SEARCH

SEARCH BY CITATION

Keywords:

  • beaked whale;
  • Navy sonar;
  • mass stranding;
  • behavioral response;
  • killer whale predation response;
  • mid-frequency active sonar;
  • Mesoplodon densirostris

Abstract

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Literature Cited
  8. Supporting Information

Increasing evidence links exposure to Navy sonar with certain mass stranding events of deep diving beaked whales. Although the cause of these strandings is unknown, one theory suggests that the animals confuse the sonar signals with vocalizations of killer whales, a known predator. Here we analyze the movement patterns of a tagged female Blainville's beaked whale in reaction to playback of killer whale predation calls. During a deep foraging dive, the whale was exposed to a playback of killer whale vocalizations with the source level slowly increased until the whale prematurely ceased foraging. The heading data from the tag were analyzed using a rotation test with a likelihood ratio calculated for a nonparametric kernel density estimate. We found a significant difference (< 0.005) in the distribution of Δheading (the change in heading averaged over 200 s) after the cessation of the killer whale playback. A test of the angular standard deviation (SD) of the Δheading showed that after the playback, the SD was significantly reduced (= 0.0064), which indicates that the animal maintained a straighter than normal course for an extended period of time. The prolonged directed avoidance response observed here suggests a behavioral reaction that could pose a risk factor for stranding.

Increasing anthropogenic noise in the ocean and its effects on marine life has become a rising concern for lawmakers and researchers alike in recent years. Of particular concern are those marine animals that utilize sound to communicate and explore their environment, such as marine mammals. Human contributions to noise in the ocean, including shipping, oil and gas development, and military activities, have greatly increased in the last 50 yr (McDonald et al. 2008). While most of the concern centers around the effects of low frequency sound on baleen whales, which can range from changes in the vocal behavior of the whales (Parks et al. 2007) to abandonment of habitat (Bryant et al. 1984), the most immediate and extreme consequences of anthropogenic sounds are the mass strandings of beaked whales associated with military mid-frequency active (MFA) sonar exercises.

Starting in the late 1990s, evidence began to accumulate that atypical mass strandings of several species of beaked whales were associated with military sonar activities (Frantzis 1998). There have been 12 mass stranding events associated with the presence of naval exercises or warships outfitted with MFA sonar, ranging in location from the Bahamas to the Mediterranean (D'Amico et al. 2009). These sonar-related mass strandings have mainly involved Cuvier's (Ziphius cavirostris) and Blainville's (Mesoplodon densirostris) beaked whales.

Beaked whales are extreme deep divers, with Blainville's beaked whales regularly conducting foraging dives to depths in excess of 1,000 m (Tyack et al. 2006). At depth they emit echolocation clicks with frequencies centered around 40 kHz and with little energy below 20 kHz (Zimmer et al. 2005). Acoustic tags have recorded echoes of these clicks from prey items, providing direct evidence that these clicks are used in foraging (Johnson et al. 2004). One study has shown that Blainville's beaked whales produce these echolocation clicks at depth for an average of 26 min and have an average total dive duration of 47 min (Tyack et al. 2006). The deep diving and infrequent surfacing behavior of beaked whales make them very difficult to study, yet they exhibit one of the most dramatic and lethal responses of marine mammals to human activities. Determining what factors cause beaked whales to mass strand is an important step in guiding regulation of sonar use in order to minimize its effects on beaked whales.

There has been extensive speculation as to what leads to the stranding and death of beaked whales during navy MFA sonar exercises. Initially it was hypothesized that the sonar caused direct physical damage to the whales, due to the presence of gas bubble lesions and subarachnoid hemorrhages observed in stranded animals (Evans and England 2001, Jepson et al. 2003) and the potential for intense sound energy to cause bubbles to grow in supersaturated tissues (Crum and Mao 1996). More recent hypotheses have focused on the possibility that sonar initiates a chain of events that lead to strandings but starts with a purely behavioral reaction. Beaked whales live in deep waters, so they must show a strong avoidance reaction to swim from their normal habitat onto the beach (Cox et al. 2006).

The frequency of MFA sonar ranges from 2.6 to 14 kHz (D'Amico et al. 2009), which is well below the best hearing range of beaked whales (Cook et al. 2006, Finneran et al. 2009). However, the sonar signals are acoustically similar to the stereotyped calls of killer whales (Orcinus orca), a primary predator of beaked whales (Zimmer and Tyack 2007). It has been hypothesized that the MFA sonar signal may initiate a predator avoidance reaction in the beaked whales, similar to the reaction elicited by killer whales, that may lead to stranding (Zimmer and Tyack 2007). Studies of killer whale predation on large baleen whales have shown that baleen whales employ two basic strategies for avoiding killer whale predation: fight and flight (Ford and Reeves 2008). Those species that employ a flight strategy attempt to outdistance the killer whales by maintaining a straight heading at high speeds over an extended time period (Ford et al. 2005, Ford and Reeves 2008). Of the flight species, both minke (Balaenoptera acutorostrata) and sei whales (Balaenoptera borealis) have been observed to strand themselves in attempts to escape predation by killer whales (Ford et al. 2005, Ford and Reeves 2008). In most cases the stranding itself leads to eventual death, but in rare cases the fleeing whale succeeded in swimming away when the tide rose and thus effectively escaped killer whale predation (Ford and Reeves 2008). It has been hypothesized that beaked whales may employ an avoidance strategy similar to these whales, and that the strandings are the result of either mistaken direction during flight, or a deliberate action taken to avoid what they may perceive as an immediate threat.

Understanding what factors lead beaked whales to strand during navy sonar exercises is an important step in determining how to reduce the risk of these activities. However, the elusive nature of these animals and the diverse factors involved in each stranding incident lead to extreme difficulty in studying this problem.

This paper utilizes a controlled exposure experiment to test one beaked whale's reaction to MFA sonar signals and the calls of mammal-eating killer whales filtered to a frequency bandwidth similar to that of MFA sonar. This experiment was designed to test the above hypothesis that beaked whales respond to killer whale predation calls with a directed prolonged avoidance reaction similar to the flight response of baleen whales. We use the heading data from a tagged beaked whale to develop a method of statistical analysis of avoidance reactions and discuss the implications of the observed reaction.

Methods

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Literature Cited
  8. Supporting Information

Field Site

To reduce some of the difficulty associated with locating beaked whales for study, the experiment was conducted on the Atlantic Undersea Test and Evaluation Center (AUTEC) near Andros Island, Bahamas. AUTEC is located in the Tongue of the Ocean (TOTO), a deep water canyon (maximum depth ~2,000 m) that runs roughly north-south with the only deep water entrance located at the northern end (Fig. 1). This area is home to three species of beaked whales, with Blainville's beaked whale being the most common (Claridge 2006). Roughly 25 Blainville's beaked whales use TOTO as a foraging ground at any one time (Marques et al. 2009). This canyon was chosen as a study site due to the presence of an array of 82 hydrophones installed by the U.S. Navy on the sea floor of the AUTEC range. A marine mammal monitoring program has been installed to localize the echolocation clicks of Blainville's beaked whales in real time (Ward et al. 2008) and this system was utilized during the study to monitor the clicking of the tagged whale.

image

Figure 1. A map of the location of the study area in the Tongue of the Ocean, near Andros Island, Bahamas. The black box indicates the area containing the AUTEC hydrophone array, and the deployment and recovery locations of the Dtag are marked. The two locations are approximately 24 km apart. Bathymetry data from Amante and Eakins (2009).

Download figure to PowerPoint

Digital Acoustic Recording Tagging

This study used a digital acoustic recording tag (Dtag), which is an archival suction cup tag that contains a pressure sensor and three-axis magnetometers and accelerometers that measure depth, pitch, roll, and heading of the tagged whale at a sample rate of 50 Hz (Johnson and Tyack 2003). In addition, two hydrophones record acoustic data at a sampling rate of 192 kHz (Johnson and Tyack 2003). The tag is designed for deployments of up to 17 h, and is attached to the whale via four suction cups. The tag releases at a preprogrammed time, and is tracked and recovered utilizing a VHF radio transmitter.

Tagged Whale

A Dtag was deployed on a female Blainville's beaked whale on 2 September 2007 within the AUTEC range. For the duration of the deployment, the whale was tracked while at the surface utilizing the VHF radio beacon on the tag. The whale was monitored during its first three foraging dives by localizing its echolocation clicks via the AUTEC hydrophone array. When possible, visual sightings of the tagged whale at the surface were utilized to augment the tracking data.

Playbacks

The tagged whale was exposed to two stimuli: an MFA sonar signal and vocalizations of marine mammal–eating killer whales. All playbacks were conducted utilizing a Naval Undersea Warfare Center (NUWC) Eryn I MFA source. The source is capable of transmitting MFA sonar signals and other broadband sounds in the 2–5 kHz band, up to a source level (SL) of 212–214 dB re 1 μPa at 1 m. The beam pattern is somewhat directional, with more of the output acoustic energy directed near to the horizontal plane of the source. For the duration of the playbacks, the transducer was deployed from the M/V Ranger at a depth of 45 m while the ship drifted at a distance of approximately 1 km from where the tagged whale began its deep foraging dives.

After the whale conducted a single preexposure dive and began a second foraging dive, an MFA sonar playback was performed. Playback was not initiated until foraging began, indicated by reception of echolocation clicks on the AUTEC array. The MFA sonar signal was designed to simulate an actual waveform transmitted by the U.S. Navy. It was composed of three sequential components: a 0.5 s linear frequency modulated upsweep from 3.2 to 3.3 kHz, a 0.5 s constant frequency tone of 3.43 kHz, a 0.1 s silent interval and a 0.3 s constant frequency tone of 3.75 kHz. The signal sequence was repeated every 25 s. The playback started at a source level (SL) of 152 dB re 1 μPa at 1 m, and was increased by 3 dB every 25 s. The playback protocol called for continual increase of the SL until echolocation clicks from the foraging whale were no longer heard on the AUTEC hydrophone array, or a maximum SL of 212 dB re 1 μPa at 1 m was achieved.

Once the tagged whale started producing echolocation clicks on the third posttagging dive, playback of the killer whale predation calls was initiated. The transmitted killer whale sounds consisted of a 10 min segment of recordings from wild marine mammal-eating killer whales recorded in southeast Alaska. The killer whale calls were band-pass filtered to a range of 2–5 kHz, in order to match the frequency range of the transducer (Fig. S1). The killer whale playback was initiated at a SL of 130–140 dB re 1 μPa at 1 m, and then increased by 5 dB every 30 s, reaching a maximum of 190–203 dB re 1 μPa at 1 m. Playback was terminated several minutes after echolocation clicks ceased to be detected on the AUTEC array. Data from the whale were recorded continuously until the Dtag detached approximately 10 h later.

Statistical Analysis

The heading data recorded on the Dtag were used to conduct a statistical analysis to test if the tagged whale's movement patterns from before either the MFA sonar or the killer whale playback were different from those after each playback. The observed headings were averaged over nonoverlapping 200 s intervals in order to filter out any small-scale variation in movements due to fluking motion, head scanning, etc. For this analysis, the change between subsequent averaged headings (Δheading), rather than the true heading of the whale, was utilized in order to test for patterns of change in movement. ΔHeading was calculated using CircStat, a circular statistics toolbox for MATLAB (Berens 2009).

Let Δ1, Δ2, …, Δτ, Δτ+1, …, Δn be the time series of heading changes where τ is the time of the cessation of the playback, which approximates initiation of the whale's response to each playback. We assume that Δ1, Δ2, …, Δτ are independent and identically distributed with unknown probability density function fB(Δ) and Δ1, Δ2,…, Δτ are also independent and identically distributed with probability density function fA(Δ). We tested the null hypothesis: H0:fB = fA = f that heading changes before and after the playback have a common distribution against the alternative hypothesis: H1:fBfA that they do not.

The Δheading data were used to conduct a nonparametric likelihood ratio (NLR) test to determine if the distributions of the data before and after the each playback were different. Under this model, the log-likelihood is given by:

  • display math(1)

and the likelihood ratio statistic for testing H0 against H1 is:

  • display math(2)

where log L1 and log L0 are the maximized values of the log-likelihood under H0 and H1, respectively. In the absence of suitable parametric models for fB and fA, a NLR can be formed using:

  • display math(3)

and

  • display math(4)

where inline image is a nonparametric kernel estimate of fB based only on Δ1, Δ2, …, Δτ, inline image is a nonparametric kernel estimate of fA based only on Δτ+1, Δτ+2, …, Δn and inline image is a nonparametric kernel estimate of f based on the entire time series. This is an example of a nonparametric likelihood ratio statistic (Cao and Van Keilegom 2006).

To assess the significance of the observed value of the NLR statistic, we used the rotation method of DeRuiter and Solow (2008). This involved transforming the time series into a circle by joining its end to its beginning and then rotating the order by one sample. The NLR statistic was then calculated for the rotated time series, using the same breakpoint position as the observed data. This process was repeated for each rotation position until we stepped through the entire time series. The observed significance level (or P-value) was then estimated by the proportion of rotated time series for which NLR exceeded the observed value. The advantage of this approach is that, except for a negligible end effect, it preserves any serial dependence in the rotated time series of Δheading. Such serial dependence can undermine the validity of a standard randomization test in which the time series is randomly scrambled (Manly 2006).

The kernel density estimate (KDE) for the angular data was calculated according to Fisher (1995). Briefly, the KDE based on observations Δ1, Δ2, …, Δm has the form:

  • display math(5)

where K is a probability density function symmetric around 0, |Δ − Δj| is the angular difference between Δ and Δj, and h is the bandwidth that controls the smoothness of inline image. Here we use the bisquare kernel:

  • display math(6)

A common choice of bandwidth is:

  • display math

where inline image is the sample angular standard deviation of the observations. For cosmetic reasons, we used = 1.5hS. However, the results are insensitive to the choice of bandwidth in this vicinity.

Standard Deviation Analysis

The nonparametric likelihood ratio test is designed to test for a general change in the distribution of Δheading. To sharpen the analysis, we focused on detecting a change in the dispersion of Δheading as measured by the angular standard deviation σang. Let inline image and inline image be the sample angular standard deviations formed from the data before and after the cessation of the killer whale playback, respectively. To test the null hypothesis H0ang,B = σang,A against the alternative hypothesis H1ang,B ≠ σang,A, we formed the absolute difference inline image. The significance of this absolute difference was assessed by the same rotation procedure outlined above. In this case, the P-value was approximated by the proportion of rotated time series for which the value of inline image exceeded the observed value.

Results

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Literature Cited
  8. Supporting Information

During August and September of 2007, we used a digital acoustic recording tag (Dtag) (Johnson and Tyack 2003) to conduct a behavioral response study of a Blainville's beaked whale. The Dtag was deployed on an adult female Blainville's beaked whale at 24.6025ºN, 77.6210ºW on 2 September 2007 (Fig. 1).

After tag attachment, the whale conducted a deep dive that we considered a preexposure baseline dive. Clicks from the tagged whale were monitored on the AUTEC hydrophone array. After the whale initiated its second deep dive and was heard producing echolocation clicks associated with foraging, the MFA playback was initiated. The whale ceased clicking 9 min after the start of playback, when the received level of the sonar signal at the tag was 138 dB re 1 μPa sound pressure level (SPL), with a cumulative sound exposure level of 142 dB re 1 μPa2s (fig. 9, Tyack et al. 2011). The whale then ascended more slowly than usual and moved away from the sound source. The whale remained in the area for around 2 h and then commenced a third foraging dive (Tyack et al. 2011).

Once foraging clicks were initiated on the third dive, the whale was exposed to playback of the killer whale calls. The killer whale playback was also slowly ramped up as the tagged whale's clicking was monitored. The whale stopped clicking about 5 min into the playback, approximately 1 min after the received level of the killer whale sound reached 98 dB re 1 μPa SPL (fig. 10, Tyack et al. 2011). The whale then again made a slow ascent, the slowest analyzed from a set of 32 deep foraging dives from six whales tagged at this site (Tyack et al. 2011). After surfacing, the whale swam away from the playback location for approximately 10 h, before the tag detached and was then recovered at 24.8136ºN, 77.6265ºW, a location approximately 24 km away from the deployment site (Fig. 1).

Utilizing speed estimation from the pitch angle and the rate of change of depth recorded on the Dtag, a rough approximation of the tagged whale's path, called a pseudo-track, was generated (Tyack et al. 2011) (Fig. 2). As seen in Fig. 2, after cessation of the MFA sonar playback, the whale briefly maintained a course heading to the north. After several hours, the whale started a deep foraging dive. After cessation of the killer whale playback, the whale maintained a heading directed to the north for the remainder of the tag attachment (Fig. 3). If the whale continued on this course after tag detachment, it would have passed through the only deep-water exit from the TOTO canyon.

image

Figure 2. The estimated tracks of four tagged beaked whales. The horizontal track of the tagged whale is indicated in dark gray, with deep foraging dives marked in dark green. The sonar playback and killer whale playback are marked in red and labeled. The tracks of three other beaked whales tagged, but not exposed to playbacks, are indicated in light gray (reproduced from Tyack et al. 2011). A significant straightening of the whale's course is apparent after the end of the killer whale playback.

Download figure to PowerPoint

image

Figure 3. A time series plot of the heading data averaged over 200 s with the breakpoint in the data (cessation of the killer whale playback) marked by the red line. The variation in heading is noticeably reduced after the breakpoint and mainly fluctuates around 0ºN.

Download figure to PowerPoint

In order to test whether the whale altered its movement patterns in response to either the MFA sonar or killer whale playback, we performed a rotation test of the heading data from the Dtag. We used a nonparametric likelihood ratio test (NLR) (Cao and Van Keilegom 2006) to determine if the distribution of Δheading was different in the two periods: before and after cessation of the MFA and killer whale playbacks. The kernel density estimate (KDE) was calculated for each of the time periods (Fig. 4) and we assessed the significance of the observed value of the NLR statistic via a discrete-time version of a rotation test (Deruiter and Solow 2008). Of 312 NLR values generated using the breakpoint defined by cessation of the MFA playback, 146 exceeded the observed value, giving a P-value of 0.468 (Fig. S2). This indicates that there is no significant change in the whale's movements after the cessation of the MFA sonar playback. Of 312 values of the NLR statistic generated in this way for the killer whale playback, none exceeded the observed value (Fig. 5) giving an estimated P-value of <0.005. Therefore, we conclude that there is a significant difference in the whale's movement behavior between these two periods, as reflected in the distribution of Δheading.

image

Figure 4. Angular histogram plots of Δheading data in blue with the kernel density estimate drawn in red. The distribution of Δheading data before the killer whale breakpoint is illustrated in A, with the maximum occurrence of greater than 0.5 for the angles, and the distribution of Δheading data after the killer whale breakpoint is shown in B, with a maximum occurrence of greater than 1.

Download figure to PowerPoint

image

Figure 5. Comparison of rotated and observed nonparametric likelihood-ratios (NLR) calculated for the killer whale playback breakpoint. The histogram shows the distribution of the NLRs calculated for rotated data order and the observed NLR is plotted in red. It is outside the range of all NLRs calculated for rotated data order (< 0.005).

Download figure to PowerPoint

In order to further our understanding of how the beaked whale responded to the killer whale playback, we tested for a difference in the dispersion of Δheading after the killer whale breakpoint, as measured by the angular standard deviation (Fisher 1995). As before, significance was assessed by rotating the order of the time series of Δheading. Of the 312 values generated this way, two exceeded the observed value, giving an estimated = 0.0064 (Fig. 6). While the distributions of Δheading both before and after the breakpoint are centered around zero, the angular standard deviation of the data after the breakpoint was 27.4º less than that before. This reduced standard deviation indicates that the tagged whale maintained a more directed course after the cessation of the killer whale playback.

image

Figure 6. Comparison of the rotated and observed difference in absolute standard deviation. The histogram shows the distribution of the difference in angular standard deviation between before and after the killer whale breakpoint for all rotations of data order. The observed difference in standard deviation is shown in red and is greater than almost all of the values calculated via the rotation test (= 0.0064).

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Literature Cited
  8. Supporting Information

This study utilized a playback experiment to test the behavioral reaction of a tagged Blainville's beaked whale to MFA sonar and the calls of killer whales that feed on marine mammals. Due to the difficult nature of finding and tagging M. densirostris, this study represents the only playback experiment to date for these whales with an extended monitoring period after exposure. Determining what features of MFA sonar cause beaked whales to strand is an important but difficult task. A whale living in deep water must swim far from its typical habitat before it is at risk of stranding. Baleen whales avoiding predation by killer whales have been observed to strand (Ford et al. 2005, Ford and Reeves 2008), suggesting that directed avoidance in reaction to predators may increase a whale's risk of stranding. Therefore, we use heading data here to study whether a beaked whale responded to playback of MFA sonar or killer whale calls with a straighter course of travel that would cause it to swim far from its foraging site, potentially raising the risk of stranding. The small sample size limits the conclusions that can be drawn from the experimental scenario. However, utilizing the heading data from the Dtag, we are able to employ a novel statistical technique to draw some basic conclusions about the data.

During exposure to each of the playback stimuli the whale ceased clicking early in the deep foraging dive at a received level of 138 dB re 1 μPa SPL for the MFA playback and 98 dB re 1 μPa SPL for the killer whale playback. In each case, after cessation of clicking, the whale initiated a slower than normal ascent to the surface (Tyack et al. 2011). While there is a temporary avoidance reaction to the MFA sonar playback, observed as a straightening of course (Fig. 2), the whale appeared to resume normal foraging about two hours after surfacing (Tyack et al. 2011). A test of the heading data before and after cessation of the MFA playback revealed that there were no significant differences in the whale's heading after this playback (Fig. S2). An extended avoidance reaction was observed only after the killer whale playback. However, because the stimuli were played in sequence, we cannot rule out the possibility that the behavioral response was cumulative, and that the MFA sonar playback only several hours earlier had a potentiating effect on the response to the killer whale playback.

When the Δheading data were split into two sections, a likelihood ratio test revealed that, when the cessation of the killer whale playback was used as the breakpoint, the distributions for these two groups were significantly different (Fig. 5). Further testing indicated a significant difference in the angular standard deviation of the Δheading data, with the SD of the Δheading distribution after the playback significantly lower than would be predicted from rotated data (Fig. 6). This indicates that the whale maintained a more directed course after the cessation of the killer whale playback (Fig. 2, 3). The whale's course heading was centered on a northerly direction (Fig. 7), which took it directly away from the source of the playback, and towards the only deep-water exit of the TOTO canyon.

image

Figure 7. Angular histogram plots illustrating the distribution of heading data before the killer whale breakpoint (A) and after the killer whale breakpoint (B). There is a maximum occurrence of greater than 5 for the heading angles in A and greater than 25 in B. Clustering around the north (0°) to north-west direction can be seen in B.

Download figure to PowerPoint

It should be noted that, while the experiment was designed to test for a change in movement patterns, as measured by heading, the angular standard deviation test was developed post hoc. The NLR tests for any change in the distribution of the Δheading data. Once it was determined that there was a significant difference between the distribution of the whale's heading before and after the killer whale playback, we then chose to focus on the variation in heading, as measured by the angular standard deviation. This decision was influenced by the observed results, and ideally, the test developed after this examination of the data would be utilized to confirm these findings in future playback experiments. However, the difficulty involved in finding and tagging beaked whales made this unfeasible in this case. One goal for this paper is to encourage similar future playback experiments to use this method to test for similar responses.

This prolonged, directed avoidance in reaction to the killer whale playback put increasing distance between the whale and the location of the playback, similar to that seen in predation avoidance by other species. Minke and sei whales, which employ this flight strategy, have been observed to beach themselves while being chased by killer whales (Ford et al. 2005, Ford and Reeves 2008). The reaction observed here may be an antipredator response similar to the flight reaction of baleen whales to killer whale predators (Ford et al. 2005, Ford and Reeves 2008) and it is possible that this sustained directed flight puts beaked whales at risk for stranding as well. It is not apparent whether the strandings of baleen whales were the result of an intentional avoidance strategy, or if the whales inadvertently ran into the shallows due to their fixed course, or were perhaps driven ashore by the pursuing whales (Ford et al. 2005, Ford and Reeves 2008). Regardless of the reason for stranding, in only one observed case was a minke whale able to work its way off the beach after the killer whales departed. Therefore, if it is an intentional strategy, it must be a last ditch very high risk effort, motivated by extreme predation pressure. If Blainville's beaked whales utilize a similar strategy, then in extreme cases this may put them at risk for stranding. In this experiment, a prolonged avoidance response was observed for playback of killer whale sounds, but not for the short, low-level playback of MFA sonar. However, if naval sonar exercises are either very loud, very extended or both, it is possible that they could elicit this same prolonged avoidance response in beaked whales that could lead to stranding.

In addition to the extreme response of mass strandings, it is possible that lower levels of MFA sonar exposure could produce lesser behavioral reactions that could still have adverse effects on the whales. The greater variation in the Δheading before the killer whale breakpoint likely represents standard foraging search patterns. These whales forage on deep dwelling prey items that may be located in discrete patches (Johnson et al. 2008), therefore they likely employ a foraging search pattern that maximizes their likelihood of encountering these patches. While we filtered out the smaller scale movements, the whales are still likely to move between feeding sites over the longer term. The light gray tracks in Figure 2 indicate the restricted area search typical for undisturbed beaked whales in the Tongue of the Ocean. The reduced variation in the Δheading of the tagged whale after the killer whale breakpoint indicates that it maintained a relatively straight course. Analysis of the acoustic record of the tag shows that the number of buzzes produced, which indicate prey capture events, was reduced during the sonar and killer whale playbacks and then increased in subsequent foraging dives (Tyack et al. 2011). These factors together may indicate that the whale was immediately reducing foraging effort in favor of directed flight from the area of playbacks. Areas with frequent sonar exercises may cause the resident population of beaked whales to abandon their preferred foraging habitat during sonar playbacks, possibly reducing their foraging intake or foraging selectivity (Tyack et al. 2011).

The whale reacted to a much lower received level for the killer whale playback than for the MFA sonar playback, however these stimuli were played in sequence so we cannot rule out the possibility that the effect of the playbacks was cumulative. Additionally, the AUTEC range is frequently used for naval sonar exercises including those utilizing MFA sonar signals. The repeated exposure to this signal may have habituated the tagged animal to these sounds, leading to the reduced reaction to the MFA playback. By contrast, killer whales are very rare in AUTEC waters. The calls of killer whales are likely a much less frequent sound heard at AUTEC than MFA sonar, so we cannot determine if the beaked whale recognized the sound as a potentially lethal predator, or whether it simply interpreted it as a novel sound, thus causing the stronger response to the killer whale playback.

Two other factors make it possible that the killer whale playback stimulus could have been interpreted as a novel sound rather than recognized as killer whales. The killer whale vocalizations used for the playback stimulus were filtered to match the frequency range of the source, reducing the bandwidth considerably (Fig. S1). The recordings are also from the population of mammal-eating killer whales residing in British Columbia, and therefore may differ from those of the whales in the Bahamas. We cannot disregard the possibility that these two alterations may have been significant enough to change the whale's perception of the stimulus, from that of a predation call to simply a novel signal.

Additionally, while the Navy MFA sonar contains frequency and timing elements similar to that of killer whale predation calls, it is not an exact match. In the MFA playback, one 1.3 s MFA sonar sound was played every 25 s, while the killer whale stimulus was an actual recordings of natural sounds, often with more than one vocalization every 25 s. However, both the MFA and killer whale sounds are below the best hearing range of those beaked whale species whose hearing has been measured (Cook et al. 2006). The lowered perception of signals in this frequency range may mean that the whales err on the side of caution and interpret the sonar signals in a natural behavioral context as similar to the sounds of a predator. The mismatch of some of the elements of the two signals may mean that the whales require either higher received levels or greater cumulative sound exposure levels in order to induce an antipredator reaction.

While it is not possible to draw a direct connection between MFA sonar and an antipredator behavioral reaction in M. densirostris due to the limited sample size and confounding factors, a definitive behavioral reaction has been quantified in this experiment. Despite the confounding factors, our results do show that Blainville's beaked whales respond to modified killer whale predation sounds with a prolonged and directed avoidance reaction. The method developed here can be applied to movement data from future controlled exposure experiments. Further experiments should focus on differentiating between the reactions to the two stimuli.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Literature Cited
  8. Supporting Information

The authors acknowledge the support and involvement of numerous field participants in this project. In particular, we acknowledge Leigh Hickmott who provided tagging support, and Walter Zimmer for analysis of tag data. In addition, we acknowledge Ian Boyd, Christopher Clark, Diane Claridge, David Moretti, and Brandon Southall for their invaluable work conceiving of, planning, and executing this project. We thank Ari Daniel Shapiro for the initiation of the data analysis. We also thank Volker Deecke for providing the recordings of killer whales used as a playback stimulus.

The authors acknowledge the support of the MASTS pooling initiative (The Marine Alliance for Science and Technology for Scotland) in the completion of this study. MASTS is funded by the Scottish Funding Council (grant reference HR09011) and contributing institutions. This research was conducted under permits for marine mammal research issued by the U.S. National Marine Fisheries Service to John Boreman (Permit 1121-1900) and to Peter Tyack (Permit 81-1578), and issued by the Government of the Bahamas to Ian Boyd (Bahamas permit 02/07). The experiment was approved by the Institutional Animal Care and Use Committees of both Woods Hole Oceanographic Institution and the Bahamas Marine Mammal Research Organisation and the Animal Welfare and Ethics Committee of the University of St Andrews.

Literature Cited

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Literature Cited
  8. Supporting Information
  • Amante, C., and B. W. Eakins. 2009. ETOPO1 1 arc-minute global relief model: Procedures, data sources and analysis. NOAA Technical Memorandum NESDIS NGDC-24. 19 pp.
  • Berens, P. 2009. CircStat: A MATLAB toolbox for circular statistics. Journal of Statistical Software 31:2527.
  • Bryant, P. J., C. M. Lafferty and S. K. Lafferty. 1984. Reoccupation of Laguna Guerrero Negro, Baja California, Mexico, by gray whales. Pages 375387 in M. L. Jones, S. L. Swartz and S. Leatherwood, eds. The gray whale (Eschrichtius robustus). Academic Press, Orlando, FA.
  • Cao, R., and I. Van Keilegom. 2006. Empirical likelihood tests for two-sample problems via nonparametric density estimation. Canadian Journal of Statistics 34:6177.
  • Claridge, D. E. 2006. Fine-scale distribution and habitat selection of beaked whales. M.Sc. thesis, University of Aberdeen, Aberdeen, Scotland. 238 pp.
  • Cook, M., R. Varela, J. Goldstein, et al. 2006. Beaked whale auditory evoked potential hearing measurements. Journal of Comparative Physiology A: Neuroethology, Sensory, Neural, and Behavioral Physiology 192:489495.
  • Cox, T., T. Ragen, A. Read, et al. 2006. Understanding the impacts of anthropogenic sound on beaked whales. Journal of Cetacean Research and Management 7:177187.
  • Crum, L. A., and Y. Mao. 1996. Acoustically enhanced bubble growth at low frequencies and its implications for human diver and marine mammal safety. The Journal of the Acoustical Society of America 99:28982907.
  • D'Amico, A., R. C. Gisiner, D. R. Ketten, J. A. Hammock, C. Johnson, P. L. Tyack and J. Mead. 2009. Beaked whale strandings and naval exercises. Aquatic Mammals 35:452472.
  • Deruiter, S. L., and A. R. Solow. 2008. A rotation test for behavioural point-process data. Animal Behaviour 76:14291434.
  • Evans, D. L., and G. R. England. 2001. Joint Interim Report, Bahamas Marine Mammal Stranding Event of 15–16 March 2000. 59 pp. Available at http://www.nmfs.noaa.gov/pr/pdfs/health/stranding_bahamas2000.pdf.
  • Finneran, J. J., D. S. Houser, B. Mase-Guthrie, R. Y. Ewing and R. G. Lingenfelser. 2009. Auditory evoked potentials in a stranded Gervais' beaked whale (Mesoplodon europaeus). The Journal of the Acoustical Society of America 126:484490.
  • Fisher, N. I. 1995. Statistical analysis of circular data. Cambridge University Press, New York, NY.
  • Ford, J. K. B., and R. R. Reeves. 2008. Fight or flight: Antipredator strategies of baleen whales. Mammal Review 38:5086.
  • Ford, J. K. B., G. M. Ellis, D. R. Matkin, K. C. Balcomb, D. Briggs and A. B. Morton. 2005. Killer whale attacks on minke whales: Prey capture and antipredator tactics. Marine Mammal Science 21:603618.
  • Frantzis, A. 1998. Does acoustic testing strand whales? Nature 392:2929.
  • Jepson, P. D., M. Arbelo, R. Deaville, et al. 2003. Gas-bubble lesions in stranded cetaceans. Nature 425:575576.
  • Johnson, M. P., and P. L. Tyack. 2003. A digital acoustic recording tag for measuring the response of wild marine mammals to sound. IEEE journal of oceanic engineering 28:312.
  • Johnson, M., P. T. Madsen, W. M. X. Zimmer, N. A. De Soto and P. L. Tyack. 2004. Beaked whales echolocate on prey. Proceedings of the Royal Society of London Series B: Biological Sciences 271:S383S386.
  • Johnson, M., L. S. Hickmott, N. Aguilar Soto and P. T. Madsen. 2008. Echolocation behaviour adapted to prey in foraging Blainville's beaked whale (Mesoplodon densirostris). Proceedings of the Royal Society Series B: Biological Sciences 275:133139.
  • Manly, B. F. J. 2006. Randomization, bootstrap and monte carlo methods in biology. Second edition. Taylor & Francis, Boca Raton, FL.
  • Marques, T. A., L. Thomas, J. Ward, N. Dimarzio and P. L. Tyack. 2009. Estimating cetacean population density using fixed passive acoustic sensors: An example with Blainville's beaked whales. The Journal of the Acoustical Society of America 125:19821994.
  • McDonald, M. A., J. A. Hildebrand, S. M. Wiggins and D. Ross. 2008. A 50 year comparison of ambient ocean noise near San Clemente Island: A bathymetrically complex coastal region off Southern California. The Journal of the Acoustical Society of America 124:19851992.
  • Parks, S. E., C. W. Clark and P. L. Tyack. 2007. Short- and long-term changes in right whale calling behavior: The potential effects of noise on acoustic communication. The Journal of the Acoustical Society of America 122:37253731.
  • Tyack, P. L., M. Johnson, N. Aguilar De Soto, A. Sturlese and P. T. Madsen. 2006. Extreme diving of beaked whales. The Journal of Experimental Biology 209:42384252.
  • Tyack, P. L., W. M. X. Zimmer, D. Moretti, et al. 2011. Beaked whales respond to simulated and actual navy sonar. PLoS ONE 6:e17009.
  • Ward, J., R. Morrissey, D. Moretti, et al. 2008. Passive acoustic detection and localization of Mesoplodon densirostris (Blainville's beaked whale) vocalizations using distributed bottom-mounted hydrophones in conjunction with a Digital Tag (DTag) recording. Canadian Acoustics 36:6066.
  • Zimmer, W. M. X., and P. L. Tyack. 2007. Repetitive shallow dives pose decompression risk in deep-diving beaked whales. Marine Mammal Science 23:888925.
  • Zimmer, W. M. X., M. P. Johnson, P. T. Madsen and P. L. Tyack. 2005. Echolocation clicks of free-ranging Cuvier's beaked whales (Ziphius cavirostris). The Journal of the Acoustical Society of America 117:39193927.

Supporting Information

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Literature Cited
  8. Supporting Information
FilenameFormatSizeDescription
mms12028-sup-0001-FigS1.epsimage/eps1051KFigure S1. Spectrograms of a 17 s segment of the killer whale playback stimulus. Top: Spectrogram of the unfiltered killer whale vocalizations. Bottom: Spectrogram of the killer whale vocalizations filtered to between 2 and 5 kHz.
mms12028-sup-0002-FigS2.epsimage/eps507KFigure S2. Comparison of rotated and observed nonparametric likelihood-ratios (NLR) calculated for the MFA sonar playback breakpoint. The histogram shows the distribution of the NLR calculated for the rotated data order, and the observed NLR is plotted in red. It is exceeded by 146 of the rotated NLR values, giving a P-value of 0.468.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.