Studies on chelonian acoustic behavior are still scarce. For over a century turtles were considered mute and deaf by scientists (Pope 1955), propagating an erroneous perception in the scientific community and constraining investments in research related to turtle acoustic communication.

There are records of chelonian hatchlings emitting sounds in different contexts: in the egg, in the nest, and in the river where they communicate with females after they reach the river (Ferrara 2012; Ferrara et al. 2013, 2014a, 2014b). Here we discuss the hypothesis that hatchling vocalizations are used to stimulate group hatching and mediate hatchlings' synchronization in digging out of the nest. This coordination is thought to enable aquatic turtles' joint dispersion along the beaches to the river and group migration to the feeding grounds (Ferrara et al. 2014c). Currently, more than 55 species of turtles have been documented emitting sounds in different behavioral contexts (Campbell and Evans 1967; Campbell 1972; Auffenberg 1977; Sacchi et al. 2003; Cook and Forrest 2005; Galeotti et al. 2005a, 2005b; Giles et al. 2009; Ferrara 2012; Ferrara et al., 2013, 2014a, 2014b, 2014c, 2017, 2018). Knowledge about the vocal repertoire of species of different taxonomic groups is essential to understand the evolution of the complexity of animal communication (Ferrara et al. 2009). According to Ferrara et al. (2013) and Iverson (1990), acoustic communication in turtles has a role related to parental care in freshwater species, suggesting that vocalizations may be much more important than previously thought (Colafrancesco and Gridi-Papp 2016). However, the taxonomic coverage of this behavioral trait is still unknown (Ferrara et al. 2017) and here we attempt to verify if this behavior is present in hawksbill turtles to elicit further inquiry about the evolution of chelonian acoustic communication

Methods. — Data collection was carried with collaboration of the TAMAR Project (Parnamirim, Rio Grande do Norte, Brazil) in an area protected by the Brazi lian Space Agency (lat 5°54′56 ″S, long 35°15′46″W). In Rio Grande do Norte state, the nesting season of Eretmochelys imbricata runs from November to May (Santos et al. 2013), with the peak of nesting activity in February (Marcovaldi et al. 2007). The duration of incubation time in this area varies from 49 to 70 d (Marcovaldi et al. 2014).

Recording effort started on the 50th incubation day of E. imbricata nests (dates varied according to the date each female laid their eggs), each recording lasting for 10 min once a day. We have analyzed 103 recordings from 26 nests with presence of eggs or eggs and hatchlings. Hatchlings were also recorded after nest emergence. Data collection was performed at night and day, according to the nests monitored by the TAMAR Project. A solid-state Tascam DR-40 PCM system (recorder with stereo microphones) with 96 kHz/24 bit acquisition rate was used in this experiment. From the 50th day on, we opened the nest every day (weather permitting) until we found eggs or hatchlings to start acquiring the sounds by placing the recording system inside the opened nest. We used a plastic container to cover the recorder, which was then covered with sand that was taken from the nest while digging to find eggs or hatchlings. Covering the recording system with plastic and covering it back again with sand (Fig. 1) reduced the interference of external noise in the recordings and ensured high-quality sound data (high signal-to-noise ratio).

View Figure

• Download as Powerpoint
• Download Figure

A second experimental methodology was used to record hatchlings. Hatchlings outside the nests were translocated from another TAMAR base situated at Pipa Beach, Rio Grande do Norte, and brought to the Parnamirim base. The hatchlings from each nest were kept together. Eight hatchlings from each of the 2 translocated nests were removed after 18 hrs of their emergence and placed inside polystyrene rectangular boxes (dimensions 30 × 20 × 25 cm) lined with cloth to avoid noise resulting from the movement of the hatchlings on the polystyrene. The recording system was then positioned in the box with the animals and covered with the lid during a 10-min recording session to ensure acoustic isolation from background noise (Fig. 2). This procedure was repeated twice for each of the 2 translocated nests, using 8 different individuals each time, and sampling a total of 32 hatchlings (16 from each nest) resulting in 4 10-min recordings made in the box.

View Figure

• Download as Powerpoint
• Download Figure

All recordings were analyzed with Raven Pro 1.5 software. The spectrographic parameters used were window type: hamming; window size (fast Fourier transformed [FFT]) varied from 256 to 512 Hz. The variables extracted from the manually selected sound signals were low frequency, high frequency, peak frequency, and signal duration (which was measured in the oscillogram). Selected sounds were classified into sound types for further analyses. To avoid the inclusion of background noise in the analyses, we calculated the signalto-noise ratio through the inband power (units) obtained in Raven. Only signals with a minimum of 10 dB above the noise levels were used. To generate the spectrographic images, R software and the SeeWave (Sueur et al. 2008) data packet were used.

Results. — A total of 107 10-min audio recordings were obtained. To make sure the sounds were sound emissions from the turtles, not background noise, we recorded a false nest, i.e., only the hole at the same mean depth of all nests (50 cm), but with no eggs or hatchlings inside, to compare with the actual recordings, and no sound signals similar to those in real nests were detected.

After a detailed manual analysis of the selected sounds (n = 575), 4 different acoustics types (Fig. 3) were determined, which were characterized according to their spectral characteristics. The window size (FFT) used in the spectrographic analyzes was 256 for Types I and II and 512 for Types III and IV and the description of each sound type is presented below (descriptive statistics are shown in Table 1).

View Figure

• Download as Powerpoint
• Download Figure

Table 1. Descriptive statistics of each type of acoustic signal recorded in embryos and hatchlings of Eretmochelys imbricata. Values are mean ± SD.

View Fullscreen

Type I (n = 86): This sound type is characterized by pulses with mean duration of 0.0015 ± 0.0005 sec SD. This type can appear as a single pulse or be composed of a series of several pulses with a maximum of 64 pulses in a series. The peak frequency ranges from 750 to 5250 Hz with mean value of 1620 Hz.

Type II (n = 330): This sound is the most common found in our sample. It is composed of 1–25 pulses, with mean duration of 0.0010 ± 0.0004 sec SD for each pulse. The peak frequency ranges from 370 to 5250 Hz with mean value of 3050 Hz.

Type III (n = 110): This sound is the longest and lowest-frequency sound type found in our sample. Sounds show a single band and low-frequency modulated tone. This sound showed variable signal duration, between 0.0028 and 0.096 sec. Peak frequency ranges from 560 to 11,600 Hz with mean value of 2600 Hz.

Type IV (n = 49): This type is characterized by tonal signals, with the presence of 2 to 5 harmonics and duration between 0.0025 and 0.031 sec. Frequency modulation can be upsweep or down-sweep, the second case being the most common. Some signals of this type have the second or third harmonics with a higher amount of energy than the first, the mean peak frequency is 1350 ± 430 Hz SD.

Discussion. — The results show that embryos and hatchlings of E. imbricata produce sound inside the nests and after emergence. Studies showed that embryos near hatching can communicate with sounds as reported for other reptiles such as Podocnemis expansa, Dermochelys coriacea (Ferrara et al. 2013, 2014b), and crocodiles (Vergne and Mathevon. 2008). Here, we described a vocal repertoire of similar complexity to other hatchlings of sea turtles (Ferrara 2014a, 2014b). Our recordings were made from the 50th day of incubation until emergence from the nest. However, the vocalizations only began to be identified from the 51th day of incubation, which coincides with the beginning of the presence of the hatchlings, which suggests that bioacoustic activity in E. imbricata comes mostly from hatchlings and not from embryos.

Giles et al. (2009) classified the vocalizations of Chelodina oblonga according to the auditory characteristics, frequency, and duration of the signals. Considering the terms used by this author and the spectral characteristics, we can classify Type I sounds as “drum rolls” and Type II sounds as “double-clicks,” with variation in signal frequencies for each species. In our sample, the majority of the signals were pulses (n = 416), which can be found alone, in pairs, trios, or in series, with different energy distribution among frequencies. Ferrara et al. (2014b) reported the presence of pulses in D. coriacea nests. Pulsed sounds can be produced within eggs by friction of animal body parts (stridulation), vibration of a part of the body in contact with some resonant part within the egg (percussion), or by gas bubbles (Wilson et al. 2004).

Sounds with high frequency that begin and break abruptly can be used to facilitate the location or identification of the sender by another individual of the same species (Morton 1977). These characteristics can be observed in the Types I and II sounds. High frequencies can be attenuated more efficiently, making it difficult to be found by a predator and to locate the source of sound emission (Forrest et al. 1993). Thus, the emission of highfrequency sounds by the turtles may function as a strategy to avoid detection by potential predators (Ferrara et al. 2014b). The study area has a high predation rate by fox (Cerdocyon thous) and armadillo (Euphractus sexcinctus). A study by Neto et al. (2010) used photographic traps in nests monitored in the Barreira do Inferno region and found that nests are frequently visited and/or predated, mainly by foxes.

The tonal sounds (n = 159) of Types III and IV were found mainly when hatchlings were already present outside the eggs or emerging from the eggs. These sounds may show harmonics (up to 5) or a single band, usually with frequency modulation. Recordings of C. insculpta adults have a vocal repertoire of sounds with different structural characteristics, including harmonic and nonharmonic structures (Ferrara et al. 2017). The Type III sound reported by Ferrara et al. (2013) resembles our Type III sounds recorded in the nests with hatchlings. However, our Type III sounds do not present harmonics.

In the vocalizations of the hatchlings and embryos of E. imbricata, we observed a repertoire that involves simple sounds and more complex sounds, with harmonics and frequency modulation. Chelonians present complex mechanisms of sound production (Galeotti et al. 2005a) similar to those used by birds and mammals (Catchpole and Slater 2008; Manser 2001). The functional morphology of chelonian sound production needs to be addressed in future research.

Recently, hypotheses about the function of sound production in chelonian nests have been discussed (Ferrara et al. 2013). Emissions of Types I and II may be related to the synchronization of the hatchlings during the nest egg eclosion. Sounds can be produced to synchronize nest eclosion and promote group digging to emerge from the nest (this may be related to diluting the predation pressure on the nests during dispersal into the water). In some freshwater species, the hatchlings produce calls to contact the females to lead them to the flooded forest feeding habitat (Ferrara et al. 2013). Our results corroborate to one of the hypotheses discussed by Ferrara et al. (2013), in which the hatchlings tend to leave the nest together. This group behavior decreases the pressure of predation. In addition, this behavior may be interesting because some chemical signals are released during incubation, attracting predators to the remaining eggs (Lack 1968; Vitt 1991).

Sound emission frequencies are inversely proportional to animal size (Martin et al. 2011). Considering that the D. coriacea body size is larger that of E. imbricata, it is expected that frequency emissions of E. imbricata should be higher than those of D. coriacea. Indeed, Ferrara et al. (2014b) found the mean peak frequency of the sound emissions of D. coriacea to be 993.7 Hz while the peak for E. imbricata is 2670 Hz.

Other studies show that high levels of anthropogenic noise may have an effect on behavior of sea turtles (Samuel et al. 2005). Anthropic noise pollution has previously not been considered important for chelonians because they were not considered to be animals that used acoustic communication. Our results highlight the need to consider noise pollution in conservation efforts to protect sea turtle nesting environments because such noise could mask sounds produced by hatchlings and reduce coordination and possibly lead to increased risk of predation.