Fin whale song evolution in the North Atlantic

  1. Miriam Romagosa  Is a corresponding author
  2. Sharon Nieukirk
  3. Irma Cascão
  4. Tiago A Marques
  5. Robert Dziak
  6. Jean-Yves Royer
  7. Joanne O'Brien
  8. David K Mellinger
  9. Andreia Pereira
  10. Arantza Ugalde
  11. Elena Papale
  12. Sofia Aniceto
  13. Giuseppa Buscaino
  14. Marianne Rasmussen
  15. Luis Matias
  16. Rui Prieto
  17. Mónica A Silva
  1. Institute of Marine Sciences - OKEANOS & Institute of Marine Research - IMAR, University of the Azores, Portugal
  2. Cooperative Institute for Marine Ecosystem and Resources Studies, Oregon State University, United States
  3. Centre for Research into Ecological and Environmental Modelling, University of St Andrews, United Kingdom
  4. Centro de Estatística e Aplicações, Departamento de Biologia, Faculdade de Ciências, Universidade de Lisboa, Portugal
  5. NOAA Pacific Marine Environmental Laboratory, Hatfield Marine Science Center, United States
  6. CNRS - UBO - UBS - Ifremer, IUEM - Lab. Geo-Ocean, France
  7. Marine and Freshwater Research Centre (MFRC), Atlantic Technological University, Ireland
  8. Instituto Dom Luiz (IDL), Universidade de Lisboa, Portugal
  9. Institute of Marine Sciences, ICM‐CSIC, Spain
  10. Institute for the Study of Anthropic Impacts and Sustainability in the Marine Environment of the National Research Council of Italy (CNR-IAS), Italy
  11. Akvaplan-niva, Norway
  12. University of Iceland Research Centre in Húsavík, Iceland

Abstract

Animal songs can change within and between populations as the result of different evolutionary processes. When these processes include cultural transmission, the social learning of information or behaviours from conspecifics, songs can undergo rapid evolutions because cultural novelties can emerge more frequently than genetic mutations. Understanding these song variations over large temporal and spatial scales can provide insights into the patterns, drivers and limits of song evolution that can ultimately inform on the species’ capacity to adapt to rapidly changing acoustic environments. Here, we analysed changes in fin whale (Balaenoptera physalus) songs recorded over two decades across the central and eastern North Atlantic Ocean. We document a rapid replacement of song INIs (inter-note intervals) over just four singing seasons, that co-occurred with hybrid songs (with both INIs), and a clear geographic gradient in the occurrence of different song INIs during the transition period. We also found gradual changes in INIs and note frequencies over more than a decade with fin whales adopting song changes. These results provide evidence of vocal learning in fin whales and reveal patterns of song evolution that raise questions on the limits of song variation in this species.

Editor's evaluation

This study is a valuable contribution to our understanding of vocal variation in acoustic displays of male baleen whales, part of a developing story about cultural change in songs in species other than the relatively well studied humpback whales. The authors present solid evidence of changes at various timescales in 20-Hz song note intervals and call center frequency over decadal time scales and large spatial scales.

https://doi.org/10.7554/eLife.83750.sa0

Introduction

Animal songs, often used as acoustic sexual displays, can change within and between populations through different evolutionary processes. These processes can be selective (i.e. sexual, cultural, or natural selection), favouring song changes that confer advantages to singers, or non-selective (i.e. cultural or genetic drift), causing random changes in songs. Both selective and non-selective processes may result in rapid and gradual population-wide shifts in the structure, complexity, frequency, and temporal properties of songs (Garland et al., 2011; Otter et al., 2020; Whiten, 2019; Williams et al., 2013), although song evolution can also be constrained by the species’ genetic variation and mechanical design (Podos et al., 2004).

The best-known examples of song evolution are found among songbirds, which field studies started decades ago and have led to extensive literature on the topic. Songs from many songbirds are culturally transmitted through vocal learning, wherein animals learn to sing by hearing and imitating conspecifics (Williams, 2021). Vocal learning and specific patterns of dispersal are largely responsible for the geographic variation found in songs of many bird species (Podos and Warren, 2007). The formation of local dialects is, in part, a consequence of certain mechanisms of song learning (i.e., copying ‘errors’) that generate vocal novelties (Podos and Warren, 2007). Learned songs may also undergo rapid evolutions within populations, basically because cultural novelties can emerge more frequently than genetic mutations (Wilkins et al., 2013). One example of rapid song evolution is found in the white-throated sparrow (Zonotrichia albicollis), in which a new doublet ending song spread across the North America continent in less than 20 years, completely replacing the established triplet-ending song. This fast spread is believed to have occurred because birds singing the old and new songs overwintered in the same grounds and learned from each other. Although it remains unclear why the new song overturned the old song, one possible explanation for this rapid song transition is that certain innovations are adopted non-randomly by all males to maintain female interest (Otter et al., 2020). Birdsong properties can also show a gradual directional evolution in response to specific evolutionary process. A clear example are birds from urban areas, which song elements increase in frequency (Hz) in response to noisy environmental conditions (Slabbekoorn, 2013). Directional song evolution can also cause directional song changes (e.g. faster trill rates, broader frequency bands, and lower frequency trills) driven by sexual selection operating through male-male interactions, mate choice by females or both (de Kort et al., 2009; Illes et al., 2006; Williams et al., 2013). Yet, all these song variations are constrained by the singers’ morphological (e.g. beak shape or body size) and neurological limitations that sometimes can hinder the animals’ adaptation to rapid human induced changes in the environment (e.g. urban noise; Luther and Derryberry, 2012; Podos et al., 2004).

A parallelism to birdsong evolution can be found in the marine realm. Songs from humpback whales (Megaptera novaeangliae) differ across ocean regions (Winn et al., 1981), evolve gradually over time (Payne et al., 1983; Payne and Payne, 1985) and can go through revolutionary changes (Noad et al., 2000). During song revolutions, a population song type is rapidly replaced by a novel song type introduced from a neighbouring population (Garland et al., 2011; Noad et al., 2000). Most authors agree that these spatial and temporal patterns in humpback whale song changes can only be explained by vocal learning (Garland et al., 2011; Janik and Knörnschild, 2021; Noad et al., 2000; Tyack, 2008). Yet, the learning capacity of novel songs in humpback whales may be limited because song complexity always decreases in each revolutionary event (Allen et al., 2018).

Compared to humpback whales and most songbirds, fin whales (Balaenoptera physalus) produce simpler songs consisting of a stereotyped repetition of a few low-frequency note types. These songs are also believed to act as mating displays (Thompson et al., 1992; Watkins et al., 1987), because they are produced by males (Croll et al., 2002) and intensify during the breeding season (Lockyer, 1984; Širović et al., 2013; Thompson et al., 1992; Watkins et al., 1987). In this species, the song inter-note interval (INI) is the most distinctive parameter between regions (Castellote et al., 2012; Delarue et al., 2009; Hatch and Clark, 2004; Širović et al., 2017; Watkins et al., 1987) and has been used to differentiate stocks and populations (Castellote et al., 2012; Delarue et al., 2009; Morano et al., 2012; Širović et al., 2017; Wood and Širović, 2022). Previous studies showed that fin whale song INIs differed between western, central, and eastern North Atlantic areas, as well as between these and the Mediterranean Sea (Castellote et al., 2012; Delarue et al., 2009; Hatch and Clark, 2004; Morano et al., 2012). These results partially agree with genetic data that indicate significant levels of heterogeneity in the mitochondrial DNA between the Mediterranean Sea, the eastern (Spain), and the western (Gulf of Maine and Gulf of St Lawrence) North Atlantic; however, samples from West Greenland and Iceland could not be assigned to either of the two North Atlantic areas, suggesting a mixture of subpopulations in these feeding grounds (Bérubé et al., 1998). Another large-scale study combining fin whale genetic and song data from the Northeast Pacific, North Atlantic, and Mediterranean Sea showed that acoustic differentiation among fin whales were not always reflected in estimates of genetic divergence (Hatch and Clark, 2004). These authors concluded that differences in songs may reflect differences in fin whale movements and/or social and vocal behaviours that occur at shorter timescales than genome evolution. In fact, fin whale song INIs can change abruptly from one year to the next in the same region (Delarue et al., 2009; Hatch and Clark, 2004; Helble et al., 2020; Morano et al., 2012; Širović et al., 2017) and have been progressively changing over time in different ocean regions (Best et al., 2022; Helble et al., 2020; Leroy et al., 2018a; Weirathmueller et al., 2017). Also, the center frequencies of two fin whale song components, the 20 Hz note and the higher frequency (~130 Hz) upsweep (hereafter HF note; Hatch and Clark, 2004), have been decreasing gradually over the last decade in different ocean basins (Leroy et al., 2018a; Weirathmueller et al., 2017; Wood and Širović, 2022).

Currently, we do not understand the mechanisms and drivers of fin whale song variations nor how these variations may be affected by the species’ physiological and morphological constraints of vocal performance. Broad-scale studies matching the known scales of fin whale natural history and ecology can shed light into the species’ population structure and demography, even before genetic differentiation is evident, and elucidate patterns, drivers and limits of song evolution that can ultimately inform on the species’ capacity to adapt to human-induced changes in their acoustic habitats (e.g. anthropogenic noise and climate change).

Our study attempts to address these issues by investigating changes over two decades of three fin whale song parameters (INIs and peak frequencies of the 20 Hz and HF note types) in a wide area of the North Atlantic Ocean. Our work provides evidence of social learning in fin whale songs and shows: (i) a rapid evolution in song INIs across a vast area of the central North Atlantic in just four singing seasons, with the existence of hybrid songs (including both INIs) and a clear geographic gradient of song INIs during the transition period; (ii) a gradual evolution of song parameters showing an increase in INIs and a decrease in frequencies of the 20 Hz and HF notes over more than a decade in the central and eastern North Atlantic; and (iii) the adoption of both rapid and gradual song changes by fin whales from a wide region. We conclude by discussing song changes under the scope of cultural transmission, song function and the limits of song variation.

Results

The processing across all acoustic data resulted in 379 songs, from which 39680 INIs and its corresponding note frequencies were measured, and 143 songs, from which 9185 HF note peak frequencies were measured (Supplementary file 1a; ‘Materials and methods’). The greatest numbers of INIs came from the SE and Azores locations in the Oceanic Northeast Atlantic (ONA) region, with ~32% and~1% respectively. Contributions from the remaining locations of the ONA region ranged from 3% to 7%, while contributions from locations outside the ONA region ranged from 0.6% to 7%. Among the locations in which the measurement of the HF note was possible, the Azores and SE Greenland locations contributed the most (~46% and~22% respectively; Table 1).

Table 1
Sampling information and effort.

For each location within each region this table shows: sampled period, duty cycle, sampling rate (Samp. rate), total number of recording hours (Rec. hours), number of measured inter-note intervals (INIs) (Num. INIs), contribution to total number of INIs measured (Contr. INIs), number of measured high frequency (HF) note peak frequencies (Num. HF note) and percent contribution to total number of measured HF note peak frequencies (Contr. HF note).

RegionLocationSampled periodDuty cycle(%)Samp. rate (Hz)Rec.hoursNum.INIsContr. INIs(%)Num.HF noteContr. HF note (%)
SE GreenlandSE Greenland01/10/2007 - 14/03/2008Cont.2000439228417.1204022.2
SE IcelandSE Iceland04/01/2007 - 31/03/2007Cont.400020882910.71691.9
Celtic SeaNorth Porcupine01/10/2015 - 03/11/2016102000520.82860.72152.3
South Porcupine01/10/2015 - 03/11/2016102000520.86741.76276.9
ONANE01/10/2002 - 31/03/2003Cont.250434426506.7NANA
NW01/10/2002 - 31/03/2003Cont.250434427547.1NANA
CE01/10/2002 - 31/03/2003Cont.250434428177.1NANA
CW01/10/2002 - 31/03/2003Cont.250434429307.3NANA
Azores01 –31/01/2006; 01-31/01/2007;01-31/01/2008Cont.200022325731.4NANA
01/10/2008 - 06/03/2011102000650.47491.8133514.5
15/10/2011 - 06/03/20124320001497.610172.56777.3
01/10/2012 - 18/10/2012292000122.4120.02162.3
23/02/2017 - 31/03/2020252000213623826203622.1
Total6638.4473311.9426446.4
SE08/02/1999 - 31/03/2005Cont.110317041299632.7NANA
SW31/12/2002 - 31/03/2003Cont.110218415173.8NANA
SW PortugalSW Portugal001/12/2007 - 29/02/2008Cont.10021848182.1NANA
01/10/2015 - 31/03/2016202000878.411953.1129414
Canary IslandsCanary Islands01/11/2014 - 29/02/2015Cont.100216019915.1NANA
Barents SeaSvalbard02/10/2014 - 31/01/2016Cont.50028822600.61651.8
Vesterålen01/01/2018 - 28/02/2018Cont.40014169272.34114.5
Total74944.4396801009185100

Transition in song INIs in the Oceanic Northeast Atlantic region

Results showed a rapid shift in song INIs in the SE location from the ONA region (Figure 1A), previously noted by Nieukirk et al., 2011, where songs with 19 s INIs were completely replaced by songs with 12 s INIs in just four singing seasons (Figure 1B and C). In 1999 and 2000, the 19s-INI song was the only song present in this location. By 2004, the 19s-INI song had disappeared from this location (Figure 1B) and was not detected in any of the sampled regions from 2006 to 2020, except from a single song in 2008 Figure 3A. During the transition period, 12s- and 19s INI-songs co-existed and there was a notable percentage of songs containing both INIs, which we refer as ‘hybrid songs’. Hybrid songs showed two INIs in variable ratios and no apparent pattern, either mixed within the same song sequences (i.e. series of consecutive 20 Hz notes separated by periods of silence; Watkins et al., 1987) or found separated in different sequences from the same song. The singing season with the most hybrid songs was 2002/2003, when there were ~30% of hybrids.

Transition in song INIs in the Oceanic Northeast Atlantic region.

(A) Map showing the SE location (red circle) of the Oceanic Northeast Atlantic region. (B) Percentage of songs with each inter-note interval (INI) type (19 s, hybrid (hyb), 12 s) in this location during the song INI shift in 1999 – 2005. (C) INIs from 1999 to 2005 for this same location. Points represent mean values per song and error bars are standard deviations.

Six locations of the ONA region, with simultaneous data from the 2002/2003 singing season, were used to analyse the spatial pattern in song INIs. In this period, the prevalence of songs with each INI type showed a clear spatial gradient across the entire ONA region. The 19s-INI song largely dominated in the SW ONA, with only 9% of hybrid songs, and no detection of 12s-INI songs. The proportion of 19s-INI songs decreased progressively to the east, reaching 0–4% in the easternmost locations (CE and NE), where the 12s-INI songs were prominent (90 and 83%). Hybrid songs were more abundant (17–23%) at central ONA (NW, CW, and SE) than in easternmost locations (NE and CE; 10–13%; Figure 2).

Map showing the percentages of fin whale songs with each inter-note interval (INI) type for six locations within the Oceanic Northeast Atlantic region during the 2002/2003 singing season.

Gradual changes in song INIs and notes frequencies

After the song transition, we found a gradual change in three fin whale song parameters over more than a decade, with most regions fitting the trend. The only exception was the Barents Sea where INIs differed from the rest of the sampled area showing a bimodal pattern. From 2006 to 2021, INIs increased at 0.21 s/yr (Adj. R-sq.=0.4; p<0.001) (Figure 3A and Figure 3—figure supplement 1). Peak frequencies of the 20 Hz note decreased at a rate of –0.06 Hz/yr (Adj. R-sq.=0.1 from 2009 to 2020; p<0.001) (Figure 3B and Figure 3—figure supplement 1) while peak frequencies of the HF note decreased at a rate of –0.35 Hz/yr (Adj. R-sq.=0.8; p<0.001) from 2007 to 2020, with all regions fitting the trend including the Barents Sea region (Figure 3C and Figure 3—figure supplement 1).

Figure 3 with 2 supplements see all
Gradual changes in song INIs and notes frequencies.

(A) Inter-note intervals (INIs) from 2006 to 2020 for all regions sampled. INIs increased at a mean rate of 0.21 s/yr. (B) Peak frequencies of the 20 Hz note for SW Portugal (2015/2016) and Azores locations (Oceanic Northeast Atlantic region) sampled with Ecologic Acoustic Recorders (Lammers et al., 2008) (2008–2020); these changed at a mean rate of –0.06 Hz/yr. (C) Peak frequencies of the High Frequency (HF) note for all regions sampled; these changed at a mean rate of –0.36 Hz/yr. Points represent average values per song, error bars are standard deviations and black lines represent the fitted linear regression model with confidence intervals in shadowed grey.

Differences in fin whale song parameters between regions

When comparing data from different regions (SE Iceland, SE Greenland, ONA, SW Portugal, Canary Islands, Barents and Celtic Sea) with simultaneous recordings (i.e. in the same singing season) results showed unimodal overlapping distributions in INIs and HF note peak frequencies (Figure 4). The only exception was the Barents Sea region, where INIs differed from the Canary Islands in 2014/2015 (Barents Sea:~9 s and ~14 s; Canary Islands:~15 s), from SW Portugal and the Celtic Sea in 2015/2016 (Barents Sea:~10 s and ~15 s; SW Portugal and Celtic Sea:~15 s) and from the ONA region in 2017/2018 (Barents Sea:~10 s and 16 s; ONA:~16 s; Figure 4A).

Histograms of (A) inter-note intervals (INIs) and (B) higher frequency (HF) note peak frequencies by singing season (Oct-Mar) from regions with concurrent data.

Discussion

The rapid and gradual changing patterns of different fin whale song parameters reported here for a wide area of the central and eastern North Atlantic provides evidence of vocal learning in this species. Decoupled variations in song INIs and frequency (i.e. INIs changed abruptly but frequencies did not) reveal the complex interplay between different selective pressures and shed some light on the potential limits of song variation.

The rapid replacement of fin whales’ song INIs (from 19s to 12s) described here for the ONA region cannot be explained by environmental causation. The shift in INIs found in the ONA region seemed to occur simultaneously at northern feeding grounds, in the so-called Northeast North Atlantic (NENA) region (Hatch and Clark, 2004). This variation in INI patterns during the same singing season between neighbouring locations within ONA, together with an identical shift in INIs documented for the same period in the environmentally distant NENA region (Hatch and Clark, 2004), strongly suggest that the transition in INIs was not a response to local acoustic environments. Fin whale song INIs are regionally distinct (Castellote et al., 2012; Constaratas et al., 2021; Delarue et al., 2009; Hatch and Clark, 2004; Morano et al., 2012; Pereira et al., 2020; Širović et al., 2017; Víkingsson and Gunnlaugsson, 2006) and the shift in INIs found in our study could have been caused by a population replacement. However, if this was the case, we would not find hybrid songs containing both INIs during the transition period, as the new song pattern would simply substitute the former, as documented for fin whale songs off Southern California (Širović et al., 2017). Multiyear and seasonal alternation of different fin whale song INIs, with presence of hybrid songs, have also been reported in two regions of the Northwest Atlantic. In both cases, authors suggest INI shifts occurred within the same population (Delarue et al., 2009; Morano et al., 2012). Thus, we suggest that the rapid turnover of fin whale song INIs along a spatial gradient in the ONA region, with males adopting the new song INI, and the existence of hybrid songs, is the result of cultural transmission, the social learning of information or behaviours from conspecifics (Rendell and Whitehead, 2001).

Our study also shows that fin whale song INIs from the distant Barents Sea region differ from the rest of the sampled area (central and eastern North Atlantic). Geographic differentiation in song INIs (Hatch and Clark, 2004) with fin whales within a certain area conforming the same INI has been widely documented (Castellote et al., 2012; Delarue et al., 2009; Hatch and Clark, 2004; Širović et al., 2017; Wood and Širović, 2022). Bird songs also vary geographically, and this variation can be largely attributed to their ability of learning to vocalize through imitation (Kroodsma, 2004; Podos and Warren, 2007). When songbirds learn their songs from models (i.e. conspecifics) inhabiting the same geographic area where they set their breeding territories, local similarities in song structure can arise (i.e. dialects). This learning can occur after dispersal with birds learning or retaining the dialects sang in the breeding grounds where they set (Nelson et al., 2001). Thus, in most species in which vocal learning occurs, the distribution of learned songs may reflect the social interactions among birds, not the genetic structure of the populations (Kroodsma, 2004). A decoupling between patterns of cultural (songs) and genetic variation has also been reported for fin whales (Hatch and Clark, 2004), further suggesting that song INIs may be socially learned in this species. Learning of novel rhythms (i.e. INIs; Vernes et al., 2021) can also be found in sperm whales (Physeter macrocephalus), which can match their clicks to the rhythm of a ship echosounder (Backus and Schevill, 1996), and use codas (i.e. rhythmic patterns of clicks) that are unique to each vocal clan and are socially learned (Rendell and Whitehead, 2003). Fin whales may also be able to learn songs from other populations that not only differ in their INIs but also in their note composition (Helble et al., 2020).

After the song transition from 1999 to 2005, we found a gradual increase in song INIs along with a decrease in peak frequencies of the 20 Hz and HF notes. These findings are in line with the gradual trends of decreasing frequencies (Best et al., 2022; Leroy et al., 2018b; Weirathmueller et al., 2017) and increasing INIs (Best et al., 2022; Morano et al., 2012; Širović et al., 2017; Weirathmueller et al., 2017) described for fin whale songs in other ocean basins and in the Mediterranean Sea. Contrarily to the rapid changes in INIs, a global-scale process of cultural transmission cannot explain these directional changes. First, changes in INIs and frequencies occur at different rates in different oceans and there is no convergence in song acoustic characteristics across populations (Leroy et al., 2018a; Širović et al., 2017; Weirathmueller et al., 2017). Second, a similar pattern of decreasing frequencies and increasing INIs has been described for blue whale (B. musculus) songs (Jolliffe et al., 2019; Malige et al., 2020; McDonald et al., 2009), and decreasing frequencies have been reported for bowhead whales (Balaena mysticetus) calls (Thode et al., 2017). Such gradual song changes in multiple species and different ocean basins suggest an adaptation to a common selective pressure, which does not mean that within-region conformity in song characteristics does not result from cultural transmission. Mathematical modelling of the linear decrease in blue whale song frequencies suggest a simultaneous effect from two selection processes: conformity and sexual selection (Malige et al., 2022). Conformity would occur because individuals would be more likely to share variants of a cultural trait with nearby individuals than with more distant ones. This could be caused either by a conformist bias, which occurs when individuals select common variants from those available more often than would be expected by chance, or by more simple processes, such as only learning from nearby individuals (Morgan and Laland, 2012). Sexual selection would drive males to sing lower frequency songs than other whales, presumably because females prefer bigger males that are able to sing lower pitch songs (Malige et al., 2022). Increased blue whale body size in a post-whaling recovery scenario has also been proposed as a potential explanation for this species’ song changes; yet blue whale body size distributions should have returned to near pre-whaling values by now and song frequencies continue to decrease (McDonald et al., 2009). Also, it is very unlikely that changes in whale body size evolved in such a straight line at this timescale (Malige et al., 2022). Fin whale songs may evolve in a similar way as blue whale songs do, but so far, none of the proposed hypotheses can convincingly explain the slow frequency song changes in these species (McDonald et al., 2009; Thode et al., 2017). Large-scale and long-term datasets would help understanding if fin whale song INIs and frequencies are constantly evolving or started changing recently in response to a new driver.

The rapid and gradual evolution of fin whale song parameters found in this, and other studies (Hatch and Clark, 2004; Širović et al., 2017; Weirathmueller et al., 2017), resemble the patterns of song evolution of some bird species and humpback whales. Evidence from songbirds suggest that these different trajectories in song evolution (rapid versus gradual) occur within certain boundaries because learned songs are subject to a combination of strong stabilizing selection and underlying genetic variation that prevent incremental change for long periods of time (McEntee et al., 2021). In humpback whales, song complexity increases as songs evolve gradually over time, but decreases when revolutions occur (i.e. periods of rapid song changes), suggesting that learning capacities in this species are limited (Allen et al., 2018). After the rapid shift in fin whale song INIs, from 19s to 12s, a gradual reset towards the 19s-INIs seems to be occurring in all sampled areas, except from the Barents Sea. In the northwest Atlantic Ocean, rapid shifts in fin whale song INIs occurred between 15 s and 9 s (Delarue et al., 2009; Morano et al., 2012). Perhaps, like in birdsongs and humpback whales, changes in fin whale song INIs are also limited by learning constrains and genetic predispositions. Our results show that fin whale song INIs from the Barents Sea region differ from the rest of the sampled area. Yet, satellite tracking data from 2015 to 2019 showed that some fin whales summering in Svalbard (Barents Sea) migrate to the SW Portugal region in fall and winter (Lydersen et al., 2020), so some degree of mixing between males from these two acoustic populations occur. Also, a recent study from Svalbard revealed that fin whale song INIs differed between singing seasons, which suggests that either fin whales from that area switch between INIs or different populations use the area sequentially (Papale et al., 2023). Investigating the changing patterns of fin whale song INIs in these two regions (Barents seas and SW Portugal) may shed some light on the learning mechanisms of song INIs and the limitations of its variability.

Compared to INIs, fin whale song frequencies of the 20 Hz and HF notes do not vary abruptly but only gradually. Fundamental frequencies of this species’ songs seem constrained by the optimisation of long-range communication in pelagic environments (Clark and Garland, 2022; Payne and Webb, 1971). This song frequency limitation may be an adaptation first, to a dispersed and open water distribution of this species during the breeding season (Edwards et al., 2015; Nieukirk et al., 2004) and second, to match a particular frequency band with low levels of noise in deep waters (i.e. a quiet window in frequency) (Clark and Garland, 2022; Curtis et al., 1999). Comparatively, humpback and right whales (Eubalaena spp.) aggregate in coastal breeding grounds (Clapham, 2018; Kenney, 2009) and use higher frequency songs and calls that transmit better in shallow environments (quiet window: 100–400 Hz) (Clark and Ellison, 2004) and do not need to reach distant conspecifics (Clark, 1982; Clark and Garland, 2022). Therefore, the acoustic environment during the mating season and the species’ breeding behaviour could constrain variation in song frequencies to keep them within the quiet window. In addition, the animals’ physiology can constrain song frequency variation. In birds, the ability to produce low-frequency songs is linked to body size (Ryan and Brenowitz, 1985). If fin whale song frequencies continue to decrease, it can potentially reach the physiological limits of sound production. These limits in song variation can compromise song function and ultimately male fitness when the acoustic habitat in which these songs evolved is changing too rapidly to adapt. For example, the vocal adaptation ability of birds in urban environments (e.g. increasing song frequencies) affect the detection by receivers. If birds are not able to avoid the masking of their songs by noise, this may difficult the establishment and defence of a territory that can ultimately affect their fitness (Habib et al., 2007; Luther and Derryberry, 2012). Similarly, the constraints in fin whale song frequency may limit adaptation to an increasingly noisy environment. Shipping noise, the major source of ocean noise, overlaps in frequency with fin whale songs and can cause a reduction of the communication space (CS) in this species (Clark et al., 2009; Erbe et al., 2019). Models estimate a reduction of CS by vessel noise of up to 80% for fin whales (Cholewiak et al., 2018; Clark et al., 2009). We ignore if fin whales use any anti-masking release mechanism when exposed to vessel noise (e.g. changing song frequencies), but if they do not, such reduction of CS could certainly disrupt communication and hinder the search of mates for reproduction, which ultimately would affect fin whale fitness.

Results from this and other studies suggest that male fin whales are in acoustic contact over vast areas and adjust their song properties to match those of conspecifics (Leroy et al., 2018a; Oleson et al., 2014; Weirathmueller et al., 2017). These acoustic communities culturally evolve more quickly and efficiently than genetic communities (Hatch and Clark, 2004) and should be considered in conservation strategies when delimiting stocks or populations. The acoustic habitat in which these songs evolved has shaped the acoustic properties and limits of variation of these signals. Understanding the cultural evolution of fin whale songs can inform us about the species’ ability to adapt to the actual scenario of rapidly changing ocean soundscapes due to anthropogenic activities. These results also have implications for cue counting approaches, which use cue rates (e.g. notes per unit time) to convert density of sounds to density of animals (Marques et al., 2013). The temporal and spatial changes in fin whale song INIs found here affect cue rates and need to be considered to avoid bias in estimating densities using passive acoustic monitoring in this species. The unique large spatial scale over which fin whales communicate, although technologically challenging for researchers, opens interesting perspectives in the processes of animal acoustic communication.

Materials and methods

Sampling locations

Request a detailed protocol

Acoustic data were compiled from 15 locations in the central and northeast Atlantic Ocean, grouped into seven regions: SE Greenland, SE Iceland, Celtic Sea (North and South Porcupine), Oceanic Northeast Atlantic (ONA) (NE, NW, CE, CW, Azores, SE and SW), SW Portugal, Canary Islands and Barents Sea (Svalbard and Vesterålen; Figure 5A).

Figure 5 with 1 supplement see all
Sampling locations and fin whale song spectrogram.

(A) Locations (stars) of acoustic recordings grouped in regions (colours in stars): SE Greenland (black), SE Iceland (turquoise), Celtic Sea (NP: North Porcupine and SP: South Porcupine; green), Oceanic Northeast Atlantic (NW, NE, CW, CE, Azores, SW and SE; blue), SW Portugal (red), Canary Islands (yellow) and Barents Sea (SV: Svalbard and VE: Vesterålen; purple). (B) Spectrogram (FFT sample duration 0.5 s, Hann window, 50% overlap) of a fin whale song showing the acoustic parameters analysed in this study (INIs and peak frequencies of the 20 Hz and HF note).

Data collection

Request a detailed protocol

Recordings from 1999 to 2020 collected by different research groups with varied objectives were compiled and standardised. Not all regions were sampled in all years and time periods. Recordings were either continuous or duty-cycled with different sampling rates (Table 1 and Figure 5—figure supplement 1). Ocean-Bottom Seismometers (OBS) were used in the Canary Islands (2014–2015) and SW Portugal (2007–2008). The hydrophone channel was selected for OBS recordings in the Canary Islands, while the seismometer channel (vertical component Z) was preferred for recordings from SW Portugal (2007–2008). Fixed autonomous recorders (AR) were used in the remaining regions (Supplementary file 1b).

Song selection criteria

Request a detailed protocol

We focused the analyses on data collected between October and March (hereafter singing season), because fin whale song parameters show less variation during this period (Hatch and Clark, 2004) and seasonal variation was outside the scope of this study. All datasets were manually inspected to identify songs composed of 20 Hz notes (Watkins et al., 1987) or 20 Hz and HF note types (Hatch and Clark, 2004; Figure 5B), except for the Azores dataset, which had been analysed for another study using a Low Frequency Detection and Classification System (LFDCS) (Baumgartner and Mussoline, 2011) (procedures described in Romagosa et al., 2020 ). In all datasets, spectrograms of days with fin whale detections were manually analysed using Adobe Audition 3.0 software (Adobe Systems Incorporated, CA, USA) to select periods with good quality notes, based on: (a) clearly distinguishable song notes in the spectrogram (Signal to noise ratio - SNR >5 dB), (b) absence of masking from noise, (c) presence of a single singer and, (d) occurrence of at least 10 notes organized in a series. The last criterion could not be applied for recordings with small duty cycles (SW Portugal 2015–2016, Azores 2008–2011 and the Celtic Sea) (Table 1); nevertheless, regularly spaced notes could still be identified as part of songs and were used for these sites. SNR was estimated for all selected 20 Hz notes by using the Inband Power measurement in Raven Pro 1.5 software (Cornell Lab of Ornithology, Ithaca, NY, USA) (Supplementary file 1c). For each selected note (Signal), a companion selection (Noise) was created and the Inband Power measured. Then we estimated SNR of each note by using the following formula (Charif et al., 2010):

SNR=Signal Inband Power-Noise Inband PowerNoise Inband Power

Song sampling

Request a detailed protocol

Selected days with detections were non-consecutive to minimize the likelihood of sampling the same individual multiple times. The number of sampled days varied depending on the quality of fin whale songs found in the recordings. The average number of days sampled per singing season was 11.4 days, and the average number of notes analysed per song was 102 (Figure 5—figure supplement 1). Recordings from the Canary Islands, SW Portugal (2007–2008), and ONA regions, except for the Azores, were excluded from the analysis of the HF note, because sampling rates were too low to enable detection of the HF note frequency (~130 Hz) (Hatch and Clark, 2004; Table 1).

Measurement of song parameters: INIs and peak frequencies

Request a detailed protocol

Selected days with good quality notes (see ‘Song selection criteria’ section) were fed into a band-limited energy detector in Raven Pro 1.5 software (Cornell Lab of Ornithology, Ithaca, NY, USA) that automatically selected all 20 Hz and HF notes in the spectrogram. All selections were checked manually by the same analyst to ensure that notes were well imbedded in the selection square. Spectrogram characteristics were adjusted to visualise all data with the same frequency and time resolution (1.25 Hz and 0.4 s). For each selected note, the software measured Begin and End time, Time of the 5% cumulative energy (Time 5%), Peak frequency and Inband Power (Supplementary file 1c). INIs were calculated by subtracting the time (Time 5%) difference between consecutive 20 Hz notes (Širović et al., 2017; Watkins et al., 1987; Figure 5B). This measurement calculates the point in time dividing the selection into two-time intervals containing 5% and 95% of the energy. Peak frequencies were measured for 20 Hz and HF notes and represent the value at which the maximum energy in the signal occurs. It is considered a robust measurement since it is based on the energy within the selection and not the time and frequency boundaries of the selection (Charif et al., 2010). Only one sequence of notes or song fragment (hereafter referred as song) was analysed per day in each location. If multiple songs were found in one day, the one with the highest SNR was selected. For each song, we calculated the mean and standard deviation of INIs and of peak frequencies of the 20 Hz and HF notes.

Analysis of fin whale song INIs and note frequencies

Request a detailed protocol

Temporal patterns in song INIs were investigated by plotting mean song INIs and standard deviations of all regions into chronological order. Due to the identification of two song INIs during the first period of data (1999–2005), belonging to the ONA region, we also calculated the percentage of each song INI per singing season in this dataset. Specifically, the SE location of the ONA region, which had the longest time series (1999–2005), was used to investigate changes in song INI percentages over this period. The other locations of the ONA region had data only for the singing season of 2002/2003 and were used to investigate the spatial patterns in song INIs across six locations (NE, NW, CE, CW, SE and SW) (Figure 5A and Figure 5—figure supplement 1).

After this first period, data from all regions were plotted in chronological order to investigate how song parameters varied over time. A linear regression model was fit to each response variable (INIs and peak frequencies of the 20 Hz and HF notes) using a Gaussian distribution and year as the explanatory variable. Model assumptions were verified by plotting residuals versus fitted values and residual QQ plots to check for homogeneity of variance and normality (Figure 3—figure supplement 1). Measurements of 20 Hz peak frequencies were greatly affected by the recording equipment (Supplementary file 1d and Figure 3—figure supplement 2). For this reason, only data from the Ecological Acoustic Recorders (EARs) (Lammers et al., 2008), which sampled the longest period (2008–2020) (Table 1 and Figure 5—figure supplement 1), were used to explore temporal variations in the peak frequencies of the 20 Hz note. All statistical analyses were performed in R (v. 4.0.2) (R Core team, 2020).

Regional comparison of song parameters

Request a detailed protocol

Given inter-annual variations in fin whale song parameters (Delarue et al., 2009; Širović et al., 2017; Weirathmueller et al., 2017), only songs recorded within the same singing season were used to compare song parameters among regions. Histograms were built for each singing season to investigate differences in the distribution of INIs and peak frequencies of the HF note per region sampled.

Data availability

All datasets and R scripts used in this study have been deposited in the Dryad Digital Repository.

The following data sets were generated

References

    1. Backus R
    2. Schevill W
    (1996)
    Whales, Dolphins and Porpoises
    510–527, Physeter clicks, Whales, Dolphins and Porpoises, Berckley, University of California Press.
  1. Book
    1. Charif RA
    2. Waack AM
    3. Strickman LM
    (2010)
    Raven Pro 1.4 User’s Manual
    New York: The Cornell Lab of Ornithology.
    1. Clapham PJ
    (2018)
    Megaptera Novaeangliae
    489–492, Humpback whale, Megaptera Novaeangliae, Academic Press, 10.1016/B978-0-12-804327-1.00154-0.
  2. Book
    1. Clark CW
    2. Ellison WT
    (2004)
    Potential use of low frequency sounds by baleen whales for probing the environment: evidence from models and empirical measurements
    In: Thomas JA, Moss CF, Vater M, editors. Advances Ion the Study of Echolocation in Bats and Dolphins. Chicago: University of Chicago Press. pp. 564–589.
  3. Book
    1. Clark CW
    2. Garland EC
    (2022) Ethology and behavioral Ecology of Mysticetes
    In: Clark CW, Garland EC, editors. Baleen Whale Acoustic Ethology BT - Ethology and Behavioral Ecology of Mysticetes In. Cham: Springer International Publishing. pp. 11–43.
    https://doi.org/10.1007/978-3-030-98449-6
  4. Conference
    1. Hatch LT
    2. Clark CW
    (2004)
    Acoustic differentiation between fin whales in both the North Atlantic and North Pacific Oceans, and integration with genetic estimates of divergence
    Paper presented to the IWC Scientific Committee.
    1. Kenney RD
    (2009)
    Right Whales: Eubalaena Glacialis
    962–972, Encyclopedia of Marine mammals, Right Whales: Eubalaena Glacialis, London, Academic Press, 10.1016/B978-0-12-373553-9.00220-0.
  5. Book
    1. Kroodsma D
    (2004) The diversity and plasticity of Birdsong
    In: Peter M, Hans S, editors. Nature’s Music: The Science of Birdsong. Cambridge: Academic Press. pp. 108–131.
    https://doi.org/10.1016/B978-012473070-0/50007-4
    1. Lockyer C
    (1984)
    Review of baleen whale (Mysticeti) reproduction and implications for management
    International Whaling Commission 6:27–50.
  6. Conference
    1. Nieukirk SL
    2. Mellinger DK
    3. Klinck R
    (2011)
    Increase in peak fin whale calling levels observed in the mid-Atlantic ocean
    Fifth International Workshop on Detection, Classification, Localization, and Density Estimation of Marine Mammals Using Passive Acoustics. 81.
  7. Book
    1. Payne K
    2. Tyack P
    3. Payne RS
    (1983)
    Progressive changes in the songs of Humpback whales Megaptera Novaeangliae: a detailed analysis of two seasons in Hawaii
    In: Payne R, editors. Communication and Behavior of Whales. Boulder, CO: Westview. pp. 9–57.
  8. Book
    1. R Core team
    (2020)
    R: A Language and Environment for Statistical Computing
    Vienna: R Foundation for Statistical Computing.
    1. Vernes SC
    2. Kriengwatana BP
    3. Beeck VC
    4. Fischer J
    5. Tyack PL
    6. Ten Cate C
    7. Janik VM
    (2021) The multi-dimensional nature of vocal learning
    Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 376:20200236.
    https://doi.org/10.1098/rstb.2020.0236
  9. Conference
    1. Víkingsson G
    2. Gunnlaugsson T
    (2006)
    Stock structure of fin whales (Balaenoptera physalus) in the North Atlantic – indications from non-genetic data
    Joint North Atlantic Marine Mammal Commission (NAMMCO) and International Whaling Commission (IWC) Scientific workshop on the catch history, stock structure and abundance of North Atlantic fin whales.
    1. Whiten A
    (2019) Cultural evolution in animals
    Annual Review of Ecology, Evolution, and Systematics 50:27–48.
    https://doi.org/10.1146/annurev-ecolsys-110218-025040

Decision letter

  1. Luke Rendell
    Reviewing Editor; University of St Andrews, United Kingdom
  2. Christian Rutz
    Senior Editor; University of St Andrews, United Kingdom
  3. Christopher W Clark
    Reviewer; Cornell University, United States

Our editorial process produces two outputs: i) public reviews designed to be posted alongside the preprint for the benefit of readers; ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Fin whale singalong: evidence of song conformity" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Christian Rutz as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Christopher W Clark (Reviewer #1).

Overall, our decision is that the manuscript is not suitable for publication in its current form, but we invite you to revise it following the feedback given below. The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission. The revisions we ask for are substantial, but hope you agree that they will improve the manuscript, and I encourage you to take them on board – failing to do so to the satisfaction of the reviewing team may lead to rejection.

Essential revisions:

1. Having looked at the data and the analysis, we are not convinced the main story here is about conformity. Isn't the picture more about the fact that there are continual gradual changes which raises the question as to how changes could keep going in the same direction once they start to hit physiological limits, and now, we have a possible answer – periodically they reset? This is a really useful piece of understanding, but it would be better if the paper centred on this as a principal question – the gradual changes were already known about in blues and to some degree in fins (e.g. Rankin /Stafford/Castellote Oleson etc. etc.), but now we see how some song features of a male fin whale singer population can remain within certain bounds and not infinitely (and impossibly) reduce or change song characteristics. Please revise in the light of these arguments.

2. We are not convinced the authors are correct in describing this change in rate as a change of song type. No definition of 'type' in fin whale song is given. Even though fin whale calls are evidently a male mating ground display, little is known about its function. Compared to humpback whales with their impressive repertoire of vocalizations, repeating themselves on the breeding grounds after some tens of minutes and therefore qualifying as a very slow 'song' similar to bird song, fin whale songs in the North Atlantic are composed from three simple note types, and they are only changing pitch and timing – it is not really a new type of song. Please revise in the light of these arguments.

3. The abstract states: "We also revealed gradual changes in INIs and note frequencies over more than a decade with all males adopting both rapid and gradual changes." The data are such that the authors could not determine if an individual singer adopted both rapid and gradual changes. A singer only contributed a data point to the overall collection of INI and frequency values. As such, if individual males could not be shown to actually adopt changes in INIs and note frequencies over more than a decade, how could all males be shown to do this? Please reflect on the nature of your data and how it relates to changes within and between individuals in the light of these arguments.

4. Please undertake a major rewrite of the introduction – currently it lacks depth and breadth – improving one of these should be sufficient – for depth, for example, what are the various motivations for understanding song variation in fin whales and in other taxa? L89-91 is not sufficient. Also, a discussion of genetic population structure in fin whales should be included – spatially this study relies on ICES ecoregions but the relevance of these for fin whales is unclear. The inclusion and treatment of avian song learning studies is simplistic and not accurate in places. Given the immense amount of multi-decadal, multi-species, impactful research on captive and free-ranging birds, the reduction of the level of understanding of avian song acquisition to "learning to vocalize through imitation" is both incorrect and unacceptable. At best, a few selected references to the most appropriate discoveries and synthesis should be cited, for example, Marler and Slabbekoorn, Nature's Music: The Science of Birdsong. Consider also Mennill's work on cultural transmission and consider also Otter et al. Curr Bio 2020.

5. It took reviewers some time reading this manuscript to figure out that the analyses were focussed on two different time periods – the ONA data showing the apparent transition from the 19s INI to the 12s INI in the early 2000s and then the broader scale data showing more gradual trends apparently paralleled across the study area. We think the presentation could go a lot further to making these 2 different analyses more clearly distinguished – one local showing abrupt change in one region and then one on a larger scale showing the gradual change over time.

6. The points about conformity on L49 could be better supported by citations – in fact, this is not a well-known definition of conformity – the cultural evolution literature is replete with them – look to papers by e.g. Morgan, Whiten, Laland – and the learning of nearby songs over distant ones can be explained simply by the physics of transmission loss under the reasonable assumption that the clearer you hear a song the more likely you are to learn it. It still leads to interesting patterns of variation and does not argue against social learning still underlying song variation, but we are less convinced by this conformity argument.

7. In places it sounds like the authors are writing from a slightly defensive posture as if they feel bad somehow that fin whale songs are less 'complex' than, say, humpbacks. We encourage the authors to be specific about what is meant by complexity (e.g. number of different song unit types, or number of phrase types in a song), reflect on whether the relationship between complexity and interpretability is as straightforward as they suggest, and finally reflect on why they apparently buy into the idea that 'complexity=interesting' assumption that they feel the need to defend their study against, and revise accordingly.

8. L75 "studies on fin whale song changes have been merely descriptive"

- how is this study not also descriptive? Reviewers disliked the descriptor 'merely' for descriptive studies – try posing a hypothesis without leaning on some form of description of your study system! This might be better focussed on some strengths of this study like temporal and spatial coverage.

9. L79-80 the phrase 'unknown mechanisms of vocal behaviour' is a bit mysterious please be more explicit in your meaning here.

10. L77 this para was vague on the motivations of the study – 'report the dynamics' is a kind of place-holder from which it is hard to predict the actual analyses that were run. Why did you suspect there might be such dynamics? What aspects did you think would be subject to geographic and temporal variation, and why?

11. The lack of proper inclusion and interpretation of previous studies documenting seasonal and multi-annual changes in fin whale INI needs to be addressed (e.g., Delarue et al. 2009, Morano et al. 2012). For example, Morano et al. 2012 shows seasonal INI variations and fluctuation between 10s and 16s.

12. L95 – Figure 1A shows a map, not changes in frequency?

13. L95 – Two 'songs' are described here as if the reader should know what they are but they haven't been defined yet. Perhaps explain here what you mean by the '19s-INI' song and the rules you used to differentiate between the songs. Reviewers would ask you to start with a strictly factual description of the results before adding interpretations such as 'we named these two different INI periods as two different song types'. We are not sure that these are different songs – there are definitely different rhythms, but actually the units and their order are exactly the same. So they might be better described as distinct 'tempos' or 'rhythms' rather than completely different songs?

14. L97 why is only the SE ONA region presented in such detail – what happens at the other recorders concurrently? And why is only 2002/2003 presented for the entire ONA region?

15. L99-100 – define a 'hybrid song' and how they were identified – would also be interesting to know how the different INIs were distributed over the sequence – first one then the other, or more mixed?

16. The text mentions acquisition duty-cycle with a reference to Table S1, but that table does not list the start date-time and end date-time of the sampling. If one goes to Figure S1 showing a color-coded listing of recording effort, the time resolution is in months, and data in this chart is not linked to sample rate or duty-cycle. A reader must be given all the necessary information by which to assess data quality, validity of analysis results and reliability of interpretation.

17. Along similar lines: L300-301 – how was the 'lowest SNR' channel selected – from measurements, estimations, sampling? How was SNR measured? Important details such as duty cycle and sampling rate are in a supplementary table, but I think they are important enough to earn their place in the main text.

18. L177 this sentence confused the reviewers – it seems to contradict itself?

19. L241-242 – isn't the adoption of a new song actually evidence against conformity? Since in a conformist scenario, the new song has to get common enough to be the majority option?

20. L259-260 – We urge the authors to think more critically about the suggestion that humpback song could function as a sonar (for example, considering the long-standing arguments of Au and Cato against this notion) but it is worth noting that if any baleen whale song were to function in some kind of sonar way then they would be better adapted in terms of their regular structure – however, the frequencies being so low really do make it an unlikely function.

21. Text concludes that there was a "complete turnover" of INI. This is a rather strong conclusion given the spatial and temporal sparseness of the data. Is this the right form of words to describe the observations?

22. The time-warping used in Figure 2A, Figure 3A, B and C is a bit misleading. Excluding time gaps in available data distorts the time series and could obscure an actual structure (tendency.) We suggest remaking Figure 2A such that the time scale is uniformly linear, not dilated as in a form of dynamic time warping. As presented, it is difficult to interpret. For example, what does the 2001-2002 period include, October 2001 through March 2002? But it could be that all these samples are from Nov 2001 through Jan 2002. Maybe monthly tic marks would help? How about horizontal lines at each of the integer INI values? This might help the reader see slight increases in INI slope over time? Figure S1 is an incomplete synthesis of the data set. It would be valuable to have more and better represented details. We suggest rearranging this figure so that the time months are based on the singing season, not the human calendar. This would have the left-to-right 6six month period go from Oct through March, thereby representing the behavior from the species' behavioral biology perspective rather than a human calendar. This simplifies the representation of when and where each of the possible 20 singing seasons is sampled. It would also be beneficial to clearly show how much each of the sites contributed to the INI and frequency samples.

23. L309-311 – given that two different methodologies were used (manual and LFDCS), was there any test of the two systems on the same recordings to check they are yield comparable results?

24. L315 specify how long a 'long series' was.

25. The text and figures are not always consistent. For example, the text refers to the E region as the "Bay of Biscay and Iberian Coast", when it contained only a single sensor deployed off the southwest of the Iberian Peninsula. This contrast between the realities of where and when acoustic samples were collected and how the aggregated results are spatially or temporally interpreted or labelled persists throughout the manuscript.

26. Supplementary material needs a table listing the details of the acoustic data recording effort (e.g., from date-time, to date-time) and the acoustic analysis effort (e.g., total recording hours per deployment, for each deployment at each site list the proportion of the total hours relative to the grand total of recording hours, INI-song dates, song time of day start, duration, number of notes.) For example in Figure S1, in January of 2003 for the SE site in the Oceanic Northeast Atlantic region, a song from each of four different days was analyzed yielding a total of 556 INI values for that month. Data from this SE site is the dominant source of recordings in this study. Section 1 in the Supplemental Material is headed in the right direction of addressing sampling aliasing influences on outcome, and this type of effort needs to be included given the large differences in sampling effort/site.

27. We could not find any details on analysis parameters except for the spectrogram figure in Figure 1, where it gives 1024-point FFT, Hann window, 50% overlap. There was no mention of time and frequency resolutions. E.g., if using 1024-point FFT, Hann window, 50% overlap, then the time resolution is 0.26 s and frequency is 1.95 Hz. These resolutions determine the ability to measure and track changes in INI and note peak frequencies. We may have missed seeing a listing of the measurements collected using Raven, but if they weren't there, then they should be included.

28. L335-337 – the measurement of INIs is insufficiently explained – the figure implies this was a separate measurement from the Raven selection boxes, yet there are robust measurement methods in Raven for measuring onsets based on energy content over time (e.g the 5% cumulative energy threshold).

29. L350 song types were known prior to this study? Otherwise, this is a result in the wrong place…

30. L360 – 'except for the Barent's Sea which showed different INIs' – this is a little problematic since it implies you excluded data not for any a-priori reason but because it didn't fit with your preferred model – it is obvious why such a process would not be conducive to rigour. If there is reason to believe the whales from that region belong to an entirely different population and thereby establish an a-priori reason for their exclusion then that should be presented and justified, but excluding data from analyses after results are known isn't a robust way to go about things.

31. The degree to which this is evidence of vocal learning may be a bit overplayed. Definitely there is change, but it is tricky to compare this to e.g. experimental demonstrations. For example, age-related changes in a changing post-whaling demographic scenario should at least be considered?

32. The document contains various grammatical and punctuation errors, and there are several errors in the reference section. Please proof-read the manuscript thoroughly.

Reviewer #1 (Recommendations for the authors):

This is a valuable piece of research and represents a tremendous amount of work. Overall, it is well written and well organized. The extent of these comments reflects the extent of my keen interest in the topic as well wanting to strengthen the manuscript's scientific integrity. With some further effort, it is certainly worthy of publication somewhere, but I have not been asked to render a decision as to where it might best be published.

These following comments are not exhaustive. They are representative. The more I continued to read, reread, and focus on certain aspects of this very interesting and well written paper, the more worried I became about of the influences on sampling aliases on the results and the interpretations of those results. For example, the text mentions acquisition duty-cycle with a reference to Table S1, but that table does not list the start date-time and end date-time of the sampling. If one goes to Figure S1 showing a color-coded listing of recording effort, the time resolution is in months, and data in this chart is not linked to sample rate or duty-cycle. A reader must be given all the necessary information by which to assess data quality, validity of analysis results and reliability of interpretation.

The inclusion and treatment of avian song learning studies is remarkably simplistic and certainly not scholarly accurate. Given the immense amount of multi-decadal, multi-species, impactful research on captive and free-ranging birds, the reduction of the level of understanding of avian song acquisition to "learning to vocalize through imitation" is both incorrect and unacceptable. At best, a few selected references to the most appropriate discoveries and synthesis should be cited, for example, Marler and Slabbekoorn, Nature's Music: The Science of Birdsong.

The lack of proper inclusion and interpretation of previous studies documenting seasonal and multi-annual changes in fin whale INI is perplexing (e.g., Delarue et al. 2009, Morano et al. 2012). For example, Morano et al. 2012 shows seasonal INI variations and fluctuation between 10s and 16s.

Text concludes that there was a "complete turnover" of INI. This is a rather strong conclusion given the spatial and temporal sparseness of the data. Is this the right form of words to describe the observations?

I've already made comments about the sampling bias resulting from the high variability in sampling dates and durations. I found time-warping used in Figure 2A, Figure 3A, B and C inappropriately misleading. Excluding time gaps in available data distorts the time series and could obscure an actual structure (tendency.) I don't doubt that the statistical evaluations in the Supplemental materials are appropriate, but Figure S1 is an incomplete synthesis of the data set. It would be valuable to have more and better represented details. I suggest rearranging this figure so that the time months are based on the singing season, not the human calendar. This would have the left-to-right six month period go from Oct through March, thereby representing the behavior from the species' behavioral biology perspective rather than a human calendar. This simplifies the representation of when and where each of the possible 20 singing season is sampled. It would also be beneficial to clearly show how much each of the sites contributed to the INI and frequency samples.

There is a tendency to slightly, and likely inadvertently, misstate or overstate a result. For example, in the Abstract: "We also revealed gradual changes in INIs and note frequencies over more than a decade with all males adopting both rapid and gradual changes." The data are such that the authors could not determine if an individual singer adopted both rapid and gradual changes. A singer only contributed a data point to the overall collection of INI and frequency values. As such, if individual males could not be shown to actually adopt changes in INIs and note frequencies over more than a decade, how could all males be shown to do this? The unit of analysis needs to be very clearly stated and adhered to throughout the manuscript. The observation of rapid and gradual changes is derived by aggregating INI results from all singers (N = sum of all monthly song samples?) and is not based on the behavior of single individual singers.

The text and figures are not always consistent. For example, the text refers to the E region as the "Bay of Biscay and Iberian Coast", when it contained only a single sensor deployed off the southwest of the Iberian Peninsula. This contrast between the realities of where and when acoustic samples were collected and how the aggregated results are spatially or temporally interpreted or labelled persists throughout the manuscript. Why are ICES ecoregions needed or useful?

Suggest remaking Figure 2A such that the time scale is uniformly linear, not dilated as in a form of dynamic time warping. As presented, it is difficult to interpret. For example, what does the 2001-2002 period include, October 2001 through March 2002? But it could be that all these samples are from Nov 2001 through Jan 2002. Maybe monthly tic marks would help? How about horizontal lines at each of the integer INI values? This might help the reader see slight increases in INI slope over time?

Supplementary material needs a table listing the details of the acoustic data recording effort (e.g., from date-time, to date-time) and the acoustic analysis effort (e.g., total recording hours per deployment, for each deployment at each site list the proportion of the total hours relative to the grand total of recording hours, INI-song dates, song time of day start, duration, number of notes.) If I read Figure S1 correctly, for example, in January of 2003 for the SE site in the Oceanic Northeast Atlantic region, a song from each of four different days was analyzed yielding a total of 556 INI values for that month. Data from this SE site is the dominant source of recordings in this study. Section 1. In Supplemental material is headed in the right direction of addressing sampling aliasing influences on outcome, and this type of effort needs to be included given the large differences in sampling effort/site.

I could not find any details on analysis parameters except for the spectrogram figure in Figure 1, where it gives 1024-point FFT, Hann window, 50% overlap. There was no mention of time and frequency resolutions. E.g., if using 1024-point FFT, Hann window, 50% overlap, then the time resolution is 0.26 s and frequency is 1.95 Hz. These resolutions determine the ability to measure and track changes in INI and note peak frequencies. I must have missed seeing a listing of the measurements collected using Raven, but if they weren't there, then they should be included.

In Figure 1B, it appears that the time of occurrence of a 20-Hz note is taken around the time at which the analyst visually determined the "begin time" of the note. This is not a reliable means of determining the time of occurrence of an acoustic event. The same issue regarding frequency resolution applies (see Fristrup and Watkins 1993, and Raven Pro User's Manual). I don't think this necessarily will make a huge difference in the basic message in the manuscript, but it does impact the resolution and uncertainty in the two major measurements in this manuscript.

The document contains various grammatical and punctuation errors, and there are several errors in the reference section. If I've misinterpreted the text, figures, or supplemental materials, these are certainly unintentional, and I extend my apologies. The paper was very interesting, and I have tried my best to understand it from a variety of perspectives, while being forthright about what I feel must be corrected.

Reviewer #2 (Recommendations for the authors):

It took me some time reading this manuscript to figure out that the analyses were focussed on two different time periods – the ONA data showing the apparent transition from the 19s INI to the 12s INI in the early 2000s and then the broader scale data showing more gradual trends apparently paralleled across the study area. I think the presentation could go a lot further to making these 2 different analyses more clearly distinguished – one local showing abrupt change in one region and then one on a larger scale showing the gradual change over time.

Having looked at the data and the analysis I am not convinced the main story here is about conformity. Isn't the picture more about the fact that there are continual gradual changes which initially raised the query as how this could keep going and now we have an answer – periodically they reset. This is a really useful piece of understanding but it would be better in my view if the paper centred this as a principal question – the gradual changes were already known about in blues and to some degree in fins (e.g. Rankin/Castellote) but now we see how the population can stay within certain bounds and not infinitely (and impossibly) reduce or change song characters. Why this happens – gradual change with periodic 'resets' – is similar but not precisely parallel to humpbacks – mainly because I don't think these are different 'songs' – the units are the same, the structure is the same, just some quantitative aspects of the song change. This doesn't diminish the importance for me, but just is more accurate – I think it would be a mistake to too easily reach for the humpback comparison – the differences are fascinating and worthy of further study – why such simple songs compared to the humpback acoustic peacock tail?? Is it related to functionality, or to different population structure and breeding patterns? And why the changes? I agree such directional changes make conformity a plausible process, and this is especially the case if whales changed their individual songs at the 2000 'reset' – fascinating to reflect on how that might be instigated or triggered, but also important to consider whether range shifts/population movements and/or demographic changes might mean it's actually different individuals or ontogenetic changes.

Introduction:

In general the introduction lacks depth and breadth – improving one of these should be sufficient – for depth, for example, what are the various motivations for understanding song variation in fin whales and in other taxa – L89-91 is not sufficient. Also, a discussion genetic population structure in fin whales should be included – spatially this study relies on ICES ecoregions but the relevance of these for fin whales is unclear.

48 used in a different context or sequence. When individuals are more likely to share song

49 variants with nearby individuals than with more distant ones, we talk about conformity

- this could be better supported by citations – in fact, this is not a definition of conformity I have met before (the cultural evolution literature is replete with them) – look to papers by e.g. Morgan, Whiten, Laland – and to me the learning of nearby songs over distant ones can be explained simply by the physics of transmission loss under the reasonable assumption that the clearer you hear a song the more likely you are to learn it. It still leads to interesting patterns of variation and does not argue against social learning still underlying song variation, but I am less convinced by this conformity argument.

62 has focused on complex songs and vocal learning of rhythm

67 whales or songbirds, offers an easier-to-interpret scenario

– I don't really get these points – it sounds like the authors are writing from a slightly defensive posture as if they feel bad somehow that fin whale songs are less 'complex' than say humpbacks. I would encourage the authors to be specific about what is meant by complexity (e.g. number of different song unit types, or number of phrase types in a song), reflect on whether the relationship between complexity and interpretability is as straightforward as they suggest, and finally reflect on why they apparently buy into the idea that 'complexity=interesting' assumption that they feel the need to defend their study against.

75 far, studies on fin whale song changes have been merely descriptive

– how is this study not also descriptive? I strongly dislike the descriptor 'merely' for descriptive studies – try posing a hypothesis without leaning on some form of description of your study system! This might be better focussed on some strengths of this study like temporal and spatial coverage.

79 scope of cultural song evolution, which provide a unique opportunity to investigate unknown

80 mechanisms of vocal behaviour in this species.

– the phrase 'unknown mechanisms of vocal behaviour' is a bit mysterious to me – could the authors be more explicit in their meaning here.

L77 this para more like an abstract? I found it vague on the motivations of the study –

'report the dynamics' is a kind of place-holder from which it is hard to predict the actual analyses that were run. Why did you suspect there might be such dynamics? What aspects did you think would be subject to geographic and temporal variation, and why?

Methods:

L300-301 – how was the 'lowest SNR' channel selected – from measurements, estimations, sampling? How was SNR measured? Important details such as duty cycle and sampling rate are in a supplementary table but I think they are important enough to earn their place in the main text.

L309-311 – given that two different methodologies were used (manual and LFDCS) was there any test of the two systems on the same recordings to check they are comparable.

L315 specify how long a ‘long series’ was.

L333-334 – specifically what time and frequency resolutions? This is important to put measured changes in context, especially the very subtle frequency ones.

L335-337 – the measurement of INIs is insufficiently explained – the figure implies this was a separate measurement from the Raven selection boxes, yet there are robust measurement methods in Raven for measuring onsets based on energy content over time (e.g the 5% cumulative energy threshold).

L340 what was the minimum number of notes or units required for–a sequence to be included in the analysis?

L342 was ‘highest SNR’ measured or estimated?

L345-346 ‘the only area with recordings during the song transition’.

I struggled with this as part of the methods – surely the song transition is a result of the study, not something that should have been used in making analysis decisions – the implied circularity ('we only looked here because here was the only place that had what we were looking for') is rather seriously undermining. It would be better to set out some clearer general objectives in the introduction and then make it clear that ONA specific analyses were conducted only after a result of interest was found in that location – this is fundamental to not ending up in the garden of forking paths where data and outcomes determine analysis decisions, rather than pre-specified hypotheses.

L350 song types were known prior to this study? Otherwise this is a again a result in the wrong place…

L360 – 'except for the Barent's Sea which showed different INIs' – this is a little problematic since it implies you excluded data not for any a-priori reason but because it didn't fit with your preferred model – I think it is obvious why such a process would not be conducive to rigour. If there is reason to believe the whales from that region belong to an entirely different population and thereby establish an a-priori reason for their exclusion then that should be presented and justified, but excluding data from analyses after results are known isn't the way to go about things.

L361 – explain more why frequency measurements were greatly affected here.

Figure S1 is quite confusing – can this be re-organised as a gant chart with time on the x axis and site on the y?

Figure S2 has some title glitches.

Results:

L95 – Figure 1A shows a map, not changes in frequency?

L95 – two 'songs' are described here as if the reader should know what they are but they haven't been defined yet. Perhaps explain here what you mean by the '19s-INI' song and the rules you used to differentiate between the songs. I think you need to start with a strictly factual description of the results before adding interpretations such as 'we named these two different INI periods as two different song types'. I am not sure I buy that these are different songs – there are definitely different rhythms, but actually the units and their order are exactly the same. So they might be better described as distinct 'tempos' or 'rhythms' rather than completely different songs?

L97 why is only the SE ONA region presented in such detail – what happens at the other recorders concurrently? And why is only 2002/2003 presented for the entire ONA region?

L98 'except FOR'.

L98 'isolated account' – you mean from the Barent's Sea – I think this deserves more than the dismissive treatment here – you are confounding location and time and should perhaps give a little more attention to this observation.

L99-100 – define a 'hybrid song' and how they were identified – would also be interesting to know how the different INIs were distributed over the sequence – first one then the other, or more mixed?

Discussion:

I think perhaps the degree to which this is evidence of vocal learning may be a bit overplayed. Definitely there is change, but it is tricky to compare this to e.g. experimental demonstrations. For example, age-related changes in a changing post-whaling demographic scenario should at least be considered?

L177 this sentence confused me – it seems to contradict itself?

L241-242 – isn't the adoption of a new song actually evidence against conformity? Since in a conformist scenario, the new song has to get common enough to be the majority option?

L259-260 – I urge the authors to think more critically about the suggestion that humpback song could function as a sonar (for example, considering the long-standing arguments of Au and Cato against this notion) but it is worth noting that if any baleen whale song were to function in some kind of sonar way then they would be better adapted in terms of their regular structure – however, the frequencies being so low really do make it an unlikely function.

Reviewer #3 (Recommendations for the authors):

I suggest the authors downplay the rather seeked take-home messages re. song learning and song conformity; just reporting the actual results, interesting as they are, should suffice, without making claims of learning et cetera.

https://doi.org/10.7554/eLife.83750.sa1

Author response

Essential revisions:

1. Having looked at the data and the analysis, we are not convinced the main story here is about conformity. Isn't the picture more about the fact that there are continual gradual changes which raises the question as to how changes could keep going in the same direction once they start to hit physiological limits, and now, we have a possible answer – periodically they reset? This is a really useful piece of understanding, but it would be better if the paper centred on this as a principal question – the gradual changes were already known about in blues and to some degree in fins (e.g. Rankin /Stafford/Castellote Oleson etc. etc.), but now we see how some song features of a male fin whale singer population can remain within certain bounds and not infinitely (and impossibly) reduce or change song characteristics. Please revise in the light of these arguments.

We have reviewed the manuscript according to the reviewers’ perspective that conformity is not the main story that can better reflect our data and analysis. First, we have changed the title from “Fin whale singalong: evidence of song conformity” to “Fin whale song evolution in the North Atlantic”. Then, we have made major changes in the introduction and discussion so that the manuscript does not focus on conformity but on cultural transmission, song evolution and the limits of song variation. We believe that fin whales show enough evidence of song cultural transmission (from this and other studies), and this is why we decided to maintain this topic in the introduction and discussion.

2. We are not convinced the authors are correct in describing this change in rate as a change of song type. No definition of 'type' in fin whale song is given. Even though fin whale calls are evidently a male mating ground display, little is known about its function. Compared to humpback whales with their impressive repertoire of vocalizations, repeating themselves on the breeding grounds after some tens of minutes and therefore qualifying as a very slow 'song' similar to bird song, fin whale songs in the North Atlantic are composed from three simple note types, and they are only changing pitch and timing – it is not really a new type of song. Please revise in the light of these arguments.

The reviewers raise an interesting point here and because we cannot prove that the different INI types here belong to different song types, we have replaced the term song type to song INI type. In the discussion, we hypothesise that these INI types may belong to different song types based on previous evidence that INIs are regionally distinct and could be considered different song types that differentiate populations.

3. The abstract states: "We also revealed gradual changes in INIs and note frequencies over more than a decade with all males adopting both rapid and gradual changes." The data are such that the authors could not determine if an individual singer adopted both rapid and gradual changes. A singer only contributed a data point to the overall collection of INI and frequency values. As such, if individual males could not be shown to actually adopt changes in INIs and note frequencies over more than a decade, how could all males be shown to do this? Please reflect on the nature of your data and how it relates to changes within and between individuals in the light of these arguments.

We acknowledge this sentence was not accurate or did not reflect the nature of our data. We have now removed the terms “all males” by just simply “sampled fin whale songs”, not specifying if all individuals or not adopt song changes.

4. Please undertake a major rewrite of the introduction – currently it lacks depth and breadth – improving one of these should be sufficient – for depth, for example, what are the various motivations for understanding song variation in fin whales and in other taxa? L89-91 is not sufficient. Also, a discussion of genetic population structure in fin whales should be included – spatially this study relies on ICES ecoregions but the relevance of these for fin whales is unclear. The inclusion and treatment of avian song learning studies is simplistic and not accurate in places. Given the immense amount of multi-decadal, multi-species, impactful research on captive and free-ranging birds, the reduction of the level of understanding of avian song acquisition to "learning to vocalize through imitation" is both incorrect and unacceptable. At best, a few selected references to the most appropriate discoveries and synthesis should be cited, for example, Marler and Slabbekoorn, Nature's Music: The Science of Birdsong. Consider also Mennill's work on cultural transmission and consider also Otter et al. Curr Bio 2020.

We have made a major rewrite of the introduction, so conformity is not the principal topic of our manuscript. We now focus on song evolution mechanisms, cultural transmission and the limits of song variation. To do so, we have reviewed a few examples of song evolution in birds, because it is the best studied taxa on the topic, and then we mentioned a few aspects of humpback whale song evolution to go straight to fin whales from there. We have further extended the fin whale paragraph on fin whale songs and added a genetic review from the North Atlantic and how this relates to the acoustic properties of songs in this species. This leads to the motivations of investigating song variations in animals and how can we improve this knowledge. Regarding the area divisions based on the ICES ecoregions, we have adopted our own divisions based on practicality. For example, we have kept the ONA region in the Mid-North Atlantic Ocean facilitate the explanation of song changes in the first sampled period. In the discussion we have added more accurate information on birds’ vocal learning, following the reviewers’ recommendations.

5. It took reviewers some time reading this manuscript to figure out that the analyses were focussed on two different time periods – the ONA data showing the apparent transition from the 19s INI to the 12s INI in the early 2000s and then the broader scale data showing more gradual trends apparently paralleled across the study area. We think the presentation could go a lot further to making these 2 different analyses more clearly distinguished – one local showing abrupt change in one region and then one on a larger scale showing the gradual change over time.

We acknowledge that the structure of the results was confusing, and the different analyses and time periods were not clearly explained. We have changed the titles of the Results section and figures to be clearer and more direct in terms of process, region and sampling period analysed. We have now separated the rapid and gradual changes in 2 sections: 1. Transition in song INIs in the Oceanic Northeast Atlantic region and 2. Gradual changes in song INIs and notes frequencies. Within the first section on rapid song changes (1999 – 2005) we have split the old version Figure 2 into two figures, now Figure 1 only referring to the temporal variation in the SE location; and Figure 2. Referring to the spatial pattern of song INIs in the ONA region (2002/2003). We have also clearly specified these different analyses in the methods, in the section now titled “Temporal and spatial analysis of fin whale song INIs.

6. The points about conformity on L49 could be better supported by citations – in fact, this is not a well-known definition of conformity – the cultural evolution literature is replete with them – look to papers by e.g. Morgan, Whiten, Laland – and the learning of nearby songs over distant ones can be explained simply by the physics of transmission loss under the reasonable assumption that the clearer you hear a song the more likely you are to learn it. It still leads to interesting patterns of variation and does not argue against social learning still underlying song variation, but we are less convinced by this conformity argument.

The manuscript is now not focused on conformity and we have removed most of the references to this process.

7. In places it sounds like the authors are writing from a slightly defensive posture as if they feel bad somehow that fin whale songs are less 'complex' than, say, humpbacks. We encourage the authors to be specific about what is meant by complexity (e.g. number of different song unit types, or number of phrase types in a song), reflect on whether the relationship between complexity and interpretability is as straightforward as they suggest, and finally reflect on why they apparently buy into the idea that 'complexity=interesting' assumption that they feel the need to defend their study against, and revise accordingly.

We have done a major rewrite of the introduction and we do not make this comparison between simple and complex. The motivation of the study has slightly changed, and we do not focus on the advantages of studying simple songs but on the study of song evolution in general (please see response to question 1).

8. L75 "studies on fin whale song changes have been merely descriptive"

- how is this study not also descriptive? Reviewers disliked the descriptor 'merely' for descriptive studies – try posing a hypothesis without leaning on some form of description of your study system! This might be better focussed on some strengths of this study like temporal and spatial coverage.

We acknowledge this statement was inaccurate and inappropriate. With the major re-writing of the introduction and modification of motivations, this statement has been removed. As the reviewer suggest, we have focused on the temporal and spatial coverage of our study as a major strength.

9. L79-80 the phrase 'unknown mechanisms of vocal behaviour' is a bit mysterious please be more explicit in your meaning here.

We have done a major rewrite of the introduction and removed the mention of 'unknown mechanisms of vocal behaviour'.

10. L77 this para was vague on the motivations of the study – 'report the dynamics' is a kind of place-holder from which it is hard to predict the actual analyses that were run. Why did you suspect there might be such dynamics? What aspects did you think would be subject to geographic and temporal variation, and why?

The introduction has been rewritten and consequently the motivation of the study modified. We do not further include “report the dynamics” in the introduction.

11. The lack of proper inclusion and interpretation of previous studies documenting seasonal and multi-annual changes in fin whale INI needs to be addressed (e.g., Delarue et al. 2009, Morano et al. 2012). For example, Morano et al. 2012 shows seasonal INI variations and fluctuation between 10s and 16s.

These studies are now included in the introduction in general terms “the song inter-note interval (INI) is the most distinctive parameter between regions(ref) and have been used to differentiate stocks and populations (ref)” and are also discussed in the first paragraph of the discussion, lines 396-399.

12. L95 – Figure 1A shows a map, not changes in frequency?

We thank the reviewers for spotting this mistake. With all the new figures and arrangements, this has been changed and corrected.

13. L95 – Two 'songs' are described here as if the reader should know what they are but they haven't been defined yet. Perhaps explain here what you mean by the '19s-INI' song and the rules you used to differentiate between the songs. Reviewers would ask you to start with a strictly factual description of the results before adding interpretations such as 'we named these two different INI periods as two different song types'. We are not sure that these are different songs – there are definitely different rhythms, but actually the units and their order are exactly the same. So they might be better described as distinct 'tempos' or 'rhythms' rather than completely different songs?

We have now changed the term song type by song INI type referring to different rhythms and leaving the term song type for discussion in the Discussion section (see response to question 2).

14. L97 why is only the SE ONA region presented in such detail – what happens at the other recorders concurrently? And why is only 2002/2003 presented for the entire ONA region?

We have now rearranged the Results section and figures to clarify the different analyses. Please see response to question 5.

15. L99-100 – define a 'hybrid song' and how they were identified – would also be interesting to know how the different INIs were distributed over the sequence – first one then the other, or more mixed?

We have included further information about hybrid songs in the Results section in lines 313-315.

16. The text mentions acquisition duty-cycle with a reference to Table S1, but that table does not list the start date-time and end date-time of the sampling. If one goes to Figure S1 showing a color-coded listing of recording effort, the time resolution is in months, and data in this chart is not linked to sample rate or duty-cycle. A reader must be given all the necessary information by which to assess data quality, validity of analysis results and reliability of interpretation.

We acknowledge that more information should be given about the sampling effort and songs. For this reason, we have made the following changes and additions: (a) added a paragraph in the Results sections about sampling effort and contributions of each region to the total gross data; (b) added a table (Table 1) showing sampled periods by region and location, the duty cycle and sampling rate and total hours of the recordings, number of INIs and HF note frequencies measures as well as their contribution to the total analysed data; (c) Modified the old supplementary Figure S1 (now Figure 5 —figure supplement 1) to show singing season instead of the normal calendar year and added two columns with days analysed and number of INIs and notes; (d) added a new table in supplementary material (Supplementary file 1a) with information on dates, start and end time, duration and number of INIs analysed for each song of each location.

17. Along similar lines: L300-301 – how was the 'lowest SNR' channel selected – from measurements, estimations, sampling? How was SNR measured? Important details such as duty cycle and sampling rate are in a supplementary table, but I think they are important enough to earn their place in the main text.

In the previous version of the manuscript, SNRs were chosen subjectively by looking at spectrograms. We have reanalysed the data, measured SNRs for each note and selected only notes with SNR > 5 dB. We have specified the methodology in the methods section “Song selection criteria” where we first mention the term SNR. We have also added a table in the main text (Table 1) with information on duty cycle and sampling rates for each sampled region and location (See response to question 16).

18. L177 this sentence confused the reviewers – it seems to contradict itself?

We have removed this sentence as it was not adding any new information.

19. L241-242 – isn't the adoption of a new song actually evidence against conformity? Since in a conformist scenario, the new song has to get common enough to be the majority option?

We have changed the focus of the manuscript and the major topic is not conformity, so we have removed most references to this process (See response to question 1).

20. L259-260 – We urge the authors to think more critically about the suggestion that humpback song could function as a sonar (for example, considering the long-standing arguments of Au and Cato against this notion) but it is worth noting that if any baleen whale song were to function in some kind of sonar way then they would be better adapted in terms of their regular structure – however, the frequencies being so low really do make it an unlikely function.

We have reviewed the literature on this topic and acknowledge this is an unlikely function for humpback whale songs. Consequently, we have removed this sentence from the discussion.

21. Text concludes that there was a "complete turnover" of INI. This is a rather strong conclusion given the spatial and temporal sparseness of the data. Is this the right form of words to describe the observations?

We have removed “complete” from the sentence because it is rather strong considering our data.

22. The time-warping used in Figure 2A, Figure 3A, B and C is a bit misleading. Excluding time gaps in available data distorts the time series and could obscure an actual structure (tendency.) We suggest remaking Figure 2A such that the time scale is uniformly linear, not dilated as in a form of dynamic time warping. As presented, it is difficult to interpret. For example, what does the 2001-2002 period include, October 2001 through March 2002? But it could be that all these samples are from Nov 2001 through Jan 2002. Maybe monthly tic marks would help? How about horizontal lines at each of the integer INI values? This might help the reader see slight increases in INI slope over time? Figure S1 is an incomplete synthesis of the data set. It would be valuable to have more and better represented details. We suggest rearranging this figure so that the time months are based on the singing season, not the human calendar. This would have the left-to-right 6six month period go from Oct through March, thereby representing the behavior from the species' behavioral biology perspective rather than a human calendar. This simplifies the representation of when and where each of the possible 20 singing seasons is sampled. It would also be beneficial to clearly show how much each of the sites contributed to the INI and frequency samples.

We completely agree with reviewers regarding the time scale in the x-axis of these figures. We have modified all figures displaying time in their x-axis scales (Figure 1C, Figures 3A, B and C, and Figure 3 —figure supplement 2) and they now have a uniform linear time scale with years in axis ticks and titles and horizontal lines at each INI value. Figure S1, now Figure 5 —figure supplement 1, has also been modified and now time is represented as singing seasons instead of the normal human calendar. Also, days and number of INIs and HF notes are displayed in two columns next to the chronogram. A new table (Table 1) has been added to the main text with information on contribution of each location to the total number of INIs and HF note peak frequencies measured.

23. L309-311 – given that two different methodologies were used (manual and LFDCS), was there any test of the two systems on the same recordings to check they are yield comparable results?

The manual and automatic methods were only used to extract days with fin whale song notes. The automatic method (LFCDS) had been used in some data of the Azores (2008-2012) for a previous study and we reused this data here. Then, a manual inspection of these potential days was done on all datasets to select songs. We have now better explained this in the methods sections.

24. L315 specify how long a 'long series' was.

Long series referred to a minimum of 10 notes, which is now specified in the methods section.

25. The text and figures are not always consistent. For example, the text refers to the E region as the "Bay of Biscay and Iberian Coast", when it contained only a single sensor deployed off the southwest of the Iberian Peninsula. This contrast between the realities of where and when acoustic samples were collected and how the aggregated results are spatially or temporally interpreted or labelled persists throughout the manuscript.

We have moved methods before results to better clarify regions and locations in the first place and be consistent throughout the manuscript when naming locations. We have now adopted our own delimitations of locations and regions and the old ICES Region E “Bay of Biscay and Iberian Coast” is now refereed as SW Iberia. Now this is clearly specified in the methods and the map’s figure caption, where each region has their locations in brackets.

26. Supplementary material needs a table listing the details of the acoustic data recording effort (e.g., from date-time, to date-time) and the acoustic analysis effort (e.g., total recording hours per deployment, for each deployment at each site list the proportion of the total hours relative to the grand total of recording hours, INI-song dates, song time of day start, duration, number of notes.) For example in Figure S1, in January of 2003 for the SE site in the Oceanic Northeast Atlantic region, a song from each of four different days was analyzed yielding a total of 556 INI values for that month. Data from this SE site is the dominant source of recordings in this study. Section 1 in the Supplemental Material is headed in the right direction of addressing sampling aliasing influences on outcome, and this type of effort needs to be included given the large differences in sampling effort/site.

We have added a new table (Table 1) in the main text with information on sampling period, recording hours and contribution of each location to the total number of INIs and HF note peak frequencies measured. We have also added a new table in supplementary material (Supplementary file 1a) with information on date, times of first and last note, duration and number of INIs measured for each fin whale song analysed.

27. We could not find any details on analysis parameters except for the spectrogram figure in Figure 1, where it gives 1024-point FFT, Hann window, 50% overlap. There was no mention of time and frequency resolutions. E.g., if using 1024-point FFT, Hann window, 50% overlap, then the time resolution is 0.26 s and frequency is 1.95 Hz. These resolutions determine the ability to measure and track changes in INI and note peak frequencies. We may have missed seeing a listing of the measurements collected using Raven, but if they weren't there, then they should be included.

The frequency and time resolution are now incorporated in the main text. In the methods section “Measurement of song parameters: inter-note intervals and peak frequencies” we have included in brackets, information on time and frequency resolution: 1.25 Hz frequency resolution and 0.4 sec.

28. L335-337 – the measurement of INIs is insufficiently explained – the figure implies this was a separate measurement from the Raven selection boxes, yet there are robust measurement methods in Raven for measuring onsets based on energy content over time (e.g the 5% cumulative energy threshold).

In the previous manuscript version, we had measured INIs by calculating the time difference between two begin times of consecutive 20-Hz notes. As the reviewer points out, this was not a robust measurement as it highly depends on the selection box. For this reason, we have reanalysed the entire dataset with Raven to calculate INIs using the 5% cumulative energy threshold, as the reviewer suggests. This analysis has not changed our results. We have also included a description of all Raven measurements in a new table (Supplementary file 1c) placed in supplementary material.

29. L350 song types were known prior to this study? Otherwise, this is a result in the wrong place…

We completely agree with the reviewer and have now renamed song types with INI types (please see response to comment 2).

30. L360 – 'except for the Barent's Sea which showed different INIs' – this is a little problematic since it implies you excluded data not for any a-priori reason but because it didn't fit with your preferred model – it is obvious why such a process would not be conducive to rigour. If there is reason to believe the whales from that region belong to an entirely different population and thereby establish an a-priori reason for their exclusion then that should be presented and justified, but excluding data from analyses after results are known isn't a robust way to go about things.

We agree with the reviewer and acknowledge this was not scientifically correct. In consequence, we have removed this sentence and fitted the linear model considering all regions sampled and recalculated the changing rate.

31. The degree to which this is evidence of vocal learning may be a bit overplayed. Definitely there is change, but it is tricky to compare this to e.g. experimental demonstrations. For example, age-related changes in a changing post-whaling demographic scenario should at least be considered?

We lowered the tone when discussing vocal learning in fin whales, but we still believe there is enough evidence from previous studies that this species is able of vocal learning. We do not affirm that our study proves vocal learning in this species, but we discuss it considering all available evidence of this process (other studies). As the reviewer suggested, we have added a discussion on the age-related changes in a changing post-whaling demographic scenario as a potential driver of song changes in fin whales.

32. The document contains various grammatical and punctuation errors, and there are several errors in the reference section. Please proof-read the manuscript thoroughly.

We have proof-read the manuscript thoroughly.

https://doi.org/10.7554/eLife.83750.sa2

Article and author information

Author details

  1. Miriam Romagosa

    Institute of Marine Sciences - OKEANOS & Institute of Marine Research - IMAR, University of the Azores, Horta, Portugal
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Visualization, Methodology, Writing – original draft
    For correspondence
    m.romagosa4@gmail.com
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2781-5528
  2. Sharon Nieukirk

    Cooperative Institute for Marine Ecosystem and Resources Studies, Oregon State University, Corvallis, United States
    Contribution
    Conceptualization, Resources, Data curation, Writing – review and editing
    Competing interests
    No competing interests declared
  3. Irma Cascão

    Institute of Marine Sciences - OKEANOS & Institute of Marine Research - IMAR, University of the Azores, Horta, Portugal
    Contribution
    Resources, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-6231-0483
  4. Tiago A Marques

    1. Centre for Research into Ecological and Environmental Modelling, University of St Andrews, St Andrews, United Kingdom
    2. Centro de Estatística e Aplicações, Departamento de Biologia, Faculdade de Ciências, Universidade de Lisboa, Lisboa, Portugal
    Contribution
    Supervision, Methodology, Writing – review and editing
    Competing interests
    No competing interests declared
  5. Robert Dziak

    NOAA Pacific Marine Environmental Laboratory, Hatfield Marine Science Center, Corvallis, United States
    Contribution
    Resources, Writing – review and editing
    Competing interests
    No competing interests declared
  6. Jean-Yves Royer

    CNRS - UBO - UBS - Ifremer, IUEM - Lab. Geo-Ocean, Plouzane, France
    Contribution
    Resources
    Competing interests
    No competing interests declared
  7. Joanne O'Brien

    Marine and Freshwater Research Centre (MFRC), Atlantic Technological University, Galway, Ireland
    Contribution
    Resources, Writing – review and editing
    Competing interests
    No competing interests declared
  8. David K Mellinger

    Cooperative Institute for Marine Ecosystem and Resources Studies, Oregon State University, Corvallis, United States
    Contribution
    Resources, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5228-0513
  9. Andreia Pereira

    Instituto Dom Luiz (IDL), Universidade de Lisboa, Lisboa, Portugal
    Contribution
    Resources, Writing – review and editing
    Competing interests
    No competing interests declared
  10. Arantza Ugalde

    Institute of Marine Sciences, ICM‐CSIC, Barcelona, Spain
    Contribution
    Resources, Writing – review and editing
    Competing interests
    No competing interests declared
  11. Elena Papale

    Institute for the Study of Anthropic Impacts and Sustainability in the Marine Environment of the National Research Council of Italy (CNR-IAS), Torretta Granitola, Italy
    Contribution
    Resources, Writing – review and editing
    Competing interests
    No competing interests declared
  12. Sofia Aniceto

    Akvaplan-niva, Tromsø, Norway
    Contribution
    Resources, Writing – review and editing
    Competing interests
    No competing interests declared
  13. Giuseppa Buscaino

    Institute for the Study of Anthropic Impacts and Sustainability in the Marine Environment of the National Research Council of Italy (CNR-IAS), Torretta Granitola, Italy
    Contribution
    Resources
    Competing interests
    No competing interests declared
  14. Marianne Rasmussen

    University of Iceland Research Centre in Húsavík, Húsavík, Iceland
    Contribution
    Resources, Writing – review and editing
    Competing interests
    No competing interests declared
  15. Luis Matias

    Instituto Dom Luiz (IDL), Universidade de Lisboa, Lisboa, Portugal
    Contribution
    Resources, Writing – review and editing
    Competing interests
    No competing interests declared
  16. Rui Prieto

    Institute of Marine Sciences - OKEANOS & Institute of Marine Research - IMAR, University of the Azores, Horta, Portugal
    Contribution
    Resources, Writing – review and editing
    Competing interests
    No competing interests declared
  17. Mónica A Silva

    Institute of Marine Sciences - OKEANOS & Institute of Marine Research - IMAR, University of the Azores, Horta, Portugal
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Validation, Writing – original draft, Project administration, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-2683-309X

Funding

Fundação para a Ciência e a Tecnologia (TRACE-PTDC/MAR/74071/2006)

  • Rui Prieto
  • Mónica A Silva

Fundo Regional para a Ciência e Tecnologia (MAPCET-M2.1.2/F/012/2011)

  • Rui Prieto
  • Mónica A Silva

Fundação para a Ciência e a Tecnologia (FCT-Exploratory-IF/00943/2013/CP1199/CT0001)

  • Rui Prieto
  • Mónica A Silva

Fundação para a Ciência e a Tecnologia (AWARENESS-PTDC/BIA-BMA/30514/2017)

  • Miriam Romagosa
  • Irma Cascão
  • Mónica A Silva

Fundação para a Ciência e a Tecnologia (UIDB/05634/2020)

  • Miriam Romagosa
  • Irma Cascão
  • Rui Prieto
  • Mónica A Silva

Fundación Carmen y Severo Ochoa (CEX2019-000928-S)

  • Arantza Ugalde

National Oceanic and Atmospheric Administration (5326)

  • Robert Dziak

Department of Communications, Climate Action and Environment (ObSERVE Programme)

  • Joanne O'Brien

Consiglio Nazionale delle Ricerche (KUAM (235878/E10))

  • Elena Papale
  • Giuseppa Buscaino

Norwegian Research Council (Calving SEIS (244196/ E10))

  • Elena Papale
  • Giuseppa Buscaino

Norwegian Research Council

  • Sofia Aniceto

Velux Fonden

  • Marianne Rasmussen

Knud Højgårds Fond

  • Marianne Rasmussen

Fundo Regional para a Ciência e Tecnologia (M3.1.a/F/028/2015)

  • Miriam Romagosa

Fundação para a Ciência e a Tecnologia (UIDP/05634/2020)

  • Irma Cascão

Fundação para a Ciência e a Tecnologia (UIDB/50019/2020-IDL)

  • Andreia Pereira

Fundação para a Ciência e a Tecnologia (UIDB/00006/2020)

  • Tiago A Marques

US Navy (Living Marine Resources N3943019C2176)

  • Tiago A Marques

Fundação para a Ciência e a Tecnologia (SFRH/BPD/108007/2015)

  • Rui Prieto

Fundação para a Ciência e a Tecnologia (IF/00943/2013)

  • Mónica A Silva

H2020 European Institute of Innovation and Technology (SUMMER H2020- EU.3.2.3.1)

  • Mónica A Silva

Operational Program AZORES 2020 (01-0145-FEDER-000140)

  • Mónica A Silva

NOAA Pacific Marine Environmental Laboratory (5025)

  • David K Mellinger

H2020 European Institute of Innovation and Technology (GA 817806)

  • Mónica A Silva

Fundação para a Ciência e a Tecnologia (UIDP/05634/2020)

  • Miriam Romagosa
  • Irma Cascão
  • Rui Prieto
  • Mónica A Silva

Governo Regional dos Açores (M1.1.A/REEQ.CIENTÍFICO UI&D/2021/010)

  • Miriam Romagosa
  • Irma Cascão
  • Rui Prieto
  • Mónica A Silva

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

Research was supported by the Portuguese Science & Technology Foundation (FCT), the Azorean Science & Technology Fund (FRCT) and the EC through research projects TRACE-PTDC/MAR/74071/2006, MAPCET-M2.1.2/F/012/2011, FCT-Exploratory-IF/00943/2013 /CP1199/CT0001, AWARENESS-PTDC/BIA-BMA/30514/2017, co-funded by FEDER, COMPETE, QREN, POPH, ESF, Lisbon and Azores Regional Operational Programme, Portuguese Ministry for Science and Education. This work also recieved national funds through the FCT – Foundation for Science and Technology, I.P., under the project UIDB/05634/2020 and UIDP/05634/2020 and through the Regional Government of the Azores through the initiative to support the Research Centers of the University of the Azores and through the project M1.1.A/REEQ.CIENTÍFICO UI&D/2021/010. OKEANOS R&D Centre is supported by FCT through the strategic fund (UIDB/05634/2020). Canary Island data was provided by the Institut de Ciències del Mar under the ‘Severo Ochoa Centre of Excellence’ accreditation (CEX2019-000928-S). Data used from the ONA region is a NOAA-PMEL contribution number 5326. The Celtic Sea data belongs to the ObSERVE Acoustic project, initiated and funded by the Department of Communications, Climate Action and Environment in partnership with the Department of Culture, Heritage and the Gaeltacht under Ireland’s ObSERVE Programme. Data collection in the Barents Sea region was made by the Italian CNR under the Arctic Field Grant project KUAM (235878/E10), funded by the Norwegian Research Council through the Svalbard Science Forum, and under the Project Calving SEIS (244196/E10) funded by the Norwegian Research Council. Vesterålen data was provided by the LoVe Ocean Observatory project, led by the Institute of Marine Research and funded by the Norwegian Research Council and Equinor. Iceland data was collected under Velux Fonden and Knud Højgårds Fond funding. MR was supported by a DRCT doctoral grant (M3.1.a/F/028/2015). IC was supported by the FCT-IP Project UIDP/05634/2020. AP was supported by project AWARENESS 'PTDC/BIABMA/30514/2017' and 'UIDB/50019/2020–IDL'. TAM by CEAUL (funded by FCT - Fundação para a Ciência e a Tecnologia, Portugal, through the project UIDB/00006/2020) and the LMR ACCURATE project (contract no. N3943019C2176). R.P. was supported by an FCT grant (SFRH/BPD/108007/2015). M.A.S. was funded by FCT (IF/00943/2013), EC (SUMMER H2020- EU.3.2.3.1, GA 817806) and the Operational Program AZORES 2020, through the Fund 01–0145-FEDER-000140 'MarAZ Researchers: Consolidate a body of researchers in Marine Sciences in the Azores' of the European Union. We are grateful to Marc Lammers, for providing the EARs and technical support, and to Sérgio Gomes, Norberto Serpa and all skilled skippers and crew that participated in the preparation and deployment of the EARs at DOP/IMAR and all other instruments used in this study. We also thank Dr. Christopher W Clark and the other two reviewers for significantly improving the original manuscript.

Senior Editor

  1. Christian Rutz, University of St Andrews, United Kingdom

Reviewing Editor

  1. Luke Rendell, University of St Andrews, United Kingdom

Reviewer

  1. Christopher W Clark, Cornell University, United States

Version history

  1. Received: September 27, 2022
  2. Preprint posted: October 10, 2022 (view preprint)
  3. Accepted: December 1, 2023
  4. Version of Record published: January 9, 2024 (version 1)
  5. Version of Record updated: January 12, 2024 (version 2)

Copyright

© 2024, Romagosa et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 670
    Page views
  • 136
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Miriam Romagosa
  2. Sharon Nieukirk
  3. Irma Cascão
  4. Tiago A Marques
  5. Robert Dziak
  6. Jean-Yves Royer
  7. Joanne O'Brien
  8. David K Mellinger
  9. Andreia Pereira
  10. Arantza Ugalde
  11. Elena Papale
  12. Sofia Aniceto
  13. Giuseppa Buscaino
  14. Marianne Rasmussen
  15. Luis Matias
  16. Rui Prieto
  17. Mónica A Silva
(2024)
Fin whale song evolution in the North Atlantic
eLife 13:e83750.
https://doi.org/10.7554/eLife.83750

Share this article

https://doi.org/10.7554/eLife.83750

Further reading

    1. Ecology
    2. Plant Biology
    Jamie Mitchel Waterman, Tristan Michael Cofer ... Matthias Erb
    Research Article

    Volatiles emitted by herbivore-attacked plants (senders) can enhance defenses in neighboring plants (receivers), however, the temporal dynamics of this phenomenon remain poorly studied. Using a custom-built, high-throughput proton transfer reaction time-of-flight mass spectrometry (PTR-ToF-MS) system, we explored temporal patterns of volatile transfer and responses between herbivore-attacked and undamaged maize plants. We found that continuous exposure to natural blends of herbivore-induced volatiles results in clocked temporal response patterns in neighboring plants, characterized by an induced terpene burst at the onset of the second day of exposure. This delayed burst is not explained by terpene accumulation during the night, but coincides with delayed jasmonate accumulation in receiver plants. The delayed burst occurs independent of day:night light transitions and cannot be fully explained by sender volatile dynamics. Instead, it is the result of a stress memory from volatile exposure during the first day and secondary exposure to bioactive volatiles on the second day. Our study reveals that prolonged exposure to natural blends of stress-induced volatiles results in a response that integrates priming and direct induction into a distinct and predictable temporal response pattern. This provides an answer to the long-standing question of whether stress volatiles predominantly induce or prime plant defenses in neighboring plants, by revealing that they can do both in sequence.

    1. Ecology
    Congnan Sun, Yoel Hassin ... Yossi Yovel
    Research Article

    Covid-19 lockdowns provided ecologists with a rare opportunity to examine how animals behave when humans are absent. Indeed many studies reported various effects of lockdowns on animal activity, especially in urban areas and other human-dominated habitats. We explored how Covid-19 lockdowns in Israel have influenced bird activity in an urban environment by using continuous acoustic recordings to monitor three common bird species that differ in their level of adaptation to the urban ecosystem: (1) the hooded crow, an urban exploiter, which depends heavily on anthropogenic resources; (2) the rose-ringed parakeet, an invasive alien species that has adapted to exploit human resources; and (3) the graceful prinia, an urban adapter, which is relatively shy of humans and can be found in urban habitats with shrubs and prairies. Acoustic recordings provided continuous monitoring of bird activity without an effect of the observer on the animal. We performed dense sampling of a 1.3 square km area in northern Tel-Aviv by placing 17 recorders for more than a month in different micro-habitats within this region including roads, residential areas and urban parks. We monitored both lockdown and no-lockdown periods. We portray a complex dynamic system where the activity of specific bird species depended on many environmental parameters and decreases or increases in a habitat-dependent manner during lockdown. Specifically, urban exploiter species decreased their activity in most urban habitats during lockdown, while human adapter species increased their activity during lockdown especially in parks where humans were absent. Our results also demonstrate the value of different habitats within urban environments for animal activity, specifically highlighting the importance of urban parks. These species- and habitat-specific changes in activity might explain the contradicting results reported by others who have not performed a habitat specific analysis.