Ecological inferences about marine mammals from passive acoustic data.

Monitoring on the basis of sound recordings, or passive acoustic monitoring, can complement or serve as an alternative to real-time visual or aural monitoring of marine mammals and other animals by human observers. Passive acoustic data can support the estimation of common, individual-level ecological metrics, such as presence, detection-weighted occupancy, abundance and density, population viability and structure, and behaviour. Passive acoustic data also can support estimation of some community-level metrics, such as species richness and composition. The feasibility of estimation and certainty of estimates is highly context dependent, and understanding the factors that affect the reliability of measurements is useful for those considering whether to use passive acoustic data. Here, we review basic concepts and methods of passive acoustic sampling in marine systems that often are applicable to marine mammal research and conservation. Our ultimate aim is to facilitate collaboration among ecologists, bioacousticians, and data analysts. Ecological applications of passive acoustics require one to make decisions about sampling design, which in turn requires consideration of sound propagation, sampling of signals, and data storage. One also must make decisions about signal detection and classification and evaluation of the performance of algorithms for these tasks. Investment in the research and development of systems that automate detection and classification, including machine learning, are increasing. Passive acoustic monitoring is more reliable for detection of species presence than for estimation of other species-level metrics. Use of passive acoustic monitoring to distinguish among individual animals remains difficult. However, information about detection probability, vocalisation or cue rate, and relations between vocalisations and the number and behaviour of animals increases the feasibility of estimating abundance or density. Most sensor deployments are fixed in space or are sporadic, making temporal turnover in species composition more tractable to estimate than spatial turnover. Collaborations between acousticians and ecologists are most likely to be successful and rewarding when all partners critically examine and share a fundamental understanding of the target variables, sampling process, and analytical methods.

Silbido profundo: An open source package for the use of deep learning to detect odontocete whistles.

This work presents an open-source matlab software package for exploiting recent advances in extracting tonal signals from large acoustic data sets. A whistle extraction algorithm published by Li, Liu, Palmer, Fleishman, Gillespie, Nosal, Shiu, Klinck, Cholewiak, Helble, and Roch [(2020). Proceedings of the International Joint Conference on Neural Networks, July 19-24, Glasgow, Scotland, p. 10] is incorporated into silbido, an established software package for extraction of cetacean tonal calls. The precision and recall of the new system were over 96% and nearly 80%, respectively, when applied to a whistle extraction task on a challenging two-species subset of a conference-benchmark data set. A second data set was examined to assess whether the algorithm generalized to data that were collected across different recording devices and locations. These data included 487 h of weakly labeled, towed array data collected in the Pacific Ocean on two National Oceanographic and Atmospheric Administration (NOAA) cruises. Labels for these data consisted of regions of toothed whale presence for at least 15 species that were based on visual and acoustic observations and not limited to whistles. Although the lack of per whistle-level annotations prevented measurement of precision and recall, there was strong concurrence of automatic detections and the NOAA annotations, suggesting that the algorithm generalizes well to new data.

Improve automatic detection of animal call sequences with temporal context.

Many animals rely on long-form communication, in the form of songs, for vital functions such as mate attraction and territorial defence. We explored the prospect of improving automatic recognition performance by using the temporal context inherent in song. The ability to accurately detect sequences of calls has implications for conservation and biological studies. We show that the performance of a convolutional neural network (CNN), designed to detect song notes (calls) in short-duration audio segments, can be improved by combining it with a recurrent network designed to process sequences of learned representations from the CNN on a longer time scale. The combined system of independently trained CNN and long short-term memory (LSTM) network models exploits the temporal patterns between song notes. We demonstrate the technique using recordings of fin whale (Balaenoptera physalus) songs, which comprise patterned sequences of characteristic notes. We evaluated several variants of the CNN + LSTM network. Relative to the baseline CNN model, the CNN + LSTM models reduced performance variance, offering a 9-17% increase in area under the precision-recall curve and a 9-18% increase in peak F1-scores. These results show that the inclusion of temporal information may offer a valuable pathway for improving the automatic recognition and transcription of wildlife recordings.

High resolution three-dimensional beam radiation pattern of harbour porpoise clicks with implications for passive acoustic monitoring.

The source properties and radiation patterns of animal vocalisations define, along with propagation and noise conditions, the active space in which these vocalisations can be detected by conspecifics, predators, prey, and by passive acoustic monitoring (PAM). This study reports the 4π (360° horizontal and vertical) beam profile of a free-swimming, trained harbour porpoise measured using a 27-element hydrophone array. The forward echolocation beam is highly directional, as predicted by a piston model, and is consistent with previous measurements. However, at off-axis angles greater than ±30°, the beam attenuates more rapidly than the piston model and no side lobes are present. A diffuse back beam is also present with levels about -30 dB relative to the source level. In PAM, up to 50% of detections can be from portions of the beam profile with distorted click spectra, although this drops substantially for higher detection thresholds. Simulations of the probability of acoustically detecting a harbour porpoise show that a traditional piston model can underestimate the probability of detection compared to the actual three-dimensional radiation pattern documented here. This highlights the importance of empirical 4π measurements of beam profiles of toothed whales, both to improve understanding of toothed whale biology and to inform PAM.

Publisher Correction: Deep neural networks for automated detection of marine mammal species.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Characterisation of underwater operational sound of a tidal stream turbine.

The underwater sound emitted during the operation of the Atlantis AR1500 turbine, a 1.5 MW three bladed horizontal axis tidal-stream turbine, was measured in the Pentland Firth, Scotland. Most sound was concentrated in the lower frequencies, ranging from 50 to 1000 Hz. Within 20 m of the turbine, third-octave band sound pressure levels were elevated by up to 40 dB relative to ambient conditions. In comparison, ambient noise at these frequencies fluctuated by about 5-10 dB between different tidal states. At the maximum recording distance of 2300 m from the turbine, median sound pressure levels when the turbine was operational were still over 5 dB higher than ambient noise levels alone. A higher frequency, tonal signal was observed at 20 000 Hz. This signal component appears at a constant level whenever the turbine is operational and did not change with turbine rotation rate. It is most likely produced by the turbine's generator. This study highlights the importance of empirical measurements of turbine underwater sound. It illustrates the utility and challenges of using drifting hydrophone systems to spatially map operational turbine signal levels with reduced flow noise artefacts when recording in high flow environments.

Passive acoustic methods for tracking the 3D movements of small cetaceans around marine structures.

A wide range of anthropogenic structures exist in the marine environment with the extent of these set to increase as the global offshore renewable energy industry grows. Many of these pose acute risks to marine wildlife; for example, tidal energy generators have the potential to injure or kill seals and small cetaceans through collisions with moving turbine parts. Information on fine scale behaviour of animals close to operational turbines is required to understand the likely impact of these new technologies. There are inherent challenges associated with measuring the underwater movements of marine animals which have, so far, limited data collection. Here, we describe the development and application of a system for monitoring the three-dimensional movements of cetaceans in the immediate vicinity of a subsea structure. The system comprises twelve hydrophones and software for the detection and localisation of vocal marine mammals. We present data demonstrating the systems practical performance during a deployment on an operational tidal turbine between October 2017 and October 2019. Three-dimensional locations of cetaceans were derived from the passive acoustic data using time of arrival differences on each hydrophone. Localisation accuracy was assessed with an artificial sound source at known locations and a refined method of error estimation is presented. Calibration trials show that the system can accurately localise sounds to 2m accuracy within 20m of the turbine but that localisations become highly inaccurate at distances greater than 35m. The system is currently being used to provide data on rates of encounters between cetaceans and the turbine and to provide high resolution tracking data for animals close to the turbine. These data can be used to inform stakeholders and regulators on the likely impact of tidal turbines on cetaceans.

Deep neural networks for automated detection of marine mammal species.

Deep neural networks have advanced the field of detection and classification and allowed for effective identification of signals in challenging data sets. Numerous time-critical conservation needs may benefit from these methods. We developed and empirically studied a variety of deep neural networks to detect the vocalizations of endangered North Atlantic right whales (Eubalaena glacialis). We compared the performance of these deep architectures to that of traditional detection algorithms for the primary vocalization produced by this species, the upcall. We show that deep-learning architectures are capable of producing false-positive rates that are orders of magnitude lower than alternative algorithms while substantially increasing the ability to detect calls. We demonstrate that a deep neural network trained with recordings from a single geographic region recorded over a span of days is capable of generalizing well to data from multiple years and across the species' range, and that the low false positives make the output of the algorithm amenable to quality control for verification. The deep neural networks we developed are relatively easy to implement with existing software, and may provide new insights applicable to the conservation of endangered species.

Time of arrival difference estimation for narrow band high frequency echolocation clicks.

Algorithms are presented for the accurate time of arrival difference estimation of high frequency narrow band echolocation clicks from Harbor Porpoise. These clicks typically have a center frequency of around 130 kHz (wavelength ∼1.2 cm) and duration of <0.1 ms. When using hydrophones spaced centimeters apart, spatial aliasing can cause large errors on inter-hydrophone timing measurements due to the incorrect peak in the cross-correlation function of two signals being selected. It is shown that at sample rates of less than about 6 times the fundamental frequency, the incorrect correlation peak will be selected in 55% of measurements leading to large errors in time of arrival estimates. For clicks with a SNR > 10 dB these errors can be reduced by over two orders of magnitude through a combination of up-sampling the data and parabolic interpolation of peaks in the cross-correlation functions.

A review of unmanned vehicles for the detection and monitoring of marine fauna.

Recent technology developments have turned present-day unmanned systems into realistic alternatives to traditional marine animal survey methods. Benefits include longer survey durations, improved mission safety, mission repeatability, and reduced operational costs. We review the present status of unmanned vehicles suitable for marine animal monitoring conducted in relation to industrial offshore activities, highlighting which systems are suitable for three main monitoring types: population, mitigation, and focal animal monitoring. We describe the technical requirements for each of these monitoring types and discuss the operational aspects. The selection of a specific sensor/platform combination depends critically on the target species and its behaviour. The technical specifications of unmanned platforms and sensors also need to be selected based on the surrounding conditions of a particular offshore project, such as the area of interest, the survey requirements and operational constraints.

Comparing methods suitable for monitoring marine mammals in low visibility conditions during seismic surveys.

Loud sound emitted during offshore industrial activities can impact marine mammals. Regulations typically prescribe marine mammal monitoring before and/or during these activities to implement mitigation measures that minimise potential acoustic impacts. Using seismic surveys under low visibility conditions as a case study, we review which monitoring methods are suitable and compare their relative strengths and weaknesses. Passive acoustic monitoring has been implemented as either a complementary or alternative method to visual monitoring in low visibility conditions. Other methods such as RADAR, active sonar and thermal infrared have also been tested, but are rarely recommended by regulatory bodies. The efficiency of the monitoring method(s) will depend on the animal behaviour and environmental conditions, however, using a combination of complementary systems generally improves the overall detection performance. We recommend that the performance of monitoring systems, over a range of conditions, is explored in a modelling framework for a variety of species.

Acoustic monitoring to document the spatial distribution and hotspots of blast fishing in Tanzania.

Destructive fishing using explosives occurs in a number of countries worldwide, negatively impacting coral reefs and fisheries on which millions of people rely. Documenting, quantifying and combating the problem has proved problematic. In March-April 2015 231h of acoustic data were collected over 2692km of systematically laid transects along the entire coast of Tanzania. A total of 318 blasts were confirmed using a combination of manual and supervised semi-autonomous detection. Blasts were detected along the entire coastline, but almost 62% were within 80km of Dar es Salaam, where blast frequency reached almost 10blasts/h. This study is one of the first to use acoustic monitoring to provide a spatial assessment of the intensity of blast fishing. This can be a useful tool that can provide reliable data to define hotspots where the activity is concentrated and determine where enforcement should be focused for maximum impact.

Passive acoustic methods for fine-scale tracking of harbour porpoises in tidal rapids.

The growing interest in generating electrical power from tidal currents using tidal turbine generators raises a number of environmental concerns, including the risk that marine mammals might be injured or killed through collision with rotating turbine blades. To understand this risk, information on how marine mammals use tidal rapid habitats and in particular, their underwater movements and dive behaviour is required. Porpoises, which are the most abundant small cetacean at most European tidal sites, are difficult animals to tag, and the limited size of tidal habitats means that any telemetered animal would be likely to spend only a small proportion of time within them. Here, an alternative approach is explored, whereby passive acoustic monitoring (PAM) is used to obtain fine scale geo-referenced tracks of harbour porpoises in tidal rapid areas. Large aperture hydrophone arrays are required to obtain accurate locations of animals from PAM data and automated algorithms are necessary to process the large quantities of acoustic data collected on such systems during a typical survey. Methods to automate localisation, including a method to match porpoise detections on different hydrophones and separate different vocalising animals, and an assessment of the localisation accuracy of the large aperture hydrophone array are presented.

Software for real-time localization of baleen whale calls using directional sonobuoys: A case study on Antarctic blue whales.

Directional frequency analysis and recording (DIFAR) sonobuoys can allow real-time acoustic localization of baleen whales for underwater tracking and remote sensing, but limited availability of hardware and software has prevented wider usage. These software limitations were addressed by developing a module in the open-source software PAMGuard. A case study is presented demonstrating that this software provides greater efficiency and accessibility than previous methods for detecting, localizing, and tracking Antarctic blue whales in real time. Additionally, this software can easily be extended to track other low and mid frequency sounds including those from other cetaceans, pinnipeds, icebergs, shipping, and seismic airguns.

Automatic detection and classification of odontocete whistles.

Methods for the fully automatic detection and species classification of odontocete whistles are described. The detector applies a number of noise cancellation techniques to a spectrogram of sound data and then searches for connected regions of data which rise above a pre-determined threshold. When tested on a dataset of recordings which had been carefully annotated by a human operator, the detector was able to detect (recall) 79.6% of human identified sounds that had a signal-to-noise ratio above 10 dB, with 88% of the detections being valid. A significant problem with automatic detectors is that they tend to partially detect whistles or break whistles into several parts. A classifier has been developed specifically to work with fragmented whistle detections. By accumulating statistics over many whistle fragments, correct classification rates of over 94% have been achieved for four species. The success rate is, however, heavily dependent on the number of species included in the classifier mix, with the mean correct classification rate dropping to 58.5% when 12 species were included.

The effects of acoustic misclassification on cetacean species abundance estimation.

To estimate the density or abundance of a cetacean species using acoustic detection data, it is necessary to correctly identify the species that are detected. Developing an automated species classifier with 100% correct classification rate for any species is likely to stay out of reach. It is therefore necessary to consider the effect of misidentified detections on the number of observed data and consequently on abundance or density estimation, and develop methods to cope with these misidentifications. If misclassification rates are known, it is possible to estimate the true numbers of detected calls without bias. However, misclassification and uncertainties in the level of misclassification increase the variance of the estimates. If the true numbers of calls from different species are similar, then a small amount of misclassification between species and a small amount of uncertainty around the classification probabilities does not have an overly detrimental effect on the overall variance. However, if there is a difference in the encounter rate between species calls and/or a large amount of uncertainty in misclassification rates, then the variance of the estimates becomes very large and this dramatically increases the variance of the final abundance estimate.

Field recordings of Gervais' beaked whales Mesoplodon europaeus from the Bahamas.

The first recordings from free-ranging Gervais' beaked whale (Mesoplodon europaeus) are presented. Nine Gervais' beaked whales were observed visually for over 6 h. Clicks were only detected over a 15 min period during the encounter, which coincided with an 88 min period during which no whales were observed at the surface. Click lengths were typically around 200 microS and their dominant energy was in the frequency range 30-50 kHz. While these clicks were broadly similar to those of Cuvier's and Blainville's beaked whales, the Gervais' beaked whale clicks were at a slightly higher frequency than those of the other species.