to complement observer‐based studies of avian ecology.

Acoustic detection range of right whale upcalls identified in near-real time from a moored buoy and a Slocum glider

The goal of this study was to characterize the detection range of a near real-time baleen whale detection system, the digital acoustic monitoring instrument/low-frequency detection and classification system (DMON/LFDCS), equipped on a Slocum glider and a moored buoy. As a reference, a hydrophone array was deployed alongside the glider and buoy at a shallow-water site southwest of Martha's Vineyard (Massachusetts, USA) over a four-week period in spring 2017. A call-by-call comparison between North Atlantic right whale upcalls localized with the array (n = 541) and those detected by the glider or buoy was used to estimate the detection function for each DMON/LFDCS platform. The probability of detection was influenced by range, ambient noise level, platform depth, detection process, review protocol, and calling rate. The conservative analysis of near real-time pitch tracks suggested that, under typical conditions, a 0.33 probability of detection of a single call occurred at 6.2 km for the buoy and 8.6-13.4 km for the glider (depending on glider depth), while a 0.10 probability of detection of a single call occurred at 14.4 m for the buoy and 22.6-27.5 km for the glider. Probability of detection is predicted to increase substantially at all ranges if more than one call is available for detection.

A machine learning pipeline for classification of cetacean echolocation clicks in large underwater acoustic datasets

Machine learning algorithms, including recent advances in deep learning, are promising for tools for detection and classification of broadband high frequency signals in passive acoustic recordings. However, these methods are generally data-hungry and progress has been limited by challenges related to the lack of labeled datasets adequate for training and testing. Large quantities of known and as yet unidentified broadband signal types mingle in marine recordings, with variability introduced by acoustic propagation, source depths and orientations, and interacting signals. Manual classification of these datasets is unmanageable without an in-depth knowledge of the acoustic context of each recording location. A signal classification pipeline is presented which combines unsupervised and supervised learning phases with opportunities for expert oversight to label signals of interest. The method is illustrated with a case study using unsupervised clustering to identify five toothed whale echolocation click types and two anthropogenic signal categories. These categories are used to train a deep network to classify detected signals in either averaged time bins or as individual detections, in two independent datasets. Bin-level classification achieved higher overall precision (>99%) than click-level classification. However, click-level classification had the advantage of providing a label for every signal, and achieved higher overall recall, with overall precision from 92 to 94%. The results suggest that unsupervised learning is a viable solution for efficiently generating the large, representative training sets needed for applications of deep learning in passive acoustics.

Estimating minke whale relative abundance in the North Atlantic using passive acoustic sensors

Estimates of abundance and their changes through time are key elements of marine mammal conservation and management. Absolute marine mammal abundance in a region of the open ocean is often difficult to attain. However, methods of estimating their abundance based on passive acoustic recordings are becoming increasingly employed. This study shows that passive acoustic monitoring of North Atlantic minke whales with a single hydrophone provides sufficient information to estimate relative population abundance. An automated detector was developed for minke whale pulse trains and an approach for converting its output into a relative abundance index is proposed by accounting for detectability as well as false positives and negatives. To demonstrate this technique, a 2 y dataset from the seven sites of the Atlantic Deepwater Ecosystem Observatory Network project on the U.S. east coast was analyzed. Resulting relative abundance indices confirm pulse train-calling minke whale presence in the deep waters of the outer continental shelf. The minkes are present December through April annually with the highest abundance near the site offshore of Savannah, Georgia.

Noninvasive Technologies for Primate Conservation in the 21st Century

Observing and quantifying primate behavior in the wild is challenging. Human presence affects primate behavior and habituation of new, especially terrestrial, individuals is a time-intensive process that carries with it ethical and health concerns, especially during the recent pandemic when primates are at even greater risk than usual. As a result, wildlife researchers, including primatologists, have increasingly turned to new technologies to answer questions and provide important data related to primate conservation. Tools and methods should be chosen carefully to maximize and improve the data that will be used to answer the research questions. We review here the role of four indirect methods—camera traps, acoustic monitoring, drones, and portable field labs—and improvements in machine learning that offer rapid, reliable means of combing through large datasets that these methods generate. We describe key applications and limitations of each tool in primate conservation, and where we anticipate primate conservation technology moving forward in the coming years.

Model-based localization of deep-diving cetaceans using towed line array acoustic data

Passive acoustic monitoring using a towed line array of hydrophones is a standard method for localizing cetaceans during line-transect cetacean abundance surveys. Perpendicular distances estimated between localized whales and the trackline are essential for abundance estimation using acoustic data. Uncertainties in the acoustic data from hydrophone movement, sound propagation effects, errors in the time of arrival differences, and whale depth are not accounted for by most two-dimensional localization methods. Consequently, location and distance estimates for deep-diving cetaceans may be biased, creating uncertainty in abundance estimates. Here, a model-based localization approach is applied to towed line array acoustic data that incorporates sound propagation effects, accounts for sources of error, and localizes in three dimensions. The whale's true distance, ship trajectory, and whale movement greatly affected localization results in simulations. The localization method was applied to real acoustic data from two separate sperm whales, resulting in three-dimensional distance and depth estimates with position bounds for each whale. By incorporating sources of error, this three-dimensional model-based approach provides a method to address and integrate the inherent uncertainties in towed array acoustic data for more robust localization.

Passive acoustic monitoring for detecting the Yellow-bellied Glider, a highly vocal arboreal marsupial

Passive acoustic monitoring (PAM) is increasingly being used for the survey of vocalising wildlife species that are otherwise cryptic and difficult to survey. Our study aimed to develop PAM guidelines for detecting the Yellow-bellied Glider, a highly vocal arboreal marsupial that occurs in native Eucalyptus forests in eastern and south-eastern Australia. To achieve this, we considered the influence of background noise, weather conditions, lunar illumination, time since sunset and season on the probability of detecting vocalisations. We deployed Autonomous Recording Units (ARUs) at 43 sites in the Central Highlands of Victoria during two periods: spring/summer (October 2018 to January 2019), and autumn/winter (May to August 2019). ARUs were programmed to record for 11 hours from sunset for 14 consecutive days during each period. Background noise resulted from inclement weather (wind and rain) and masked vocalisations in spectrograms of the recordings, thus having the greatest influence on detection probability. Vocalisations were most common in the four hours after sunset. Rainfall negatively influenced detection probability, especially during the autumn/winter sampling period. Detection of Yellow-bellied Gliders with PAM requires deploying ARUs programmed to record for four hours after sunset, for a minimum of six nights with minimal inclement weather (light or no wind or rain). The survey period should be extended to 12 nights when rain or wind are forecast. Because PAM is less labour intensive than active surveys (i.e., spotlighting and call playbacks with multiple observers and several nights’ survey per site), its use will facilitate broad-scale surveys for Yellow-bellied Gliders.

Using context to train time-domain echolocation click detectors

This work demonstrates the effectiveness of using humans in the loop processes for constructing large training sets for machine learning tasks. A corpus of over 57 000 toothed whale echolocation clicks was developed by using a permissive energy-based echolocation detector followed by a machine-assisted quality control process that exploits contextual cues. Subsets of these data were used to train feed forward neural networks that detected over 850 000 echolocation clicks that were validated using the same quality control process. It is shown that this network architecture performs well in a variety of contexts and is evaluated against a withheld data set that was collected nearly five years apart from the development data at a location over 600 km distant. The system was capable of finding echolocation bouts that were missed by human analysts, and the patterns of error in the classifier consist primarily of anthropogenic sources that were not included as counter-training examples. In the absence of such events, typical false positive rates are under ten events per hour even at low thresholds.

Decadal Assessment of Sperm Whale Site-Specific Abundance Trends in the Northern Gulf of Mexico Using Passive Acoustic Data

Passive acoustic monitoring has been successfully used to study deep-diving marine mammal populations. To assess regional population trends of sperm whales in the northern Gulf of Mexico (GoM), including impacts of the Deepwater Horizon platform oil spill in 2010, the Littoral Acoustic Demonstration Center-Gulf Ecological Monitoring and Modeling (LADC-GEMM) consortium collected broadband acoustic data in the Mississippi Valley/Canyon area between 2007 and 2017 using bottom-anchored moorings. These data allow the inference of short-term and long-term variations in site-specific abundances of sperm whales derived from their acoustic activity. A comparison is made between the abundances of sperm whales at specific sites in different years before and after the oil spill by estimating the regional abundance density. The results show that sperm whales were present in the region throughout the entire monitoring period. A habitat preference shift was observed for sperm whales after the 2010 oil spill with higher activities at sites farther away from the spill site. A comparison of the 2007 and 2015 results shows that the overall regional abundance of sperm whales did not recover to pre-spill levels. The results indicate that long-term spatially distributed acoustic monitoring is critical in characterizing sperm whale population changes and in understanding how environmental stressors impact regional abundances and the habitat use of sperm whales.

Cetacean distribution models based on visual and passive acoustic data

Distribution models are needed to understand spatiotemporal patterns in cetacean occurrence and to mitigate anthropogenic impacts. Shipboard line-transect visual surveys are the standard method for estimating abundance and describing the distributions of cetacean populations. Ship-board surveys provide high spatial resolution but lack temporal resolution and seasonal coverage. Stationary passive acoustic monitoring (PAM) employs acoustic sensors to sample point locations nearly continuously, providing high temporal resolution in local habitats across days, seasons and years. To evaluate whether cross-platform data synthesis can improve distribution predictions, models were developed for Cuvier’s beaked whales, sperm whales, and Risso’s dolphins in the oceanic Gulf of Mexico using two different methods: generalized additive models and neural networks. Neural networks were able to learn unspecified interactions between drivers. Models that incorporated PAM datasets out-performed models trained on visual data alone, and joint models performed best in two out of three cases. The modeling results suggest that, when taken together, multiple species distribution models using a variety of data types may support conservation and management of Gulf of Mexico cetacean populations by improving the understanding of temporal and spatial species distribution trends.

Marine mammal conservation: over the horizon

Marine mammals can play important ecological roles in aquatic ecosystems, and their presence can be key to community structure and function. Consequently, marine mammals are often considered indicators of ecosystem health and flagship species. Yet, historical population declines caused by exploitation, and additional current threats, such as climate change, fisheries bycatch, pollution and maritime development, continue to impact many marine mammal species, and at least 25% are classified as threatened (Critically Endangered, Endangered or Vulnerable) on the IUCN Red List. Conversely, some species have experienced population increases/recoveries in recent decades, reflecting management interventions, and are heralded as conservation successes. To continue these successes and reverse the downward trajectories of at-risk species, it is necessary to evaluate the threats faced by marine mammals and the conservation mechanisms available to address them. Additionally, there is a need to identify evidence-based priorities of both research and conservation needs across a range of settings and taxa. To that effect we: (1) outline the key threats to marine mammals and their impacts, identify the associated knowledge gaps and recommend actions needed; (2) discuss the merits and downfalls of established and emerging conservation mechanisms; (3) outline the application of research and monitoring techniques; and (4) highlight particular taxa/populations that are in urgent need of focus.

Echolocation click parameters of short-finned pilot whales (Globicephala macrorhynchus) in the wild

Short-finned pilot whales (Globicephala macrorhynchus) are large, deep-diving predators with diverse foraging strategies, but little is known about their echolocation. To quantify the source properties of short-finned pilot whale clicks, we made 15 deployments off the coast of Tenerife of a deep-water hydrophone array consisting of seven autonomous time-synced hydrophone recorders (SoundTraps), enabling acoustic localization and quantification of click source parameters. Of 8185 recorded pilot whale clicks, 47 were classified as being recorded on-axis, with a mean peak-to-peak source level (SL) of 181 ± 7 dB re 1 μPa, a centroid frequency of 40 ± 4 kHz, and a duration of 57 ± 23 μs. A fit to a piston model yielded an estimated half-power (-3 dB) beam width of 13.7° [95% confidence interval (CI) 13.2°-14.5°] and a mean directivity index (DI) of 22.6 dB (95% CI 22.5-22.9 dB). These measured SLs and DIs are surprisingly low for a deep-diving toothed whale, suggesting we sampled the short-finned pilot whales in a context with little need for operating a long-range biosonar. The substantial spectral overlap with beaked whale clicks emitted in similar deep-water habitats implies that pilot whale clicks may constitute a common source of false detections in beaked whale passive acoustic monitoring efforts.

Deep-diving beaked whales dive together but forage apart

Echolocating animals that forage in social groups can potentially benefit from eavesdropping on other group members, cooperative foraging or social defence, but may also face problems of acoustic interference and intra-group competition for prey. Here, we investigate these potential trade-offs of sociality for extreme deep-diving Blainville's and Cuvier's beaked whales. These species perform highly synchronous group dives as a presumed predator-avoidance behaviour, but the benefits and costs of this on foraging have not been investigated. We show that group members could hear their companions for a median of at least 91% of the vocal foraging phase of their dives. This enables whales to coordinate their mean travel direction despite differing individual headings as they pursue prey on a minute-by-minute basis. While beaked whales coordinate their echolocation-based foraging periods tightly, individual click and buzz rates are both independent of the number of whales in the group. Thus, their foraging performance is not affected by intra-group competition or interference from group members, and they do not seem to capitalize directly on eavesdropping on the echoes produced by the echolocation clicks of their companions. We conclude that the close diving and vocal synchronization of beaked whale groups that quantitatively reduces predation risk has little impact on foraging performance.

Acoustic detection range and population density of Cuvier's beaked whales estimated from near-surface hydrophones

The population density of Cuvier's beaked whales is estimated acoustically with drifting near-surface hydrophone recorders in the Catalina Basin. Three empirical approaches (trial-based, distance-sampling, and spatially explicit capture-recapture) are used to estimate the probability of detecting the echolocation pulses as a function of range. These detection functions are used with two point-transect methods (snapshot and dive-cue) to estimate density. Measurement errors result in a small range of density estimates (3.9-5.4 whales per 1000 km²). Use of multiple approaches and methods allows comparison of the required information and assumptions of each. The distance-sampling approach with snapshot-based density estimates has the most stringent assumptions but would be the easiest to implement for large scale surveys of beaked whale density. Alternative approaches to estimating detection functions help validate this approach. The dive cue method of density estimation has promise, but additional work is needed to understand the potential bias caused by animal movement during a dive. Empirical methods are a viable alternative to the theoretical acoustic modeling approaches that have been used previously to estimate beaked whale density.

Automated extraction of dolphin whistles-A sequential Monte Carlo probability hypothesis density approach

The need for automated methods to detect and extract marine mammal vocalizations from acoustic data has increased in the last few decades due to the increased availability of long-term recording systems. Automated dolphin whistle extraction represents a challenging problem due to the time-varying number of overlapping whistles present in, potentially, noisy recordings. Typical methods utilize image processing techniques or single target tracking, but often result in fragmentation of whistle contours and/or partial whistle detection. This study casts the problem into a more general statistical multi-target tracking framework and uses the probability hypothesis density filter as a practical approximation to the optimal Bayesian multi-target filter. In particular, a particle version, referred to as a sequential Monte Carlo probability hypothesis density (SMC-PHD) filter, is adapted for frequency tracking and specific models are developed for this application. Based on these models, two versions of the SMC-PHD filter are proposed and the performance of these versions is investigated on an extensive real-world dataset of dolphin acoustic recordings. The proposed filters are shown to be efficient tools for automated extraction of whistles, suitable for real-time implementation.

MeLa: A Programming Language for a New Multidisciplinary Oceanographic Float

At 2000 m depth in the oceans, one can hear biological, seismological, meteorological, and anthropogenic activity. Acoustic monitoring of the oceans at a global scale and over long periods of time could bring important information for various sciences. The Argo project monitors the physical properties of the oceans with autonomous floats, some of which are also equipped with a hydrophone. These have a limited transmission bandwidth requiring acoustic data to be processed on board. However, developing signal processing algorithms for these instruments requires one to be an expert in embedded software. To reduce the need of such expertise, we have developed a programming language, called MeLa. The language hides several aspects of embedded software with specialized programming concepts. It uses models to compute energy consumption, processor usage, and data transmission costs early during the development of applications; this helps to choose a strategy of data processing that has a minimum impact on performances. Simulations on a computer allow for verifying the performance of the algorithms before their deployment on the instrument. We have implemented a seismic P wave detection and a blue whales D call detection algorithm with the MeLa language to show its capabilities. These are the first efforts toward multidisciplinary monitoring of the oceans, which can extend beyond acoustic applications.

Validation prediction: a flexible protocol to increase efficiency of automated acoustic processing for wildlife research

Automated recognition is increasingly used to extract species detections from audio recordings; however, the time required to manually review each detection can be prohibitive. We developed a flexible protocol called "validation prediction" that uses machine learning to predict whether recognizer detections are true or false positives and can be applied to any recognizer type, ecological application, or analytical approach. Validation prediction uses a predictable relationship between recognizer score and the energy of an acoustic signal but can also incorporate any other ecological or spectral predictors (e.g., time of day, dominant frequency) that will help separate true from false-positive recognizer detections. First, we documented the relationship between recognizer score and the energy of an acoustic signal for two different recognizer algorithm types (hidden Markov models and convolutional neural networks). Next, we demonstrated our protocol using a case study of two species, the Common Nighthawk (Chordeiles minor) and Ovenbird (Seiurus aurocapilla). We reduced the number of detections that required validation by 75.7% and 42.9%, respectively, while retaining at least 98% of the true-positive detections. Validation prediction substantially improves the efficiency of using automated recognition on acoustic data sets. Our method can be of use to wildlife monitoring and research programs and will facilitate using automated recognition to mine bioacoustic data sets.

New technologies in the mix: Assessing N-mixture models for abundance estimation using automated detection data from drone surveys

Reliable estimates of abundance are critical in effectively managing threatened species, but the feasibility of integrating data from wildlife surveys completed using advanced technologies such as remotely piloted aircraft systems (RPAS) and machine learning into abundance estimation methods such as N-mixture modeling is largely unknown due to the unique sources of detection errors associated with these technologies.We evaluated two modeling approaches for estimating the abundance of koalas detected automatically in RPAS imagery: (a) a generalized N-mixture model and (b) a modified Horvitz-Thompson (H-T) estimator method combining generalized linear models and generalized additive models for overall probability of detection, false detection, and duplicate detection. The final estimates from each model were compared to the true number of koalas present as determined by telemetry-assisted ground surveys.The modified H-T estimator approach performed best, with the true count of koalas captured within the 95% confidence intervals around the abundance estimates in all 4 surveys in the testing dataset (n = 138 detected objects), a particularly strong result given the difficulty in attaining accuracy found with previous methods.The results suggested that N-mixture models in their current form may not be the most appropriate approach to estimating the abundance of wildlife detected in RPAS surveys with automated detection, and accurate estimates could be made with approaches that account for spurious detections.

Listening forward: approaching marine biodiversity assessments using acoustic methods

Ecosystems and the communities they support are changing at alarmingly rapid rates. Tracking species diversity is vital to managing these stressed habitats. Yet, quantifying and monitoring biodiversity is often challenging, especially in ocean habitats. Given that many animals make sounds, these cues travel efficiently under water, and emerging technologies are increasingly cost-effective, passive acoustics (a long-standing ocean observation method) is now a potential means of quantifying and monitoring marine biodiversity. Properly applying acoustics for biodiversity assessments is vital. Our goal here is to provide a timely consideration of emerging methods using passive acoustics to measure marine biodiversity. We provide a summary of the brief history of using passive acoustics to assess marine biodiversity and community structure, a critical assessment of the challenges faced, and outline recommended practices and considerations for acoustic biodiversity measurements. We focused on temperate and tropical seas, where much of the acoustic biodiversity work has been conducted. Overall, we suggest a cautious approach to applying current acoustic indices to assess marine biodiversity. Key needs are preliminary data and sampling sufficiently to capture the patterns and variability of a habitat. Yet with new analytical tools including source separation and supervised machine learning, there is substantial promise in marine acoustic diversity assessment methods.

Estimating spatial distribution and density of vocalizing marine animals using compact arrays

The problem of estimating spatial distribution and density of vocalizing marine animals is addressed. The proposed solution is based on using a fixed compact array of synchronized hydrophones and statistically optimal detection and estimation algorithms. The closed-form representations of the practical algorithms are presented. The performance of the proposed technique is evaluated analytically and using statistical simulations. The case study involved identifying an area of high residency and estimating the density of vocalizing beluga whales in the St. Lawrence Estuary. The advantages and disadvantages of the proposed technique are demonstrated and the future steps are discussed.

Fear of Killer Whales Drives Extreme Synchrony in Deep Diving Beaked Whales

Fear of predation can induce profound changes in the behaviour and physiology of prey species even if predator encounters are infrequent. For echolocating toothed whales, the use of sound to forage exposes them to detection by eavesdropping predators, but while some species exploit social defences or produce cryptic acoustic signals, deep-diving beaked whales, well known for mass-strandings induced by navy sonar, seem enigmatically defenceless against their main predator, killer whales. Here we test the hypothesis that the stereotyped group diving and vocal behaviour of beaked whales has benefits for abatement of predation risk and thus could have been driven by fear of predation over evolutionary time. Biologging data from 14 Blainville’s and 12 Cuvier’s beaked whales show that group members have an extreme synchronicity, overlapping vocal foraging time by 98% despite hunting individually, thereby reducing group temporal availability for acoustic detection by killer whales to <25%. Groups also perform a coordinated silent ascent in an unpredictable direction, covering a mean of 1 km horizontal distance from their last vocal position. This tactic sacrifices 35% of foraging time but reduces by an order of magnitude the risk of interception by killer whales. These predator abatement behaviours have likely served beaked whales over millions of years, but may become maladaptive by playing a role in mass strandings induced by man-made predator-like sonar sounds.

Lombard effect: Minke whale boing call source levels vary with natural variations in ocean noise

Minke whales were acoustically detected, localized, and tracked on the U.S. Navy's Pacific Missile Range Facility from 2012 to 2017. Animal source levels (SLs) were estimated by adding transmission loss estimates to measured received levels of 42 159 individual minke whale boings. Minke whales off Hawaii exhibited the Lombard effect in that they increased their boing call intensity in increased background noise. Minke whales also decreased the variance of the boing call SL in higher background noise levels. Although the whales partially compensated for increasing background noise, they were unable or unwilling to increase their SLs by the same amount as the background noise. As oceans become louder, this reduction in communication space could negatively impact the health of minke whale populations. The findings in this study also have important implications for acoustic animal density studies, which may use SL to estimate probability of detection.

Comparison of fin whale 20 Hz call detections by deep-water mobile autonomous and stationary recorders

Acoustically equipped deep-water mobile autonomous platforms can be used to survey for marine mammals over intermediate spatiotemporal scales. Direct comparisons to fixed recorders are necessary to evaluate these tools as passive acoustic monitoring platforms. One glider and two drifting deep-water floats were simultaneously deployed within a deep-water cabled hydrophone array to quantitatively assess their survey capabilities. The glider was able to follow a pre-defined track while float movement was somewhat unpredictable. Fin whale (Balaenoptera physalus) 20 Hz pulses were recorded by all hydrophones throughout the two-week deployment. Calls were identified using a template detector, which performed similarly across recorder types. The glider data contained up to 78% fewer detections per hour due to increased low-frequency flow noise present during glider descents. The glider performed comparably to the floats and fixed recorders at coarser temporal scales; hourly and daily presence of detections did not vary by recorder type. Flow noise was related to glider speed through water and dive state. Glider speeds through water of 25 cm/s or less are suggested to minimize flow noise and the importance of glider ballasting, detector characterization, and normalization by effort when interpreting glider-collected data and applying it to marine mammal density estimation are discussed.

Machine learning in acoustics: Theory and applications

Acoustic data provide scientific and engineering insights in fields ranging from biology and communications to ocean and Earth science. We survey the recent advances and transformative potential of machine learning (ML), including deep learning, in the field of acoustics. ML is a broad family of techniques, which are often based in statistics, for automatically detecting and utilizing patterns in data. Relative to conventional acoustics and signal processing, ML is data-driven. Given sufficient training data, ML can discover complex relationships between features and desired labels or actions, or between features themselves. With large volumes of training data, ML can discover models describing complex acoustic phenomena such as human speech and reverberation. ML in acoustics is rapidly developing with compelling results and significant future promise. We first introduce ML, then highlight ML developments in four acoustics research areas: source localization in speech processing, source localization in ocean acoustics, bioacoustics, and environmental sounds in everyday scenes.

Comparison of estimated 20-Hz pulse fin whale source levels from the tropical Pacific and Eastern North Atlantic Oceans to other recorded populations

Passive acoustic monitoring, mitigation, animal density estimation, and comprehensive understanding of the impact of sound on marine animals all require accurate information on vocalization source level to be most effective. This study focused on examining the uncertainty related to passive sonar equation terms that ultimately contribute to the variability observed in estimated source levels of fin whale calls. Differences in hardware configuration, signal detection methods, sample size, location, and time were considered in interpreting the variability of estimated fin whale source levels. Data from Wake Island in the Pacific Ocean and off Portugal in the Atlantic Ocean provided the opportunity to generate large datasets of estimated source levels to better understand sources of uncertainty leading to the observed variability with and across years. Average seasonal source levels from the Wake Island dataset ranged from 175 to 188 dB re 1 μPa m, while the 2007-2008 seasonal average detected off Portugal was 189 dB re 1 μPa m. Owing to the large inherent variability within and across this and other studies that potentially masks true differences between populations, there is no evidence to conclude that the source level of 20-Hz fin whale calls are regionally or population specific.

Migrating eastern North Pacific gray whale call and blow rates estimated from acoustic recordings, infrared camera video, and visual sightings

During the eastern North Pacific gray whale 2014–2015 southbound migration, acoustic call recordings, infrared blow detections, and visual sightings were combined to estimate cue rates, needed to convert detections into abundance. The gray whale acoustic call rate ranged from 2.3–24 calls/whale/day during the peak of the southbound migration with an average of 7.5 calls/whale/day over both the southbound and northbound migrations. The average daily calling rate increased between 30 December–13 February. With a call rate model, we estimated that 4,340 gray whales migrated south before visual observations began on 30 December, which is 2,829 more gray whales than used in the visual estimate, and would add approximately 10% to the abundance estimate. We suggest that visual observers increase their survey effort to all of December to document gray whale presence. The infrared camera blow rate averaged 49 blows/whale/hour over 5–8 January. Probability of detection of a whale blow by the infrared camera was the same at night as during the day. However, probability of detection decreased beyond 2.1 km offshore, whereas visual sightings revealed consistent whale densities up to 3 km offshore. We suggest that future infrared camera surveys use multiple cameras optimised for different ranges offshore.

Impact of pile-driving on Hector's dolphin in Lyttelton Harbour, New Zealand

Several dolphin species occur close inshore and in harbours, where underwater noise generated by pile-driving used in wharf construction may constitute an important impact. Such impacts are likely to be greatest on species such as the endangered Hector's dolphin (Cephalorhynchus hectori), which has small home ranges and uses this habitat type routinely. Using automated echolocation detectors in Lyttelton Harbour (New Zealand), we studied the distribution of Hector's dolphins using a gradient sampling design over 92 days within which pile-driving occurred on 46 days. During piling operations, dolphin positive minutes per day decreased at the detector closest to the piling but increased at the mid-harbour detector. Finer-grained analyses showed that close to the piling operation, detections decreased with increasing sound exposure level, that longer piling events were associated with longer reductions in detections, and that effects were long-lasting - detection rates took up to 83 h to return to pre-piling levels.

Predicting acoustic dose associated with marine mammal behavioural responses to sound as detected with fixed acoustic recorders and satellite tags

To understand the consequences of underwater noise exposure for cetaceans, there is a need for assessments of behavioural responses over increased spatial and temporal scales. Bottom-moored acoustic recorders and satellite tags provide such long-term and large spatial coverage of behaviour compared to short-duration acoustic-recording tags. However, these tools result in a decreased resolution of data from which an animal response can be inferred, and no direct recording of the sound received at the animal. This study discusses the consequence of the decreased resolution of data from satellite tags and fixed acoustic recorders on the acoustic dose estimated by propagation modelling and presents a method for estimating the range of sound levels that animals observed with these methods have received. This problem is illustrated using experimental results obtained during controlled exposures of northern bottlenose whales (Hyperoodon ampullatus) exposed to naval sonar, carried out near Jan Mayen, Norway. It is shown that variability and uncertainties in the sound field, resulting from limited sampling of the acoustic environment, as well as decreased resolution in animal locations, can lead to quantifiable uncertainties in the estimated acoustic dose associated with the behavioural response (in this case avoidance and cessation of foraging).

Using on-board sound recordings to infer behaviour of free-moving wild animals

Technological advances in the last 20 years have enabled researchers to develop increasingly sophisticated miniature devices (tags) that record an animal's behaviour not from an observational, external viewpoint, but directly on the animals themselves. So far, behavioural research with these tags has mostly been conducted using movement or acceleration data. But on-board audio recordings have become more and more common following pioneering work in marine mammal research. The first questions that come to mind when recording sound on-board animals concern their vocal behaviour. When are they calling? How do they adjust their behaviour? What acoustic parameters do they change and how? However, other topics like foraging behaviour, social interactions or environmental acoustics can now be addressed as well and offer detailed insight into the animals' daily life. In this Review, we discuss the possibilities, advantages and limitations of on-board acoustic recordings. We focus primarily on bats as their active-sensing, echolocating lifestyle allows many approaches to a multi-faceted acoustic assessment of their behaviour. The general ideas and concepts, however, are applicable to many animals and hopefully will demonstrate the versatility of on-board acoustic recordings and stimulate new research.