Measurements of underwater sound are still scarce in the rapidly changing Arctic. Tele-seismically detectable glacial earthquakes caused by iceberg calving have been known for nearly two decades but their underwater sound levels remain undocumented. Here, we present near-source underwater sound records from a kilometer-scale iceberg calving associated with a glacial earthquake. Records were obtained using an ocean-bottom lander deployed near the calving front of a Greenlandic tidewater Bowdoin Glacier in July 2019. An underwater-detonation-like signal with an overall duration of 30 min and two major phases owing to iceberg detachment and disintegration corresponded to extreme source sound levels (225 ± 10 dBp2p re 1 μPa) and acoustic energy on the order of 108–10 J or 0.1–7.6 tonnes TNT-equivalent. Our estimates and comparison with other anthropogenic and natural sources suggest that this type of geophysical event is among the loudest sounds in the Arctic. Such high sound levels are important for estimating the noise budget of the ocean and possible impacts on endemic Arctic species exposed to such sounds. The sound of calving may cause direct mechanical damage to the hearing of marine mammals such as narwhals and seals present in the glacial fjord.
Narwhals (Monodon monoceros) are attracted to glacial fjords during the Arctic summer, the reasons for which are unclear (Heide-Jørgensen et al., 2020; Laidre et al., 2016). These relatively small whales are endemic to the Arctic and famous for their spiral tusk. They rely on high-frequency vocalization for echolocation, foraging, and communication. Most narwhal sounds have high frequencies of >10 kHz, even extending into the ultrasonic range above the environmental noise of a glacial fjord (Podolskiy and Sugiyama, 2020), as might be expected for evolution-optimized use of the local soundscape (Boebel et al., 2016). Details of narwhal activity near the calving fronts of glaciers are poorly documented, but it is remarkable that these marine mammals seem to spend summer in one of the noisiest areas of the ocean (Pettit et al., 2015; Podolskiy and Sugiyama, 2020). Due to its remoteness and difficulty of access, the acoustic habitat of narwhals has barely been characterized, limiting our understanding of the noise field within which these species live. Only relatively recently has it been recognized that autonomous passive acoustic recorders are useful in the study of such glacial processes as iceberg calving and subaqueous melting (e.g., Deane et al., 2019; Schulz et al., 2008).
Listening in the ocean and the Arctic is not novel, beginning in the early 1950s during the Cold War (Au and Lammers, 2016), and several reviews are available (e.g., Howe et al., 2019; Mikhalevsky et al., 2015; PAME, 2019). However, on the one hand, previous knowledge of the ever-changing ocean is becoming obsolete due to sudden climate change, which is modifying sound-speed profiles and eliminating sea ice (Howe et al., 2019; Worcester et al., 2020). On the other hand, climate change has motivated multi-purpose acoustic observations, which are particularly important due to: (1) increasing anthropogenic noise (Duarte et al., 2021; Jones, 2019); (2) the scarcity of studies of underwater noise and its impact on marine mammals in Baffin Bay and the Arctic (PAME, 2019); and (3) poor understanding of glacial-fjord acoustics (e.g., ambient noise levels, source types and mechanisms, and sound propagation), which is the subject of this study.
To learn more about the underwater sounds of both glacial fjords and narwhals, we analyzed records made by an autonomous bottom-mounted hydrophone in a narwhal summering ground in a glacial fjord in northwest Greenland, representing the closest underwater observations made to date to a marine-terminating glacier in Greenland, and documented a major calving event. These near-source records provide direct and indirect insights into underwater noise and marine life in the Arctic, which nowadays is exposed more frequently to such major acoustic phenomena, particularly since 2010 in west Greenland (Sergeant et al., 2019).
II. STUDY SITE AND METHODS
An ocean-bottom seismometer (OBS) with a hydrophone and thermometer was deployed at ∼243 m depth in front of tidewater Bowdoin Glacier in July 2019, 640 m from the center of the calving front (Fig. 1). The glacier is also known as Kangerluarsuup Sermia in Greenlandic; it is ∼3 km wide and extends 250 m below the waterline at the center of the calving front (more than ∼90% of the full ice thickness) (Sugiyama et al., 2015). The glacier flows into the Bowdoin Fjord which spans ∼18 km, is 3–5 km wide, and is up to 500 m deep. The fjord is a tributary of the larger Inglefield Bredning Fjord (Podolskiy and Sugiyama, 2020). Technical details of this multi-purpose experiment and comprehensive site descriptions were provided previously (Podolskiy et al., 2021b, 2021c). Here, we focus on continuous hydroacoustic data collected during a major full-depth calving event on July 29, 2019. This energetic event is particularly interesting because it generated a glacial earthquake detectable by seismometers up to 500 km away in Greenland and Canada (Podolskiy et al., 2021b, 2021c).
Data were collected by a SoundTrap STD300 Digital Sound Recorder (Ocean Instruments, New Zealand) at a sampling rate of 96 kHz with 16-bit resolution. The self-noise floor of the hydrophone was <34 dB re 1 μPa above 2 kHz, with a working bandwidth of 20 Hz–60 kHz. We converted sound data to acoustic pressure (using the calibration sheet from the manufacturer) and explored it using basic signal-processing tools including filtering and fast Fourier transform (FFT) methods. For consistency with the commonly used acoustic terms, we followed Erbe (2011) and Merchant et al. (2015). Although the duration of our deployment was relatively short (16 days), a drift in the internal clock within the recorder was possible as in most similar devices (Au and Lammers, 2016). However, the lack of GPS timing information is not considered a major issue because the results presented here are not centered on absolute time with second-scale time precision.
III. RESULTS AND DISCUSSION
A. Iceberg calving and disintegration sounds
Waveforms recorded during the calving event of July 29, 2019, are provided in Fig. 2(a) and a multimedia audio file (Mm. 1), showing ocean pressure perturbations relative to background hydrostatic pressure. From time-lapse images (Fig. 1) and earlier seismic analysis (Podolskiy et al., 2021b, 2021c), we understand that this event had two major phases, 1 and 2, beginning in a nearly ice-free fjord: Phase 1, during detachment and capsizing of the iceberg (03:45–03:50 UTC); and Phase 2, during disintegration of the main iceberg directly above the OBS station (04:02–04:07 UTC). Both episodes are clearly visible in the waveforms as explosion-like impulsive sounds with pressure amplitudes (zero-to-peak) exceeding 625 Pa [Fig. 2(a)] and clipping the sensor. The total duration of the elevated noise was approximately 30 min.
The temporal change in the partition of power by frequency band is illustrated by cumulative sums of pressure squared, normalized to unity and proportional to the cumulative acoustic energy [Fig. 2(a)]. Two separate bursts occurred at frequencies below 5 kHz. Due to the emergent nature of each tremor-like burst, their durations were ∼5 min, with the most energetic radiation of sound lasting 3 and 2 min for phases 1 and 2, respectively (i.e., defined by the 90% energy envelope). This duration is similar to that of an iceberg capsizing event recorded on July 8, 2017, at Bowdoin Glacier (Podolskiy, 2018) and roughly consistent with characteristic timescales of major iceberg capsizes in Greenland (Sergeant et al., 2019).
Interestingly, the highest-frequency (>5 kHz) sound was not fully contained by the two bursts. Rather, it was initiated by the first burst and emitted continuously between bursts (i.e., while the ice was drifting towards the OBS station) and after the second burst. This can be seen in the spectrogram as a step after the first burst with frequencies ∼5 kHz [Fig. 2(a)].
Overall, the event was broadband, spanning the entire infrasound to ultrasound range [Fig. 2(b)]. The greatest increase in power, at 20–35 dB above preceding noise background levels, was observed at frequencies below 1 kHz. In the ultrasonic range, calving induced an ∼18 dB increase in power. Taking into account the dominant period of at least 25 s detected by the seismometer co-located with the hydrophone (Podolskiy et al., 2021b), the bandwidth of the source was at least 20.2 octaves (upper bound limited by the Nyquist frequency of the hydrophone). Finally, we note that some minor peaks near 1.5 kHz could be seen most clearly in the background noise spectrum [Fig. 2(b)], in the band associated with a submarine melt by previous studies (Deane et al., 2014; Glowacki et al., 2018; Pettit et al., 2015).
B. Narwhal acoustic presence
During the OBS campaign, the presence of narwhals in Bowdoin Fjord was confirmed directly by observation only early in the morning of 28 July 2019, more than 23 h before the calving event (Podolskiy and Sugiyama, 2020).
Visual inspection of arbitrary chosen audio segments from 28 July reveals narwhal sounds, such as quasi-regular high-frequency (>20 kHz) echolocation clicks repeated with median inter-click intervals of ∼122 ms [Fig. 3(a)]. Narwhals are known among cetaceans for their very high source levels of vocalisation. The received sound pressure of ∼50 Pa corresponds to a range from 15 to 325 m between the OBS and a narwhal (the large range is due to uncertainty in beam directivity relative to the receiver; Podolskiy and Sugiyama, 2020). Narwhals thus closely approached the calving front, as confirmed earlier visually (Podolskiy and Sugiyama, 2020).
Closer inspection of the calving waveforms [Fig. 2(a)] confirmed the presence of narwhal vocalizations before and at the onset of iceberg calving (Fig. 3). Examples [Fig. 3(b)] include a mid-frequency (<25 kHz) whistle with overtones known to be used for communication, and a click burst with inter-click intervals decreasing from >100 to 7.5 ms, presumably during foraging (Podolskiy and Sugiyama, 2020).
These observations indicate the presence of narwhals during the calving event. More comprehensive analysis of OBS hydroacoustic data for the acoustic presence of narwhals is beyond the scope of this paper and will be published elsewhere. Regarding other marine mammals, seals were frequently seen near the calving front of Bowdoin Glacier in a surface expression of a subglacial discharge plume, and local Inuit come to the fjord to hunt them (Podolskiy and Sugiyama, 2020).
C. Loudness of iceberg calving
Fifty years ago, an eminent hydroacoustician, Urick (1971), wrote that “Icebergs seem to have escaped serious attention as sources of underwater sound.” More recently, some studies have been undertaken concerning iceberg sounds in the Arctic and Antarctic, with large distances from floating ice shelves and icebergs (Dziak et al., 2013; Dziak et al., 2019; Matsumoto et al., 2014, and references therein), relatively small icebergs during serac collapse at grounded glaciers (Glowacki and Deane, 2020; Köhler et al., 2019), or free-floating icebergs (Glowacki et al., 2018; MacAyeal et al., 2008; Urick, 1971).
Studies of large Antarctic icebergs (>10 km long) have repeatedly reported so-called “iceberg singing”, with iceberg harmonic tremor having multiple overtones (Dziak et al., 2013; MacAyeal et al., 2008). No harmonic tremor was observed in the studied case, which indicates a lack of prolonged ice-to-ice contact or iceberg grounding, consistent with a disintegration process (Dziak et al., 2013).
The event reported here produced outstandingly high median noise levels, exceeding 120 dB and reaching 140 dB (re 1 μPa2 Hz−1) at frequencies below 200 Hz [Figs. 2(a) and 2(b)]. We therefore speculate that calving events associated with glacial earthquakes produce one of the loudest and broadest natural sounds in the Arctic.
However, it is not a simple matter to directly compare received acoustic energy to source levels of biological, anthropogenic, and environmental sounds reported in previous non-Arctic and Arctic reviews (e.g., Erbe, 2011; Jones, 2019; PAME, 2019), for the following three reasons:
The language used in the field of hydroacoustics is in disarray because the commonly used pseudo-unit of “dB” does not represent any physical metric (there are at least nine different ways of relating it to sound pressure, μPa), and it usually appears with incomplete description (Boebel et al., 2016). This leads to confusion when reproduction of summary plots apparently continues with difficult-to-trace original literature sources, lost captions, and modified units.
Back-calculated source levels (i.e., accounting for acoustic transmission loss) depends on many assumptions related to the estimation of sound propagation from a given source to a receiver, including ray-launch angles, spreading and absorption losses, sound-speed profiles (and their smoothing), boundary conditions (air–water, ice–water, and water–floor interfaces), reflection and scattering characteristics, and interface roughness (Erbe, 2011; Zeh et al., 2019). Furthermore, it is common for back-calculation procedures not to be described explicitly, with only results being provided (Dziak et al., 2013).
The amplitudes of continuous (e.g., wind, ship) and impulsive sounds (e.g., echolocation click, seismic airgun) are measured in different ways (Merchant et al., 2015).
Considering the impulsive nature of the calving signal [Fig. 2(a)], the amplitude of the sound was determined using pressure waveforms rather than frequency spectra (Merchant et al., 2015). Our frame of reference is therefore peak-to-peak sound pressure expressed in dB re 1 μPa, denoted as dBp2p [other metrics, such as root-mean-square (rms), median, and zero-to-peak are not always comparable; PAME, 2019]. Based on this metric and the full-bandwidth signal of Phase 1 of the calving event (i.e., before the newly formed iceberg could drift a significant distance from the terminus), the received sound pressure level was 182 dBp2p. Measured sound-pressure levels as a function of frequency range are provided in Fig. 4(a). These levels were quantified using one-octave-wide frequency bands with the corresponding center frequencies computed as the geometric means. Figure 4(a) shows pressures above ∼160 dBp2p at frequencies as high as 20 kHz. At such frequencies pressures are lower due to physical processes (e.g., increased absorption, scattering), and instrument sampling (amplitude losses near the Nyquist frequency).
1. Back-calculating source levels of calving sound
According to Sentinel satellite imagery acquired on July 27 and 29, 2019 (Fig. 1), the horizontal dimensions of the iceberg were approximately 1050 m by 110 m. Assuming a full ice thickness of 250 m (justified by the presence of icebergs with debris-laden basal ice after calving; Podolskiy et al., 2021b), yields 2.9 m3 (∼0.03 km3) as the upper limit on ice volume. This estimate exceeds the maximum volume studied by Glowacki and Deane (2020) ( m3) by as much as three orders of magnitude and remains at the lower end of iceberg volumes investigated by Sergeant et al. (2019). Nevertheless, the corresponding cross-sectional area of fracture would be around 27 500 m2, and representing it as a point-source (instead of a plane source) could be problematic. However, according to numerical simulations and laboratory experiments on glacial-earthquake source mechanisms (Murray et al., 2015; Sergeant et al., 2019), the most powerful sound energy is emitted by the impact between the iceberg edge and the newly formed ice cliff. We note that such a zone is at least 100 m farther from the receiver than the shortest horizontal distance to the original calving front (i.e., at the range of 740 m instead of 640 m). For a two-dimensional (2-D) ice block, such contact can be represented as a point-source. For a three-dimensional (3-D) iceberg, such contact might well be represented as a line source of a given length L, which the across-glacier length of the iceberg will limit. The maximum physically possible length of the iceberg, in its turn, is determined by the width of the glacier terminus (∼2.8 km). Sound level reduction (dB) for a finite line source can be approximated using Eq. (10) of Maekawa (1970),
where d is the distance to the source of length L. According to it, the sound level decreases nearly cylindrically at short distances and nearly spherically at distances larger than L. Homogeneous partition of the impact force along the full length of the iceberg is unlikely and the strongest impact zone could be shorter. With decreasing L, the sound level reduction starts to remind spherical spreading from a point-source. This implies that the finite-line-source estimate can be considered the lowest level bound and spherical spreading model as the highest level bound.
To back-calculate the source level (SL) of calving from the received sound pressure level (RL) measured with a bottom-mounted hydrophone in shallow water, we also made the following assumptions to calculate transmission loss, TL (Erbe, 2011):
Sound propagates spherically near the source (–20 dB/decade) and cylindrically at a distance greater than the water depth, bound by the fjord floor and sea surface (–10 dB/decade). For reference, we included spreading loss separately for spherical and cylindrical cases, which may be seen as over- and under-estimate bounds, respectively.
For dominant frequencies below 10 kHz and for the short distances considered here, any frequency-dependent absorption loss (due to the molecular relaxation of boric acid and magnesium sulphate) is negligible. Additional losses could be produced by air bubbles entrained during calving (Glowacki, 2020). However, the contribution of this “bubble-curtain” effect to the absorption of sound in seawater during full-depth buoyancy-driven calving is unknown and therefore difficult to consider.
Source levels are commonly estimated in a relatively simple manner, facilitating literature comparisons (Erbe, 2011; Kyhn et al., 2019). More sophisticated modelling of sound propagation in a glacier fjord (e.g., Glowacki and Deane, 2020; Zeh et al., 2019) requires knowledge of the sound-speed profile, bathymetry, and conditions at the sea-floor/surface, which were not readily available for this study.
The argument that the most energetic source could be at the impact zone between the main iceberg and the newly formed ice cliff, “curtained” from the receiver by the iceberg implies that the source levels could be underestimated. An upward-refracting sound-speed profile is expected in the Arctic (Worcester et al., 2020), so further reductions in received levels are possible but expected to be minor due to the short distance to the source (Glowacki et al., 2016). Furthermore, the sensor was clipped.
2. Calving versus other sound sources
For a full bandwidth, our estimated source level is at least 215 dBp2p for a finite-length source (SLfl) [Fig. 4(b)] with an upper bound of 235 dBp2p (conservatively bounded by the combined spreading model for a point source, SLp).
Pile driving (216 dBp2p), seismic airgun arrays (211–258 dBp2p), and underwater blasting (268 dBp2p) may be close to or exceed the source levels of the calving event reported here (Kyhn et al., 2019; PAME, 2019). Reviews of various source levels suggest that blasting is the loudest anthropogenic source, with the sound of calving being louder than any vessels (including supertankers and icebreakers), drilling rigs, and naval mid-frequency sonar (Jones, 2019; PAME, 2019). This is consistent with previous reports indicating that iceberg-breakup sounds from Antarctica are detectable at basin-scale ranges (e.g., at the Equator) with source sound levels of ∼220 dBrms, significantly greater than anthropogenic sources (Dziak et al., 2013).
This loudness also places calving in Greenland in the category of the most powerful natural sounds possible in the ocean, together with sea-floor volcanic eruptions, undersea earthquakes (Jones, 2019; Pettit et al., 2015), and possibly lightning strikes. Due to complex geometry and bathymetry of Bowdoin Fjord and Inglefield Bredning Fjord it is a part of, estimation of the detection range of this event is not straightforward and can be frequency-dependent (e.g., shallower regions would cutoff low frequencies). Assuming high background noise levels of 140 dBp2p and taking into account steep transmission loss owing to spherical spreading [Fig. 4(b)], we can expect distances of at least ∼70 km (and greater as noise levels drop in ice-free open water). This corresponds to the intersection between the x axis and the dash-dot line in Fig. 4(b).
3. Sound exposure level and acoustic energy
As an additional metric, the acoustic energy of the signal (or total energy flux) was estimated from the cumulative sound exposure level (SEL) (Merchant et al., 2015). For this, we integrated the squared pressure over the duration of the calving signal, T, defined by the 90% energy window of Phase 1 (Merchant et al., 2015), subtracted background noise energy [computed from a time segment of the same duration preceding the calving event (Erbe, 2011; Glowacki and Deane, 2020)], and expressed the result in dB (re 1 μPa2s). This yields a received SEL of 229.12 dB90% over the 90% energy window; i.e., 3 min (re 1 μPa2s).
The corresponding received acoustic energy (with background noise energy removed) is also commonly expressed in Joules (Glowacki and Deane, 2020):
Where ρ is the density of sea water (1027 kg m−3; Podolskiy et al., 2021a) and c is the speed of sound (1450 m s−1; Glowacki and Deane, 2020). This gives 6.9 J for the full-band pressure signal, or 3.5 J for the low-pass filtered pressure signal (<100 Hz; Glowacki and Deane, 2020), which, without consideration of transmission loss, is two orders of magnitude greater than the source levels of regular calving events in Svalbard (Glowacki and Deane, 2020).
Regarding transmission losses, the total sound energy of the calving can be estimated following Glowacki and Deane (2020) as
For total TL, Glowacki and Deane (2020) also considered a possible contribution of terminus-reflected noise to the total acoustic energy, which has the opposite sign to propagation loss. For the closest receiver range of 700 m and depth of 40 m, they expect a maximum increase in sound energy of 3 dB as a “worse-case” scenario. Adopting such a value, and using the above TLs [Fig. 4(b)], the total energy loss is in the range from –29.7 to –49.6 dB. This translates to a loss of 3–5 orders of magnitude in acoustic energy of the source. For the frequencies below 100 Hz, the total acoustic energy of the calving event is therefore 3 J, with a power of 2–176 MW. The released energy corresponds to 0.1–7.6 tonnes TNT-equivalent. Using the above TLs, we estimate that the SEL at the source was 258.8–278.7 dB (re 1 μPa2s).
In comparison, for a 20-min-long disintegration of a tabular iceberg ( 25 km2) in Antarctica, Dziak et al. (2013) reported a total energy flux of 252 dB re 1 μPa2s, corresponding to 6.9 J (0.06 MW). To compute the total acoustic energy in Joules, they used a wider frequency band (1–440 Hz) and a factor of 2π. Repeating our calculation using the same properties did not change the orders of magnitude 108–10 J).
D. Biological significance
The risk of auditory injury (permanent threshold shifts, PTS) induced by explosive and impulsive sounds was estimated following Kyhn et al. (2019) and referring to the technical guidance of the US National Marine Fisheries Service (NMFS, 2016). NMFS (2016) is classified as a Highly Influential Scientific Assessment and provides the most state-of-the-art knowledge regarding the impact of anthropogenic sounds on marine mammal hearing. No document of equivalent quality is available specifically for acoustic exposure to natural sounds. For impulsive sounds, NMFS (2016) recommends dual metrics based either on peak sound pressure level (Sec. III D 1) or M-weighted cumulative sound exposure level (SEL M-weighted; Sec. III D 2, with the lowest being applied.
1. Peak-pressure acoustic threshold
Regarding peak pressure, the NMFS defines PTS thresholds of 218 and 230 dBzero-to-peak for groups of Phocid pinnipeds (e.g., ringed seal) and mid-frequency (MF) cetaceans, respectively. MF-cetaceans include beluga whales, considered here as surrogate species as the closest relatives of narwhals because no empirical hearing thresholds (i.e., audiograms) have been measured for narwhals living only in the wild (PAME, 2019). In terms of peak-to-peak pressure (i.e., +6 dB), these NMFS thresholds correspond to 224 and 236 dBp2p, respectively. As seen from Fig. 4(b), it is highly unlikely that seals with the lowest threshold were exposed to dangerous sound levels (e.g., to exceed the threshold a seal would have to be closer than m to the hypothetical point source).
Similarly, even if temporary threshold shift (TTS) is considered with lower onset thresholds of 218 and 230 dBp2p for pinnipeds and cetaceans, respectively (Erbe, 2011), such levels were reached only within an unrealistic range of <10 m from the point source. Therefore, from the perspective of peak-pressure metrics alone, and notwithstanding the high peak sound pressure of calving, it is very unlikely that it posed a threat to the hearing of marine mammals such as seals and narwhals in Bowdoin Fjord.
2. SEL M-weighted acoustic threshold
The M-weighted SEL accounts for differences in sound perception by different species (NMFS, 2016). M-weighting is more elaborate than peak-pressure and corresponds to the auditory frequency weightings of sounds, emphasizing frequencies with potential auditory effects by adjusting the sound with a curve appropriate for each hearing group.
Spectra of the calving event (Phase 1) weighted according to underwater auditory weighting functions recommended by NMFS (2016) are shown in Fig. 5 for hearing groups of MF-cetaceans and Phocid pinnipeds. It is clear that the low frequencies of calving spectra become less important, especially for MF-cetaceans with lower sensitivity than seals to low frequencies.
The net effect of frequency weighting was investigated following Kyhn et al. (2019) to calculate the difference between unweighted sound pressure levels and corresponding M-weighted levels using third-octave frequency spectra (Fig. 5). In each case, the total sound pressure level was represented by the sum of third-octave levels (TOLs) across all frequencies (L = 10 log ). For MF-cetaceans and Phocid pinnipeds, the net effects of M-weighting were –28.2 dB and –15.8 dB, respectively. These weighting terms were then applied to the received SEL computed from the calving signal and combined with estimated transmission losses, as shown in Fig. 6.
The received M-weighted SELs of calving are similar to source M-weighted SELs of a seismic airgun pulse (Kyhn et al., 2019). For a received M-weighted SEL, NMFS (2016) defines the PTS onset threshold of 185 dB re 1 μPa2s for impulsive sounds, for both MF-cetaceans and Phocid pinnipeds. This implies that the threshold is exceeded for both groups within∼3 km of the source (Fig. 6).
In addition to more injurious impulsive sounds with high peak sound pressure and rapid rise, NMFS (2016) provides PTS onset thresholds for non-impulsive sounds, which lack a high peak pressure and rapid rise; specifically 198 and 201 dB (re 1 μPa2s) for MF-cetaceans and Phocid pinnipeds, respectively. Even when it is difficult to strictly categorize the observed calving signal with multiple high-pressure peaks as non-impulsive, we note that using such higher thresholds for narwhals reduces the impact range to <600 m. For any adopted sound classification, the impact on seals is expected to be greater due to higher M-weighted SEL, as seals are more sensitive to low-frequency sounds.
Overall, this suggests that due to the high cumulative sound exposure level, calving sound may lead to the onset of PTS and thus hearing loss in narwhals and seals present within hundreds to a few thousand meters of the calving front.
3. Masking, behaviour, and other effects
Temporary interference with calving noise by marine-mammal vocalizations and the masking of sounds used by them (i.e., decrease in signal-to-noise ratio) are also possible. Characteristic narwhal sounds (Fig. 3) are mainly echolocation clicks and communication signals, such as whistles with lower frequency and amplitudes than clicks, with communication likely to be temporarily affected by masking with calving noise. However, the degree of masking is a complex issue that depends on the context and structure of the noise and signal (Erbe, 2011). For example, Erbe and Farmer (1998) experimentally studied the effects of masking on beluga whales and concluded that natural sea-ice cracking due to thermal stress and pressure variations had the lowest interference due to its irregular temporal structure, unlike the continuous noise of an icebreaker. However, the noise levels of sea-ice cracking are much lower than that of calving. As a simple assessment of masking likelihood, we considered how the signal-to-noise ratio of narwhal signals would be reduced by calving if our receiver was substituted with a whale (Fig. 3). For this, we took the received amplitudes (A) of echolocation, foraging, and whistle signals, A = [50, 1.5, 0.1] Pa, respectively, and estimated the signal-to-noise-ratio, SNRdB, as (20 × log10(2A) – 155 dBp2p), where 155 dBp2p is the assumed noise level of calving at ultrasonic frequencies (Fig. 4a), yielding approximately [5, –25, –49] dB. This suggests that only echolocation clicks may have signals greater than noise (>0 dB), and that the generally weaker communication signals are more likely to be incomprehensible, as recognized in previous studies on belugas (e.g., Erbe and Farmer, 1998).
Behavioural responses of narwhals, known for their timid nature (Kyhn et al., 2019; Podolskiy and Sugiyama, 2020) and extreme auditory sensitivity (Heide-Jørgensen et al., 2021; Tervo et al., 2021), likely involve movement away from the calving front. This is difficult to verify acoustically because signal detection before and after calving can be strongly biased by the general increase in background noise.
Calving events of such large scale occur in Bowdoin Fjord only once every several weeks during summer (van Dongen et al., 2021). Smaller calving events are more frequent, averaging about ∼13 day−1 (Minowa et al., 2019). Such episodic increases in noise near the calving front are difficult to consider as being chronic, so there is likely to be recovery from hearing loss. If our suggestion of likely PTS onset is correct, there could be some benefits for marine mammals in taking this risk. For example, it is known that seals are attracted to subglacial discharge plumes for foraging, but for narwhals there is no such obvious conclusion. There may be other advantages, such as the masking of baleen whale mother–calf communication, preventing detection by orca (or killer whales) (Boebel et al., 2016); existence of stereotyped narwhal mother–calf calls were recently reported (Ames et al., 2021). Masking was also suggested for seals in glacial fjords (Pettit et al., 2015), but remains unverified. Sightings of killer whales in Baffin Bay may be increasing (Breed et al., 2017), with sightings near Qaanaaq, close to the study area (Oshima, 2019). The episodic deterrence of fish (e.g., Arctic cod, Polar cod) from the inner fjord may be a factor, just as birds are deterred by calving. However, how important such factors may be for foraging predators is unknown.
E. Elevated high-frequency background
Following the calving event (Phase 1), continuous radiation of high-frequency (>5 kHz) noise [Fig. 2(a)] was observed. For example, at 10–20 kHz, a cigar-shaped tremor-like signal lasting several minutes can be recognized [Fig. 2(a)]. Possible mechanisms for such broadband noise include moving/cracking/melting ice (Pettit et al., 2015) and/or sediment transport associated with bursts of current or deposition from an iceberg.
After iceberg detachment, the water surface was covered with ice debris (Podolskiy et al., 2021c), which continued to disintegrate, crack, and rub, and also substantially increase the surface area of ice in contact with water, explosively releasing pressurized air bubbles. With capsizing, there is a sudden movement of basal ice from cold water (the mean water temperature at the bottom of the fjord was C; Podolskiy et al., 2021c) and high pressure to warmer water and lower hydrostatic pressure. This corresponds to an increase in cracking and noise related to air-bubble release, with higher pressures suppressing such bubble sounds, which are more common in shallow, near-surface water (<20 m depth; Deane et al., 2019; Vishnu, 2020). The drop in hydrostatic pressure was rapid, at up to ∼13.6 kPa s−1, as it took < 3 min for the iceberg to capsize (in comparison, it took 3 min for our free-falling OBS to reach the fjord floor at deployment; Podolskiy et al., 2021c). The second burst had more ultrasound power (>20 kHz) than the first, as expected for the shorter distance between the receiver and the main iceberg (i.e., ∼240 m overhead at ∼04:05 versus 640 m horizontally at ∼03:48; Fig. 1).
There is, however, an alternative interpretation of the high-frequency background. Seabed ultrasonic noise from sediments in suspension was reported by Tregenza et al. (2016), who found that sandy seabed (contrary to mud) could complicate dolphin click detection through tonal ultrasonic interference. Experimental work by Thorne (1990, 1986), exploring the use of a hydrophone to monitor sediment transport, indicated that rigid-body collisions between sand particles can be a source of ultrasound, with particle diameter, D (m), being estimated from the acoustic spectrum (specifically, from the characteristic frequency, fc (Hz), using ). This relationship implies increasing frequency with decreasing particle size. We did not observe clear characteristic frequencies, but the elevated high-frequency noise of 15 to 25 kHz, corresponds to particle diameters of between 8 and 4 mm, respectively.
Coring near the deployment site indicates that the fjord floor is muddy (Podolskiy et al., 2021c), and to bring such sediments into suspension, local currents are required, which the rotational moment of calving could generate. This would correspond to a mass of agitated material coming very close to the sensor (for example, Thorne, 1986 measured amplitudes of up to 5–8 Pa from a range of 0.3 m). Another source of sedimentation could be the iceberg releasing sediment while drifting overhead. In neither case would agitated and precipitating materials occur with instantaneous calving signals, requiring some time to reach the vicinity of the receiver. The passage of such a burst would thus be delayed (for 5–10 min assuming extreme current speeds of 1–2 m s−1) and might account for the cigar-shaped tremor ∼10 min after the calving [Fig. 2(a)]. At about this time (04:00 UTC), a time-lapse camera (Fig. 1) showed that the main iceberg was overhead the sensor (Podolskiy et al., 2021c), with depositing particles and a short distance to the cracking iceberg being more straightforward explanations than the current burst.
In summary, we acknowledge that high-frequency sound production by sediment transport and ice cracking are difficult to distinguish, and in our opinion are one of the least explored realms of glacier-fjord acoustics, requiring more thorough studies. Due to the lack of a clear characteristic (tonal) frequency (Thorne, 1986; Tregenza et al., 2016), and simultaneous with the calving initiation of high-frequency emissions [Fig. 2(a)], we suggest that the broadband impulsive cracking from moving and melting ice generally saturates the noise field, with some sediment bursts occurring as transients.
IV. CONCLUDING REMARKS
We report extreme near-source records of underwater sound produced by a kilometer-scale iceberg calving event obtained with the first-of-its-kind OBS experiment in Greenland. This event generated a glacial earthquake detected by regional seismic stations up to 500 km away and occurred when narwhals were present in the fjord. The estimated source levels were among the highest ever recorded in the Arctic, with likely detrimental effects on the hearing of marine mammals in close proximity to the calving front. To our knowledge, this is the first quantitative attempt to assess the effects of iceberg sound on marine mammals.
A general lack of source level measurements is recognized as a major knowledge gap in the Arctic soundscape (PAME, 2019). Considering the recent recognition of a need to establish sustained ocean-acoustic and Greenland observations (Howe et al., 2019; Straneo et al., 2019), and the societal importance of the local ecosystems, particularly as hunting and fishing grounds (Podolskiy and Sugiyama, 2020; Sugiyama et al., 2021), we believe this study is timely in providing a foundation for further studies with bottom landers, hopefully inspiring more interest in the response of animals to iceberg calving.
We thank colleagues (I. Asaji, T. Ando, Y. Sakuragi, A. Mangeney, E. van Dogen, A. Bauder, and O. Castelnau) and local guides (T. Oshima, K. Petersen, Q. Aladaq, S. Aladaq, and R. Daorana) for their assistance during fieldwork. The Ministry for Nature, Environment and Justice (Government of Greenland) granted permission (C-19–37) for this fieldwork. This work and N.K. were supported by an Arctic Challenge for Sustainability research projects (ArCS and ArCS-II; JPMXD1300000000 and JPMXD1420318865, respectively), funded by the Ministry of Education, Culture, Sports, Science and Technology of Japan (MEXT), J-ARC Net, Grants-in-Aid for Scientific Research “KAKENHI” No. 18K18175, and the Second Earthquake and Volcano Hazards Observation and Research Program (Earthquake and Volcano Hazard Reduction Research funded by MEXT). We also thank the editor M. Badiey and the three anonymous reviewers for their help in improving this manuscript. Hydroacoustic data are publicly available through the Arctic Data archive System website (https://ads.nipr.ac.jp/dataset/; A20200108-002). The main calving signal is presented as multimedia audio file (Mm. 1). Standard functions of Matlab R2018b were used for analysis (https://mathworks.com/products/matlab.html). Audio files were inspected in publicly available software Raven Lite 2.0.0 (Cornell Lab of Ornithology).