Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=tbio20
Bioacoustics
The International Journal of Animal Sound and its Recording
ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/tbio20
Observations on mechanisms and phenomena
underlying underwater and surface vocalisations
of grey seals
Lukasz J. Nowak
To cite this article: Lukasz J. Nowak (2020): Observations on mechanisms and phenomena underlying underwater and surface vocalisations of grey seals, Bioacoustics, DOI:
10.1080/09524622.2020.1851298
To link to this article: https://doi.org/10.1080/09524622.2020.1851298
© 2020 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.
Published online: 07 Dec 2020.
Submit your article to this journal
Article views: 470
View related articles
Observations on mechanisms and phenomena underlying
underwater and surface vocalisations of grey seals
Lukasz J. Nowak
Biomedical Photonic Imaging Group, University of Twente, Enschede, The Netherlands ABSTRACT
Air and water constitute acoustic propagation media with very different physical properties. Consequently, mechanisms for air- and water-borne sound generation observed in terrestrial and aquatic animals are also significantly different. Notwithstanding, some amphibious species are capable of making efficient use of their vocal tracts in both environments. The present study investi-gates such a phenomenon on the example of grey seals Halichoerus
grypus. Grey seals vocalise both underwater and above the water
surface emitting variety of sounds. Both kinds of vocalisations were recorded using microphones and hydrophones, and analysed for their acoustic parameters which could be related to the underlying generation mechanisms. Also, video with synchronised underwater audio recordings were captured, allowing to link the acoustic phe-nomena with specific behaviour of the animals. The vocalisations are categorised into three groups based on differences in acoustic characteristics. Temporal and spectral parameters of the sounds belonging to each group are determined and discussed in terms of potential underlying generation mechanisms. Based on the adopted criteria surface vocalisations do not constitute a separate group but are categorised together with other tonal sounds pro-duced underwater. Other two vocalisation groups include pulsed sounds with different bandwidths, duration times, repetition rates and associated behavioural signatures.
ARTICLE HISTORY
Received 15 June 2020 Accepted 23 October 2020
KEYWORDS
Grey seals; sound production; underwater vocalisations; vocalisation mechanisms
Introduction
Efficient acoustic communication is possible both in air and in water, however acoustic properties of both media are significantly different. Density of water is approximately eight hundred times greater than air and acoustic wave propagation velocity is over four times greater. Water is also almost incompressible compared to air. Those differences are reflected in mechanisms and phenomena underlying sound production in terrestrial and aquatic animals.
Many air-breathing terrestrial species, including humans, utilise so-called myoelastic- aerodynamic (MEAD) mechanism (Titze Ingo 1980) for direct excitation of air-borne sounds during inhalation or exhalation. This mechanism, based on compression and modulation of the airflow is observed in amphibians (Narins et al. 2006), reptiles (Fitch and Suthers 2016), birds, and mammals (Elemans et al. 2015). MEAD in principle
CONTACT Lukasz J. Nowak l.j.nowak@utwente.nl
© 2020 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits non-commercial re-use, distribution, and reproduction in any med-ium, provided the original work is properly cited, and is not altered, transformed, or built upon in any way.
includes also structural tissue vibrations, however generation of tissue-borne sound is in this case rather side effect, as the primary acoustic radiation path is through the outlet of the respiratory/vocal tract. Some of air-borne vocalisation sounds (as in, for instance, human speech) can be also generated purely by exploiting airflow turbulences and cavity resonances (Small 2019). The described mechanisms can be efficiently used only for surface sound production. Due to the high impedance contrast air-borne acoustic waves hardly penetrate water interface (Leighton 2012).
Body surface vibrations can be used to generate sound in any acoustic medium, acting on the same principle as a loudspeaker. However, just as not every loudspeaker will work underwater, several conditions have to be met for this mechanism to be efficient. Structures vibrating in air, if their dimensions are relatively low compared to the wavelength of radiated acoustic waves, encounter low damping and their vibrational behaviour can be approximately modelled neglecting the impact of the surrounding medium (Nowak and Zielinski 2013). In other words, they can be easily excited to vibrate at their natural resonant frequencies. Exciting structural vibrations of specialised body parts for sound production is widely exploited by many insects (Bennet-Clark 1975). Water introduces high inertial loading and damping, which significantly alter vibroa-coustic properties of submerged structures (Nowak and Zielinski 2012). This also makes excitation of free vibrations not feasible, requiring some intermediate sound transmis-sion mechanisms. Dolphins produce sounds underwater inducing vibrations of tissues within pneumatic cavities by air flowing through nasal sacs (Au 1993). The generated tissue-borne acoustic waves follow anatomical transmission paths to be radiated to the ambient space by body surface vibrations. The transmission efficiency is ensured due to the match in acoustic impedances between tissues and the surrounding water. Similar mechanism is utilised by other marine mammal species for underwater communication
and echolocation (Au and Hastings 2009).
Despite the substantial differences in operating conditions imposed on acoustic sources by air and water, various amphibious species are capable of effective acoustic communication in the both environments. The present study investigates this phenom-enon on the example of grey seals Halichoerus grypus), focusing on acoustic and behavioural signatures that could be linked with mechanisms underlying sound produc-tion below and above the water surface.
Grey seals vocalise both in air and underwater emitting variety of sounds (Au and Hastings 2009). Those sounds were associated with different social interactions between animals and in this context were the subject of numerous studies (see, for instance, Schneider 1974; Schusterman 1978; Boness and James 1979; Caudron et al. 1998; McCulloch and Boness 2000; Sauve et al. 2015). Possibility of using echoloca-tion by seals was also investigated, however without conclusive results supporting this claim (Schevill et al. 1963; Oliver 1978; Asselin et al. 1993; Schusterman et al. 2000). Asselin et al. (1993) presented broad overview of grey seals’ vocal repertoire, based on surface and underwater recordings of vocalisations of wild animals. The sounds were classified into seven categories, based on the differences in fundamental and maximum frequencies, duration, and presence of harmonics. The study aimed at relating the vocalisations with changes in animal behaviour. Another broad overview of vocal repertoire of various pinnipeds was presented by Miksis-Olds et al. (2016). The study specifies the number of different aquatic call types of grey
seals as ten (referring to the state of knowledge until 2000). It was also shown, that vocalisations of different seal species reveal distinctive individual variations (Hanggi and Schusterman 1994; McCulloch et al. 1999). Reported frequency range of the emitted sounds falls within the audible bandwidth, ranging from approximately 100 Hz up to 8,6 kHz (Asselin et al. 1993). In this regard amphibious pinnipeds are not different from many terrestrial animals (Martin et al. 2017). Recently, observation of percussive underwater signalling in wild grey seals was also reported (Hocking et al. 2020).
The present study investigates grey seals as complex acoustic sources capable of efficiently emitting variety of sounds both in air and in water. It aims at drawing consistent conclusions regarding mechanisms and phenomena under-lying those capabilities, based on behavioural and acoustic signatures asso-ciated with specific vocalisations. Thus, quite an opposite link is attempted to be made compared to the other studies assigning social interaction functions to the emitted sounds. Conducting research on captive animals allowed to sepa-rate the individuals and obtain detailed audio-visual information regarding the vocalisation process. The considered behavioural signatures include air exhala-tion (manifested by the presence of bubbles exhaled underwater), degree of immersion in water and body postures during sound production. The con-ducted analysis of the recorded sounds had focussed on frequency content and its temporal variations, time duration, and relation of the acoustic character-istics with the accompanying animal behaviour. Differences between the para-meters of air- and water-borne sounds were investigated. Among hundreds of recorded vocalisations some revealed distinctive, unusual features. Such acous-tic events were given a special attention, as potential carriers of additional cues regarding the sound production mechanisms.
Understanding of phenomena exploited by animals for sound emission in two, radically different acoustic media could prove useful not only for biologically related studies, but also for the development of new biomimetic acoustic transducers. This constitutes interesting research opportunities for instance in the field of underwater
communication (Nowak 2016).
Materials and methods
Vocalisations of six captive grey seals, four females and two males, were recorded at Hel Marine Station of the University of Gdansk. The recordings took place during the breeding seasons 2013 and 2015, February – March. The facility consists of three irregularly shaped water reservoirs, connected with narrow passages which can be closed enabling separation of animals. The depth of the pools varied between approximately 1 and 3 metres. Surface recordings, conducted in March 2015, included additionally three newborn pups. Attempts were also made to record underwater vocalisations during different times of a year, out of breeding season, however the animals produced sounds very sporadically, causing the recording and extraction of acoustic events to be ineffec-tive. This observation is consistent with other reports on grey seals’ behaviour (Asselin et al. 1993).
The underwater recordings were conducted using Aquarian Audio H2a hydrophones (Aquarian Audio & Scientific, Anacortes, USA) and Zoom H2N audio recorder (ZOOM Corp., Tokyo, Japan). For long-time sound acquisitions of over several hours the hydrophones were attached by cables to ropes stretched over the water reservoirs and submerged approximately one metre below the surface. During other measurements, the hydrophones were mounted on a long rod and operated manually, following the animals in water. Real-time acoustic monitoring and visual observations allowed, in specific cases (e.g. when only one animal was present in water) to assign vocalisations to indivi-duals and specific behaviour and phenomena (e.g. presence or absence of exhaled air bubbles during vocalisation). Synchronised audio and video record-ings were also conducted using the hydrophones connected to a Pentax K-5 camera (Pentax Corporation, Tokyo, Japan). For surface measurements, built- in microphones of the audio recorder were used to capture sound. The recor-der was installed on a tripod and positioned close to (several metres away) the place where three mothers with newborn pups were located. This was dictated by the fact, that the animals vocalised on surface almost exclusively in this area.
Approximately 130 hours of underwater audio recordings, 6 hours of surface record-ings, and one hour of video with synchronised audio recordings were collected. The data were first visually analysed using waveform visualisation software (Sonic visualiser and Audacity) in order to pre-select parts containing acoustic events of interest. All the pre- selected fragments were extracted as separate audio or video files and listened to for subjective evaluation. Files containing noise or other unwanted sounds were rejected, together with vocalisations during which the signal was overdriven (caused by close proximity of hydrophone to the vocalising animal and relatively high sound levels) or where multiple sounds overlapped. The remaining set of data had then undergone detailed acoustic analysis conducted using Sonic Visualiser and in-house Matlab scripts. Fundamental frequencies, presence and amplitude ratios of harmonics, frequency change (for tonal sounds), total numbers and repetition rates (for pulsed sounds), occupied bandwidths and duration times were determined. Special attention was given to all unusual events, such as sudden shift in harmonic amplitude ratio accompanying fre-quency or sound level change. Recordings including such events were marked and additionally described, as carriers of potentially important cues on the character of the underlying acoustic phenomena. Based on the obtained results, the sounds were divided into three groups associated with the assumed differences in sound production mechan-isms, as it will be described in the following parts of the present manuscript. Parameter range and variability for each group were determined. The results of acoustic analysis were complemented by the visual observations associated with specific events, provided either in the form of notes taken during the measurements or as results of analysis of acoustic events extracted from video files with synchronised underwater audio. The video recordings had also undergone additional selection based on the image quality. Due to the relatively high water turbidity, limited field of view and automatic focus operation not all the recordings with proper acoustic events contained clear visual information. Such files were also rejected, leaving only videos in which the vocalising animals were clearly visible.
Results
Extraction and classification
Following the procedure described in the previous section a total number of 394 acoustic events were extracted as separate files and individually analysed. This includes 202 audio files with underwater recordings, 183 audio files with surface recordings and 9 files with video recordings. Each file contains either a single vocalisation or a series of vocalisations. All the data files are stored in an online repository and can be freely accessed (Nowak 2020a, 2020b, 2020c, 2020d, 2020e).
The collected recordings were divided into three groups. Two categorisation criteria related to temporal and spectral characteristics of vocalisations were selected. The first parameter is a binary variable describing vocalisation type: tonal (with clearly determin-able dominant harmonic component) or pulsed (series of short pulses without dominant harmonic components). The other parameter is a pulse rate, defined as the ratio of the number of pulses or separate acoustic events to the total vocalisation time. The pulse rates were normalised with respect to the highest determined value to balance the division criteria. K-means clustering was applied for categorisation. The results are presented in Figure 1.
The vocalisations were divided into three groups, referred to as ‘‘S1ʹ’, ‘‘S2ʹ’, and ‘‘S3ʹ’, with distinct separation of the clusters based on the adopted categorisation criteria. The introduced vocalisation types together with their basic acoustic characteristics underlying categorisation procedure are presented in Table 1.
The detailed characteristics and determined features of sounds from each group are presented below, followed by the discussion justifying the division criteria and hypoth-eses linking the observed phenomena with possible underlying sound generation mechanisms.
Figure 1. K-means clustering of the recorded vocalisations: S1, S2, and S3 represent the determined clusters with markers indicating individual audio recordings.
S1 vocalisations
The S1 group contains tonal vocalisations with multiple harmonic components. The sounds were produced by completely and partially submerged seals, and also on the surface. The S1 vocalisations outnumbered other recorded sounds with 174 underwater and 183 surface acoustic events extracted as separate audio files. The files can be accessed through the online repository (Nowak 2020a, 2020b, 2020e). The underwater vocalisa-tions were emitted by adult males and females, while sounds produced on surface are primarily communication between mothers and their pups.
For every single vocalisation, the fundamental frequency and dominant harmonic number were determined. As these values in many cases varied in time, the changes were also noted. Most of the vocalisations involved amplitude and/or frequency modulation. Some basic parameters of the investigated S1 vocalisations are presented in Table 2. There are no large discrepancies between the recorded air- and waterborne sounds: the deter-mined median of fundamental frequencies is exactly the same, while the observed frequency modulation range was slightly greater on surface. The presented differences, however, might be also caused by uneven animal representation: the newborn pups did not enter the water during the recordings sessions, spending most of the time with their mothers.
Figure 2 presents the determined frequency shifts of all the recorded vocalisations classified as S1 type. The plots were created separately for surface and underwater sounds.
Table 1. Vocalisation types and their basic acoustic characteristics.
S1 S2 S3
Vocalisation type Tonal Pulsed Pulsed
No. of underwater recordings 174 15 12
No. of surface recordings 183 0 0
Pulse rate, mean [Hz] 0.95 45.3 3.94
Pulse rate, max [Hz] 3.75 50 5.75
Pulse rate, min [Hz] 0.11 45 1.44
Pulse rate, median [Hz] 0.81 5 3.95
Vocalisation time, mean [s] 1.74 1.68 2.72
Vocalisation time, max [s] 18.85 1.99 11.88
Vocalisation time, min [s] 0.27 0.85 0.37
Vocalisation time, median [s] 1.3 1.74 1.12
Number of acoustic events, mean 1.14 73.73 10
Number of acoustic events, min 1 37 2
Number of acoustic events, max 7 88 38
Number of acoustic events, median 1 77 4
Table 2. Acoustic parameters typical for the S1 type vocalisations.
Recording type Underwater Surface
Number of files 174 183
Fundamental frequency, mean [Hz] 326 380
Fundamental frequency, median [Hz] 380 380
Fundamental frequency, min [Hz] 50 80
Fundamental frequency, max [Hz] 710 750
No. of recordings with frequency modulation 128 156
Modulation range, mean [Hz] 63 81
Modulation range, median [Hz] 40 60
Modulation range, max [Hz] 320 360
In numerous cases in which the change of fundamental frequency was observed during a single S1 type vocalisation, it had a form of a continuous modulation. The ranges of these changes are indicated in the plots in Figure 2 as vertical lines.
The duration of a single vocalisation classified here as S1 type varied from 0,27 s to 18,85 s (mean: 1,74 s; median: 1,3 s). The longest recorded vocalisations consisted of several subsequent acoustic events separated with pauses (maximum 7), however in most of the cases in consisted of a single acoustic event (minimum: 1; mean: 1,14; median: 1, N = 357). In all cases in which it was possible to observe, the vocalisations were accompanied by air exhalation through nostrils – visible as air bubbles in water. Observations of partially submerged, vocalising animals provided important cues on the sound production mechanisms and acoustic transmission paths. Some of such events were captured as video recordings with synchronised underwater audio. Figures 3 and 4 present examples illustrating such recordings. The vocalising animal kept its nostrils above water for most of the time during sound production. Within this period the sound could be clearly heard on the surface, and also captured by the hydrophone underwater. As it can be seen in Figure 3, the seal keeps its nostrils just above the surface, and the moment of complete immersion coincides with a sudden vanishing of sound. In other case, as introduced in Figure 4, the vocalisation continues after the nostrils are sub-merged, and water bubbles can be seen until the sound emission stops. The air-water interface is hardly penetrable by acoustic waves propagating in any direction, thus the described observations provide a strong indication on the existence of two independent sound transmission paths in grey seals.
As presented in Table 2, 100 out of 174 recorded underwater vocalisations and 77 out of 183 surface vocalisations contained a single dominant harmonic – i.e. the number of a highest level harmonic component did not change over the whole sound production period. In the remaining recordings a shift in the dominant harmonic was observed. This phenomenon is illustrated in Figures 5 and 6, presenting waveforms and spectrograms of two S1 type Figure 2. Fundamental frequency ranges of the recorded surface and underwater S1 vocalisations.
vocalisations recorded underwater and on the surface, respectively. In Figure 5 almost seven seconds long vocalisation begins with a part with dominant fundamental frequency compo-nent at around 400 Hz. After approximately 4.2 seconds the timbre is altered, and the second harmonic takes over the dominant level. The change is preceded with a short decay. Figure 3. Frames from a video recording (top) captured with synchronised underwater audio record-ing (represented as a waveform plot in the middle) presentrecord-ing partially submerged, vocalisrecord-ing grey seal. Bottom plot presents spectrogram of the audio signal. Dashed vertical lines on the waveform plot mark the moments of time corresponding to the video frames.
Figure 4. Frames from a video recording (top) captured with synchronised underwater audio record-ing (represented as a waveform plot in the middle) presentrecord-ing partially submerged, vocalisrecord-ing grey seal. Bottom plot presents spectrogram of the audio signal. Dashed vertical lines on the waveform plot mark the moments of time corresponding to the video frames.
Approximately 5 seconds after the beginning of the sound another rapid shift back to the dominant fundamental frequency is observed. Figure 6 presents a S1 type vocalisation recorded on the surface, consisting of two parts divided by approximately 200 ms long pause. The first part includes dominant harmonic component at fundamental frequency of approximately 420 Hz. After the pause vocalisation continues, this time with third harmonic component at the highest level. What is also typical of the S1 vocalisations, fundamental frequency remains almost constant during most of the sound emission period, while complex amplitude modulation is observed.
Figure 5. A waveform (top) and the corresponding spectrogram (bottom) of a S1 type vocalisation recorded underwater.
Figure 6. A waveform (top) and the corresponding spectrogram (bottom) of a S1 type vocalisation recorded on the surface.
S2 vocalisations
The S2 vocalisations are defined as series of short bursts without any distinct tonal components present. A typical waveform and spectrogram of such sound are presented in Figure 7. Emission of the S2 type vocalisations was observed only underwater, and only by the male adult seals – the sounds are thus, most presumably, the mating calls, as also described in the literature (Asselin et al. 1993). 15 such vocalisations were extracted as separate audio files, together with 3 video recordings with underwater audio. The files can be accessed through the online repository (Nowak 2020c, 2020e).
In all cases in which visual observation was possible, emission of S2 type vocalisa-tions was accompanied by air exhalation manifested in the form of air bubbles under-water. This was also captured on a video recording, as illustrated in the Figure 8. The presented sequence of video frames shows also typical behaviour associated with emission of the described sounds: the animal swimming initially at the surface takes a dive (t = 4 s), then turns around taking arched, chest-up posture and vocalises exhaling air (approx. 7.7–9.6 s), and then returns to normal swimming.
Although the described behaviour illustrated in Figure 9 was common for both observed male seals it was also noted that the reversed, chest-up posture was not required to actually produce the S2 type sound. A counterexample was also captured on a video, as presented in Figure 9. In this situation, a male was approaching a female in a shallow part of the water reservoir, where taking a dive and turning around was not feasible. Still, the animal emitted almost 2 seconds long vocalisation without any special, visible behavioural cues.
Each of the 17 vocalisations extracted in the separate audio files included between 37 and 88 pulses (average 74, median 77). The repetition rate was in almost all the cases constant and equal 45 Hz, with only one observed exception at 50 Hz. No significant fluctuations of the pulse duration during vocalisation process were noted.
Figure 7. A waveform (top) and the corresponding spectrogram (bottom) of a S2 type vocalisation recorded underwater.
Figure 8. Frames from a video recording (top) captured with synchronised underwater audio record-ing (represented as a waveform plot in the middle) illustratrecord-ing typical behaviour related to production of S2 type vocalisations. Bottom plot presents spectrogram of the audio signal. Dashed vertical lines on the waveform plot mark the moments of time corresponding to the video frames.
Figure 9. Frames from a video recording (top) captured with synchronised underwater audio record-ing (represented as a waveform plot in the middle) illustratrecord-ing observed situation in which emission of S2 vocalisation took place in a shallow water environment, without taking typical reversed posture by the vocalising animal (dark back of the seal constantly visible). Bottom plot presents spectrogram of the audio signal. Dashed vertical lines on the waveform plot mark the moments of time corresponding to the video frames.
S2 vocalisations were presumably the loudest sounds emitted by the observed animals. Although direct sound level measurements were not feasible due to the complex acoustic environment and animal movement, such conclusion was drawn based on the fact that the sounds emitted underwater by entirely submerged seals could also be heard on the surface, despite high transmission losses at the air-water interface. Such phenomenon was not observed in any other of the described acoustic signals.
S3 vocalisations
The S3 vocalisations are defined as a series of short, pulsed sounds with several distinct harmonic components. Such acoustic signals were observed only as water-borne sounds, and captured only with a hydrophone. After analysis 12 audio files with the correspond-ing sounds were extracted. They were also present in three selected video recordcorrespond-ings. The files can be accessed through the online repository (Nowak 2020d, 2020e).
Example waveforms and spectrograms of such acoustic signals are presented in Figures 10 and 11. In the recorded signals the lowest harmonic components had frequencies between approximately 50 Hz and 160 Hz, and most of the acoustic energy was contained within the bandwidth up to approximately 3 kHz. A characteristic feature of the acquired S3 sounds was that they appeared either as fast repeating (dozens of millisecond separation) pairwise similar sounding pulses (as in Figure 10), or as sub-sequent similar pulses separated with longer (hundreds of milliseconds, up to over a second) pauses (as in Figure 11). The pulses also often included frequency modulation (monotonic increase or decrease of all the harmonic components), as it can be seen in Figure 10 (last part of the vocalisation).
The duration of the recorded pulses varied between 24 ms and 425 ms (mean: 126 ms; median: 105 ms; total number of pulses in all the recorded S3 vocalisations N = 140). In
Figure 10. A waveform (top) and the corresponding spectrogram (bottom) of a S3 type vocalisation recorded underwater.
all the cases when visual identification and observation of an vocalising animal was possible no air bubbles were noticed at the time of S3 sound emission. This can be also seen in the captured video recordings, as illustrated in Figure 12. Vocalising female seal was keeping head and nostrils underwater while emitting series of pulsed, tonal sounds without any visible signs of air exhalation. Such behaviour is in this regard different from the one observed together with S1 vocalisations. Despite the fact that in some observed Figure 11. An example of a S3 vocalisation created by a sequence of singular pulses.
Figure 12. Frames from a video recording (top) captured with synchronised underwater audio recording (represented as a waveform plot in the middle) presenting emission of a series of S3 type vocalisations by a partially submerged animal with head and nostrils underwater. Bottom plot presents spectrogram of the audio signal. Dashed vertical lines on the waveform plot mark the moments of time corresponding to the video frames.
cases, as illustrated for example in Figure 12, vocalising animals were only partially submerged, the produced S3 sounds could only be heard underwater, and not on the surface.
Discussion
The classification of the recorded underwater and surface vocalisations into three groups, herein referred to as S1, S2, and S3, was based on the differences in temporal and spectral acoustic characteristics. The division criteria were selected in such a way, that all the recorded vocalisations could be unambiguously categorised, with no sounds rejected or matching only partially. Potential social functions of the emitted acoustic signals did not fall within the scope of the present study, and were not considered as a classification factor. Such an approach is different than the ones presented in other studies, and thus the number of indicated vocalisation types is also different, compared to, for instance, Asselin et al. (1993); Miksis-Olds et al. (2016).
Asselin et al. described six different underwater tonal vocalisations with mean funda-mental frequencies ranging from 200 to 400 Hz and average duration times between 1,2 and 2,74 s (Asselin et al. (1993)). The S1 type tonal vocalisations (underwater) described in the present study are characterised with overall mean fundamental frequency 326 Hz and mean duration 1,74 s, which is consistent with the previously reported values. Also, the mean frequency bandwidths of the pulsed sounds described by Asselin et al. as trrots and clicks (3,3 and 3,5 kHz), and their mean durations (1,21 and 2,09 s, respectively), are close to the values reported herein for S2 and S3 type vocalisations (as presented in Table 1 and, e.g., Figures 9 and 12). The percussive sounds described recently by Hocking et al. (2020) were not observed in the investigated animals and are not included in the classification.
The study was conducted on captive animals, within contained and controlled envir-onment, adapted also for general audience. The close access to the water reservoirs, with good visibility of vocalising animals enabled tracking them with the hydrophone, and in many cases to note or capture on video important behavioural cues related to the sound production process – such as, for instance, presence of air bubbles indicating air exhalation.
The first introduced type of vocalisations, referred to as S1, was defined as tonal sounds with single fundamental frequency and higher order harmonics. This determined fundamental frequency values within the collected recordings ranged between 50 and 750 Hz, which is significant difference and might rise questions if all those acoustic signals should be considered similar. In order to justify the adopted clustering criteria in this regard the frequency modulation ranges were determined for every extracted vocalisation and put together on a plot, as presented in Figure 2. As it can be seen, the vertical lines (indicating frequency ranges covered by every single recorded vocalisation) overlap, covering the whole considered bandwidth without any empty spaces between. Such spaces could indicate the legitimacy of further division of the acquired data into more call types. In the present case, the observed parameter variability might be explained by individual variations and tunability of the underlying sound production mechanisms.
The S1 vocalisations were the only sounds which could be simultaneously heard both on the surface and underwater when emitted by a partially submerged animal. The sound
production process was accompanied by air exhalation in all the cases in which such an observation was possible. The air-borne sounds could only be heard as long as the nostrils of the vocalising animal were above the water surface – examples of such an apparently intentional behaviour were observed and captured on video recordings, as illustrated in Figures 3 and 4. The underwater sounds captured simultaneously with the hydrophone were far to intense and clear to suppose that they might originate from acoustic waves partially penetrating the air-water interface. Thus, it is hypothesised that the production of S1 vocalisations is based on double sound transmission path from inside the body to the ambient space. The tonal character of the sound with multiple harmonic components resembles the MEAD mechanism observed in many other terres-trial and amphibious species. Assuming this is the case, vibrations of vocal folds (or similar anatomical structures) excited by an airflow in grey seals would be the source of air-borne sound emitted to the ambient space through the nostrils, but would also be efficiently transmitted as tissue-borne sounds. When an animal is on the surface, the former mechanism would be dominant, while in the water the latter would take over the primary role. This would also explain efficient simultaneous transmission into both environments when the animal is partially submerged. Again, it should be emphasised that the description presented above is only a hypothesis developed based on the observed phenomena, and should be tested against further experimental data.
Series of loud, short pulses – referred herein as S2 type vocalisations – were emitted only by two adult male seals, and only underwater. The determined time period between the subsequent pulses (approximately 20 ms) was constant and repeatable with no clear individual differences between the recorded animals. The observed behavioural signatures indicate that the sound production process in this case is powered by flow of the exhaled air. The specific upside-down posture taken by the vocalising animal (presented in Figure 8) was observed for most of the corresponding recorded acoustic events, suggesting that such body orientation could in some way influence the sound production process. However, it was also noted that S2 vocalisations can be emitted without reversing and stretching chest region, as presented in Figure 9. In order to determine if the upside-down position has any direct effect on pulse repetition rate and duration, the acoustic signals were extracted from both video files illustrated in Figures 8 and 9, and compared. Waveforms of the extracted fragments of these signals are presented in Figure 13. The pulse count was identical in both cases, with 26 repetitions over the period of 520 ms. There was a significant difference in the maximal values of sound pressure levels (with sound emitted in upside-down position being louder). Although it seems probable that the observed behaviour should be somehow beneficial in terms of the sound emission, and that those benefits could be related to the achievable sound pressure level, the obtained results are not conclusive in this regard. This is due to the fact that the distance from the hydrophone to the vocalising animals and acoustic propagation conditions were variable and out of control. Still, the obtained results and analysis of the synchronised audio and video recordings allow to state that the body position does not have a direct effect on the frequency of the emitted acoustic pulses in the S2 vocalisations.
S3 vocalisations were defined as series of short, pulsed sounds with distinct har-monic components. The bandwidth of the recorded signals of this kind was approxi-mately the same as in the case of S1 vocalisations, as illustrated in Figures 3–5 and 10–12. The most important difference between S1 and S3 type vocalisations is
considered to lie in the absence of air exhalation during emission of the latter. S3 pulses are also significantly shorter – reaching usually 100 ms up to maximum 425 ms in the observed cases – than the tonal S1 sounds with recorded duration times of several seconds (shorter S1 calls were also observed, but they did not occur in series). Based on these observations it is hypothesised that the anatomical structures involved in generation of S3 sounds are the same as in the case of S1 vocalisations, except that in the former case the nostrils remain closed and the air is subsequently pumped back and forth through the vocal tract without being exhaled. Thus, the duration of a single pulse would be limited by the amount of air that can be compressed above the tissues excited to vibrate by the flow. Another implication resulting from the presented hypothesis would be a pairwise similarity within pulse sequences that should be observed due to the subsequent airflow direction switching and related variances in excited tissue vibrations. Such a phenomenon can indeed be clearly heard in many among the recorded S3 vocalisations (available from Nowak (2020d)). It can be also illustrated on an example of a pulse sequence extracted from the longer pulse train (depicted in Figure 10) for better visibility, as presented in Figure 14. In some recordings, sequences of singular, similar pulses were also observed. The pulses, however, were separated by longer pauses lasting several hundreds of milliseconds. An example of such a recording is presented in Figure 11. It is hypothesised that in such a case one of the S3 vocalisation phases (i.e. for a given airflow direction) was mute. The muting could be achieved by slowing down the air compression process.
During the pause periods visible in Figure 11 some residual sound pulses (at
approxi-mately 0,8 and 1,8 s) can also be noted, which would in fact also support the pairwise similarity hypothesis.
Figure 13. Waveforms of S2 vocalisations recorded underwater when vocalising seal took character-istic upside-down position during sound production process (top plot), and without such behaviour, in shallow water environment (bottom plot).
The extent of different parameters of the underwater and surface vocalisa-tions of grey seals presented herein describes what is certainly achievable for the investigated animals, however the introduced limits cannot be interpreted as the vocal ability limits in general. This refers to maximum and minimum fundamental frequencies, modulation ranges, and duration times. Long-term observations of larger groups of grey seals could push some of these numbers further, and thus could be potentially interesting follow-up of this study. The biggest advantages of the introduced approach (i.e., investigations of captive animals in the confined environment) were the possibilities to link acoustic events with visual cues and to keep microphones and hydrophones relatively close to the vocalising animals. The gathered data allowed to formulate the presented hypotheses regarding the mechanisms and phenomena underlying sound production and to introduce the corresponding classification criteria. Another interesting continuation of the present study would be an attempt to link the described results with the anatomical data. The formulated statements could be also used as a basis and a reference for similar investigations on other amphibious species.
Acknowledgements
I would like to acknowledge the invaluable help and support obtained from prof. Krzysztof E. Skora, and other employees of the Hel Marine Station of the University of Gdansk.
Disclosure statement
The author declares no potential conflict of interest.
Figure 14. An example of a sequence of pairwise similar pulses extracted from a recorded S3 vocalisation.
Funding
The data acquisition for this study was possible thanks to the funds granted by the Foundation for Polish Science, grant no. [175/UD/SKILLS/2012], realized within the framework of SKILLS programme, co-financed from European Union, European Social Fund.
Data availability
The audio and video recordings containing all the data acquired within the present study are available through the online repository:
https://figshare.com/articles/Underwater\_vocalizations\_of\_grey\_seals\_-\_video\_with \_synchronized\_underwater\_audio\_recordings/12033849 https://figshare.com/articles/S1\_surface\_vocalizations\_of\_grey\_seals/12033828 https://figshare.com/articles/S1\_underwater\_vocalizations\_of\_grey\_seals/12033807 https://figshare.com/articles/S2\_underwater\_vocalizations\_of\_grey\_seals/12033834 https://figshare.com/articles/S3\_underwater\_vocalizations\_of\_grey\_seals/12033837 ORCID
Lukasz J. Nowak http://orcid.org/0000-0003-1155-3447
References
Asselin S, Hammill MO, Barrette C. 1993. Underwater vocalizations of ice breeding grey seals. Can J Zool. 71(11):2211–2219. [accessed 2016 Jan 25]. http://www.nrcresearchpress.com/doi/abs/10. 1139/z93-310.
Au WWL. 1993. The Sonar of Dolphins. New York: Springer-Verlag. [accessed 2020 Apr 28].
https://www.springer.com/gp/book/9780387978352.
Au WWL, Hastings MC. 2009. Principles of marine bioacoustics. New York, NY: Springer Science & Business Media.
Bennet-Clark HC. 1975. Sound production in insects. Sci Prog. 62(246):263–283. Publisher: Sage Publications, Ltd.
Boness DJ, James H. 1979. Reproductive behaviour of the grey seal (Halichoerus grypus) on Sable Island, Nova Scotia. J Zool. 188(4):477–500. eprint: https://zslpublications.onlinelibrary.wiley. com/doi/pdf/10.1111/j.1469-7998.1979.tb03430.x; [accessed 2020 May 5]. https://zslpublica tions.onlinelibrary.wiley.com/doi/abs/10.1111/j.1469-7998.1979.tb03430.x.
Caudron AK, Kondakov AA, Siryanov SV. 1998. Acoustic structure and individual variation of grey seal (Halichoerus Grypus) pup calls. J Mar Biol Assoc UK. 78(2):651–658.
Elemans CPH, Rasmussen JH, Herbst CT, During DN, Zollinger SA, Brumm H, Srivastava K, Svane N, Ding M, Larsen ON, et al. 2015. Universal mechanisms of sound production and control in birds and mammals. Nat Commun. 6(1):1–13.Number: 1 Publisher: Nature Publishing Group; [accessed 2020 Apr 22]. https://www.nature.com/articles/ncomms9978. Fitch WT, Suthers RA. 2016. Vertebrate vocal production: an introductory overview. In:
Suthers RA, Fitch WT, Fay RR, Popper AN, editors. Vertebrate sound production and acoustic communication. Cham: Springer International Publishing; p. 1–18. Springer Handbook of Auditory Research.
Hanggi EB, Schusterman RJ. 1994. Underwater acoustic displays and individual variation in male harbour seals, Phoca vitulina. Anim Behav. 48(6):1275–1283. [accessed 2020 May 2]. http:// www.sciencedirect.com/science/article/pii/S0003347284713637.
Hocking DP, Burville B, Parker WMG, Evans AR, Park T, Marx FG. 2020. Percussive underwater signaling in wild gray seals. Mar Mam Sci. 36(2):728–732. eprint: https://onlinelibrary.wiley.
com/doi/pdf/10.1111/mms.12666; [accessed 2020 Apr 29]. https://onlinelibrary.wiley.com/doi/ abs/10.1111/mms.12666.
Leighton TG. 2012. How can humans, in air, hear sound generated underwater (and can goldfish hear their owners talking)? J Acoust Soc Am. 131(3):2539–2542. Publisher: Acoustical Society of America; [accessed 2020 Apr 28]. https://asa.scitation.org/doi/10.1121/1.3681137.
Martin K, Tucker MA, Rogers TL. 2017. Does size matter? Examining the drivers of mammalian vocalizations. Evolution. 71(2):249–260. eprint: https://onlinelibrary.wiley.com/doi/pdf/10. 1111/evo.13128; [accessed 2020 Mar 14]. https://onlinelibrary.wiley.com/doi/abs/10.1111/evo. 13128.
McCulloch S, Boness DJ. 2000. Mother-pup vocal recognition in the grey seal (Halichoerus grypus) of Sable Island, Nova Scotia, Canada. J Zool. 251(4):449–455. [accessed 2016 Jan 25].
http://onlinelibrary.wiley.com/doi/10.1111/j.1469-7998.2000.tb00800.x/abstract.11.
McCulloch S, Pomeroy PP, Slater PJ. 1999. Individually distinctive pup vocalizations fail to prevent allo-suckling in grey seals. Can J Zool. 77(5):716–723. [accessed 2016 Jan 25]. http:// www.nrcresearchpress.com/doi/abs/10.1139/z99-023.
Miksis-Olds JL, Van Opzeeland IC, Van Parijs SM, Jones J. 2016. Pinniped sounds in the Polar Oceans. In: Au WWL, Lammers MO, editors. Listening in the Ocean. New York (NY): Springer; p. 257–308. Modern Acoustics and Signal Processing.
Narins PM, Feng AS, Fay RR. 2006. Hearing and sound communication in amphibians. New York, NY: Springer Science & Business Media. Google-Books-ID: UPJcyY7TuiIC.
Nowak L. 2020a. S2 underwater vocalizations of grey seals; Mar. [accessed 2020 May 6]. https:// figshare.com/articles/S2_underwater_vocalizations_of_grey_seals/12033834.
Nowak L. 2020b. Underwater vocalizations of grey seals - video with synchronized underwater audio recordings; Mar. [accessed 2020 May 6]. https://figshare.com/articles/Underwater_ vocalizations_of_grey_seals_-_video_with_synchronized_underwater_audio_recordings/ 12033849.
Nowak LJ. 2016. Underwater communication for divers. Scubaphone inspired by Marine Mammal Communication. Sea Technol. 57(12):26-+.
Nowak LJ. 2020c. S1 surface vocalizations of grey seals; Mar. [accessed 2020 May 6]. https:// figshare.com/articles/S1_surface_vocalizations_of_grey_seals/12033828.
Nowak LJ. 2020d. S1 underwater vocalizations of grey seals; Mar. [accessed 2020 May 6]. https:// figshare.com/articles/S1_underwater_vocalizations_of_grey_seals/12033807.
Nowak LJ. 2020e. S3 underwater vocalizations of grey seals; Mar. [accessed 2020 May 6]. https:// figshare.com/articles/S3_underwater_vocalizations_of_grey_seals/12033837.
Nowak LJ, Zielinski TG. 2012. Acoustic radiation of vibrating plate structures submerged in water. Hydroacoustics. 15:163–170. [accessed 2015 Mar 9]. http://yadda.icm.edu.pl/yadda/element/ bwmeta1.element.baztech-article-BWM1-0010-0019.
Nowak LJ, Zielinski TG. 2013. Modal sensitivity and selectivity of small, rectangle-shaped piezo-electric transducers used as sensors and actuators in active vibroacoustic control systems. J Low Freq Noise Vibr Act Control. 32(4):253–272.
Oliver GW. 1978. Navigation in mazes by a grey seal, Halichoer Us Gr Yp Us (Fabricius). Behaviour. 67(1):97–114. [accessed 2016 Jan 25]. http://booksandjournals.brillonline.com/con tent/journals/10.1163/156853978x00279.
Sauve CC, Beauplet G, Hammill MO, Charrier I. 2015. Acoustic analysis of airborne, underwater, and amphibious mother attraction calls by wild harbor seal pups (Phoca vit- ulina). J Mammal. 96(3):591–602. [accessed 2016 Jan 25]. http://jmammal.oxfordjournals.org/content/96/3/591. Schevill WE, Watkins WA, Ray C. 1963. Underwater sounds of pinnipeds. Science. 141
(3575):50–53. [accessed 2016 Jan 25]. http://science.sciencemag.org/content/141/3575/50. Schneider J. 1974. Description and probable behavioral significance of grey seal (Halichoerus
grypus) vocalizations [master’s thesis]. University of Rhode Island.
Schusterman R. 1978. Vocal communication in pinnipeds. Behavior of captive wild animals. Chicago: Nelson-Hall. p. 247–308.
Schusterman RJ, Kastak D, Levenson DH, Reichmuth CJ, Southall BL. 2000. Why pinnipeds don’t echolocate. J Acoust Soc Am. 107(4):2256–2264. [accessed 2016 Jan 25]. http://scitation.aip.org/ content/asa/journal/jasa/107/4/10.1121/1.428506.
Small LH. 2019. Fundamentals of phonetics: a practical guide for students. 5th ed. Boston: Pearson. Titze Ingo R. 1980. Comments on the myoelastic - aerodynamic theory of phonation. J Speech
Lang Hear Res. 23(3):495–510. Publisher: American Speech-Language-Hearing Association; [accessed 2020 Apr 22]. https://pubs.asha.org/doi/10.1044/jshr.2303.495.
