Whistle comparison of four delphinid species in Southeastern Brazil

Whistle repertoires of delphinds have interspecific variations that may be influenced by many factors: differences in body sizes among species (May-Collado et al., 2007); degree of social fluidity (Bazúa-Durán, 2004); phylogenetic relationships and zoogeographic relationships (May-Collado et al., 2007).

Studies have investigated whistle interspecific variations as a way to develop methods of species identification in the Pacific Ocean (Oswald et al., 2004; Oswald et al., 2007). However, in the Atlantic Ocean, information on interspecific whistle differences and similarities is scarce. Comprehension of whistle variations may contribute to future acoustic monitoring of the species characterized and to our knowledge of the factors promoting differences. In this study, differences and similarities among whistles of four delphinid species in the Rio de Janeiro State Coast, Southeastern Brazil were investigated: Atlantic-spotted dolphin (Stenella frontalis), rough-toothed dolphin (Steno bredanensis), Guiana dolphin (Sotalia guianensis) and common bottlenose dolphin (Tursiops truncatus).

Acoustic recordings were made in Rio de Janeiro State Coast during 44 days between April 2007 and March 2014. Recordings of Guiana dolphins (S. guianensis) were made in Guanabara Bay (22°50′S, 43°10′W), Ilha Grande Bay (23°09′S, 44°28′W), and Sepetiba Bay (23°00′S, 43°52′W); recordings of Atlantic-spotted dolphins (S. frontalis) were made in Ilha Grande Bay; recordings of rough-toothed dolphins (S. bredanensis) were made in Guanabara Bay and adjacent waters; and recordings of common bottlenose dolphins (T. truncatus) were made in adjacent waters of Guanabara Bay. All recordings were made aboard outboard-powered boats (5.5 m to 7.5 m in length) with engines turned off and under similar weather conditions (Beaufort sea states ≤ 2). Prior to each acoustic recording session, the area was scanned to confirm that there was no other cetacean species around. Two systems with 96 kHz sample rate were used for the recordings. The first consisted of a C54XRS hydrophone (−165 dB re: 1 V/μPa, flat frequency response from 0.009 to 44 kHz; Cetacean Research Technology, Seattle, WA), connected to a Fostex FR-2 Recorder (Foster Electric USA, Inc., Norwalk, CA). The second consisted of an HTI-96-MIN hydrophone (−170.5 dB re: 1 V/μPa, flat frequency response from 5 Hz to 30 kHz; HighTech, Inc., Long Beach, MS) and a PMD 671 Marantz recorder (Marantz America LLC., Mahwah, NJ).

Whistles with complete clear spectral contour with low interference from background noise were manually selected and analyzed in Raven 1.4 (Cornell Lab of Ornithology), with a 512 sample Hanning window, a fast Fourier transform size of 512 points, and 50% overlap. To avoid oversampling groups/individuals, the number of whistles analyzed for each group was up to twice the number of individuals in the group for every species to minimize oversampling. The following acoustic parameters (Azevedo and Van Sluys, 2005; Andrade et al., 2015a) were measured: starting frequency (SF), ending frequency (EF), minimum frequency (MinF), maximum frequency (MaxF), delta frequency [MaxF - MinF] (DF), frequency at 1/4 of duration (F1/4), frequency at 1/2 of duration (F1/2), frequency at 3/4 of duration (F3/4), mean frequency (MF), duration (DUR), and number of inflection points (INF). All of the frequency variables were measured in kHz and the durations were measured in milliseconds.

For analysis, the whistles of the different populations of S. guianensis were grouped together. This was done because a previous comparison among the repertoire of these populations and among other populations of this species recorded in adjacent sites in the Brazilian coast showed a high value of Wilks' Lambda for discriminant function analysis (DFA; Andrade et al., 2015a) and a low correct classification score (Azevedo and Van Sluys, 2005). Statistica 7.1 (Stasoft, Inc., Tulsa, OK) was used for statistical analyses. Descriptive statistics for all acoustic parameters measured for each species included mean, standard deviation, median, minimum, and maximum values. Kruskal-Wallis test (P < 0.05) was applied to verify if there were differences among the whistles of the species recorded. The data were rank-ordered for the application of the post hoc Tukey test to identify which parameters were different in each species comparison (Zar, 1999). Bonferroni procedure was applied to adjust the levels of significance to the Tukey test (P < 0.008) in order to minimize the data dependence problem. Step-wise DFA was applied to classify whistles to species (Zar, 1999).

A total of 487 whistles were analyzed from 33 h and 26 min of recording. There were differences among species across all parameters (Kruskal-Wallis test, H(3, N = 487); P < 0.05; Table 1). Whistles of Guiana dolphins had the highest values for most frequency parameters (Tukey post hoc test, P < 0.001). The whistles of this species only had lower values for starting, minimum frequency, and frequency at 1/4 of duration than whistles of common bottlenose dolphins (Tukey post hoc test, P < 0.001). Rough-toothed dolphins emitted whistles with lower frequency values than the ones of other species (Tukey post hoc test, P < 0.001). Whistles of common bottlenose dolphins were longer and had larger numbers of inflection points than whistles of other species (Tukey post hoc test, P < 0.001). Seven frequency parameters of common bottlenose dolphin whistles (SF, MinF, MaxF, F1/4, F1/2, F3/4, and MF) had higher values than the whistles of Atlantic-spotted dolphins (Tukey post hoc test, P < 0.001; Table 1).

DFA (Wilks' Lambda = 0.20429; F(27.14) = 37.141 P < 1.0 × 10⁻⁵) indicated largest differences between whistles of rough-toothed dolphins and Guiana dolphins (D²= 9.81; F = 69.77), followed by differences between whistles of Guiana and common bottlenose dolphins (D²= 9.54; F = 69.06), and between rough-toothed and common bottlenose dolphins (D²= 9.27; F = 51.45). Whistles of Atlantic-spotted dolphins shared more similarities with whistles of other species, especially with rough-toothed dolphins (D²= 1.53; F = 10.22). Overall, 72.5% of the whistles were classified to the correct species. Whistles of rough-toothed dolphins presented the largest correct classification score (88.7%), followed by common bottlenose (74.7%) and Guiana dolphin whistles (70.7%; Table 2). Whistles of Atlantic-spotted dolphins presented the lowest correct classification percentage (58.8%). Nine acoustic variables were included by the DFA in the model (DUR, SF, MaxF, DF, F1/4, F1/2, F3/4, MF, and INF).

The results indicated differences among whistles of four delphinid species normally found in the Rio de Janeiro State Coast. The high frequency range of the whistles of Guiana dolphins seems to play an important role in the communication of Guiana dolphins (Andrade et al., 2015b) and was therefore important in interspecific comparisons. Discrimination of whistles of rough-toothed dolphins was facilitated by their overall lower frequency values and modulation, which have also been reported for this species in the Pacific Ocean (Rankin et al., 2015). Whistles of common bottlenose dolphins showed high signal specificity and are normally among the best classified in other comparison studies, with correct classification scores varying from 60% to 86.7% (Oswald et al., 2007; May-Collado and Wartzok, 2009). Whistles of Atlantic-spotted dolphins had high overlap in time and frequency variables with the ones of other species, and may be better discriminated by other signal characteristics that were not considered in this study.

The use of recording systems with 96 kHz sample rate was important for whistle comparison (Oswald et al., 2004), especially for adequate measurement of whistles of Guiana dolphins (May-Collado and Wartzok, 2009; Andrade et al., 2015b). Comparisons among species with whistles that have higher frequency components than the frequency limit of the recording system may lead to overlooked differences (Oswald et al., 2004).

The largest differences obtained in this study were found between whistles of species that were closest phylogenetically (Cunha et al., 2012), which highlights the fact that interspecific differences may not be exclusively attributed to phylogenetic relationships. Differences may be influenced by divergences in ecological characteristics such as adaptations to different environments. Whistles of Guiana dolphins may be adapted to conditions in specific coastal environments (Andrade et al., 2015a) due to high degree of residence in costal environments throughout their distribution (Azevedo et al., 2005), while whistles of the other species may be adapted to a variety of environments, including coastal and oceanic regions. The relatively small sample size and the fact that only one of the species was recorded in different sites may have reduced the explanatory power of the analysis presented in this study. Nevertheless, this is the first published study that provides evidence of interspecific differences among species recorded in Southeastern Brazil. Studies that characterize the repertoire of each species in different environments and social contexts are necessary in order to investigate which factors influence variation among delphinid species' whistle repertoires.

We particularly thank the Faculdade de Oceanografia (UERJ), the Iate Clube Jardim Guanabara, ESEC-Tamoios, and APA de Guapimirim for their support. We also thank engineer Orlando Afonso, from the Brazilian Navy Research Institute (IPqM), for technical support. We thank the anonymous reviewer who made useful suggestions for improving the manuscript. We are grateful to MAQUA team for field assistance and insightful advice. This work was supported by Rio de Janeiro State Government Research Agency (FAPERJ), Brazilian Research Council (CNPq) and Cetacean Society International have supported research developed by MAQUA. A.F.A and J.L.B have research grant from CNPq (PQ-1D), FAPERJ (JCNE), and UERJ (Prociência). I.M.S.L. received a scholarship from the Coordination for Improvement of Higher Education Personnel (CAPES). This study was developed at the Graduate Program of Oceanography (UERJ).