Intra-Seasonal Acoustic Variation in Humpback Whale Songs in the North Colombian Pacific

Abstract

Humpback whales (Megaptera novaeangliae) are well known for their complex acoustic communication, which plays a critical role in social interactions and reproduction. Understanding the variability in humpback whale songs is crucial to deciphering their communication strategies and the factors that influence these changes, which may affect reproductive success and population dynamics. While most studies of humpback whale song behavior have focused on annual variation, intra-seasonal changes remain underexplored. This study investigates intra-seasonal song variation in the Colombian Pacific humpback whale population, a unique and diverse breeding stock. We analyzed 37 h of recordings collected during two distinct periods of the 2019 breeding season (July and August–September) in the northern Colombian Pacific. Song repertoires were compared between periods, and the acoustic structure of a common song unit (Unit1) was analyzed using spectrographic cross-correlation. Results revealed a decrease in repertoire diversity over the course of the season, along with an increase in the song rate and the acoustic consistency of Unit1 during the second period. These findings highlight the dynamic nature of humpback whale song production and suggest potential influences of social learning and hormonal modulation. Such insights may be useful for the conservation and monitoring of humpback whale populations in breeding areas.

1. Introduction

Acoustic sensory systems are fundamental for marine mammals, enabling navigation, foraging, and social coordination. Among these, humpback whales (Megaptera novaeangliae) are particularly notable for their elaborate songs, which play a pivotal role in reproductive and social behaviors. These vocalizations are considered to be among the most complex acoustic phenomena in the animal kingdom, making them a central focus of research in marine bioacoustics. Understanding the mechanisms driving song variability is critical for elucidating the ecological and social dynamics within whale populations and for exploring how such variability may influence reproductive success and survival [1].

Male humpback whales produce songs with frequencies ranging from 20 Hz to 8 kHz [2,3]. These songs are long, complex, and structured, consisting of sound units that form phrases, which in turn, are repeated in different themes [4]. All males within a population generally sing the same songs, a population-specific dialect that converges closely in the same locality [5,6]. While cultural transmission is a widely accepted mechanism for song variation, it is important to consider other factors such as environmental conditions, social interactions, and individual innovation. These factors may interact with learning processes to shape the structure and dynamics of whale songs across populations and seasons.

Examples of song convergence have been documented between geographically distant populations, such as eastern Australia and French Polynesia [7,8], and northeastern Brazil and Gabon [9]. These events have been attributed to migration overlaps or exposure to songs in shared feeding grounds, reflecting the dynamic nature of cultural transmission. However, an alternative hypothesis suggests that humpback whale songs may serve an ecological function, acting as a tool for long-range navigation and providing spatial orientation cues that help whales navigate in complex marine environments [10,11].

Humpback whale songs are variable over time, with gradual divergence occurring over years within a population, and marked variation observed within the same year between populations [5,7,8]. In addition, both inter- and intra-seasonal variation in spectral and temporal traits have been reported [9]. Despite these findings, the mechanisms are poorly understood, and it remains unclear whether this is a widespread phenomenon across populations. Seasonal changes in song structure, repertoire, and activity have been widely studied and documented in birds. In species such as the song sparrow (Melospiza melodia) and the tui (Prosthemadera novaeseelandiae), increases in testosterone (T) levels during the breeding season led to greater song consistency and crystallization [10,11]. Similarly, seasonal changes in T levels are correlated with song activity in great tits (Parus major) [12]. In addition, the maximum number of different syllables is reached in canaries (Serinus canaria) as T levels increase [13]. These phenomena suggest that endocrine regulation plays a critical role in shaping vocal behavior, at least in birds [14,15]. Similar mechanisms may influence intra-seasonal variation in humpback whale song, although this remains to be thoroughly tested.

Despite the abundant breeding population of humpback whales that overwinters along the Colombian Pacific coast, their acoustic behavior remains understudied. The Colombian Pacific population of humpback whales belongs to Breeding Stock G and migrates annually between Antarctic feeding grounds and equatorial breeding areas off Colombia [7,16,17]. This population is considered biologically important due to its high seasonal abundance [18], migratory fidelity [17], and relatively limited acoustic monitoring compared to other well-studied breeding grounds [19,20]. Understanding the intra-seasonal variability in the song of this population is essential for exploring the ecological and social dynamics that shape their communication strategies. Only two published studies have investigated the songs of the Colombian Pacific population. The first described the temporal structure of songs during the 2013 breeding season [20], while the second analyzed specific frequency bands preferred by the population between 2013 and 2016 [19]. Building on this foundation, our study examined humpback whale songs recorded at the beginning and end of the 2019 breeding season in the northern Colombian Pacific. By comparing song repertoires and the acoustic structure of a common song unit across these two periods, this study aimed to assess intra-seasonal song variability within this population and contribute to a broader understanding of their acoustic behavior. Given the well-documented effects of seasonal changes in song activity, repertoire, and consistency in birds [10,11,12,13], we hypothesize that similar changes may occur in humpback whales, leading to an increase in song repertoire, song activity and song consistency as the breeding season progresses.

2. Methods

2.1. Study Area

The study was conducted in two sites, from south to north, the Gulf of Tribugá and Bahía Solano, located in the northern zone of the Colombian Pacific in the department of Chocó (Figure 1), a little-known region with a large number of marine ecosystems and a wide bathymetric gradient with shallow areas and cliffs [21]. This area is characterized by a tropical humid climate, with average temperatures around 28 °C (82 °F) and high annual precipitation, particularly during the rainy season [22]. The Gulf of Tribugá is part of the breeding grounds of the G Stock of humpback whales, and this ecosystem remains relatively undisturbed by large-scale human activities [22]. The Colombian Pacific coast is a critical breeding area for humpback whales (Megaptera novaeangliae), with individuals migrating to this region annually during their reproductive period from late May to mid-October [23]. Since the Colombian humpback whale population migrates between the Antarctic feeding grounds and the equatorial breeding areas off Colombia, moving from south to north, our recordings were performed at two sites, trying to follow the natural migration route of the individuals, recording first at the southern site, and then at the northern site. This seasonal presence provides a unique opportunity to study the acoustic behavior of these whales within an ecologically important habitat.

2.2. Boat Surveys and Acoustic Recordings

The boat surveys took place during the 2019 breeding season, focusing on two non-consecutive periods within the same season. The first sampling period was conducted between 1 and 31 July in the Gulf of Tribugá, resulting in a total of 15 h and 36 min of recordings. The second period took place between 15 August and 9 September in Bahía Solano, where 22 h and 10 min of recordings were obtained. The Gulf of Tribugá and Bahía Solano are known to be key breeding areas for humpback whales along the Colombian Pacific coast. These areas have been considered part of the breeding grounds for Breeding Stock G (G Stock) of humpback whales, based on migratory patterns and known reproductive behavior in the region [16]. While there are genetic and GPS tracking studies that indicate the presence of distinct genetic populations in the Colombian Pacific [16,24], further research is needed to conclusively delineate these populations.

Acoustic recordings were made opportunistically whenever singing males were detected. To detect singing individuals, an acoustic inspection was carried out every 20 min. Once a song was detected, it was recorded for the entire song cycle or for as long as environmental conditions, such as weather and noise, permitted. Recordings were made using an SQ26-08 hydrophone from Cetacean Research Technology, connected to a ZOOM H1 digital recorder configured with a sampling frequency of 44 kHz and 24-bit resolution. Individuals were not visually identified during the recordings, but each recording session was conducted continuously to ensure that each song recorded corresponded to a single individual.

During the recording periods, typical environmental conditions included warm daytime temperatures around 29 °C and very high precipitation, as the region is known for its heavy rainfall. Given these conditions, surveys were conducted only on days with favorable weather, specifically those without rain and with low wind. This ensured that sea conditions were suitable for acoustic recordings, minimizing ambient noise interference and ensuring data quality.

2.3. Repertoire Analysis

The acoustic structure and song repertoires were analyzed using the RAVEN PRO 1.6.1 software [25]. A sound unit was defined as the shortest continuous sound perceptible to the ear in real time [4]. For each recorded song, all sound units were extracted and saved as individual files. A spectrographic cross-correlation analysis (SPCC) was performed between the song units of each recording. SPCC is a technique described by Clark et al. [26] that simultaneously analyzes the frequency, amplitude, and temporal components of a signal and produces a paired matrix of correlation coefficients between the analyzed units. Following the method by Dalisio et al. [27], we performed a cluster analysis using Euclidean distances on the correlation matrix. The analysis was performed using RStudio (version 1.3.1093), R (version 4.0.3, R Core Team, 2024), and the ‘stats’ package. The clustering allowed us to visualize groups of similar sound units [27].

In this study, a unit refers to a discrete acoustic element within the song, characterized by specific parameters such as frequency, duration, and modulation pattern [2,4]. These units are the fundamental building blocks of humpback whale songs, and their acoustic properties may vary depending on the context or stage of the breeding season [17,28]. A unit type was defined as a set of acoustic units with a spectrographic correlation greater than 0.6, ensuring a high degree of similarity among the units. This classification allows for consistent grouping of similar units within a single recording. An individual repertoire refers to the complete set of unit types produced during a single recording, providing an overview of the diversity of the whale’s song. Once the individual repertoires were established, we performed a second SPCC analysis across different recordings to create two combined repertoires, gathering all songs recorded either at the beginning or the end of the breeding season.

To quantify the similarity between song repertoires recorded at the beginning and end of the breeding season, we used the Jaccard index. This index measures the proportion of shared unit types between two sets, ranging from 0 (no similarity) to 1 (complete similarity). Similarity indices, such as the Jaccard index, are commonly used in the study of acoustic patterns in animal vocalizations [29].

2.4. Acoustic Analysis

To compare acoustic parameters between recordings, a highly repetitive unit type present in all recordings was selected (Figure 2). Spectrographic analysis of this consistent unit type allows for a temporal comparison of spectral changes while avoiding potential biases that could arise from including all different unit types, which may obscure significant differences or confound the results. This unit was selected not only for its ubiquity but also because its stereotyped structure across recordings made it particularly suitable for reliable SPCC analysis. The unit type selection process involved two steps: First, the frequency contour of all units was analyzed to identify similar patterns across recordings. Second, Raven’s Band-Limited Energy Detector tool was used to refine the selection based on frequency and duration ranges. The detector was adjusted until a similarity threshold of at least 60% was reached, as determined by an SPCC. The unit type that met this criterion was designated as Unit type 1 for all subsequent analyses.

Once Unit type 1 was extracted from all recordings, we measured robust acoustic parameters that minimize inter-observer variability and provide reliable, quantitative measurements. The parameters included the following:

• Peak Frequency (Hz) and its contour (PFC), which include maximum and minimum frequencies as well as maximum, minimum, and average slopes, giving insights into the dominant frequencies and their temporal evolution within each unit.
• Delta Frequency, Center Frequency, and Delta Time, which capture variations in the acoustic structure of the units.
• Bandwidth 90% (Hz) and Duration 90% (s), focusing on the core energy of the signal to ensure consistency across recordings.

These parameters are critical as they reduce subjective interpretation, ensuring that comparisons between recordings and locations are objective and reliable.

2.5. Statistical Analysis

Statistical analyses were performed using R 3.5.0 software [31]. Since individual whales were not identified, we could not perform tests for dependent variables, such as repeated measures from the same individual. Each recording was treated as independent. All tests were two-tailed, and the alpha level was set at 0.05. A comparison of unit production between the first and second recording periods was made, adjusting for recording time in each period. A Chi-squared test was applied to the rate of unit production per hour between the two periods. Then, the values of the acoustic parameters of Unit1 were statistically compared between the two recording periods. For each parameter, the normality of the data was assessed graphically using Q-Q plots. For normally distributed variables, the effect of the period on parameter values was analyzed using mixed linear models, with individual recording included as a random factor. An ANOVA (F-test) was applied to these models. Model quality was assessed by checking the normality of the residuals of the distribution model. This was performed using the check-model function from the RVAideMemoire package. For non-normally distributed variables, Mann–Whitney tests were used to compare the two periods. To control for the effect of individual recordings, these tests were applied to the mean values calculated for each recording (N = 9 in July and N = 12 in August–September). p-values were adjusted for multiple comparisons using the “false discovery rate” (FDR) method [32], which controls the expected proportion of false positives among the rejected hypotheses.

3. Results

3.1. Repertoire Analysis

A total of 1293 units were recorded in the first period and 2563 units in the second period, with recording times of 4.6 h and 7.3 h, respectively. The production rate of units per hour was significantly higher during the second period (X² = 42.683, df = 1, p < 0.001). Spectral cross-correlation (SPCC) analysis of the vocal units identified a total of 819 unit types. These included 474 types unique to the first recording period, 271 types unique to the second period, and 74 types shared between both periods (

Table 1. Summary of Vocal Repertoire Analysis.

Category	Unit Types Count
First recording period (unique)	474
Second recording period (unique)	271
Shared units	74
Total	819
Jaccard Index	0.0904

; see Supplementary Material Figures S1–S3 for detailed repertoires). The Jaccard index value obtained was 0.0904, indicating low similarity between the song repertoires recorded during the two periods of the breeding season. This indicates that whales in the second period produced a higher rate of song units, but with a lower diversity, compared to the beginning of the season.

3.2. Acoustic Analysis

We found no significant differences in the acoustic parameters of Unit1 between recordings in July and August–September, i.e., for Peak Frequency (F = 1.625, p = 0.341), Delta Frequency (W = 34, p = 0.341), Centre Frequency (F = 0.788, p = 0.459), PFC Maximum Frequency (W = 87, p = 0.101), PFC Minimum Frequency (W = 51, p = 0.862), PFC Average Slope (F = 0.695, p = 0.459), PFC Maximum Slope (W = 75, p = 0.341), PFC Minimum Slope (W = 41, p = 0.459), Duration 90% (F = 3.276, p = 0.312), and Delta Time (F = 6.586, p = 0.341), although the Bandwidth 90% tended to decrease between July and August (W = 91, p = 0.080) (Figure 3).

In addition to the individual acoustic parameters, we compared the average correlation values (obtained by SPCC) of Unit1 across recordings from the two periods. A Mann–Whitney U test revealed a highly significant difference in correlation values between the July and August–September recordings (W = 6,498,426, p < 0.001). Specifically, the correlation values were higher in the second period, indicating increased consistency in Unit1 as the breeding season progressed (Figure 4).

4. Discussion

Understanding intra-seasonal changes in humpback whale songs is essential for unraveling the mechanisms that drive song variation. Our study revealed two key patterns in the Colombian Pacific breeding population: (1) a significant increase in the number of acoustic units produced per hour in the second recording period (August–September) and (2) greater acoustic consistency in a common song unit (Unit1) during the later phase of the breeding season. These findings align with previous studies documenting seasonal changes in whale song structure [33,34] and provide new insights into how song variability unfolds within a single breeding season.

A striking finding of this study was the significant increase in unit production rate during the second sampling period. This result is consistent with our prediction and previous findings in birds, where an increase in singing activity was recorded as the breading season progressed [12]. In birds, increases in singing activity have been explained by elevated T levels, both correlational [12] and experimentally [35]. The role of testosterone in modulating song activity in birds is well documented [36]. T levels correlate with testes size in vertebrates [37], song activity [15,36] and territoriality [36,38] among other traits. Circulating T varies throughout the breeding season, increasing as the season progresses [36]. T levels also vary throughout the season in humpback whales [39], suggesting that a similar mechanism may drive variation in song rate in this species.

While a higher number of acoustic units were recorded in August–September, the diversity of unit types decreased, indicating a reduction in repertoire size. This result is in contrast to our initial prediction, which expected a broader repertoire later in the season. The observed decline in unit diversity suggests that, rather than continuously introducing novel elements, singers may favor a more selective subset of units as the season progresses. Interestingly, this pattern also suggests a shift toward a more repetitive and consistent song structure. Instead of introducing new elements over time, as observed in long-term song evolution [33,34], our findings highlight a process of song simplification or crystallization within a single breeding season. This intra-seasonal trend may reflect functional constraints favoring stereotyped sequences, which could enhance signal efficiency for communication or reproductive success.

Another interesting finding was that our humpback whale population displayed higher song consistency as the season progressed. This intra-seasonal shift in song consistency is reminiscent of patterns observed in other vocal learning species, such as songbirds, where hormonal regulation influences vocal plasticity [10,11]. In birds, elevated testosterone levels during the breeding season promote song stability and stereotypy, thereby increasing signal reliability for conspecific communication [10,36]. Studies in other bird species have shown that testosterone influences not only song stability but also aggressiveness and territorial behaviors during the breeding season [38]. In great tits, seasonal testosterone fluctuations modulate both song output and parental behaviors [12,35], while in other species, testosterone has been linked to individual differences in singing effort and response to social challenges [36,38]. These findings highlight the multifaceted effects of hormonal modulation on vocal performance and social interactions and may offer valuable parallels for interpreting the seasonal acoustic shifts observed in humpback whales. If similar endocrine mechanisms operate in humpback whales, hormonal fluctuations may play a role in shaping seasonal song consistency. However, further studies incorporating hormone measurements would be needed to explore this hypothesis.

The observed increase in Unit1 consistency raises questions regarding the mechanisms underlying this change. The cultural transmission hypothesis posits that song modifications arise through social learning, with males incorporating and propagating new song elements over time [33,34]. If this were the dominant mechanism, we would expect continuous innovation and divergence in song structure throughout the season. Instead, we observed a reduction in Unit1 diversity, which is not fully consistent with predictions of progressive song modification. An alternative explanation is that changes in song structure are influenced by external factors, such as environmental conditions or shifts in male competitive dynamics. In species where vocal displays play a role in sexual selection, stereotyped signals may improve communication efficiency, facilitating mate attraction and male–male assessment [40]. The increased consistency in Unit1 may thus reflect selection for stable vocal signals, which could enhance their effectiveness in reproductive contexts [41].

Another possible mechanism is the sonar hypothesis, which suggests that humpback whale songs serve an echolocation-like function to aid navigation [17,18]. According to this hypothesis, maintaining consistent acoustic features could optimize the generation of predictable echoes that aid spatial orientation in complex marine environments [18,42]. However, the variability in conditions between the first and second recording periods—such as changes in group composition and the intensity of social interactions—could influence the whales’ navigational needs and song behavior. While this hypothesis provides a plausible framework for understanding the observed increase in song consistency, further evidence is needed to clarify how these environmental and social factors contribute to the variation in acoustic behavior across the breeding season.

Our results provide valuable insights into intra-seasonal song dynamics; however, certain limitations should be acknowledged. A major limitation is the inability to follow individual singers across recording periods, which prevents us from determining whether the observed changes reflect true within-individual song modification or shifts in the composition of the singing population. Male humpback whales are known to move in and out of the breeding grounds throughout the season [28], meaning that the observed variation may be due in part to inter-individual differences rather than progressive changes within the same individuals.

Given the challenges in identifying and recording the same individual across different sampling periods and the low probability of capturing the same singers multiple times, the focus of the analysis has been shifted to broader changes observed at the stock level. The observed variation in song characteristics reflects the acoustic dynamics of the population across the breeding season, rather than individual-level vocal changes. This broader perspective emphasizes population-level patterns in vocal behavior, providing insights into how humpback whale songs evolve throughout the season.

Another limitation is the high vocal activity observed between recording periods (early August), which prevented us from obtaining clean, isolated recordings of individual singers. During this peak period, overlapping songs created a cacophony of sound, making it difficult to analyze fine-scale vocal variation. While this intense acoustic environment may itself have ecological significance, potentially indicating increased social interactions or male competition, it also limited our ability to assess individual-level song modifications.

In addition to the acoustic analysis, the genetic structure of the humpback whale population in the Colombian Pacific remains unclear. While our study assumes that the population under consideration belongs to a single breeding stock, recent studies provide evidence of a more complex structure. Caballero et al. [16] identified three distinct genetic groupings in Colombian waters, suggesting that the population may not be as homogeneous as previously thought. Additionally, a recent study using satellite tracking indicates that tracks of tagged whales from Costa Rica and Panama passed close to these study sites, suggesting overlap in migration routes [24]. However, the migration of the northern individuals began between 23 August and 7 October each year [24], reducing the likelihood of overlap with our study population. These findings highlight the need for further genetic studies to clarify the population structure and to better interpret the implications of our acoustic observations. Therefore, future research integrating genetic analyses and individual identification methods will be crucial to disentangle the contributions of inter-individual variability from seasonal trends in song modification.

Despite these challenges, this study provides the first detailed analysis of intra-seasonal song variation in the Colombian Pacific humpback whale population. Future research integrating individual identification methods (e.g., acoustic tracking and photo-identification) and physiological measures (e.g., hormone levels) will be essential to disentangle the contributions of inter-individual variability from true seasonal trends in song modification.

These findings also have important implications for the conservation and management of humpback whales in the Colombian Pacific. Understanding how acoustic behavior varies within a breeding season can inform the timing and spatial regulation of human activities such as whale-watching or maritime traffic. For example, the observed increase in vocal activity and Unit1 consistency later in the season may reflect periods of heightened social or reproductive interactions, during which whales could be more sensitive to acoustic disturbances. Incorporating knowledge of intra-seasonal acoustic dynamics into management strategies could help minimize potential disruptions during critical phases of the breeding season and support more effective long-term monitoring of population health and behavior.

This study reveals significant intra-seasonal changes in humpback whale song structure, with increased unit production rates and enhanced Unit1 consistency later in the breeding season. These findings contribute to our understanding of how whale songs vary over short timescales and suggest that factors beyond cultural transmission—such as hormonal regulation, sexual selection, or navigation—may shape song variability. By further investigating these mechanisms, we can refine our understanding of the ecological and evolutionary forces driving acoustic communication in marine mammals.

5. Conclusions

This study provides the first detailed analysis of intra-seasonal song variation in the Colombian Pacific humpback whale population. By comparing recordings from the beginning and end of the 2019 breeding season, we observed a significant increase in the rate of unit production and a shift toward greater acoustic consistency in a common unit type. Contrary to our initial hypothesis, the repertoire diversity decreased over time, suggesting a process of song simplification rather than progressive innovation. These findings suggest that factors beyond cultural transmission, such as hormonal regulation, sexual selection, or navigational functions, may shape the structure of humpback whale songs over short timescales. Future studies incorporating individual identification and physiological measures will be critical for disentangling the mechanisms driving these seasonal changes.