Seasonal and diel patterns in singing activity of humpback whales migrating through Bermuda

Narganes Homfeldt, Tamara; Risch, Denise; Stevenson, Andrew; Henry, Lea-Anne

doi:10.3389/fmars.2022.941793

ORIGINAL RESEARCH article

Front. Mar. Sci., 13 September 2022

Sec. Deep-Sea Environments and Ecology

Volume 9 - 2022 | https://doi.org/10.3389/fmars.2022.941793

This article is part of the Research Topic Managing Deep-sea and Open Ocean Ecosystems at Ocean Basin Scale - Volume 2 View all 8 articles

Seasonal and diel patterns in singing activity of humpback whales migrating through Bermuda

Tamara Narganes Homfeldt^1,2

Denise Risch³

Andrew Stevenson⁴

Lea-Anne Henry^2*

¹Department of Migration, Max Planck Institute of Animal Behavior, Radolfzell, Germany
²School of GeoSciences, University of Edinburgh, Edinburgh, United Kingdom
³Scottish Association for Marine Science, Oban, United Kingdom
⁴Whales Bermuda, Pembroke, Bermuda

Humpback whales (Megaptera novaeangliae) produce song and non-song vocalisations, which allows their presence to be detected through passive acoustic monitoring. To determine the seasonal and diel acoustic presence and acoustic behaviour of humpback whales at the migratory stopover site off Bermuda, three hydrophones were deployed between March 2018 and April 2019 on Challenger Bank and the Bermuda platform. Song was the predominant vocalisation type encountered, with 65% of song recordings containing whale chorus and a clear seasonal trend of humpback whale occurrence in the spring and winter months from late December to mid-May. A strong diel pattern in singing activity was detected. Singing activity significantly increased at night relative to the daytime (p<0.01), whilst twilight periods were characterised by intermediate levels of singing. The song structure encountered in spring 2018 consisted of 18 units, 6 themes and 5 transitional phrases. The high occurrence of whale chorus and the strong seasonal and diel patterns of male humpback whale singing activity highlights the importance of Bermuda not just on their northward migration during spring, as described historically, but also on their southward migration during winter. Bermuda therefore constitutes a two-way migratory stopover site for humpback whales. The present study also provides Bermuda’s planning authorities with better constraints on the duration and intensity of anthropogenic activities in these waters.

Introduction

Humpback whales (Megaptera novaeangliae) are one of the large baleen whales best known for their extremely variable vocal behaviour. “Song of the Humpback Whale”, recorded off Bermuda in the 1950s by Frank Watlington, was the first recording of humpback whale song worldwide and initiated an era of humpback whale research off Bermuda throughout the 1960s and 1970s (Payne and Payne, 1985). While humpback whales produce unstructured non-song vocalisations year-round in different behavioural contexts (Silber, 1986; Dunlop et al., 2007; Rekdahl et al., 2015), humpback whale song is the most dominant vocal display of the species. Humpback whale song is displayed exclusively by males (Herman et al., 2013), and is thought to be a multi-message reproductive display (Murray et al., 2018) involved in both inter- (Smith et al., 2008) and intra-sexual interactions (Darling and Berube, 2001; Cholewiak et al., 2018b). However, the exact function of the song remains uncertain (Herman, 2017).

Humpback whale songs differ across ocean basins, due to the geographic isolation of the three recognised humpback whale subspecies (Baker et al., 1990; Bettridge et al., 2015; Cooke, 2018), and to some extent within an ocean basin, but are the same within a breeding population (Murray et al., 2012; Garland et al., 2013; Nikšić, 2014; Darling et al., 2019). Song is defined as a repetitive, stereotyped vocal display with a hierarchical structure (Payne and McVay, 1971; Cholewiak et al., 2013). A humpback whale “song” or “song cycle” is repeated for the duration of a “song session”, i.e., the time period an individual whale sings continuously, which has been shown to last up to 22 hours (Winn and Winn, 1978). A song can be subdivided into a hierarchical order of several distinct “themes”, each of which consists of repeated “phrases”, which in turn consist of a sequence of individual “units” (Payne and McVay, 1971; Cholewiak et al., 2013). Along with songs produced by bowhead whales (Balaena mysticetus), humpback whale song is considered the most complex (Janik, 2009; Stafford et al., 2018). A population’s song undergoes constant and progressive changes through time, a phenomenon referred to as song evolution (or song revolution in the case of sudden changes) (Winn and Winn, 1978; Payne and Payne, 1985; Noad et al., 2000; Garland et al., 2011; Garland et al., 2017).

North Atlantic humpback whales undertake extensive seasonal migrations between high latitude summer feeding grounds – off northern Norway and Iceland (referred to as eastern feeding grounds), as well as western Greenland, eastern Canada and the northeastern United States (referred to as western feeding grounds) – and low latitude winter breeding grounds around the West Indies and Cape Verde (Stevick et al., 2003; Reeves et al., 2004; Wenzel et al., 2009; Ruegg et al., 2013; Bettridge et al., 2015). However, fluke identification matches now suggest that the Caribbean breeding ground might be further subdivided, as humpback whales wintering in the southeast Caribbean are behaviourally distinct from those wintering in the northwest Caribbean, in two ways (Stevick et al., 2016). First, humpback whales winter in the northwestern Caribbean between January–April with peaks between February–March, while those wintering in the southeastern Caribbean do so a bit later between March–May with peaks in April (Charif et al., 2001; Stevick et al., 2003; Gandilhon, 2012; Stevick et al., 2018; Heenehan et al., 2019). Second, re-sightings of individuals revealed a strong tendency for southeastern Caribbean humpback whales to migrate to eastern North Atlantic feeding grounds, while whales from northwestern Caribbean breeding grounds tend to migrate to western feeding grounds (MacKay, 2015; Stevick et al., 2018), causing some genetical differentiation (partly due to feeding ground destination in humpback whales showing strong maternally-directed fidelity) (Baker et al., 1990; Palsboll et al., 1995; Weinrich, 1998). However, some individuals have been matched between Cape Verde and the southeastern Caribbean, as well as between the southeastern and northwestern Caribbean, demonstrating that the population units and boundaries are not as clear as previously thought (Stevick et al., 2016; MacKay et al., 2019). While the western North Atlantic humpback whale population has been increasing in recent years after the cessation of whaling, the Cape Verde population is still of considerable concern (Wenzel et al., 2020).

Although historically believed to exclusively occur on breeding grounds (Winn and Winn, 1978), humpback whale song has been increasingly recorded on their feeding grounds during the breeding season, which suggests that some males may not migrate at all but instead remain year-round in their feeding grounds (Mattila et al., 1987; Vu et al., 2012; Baumgartner et al., 2019; Magnúsdóttir and Lim, 2019; Kowarski et al., 2021; Tyarks et al., 2021). In the North Atlantic, humpback whales start singing in early autumn (around September) and continue singing through winter, stopping in late spring (around June) (Mattila et al., 1987; Vu et al., 2012; Kowarski et al., 2019; Kowarski et al., 2021). Transitions between song and “non-song” periods at the start and end of summer are dominated by “song fragments”, i.e., only a short part of the complete song is sung (Mattila et al., 1987; Vu et al., 2012; Kowarski et al., 2019; Kowarski et al., 2021). The seasonal singing behaviour displayed by males is thought to underlie a hormonally triggered physiological mechanism (Wright and Walsh, 2010; Vu et al., 2012), as the males’ testosterone levels are the lowest during the summer months and highest during the winter months, i.e., during the breeding season (Cates et al., 2019). Thus, song fragments could be the result of spring decreases and autumnal increases in testosterone levels (Cates et al., 2019; Kowarski et al., 2019).

Notably, most of what is known about North Atlantic humpback whales and their vocalisations has come from coastal studies on feeding and breeding grounds. Thus, their migration routes and mid-ocean behaviours, including vocalisations, with the exception of a few studies tracking individual whales with satellite tags (Kennedy et al., 2014; Kennedy and Clapham, 2017), remain vastly understudied or unknown (Reeves et al., 2004; MacKay, 2015; Kowarski et al., 2018). Bermuda, being an oceanic migratory stopover site for North Atlantic humpback whales on their northward migration (Payne and McVay, 1971; Stone et al., 1987; Stevenson and Stevick, 2009), provides a unique opportunity to study vocalisations of migrating humpback whales. Individuals observed off Bermuda have been re-sighted in the northwestern Caribbean breeding grounds and to a much lesser extent in the southeastern Caribbean, as well as, in all major feeding grounds (with the exception of Norway) (Stone et al., 1987; Beaudette et al., 2009; Stevenson, 2010; MacKay, 2015; Stevenson, unpublished data), but predominantly in western North Atlantic feeding grounds (Stone et al., 1987; Beaudette et al., 2009; Kennedy et al., 2014). Thus, the waters around Bermuda most likely represent an oceanic migratory stopover site between the northwestern Caribbean breeding grounds and higher latitude western feeding grounds. While in Bermuda, humpback whales have been observed to linger for several days whilst aggregating into large groups, accompanied by male singing (Payne and McVay, 1971; Payne and Payne, 1985; Stevenson, 2011), before continuing their northward migration.

Recordings of humpback whales off Bermuda led to the first formal description and definition of humpback whale song (Payne and McVay, 1971), which is now fundamental to the field of humpback whale song research. However, there has been no acoustic recording or analysis of whale vocalisations in Bermuda since 1976 (Winn and Winn, 1978; Payne and Payne, 1985) and these initial studies did not use long-term Passive Acoustic Monitoring (PAM) deployments that permit year-round data collection (under all weather conditions and overnight) of the marine soundscape and therefore year-round acoustic detection of vocal species like humpback whales (Johnson et al., 2009; Rafter et al., 2018). Thus, in stark contrast to a good baseline knowledge of the seasonal occurrence of humpback whale vocalisations from North Atlantic feeding and breeding grounds, the acoustic presence of humpback whales at their stopover site off Bermuda has not yet been analysed.

Such knowledge on the temporal presence of whales off Bermuda is urgently needed to address potential threats posed by increasing human activities. The Government of Bermuda has enacted some protection for humpback whales under its Fisheries Act 1972 (Government of Bermuda, 1978) and its Protected Species Act 2003 (Government of Bermuda, 2003; Minister of Health Seniors and Environment, 2016) and there are voluntary whale-watching guidelines (Department of Environment and Natural Resources, 2017). Bermuda’s Exclusive Economic Zone (EEZ) is also an important migratory corridor and stopover location for various cetacean species (Klatsky et al., 2007; Hallett, 2011; Hoyt, 2011) and was designated as a Marine Mammal Sanctuary but this designation only offers data-sharing opportunities and comes with no management or protection measures (NOAAGovernment of Bermuda, 2012). With Bermuda becoming popular as a megafauna “hotspot”, as a cruise destination and sites for various international sporting events, the tourism industry including the whale-watching industry (O’Connor et al., 2009) is anticipated to further develop in line with Bermuda’s six-year National Tourism Plan 2015–2023 (Bermuda Tourism Authority, 2019). New legislation under the Superyachts and Other Vessel (Miscellaneous) Act 2019 now also allows more large yachts (>24 meters length) to secure cruising and charter permits. Thus, growth in Bermuda’s tourism industry will increase vessel traffic of all kinds (cruise liners, freight, superyachts, whale-watching boats and smaller recreational craft).

Increased vessel traffic and marine tourism can have various negative impacts on humpback whales, from behavioural disturbance, increased stress levels and physical injuries, to disturbing their crucial auditory sensory system and communication (Au and Green, 2000; Cholewiak et al., 2018a; Fiori et al., 2020; Sprogis et al., 2020; Currie et al., 2021). Humpback whales produce low to mid frequency vocalisations (ranging from 0.01-28 kHz), but like all baleen whales most energy is produced in the lower frequencies (below 2 kHz), which can propagate across an entire ocean basin (Payne and Webb, 1971; Hannay et al., 2013; Cerchio et al., 2014; Huang et al., 2016; Cholewiak et al., 2018a; Davis et al., 2020). However, vessel-generated noise, the most prevalent anthropogenic underwater noise (Cato, 2014), overlaps in frequency (Clark et al., 2009; Rolland et al., 2012) and thus interferes with the acoustic detection of mysticetes non-song and song vocalisations (“masking”) (André, 2018; Dooling and Leek, 2018) and reduces the distance over which they are able to acoustically communicate (Clark et al., 2009; Rolland et al., 2012; Cato, 2014; Cholewiak et al., 2018a; Dunlop, 2019). Thus, to mitigate impacts of increased anthropogenic noise levels on humpback whales migrating through Bermuda, knowledge of spatiotemporal patterns of humpback whale presence and vocalisations needs to be gathered and integrated by ocean planners and authorities into planning and management scenarios and decisions for sustainable developments.

The present study is the first long-term PAM study of humpback whale vocalisations in Bermuda. The aim is to investigate their seasonal and diel acoustic presence and acoustic behaviour at this migratory stopover site. To facilitate future comparisons of song structures across the North Atlantic, as well as within Bermuda (determining inter-annual song variation), the song structure encountered in spring 2018 will be described in detail at the unit, phrase and theme level.

Materials and methods

Study area and acoustic data collection

Bermuda forms part of a small mid-ocean seamount chain of volcanic origin rising abruptly from the deep abyssal plain of the Sargasso Sea (Figure 1) (Stone et al., 1987; Vogt and Jung, 2007). Besides the topographic highs of the inhabited Bermuda platform, Bermuda’s EEZ has three large submerged seamounts, which are known for their high biodiversity: Bowditch Seamount, Challenger Bank (CB) and Plantagenet Bank (Figure 1) (Vogt and Jung, 2007; Hallett, 2011). CB and Sally Tucker (ST; located at the southwest edge of the Bermuda platform), which are 13 km apart, were chosen as recording sites for the present study (Figure 1).

FIGURE 1

Figure 1 Map of the study area, showing Bermuda (in yellow) and the three seamounts Bowditch Seamount, Challenger Bank and Plantagenet Bank. The three hydrophones were deployed on Challenger Bank (red rhombus) and Sally Tucker (red circle), which are 13 km apart. The map was created in QGIS 3.2 software (QGIS Development Team, 2016) using the coordinate reference system WGS 84, EPSG: 4326 and a world (source: Natural Earth) and bathymetry map (source: Open Digital Elevation Model, 2019).

Two Autonomous Multichannel Acoustic Recorders (AMAR G3A; JASCO Applied Sciences) equipped with M36-V35-100 omnidirectional hydrophones (-165 ± 4 dB re 1 V/μPa sensitivity) (Supplementary Figure 1) were deployed from 31 March to 6 September 2018 on CB at a water depth of 45.7 m at 32.08746, -65.05373 and from 31 March to 10 September 2018 on ST at a water depth of 40.2 m at 32.19605, -64.99133. The AMARs were programmed to record 30 minutes of every hour. Another AMAR equipped with M36-V35-900 omnidirectional hydrophone (-165 ± 4 dB re 1 V/μPa sensitivity) was deployed on CB at a water depth of 47.7 m from 10 September 2018 to 23 April 2019 at 32.08725, -65.05386, and programmed to record 30 minutes every 75 minutes. All three hydrophones located on the seafloor recorded 29 minutes at a sampling rate of 16 kHz (24 bit resolution) and 1 minute at a sampling rate of 250 kHz (16 bit resolution). These sampling rates were chosen to detect the lower frequency vocalisations of baleen whales (Kowarski et al., 2018), as well as high-frequency clicks and whistles produced by toothed whale (Edds-Walton, 1997). For the present study, only the 29-minute recordings were analysed. Due to technical failure of the recording device, recordings from 20 September 2018 to 1 November 2018 were not usable. The AMARs were retrieved from anchors using an acoustic release (Supplementary Figure 1).

Acoustic analysis

All recordings were manually scanned for humpback whale sounds using spectrograms generated with Raven Pro 1.6 sound analysis software (fast Fourier transformation [FFT] size: 2048 points, 75% overlap, Hann window, frequency resolution: 7.8 Hz, time resolution: 32 ms) (Center for Conservation Bioacoustics, 2019). As most humpback whale vocalisations are detected below 2 kHz (Silber, 1986; Cerchio et al., 2014; Huang et al., 2016), the 29-minute-long recordings were viewed zooming into the frequency band of 0–2 kHz.

Every 29-minute recording was first scanned for humpback whale song. If no song was detected then the recording was re-scanned for other humpback whale vocalisations (non-song vocalisations) to capture the acoustic presence of whales, in absence of song. Each recording was classified into one of three whale sound categories: song (song fragments were also allocated to this category), calls, and no vocalisation.

Second, all recordings allocated to the song category were re-analysed to assess the minimum number of singing whales (“singers”) per hour. Spectrograms of song-containing recordings were scanned for the first 10 minutes (Cerchio et al., 2014; Cholewiak et al., 2018b) of song detection to determine the highest number of overlapping singers in that period. Number of singers was determined by visually counting (i) overlapping units, (ii) overlapping phrases and (iii) differences in sound intensity (Cerchio et al., 2014; Magnúsdóttir and Lim, 2019) from the spectrogram (Figure 2). This is because song structures overlapping in time cannot be produced by the same whale and therefore indicate the number of simultaneously singing whales at one moment in time (Magnúsdóttir and Lim, 2019), while sounds of different intensities indicate multiple vocalising whales at different distances from the hydrophone (Cerchio et al., 2014). Up to five simultaneously singing whales could be differentiated confidently in the present dataset. Therefore, every recording was allocated the minimum number of simultaneous singers, ranging from 0 to 5 (Figure 2). In instances where five or more singers were detected, they could not be further differentiated and were allocated the category “5”.

FIGURE 2

Figure 2 Spectrograms of multiple humpback whales signing, showing five different recordings that vary in their number of simultaneously singing humpback whales. In the lower panel, every singer has been allocated a colour to visualise overlapping units and/or overlapping phrases. Only the fundamental units were coloured. Differences in sound intensity can be seen in the upper panel by the darkness of a unit. Spectrogram parameters: fast Fourier transform (FFT) size = 2048 points, overlap = 75%, sample rate = 16000 Hz, frequency resolution = 7.8 Hz and time resolution = 32 ms.

Third, all 29-minute recordings where only one whale was singing one entire song cycle were graded into low, medium, and high-quality recordings (Supplementary Figure 2). Only recordings in the high-quality category were considered for the detailed 2018 Bermuda song description. To ensure consistency throughout the acoustic analysis, all recordings were analysed and categorised into vocalisation type, number of singers and quality by a single person (the first author).

Analysis of seasonal and diel patterns in singing activity

Temporal patterns of humpback whale song based on the numbers of singing whales were statistically analysed in R version 3.6.2 (R Core Team, 2019), with figures created using the R-package ggplot2 (Wickham, 2016). To explore diel patterns, the entire dataset was shifted to Atlantic Standard Time (UTC -4, Bermudas standard time zone) and every day was split into four light conditions in accordance with previous studies (Cholewiak, 2008; Kowarski et al., 2018; Ryan et al., 2019): nighttime, dawn, daytime and dusk. To accommodate for the variation in day and night length over the study period, the corresponding time intervals defined by nautical dawn, sunrise, sunset and nautical dusk were determined separately for each day. The four variables were obtained from https://www.timeanddate.com/ (Time and Date, 2019a) and uniformly transformed to UTC -4. Then, the 29-minute-long audio files recorded during the around one-hour long dawn and dusk periods were allocated to the appropriate light level and the remaining recordings to the daytime or nighttime category accordingly. To further explore seasonal patterns, the entire study period was split into astronomical seasons, i.e., as defined by equinoxes and solstices (Time and Date, 2019b), giving rise to four datasets: ST-spring-2018, CB-spring-2018, CB-winter-2018/19 and CB-spring-2019.

Statistical analysis of diel patterns in singing activity focussed exclusively on days with song-containing recordings. As humpback whale songs can be detected over greater distances than the distance between CB and ST (Cholewiak et al., 2018a), we cannot exclude that individual whales were recorded simultaneously during overlapping time frames between the two sites. Given the unequal sample sizes between the light categories and seasons, the data not being normally distributed and the spring 2018 datasets potentially being dependent of each other, each dataset was analysed individually with a non-parametric Kruskal-Wallis test to determine if the mean number of singing whales significantly differed between different light conditions, to determine any diel variation. Consequently, Bonferroni post-hoc tests were conducted to identify specific pairs of conditions that differed.

Humpback whale song description

Phrase and theme allocation in humpback whale song analysis is most often conducted through the subjective manual analysis of spectrograms (Cholewiak et al., 2013; Nikšić, 2014; Garland et al., 2017; Magnúsdóttir, 2017; Hauer-Jensen, 2018). Thus, it is recommended to base the song description on the review of multiple song recordings from different individuals (Cholewiak et al., 2013). Only the CB recordings from spring 2018 were used for the song description. By going through every song-containing recording (n = 777) in detail to determine the number of singers, the typical song structure became rapidly evident. Seven high-quality song recordings (Narganes Homfeldt et al., 2022), each around a week apart from the next, were chosen to describe the song in more detail. This interval was chosen to minimise the chances of describing the vocal display of the same individual (Kowarski et al., 2019; Magnúsdóttir and Lim, 2019), given that some individuals are sighted for eight consecutive days off Bermuda (Stevenson, 2011). The methodology to describe the songs generally followed approaches adopted in previous humpback whale song analyses (Nikšić, 2014; Garland et al., 2017; Magnúsdóttir, 2017; Darling et al., 2019; Kowarski et al., 2019) and is summarised below.

The seven songs were viewed as spectrograms (FFT size: 2048 points, 75% overlap, Hann window, frequency resolution: 7.8 Hz, time resolution: 32 ms) and each full song cycle was delineated at the unit and phrase level, based on aural and visual spectrographic characteristics. Distinct units and phrases were identified and allocated an alphanumeric code: units (a,b,c); phrases (1,2,3). As a theme is defined as a repeated sequence of the same or similar phrases (Payne and McVay, 1971; Cholewiak et al., 2013; Magnúsdóttir, 2017), themes were allocated the same numerical code as the phrase type it contained, e.g., phrase 1 is repeated within theme 1. The identified characteristic unit sequences were then grouped into phrases and phrase types into themes. Any natural variation in the characteristic unit sequences through addition, deletion or replacement of individual units within a phrase that would generate imperfect replicas of a phrase type (Payne and McVay, 1971; Cholewiak et al., 2013) were considered to be the same phrase type. In addition, there were “transitional phrases”, which are sometimes encountered at the transition between two subsequent themes and contain units from both themes (Payne and Payne, 1985; Garland et al., 2017). For example, a transitional phrase between themes 2 and 3 would be referred to as phrase 2/3. All seven songs were transcribed into alphanumerical sequences to facilitate the identification of the songs’ theme order.

Given the subjective nature of the song delineation process, Cholewiak et al. (2013) advised to support the alphanumerical classification with exemplar spectrograms to illustrate at least part of the song structure to the humpback whale song research field. Thus, in the present study, exemplars of the identified unit, phrase and theme order were illustrated (see Results).

Results

The present study analysed a total of 347 days of continuous recording [excluding a 43-day period with no available recordings (Figure 3)] from 2018 and 2019 across two locations off Bermuda, scanning a total of 5417 hours of acoustic data for humpback whale vocalisations. Besides humpback whale song and non-song vocalisations, an unusually long, tonal baleen whale vocalisation, lasting for 18-seconds, was detected in the presence of multiple humpback whale singers (Supplementary Figure 3). Also, other biotic sounds emitted by dolphins (and possibly other cetacean species), fish and invertebrates, as well as anthropogenic sounds from vessels and echosounders, were regularly detected in recordings.

FIGURE 3

Figure 3 Seasonal occurrence and diel pattern of humpback whale song on Challenger Bank encountered across 55 weeks between 2018 and 2019. Sunset to sunrise (yellow lines) defines daytime, dusk to dawn (black lines) defines nighttime and the two remaining time intervals define the twilight periods. Grey shaded fields indicate no available recordings due to technical failure of the recording device.

Seasonality of vocalisations and singing activity

Humpback whale vocalisations were heard in 32% (1733 h) of recordings. All acoustic detections occurred on 48% (166 days) of recording days, whereby social calls were only detected on days when song, the predominant vocalisation type (97%), was present too. Thus, non-song vocalisations were neglected for determining the seasonality of humpback whale vocalisations. Humpback whales were exclusively heard between 31 March (start of study) and 19 May 2018 (across both sites), as well as 26 December 2018 and 23 April 2019 (end of study) (Figure 4). There was one exception on 31 August 2018 (Figure 3) when a single song fragment was documented on CB. Overall, acoustic presences across the three deployments indicated a clear seasonal trend of humpback whale occurrence off Bermuda in the spring and winter months, ranging from late December to mid-May, hereafter referred to as “whale season” (Figures 3, 4). Notably, 65% of song recordings contained whale chorus (≥ 2 singers).

FIGURE 4

Figure 4 Daily mean number of singers during the whale season at Sally Tucker (A) and Challenger Bank between 31 March and 19 May 2018 (B) and between 26 December 2018 and 23 April 2019 (C). Error bars are ± 1 standard error. The average number of singers across the entire whale season (1.73 singers; blue line) is illustrated as a reference.

Singing activity, as quantified by the numbers of simultaneously singing whales, starts in late December and reaches the daily average across the whale season (1.7 ± 1.5 singers) quite quickly (Figure 4). The daily averages of simultaneously singing whales recorded at CB during the two spring seasons (1.921 ± 1.385 singers) and the winter season (1.926 ± 1.388 singers) were nearly identical, indicating that the intra-seasonal variation might be bigger than inter-seasonal variation. By mid-February to early March, fewer male humpback whales were recorded singing (Figures 3, 4). The highest average number of singers was detected in January, March and early April (in both 2018 and 2019 data) (Figure 4). In May, consecutive recordings were often characterised by a single singer and similar intensities, which likely represent a single male’s song session lasting several hours. Thus, although still singing for most hours of the day, the acoustic density of singing males fell below daily average in late April and ceased entirely by mid-May (Figures 3, 4). Days without song presence during the whale season only occurred at the start (27 December 2018) and end of the whale season at both CB (5-7 and 19 May 2018) and ST (5-7, 10, 13, 15, 18 May 2018) (Figure 4).

ST and CB showed broadly similar patterns in singing activity in spring 2018 (Figure 4). The concentration of singers was higher at CB, with the last humpback whale song being recorded on 18 May 2018, whilst singing activity at ST had already decreased below average by mid-April, with the last humpback whale song being recorded on 19 May 2018 (Figure 4). Therefore, humpback whale singing activity in Bermuda showed a strong seasonal pattern in the spring and winter months but with reduced activity towards the start and end of the whale season, as well as mid-season from February to early March.

Diel patterns in singing activity

Throughout the whale season moderate singing activity was detected during daylight hours and high levels of whale chorusing during the night (Figure 3). A statistically significant difference in the mean number of singing whales was detected between light conditions for all four datasets (Kruskal-Wallis; ST-spring-2018: χ² = 13.9; df = 3; p = 0.003; CB-spring-2018: χ² = 27.2; df = 3;

p = 5.4 x 10 ^-6; CB-winter-2018/19: χ² = 26.7; df = 3; p = 6.7 x 10 ^-6; CB-spring-2019: χ² = 55.0;

df = 3; p = 6.9 x 10 ^-12), with the number of singers being significantly lower during daytime than nighttime (ST-spring-2018: p = 0.0011; CB-spring-2018: p = 3.5 x 10 ^-5; CB-winter-2018/19:

p = 7.7 x 10 ^-7; CB-spring-2019: p = 2.1 x 10 ^-11) (Figure 5). At CB, the number of singers across the seasons was also significantly lower during dawn than nighttime (CB-spring-2018: p = 0.013; CB-winter-2018/19: p = 0.013; CB-spring-2019: p = 0.0053) (Figure 5). In both spring seasons at CB there was a significantly lower number of singers during daytime than dusk (CB-spring-2018:

FIGURE 5

Figure 5 Diel patterns of humpback whale singing activity at Sally Tucker in spring 2018 (A) and Challenger Bank in spring 2018 (B), winter 2018/19 (C) and spring 2019 (D). The median number of singers (horizontal black line) varied significantly between daytime and nighttime in all four datasets. Boxes indicate the interquartile range and the vertical lines indicate the range of daily averages in the number of singers. Significance levels are illustrated with stars (*p < 0.05; **p < 0.01; ***p < 0.001).

p = 6.5 x 10 ^-4; CB-spring-2019: p = 6.5 x 10 ^-6) and in spring 2019 during daytime than dawn

(p = 3.3 x 10 ^-5) (Figure 5). Across all datasets the mean number of singers during the twilight periods did not differ significantly from each other (Figure 5). In addition, the single song fragment encountered in late August was also recorded during nighttime (Figure 3). Therefore, humpback whale singing activity in Bermuda showed a diel pattern across spring and winter months with significantly increased singing at night relative to the daytime and with twilight periods characterised by intermediate levels of singing (Figures 3, 5).

Song structure

The detailed analysis of the seven humpback whale songs that were transcribed at the unit and phrase level contained 14 full song cycles (Supplementary Table 1). The coded songs revealed the stereotypical song structure of the whales migrating through Bermuda in spring 2018, consisting of: 18 distinctive unit types (Supplementary Figure 4), making up 6 phrase types and consequently 6 themes, and 5 transitional phrases (Table 1; Figure 6).

TABLE 1

Table 1 Unit sequences of phrase types.

FIGURE 6

Figure 6 Humpback whale song type encountered in spring 2018 at Challenger Bank, Bermuda. A representative phrase for each theme and transitional phrase (/) is shown. The phrases are in cyclical order 5/1-1-1/2-2-2/3-3-3/4-4-6-4/5-5. Phrase 6 did not occur in every analysed song cycle. The vertical blue lines indicate divisions between phrase types. Spectrogram parameters: fast Fourier transform (FFT) size = 2048 points, overlap = 75%, sample rate = 16000 Hz, frequency resolution = 7.8 Hz and time resolution = 32 ms.

Every song cycle, even from the same individual, showed small unit variations within the same phrase type. These variations were still considered the same phrase and thus part of the same theme (Table 1). In particular, units b, c, f, p, q seemed to act as synonyms and be interchangeable to some extent within various phrase types (Table 1). Also, the amount of unit repetitions within the same phrase sequence varied (Supplementary Table 1).

All seven whales sung the theme order: 1-1/2-2-2/3-3-3/4-4-4/5-5-5/1-1 (Figure 6), which repeated itself through various song cycles (Supplementary Table 1). However, in 4 of the 14 analysed song cycles, although the transitional phrases 2/3 and 3/4 occurred, phrase 3 was not sung (Supplementary Table 1). Every song cycle contained the five transitional phrases once, while the non-transitional phrase types were repeated to a varying extent, even within an individual’s song session (Supplementary Table 1). This variation in phrase repetitions resulted in a large range of song cycle duration, ranging from 4.05 to 16.15 min for the same theme frequency (Supplementary Table 1).

Notably, phrase 6, the only phrase containing unit m (Table 1), was only sung by three of the seven whales (on 27 April, 3 and 18 May), and two of these did not sing phrase 6 in every song cycle either (Supplementary Table 1). Phrase 6 either replaced phrase 4 or occurred after phrase 4, and was always followed by the transitional phrase 4/5 (Figure 6; Supplementary Table 1). All other phrases remained stable throughout the whale season.

Discussion

Vocalisations at Challenger Bank and Sally Tucker's

As all recordings obtained from the fixed AMARs contained high levels of singing activity throughout the whale season (Figures 3, 4), CB and ST seem to be an important singing habitat and stopover site for male humpback whales on their annual migrations. Seamounts with shallow summits like CB are generally known as hotspots for aggregations of migratory megafauna (Morato et al., 2008) and are considered important offshore habitats for humpback whales worldwide, on their breeding grounds but also as key sites along their trans-oceanic migration routes (Garrigue et al., 2015; MacKay, 2015; Derville et al., 2020). Garrigue et al. (2015) suggests that humpback whales visit seamounts on their migration to rest, to use them as navigational landmarks, and potentially as opportunistic feeding areas. These habitat uses could all be applicable to Bermuda (Stevenson, 2011). The Bermuda platform (ST), which is shallower and has calmer waters than CB, represents the typical protected nearshore habitat in which mother-calf pairs are frequently encountered in other regions (MacKay, 2015; Indeck et al., 2021). Although ST and CB thus might form different habitats for humpback whales, both locations showed a very similar pattern in singing activity in spring 2018 (Figure 4). However, singing activity was most prevalent at CB, which might indicate that the more sheltered waters of ST are visited less by male singers and might in fact be used more by mothers and calves (Stevenson, unpublished data).

Finally, the present study detected a long tonal vocalisation of unknown origin on 8 April 2018 within the presence of humpback whale chorus (Supplementary Figure 3). This type of vocalisation is not part of humpback whales’ described call repertoire in either the North Pacific (Fournet, 2018), Southern Hemisphere (Dunlop et al., 2007; Recalde-Salas et al., 2020; Ross-Marsh et al., 2022) or the North Atlantic (Stimpert et al., 2011). The “long cry” as defined in the repertoire of humpback whales in Western Australia is typically only 3 seconds long (Recalde-Salas et al., 2020), compared to the 18-second-long cry documented in this present study (Supplementary Figure 3). Therefore, if produced by humpback whales, this very long cry could be a rare vocalisation (from calves, females and/or males), and may serve a highly specialised, but rarely required, function. Alternatively, it could have been emitted by another baleen species, such as Bryde’s whale (Balaenoptera edeni), who are known for long moans that are in this range of duration (Rice et al., 2014) but who have not so far been recorded off Bermuda (Stevenson, unpublished data).

Seasonality and migration pattern

The present study identified a strong seasonal pattern of humpback whale singing activity, with a moderately high occurrence of chorusing whales, off Bermuda in the spring and winter months, ranging from late December to mid-May (Figure 3). This is in line with boat-based observations from Whales Bermuda (Stevenson and Stevick, 2009; Stevenson, 2011) and the historic records of humpback whales (Stone et al., 1987) and their songs being recorded off Bermuda in January, April and May (Payne and McVay, 1971; Payne and Payne, 1985). Fluctuations in numbers of singers throughout the spring and winter months (Figure 4) suggest that male humpback whales migrate through Bermuda in waves and do not stay for longer periods of time. This theory is supported by visual boat-based observations made by Whales Bermuda and studies showing North Atlantic humpback whales lingering for several days up to two weeks and aggregating around Bermuda before continuing their northward migrations in large groups (Payne and Payne, 1985; Stevenson, 2008; Stevenson and Stevick, 2009; Stevenson, 2011). Humpback whales overwintering in the northwestern Caribbean are sighted at their breeding grounds between January–April with peaks between February–March, while those wintering in the southeastern Caribbean do so a bit later between March–May with peaks in April (Charif et al., 2001; Stevick et al., 2003; Gandilhon, 2012; Stevick et al., 2018; Heenehan et al., 2019). In Bermuda, reduced singing activity was observed between mid-February to early March (Figures 3, 4), which is in line with the northwestern Caribbean peak. Detecting humpback whale song from late December onwards and to a lesser extent during the northwestern Caribbean peak breeding season (Figure 3), as well as the observed higher intra- than inter-seasonal variation in average number of singers, suggests that Bermuda, previously described as a one-way stopover on their northward migration during the spring months (Payne and McVay, 1971; Stone et al., 1987; Stevenson and Stevick, 2009) also acts as a migratory stopover site on their southward migration during the winter months.

The humpback whales recorded in the present study could have migrated to and from either Caribbean breeding ground, given the temporal overlap in both breeding grounds. However, as Bermuda’s whale season starts just before the northwestern Caribbean whale season and the northwestern peak aligns with reduced singing in Bermuda (Figure 3), it is more likely that humpback whales migrate to and from the northwestern Caribbean through Bermuda. Both, satellite-tagging (Kennedy et al., 2014) and fluke identification matches support this theory (Stone et al., 1987; Beaudette et al., 2009; MacKay, 2015). However, in 2017, humpback whales were heard singing until 13 May on the northwestern Caribbean breeding ground and until 27 May on the southeastern Caribbean breeding ground (Heenehan et al., 2019). Thus, it is possible, that humpback whales recorded towards the end of Bermuda’s whale season (Figure 3) could also be originating from the southeastern breeding ground.

The single song fragment heard on 31 August 2018 (Figure 3) matches temporarily with song fragments only starting to become more frequent from late August onwards in feeding grounds (Vu et al., 2012; Baumgartner et al., 2019). However, North Atlantic humpback whales are thought to leave their feeding grounds and start their southward migrations much later than August, usually in late autumn to early winter, or, for some whales from the eastern feeding grounds, even in late winter (Stevick et al., 2018; Heenehan et al., 2019; Kowarski et al., 2019; Magnúsdóttir and Lim, 2019). Bermudian fishermen have reported sightings of humpback whales in September (Stevenson, unpublished data). This suggests that a few individuals in some years could start their migration earlier than previously expected and the present acoustic study confirms the presence of such an individual at this time in 2018. Although no acoustic presence of humpback whales was evident in June 2018 (Figure 3), a humpback whale was observed breaching at CB two years later on 15 June 2020 (Stevenson, unpublished data). Thus, inter-annual variability of occasional single individuals migrating through Bermuda seems to occur off-season, e.g., on 31 August 2018 and 15 June 2020.

Diel pattern in singing activity

The diel singing pattern observed in our study, with peak singing activity at nighttime (Figure 3; 5), has been observed in all three humpback whale subspecies on feeding (Magnúsdóttir et al., 2014; Huang et al., 2016; Kowarski et al., 2018) and breeding (Au et al., 2000; Cholewiak, 2008; Cerchio et al., 2014; Kobayashi et al., 2021) grounds, as well as migration routes (Ryan et al., 2019; Shabangu and Kowarski, 2022), and is thus suggested to be a species-wide characteristic (Kowarski et al., 2019). Au et al. (2000) suggested that the diel pattern results from humpback whales relying on visual and acoustic cues during the day, but solely on acoustic cues during the night. For the singing whale, this means that nighttime is the most efficient time to send out the signal and may explain why the diel trend is observed across different breeding, feeding and migratory habitats and why even most non-song vocalisations increase at night (Parks et al., 2014; Huang et al., 2016; Kowarski et al., 2018). In other baleen species, diel variations in vocal activity have been linked to diel distribution in their prey: when prey aggregate, foraging becomes most efficient and whales vocalise less, unless the vocalisation is directly associated to foraging (Wiggins et al., 2005; Baumgartner and Fratantoni, 2008; Širović et al., 2013). However, this hypothesis would not explain why humpback whale singing activity also peaks at nighttime on breeding grounds where they are not exhibiting any feeding behaviour (Cholewiak, 2008). Thus, reliance on visual cues during daytime, as suggested by Au et al. (2000) is currently considered the most plausible explanation for the observed diel patterns in humpback whale singing activity.

Future anthropogenic noise mitigation measures

The present study reveals the importance of Bermuda’s waters for migrating humpback whales throughout the spring and winter months. However, the high volume of vessel traffic and underwater noise that accompanies expansion of Bermuda’s tourism industry could become a key issue (Jones, 2011; Lester et al., 2016; Bermuda Tourism Authority, 2019). Bermuda’s lucrative cruise tourism season currently operates from May to October (Jones, 2011) but could be extended to last from April to December (Bermuda Tourism Authority, 2019). The presence of humpback whales, being continuously present from late December until mid-May (Figure 3), already coincides with the start of the current cruise season; should the new cruise season be rolled out, this temporal overlap will increase. Besides increased risks from ship strikes and behavioural disturbance from prolonged and more intense whale-watching, cruise and freight passages through Bermuda, growth in the tourism sector will further elevate ambient noise levels (e.g., through coastal infrastructure construction and increased vessel noise) which reduce humpback whales’ communication space (Au and Green, 2000; Jones, 2011; Micheli et al., 2012; Dunlop, 2016; Lester et al., 2016; Cholewiak et al., 2018a; Gabriele et al., 2018; Sprogis et al., 2020; Currie et al., 2021). In response to elevated noise levels male humpback whales have been shown to cease their song (Sousa-Lima and Clark, 2008; Risch et al., 2012; Cerchio et al., 2014; Tsujii et al., 2018), lengthen their song cycle (Miller et al., 2000) and amplify their vocalisation display, exhibiting the Lombard effect (Guazzo et al., 2020). Given that the present study detected humpback whales frequently chorusing, the masking of these displays by anthropogenic noise could negatively impact this behaviour in Bermuda, with currently unknown consequences on the population.

In light of this, we advocate authorities to consider precautionary and mitigation measures (Currie et al., 2021; Pires et al., 2021; Risch et al., 2021) during the whale season (December–May). The heavy cruise and shipping traffic lanes operating near CB and the Bermuda platform (ST) could be re-routed given the authorisation by the International Maritime Organization (IMO, 2014), similar to the implementation at the sister Marine Mammal Sanctuary around the Dominican Republic (Kennedy and Clapham, 2017; Heenehan et al., 2019). Dredging and construction of critical marine infrastructure to support the shipping and tourism sectors could be avoided during the night, when humpback whale vocal activity is at its highest (Figure 5), or during the whale season entirely. In addition, moving towards more sustainable noise-reduction vessel designs (Arranz et al., 2021) and implementing a 10 knot vessel speed limit across Bermuda’s EEZ for the duration of the whale season could significantly reduce noise levels and reduce masking for both mysticetes and odontocetes (IMO, 2014; Pensieri and Bozzano, 2017; Williams et al., 2019; Aschettino et al., 2020; ZoBell et al., 2021). The latter measure, would also reduce the risk of vessel collisions and cut down on the vessel’s carbon emissions (Laist et al., 2001; Lack and Corbett, 2012).

Although these precautionary measures will reduce anthropogenic noise for humpback whales when migrating through Bermuda, year-round implementation of any measure to reduce cetacean disturbance, collision risk and anthropogenic noise would benefit other marine species in Bermuda’s EEZ including the resident bottlenose dolphin (Tursiops truncatus) population (Klatsky et al., 2007), Cuvier’s beaked whales (Ziphius cavirostris) and occasional passing sperm whales (Physeter macrocephalus) (Hallett, 2011; Stevenson, unpublished data).

Bermuda’s song and future song comparisons

The present study characterised Bermuda’s humpback whale song structure, encountered in spring 2018, for the first time since 1976. As phrase 6 occurred less frequently than other phrase types and was increasingly present at the end of the 2018 spring season (Supplementary Table 1), it may represent a new phrase type that evolved in late April from phrase 4 and was slowly being introduced into the song repertoire of the predominant breeding population (northwestern Caribbean) migrating through Bermuda across the whale season. Alternatively, phrase 6 could represent the song repertoire of the southeastern breeding population migrating through Bermuda later in the spring season. Therefore, further research should compare the song type described in the present study to acoustic recordings obtained between 2017 and 2019 across the full North Atlantic humpback whale range including off Cape Verde, the southeastern and northwestern Caribbean, Bermuda, the migratory corridor off the British Isles and eastern and western feeding grounds. Analysing and identifying similarities and differences in song structure of all the above listed habitats within the same song season would help elucidate North Atlantic humpback whales’ population structure (Murray et al., 2012; Archer et al., 2020), migration paths and the role of Bermuda as a migratory stopover for both the northwestern and southeastern breeding population, which is important information required for the conservation management of this migratory species.

Conclusion

The present acoustic study represents the first long-term PAM study of humpback whale vocalisations off Bermuda. Our results highlight the importance of Bermuda as a key two-way migration stopover site for male North Atlantic humpback whales. They primarily display nocturnal singing activity in the spring and winter months from late December until mid-May. The strong seasonal and diel pattern of whale chorus observed in this study provides new evidence to aid Bermuda’s planning authorities with sustainable marine development around Bermuda and the wider Sargasso Sea.

Data availability statement

The original contributions presented in the study are included in the article/Supplementary Material and an additional data publication is archived with PANGAEA at https://doi.pangaea.de/10.1594/PANGAEA.946517. Further inquiries can be directed to the corresponding author.

Ethics statement

The animal study was reviewed and approved by School of GeoSciences, University of Edinburgh.

Author contributions

AS and L-AH contributed to the conception of the study. AS conducted all fieldwork. TNH and DR designed the study. TNH processed the data, performed the analysis, created the figures and took the lead in writing the manuscript. DR and L-AH supervised the project and provided critical feedback throughout the study. All authors contributed to the article and approved the submitted version.

Funding

This study has received funding from the Atlantic Conservation Partnership and the European Union’s Horizon 2020 research and innovation programme under grant agreement No. 818123 for the iAtlantic project. This output reflects only the authors’ views and the European Union cannot be held responsible for any use that may be made of the information contained therein.

Acknowledgments

We thank JASCO Applied Sciences (Canada) Ltd. for providing the measurement equipment for the data collection and to Katie Kowarksi for discussions on data analysis and interpretation of humpback song. Data for this study were collected under a Protected Species Licence (License no: 18-12-18-73) for Scientific Research Activities issued by the Department of Environment and Natural Resources, Government of Bermuda.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmars.2022.941793/full#supplementary-material

References

André M. (2018). Ocean noise: Making sense of sounds. Soc Sci. Inf. 57, 483–493. doi: 10.1177/0539018418793052