Evaluating and Assessing the Relative Effectiveness of Acoustic Deterrent Devices and other Non-Lethal Measures on Marine Mammals

4 Acoustic Devices to Reduce Risks Associated with Pile-Driving

4.1 Introduction

Some anthropogenic maritime activities bring with them acute and short term risks of damage to marine mammals. Two examples in Scottish waters are the use of explosives and pile driving. Explosives have been widely used in Scottish waters, for example during oil field decommissioning, to sever well heads. The use of explosives is believed to be declining as alternative mechanical cutting technologies are introduced. By contrast, the extent of offshore pile driving has grown enormously over the last decade or so because it is the favoured method of sinking the monopile foundations for offshore wind farms. This trend is set to accelerate; not only are more piles being driven but the diameter of piles and with it the level of acoustic energy output during piling has also increased substantially. Explosives certainly have the potential to cause severe injury or death to marine life including marine mammals. While pile driving is unlikely to cause mortality or tissue damage it is generally believed that pile driving could result in permanent hearing damage for marine mammals at substantial ranges. The traditional approach to mitigation for these activities is to determine an exclusion zone within which animals are thought to be at risk and then to search in this area using visual monitoring (sometimes supplemented by passive acoustic monitoring methods), to determine that no animals are present before the activity (explosive detonation or pile driving) can take place. It is well known that neither visual nor acoustic monitoring is entirely effective at detecting animals, especially small or shy marine mammals such as seals or harbour porpoises, and visual monitoring will be particularly ineffective during poor weather conditions and at night. Providing mitigation monitoring offshore can be a very expensive undertaking often involving a team of several qualified marine mammal observers ( MMOs), specialist acoustic monitoring equipment and a dedicated vessel.

One approach which has the potential to mitigate risk by directly reducing the noise created by pile-driving is the bubble curtain. This technique has been tested in a number of projects ( e.g. Lucke et al., 2011; Matuschek and Betke, 2009; Reyff, 2003; Würsig et al., 2000) and has been shown to reduce broadband noise emissions by approximately 5 - 10 dB re 1 µPa.

Another approach to mitigating such acute short-term risks, which could be used as an alternative to, or enhancement of, monitoring, is to use an aversive sound to move animals to a "safe distance" before pile driving or explosions take place. Ideally, a signal of this type would reliably elicit the desired behavioural response without adding significantly to the subjects' acoustic dose. Aversive sound mitigation has the potential to offer a greater degree of risk reduction and to be very much more cost-effective than traditional monitoring mitigation. The methodology has still to be developed, however, and if it is to offer reliable and quantifiable risk reduction the performance of any system needs to be carefully measured (as, for example, by Brandt et al., 2012b). It is also important that any induced behavioural responses should not be so severe as to lead to adverse consequences such as mothers and calves losing contact with each other, nor animals being caused to strand, nor that widespread and intensive use should clear animals out of large parts of their foraging habitat. Recognising the potential of this approach, the Collaborative Offshore Wind Research in the Environment ( COWRIE) commissioned a review to explore the potential of the approach (Gordon et al., 2007). Below we briefly summarise the main findings of that review and update it with some recent research.

4.2 Risks and Required Mitigation Performance

The first requirement in assessing a mitigation system is to understand the circumstances and ranges in which different types of marine mammals might be at risk, because it is this information that determines the specification and the performance that any mitigation procedure needs to achieve. UK and European regulators have not put forward any clear criteria for unacceptable physical effects or for the exposure thresholds at which these might take place. US regulators, however, have been more proactive and have provided a series of thresholds for unsafe exposure. In their review, Gordon et al. (2007) proposed thresholds for exposure based on values available from US regulators in 2007, and also on their own interpretation of marine mammal research on Temporary Threshold Shifts ( TTS). A very useful initiative from the US has been completed and published since the report by Gordon et al. (2007), namely a series of workshops of North American experts which critically reviewed available information, explained a procedure for deriving thresholds for exposure for different classes of marine mammals and proposed thresholds for exposure based on the best available evidence. This process is described in an important peer-reviewed publication (Southall et al., 2007). This work is relevant to many different parts of the current review and is discussed at greater length in section 7.2 of this report.

Southall et al. (2007) put forward two different types of criteria for hearing damage. One was based on sound level, the peak pressure level (in dB re 1 µPa) of a sound, however short, that should be considered unacceptably likely to cause hearing damage. The second was based on received acoustic energy or sound exposure level ( SEL) in units of dB re 1 µPa² s ^-1. This is a measure of the acoustic dose received over a period of time (nominally 24 hours).

The received pressure level for an animal at a given range from a source can be estimated by applying a propagation loss to a source level. Propagation loss is typically a function of range but the precise nature of the relationship with range varies with environmental factors. Acoustic propagation in the marine environment is a topic of very general relevance. It is of great interest to the military for example and it has consequently been an area of extensive research. Models of propagation loss perform well and are able to make reliable predictions if the environmental parameters (such as water depth, bottom types, sound speed profile etc.) are accurately known. In addition, sound propagation can be readily measured in the field. Thus, ranges at which a certain peak pressure will be expected to occur, required for the Southall et al.'s peak pressure criteria, are relatively easy to predict. Sound exposure level for a mobile receiver is more difficult to calculate however. Source level and propagation loss are of course important for defining a sound field around the noise source, but the length of time that the source is active and, critically, the way in which the receiver moves within the sound field are additional and highly influential parameters. Animal movements during pile driving exercises have not been measured. It is assumed that sensitive animals move away from the sound source. Observations of reduced densities over extensive ranges after pile driving ( e.g. Brandt et al., 2012a) certainly indicate movement away from pile driving activity, but they don't provide information on the speed or directivity of movements.

Gordon et al. (2007) used a simple movement model (an animal moving directly away from a sound source at plausible swimming speeds) and values of source level and propagation loss from published field observations of pile-driving operations, to determine the starting range at which various thresholds for exposure might be exceeded and thus, the range at which mitigation systems would need to be effective. For this report we have repeated these calculations using the threshold values proposed by Southall et al. (2007) and also new thresholds that we have derived by applying the Southall methodology to new research findings. The most important of these newly reported research findings in this context are contained in a paper by Klaus Lucke and colleagues (Lucke et al., 2009). This reports on a series of trials which measured TTS caused by exposing a harbour porpoise to airgun pulses. An airgun was used in these experiments as a surrogate for pile driving: Airguns produce intense low frequency signals with similar acoustic characteristics to pile driving pulses but they are relatively easy to operate close to captive animals. Lucke et al., (2009) showed that TTS was induced in harbour porpoises at an unexpectedly low sound exposure level of 164 dB re 1 µPa² s ^-1. Applying Southall et al.'s logic and methodology (which assumed that the injury threshold was 15 dB above the threshold for TTS) provides a predicted SEL injury threshold for porpoises of 179 dB re 1 µPa² s ^-1. This was much lower (some 19 dB less) than the levels suggested by Southall et al. (2007) for high frequency cetaceans, and which had been based on experiments with bottlenose dolphins and beluga whales. Lucke et al.'s findings were also unexpected because, airgun pulses, like pile driving noise, are dominated by low frequency sound, which is far below the frequencies at which porpoises are most sensitive and also the frequencies (4 kHz) at which TTS was measured by Lucke et al (2009). Although these trials were limited to one animal, they have the very significant advantage of having tested the right type of sound (low frequency pulses) on the cetacean species most likely to be affected by pile driving in Scottish waters (the harbour porpoise).

Figures 8 - 11 summarise our modelling results. They show plots of the maximum "starting range" at which cumulative SEL thresholds for PTS would be exceeded for animals that moved directly away from the piling noise in a straight line for the duration of piling. The plots are for a 4.7 and 6.5 m diameter pile (less than the maximum likely for some Scottish sites). Values for a range of likely "escape speeds" and propagation loss relationships are shown. It is clear that the "mitigation range" to which animals should be moved before piling starts, varies greatly with propagation conditions, speed of movement and source level. In addition, mitigation range varies between species groups, and those with better low frequency hearing such as seals having the greatest ranges.

KG No.	Knowledge Gap
40	Empirical measures of displacement movement rates are required in order to improve the TTS risk modelling approach. Appropriate movement models are the limiting factor in predicting risk of TTS.

Figure 8 High Frequency Cetacean, 6.5m Pile, Southall et al., SEL 198 dB ( MHFC)

Figure 9 Seals 6.5m Pile, Southall et al., SEL 186 dB (Mpin)

Figure 10 Porpoise 4.7 m Pile, extrapolation from Lucke et al., SEL

Figure 11 Porpoise 6.5 m Pile, extrapolation from Lucke et al., SEL

Figures 8 - 11 show plots of maximum "starting range" at which thresholds for PTS would be exceeded. In each case cumulative exposure was modelled for an animal as a set "starting range" at the start of piling. The animal moved directly away from the piling at a fixed speed throughout the piling episode. The sound field around the pile was determined by a simple propagation loss model PL = TLF Log(range). Plots are shown for a range of transmission loss factors ( TLFs) from 10 dB (cylindrical spreading), through 20 dB (cylindrical spreading) up to 22 dB. Scenarios were modelled for a 6.5 m diameter pile with 3000 impacts at full power and a 600 impact "soft start" source level energy flux density of 226 dB 1 µPa ² s ^-1 for seals and for high frequency cetaceans. See Gordon et al. (2007) for additional details.

This is in line with some other similar modelling exercises. For example, a recent environmental impact assessment for a wind farm construction (SmartWind, 2012) suggested that for an 8.5 m diameter pile, a typical range at which the Southall et al (2007). SEL threshold for Permanent Threshold Shifts ( PTS) would be exceeded for a fleeing porpoise would be 850 m, while for seals the range for exceeding the SEL PTS threshold from a single hammer strike would be 220 m. In this case presumably the range for an entire pile driving exercise would be very much greater.

It is clear there that there are large uncertainties and that mitigation ranges will vary between species and with conditions. However, these modelling exercises do at least indicate that there is a real risk of hearing damage from pile driving and also serve to roughly quantify the performance that an aversive mitigation system needs to achieve: it must be capable of reliably moving animals to ranges of many hundreds or even thousands of meters from a piling location before piling begins and ideally it should do so while adding a minimum additional acoustic dose.

4.3 Candidate Aversive Sound Types

Two different classes of sound might be considered as likely candidates for mitigating pile-driving interactions; those with biological significance (either innate or learned), such as alarm calls or the vocalisations of predators, and those without biological significance, or "inherently" aversive sounds.

In humans, sounds with certain characteristics seem to be inherently unpleasant. Zwicker and Fastl (2004) reviewed psycho-acoustic work in this area and found that measures of annoyance were best explained by a sound's loudness, fluctuation strength and "sharpness" (sharpness describes a sound having a narrow frequency emphasis within a critical frequency band - higher frequency sounds exhibit this quality more strongly). Other general properties of sounds that are "inherently" aversive to humans include unpredictability (such as randomly modulating amplitude) and dissonance. Dissonant sounds are composed of tones which are not simple ratios of each other. By contrast, humans prefer combinations of tones varying by simple ratios, such as whole octaves: called consonant sounds. Attempts to find a similar preference for consonant sounds in another primate, the cotton-top tamarin ( Saguinus oedipus), were not successful (McDermott and Hauser, 2004). Thus, findings from human research may be difficult to transfer to other species and these sounds are unlikely to be sufficiently aversive to induce movements over the ranges required.

KG No.	Knowledge Gap
41	It is unclear whether auditory preference/aversion is transferable between species.

The startle reflex is a widely exhibited response elicited by loud sounds that have a sufficiently rapid onset. This response is likely to be exhibited by all mammals and is experienced as being unpleasant. Stimuli that induce the startle response are therefore avoided after multiple exposures. Götz (2008) and Gotz and Janik (2010) have explored the use of the startle response sounds as the basis for more effective acoustic deterrent devices. However, elicitation of the startle response has not been shown to induce movements over the sort of ranges required for this application. Further, to be effective in eliciting a startle reflex, a sound must be received at a high level and with a rapid onset time. Both of these characteristics will be reduced with range from a sound source as received levels fall and reverberation and absorption reduce the signal rise time. Thus it seems unlikely that it will be practical to induce a startle reflex at substantial ranges except with very powerful noise sources, which might in themselves be considered likely to cause hearing damage.

Absorption and reverberation will affect any and all signal types, such that beyond a given range they may lose their aversive sound characteristics. The range at which aversive characteristics are lost will depend on the complexity of the signal as well as the source level. In this respect, it is the simplest signal types, such as the pure tones produced by the Lofitech device, which will be least affected by increased range.

KG No.	Knowledge Gap
42	Effects of absorption and reverberation on different signal types have not been shown, and could be prohibitive to long-range effectiveness of complex signals.

There are many examples of aversive sounds being used to frighten terrestrial animals away from agriculture. Gordon et al. (2007) reviewed several case studies from which a number of useful insights can be taken. Generally, animals habituate quickly to sounds which are repeatedly presented and are not reinforced (with a negative association). Thus, signals should be broadcast for the minimum length of time and if possible should be associated with reinforcement. Typically, sounds that had biological significance, such as alarm calls or predator vocalisations, proved more effective in keeping animals away from crops and were less readily habituated to. A typical course of events was for manufacturers to quickly bring acoustic devices to market, based on initially encouraging results, only to find their effectiveness waned as habituation occurred. Thus, it is important that devices are extensively tested before they are widely introduced.

4.4 Examples of Marine Mammals Moving in Responses to Aversive Sounds

4.4.1 Predator Vocalisations

The killer whale is a predator of most marine mammals, and a number of studies have reported marine mammal responses to broadcast of killer whale calls. For example, Fish and Vania (1971) played killer whale vocalisations to reduce beluga predation of salmon smolts. Beluga showed a strong avoidance. They also avoided playback of 2.5 kHz pulsed tones, however, suggesting a more general avoidance of sounds with acoustics characteristics similar to those of killer whale calls, or simple aversion of unfamiliar sound-types (neophobia). Gray whales ( Eschrichtius robustus) have also been observed to react to playback of killer whale calls (Cummings and Thompson, 1971; Dahlheim, 1987).

Grey and harbour seals in the Baltic showed strong responses to playback of killer whale calls by either swimming to a resting site and hauling out or moving away to a range of ~1km from the playback (Anon., 2002b; De La Croix, 2010). Similar apparently adaptive anti-predator responses have been observed after opportunistic playback of killer whale calls to grey seals in open water (David Thompson, pers. comm.). Deecke et al. (2002) reported a more nuanced response. They broadcast calls of both fish-eating and mammal-eating killer whale pods to seals swimming in the water at haul out sites and measured a smaller behavioural response to the call types from fish eating than from mammal eating killer whales. This suggests that in areas where killer whale pods that do not feed on marine mammals are present it will be necessary to avoid using local call types as aversive mitigation signals.

4.4.2 Sonar

There are a number of reports of strong responses to different types of mid-frequency sonar. After the Second World War whalers hunting baleen whales started to experiment with adapted military sonar units. They soon found that these military units were of marginal value for detecting and localising whales but they scared baleen whales so consistently that the animals would flee on the surface, making their movements predictable and rendering them easier to catch. Special versions of the sonar called "whale starters" were later developed to maximise these aversive effects (Mitchell et al., 1981).

Over the last decade or so it has become clear that mid-frequency military sonar can cause mass stranding and mortality in cetaceans. Beaked whales seem to be the most vulnerable group (Cox et al., 2006). It is now widely believed that the mechanism behind these stranding events involves dramatic behavioural responses of marine mammals to sonar signals received at relatively low levels (Tyack et al., 2011), which, in the case of beaked whales at least, may lead to decompression sickness and mortality (Fernandez et al., 2005). Field observations of responses to sonar trials, analysis of stranding events and some extensive on-going behavioural response studies, show that a range of cetacean species including beaked whales (Tyack et al., 2011), sperm whales (Watkins et al., 1985), minke whales ( Balaenoptera acutorostrata) ( NOAA/ DON, 2001), killer whales (Fromm, 2006; NMFS, 2005), melon-headed whales ( Peponocephala electra) (Southall et al., 2006), pilot whales (Rendell and Gordon, 1999) and a variety of dolphins ( NOAA/ DON, 2001) all respond strongly to mid-frequency sonar signals. Some have suggested that one of the reasons for this strong avoidance might lie in the acoustic similarities between mid-frequency military sonar and some killer whale vocalisations.

4.4.3 Airguns

Airguns are devices used during seismic surveys to produce very powerful, predominantly low frequency, sound pulses. Concerns about the damaging effects that airguns could have on marine mammals has led to a large amount of research measuring behavioural response ( e.g. see review by Gordon et al., 2004). Most marine mammals do show avoidance responses, though there are few cases where dramatic behavioural change or large-scale exclusion has been measured. With most of their energy below 200 Hz, airgun pulses lie outside the range of most sensitive hearing for most odontocetes. One of the clearest examples of a short-term behavioural avoidance of a small airgun comes from a series of controlled exposure experiments which were carried out on seals in Scottish, Norwegian and Swedish waters (Thompson et al., 1998). A single airgun or small array was activated for an hour at ranges of between 1.5 and 2.5 km and the movements and behaviour of seals was monitored using VHF and acoustic telemetry. Strong avoidance behaviour was exhibited during six out of eight trials with harbour seals, while one animal showed no response. Clear avoidance was also shown by grey seals. Some animals close to land hauled out, while others moved away. As part of the same project, however, Gordon et al. (1998) investigated responses of harbour porpoises to both the same small airguns and to full scale commercial arrays by comparing detection rates on towed hydrophones. They were not able to show any statistically significant effects.

Airguns require compressed air to operate so there may be some practical and safety issues with deploying them at sea. Airguns produce powerful acoustic pulses which may contribute significantly to sound exposure, potentially leading to TTS. For example, Lucke et al. (2009) used a small airgun to induce TTS in porpoises. The apparently high potential for inducing hearing damage, combined with the lack of clear aversive behavioural responses shown by some species and practical difficulties of deployment in the field, argue against the use of airguns as a mitigation tool.

4.4.4 Acoustic Deterrent Devices

Examples of harbour porpoise and killer whales avoiding acoustic deterrent devices designed for use at aquaculture sites are reviewed in detail elsewhere in this report ( section 7.3). In many cases cetaceans have been shown to be deterred at ranges that would be useful for pile driving mitigation ( e.g. Olesiuk et al., 1995). Based on these results the use of ADDs has been recommended under the Joint Nature Conservation Committee protocol for mitigation of pile driving ( JNCC, 2010). Several recent studies have explored the use of ADDs in this context with promising results.

4.4.4.1 Bioconsult Trials with Wild Harbour Porpoises

One of the most recent and complete set of trials of acoustic deterrents for mitigation to have been conducted since the Gordon et al. (2007) report was carried out by the German environmental consultancy, Bioconsult, with funding from the German Federal Ministry of the Environment, Nature Conservation and Nuclear Safety and the Danish Offshore Demonstration Program for Large-scale wind farms. This study was designed to specifically investigate the use of an ADD (a Lofitech seal scarer) for pile driving mitigation with harbour porpoises. Research was carried out at two contrasting field sites. One was an inshore site in Danish Baltic waters where observations of porpoise locations and movements could be made from cliff-based shore station and pods could be used to monitor porpoise presence (Brandt et al., 2013). The other site was offshore in the German North Sea. Here, an array of PODs was used to monitor porpoise presence and aerial surveys were also conducted. Results from both are presented in a project report (Brandt et al., 2012b) and the offshore work has also been published in the primary literature (Brandt et al., 2012c).

At the offshore site, PODs were deployed along three "transect" lines extending from the central location where the seal scarer was positioned during trials. Each line was oriented at approximately 120 degrees to the others with five PODs located along each transect at ranges of 750 m, 1500 m, 3000 m, 5000 m and 7500 m from the central ADD location. A single pod was located in the centre close to the ADD location. Thus, in total there were 16 PODs with three PODs at each of five ranges and one at zero range. Ten ADD broadcast trials were completed. During each trial a boat motored out to the central location, turned off its engine, and then activated the ADD for 4 hours. POD data for 19 hours before and 18 hours after the start of each trial were analysed, grouped into three hour blocks around the start time of each trial. The first hour of each trial was not analysed because it was feared it might be confounded by the influence of the boat arriving. The proportion of porpoise positive minutes in the different three hour blocks were compared for each location. Porpoise detection rates were lower when the ADD was active at all ranges from zero to 7500 m. However, this effect was not significant at two of the five ranges (1500 m and 5000 m). These were two ranges at which porpoise detections were always low and thus would have had lower power to show a change. At zero range, porpoise were almost completely absent when the ADD was active and detection rates were 86% lower at 750 m and 96% lower at 7500 m. Effects seemed to be greatest at locations that had the highest detection rates during control periods. There was no clear evidence of a reduction in the exclusion effect with range and therefore no indication that 7500 m was the maximum range at which reduced densities should be expected. Detection rates "recovered" after the cessation of the trials and in the three hour analysis block between 9 and 12 hours after the trial they were no longer significantly different from pre-trial controls. Brandt et al. (2012b) also measured sound levels at different ranges in this environment and calculated propagation loss. These measurements showed a good fit to the "Theile approximation" for propagation loss in this type of environment with a 194 dB re 1µPa ( RMS) source level, which gives a received level of ~142 dB re 1µPa at 750m and of 115 dB re 1µPa at 7500 m.

On one occasion, visual aerial surveys were carried out immediately before and during ADD activity allowing porpoise densities at the site to be compared. Aerial surveys covered an approximately 30km x 30km square survey block centred on the ADD location and followed a standard line transect protocol. Overall density was significantly reduced, by some 80%, during ADD playback. Before the ADD was active nine porpoises were sighted within 7500 m of the ADD location, while during ADD broadcast one porpoise was sighted in this area (at a range of 6300 m from the ADD). Although the aerial survey data were less extensive than the static acoustic monitoring data, they were very useful in showing that the reduced acoustic detection rates are probably due to animals leaving the area rather than simply changing their vocal behaviour.

One short-coming of these observations from the perspective of assessing ADD use in mitigation is that they measured responses integrated over a 3 - 6 hour period. For practical reasons, and to minimise disruption and the risk of habituation, one would hope to be able to achieve the desired effect of moving animals out of the mitigation zone over as short period as possible. Thus, it is animal movements and density changes in the first hour or so of operation, data which were discounted from this analysis, that are of prime relevance.

Brandt et al.'s second study (2013) was conducted at Fyns Hoved, a site in Danish Baltic waters where an elevated vantage point (a 20m cliff) overlooks sheltered waters with a high porpoise density. Visual observations were made from a shore base to a range of approximately 1 km. Porpoise sightings were recorded during standardised 10 minute scans of the area (out to a range of approximately 1000 m) and a theodolite was used to fix the location of porpoises observed at the surface at ranges out to 800 m. A mooring for a broadcast vessel was located below the shore station and 200 m from the shore and a small boat with an ADD was moored here during trials. Four PODs were also deployed: one at the broadcast mooring, and three at a range of 450 m from the broadcast mooring (Brandt et al., 2012b). Another one was positioned directly offshore from the mooring and the other two either side of it on a line parallel to the shore. On each trial day baseline observations were made for about an hour after which the shore team would find a group of porpoises within range and track them for long enough to obtain at least five locations before deciding that the trial should start. The skipper would then either activate the ADD for four hours or not activate it. The decision as to whether to activate the ADD was based on the flip of a coin and the shore observers were blind as to whether the ADD was active or not. These trials were supplemented by four days of "response observations" during which the responses of animals to the ADD deployed at ranges beyond 1000 m were observed.

Three types of data were examined for effects of ADD activity: visual sighting rates, acoustic detection rates from the PODs and tracks of animals from theodolite fixes. Sighting rates over the whole observation area, out to 1000 m and analysed in 4 hour blocks, fell from 31 sightings per 4 hours when the ADD was not active to 0.3 sighting per 4 hours when it was active (1.2% of the control value). Both of the sightings during ADD activity occurred at the edge of detection range and occurred after 85 and 21 minutes of ADD activity. To investigate effects on sighting rate statistically, Brandt et al. (2012b) constructed a model to predict sighting rate; the hour of the day and sea state were included as significant predictors. When data from the time when ADDs were active were included in the model it became the factor explaining most of the variation. Acoustic detection rates on PODs also fell dramatically and significantly, decreasing from a median of 0.83 porpoise positive minutes in the four hours before ADD activity to zero during the 4 hours that the ADD was active. On average there was a gap of 131 minutes before porpoises were again detected once the ADD was turned off.

Even though porpoises were being tracked before the ADD was activated, it proved very difficult to follow their movements once the ADD was active. In fact the focal animal immediately disappeared from the surface in six out of seven active trials and was not seen again. One animal was seen for four separate surfacings after ADD activation on one occasion and was obviously swimming away from it. During fifteen "response trials" the ADD was activated after porpoises had been tracked at greater ranges. Avoidance was assessed by direct observation and more objectively by combining a range of movement parameters. A porpoise being tracked at a range of 1.1 km immediately disappeared. During four trials at ranges between 1.6 and 2.4 km the animal turned away from the ADD and swam away in a more directional manner. Animals in six trials at ranges of 2.1 - 3.3 km showed no clear avoidance. Animals showing avoidance whose tracks could be determined moved with a mean speed of 1.6 m/s.

Propagation loss of the ADD signal was measured at the Baltic site and found to be much higher than was the case at the North Sea site. Brandt et al. (2012b) compared the predicted source levels for avoidance ranges seen at both the Baltic and North Sea sites, fairly complete deterrence appeared to occur at levels of 132 dB re 1 µPa and higher and there seemed to be not clear indications of avoidance at levels below 119 dB re 1 µPa.

These studies complement each other and, taken together, they provide a rather complete picture. The North Sea data were collected in an offshore location typical of those for which mitigation will most often be required in the real world. Here, the POD data revealed effects over large spatial and temporal scales while the aerial survey was very valuable in showing that reduced POD detections reflect a much lower porpoise density, rather than a change in acoustic behaviour. The data collected from the Baltic complements this by revealing the behaviour and movements that lead to exclusion and the speed with which these occur. The Baltic data provided higher resolution information on acoustic received level thresholds for response and it is encouraging that these broadly agree with those from the North Sea.

It is unfortunate however that the research used an aquaculture ADD as their aversive sound source. As noted elsewhere in this report ( section 2.2.3) there is little evidence that seals are excluded completely from fish farms by acoustic deterrent devices, and some seals are reported to tolerate ADDs even at farms where they are not depredating the farmed fish. Reports of initial success of ADDs when first introduced do suggest some deterrence effects on naïve animals but, as far as we are aware, no one has measured avoidance in seals over the ranges required for pile driving mitigation. Given seals' propensity to range over wide areas and the near ubiquitous use of ADDs in some regions (particularly in Scottish waters for aquaculture) it is quite possible that seals have learned not to avoid particular ADD signals and their use at pile driving sites may therefore be ineffective. Existing 'aquaculture' ADDs then could be an unwise choice as a tool for mitigating effects of pile driving and other high risk activities on seals.

KG No.	Knowledge Gap
43	To what extent can a learnt response to a specific signal ( e.g. non-response to aquaculture ADDs) be transferred to a different context?

4.4.4.2 Research with Captive Harbour Porpoise and Harbour Seals

A second substantial piece of work that was intended to investigate the feasibility of using aversive sounds for mitigation of pile driving, and funded by COWRIE (The Crown Estates Offshore Wind Research Fund), was a set of captive studies by Kastelein et al. (2010) of responses of a harbour porpoise and two female harbour seals to signals from acoustic deterrents, including a pinger (Netmark 100) and two commercially available seal scarers (the Ace Aquatec and the Lofitech). This team measured the detection level for signals from the three devices in both harbour seals and the porpoise. They found them to be very similar to those predicted on the basis of the animals' audiograms. To measure the signals' capacity to deter animals from an area and thus assess their potential as mitigation tools the researchers allowed the subjects (seals or porpoises) to move freely in a pool which contained a speaker broadcasting recordings of ADD signals. Using a speaker rather than the devices themselves allowed broadcasts at lower intensities. One signal was used in each trial and was broadcast at one of three source levels, a level just below that at which a clear response had been observed, a level with a moderate response and a level at which high level of response was expected (in the case of seals this response often involved hauling out to remove themselves completely from the underwater noise). These levels differed somewhat between devices. During trials surfacing rate and an assessment of swim speed were scored and specific behavioural responses, including leaping and holding their heads out of the water in the porpoises, and hauling out for seals, were scored. Trials lasted for 30 minutes and mean received level for sessions was assessed by noting the location of surfacing and relating these to sound levels previously measured in different areas of the tank. Porpoises swam significantly faster, showed more leaping behaviour and had a greater mean distance from the device at higher broadcast levels. Seals also behaved in ways that distanced them from the sound source as levels increased. Kastelein et al. (2010) concluded that the Netmark 100 would be unlikely to deter porpoises while the Ace Aquatec and Lofitech would be likely to deter porpoises at ranges between 0.2 and 1.2 km. For seals they suggested the Netmark 100 would be unlikely to have a deterrent effect but the Ace Aquatec and Lofitech should be effective at ranges of between 0.2 and 4.1 km.

It is very difficult to relate observations of trained captive animals in the highly constrained and artificial environment of a test tank to behaviours that might be expected to be elicited from the same species in the wild. In the first place, the received signal itself might be quite different in a constrained reverberant tank. In addition, the context for a trained captive animal will be quite different to that in the wild. The captive animal is probably well-fed and cared for and should be used to being exposed to experimental procedures that do not ultimately harm it. Fear of predation is also unlikely to be a strong motivation. Furthermore, in this case, the animals' ability to display the very behaviour of interest, removing itself from an aversive sound by swimming to a significant distance, was eliminated by the pool size. It is informative therefore to compare Kastelein et al.'s (2010) prediction of maximum deterrence range for the Lofitech ADD for harbour porpoises (0.2-1.2 km) with the real world, field observations of (2.4 to 7.5 km+) made by Brandt et al. (2012). The work described by Kastelein et al. (2010) was carried out by SEAMARCO (Sea Mammal Research Company), one of the world's premier dedicated captive marine research facilities and there is no question that the research was carried out to a very high standard. It is clear, however, that results from captive studies, no matter how well executed, do not provide useful predictions of behaviour in the wild. Behaviour of wild marine mammals, especially straight-forward activities such as movements, can be measured directly and it is only this approach that we believe will provide reliable results.

4.5.5 Discussion

Aversive sound mitigation is essentially a straight-forward concept and we judge the potential for success to be high.

The examples reviewed above provide many instances of marine mammals moving substantial distances in response to sounds well below the level at which hearing damage would be a concern. Indeed, marine mammals seem more likely to move significant distances in response to sound than do terrestrial mammals. Several biological factors may contribute to this. In the first place marine mammals are acoustically oriented animals; they have very sensitive hearing, sound travels very efficiently underwater and their primary sensory modality is acoustic. They also live in a habitat that might favour flight as a strategy for avoiding threats. For many terrestrial animals, crypsis and/or hiding represent good strategies to adopt in response to a frightening signal and to avoid predators. For marine mammals in the open sea this is unlikely to be an option for two reasons: there are few refuges or hiding places at the surface in the open sea, and the length of time they can remain still and hidden underwater, near the bottom for example, is limited by the need to return to the surface to breathe. For these reasons, aversive signals may be more likely to cause displacement in marine mammals than in many terrestrial animals and we might expect to achieve better results than have been shown in some terrestrial studies.

Many attempts to use aversive sounds to keep pests or predators away from sensitive areas such as agricultural sites (including the use of ADDs at fish farms) have shown reduced success over time because of declining responsiveness of animals that may be highly motivated to feed on the food resource being protected. Neither of these concerns should apply in this case. Clearly, a food resource is not being protected so there should be no strong motivation for the animals to remain in the area. Habituation is also unlikely because the signal need only be broadcast for a short period of time (this is one of the reasons why it is important not to use signals such as ADDs that other marine uses may leave activated for long periods). In addition, in this case, if any learning does take place, we might expect it to result in sensitisation because the animal may learn to associate the ADD signals with the unpleasant sensation of pile driving that follows.

While the concept of aversive sound mitigation is straight-forward many details still need to be explored. In particular, a range of appropriate signals must be tested in field conditions on the range of species of concern to provide regulators with reliable and auditable foundation of knowledge with which to justify amended mitigation protocols.