Case study: Evidence of long-term stability in a stereotyped whistle in a single free-ranging humpback dolphin (Sousa plumbea) found in sympatry (Tursiops aduncus)

The acoustic repertoire of Sousa spp. comprises acoustic signals common in delphinids, typically grouped into echolocation clicks, burst pulses, and whistles (Herzing, 2014; Van Parijs , 2000; Yang , 2020). Whistles are frequency-modulated tonal calls strongly associated with social contexts (Tyack, 1997), and humpback dolphins seem to present anatomical adaptations for high frequency communication sounds (Frainer , 2019). Whistle frequency parameters vary across the genus, with Atlantic humpback dolphins (S. teuszii) producing whistles with the lowest fundamental frequency across the genus, with a mean maximum of 8.2 kHz (Weir, 2010). Australian humpback dolphins (S. sahulensis) produce whistles with a higher mean maximum of 10.58 kHz and a maximum recorded fundamental frequency of 22 kHz (although this maximum frequency may have been limited by recording equipment) (Van Parijs and Corkeron, 2001b). The whistle repertoire of Indian Ocean humpback dolphins (Sousa plumbea) has only been reported from India, where mean minimum and maximum frequencies ranged from 7.6–10.2 kHz, respectively, and an absolute maximum frequency of 33 kHz was reported (Bopardikar , 2018).

Indian Ocean humpback dolphins (Sousa plumbea) are listed as endangered on the International Union for the Conservation of Nature Red List (Braulik , 2017), and effective methods to monitor populations at an individual level are needed. Passive acoustic monitoring (PAM) is a well-established research tool for the study of wild cetacean populations through the identification of species-specific calls in long-term acoustic recordings. More recently, the use of PAM for studying delphinids has been extended to individual capture–recapture using “signature whistles” (Erbs , 2017; Fearey , 2022; Longden , 2020; Romeu , 2024). These are individually distinctive calls used to communicate identity information and maintain group cohesion and are well demonstrated in bottlenose dolphins (Tursiops truncatus) (Janik and Sayigh, 2013). Since the 1960s, evidence has supported likely signature whistle use across other odontocete species (Tursiops aduncus) (Gridley , 2014), Stenella longirostris (Rio, 2023), Delphinus delphis (Fearey , 2019), Sousa sahulensis (formally S. chinensis) (Van Parijs and Corkeron, 2001a), Sousa chinensis (Cheng , 2017), Pseudorca crassidens (Rio, 2023), Legenorhynchus obliquidens (Caldwell , 1971), Sotalia guianensis (Lima and Le Pendu, 2014), and Monodon monoceros (Shapiro, 2006). Evidence for signature whistle use has been presented in Sousa spp. under instances of stress, including an injured Indo-Pacific humpback dolphin (S. chinensis) (Cheng , 2017) and from a live stranded Australian humpback dolphin (S. sahulensis, formally described as S. chinensis) (Van Parijs and Corkeron, 2001a). Evidence for signature whistle use has not yet been demonstrated in free-swimming S. plumbea.

To reliably communicate identity information, the acoustic signal should have high inter-individual (i.e., distinctive) and low intra-individual (i.e., stereotyped) variability (Charrier , 2003; Janik , 2006; Sayigh , 1990). This can be achieved through inter-individual differences in the vocal apparatus, resulting in by-product distinctiveness (so-called voice features) (Taylor and Reby, 2010). Provided that these differences remain stable over time and the receiver can perceive these differences, identity information can accurately be encoded and decoded by the signaler and receiver, respectively. Alternatively, and more cognitively complex, is the use of “designed individual signatures,” which can develop through a process of production learning. For bottlenose dolphins, there is strong evidence that signature whistles are developed through a process of vocal production learning (Janik, 2009), where the acoustic environment experienced by the dolphin can influence the overall frequency modulation pattern of the “crystalised” signature whistle (Fripp, 2005; Kershenbaum , 2013). Signature whistles are established in calves as early as the first 2 months of life, with high production rates in mother-calf pairs who use signature whistles to facilitate reunions after short separation periods (Janik, 2009; King , 2016; Smolker , 1993). As cohesion calls, signature whistles are produced at elevated rates during separation or isolation events and high arousal states, including times of stress (Caldwell , 1990; Esch , 2009; Probert , 2021; Probert , 2023). In these contexts, emissions may become less stereotyped and more variable in nature, occasionally with the production of non-linear phenomena (Sportelli , 2022).

Like other species within the genus, the habitat preference of S. plumbea includes estuaries and murky rivers, and they are rarely held in captivity; therefore, underwater observations of humpback dolphins are rare, and the context of sound production is often unknown. It is generally accepted that the acoustic signals produced by dolphins originate from the monkey-lips dorsal bursa complex (Cranford , 1996), and that bi-phonation is possible, whereby dolphins can produce whistles and pulsed calls (clicks and burst pulses) simultaneously (Kriesell , 2014; Madsen , 2013). Unlike other wild dolphin populations, information on individual acoustic behaviour in S. plumbea, including signature whistle production, and bi-phonation, is restricted (Bonato , 2015; Herzing, 1996). However, Ponta do Ouro, Mozambique, is one of the few places where long-term in-water interactions with this species have been recorded. Using historical data, we investigate the whistling behaviour of S. plumbea using a case study of a resident individual in Ponta do Ouro, Mozambique, named Herme, using in-water video and acoustic recordings.

Ponta do Ouro Partial Marine Reserve (26.84°S, 32.88°E) is Mozambique's southernmost conservation area. Boat-based research and dolphin tourism involving swim-with-dolphin experiences have been taking place in Ponta do Ouro since 1994, mainly targeting inshore Indian Ocean bottlenose dolphins (Tursiops aduncus). Over 250 bottlenose dolphins have been catalogued within this population, of which 60 are considered “resident,” and live sympatrically with S. plumbea (Gullen, 2020a). Comparatively, only a small number of S. plumbea (n = 6) have been identified and catalogued during this time (Gullen, 2020b). One humpback dolphin, locally named Herme, was born in 2006 and is the most commonly sighted S. plumbea in the study area and is exclusively encountered with T. aduncus groups.

Acoustic and video data were opportunistically collected between 2009 and 2019 during commercial swim-with-dolphin activities in the Ponta do Ouro Bay. During 2020, additional acoustic data were collected with a focus on S. plumbea. Acoustic data were collected by permitted tour operators and/or under research permits from an 8 m dive boat with two 90 hp four stroke engines. Encounters were limited to 30 min, and a maximum of two in-water encounters were permitted per launch (up to three launches/day).

Once dolphins were sighted and before they were approached, the group was observed for 5 min, and swimmers did not enter the water if the dolphins showed restless behaviour, if they had newborns in the group, or if they were travelling. If conditions were suitable, dolphins were close (<20 m), and it was considered safe to do so, swimmers entered the water to the side or ahead of the group, according to strict regulations to minimise disturbance. During underwater encounters, audio and visual data were collected in close proximity to both T. aduncus and S. plumbea using the equipment summarised in Table 1.

Videos of encounters with Herme present were confirmed by author A.G. through visual confirmation of Herme's identifiable features. The videos were scanned for periods of whistles produced by Herme, whereby the individual was the only humpback dolphin present within the mixed species group, and the dolphin was both audibly and visually (through the production of concurrent blowhole movements and/or bubblestream emissions) calling. The acoustic tracks of the video files from 2009–2019 were extracted as mono.wav files (see Table 1) using the behavioural software BORIS (Behavioural Observation Research Interactive Software, BORIS, University of Torino, Torino, Italy) (Friard and Gamba, 2016). Acoustic data collected in 2020 from the SoundTrap was matched to the concurrent GoPro video using time stamps, and then decimated to 96 kHz to facilitate ease of whistle analysis.

Whistles were manually annotated and visually classified using Raven Pro v 1.6 (Center for Conservation Bioacoustics, Cornell Lab of Ornithology, Ithaca, NY). All whistles of a particular frequency modulation pattern or contour (see Janik , 2013) were manually grouped into whistle types (SWT) by two observers (T.G., S.D.). The inter-whistle interval (IWI) of stereotyped contours was measured, and a bout analysis was applied by identifying periods where at least three out of four whistles are produced in a bout with an IWI of 1–10 s [for the SIGnature IDentification (SIGID) method, see Janik , 2013]. This method was applied to investigate whether the temporal production of this SWT was similar to that observed in common bottlenose dolphins (T. truncatus). Due to data limitations (i.e., the duration of continuous recording), SIGID was only applied to calls recorded during research effort in 2020, comprising 15 min and 55 s of continuous recording with dolphins (Table 1). Standard time–frequency parameters of the whistles were measured [start frequency (startF), end frequency (endF), minimum frequency (minF), maximum frequency (maxF), and duration], together with simultaneous burst pulse calls using the selection function in Raven Pro. The number of inflection points were manually counted. For the burst pulse signals, the peak frequency (peakF) was calculated within a constrained bandwidth between 2–5 kHz, as this was where most of the energy was focused and prevented erroneous measurement due to the simultaneous high amplitude whistles, which occupied the frequency bandwidth above 5 kHz.

Across the 31 encounters with Herme over the 11 years, the individual produced a stereotyped whistle in 13 encounters, with over 110 emissions of this whistle documented. This whistle was only recorded in encounters with Herme present.

This stereotyped whistle was commonly produced in bouts and met the criteria of SIGID on seven out of eight bouts within the 2020 encounter (Janik , 2013). The distribution of inter-whistle intervals and log survivorship plot demonstrated a clear pattern in the temporal production of whistles. Across the 21 stereotyped whistles, with a peak at 1–3 s, 65% of whistles in a bout occurred within 0.25–10 s of another whistle of the same type (Fig. 1).

This stereotyped whistle produced by Herme was higher in minF (5.94 ± 1.00 kHz), lower in mean maxF (14.64 ± 1.28 kHz), and longer in duration (1.7 ± 0.7 s), compared to other whistles recorded from the same encounter, which were most likely emitted by T. aduncus. The contour shape of Herme's whistle type showed a relatively stable frequency modulation pattern over the time span [Fig. 2(A)] and was highly distinctive from the other whistle types, including from T. aduncus in the area [Fig. 2(B)], which in the most part produced SWT, which conform to an upsweep multi-loop pattern (see Probert , 2023). Hermes's whistle was more complex, having multiple inflection points (mean 9.2 ± 4.9, compared to T. aduncus mean 3.6 ± 3.3), and an introductory loop that differed from successive loops, with distinct periods of bi-phonation and other non-linear phenomena, such as frequency jumps [see Fig. 2(A)].

Although this whistle was stereotyped, there were variations in time–frequency components over the 11-year time frame [coefficient of variation (CoV): minF = 17, maxF = 7, dur = 42, Frange = 16, inflection points = 53). The introductory loop was most consistently produced, with greater variability in the occurrence, duration, and number of inflection points of the subsequent tail of the whistle contour [Fig. 2(A)]. The whistle was mostly emitted as a continuous contour, with a maximum duration of 3.6 s. However, a disconnected version of the whistle, including periods of silence, was produced on three occasions.

Non-linear phenomena, such as frequency jumps and biphonation, were also present in Herme's whistle type (63/110 whistles, 57%). Frequency jumps were present in 31 whistle emissions. One to seven discrete burst pulse sounds were recorded during 43% (47/110) of Herme's whistle emissions, providing evidence of biphonation. These burst pulses were highly stereotyped in terms of duration (0.2 ± 0.07 s, CoV = 33) and peakF (3.43 ± 0.43 kHz, CoV = 13). On only one occasion was a burst pulse (with the same characteristics) produced immediately preceding the whistle type. In all other emissions, the burst pulses were produced at intervals separated by 0.1–0.2 s throughout the whistle duration.

This case study provides insight into the acoustic behaviour of a single Sousa plumbea individual encountered exclusively in a free-swimming mixed species group with Tursiops aduncus over an 11-year period. The production of a single stereotyped call type in bouts by Herme shows evidence of both individual distinctiveness and temporal stability. In addition, the whistle type presents evidence of non-linear phenomena and variability in whistling behaviour possibly linked to arousal.

Signature whistle types (SWTs) are well-documented in bottlenose dolphins (Caldwell , 1990; Gridley , 2014; Janik and Sayigh, 2013) and represent the best example of an individual acoustic label formed through vocal learning (Sayigh , 2007). Evidence for signature whistle use has been documented in other Sousa species through opportunistic recordings of isolated or live stranded individuals (Cheng , 2017; Van Parijs and Corkeron, 2001a) Sousa are therefore often included as a genus in which signature whistles are used, based on this limited body of evidence (Fearey , 2019; Janik, 2009). The formation of signature whistles through vocal learning allows the individual to modify the acoustic structure of the sound based on a model sound, crystalising these SWTs over time, but can also be attributed to high levels of interspecific similarity between whistles in mixed species groups (Favaro , 2016). In this case study, the degree of dissimilarity of Herme's whistle to sympatric Tursiops aduncus indicates a lack of whistle convergence occurring between the two species. As well as this, the bout production of Hermes stereotyped whistle fit SIGID criteria, indicating these whistles follow expected signature whistle temporal production patterns. Although this whistle type produced by Herme cannot be classed as a signature whistle based on this limited dataset, this case study contributes to a growing body of evidence that signature whistles are produced within the Sousa genus.

The stereotyped whistle produced by Herme varied in duration and loop structure, which could be indicative of the recording context. In high arousal contexts, where signalling becomes strained or tremulous, signal production (Esch , 2009; Probert , 2021) and variability in call structure (Briefer, 2012) can increase. The combination of human activity (boat presence, engine noise, swimmers in the water) and mixed species assemblage will have likely increased arousal of the dolphins during the recording context (Scarpaci , 2000). Australian bottlenose dolphins (T. australis) are known to change their behavioural pattern after being exposed to “swim-with-dolphin” tourism occasions (Filby , 2017) and even after the interactions, behaviour patterns did not return to the same levels they were before the interaction (Peters , 2013). Mixed species assemblages can elicit behavioural modifications, for example, Guyana dolphins (Sotalia guianensis) change the frequency of their whistles when interacting with bottlenose dolphins (T. truncatus) at the southern Caribbean coast of Costa Rica (May-Collado, 2010), although these changes are linked to harassment behaviour from bottlenose dolphins. Aggressive behaviour between Herme and T. aduncus individuals was not observed in this study, although he was often observed on the fringes of the social group and rarely involved in social behaviour with T. aduncus (Fig. 3). S. plumbea in Algoa Bay, South Africa, and Zanzibar Island, Tanzania, tend to avoid bottlenose dolphins, potentially to minimise costly interactions (Koper and Plön, 2016; Stensland , 2003). Thus, the variations observed in Herme's whistle type could be related to behavioural changes due to unknown, potentially complex interspecific interactions or to the nature of data collection (i.e., during swim-with-dolphin activities).

Biphonation of whistles and burst pulses were prevalent in Herme's recorded whistles indicating that the stereotypic use of biphonation might be a purposeful addition of information encoding (Papale , 2015; Sportelli , 2022). Biphonation can be categorized as the production of two fundamental frequencies, harmonically unrelated to each other, from a single sound source (Kriesell , 2014; Papale , 2015), or the production of two call types using different vocal structures, which is often a frequency modulated whistle with an overlapping burst pulse (Lilly and Miller, 1961). In other species, the production of two fundamental frequencies has been suggested to increase the probability of caller identification (Filatova , 2009; Filatova , 2012; Volodina , 2006). These can be used as markers of a pod or a matriline association (Filatova , 2009) or vocal cues for mate–mate and/or parent–chick recognition (Aubin , 2000; Jouventin and Aubin, 2002; Lengagne , 2001; Volodina , 2006). As burst pulses are more directional than whistles, the addition of these to stereotyped whistles might help to strengthen individual identification through directionality and amplification. Alternatively, this could be a result of simultaneous navigation and communication (multitasking) (Sportelli , 2022).

Opportunistically collected acoustic data using low-cost action cameras has previously been validated (Chapuis , 2021; Probert , 2023) Although most of the extracted acoustic data had a maximum usable frequency of 15 kHz, the proximity of the animals to the videographer combined with visual cues of whistle production makes this a useful dataset. Although the use of visual cues, particularly bubblestream emissions, to identify call producer has been critiqued (see Fripp, 2005), this method has previously been used to identify call producers in other dolphin species (Miles and Herzing, 2003; Slack, 2018; van der Woude, 2009), including long-term signature whistle stability in T. aduncus in the same region (Probert , 2023). Furthermore, Probert (2023) suggest that bubblestream production is an unbiased cue for identifying both signature and non-signature whistles in T. aduncus.

In conclusion, this case study documents a single stereotyped whistle that dominated recordings during encounters with a single free-swimming humpback dolphin over 11 years, whilst in a mixed-species group. Through our interrogation of the acoustic behaviour displayed, this case study could support further evidence that signature whistles are produced by Sousa spp. and provides temporal evidence of stereotyped whistle use in a free-swimming S. plumbea. Although this case study contributes to the growing body of evidence for possible signature whistle use, further data are required to fully confirm this, including single species recordings of S. plumbea, and warrants further investigation.

We would like to thank the Maputo National Park (previously known as the Ponta do Ouro Partial Marine Reserve), the Department of Conservation Areas (ANAC) in Mozambique, Natural History Museum Maputo, and previous volunteers and guests with DolphinEncountours.org. This research was funded by the National Research Foundation Marine and Coastal grant (2019–2021).

The authors have no conflicts to disclose.

All research was passive and observational, and all research activities were within the scope of normal activities carried out by Dolphin Encountours Research Center's memorandum of understanding with the Department of Conservation Areas (ANAC). Acoustic monitoring of these animals was non-invasive and carried out under ethics permits issued by Stellenbosch University, University of KwaZulu-Natal, University of Cape Town, and Ponta do Ouro Partial Marine Reserve.

The data that support the findings of this study are available from the corresponding author upon reasonable request.