First indications that northern bottlenose whales are sensitive to behavioural disturbance from anthropogenic noise

Although northern bottlenose whales were the most heavily hunted beaked whale, we have little information about this species in its remote habitat of the North Atlantic Ocean. Underwater anthropogenic noise and disruption of their natural habitat may be major threats, given the sensitivity of other beaked whales to such noise disturbance. We attached dataloggers to 13 northern bottlenose whales and compared their natural sounds and movements to those of one individual exposed to escalating levels of 1–2 kHz upsweep naval sonar signals. At a received sound pressure level (SPL) of 98 dB re 1 μPa, the whale turned to approach the sound source, but at a received SPL of 107 dB re 1 μPa, the whale began moving in an unusually straight course and then made a near 180° turn away from the source, and performed the longest and deepest dive (94 min, 2339 m) recorded for this species. Animal movement parameters differed significantly from baseline for more than 7 h until the tag fell off 33–36 km away. No clicks were emitted during the response period, indicating cessation of normal echolocation-based foraging. A sharp decline in both acoustic and visual detections of conspecifics after exposure suggests other whales in the area responded similarly. Though more data are needed, our results indicate high sensitivity of this species to acoustic disturbance, with consequent risk from marine industrialization and naval activity.

1. Introduction

Underwater noise generated by human activities such as shipping, exploration, naval sonar and other sources is considered by many international agencies to be marine pollution [1]. The extent to which noise pollution degrades the quality of marine habitats depends crucially upon the effects of noise on animals in the exposed environment. Concern for effects of underwater noise on cetaceans is marked because of their dependence on sound for communication, foraging and sensing the environment, and the tendency of some species (particularly beaked whales) to strand when exposed to intense sounds produced by naval sonars [2,3]. Spurred by these concerns, a number of recent studies have described and quantified changes in functional behaviour of free-ranging cetaceans to both simulated and real navy sonar signals [4–11]. The acoustic thresholds and types of behavioural responses vary across and within species [5,9], and also appear to be strongly affected by the context in which animals are exposed [6,12].

The species group most clearly associated with sonar-linked stranding events, beaked whales (Family: Ziphiidae), have been shown to be sensitive to simulated and real active sonar transmissions in key studies conducted near US Navy ranges: in the Bahamas Atlantic Undersea Test and Evaluation Center (AUTEC: Blainville's beaked whale, Mesoplodon densirostris [4,10]) and near the Southern California Offshore Range (SCORE: Cuvier's beaked whale, Ziphius cavirostris [7] and Baird's beaked whale, Berardius bairdii [11]). All five beaked whales exposed to sonar while tagged exhibited strong reactions, including cessation of foraging, avoidance and heightened swimming speed, that started at rather low received levels and persisted for several hours after the end of exposure periods [4,7]. The response of an individual B. bairdii was similar to that of the other two species, but was of shorter duration after sound transmission ceased [11]. Responses of individuals tagged with high-resolution DTAGs were consistent with larger scale trends in movement [4] and echolocation behaviour of whales during actual exercises on the AUTEC range [4,10]. These small but crucial datasets indicate that beaked whales are particularly sensitive to sonar, supporting the suggestion by Cox et al. [3] that behavioural responses may be an important factor leading to their stranding. Thus far, such studies have been carried out exclusively in the vicinity of navy underwater training ranges, and the high level of sonar activity in such areas is clearly an important contextual feature limiting our ability to predict responses of beaked whales in more pristine habitats.

The northern bottlenose whale, Hyperoodon ampullatus, is a beaked whale that ranges over high latitudes of the north Atlantic Ocean [13]. The current population of northern bottlenose whales is thought to be much-reduced due to intense historical whaling [14,15]. These whales were noted for their curiosity towards unusual sounds and their allegiance towards wounded companions, which, unusually among whaling vessels, led to harpoon guns mounted both fore and aft, and allowed whalers to capture entire groups [16,17]. Underwater noise, which is expected to increase in the species' habitat as human activities expand into deep Arctic waters [18], is a potential threat to the recovery of the species. One northern bottlenose whale stranded in 1988 on the island of Fuerteventura during a sonar exercise [19], but to date there have been no direct observations of how this species might respond to noise exposure. While many aspects of the diving and foraging behaviour of the northern bottlenose whale appear to be similar to those of other beaked whales [20–22], the observations of whalers that northern bottlenose whales have social defences against threats suggests that they might be less at risk of flight responses that could lead to stranding. Here, using a dose-escalation experiment, we tested how bottlenose whales in a habitat far from any naval testing range responded to naval sonar.

2. Material and methods

We conducted a controlled exposure of naval sonar to an aggregation of bottlenose whales in which one individual was tagged with a sound and movement-recording datalogger attached to the whale by suction cups (DTAG [23]). The protocol for sonar exposure specified that the sonar source was to start transmissions at a planned distance of approximately 1 km from the tagged whale after at least 8 h of baseline tag data were recorded. The tag was programmed to release after being attached to the whale for 18 h.

During the sonar exposure, the source vessel (RV HU Sverdrup II) followed a pre-determined course, independent of the movement of the tagged whale or other whales, while towing a sonar sound source at 90–100 m depth. The sonar transmission consisted of one hundred and four 1 s duration 1–2 kHz hyperbolic upsweep pulses at 20 s intervals. The source level of the sonar pulses increased by 1 dB per pulse from 152 to 214 dB re 1 μPa m over 20 min (61 pulses), and the remaining pulses were transmitted for 15 min at a source level of 214 dB re 1 μPa m.

Before, during and after the sonar exposure, observers on the source vessel recorded visual and acoustic detections of northern bottlenose whales. Whales were sighted using the naked eye and 15×80 mounted binoculars, and clicks were detected using a hydrophone array [24] towed at 90–100 m depth. All visual detections were classified as new sightings, unless they were recognized as having been sighted previously (based upon group location, size and natural markings), in which case they were classified as re-sightings. Group sizes and the number of whales within 200 m of the tagged whale group were recorded using the protocol specified in Visser et al. [25]. Acoustic detections were scored as presence or absence within 5 min intervals. Representative received levels of the sonar pulses near the experiment were computed using the BELLHOP propagation model [26], including a range-dependent bathymetry and sound speed profile measured prior to the exposure experiment. Range- and depth-dependent incoherent transmission loss was computed, using the location of the source at the start of the exposure experiment. The transmission loss was computed in 360° around the source using steps of 2°. The source was modelled as a point source, with a source depth of 97 m (based on the value of the depth sensor on the source), an opening angle of ±60° to account for the source directionality, and a source level of 214 dB re 1 μPa m (the maximum RMS source level during the experiment). We ran BELLHOP at the centre frequency of 1500 Hz, using 500 rays in order to achieve a convergence of the computed transmission loss.

Data from the tag were converted to pressure, acceleration, magnetic field strength, and pitch, roll and heading in the whale-frame axis using standard methods [23]. The calibrated received sound pressure level (SPL) was measured over a 200 ms RMS averaging window for each sonar ping [5]. This received SPL was calculated by summation of the acoustic power over the 1–40 kHz third-octave bands that had a signal-to-noise level more than 10 dB. The position of the tagged whale at the start of the experimental sonar transmissions was geo-referenced by finding the position with the smallest RMS difference between observed sonar arrival times and predicted arrival times based upon the dead-reckoned track. The time-synchronization between the tag and the GPS-synched sonar transmission clock was accurate to less than ±0.5 s, so we are confident of our whale track locations to approximately ±0.75 km during the sonar transmission period.

Audio files recovered from the experiment and baseline tags were examined aurally, and spectrograms were inspected visually to identify the start and stop times of foraging sounds (foraging echolocation clicks and buzzes, which are likely to represent prey-capture attempts) produced by the tagged whale (figure 1a), as well as those produced by other whales [21]. Sounds were ascribed to the tagged whale or not based upon their intensity in the tag recording and the timing of arrival of the sounds on stereo tags (for baseline records only). Flow noise (less than 500 Hz) recorded on the audio channel was correlated with speed through the water [27] measured during steep (more than 60° pitch) transit periods to estimate speed throughout the tag record (except for depths less than 10 m, where 2 and 3 m s⁻¹ were each modelled to bracket expected speeds near the surface). Estimated speed was combined with pitch and heading data to estimate a dead-reckoned track of the whale [23,28]. Sonar arrival times were marked in the audio record by cross-correlation with the transmitted waveform.

Behaviour of the tagged whale during the pre-exposure baseline period was summarized using two sets of quantitative variables calculated at a common sampling rate of 5 Hz. The first set of variables, movement parameters, was designed to detect predicted movements during avoidance and consisted of dive profile wiggliness (proportion zero crossings in the first difference of the depth time series), whale heading (decomposed into sine and cosine components) and variability of animal pitch and heading. The second set, energetic parameters, was designed to track locomotion effort of the whale and consisted of overall dynamic body acceleration (ODBA [29]) and pitching movements relative to the body axis. As these energetic parameters can be sensitive to tag position on the animal, we confirmed that tag movements on the body did not influence the outcome of this analysis by inspection of the data to confirm that overall ODBA values were consistent over the time period of the tag deployment. The first dive after tagging was not included in the baseline period to reduce possible short-term influences of the tagging procedure. For both sets of variables, we calculated the Mahalanobis distance between the baseline-period average value and the averages of 15 min windows centred at 1 min intervals [11]. We set a threshold for change-point detection at the 95th percentile of expected Mahalanobis distance. This threshold was derived from resampling (from baseline) 100 000 periods of the same duration as the 35 min exposure, and setting the response threshold at the 95th percentile of the maxima of the resampled periods (figure 1b,c).

To compare dives of the exposed whale with those of other tagged northern bottlenose whales, we analysed data from eight other tag records obtained from the Gully, eastern Canada, and five other tag records from Jan Mayen. All of the Jan Mayen records and two of the Gully records used DTAGs. Four Gully records used Little Leonardo 3MPD3GT loggers [30] and two were time-depth recorder data previously reported by Hooker & Baird [20]. All dives greater than 10 m depth were extracted from time-depth records. Dives were segmented into three phases: descent (from the surface until first excursion greater than 85% of maximum depth), ascent (from the final excursion greater than 85% of maximum depth to surfacing) and bottom (from end of descent to start of ascent); we also measured the surface interval after each dive (from surfacing until the start of the next dive) [31]. For each dive across all tag types, basic dive parameters were calculated, including: maximum depth, dive duration, descent rate, ascent rate, bottom duration and surface interval [31]. For tag types including acceleration sensors (DTAGs and 3MPD3GT), kinematic dive parameters included ascent pitch, descent pitch, variability of heading, variability of pitch and ODBA. ODBA values were normalized by the whale-specific median value to help account for the effects of tag locations on the animals. Dives were clustered into short-shallow or long-deep dive types by k-means clustering based upon dive duration, maximum depth, and descent and ascent rates. For basic dive parameters, all dives by all tagged whales were included in the analysis: 477 shallow dives and 79 deep dives. For the kinematic dive parameters, only data from whales tagged with DTAGs and 3MPD3GTs were included: 379 shallow dives and 57 deep dives. The Mahalanobis distance between each dive and the dive-type-specific mean value were calculated for both basic and kinematic dive parameters. We also estimated the probability of observing a Mahalanobis distance as extreme as that of the deep exposure dive (under the null hypothesis that distances were normally distributed and not dependent on exposure).

3. Results

We surveyed the waters near Jan Mayen from 22 June to 10 July 2013. During that period, 220 groups of Hyperoodon were sighted, with group sizes ranging from 1 to 10 animals, with a mean (s.d.) group size of 3.5 (1.7) individuals. On 25 June, we tagged an individual (ha13_176a) within a group of six animals that approached the tag boat. The tagged whale was later seen in a group of three to four animals. During the 10.5 h pre-exposure period, the tagged whale exhibited a diverse set of shallow and deep dives. Clicking and buzzing foraging sounds at amplitudes indicating they were most likely produced by the tagged whale were recorded, primarily during the deeper dives (figure 1). A large number of fainter foraging sounds, judged to be produced by other whales, were also recorded.

The sonar exposure started when the tagged whale had been on a dive to 100 m for a duration of 7.8 min (figure 1, inset). Based upon the geo-referenced track, the whale was approximately 5 km away, almost due north of the source when sonar transmissions started (figure 2). The whale returned to the surface for 3.6 min, taking 15 breaths. At the start of the next dive, the whale had turned towards the source with SPL of the sonar at a received level of 98 dB re 1 μPa (figure 1a, inset box top), which was, according to our criteria, at least 10 dB above the noise level in at least one 1/3-octave band between 1 and 2 kHz. Noise recorded by the DTAG can be affected by flow noise, resulting in elevation of the noise levels above ambient noise. However, the lowest 1/3-octave noise levels within the sonar band were comparable with those predicted by the Wenz curves [32] for deep sea ambient noise levels (approx. 90 dB re 1 μPa) for a Sea State=3 (representative of the weather conditions during the exposure experiment), although some increase in ambient noise level is expected when the animal was closer to the surface.

The received SPL continued to increase as the whale made an unusually smooth descent to approximately 90 m. The time-series Mahalanobis distance analysis for movement detected the onset of a behavioural response as the whale began swimming along an unusually straight course as it passed 60 m depth, still approaching the source, when received SPL of the sonar was at 107 dB re 1 μPa (figure 1). Next, the whale made a sharp change to greater depth turning away from the source as the received SPL of the sonar was 130 dB re 1 μPa.