Current State of Knowledge of Behavioral and Physiological Effects of Noise on Marine Mammals
BEHAVIORAL RESPONSES TO ACOUSTIC STIMULI
Various approaches have been used to study behavioral responses of marine mammals to acoustic exposure. Observational studies have been used to correlate distributional or behavioral effects on uncontrolled human activities. That approach is particularly suited to the large spatial or temporal scales over which there may be consistent variation in human activities. For example, Bryant et al. (1984) collated sighting data from surveys of gray whales in one of their breeding lagoons. They reported fewer gray whales sighted after a saltworks started dredging and shipping in the lagoon. Gray whales apparently abandoned the lagoon during this activity, and took several years to start using the lagoon again after the saltworks ceased operating. Although long-term abandonment of critical gray whale breeding habitat clearly reaches the threshold of biological significance, it has not been demonstrated that it impeded the recovery of the population. Morton and Symonds (2002) report a significant decline in sightings of killer whales during a 5-year period when acoustic-harassment devices were operated in an area of water about 10 km × 10 km in an archipelago. The acoustic-harassment devices have a source level of about 194 dB re 1 μPa at 1m and are designed to be loud enough to deter pinnipeds from breaking into fish farms to feed, but they have unintended consequences for inshore cetaceans. The exposures that caused an avoidance reaction in the killer whales are not known—a common problem in
correlational studies when precise relationships between acoustic stimuli and behavioral responses are obscure.
Researchers have addressed concerns that marine mammals might avoid intense sounds. Some census studies have towed hydrophones through areas with commercial seismic surveys. Rankin (1999) and Norris, et al. (2000) found no association between the signal-to-noise ratio of seismic impulses from airgun arrays and sighting rates of cetaceans, but they caution that their analysis was so crude that it was unable to detect changes in distribution of less than 100 km. Their study exemplifies the critical point that a reported lack of an effect must carefully specify the statistical power of a study to detect specific effects. Other studies sighting marine mammals closer to sound sources have found avoidance at several hundred to thousands of meters (e.g., Goold, 1996). And some studies have shown no displacements. Ringed seals (Phoca hispida) near an artificial-island drilling site were monitored before and during development of the site. Although in-air and underwater sound was audible to the seals for up to 5 km, there was no change in their density in that area between breeding seasons before and breeding seasons after development began (Moulton et al., 2003).
The last few decades have seen the development of experiments designed to study the causal relationship between exposure to sound and behavior. As Tyack et al. (2004) discuss, these controlled-exposure experiments (CEEs) are similar to playback experiments that are commonly used to study animal communication. The primary difference is that CEEs carefully titrate the acoustic exposure required to elicit a specific behavioral reaction. In few studies have responses of marine mammals been related to levels of anthropogenic sounds. Playback of sounds associated with oil-industry activities indicated a clear relationship between the received-sound pressure level and the probability that migrating gray whales will deviate from their migration path. For continuous sounds, about 50% of the whales avoided exposure to about 120 dB rms re 1 μPa; for short impulses from airguns (about 0.01 sec every 10 sec), 50% avoidance occurred at about 170 dB re 1 μPa (Malme et al., 1983, 1984; Tyack, 1998; airgun levels are average pulse pressures). Tyack and Clark (1998) replicated the earlier experiments of Malme and colleagues by using Surveillance Towed Array Sensor System-Low Frequency Active (SURTASS-LFA) sonar sounds transmitted for 42 sec every 6 min and found that course deflection occurred when the received levels were about 140 dB rms re 1 μPa. Not only was there a steady increase in avoidance with increasing received level of each
stimulus type, but there also was a clear pattern in which higher levels were required to achieve the same avoidance when signals were of shorter duration and lower duty cycle. Similar relationships between temporary threshold shift (TTS) and duration of the sound have been shown in laboratory studies (see below in the discussion of physiological effects).
Other CEEs have found a relationship between received level of sound and probability of some responses and less relationship for others. In a playback experiment involving the SURTASS-LFA sound and singing humpback whales (Megaptera novaeangliae), Fristrup et al. (2003) analyzed 378 songs recorded before, during, and after playback. They found that the songs of the humpback whales were longer when the playback was louder (they could not determine received level at the whale). Miller et al. (2000) followed 16 singers during 18 of the same playbacks. During 18 playbacks, nine of the whales stopped singing. Of the nine, four stopped when they joined with another whale (a normal baseline behavior), so, there were five cessations of song potentially in response to the sonar (although whales stop singing without joining even under baseline conditions). The received levels measured next to the whales were 120-150 dB rms re 1 μPa, and there was no relationship between received level and the probability of cessation of singing. For six whales in which at least one complete song was recorded during the playback, the songs were an average of 29% longer. Miller et al. (2000) did not find a significant increase in song length with received playback level, probably because their study was less powerful than that of a larger sample analyzed in Fristrup et al. (2003). A similar CEE with responses of right whales to three 2-min stimuli, 60% duty cycles, and energy of 500-4,500 Hz showed no relation between probability or strength of response and received level, which was 133-148 dB rms (Nowacek et al., 2003), but this result is also limited by the small sample.
Both observational studies and CEEs demonstrate that behavioral context can have a substantial effect on relationships of acoustic dosage to behavioral response. For example, Tyack and Clark (1998) report that the avoidance reaction found when the SURTASS-LFA sound source was placed in the middle of the migration path apparently disappeared when the sound source was placed just offshore of the main migration path, even if the whales passed close to the source. On a larger scale, beluga whales in the Canadian high arctic show intense and prolonged reactions to the propulsion sounds of icebreakers (Cosens and Dueck, 1988; Finley et al., 1990), whereas beluga whales in Bristol Bay, Alaska, continued to feed when surrounded by fishing vessels and resisted dispersal even when pur-
posely harassed by motorboats (Fish and Vania, 1971). This context specificity of behavioral reactions to sound raises questions about the ecological validity of extrapolating data from captive animals to the wild.
The behavioral responses of marine mammals to acoustic stimuli vary widely, depending on the species, the context, the properties of the stimuli, and prior exposure of the animals (Wartzok et al., 2004). Species variation in auditory processing is so important that a distinction should certainly be made between taxonomic groups that have widely different hearing and sensitivity frequencies. For example, pinnipeds have lower maximal frequency of hearing and maximal sensitivity of hearing than odontocetes (toothed whales). They typically have a high-frequency cutoff in their underwater hearing between 30 and 60 kHz, and maximal sensitivity of about 60 dB re 1 μPa, and odontocetes have best frequency of hearing between 80 and 150 kHz and maximum sensitivity between 40-50 dB. Therefore, odontocetes can hear over a wider frequency range and have keener hearing than pinnipeds, so they could potentially be affected by a wider variety of sounds. Little is known about the frequency range of hearing and sensitivity of some marine mammal taxa, such as baleen whales, but several attempts have been made to divide marine mammals into functional categories on the basis of hearing (e.g., Ketten, 1994).
As mentioned above, some of the variation in responses between species or individuals may stem from differences in audition. Not only do different species have different hearing capabilities but there is considerable variation in hearing among conspecifics. One of the most predictable patterns in mammals involves age-related hearing loss, which particularly affects high frequencies and is more common in males than females (Willott et al., 2001).
Auditory processing is less likely than behavior to differ between captive and wild animals, and captive data on behavioral reactions closely linked to audition may be relevant to other settings. For example, Schlundt et al. (2000) noted disturbance reactions of captive bottlenose dolphins (Tursiops truncatus) and beluga whales during TTS experiments. The behavioral reactions involved avoidance of the source, refusal of participation in the test, aggressive threats, or attacks on the equipment. Finneran and Schlundt (2004) showed that the probability of those reactions increased with increasing received level from 160 to 200 dB rms re 1 μPa at 1m except for low-frequency (400-Hz) stimuli near the low-frequency boundary of auditory sensitivity. The kinds of reactions observed and how they scale with intense exposures near the level that provoked TTS suggest that the signals were perceived as annoyingly loud.
Some of the variation in responses to sound may stem from experience. There are several well-known mechanisms by which an animal modifies its responses to a sound stimulus, depending upon reinforcement correlated with exposure. The response of animals to an innocuous stimulus often wanes after repeated exposure—a process called habituation. The National Research Council (NRC, 1993) recommended studies on habituation of marine mammals to repeated human-made sounds. In one of few experimental studies of habituation in marine mammals, Cox et al. (2001) showed that porpoises tended to avoid at a distance of 208 m upon initial exposure to a 10-kHz pinger with a source level of 132 dB peak to peak re 1 μPa at 1m. This avoidance distance dropped by 50% within 4 days, and sightings within 125 m equaled control values within 10-11 days. The pingers are used on nets to prevent porpoises from becoming entangled in them, so evaluations of their effectiveness must take habituation into account.
Kastelein et al. (1997) report that a captive harbor porpoise (Phocoena phocoena) avoided exposure to high-frequency pingers with source levels of 103-117 dB rms re 1 μPa at 1m and received levels of 78-90 dB rms re 1 μPa. When exposed to a source with a level of 158 dB rms re 1 μPa at 1m, the porpoise swam as far away as possible in the enclosure and made shallow rapid dives. Those results combine with the results of Cox et al. (2001) to suggest that porpoises react to sound at much lower levels than the captive delphinids studied by Finneran and Schlundt (2004). However, the context of the captive studies was quite different: the dolphins and belugas studied by Finneran and Schlundt were being rewarded for submitting to exposure to intense sounds, whereas the porpoise was not being rewarded for remaining in the sound field.
If an animal in captivity or the wild is conditioned to associate a sound with a food reward, it may become more tolerant of the sound and may become sensitized and use the sound as a cue for foraging. Several large-scale studies have shown that the distribution of feeding baleen whales correlates with prey but not with loud sonar or industrial activities (Croll et al., 2001); but the studies were unable to test for potentially more subtle effects on feeding, such as reduced prey capture per unit effort and reduced time engaged in feeding.
Some of the strongest reactions of marine mammals to human-generated noise may occur when the sound happens to match their general template for predator sounds. The risk-benefit relationship is very different for predator defense and foraging. An animal may lose a meal if it fails to
recognize a foraging opportunity, but it may die if it fails to detect predators. Animals do not have the luxury of learning to detect predators through experience with them. Deecke et al. (2002) showed that harbor seals responded strongly to playbacks of the calls of mammal-eating killer whales and unfamiliar fish-eating killer whales but not to familiar calls of local fish-eating killer whales. That suggests that, like birds studied with visual models of predators (Schleidt, 1961a; 1961b), these animals inherit diffuse templates for predators. They initially respond to any stimulus similar to the predator template but learn through habituation to cease responding to harmless variants of the general predator image.
It would make sense for animals to show strong reactions to novel sounds that fit within the predator template, whatever the received level. Indeed, the behavioral reactions of belugas to ice breaker noise match the local Inuit description of their responses to killer whales, a dangerous predator. Some of those strong reactions to novel sounds, such as the responses of diving right whales to an artificial alarm stimulus as reported by Nowacek et al. (2003), might be expected to habituate if the stimuli are distinguishable from real predators and are not associated with aversive effects. In fact, the only right whale subject not to respond was the last of six whales tested, and it may have heard the stimulus up to five times before. Beluga whales that fled icebreaker noise at received levels of 94-105 dB rms re 1 μPa returned in 1-2 days to the area where received icebreaker noise was 120 dB rms re 1 μPa (Finley et al., 1990). In contrast, Kastak and Schusterman (1996) reported that a captive elephant seal not only did not habituate but was sensitized to a broadband pulsed stimulus somewhat similar to killer whale echolocation clicks even though nothing dangerous or aversive was associated with the noise.
The low sound levels that stimulate intense responses of Arctic beluga whales (Frost et al., 1984; LGL and Greeneridge, 1986; Cosens and Dueck, 1988) contrast sharply with the high levels required to evoke responses in captive beluga whales (Finneran and Schlundt, 2004). This difference highlights that there are likely to be several kinds of response, depending on whether the animal is captive and whether the noise resembles that of a known predator. Annoyance responses may require levels of sound well above levels that may stimulate strong antipredator responses. If animals in the wild hear a sound that matches their auditory template for a predator, they may avoid exposures to sound levels much lower than those required to elicit the disturbance responses observed by Finneran and Schlundt (2004). If learning can modify the predator template, as suggested by
Deecke et al. (2002), it is essential to conduct studies of behavioral responses of animals to human-made stimuli in habitats resembling those encountered by wild populations.
An important property of most anthropogenic sound is that high-intensity levels are typically confined to the immediate location of the sound source (an exception is high-intensity, low-frequency sound), so any effects caused by exposure to high levels are reduced as animals move away from the source. However, high-intensity low-frequency sound travels well enough underwater that animals can detect signals at ranges of tens to hundreds of kilometers from the source. If, as in the case of Arctic belugas hearing icebreaker noise, exposure to low received levels can still trigger an intense response, a few sources may affect a large fraction of a population.
Even in the absence of a strong response, low received levels of sound can affect a large fraction of a population if the sound results in a masking of normal stimuli. Marine mammals show exquisite adaptations to overcome masking, but they may not be effective in the presence of pervasive anthropogenic sounds (reviewed in NRC, 2003b; Wartzok et al., 2004).
PHYSIOLOGICAL RESPONSES TO ACOUSTIC STIMULI
Most discussions of physiological effects of noise have centered on the auditory system. Audition has evolved for sensitivity to sound, so it is likely to be the physiological system most sensitive to acoustic stimuli that are within the frequency range of hearing. When the mammalian auditory system is exposed to a high level of sounds for a specific duration, the hair cells in the cochlea begin to fatigue and do not immediately return to their normal shape. When the hair cells fatigue in that way, the animal’s hearing becomes less sensitive. If the exposure is below some critical energy flux density limit, the hair cells will eventually return to their normal shape; the hearing loss will be temporary, and the effect is termed a temporary threshold shift in hearing sensitivity, or TTS. If the sound exposure exceeds a higher limit, the hair cells in the cochlea become permanently damaged and will eventually die; the hearing loss will be permanent, and the effect is termed a permanent threshold shift in sensitivity, or PTS. TTS and PTS limits vary among individuals in a population, so they need to be characterized statistically. A relationship between the TTS limit and the PTS limit has been
determined for laboratory animals; the appropriateness of extrapolating of such a relationship to marine mammals is untested.
A major recommendation of the National Research Council 1994 report supported the development of TTS studies in marine mammals. Since then, TTS experiments have been conducted in two species of odontocetes (Tursiops truncatus and Delphinapterus leucas) with both behavioral and electrophysiological techniques (Finneran et al., 2000; Schlundt et al., 2000; Nachtigall et al., 2003, 2004) and three species of pinnipeds (Phoca vitulina, Zalophus californianus, and Mirounga augustirostris) with behavioral techniques (Kastak et al., 1999; Finneran et al., 2002). Those experiments were conducted at three centers for research on marine mammals that have facilities to hold their own animals: the Hawaii Institute of Marine Biology of the University of Hawaii, Long Marine Laboratory of the University of California, Santa Cruz, and the Space and Naval Warfare Systems Center (SPAWAR) of the US Navy in San Diego. The scientists at the Hawaii Institute of Marine Biology used continuous random noise with a bandwidth slightly greater than 1 octave as the fatiguing stimulus and both behavioral and electrophysiological techniques to measure TTS in the bottlenose dolphin. The fatiguing stimulus had a broadband received level of 179 dB rms re 1 μPa, which was about 99 dB above the animal’s pure-tone threshold of 80 dB at the test-tone frequency of 7.5 kHz (Nachtigall et al., 2003). Exposure to 50 min of the fatiguing stimulus resulted in a TTS of 2-18 dB. Recovery from the TTS occurred within 20 minutes after the cessation of the fatiguing stimulus. More recent studies (Nachtigall et al., 2004) that used an auditory brainstem response showed a TTS of 5-8 dB in response to 30 minutes of a 160-dB rms re 1 μPa fatiguing stimulus. Although the intensity of the fatiguing stimulus fell rapidly above 11 kHz, the greatest TTS was shown at 16 kHz. This pattern of TTS being more prominent at a frequency above the frequency of the fatiguing stimulus matches results for humans (Ward, 1963). The recovery occurred at 1.5 dB per doubling time with complete recovery within 45 min. The 1.5 dB recovery per doubling time was also found for recovery from the more intense 179 dB fatiguing stimulus used in the earlier study (Nachtigall et al., 2003). Researchers at Long Marine Laboratory used continuous random noise of 1-octave bandwidth as the fatiguing stimulus and a behavioral technique to measure TTS in the harbor seal (Phoca vitulina), California sea lion (Zalophus californianus), and elephant seal (Mirounga augustirostris). They exposed the subjects to 20-22 min of the fatiguing stimulus and found that it only had to be 60-75 dB above the
hearing threshold to induce a TTS of 4-5 dB for test signals at frequencies between 100 Hz and 2 kHz. Threshold measurements conducted 24 hours after the cessation of the fatiguing stimulus indicated complete recovery from the TTS (Kastak et al., 1999). Researchers at SPAWAR used impulse sounds from a seismic watergun as the fatiguing stimulus and a behavioral technique to measure the TTS (Finneran et al., 2002). The fatiguing stimulus had a variable duration of about 1 ms, peak pressure of 160 kPa, a sound pressure of 226 dB peak-to-peak re 1 μPa at 1m, and an energy flux density of 186 dB re 1 μPa2s, which produced a TTS of 7 and 6 dB at 0.4 and 30 kHz respectively in beluga whales but not at the other tested frequency of 4 kHz. In dolphins, no TTS could be demonstrated at 0.4, 4 or 30 kHz in spite of raising the fatiguing stimulus to its maximum intensity of 228 dB (Finneran et al., 2002). Each of these experiments used different durations of fatiguing stimuli. When the sound pressure required to produce a TTS is plotted against the duration of the stimulus for all these experiments, the result is a line with a slope of -3 dB per doubling of stimulus duration, that is, a line showing that the TTS occurred at about an equal energy in all cases tested to date.
Changes in hearing threshold, even TTSs, have the potential to affect population vital rates through increased predation or decreased foraging sources of individual animals that experience a TTS as they use sound for these tasks. A TTS also has the potential to decrease the range over which socially significant communication takes place, for example, between competing males, between males and females during mating season, and between mothers and offspring. Unless a critical opportunity is available only during a narrow time window, the potential effects on vital rates are important only if exposures and any resulting TTS are prolonged. In spite of the importance of sound for marine mammals, there is considerable variability in hearing sensitivity within a species, and there is evidence of age-related hearing loss.
Nonauditory Effects of Sound
A marine mammal has many airspaces and gas-filled tissues that could theoretically be driven into resonance by impinging acoustic energy. The lungs, air-filled sinuses that include those of the middle ears, and in the intestines, where there can be small gas bubbles, are among the areas that
may be susceptible to resonance induced by acoustic sources. However, there were no published measurements of resonance in a marine mammal until the work of Finneran (2003), who measured the resonance of the lungs of a bottlenose dolphin and a beluga whale. Before Finneran’s work, most studies of acoustic damage in marine mammals concentrated on the effects of shock waves, including blast-related phenomena.
Finneran (2003) used a backscatter technique to measure the resonance of the lungs of a 280-kg bottlenose dolphin and a 540-kg beluga whale. He obtained resonance frequencies of 30 Hz for the larger white whale and 36 Hz for the bottlenose dolphin. However, the resonance was highly damped and far less intense than predicted by a free-standing bubble model. The lungs experience a symmetric expansion and contraction when ensonified. How intensely a structure resonates at its resonant frequency can be quantified, and is represented by Q. The higher the Q, the more resonant the structure. The Q values measured in marine mammals are low. The Q of the lungs of the beluga whale was found to be 2.5, and of the bottlenose dolphin 3.1. Those Q values suggest a broad resonance property that is highly damped. Apparently, the tissue and other mass surrounding the lungs dampen the susceptibility of the lungs and probably other structures to resonate intensely.
Although other gas-filled structures will resonate at different frequencies, the probable low Q values, as in Finneran’s study, suggest that resonance of air spaces is not likely to lead to detrimental physiological effects on marine mammals. That was also the conclusion of a panel of experts convened by National Oceanic and Atmospheric Administration (NOAA) Fisheries (NOAA, 2002).
Rectified diffusion is a physical phenomenon that leads to the growth of microscopic bubble nuclei in the presence of high-intensity sound. It has been demonstrated only in laboratory preparations, but it is theoretically possible that exposure to high-intensity sound could enhance bubble growth in humans and marine mammals (Crum and Mao, 1996). Rectified diffusion might be a possible mechanism of nonauditory acoustic trauma in human divers and marine mammals, in that bubbles in tissue or blood can lead to injury or death. Calculations by Crum and Mao (1996) suggest that, given a modest degree of nitrogen (N2) supersaturation of biological tissues (for example, between 100% and 200%), the growth of
normally stabilized nuclei would be unlikely to occur at sound pressures below 190 dB rms re 1 μPa. However, at sound pressures above 210 dB, significant bubble growth could occur. As nitrogen supersaturation increases, the exposure threshold of activation should decrease, and the growth rate of bubbles should increase.
Houser et al. (2001) modeled the accumulation of N2 in the muscle of diving cetaceans on the basis of dive profiles of deep-diving odontocetes and data on N2 accumulation previously measured in the muscle of diving bottlenose dolphins (Ridgway and Howard, 1979). The model necessarily assumed that N2 kinetics were the same between species and that lung collapse occurred at 70 m—a prediction made by Ridgway and Howard for bottlenose dolphins. The conclusions of the model were that slow deep-diving cetaceans (diving beyond the depth of lung collapse), which had few extended surface intervals, would accumulate the greatest amount of N2 in their tissues while diving. The slower the dive in water shallower than lung collapse, the longer the time the animal experiences pressure that drives the accumulation of gas in the tissues; short surface intervals between deep dives would limit the time the animal has to clear accumulated N2 from its body.
The magnitude of tissue N2 supersaturation—and thus the possibility of rectified diffusion—depends on dive behavior as described above. Records of dive behavior of beaked whales—Cuvier’s beaked whale (Ziphius cavirostris) and Blainville’s beaked whale (Mesoplodon densirostris)—presented at a recent workshop (Marine Mammal Commission, 2004) indicate that these animals have long deep dives followed by a short surfacing and than a series of shallow dives primarily within the region in which gas exchange occurs in the lung. The short surfacings and the repeated “bounce” dives near the surface could lead to high tissue N2 pressure and the possibility of bubble formation. Those are the predominant species of beaked whales that have stranded in association with naval sonar activity, although other beaked whale species have also been involved.
Evidence of deleterious bubble formation in diving cetaceans and the putative causative mechanisms (acoustically and behaviorally mediated) remain open to debate. Jepson et al. (2003) conducted necropsies of stranded cetaceans and reported on signs of bubble-related injury, but their interpretation has been challenged (Piantadosi and Thalmann, 2004). No experimental evidence has been collected on the feasibility of the putative mechanisms of bubble formation in breath-hold divers. More research is needed to understand the role of rectified diffusion in marine mammals, but our current understanding suggests that it would be relevant only for animals
exposed to sound substantially above 180 dB re 1 μPa, which is already considered by regulators to be a threshold for risk of other forms of injury.
PROGRESS ON EARLIER NATIONAL RESEARCH COUNCIL RECOMMENDATIONS
Three previous National Research Council reports recommended research to resolve critical uncertainties about the effects of noise on marine mammals (1994, 2000, 2003b). All three highlight the need for research in behavioral ecology, auditory physiology and anatomy, nonauditory effects of sound, effects of sound on prey of marine mammals, and development of new techniques. The 2003 report also recommended research on sources and modeling of ocean noise. Some of the recommendations have led to research that has greatly reduced the data gap. For example, the 1994 and 2000 reports recommended experiments to determine acoustic exposures that would lead to temporary shifts in the threshold of hearing of marine mammals. In the last decade, several laboratories have succeeded in conducting the experiments; as a result, the uncertainty involved in modeling the noise exposures that start to cause physiological effects on hearing has been reduced.
Progress has also been made on the recommendation with respect to development of new technology. For example, the 1994 report recommended the development of tags to record physiology, behavior, location, and sound exposure. In the last decade, tags have been developed to record all but physiological characteristics (Johnson and Tyack, 2003).
For many of the other research recommendations, research is being conducted, but progress has been slow enough over the last decade to argue for the establishment of a targeted research program. The 2000 and 2003 reports recommended better coordination between federal regulatory agencies and science-funding agencies to develop a multidisciplinary research program. It was recommended that the research program operate like that of the National Science Foundation and the Office of Naval Research, issuing targeted requests for proposals and judging the quality of proposals with peer review. Although some progress has been made, it is worth reiterating that progress on critical research requires that the federal government develop and fund a dedicated research program.