The Antillean manatee produces broadband vocalizations with ultrasonic frequencies

The two subspecies of the West Indian manatee (Trichechus manatus)—the Antillean (T. m. manatus) and Florida manatees (T. m. latirostris)—produce vocalizations previously shown to be important for underwater communication (Bengtson and Fitzgerald, 1985; O'Shea and Poché, 2006). Manatees produce a variety of short-duration (200–800 ms) signals that include tonal harmonic sounds referred to as squeaks, screeches, whines, and trills (Sousa-Lima et al., 2008; Umeed et al., 2017). These calls serve a range of social functions, such as maintaining contact between mothers and calves and as affiliative calls between conspecifics (Bengtson and Fitzgerald, 1985). Individual variation in calls has been shown to convey information about sex and identity (Reynolds, 1981; Sousa-Lima et al., 2008). Manatee groups have been documented to increase their vocal production rates during social interactions (Bengtson and Fitzgerald, 1985; Miksis-Olds and Tyack, 2009) and in response to playbacks of conspecific calls (Sousa-Lima et al., 2008; Umeed et al., 2017). Many of these previous studies have examined the acoustic behavior of Florida manatees; few have explored the vocal behavior of wild Antillean manatees.

Antillean manatees produce vocalizations that share similar acoustic structures and parameters across various populations. Their calls have been reported with mean fundamental frequencies of 0.64–5.90 kHz (Nowacek et al., 2003; Umeed et al., 2017) and frequency ranges of 2–22 kHz (Mann et al., 2006; Landrau-Giovannetti et al., 2014). Assessments of Florida manatee hearing sensitivity indicate their hearing may be best between 8 and 20 kHz (Gerstein et al., 1999) or 8 and 32 kHz (Gaspard et al., 2012). However, extended hearing sensitivity to tonal sounds at 76 kHz (Gerstein and Gerstein, 1997) and up to 90.5 kHz (Gaspard et al., 2012) suggest ultrasonic frequencies could be important to the detection of conspecific sounds in noisy habitats.

Manatees of both subspecies of the West Indian manatee are reported to produce acoustic signals with energies in the ultrasonic range above 20 kHz (Mann et al., 2006; Landrau-Giovannetti et al., 2014; Castro et al., 2015; Rivera-Chavarría et al., 2015; Merchan et al., 2019). The production of ultrasonic vocalizations with significant energy (103 dB re: 1 μPa at 46 kHz) has been reported between mothers and calves in Florida manatees (Gerstein and Gerstein, 1997; Gerstein et al., 2008). In the Antillean subspecies, vocalizations of wild manatees with ultrasonic energy content were noted in a few studies (Castro et al., 2015; Rivera-Chavarría et al., 2015) and illustrated with energy in harmonics extending up to ∼47 kHz (Merchan et al., 2019). However, the proportion of their calls with ultrasonic energy content has not been investigated, and the upper limits of this energy content and possible function for wild manatee communication are unknown. This information can provide further insight about the functional use of sounds in wild manatees, as well as identify factors for effective detection of their calls through passive acoustic monitoring (PAM) (Niezrecki et al., 2003; Merchan et al., 2019) and assess the effects of anthropogenic noise on their behavior and welfare (Miksis-Olds et al., 2007; Miksis-Olds and Tyack, 2009).

Here, we describe characteristics of the vocalizations of Antillean manatees recorded in Belize. We examined the spectral characteristics of wild manatee calls recorded during deployments of a passive acoustic recorder in a seagrass channel and from recordings of a recently reintroduced sub-adult female at the edge of the lagoon-side entrance of a wildlife rehabilitation center.

Vocalizations of wild manatees were recorded on the leeward side of St. George's Caye (SGC), a small crescent-shaped island located 9.5 km east of mainland Belize near the Belize Barrier Reef [Figs. 1(a) and 1(b)]. Recordings were made from 1015 h on July 8 to 1410 h on July 10, 2017. The caye is surrounded by expansive seagrass flats, sand patches, and deep channels and holes, and the area is regularly inhabited by manatees of all ages and sex classes (Ramos et al., 2017). The presence and vocalizations of at least eight different manatees, and likely more, were documented at this site during recordings that were part of several ongoing studies (e.g., Ramos et al., 2017; Ramos et al., 2018).

Acoustic recordings of wild manatees were made with a calibrated SoundTrap 300 HF (Ocean Instruments, New Zealand) that sampled sounds continuously at a 288 kHz sample rate in 16 bit resolution (flat frequency response: 0.02–150 kHz [±2 dB], clip level: 172 dB re: 1 μPa) with the preamplifier gain on. The SoundTrap was anchored by rope to the seafloor with a cinderblock and suspended in the water column at a depth of 1 m above the seafloor in water 1.5 m deep. The device was placed at the edge of a seagrass bed in a human-dredged channel, near a large hole often used by manatees as a resting hole [Figs. 1(b) and 1(d)]. In Belize, manatees are often observed using resting holes, which are deep depressions in the coastal marine substrate where manatees appear to rest (Bacchus et al., 2009).

The vocalizations of a 4.5 yr old female manatee (“Khaleesi”) were recorded for 90 min during the hours of 1110 to 1240 on March 12, 2018. Prior to these recordings, Khaleesi was rehabilitated at Wildtracks, an NGO that operates an animal rehabilitation center in Corozal, Northern Belize [Figs. 1(a) and 1(c)]. Khaleesi had entered the facility on July 29, 2013 as an orphaned newborn calf (estimate of 3–4 days old) and was housed alone for her first 2 years, and then integrated with several other young manatees at ∼3 yr old. Her vocalizations were recorded while she was alone at the lagoon-side edge of the center, several weeks following her successful soft release in February 2018. Soft release involves release into a relatively controlled, safe environment (e.g., the lagoon here) with natural food resources (e.g., seagrass), where young manatees can learn the skills they need in the wild. During this time, manatees receive constant monitoring and a continued feeding routine that is steadily reduced over a year to until they maintain their bodyweight with no supplemental feeding and are considered independent.

Recordings were made with the same model SoundTrap previously described, but at a higher sampling rate (576 kHz) to optimize recording the full frequency range of manatee calls as the wild recordings were restricted by memory limitations of the device. The device was suspended at 1 m in water 1.5 m deep near a limestone wall at the edge of the adjacent lagoon, at a horizontal distance of 1–3 m from Khaleesi during stationary rest and feeding while she was partially oriented with her head facing the recorder. Her presence was visibly confirmed by sighting the floating GPS tag attached to her with a peduncle belt. Aerial observations using a small quadcopter drone confirmed that she was the only manatee in the area during recording.

To examine the characteristics of manatee vocal production and signal parameters, we used Raven 1.5 (Bioacoustics Research Program, 2014) to visually and aurally review all sound recordings in spectrograms with different parameters to account for the different sampling rates of recorded sounds (wild manatees/rehabilitated manatee: sample rate: 288/576 kHz; DFT size: 2048/4096 samples; Hann window; overlap: 90%; time resolution: 10.7/23.1 ms). Individual manatee vocalizations were identified on the discrete time frequency grid and measured with the selection tool to quantify their time and frequency parameters up to 150 kHz. The signal-to-noise ratios (SNRs) of each manatee call was determined by selecting a 10 ms section and comparing the energy values of the section in dB to a 10 ms section of ambient noise without manatee signals in each recording. Calls that had a SNR ≥6 dB were used for analysis. Calls obtained from Khaleesi with a SNR ≥10 dB were selected for analysis of their ultrasonic energy content; this more selective procedure was due to the abundance of high-quality recordings of her sounds.

Five vocal parameters of each call were measured from the entire call: total duration (s); minimum frequency; maximum frequency; peak frequency; and fundamental frequency. A 1024-point section extracted from the middle of each signal was measured in the spectrogram to determine peak frequency, defined as the frequency with the most energy. The same section was used to determine the fundamental frequency by measuring the harmonic spacing between frequency bands of the call in the spectrogram (Nowacek et al., 2003).

The energy of individual ultrasonic bands was extracted from the calibrated power spectrum. Source levels of ultrasonic bands were estimated using a cylindrical spreading model for transmission loss and graphed in SPSS version 21 (SPSS Inc., Chicago, IL).

Descriptive statistics characterizing acoustic signal parameters were compared to previous studies of Antillean manatee vocalizations to determine similarities and differences of call parameters reported here compared to other regions (Nowacek et al., 2003; Landrau-Giovannetti et al., 2014; Rivera-Chavarría et al., 2015).

A total of 2573 wild manatee vocalizations were detected in 52 h of sound recordings. Of these calls, 335 vocalizations with a SNR ≥6 dB were selected for analysis of signal parameters [Figs. 2(a) and 2(b); See MM1.WAV for a WAV file with several example calls].

Vocalizations of wild manatees were of short average duration (mean ± standard deviation (SD): 245 ms ± 88) and primarily consisted of tonal harmonic calls (92.2% of total; n = 309). Ultrasonic frequency content was detected in 69.6% of vocalizations and signals contained maximum frequencies of up to 88.3 kHz (Table 1). The mean fundamental frequency of calls produced by wild manatees was lower than previous reports of manatees in Belize and throughout Central and South America, but higher than reports from Puerto Rico (Table 1). Call duration was consistent with most reports, but the peak and fundamental frequencies were more variable.

Signal parameters were analyzed from a sample of vocalizations (n = 100) produced by Khaleesi, all with a SNR ≥10 dB [Figs. 2(c) and 2(d), See MM2.WAV for a WAV file with several example calls]. Khaleesi's calls were generally higher in frequency than the wild manatees; all analyzed calls contained ultrasonic frequency bands and their maximum frequencies were higher than that of all other calls observed from other manatees in this study (Table 1). While minimum and peak frequencies of wild manatees were lower than vocalizations from Khaleesi, maximum frequencies of Khaleesi's signals were more variable and ranged from 42.7 to 150.0 kHz compared to wild manatees that ranged from 9.0 to 88.3 kHz (Table 1).

Spectral analysis of ultrasonic frequency content in Khaleesi's vocalizations revealed 18 calls that had significant energy in frequency bands >24 kHz (Fig. 3). Most of the energy observed was contained in frequencies ≤40 kHz. Estimated source levels for all ultrasonic bands had a mean of 61.4 dB re: 1 μPa at 1 m and ranged from 52.7 to 67.7 dB. The highest ultrasonic frequency with significant energy content was 54.3 kHz with an estimated source level of 57 dB (Fig. 3).

In this study, characteristics of vocalizations are described for Antillean manatees observed in Belize. For the most part, acoustic characteristics noted within this study were consistent with previous studies on manatee acoustic signals (e.g., Nowacek et al., 2003; Landrau-Giovannetti et al., 2014; Rivera-Chavarría et al., 2015; Umeed et al., 2017). However, using higher sampling rates, we discovered that the majority of manatee vocalizations included ultrasonic harmonics. This evidence suggests ultrasonic calling comprises a large proportion of the acoustic repertoire of Antillean manatees. Khaleesi, a young rehabilitated and released female manatee produced vocalizations with maximum frequencies more than twice as high as those observed in the wild. These calls exceed the reported frequency range of vocalizations of the West Indian manatee by 92 kHz (previous highest frequency: 46 kHz in T. m. latirostris; Gerstein et al., 2008).

It is unknown if her vocalizations developed as a result of: behavioral factors associated with rehabilitation in an artificial habitat (e.g., separation from mother, isolation from conspecifics during early life); age and/or sex-related factors (e.g., body size); or simply a variant of this subspecies of manatee in this region. Recordings of her relatively higher and stronger ultrasonic signals could also be attributed to site-specific propagation factors (no seagrass, a nearby limestone wall), and relatively shorter distances to the hydrophone than those of the wild manatees recorded. Both the detection and received levels of highly directional ultrasonic components would have been dependent upon the alignment of manatee callers and whether they were near or in-line with the acoustic center of a single hydrophone. Khaleesi may have been better aligned with the hydrophone and directional calls would have been more effectively recorded than those of the wild manatees detected near stationary deployments. All the ultrasonic levels still fell below the measured hearing level for manatees (Gerstein et al., 1999; Gaspard et al., 2012), which may be in part attributable to propagation in the recording environments, distances to the receiver, and orientation with the acoustic center of a single hydrophone. Using a static four hydrophone array with a sampling rate of 100 kHz, Gerstein et al. (2008) recorded directional ultrasonic calls between wild mothers and calves with significant energy (103 dB) at 46 kHz. Orientation was key, as the calls and received levels were only documented when mother–calf manatee pairs were positioned on axis within the array.

Mother–calf pairs were present each day during this study. Concurrent and ongoing research has documented high seasonal site-fidelity of numerous individuals, including several mother-calf pairs, at SGC (Ramos et al., 2017; Ramos et al., 2018; Landeo et al., 2019). Although it could not be reliably determined which individuals were vocalizing in the resting holes, manatee mothers and their calves may have produced ultrasonic calls as these high-frequency spectral components may be important for directional hearing (Gerstein et al., 2008).

Vocalizations with ultrasonic frequencies may have evolved as they improved the ability of manatee's to communicate and locate each other effectively in shallow habitats. High-frequency manatee vocalizations may be easier for manatees to localize with greater differences in interaural and intensity cues than those afforded by lower frequency calls (Gerstein, 1999; Colbert et al., 2009). Similarly, high frequency calling may be advantageous in shallow water where lower frequencies have reduced propagation ranges due to boundary limits and are also subject to cancelation near the water's surface via the Lloyd's mirror effect (Gerstein et al., 1999).

Previous studies of the acoustic behavior of manatees used significantly lower sampling rates (44 kHz for Landrau-Giovannetti et al., 2014; 48 kHz for Umeed et al., 2017; and 100 kHz for Gerstein et al., 2008) than the 288 and 576 kHz used for this study. Detection of a high proportion of ultrasonic manatee calls supports the need for broadband acoustic recording systems (which sample at frequencies up to and above 150 kHz) deployed over days to months in order to reliably document and characterize the vocal repertoire of each subspecies. This information will provide an improved understanding of the frequency range of manatee calls. Studies using PAM to detect manatees have demonstrated the capacity for the detection of individual manatee vocalizations (Merchan et al., 2019) and algorithms designed to more effectively automate their detection (Niezrecki et al., 2003). Future studies using acoustic playback of ultrasonic calls to manatees, detailed investigations of ambient sound conditions, and experiments of sound transmissibility of manatee calls are needed to further elucidate the function of high-frequency harmonic bands and to better identify the factors which promote their evolution. Systems with the capacity for recording broadband sounds enable the detection of features such as high-frequency harmonics of manatee calls, which may play a significant role in manatee communication.

Thank you to Linda and John Searle of ECOMAR for facilitating research with manatees at SGC. Thank you to Kimberly E. McCabe and Amelia Bradley for their help with acoustic analysis. Thank you to Wildtracks for their collaboration and letting us record manatees at their facilities. Thanks to Oceanic Society, SEE Turtles, and their volunteers for supporting ongoing research activities in Belize. Big thanks to Megan McGrath for editing this manuscript and dramatically improving its quality. This research was permitted by the Belize Fisheries Department, Belize Forest Department, and the Civil Aviation Authority of Belize.