The direct comparison of monthly losses between the pre-deployment period, test period and control sites showed that the startle method was capable of significantly reducing predation losses throughout the one year deployment period (Fig 3). This is confirmed by the predation model which showed that sound exposure was the most important explanatory factor with respect to variation in seal predation. The model also revealed that predation varied throughout the year, although different sites had similar fish losses during different times of year with overall predation levels similar across sites. This shows that the low level of predation on the test site during sound exposure was not due to seasonality of predation but was due to the deterrence equipment.
Furthermore, predation was completely absent for 8 consecutive months (June 2011 to January 2012). Throughout the 13 months period of systems operation, there were only 5 events of predation on the test site, three of which were negligible. The most likely explanation for the occurrence of predation while the equipment was operating is that the predating individuals had compromised hearing which could be either the result of genetic predisposition, old age or previous exposure to anthropogenic noise source (such as commercially available seal scarers).
The experiment at Ardmaddy showed that the startle method is capable of reducing predation where losses prior to deployment were high. The experiment in Orkney at the Quanterness fish farm remains inconclusive. While predation was absent in the week after deployment, we do not know when it was operational in the following week when predation returned to pre-deployment levels. The exact breakdown date is unknown making further quantitative analysis impossible.
Animal abundance and movement
A unexpected result was that seals were frequently seen in the vicinity of the cages during sound exposure. This is in contrast to results from our earlier report (Janik & Götz 2008) where surface positions proved to be a good predictor for movement responses. In these earlier studies seals were excluded from a zone of about 25m around the transducer and seal numbers were lower up to 60m from a transducer operating at a source level of 180 dB re 1µPa.
The main difference between these studies was that the current study involved continuous sound exposure for 24 hours a day over several months while the previous studies used limited exposure times in a controlled experimental setup with several repetitions of different treatments. There are two possible interpretations for these results. First, it is possible that seals habituated to the deterrent sound pulses and were therefore not affected in their movement responses. However, this seems unlikely given that predation levels were low and predation was virtually absent for the second half of the test period. The alternative and more likely explanation is that seals sensitised to the sound and increasingly reduced dive times. Through extended surface swimming seals could have removed themselves from sound exposure making them more likely to be detected by human observers on the barge. This second scenario is consistent with previous captive studies that showed that repeated exposure to startling pulses result in a decrease in diving behaviour (Götz and Janik 2011).
Further observations to investigate the diving behaviour of seals were conducted on a single day when we measured 7 seal tracks underwater using a Tritech Gemini 720 sonar. In only one of these tracks did a seal approach the speaker to a distance of less than 20m, slightly less than the exclusion zone found in our previous study. There is an additional factor that should be considered regarding seal movement behaviour. A large percentage of the close tracks occurred in August/September 2011, approximately 2-3 month following the harbour seal pupping season. Juvenile seals are often positively buoyant (due to high fat content after weaning) and juveniles accounted for most tracks in our study. Thus, energetic constraints may have reduced dive times and led to increased surface swimming.
Harbour porpoise movement behaviour and abundance throughout the year was not affected by the deterrence setup with porpoises regularly seen within 15m of the loudspeakers. This confirms findings from our previous report which showed that the startle method can be tuned towards the hearing thresholds of target species, so that it affects seals but not porpoises. Furthermore, the deterrence method can be adjusted for other species so that it targets porpoises but not seals, or deters both taxa. This could be useful when trying to clear an area of marine mammals during industrial operations such as wind farm construction.
Distribution of otters around the farm did not seem to be affected by the deterrence system. Although the closest observed approach distance was smaller during control periods the data showed that this species is not negatively affected in areas beyond 50m from the farm. Therefore, the startle method can be used safely for seal deterrence in areas where otters are present.
The startle reflex in echolocating toothed whales
Bottlenose dolphins use brief high-intensity sound pulses for echolocation and are known to possess the ability to regulate their auditory sensitivity when solving echolocation tasks (Nachtigall & Supin 2008). This has posed the question whether dolphins are capable of suppressing the startle reflex under certain circumstances. However, the data from this study showed that external sound pulses at sound pressure levels much lower than those used by dolphins for echolocation are capable of eliciting the startle reflex. One contributing factor may be that the time of the playback could not easily be predicted by the dolphins due to the randomly selected sound presentation times.
The fact that the startle threshold roughly followed the hearing threshold of the dolphins across a range of different frequencies is consistent with data on terrestrial mammals (Pilz et al. 1987). We found that the difference between the auditory threshold and startle threshold appeared to be only 45 dB. This would be considerably lower than for any other mammal. However, the hearing thresholds obtained from the dolphins were almost 40 dB higher than previously measured thresholds for this species (Johnson 1967). The reasons for this are: a) the hearing threshold was a masked threshold due to the noise caused by snapping shrimp in the test pens, b) the thresholds were obtained with an electro-physiological method which typically yields approximately 20 dB higher thresholds than traditional psycho-physical methods, and c) the animals may have had some degree of age-related hearing loss at higher frequencies. Hence, even when excluding the possibility of mild hearing loss it seems fair to assume that actual unmasked hearing thresholds obtained with a psycho-physical method would have been at least 35 dB lower. This means that sensation levels needed to trigger the startle reflex in dolphins are more likely to be in the order 80 dB which is much closer to values previously reported for other species (Stoddart et al. 2008).
The startle reflex has previously been shown to induce flight responses, interrupt foraging behaviour and cause sensitisation of subsequent avoidance behaviour in pinnipeds (Götz and Janik, 2011). One dolphin tested in this study also backed out of the hoop station during the first exposures. This behaviour in a highly trained animal suggests that a startle response is likely to be followed by flight and avoidance behaviour in wild untrained dolphins. Thus, we think it is possible to develop a deterrence system for dolphins and porpoises by choosing a startle sound at a frequency of around 40 kHz. This means that a startle-based deterrence system could potentially be used to keep porpoises and dolphins away from areas of harm such as noisy marine construction sites. Further applications could involve guiding marine mammals around tidal turbines to mitigate collision risk or deterring porpoises from gillnets more reliably than can be achieved with current pingers.
While our study confirmed that startle thresholds are frequency specific, it also demonstrated that startle responses can be elicited at moderate levels outside the most sensitive hearing range of dolphins. This means that many anthropogenic pulsed noise sources also have the potential to startle animals and therefore cause strong behavioural responses. If the startle reflex is the underlying mechanism for such strong aversion responses then noise effects could be mitigated by increasing the onset-time of the sounds that cause the reactions.
Implications for regulators and management
The deterrent system tested in this study operated at a duty cycle of less than 1% which is between one and two orders of magnitude lower than the current commercially available deterrent devices. The fact that brief, isolated pulses were emitted at only moderately loud source levels means that noise pollution was greatly reduced and the potential for masking communication signals or hearing damage is low. This is in contrast to current commercially available seal scarers, which emit sound at high duty cycles and high source levels (Lepper et al. 2004).
Noise pollution for cetaceans however is a concern. Long-term and large-scale habitat exclusion has been found for odontocetes around operating ADDs at relatively low received levels ( e.g. Morton and Symonds 2002; Olesiuk et al. 2002). While most of these reactions were reported for Airmar devices, ADDs of other manufacturers produce even more energy at high frequencies where odontocetes are most sensitive, and may therefore have an even more severe effect. Furthermore, porpoises have been shown to suffer temporary hearing damage at relatively low levels (Lucke 2007) and current commercially available acoustic deterrent devices may have the potential to damage the ears of odontocetes (Götz and Hastie 2009). The use of lower source levels, low duty cycles and large gaps between brief acoustic emissions as tested in this study will remove the risk of hearing damage. We recommend a scientific evaluation of the potential for damage to wildlife before a device is approved for use in the marine environment.
Apart from decreasing noise pollution and effects on non-target species, the startle method also appeared to be more effective at reducing seal predation than existing commercial ADDs (Götz & Janik 2010). The 13 months trial reported here confirmed the high effectiveness of the startle method using seal predation (rather than approaches) as a measure. This creates benefits for the farm and for seal populations. While farmers will experience less predation, the requirement for lethal removal of seals may be reduced. Furthermore, our comprehensive theodolite tracking data set showed that seals spent significant amounts of time in the vicinity of the cages but did not predate on farmed fish. Hence, removal of seals close to the farm does not necessarily solve the problem of fish damage since close approaches do not necessarily mean predation takes place. In relation to seal conservation and the decline of harbour seal populations in Scotland (Lonergan et all, 2007), efforts to reduce shooting and encourage the use of alternative technologies such as the startle method are important.
Our results on the startle threshold in bottlenose dolphins show that the startle method can be finely tuned to deter only seals, only toothed whales or both, depending on the exact design of the signal. Further tests on the potential for sensitization in delphinids and porpoises are needed. If successful, the startle method could be used for the control of marine mammal movement in fields beyond aquaculture such as marine construction and the operation of marine renewable energy devices.