A statistical acoustics approach for estimating population-scale bowhead whale migration speed and direction.

Bowhead whales vocalize during their annual fall migration from the Beaufort Sea to the Bering Sea, but the calling rates of individual animals are so low that tracking an individual trajectory is impractical using passive acoustic methods. However, the travel speed and direction of the migrating population can be inferred on a statistical basis by cross-correlating time sequences of call density measured at two locations spaced several kilometers apart. By using the triangulation abilities of a set of vector sensors deployed offshore the Alaskan North Slope between 2008 and 2014, call density time sequences were generated from 1-km wide and 40-km tall rectangular "zones" that were separated by distances ranging from 3.5 to 15 km. The cross-covariances between the two sequences generate a peak corresponding to the average time it takes for whales to travel between the zones. Consistent westward travel speeds of ∼5 km/h were obtained from four different locations on 6 of the 7 years of the study, independent of whether the zones were separated by 3.5, 7, or 15 km. Some sites, however, also revealed a less prominent eastern movement of whales, and shifts in migration speed were occasionally detectable over week-long time scales.

Diel spatio-temporal patterns of humpback whale singing on a high-density breeding ground.

Humpback whale song chorusing dominates the marine soundscape in Hawai'i during winter months, yet little is known about spatio-temporal habitat use patterns of singers. We analysed passive acoustic monitoring data from five sites off Maui and found that ambient noise levels associated with song chorusing decreased during daytime hours nearshore but increased offshore. To resolve whether these changes reflect a diel offshore-onshore movement or a temporal difference in singing activity, data from 71 concurrently conducted land-based theodolite surveys were analysed. Non-calf pods (n = 3082), presumably including the majority of singers, were found further offshore with increasing time of the day. Separately, we acoustically localized 217 nearshore singers using vector-sensors. During the day, distances to shore and minimum distances among singers increased, and singers switched more between being stationary and singing while travelling. Together, these findings suggest that the observed diel trends in humpback whale chorusing off Maui represent a pattern of active onshore-offshore movement of singers. We hypothesize that this may result from singers attempting to reduce intraspecific acoustic masking when densities are high nearshore and avoidance of a loud, non-humpback, biological evening chorus offshore, creating a dynamic of movement of singers aimed at increasing the efficiency of their acoustic display.

Automating multi-target tracking of singing humpback whales recorded with vector sensors.

Passive acoustic monitoring is widely used for detection and localization of marine mammals. Typically, pressure sensors are used, although several studies utilized acoustic vector sensors (AVSs), that measure acoustic pressure and particle velocity and can estimate azimuths to acoustic sources. The AVSs can localize sources using a reduced number of sensors and do not require precise time synchronization between sensors. However, when multiple animals are calling concurrently, automated tracking of individual sources still poses a challenge, and manual methods are typically employed to link together sequences of measurements from a given source. This paper extends the method previously reported by Tenorio-Hallé, Thode, Lammers, Conrad, and Kim [J. Acoust. Soc. Am. 151(1), 126-137 (2022)] by employing and comparing two fully-automated approaches for azimuthal tracking based on the AVS data. One approach is based on random finite set statistics and the other on message passing algorithms, but both approaches utilize the underlying Bayesian statistical framework. The proposed methods are tested on several days of AVS data obtained off the coast of Maui and results show that both approaches successfully and efficiently track multiple singing humpback whales. The proposed methods thus made it possible to develop a fully-automated AVS tracking approach applicable to all species of baleen whales.

Multi-target 2D tracking method for singing humpback whales using vector sensors.

Acoustic vector sensors allow estimating the direction of travel of an acoustic wave at a single point by measuring both acoustic pressure and particle motion on orthogonal axes. In a two-dimensional plane, the location of an acoustic source can thus be determined by triangulation using the estimated azimuths from at least two vector sensors. However, when tracking multiple acoustic sources simultaneously, it becomes challenging to identify and link sequences of azimuthal measurements between sensors to their respective sources. This work illustrates how two-dimensional vector sensors, deployed off the coast of western Maui, can be used to generate azimuthal tracks from individual humpback whales singing simultaneously. Incorporating acoustic transport velocity estimates into the processing generates high-quality azimuthal tracks that can be linked between sensors by cross-correlating features of their respective azigrams, a particular time-frequency representation of sound directionality. Once the correct azimuthal track associations have been made between instruments, subsequent localization and tracking in latitude and longitude of simultaneous whales can be achieved using a minimum of two vector sensors. Two-dimensional tracks and positional uncertainties of six singing whales are presented, along with swimming speed estimates derived from a high-quality track.

Measurements of open-water arctic ocean noise directionality and transport velocity.

Measurements from bottom-mounted acoustic vector sensors, deployed seasonally between 2008 and 2014 on the shallow Beaufort Sea shelf along the Alaskan North Slope, are used to estimate the ambient sound pressure power spectral density, acoustic transport velocity of energy, and dominant azimuth between 25 and 450 Hz. Even during ice-free conditions, this region has unusual acoustic features when compared against other U.S. coastal regions. Two distinct regimes exist in the diffuse ambient noise environment: one with high pressure spectral density levels but low directionality, and another with lower spectral density levels but high directionality. The transition between the two states, which is invisible in traditional spectrograms, occurs between 73 and 79 dB re 1 µPa2/Hz at 100 Hz, with the transition region occurring at lower spectral levels at higher frequencies. Across a wide bandwidth, the high-directionality ambient noise consistently arrives from geographical azimuths between 0° and 30° from true north over multiple years and locations, with a seasonal interquartile range of 40° at low frequencies and high transport velocities. The long-term stability of this directional regime, which is believed to arise from the dominance of wind-driven sources along an east-west coastline, makes it an important feature of arctic ambient sound.

Estimating acoustic cue rates in bowhead whales, Balaena mysticetus, during their fall migration through the Alaskan Beaufort Sea.

Eight years of passive acoustic data (2007-2014) from the Beaufort Sea were used to estimate the mean cue rate (calling rate) of individual bowhead whales (Balaena mysticetus) during their fall migration along the North Slope of Alaska. Calls detected on directional acoustic recorders (DASARs) were triangulated to provide estimates of locations at times of call production, which were then translated into call densities (calls/h/km2). Various assumptions were used to convert call density into animal cue rates, including the time for whales to cross the arrays of acoustic recorders, the population size, the fraction of the migration corridor missed by the localizing array system, and the fraction of the seasonal migration missed because recorders were retrieved before the end of the migration. Taking these uncertainties into account in various combinations yielded up to 351 cue rate estimates, which summarize to a median of 1.3 calls/whale/h and an interquartile range of 0.5-5.4 calls/whale/h.

Using anisotropic and narrowband ambient noise for continuous measurements of relative clock drift between independent acoustic recorders.

Relative clock drift between instruments can be an issue for coherent processing of acoustic signals, which requires data to be time-synchronized between channels. This work shows how cross correlation of anisotropic narrowband ambient noise allows continuous estimation of the relative clock drift between independent acoustic recorders, under the assumption that the spatial distribution of the coherent noise sources is stationary. This method is applied to two pairs of commercial passive acoustic recorders deployed up to 14 m apart at 6 and 12 m depth, respectively, over a period of 10 days. Occasional calibration signals show that this method allows time-synchronizing the instruments to within ±1 ms. In addition to a large linear clock drift component on the order of tens of milliseconds per hour, the results reveal for these instruments non-linear excursions of up to 50 ms that cannot be measured by standard methods but are crucial for coherent processing. The noise field displays the highest coherence between 50 and 100 Hz, a bandwidth dominated by what are believed to be croaker fish, which are particularly vocal in the evenings. Both the passive and continuous nature of this method provide advantages over time-synchronization using active sources.

Deep embedded clustering of coral reef bioacoustics.

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

Automated two-dimensional localization of underwater acoustic transient impulses using vector sensor image processing (vector sensor localization).

Detecting acoustic transients by signal-to-noise ratio (SNR) becomes problematic in nonstationary ambient noise environments characteristic of coral reefs. An alternate approach presented here uses signal directionality to automatically detect and localize transient impulsive sounds collected on underwater vector sensors spaced tens of meters apart. The procedure, which does not require precise time synchronization, first constructs time-frequency representations of both the squared acoustic pressure (spectrogram) and dominant directionality of the active intensity (azigram) on each sensor. Within each azigram, sets of time-frequency cells associated with transient energy arriving from a consistent azimuthal sector are identified. Binary image processing techniques then link sets that share similar duration and bandwidth between different sensors, after which the algorithm triangulates the source location. Unlike most passive acoustic detectors, the threshold criterion for this algorithm is bandwidth instead of pressure magnitude. Data collected from shallow coral reef environments demonstrate the algorithm's ability to detect SCUBA bubble plumes and consistent spatial distributions of somniferous fish activity. Analytical estimates and direct evaluations both yield false transient localization rates from 3% to 6% in a coral reef environment. The SNR distribution of localized pulses off Hawaii has a median of 7.7 dB and interquartile range of 7.1 dB.

Nonlinear time-warping made simple: A step-by-step tutorial on underwater acoustic modal separation with a single hydrophone.

Classical ocean acoustic experiments involve the use of synchronized arrays of sensors. However, the need to cover large areas and/or the use of small robotic platforms has evoked interest in single-hydrophone processing methods for localizing a source or characterizing the propagation environment. One such processing method is "warping," a non-linear, physics-based signal processing tool dedicated to decomposing multipath features of low-frequency transient signals (frequency f < 500 Hz), after their propagation through shallow water (depth D < 200 m) and their reception on a distant single hydrophone (range r > 1 km). Since its introduction to the underwater acoustics community in 2010, warping has been adopted in the ocean acoustics literature, mostly as a pre-processing method for single receiver geoacoustic inversion. Warping also has potential applications in other specialties, including bioacoustics; however, the technique can be daunting to many potential users unfamiliar with its intricacies. Consequently, this tutorial article covers basic warping theory, presents simulation examples, and provides practical experimental strategies. Accompanying supplementary material provides matlab code and simulated and experimental datasets for easy implementation of warping on both impulsive and frequency-modulated signals from both biotic and man-made sources. This combined material should provide interested readers with user-friendly resources for implementing warping methods into their own research.

Roaring and repetition: How bowhead whales adjust their call density and source level (Lombard effect) in the presence of natural and seismic airgun survey noise.

Over 500 000 automated and manual acoustic localizations, measured over seven years between 2008 and 2014, were used to examine how natural wind-driven noise and anthropogenic seismic airgun survey noise influence bowhead whale call densities (calls/km2/min) and source levels during their fall migration in the Alaskan Beaufort Sea. Noise masking effects, which confound measurements of behavioral changes, were removed using a modified point transect theory. The authors found that mean call densities generally rose with increasing continuous wind-driven noise levels. The occurrence of weak airgun pulse sounds also prompted an increase in call density equivalent to a 10-15 dB change in natural noise level, but call density then dropped substantially with increasing cumulative sound exposure level (cSEL) from received airgun pulses. At low in-band noise levels the mean source level of the acoustically-active population changed to nearly perfectly compensate for noise increases, but as noise levels increased further the mean source level failed to keep pace, reducing the population's communication space. An increase of >40 dB cSEL from seismic airgun activity led to an increase in source levels of just a few decibels. These results have implications for bowhead acoustic density estimation, and evaluations of the masking impacts of anthropogenic noise.

Displaying bioacoustic directional information from sonobuoys using "azigrams".

The AN/SSQ-53 Directional Frequency Analysis and Recording (DIFAR) sonobuoy is an expendable device that can derive acoustic particle velocity along two orthogonal horizontal axes, along with acoustic pressure. This information enables computation of azimuths of low-frequency acoustic sources from a single compact sensor. The standard approach for estimating azimuth from these sensors is by conventional beamforming (i.e., adding weighted time series), but the resulting "cardioid" beampattern is imprecise, computationally expensive, and vulnerable to directional noise contamination for weak signals. Demonstrated here is an alternative multiplicative processing scheme that computes the "active intensity" of an acoustic signal to obtain the dominant directionality of a noise field as a function of time and frequency. This information is conveniently displayed as an "azigram," which is analogous to a spectrogram, but uses color to indicate azimuth instead of intensity. Data from several locations demonstrate this approach, which can be computed without demultiplexing the raw signal. Azigrams have been used to help diagnose sonobuoy issues, improve detectability, and estimate bearings of low signal-to-noise ratio signals. Azigrams may also enhance the detection and potential classification of signals embedded in directional noise fields.

Effects of tones associated with drilling activities on bowhead whale calling rates.

During summer 2012 Shell performed exploratory drilling at Sivulliq, a lease holding located in the autumn migration corridor of bowhead whales (Balaena mysticetus), northwest of Camden Bay in the Beaufort Sea. The drilling operation involved a number of vessels performing various activities, such as towing the drill rig, anchor handling, and drilling. Acoustic data were collected with six arrays of directional recorders (DASARs) deployed on the seafloor over ~7 weeks in Aug-Oct. Whale calls produced within 2 km of each DASAR were identified and localized using triangulation. A "tone index" was defined to quantify the presence and amplitude of tonal sounds from industrial machinery. The presence of airgun pulses originating from distant seismic operations was also quantified. For each 10-min period at each of the 40 recorders, the number of whale calls localized was matched with the "dose" of industrial sound received, and the relationship between calling rates and industrial sound was modeled using negative binomial regression. The analysis showed that with increasing tone levels, bowhead whale calling rates initially increased, peaked, and then decreased. This dual behavioral response is similar to that described for bowhead whales and airgun pulses in earlier work. Increasing call repetition rates can be a viable strategy for combating decreased detectability of signals arising from moderate increases in background noise. Meanwhile, as noise increases, the benefits of calling may decrease because information transfer becomes increasingly error-prone, and at some point calling may no longer be worth the effort.

Eastern Arctic ambient noise on a drifting vertical array.

Ambient noise in the eastern Arctic was studied from April to September 2013 using a 22 element vertical hydrophone array as it drifted from near the North Pole (89° 23'N, 62° 35'W) to north of Fram Strait (83° 45'N, 4° 28'W). The hydrophones recorded for 108 min/day on six days per week with a sampling rate of 1953.125 Hz. After removal of data corrupted by non-acoustic transients, 19 days throughout the transit period were analyzed. Noise contributors identified include broadband and tonal ice noises, bowhead whale calling, seismic airgun surveys, and earthquake T phases. The bowhead whale or whales detected are believed to belong to the endangered Spitsbergen population, and were recorded when the array was as far north as 86° 24'N. Median power spectral estimates and empirical probability density functions along the array transit show a change in the ambient noise levels corresponding to seismic survey airgun occurrence and received level at low frequencies and transient ice noises at high frequencies. Median power for the same periods across the array shows that this change is consistent in depth. The median ambient noise for May 2013 was among the lowest of the sparse reported observations in the eastern Arctic but comparable to the more numerous observations of western Arctic noise levels.

Decadal-scale frequency shift of migrating bowhead whale calls in the shallow Beaufort Sea.

Automated and manual acoustic localizations of bowhead whale calls in the Beaufort Sea were used to estimate the minimum frequency attained by their highly variable FM-modulated call repertoire during seven westerly fall migrations. Analyses of 13 355 manual and 100 009 automated call localizations found that between 2008 and 2014 the proportion of calls that dipped below 75 Hz increased from 27% to 41%, shifting the mean value of the minimum frequency distribution from 94 to 84 Hz. Multivariate regression analyses using both generalized linear models and generalized estimating equations found that this frequency shift persisted even when accounting for ten other factors, including calling depth, call range, call type, noise level, signal-to-noise ratio, local water depth (site), airgun activity, and call spatial density. No single call type was responsible for the observed shift, but so-called "complex" calls experienced larger percentage downward shifts. By contrast, the call source level distribution remained stable over the same period. The observed frequency shift also could not be explained by migration corridor shifts, relative changes in call detectability between different frequency bands, long-term degradation in the automated airgun detector, physiological growth in the population, or behavioral responses to increasing population density (estimated via call density).

Using nonlinear time warping to estimate North Pacific right whale calling depths in the Bering Sea.

Calling depth distributions are estimated for two types of calls produced by critically endangered eastern North Pacific right whales (NPRWs) in the Bering Sea, using passive acoustic data collected with bottom-mounted hydrophone recorders. Nonlinear time resampling of 12 NPRW "upcalls" and 20 "gunshots" recorded in a critical NPRW habitat isolated individual normal mode arrivals from each call. The relative modal arrival times permitted range estimates between 1 and 40 km, while the relative modal amplitudes permitted call depth estimates, provided that environmental inversions were obtained from high signal-to-noise ratio calls. Gunshot sounds were generally only produced at a few meters depth, while upcall depths clustered between 10 and 25 m, consistent with previously published bioacoustic tagging results from North Atlantic right whales. A Wilcoxon rank sum test rejected the null hypothesis that the mean calling depths of the two call types were the same (p = 2.9 × 10-5); the null hypothesis was still rejected if the sample set was restricted to one call per acoustic encounter (p = 0.02). Propagation modeling demonstrates that deeper depths enhance acoustic propagation and that source depth estimates impact both NPRW upcall source level and detection range estimates.

Waveguide mode amplitude estimation using warping and phase compensation.

In shallow water, low-frequency propagation can be described by modal theory. Acoustical oceanographic measurements under this situation have traditionally relied on spatially filtering signals with arrays of synchronized hydrophones. Recent work has demonstrated how a method called warping allows isolation of individual mode arrivals on a single hydrophone, a discovery that subsequently opened the door for practical single-receiver source localization and geoacoustic inversion applications. Warping is a non-linear resampling of the signal based on a simplistic waveguide model. Because warping is robust to environmental mismatch, it provides accurate estimates of the mode phase even when the environment is poorly known. However, the approach has issues with mode amplitude estimation, particularly for the first arriving mode. As warping is not invariant to time shifting, it relies on accurate estimates of the signal's time origin, which in turn heavily impacts the first mode's amplitude estimate. Here, a revised warping operator is proposed that incorporates as much prior environmental information as possible, and is actually equivalent to compensating the relative phase of each mode. Warping and phase compensation are applied to both simulated and experimental data. The proposed methods notably improve the amplitude estimates of the first arriving mode.

A double-difference method for high-resolution acoustic tracking using a deep-water vertical array.

Ray-tracing is typically used to estimate the depth and range of an acoustic source in refractive deep-water environments by exploiting multipath information on a vertical array. However, mismatched array inclination and uncertain environmental features can produce imprecise trajectories when ray-tracing sequences of individual acoustic events. "Double-difference" methods have previously been developed to determine fine-scale relative locations of earthquakes along a fault [Waldhauser and Ellsworth (2000). Bull. Seismolog. Soc. Am. 90, 1353-1368]. This technique translates differences in travel times between nearby seismic events, recorded at multiple widely separated stations, into precise relative displacements. Here, this method for acoustic multipath measurements on a single vertical array of hydrophones is reformulated. Changes over time in both the elevation angles and the relative arrival times of the multipath are converted into relative changes in source position. This approach is tested on data recorded on a 128-element vertical array deployed in 4 km deep water. The trajectory of a controlled towed acoustic source was accurately reproduced to within a few meters at nearly 50 km range. The positional errors of the double-difference approach for both the towed source and an opportunistically detected sperm whale are an order of magnitude lower than those produced from ray-tracing individual events.

Using line acceleration to measure false killer whale (Pseudorca crassidens) click and whistle source levels during pelagic longline depredation.

False killer whales (Pseudorca crassidens) depredate pelagic longlines in offshore Hawaiian waters. On January 28, 2015 a depredation event was recorded 14 m from an integrated GoPro camera, hydrophone, and accelerometer, revealing that false killer whales depredate bait and generate clicks and whistles under good visibility conditions. The act of plucking bait off a hook generated a distinctive 15 Hz line vibration. Two similar line vibrations detected at earlier times permitted the animal's range and thus signal source levels to be estimated over a 25-min window. Peak power spectral density source levels for whistles (4-8 kHz) were estimated to be between 115 and 130 dB re 1 µPa2/Hz @ 1 m. Echolocation click source levels over 17-32 kHz bandwidth reached 205 dB re 1 µPa @ 1 m pk-pk, or 190 dB re 1 µPa @ 1 m (root-mean-square). Predicted detection ranges of the most intense whistles are 10 to 25 km at respective sea states of 4 and 1, with click detection ranges being 5 times smaller than whistles. These detection range analyses provide insight into how passive acoustic monitoring might be used to both quantify and avoid depredation encounters.

Humpback whale-generated ambient noise levels provide insight into singers' spatial densities.

Baleen whale vocal activity can be the dominant underwater ambient noise source for certain locations and seasons. Previous wind-driven ambient-noise formulations have been adjusted to model ambient noise levels generated by random distributions of singing humpback whales in ocean waveguides and have been combined to a single model. This theoretical model predicts that changes in ambient noise levels with respect to fractional changes in singer population (defined as the noise "sensitivity") are relatively unaffected by the source level distributions and song spectra of individual humpback whales (Megaptera novaeangliae). However, the noise "sensitivity" does depend on frequency and on how the singers' spatial density changes with population size. The theoretical model was tested by comparing visual line transect surveys with bottom-mounted passive acoustic data collected during the 2013 and 2014 humpback whale breeding seasons off Los Cabos, Mexico. A generalized linear model (GLM) estimated the noise "sensitivity" across multiple frequency bands. Comparing the GLM estimates with the theoretical predictions suggests that humpback whales tend to maintain relatively constant spacing between one another while singing, but that individual singers either slightly increase their source levels or song duration, or cluster more tightly as the singing population increases.