Buzzing during biosonar-based interception of prey in the delphinids Tursiops truncatus and Pseudorca crassidens

Echolocating bats and toothed whales emit powerful sound pulses and listen for returning echoes to navigate and capture prey under poor lighting (Griffin et al., 1960; Au, 1993). The pulse repetition rate and the output energy together delimit the depth of an echolocator's acoustic gaze for unambiguous echo ranging (Wisniewska et al., 2012; Madsen and Surlykke, 2013). Early on, Griffin et al. (Griffin et al., 1960) identified distinct acoustic behaviours of bats that define the three foraging phases of search, approach and capture. In the search phase, most aerial hunting bats emit long and powerful cries. Upon the detection of a prey item, bats initiate the approach phase in which they call faster and gradually reduce call levels as they close in on prey. Just before capture, bats enter the terminal phase in which they emit a fast repetition rate buzz of short cries at low output levels. Thus, bats manipulate the rate and level of their calls to achieve dramatic acoustic gaze changes that accommodate the changing spatial relationship with their prey during capture attempts.

Although toothed whales must go through the same echo-guided tasks of search, approach and capture when hunting with echolocation, much less is known about the biosonar behaviour of most species during foraging. Early work (Norris et al., 1961; Morozov et al., 1972) showed that an echolocating bottlenose dolphin would reduce its inter-click intervals (ICIs) as it approached a dead fish, producing click repetition rates of ∼300 Hz when close to the fish, but no functional links were made to buzzing in bats, and this clicking mode received little further attention in captive studies. The next decades of biosonar research on dolphins had a strong biomimetic focus in order to understand the remarkable performance of toothed whale echolocation that may surpass that of man-made sonars (Au, 1993). All of the basic concepts and biosonar parameters used today stem from this research effort (Au, 1993). However, most of these studies involved a stationed delphinid performing an echolocation task with man-made targets at relatively long and static ranges. Remarkably, to our knowledge, no studies have been conducted to uncover in detail how a dolphin would use its sonar to do what it actually evolved for: to find and catch prey.

Contemporaneously with these captive studies, acoustic observations of narwhals in the wild producing fast click series prompted Miller et al. (Miller et al., 1995) to make a functional connection with the buzz of bats. However, it was not until the development of sound recording tags that this interpretation could be substantiated. It is now clear that buzzes are associated with prey capture attempts in sperm whales (Miller et al., 2004), beaked whales (Johnson et al., 2004), pilot whales (Aguilar Soto et al., 2008), porpoises (DeRuiter et al., 2009) and belugas (Ridgway et al., 2014).

Here, we report the biosonar behaviour during prey interception in two trained specimens of the most studied delphinid species, a false killer whale, Pseudorca crassidens (Owen 1846), and a bottlenose dolphin, Tursiops truncatus (Montagu 1821), and test the hypothesis that fast repetition rate buzzing is an integral part of echolocation for prey in these species. We show that these delphinids employ range-dependent output changes when approaching prey as found in non-prey target experiments (Au and Benoit-Bird, 2003). However, most prey capture attempts occur with a radical output change in the form of a high repetition rate, low output buzz, suggesting a universal and strong functional convergence in the biosonar of bats and toothed whales.

A total of 116 (63 herring, 53 capelin) prey capture trials with a false killer whale, Kina, and 103 (42 herring, 61 capelin) trials with a bottlenose dolphin, BJ, were recorded over the course of four recording sessions per animal. Of these, 40 (20 herring, 20 capelin) trials for Kina and 72 (33 herring, 39 capelin) trials for BJ generated echograms (see Materials and methods) (see also Johnson et al., 2004) with discernible prey echoes (Fig. 1). In general, the ability to record echoes on a tag will depend on the placement of the tag with respect to the axis of the animal's sonar beam and the sound-shading structures of its body. A false killer whale has a larger melon than does a bottlenose dolphin (see supplementary material Fig. S1) and this may have impacted on the quality of the recordings. A number of other factors may have contributed to the variable quality of the received echo signals including: the amount of air inside the fish, and hence its target strength; the angle of approach; and the proximity of the fish to other targets and the water surface, and hence the signal-to-clutter ratio. Finally, the net-pen was not separated in any way from the natural environment of Kaneohe Bay, which is abundant in snapping shrimps. The signal-to-noise ratio between sessions, or trials, could have therefore varied.

When discernible, the echoes could be tracked until the prey were 45–55 cm ahead of the blowhole for the false killer whale and 25–40 cm for the bottlenose dolphin (Fig. 1), corresponding approximately to the tip of the rostra of the animals (see supplementary material Fig. S1), and allowing verification of the exact time of prey interception. As in free-ranging beaked whales (Johnson et al., 2004), prey engulfment by the delphinids was accompanied by a rapid change in acceleration (Fig. 1) that most likely reflected movements of the musculature in the gular region, giving rise to a high magnitude jerk signature (Ydesen et al., 2014). The signature was used to pinpoint the exact time of prey interception in trials with no discernible prey echoes in the echograms.

Roughly half of prey interceptions (56% and 39% of trials with Kina and BJ, respectively) were followed by distinct, isolated sequences of rapid pulsed sounds. These were short click trains of low (∼2 ms for BJ and 2–5 ms for Kina), nearly constant or slightly increasing ICIs (Fig. 1) that may be analogous to the recently described victory squeals (VS) produced by trained belugas and bottlenose dolphins elsewhere (Ridgway et al., 2014). Because of their post-capture occurrence and their apparent communicative, rather than echolocation, function, the VS were disregarded in further ICI analyses (e.g. Fig. 2; supplementary material Figs S2, S3).

Both delphinids went through a series of acoustic gaze changes whilst approaching prey, culminating in a transition from regular clicking to a high click rate, low output buzz during the terminal approach and capture (Figs 1, 2; supplementary material Fig. S2). In 23 of the trials with Kina (15 with herring, eight with capelin), the whale did not produce any buzz clicks (ICI<14 ms; see supplementary material Fig. S3) within a body length of the prey [i.e. up to 2.4 s prior to jerk peak under the assumption of Kina moving at a normal odontocete swimming speed of 2 m s^–1 (Madsen and Surlykke, 2013)]. Similarly, on 12 occasions (four with herring and eight with capelin), BJ captured a fish without transitioning into a buzz within 1.2 m from jerk peak. However, the majority (>75%) of prey interceptions in both animals were accompanied by buzzing.

During most prey approaches, both the bottlenose dolphin and the false killer whale gradually reduced their ICIs and output levels roughly proportionate with reducing target range (Figs 1, 2; supplementary material Fig. S2), as has been seen for free-ranging delphinids approaching hydrophone arrays (Au and Benoit-Bird, 2003). Thus, the concept of automatic gain control to compensate for a reduced transmission loss with decreasing target range also seems to be applicable to some degree in delphinids approaching small prey items. Yet, both animals exercised a degree of flexibility in their biosonar sampling; either they gradually decreased ICI with decreasing target range, like harbour porpoises (Fig. 2) (Madsen and Surlykke, 2013), or kept a wide auditory scene by maintaining relatively constant clicking rates prior to buzzes, akin to beaked whales (Fig. 2) (Madsen and Surlykke, 2013). This suggests that ICI adjustments in the approach may not always be necessary or advantageous. These delphinids, and perhaps all toothed whales, seem to have a preferred strategy, but they may switch between, and perhaps combine, different degrees of adjustments, depending on the environment and the echolocation task at hand.

A much more dramatic change in the acoustic gaze, featuring increased click repetition rates and >10-fold output level reduction, occurred at the transition to a buzz (Figs 1, 2), which was initiated at a median range of 0.6 m (interquartile range IQR 0.4–0.1, Kina) and 1.5 m (IQR 0.8–1.9, BJ) from the prey [assuming an upper ICI limit for the onset of buzz of 14 ms for Kina and 16 ms for BJ (see supplementary material Fig. S3)]. Both delphinids buzzed with click repetition rates of the order of 250–500 clicks s^–1 (Fig. 2; supplementary material Fig. S2), and used short buzz durations of 1 s (0.6–1.5 s, BJ) and 0.6 s (0.1–1.3 s, Kina) with a tendency toward longer buzzes in trials with preceding gradual ICI changes (Fig. 2). The high sampling rates during the buzz are comparable to the high rate click trains, variously coined ‘mews’ or ‘creaks’, reported in early studies (Norris et al., 1961; Morozov et al., 1972). These rates are similar to those found in buzzing porpoises (DeRuiter et al., 2009) but slightly faster than those of similar-sized beaked whales (Fig. 2). They therefore deviate from a simple size scaling, but support the recently proposed hypothesis that biosonar sampling rates during capture are scaled with manoeuvrability of toothed whales (Madsen and Surlykke, 2013) as well as perhaps the agility of their preferred prey.

During both the approach and capture phases, the ICIs were consistently longer than the two-way travel time to the prey, but in the buzz the lag time, i.e. the time between the arrival of the prey echo and the subsequent click, was reduced to 1.5–4 ms just before target interception (Fig. 2). This suggests that ICI is adjusted throughout the buzz to avoid target echo ambiguity (Madsen and Surlykke, 2013). Such adjustments reveal an acute vocal-motor control that is finely tuned to the echo delays to targets of interest (Wisniewska et al., 2012). The ICI dynamics shown here (Fig. 2) further imply that the concept of a fixed lag time of ∼20 ms for echo processing, found for stationed dolphins performing target detection tasks (Au, 1993), is unsupported for free-swimming delphinids targeting prey. The very short lag times found during buzzes and their wide variability suggest that the lag time may have little connection with cognitive processing time in a buzz.

Despite having been in captivity for many years and being faced with an easy task of intercepting a dead fish in the well-lit waters of a familiar net-pen, the trained delphinids studied here displayed stereotyped buzz behaviour very similar to that reported from diverse toothed whale species when catching prey in the wild. This suggests that high-rate biosonar sampling is a deeply integrated part of close-range target interception in a toothed whale. We propose that buzzing, although hitherto largely overlooked in captive animals, is a universal and critical key to biosonar-based interception of prey in toothed whales, as is the case for bats hunting prey on the wing.

Experiments were conducted in a 12×12 m net-pen at the University of Hawaii Marine Mammal Facility with a 28 year old female bottlenose dolphin, BJ, and a >26 year old female false killer whale, Kina. Animals were trained to wear a DTAG-3 multi-sensor tag (Woods Hole Oceanographic Institution, Woods Hole, MA, USA) that simultaneously records stereo sound at a sampling rate of 500 kHz (16-bit resolution, 178 dB re. 1 μPa clip level) and samples tri-axial accelerometers at 500 Hz. The suction cup-attached tag was placed dorsally behind the blowhole (supplementary material Fig. S1) to ensure good quality recordings of outgoing clicks and to maximize the chance of recording faint echoes. Each session started with the animal stationed at the trainer, after which three dead fish (10–15 cm capelin or 20–25 cm herring, estimated target strength of –35 to –40 dB) were introduced in opposite ends of the pen. The tagged animal was then sent towards the fish for interception at depths between 0.5 and 2 m. Each session was filmed using an in-air Sony camcorder, synchronized with the DTAG-3 by tapping the tag hydrophones in front of the camera, allowing for visual verification of the time of prey ingestion on the DTAG-3 with <1 s of error.

Analyses were performed using MATLAB (The MathWorks, Natick, MA, USA). Clicks from tag recordings were extracted with a supervised click detector. Relative apparent output level (RAOL) of the clicks (Wisniewska et al., 2012) was computed as peak–peak received level in dB re. maximum level in the trial, and ICIs were calculated as time intervals between each click and the preceding click. For each species, we used a change in the distributions of ICIs of all trials as the border value between buzz and regular clicks (see supplementary material Fig. S3). Clicks with ICIs <14 ms for Kina and <16 ms for BJ were accordingly classified as being part of a buzz. Time range echograms (Johnson et al., 2004) were formed by time aligning the signal recorded in a time window from –0.5 to +6.5 ms around each click, and subsequent colour coding by intensity. The width of each time bin in the echogram was adjusted to the ICI of that click to provide a time resolution matching the animal's sampling rate and to form a time axis compatible with the video recording (Johnson et al., 2004). Echo range was computed as half the time delay between the emitted click and the returning echo divided by a sound speed of 1534 m s^–1. The range estimates were corrected for the distance between the sound source and the tip of the animal's rostrum. To investigate possible changes in the animals' movement patterns accompanying prey interception, we computed acceleration rate, or jerk, of the whales for the duration of the trials (Ydesen et al., 2014). The accelerometer data were low-pass filtered at 30 Hz (6th order Butterworth filter) and the total jerk was computed at each time instant as the norm of the differential of the acceleration for each axis (Ydesen et al., 2014).

S. Vlachos, C. Quintos, M. Breese, students and staff are thanked for assistance in the experiments conducted under University of Hawaii IACUC protocol 09-712 and US Marine Mammal Permit no.978-1567 issued to P.E.N.