Whistles of Atlantic spotted dolphin from a coastal area in the southwestern Atlantic Ocean Open Access

Currently, there is a data gap on the acoustic repertoire of several odontocete species in the southwestern Atlantic Ocean (SWAO). Although there are efforts to differentiate such species through their vocal repertoire (Amorim et al., 2019; Lima et al., 2016), more systematic studies on the characteristics of delphinid vocalizations are needed in the area in order to efficiently perform such differentiation. In addition, increasing anthropogenic activities expand the exposure of marine mammals to noise pollution in the area, highlighting the necessity of obtaining baseline bioacoustics data in order to understand potential effects of noise pollution on each species.

At the Brazilian continental shelf many species of the Stenella genus are sympatric (Moreno et al., 2005). The Atlantic spotted dolphin, Stenella frontalis, is one of the most sighted, since this species has a preference for shallow waters in comparison to other Stenella spp., occurring from coastal areas to the continental slope (Moreno et al., 2005). Despite being one of the most sighted Stenella spp. in the SWAO, only one study has focused solely on this species' whistles, but they were recorded with a limited sampling rate of 48 kHz (Azevedo et al., 2010), and two others used this species in an inter-species comparison (Amorim et al., 2019; Lima et al., 2016). In contrast, the acoustic repertoire of S. frontalis has been well studied in multiple populations of the North Atlantic Ocean (Gannier et al., 2020; Lammers et al., 2003; Papale et al., 2015; Papale et al., 2016), where acoustic characterization has allowed the identification of a broad frequency range in their whistles that varies according to underwater noise and social context. Therefore, the acoustic behavior of this species still demands investigation to be used for managing and conservation efforts.

Field surveys were conducted between the years of 2017 and 2020 in Ilha Grande Bay (23°09′S; 44°28′W), southeastern Brazil, on outboard motorboats. Survey scans were made in different areas of the bay in different times of day until a delphinid sighting occurred. In this three-year period, there were thirteen sightings, totalizing 4 h and 53 min of sound recordings (Table 1). During recordings, three behavioral states were observed (travel, feeding, and socializing), classified following the categories described by Shane (1990). Group size varied between 9 and 200 individuals (108.8 ± 46.6), and in one occasion there was one sighting and recording of a solitary adult individual. No information is known on the identification of specific individuals or the composition of the groups sighted. The recording system consisted of an omnidirectional hydrophone model C54XRS (−165.0 ± 3 dB/V re 1 μPa, 0.009–100 kHz) coupled to a digital recorder model Fostex FR22 operating at 192 kHz sampling rate and 24-bit to create .wav files. Recordings lasted from two to five minutes. During recordings, the hydrophone was placed approximately 4 m below the surface with the boat's engine turned off.

Acoustic analyses were carried out in raven 1.5 (BRP, 2011). First, all acoustic files were manually visualized to identify whistles in the recordings. As no recorded whistle contained energy in the fundamental frequency band above 30 kHz, acoustic files were downsampled to a 96 kHz sampling rate. Spectrograms were then generated with 50% overlap, and a 1024 point Hann window, and visualized in 2 s time periods. Subsequently, all sounds clearly identifiable as whistles in the recordings were counted for the calculation of the whistling rate (total number of whistles in the recording divided by the recording duration). This analysis did not exclude weak or overlapped whistles, since the goal was to estimate whistling activity, independent of the situation (e.g., whistling bouts, quiet moments from the animals, behavioral state, etc.). The overall whistling rate was estimated, as well as the whistling rate for each behavioral state.

Next, we performed a selection among the counted whistles in order to extract acoustic parameters. Whistles with a strong contour, for which the start and end could be clearly identified, were selected for analysis. For this analysis, whistles that had the exact same contour were not selected more than three times in each recording to avoid oversampling (Oswald et al., 2004). Seven acoustic parameters were measured for all whistles selected for analysis, based on previous acoustic studies of delphinids (Bazúa-Durán and Au, 2004; Lima et al., 2016; Oswald et al., 2004): duration (ms), number of inflection points (an inflection being the point where there is a change in the whistle contour from ascending to descending or vice-versa), start frequency (kHz), end frequency (kHz), minimum frequency (kHz), maximum frequency (kHz), and delta frequency (kHz). We also calculated a mean frequency (kHz) as the mean of four of the measured frequency parameters, excluding delta frequency, following Azevedo et al. (2010). The selected whistles were also classified as belonging to a major contour category: ascending, ascending-descending, descending, descending-ascending, multiple, or constant (Steiner, 1981). These data are referred to as the “present study dataset.”

In addition, a recording was made of a solitary individual sighted on one occasion, containing a total of approximately 103 whistles, of which only 10 had a clear and strong contour. The other 93 whistles were very faint and difficult to recognize. No other animals were sighted during this recording, indicating that the weak whistles probably belonged to a group outside the viewing range of the observers. Therefore, these ten good quality whistles were assigned to the lone individual and clearly had the same contour, confirmed separately by two experienced observers classifying whistles, indicating they were stereotyped whistles. For these ten whistles the time interval between repetitions was measured, as well as the eight acoustic parameters previously described. This procedure was performed only for this situation due to the unique opportunity of associating the whistling behavior and a signature whistle to the lone individual. The behavioral state for this solitary adult could not be determined in the field, therefore we did not use its whistles in behavioral comparisons.

In order to investigate whistle variation, we performed comparisons between behavioral states. First, to verify if the composition of contour categories of whistles were similar among the three observed behavioral states, a Pearson's chi-square test was employed. Then, a Mann-Whitney U test with a significance level set at p < 0.05 in software statistica 7.0 was used to compare whistles' acoustic parameters from travel and feeding behavioral states. Due to large differences in sample size, we did not use whistles from social behavior in this comparison. For these procedures, only the present study dataset was employed.

Aiming to compare whistles collected in Ilha Grande Bay with those from other populations, we also used data collected by Azevedo et al. (2010), which recorded S. frontalis in the same area between 2007 and 2008 with a 48 kHz sampling rate for a total of 4 h and 39 min (Table 1). Due to the larger sample size of 1092 whistles gathered by Azevedo et al. (2010), we randomly selected 300 whistles from their work in order to reach a comparable N to ours (Table 1), which we referred to as the “Azevedo dataset.” This dataset was used separately from the present study dataset. First, we compared the acoustic parameters of whistles from the present study and the Azevedo datasets employing a Mann-Whitney U Test with a significance level set at p < 0.05. Then, in an effort to further investigate S. frontalis whistles from Ilha Grande Bay in relation to other areas, we applied a two-sided T test with a significance level set at p < 0.01 (Bazúa-Durán and Au, 2002) to compare the means of acoustic parameters with data from three S. frontalis populations inhabiting the North Atlantic: Little Bahama Bank (Lammers et al., 2003), Canary Archipelago (Papale et al., 2015), and Azores (Gannier et al., 2020).

A total of 6876 whistles were counted; 46.9% occurred during travel, 43.4% during socializing, and 9.7% during feeding. Groups emitted on average 33.5 whistles/min, with situations varying from 0.5 to 318 whistles/min. The groups emitted a mean of 26.8 whistles/min during travel behavior, 248.3 whistles/min during socializing, and 9.8 whistles/min during feeding. Of all whistles, 330 fitted our selection criteria. Ascending whistles represented the majority of whistles (37.1%), followed by multiple (26.6%), ascending-descending (22.2%), and descending-ascending (12.9%). Whistles with constant or descending contours were not observed. This represented a change from the Azevedo dataset 300 whistles, in which all categories were observed: ascending (56%), descending-ascending (13%), multiple (12.7%), ascending-descending (10%), descending (4.3%), and constant (3.7%).

The use of whistle categories in the present study differed between behaviors [χ²(4, N = 330) = 29.6, p = 0.00005), with ascending whistles being the majority during travel, multiple whistles being the majority during feeding, and ascending-descending being the majority during socializing (Table 2). The acoustic parameters of present study whistles varied according to the behavioral state and are presented in Table 2. Although socializing behavior was not used in the statistical comparison, their descriptive values are also presented. During travel whistles had shorter duration (MW, p < 0.01) and higher start (MW, p = 0.000295), end (MW, p = 0.003707), and minimum frequencies (MW, p = 0.000058) than during feeding.

The present study dataset (N = 330) and the Azevedo dataset (N = 300) had differences in all acoustic parameters (Table 3). Present study whistles were longer with more inflection points, and had higher end, maximum, mean, and delta frequencies (MW, p < 0.01). The Azevedo dataset had whistles with higher start and minimum frequencies (MW, p < 0.01). Comparisons indicated differences and similarities between Ilha Grande Bay S. frontalis from both datasets and other populations, which are detailed in Table 3. Frequency parameters from the Azevedo dataset had more differences than similarities with North Atlantic whistles, having lower means of start, end, maximum and delta (two-sided T test, df = 1000, p < 0.01) frequencies than the other three populations. Whistles from the Azevedo dataset were also shorter and had less inflection points than those from Azores (two-sided T test, df = 1000, p < 0.01), but similar in duration to those from Bahama Bank (2003). Many of these differences shifted when comparing present study whistles with the others. Maximum frequency, for example, had a higher mean in Ilha Grande Bay than in the Little Bahama Bank (two-sided T test, df = 1000, p < 0.01) and reached a similar mean to the Azores whistles. In contrast, the means of start, minimum and maximum frequencies were still lower than those of the Canary Archipelago (two-sided T test, df = 1000, p < 0.01).

The solitary individual emitted ten ascending-descending whistles over 10 min that had the exact same contour (Fig. 1). The average interval between these ten whistle emissions was of 80 s; with the shortest interval of 8.3 s and the longest of 200 s. The acoustic parameters of these whistles had little variation (Table 3), thus are stereotyped and may be the signature whistle of this individual.

The selected whistles from this study had a wide frequency range, from 1.3 to 29.0 kHz; a range that is similar to what is reported by other whistle studies with S. frontalis that also used recording sampling rates larger than 48 kHz [e.g., Lammers et al. (2003), Papale et al. (2015), and Gannier et al. (2020)]. The maximum frequency values reported are significantly larger than those reported for the previous study in Ilha Grande Bay in the SWAO (Azevedo et al., 2010) and for the continental slope of southern Brazil [maximum of 19.6 kHz, Amorim et al. (2019)], which both used a sampling rate of 48 kHz. The observed differences between populations from the South and North Atlantic in start and minimum frequency suggest that S. frontalis in the South Atlantic has a broader frequency range than previously thought. The use of sampling rates of at least 48 kHz is important for adequate characterization of delphinid whistles (Oswald et al., 2004), and it is possible that to differentiate populations of the same species through their whistles this sampling rate value must be higher, such as 96 kHz.

Whistle duration increased significantly from the Azevedo dataset to the present study, and was similar to the longer whistles of S. frontalis reported by Gannier et al. (2020). However, the number of inflection points was lower than that reported for S. frontalis in other studies [e.g., Gannier et al. (2020)]. Therefore, the variation in these parameters between datasets and among populations may be due to other factors such as behavior (Bazúa-Durán and Au, 2004; Papale et al., 2016). Azevedo et al. (2010) showed that agitation state influenced the acoustic parameters of S. frontalis in Ilha Grande Bay, and now our results bring further insight on this subject for the area. At the Canary Islands, signal frequency and duration did not vary with behavioral state (Papale et al., 2016) while in the present study these differences appeared. Also, we have not found differences in number of inflection points while Papale et al. (2016) observed differences in number of maxima and minima (which are analogous to inflection points in our methods). Although these comparisons were performed for a limited sample size, these findings suggest that each population of S. frontalis could show unique variation according to behavioral state and that the differences observed between populations could be related to the behavioral contexts in which each population was recorded. The results for Ilha Grande Bay indicate that this is an important factor that should be addressed in future studies of S. frontalis in the SWAO.

We registered the emission of the same whistle contour by a solitary individual indicating that signature whistles may play an important role for maintaining contact with distant groups in S. frontalis populations in the SWAO. The occurrence of signature whistles in different behavioral contexts has also been noted in North Atlantic populations of the species (Herzing, 1996; Lammers et al., 2003; Papale et al., 2016). Therefore, future studies with SWAO S. frontalis populations should focus on detailed investigation of whistle types and their repetition in large vocal groups to better understand the function of stereotyped whistles in wild dolphins. The repetition of this whistle contour from the same animal shows that in order to study this emission behavior all of the emitted whistles should be used in the analysis, since they may be part of an important communication sequence.

The present study adds new data to the acoustical knowledge about this species in the SWAO and demonstrates that the studied population produces whistles with a frequency range similar to North Atlantic populations. The whistle repertoire consistently contained fundamental frequencies below 30 kHz, which is 10 kHz higher than the limit posed by Azevedo et al. (2010) and within the same range of the Azores population (Gannier et al., 2020). Other factors that can influence delphinid acoustic behavior, such as underwater noise (May-Colado and Wartzok, 2008) have not been addressed in this work, but previous research with North Atlantic S. frontalis has shown that they affect this species' communication (Papale et al., 2015). More research in the area sampling acoustically with a broad frequency range should investigate how multiple factors affect the communication behavior of S. frontalis, as well as to address other phonation types, like burst-pulses [e.g., Herzing (1996) and Lammers et al. (2003)], in order to expand our understanding of this species in the southwestern Atlantic Ocean.

We are grateful to the MAQUA team for field assistance. The authors would like to thank Isabela Lima for insightful advice. The Tamoios Ecological Station has collaborated in the conduction of this study. The Rio de Janeiro State Government Research Agency (FAPERJ) and Brazilian Research Council (CNPq) have supported research developed by MAQUA in the Rio de Janeiro coastal area. L.B. had a post-doctoral scholarship from FAPERJ (PDR-10). T.L.B., A.F.A., and J.L.-B. had research grants from CNPq (PQ), FAPERJ (CNE and JCNE), and UERJ (Prociência). This study was financed in part by Transpetro “Programa de Conservação dos botos-cinza (Sotalia guianensis) e outros cetáceos das baías da Ilha Grande e de Sepetiba” (MAQUA/UERJ, Associação Cultural e de Pesquisa Noel Rosa, INEA, Transpetro) (Grant No. TAC-4600012708).